Re: [PATCH] gfs2: reject undersized EA records during xattr walks

Hello ZhengYuan Huang,

On Tue, Apr 28, 2026 at 6:05 AM ZhengYuan Huang <[email protected]> wrote:
> [BUG]
> A fuzzed GFS2 image can crash in `ea_split_ea()` while servicing setxattr:
>
> BUG: KASAN: slab-out-of-bounds in ea_split_ea+0x1f3/0x230 fs/gfs2/xattr.c:835
> Write of size 4 at addr ffff88801550a1c0 by task syz.0.379/950
>
> Call Trace:
>  ...
>  ea_split_ea+0x1f3/0x230 fs/gfs2/xattr.c:835
>  ea_set_simple_noalloc fs/gfs2/xattr.c:888 [inline]
>  ea_set_simple+0x66f/0x7c0 fs/gfs2/xattr.c:951
>  ea_foreach_i+0x2a1/0x560 fs/gfs2/xattr.c:112
>  ea_foreach+0x3cd/0x5d0 fs/gfs2/xattr.c:140
>  ea_set_i+0x176/0x3d0 fs/gfs2/xattr.c:1063
>  __gfs2_xattr_set+0x8de/0xc60 fs/gfs2/xattr.c:1215
>  gfs2_xattr_set+0x2ed/0x450 fs/gfs2/xattr.c:1258
>  __vfs_setxattr+0x14f/0x1c0 fs/xattr.c:200
>  __vfs_setxattr_noperm+0x10b/0x5c0 fs/xattr.c:234
>  __vfs_setxattr_locked+0x172/0x240 fs/xattr.c:295
>  vfs_setxattr+0x167/0x390 fs/xattr.c:321
>  do_setxattr+0x13c/0x180 fs/xattr.c:636
>  filename_setxattr+0x16b/0x1c0 fs/xattr.c:665
>  path_setxattrat+0x1d8/0x280 fs/xattr.c:713
>  __do_sys_setxattr fs/xattr.c:747 [inline]
>  __se_sys_setxattr fs/xattr.c:743 [inline]
>  __x64_sys_setxattr+0xcd/0x150 fs/xattr.c:743
>  ...
>
> Allocated by task 893:
>  kasan_save_stack+0x39/0x70 mm/kasan/common.c:56
>  kasan_save_track+0x14/0x40 mm/kasan/common.c:77
>  kasan_save_alloc_info+0x37/0x60 mm/kasan/generic.c:573
>  unpoison_slab_object mm/kasan/common.c:342 [inline]
>  __kasan_slab_alloc+0x9d/0xa0 mm/kasan/common.c:368
>  kasan_slab_alloc include/linux/kasan.h:252 [inline]
>  slab_post_alloc_hook mm/slub.c:4978 [inline]
>  slab_alloc_node mm/slub.c:5288 [inline]
>  kmem_cache_alloc_noprof+0x1f0/0x7d0 mm/slub.c:5295
>  __do_sys_getcwd+0xe0/0x730 fs/d_path.c:416
>  __se_sys_getcwd fs/d_path.c:412 [inline]
>  __x64_sys_getcwd+0x59/0x80 fs/d_path.c:412
>  x64_sys_call+0x15ac/0x26a0 arch/x86/include/generated/asm/syscalls_64.h:80
>  do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
>  do_syscall_64+0x93/0xf80 arch/x86/entry/syscall_64.c:94
>  entry_SYSCALL_64_after_hwframe+0x76/0x7e
>
> Freed by task 893:
>  kasan_save_stack+0x39/0x70 mm/kasan/common.c:56
>  kasan_save_track+0x14/0x40 mm/kasan/common.c:77
>  __kasan_save_free_info+0x3b/0x60 mm/kasan/generic.c:587
>  kasan_save_free_info mm/kasan/kasan.h:406 [inline]
>  poison_slab_object mm/kasan/common.c:252 [inline]
>  __kasan_slab_free+0x6f/0xa0 mm/kasan/common.c:284
>  kasan_slab_free include/linux/kasan.h:234 [inline]
>  slab_free_hook mm/slub.c:2543 [inline]
>  slab_free mm/slub.c:6642 [inline]
>  kmem_cache_free+0x384/0x7a0 mm/slub.c:6752
>  __do_sys_getcwd+0x3f8/0x730 fs/d_path.c:446
>  __se_sys_getcwd fs/d_path.c:412 [inline]
>  __x64_sys_getcwd+0x59/0x80 fs/d_path.c:412
>  x64_sys_call+0x15ac/0x26a0 arch/x86/include/generated/asm/syscalls_64.h:80
>  do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
>  do_syscall_64+0x93/0xf80 arch/x86/entry/syscall_64.c:94
>  entry_SYSCALL_64_after_hwframe+0x76/0x7e
>
> and also as:
>
> BUG: KASAN: use-after-free in ea_split_ea+0x1f3/0x230 fs/gfs2/xattr.c:835
> Write of size 4 at addr ffff888023742118 by task syz.0.27/395
>
> Call Trace:
>  __dump_stack lib/dump_stack.c:94 [inline]
>  dump_stack_lvl+0xbe/0x130 lib/dump_stack.c:120
>  print_address_description mm/kasan/report.c:378 [inline]
>  print_report+0xd1/0x650 mm/kasan/report.c:482
>  kasan_report+0xfb/0x140 mm/kasan/report.c:595
>  __asan_report_store4_noabort+0x17/0x30 mm/kasan/report_generic.c:385
>  ea_split_ea+0x1f3/0x230 fs/gfs2/xattr.c:835
>  ea_set_simple_noalloc fs/gfs2/xattr.c:888 [inline]
>  ea_set_simple+0x66f/0x7c0 fs/gfs2/xattr.c:951
>  ea_foreach_i+0x2a1/0x560 fs/gfs2/xattr.c:112
>  ea_foreach+0x334/0x5d0 fs/gfs2/xattr.c:162
>  ea_set_i+0x176/0x3d0 fs/gfs2/xattr.c:1063
>  __gfs2_xattr_set+0x8de/0xc60 fs/gfs2/xattr.c:1215
>  gfs2_xattr_set+0x2ed/0x450 fs/gfs2/xattr.c:1258
>  __vfs_setxattr+0x14f/0x1c0 fs/xattr.c:200
>  __vfs_setxattr_noperm+0x10b/0x5c0 fs/xattr.c:234
>  __vfs_setxattr_locked+0x172/0x240 fs/xattr.c:295
>  vfs_setxattr+0x167/0x390 fs/xattr.c:321
>  do_setxattr+0x13c/0x180 fs/xattr.c:636
>  filename_setxattr+0x16b/0x1c0 fs/xattr.c:665
>  path_setxattrat+0x1d8/0x280 fs/xattr.c:713
>  __do_sys_setxattr fs/xattr.c:747 [inline]
>  __se_sys_setxattr fs/xattr.c:743 [inline]
>  __x64_sys_setxattr+0xcd/0x150 fs/xattr.c:743
>  ...
>
> [CAUSE]
> `ea_foreach_i()` validates only that each EA record has a non-zero
> `ea_rec_len`, stays within the block, and has a valid type. It never checks
> that the record span actually covers the layout encoded by
> `ea_name_len`, `ea_data_len`, and `ea_num_ptrs`.
>
> If a malformed record reaches `ea_set_simple()` with
> `GFS2_EA_REC_LEN(ea) < GFS2_EA_SIZE(ea)`, the unsigned subtraction in the
> split test underflows and falsely claims there is enough free space.
> `ea_split_ea()` then writes the new EA header outside the real record.
>
> [FIX]
> Reject EA records whose reserved record length is smaller than their
> computed layout size in `ea_foreach_i()`, which is the common parse
> boundary for existing xattr metadata. This stops malformed records before
> any update path can split or otherwise trust the broken geometry.
>
> Signed-off-by: ZhengYuan Huang <[email protected]>
> ---
> fs/gfs2/xattr.c | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/fs/gfs2/xattr.c b/fs/gfs2/xattr.c
> index b9f48d6f10a9..16cf31bde99a 100644
> --- a/fs/gfs2/xattr.c
> +++ b/fs/gfs2/xattr.c
> @@ -96,6 +96,8 @@ static int ea_foreach_i(struct gfs2_inode *ip, struct buffer_head *bh,
>                 return -EIO;
>
>         for (ea = GFS2_EA_BH2FIRST(bh);; prev = ea, ea = GFS2_EA2NEXT(ea)) {
> +               u32 ea_size;
> +
>                 if (!GFS2_EA_REC_LEN(ea)) {
>                         gfs2_consist_inode(ip);
>                         return -EIO;
> @@ -109,6 +111,11 @@ static int ea_foreach_i(struct gfs2_inode *ip, struct buffer_head *bh,
>                         gfs2_consist_inode(ip);
>                         return -EIO;
>                 }
> +               ea_size = GFS2_EA_SIZE(ea);
> +               if (GFS2_EA_REC_LEN(ea) < ea_size) {
> +                       gfs2_consist_inode(ip);
> +                       return -EIO;
> +               }
>                 error = ea_call(ip, bh, ea, prev, data);
>                 if (error)
>                         return error;
> --
> 2.43.0
>

the '!GFS2_EA_REC_LEN(ea)' and 'GFS2_EA_REC_LEN(ea) < ea_size' checks
can be combined here. Also, nothing prevents GFS2_EA_SIZE() from
overflowing.

Can you fix that?

Thanks,
Andreas