Temp directory recursion in ClamAV is when each layer of a scan gets its
own temp directory in the parent layer's temp directory.
In addition to temp directory recursion, ClamAV has been creating a new
subdirectory for each file scan as a risk-adverse method to ensure
no temporary file leaks fill up the disk.
Creating a directory is relatively slow on Windows in particular if
scanning a lot of very small files.
This commit:
1. Separates the temp directory recursion feature from the leave-temps
feature so that libclamav can leave temp files without making
subdirectories for each file scanned.
2. Makes it so that when temp directory recursion is off, libclamav
will just use the configure temp directory for all files.
The new option to enable temp directory recursion is for libclamav-only
at this time. It is off by default, and you can enable it like this:
```c
cl_engine_set_num(engine, CL_ENGINE_TMPDIR_RECURSION, 1);
```
For the `clamscan` and `clamd` programs, temp directory recursion will
be enabled when `--leave-temps` / `LeaveTemporaryFiles` is enabled.
The difference is that when disabled, it will return to using the
configured temp directory without making a subdirectory for each file
scanned, so as to improve scan performance for small files, mostly on
Windows.
Under the hood, this commit also:
1. Cleans up how we keep track of tmpdirs for each layer.
The goal here is to align how we keep track of layer-specific stuff
using the scan_layer structure.
2. Cleans up how we record metadata JSON for embedded files.
Note: Embedded files being different from Contained files, as they
are extracted not with a parser, but by finding them with
file type magic signatures.
CLAM-1583
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
We add the _OR_GOTO_DONE suffix to the macros that go to done if the
allocation fails. This makes it obvious what is different about the
macro versus the equivalent function, and that error handling is
built-in.
Renamed the cli_strdup to safer_strdup to make it obvious that it exists
because it is safer than regular strdup. Regular strdup doesn't have the
NULL check before trying to dup, and so may result in a NULL-deref
crash.
Also remove unused STRDUP (_OR_GOTO_DONE) macro, since the one with the
NULL-check is preferred.
We have some special functions to wrap malloc, calloc, and realloc to
make sure we don't allocate more than some limit, similar to the
max-filesize and max-scansize limits. Our wrappers are really only
needed when allocating memory for scans based on untrusted user input,
where a scan file could have bytes that claim you need to allocate
some ridiculous amount of memory. Right now they're named:
- cli_malloc
- cli_calloc
- cli_realloc
- cli_realloc2
... and these names do not convey their purpose
This commit renames them to:
- cli_max_malloc
- cli_max_calloc
- cli_max_realloc
- cli_max_realloc2
The realloc ones also have an additional feature in that they will not
free your pointer if you try to realloc to 0 bytes. Freeing the memory
is undefined by the C spec, and only done with some realloc
implementations, so this stabilizes on the behavior of not doing that,
which should prevent accidental double-free's.
So for the case where you may want to realloc and do not need to have a
maximum, this commit adds the following functions:
- cli_safer_realloc
- cli_safer_realloc2
These are used for the MPOOL_REALLOC and MPOOL_REALLOC2 macros when
MPOOL is disabled (e.g. because mmap-support is not found), so as to
match the behavior in the mpool_realloc/2 functions that do not make use
of the allocation-limit.
There are a large number of allocations for fix sized buffers using the
`cli_malloc` and `cli_calloc` calls that check if the requested size is
larger than our allocation threshold for allocations based on untrusted
input. These allocations will *always* be higher than the threshold, so
the extra stack frame and check for these calls is a waste of CPU.
This commit replaces needless calls with A -> B:
- cli_malloc -> malloc
- cli_calloc -> calloc
- CLI_MALLOC -> MALLOC
- CLI_CALLOC -> CALLOC
I also noticed that our MPOOL_MALLOC / MPOOL_CALLOC are not limited by
the max-allocation threshold, when MMAP is found/enabled. But the
alternative was set to cli_malloc / cli_calloc when disabled. I changed
those as well.
I didn't change the cli_realloc/2 calls because our version of realloc
not only implements a threshold but also stabilizes the undefined
behavior in realloc to protect against accidental double-free's.
It may be worth implementing a cli_realloc that doesn't have the
threshold built-in, however, so as to allow reallocaitons for things
like buffers for loading signatures, which aren't subject to the same
concern as allocations for scanning possible malware.
There was one case in mbox.c where I changed MALLOC -> CLI_MALLOC,
because it appears to be allocating based on untrusted input.
PDB, WDB, and CDB signatures can use a regex feature to match
domain names. At one time in the process we ran a filter search
to speed up filtering out non-matching static patterns but were
accidentally discarding the result. When tested, it turns out
it wasn't working correctly anyways.
Since then, we've fixed some bugs and upgraded the regex
implementation to the latest version. After re-testing, I have
found that the filter_search() appears to be working correctly
now, both in the unit tests and with the existing CDB signatures.
The header parsing / executable metadata collecting functions for the
PE, ELF, and Mach-O file types were using `int` for the return type.
Mostly they were returning 0 for success and -1, -2, -3, or -4 for
failure. But in some cases they were returning cl_error_t enum values
for failure. Regardless, the function using them was treating 0 as
success and non-zero as failure, which it stored as -1 ... every time.
This commit switches them all to use cl_error_t. I am continuing to
storeo the final result as 0 / -1 in the `peinfo` struct, but outside of
that everything has been made consistent.
While I was working on that, I got a tad side tracked. I noticed that
the target type isn't an enum, or even a set of #defines. So I made an
enum and then changed the code that uses target types to use the enum.
I also removed the `target` parameter from a number of functions that
don't actually use it at all. Some recursion was masking the fact that
it was an unused parameter which is why there was no warning about it.
During wdb load, it was possible to go beyond the bounds
of the pattern buffer due to two subsequent increment ops
with no bounds checking in between.
This issue was reported by external researchers and
they provided the fix as well.
Based on our own research, this is a defect but not a vulnerability.
Co-authored-by: Mickey Sola <micksola@cisco.com>
The check of pattern_len against FILEBUF is largely meaningless since
pattern is derived from a strchr() call against buffer (with length FILEBUFF).
This fix ensures that the relative size is checked against max buffer size
which prevents overwriting stack memory with a single null byte.
Resolves: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=45247
A heap buffer overflow could occur during resource cleanup if a
malloc fails when adding a regex pattern to the phishing suffix tree.
The solution is to increment suffix_cnt after cli_realloc succeeds.
The issue was identified using fault injection and is not a vulnerability.
Resolves: https://github.com/Cisco-Talos/clamav/issues/429
Corrected buffer size check in the regex signature parsing code.
This resolves a possible 1-byte stack buffer overwrite (NULL byte).
We determined that this issue is not a vulnerability.
Fixes: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=43869
* Added loglevel parameter to logg()
* Fix logg and mprintf internals with new loglevels
* Update all logg calls to set loglevel
* Update all mprintf calls to set loglevel
* Fix hidden logg calls
* Executed clam-format
Improvements to use modern block list and allow list verbiage.
blacklist -> block list
whitelist -> allow listed
blacklisted -> blocked
whitelisted -> allowed
In the case of certificate verification, use "trust" or "verify" when
something is allowed.
Also changed domainlist -> domain list (or DomainList) to match.
An ENABLE_TESTS CMake option is provided so that users can disable
testing if they don't want it. Instructions for how to use this
included in the INSTALL.cmake.md file.
If you run `ctest`, each testcase will write out a log file to the
<build>/unit_tests directory.
As with Autotools' make check, the test files are from test/.split
and unit_tests/.split files, but for CMake these are generated at
build time instead of at test time.
On Posix systems, sets the LD_LIBRARY_PATH so that ClamAV-compiled
libraries can be loaded when running tests.
On Windows systems, CTest will identify and collect all library
dependencies and assemble a temporarily install under the
build/unit_tests directory so that the libraries can be loaded when
running tests.
The same feature is used on Windows when using CMake to install to
collect all DLL dependencies so that users don't have to install them
manually afterwards.
Each of the CTest tests are run using a custom wrapper around Python's
unittest framework, which is also responsible for finding and inserting
valgrind into the valgrind tests on Posix systems.
Unlike with Autotools, the CMake CTest Valgrind-tests are enabled by
default, if Valgrind can be found. There's no need to set VG=1.
CTest's memcheck module is NOT supported, because we use Python to
orchestrate our tests.
Added a bunch of Windows compatibility changes to the unit tests.
These were primarily changing / to PATHSEP and making adjustments
to use Win32 C headers and ifdef out the POSIX ones which aren't
available on Windows. Also disabled a bunch of tests on Win32
that don't work on Windows, notably the mmap ones and FD-passing
(i.e. FILEDES) ones.
Add JSON_C_HAVE_INTTYPES_H definition to clamav-config.h to eliminate
warnings on Windows where json.h is included after inttypes.h because
json-c's inttypes replacement relies on it.
This is a it of a hack and may be removed if json-c fixes their
inttypes header stuff in the future.
Add preprocessor definitions on Windows to disable MSVC warnings about
CRT secure and nonstandard functions. While there may be a better
solution, this is needed to be able to see other more serious warnings.
Add missing file comment block and copyright statement for clamsubmit.c.
Also change json-c/json.h include filename to json.h in clamsubmit.c.
The directory name is not required.
Changed the hash table data integer type from long, which is poorly
defined, to size_t -- which is capable of storing a pointer. Fixed a
bunch of casts regarding this variable to eliminate warnings.
Fixed two bugs causing utf8 encoding unit tests to fail on Windows:
- The in_size variable should be the number of bytes, not the character
count. This was was causing the SHIFT_JIS (japanese codepage) to UTF8
transcoding test to only transcode half the bytes.
- It turns out that the MultiByteToWideChar() API can't transcode
UTF16-BE to UTF16-LE. The solution is to just iterate over the buffer
and flip the bytes on each uint16_t. This but was causing the UTF16-BE
to UTF8 tests to fail.
I also split up the utf8 transcoding tests into separate tests so I
could see all of the failures instead of just the first one.
Added a flags parameter to the unit test function to open testfiles
because it turns out that on Windows if a file contains the \r\n it will
replace it with just \n if you opened the file as a text file instead of
as binary. However, if we open the CBC files as binary, then a bunch of
bytecode tests fail. So I've changed the tests to open the CBC files in
the bytecode tests as text files and open all other files as binary.
Ported the feature tests from shell scripts to Python using a modified
version of our QA test-framework, which is largely compatible and will
allow us to migrate some QA tests into this repo. I'd like to add GitHub
Actions pipelines in the future so that all public PR's get some testing
before anyone has to manually review them.
The clamd --log option was missing from the help string, though it
definitely works. I've added it in this commit.
It appears that clamd.c was never clang-format'd, so this commit also
reformats clamd.c.
Some of the check_clamd tests expected the path returned by clamd to
match character for character with original path sent to clamd. However,
as we now evaluate real paths before a scan, the path returned by clamd
isn't going to match the relative (and possibly symlink-ridden) path
passed to clamdscan. I fixed this test by changing the test to search
for the basename: <signature> FOUND within the response instead of
matching the exact path.
Autotools: Link check_clamd with libclamav so we can use our utility
functions in check_clamd.c.
290424 Missing break in switch - In hash_match: Missing break
statement between cases in switch statement
290414 Resource leak - In cli_scanishield_msi: Leak of memory or
pointers to system resources. Memory leak in a fail case
288197 Resource leak - In decrypt_any: Leak of memory or pointers
to system resources. Memory leak in a fail case
290426 Resource leak - In cli_magic_scan: Leak of memory or pointers
to system resources. Leaked a file prefix when running with
--save-temps
192923 Resource leak - In cli_scanrar: Leak of memory or pointers to
system resources. Leaked a file descriptor if a virus was found in
a RAR file comment
225146 Resource leak - In cli_scanegg: Leak of memory or pointers
to system resources. Leaked a file descriptor if unable to write
a comment file to disk
290425 Resource leak - In scan_common: Leak of memory or pointers
to system resources. Memory leaks in various fail cases.
Also changes cli_scanrar to write out the file comment only if
--leave-temps is specified and scan the buffer (like what is done
in cli_scanegg) instead of writing the file out, scanning that,
and then deleting the file if --leave-temps is not specified.
The unit tests stopped working when correcting an issue with a
switch statement that determined what type of signature had matched
on a Google SafeBrowsing GDB rule. Looking into the unit tests, it
looks like the code had always assumed that the test cases would be
detected by a malware test rule in unit_tests/input/daily.gdb, but
now some of the tests get matched on the phishing test rule.
I updated the test logic to be more clear, and added tests for both
cases now.
