Re: [PATCH V3] ocfs2: add a sanity check for i_size

[Joseph Qi <joseph.qi@linux.alibaba.com>]

On 10/14/24 7:47 PM, Edward Adam Davis wrote:
> Syzbot reported a BUG in ocfs2_read_virt_blocks.
> ------------[ cut here ]------------
> kernel BUG at fs/ocfs2/extent_map.c:971!
> Oops: invalid opcode: 0000 [#1] PREEMPT SMP KASAN PTI
> CPU: 1 UID: 0 PID: 5221 Comm: syz-executor117 Not tainted 6.12.0-rc2-syzkaller-00006-g87d6aab2389e #0
> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 09/13/2024
> RIP: 0010:ocfs2_read_virt_blocks+0xaa3/0xb20 fs/ocfs2/extent_map.c:971
> Code: 8b 54 24 30 8b 4c 24 1c 4c 8b 44 24 38 45 89 e1 ff 74 24 48 e8 7e 0a 0c 00 48 83 c4 08 eb 25 e8 43 d2 34 08 e8 ee d1 0c fe 90 <0f> 0b e8 e6 d1 0c fe 4c 89 e8 45 89 e5 49 89 c4 e9 01 f7 ff ff e8
> RSP: 0018:ffffc90002e2e820 EFLAGS: 00010293
> RAX: ffffffff83881432 RBX: 0000000000000000 RCX: ffff88802d3e9e00
> RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
> RBP: ffffc90002e2e9f0 R08: ffffffff83880eae R09: 1ffffffff203781d
> R10: dffffc0000000000 R11: fffffbfff203781e R12: ffff8880764d8878
> R13: 0000000000000000 R14: 0000000000000000 R15: 1ffff920005c5d18
> FS:  000055557cd55380(0000) GS:ffff8880b8700000(0000) knlGS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> CR2: 00005628fd41e058 CR3: 000000002d2e6000 CR4: 00000000003526f0
> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> Call Trace:
>  <TASK>
>  ocfs2_read_dir_block+0x106/0x5c0 fs/ocfs2/dir.c:508
>  ocfs2_find_dir_space_el fs/ocfs2/dir.c:3427 [inline]           // status = ocfs2_read_dir_block(dir, 0, &bh, 0);

As I replied before, the comment behind seems unneeded.

>  ocfs2_prepare_dir_for_insert+0x3f2/0x5c60 fs/ocfs2/dir.c:4274
>  ocfs2_mknod+0xcaf/0x2b40 fs/ocfs2/namei.c:292
>  ocfs2_create+0x1ab/0x480 fs/ocfs2/namei.c:672
>  lookup_open fs/namei.c:3595 [inline]
>  open_last_lookups fs/namei.c:3694 [inline]
>  path_openat+0x1c03/0x3590 fs/namei.c:3930
>  do_filp_open+0x235/0x490 fs/namei.c:3960
>  do_sys_openat2+0x13e/0x1d0 fs/open.c:1415
>  do_sys_open fs/open.c:1430 [inline]
>  __do_sys_openat fs/open.c:1446 [inline]
>  __se_sys_openat fs/open.c:1441 [inline]
>  __x64_sys_openat+0x247/0x2a0 fs/open.c:1441
>  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
> 
> In this path, v_block(0), flags(0), and nr(1) are fixed values that are
> hard coded, so if i_size is 0, the bug will be triggered.
> 

This is not a proper commit log.
You have to convert to the user scenario, but not simply list the
parameters value that cause the issue.

> Reported-and-tested-by: syzbot+797d4829dafe3f11dce7@syzkaller.appspotmail.com
> Closes: https://syzkaller.appspot.com/bug?extid=797d4829dafe3f11dce7
> Signed-off-by: Edward Adam Davis <eadavis@qq.com>
> ---
> V1 -> V2: keep rc to 0 when falgs contains READHEAD
> V2 -> V3: check i_size only and alert subject and comments
> 
>  fs/ocfs2/extent_map.c | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> #syz test
> 
> diff --git a/fs/ocfs2/extent_map.c b/fs/ocfs2/extent_map.c
> index f7672472fa82..29d27a70dbdd 100644
> --- a/fs/ocfs2/extent_map.c
> +++ b/fs/ocfs2/extent_map.c
> @@ -961,13 +961,17 @@ int ocfs2_read_virt_blocks(struct inode *inode, u64 v_block, int nr,
>  	int rc = 0;
>  	u64 p_block, p_count;
>  	int i, count, done = 0;
> +	loff_t i_size = i_size_read(inode);
>  
>  	trace_ocfs2_read_virt_blocks(
>  	     inode, (unsigned long long)v_block, nr, bhs, flags,
>  	     validate);
>  
> +	if (!i_size)
> +		return -EINVAL;
> +

Take a more consideration, inode size 0 doesn't mean it has no blocks,
since we have a case that fallocate with KEEP_SIZE.
Could you please check inode->i_blocks in above coredump?

Thanks,
Joseph

>  	if (((v_block + nr - 1) << inode->i_sb->s_blocksize_bits) >=
> -	    i_size_read(inode)) {
> +	    i_size) {
>  		BUG_ON(!(flags & OCFS2_BH_READAHEAD));
>  		goto out;
>  	}