Re: [PATCH v4 2/3] ext4: record error when detect abnormal 'i_reserved_data_blocks'

From: "Theodore Ts'o" <tytso@mit.edu>
To: Ye Bin <yebin@huaweicloud.com>
Cc: adilger.kernel@dilger.ca, linux-ext4@vger.kernel.org,
	linux-kernel@vger.kernel.org, jack@suse.cz,
	Ye Bin <yebin10@huawei.com>, Eric Whitney <enwlinux@gmail.com>
Subject: Re: [PATCH v4 2/3] ext4: record error when detect abnormal 'i_reserved_data_blocks'
Date: Fri, 9 Dec 2022 00:50:26 -0500	[thread overview]
Message-ID: <Y5LMosWupn5a1orF@mit.edu> (raw)
In-Reply-To: <20221208033426.1832460-3-yebin@huaweicloud.com>

On Thu, Dec 08, 2022 at 11:34:25AM +0800, Ye Bin wrote:
> From: Ye Bin <yebin10@huawei.com>
> 
> If 'i_reserved_data_blocks' is not cleared which mean something wrong with
> code, free space accounting is likely wrong, according to Jan Kara's advice
> use ext4_error() to record this abnormal let fsck to repair and also we can
> capture this issue.

If i_reserved_data_block, it means that there is something wrong with
our delayed allocation accounting.  However, this accounting is
usually only an in-memory error.  The one exception is if quota is
enabled, in which case the space is decremented from the
user/group/project quota.

However, if quota is not in use, which is the overwhelmingly common
case, there will be nothing for fsck to repair.  (It does mean that df
will show a slightly smaller free space value due to the messed up
delayed allocation accounting, but that disappears after a reboot.)

Marking the file system as in need of repair when nothing is actually
wrong with the on-disk file system can backfire, in that it confuses
users when they see the ext4 error but then when they run fsck, fsck
reports nothing wrong --- at which point they file a bug.

We could use ext4_error if quotas are enabled, and ext4_msg if not,
but it might not worth the extra complexity.

Cheers,

					- Ted