Re: [Devel] [PATCH hci-8.1 v9 05/10] dm-qcow2: flush lower metadata before clearing its dirty bit on merge failure

From: Alexander Ivanov <alexander.ivanov@virtuozzo.com>
To: Konstantin Khorenko <khorenko@virtuozzo.com>
Cc: devel@openvz.org
Subject: Re: [Devel] [PATCH hci-8.1 v9 05/10] dm-qcow2: flush lower metadata before clearing its dirty bit on merge failure
Date: Mon, 9 Nov 2026 10:16:11 +0100	[thread overview]
Message-ID: <4477e479-afc9-4159-9269-2380458cab2b@virtuozzo.com> (raw)
In-Reply-To: <179372824968.2555502.16948569421120815187.b4-review@b4>

Rewrote.

On 11/3/26 18:50, Konstantin Khorenko wrote:
> I won't apply this version, please send v10 with the changes below.
>
> On 10/28/26 16:06, Alexander Ivanov wrote:
>> leaves the lower image marked clean on disk with L2 tables and refblocks
>> that lag behind the merged data.
>>
>> A failed md writeback of the lower image is also missed:
>> md_writeback_error is set only by qcow2_flush_deferred_activity(), so
>> the next resume is not refused.
> - Drop the second paragraph.
> - Say refblocks only, not "L2 tables and refblocks".
> - Shorten the subject, it is 77 columns.
>
> The next postsuspend flushes the lower image as well:
> qcow2_flush_deferred_activity_all() walks tgt->top and every ->lower
> and sets md_writeback_error on a failed write. So that resume is
> refused already. What the patch adds is the order: the flush now comes
> before the header write that clears the dirty bit.
>
> Merge writes wait for their L2 writeback before they complete, so L2
> does not lag. Only the refblocks are in the delayed writeback.
>
_______________________________________________
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel