clamav/libclamav/matcher.c

1528 lines
56 KiB
C
Raw Permalink Normal View History

2003-07-29 15:48:06 +00:00
/*
2025-02-14 10:24:30 -05:00
* Copyright (C) 2013-2025 Cisco Systems, Inc. and/or its affiliates. All rights reserved.
* Copyright (C) 2007-2013 Sourcefire, Inc.
*
* Authors: Tomasz Kojm
2003-07-29 15:48:06 +00:00
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as
* published by the Free Software Foundation.
2003-07-29 15:48:06 +00:00
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,
* MA 02110-1301, USA.
2003-07-29 15:48:06 +00:00
*/
#if HAVE_CONFIG_H
#include "clamav-config.h"
#endif
2003-07-29 15:48:06 +00:00
#include <string.h>
#include <ctype.h>
#include <sys/types.h>
#include <sys/stat.h>
#ifdef HAVE_UNISTD_H
#include <unistd.h>
#endif
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
#include <stdbool.h>
2003-07-29 15:48:06 +00:00
#include "clamav.h"
#include "clamav_rust.h"
2003-07-29 15:48:06 +00:00
#include "others.h"
#include "matcher-ac.h"
#include "matcher-bm.h"
2014-10-01 17:16:09 -04:00
#include "matcher-pcre.h"
#include "filetypes.h"
#include "matcher.h"
#include "pe.h"
#include "elf.h"
#include "execs.h"
#include "special.h"
#include "scanners.h"
#include "str.h"
#include "default.h"
#include "macho.h"
#include "fmap.h"
#include "pe_icons.h"
#include "regex/regex.h"
#include "filtering.h"
#include "perflogging.h"
2010-08-02 17:04:35 +03:00
#include "bytecode_priv.h"
#include "bytecode_api_impl.h"
#ifdef HAVE_YARA
#include "yara_clam.h"
2015-06-01 14:32:04 -04:00
#include "yara_exec.h"
#endif
#ifdef CLI_PERF_LOGGING
Record names of extracted files A way is needed to record scanned file names for two purposes: 1. File names (and extensions) must be stored in the json metadata properties recorded when using the --gen-json clamscan option. Future work may use this to compare file extensions with detected file types. 2. File names are useful when interpretting tmp directory output when using the --leave-temps option. This commit enables file name retention for later use by storing file names in the fmap header structure, if a file name exists. To store the names in fmaps, an optional name argument has been added to any internal scan API's that create fmaps and every call to these APIs has been modified to pass a file name or NULL if a file name is not required. The zip and gpt parsers required some modification to record file names. The NSIS and XAR parsers fail to collect file names at all and will require future work to support file name extraction. Also: - Added recursive extraction to the tmp directory when the --leave-temps option is enabled. When not enabled, the tmp directory structure remains flat so as to prevent the likelihood of exceeding MAX_PATH. The current tmp directory is stored in the scan context. - Made the cli_scanfile() internal API non-static and added it to scanners.h so it would be accessible outside of scanners.c in order to remove code duplication within libmspack.c. - Added function comments to scanners.h and matcher.h - Converted a TDB-type macros and LSIG-type macros to enums for improved type safey. - Converted more return status variables from `int` to `cl_error_t` for improved type safety, and corrected ooxml file typing functions so they use `cli_file_t` exclusively rather than mixing types with `cl_error_t`. - Restructured the magic_scandesc() function to use goto's for error handling and removed the early_ret_from_magicscan() macro and magic_scandesc_cleanup() function. This makes the code easier to read and made it easier to add the recursive tmp directory cleanup to magic_scandesc(). - Corrected zip, egg, rar filename extraction issues. - Removed use of extra sub-directory layer for zip, egg, and rar file extraction. For Zip, this also involved changing the extracted filenames to be randomly generated rather than using the "zip.###" file name scheme.
2020-03-19 21:23:54 -04:00
static inline void perf_log_filter(int32_t pos, int32_t length, int8_t trie)
{
cli_perf_log_add(RAW_BYTES_SCANNED, length);
cli_perf_log_add(FILTER_BYTES_SCANNED, length - pos);
cli_perf_log_count2(TRIE_SCANNED, trie, length - pos);
}
Record names of extracted files A way is needed to record scanned file names for two purposes: 1. File names (and extensions) must be stored in the json metadata properties recorded when using the --gen-json clamscan option. Future work may use this to compare file extensions with detected file types. 2. File names are useful when interpretting tmp directory output when using the --leave-temps option. This commit enables file name retention for later use by storing file names in the fmap header structure, if a file name exists. To store the names in fmaps, an optional name argument has been added to any internal scan API's that create fmaps and every call to these APIs has been modified to pass a file name or NULL if a file name is not required. The zip and gpt parsers required some modification to record file names. The NSIS and XAR parsers fail to collect file names at all and will require future work to support file name extraction. Also: - Added recursive extraction to the tmp directory when the --leave-temps option is enabled. When not enabled, the tmp directory structure remains flat so as to prevent the likelihood of exceeding MAX_PATH. The current tmp directory is stored in the scan context. - Made the cli_scanfile() internal API non-static and added it to scanners.h so it would be accessible outside of scanners.c in order to remove code duplication within libmspack.c. - Added function comments to scanners.h and matcher.h - Converted a TDB-type macros and LSIG-type macros to enums for improved type safey. - Converted more return status variables from `int` to `cl_error_t` for improved type safety, and corrected ooxml file typing functions so they use `cli_file_t` exclusively rather than mixing types with `cl_error_t`. - Restructured the magic_scandesc() function to use goto's for error handling and removed the early_ret_from_magicscan() macro and magic_scandesc_cleanup() function. This makes the code easier to read and made it easier to add the recursive tmp directory cleanup to magic_scandesc(). - Corrected zip, egg, rar filename extraction issues. - Removed use of extra sub-directory layer for zip, egg, and rar file extraction. For Zip, this also involved changing the extracted filenames to be randomly generated rather than using the "zip.###" file name scheme.
2020-03-19 21:23:54 -04:00
static inline int perf_log_tries(int8_t acmode, int8_t bm_called, int32_t length)
{
if (bm_called)
cli_perf_log_add(BM_SCANNED, length);
if (acmode)
cli_perf_log_add(AC_SCANNED, length);
return 0;
}
#else
Record names of extracted files A way is needed to record scanned file names for two purposes: 1. File names (and extensions) must be stored in the json metadata properties recorded when using the --gen-json clamscan option. Future work may use this to compare file extensions with detected file types. 2. File names are useful when interpretting tmp directory output when using the --leave-temps option. This commit enables file name retention for later use by storing file names in the fmap header structure, if a file name exists. To store the names in fmaps, an optional name argument has been added to any internal scan API's that create fmaps and every call to these APIs has been modified to pass a file name or NULL if a file name is not required. The zip and gpt parsers required some modification to record file names. The NSIS and XAR parsers fail to collect file names at all and will require future work to support file name extraction. Also: - Added recursive extraction to the tmp directory when the --leave-temps option is enabled. When not enabled, the tmp directory structure remains flat so as to prevent the likelihood of exceeding MAX_PATH. The current tmp directory is stored in the scan context. - Made the cli_scanfile() internal API non-static and added it to scanners.h so it would be accessible outside of scanners.c in order to remove code duplication within libmspack.c. - Added function comments to scanners.h and matcher.h - Converted a TDB-type macros and LSIG-type macros to enums for improved type safey. - Converted more return status variables from `int` to `cl_error_t` for improved type safety, and corrected ooxml file typing functions so they use `cli_file_t` exclusively rather than mixing types with `cl_error_t`. - Restructured the magic_scandesc() function to use goto's for error handling and removed the early_ret_from_magicscan() macro and magic_scandesc_cleanup() function. This makes the code easier to read and made it easier to add the recursive tmp directory cleanup to magic_scandesc(). - Corrected zip, egg, rar filename extraction issues. - Removed use of extra sub-directory layer for zip, egg, and rar file extraction. For Zip, this also involved changing the extracted filenames to be randomly generated rather than using the "zip.###" file name scheme.
2020-03-19 21:23:54 -04:00
static inline void perf_log_filter(int32_t pos, uint32_t length, int8_t trie)
{
UNUSEDPARAM(pos);
UNUSEDPARAM(length);
UNUSEDPARAM(trie);
}
Record names of extracted files A way is needed to record scanned file names for two purposes: 1. File names (and extensions) must be stored in the json metadata properties recorded when using the --gen-json clamscan option. Future work may use this to compare file extensions with detected file types. 2. File names are useful when interpretting tmp directory output when using the --leave-temps option. This commit enables file name retention for later use by storing file names in the fmap header structure, if a file name exists. To store the names in fmaps, an optional name argument has been added to any internal scan API's that create fmaps and every call to these APIs has been modified to pass a file name or NULL if a file name is not required. The zip and gpt parsers required some modification to record file names. The NSIS and XAR parsers fail to collect file names at all and will require future work to support file name extraction. Also: - Added recursive extraction to the tmp directory when the --leave-temps option is enabled. When not enabled, the tmp directory structure remains flat so as to prevent the likelihood of exceeding MAX_PATH. The current tmp directory is stored in the scan context. - Made the cli_scanfile() internal API non-static and added it to scanners.h so it would be accessible outside of scanners.c in order to remove code duplication within libmspack.c. - Added function comments to scanners.h and matcher.h - Converted a TDB-type macros and LSIG-type macros to enums for improved type safey. - Converted more return status variables from `int` to `cl_error_t` for improved type safety, and corrected ooxml file typing functions so they use `cli_file_t` exclusively rather than mixing types with `cl_error_t`. - Restructured the magic_scandesc() function to use goto's for error handling and removed the early_ret_from_magicscan() macro and magic_scandesc_cleanup() function. This makes the code easier to read and made it easier to add the recursive tmp directory cleanup to magic_scandesc(). - Corrected zip, egg, rar filename extraction issues. - Removed use of extra sub-directory layer for zip, egg, and rar file extraction. For Zip, this also involved changing the extracted filenames to be randomly generated rather than using the "zip.###" file name scheme.
2020-03-19 21:23:54 -04:00
static inline int perf_log_tries(int8_t acmode, int8_t bm_called, int32_t length)
{
UNUSEDPARAM(acmode);
UNUSEDPARAM(bm_called);
UNUSEDPARAM(length);
return 0;
}
#endif
static inline cl_error_t matcher_run(const struct cli_matcher *root,
const unsigned char *buffer, uint32_t length,
const char **virname, struct cli_ac_data *mdata,
uint32_t offset,
const struct cli_target_info *tinfo,
cli_file_t ftype,
struct cli_matched_type **ftoffset,
unsigned int acmode,
unsigned int pcremode,
struct cli_ac_result **acres,
fmap_t *map,
struct cli_bm_off *offdata,
struct cli_pcre_off *poffdata,
cli_ctx *ctx)
{
cl_error_t ret, saved_ret = CL_CLEAN;
int32_t pos = 0;
struct filter_match_info info;
uint32_t orig_length, orig_offset;
const unsigned char *orig_buffer;
if (root->filter) {
if (filter_search_ext(root->filter, buffer, length, &info) == -1) {
/* for safety always scan last maxpatlen bytes */
pos = length - root->maxpatlen - 1;
if (pos < 0) pos = 0;
Record names of extracted files A way is needed to record scanned file names for two purposes: 1. File names (and extensions) must be stored in the json metadata properties recorded when using the --gen-json clamscan option. Future work may use this to compare file extensions with detected file types. 2. File names are useful when interpretting tmp directory output when using the --leave-temps option. This commit enables file name retention for later use by storing file names in the fmap header structure, if a file name exists. To store the names in fmaps, an optional name argument has been added to any internal scan API's that create fmaps and every call to these APIs has been modified to pass a file name or NULL if a file name is not required. The zip and gpt parsers required some modification to record file names. The NSIS and XAR parsers fail to collect file names at all and will require future work to support file name extraction. Also: - Added recursive extraction to the tmp directory when the --leave-temps option is enabled. When not enabled, the tmp directory structure remains flat so as to prevent the likelihood of exceeding MAX_PATH. The current tmp directory is stored in the scan context. - Made the cli_scanfile() internal API non-static and added it to scanners.h so it would be accessible outside of scanners.c in order to remove code duplication within libmspack.c. - Added function comments to scanners.h and matcher.h - Converted a TDB-type macros and LSIG-type macros to enums for improved type safey. - Converted more return status variables from `int` to `cl_error_t` for improved type safety, and corrected ooxml file typing functions so they use `cli_file_t` exclusively rather than mixing types with `cl_error_t`. - Restructured the magic_scandesc() function to use goto's for error handling and removed the early_ret_from_magicscan() macro and magic_scandesc_cleanup() function. This makes the code easier to read and made it easier to add the recursive tmp directory cleanup to magic_scandesc(). - Corrected zip, egg, rar filename extraction issues. - Removed use of extra sub-directory layer for zip, egg, and rar file extraction. For Zip, this also involved changing the extracted filenames to be randomly generated rather than using the "zip.###" file name scheme.
2020-03-19 21:23:54 -04:00
perf_log_filter(pos, length, root->type);
} else {
/* must not cut buffer for 64[4-4]6161, because we must be able to check
Record names of extracted files A way is needed to record scanned file names for two purposes: 1. File names (and extensions) must be stored in the json metadata properties recorded when using the --gen-json clamscan option. Future work may use this to compare file extensions with detected file types. 2. File names are useful when interpretting tmp directory output when using the --leave-temps option. This commit enables file name retention for later use by storing file names in the fmap header structure, if a file name exists. To store the names in fmaps, an optional name argument has been added to any internal scan API's that create fmaps and every call to these APIs has been modified to pass a file name or NULL if a file name is not required. The zip and gpt parsers required some modification to record file names. The NSIS and XAR parsers fail to collect file names at all and will require future work to support file name extraction. Also: - Added recursive extraction to the tmp directory when the --leave-temps option is enabled. When not enabled, the tmp directory structure remains flat so as to prevent the likelihood of exceeding MAX_PATH. The current tmp directory is stored in the scan context. - Made the cli_scanfile() internal API non-static and added it to scanners.h so it would be accessible outside of scanners.c in order to remove code duplication within libmspack.c. - Added function comments to scanners.h and matcher.h - Converted a TDB-type macros and LSIG-type macros to enums for improved type safey. - Converted more return status variables from `int` to `cl_error_t` for improved type safety, and corrected ooxml file typing functions so they use `cli_file_t` exclusively rather than mixing types with `cl_error_t`. - Restructured the magic_scandesc() function to use goto's for error handling and removed the early_ret_from_magicscan() macro and magic_scandesc_cleanup() function. This makes the code easier to read and made it easier to add the recursive tmp directory cleanup to magic_scandesc(). - Corrected zip, egg, rar filename extraction issues. - Removed use of extra sub-directory layer for zip, egg, and rar file extraction. For Zip, this also involved changing the extracted filenames to be randomly generated rather than using the "zip.###" file name scheme.
2020-03-19 21:23:54 -04:00
* 64! */
pos = info.first_match - root->maxpatlen - 1;
if (pos < 0) pos = 0;
Record names of extracted files A way is needed to record scanned file names for two purposes: 1. File names (and extensions) must be stored in the json metadata properties recorded when using the --gen-json clamscan option. Future work may use this to compare file extensions with detected file types. 2. File names are useful when interpretting tmp directory output when using the --leave-temps option. This commit enables file name retention for later use by storing file names in the fmap header structure, if a file name exists. To store the names in fmaps, an optional name argument has been added to any internal scan API's that create fmaps and every call to these APIs has been modified to pass a file name or NULL if a file name is not required. The zip and gpt parsers required some modification to record file names. The NSIS and XAR parsers fail to collect file names at all and will require future work to support file name extraction. Also: - Added recursive extraction to the tmp directory when the --leave-temps option is enabled. When not enabled, the tmp directory structure remains flat so as to prevent the likelihood of exceeding MAX_PATH. The current tmp directory is stored in the scan context. - Made the cli_scanfile() internal API non-static and added it to scanners.h so it would be accessible outside of scanners.c in order to remove code duplication within libmspack.c. - Added function comments to scanners.h and matcher.h - Converted a TDB-type macros and LSIG-type macros to enums for improved type safey. - Converted more return status variables from `int` to `cl_error_t` for improved type safety, and corrected ooxml file typing functions so they use `cli_file_t` exclusively rather than mixing types with `cl_error_t`. - Restructured the magic_scandesc() function to use goto's for error handling and removed the early_ret_from_magicscan() macro and magic_scandesc_cleanup() function. This makes the code easier to read and made it easier to add the recursive tmp directory cleanup to magic_scandesc(). - Corrected zip, egg, rar filename extraction issues. - Removed use of extra sub-directory layer for zip, egg, and rar file extraction. For Zip, this also involved changing the extracted filenames to be randomly generated rather than using the "zip.###" file name scheme.
2020-03-19 21:23:54 -04:00
perf_log_filter(pos, length, root->type);
}
} else {
Record names of extracted files A way is needed to record scanned file names for two purposes: 1. File names (and extensions) must be stored in the json metadata properties recorded when using the --gen-json clamscan option. Future work may use this to compare file extensions with detected file types. 2. File names are useful when interpretting tmp directory output when using the --leave-temps option. This commit enables file name retention for later use by storing file names in the fmap header structure, if a file name exists. To store the names in fmaps, an optional name argument has been added to any internal scan API's that create fmaps and every call to these APIs has been modified to pass a file name or NULL if a file name is not required. The zip and gpt parsers required some modification to record file names. The NSIS and XAR parsers fail to collect file names at all and will require future work to support file name extraction. Also: - Added recursive extraction to the tmp directory when the --leave-temps option is enabled. When not enabled, the tmp directory structure remains flat so as to prevent the likelihood of exceeding MAX_PATH. The current tmp directory is stored in the scan context. - Made the cli_scanfile() internal API non-static and added it to scanners.h so it would be accessible outside of scanners.c in order to remove code duplication within libmspack.c. - Added function comments to scanners.h and matcher.h - Converted a TDB-type macros and LSIG-type macros to enums for improved type safey. - Converted more return status variables from `int` to `cl_error_t` for improved type safety, and corrected ooxml file typing functions so they use `cli_file_t` exclusively rather than mixing types with `cl_error_t`. - Restructured the magic_scandesc() function to use goto's for error handling and removed the early_ret_from_magicscan() macro and magic_scandesc_cleanup() function. This makes the code easier to read and made it easier to add the recursive tmp directory cleanup to magic_scandesc(). - Corrected zip, egg, rar filename extraction issues. - Removed use of extra sub-directory layer for zip, egg, and rar file extraction. For Zip, this also involved changing the extracted filenames to be randomly generated rather than using the "zip.###" file name scheme.
2020-03-19 21:23:54 -04:00
perf_log_filter(0, length, root->type);
}
orig_length = length;
orig_buffer = buffer;
orig_offset = offset;
length -= pos;
buffer += pos;
offset += pos;
if (!root->ac_only) {
Record names of extracted files A way is needed to record scanned file names for two purposes: 1. File names (and extensions) must be stored in the json metadata properties recorded when using the --gen-json clamscan option. Future work may use this to compare file extensions with detected file types. 2. File names are useful when interpretting tmp directory output when using the --leave-temps option. This commit enables file name retention for later use by storing file names in the fmap header structure, if a file name exists. To store the names in fmaps, an optional name argument has been added to any internal scan API's that create fmaps and every call to these APIs has been modified to pass a file name or NULL if a file name is not required. The zip and gpt parsers required some modification to record file names. The NSIS and XAR parsers fail to collect file names at all and will require future work to support file name extraction. Also: - Added recursive extraction to the tmp directory when the --leave-temps option is enabled. When not enabled, the tmp directory structure remains flat so as to prevent the likelihood of exceeding MAX_PATH. The current tmp directory is stored in the scan context. - Made the cli_scanfile() internal API non-static and added it to scanners.h so it would be accessible outside of scanners.c in order to remove code duplication within libmspack.c. - Added function comments to scanners.h and matcher.h - Converted a TDB-type macros and LSIG-type macros to enums for improved type safey. - Converted more return status variables from `int` to `cl_error_t` for improved type safety, and corrected ooxml file typing functions so they use `cli_file_t` exclusively rather than mixing types with `cl_error_t`. - Restructured the magic_scandesc() function to use goto's for error handling and removed the early_ret_from_magicscan() macro and magic_scandesc_cleanup() function. This makes the code easier to read and made it easier to add the recursive tmp directory cleanup to magic_scandesc(). - Corrected zip, egg, rar filename extraction issues. - Removed use of extra sub-directory layer for zip, egg, and rar file extraction. For Zip, this also involved changing the extracted filenames to be randomly generated rather than using the "zip.###" file name scheme.
2020-03-19 21:23:54 -04:00
perf_log_tries(0, 1, length);
if (root->bm_offmode) {
/* Don't use prefiltering for BM offset mode, since BM keeps tracks
Record names of extracted files A way is needed to record scanned file names for two purposes: 1. File names (and extensions) must be stored in the json metadata properties recorded when using the --gen-json clamscan option. Future work may use this to compare file extensions with detected file types. 2. File names are useful when interpretting tmp directory output when using the --leave-temps option. This commit enables file name retention for later use by storing file names in the fmap header structure, if a file name exists. To store the names in fmaps, an optional name argument has been added to any internal scan API's that create fmaps and every call to these APIs has been modified to pass a file name or NULL if a file name is not required. The zip and gpt parsers required some modification to record file names. The NSIS and XAR parsers fail to collect file names at all and will require future work to support file name extraction. Also: - Added recursive extraction to the tmp directory when the --leave-temps option is enabled. When not enabled, the tmp directory structure remains flat so as to prevent the likelihood of exceeding MAX_PATH. The current tmp directory is stored in the scan context. - Made the cli_scanfile() internal API non-static and added it to scanners.h so it would be accessible outside of scanners.c in order to remove code duplication within libmspack.c. - Added function comments to scanners.h and matcher.h - Converted a TDB-type macros and LSIG-type macros to enums for improved type safey. - Converted more return status variables from `int` to `cl_error_t` for improved type safety, and corrected ooxml file typing functions so they use `cli_file_t` exclusively rather than mixing types with `cl_error_t`. - Restructured the magic_scandesc() function to use goto's for error handling and removed the early_ret_from_magicscan() macro and magic_scandesc_cleanup() function. This makes the code easier to read and made it easier to add the recursive tmp directory cleanup to magic_scandesc(). - Corrected zip, egg, rar filename extraction issues. - Removed use of extra sub-directory layer for zip, egg, and rar file extraction. For Zip, this also involved changing the extracted filenames to be randomly generated rather than using the "zip.###" file name scheme.
2020-03-19 21:23:54 -04:00
* of offsets itself, and doesn't work if we skip chunks of input
* data */
ret = cli_bm_scanbuff(orig_buffer, orig_length, virname, NULL, root, orig_offset, tinfo, offdata, ctx);
} else {
ret = cli_bm_scanbuff(buffer, length, virname, NULL, root, offset, tinfo, offdata, ctx);
}
if (ret != CL_SUCCESS) {
if (ret != CL_VIRUS)
return ret;
/* else (ret == CL_VIRUS) */
ret = cli_append_virus(ctx, *virname);
if (ret != CL_SUCCESS)
return ret;
}
}
Record names of extracted files A way is needed to record scanned file names for two purposes: 1. File names (and extensions) must be stored in the json metadata properties recorded when using the --gen-json clamscan option. Future work may use this to compare file extensions with detected file types. 2. File names are useful when interpretting tmp directory output when using the --leave-temps option. This commit enables file name retention for later use by storing file names in the fmap header structure, if a file name exists. To store the names in fmaps, an optional name argument has been added to any internal scan API's that create fmaps and every call to these APIs has been modified to pass a file name or NULL if a file name is not required. The zip and gpt parsers required some modification to record file names. The NSIS and XAR parsers fail to collect file names at all and will require future work to support file name extraction. Also: - Added recursive extraction to the tmp directory when the --leave-temps option is enabled. When not enabled, the tmp directory structure remains flat so as to prevent the likelihood of exceeding MAX_PATH. The current tmp directory is stored in the scan context. - Made the cli_scanfile() internal API non-static and added it to scanners.h so it would be accessible outside of scanners.c in order to remove code duplication within libmspack.c. - Added function comments to scanners.h and matcher.h - Converted a TDB-type macros and LSIG-type macros to enums for improved type safey. - Converted more return status variables from `int` to `cl_error_t` for improved type safety, and corrected ooxml file typing functions so they use `cli_file_t` exclusively rather than mixing types with `cl_error_t`. - Restructured the magic_scandesc() function to use goto's for error handling and removed the early_ret_from_magicscan() macro and magic_scandesc_cleanup() function. This makes the code easier to read and made it easier to add the recursive tmp directory cleanup to magic_scandesc(). - Corrected zip, egg, rar filename extraction issues. - Removed use of extra sub-directory layer for zip, egg, and rar file extraction. For Zip, this also involved changing the extracted filenames to be randomly generated rather than using the "zip.###" file name scheme.
2020-03-19 21:23:54 -04:00
perf_log_tries(acmode, 0, length);
ret = cli_ac_scanbuff(buffer, length, virname, NULL, acres, root, mdata, offset, ftype, ftoffset, acmode, ctx);
if (ret != CL_SUCCESS) {
if (ret == CL_VIRUS) {
ret = cli_append_virus(ctx, *virname);
if (ret != CL_SUCCESS)
return ret;
} else if (ret > CL_TYPENO && acmode & AC_SCAN_VIR) {
saved_ret = ret;
} else {
return ret;
}
}
if (root->bcomp_metas) {
Fix byte-compare subsignature premature alert The byte compare feature in logical signatures will cause the rule to alert if it successfully matches regardless of the rest of the logical signature. An easy way to test this is with a logical signature that has two bcomp subsignatures and requires both to match for the rule to alert. In the following example, we have 4 signatures where - the first will match both bcomp subsigs. - the second will match neither. - the last two match just one bcomp subsig. In an --allmatch test, you'll find that the 3 of these match, with the first one matching *twice*, once for each bcomp subsig. test.ldb: ``` bcomp.both;Engine:51-255,Target:0;0&1&2&3;4141;0(>>5#hb2#=123);4242;2(>>5#hb2#=255) bcomp.neither;Engine:51-255,Target:0;0&1&2&3;4141;0(>>5#hb2#=124);4242;2(>>5#hb2#=254) bcomp.second;Engine:51-255,Target:0;0&1&2&3;4141;0(>>5#hb2#=124);4242;2(>>5#hb2#=255) bcomp.first;Engine:51-255,Target:0;0&1&2&3;4141;0(>>5#hb2#=123);4242;2(>>5#hb2#=254) ``` test.sample: ``` AA = 7B; BB = FF ``` You can also try a similar test to compare the behavior with regular ac-pattern-match subsigs with this lsig-test.ldb: ``` pattern.both;Engine:51-255,Target:0;0&1;4141;4242 pattern.neither;Engine:51-255,Target:0;0&1;4140;4241 pattern.second;Engine:51-255,Target:0;0&1;4140;4242 pattern.first;Engine:51-255,Target:0;0&1;4141;4241 ``` This commit fixes the issue by incrementing the logical subsignature count for each bcomp subsig match instead of appending an alert for each bcomp match. Also removed call to `lsig_sub_matched()` that didn't do anything.
2022-02-03 16:06:05 -08:00
ret = cli_bcomp_scanbuf(orig_buffer, orig_length, acres, root, mdata, ctx);
if (ret != CL_CLEAN) {
if (ret > CL_TYPENO && acmode & AC_SCAN_VIR) {
saved_ret = ret;
} else {
return ret;
}
}
}
switch (ftype) {
case CL_TYPE_GIF:
case CL_TYPE_TIFF:
case CL_TYPE_JPEG:
case CL_TYPE_PNG:
case CL_TYPE_GRAPHICS: {
if (ctx->recursion_stack[ctx->recursion_level].calculated_image_fuzzy_hash &&
!fuzzy_hash_check(root->fuzzy_hashmap, mdata, ctx->recursion_stack[ctx->recursion_level].image_fuzzy_hash)) {
cli_errmsg("Unexpected error when checking for fuzzy hash matches.\n");
return CL_ERROR;
}
}
default:
break;
}
/* due to logical triggered, pcres cannot be evaluated until after full subsig matching */
/* cannot save pcre execution state without possible evasion; must scan entire buffer */
/* however, scanning the whole buffer may require the whole buffer being loaded into memory */
if (root->pcre_metas) {
int rc;
uint64_t maxfilesize;
2014-10-28 13:37:42 -04:00
if (map && (pcremode == PCRE_SCAN_FMAP)) {
if (offset + length >= map->len) {
/* check that scanned map does not exceed pcre maxfilesize limit */
maxfilesize = (uint64_t)cl_engine_get_num(ctx->engine, CL_ENGINE_PCRE_MAX_FILESIZE, &rc);
if (rc != CL_SUCCESS)
return rc;
if (maxfilesize && (map->len > maxfilesize)) {
2016-02-22 13:26:15 -05:00
cli_dbgmsg("matcher_run: pcre max filesize (map) exceeded (limit: %llu, needed: %llu)\n",
(long long unsigned)maxfilesize, (long long unsigned)map->len);
return CL_EMAXSIZE;
}
cli_dbgmsg("matcher_run: performing regex matching on full map: %u+%u(%u) >= %zu\n", offset, length, offset + length, map->len);
buffer = fmap_need_off_once(map, 0, map->len);
if (!buffer)
return CL_EMEM;
/* scan the full buffer */
ret = cli_pcre_scanbuf(buffer, map->len, virname, acres, root, mdata, poffdata, ctx);
}
} else if (pcremode == PCRE_SCAN_BUFF) {
/* check that scanned buffer does not exceed pcre maxfilesize limit */
maxfilesize = (uint64_t)cl_engine_get_num(ctx->engine, CL_ENGINE_PCRE_MAX_FILESIZE, &rc);
if (rc != CL_SUCCESS)
return rc;
if (maxfilesize && (length > maxfilesize)) {
2016-02-22 13:26:15 -05:00
cli_dbgmsg("matcher_run: pcre max filesize (buf) exceeded (limit: %llu, needed: %u)\n", (long long unsigned)maxfilesize, length);
return CL_EMAXSIZE;
}
cli_dbgmsg("matcher_run: performing regex matching on buffer with no map: %u+%u(%u)\n", offset, length, offset + length);
/* scan the specified buffer */
ret = cli_pcre_scanbuf(buffer, length, virname, acres, root, mdata, poffdata, ctx);
}
}
/* end experimental fragment */
if (ctx && ret == CL_VIRUS) {
ret = cli_append_virus(ctx, *virname);
if (ret != CL_SUCCESS)
return ret;
}
if (saved_ret && ret == CL_CLEAN) {
return saved_ret;
}
return ret;
}
cl_error_t cli_scan_buff(const unsigned char *buffer, uint32_t length, uint32_t offset, cli_ctx *ctx, cli_file_t ftype, struct cli_ac_data **acdata)
{
cl_error_t ret = CL_CLEAN;
unsigned int i = 0, j = 0;
struct cli_ac_data matcher_data;
struct cli_matcher *generic_ac_root, *target_ac_root = NULL;
const char *virname = NULL;
const struct cl_engine *engine = ctx->engine;
if (!engine) {
cli_errmsg("cli_scan_buff: engine == NULL\n");
return CL_ENULLARG;
}
generic_ac_root = engine->root[0]; /* generic signatures */
if (ftype != CL_TYPE_ANY) {
// Identify the target type, to find the matcher root for that target.
for (i = 1; i < CLI_MTARGETS; i++) {
for (j = 0; j < cli_mtargets[i].target_count; ++j) {
if (cli_mtargets[i].target[j] == ftype) {
// Identified the target type, now get the matcher root for that target.
target_ac_root = ctx->engine->root[i];
break; // Break out of inner loop
}
}
if (target_ac_root) break;
}
}
if (target_ac_root) {
/* If a target-specific specific signature root was found for the given file type, match with it. */
if (!acdata) {
// no ac matcher data was provided, so we need to initialize our own.
ret = cli_ac_initdata(&matcher_data, target_ac_root->ac_partsigs, target_ac_root->ac_lsigs, target_ac_root->ac_reloff_num, CLI_DEFAULT_AC_TRACKLEN);
if (CL_SUCCESS != ret) {
return ret;
}
}
ret = matcher_run(target_ac_root, buffer, length, &virname,
acdata ? (acdata[0]) : (&matcher_data),
offset, NULL, ftype, NULL, AC_SCAN_VIR, PCRE_SCAN_BUFF, NULL, ctx->fmap, NULL, NULL, ctx);
if (!acdata) {
// no longer need our AC local matcher data (if using)
cli_ac_freedata(&matcher_data);
}
if (ret == CL_EMEM || ret == CL_VIRUS) {
return ret;
}
// reset virname back to NULL for matching with the generic AC root.
virname = NULL;
}
if (!acdata) {
// no ac matcher data was provided, so we need to initialize our own.
ret = cli_ac_initdata(&matcher_data, generic_ac_root->ac_partsigs, generic_ac_root->ac_lsigs, generic_ac_root->ac_reloff_num, CLI_DEFAULT_AC_TRACKLEN);
if (CL_SUCCESS != ret) {
return ret;
}
}
2003-07-29 15:48:06 +00:00
ret = matcher_run(generic_ac_root, buffer, length, &virname,
acdata ? (acdata[1]) : (&matcher_data),
offset, NULL, ftype, NULL, AC_SCAN_VIR, PCRE_SCAN_BUFF, NULL, ctx->fmap, NULL, NULL, ctx);
if (!acdata) {
// no longer need our AC local matcher data (if using)
cli_ac_freedata(&matcher_data);
}
return ret;
2003-07-29 15:48:06 +00:00
}
/*
* offdata[0]: type
* offdata[1]: offset value
* offdata[2]: max shift
* offdata[3]: section number
*/
cl_error_t cli_caloff(const char *offstr, const struct cli_target_info *info, cli_target_t target, uint32_t *offdata, uint32_t *offset_min, uint32_t *offset_max)
{
char offcpy[65] = {0};
unsigned int n = 0, val = 0;
char *pt = NULL;
if (!info) { /* decode offset string */
if (!offstr) {
cli_errmsg("cli_caloff: offstr == NULL\n");
return CL_ENULLARG;
}
if (!strcmp(offstr, "*")) {
offdata[0] = *offset_max = *offset_min = CLI_OFF_ANY;
return CL_SUCCESS;
}
if (strlen(offstr) > 64) {
cli_errmsg("cli_caloff: Offset string too long\n");
return CL_EMALFDB;
}
strcpy(offcpy, offstr);
if ((pt = strchr(offcpy, ','))) {
if (!cli_isnumber(pt + 1)) {
cli_errmsg("cli_caloff: Invalid offset shift value\n");
return CL_EMALFDB;
}
offdata[2] = atoi(pt + 1);
*pt = 0;
} else {
offdata[2] = 0;
}
*offset_max = *offset_min = CLI_OFF_NONE;
if (!strncmp(offcpy, "EP+", 3) || !strncmp(offcpy, "EP-", 3)) {
if (offcpy[2] == '+')
offdata[0] = CLI_OFF_EP_PLUS;
else
offdata[0] = CLI_OFF_EP_MINUS;
if (!cli_isnumber(&offcpy[3])) {
cli_errmsg("cli_caloff: Invalid offset value\n");
return CL_EMALFDB;
}
offdata[1] = atoi(&offcpy[3]);
} else if (offcpy[0] == 'S') {
if (offcpy[1] == 'E') {
if (!cli_isnumber(&offcpy[2])) {
cli_errmsg("cli_caloff: Invalid section number\n");
return CL_EMALFDB;
}
offdata[0] = CLI_OFF_SE;
offdata[3] = atoi(&offcpy[2]);
} else if (!strncmp(offstr, "SL+", 3)) {
offdata[0] = CLI_OFF_SL_PLUS;
if (!cli_isnumber(&offcpy[3])) {
cli_errmsg("cli_caloff: Invalid offset value\n");
return CL_EMALFDB;
}
offdata[1] = atoi(&offcpy[3]);
} else if (sscanf(offcpy, "S%u+%u", &n, &val) == 2) {
offdata[0] = CLI_OFF_SX_PLUS;
offdata[1] = val;
offdata[3] = n;
} else {
cli_errmsg("cli_caloff: Invalid offset string\n");
return CL_EMALFDB;
}
} else if (!strncmp(offcpy, "EOF-", 4)) {
offdata[0] = CLI_OFF_EOF_MINUS;
if (!cli_isnumber(&offcpy[4])) {
cli_errmsg("cli_caloff: Invalid offset value\n");
return CL_EMALFDB;
}
offdata[1] = atoi(&offcpy[4]);
} else if (!strncmp(offcpy, "VI", 2)) {
/* versioninfo */
offdata[0] = CLI_OFF_VERSION;
} else if (strchr(offcpy, '$')) {
if (sscanf(offcpy, "$%u$", &n) != 1) {
cli_errmsg("cli_caloff: Invalid macro($) in offset: %s\n", offcpy);
return CL_EMALFDB;
}
if (n >= 32) {
cli_errmsg("cli_caloff: at most 32 macro groups supported\n");
return CL_EMALFDB;
}
offdata[0] = CLI_OFF_MACRO;
offdata[1] = n;
} else {
offdata[0] = CLI_OFF_ABSOLUTE;
if (!cli_isnumber(offcpy)) {
cli_errmsg("cli_caloff: Invalid offset value\n");
return CL_EMALFDB;
}
*offset_min = offdata[1] = atoi(offcpy);
*offset_max = *offset_min + offdata[2];
}
if (offdata[0] != CLI_OFF_ANY && offdata[0] != CLI_OFF_ABSOLUTE &&
offdata[0] != CLI_OFF_EOF_MINUS && offdata[0] != CLI_OFF_MACRO) {
if (target != TARGET_PE && target != TARGET_ELF && target != TARGET_MACHO) {
cli_errmsg("cli_caloff: Invalid offset type for target %u\n", target);
return CL_EMALFDB;
}
}
} else {
/* calculate relative offsets */
*offset_min = CLI_OFF_NONE;
2012-07-12 10:21:00 -04:00
if (offset_max)
*offset_max = CLI_OFF_NONE;
PE parsing code improvements, db loading bug fixes Consolidate the PE parsing code into one function. I tried to preserve all existing functionality from the previous, distinct implementations to a large extent (with the exceptions mentioned below). If I noticed potential bugs/improvements, I added a TODO statement about those so that they can be fixed in a smaller commit later. Also, there are more TODOs in places where I'm not entirely sure why certain actions are performed - more research is needed for these. I'm submitting a pull request now so that regression testing can be done, and because merging what I have thus far now will likely have fewer conflicts than if I try to merge later PE parsing code improvements: - PEs without all 16 data directories are parsed more appropriately now - Added lots more debug statements Also: - Allow MAX_BC and MAX_TRACKED_PCRE to be specified via CFLAGS When doing performance testing with the latest CVD, MAX_BC and MAX_TRACKED_PCRE need to be raised to track all the events. Allow these to be specified via CFLAGS by not redefining them if they are already defined - Fix an issue preventing wildcard sizes in .MDB/.MSB rules I'm not sure what the original intent of the check I removed was, but it prevents using wildcard sizes in .MDB/.MSB rules. AFAICT these wildcard sizes should be handled appropriately by the MD5 section hash computation code, so I don't think a check on that is needed. - Fix several issues related to db loading - .imp files will now get loaded if they exist in a directory passed via clamscan's '-d' flag - .pwdb files will now get loaded if they exist in a directory passed via clamscan's '-d' flag even when compiling without yara support - Changes to .imp, .ign, and .ign2 files will now be reflected in calls to cl_statinidir and cl_statchkdir (and also .pwdb files, even when compiling without yara support) - The contents of .sfp files won't be included in some of the signature counts, and the contents of .cud files will be - Any local.gdb files will no longer be loaded twice - For .imp files, you are no longer required to specify a minimum flevel for wildcard rules, since this isn't needed
2019-01-08 00:09:08 -05:00
if (info->status == -1) {
// If the executable headers weren't parsed successfully then we
// can't process any ndb/ldb EOF-n/EP+n/EP-n/Sx+n/SEx/SL+n subsigs
return CL_SUCCESS;
PE parsing code improvements, db loading bug fixes Consolidate the PE parsing code into one function. I tried to preserve all existing functionality from the previous, distinct implementations to a large extent (with the exceptions mentioned below). If I noticed potential bugs/improvements, I added a TODO statement about those so that they can be fixed in a smaller commit later. Also, there are more TODOs in places where I'm not entirely sure why certain actions are performed - more research is needed for these. I'm submitting a pull request now so that regression testing can be done, and because merging what I have thus far now will likely have fewer conflicts than if I try to merge later PE parsing code improvements: - PEs without all 16 data directories are parsed more appropriately now - Added lots more debug statements Also: - Allow MAX_BC and MAX_TRACKED_PCRE to be specified via CFLAGS When doing performance testing with the latest CVD, MAX_BC and MAX_TRACKED_PCRE need to be raised to track all the events. Allow these to be specified via CFLAGS by not redefining them if they are already defined - Fix an issue preventing wildcard sizes in .MDB/.MSB rules I'm not sure what the original intent of the check I removed was, but it prevents using wildcard sizes in .MDB/.MSB rules. AFAICT these wildcard sizes should be handled appropriately by the MD5 section hash computation code, so I don't think a check on that is needed. - Fix several issues related to db loading - .imp files will now get loaded if they exist in a directory passed via clamscan's '-d' flag - .pwdb files will now get loaded if they exist in a directory passed via clamscan's '-d' flag even when compiling without yara support - Changes to .imp, .ign, and .ign2 files will now be reflected in calls to cl_statinidir and cl_statchkdir (and also .pwdb files, even when compiling without yara support) - The contents of .sfp files won't be included in some of the signature counts, and the contents of .cud files will be - Any local.gdb files will no longer be loaded twice - For .imp files, you are no longer required to specify a minimum flevel for wildcard rules, since this isn't needed
2019-01-08 00:09:08 -05:00
}
switch (offdata[0]) {
case CLI_OFF_EOF_MINUS:
*offset_min = info->fsize - offdata[1];
break;
case CLI_OFF_EP_PLUS:
*offset_min = info->exeinfo.ep + offdata[1];
break;
case CLI_OFF_EP_MINUS:
*offset_min = info->exeinfo.ep - offdata[1];
break;
case CLI_OFF_SL_PLUS:
PE parsing code improvements, db loading bug fixes Consolidate the PE parsing code into one function. I tried to preserve all existing functionality from the previous, distinct implementations to a large extent (with the exceptions mentioned below). If I noticed potential bugs/improvements, I added a TODO statement about those so that they can be fixed in a smaller commit later. Also, there are more TODOs in places where I'm not entirely sure why certain actions are performed - more research is needed for these. I'm submitting a pull request now so that regression testing can be done, and because merging what I have thus far now will likely have fewer conflicts than if I try to merge later PE parsing code improvements: - PEs without all 16 data directories are parsed more appropriately now - Added lots more debug statements Also: - Allow MAX_BC and MAX_TRACKED_PCRE to be specified via CFLAGS When doing performance testing with the latest CVD, MAX_BC and MAX_TRACKED_PCRE need to be raised to track all the events. Allow these to be specified via CFLAGS by not redefining them if they are already defined - Fix an issue preventing wildcard sizes in .MDB/.MSB rules I'm not sure what the original intent of the check I removed was, but it prevents using wildcard sizes in .MDB/.MSB rules. AFAICT these wildcard sizes should be handled appropriately by the MD5 section hash computation code, so I don't think a check on that is needed. - Fix several issues related to db loading - .imp files will now get loaded if they exist in a directory passed via clamscan's '-d' flag - .pwdb files will now get loaded if they exist in a directory passed via clamscan's '-d' flag even when compiling without yara support - Changes to .imp, .ign, and .ign2 files will now be reflected in calls to cl_statinidir and cl_statchkdir (and also .pwdb files, even when compiling without yara support) - The contents of .sfp files won't be included in some of the signature counts, and the contents of .cud files will be - Any local.gdb files will no longer be loaded twice - For .imp files, you are no longer required to specify a minimum flevel for wildcard rules, since this isn't needed
2019-01-08 00:09:08 -05:00
*offset_min = info->exeinfo.sections[info->exeinfo.nsections - 1].raw + offdata[1];
break;
case CLI_OFF_SX_PLUS:
if (offdata[3] >= info->exeinfo.nsections)
*offset_min = CLI_OFF_NONE;
else
PE parsing code improvements, db loading bug fixes Consolidate the PE parsing code into one function. I tried to preserve all existing functionality from the previous, distinct implementations to a large extent (with the exceptions mentioned below). If I noticed potential bugs/improvements, I added a TODO statement about those so that they can be fixed in a smaller commit later. Also, there are more TODOs in places where I'm not entirely sure why certain actions are performed - more research is needed for these. I'm submitting a pull request now so that regression testing can be done, and because merging what I have thus far now will likely have fewer conflicts than if I try to merge later PE parsing code improvements: - PEs without all 16 data directories are parsed more appropriately now - Added lots more debug statements Also: - Allow MAX_BC and MAX_TRACKED_PCRE to be specified via CFLAGS When doing performance testing with the latest CVD, MAX_BC and MAX_TRACKED_PCRE need to be raised to track all the events. Allow these to be specified via CFLAGS by not redefining them if they are already defined - Fix an issue preventing wildcard sizes in .MDB/.MSB rules I'm not sure what the original intent of the check I removed was, but it prevents using wildcard sizes in .MDB/.MSB rules. AFAICT these wildcard sizes should be handled appropriately by the MD5 section hash computation code, so I don't think a check on that is needed. - Fix several issues related to db loading - .imp files will now get loaded if they exist in a directory passed via clamscan's '-d' flag - .pwdb files will now get loaded if they exist in a directory passed via clamscan's '-d' flag even when compiling without yara support - Changes to .imp, .ign, and .ign2 files will now be reflected in calls to cl_statinidir and cl_statchkdir (and also .pwdb files, even when compiling without yara support) - The contents of .sfp files won't be included in some of the signature counts, and the contents of .cud files will be - Any local.gdb files will no longer be loaded twice - For .imp files, you are no longer required to specify a minimum flevel for wildcard rules, since this isn't needed
2019-01-08 00:09:08 -05:00
*offset_min = info->exeinfo.sections[offdata[3]].raw + offdata[1];
break;
case CLI_OFF_SE:
if (offdata[3] >= info->exeinfo.nsections) {
*offset_min = CLI_OFF_NONE;
} else {
PE parsing code improvements, db loading bug fixes Consolidate the PE parsing code into one function. I tried to preserve all existing functionality from the previous, distinct implementations to a large extent (with the exceptions mentioned below). If I noticed potential bugs/improvements, I added a TODO statement about those so that they can be fixed in a smaller commit later. Also, there are more TODOs in places where I'm not entirely sure why certain actions are performed - more research is needed for these. I'm submitting a pull request now so that regression testing can be done, and because merging what I have thus far now will likely have fewer conflicts than if I try to merge later PE parsing code improvements: - PEs without all 16 data directories are parsed more appropriately now - Added lots more debug statements Also: - Allow MAX_BC and MAX_TRACKED_PCRE to be specified via CFLAGS When doing performance testing with the latest CVD, MAX_BC and MAX_TRACKED_PCRE need to be raised to track all the events. Allow these to be specified via CFLAGS by not redefining them if they are already defined - Fix an issue preventing wildcard sizes in .MDB/.MSB rules I'm not sure what the original intent of the check I removed was, but it prevents using wildcard sizes in .MDB/.MSB rules. AFAICT these wildcard sizes should be handled appropriately by the MD5 section hash computation code, so I don't think a check on that is needed. - Fix several issues related to db loading - .imp files will now get loaded if they exist in a directory passed via clamscan's '-d' flag - .pwdb files will now get loaded if they exist in a directory passed via clamscan's '-d' flag even when compiling without yara support - Changes to .imp, .ign, and .ign2 files will now be reflected in calls to cl_statinidir and cl_statchkdir (and also .pwdb files, even when compiling without yara support) - The contents of .sfp files won't be included in some of the signature counts, and the contents of .cud files will be - Any local.gdb files will no longer be loaded twice - For .imp files, you are no longer required to specify a minimum flevel for wildcard rules, since this isn't needed
2019-01-08 00:09:08 -05:00
*offset_min = info->exeinfo.sections[offdata[3]].raw;
if (offset_max)
PE parsing code improvements, db loading bug fixes Consolidate the PE parsing code into one function. I tried to preserve all existing functionality from the previous, distinct implementations to a large extent (with the exceptions mentioned below). If I noticed potential bugs/improvements, I added a TODO statement about those so that they can be fixed in a smaller commit later. Also, there are more TODOs in places where I'm not entirely sure why certain actions are performed - more research is needed for these. I'm submitting a pull request now so that regression testing can be done, and because merging what I have thus far now will likely have fewer conflicts than if I try to merge later PE parsing code improvements: - PEs without all 16 data directories are parsed more appropriately now - Added lots more debug statements Also: - Allow MAX_BC and MAX_TRACKED_PCRE to be specified via CFLAGS When doing performance testing with the latest CVD, MAX_BC and MAX_TRACKED_PCRE need to be raised to track all the events. Allow these to be specified via CFLAGS by not redefining them if they are already defined - Fix an issue preventing wildcard sizes in .MDB/.MSB rules I'm not sure what the original intent of the check I removed was, but it prevents using wildcard sizes in .MDB/.MSB rules. AFAICT these wildcard sizes should be handled appropriately by the MD5 section hash computation code, so I don't think a check on that is needed. - Fix several issues related to db loading - .imp files will now get loaded if they exist in a directory passed via clamscan's '-d' flag - .pwdb files will now get loaded if they exist in a directory passed via clamscan's '-d' flag even when compiling without yara support - Changes to .imp, .ign, and .ign2 files will now be reflected in calls to cl_statinidir and cl_statchkdir (and also .pwdb files, even when compiling without yara support) - The contents of .sfp files won't be included in some of the signature counts, and the contents of .cud files will be - Any local.gdb files will no longer be loaded twice - For .imp files, you are no longer required to specify a minimum flevel for wildcard rules, since this isn't needed
2019-01-08 00:09:08 -05:00
*offset_max = *offset_min + info->exeinfo.sections[offdata[3]].rsz + offdata[2];
// TODO offdata[2] == MaxShift. Won't this make offset_max
// extend beyond the end of the section? This doesn't seem like
// what we want...
}
break;
case CLI_OFF_VERSION:
if (offset_max)
*offset_min = *offset_max = CLI_OFF_ANY;
break;
default:
cli_errmsg("cli_caloff: Not a relative offset (type: %u)\n", offdata[0]);
return CL_EARG;
}
if (offset_max && *offset_max == CLI_OFF_NONE && *offset_min != CLI_OFF_NONE)
*offset_max = *offset_min + offdata[2];
}
return CL_SUCCESS;
}
PE parsing code improvements, db loading bug fixes Consolidate the PE parsing code into one function. I tried to preserve all existing functionality from the previous, distinct implementations to a large extent (with the exceptions mentioned below). If I noticed potential bugs/improvements, I added a TODO statement about those so that they can be fixed in a smaller commit later. Also, there are more TODOs in places where I'm not entirely sure why certain actions are performed - more research is needed for these. I'm submitting a pull request now so that regression testing can be done, and because merging what I have thus far now will likely have fewer conflicts than if I try to merge later PE parsing code improvements: - PEs without all 16 data directories are parsed more appropriately now - Added lots more debug statements Also: - Allow MAX_BC and MAX_TRACKED_PCRE to be specified via CFLAGS When doing performance testing with the latest CVD, MAX_BC and MAX_TRACKED_PCRE need to be raised to track all the events. Allow these to be specified via CFLAGS by not redefining them if they are already defined - Fix an issue preventing wildcard sizes in .MDB/.MSB rules I'm not sure what the original intent of the check I removed was, but it prevents using wildcard sizes in .MDB/.MSB rules. AFAICT these wildcard sizes should be handled appropriately by the MD5 section hash computation code, so I don't think a check on that is needed. - Fix several issues related to db loading - .imp files will now get loaded if they exist in a directory passed via clamscan's '-d' flag - .pwdb files will now get loaded if they exist in a directory passed via clamscan's '-d' flag even when compiling without yara support - Changes to .imp, .ign, and .ign2 files will now be reflected in calls to cl_statinidir and cl_statchkdir (and also .pwdb files, even when compiling without yara support) - The contents of .sfp files won't be included in some of the signature counts, and the contents of .cud files will be - Any local.gdb files will no longer be loaded twice - For .imp files, you are no longer required to specify a minimum flevel for wildcard rules, since this isn't needed
2019-01-08 00:09:08 -05:00
void cli_targetinfo_init(struct cli_target_info *info)
{
if (NULL == info) {
return;
}
info->status = 0;
cli_exe_info_init(&(info->exeinfo), 0);
PE parsing code improvements, db loading bug fixes Consolidate the PE parsing code into one function. I tried to preserve all existing functionality from the previous, distinct implementations to a large extent (with the exceptions mentioned below). If I noticed potential bugs/improvements, I added a TODO statement about those so that they can be fixed in a smaller commit later. Also, there are more TODOs in places where I'm not entirely sure why certain actions are performed - more research is needed for these. I'm submitting a pull request now so that regression testing can be done, and because merging what I have thus far now will likely have fewer conflicts than if I try to merge later PE parsing code improvements: - PEs without all 16 data directories are parsed more appropriately now - Added lots more debug statements Also: - Allow MAX_BC and MAX_TRACKED_PCRE to be specified via CFLAGS When doing performance testing with the latest CVD, MAX_BC and MAX_TRACKED_PCRE need to be raised to track all the events. Allow these to be specified via CFLAGS by not redefining them if they are already defined - Fix an issue preventing wildcard sizes in .MDB/.MSB rules I'm not sure what the original intent of the check I removed was, but it prevents using wildcard sizes in .MDB/.MSB rules. AFAICT these wildcard sizes should be handled appropriately by the MD5 section hash computation code, so I don't think a check on that is needed. - Fix several issues related to db loading - .imp files will now get loaded if they exist in a directory passed via clamscan's '-d' flag - .pwdb files will now get loaded if they exist in a directory passed via clamscan's '-d' flag even when compiling without yara support - Changes to .imp, .ign, and .ign2 files will now be reflected in calls to cl_statinidir and cl_statchkdir (and also .pwdb files, even when compiling without yara support) - The contents of .sfp files won't be included in some of the signature counts, and the contents of .cud files will be - Any local.gdb files will no longer be loaded twice - For .imp files, you are no longer required to specify a minimum flevel for wildcard rules, since this isn't needed
2019-01-08 00:09:08 -05:00
}
void cli_targetinfo(struct cli_target_info *info, cli_target_t target, cli_ctx *ctx)
{
cl_error_t (*einfo)(cli_ctx *, struct cli_exe_info *) = NULL;
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
info->fsize = ctx->fmap->len;
switch (target) {
case TARGET_PE:
einfo = cli_pe_targetinfo;
break;
case TARGET_ELF:
einfo = cli_elfheader;
break;
case TARGET_MACHO:
einfo = cli_machoheader;
break;
default:
return;
}
if (CL_SUCCESS != einfo(ctx, &info->exeinfo))
info->status = -1;
else
info->status = 1;
}
PE parsing code improvements, db loading bug fixes Consolidate the PE parsing code into one function. I tried to preserve all existing functionality from the previous, distinct implementations to a large extent (with the exceptions mentioned below). If I noticed potential bugs/improvements, I added a TODO statement about those so that they can be fixed in a smaller commit later. Also, there are more TODOs in places where I'm not entirely sure why certain actions are performed - more research is needed for these. I'm submitting a pull request now so that regression testing can be done, and because merging what I have thus far now will likely have fewer conflicts than if I try to merge later PE parsing code improvements: - PEs without all 16 data directories are parsed more appropriately now - Added lots more debug statements Also: - Allow MAX_BC and MAX_TRACKED_PCRE to be specified via CFLAGS When doing performance testing with the latest CVD, MAX_BC and MAX_TRACKED_PCRE need to be raised to track all the events. Allow these to be specified via CFLAGS by not redefining them if they are already defined - Fix an issue preventing wildcard sizes in .MDB/.MSB rules I'm not sure what the original intent of the check I removed was, but it prevents using wildcard sizes in .MDB/.MSB rules. AFAICT these wildcard sizes should be handled appropriately by the MD5 section hash computation code, so I don't think a check on that is needed. - Fix several issues related to db loading - .imp files will now get loaded if they exist in a directory passed via clamscan's '-d' flag - .pwdb files will now get loaded if they exist in a directory passed via clamscan's '-d' flag even when compiling without yara support - Changes to .imp, .ign, and .ign2 files will now be reflected in calls to cl_statinidir and cl_statchkdir (and also .pwdb files, even when compiling without yara support) - The contents of .sfp files won't be included in some of the signature counts, and the contents of .cud files will be - Any local.gdb files will no longer be loaded twice - For .imp files, you are no longer required to specify a minimum flevel for wildcard rules, since this isn't needed
2019-01-08 00:09:08 -05:00
void cli_targetinfo_destroy(struct cli_target_info *info)
{
if (NULL == info) {
PE parsing code improvements, db loading bug fixes Consolidate the PE parsing code into one function. I tried to preserve all existing functionality from the previous, distinct implementations to a large extent (with the exceptions mentioned below). If I noticed potential bugs/improvements, I added a TODO statement about those so that they can be fixed in a smaller commit later. Also, there are more TODOs in places where I'm not entirely sure why certain actions are performed - more research is needed for these. I'm submitting a pull request now so that regression testing can be done, and because merging what I have thus far now will likely have fewer conflicts than if I try to merge later PE parsing code improvements: - PEs without all 16 data directories are parsed more appropriately now - Added lots more debug statements Also: - Allow MAX_BC and MAX_TRACKED_PCRE to be specified via CFLAGS When doing performance testing with the latest CVD, MAX_BC and MAX_TRACKED_PCRE need to be raised to track all the events. Allow these to be specified via CFLAGS by not redefining them if they are already defined - Fix an issue preventing wildcard sizes in .MDB/.MSB rules I'm not sure what the original intent of the check I removed was, but it prevents using wildcard sizes in .MDB/.MSB rules. AFAICT these wildcard sizes should be handled appropriately by the MD5 section hash computation code, so I don't think a check on that is needed. - Fix several issues related to db loading - .imp files will now get loaded if they exist in a directory passed via clamscan's '-d' flag - .pwdb files will now get loaded if they exist in a directory passed via clamscan's '-d' flag even when compiling without yara support - Changes to .imp, .ign, and .ign2 files will now be reflected in calls to cl_statinidir and cl_statchkdir (and also .pwdb files, even when compiling without yara support) - The contents of .sfp files won't be included in some of the signature counts, and the contents of .cud files will be - Any local.gdb files will no longer be loaded twice - For .imp files, you are no longer required to specify a minimum flevel for wildcard rules, since this isn't needed
2019-01-08 00:09:08 -05:00
return;
}
cli_exe_info_destroy(&(info->exeinfo));
info->status = 0;
PE parsing code improvements, db loading bug fixes Consolidate the PE parsing code into one function. I tried to preserve all existing functionality from the previous, distinct implementations to a large extent (with the exceptions mentioned below). If I noticed potential bugs/improvements, I added a TODO statement about those so that they can be fixed in a smaller commit later. Also, there are more TODOs in places where I'm not entirely sure why certain actions are performed - more research is needed for these. I'm submitting a pull request now so that regression testing can be done, and because merging what I have thus far now will likely have fewer conflicts than if I try to merge later PE parsing code improvements: - PEs without all 16 data directories are parsed more appropriately now - Added lots more debug statements Also: - Allow MAX_BC and MAX_TRACKED_PCRE to be specified via CFLAGS When doing performance testing with the latest CVD, MAX_BC and MAX_TRACKED_PCRE need to be raised to track all the events. Allow these to be specified via CFLAGS by not redefining them if they are already defined - Fix an issue preventing wildcard sizes in .MDB/.MSB rules I'm not sure what the original intent of the check I removed was, but it prevents using wildcard sizes in .MDB/.MSB rules. AFAICT these wildcard sizes should be handled appropriately by the MD5 section hash computation code, so I don't think a check on that is needed. - Fix several issues related to db loading - .imp files will now get loaded if they exist in a directory passed via clamscan's '-d' flag - .pwdb files will now get loaded if they exist in a directory passed via clamscan's '-d' flag even when compiling without yara support - Changes to .imp, .ign, and .ign2 files will now be reflected in calls to cl_statinidir and cl_statchkdir (and also .pwdb files, even when compiling without yara support) - The contents of .sfp files won't be included in some of the signature counts, and the contents of .cud files will be - Any local.gdb files will no longer be loaded twice - For .imp files, you are no longer required to specify a minimum flevel for wildcard rules, since this isn't needed
2019-01-08 00:09:08 -05:00
}
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
cl_error_t cli_check_fp(cli_ctx *ctx, const char *vname)
{
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
cl_error_t status = CL_VIRUS;
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
cl_error_t ret;
size_t i;
const char *virname = NULL;
2014-01-20 15:36:11 -05:00
fmap_t *map;
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
int32_t stack_index;
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
uint8_t *hash;
cli_hash_type_t hash_type;
char hash_string[SHA256_HASH_SIZE * 2 + 1];
bool need_hash[CLI_HASH_AVAIL_TYPES] = {false};
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
stack_index = (int32_t)ctx->recursion_level;
char *source = NULL;
size_t source_len;
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
while (stack_index >= 0) {
map = ctx->recursion_stack[stack_index].fmap;
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
need_hash[CLI_HASH_MD5] = cli_hm_have_size(ctx->engine->hm_fp, CLI_HASH_MD5, map->len) ||
cli_hm_have_wild(ctx->engine->hm_fp, CLI_HASH_MD5);
2014-01-20 15:36:11 -05:00
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
need_hash[CLI_HASH_SHA1] = cli_hm_have_size(ctx->engine->hm_fp, CLI_HASH_SHA1, map->len) ||
cli_hm_have_wild(ctx->engine->hm_fp, CLI_HASH_SHA1) ||
cli_hm_have_size(ctx->engine->hm_fp, CLI_HASH_SHA1, 1);
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
need_hash[CLI_HASH_SHA2_256] = cli_hm_have_size(ctx->engine->hm_fp, CLI_HASH_SHA2_256, map->len) ||
cli_hm_have_wild(ctx->engine->hm_fp, CLI_HASH_SHA2_256) ||
Reduce unnecessary scanning of embedded file FPs (#1571) When embedded file type recognition finds a possible embedded file, it is being scanned as a new embedded file even if it turns out it was a false positive and parsing fails. My solution is to pre-parse the file headers as little possible to determine if it is valid. If possible, also determine the file size based on the headers. That will make it so we don't have to scan additional data when the embedded file is not at the very end. This commit adds header checks prior to embedded ZIP, ARJ, and CAB scanning. For these types I was also able to use the header checks to determine the object size so as to prevent excessive pattern matching. TODO: Add the same for RAR, EGG, 7Z, NULSFT, AUTOIT, IShield, and PDF. This commit also removes duplicate matching for embedded MSEXE. The embedded MSEXE detection and scanning logic was accidentally creating an extra duplicate layer in between scanning and detection because of the logic within the `cli_scanembpe()` function. That function was effectively doing the header check which this commit adds for ZIP, ARJ, and CAB but minus the size check. Note: It is unfortunately not possible to get an accurage size from PE file headers. The `cli_scanembpe()` function also used to dump to a temp file for no reason since FMAPs were extended to support windows into other FMAPs. So this commit removes the intermediate layer as well as dropping a temp file for each embedded PE file. Further, this commit adds configuration and DCONF safeguards around all embedded file type scanning. Finally, this commit adds a set of tests to validate proper extraction of embedded ZIP, ARJ, CAB, and MSEXE files. CLAM-2862 Co-authored-by: TheRaynMan <draynor@sourcefire.com>
2025-09-23 15:57:28 -04:00
cli_hm_have_size(ctx->engine->hm_fp, CLI_HASH_SHA2_256, 1) ||
// If debug logging is enabled, we want to calculate SHA256 hashes for all layers.
// Some users rely on the debug log output to create new FP signatures.
cli_debug_flag;
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
/* Set fmap to need hash later if required.
* This is an optimization so we can calculate all needed hashes in one pass. */
for (hash_type = CLI_HASH_MD5; hash_type < CLI_HASH_AVAIL_TYPES; hash_type++) {
if (need_hash[hash_type]) {
ret = fmap_will_need_hash_later(map, hash_type);
if (CL_SUCCESS != ret) {
cli_dbgmsg("cli_check_fp: Failed to set fmap to need MD5 hash later\n");
status = CL_VIRUS;
goto done;
}
2014-01-20 15:36:11 -05:00
}
}
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
for (hash_type = CLI_HASH_MD5; hash_type < CLI_HASH_AVAIL_TYPES; hash_type++) {
if (need_hash[hash_type]) {
size_t hash_len = cli_hash_len(hash_type);
2014-01-20 15:36:11 -05:00
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
/* If we need a hash, we will calculate it now */
ret = fmap_get_hash(map, &hash, hash_type);
if (CL_SUCCESS != ret) {
cli_dbgmsg("cli_check_fp: Failed to get hash for the map at stack index # %u\n", stack_index);
status = CL_VIRUS;
goto done;
}
Reduce unnecessary scanning of embedded file FPs (#1571) When embedded file type recognition finds a possible embedded file, it is being scanned as a new embedded file even if it turns out it was a false positive and parsing fails. My solution is to pre-parse the file headers as little possible to determine if it is valid. If possible, also determine the file size based on the headers. That will make it so we don't have to scan additional data when the embedded file is not at the very end. This commit adds header checks prior to embedded ZIP, ARJ, and CAB scanning. For these types I was also able to use the header checks to determine the object size so as to prevent excessive pattern matching. TODO: Add the same for RAR, EGG, 7Z, NULSFT, AUTOIT, IShield, and PDF. This commit also removes duplicate matching for embedded MSEXE. The embedded MSEXE detection and scanning logic was accidentally creating an extra duplicate layer in between scanning and detection because of the logic within the `cli_scanembpe()` function. That function was effectively doing the header check which this commit adds for ZIP, ARJ, and CAB but minus the size check. Note: It is unfortunately not possible to get an accurage size from PE file headers. The `cli_scanembpe()` function also used to dump to a temp file for no reason since FMAPs were extended to support windows into other FMAPs. So this commit removes the intermediate layer as well as dropping a temp file for each embedded PE file. Further, this commit adds configuration and DCONF safeguards around all embedded file type scanning. Finally, this commit adds a set of tests to validate proper extraction of embedded ZIP, ARJ, CAB, and MSEXE files. CLAM-2862 Co-authored-by: TheRaynMan <draynor@sourcefire.com>
2025-09-23 15:57:28 -04:00
if (cli_debug_flag ||
((CLI_HASH_MD5 == hash_type) && (ctx->engine->cb_hash))) {
/* Convert hash to string */
for (i = 0; i < hash_len; i++) {
sprintf(hash_string + i * 2, "%02x", hash[i]);
}
hash_string[hash_len * 2] = 0;
2014-01-20 15:36:11 -05:00
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
const char *name = ctx->recursion_stack[stack_index].fmap->name;
const char *type = cli_ftname(ctx->recursion_stack[stack_index].type);
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
cli_dbgmsg("FP SIGNATURE: %s:%u:%s # Name: %s, Type: %s\n",
hash_string, (unsigned int)map->len, vname ? vname : "Name", name ? name : "n/a", type);
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
}
2014-01-20 15:36:11 -05:00
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
if (CLI_HASH_MD5 == hash_type) {
/* Run legacy callbacks that include MD5 hash */
if (ctx->engine->cb_hash) {
ctx->engine->cb_hash(fmap_fd(ctx->fmap), map->len, hash_string, vname ? vname : "noname", ctx->cb_ctx);
}
if (ctx->engine->cb_stats_add_sample) {
stats_section_t sections;
memset(&sections, 0x00, sizeof(stats_section_t));
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
if (!(ctx->engine->engine_options & ENGINE_OPTIONS_DISABLE_PE_STATS) &&
!(ctx->engine->dconf->stats & (DCONF_STATS_DISABLED | DCONF_STATS_PE_SECTION_DISABLED))) {
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
cli_genhash_pe(ctx, CL_GENHASH_PE_CLASS_SECTION, 1, &sections);
}
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
// TODO We probably only want to call cb_stats_add_sample when
// sections.section != NULL... leaving as is for now
ctx->engine->cb_stats_add_sample(vname ? vname : "noname", hash, map->len, &sections, ctx->engine->stats_data);
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
if (sections.sections) {
free(sections.sections);
}
}
}
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
if (cli_hm_scan(hash, map->len, &virname, ctx->engine->hm_fp, hash_type) == CL_VIRUS) {
cli_dbgmsg("cli_check_fp: Found false positive detection for %s (fp sig: %s)\n", cli_hash_name(hash_type), virname);
source_len = strlen(virname) + strlen("false positive signature match: ") + 1;
source = malloc(source_len);
if (source) {
snprintf(source, source_len, "false positive signature match: %s", virname);
}
// Remove any evidence and set the verdict to trusted for the layer where the FP hash matched, and for all contained layers.
(void)cli_trust_layers(ctx, (uint32_t)stack_index, ctx->recursion_level, source);
free(source);
source = NULL;
status = CL_VERIFIED;
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
goto done;
}
if (cli_hm_scan_wild(hash, &virname, ctx->engine->hm_fp, hash_type) == CL_VIRUS) {
cli_dbgmsg("cli_check_fp: Found false positive detection for %s (fp sig: %s)\n", cli_hash_name(hash_type), virname);
source_len = strlen(virname) + strlen("false positive signature match: ") + 1;
source = malloc(source_len);
if (source) {
snprintf(source, source_len, "false positive signature match: %s", virname);
}
// Remove any evidence and set the verdict to trusted for the layer where the FP hash matched, and for all contained layers.
(void)cli_trust_layers(ctx, (uint32_t)stack_index, ctx->recursion_level, source);
free(source);
source = NULL;
status = CL_VERIFIED;
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
goto done;
}
if (CLI_HASH_MD5 != hash_type) {
/* See whether the hash matches those loaded in from .cat files
* (associated with the .CAB file type) */
if (cli_hm_scan(hash, 1, &virname, ctx->engine->hm_fp, hash_type) == CL_VIRUS) {
cli_dbgmsg("cli_check_fp: Found .CAB false positive detection for %s via catalog file\n", cli_hash_name(hash_type));
source_len = strlen(virname) + strlen("false positive signature match: ") + 1;
source = malloc(source_len);
if (source) {
snprintf(source, source_len, "false positive signature match: %s", virname);
}
// Remove any evidence and set the verdict to trusted for the layer where the FP hash matched, and for all contained layers.
(void)cli_trust_layers(ctx, (uint32_t)stack_index, ctx->recursion_level, source);
free(source);
source = NULL;
status = CL_VERIFIED;
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
goto done;
}
}
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
}
}
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
stack_index -= 1;
}
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
done:
if (NULL != source) {
free(source);
}
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
return status;
}
static cl_error_t matchicon(cli_ctx *ctx, struct cli_exe_info *exeinfo, const char *grp1, const char *grp2)
{
2014-03-05 12:47:41 -05:00
icon_groupset iconset;
if (!ctx ||
!ctx->engine ||
!ctx->engine->iconcheck ||
!ctx->engine->iconcheck->group_counts[0] ||
!ctx->engine->iconcheck->group_counts[1] ||
!exeinfo->res_addr) return CL_CLEAN;
2014-03-05 12:47:41 -05:00
if (!(ctx->dconf->pe & PE_CONF_MATCHICON))
return CL_CLEAN;
cli_icongroupset_init(&iconset);
cli_icongroupset_add(grp1 ? grp1 : "*", &iconset, 0, ctx);
cli_icongroupset_add(grp2 ? grp2 : "*", &iconset, 1, ctx);
PE parsing code improvements, db loading bug fixes Consolidate the PE parsing code into one function. I tried to preserve all existing functionality from the previous, distinct implementations to a large extent (with the exceptions mentioned below). If I noticed potential bugs/improvements, I added a TODO statement about those so that they can be fixed in a smaller commit later. Also, there are more TODOs in places where I'm not entirely sure why certain actions are performed - more research is needed for these. I'm submitting a pull request now so that regression testing can be done, and because merging what I have thus far now will likely have fewer conflicts than if I try to merge later PE parsing code improvements: - PEs without all 16 data directories are parsed more appropriately now - Added lots more debug statements Also: - Allow MAX_BC and MAX_TRACKED_PCRE to be specified via CFLAGS When doing performance testing with the latest CVD, MAX_BC and MAX_TRACKED_PCRE need to be raised to track all the events. Allow these to be specified via CFLAGS by not redefining them if they are already defined - Fix an issue preventing wildcard sizes in .MDB/.MSB rules I'm not sure what the original intent of the check I removed was, but it prevents using wildcard sizes in .MDB/.MSB rules. AFAICT these wildcard sizes should be handled appropriately by the MD5 section hash computation code, so I don't think a check on that is needed. - Fix several issues related to db loading - .imp files will now get loaded if they exist in a directory passed via clamscan's '-d' flag - .pwdb files will now get loaded if they exist in a directory passed via clamscan's '-d' flag even when compiling without yara support - Changes to .imp, .ign, and .ign2 files will now be reflected in calls to cl_statinidir and cl_statchkdir (and also .pwdb files, even when compiling without yara support) - The contents of .sfp files won't be included in some of the signature counts, and the contents of .cud files will be - Any local.gdb files will no longer be loaded twice - For .imp files, you are no longer required to specify a minimum flevel for wildcard rules, since this isn't needed
2019-01-08 00:09:08 -05:00
return cli_scanicon(&iconset, ctx, exeinfo);
}
int32_t cli_bcapi_matchicon(struct cli_bc_ctx *ctx, const uint8_t *grp1, int32_t grp1len,
const uint8_t *grp2, int32_t grp2len)
2010-08-02 17:04:35 +03:00
{
cl_error_t ret;
2010-08-02 17:04:35 +03:00
char group1[128], group2[128];
struct cli_exe_info info;
// TODO This isn't a good check, since EP will be zero for DLLs and
// (assuming pedata->ep is populated from exeinfo->pe) non-zero for
// some MachO and ELF executables
if (!ctx->hooks.pedata->ep) {
cli_dbgmsg("bytecode: matchicon only works with PE files\n");
return -1;
2010-08-02 17:04:35 +03:00
}
if ((size_t)grp1len > sizeof(group1) - 1 ||
(size_t)grp2len > sizeof(group2) - 1)
return -1;
2010-08-02 17:04:35 +03:00
memcpy(group1, grp1, grp1len);
memcpy(group2, grp2, grp2len);
group1[grp1len] = 0;
group2[grp2len] = 0;
memset(&info, 0, sizeof(info));
2010-09-02 18:04:00 +03:00
if (ctx->bc->kind == BC_PE_UNPACKER || ctx->bc->kind == BC_PE_ALL) {
if (le16_to_host(ctx->hooks.pedata->file_hdr.Characteristics) & 0x2000 ||
!ctx->hooks.pedata->dirs[2].Size)
info.res_addr = 0;
else
info.res_addr = ctx->hooks.pedata->dirs[2].VirtualAddress;
} else
info.res_addr = ctx->resaddr; /* from target_info */
PE parsing code improvements, db loading bug fixes Consolidate the PE parsing code into one function. I tried to preserve all existing functionality from the previous, distinct implementations to a large extent (with the exceptions mentioned below). If I noticed potential bugs/improvements, I added a TODO statement about those so that they can be fixed in a smaller commit later. Also, there are more TODOs in places where I'm not entirely sure why certain actions are performed - more research is needed for these. I'm submitting a pull request now so that regression testing can be done, and because merging what I have thus far now will likely have fewer conflicts than if I try to merge later PE parsing code improvements: - PEs without all 16 data directories are parsed more appropriately now - Added lots more debug statements Also: - Allow MAX_BC and MAX_TRACKED_PCRE to be specified via CFLAGS When doing performance testing with the latest CVD, MAX_BC and MAX_TRACKED_PCRE need to be raised to track all the events. Allow these to be specified via CFLAGS by not redefining them if they are already defined - Fix an issue preventing wildcard sizes in .MDB/.MSB rules I'm not sure what the original intent of the check I removed was, but it prevents using wildcard sizes in .MDB/.MSB rules. AFAICT these wildcard sizes should be handled appropriately by the MD5 section hash computation code, so I don't think a check on that is needed. - Fix several issues related to db loading - .imp files will now get loaded if they exist in a directory passed via clamscan's '-d' flag - .pwdb files will now get loaded if they exist in a directory passed via clamscan's '-d' flag even when compiling without yara support - Changes to .imp, .ign, and .ign2 files will now be reflected in calls to cl_statinidir and cl_statchkdir (and also .pwdb files, even when compiling without yara support) - The contents of .sfp files won't be included in some of the signature counts, and the contents of .cud files will be - Any local.gdb files will no longer be loaded twice - For .imp files, you are no longer required to specify a minimum flevel for wildcard rules, since this isn't needed
2019-01-08 00:09:08 -05:00
info.sections = (struct cli_exe_section *)ctx->sections;
2010-08-02 17:04:35 +03:00
info.nsections = ctx->hooks.pedata->nsections;
info.hdr_size = ctx->hooks.pedata->hdr_size;
cli_dbgmsg("bytecode matchicon %s %s\n", group1, group2);
ret = matchicon(ctx->ctx, &info, group1[0] ? group1 : NULL,
group2[0] ? group2 : NULL);
return (int32_t)ret;
2010-08-02 17:04:35 +03:00
}
cl_error_t cli_scan_desc(int desc, cli_ctx *ctx, cli_file_t ftype, bool filetype_only, struct cli_matched_type **ftoffset, unsigned int acmode, struct cli_ac_result **acres, const char *name, const char *path, uint32_t attributes)
2003-07-29 15:48:06 +00:00
{
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
cl_error_t status = CL_CLEAN;
int empty;
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
fmap_t *new_map = NULL;
new_map = fmap_check_empty(desc, 0, 0, &empty, name, path);
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
if (NULL == new_map) {
if (!empty) {
cli_dbgmsg("cli_scan_desc: Failed to allocate new map for file descriptor scan.\n");
status = CL_EMEM;
}
goto done;
}
status = cli_recursion_stack_push(ctx, new_map, ftype, true, attributes); /* Perform scan with child fmap */
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
if (CL_SUCCESS != status) {
cli_dbgmsg("cli_scan_desc: Failed to scan fmap.\n");
goto done;
}
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
status = cli_scan_fmap(ctx, ftype, filetype_only, ftoffset, acmode, acres);
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
(void)cli_recursion_stack_pop(ctx); /* Restore the parent fmap */
done:
if (NULL != new_map) {
fmap_free(new_map);
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
}
return status;
}
2009-08-31 04:41:06 +02:00
static int intermediates_eval(cli_ctx *ctx, struct cli_ac_lsig *ac_lsig)
{
uint32_t i, icnt = ac_lsig->tdb.intermediates[0];
// -1 is the deepest layer (the current layer), so we start at -2, which is the first ancestor
int32_t j = -2;
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
if (ctx->recursion_level < icnt)
return 0;
for (i = icnt; i > 0; i--) {
if (ac_lsig->tdb.intermediates[i] == CL_TYPE_ANY)
continue;
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
if (ac_lsig->tdb.intermediates[i] != cli_recursion_stack_get_type(ctx, j--))
return 0;
}
return 1;
}
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
static cl_error_t lsig_eval(cli_ctx *ctx, struct cli_matcher *root, struct cli_ac_data *acdata, struct cli_target_info *target_info, uint32_t lsid)
{
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
cl_error_t status = CL_CLEAN;
unsigned evalcnt = 0;
uint64_t evalids = 0;
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
fmap_t *new_map = NULL;
struct cli_ac_lsig *ac_lsig = root->ac_lsigtable[lsid];
char *exp = ac_lsig->u.logic;
char *exp_end = exp + strlen(exp);
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
status = cli_ac_chkmacro(root, acdata, lsid);
if (status != CL_SUCCESS)
return status;
if (cli_ac_chklsig(exp, exp_end, acdata->lsigcnt[lsid], &evalcnt, &evalids, 0) != 1) {
// Logical expression did not match.
goto done;
}
// Logical expression matched.
// Need to check the other conditions, like target description block, icon group, bytecode, etc.
// If the lsig requires a specific container type, check if check that it matches
if (ac_lsig->tdb.container &&
ac_lsig->tdb.container[0] != cli_recursion_stack_get_type(ctx, -2)) {
// So far the match is good, but the container type doesn't match.
// Because this may need to match in a different scenario where the
// container does match, we do not want to cache this result.
ctx->fmap->dont_cache_flag = 1;
goto done;
}
// If the lsig has intermediates, check if they match the current recursion stack
if (ac_lsig->tdb.intermediates &&
!intermediates_eval(ctx, ac_lsig)) {
// So far the match is good, but the intermediates type(s) do not match.
// Because this may need to match in a different scenario where the
// intermediates do match, we do not want to cache this result.
ctx->fmap->dont_cache_flag = 1;
goto done;
}
// If the lsig has filesize requirements, check if they match
if (ac_lsig->tdb.filesize && (ac_lsig->tdb.filesize[0] > ctx->fmap->len || ac_lsig->tdb.filesize[1] < ctx->fmap->len)) {
goto done;
}
if (ac_lsig->tdb.ep || ac_lsig->tdb.nos) {
if (!target_info || target_info->status != 1)
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
goto done;
if (ac_lsig->tdb.ep && (ac_lsig->tdb.ep[0] > target_info->exeinfo.ep || ac_lsig->tdb.ep[1] < target_info->exeinfo.ep))
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
goto done;
if (ac_lsig->tdb.nos && (ac_lsig->tdb.nos[0] > target_info->exeinfo.nsections || ac_lsig->tdb.nos[1] < target_info->exeinfo.nsections))
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
goto done;
}
2010-07-29 16:58:27 +02:00
if (ac_lsig->tdb.handlertype) {
// This logical signature has a handler type, which means it's effectively a complex file type signature.
// Instead of alerting, we'll make a duplicate fmap (add recursion depth, to prevent infinite loops) and
// scan the file with the handler type.
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
/*
* If the current layer was re-typed already, then prevent HandlerType from being applied again.
*/
if (!(ctx->recursion_stack[ctx->recursion_level].attributes & LAYER_ATTRIBUTES_RETYPED)) {
/*
* Create an fmap window into our current fmap using the original offset & length, and rescan as the new type
*
* TODO: Unsure if creating an fmap is the right move, or if we should rescan with the current fmap as-is,
2023-11-26 15:01:19 -08:00
* since it's not really a container so much as it is type reassignment. This new fmap layer protect against
* a possible infinite loop by applying the scan recursion limit, but maybe there's a better way?
* Testing with both HandlerType type reassignment sigs + Container/Intermediates sigs should indicate if
* a change is needed.
*/
new_map = fmap_duplicate(ctx->fmap, 0, ctx->fmap->len, ctx->fmap->name);
if (NULL == new_map) {
status = CL_EMEM;
cli_dbgmsg("Failed to duplicate the current fmap for a re-scan as a different type.\n");
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
goto done;
}
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
status = cli_recursion_stack_push(ctx, new_map, ac_lsig->tdb.handlertype[0], true, LAYER_ATTRIBUTES_RETYPED); /* Perform scan with child fmap */
if (CL_SUCCESS != status) {
cli_dbgmsg("Failed to re-scan fmap as a new type.\n");
goto done;
}
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
status = cli_magic_scan(ctx, ac_lsig->tdb.handlertype[0]);
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
(void)cli_recursion_stack_pop(ctx); /* Restore the parent fmap */
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
goto done;
}
}
if (ac_lsig->tdb.icongrp1 || ac_lsig->tdb.icongrp2) {
// Logical sig depends on icon match. Check for the icon match.
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
if (!target_info || target_info->status != TARGET_PE) {
// Icon group feature only applies to PE files, so target description must match a PE file.
// This is a signature issue and should have been caught at load time, but just in case, we're checking again here.
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
goto done;
}
if (CL_VIRUS != matchicon(ctx, &target_info->exeinfo, ac_lsig->tdb.icongrp1, ac_lsig->tdb.icongrp2)) {
// No icon match!
goto done;
}
}
if (!ac_lsig->bc_idx) {
// Logical sig does not depend on bytecode match. Report the virus.
status = cli_append_virus(ctx, ac_lsig->virname);
if (status != CL_SUCCESS) {
goto done;
}
} else {
// Logical sig depends on bytecode match. Check for the bytecode match.
status = cli_bytecode_runlsig(ctx, target_info, &ctx->engine->bcs, ac_lsig->bc_idx, acdata->lsigcnt[lsid], acdata->lsigsuboff_first[lsid], ctx->fmap);
if (CL_SUCCESS != status) {
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
goto done;
}
// Check time limit here, because bytecode functions may take a while.
status = cli_checktimelimit(ctx);
if (CL_SUCCESS != status) {
goto done;
}
}
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
done:
if (NULL != new_map) {
free_duplicate_fmap(new_map);
}
return status;
}
#ifdef HAVE_YARA
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
static cl_error_t yara_eval(cli_ctx *ctx, struct cli_matcher *root, struct cli_ac_data *acdata, struct cli_target_info *target_info, uint32_t lsid)
{
struct cli_ac_lsig *ac_lsig = root->ac_lsigtable[lsid];
cl_error_t rc;
YR_SCAN_CONTEXT context;
memset(&context, 0, sizeof(YR_SCAN_CONTEXT));
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
context.fmap = ctx->fmap;
context.file_size = ctx->fmap->len;
if (target_info != NULL) {
if (target_info->status == 1)
context.entry_point = target_info->exeinfo.ep;
}
rc = yr_execute_code(ac_lsig, acdata, &context, 0, 0);
if (rc == CL_VIRUS) {
if (ac_lsig->flag & CLI_LSIG_FLAG_PRIVATE) {
rc = CL_CLEAN;
} else {
rc = cli_append_virus(ctx, ac_lsig->virname);
}
}
return rc;
}
#endif
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
cl_error_t cli_exp_eval(cli_ctx *ctx, struct cli_matcher *root, struct cli_ac_data *acdata, struct cli_target_info *target_info)
{
uint32_t i;
cl_error_t status = CL_SUCCESS;
for (i = 0; i < root->ac_lsigs; i++) {
if (root->ac_lsigtable[i]->type == CLI_LSIG_NORMAL) {
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
status = lsig_eval(ctx, root, acdata, target_info, i);
}
#ifdef HAVE_YARA
else if (root->ac_lsigtable[i]->type == CLI_YARA_NORMAL || root->ac_lsigtable[i]->type == CLI_YARA_OFFSET) {
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
status = yara_eval(ctx, root, acdata, target_info, i);
}
#endif
if (CL_SUCCESS != status) {
break;
}
if (i % 10 == 0) {
// Check the time limit every n'th lsig.
// In testing with a large signature set, we found n = 10 to be just as fast as 100 or
// 1000 and has a significant performance improvement over checking with every lsig.
status = cli_checktimelimit(ctx);
if (CL_SUCCESS != status) {
cli_dbgmsg("Exceeded scan time limit while evaluating logical and yara signatures (max: %u)\n", ctx->engine->maxscantime);
break;
}
}
}
return status;
}
2009-08-31 04:41:06 +02:00
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
cl_error_t cli_scan_fmap(cli_ctx *ctx, cli_file_t ftype, bool filetype_only, struct cli_matched_type **ftoffset, unsigned int acmode, struct cli_ac_result **acres)
2009-08-31 04:41:06 +02:00
{
2014-02-08 00:31:12 -05:00
const unsigned char *buff;
cl_error_t ret = CL_CLEAN, type = CL_CLEAN;
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
cli_hash_type_t hash_type;
bool need_hash[CLI_HASH_AVAIL_TYPES] = {false};
void *hashctx[CLI_HASH_AVAIL_TYPES] = {NULL};
unsigned char digest[CLI_HASH_AVAIL_TYPES][CLI_HASHLEN_MAX];
unsigned int i = 0, j = 0;
uint32_t maxpatlen, bytes, offset = 0;
struct cli_ac_data generic_ac_data;
bool gdata_initialized = false;
struct cli_ac_data target_ac_data;
bool tdata_initialized = false;
struct cli_bm_off bm_offsets_table;
bool bm_offsets_table_initialized = false;
struct cli_pcre_off generic_pcre_offsets_table;
bool generic_pcre_offsets_table_initialized = false;
struct cli_pcre_off target_pcre_offsets_table;
bool target_pcre_offsets_table_initialized = false;
struct cli_matcher *generic_ac_root = NULL, *target_ac_root = NULL;
2014-02-08 00:31:12 -05:00
struct cli_target_info info;
bool info_initialized = false;
2014-02-08 00:31:12 -05:00
struct cli_matcher *hdb, *fp;
if (!ctx->engine) {
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
cli_errmsg("cli_scan_fmap: engine == NULL\n");
ret = CL_ENULLARG;
goto done;
2014-02-08 00:31:12 -05:00
}
if (!filetype_only) {
generic_ac_root = ctx->engine->root[0]; /* generic signatures */
}
if (ftype != CL_TYPE_ANY) {
// Identify the target type, to find the matcher root for that target.
for (i = 1; i < CLI_MTARGETS; i++) {
for (j = 0; j < cli_mtargets[i].target_count; ++j) {
if (cli_mtargets[i].target[j] == ftype) {
// Identified the target type, now get the matcher root for that target.
target_ac_root = ctx->engine->root[i];
break; // Break out of inner loop
}
2014-02-08 00:31:12 -05:00
}
if (target_ac_root) break;
2014-02-08 00:31:12 -05:00
}
}
if (!generic_ac_root) {
if (!target_ac_root) {
// Don't have a matcher root for either generic signatures or target-specific signatures.
// Nothing to do!
ret = CL_CLEAN;
goto done;
}
// Only have a matcher root for target-specific signatures.
maxpatlen = target_ac_root->maxpatlen;
} else {
if (target_ac_root) {
// Have both generic and target-specific signatures.
maxpatlen = MAX(target_ac_root->maxpatlen, generic_ac_root->maxpatlen);
} else {
// Only have generic signatures.
maxpatlen = generic_ac_root->maxpatlen;
}
}
PE parsing code improvements, db loading bug fixes Consolidate the PE parsing code into one function. I tried to preserve all existing functionality from the previous, distinct implementations to a large extent (with the exceptions mentioned below). If I noticed potential bugs/improvements, I added a TODO statement about those so that they can be fixed in a smaller commit later. Also, there are more TODOs in places where I'm not entirely sure why certain actions are performed - more research is needed for these. I'm submitting a pull request now so that regression testing can be done, and because merging what I have thus far now will likely have fewer conflicts than if I try to merge later PE parsing code improvements: - PEs without all 16 data directories are parsed more appropriately now - Added lots more debug statements Also: - Allow MAX_BC and MAX_TRACKED_PCRE to be specified via CFLAGS When doing performance testing with the latest CVD, MAX_BC and MAX_TRACKED_PCRE need to be raised to track all the events. Allow these to be specified via CFLAGS by not redefining them if they are already defined - Fix an issue preventing wildcard sizes in .MDB/.MSB rules I'm not sure what the original intent of the check I removed was, but it prevents using wildcard sizes in .MDB/.MSB rules. AFAICT these wildcard sizes should be handled appropriately by the MD5 section hash computation code, so I don't think a check on that is needed. - Fix several issues related to db loading - .imp files will now get loaded if they exist in a directory passed via clamscan's '-d' flag - .pwdb files will now get loaded if they exist in a directory passed via clamscan's '-d' flag even when compiling without yara support - Changes to .imp, .ign, and .ign2 files will now be reflected in calls to cl_statinidir and cl_statchkdir (and also .pwdb files, even when compiling without yara support) - The contents of .sfp files won't be included in some of the signature counts, and the contents of .cud files will be - Any local.gdb files will no longer be loaded twice - For .imp files, you are no longer required to specify a minimum flevel for wildcard rules, since this isn't needed
2019-01-08 00:09:08 -05:00
cli_targetinfo_init(&info);
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
cli_targetinfo(&info, i, ctx);
info_initialized = true;
if (-1 == info.status) {
cli_dbgmsg("cli_scan_fmap: Failed to successfully parse the executable header. "
"Scan features will be disabled, such as "
"NDB/LDB subsigs using EOF-n/EP+n/EP-n/Sx+n/SEx/SL+n, "
"fuzzy icon matching, "
"MDB/IMP sigs, "
"and bytecode sigs that require exe metadata\n");
}
/* If it's a PE, check the Authenticode header. This would be more
* appropriate in cli_scanpe, but scanraw->cli_scan_fmap gets
* called first for PEs, and we want to determine the trust/block
* status early on so we can skip things like embedded PE extraction
* (which is broken for signed binaries within signed binaries).
*
* If we want to add support for more signature parsing in the future
* (Ex: MachO sigs), do that here too.
*
* One benefit of not continuing on to scan files with trusted signatures
* is that the bytes associated with the exe won't get counted against the
* scansize limits, which means we have an increased chance of catching
* malware in container types (NSIS, iShield, etc.) where the file size is
* large. A common case where this occurs is installers that embed one
* or more of the various Microsoft Redistributable Setup packages. These
* can easily be 5 MB or more in size, and might appear before malware
* does in a given sample.
*/
if (1 == info.status && i == 1) {
ret = cli_check_auth_header(ctx, &(info.exeinfo));
if (ret == CL_VIRUS || ret == CL_VERIFIED) {
goto done;
}
ret = CL_CLEAN;
}
if (!filetype_only) {
/* If we're not doing a filetype-only scan, so we definitely need to include generic signatures.
So initialize the ac data for the generic signatures root. */
ret = cli_ac_initdata(&generic_ac_data, generic_ac_root->ac_partsigs, generic_ac_root->ac_lsigs, generic_ac_root->ac_reloff_num, CLI_DEFAULT_AC_TRACKLEN);
if (CL_SUCCESS != ret) {
goto done;
2014-02-08 00:31:12 -05:00
}
gdata_initialized = true;
/* Recalculate the relative offsets in ac sigs (e.g. those that are based on pe/elf/macho section start/end). */
ret = cli_ac_caloff(generic_ac_root, &generic_ac_data, &info);
if (CL_SUCCESS != ret) {
goto done;
}
/* Recalculate the pcre offsets.
This does an allocation, that we will need to free later. */
ret = cli_pcre_recaloff(generic_ac_root, &generic_pcre_offsets_table, &info, ctx);
if (CL_SUCCESS != ret) {
goto done;
}
generic_pcre_offsets_table_initialized = true;
2014-02-08 00:31:12 -05:00
}
2003-07-29 15:48:06 +00:00
if (target_ac_root) {
/* We have to match against target-specific signatures.
So initialize the ac data for the target-specific signatures root. */
ret = cli_ac_initdata(&target_ac_data, target_ac_root->ac_partsigs, target_ac_root->ac_lsigs, target_ac_root->ac_reloff_num, CLI_DEFAULT_AC_TRACKLEN);
if (CL_SUCCESS != ret) {
goto done;
2014-02-08 00:31:12 -05:00
}
tdata_initialized = true;
2014-02-08 00:31:12 -05:00
/* Recalculate the relative offsets in ac sigs (e.g. those that are based on pe/elf/macho section start/end). */
ret = cli_ac_caloff(target_ac_root, &target_ac_data, &info);
if (CL_SUCCESS != ret) {
goto done;
}
2014-02-08 00:31:12 -05:00
if (target_ac_root->bm_offmode) {
if (ctx->fmap->len >= CLI_DEFAULT_BM_OFFMODE_FSIZE) {
/* Recalculate the relative offsets in boyer-moore signatures (e.g. those that are based on pe/elf/macho section start/end). */
ret = cli_bm_initoff(target_ac_root, &bm_offsets_table, &info);
if (CL_SUCCESS != ret) {
goto done;
}
bm_offsets_table_initialized = true;
2014-02-08 00:31:12 -05:00
}
}
/* Recalculate the pcre offsets.
This does an allocation, that we will need to free later. */
ret = cli_pcre_recaloff(target_ac_root, &target_pcre_offsets_table, &info, ctx);
if (CL_SUCCESS != ret) {
goto done;
}
target_pcre_offsets_table_initialized = true;
}
hdb = ctx->engine->hm_hdb;
fp = ctx->engine->hm_fp;
if (!filetype_only && hdb) {
/* We're not just doing file typing, we're checking for viruses.
So we need to compute the hash sigs, if there are any.
Computing the hash in chunks the same size and time that we do for
matching with the AC & BM pattern matchers is an optimization so we
we can do both processes while the cache is still hot. */
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
need_hash[CLI_HASH_MD5] = cli_hm_have_size(hdb, CLI_HASH_MD5, ctx->fmap->len) ||
cli_hm_have_wild(hdb, CLI_HASH_MD5) ||
cli_hm_have_size(fp, CLI_HASH_MD5, ctx->fmap->len) ||
cli_hm_have_wild(fp, CLI_HASH_MD5);
2014-02-08 00:31:12 -05:00
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
need_hash[CLI_HASH_SHA1] = cli_hm_have_size(hdb, CLI_HASH_SHA1, ctx->fmap->len) ||
cli_hm_have_wild(hdb, CLI_HASH_SHA1) ||
cli_hm_have_size(fp, CLI_HASH_SHA1, ctx->fmap->len) ||
cli_hm_have_wild(fp, CLI_HASH_SHA1);
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
need_hash[CLI_HASH_SHA2_256] = cli_hm_have_size(hdb, CLI_HASH_SHA2_256, ctx->fmap->len) ||
cli_hm_have_wild(hdb, CLI_HASH_SHA2_256) ||
cli_hm_have_size(fp, CLI_HASH_SHA2_256, ctx->fmap->len) ||
cli_hm_have_wild(fp, CLI_HASH_SHA2_256);
/*
* Initialize hash contexts for the hashes that we need to compute.
*/
for (hash_type = CLI_HASH_MD5; hash_type < CLI_HASH_AVAIL_TYPES; hash_type++) {
if (need_hash[hash_type] && !ctx->fmap->have_hash[hash_type]) {
const char *hash_name = cli_hash_name(hash_type);
hashctx[hash_type] = cl_hash_init(hash_name);
if (NULL == hashctx[hash_type]) {
cli_errmsg("cli_scan_fmap: Error initializing %s hash context\n", hash_name);
ret = CL_EARG;
goto done;
}
}
2014-02-08 00:31:12 -05:00
}
}
2003-07-29 15:48:06 +00:00
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
while (offset < ctx->fmap->len) {
if (cli_checktimelimit(ctx) != CL_SUCCESS) {
cli_dbgmsg("Exceeded scan time limit while scanning fmap (max: %u)\n", ctx->engine->maxscantime);
ret = CL_ETIMEOUT;
goto done;
}
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
bytes = MIN(ctx->fmap->len - offset, SCANBUFF);
if (!(buff = fmap_need_off_once(ctx->fmap, offset, bytes)))
2014-02-08 00:31:12 -05:00
break;
if (ctx->scanned)
*ctx->scanned += bytes;
2014-02-08 00:31:12 -05:00
if (target_ac_root) {
const char *virname = NULL;
2014-02-08 00:31:12 -05:00
ret = matcher_run(target_ac_root, buff, bytes, &virname, &target_ac_data, offset,
&info, ftype, ftoffset, acmode, PCRE_SCAN_FMAP, acres, ctx->fmap,
bm_offsets_table_initialized ? &bm_offsets_table : NULL,
&target_pcre_offsets_table, ctx);
if (ret == CL_VIRUS || ret == CL_EMEM) {
goto done;
2014-02-08 00:31:12 -05:00
}
}
2003-07-29 15:48:06 +00:00
if (!filetype_only) {
const char *virname = NULL;
2005-12-12 21:05:26 +00:00
ret = matcher_run(generic_ac_root, buff, bytes, &virname, &generic_ac_data, offset,
&info, ftype, ftoffset, acmode, PCRE_SCAN_FMAP, acres, ctx->fmap,
NULL,
&generic_pcre_offsets_table, ctx);
if (ret == CL_VIRUS || ret == CL_EMEM) {
goto done;
2019-08-22 18:09:18 -04:00
} else if ((acmode & AC_SCAN_FT) && ((cli_file_t)ret >= CL_TYPENO)) {
if (ret > type)
2014-02-08 00:31:12 -05:00
type = ret;
}
2003-07-29 15:48:06 +00:00
/* if (bytes <= (maxpatlen * (offset!=0))), it means the last window finished the file hashing *
* since the last window is responsible for adding intersection between windows (maxpatlen) */
if (hdb && (bytes > (maxpatlen * (offset != 0)))) {
const void *data = buff + maxpatlen * (offset != 0);
uint32_t data_len = bytes - maxpatlen * (offset != 0);
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
for (hash_type = CLI_HASH_MD5; hash_type < CLI_HASH_AVAIL_TYPES; hash_type++) {
/*
* Compute the hash for the current data chunk, if we need to.
*/
if (need_hash[hash_type] && !ctx->fmap->have_hash[hash_type]) {
if (cl_update_hash(hashctx[hash_type], data, data_len)) {
const char *hash_name = cli_hash_name(hash_type);
cli_errmsg("cli_scan_fmap: Error calculating %s hash!\n", hash_name);
ret = CL_EREAD;
goto done;
}
}
}
2014-02-08 00:31:12 -05:00
}
}
if (bytes < SCANBUFF)
2014-02-08 00:31:12 -05:00
break;
offset += bytes - maxpatlen;
}
if (!filetype_only && hdb) {
/* We're not just doing file typing, we're scanning for malware.
So we need to check the hash sigs, if there are any. */
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
for (hash_type = CLI_HASH_MD5; hash_type < CLI_HASH_AVAIL_TYPES; hash_type++) {
/*
* Compute the hash for the current data chunk, if we need to.
*/
if (need_hash[hash_type] && !ctx->fmap->have_hash[hash_type]) {
cl_finish_hash(hashctx[hash_type], digest[hash_type]);
hashctx[hash_type] = NULL;
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
fmap_set_hash(ctx->fmap, digest[hash_type], hash_type);
}
}
2014-02-08 00:31:12 -05:00
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
for (hash_type = CLI_HASH_MD5; hash_type < CLI_HASH_AVAIL_TYPES; hash_type++) {
const char *virname = NULL;
const char *virname_w = NULL;
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
uint8_t *hash = NULL;
2014-02-08 00:31:12 -05:00
/* If no hash, skip to next type */
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
if (!need_hash[hash_type]) {
2014-02-08 00:31:12 -05:00
continue;
}
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
/* Get the hash for the current type.
* We already calculated all the needed ones, so this is a simple lookup.
* Yes, I know there is the digest[] array, but that one may be hashes calculated before this function. */
ret = fmap_get_hash(ctx->fmap, &hash, hash_type);
if (CL_SUCCESS != ret) {
cli_dbgmsg("cli_scan_fmap: Error getting hash for type %d\n", hash_type);
goto done;
}
/* Do hash scan checking hash sigs with specific size.
* This part is fast, so we aren't checking if there are any of hash sigs for this type of hash at this file size */
ret = cli_hm_scan(hash, ctx->fmap->len, &virname, hdb, hash_type);
if (ret == CL_VIRUS) {
/* Matched with size-based hash ... */
ret = cli_append_virus(ctx, virname);
if (ret != CL_SUCCESS) {
goto done;
2014-02-08 00:31:12 -05:00
}
}
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
/* Do hash scan checking hash sigs with wildcard size.
* This part is fast, so we aren't checking if there are any hash sigs for this type of hash with wildcard size */
ret = cli_hm_scan_wild(hash, &virname_w, hdb, hash_type);
if (ret == CL_VIRUS) {
/* Matched with size-agnostic hash ... */
ret = cli_append_virus(ctx, virname_w);
if (ret != CL_SUCCESS) {
goto done;
}
}
2014-02-08 00:31:12 -05:00
}
}
/*
* Evaluate the logical expressions for clamav logical signatures and YARA rules.
*/
2023-11-26 15:01:19 -08:00
// Evaluate for the target-specific signature AC matches.
if (NULL != target_ac_root) {
if (ret != CL_VIRUS) {
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
ret = cli_exp_eval(ctx, target_ac_root, &target_ac_data, &info);
}
}
2023-11-26 15:01:19 -08:00
// Evaluate for the generic signature AC matches.
if (NULL != generic_ac_root) {
if (ret != CL_VIRUS) {
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
ret = cli_exp_eval(ctx, generic_ac_root, &generic_ac_data, &info);
}
}
done:
Swap clean cache from MD5 to SHA2-256 Change the clean-cache to use SHA2-256 instead of MD5. Note that all references are changed to specify "SHA2-256" now instead of "SHA256", for clarity. But there is no plan to add support for SHA3 algorithms at this time. Significant code cleanup. E.g.: - Implemented goto-done error handling. - Used `uint8_t *` instead of `unsigned char *`. - Use `bool` for boolean checks, rather than `int. - Used `#defines` instead of magic numbers. - Removed duplicate `#defines` for things like hash length. Add new option to calculate and record additional hash types when the "generate metadata JSON" feature is enabled: - libclamav option: `CL_SCAN_GENERAL_STORE_EXTRA_HASHES` - clamscan option: `--json-store-extra-hashes` (default off) - clamd.conf option: `JsonStoreExtraHashes` (default 'no') Renamed the sigtool option `--sha256` to `--sha2-256`. The original option is still functional, but is deprecated. For the "generate metadata JSON" feature, the file hash is now stored as "sha2-256" instead of "FileMD5". If you enable the "extra hashes" option, then it will also record "md5" and "sha1". Deprecate and disable the internal "SHA collect" feature. This option had been hidden behind C #ifdef checks for an option that wasn't exposed through CMake, so it was basically unavailable anyways. Changes to calculate file hashes when they're needed and no sooner. For the FP feature in the matcher module, I have mimiced the optimization in the FMAP scan routine which makes it so that it can calculate multiple hashes in a single pass of the file. The `HandlerType` feature stores a hash of the file in the scan ctx to prevent retyping the exact same data more than once. I removed that hash field and replaced it with an attribute flag that is applied to the new recursion stack layer when retyping a file. This also closes a minor bug that would prevent retyping a file with an all-zero hash. :) The work upgrading cache.c to support SHA2-256 sized hashes thanks to: https://github.com/m-sola CLAM-255 CLAM-1858 CLAM-1859 CLAM-1860
2025-06-03 19:03:20 -04:00
for (hash_type = CLI_HASH_MD5; hash_type < CLI_HASH_AVAIL_TYPES; hash_type++) {
if (NULL != hashctx[hash_type]) {
cl_hash_destroy(hashctx[hash_type]);
}
}
2014-02-08 00:31:12 -05:00
if (gdata_initialized) {
cli_ac_freedata(&generic_ac_data);
}
if (tdata_initialized) {
cli_ac_freedata(&target_ac_data);
}
if (generic_pcre_offsets_table_initialized) {
cli_pcre_freeoff(&generic_pcre_offsets_table);
}
if (target_pcre_offsets_table_initialized) {
cli_pcre_freeoff(&target_pcre_offsets_table);
}
if (info_initialized) {
cli_targetinfo_destroy(&info);
}
2010-07-29 16:58:27 +02:00
if (bm_offsets_table_initialized) {
cli_bm_freeoff(&bm_offsets_table);
}
if (ret != CL_SUCCESS) {
return ret;
}
return (acmode & AC_SCAN_FT) ? type : CL_SUCCESS;
}
Fix issues reading from uncompressed nested files The fmap module provides a mechanism for creating a mapping into an existing map at an offset and length that's used when a file is found with an uncompressed archive or when embedded files are found with embedded file type recognition in scanraw(). This is the "fmap_duplicate()" function. Duplicate fmaps just reference the original fmap's 'data' or file handle/descriptor while allowing the caller to treat it like a new map using offsets and lengths that don't account for the original/actual file dimensions. fmap's keep track of this with m->nested_offset & m->real_len, which admittedly have confusing names. I found incorrect uses of these in a handful of locations. Notably: - In cli_magic_scan_nested_fmap_type(). The force-to-disk feature would have been checking incorrect sizes and may have written incorrect offsets for duplicate fmaps. - In XDP parser. - A bunch of places from the previous commit when making dupe maps. This commit fixes those and adds lots of documentation to the fmap.h API to try to prevent confusion in the future. nested_offset should never be referenced outside of fmap.c/h. The fmap_* functions for accessing or reading map data have two implementations, mem_* or handle_*, depending the data source. I found issues with some of these so I made a unit test that covers each of the functions I'm concerned about for both types of data sources and for both original fmaps and nested/duplicate fmaps. With the tests, I found and fixed issues in these fmap functions: - handle_need_offstr(): must account for the nested_offset in dupe maps. - handle_gets(): must account for nested_offset and use len & real_len correctly. - mem_need_offstr(): must account for nested_offset in dupe maps. - mem_gets(): must account for nested_offset and use len & real_len correctly. Moved CDBRANGE() macro out of function definition so for better legibility. Fixed a few warnings.
2021-10-03 14:13:55 -07:00
#define CDBRANGE(field, val) \
if (field[0] != CLI_OFF_ANY) { \
if (field[0] == field[1] && field[0] != val) \
continue; \
else if (field[0] != field[1] && ((field[0] && field[0] > val) || \
(field[1] && field[1] < val))) \
continue; \
}
cl_error_t cli_matchmeta(cli_ctx *ctx, const char *fname, size_t fsizec, size_t fsizer, int encrypted, unsigned int filepos, int res1)
{
const struct cli_cdb *cdb;
cl_error_t ret = CL_SUCCESS;
cli_dbgmsg("CDBNAME:%s:%llu:%s:%llu:%llu:%d:%u:%u\n",
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
cli_ftname(cli_recursion_stack_get_type(ctx, -1)), (long long unsigned)fsizec, fname, (long long unsigned)fsizec, (long long unsigned)fsizer,
encrypted, filepos, res1);
2011-08-22 15:22:55 +03:00
if (ctx->engine && ctx->engine->cb_meta) {
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
if (ctx->engine->cb_meta(cli_ftname(cli_recursion_stack_get_type(ctx, -1)), fsizec, fname, fsizer, encrypted, filepos, ctx->cb_ctx) == CL_VIRUS) {
cli_dbgmsg("inner file blocked by callback: %s\n", fname);
ret = cli_append_virus(ctx, "Detected.By.Callback");
if (ret != CL_SUCCESS) {
return ret;
}
}
}
2011-08-22 15:22:55 +03:00
if (NULL == ctx->engine || (NULL == (cdb = ctx->engine->cdb))) {
return CL_CLEAN;
}
do {
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
if (cdb->ctype != CL_TYPE_ANY && cdb->ctype != cli_recursion_stack_get_type(ctx, -1))
continue;
if (cdb->encrypted != 2 && cdb->encrypted != encrypted)
continue;
if (cdb->res1 && (cdb->ctype == CL_TYPE_ZIP || cdb->ctype == CL_TYPE_RAR) && cdb->res1 != res1)
continue;
libclamav: Fix scan recursion tracking Scan recursion is the process of identifying files embedded in other files and then scanning them, recursively. Internally this process is more complex than it may sound because a file may have multiple layers of types before finding a new "file". At present we treat the recursion count in the scanning context as an index into both our fmap list AND our container list. These two lists are conceptually a part of the same thing and should be unified. But what's concerning is that the "recursion level" isn't actually incremented or decremented at the same time that we add a layer to the fmap or container lists but instead is more touchy-feely, increasing when we find a new "file". To account for this shadiness, the size of the fmap and container lists has always been a little longer than our "max scan recursion" limit so we don't accidentally overflow the fmap or container arrays (!). I've implemented a single recursion-stack as an array, similar to before, which includes a pointer to each fmap at each layer, along with the size and type. Push and pop functions add and remove layers whenever a new fmap is added. A boolean argument when pushing indicates if the new layer represents a new buffer or new file (descriptor). A new buffer will reset the "nested fmap level" (described below). This commit also provides a solution for an issue where we detect embedded files more than once during scan recursion. For illustration, imagine a tarball named foo.tar.gz with this structure: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | But suppose baz.exe embeds a ZIP archive and a 7Z archive, like this: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | baz.exe | PE | 0 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | │   └── hello.txt | ASCII | 2 | 0 | | └── sfx.7z | 7Z | 1 | 1 | |    └── world.txt | ASCII | 2 | 0 | (A) If we scan for embedded files at any layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | ├── foo.tar | TAR | 1 | 0 | | │ ├── bar.zip | ZIP | 2 | 1 | | │ │   └── hola.txt | ASCII | 3 | 0 | | │ ├── baz.exe | PE | 2 | 1 | | │ │ ├── sfx.zip | ZIP | 3 | 1 | | │ │ │   └── hello.txt | ASCII | 4 | 0 | | │ │ └── sfx.7z | 7Z | 3 | 1 | | │ │    └── world.txt | ASCII | 4 | 0 | | │ ├── sfx.zip | ZIP | 2 | 1 | | │ │   └── hello.txt | ASCII | 3 | 0 | | │ └── sfx.7z | 7Z | 2 | 1 | | │   └── world.txt | ASCII | 3 | 0 | | ├── sfx.zip | ZIP | 1 | 1 | | └── sfx.7z | 7Z | 1 | 1 | (A) is bad because it scans content more than once. Note that for the GZ layer, it may detect the ZIP and 7Z if the signature hits on the compressed data, which it might, though extracting the ZIP and 7Z will likely fail. The reason the above doesn't happen now is that we restrict embedded type scans for a bunch of archive formats to include GZ and TAR. (B) If we scan for embedded files at the foo.tar layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | ├── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 2 | 1 | | │   └── hello.txt | ASCII | 3 | 0 | | └── sfx.7z | 7Z | 2 | 1 | |    └── world.txt | ASCII | 3 | 0 | (B) is almost right. But we can achieve it easily enough only scanning for embedded content in the current fmap when the "nested fmap level" is 0. The upside is that it should safely detect all embedded content, even if it may think the sfz.zip and sfx.7z are in foo.tar instead of in baz.exe. The biggest risk I can think of affects ZIPs. SFXZIP detection is identical to ZIP detection, which is why we don't allow SFXZIP to be detected if insize of a ZIP. If we only allow embedded type scanning at fmap-layer 0 in each buffer, this will fail to detect the embedded ZIP if the bar.exe was not compressed in foo.zip and if non-compressed files extracted from ZIPs aren't extracted as new buffers: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.zip | ZIP | 0 | 0 | | └── bar.exe | PE | 1 | 1 | | └── sfx.zip | ZIP | 2 | 2 | Provided that we ensure all files extracted from zips are scanned in new buffers, option (B) should be safe. (C) If we scan for embedded files at the baz.exe layer, we may detect: | description | type | rec level | nested fmap level | | ------------------------- | ----- | --------- | ----------------- | | foo.tar.gz | GZ | 0 | 0 | | └── foo.tar | TAR | 1 | 0 | | ├── bar.zip | ZIP | 2 | 1 | | │   └── hola.txt | ASCII | 3 | 0 | | └── baz.exe | PE | 2 | 1 | | ├── sfx.zip | ZIP | 3 | 1 | | │   └── hello.txt | ASCII | 4 | 0 | | └── sfx.7z | 7Z | 3 | 1 | |    └── world.txt | ASCII | 4 | 0 | (C) is right. But it's harder to achieve. For this example we can get it by restricting 7ZSFX and ZIPSFX detection only when scanning an executable. But that may mean losing detection of archives embedded elsewhere. And we'd have to identify allowable container types for each possible embedded type, which would be very difficult. So this commit aims to solve the issue the (B)-way. Note that in all situations, we still have to scan with file typing enabled to determine if we need to reassign the current file type, such as re-identifying a Bzip2 archive as a DMG that happens to be Bzip2- compressed. Detection of DMG and a handful of other types rely on finding data partway through or near the ned of a file before reassigning the entire file as the new type. Other fixes and considerations in this commit: - The utf16 HTML parser has weak error handling, particularly with respect to creating a nested fmap for scanning the ascii decoded file. This commit cleans up the error handling and wraps the nested scan with the recursion-stack push()/pop() for correct recursion tracking. Before this commit, each container layer had a flag to indicate if the container layer is valid. We need something similar so that the cli_recursion_stack_get_*() functions ignore normalized layers. Details... Imagine an LDB signature for HTML content that specifies a ZIP container. If the signature actually alerts on the normalized HTML and you don't ignore normalized layers for the container check, it will appear as though the alert is in an HTML container rather than a ZIP container. This commit accomplishes this with a boolean you set in the scan context before scanning a new layer. Then when the new fmap is created, it will use that flag to set similar flag for the layer. The context flag is reset those that anything after this doesn't have that flag. The flag allows the new recursion_stack_get() function to ignore normalized layers when iterating the stack to return a layer at a requested index, negative or positive. Scanning normalized extracted/normalized javascript and VBA should also use the 'layer is normalized' flag. - This commit also fixes Heuristic.Broken.Executable alert for ELF files to make sure that: A) these only alert if cli_append_virus() returns CL_VIRUS (aka it respects the FP check). B) all broken-executable alerts for ELF only happen if the SCAN_HEURISTIC_BROKEN option is enabled. - This commit also cleans up the error handling in cli_magic_scan_dir(). This was needed so we could correctly apply the layer-is-normalized-flag to all VBA macros extracted to a directory when scanning the directory. - Also fix an issue where exceeding scan maximums wouldn't cause embedded file detection scans to abort. Granted we don't actually want to abort if max filesize or max recursion depth are exceeded... only if max scansize, max files, and max scantime are exceeded. Add 'abort_scan' flag to scan context, to protect against depending on correct error propagation for fatal conditions. Instead, setting this flag in the scan context should guarantee that a fatal condition deep in scan recursion isn't lost which result in more stuff being scanned instead of aborting. This shouldn't be necessary, but some status codes like CL_ETIMEOUT never used to be fatal and it's easier to do this than to verify every parser only returns CL_ETIMEOUT and other "fatal status codes" in fatal conditions. - Remove duplicate is_tar() prototype from filestypes.c and include is_tar.h instead. - Presently we create the fmap hash when creating the fmap. This wastes a bit of CPU if the hash is never needed. Now that we're creating fmap's for all embedded files discovered with file type recognition scans, this is a much more frequent occurence and really slows things down. This commit fixes the issue by only creating fmap hashes as needed. This should not only resolve the perfomance impact of creating fmap's for all embedded files, but also should improve performance in general. - Add allmatch check to the zip parser after the central-header meta match. That way we don't multiple alerts with the same match except in allmatch mode. Clean up error handling in the zip parser a tiny bit. - Fixes to ensure that the scan limits such as scansize, filesize, recursion depth, # of embedded files, and scantime are always reported if AlertExceedsMax (--alert-exceeds-max) is enabled. - Fixed an issue where non-fatal alerts for exceeding scan maximums may mask signature matches later on. I changed it so these alerts use the "possibly unwanted" alert-type and thus only alert if no other alerts were found or if all-match or heuristic-precedence are enabled. - Added the "Heuristics.Limits.Exceeded.*" events to the JSON metadata when the --gen-json feature is enabled. These will show up once under "ParseErrors" the first time a limit is exceeded. In the present implementation, only one limits-exceeded events will be added, so as to prevent a malicious or malformed sample from filling the JSON buffer with millions of events and using a tonne of RAM.
2021-09-11 14:15:21 -07:00
CDBRANGE(cdb->csize, cli_recursion_stack_get_size(ctx, -1));
CDBRANGE(cdb->fsizec, fsizec);
CDBRANGE(cdb->fsizer, fsizer);
CDBRANGE(cdb->filepos, filepos);
if (cdb->name.re_magic && (!fname || cli_regexec(&cdb->name, fname, 0, NULL, 0) == REG_NOMATCH))
continue;
ret = cli_append_virus(ctx, cdb->virname);
if (ret != CL_SUCCESS) {
return ret;
}
} while ((cdb = cdb->next));
return ret;
}