Re: [PATCH] ext4: fix WARNING in ext4_update_inline_data

[Jan Kara <jack@suse.cz>]

On Thu 05-01-23 17:15:26, Ye Bin wrote:
> From: Ye Bin <yebin10@huawei.com>
> 
> Syzbot found the following issue:
> EXT4-fs (loop0): mounted filesystem 00000000-0000-0000-0000-000000000000 without journal. Quota mode: none.
> fscrypt: AES-256-CTS-CBC using implementation "cts-cbc-aes-aesni"
> fscrypt: AES-256-XTS using implementation "xts-aes-aesni"
> ------------[ cut here ]------------
> WARNING: CPU: 0 PID: 5071 at mm/page_alloc.c:5525 __alloc_pages+0x30a/0x560 mm/page_alloc.c:5525
> Modules linked in:
> CPU: 1 PID: 5071 Comm: syz-executor263 Not tainted 6.2.0-rc1-syzkaller #0
> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/26/2022
> RIP: 0010:__alloc_pages+0x30a/0x560 mm/page_alloc.c:5525
> RSP: 0018:ffffc90003c2f1c0 EFLAGS: 00010246
> RAX: ffffc90003c2f220 RBX: 0000000000000014 RCX: 0000000000000000
> RDX: 0000000000000028 RSI: 0000000000000000 RDI: ffffc90003c2f248
> RBP: ffffc90003c2f2d8 R08: dffffc0000000000 R09: ffffc90003c2f220
> R10: fffff52000785e49 R11: 1ffff92000785e44 R12: 0000000000040d40
> R13: 1ffff92000785e40 R14: dffffc0000000000 R15: 1ffff92000785e3c
> FS:  0000555556c0d300(0000) GS:ffff8880b9800000(0000) knlGS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> CR2: 00007f95d5e04138 CR3: 00000000793aa000 CR4: 00000000003506f0
> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> Call Trace:
>  <TASK>
>  __alloc_pages_node include/linux/gfp.h:237 [inline]
>  alloc_pages_node include/linux/gfp.h:260 [inline]
>  __kmalloc_large_node+0x95/0x1e0 mm/slab_common.c:1113
>  __do_kmalloc_node mm/slab_common.c:956 [inline]
>  __kmalloc+0xfe/0x190 mm/slab_common.c:981
>  kmalloc include/linux/slab.h:584 [inline]
>  kzalloc include/linux/slab.h:720 [inline]
>  ext4_update_inline_data+0x236/0x6b0 fs/ext4/inline.c:346
>  ext4_update_inline_dir fs/ext4/inline.c:1115 [inline]
>  ext4_try_add_inline_entry+0x328/0x990 fs/ext4/inline.c:1307
>  ext4_add_entry+0x5a4/0xeb0 fs/ext4/namei.c:2385
>  ext4_add_nondir+0x96/0x260 fs/ext4/namei.c:2772
>  ext4_create+0x36c/0x560 fs/ext4/namei.c:2817
>  lookup_open fs/namei.c:3413 [inline]
>  open_last_lookups fs/namei.c:3481 [inline]
>  path_openat+0x12ac/0x2dd0 fs/namei.c:3711
>  do_filp_open+0x264/0x4f0 fs/namei.c:3741
>  do_sys_openat2+0x124/0x4e0 fs/open.c:1310
>  do_sys_open fs/open.c:1326 [inline]
>  __do_sys_openat fs/open.c:1342 [inline]
>  __se_sys_openat fs/open.c:1337 [inline]
>  __x64_sys_openat+0x243/0x290 fs/open.c:1337
>  do_syscall_x64 arch/x86/entry/common.c:50 [inline]
>  do_syscall_64+0x3d/0xb0 arch/x86/entry/common.c:80
>  entry_SYSCALL_64_after_hwframe+0x63/0xcd
> 
> Above issue happens as follows:
> ext4_iget
>    ext4_find_inline_data_nolock ->i_inline_off=164 i_inline_size=60
> ext4_try_add_inline_entry
>    __ext4_mark_inode_dirty
>       ext4_expand_extra_isize_ea ->i_extra_isize=32 s_want_extra_isize=44
>          ext4_xattr_shift_entries
> 	 ->after shift i_inline_off is incorrect, actually is change to 176
> ext4_try_add_inline_entry
>   ext4_update_inline_dir
>     get_max_inline_xattr_value_size
>       if (EXT4_I(inode)->i_inline_off)
> 	entry = (struct ext4_xattr_entry *)((void *)raw_inode +
> 			EXT4_I(inode)->i_inline_off);
>         free += EXT4_XATTR_SIZE(le32_to_cpu(entry->e_value_size));
> 	->As entry is incorrect, then 'free' may be negative
>    ext4_update_inline_data
>       value = kzalloc(len, GFP_NOFS);
>       -> len is unsigned int, maybe very large, then trigger warning when
>          'kzalloc()'
> To resolve above issue there's need to update 'i_inline_off' after
> 'ext4_xattr_shift_entries()'.
> 
> Reported-by: syzbot+d30838395804afc2fa6f@syzkaller.appspotmail.com
> Signed-off-by: Ye Bin <yebin10@huawei.com>

...

> diff --git a/fs/ext4/xattr.c b/fs/ext4/xattr.c
> index 7decaaf27e82..4cb9fd5bb622 100644
> --- a/fs/ext4/xattr.c
> +++ b/fs/ext4/xattr.c
> @@ -2778,6 +2778,9 @@ int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
>  			(void *)header, total_ino);
>  	EXT4_I(inode)->i_extra_isize = new_extra_isize;
>  
> +	if (ext4_has_inline_data(inode))
> +		error = ext4_find_inline_data_nolock(inode);
> +

So this would be fine except that ext4_expand_extra_isize_ea() may be
racing e.g. with pending ext4_do_writepages() call trying to remove inline
data and thus cause an assertion failure there because
ext4_expand_extra_isize_ea() sets EXT4_STATE_MAY_INLINE_DATA. So I think
you need to move setting of EXT4_STATE_MAY_INLINE_DATA out from
ext4_find_inline_data_nolock() into the callers.

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR