ClamAV is an open source (GPLv2) anti-virus toolkit.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
clamav/libclamav/ole2_extract.c

1932 lines
66 KiB

/*
* Copyright (C) 2013-2022 Cisco Systems, Inc. and/or its affiliates. All rights reserved.
* Copyright (C) 2007-2013 Sourcefire, Inc.
*
* Authors: Trog
*
* Summary: Extract component parts of OLE2 files (e.g. MS Office Documents).
*
* Acknowledgements: Some ideas and algorithms were based upon OpenOffice and libgsf.
*
* 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.
*
* 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.
*/
#if HAVE_CONFIG_H
#include "clamav-config.h"
#endif
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <stdio.h>
#include <string.h>
#include <ctype.h>
#include <stdlib.h>
#include <errno.h>
#include <conv.h>
#include <zlib.h>
#ifdef HAVE_UNISTD_H
#include <unistd.h>
#endif
#include <stdbool.h>
#include "clamav.h"
#include "others.h"
#include "hwp.h"
#include "ole2_extract.h"
#include "xlm_extract.h"
#include "scanners.h"
#include "fmap.h"
#include "json_api.h"
#if HAVE_JSON
#include "msdoc.h"
#endif
#ifdef DEBUG_OLE2_LIST
#define ole2_listmsg(...) cli_dbgmsg(__VA_ARGS__)
#else
#define ole2_listmsg(...) ;
#endif
#define ole2_endian_convert_16(v) le16_to_host((uint16_t)(v))
#define ole2_endian_convert_32(v) le32_to_host((uint32_t)(v))
#ifndef HAVE_ATTRIB_PACKED
#define __attribute__(x)
#endif
#ifdef HAVE_PRAGMA_PACK
#pragma pack(1)
#endif
#ifdef HAVE_PRAGMA_PACK_HPPA
#pragma pack 1
#endif
12 years ago
typedef struct ole2_header_tag {
unsigned char magic[8]; /* should be: 0xd0cf11e0a1b11ae1 */
unsigned char clsid[16];
12 years ago
uint16_t minor_version __attribute__((packed));
uint16_t dll_version __attribute__((packed));
int16_t byte_order __attribute__((packed)); /* -2=intel */
uint16_t log2_big_block_size __attribute__((packed)); /* usually 9 (2^9 = 512) */
uint32_t log2_small_block_size __attribute__((packed)); /* usually 6 (2^6 = 64) */
12 years ago
int32_t reserved[2] __attribute__((packed));
12 years ago
int32_t bat_count __attribute__((packed));
int32_t prop_start __attribute__((packed));
uint32_t signature __attribute__((packed));
uint32_t sbat_cutoff __attribute__((packed)); /* cutoff for files held
* in small blocks
* (4096) */
12 years ago
int32_t sbat_start __attribute__((packed));
int32_t sbat_block_count __attribute__((packed));
int32_t xbat_start __attribute__((packed));
int32_t xbat_count __attribute__((packed));
int32_t bat_array[109] __attribute__((packed));
12 years ago
/*
* The following is not part of the ole2 header, but stuff we need in
* order to decode.
*
* IMPORANT: These must take account of the size of variables below here
* when calculating hdr_size to read the header.
*
* See the top of cli_ole2_extract().
12 years ago
*/
int32_t sbat_root_start __attribute__((packed));
uint32_t max_block_no;
off_t m_length;
bitset_t *bitset;
struct uniq *U;
fmap_t *map;
bool has_vba;
bool has_xlm;
bool has_image;
hwp5_header_t *is_hwp;
} ole2_header_t;
12 years ago
typedef struct property_tag {
char name[64]; /* in unicode */
12 years ago
uint16_t name_size __attribute__((packed));
unsigned char type; /* 1=dir 2=file 5=root */
unsigned char color; /* black or red */
uint32_t prev __attribute__((packed));
uint32_t next __attribute__((packed));
uint32_t child __attribute__((packed));
12 years ago
unsigned char clsid[16];
12 years ago
uint32_t user_flags __attribute__((packed));
uint32_t create_lowdate __attribute__((packed));
uint32_t create_highdate __attribute__((packed));
uint32_t mod_lowdate __attribute__((packed));
uint32_t mod_highdate __attribute__((packed));
uint32_t start_block __attribute__((packed));
uint32_t size __attribute__((packed));
unsigned char reserved[4];
} property_t;
12 years ago
struct ole2_list_node;
typedef struct ole2_list_node {
uint32_t Val;
struct ole2_list_node *Next;
} ole2_list_node_t;
typedef struct ole2_list {
uint32_t Size;
ole2_list_node_t *Head;
} ole2_list_t;
int ole2_list_init(ole2_list_t *list);
int ole2_list_is_empty(ole2_list_t *list);
uint32_t ole2_list_size(ole2_list_t *list);
int ole2_list_push(ole2_list_t *list, uint32_t val);
uint32_t ole2_list_pop(ole2_list_t *list);
int ole2_list_delete(ole2_list_t *list);
int ole2_list_init(ole2_list_t *list)
{
12 years ago
list->Head = NULL;
list->Size = 0;
return CL_SUCCESS;
}
int ole2_list_is_empty(ole2_list_t *list)
{
12 years ago
return (list->Head == NULL);
}
12 years ago
uint32_t
ole2_list_size(ole2_list_t *list)
{
12 years ago
return (list->Size);
}
int ole2_list_push(ole2_list_t *list, uint32_t val)
{
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.
4 years ago
ole2_list_node_t *new_node = NULL;
int status = CL_EMEM;
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.
4 years ago
CLI_MALLOC(new_node, sizeof(ole2_list_node_t),
cli_dbgmsg("OLE2: could not allocate new node for worklist!\n"));
new_node->Val = val;
12 years ago
new_node->Next = list->Head;
12 years ago
list->Head = new_node;
(list->Size)++;
status = CL_SUCCESS;
done:
return status;
}
12 years ago
uint32_t
ole2_list_pop(ole2_list_t *list)
{
uint32_t val;
ole2_list_node_t *next;
12 years ago
if (ole2_list_is_empty(list)) {
cli_dbgmsg("OLE2: work list is empty and ole2_list_pop() called!\n");
return -1;
}
val = list->Head->Val;
next = list->Head->Next;
12 years ago
free(list->Head);
list->Head = next;
12 years ago
(list->Size)--;
return val;
}
int ole2_list_delete(ole2_list_t *list)
{
12 years ago
while (!ole2_list_is_empty(list))
ole2_list_pop(list);
return CL_SUCCESS;
}
#ifdef HAVE_PRAGMA_PACK
#pragma pack()
#endif
#ifdef HAVE_PRAGMA_PACK_HPPA
#pragma pack
#endif
12 years ago
static unsigned char magic_id[] = {0xd0, 0xcf, 0x11, 0xe0, 0xa1, 0xb1, 0x1a, 0xe1};
char *
cli_ole2_get_property_name2(const char *name, int size)
{
int i, j;
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.
4 years ago
char *newname = NULL;
12 years ago
if ((name[0] == 0 && name[1] == 0) || size <= 0 || size > 128) {
12 years ago
return 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.
4 years ago
CLI_MALLOC(newname, size * 7,
cli_errmsg("OLE2 [cli_ole2_get_property_name2]: Unable to allocate memory for newname: %u\n", size * 7));
12 years ago
j = 0;
/* size-2 to ignore trailing NULL */
for (i = 0; i < size - 2; i += 2) {
if ((!(name[i] & 0x80)) && isprint(name[i]) && name[i + 1] == 0) {
12 years ago
newname[j++] = tolower(name[i]);
} else {
if (name[i] < 10 && name[i] >= 0 && name[i + 1] == 0) {
12 years ago
newname[j++] = '_';
newname[j++] = name[i] + '0';
} else {
const uint16_t x = (((uint16_t)name[i]) << 8) | name[i + 1];
12 years ago
newname[j++] = '_';
newname[j++] = 'a' + ((x & 0xF));
newname[j++] = 'a' + ((x >> 4) & 0xF);
newname[j++] = 'a' + ((x >> 8) & 0xF);
newname[j++] = 'a' + ((x >> 16) & 0xF);
newname[j++] = 'a' + ((x >> 24) & 0xF);
}
newname[j++] = '_';
}
}
newname[j] = '\0';
if (strlen(newname) == 0) {
free(newname);
newname = NULL;
12 years ago
}
done:
12 years ago
return newname;
}
static char *
12 years ago
get_property_name(char *name, int size)
{
const char *carray = "0123456789abcdefghijklmnopqrstuvwxyzabcdefghijklmnopqrstuvwxyz._";
int csize = size >> 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.
4 years ago
char *newname = NULL;
char *cname = NULL;
char *oname = name;
12 years ago
if (csize <= 0) {
12 years ago
return 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.
