Re: [PATCH v7 26/31] ext4: clear DISKSIZE_GROW_PENDING on truncate or error

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] fs: missing DISKSIZE_GROW_PENDING clear during nlink > 0 eviction
--

commit b90813716147a7864ab2f3a3c46c2814d0084c2b
Author: Zhang Yi <yi.zhang@huawei.com>

ext4: clear DISKSIZE_GROW_PENDING on truncate or error

This commit clears the disksize-grow-pending state on various paths where a
zeroed EOF block is discarded before writeback completes. It also adds a
warning in destroy_inode to catch any inodes that missed clearing the state.

> diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
> --- a/fs/ext4/inode.c
> +++ b/fs/ext4/inode.c
[ ... ]
> @@ -266,6 +266,8 @@ void ext4_evict_inode(struct inode *inode)
>  
>  	if (ext4_should_order_data(inode))
>  		ext4_begin_ordered_truncate(inode, 0);
> +	if (ext4_inode_buffered_iomap(inode))
> +		ext4_iomap_clear_disksize_pending(inode);
>  	truncate_inode_pages_final(&inode->i_data);

[Severity: Medium]
When an inode with nlink > 0 is evicted and still has dirty pages (such as
during a forced unmount or after a failed sync_filesystem), ext4_evict_inode()
logs a data loss warning and forcefully truncates the page cache using
truncate_inode_pages_final() before jumping to the no_delete label.

Because the pages are discarded immediately without initiating writeback, the
writeback error path is never invoked, meaning the
EXT4_STATE_DISKSIZE_GROW_PENDING bit is never cleared.

Since the added state clearing logic is placed after the nlink > 0 block, does
this bypass clearing the bit for those inodes entirely?

[ ... ]
> diff --git a/fs/ext4/super.c b/fs/ext4/super.c
> --- a/fs/ext4/super.c
> +++ b/fs/ext4/super.c
> @@ -1497,6 +1497,14 @@ static void ext4_destroy_inode(struct inode *inode)
>  			 "Inode %llu (%p): i_reserved_data_blocks (%u) not cleared!",
>  			 inode->i_ino, EXT4_I(inode),
>  			 EXT4_I(inode)->i_reserved_data_blocks);
> +
> +	if (!(EXT4_SB(inode->i_sb)->s_mount_state & EXT4_ERROR_FS) &&
> +	    !ext4_emergency_state(inode->i_sb) &&
> +	    WARN_ON_ONCE(ext4_test_inode_state(inode,
> +				EXT4_STATE_DISKSIZE_GROW_PENDING)))

[Severity: Medium]
Can this assertion be reliably triggered when unmounting a filesystem while
writeback is blocked or fails, since the pending bit is not cleared during
nlink > 0 eviction? This might potentially crash systems with panic_on_warn
enabled.

> +		ext4_msg(inode->i_sb, KERN_ERR,
> +			 "Inode %llu (%p): EXT4_STATE_DISKSIZE_GROW_PENDING not cleared!",
> +			 inode->i_ino, EXT4_I(inode));
>  }
>  
>  static void ext4_shutdown(struct super_block *sb)

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261009103033.2920530-1-yi.zhang@huaweicloud.com?part=26