Fix some memory leaks in libclamav/scanners.c
Reviewing Coverity bug reports we found that the return value to this
filter_search call was effectively being ignored, causing no filtering
to occur. Fixing this issue resulted in a unit test that uses the
following match list regex to fail when searching for `ebay.com`.:
.+\\.paypal\\.(com|de|fr|it)([/?].*)?:.+\\.ebay\\.(at|be|ca|ch|co\\.uk|de|es|fr|ie|in|it|nl|ph|pl|com(\\.(au|cn|hk|my|sg))?)/
After investigating further, this is because the regex_list_add_pattern
call, which parses the regex for suffixes and attempts to add these to
the filter, can't handle the `com(\\.(au|cn|hk|my|sg))?` portion of
the regex. As a result, it only adds `ebay.at`, `ebay.be`, `ebay.ca`, up
through `ebay.pl` into the filter). With the code returning if no filter match
is found, the `ebay.com` suffix not existing in the filter causes incoming URLs
to be treated as if there are no corresponding regexes for ebay.com, which results
in no regex rules being evaluated against it.
We should get the regex parsing code working (and ensure it handles any
other complex cases in daily.cdb) before re-enabling this code. The code
has had no effect for 12+ years at this point, though, so it's probably
safe to wait a bit longer without it.
Fixed the following Coverity issues:
- 225236 - In cli_egg_extract_file: Dereference of an explicit
null value (CWE-476). The first fail case checked handle for
NULL and then dereferenced it in the done block
- 225209 - In executeIfNewVersion: Leak of memory or pointers
to system resources (CWE-404). modifiedCommand was defined
twice, with the inner instance being assigned to and the
outer instance being freed
- 225201 - In regex_list_match: Code can never be reached
because of a logical contradiction (CWE-561). The code had
logic off to the side that may have been missed:
filter_search_rc = filter_search(&matcher->filter, (const unsigned char *)bufrev, buffer_len) != -1;
if (filter_search_rc == -1) {
- 225198 - In phishingCheck: Leak of memory or pointers to
system resources (CWE-404). A fail case caused by malloc
failing would leak previously allocated memory.
- 225197 - In updatecustomdb: A pointer to freed memory
is dereferenced, used as a function argument, or otherwise
used (CWE-416). In a fail case, a pointer was freed and
then used in a debug print statement
- 225190 - In updatedb: A pointer to freed memory is
dereferenced, used as a function argument, or otherwise used
(CWE-416). In a fail case, a pointer was freed and then used
in a debug print statement
- 225195 - In cli_egg_open: The sizeof operator is used on a
wrong argument that incidentally has the same size (CWE-467).
sizeof(char **) was being used instead of sizeof(char *)
- 225193 - In egg_parse_comment_header: Code can never be
reached because of a logical contradiction (CWE-561).
A cleanup case for variable comment was unnecessary, and
to fix comment was removed entirely.
- 225147 - In get_server_node: Code can never be reached
because of a logical contradiction (CWE-561). A cleanup
case for variable url was unnecessary
- 225168 - In download_complete_callback: Missing break
statement between cases in switch statement (CWE-484).
In the case where forking failed, freshclam would check
the database without forking but then continue on to
execute the code intended to be done in the child process
because of a missing break statement
- 225152 - In cli_egg_lzma_decompress: Use of an
uninitialized variable (CWE-457). Certain fail cases
would call cli_LzmaShutdown on an uninitialized stream.
Now it’s only called after initialization occurs.
Some detections, like phishing, are considered heuristic alerts because
they match based on behavior more than on content. A subset of these
are considered "potentially unwanted" (low-severity). These
low-severity alerts include:
- phishing
- PDFs with obfuscated object names
- bytecode signature alerts that start with "BC.Heuristics"
The concept is that unless you enable "heuristic precedence" (a method
of lowing the threshold to immediateley alert on low-severity
detections), the scan should continue after a match in case a higher
severity match is found. Only at the end will it print the low-severity
match if nothing else was found.
The current implementation is buggy though. Scanning of archives does
not correctly bail out for the entire archive if one email contains a
phishing link. Instead, it sets the "heuristic found" flag then and
alerts for every subsequent file in the archive because it doesn't know
if the heuristic was found in an embedded file or the target file.
Because it's just a heuristic and the status is "clean", it keeps
scanning.
This patch corrects the behavior by checking if a low-severity alerts
were found at the end of scanning the target file, instead of at the end
of each embedded file.
Additionally, this patch fixes an in issue with phishing alerts wherein
heuristic precedence mode did not cause a scan to stop after the first
alert.
The above changes required restructuring to create an fmap inside of
cl_scandesc_callback() so that scan_common() could be modified to
require an fmap and set up so that the current *ctx->fmap pointer is
never NULL when scan_common() evaluates match results.
Also fixed a couple minor bugs in the phishing unit tests and cleaned up
the test code for improved legitibility and type safety.
