`clamscan/manager.c`: Fix double-free in an error condition in `scanfile()`.
`common/optparser.c`: Fix uninitialized use of the `numarg` variable when
`arg` is `NULL`.
`libclamav/cache.c`: Don't check if `ctx-fmap` is `NULL` when we've
already dereferenced it.
`libclamav/crypto.c`: The `win_exception` variable and associated logic
is Windows-specific and so needs preprocessor platform checks. Otherwise
it generates unused variable warnings.
`libclamav/crypto.c`: Check for `size_t` overflow of the `byte_read`
variable in the `cl_hash_file_fd_ex()` function.
`libclamav/crypto.c`: Fix a memory leak in the `cl_hash_file_fd_ex()`
function.
`libclamav/fmap.c`: Correctly the `name` and `path` pointer if
`fmap_duplicate()` fails. Also need to clear those variables when
duplicating the parent `map` so that on error it does not free the wrong
`name` or `path`.
`libclamav/fmap.c`: Refine error handling for `hash_string` cleanup in
`cl_fmap_get_hash()`. Coverity's complaint was that `hash_string` could
never be non-NULL if `status` is not `CL_SUCCESS`. I.e., the cleanup is
dead code. I don't think my cleanup actually "fixes" that though it is
definitely a better way to do the error handling.
The `if (NULL != hash_string) {` check is still technically dead code.
It safeguards against future changes that may `goto done` between the
allocation and transfering ownership from `hash_string` to `hash_out`.
`libclamav/others.c`: Fix possible memory leak in `cli_recursion_stack_push()`.
`libclamav/others.c`: Refactor an if/else + switch statement inside
`cli_dispatch_scan_callback()` so that the `CL_SCAN_CALLBACK_ALERT` case
is not dead-code. It's also easier to read now.
`libclamav/pdfdecode.c`: For logging, use the `%zu` to format `size_t`
instead of casting to `long long` and using `%llu`. Simiularly use the
`STDu32` format string macro for `uint32_t`.
`libclamav/pdfdecode.c`: Fix a possible double-free for the `decoded`
pointer in `filter_lzwdecode()`.
`libclamav/pdfdecode.c`: Remove the `if (capacity > UINT_MAX) {`
overflow check inside `filter_lzwdecode()`, which didn't do anything.
The `capacity` variable this point is a fixed value and so I also changed
the `avail_out` to be that fixed `INFLATE_CHUNK_SIZE` value rather than
using `capacity`. It is more straightforward and replicates how similar
logic works later in the file.
I also removed the copy-pasted `(Bytef *)` cast which didn't reaaally do
anything, and was a copypaste from a different algorihm. The lzw
implementation interface doesn't use `Bytef`.
`libclamav/readdb.c`: Fix a possible NULL-deref on the `matcher` variable
in the error handling/cleanup code if the function fails.
`libclamav/scanners.c`: Fix an issue where the return value from some of
the parsers may be lost/overridden by the call to
`cli_dispatch_scan_callback()` just after the `done:` label in
`cli_magic_scan()`.
`libclamav/scanners.c`: Silence an unused-return value warning when
calling `cli_basename()`.
`sigtool/sigtool.c` and `unit_tests/check_regex.c`:
Fix possible NULL-derefs of the `ctx.recursion_stack` pointer in the error
handling for several functions.
Also, and this isn't a Coverity thing:
`libclamav/json_api.c` and `libclamav/others.c`:
Fix support for libjson-c version 0.13 and older.
I don't think we *should* be using the old version, but some environments
such as the current OSS-Fuzz base image are older and still use it.
The issue is that `json_object_new_uint64()` was introduced in a later
libjson-c version, so we have to fallback to use `json_object_new_int64()`
with older libjson-c, provided the int were storing isn't too big.
CLAM-2768
If the outermost layer is trusted (e.g. using an FP signature), the verdict passed
back by the `cl_scan*_ex()` functions should be CL_VERDICT_TRUSTED.
To ensure this, and other correct verdicts, I moved the logic setting the verdict
to occur when adding indicators, or trusting a given layer. Then at the end of a
scan, it will set the output verdict parameter to the top level verdict.
This commit also:
* Fixes a bug in the `ex_scan_callbacks` program where a crash would happen when
a hash was retrieved for an inner layer, but isn't for the container.
* Added debug logs whenever a hash is calculated or set, printing the hash type
and hash string.
* When a layer is trusted, in addition to removing evidence for that layer, it
will also index the metadata JSON (if that feature is enabled) and will rename
any "Viruses" to "IgnoredAlerts", and rename "ContainedIndicators" to
"IgnoredContainedIndicators".
* Fixed an issue where setting the hash algorithm with extra characters, like
setting to "sha256789" would ignore the extra characters, and report the hash
type as the same. It will now fail if the string length differs from the known
hash algorithm.
Add the following scan callbacks:
```c
cl_engine_set_scan_callback(engine, &pre_hash_callback, CL_SCAN_CALLBACK_PRE_HASH);
cl_engine_set_scan_callback(engine, &pre_scan_callback, CL_SCAN_CALLBACK_PRE_SCAN);
cl_engine_set_scan_callback(engine, &post_scan_callback, CL_SCAN_CALLBACK_POST_SCAN);
cl_engine_set_scan_callback(engine, &alert_callback, CL_SCAN_CALLBACK_ALERT);
cl_engine_set_scan_callback(engine, &file_type_callback, CL_SCAN_CALLBACK_FILE_TYPE);
```
Each callback may alter scan behavior using the following return codes:
* CL_BREAK
Scan aborted by callback (the rest of the scan is skipped).
This does not mark the file as clean or infected, it just skips the rest of the scan.
* CL_SUCCESS / CL_CLEAN
File scan will continue.
This is different than CL_VERIFIED because it does not affect prior or future alerts.
Return CL_VERIFIED instead if you want to remove prior alerts for this layer and skip
the rest of the scan for this layer.
* CL_VIRUS
This means you don't trust the file. A new alert will be added.
For CL_SCAN_CALLBACK_ALERT: Means you agree with the alert (no extra alert needed).
* CL_VERIFIED
Layer explicitly trusted by the callback and previous alerts removed FOR THIS layer.
You might want to do this if you trust the hash or verified a digital signature.
The rest of the scan will be skipped FOR THIS layer.
For contained files, this does NOT mean that the parent or adjacent layers are trusted.
Each callback is given a pointer to the current scan layer from which
they can get previous layers, can get the the layer's fmap, and then
various attributes of the layer and of the fmap such as:
- layer recursion level
- layer object id
- layer file type
- layer attributes (was decerypted, normalized, embedded, or re-typed)
- layer last alert
- fmap name
- fmap hash (md5, sha1, or sha2-256)
- fmap data (pointer and size)
- fmap file descriptor, if any (fd, offset, size)
- fmap filepath, if any (filepath, offset, size)
To make this possible, this commits introduced a handful of new APIs to
query scan-layer details and fmap details:
- `cl_error_t cl_fmap_set_name(cl_fmap_t *map, const char *name);`
- `cl_error_t cl_fmap_get_name(cl_fmap_t *map, const char **name_out);`
- `cl_error_t cl_fmap_set_path(cl_fmap_t *map, const char *path);`
- `cl_error_t cl_fmap_get_path(cl_fmap_t *map, const char **path_out, size_t *offset_out, size_t *len_out);`
- `cl_error_t cl_fmap_get_fd(const cl_fmap_t *map, int *fd_out, size_t *offset_out, size_t *len_out);`
- `cl_error_t cl_fmap_get_size(const cl_fmap_t *map, size_t *size_out);`
- `cl_error_t cl_fmap_set_hash(const cl_fmap_t *map, const char *hash_alg, char hash);`
- `cl_error_t cl_fmap_have_hash(const cl_fmap_t *map, const char *hash_alg, bool *have_hash_out);`
