Re: [PATCH] ext4: return -EINVAL on hole in ext4_convert_unwritten_extents_atomic()
From: Jan Kara
Date: Fri Oct 23 2026 - 07:58:07 EST
On Fri 23-10-26 15:53:45, Ojaswin Mujoo wrote:
> We should ideally never encounter a hole during the endio conversion as
> that indicates something is wrong with our extent handling. Still if we do
> find a hole at atomic write endio, ext4_map_blocks() returns 0 which would
> be returned back to callers by ext4_convert_unwritten_extents_atomic().
> This can be wrongly interpreted as a success. Fix this by bailing out with
> -EINVAL instead.
>
> Signed-off-by: Ojaswin Mujoo <ojaswin@xxxxxxxxxxxxx>
Makes sense! Feel free to add:
Reviewed-by: Jan Kara <jack@xxxxxxx>
Honza
> ---
> fs/ext4/extents.c | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> index 7193429a05e1..b54d8227786b 100644
> --- a/fs/ext4/extents.c
> +++ b/fs/ext4/extents.c
> @@ -5107,6 +5107,14 @@ int ext4_convert_unwritten_extents_atomic(handle_t *handle, struct inode *inode,
> map.m_lblk += ret;
> map.m_len = (max_blocks -= ret);
> ret = ext4_map_blocks(handle, inode, &map, flags);
> + if (ret == 0) {
> + ext4_msg(inode->i_sb, KERN_ERR,
> + "inode #%llu: block %u: len %u: "
> + "hole found while converting extents for atomic write",
> + inode->i_ino, map.m_lblk, map.m_len);
> + ret = -EINVAL;
> + break;
> + }
> if (ret != max_blocks)
> ext4_msg(inode->i_sb, KERN_INFO,
> "inode #%llu: block %u: len %u: "
>
> ---
> base-commit: 8c4f6e3376348fc3162eed86a05b933921c8f9ed
> change-id: 20261023-ext4-atomic-endio-fix-b4-9e1ead9e1aca
>
> Best regards,
> --
> Ojaswin Mujoo <ojaswin@xxxxxxxxxxxxx>
>
--
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR