f2fs: fix to do sanity check correctly on i_inline_xattr_size
Chao Yu
maintainerPATCH v2 1/1 f2fs: fix to do sanity check correctly on i_inline_xattr_size
syzbot reported an out-of-range access issue as below:
UBSAN: array-index-out-of-bounds in fs/f2fs/f2fs.h:3292:19 index 18446744073709550491 is out of range for type '__le32[923]' (aka 'unsigned int[923]') CPU: 0 UID: 0 PID: 5338 Comm: syz.0.0 Not tainted 6.12.0-syzkaller-10689-g7af08b57bcb9 #0 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2~bpo12+1 04/01/2014 Call Trace: <TASK> __dump_stack lib/dump_stack.c:94 [inline] dump_stack_lvl+0x241/0x360 lib/dump_stack.c:120 ubsan_epilogue lib/ubsan.c:231 [inline] __ubsan_handle_out_of_bounds+0x121/0x150 lib/ubsan.c:429 read_inline_xattr+0x273/0x280 lookup_all_xattrs fs/f2fs/xattr.c:341 [inline] f2fs_getxattr+0x57b/0x13b0 fs/f2fs/xattr.c:533 vfs_getxattr_alloc+0x472/0x5c0 fs/xattr.c:393 ima_read_xattr+0x38/0x60 security/integrity/ima/ima_appraise.c:229 process_measurement+0x117a/0x1fb0 security/integrity/ima/ima_main.c:353 ima_file_check+0xd9/0x120 security/integrity/ima/ima_main.c:572 security_file_post_open+0xb9/0x280 security/security.c:3121 do_open fs/namei.c:3830 [inline] path_openat+0x2ccd/0x3590 fs/namei.c:3987 do_file_open_root+0x3a7/0x720 fs/namei.c:4039 file_open_root+0x247/0x2a0 fs/open.c:1382 do_handle_open+0x85b/0x9d0 fs/fhandle.c:414 do_syscall_x64 arch/x86/entry/common.c:52 [inline] do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83 entry_SYSCALL_64_after_hwframe+0x77/0x7f
index: 18446744073709550491 (decimal, unsigned long long)
= 0xfffffffffffffb9b (hexadecimal) = -1125 (decimal, long long) UBSAN detects that inline_xattr_addr() tries to access .i_addr[-1125].
w/ below testcase, it can reproduce this bug easily:
- mkfs.f2fs -f -O extra_attr,flexible_inline_xattr /dev/sdb
- mount -o inline_xattr_size=512 /dev/sdb /mnt/f2fs
- touch /mnt/f2fs/file
- umount /mnt/f2fs
- inject.f2fs --node --mb i_inline --nid 4 --val 0x1 /dev/sdb
- inject.f2fs --node --mb i_inline_xattr_size --nid 4 --val 2048 /dev/sdb
- mount /dev/sdb /mnt/f2fs
- getfattr /mnt/f2fs/file
The root cause is if metadata of filesystem and inode were fuzzed as below:
- extra_attr feature is enabled
- flexible_inline_xattr feature is enabled
- ri.i_inline_xattr_size = 2048
- F2FS_EXTRA_ATTR bit in ri.i_inline was not set
sanity_check_inode() will skip doing sanity check on fi->i_inline_xattr_size, result in using invalid inline_xattr_size later incorrectly, fix it.
Meanwhile, let's fix to check lower boundary for .i_inline_xattr_size w/ MIN_INLINE_XATTR_SIZE like we did in parse_options().
There is a related issue reported by syzbot, Qasim Ijaz has anlyzed and fixed it w/ very similar way [1], as discussed, we all agree that it will be better to do sanity check in sanity_check_inode() for fix, so finally, let's fix these two related bugs w/ current patch.
Including commit message from Qasim's patch as below, thanks a lot for his contribution.
"In f2fs_getxattr(), the function lookup_all_xattrs() allocates a 12-byte (base_size) buffer for an inline extended attribute. However, when __find_inline_xattr() calls __find_xattr(), it uses the macro "list_for_each_xattr(entry, addr)", which starts by calling XATTR_FIRST_ENTRY(addr). This skips a 24-byte struct f2fs_xattr_header at the beginning of the buffer, causing an immediate out-of-bounds read in a 12-byte allocation. The subsequent !IS_XATTR_LAST_ENTRY(entry) check then dereferences memory outside the allocated region, triggering the slab-out-of bounds read.
This patch prevents the out-of-bounds read by adding a check to bail out early if inline_size is too small and does not account for the header plus the 4-byte value that IS_XATTR_LAST_ENTRY reads."
[1]: https://lore.kernel.org/linux-f2fs-devel/[email protected]/
---
v2:
- including commit message from Qasim's patch.
fs/f2fs/inode.c | 19 ++++++++++--------- 1 file changed, 10 insertions(+), 9 deletions(-)
fs/f2fs/inode.c +10 -9
@@ -302,15 +302,6 @@ static bool sanity_check_inode(struct inode *inode, struct page *node_page) F2FS_TOTAL_EXTRA_ATTR_SIZE);return false; } -if (f2fs_sb_has_flexible_inline_xattr (sbi) && -f2fs_has_inline_xattr (inode) && - (!fi->i_inline_xattr_size || - fi->i_inline_xattr_size >MAX_INLINE_XATTR_SIZE )) { -f2fs_warn (sbi,"%s: inode (ino=%lx) has corrupted i_inline_xattr_size: %d, max: %lu" , - __func__, inode->i_ino, fi->i_inline_xattr_size, -MAX_INLINE_XATTR_SIZE ); -return false; - }if (f2fs_sb_has_compression (sbi) && fi->i_flags &F2FS_COMPR_FL && F2FS_FITS_IN_INODE(ri, fi->i_extra_isize,
@@ -320,6 +311,16 @@ static bool sanity_check_inode(struct inode *inode, struct page *node_page) } } +if (f2fs_sb_has_flexible_inline_xattr (sbi) && +f2fs_has_inline_xattr (inode) && + (fi->i_inline_xattr_size <MIN_INLINE_XATTR_SIZE || + fi->i_inline_xattr_size >MAX_INLINE_XATTR_SIZE )) { +f2fs_warn (sbi,"%s: inode (ino=%lx) has corrupted i_inline_xattr_size: %d, min: %lu, max: %lu" , + __func__, inode->i_ino, fi->i_inline_xattr_size, +MIN_INLINE_XATTR_SIZE ,MAX_INLINE_XATTR_SIZE ); +return false; + } +if (!f2fs_sb_has_extra_attr (sbi)) {if (f2fs_sb_has_project_quota (sbi)) { f2fs_warn(sbi, "%s: corrupted inode ino=%lx, wrong feature flag: %u, run fsck to fix.",
Signature · 1 line
2.48.0.rc2.279.g1de40edade-goog
This patch was applied to jaegeuk/f2fs.git (dev) by Jaegeuk Kim <[email protected]>:
Hello:
This patch was applied to jaegeuk/f2fs.git (dev)
by Jaegeuk Kim <[email protected]>:
Quoted 29 lines · On Tue, 14 Jan 2025 20:34:10 +0800 you wrote:
syzbot reported an out-of-range access issue as below:
UBSAN: array-index-out-of-bounds in fs/f2fs/f2fs.h:3292:19 index 18446744073709550491 is out of range for type '__le32[923]' (aka 'unsigned int[923]') CPU: 0 UID: 0 PID: 5338 Comm: syz.0.0 Not tainted 6.12.0-syzkaller-10689-g7af08b57bcb9 #0 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2~bpo12+1 04/01/2014 Call Trace: <TASK> __dump_stack lib/dump_stack.c:94 [inline] dump_stack_lvl+0x241/0x360 lib/dump_stack.c:120 ubsan_epilogue lib/ubsan.c:231 [inline] __ubsan_handle_out_of_bounds+0x121/0x150 lib/ubsan.c:429 read_inline_xattr+0x273/0x280 lookup_all_xattrs fs/f2fs/xattr.c:341 [inline] f2fs_getxattr+0x57b/0x13b0 fs/f2fs/xattr.c:533 vfs_getxattr_alloc+0x472/0x5c0 fs/xattr.c:393 ima_read_xattr+0x38/0x60 security/integrity/ima/ima_appraise.c:229 process_measurement+0x117a/0x1fb0 security/integrity/ima/ima_main.c:353 ima_file_check+0xd9/0x120 security/integrity/ima/ima_main.c:572 security_file_post_open+0xb9/0x280 security/security.c:3121 do_open fs/namei.c:3830 [inline] path_openat+0x2ccd/0x3590 fs/namei.c:3987 do_file_open_root+0x3a7/0x720 fs/namei.c:4039 file_open_root+0x247/0x2a0 fs/open.c:1382 do_handle_open+0x85b/0x9d0 fs/fhandle.c:414 do_syscall_x64 arch/x86/entry/common.c:52 [inline] do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83 entry_SYSCALL_64_after_hwframe+0x77/0x7f
Quoted 1 line
[...]
Here is the summary with links:
- [f2fs-dev,v2] f2fs: fix to do sanity check correctly on i_inline_xattr_size
https://git.kernel.org/jaegeuk/f2fs/c/5c1768b67250
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html