- `cl_error_t cl_fmap_will_need_hash_later(const cl_fmap_t *map, const char *hash_alg);`
- `cl_error_t cl_fmap_get_hash(const cl_fmap_t *map, const char *hash_alg, const char **hash_out);`
- `cl_error_t cl_fmap_get_data(const cl_fmap_t *map, size_t offset, size_t len, const uint8_t **data_out, size_t *data_len_out);`
- `cl_error_t cl_scan_layer_get_fmap(cl_scan_layer_t *layer, cl_fmap_t **fmap_out);`
- `cl_error_t cl_scan_layer_get_parent_layer(cl_scan_layer_t *layer, cl_scan_layer_t **parent_layer_out);`
- `cl_error_t cl_scan_layer_get_type(cl_scan_layer_t *layer, const char **type_out);`
- `cl_error_t cl_scan_layer_get_recursion_level(cl_scan_layer_t *layer, uint32_t *recursion_level_out);`
- `cl_error_t cl_scan_layer_get_object_id(cl_scan_layer_t *layer, uint64_t *object_id_out);`
- `cl_error_t cl_scan_layer_get_last_alert(cl_scan_layer_t *layer, const char **alert_name_out);`
- `cl_error_t cl_scan_layer_get_attributes(cl_scan_layer_t *layer, uint32_t *attributes_out);`
This commit deprecates but does not remove the existing scan callbacks:
- `void cl_engine_set_clcb_pre_cache(struct cl_engine *engine, clcb_pre_cache callback);`
- `void cl_engine_set_clcb_file_inspection(struct cl_engine *engine, clcb_file_inspection callback);`
- `void cl_engine_set_clcb_pre_scan(struct cl_engine *engine, clcb_pre_scan callback);`
- `void cl_engine_set_clcb_post_scan(struct cl_engine *engine, clcb_post_scan callback);`
- `void cl_engine_set_clcb_virus_found(struct cl_engine *engine, clcb_virus_found callback);`
- `void cl_engine_set_clcb_hash(struct cl_engine *engine, clcb_hash callback);`
This commit also adds an interactive test program to demonstrate the callbacks.
See: `examples/ex_scan_callbacks.c`
CLAM-255
CLAM-2485
CLAM-2626
Every time we push a new map onto the scanning recursion context, give
it a unique object id number, which counts from zero.
Moved the location where we add metadata for each file from the
"cli_magic_scan" function over to the "recursion stack push" function.
Include a "path" as a parameter for creating a new fmap, and rename some
related variables and functions to be more intuitive.
CLAM-2796
See also: CLAM-2485, CLAM-2626
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
On Windows, the cli_basename function should treat both '/' and '\' as path
separators. Most Windows APIs also accept both.
On Linux/Unix, it makes sense when using a filepath that is more for
informational purposes or where it may have come from a Windows system,
to treat the '\' as a path separator.
But in situations where the the path is needed for some critical action,
like moving or deleting a file, we can't treat it as a path separator.
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.
The fmap structure has some stuff that differs in size in memory between
Linux and Windows, and between 32bit and 64bit architectures.
Notably, `time_t` appears to be defined by the Rust bindgen module as
`ulong` which may be either 8 bytes or 4 bytes, depending architecture
(thanks, C). To resolve this, we'll store time as a uint64_t instead.
The other problem in the fmap structure is the windows file and map
handles should always be exist, and may only be used in Windows, for
consistency in sizing of the structure.
Significantly tidy the `cli_scan_fmap()` function, and add comments to
explain how it all works.
Add SHA1 and SHA256 digest variables to the FMAP structure in addition
to the existing MD5. Add a function to set the hash so that when we
calculate the hashes for hash matching, we save them for subsequent FP
matching. This enabled me to remove the extra "hash-only" FP check from
`cli_scan_fmap()`. This will also make it easier to switch the clean
cache hash algorithm to SHA256 in the future.
Remove extra allmatch checks that are no longer required.
Add a new header to prevent #include order issues.
Rework the append_virus mechanism to store evidence (strong indicators,
pua indicators, and eventually weak indicators) in vectors. When
appending a "virus", we will return CLEAN when in allmatch-mode, and
simply add the indicator to the appropriate vector.
Later we can check if there were any alerts to return a vector by
summing the lengths of the strong and pua indicator vectors.
This does away with storing the latest "virname" in the scan context.
Instead, we can query for the last indicator in the evidence, giving
priority to strong indicators.
When heuristic-precendence is enabled, add PUA as Strong instead of
as PotentiallyUnwanted. This way, they will be treated equally and
reported in order in allmatch mode.
Also document reason for disabling cache with metadata JSON enabled
* 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
The fixes to the fmap bounds for nested (duplicate) fmaps added recently
introduced a subtle arithmetic bug that was detected by OSS-Fuzz:
```c
scanat = m->nested_offset + *at % m->pgsz;
```
should have been:
```c
scanat = (m->nested_offset + *at) % m->pgsz;
```
Without the parenthesis, `scanat` could be > `m->pgsz`, which would
overflow in the subsequent `memchr()` call.
See:
- https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=40452
- https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=40455
This commit also tightens up some of the other bounds checks done with
`CLI_ISCONTAINED()` macro so the check limits the bounds to the nested
fmap and not the original map.
In addition, I've added a `CLI_ISCONTAINED_0_TO()` macro that removes
checks when the "bigger" buffer starts at offset 0. This should silence
a bunch of (benign) warnings and medium severity Coverity issues.
There is also a possible use of an uninitialized variable
(`old_hook_lsig_matches`) in `cli_magic_scan()`.
Finally, I also removed an unecessary NULL-check on `filebase` in
`fmap_dup_to_file()` that Coverity was unhappy with.
CID 361074: fmap.c: Possible invalid dereference if status != success
and the new map was not yet allocated.
CID 361077: others.c: Structurally dead code revealed a bug in the
cli_recursion_stack_get_size() function.
CID 361080, 361078, 361083: sigtool.c: Inverted check for if engine
needs to be free'd, could leak the engine structure.
CID 361075: sigtool.c: Missed a `return -1` that should've been `goto
done;` and would leak the new_map buffer.
CID 361079: sigtool/vba.c: Checking if we should free the new_map on
failure only if ctx also needs to be free'd, which would leak the
new_map if ctx was not allocated yet.
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.
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.
This commit resolves https://bugzilla.clamav.net/show_bug.cgi?id=12673
Changes in 0.103 to order of operations for creating fmaps and
performaing hashes of fmaps resulted errors when scanning files that are
4096M and a different (but related) error when scanning files > 4096M.
This is despite the fact that scanning is supposed to be limited to
--max-scansize (MaxScanSize) and was also apparently limited to
INT_MAX - 2 (aka ~1.999999G) back in 2014 to alleviate reported crashes
for a few large file formats.
(see https://bugzilla.clamav.net/show_bug.cgi?id=10960)
This last limitation was not documented, so I added it to the sample
clamd.conf.
Anyways, the main issue is that the fmap module was using "unsigned int"
and was then enforcing a limitation (verbose error messages) when that
a map length exceeded the capapacity of an unsigned int. This commit
switches the associated variables over to uint64_t, and while fmaps are
still limited to size_t in other places, the fmap module will at least
work with files > 4G on 64bit systems.
In testing this, I found that the time to hash a file, particularly when
hashing a file on an NTFS partition from Linux was really slow because
we were hashing in FILEBUFF chunks (about 8K) at a time. Increasing
this to 10MB chunks speeds up scanning of large files.
Finally, now that hashing is performed immediately when an fmap is
created for a file, hashing of files larger than max-scansize was
occuring. This commit adds checks to bail out early if the file size
exceeds the maximum before creating an fmap. It will alert with the
Heuristics.Limits.Exceeded name if the heuristic is enabled.
Also fixed CheckFmapFeatures.cmake module that detects if
sysconf(_SC_PAGESIZE) is available.
The fmap_duplicate function is used create a new fmap with a view into
an existing fmap. When the new view is a different size than the old
fmap, a new hash must be calculated for the duplicate fmap. However,
when the duplicated fmap is the same size as the original fmap, the hash
will be the same and there's no point recalculating.
The issue is apparent when scanning large EXE files because the hash was
being calculated at the beginning and end of the scan.
Digging into this issue revealed that hash calculations for fmaps were
also being performed at the wrong place. For scans of maps we use
fmap_duplicate() early in the process to apply the name API argument to
the duplicate fmap. Fixing the logic so we doing recalculate the hash
revealed that we never calculated hashes for fmap's created from buffers
in the first place, so that also had to be fixed be relocating where the
hash is calculated.
I also found that fmap_duplicate()'s offset argument used an off_t,
though it and all caller offsets are not allowed to be negative. This
was a bit of tangent to fix a bunch of off_t variables and paramters
that should've been size_t.
Added a couple unit tests to verify that making duplicate fmaps, and
duplicate-duplicate fmaps works as expected after the change.
Changed CLI_ISCONTAINED() and CLI_ISCONTAINED2() macros to cast to
size_t, because pointers and buffer sizes may not be negative, and these
two macros do not rely on substraction.
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.
If using a bytecode signature that makes use of the BC_PRECLASS hook and
if it alerts, a NULL dereference may occur. This change fixes that.
Also fixed unrelated memory leaks introduced recently when adding file
name extraction to the zip parser and rar parser.
Fixed copypaste bug with duplicated fmap names being assigned to the
parent instead of the dup/child fmap.
Fixed file descriptor initialization issue in the HTML normalizer.
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.
Specifically this fixes use of cli_map_scandesc().
The cli_map_scandesc() function used to override the current fmap
settings with a new size and offset, performing a scan of the embedded
content. This broke the ability to iterate backwards through the fmap
recursion array when an alert occurs to check each map's hash for
whitelist matches.
In order to fix this issue, it needed to be possible to duplicate an
fmap header for the scan of the embedded file without duplicating the
actual map/data. This wasn't feasible with the posix fmap handle
implementation where the fmap header, bitmap array, and memory map
were all contiguouus. This commit makes it possible by extracting the
fmap header and bitmap array from the mmap region, using instead a
pointer for both the bitmap array and mmap/data. The resulting posix
fmap handle implementation as a result ended up working more similarly
to existing the Windows implementation.
In addition to the above changes, this commit fixes:
- fmap recursion tracking for cli_scandesc()
- a recursion tracking issue in cli_scanembpe() error handling
Signature alerts on content extracted into a new fmap such as normalized
HTML resulted in checking FP signatures against the fmap's hash value
that was initialized to all zeroes, and never computed.
This patch seeks will enable FP signatures of normalized HTML files or
other content that is extracted to a new fmap to work. This patch
doesn't resolve the issue that normal people will write FP signatures
targeting the original file, not the normalized file and thus won't
really see benefit from this bug-fix.
Additional work is needed to traverse the fmap recursion lists and
FP-check all parent fmaps when an alert occurs. In addition, the HTML
normalization method of temporarily overriding the ctx->fmap instead of
increasing the recursion depth and doing ctx->fmap++/-- will need to be
corrected for fmap reverse recursion traversal to work.
Fix for minor memory leak in fmap_dump_to_file().
Fix to PDF object stream logic, accounting for a realloc() issue when the only pdf object stream fails to parse, and for when pdf objects in a stream appear to extend further than the size of the stream.
Fix for memory leak cleaning up PDF object stream buffer in error condition.
Fix to bug in pdf_decodestream wherein objects were found in an object stream, but the object stream could later be free'd if max scansize was exceeded, resulting in a NULL dereference.
General cleanup of pdf_decodestream/pdf_decodestream_internal exit code logic.