4 years ago
CLI_MALLOC(newname, size,
cli_errmsg("OLE2 [get_property_name]: Unable to allocate memory for newname %u\n", size));
cname = newname;
12 years ago
while (--csize) {
uint16_t lo, hi, u = cli_readint16(oname) - 0x3800;
12 years ago
oname += 2;
if (u > 0x1040) {
FREE(newname);
return cli_ole2_get_property_name2(name, size);
12 years ago
}
lo = u % 64;
u >>= 6;
hi = u % 64;
12 years ago
*cname++ = carray[lo];
if (csize != 1 || u != 64) {
12 years ago
*cname++ = carray[hi];
}
12 years ago
}
*cname = '\0';
done:
12 years ago
return newname;
}
12 years ago
static void
print_ole2_property(property_t *property)
{
char spam[128], *buf;
12 years ago
if (property->name_size > 64) {
cli_dbgmsg("[err name len: %d]\n", property->name_size);
return;
}
buf = get_property_name(property->name, property->name_size);
snprintf(spam, sizeof(spam), "OLE2: %s ", buf ? buf : "<noname>");
spam[sizeof(spam) - 1] = '\0';
if (buf)
free(buf);
switch (property->type) {
case 2:
strncat(spam, " [file] ", sizeof(spam) - 1 - strlen(spam));
break;
case 1:
strncat(spam, " [dir ] ", sizeof(spam) - 1 - strlen(spam));
break;
case 5:
strncat(spam, " [root] ", sizeof(spam) - 1 - strlen(spam));
break;
default:
strncat(spam, " [unkn] ", sizeof(spam) - 1 - strlen(spam));
12 years ago
}
spam[sizeof(spam) - 1] = '\0';
switch (property->color) {
case 0:
strncat(spam, " r ", sizeof(spam) - 1 - strlen(spam));
break;
case 1:
strncat(spam, " b ", sizeof(spam) - 1 - strlen(spam));
break;
default:
strncat(spam, " u ", sizeof(spam) - 1 - strlen(spam));
12 years ago
}
spam[sizeof(spam) - 1] = '\0';
cli_dbgmsg("%s size:0x%.8x flags:0x%.8x\n", spam, property->size, property->user_flags);
}
12 years ago
static void
print_ole2_header(ole2_header_t *hdr)
{
12 years ago
if (!hdr || !cli_debug_flag) {
return;
}
cli_dbgmsg("\n");
cli_dbgmsg("Magic:\t\t\t0x%x%x%x%x%x%x%x%x\n",
hdr->magic[0], hdr->magic[1], hdr->magic[2], hdr->magic[3],
hdr->magic[4], hdr->magic[5], hdr->magic[6], hdr->magic[7]);
12 years ago
cli_dbgmsg("CLSID:\t\t\t{%x%x%x%x-%x%x-%x%x-%x%x-%x%x%x%x%x%x}\n",
hdr->clsid[0], hdr->clsid[1], hdr->clsid[2], hdr->clsid[3],
hdr->clsid[4], hdr->clsid[5], hdr->clsid[6], hdr->clsid[7],
hdr->clsid[8], hdr->clsid[9], hdr->clsid[10], hdr->clsid[11],
hdr->clsid[12], hdr->clsid[13], hdr->clsid[14], hdr->clsid[15]);
12 years ago
cli_dbgmsg("Minor version:\t\t0x%x\n", hdr->minor_version);
cli_dbgmsg("DLL version:\t\t0x%x\n", hdr->dll_version);
cli_dbgmsg("Byte Order:\t\t%d\n", hdr->byte_order);
cli_dbgmsg("Big Block Size:\t%i\n", hdr->log2_big_block_size);
12 years ago
cli_dbgmsg("Small Block Size:\t%i\n", hdr->log2_small_block_size);
cli_dbgmsg("BAT count:\t\t%d\n", hdr->bat_count);
cli_dbgmsg("Prop start:\t\t%d\n", hdr->prop_start);
cli_dbgmsg("SBAT cutoff:\t\t%d\n", hdr->sbat_cutoff);
cli_dbgmsg("SBat start:\t\t%d\n", hdr->sbat_start);
cli_dbgmsg("SBat block count:\t%d\n", hdr->sbat_block_count);
cli_dbgmsg("XBat start:\t\t%d\n", hdr->xbat_start);
cli_dbgmsg("XBat block count:\t%d\n", hdr->xbat_count);
cli_dbgmsg("\n");
12 years ago
return;
}
12 years ago
static int
ole2_read_block(ole2_header_t *hdr, void *buff, unsigned int size, int32_t blockno)
{
off_t offset, offend;
const void *pblock;
12 years ago
if (blockno < 0) {
return FALSE;
}
/* other methods: (blockno+1) * 512 or (blockno * block_size) + 512; */
if (((uint64_t)blockno << hdr->log2_big_block_size) < (INT32_MAX - MAX(512, (uint64_t)1 << hdr->log2_big_block_size))) {
/* 512 is header size */
offset = (blockno << hdr->log2_big_block_size) + MAX(512, 1 << hdr->log2_big_block_size);
offend = offset + size;
} else {
offset = INT32_MAX - size;
offend = INT32_MAX;
}
12 years ago
if ((offend <= 0) || (offset < 0) || (offset >= hdr->m_length)) {
12 years ago
return FALSE;
} else if (offend > hdr->m_length) {
/* bb#11369 - ole2 files may not be a block multiple in size */
memset(buff, 0, size);
size = hdr->m_length - offset;
12 years ago
}
if (!(pblock = fmap_need_off_once(hdr->map, offset, size))) {
return FALSE;
}
memcpy(buff, pblock, size);
return TRUE;
}
static int32_t
ole2_get_next_bat_block(ole2_header_t *hdr, int32_t current_block)
{
int32_t bat_array_index;
uint32_t bat[128];
12 years ago
if (current_block < 0) {
return -1;
}
bat_array_index = current_block / 128;
if (bat_array_index > hdr->bat_count) {
cli_dbgmsg("bat_array index error\n");
return -10;
}
if (!ole2_read_block(hdr, &bat, 512,
ole2_endian_convert_32(hdr->bat_array[bat_array_index]))) {
12 years ago
return -1;
}
return ole2_endian_convert_32(bat[current_block - (bat_array_index * 128)]);
}
static int32_t
ole2_get_next_xbat_block(ole2_header_t *hdr, int32_t current_block)
{
int32_t xbat_index, xbat_block_index, bat_index, bat_blockno;
uint32_t xbat[128], bat[128];
12 years ago
if (current_block < 0) {
return -1;
}
xbat_index = current_block / 128;
/*
* NB: The last entry in each XBAT points to the next XBAT block.
* This reduces the number of entries in each block by 1.
*/
xbat_block_index = (xbat_index - 109) / 127;
bat_blockno = (xbat_index - 109) % 127;
12 years ago
bat_index = current_block % 128;
if (!ole2_read_block(hdr, &xbat, 512, hdr->xbat_start)) {
return -1;
}
/* Follow the chain of XBAT blocks */
while (xbat_block_index > 0) {
if (!ole2_read_block(hdr, &xbat, 512,
ole2_endian_convert_32(xbat[127]))) {
return -1;
}
xbat_block_index--;
}
if (!ole2_read_block(hdr, &bat, 512, ole2_endian_convert_32(xbat[bat_blockno]))) {
return -1;
}
return ole2_endian_convert_32(bat[bat_index]);
}
static int32_t
ole2_get_next_block_number(ole2_header_t *hdr, int32_t current_block)
{
12 years ago
if (current_block < 0) {
return -1;
}
if ((current_block / 128) > 108) {
return ole2_get_next_xbat_block(hdr, current_block);
} else {
return ole2_get_next_bat_block(hdr, current_block);
}
}
static int32_t
ole2_get_next_sbat_block(ole2_header_t *hdr, int32_t current_block)
{
int32_t iter, current_bat_block;
uint32_t sbat[128];
12 years ago
if (current_block < 0) {
return -1;
}
current_bat_block = hdr->sbat_start;
iter = current_block / 128;
12 years ago
while (iter > 0) {
current_bat_block = ole2_get_next_block_number(hdr, current_bat_block);
iter--;
}
if (!ole2_read_block(hdr, &sbat, 512, current_bat_block)) {
return -1;
}
return ole2_endian_convert_32(sbat[current_block % 128]);
}
/* Retrieve the block containing the data for the given sbat index */
static int32_t
ole2_get_sbat_data_block(ole2_header_t *hdr, void *buff, int32_t sbat_index)
{
int32_t block_count, current_block;
12 years ago
if (sbat_index < 0) {
return FALSE;
}
if (hdr->sbat_root_start < 0) {
cli_dbgmsg("No root start block\n");
return FALSE;
}
block_count = sbat_index / (1 << (hdr->log2_big_block_size - hdr->log2_small_block_size));
12 years ago
current_block = hdr->sbat_root_start;
while (block_count > 0) {
current_block = ole2_get_next_block_number(hdr, current_block);
block_count--;
}
/*
* current_block now contains the block number of the sbat array
* containing the entry for the required small block
*/
return (ole2_read_block(hdr, buff, 1 << hdr->log2_big_block_size, current_block));
}
/**
* @brief File handler for use when walking ole2 property trees.
*
* @param hdr The ole2 header metadata
* @param prop The property
* @param dir (optional) directory to write temp files to.
* @param ctx The scan context
* @return cl_error_t
*/
typedef cl_error_t ole2_walk_property_tree_file_handler(ole2_header_t *hdr, property_t *prop, const char *dir, cli_ctx *ctx);
/**
* @brief Walk an ole2 property tree, calling the handler for each file found
*
* @param hdr The ole2 header metadata (an ole2-specific context struct)
* @param dir (optional) directory to write temp files to, passed to the handler.
* @param prop_index Index of the property being walked, to be recorded with a pointer to the root node in an ole2 node list.
* @param handler The file handler to call when a file is found.
* @param rec_level The recursion level. Max is 100.
* @param[in,out] file_count A running count of the total # of files. Max is 100000.
* @param ctx The scan context
* @param[in,out] scansize A running sum of the file sizes processed.
* @return int
*/
static int ole2_walk_property_tree(ole2_header_t *hdr, const char *dir, int32_t prop_index,
ole2_walk_property_tree_file_handler handler,
unsigned int rec_level, unsigned int *file_count, cli_ctx *ctx, unsigned long *scansize)
{
property_t prop_block[4];
int32_t idx, current_block, i, curindex;
char *dirname;
ole2_list_t node_list;
int ret, func_ret;
#if HAVE_JSON
char *name;
int toval = 0;
#endif
12 years ago
ole2_listmsg("ole2_walk_property_tree() called\n");
func_ret = CL_SUCCESS;
12 years ago
ole2_list_init(&node_list);
12 years ago
ole2_listmsg("rec_level: %d\n", rec_level);
ole2_listmsg("file_count: %d\n", *file_count);
12 years ago
if ((rec_level > 100) || (*file_count > 100000)) {
return CL_SUCCESS;
}
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.
4 years ago
if (ctx && ctx->engine->max_recursion_level && (rec_level > ctx->engine->max_recursion_level)) {
// Note: engine->max_recursion_level is re-purposed here out of convenience.
// ole2 recursion does not leverage the ctx->recursion_stack stack.
cli_dbgmsg("OLE2: Recursion limit reached (max: %d)\n", ctx->engine->max_recursion_level);
cli_append_virus_if_heur_exceedsmax(ctx, "Heuristics.Limits.Exceeded.MaxRecursion");
return CL_EMAXREC;
12 years ago
}
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.
4 years ago
// push the 'root' node for the level onto the local list
if ((ret = ole2_list_push(&node_list, prop_index)) != CL_SUCCESS) {
ole2_list_delete(&node_list);
return ret;
}
12 years ago
while (!ole2_list_is_empty(&node_list)) {
ole2_listmsg("within working loop, worklist size: %d\n", ole2_list_size(&node_list));
#if HAVE_JSON
if (cli_json_timeout_cycle_check(ctx, &toval) != CL_SUCCESS) {
ole2_list_delete(&node_list);
return CL_ETIMEOUT;
}
#endif
12 years ago
current_block = hdr->prop_start;
// pop off a node to work on
curindex = ole2_list_pop(&node_list);
12 years ago
ole2_listmsg("current index: %d\n", curindex);
if ((curindex < 0) || (curindex > (int32_t)hdr->max_block_no)) {
12 years ago
continue;
}
// read in the sector referenced by the current index
idx = curindex / 4;
12 years ago
for (i = 0; i < idx; i++) {
current_block = ole2_get_next_block_number(hdr, current_block);
if (current_block < 0) {
continue;
12 years ago
}
}
12 years ago
idx = curindex % 4;
if (!ole2_read_block(hdr, prop_block, 512, current_block)) {
continue;
12 years ago
}
if (prop_block[idx].type <= 0) {
continue;
12 years ago
}
ole2_listmsg("reading prop block\n");
prop_block[idx].name_size = ole2_endian_convert_16(prop_block[idx].name_size);
prop_block[idx].prev = ole2_endian_convert_32(prop_block[idx].prev);
prop_block[idx].next = ole2_endian_convert_32(prop_block[idx].next);
prop_block[idx].child = ole2_endian_convert_32(prop_block[idx].child);
prop_block[idx].user_flags = ole2_endian_convert_32(prop_block[idx].user_flags);
prop_block[idx].create_lowdate = ole2_endian_convert_32(prop_block[idx].create_lowdate);
12 years ago
prop_block[idx].create_highdate = ole2_endian_convert_32(prop_block[idx].create_highdate);
prop_block[idx].mod_lowdate = ole2_endian_convert_32(prop_block[idx].mod_lowdate);
prop_block[idx].mod_highdate = ole2_endian_convert_32(prop_block[idx].mod_highdate);
prop_block[idx].start_block = ole2_endian_convert_32(prop_block[idx].start_block);
prop_block[idx].size = ole2_endian_convert_32(prop_block[idx].size);
12 years ago
ole2_listmsg("printing ole2 property\n");
if (dir)
print_ole2_property(&prop_block[idx]);
ole2_listmsg("checking bitset\n");
/* Check we aren't in a loop */
if (cli_bitset_test(hdr->bitset, (unsigned long)curindex)) {
/* Loop in property tree detected */
cli_dbgmsg("OLE2: Property tree loop detected at index %d\n", curindex);
ole2_list_delete(&node_list);
return CL_BREAK;
}
ole2_listmsg("setting bitset\n");
if (!cli_bitset_set(hdr->bitset, (unsigned long)curindex)) {
continue;
12 years ago
}
ole2_listmsg("prev: %d next %d child %d\n", prop_block[idx].prev, prop_block[idx].next, prop_block[idx].child);
ole2_listmsg("node type: %d\n", prop_block[idx].type);
switch (prop_block[idx].type) {
case 5: /* Root Entry */
ole2_listmsg("root node\n");
if ((curindex != 0) || (rec_level != 0) ||
12 years ago
(*file_count != 0)) {
/* Can only have RootEntry as the top */
cli_dbgmsg("ERROR: illegal Root Entry\n");
continue;
}
hdr->sbat_root_start = prop_block[idx].start_block;
if ((int)(prop_block[idx].child) != -1) {
ret = ole2_walk_property_tree(hdr, dir, prop_block[idx].child, handler, rec_level + 1, file_count, ctx, scansize);
if (ret != CL_SUCCESS) {
if (SCAN_ALLMATCHES && (ret == CL_VIRUS)) {
func_ret = ret;
} else {
ole2_list_delete(&node_list);
return ret;
}
}
}
if ((int)(prop_block[idx].prev) != -1) {
if ((ret = ole2_list_push(&node_list, prop_block[idx].prev)) != CL_SUCCESS) {
ole2_list_delete(&node_list);
return ret;
}
}
if ((int)(prop_block[idx].next) != -1) {
if ((ret = ole2_list_push(&node_list, prop_block[idx].next)) != CL_SUCCESS) {
ole2_list_delete(&node_list);
return ret;
}
}
break;
case 2: /* File */
ole2_listmsg("file node\n");
if (ctx && ctx->engine->maxfiles && ((*file_count > ctx->engine->maxfiles) || (ctx->scannedfiles > ctx->engine->maxfiles - *file_count))) {
cli_dbgmsg("OLE2: files limit reached (max: %u)\n", ctx->engine->maxfiles);
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.
4 years ago
cli_append_virus_if_heur_exceedsmax(ctx, "Heuristics.Limits.Exceeded.MaxFiles");
ole2_list_delete(&node_list);
return CL_EMAXFILES;
}
if (!ctx || !(ctx->engine->maxfilesize) || prop_block[idx].size <= ctx->engine->maxfilesize || prop_block[idx].size <= *scansize) {
(*file_count)++;
*scansize -= prop_block[idx].size;
ole2_listmsg("running file handler\n");
ret = handler(hdr, &prop_block[idx], dir, ctx);
if (ret != CL_SUCCESS) {
if (SCAN_ALLMATCHES && (ret == CL_VIRUS)) {
func_ret = ret;
} else {
ole2_listmsg("file handler returned %d\n", ret);
ole2_list_delete(&node_list);
return ret;
}
}
} else {
cli_dbgmsg("OLE2: filesize exceeded\n");
}
if ((int)(prop_block[idx].child) != -1) {
ret = ole2_walk_property_tree(hdr, dir, prop_block[idx].child, handler, rec_level, file_count, ctx, scansize);
if (ret != CL_SUCCESS) {
if (SCAN_ALLMATCHES && (ret == CL_VIRUS)) {
func_ret = ret;
} else {
ole2_list_delete(&node_list);
return ret;
}
}
}
if ((int)(prop_block[idx].prev) != -1) {
if ((ret = ole2_list_push(&node_list, prop_block[idx].prev)) != CL_SUCCESS) {
ole2_list_delete(&node_list);
return ret;
}
}
if ((int)(prop_block[idx].next) != -1) {
if ((ret = ole2_list_push(&node_list, prop_block[idx].next)) != CL_SUCCESS) {
ole2_list_delete(&node_list);
return ret;
}
}
break;
case 1: /* Directory */
ole2_listmsg("directory node\n");
if (dir) {
#if HAVE_JSON
if (SCAN_COLLECT_METADATA && (ctx->wrkproperty != NULL)) {
if (!json_object_object_get_ex(ctx->wrkproperty, "DigitalSignatures", NULL)) {
name = cli_ole2_get_property_name2(prop_block[idx].name, prop_block[idx].name_size);
if (name) {
if (!strcmp(name, "_xmlsignatures") || !strcmp(name, "_signatures")) {
cli_jsonbool(ctx->wrkproperty, "HasDigitalSignatures", 1);
}
free(name);
}
}
}
#endif
dirname = (char *)cli_malloc(strlen(dir) + 8);
if (!dirname) {
ole2_listmsg("OLE2: malloc failed for dirname\n");
ole2_list_delete(&node_list);
return CL_EMEM;
}
snprintf(dirname, strlen(dir) + 8, "%s" PATHSEP "%.6d", dir, curindex);
if (mkdir(dirname, 0700) != 0) {
ole2_listmsg("OLE2: mkdir failed for directory %s\n", dirname);
free(dirname);
ole2_list_delete(&node_list);
return CL_BREAK;
}
cli_dbgmsg("OLE2 dir entry: %s\n", dirname);
} else
dirname = NULL;
if ((int)(prop_block[idx].child) != -1) {
ret = ole2_walk_property_tree(hdr, dirname, prop_block[idx].child, handler, rec_level + 1, file_count, ctx, scansize);
if (ret != CL_SUCCESS) {
if (SCAN_ALLMATCHES && (ret == CL_VIRUS)) {
func_ret = ret;
} else {
ole2_list_delete(&node_list);
if (dirname)
free(dirname);
return ret;
}
}
}
if (dirname) {
free(dirname);
dirname = NULL;
}
if ((int)(prop_block[idx].prev) != -1) {
if ((ret = ole2_list_push(&node_list, prop_block[idx].prev)) != CL_SUCCESS) {
ole2_list_delete(&node_list);
return ret;
}
}
if ((int)(prop_block[idx].next) != -1) {
if ((ret = ole2_list_push(&node_list, prop_block[idx].next)) != CL_SUCCESS) {
ole2_list_delete(&node_list);
return ret;
}
}
break;
default:
cli_dbgmsg("ERROR: unknown OLE2 entry type: %d\n", prop_block[idx].type);
break;
12 years ago
}
ole2_listmsg("loop ended: %d %d\n", ole2_list_size(&node_list), ole2_list_is_empty(&node_list));
}
ole2_list_delete(&node_list);
return func_ret;
}
/* Write file Handler - write the contents of the entry to a file */
static cl_error_t handler_writefile(ole2_header_t *hdr, property_t *prop, const char *dir, cli_ctx *ctx)
{
cl_error_t ret = CL_BREAK;
char newname[1024];
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.
4 years ago
char *name = NULL;
unsigned char *buff = NULL;
int32_t current_block = 0;
size_t len = 0, offset = 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.
4 years ago
int ofd = -1;
char *hash = NULL;
bitset_t *blk_bitset = NULL;
uint32_t cnt = 0;
12 years ago
UNUSEDPARAM(ctx);
12 years ago
if (prop->type != 2) {
/* Not a file */
ret = CL_SUCCESS;
goto done;
12 years ago
}
12 years ago
if (prop->name_size > 64) {
cli_dbgmsg("OLE2 [handler_writefile]: property name too long: %d\n", prop->name_size);
ret = CL_SUCCESS;
goto done;
12 years ago
}
name = cli_ole2_get_property_name2(prop->name, prop->name_size);
if (name) {
cli_dbgmsg("Storing %s in uniq\n", name);
if (CL_SUCCESS != uniq_add(hdr->U, name, strlen(name), &hash, &cnt)) {
cli_dbgmsg("OLE2 [handler_writefile]: too many property names added to uniq store.\n");
goto done;
}
} else {
if (CL_SUCCESS != uniq_add(hdr->U, NULL, 0, &hash, &cnt)) {
cli_dbgmsg("OLE2 [handler_writefile]: too many property names added to uniq store.\n");
goto done;
}
}
12 years ago
snprintf(newname, sizeof(newname), "%s" PATHSEP "%s_%u", dir, hash, cnt);
newname[sizeof(newname) - 1] = '\0';
cli_dbgmsg("OLE2 [handler_writefile]: Dumping '%s' to '%s'\n", name ? name : "<empty>", newname);
ofd = open(newname, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, S_IRUSR | S_IWUSR);
12 years ago
if (ofd < 0) {
cli_errmsg("OLE2 [handler_writefile]: failed to create file: %s\n", newname);
ret = CL_SUCCESS;
goto done;
12 years ago
}
12 years ago
current_block = prop->start_block;
len = prop->size;
12 years ago
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.
4 years ago
CLI_MALLOC(buff, 1 << hdr->log2_big_block_size,
cli_errmsg("OLE2 [handler_writefile]: Unable to allocate memory for buff: %u\n", 1 << hdr->log2_big_block_size);
ret = CL_EMEM);
12 years ago
blk_bitset = cli_bitset_init();
if (!blk_bitset) {
cli_errmsg("OLE2 [handler_writefile]: init bitset failed\n");
goto done;
12 years ago
}
12 years ago
while ((current_block >= 0) && (len > 0)) {
if (current_block > (int32_t)hdr->max_block_no) {
12 years ago
cli_dbgmsg("OLE2 [handler_writefile]: Max block number for file size exceeded: %d\n", current_block);
break;
12 years ago
}
12 years ago
/* Check we aren't in a loop */
if (cli_bitset_test(blk_bitset, (unsigned long)current_block)) {
/* Loop in block list */
cli_dbgmsg("OLE2 [handler_writefile]: Block list loop detected\n");
break;
12 years ago
}
12 years ago
if (!cli_bitset_set(blk_bitset, (unsigned long)current_block)) {
break;
12 years ago
}
if (prop->size < (int64_t)hdr->sbat_cutoff) {
12 years ago
/* Small block file */
if (!ole2_get_sbat_data_block(hdr, buff, current_block)) {
cli_dbgmsg("OLE2 [handler_writefile]: ole2_get_sbat_data_block failed\n");
break;
12 years ago
}
12 years ago
/* buff now contains the block with N small blocks in it */
offset = (1 << hdr->log2_small_block_size) * (current_block % (1 << (hdr->log2_big_block_size - hdr->log2_small_block_size)));
if (cli_writen(ofd, &buff[offset], MIN(len, 1 << hdr->log2_small_block_size)) != MIN(len, 1 << hdr->log2_small_block_size)) {
goto done;
12 years ago
}
12 years ago
len -= MIN(len, 1 << hdr->log2_small_block_size);
current_block = ole2_get_next_sbat_block(hdr, current_block);
} else {
/* Big block file */
if (!ole2_read_block(hdr, buff, 1 << hdr->log2_big_block_size, current_block)) {
break;
12 years ago
}
if (cli_writen(ofd, buff, MIN(len, (1 << hdr->log2_big_block_size))) != MIN(len, (1 << hdr->log2_big_block_size))) {
ret = CL_EWRITE;
goto done;
12 years ago
}
12 years ago
current_block = ole2_get_next_block_number(hdr, current_block);
len -= MIN(len, (1 << hdr->log2_big_block_size));
}
}
/*
* Unlike w/ handler_otf(), the ole2 summary JSON will be recorded
* when we re-ingest the files we wrote above when we scan the directory.
* See cli_ole2_tempdir_scan_vba()
*/
ret = CL_SUCCESS;
done:
FREE(name);
if (-1 != ofd) {
close(ofd);
}
FREE(buff);
if (NULL != blk_bitset) {
cli_bitset_free(blk_bitset);
}
return ret;
}
enum biff_parser_states {
BIFF_PARSER_INITIAL,
BIFF_PARSER_EXPECTING_2ND_TAG_BYTE,
BIFF_PARSER_EXPECTING_1ST_LENGTH_BYTE,
BIFF_PARSER_EXPECTING_2ND_LENGTH_BYTE,
BIFF_PARSER_NAME_RECORD,
BIFF_PARSER_BOUNDSHEET_RECORD,
BIFF_PARSER_MSODRAWINGGROUP_RECORD,
BIFF_PARSER_DATA,
};
struct biff_parser_state {
enum biff_parser_states state;
uint16_t opcode;
uint16_t length;
uint16_t data_offset;
uint8_t tmp;
};
/**
* Scan through a buffer of BIFF records and find PARSERNAME, BOUNDSHEET records (Which indicate XLM macros).
* BIFF streams follow the format OOLLDDDDDDDDD..., where OO is the opcode (little endian 16 bit value),
* LL is the data length (little endian 16 bit value), followed by LL bytes of data. Records are defined in
* the MICROSOFT OFFICE EXCEL 97-2007 BINARY FILE FORMAT SPECIFICATION.
*
* \param state The parser state.
* \param buff The buffer.
* \param len The buffer's size in bytes.
* \param ctx The ClamAV context for emitting JSON about the document.
* \returns true if a macro has been found, false otherwise.
*/
static cl_error_t scan_biff_for_xlm_macros_and_images(
struct biff_parser_state *state,
unsigned char *buff,
size_t len,
cli_ctx *ctx,
bool *found_macro,
bool *found_image)
{
cl_error_t status = CL_EFORMAT;
size_t i;
for (i = 0; i < len; ++i) {
switch (state->state) {
case BIFF_PARSER_INITIAL:
state->opcode = buff[i];
state->state = BIFF_PARSER_EXPECTING_2ND_TAG_BYTE;
break;
case BIFF_PARSER_EXPECTING_2ND_TAG_BYTE:
state->opcode |= buff[i] << 8;
state->state = BIFF_PARSER_EXPECTING_1ST_LENGTH_BYTE;
break;
case BIFF_PARSER_EXPECTING_1ST_LENGTH_BYTE:
state->length = buff[i];
state->state = BIFF_PARSER_EXPECTING_2ND_LENGTH_BYTE;
break;
case BIFF_PARSER_EXPECTING_2ND_LENGTH_BYTE:
state->length |= buff[i] << 8;
state->data_offset = 0;
switch (state->opcode) {
case OPC_BOUNDSHEET:
state->state = BIFF_PARSER_BOUNDSHEET_RECORD;
break;
case OPC_NAME:
state->state = BIFF_PARSER_NAME_RECORD;
break;
case OPC_MSODRAWINGGROUP:
state->state = BIFF_PARSER_MSODRAWINGGROUP_RECORD;
break;
default:
state->state = BIFF_PARSER_DATA;
break;
}
if (state->length == 0) {
state->state = BIFF_PARSER_INITIAL;
}
break;
default:
switch (state->state) {
case BIFF_PARSER_NAME_RECORD:
#if HAVE_JSON
if (state->data_offset == 0) {
state->tmp = buff[i] & 0x20;
} else if ((state->data_offset == 14 || state->data_offset == 15) && state->tmp) {
if (buff[i] == 1 || buff[i] == 2) {
if (SCAN_COLLECT_METADATA && (ctx->wrkproperty != NULL)) {
json_object *indicators = cli_jsonarray(ctx->wrkproperty, "MacroIndicators");
if (indicators) {
cli_jsonstr(indicators, NULL, "autorun");
} else {
cli_dbgmsg("[scan_biff_for_xlm_macros_and_images] Failed to add \"autorun\" entry to MacroIndicators JSON array\n");
}
}
}
if (buff[i] != 0) {
state->tmp = 0;
}
}
#endif
break;
case BIFF_PARSER_BOUNDSHEET_RECORD:
if (state->data_offset == 4) {
state->tmp = buff[i];
} else if (state->data_offset == 5 && buff[i] == 1) { // Excel 4.0 macro sheet
cli_dbgmsg("[scan_biff_for_xlm_macros_and_images] Found XLM macro sheet\n");
#if HAVE_JSON
if (SCAN_COLLECT_METADATA && (ctx->wrkproperty != NULL)) {
cli_jsonbool(ctx->wrkproperty, "HasMacros", 1);
json_object *macro_languages = cli_jsonarray(ctx->wrkproperty, "MacroLanguages");
if (macro_languages) {
cli_jsonstr(macro_languages, NULL, "XLM");
} else {
cli_dbgmsg("[scan_biff_for_xlm_macros_and_images] Failed to add \"XLM\" entry to MacroLanguages JSON array\n");
}
if (state->tmp == 1 || state->tmp == 2) {
json_object *indicators = cli_jsonarray(ctx->wrkproperty, "MacroIndicators");
if (indicators) {
cli_jsonstr(indicators, NULL, "hidden");
} else {
cli_dbgmsg("[scan_biff_for_xlm_macros_and_images] Failed to add \"hidden\" entry to MacroIndicators JSON array\n");
}
}
}
#endif
*found_macro = true;
}
break;
case BIFF_PARSER_DATA:
break;
case BIFF_PARSER_MSODRAWINGGROUP_RECORD:
// Embedded image found
if (true != *found_image) {
*found_image = true;
cli_dbgmsg("[scan_biff_for_xlm_macros_and_images] Found image in sheet\n");
}
break;
default:
// Should never arrive here
cli_dbgmsg("[scan_biff_for_xlm_macros_and_images] Unexpected state value %d\n", (int)state->state);
break;
}
state->data_offset += 1;
if (state->data_offset >= state->length) {
state->state = BIFF_PARSER_INITIAL;
}
}
}
status = CL_SUCCESS;
return status;
}
/**
* @brief Scan for XLM (Excel 4.0) macro sheets and images in an OLE2 Workbook stream.
*
* The stream should be encoded with <= BIFF8.
* The found_macro and found_image out-params should be checked even if an error occured.
*
* @param hdr
* @param prop
* @param ctx
* @param found_macro [out] If any macros were found
* @param found_image [out] If any images were found
* @return cl_error_t CL_EPARSE if an error was encountered
* @return cl_error_t CL_EMEM if a memory issue was encountered.
* @return cl_error_t CL_SUCCESS if no errors were encountered.
*/
static cl_error_t scan_for_xlm_macros_and_images(ole2_header_t *hdr, property_t *prop, cli_ctx *ctx, bool *found_macro, bool *found_image)
{
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.
4 years ago
cl_error_t status = CL_EPARSE;
unsigned char *buff = NULL;
int32_t current_block = 0;
size_t len = 0, offset = 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.
4 years ago
bitset_t *blk_bitset = NULL;
struct biff_parser_state state = {0};
if (prop->type != 2) {
/* Not a file */
goto done;
}
memset(&state, 0, sizeof(state));
state.state = BIFF_PARSER_INITIAL;
current_block = prop->start_block;
len = prop->size;
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.
4 years ago
CLI_MALLOC(buff, 1 << hdr->log2_big_block_size,
cli_errmsg("OLE2 [scan_for_xlm_macros_and_images]: Unable to allocate memory for buff: %u\n", 1 << hdr->log2_big_block_size);
status = CL_EMEM);
blk_bitset = cli_bitset_init();
if (!blk_bitset) {
cli_errmsg("OLE2 [scan_for_xlm_macros_and_images]: init bitset failed\n");
goto done;
}
while ((current_block >= 0) && (len > 0)) {
if (current_block > (int32_t)hdr->max_block_no) {
cli_dbgmsg("OLE2 [scan_for_xlm_macros_and_images]: Max block number for file size exceeded: %d\n", current_block);
goto done;
}
/* Check we aren't in a loop */
if (cli_bitset_test(blk_bitset, (unsigned long)current_block)) {
/* Loop in block list */
cli_dbgmsg("OLE2 [scan_for_xlm_macros_and_images]: Block list loop detected\n");
goto done;
}
if (!cli_bitset_set(blk_bitset, (unsigned long)current_block)) {
goto done;
}
if (prop->size < (int64_t)hdr->sbat_cutoff) {
/* Small block file */
if (!ole2_get_sbat_data_block(hdr, buff, current_block)) {
cli_dbgmsg("OLE2 [scan_for_xlm_macros_and_images]: ole2_get_sbat_data_block failed\n");
goto done;
}
/* buff now contains the block with N small blocks in it */
offset = (1 << hdr->log2_small_block_size) * (current_block % (1 << (hdr->log2_big_block_size - hdr->log2_small_block_size)));
(void)scan_biff_for_xlm_macros_and_images(&state, &buff[offset], MIN(len, 1 << hdr->log2_small_block_size), ctx, found_macro, found_image);
len -= MIN(len, 1 << hdr->log2_small_block_size);
current_block = ole2_get_next_sbat_block(hdr, current_block);
} else {
/* Big block file */
if (!ole2_read_block(hdr, buff, 1 << hdr->log2_big_block_size, current_block)) {
goto done;
}
(void)scan_biff_for_xlm_macros_and_images(&state, buff, MIN(len, (1 << hdr->log2_big_block_size)), ctx, found_macro, found_image);
current_block = ole2_get_next_block_number(hdr, current_block);
len -= MIN(len, (1 << hdr->log2_big_block_size));
}
}
status = CL_SUCCESS;
done:
FREE(buff);
if (blk_bitset) {
cli_bitset_free(blk_bitset);
}
return status;
}
/**
* @brief enum file Handler - checks for VBA presence
*
* @param hdr
* @param prop
* @param dir
* @param ctx the scan context
* @return cl_error_t
*/
static cl_error_t handler_enum(ole2_header_t *hdr, property_t *prop, const char *dir, cli_ctx *ctx)
{
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.
4 years ago
cl_error_t status = CL_EREAD;
char *name = NULL;
unsigned char *hwp_check = NULL;
int32_t offset = 0;
#if HAVE_JSON
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.
4 years ago
json_object *arrobj = NULL;
json_object *strmobj = NULL;
name = cli_ole2_get_property_name2(prop->name, prop->name_size);
if (name) {
if (SCAN_COLLECT_METADATA && ctx->wrkproperty != NULL) {
arrobj = cli_jsonarray(ctx->wrkproperty, "Streams");
if (NULL == arrobj) {
cli_warnmsg("ole2: no memory for streams list or streams is not an array\n");
} else {
strmobj = json_object_new_string(name);
json_object_array_add(arrobj, strmobj);
}
if (!strcmp(name, "powerpoint document")) {
cli_jsonstr(ctx->wrkproperty, "FileType", "CL_TYPE_MSPPT");
}
if (!strcmp(name, "worddocument")) {
cli_jsonstr(ctx->wrkproperty, "FileType", "CL_TYPE_MSWORD");
}
if (!strcmp(name, "workbook")) {
cli_jsonstr(ctx->wrkproperty, "FileType", "CL_TYPE_MSXL");
}
}
}
#else
UNUSEDPARAM(ctx);
#endif
UNUSEDPARAM(dir);
if (!hdr->has_vba) {
if (!name)
name = cli_ole2_get_property_name2(prop->name, prop->name_size);
if (name) {
12 years ago
if (!strcmp(name, "_vba_project") || !strcmp(name, "powerpoint document") || !strcmp(name, "worddocument") || !strcmp(name, "_1_ole10native"))
hdr->has_vba = 1;
}
}
/*
* if we can find a root entry fileheader, it may be a HWP file
* identify the HWP signature "HWP Document File" at offset 0 stream
*/
if (!hdr->is_hwp) {
if (!name) {
name = cli_ole2_get_property_name2(prop->name, prop->name_size);
}
if (name) {
if (!strcmp(name, "fileheader")) {
CLI_CALLOC(hwp_check, 1, 1 << hdr->log2_big_block_size, status = CL_EMEM);
/* reading safety checks; do-while used for breaks */
do {
if (prop->size == 0)
break;
if (prop->start_block > hdr->max_block_no)
break;
/* read the header block (~256 bytes) */
offset = 0;
if (prop->size < (int64_t)hdr->sbat_cutoff) {
if (!ole2_get_sbat_data_block(hdr, hwp_check, prop->start_block)) {
break;
}
offset = (1 << hdr->log2_small_block_size) *
(prop->start_block % (1 << (hdr->log2_big_block_size - hdr->log2_small_block_size)));
/* reading safety */
if (offset + 40 >= 1 << hdr->log2_big_block_size)
break;
} else {
if (!ole2_read_block(hdr, hwp_check, 1 << hdr->log2_big_block_size, prop->start_block)) {
break;
}
}
/* compare against HWP signature; we could add the 15 padding NULLs too */
if (!memcmp(hwp_check + offset, "HWP Document File", 17)) {
hwp5_header_t *hwp_new;
#if HAVE_JSON
cli_jsonstr(ctx->wrkproperty, "FileType", "CL_TYPE_HWP5");
#endif
CLI_CALLOC(hwp_new, 1, sizeof(hwp5_header_t), status = CL_EMEM);
/*
* Copy the header information into our header struct.
*/
memcpy(hwp_new, hwp_check + offset, sizeof(hwp5_header_t));
hwp_new->version = ole2_endian_convert_32(hwp_new->version);
hwp_new->flags = ole2_endian_convert_32(hwp_new->flags);
hdr->is_hwp = hwp_new;
}
} while (0);
}
}
}
/* If we've already found a macro and an image, we can skip this initial check.
This scan step is to save a little time so we don't have to fully parse it
later if never find anything.. */
if (!hdr->has_xlm || !hdr->has_image) {
if (!name) {
name = cli_ole2_get_property_name2(prop->name, prop->name_size);
}
if (name && (strcmp(name, "workbook") == 0 || strcmp(name, "book") == 0)) {
(void)scan_for_xlm_macros_and_images(hdr, prop, ctx, &hdr->has_xlm, &hdr->has_image);
}
}
status = CL_SUCCESS;
done:
FREE(name);
FREE(hwp_check);
return status;
}
static int
likely_mso_stream(int fd)
{
off_t fsize;
unsigned char check[2];
fsize = lseek(fd, 0, SEEK_END);
if (fsize == -1) {
cli_dbgmsg("likely_mso_stream: call to lseek() failed\n");
return 0;
} else if (fsize < 6) {
return 0;
}
if (lseek(fd, 4, SEEK_SET) == -1) {
cli_dbgmsg("likely_mso_stream: call to lseek() failed\n");
return 0;
}
if (cli_readn(fd, check, 2) != 2) {
cli_dbgmsg("likely_mso_stream: reading from fd failed\n");
return 0;
}
if (check[0] == 0x78 && check[1] == 0x9C)
return 1;
return 0;
}
static cl_error_t scan_mso_stream(int fd, cli_ctx *ctx)
{
int zret, ofd;
cl_error_t ret = CL_SUCCESS;
fmap_t *input;
off_t off_in = 0;
size_t count, outsize = 0;
z_stream zstrm;
char *tmpname;
uint32_t prefix;
unsigned char inbuf[FILEBUFF], outbuf[FILEBUFF];
/* fmap the input file for easier manipulation */
if (fd < 0) {
cli_dbgmsg("scan_mso_stream: Invalid file descriptor argument\n");
return CL_ENULLARG;
} else {
STATBUF statbuf;
if (FSTAT(fd, &statbuf) == -1) {
cli_dbgmsg("scan_mso_stream: Can't stat file descriptor\n");
return CL_ESTAT;
}
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.
6 years ago
input = fmap(fd, 0, statbuf.st_size, NULL);
if (!input) {
cli_dbgmsg("scan_mso_stream: Failed to get fmap for input stream\n");
return CL_EMAP;
}
}
/* reserve tempfile for output and scanning */
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.
6 years ago
if ((ret = cli_gentempfd(ctx->sub_tmpdir, &tmpname, &ofd)) != CL_SUCCESS) {
cli_errmsg("scan_mso_stream: Can't generate temporary file\n");
funmap(input);
return ret;
}
/* initialize zlib inflation stream */
memset(&zstrm, 0, sizeof(zstrm));
zstrm.zalloc = Z_NULL;
zstrm.zfree = Z_NULL;
zstrm.opaque = Z_NULL;
zstrm.next_in = inbuf;
zstrm.next_out = outbuf;
zstrm.avail_in = 0;
zstrm.avail_out = FILEBUFF;
zret = inflateInit(&zstrm);
if (zret != Z_OK) {
cli_dbgmsg("scan_mso_stream: Can't initialize zlib inflation stream\n");
ret = CL_EUNPACK;
goto mso_end;
}
/* extract 32-bit prefix */
if (fmap_readn(input, &prefix, off_in, sizeof(prefix)) != sizeof(prefix)) {
cli_dbgmsg("scan_mso_stream: Can't extract 4-byte prefix\n");
ret = CL_EREAD;
goto mso_end;
}
/* RFC1952 says numbers are stored with least significant byte first */
prefix = le32_to_host(prefix);
off_in += sizeof(uint32_t);
cli_dbgmsg("scan_mso_stream: stream prefix = %08x(%d)\n", prefix, prefix);
/* inflation loop */
do {
if (zstrm.avail_in == 0) {
size_t bytes_read;
zstrm.next_in = inbuf;
bytes_read = fmap_readn(input, inbuf, off_in, FILEBUFF);
if (bytes_read == (size_t)-1) {
cli_errmsg("scan_mso_stream: Error reading MSO file\n");
ret = CL_EUNPACK;
goto mso_end;
}
if (bytes_read == 0)
break;
zstrm.avail_in = bytes_read;
off_in += bytes_read;
}
zret = inflate(&zstrm, Z_SYNC_FLUSH);
count = FILEBUFF - zstrm.avail_out;
if (count) {
if (cli_checklimits("MSO", ctx, outsize + count, 0, 0) != CL_SUCCESS)
break;
if (cli_writen(ofd, outbuf, count) != count) {
cli_errmsg("scan_mso_stream: Can't write to file %s\n", tmpname);
ret = CL_EWRITE;
goto mso_end;
}
outsize += count;
}
zstrm.next_out = outbuf;
zstrm.avail_out = FILEBUFF;
} while (zret == Z_OK);
/* post inflation checks */
if (zret != Z_STREAM_END && zret != Z_OK) {
if (outsize == 0) {
cli_infomsg(ctx, "scan_mso_stream: Error decompressing MSO file. No data decompressed.\n");
ret = CL_EUNPACK;
goto mso_end;
}
cli_infomsg(ctx, "scan_mso_stream: Error decompressing MSO file. Scanning what was decompressed.\n");
}
cli_dbgmsg("scan_mso_stream: Decompressed %llu bytes to %s\n", (long long unsigned)outsize, tmpname);
if (outsize != prefix) {
cli_warnmsg("scan_mso_stream: declared prefix != inflated stream size, %llu != %llu\n",
(long long unsigned)prefix, (long long unsigned)outsize);
} else {
cli_dbgmsg("scan_mso_stream: declared prefix == inflated stream size, %llu == %llu\n",
(long long unsigned)prefix, (long long unsigned)outsize);
}
/* scanning inflated stream */
ret = cli_magic_scan_desc(ofd, tmpname, ctx, NULL);
/* clean-up */
mso_end:
zret = inflateEnd(&zstrm);
if (zret != Z_OK)
ret = CL_EUNPACK;
close(ofd);
if (!ctx->engine->keeptmp)
if (cli_unlink(tmpname))
ret = CL_EUNLINK;
free(tmpname);
funmap(input);
return ret;
}
static cl_error_t handler_otf(ole2_header_t *hdr, property_t *prop, const char *dir, cli_ctx *ctx)
{
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.
4 years ago
cl_error_t ret = CL_BREAK;
char *tempfile = NULL;
char *name = NULL;
unsigned char *buff = NULL;
int32_t current_block = 0;
size_t len = 0, offset = 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.
4 years ago
int ofd = -1;
int is_mso = 0;
bitset_t *blk_bitset = NULL;
12 years ago
UNUSEDPARAM(dir);
12 years ago
if (prop->type != 2) {
/* Not a file */
ret = CL_SUCCESS;
goto done;
12 years ago
}
print_ole2_property(prop);
if (!(tempfile = cli_gentemp(ctx ? ctx->sub_tmpdir : NULL))) {
ret = CL_EMEM;
goto done;
}
if ((ofd = open(tempfile, O_RDWR | O_CREAT | O_TRUNC | O_BINARY, S_IRUSR | S_IWUSR)) < 0) {
cli_dbgmsg("OLE2 [handler_otf]: Can't create file %s\n", tempfile);
ret = CL_ECREAT;
goto done;
12 years ago
}
12 years ago
current_block = prop->start_block;
len = prop->size;
if (cli_debug_flag) {
if (!name) {
name = cli_ole2_get_property_name2(prop->name, prop->name_size);
}
cli_dbgmsg("OLE2 [handler_otf]: Dumping '%s' to '%s'\n", name, tempfile);
}
CLI_MALLOC(buff, 1 << hdr->log2_big_block_size, ret = CL_EMEM);
12 years ago
blk_bitset = cli_bitset_init();
12 years ago
if (!blk_bitset) {
cli_errmsg("OLE2 [handler_otf]: init bitset failed\n");
goto done;
12 years ago
}
12 years ago
while ((current_block >= 0) && (len > 0)) {
if (current_block > (int32_t)hdr->max_block_no) {
cli_dbgmsg("OLE2 [handler_otf]: Max block number for file size exceeded: %d\n", current_block);
12 years ago
break;
}
12 years ago
/* Check we aren't in a loop */
if (cli_bitset_test(blk_bitset, (unsigned long)current_block)) {
/* Loop in block list */
cli_dbgmsg("OLE2 [handler_otf]: Block list loop detected\n");
12 years ago
break;
}
12 years ago
if (!cli_bitset_set(blk_bitset, (unsigned long)current_block)) {
break;
}
if (prop->size < (int64_t)hdr->sbat_cutoff) {
12 years ago
/* Small block file */
if (!ole2_get_sbat_data_block(hdr, buff, current_block)) {
cli_dbgmsg("OLE2 [handler_otf]: ole2_get_sbat_data_block failed\n");
12 years ago
break;
}
12 years ago
/* buff now contains the block with N small blocks in it */
offset = (1 << hdr->log2_small_block_size) * (current_block % (1 << (hdr->log2_big_block_size - hdr->log2_small_block_size)));
12 years ago
if (cli_writen(ofd, &buff[offset], MIN(len, 1 << hdr->log2_small_block_size)) != MIN(len, 1 << hdr->log2_small_block_size)) {
goto done;
12 years ago
}
12 years ago
len -= MIN(len, 1 << hdr->log2_small_block_size);
current_block = ole2_get_next_sbat_block(hdr, current_block);
} else {
/* Big block file */
if (!ole2_read_block(hdr, buff, 1 << hdr->log2_big_block_size, current_block)) {
break;
}
if (cli_writen(ofd, buff, MIN(len, (1 << hdr->log2_big_block_size))) != MIN(len, (1 << hdr->log2_big_block_size))) {
ret = CL_EWRITE;
goto done;
12 years ago
}
12 years ago
current_block = ole2_get_next_block_number(hdr, current_block);
len -= MIN(len, (1 << hdr->log2_big_block_size));
}
}
/* defragmenting of ole2 stream complete */
is_mso = likely_mso_stream(ofd);
12 years ago
if (lseek(ofd, 0, SEEK_SET) == -1) {
ret = CL_ESEEK;
goto done;
12 years ago
}
#if HAVE_JSON
/* JSON Output Summary Information */
if (SCAN_COLLECT_METADATA && (ctx->properties != NULL)) {
if (!name) {
name = cli_ole2_get_property_name2(prop->name, prop->name_size);
}
if (name) {
if (!strncmp(name, "_5_summaryinformation", 21)) {
cli_dbgmsg("OLE2: detected a '_5_summaryinformation' stream\n");
/* JSONOLE2 - what to do if something breaks? */
if (cli_ole2_summary_json(ctx, ofd, 0) == CL_ETIMEOUT) {
ret = CL_ETIMEOUT;
goto done;
}
}
if (!strncmp(name, "_5_documentsummaryinformation", 29)) {
cli_dbgmsg("OLE2: detected a '_5_documentsummaryinformation' stream\n");
/* JSONOLE2 - what to do if something breaks? */
if (cli_ole2_summary_json(ctx, ofd, 1) == CL_ETIMEOUT) {
ret = CL_ETIMEOUT;
goto done;
}
}
}
}
#endif
if (hdr->is_hwp) {
if (!name) {
name = cli_ole2_get_property_name2(prop->name, prop->name_size);
}
ret = cli_scanhwp5_stream(ctx, hdr->is_hwp, name, ofd, tempfile);
} else if (is_mso < 0) {
ret = CL_ESEEK;
} else if (is_mso) {
/* MSO Stream Scan */
ret = scan_mso_stream(ofd, ctx);
} else {
/* Normal File Scan */
ret = cli_magic_scan_desc(ofd, tempfile, ctx, NULL);
}
ret = ret == CL_VIRUS ? CL_VIRUS : CL_SUCCESS;
done:
FREE(name);
if (-1 != ofd) {
close(ofd);
}
FREE(buff);
if (NULL != blk_bitset) {
cli_bitset_free(blk_bitset);
}
if (NULL != tempfile) {
if (ctx && !ctx->engine->keeptmp) {
if (cli_unlink(tempfile)) {
ret = CL_EUNLINK;
}
12 years ago
}
free(tempfile);
tempfile = NULL;
}
return ret;
}
#if !defined(HAVE_ATTRIB_PACKED) && !defined(HAVE_PRAGMA_PACK) && !defined(HAVE_PRAGMA_PACK_HPPA)
12 years ago
static int
ole2_read_header(int fd, ole2_header_t *hdr)
{
int i;
12 years ago
if (cli_readn(fd, &hdr->magic, 8) != 8) {
return FALSE;
}
if (cli_readn(fd, &hdr->clsid, 16) != 16) {
return FALSE;
}
if (cli_readn(fd, &hdr->minor_version, 2) != 2) {
return FALSE;
}
if (cli_readn(fd, &hdr->dll_version, 2) != 2) {
return FALSE;
}
if (cli_readn(fd, &hdr->byte_order, 2) != 2) {
return FALSE;
}
if (cli_readn(fd, &hdr->log2_big_block_size, 2) != 2) {
return FALSE;
}
if (cli_readn(fd, &hdr->log2_small_block_size, 4) != 4) {
return FALSE;
}
if (cli_readn(fd, &hdr->reserved, 8) != 8) {
return FALSE;
}
if (cli_readn(fd, &hdr->bat_count, 4) != 4) {
return FALSE;
}
if (cli_readn(fd, &hdr->prop_start, 4) != 4) {
return FALSE;
}
if (cli_readn(fd, &hdr->signature, 4) != 4) {
return FALSE;
}
if (cli_readn(fd, &hdr->sbat_cutoff, 4) != 4) {
return FALSE;
}
if (cli_readn(fd, &hdr->sbat_start, 4) != 4) {
return FALSE;
}
if (cli_readn(fd, &hdr->sbat_block_count, 4) != 4) {
return FALSE;
}
if (cli_readn(fd, &hdr->xbat_start, 4) != 4) {
return FALSE;
}
if (cli_readn(fd, &hdr->xbat_count, 4) != 4) {
return FALSE;
}
for (i = 0; i < 109; i++) {
if (cli_readn(fd, &hdr->bat_array[i], 4) != 4) {
return FALSE;
}
}
return TRUE;
}
#endif
/**
* @brief Extract macros and images from an ole2 file
*
* @param dirname A temp directory where we should store extracted content
* @param ctx The scan context
* @param files [out] A store of file names of extracted things to be processed later.
* @param has_vba [out] If the ole2 contained 1 or more VBA macros
* @param has_xlm [out] If the ole2 contained 1 or more XLM macros
* @param has_image [out] If the ole2 contained 1 or more images
* @return cl_error_t
*/
cl_error_t cli_ole2_extract(const char *dirname, cli_ctx *ctx, struct uniq **files, int *has_vba, int *has_xlm, int *has_image)
{
ole2_header_t hdr;
cl_error_t ret = CL_CLEAN;
size_t hdr_size;
unsigned int file_count = 0;
unsigned long scansize, scansize2;
const void *phdr;
12 years ago
cli_dbgmsg("in cli_ole2_extract()\n");
if (!ctx) {
14 years ago
return CL_ENULLARG;
}
hdr.is_hwp = NULL;
12 years ago
hdr.bitset = NULL;
if (ctx->engine->maxscansize) {
if (ctx->engine->maxscansize > ctx->scansize) {
12 years ago
scansize = ctx->engine->maxscansize - ctx->scansize;
} else {
12 years ago
return CL_EMAXSIZE;
}
} else {
12 years ago
scansize = -1;
}
12 years ago
scansize2 = scansize;
/* size of header - size of other values in struct */
hdr_size = sizeof(struct ole2_header_tag) -
sizeof(int32_t) - // sbat_root_start
sizeof(uint32_t) - // max_block_no
sizeof(off_t) - // m_length
sizeof(bitset_t *) - // bitset
sizeof(struct uniq *) - // U
sizeof(fmap_t *) - // map
sizeof(bool) - // has_vba
sizeof(bool) - // has_xlm
sizeof(bool) - // has_image
sizeof(hwp5_header_t *); // is_hwp
12 years ago
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.
4 years ago
if ((size_t)(ctx->fmap->len) < (size_t)(hdr_size)) {
12 years ago
return CL_CLEAN;
}
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.
4 years ago
hdr.map = ctx->fmap;
12 years ago
hdr.m_length = hdr.map->len;
phdr = fmap_need_off_once(hdr.map, 0, hdr_size);
12 years ago
if (phdr) {
memcpy(&hdr, phdr, hdr_size);
} else {
cli_dbgmsg("cli_ole2_extract: failed to read header\n");
goto done;
12 years ago
}
hdr.minor_version = ole2_endian_convert_16(hdr.minor_version);
hdr.dll_version = ole2_endian_convert_16(hdr.dll_version);
hdr.byte_order = ole2_endian_convert_16(hdr.byte_order);
hdr.log2_big_block_size = ole2_endian_convert_16(hdr.log2_big_block_size);
12 years ago
hdr.log2_small_block_size = ole2_endian_convert_32(hdr.log2_small_block_size);
hdr.bat_count = ole2_endian_convert_32(hdr.bat_count);
hdr.prop_start = ole2_endian_convert_32(hdr.prop_start);
hdr.sbat_cutoff = ole2_endian_convert_32(hdr.sbat_cutoff);
hdr.sbat_start = ole2_endian_convert_32(hdr.sbat_start);
hdr.sbat_block_count = ole2_endian_convert_32(hdr.sbat_block_count);
hdr.xbat_start = ole2_endian_convert_32(hdr.xbat_start);
hdr.xbat_count = ole2_endian_convert_32(hdr.xbat_count);
12 years ago
hdr.sbat_root_start = -1;
hdr.bitset = cli_bitset_init();
if (!hdr.bitset) {
ret = CL_EMEM;
goto done;
12 years ago
}
if (memcmp(hdr.magic, magic_id, 8) != 0) {
cli_dbgmsg("OLE2 magic failed!\n");
ret = CL_EFORMAT;
goto done;
12 years ago
}
if (hdr.log2_big_block_size < 6 || hdr.log2_big_block_size > 30) {
cli_dbgmsg("CAN'T PARSE: Invalid big block size (2^%u)\n", hdr.log2_big_block_size);
goto done;
12 years ago
}
if (!hdr.log2_small_block_size || hdr.log2_small_block_size > hdr.log2_big_block_size) {
cli_dbgmsg("CAN'T PARSE: Invalid small block size (2^%u)\n", hdr.log2_small_block_size);
goto done;
12 years ago
}
if (hdr.sbat_cutoff != 4096) {
cli_dbgmsg("WARNING: Untested sbat cutoff (%u); data may not extract correctly\n", hdr.sbat_cutoff);
}
if (hdr.map->len > INT32_MAX) {
cli_dbgmsg("OLE2 extract: Overflow detected\n");
ret = CL_EFORMAT;
goto done;
}
12 years ago
/* 8 SBAT blocks per file block */
hdr.max_block_no = (hdr.map->len - MAX(512, 1 << hdr.log2_big_block_size)) / (1 << hdr.log2_small_block_size);
print_ole2_header(&hdr);
cli_dbgmsg("Max block number: %lu\n", (unsigned long int)hdr.max_block_no);
/* PASS 1 : Count files and check for VBA */
hdr.has_vba = false;
hdr.has_xlm = false;
hdr.has_image = false;
ret = ole2_walk_property_tree(&hdr, NULL, 0, handler_enum, 0, &file_count, ctx, &scansize);
12 years ago
cli_bitset_free(hdr.bitset);
hdr.bitset = NULL;
if (!file_count || !(hdr.bitset = cli_bitset_init())) {
goto done;
}
12 years ago
if (hdr.is_hwp) {
cli_dbgmsg("OLE2: identified HWP document\n");
cli_dbgmsg("OLE2: HWP version: 0x%08x\n", hdr.is_hwp->version);
cli_dbgmsg("OLE2: HWP flags: 0x%08x\n", hdr.is_hwp->flags);
ret = cli_hwp5header(ctx, hdr.is_hwp);
if (ret != CL_SUCCESS) {
goto done;
}
}
12 years ago
/* If there's no VBA we scan OTF */
if (hdr.has_vba || hdr.has_xlm || hdr.has_image) {
12 years ago
/* PASS 2/A : VBA scan */
cli_dbgmsg("OLE2: VBA project found\n");
if (!(hdr.U = uniq_init(file_count))) {
cli_dbgmsg("OLE2: uniq_init() failed\n");
ret = CL_EMEM;
goto done;
12 years ago
}
file_count = 0;
ole2_walk_property_tree(&hdr, dirname, 0, handler_writefile, 0, &file_count, ctx, &scansize2);
ret = CL_CLEAN;
*files = hdr.U;
if (has_vba) {
*has_vba = hdr.has_vba;
}
if (has_xlm) {
*has_xlm = hdr.has_xlm;
}
if (has_image) {
*has_image = hdr.has_image;
}
12 years ago
} else {
cli_dbgmsg("OLE2: no VBA projects found\n");
/* PASS 2/B : OTF scan */
file_count = 0;
ret = ole2_walk_property_tree(&hdr, NULL, 0, handler_otf, 0, &file_count, ctx, &scansize2);
12 years ago
}
done:
if (hdr.bitset) {
12 years ago
cli_bitset_free(hdr.bitset);
}
if (hdr.is_hwp) {
free(hdr.is_hwp);
}
12 years ago
return ret == CL_BREAK ? CL_CLEAN : ret;
}