Re: [RFT PATCH] ocfs2: Invalidate inode if i_mode is zero after block read

[Heming Zhao]

On Sat, Nov 08, 2025 at 03:01:35PM +0300, Ahmet Eray Karadag wrote:
> A panic occurs in ocfs2_unlink due to WARN_ON(inode->i_nlink == 0) when
> handling a corrupted inode with i_mode=0 and i_nlink=0 in memory.
> 
> This "zombie" inode is created because ocfs2_read_locked_inode proceeds
> even after ocfs2_validate_inode_block successfully validates a block
> that structurally looks okay (passes checksum, signature etc.) but
> contains semantically invalid data (specifically i_mode=0). The current
> validation function doesn't check for i_mode being zero.
> 
> This results in an in-memory inode with i_mode=0 being added to the VFS
> cache, which later triggers the panic during unlink.
> 
> Prevent this by adding an explicit check for (i_mode == 0, i_nlink == 0, non-orphan) 
> within ocfs2_validate_inode_block. If the check is true, return -EFSCORRUPTED to signal
> corruption. This causes the caller (ocfs2_read_locked_inode) to invoke
> make_bad_inode(), correctly preventing the zombie inode from entering
> the cache.
> 
> Reported-by: syzbot+55c40ae8a0e5f3659f2b@xxxxxxxxxxxxxxxxxxxxxxxxx
> Fixes: https://syzkaller.appspot.com/bug?extid=55c40ae8a0e5f3659f2b
> Co-developed-by: Albin Babu Varghese <albinbabuvarghese20@xxxxxxxxx>
> Signed-off-by: Albin Babu Varghese <albinbabuvarghese20@xxxxxxxxx>
> Signed-off-by: Ahmet Eray Karadag <eraykrdg1@xxxxxxxxx>
> Previous link: https://lore.kernel.org/all/20251022222752.46758-2-eraykrdg1@xxxxxxxxx/T/
> ---
>  fs/ocfs2/inode.c | 9 ++++++++-
>  1 file changed, 8 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/ocfs2/inode.c b/fs/ocfs2/inode.c
> index 14bf440ea4df..d966df3aa605 100644
> --- a/fs/ocfs2/inode.c
> +++ b/fs/ocfs2/inode.c
> @@ -1455,7 +1455,14 @@ int ocfs2_validate_inode_block(struct super_block *sb,
>  		     (unsigned long long)bh->b_blocknr);
>  		goto bail;
>  	}
> -
> +	if (!le16_to_cpu(di->i_links_count) && !le16_to_cpu(di->i_mode) &&
> +		!(le32_to_cpu(di->i_flags) & OCFS2_ORPHANED_FL)) {

1.
di->i_links_count is the low 16 bits of i_nlink. We need to check both
di->i_link_count and di->i_link_count_hi.

2.
We are only interested the the di members' value being ZERO.
Therefore, calling le16_to_cpu() to perform the conversion is wasting CPU cycles.

3.
The ocfs2_dinode structure has dozens of members, and syzbot could potentially
corrupt any one of them. At the same time, the probability of this type of syzbot
error occurring in a real-world scenario is practically zero. My idea is to
minimize unnecessary checks.
I prefer to remove the check for OCFS2_ORPHANED_FL.

4.
It is an error/exception if either i_links_count or i_mode is zero.
The code "!di->i_links_count && !di->i_mode" need to be changed to '||'. (I
didn't test it)

Thanks,
Heming

> +		mlog(ML_ERROR, "Invalid dinode #%llu: "
> +			"Corrupt state (nlink=0, mode=0, !orphan) detected!\n",
> +		        (unsigned long long)bh->b_blocknr);
> +		rc = -EFSCORRUPTED;
> +		goto bail;
> +	}
>  	/*
>  	 * Errors after here are fatal.
>  	 */
> -- 
> 2.43.0
> 
>