Virtuozzo QEMU development (svt-core@virtuozzo.com)
 help / color / mirror / Atom feed
From: Andrey Drobyshev <andrey.drobyshev@virtuozzo.com>
To: "Denis V. Lunev" <den@openvz.org>, svt-core@virtuozzo.com
Cc: den@virtuozzo.com
Subject: Re: [PATCH hci-8.0 0/9] block: cheaper zero handling in backup and commit #VSTOR-134085
Date: Wed, 30 Sep 2026 15:49:52 +0300	[thread overview]
Message-ID: <8e3e6894-0f95-4dfd-ab9c-bad9f2f5dc5d@virtuozzo.com> (raw)
In-Reply-To: <20260930111739.3866427-1-den@openvz.org>

No real findings from Sashiko
Ack, applied to hci-dev

Andrey

On 9/30/26 2:17 PM, Denis V. Lunev wrote:
> Backup and commit re-query the source block status for every task and
> size a task by the copy buffer, so a long run of zeroes turns into a
> crowd of small write-zeroes requests. This series reuses what the
> up-front scan already learned and lets one write-zeroes task cover a
> whole run.
> 
> This is the series posted upstream as v3, all nine patches with
> Andrey's Reviewed-by:
> https://lore.kernel.org/qemu-devel/20260929155125.3151111-1-den@openvz.org/
> 
> hci-8.0 carries 6b1a4754713 ("backup: skip copying unallocated for full
> mode"): sync=full runs the same up-front scan as sync=top, drops the
> clusters the source never allocated, and keeps skip_unallocated set
> afterwards, taking the block-status base from a separate top_mode flag.
> So the upstream cover letter case, a mostly unallocated image, is
> already cheap here. What the series buys on hci-8.0 is allocated zero
> clusters, the ones a guest write-zeroes or detect-zeroes leaves behind,
> in sync=full and in sync=bitmap, and the commit job.
> 
> Where the port differs from upstream, 4/9, 7/9, 8/9 and 9/9 say so in
> an "hci-8.0:" paragraph:
> - 3/9: only a context conflict. hci-8.0 lacks upstream a6c131217d,
>   which adds coroutine_fn to commit_iteration(), so the signature stays
>   as it is here.
> - 4/9: iotest 124 expects a sync=full target to keep the source's
>   holes, since the unallocated clusters are skipped rather than written
>   as zeroes.
> - 7/9: block_copy_status_base() keys on top_mode instead of
>   skip_unallocated, and the scan uses it too instead of computing the
>   base on its own.
> - 8/9: sync=full stays on block_copy_reset_unallocated(); only
>   sync=bitmap moves to block_copy_calculate_zero_bitmap().
> - 9/9: 6b1a4754713 also made the 056 copy-before-write source read as
>   zeroes. One widened write-zeroes task now covers all of it before
>   speed=1 can hold the job, so the test caps the task with x-perf
>   max-chunk. Andrey, the 4/9 and 9/9 test changes are new with this
>   port and not covered by your upstream review.
> 
> Every patch builds on its own. On dev007 with qcow2, iotests 040 055 056
> 124 129 141 185 219 257 image-fleecing backup-discard-source
> copy-before-write commit-zero-blocks iothreads-commit-active and
> migrate-during-backup pass. vz-kernel-cbt fails there with and without
> the series, the host kernel has no vz CBT.
> 
> On the same host, a 16G qcow2 source holding 1G of data, raw target on
> ext4, counting block_copy_process and commit_one_iteration trace events:
> 
>                                          before           after
>     sync=full, 15G unallocated        1024  0.44s     1024  0.44s
>     sync=full, 15G zero clusters     16384  0.66s     1032  0.46s
>     sync=bitmap, all dirty, same     16384  0.65s     1032  0.46s
>     commit, 15G zero clusters        32768  2.84s     2288  2.70s
> 
> Commit time is dominated by copying the 1G of data.
> 
> Denis V. Lunev (9):
>   block/commit: pass BDRV_WANT_PRECISE to block-status #VSTOR-134085
>   iotests/040: cover large and fragmented commit runs #VSTOR-134085
>   block/commit: batch block-status queries #VSTOR-134085
>   iotests/124: cover backup of zero clusters and holes #VSTOR-134085
>   block/block-copy: don't reserve memory for zero tasks #VSTOR-134085
>   block/block-copy: extract block_copy_set_task_method() #VSTOR-134085
>   block/block-copy: track known-zero source clusters #VSTOR-134085
>   block/backup: pre-fill zero_bitmap for full/bitmap #VSTOR-134085
>   block/block-copy: coalesce write-zeroes tasks #VSTOR-134085
> 
>  block/backup.c             |  94 ++++++----
>  block/block-copy.c         | 309 +++++++++++++++++++++++++++-----
>  block/commit.c             |  59 +++++--
>  include/block/block-copy.h |   5 +
>  tests/qemu-iotests/040     | 102 ++++++++++-
>  tests/qemu-iotests/040.out |   4 +-
>  tests/qemu-iotests/056     |   3 +-
>  tests/qemu-iotests/124     | 350 ++++++++++++++++++++++++++++++++++++-
>  tests/qemu-iotests/124.out |   4 +-
>  9 files changed, 830 insertions(+), 100 deletions(-)
> 
> 
> base-commit: f2f4b3de1a1be3360a77ca6ab79762466d238804


      parent reply	other threads:[~2026-09-30 12:49 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-09-30 11:17 Denis V. Lunev
2026-09-30 11:17 ` [PATCH hci-8.0 1/9] block/commit: pass BDRV_WANT_PRECISE to block-status #VSTOR-134085 Denis V. Lunev
2026-09-30 11:20   ` no-reply
2026-09-30 11:17 ` [PATCH hci-8.0 2/9] iotests/040: cover large and fragmented commit runs #VSTOR-134085 Denis V. Lunev
2026-09-30 11:19   ` no-reply
2026-09-30 11:17 ` [PATCH hci-8.0 3/9] block/commit: batch block-status queries #VSTOR-134085 Denis V. Lunev
2026-09-30 11:21   ` no-reply
2026-09-30 11:17 ` [PATCH hci-8.0 4/9] iotests/124: cover backup of zero clusters and holes #VSTOR-134085 Denis V. Lunev
2026-09-30 11:31   ` no-reply
2026-09-30 11:17 ` [PATCH hci-8.0 5/9] block/block-copy: don't reserve memory for zero tasks #VSTOR-134085 Denis V. Lunev
2026-09-30 11:22   ` no-reply
2026-09-30 11:17 ` [PATCH hci-8.0 6/9] block/block-copy: extract block_copy_set_task_method() #VSTOR-134085 Denis V. Lunev
2026-09-30 11:22   ` no-reply
2026-09-30 11:17 ` [PATCH hci-8.0 7/9] block/block-copy: track known-zero source clusters #VSTOR-134085 Denis V. Lunev
2026-09-30 11:30   ` no-reply
2026-09-30 11:17 ` [PATCH hci-8.0 8/9] block/backup: pre-fill zero_bitmap for full/bitmap #VSTOR-134085 Denis V. Lunev
2026-09-30 11:33   ` no-reply
2026-09-30 11:17 ` [PATCH hci-8.0 9/9] block/block-copy: coalesce write-zeroes tasks #VSTOR-134085 Denis V. Lunev
2026-09-30 11:34   ` no-reply
2026-09-30 12:49 ` Andrey Drobyshev [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=8e3e6894-0f95-4dfd-ab9c-bad9f2f5dc5d@virtuozzo.com \
    --to=andrey.drobyshev@virtuozzo.com \
    --cc=den@openvz.org \
    --cc=den@virtuozzo.com \
    --cc=svt-core@virtuozzo.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox