From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
To: Paolo Bonzini <pbonzini@redhat.com>,
Emanuele Giuseppe Esposito <eesposit@redhat.com>,
qemu-block@nongnu.org
Cc: John Snow <jsnow@redhat.com>, Kevin Wolf <kwolf@redhat.com>,
Max Reitz <mreitz@redhat.com>,
Stefan Hajnoczi <stefanha@redhat.com>,
qemu-devel@nongnu.org
Subject: Re: [PATCH v2 1/7] block-copy: streamline choice of copy_range vs. read/write
Date: Fri, 21 May 2021 18:48:32 +0300 [thread overview]
Message-ID: <5f93d053-0370-a179-0c81-27434430fd43@virtuozzo.com> (raw)
In-Reply-To: <5a4cc9ae-f419-ab40-cd7a-d30e39f91e54@redhat.com>
21.05.2021 18:10, Paolo Bonzini wrote:
> On 20/05/21 16:42, Vladimir Sementsov-Ogievskiy wrote:
>> 18.05.2021 13:07, Emanuele Giuseppe Esposito wrote:
>>> From: Paolo Bonzini <pbonzini@redhat.com>
>>>
>>> Put the logic to determine the copy size in a separate function, so
>>> that there is a simple state machine for the possible methods of
>>> copying data from one BlockDriverState to the other.
>>
>> Honestly, for me 4-state state-maching + function to determine copy-size doesn't seem better than two simple variables copy_size and use_copy_range.
>
> There were six states before (2 for s->use_copy_range, three for s->copy_size),
> of which two were unused. The heuristics for going between copy and read/write
> were quite illegible.
>
>> What's the benefit of it?
>
> Less duplication here, for example:
>
> + if (s->max_transfer < cluster_size) {
> /*
> * copy_range does not respect max_transfer. We don't want to bother
> * with requests smaller than block-copy cluster size, so fallback to
> * buffered copying (read and write respect max_transfer on their
> * behalf).
> */
> - s->use_copy_range = false;
> - s->copy_size = cluster_size;
> + s->method = COPY_READ_WRITE_CLUSTER;
> } else if (write_flags & BDRV_REQ_WRITE_COMPRESSED) {
> /* Compression supports only cluster-size writes and no copy-range. */
> - s->use_copy_range = false;
> - s->copy_size = cluster_size;
> + s->method = COPY_READ_WRITE_CLUSTER;
>
> and here:
>
> trace_block_copy_copy_range_fail(s, offset, ret);
> - s->use_copy_range = false;
> - s->copy_size = MAX(s->cluster_size, BLOCK_COPY_MAX_BUFFER);
> + s->method = COPY_READ_WRITE;
> ...
> /*
> * We enable copy-range, but keep small copy_size, until first
> * successful copy_range (look at block_copy_do_copy).
> */
> - s->use_copy_range = use_copy_range;
> - s->copy_size = MAX(s->cluster_size, BLOCK_COPY_MAX_BUFFER);
> + s->method = use_copy_range ? COPY_RANGE_SMALL : COPY_READ_WRITE;
>
> where it's not obvious that the two assignments to copy_size should be
> the same (and they're suboptimal, too, since they don't obey max_transfer).
>
> ... plus...
>
>>> While at it, store the common computation of block_copy_max_transfer
>>> into a new field of BlockCopyState, and make sure that we always
>>> obey max_transfer; that's more efficient even for the
>>> COPY_RANGE_READ_WRITE case.
>
> having a function makes it easier to spot slight differences that are
> just bugs, such as this one.
>
>>> Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
>>> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
>>> ---
>>> block/block-copy.c | 59 ++++++++++++++++++++++++++++++----------------
>>
>> stats agree with me, that its' not a simplification.
>
> Stats don't say everything. Not having something like this:
>
> s->copy_size =
> MIN(MAX(s->cluster_size, BLOCK_COPY_MAX_COPY_RANGE),
> QEMU_ALIGN_DOWN(block_copy_max_transfer(s->source,
> s->target),
> s->cluster_size));
>
> in the inner loop is already worth the extra lines for the
> function declaration, for example.
After my "[PATCH v2 00/33] block: publish backup-top filter" copy_range path becomes unused. I keep it thinking about further moving qemu-img convert to block-copy. But I don't even have a plan when to start this work. So, if we want to do something around copy_range here to prepare for thread-safety, let's just drop it for now as unused on top of "[PATCH v2 00/33] block: publish backup-top filter" (you can take one patch that drop copy_range support in backup to your series to not make a dependency). It's not difficult to reimplement it later.
--
Best regards,
Vladimir
next prev parent reply other threads:[~2021-05-21 15:50 UTC|newest]
Thread overview: 48+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-05-18 10:07 [PATCH v2 0/7] block-copy: protect block-copy internal structures Emanuele Giuseppe Esposito
2021-05-18 10:07 ` [PATCH v2 1/7] block-copy: streamline choice of copy_range vs. read/write Emanuele Giuseppe Esposito
2021-05-20 14:42 ` Vladimir Sementsov-Ogievskiy
2021-05-20 15:06 ` Emanuele Giuseppe Esposito
2021-05-20 15:24 ` Vladimir Sementsov-Ogievskiy
2021-05-21 15:10 ` Paolo Bonzini
2021-05-21 15:48 ` Vladimir Sementsov-Ogievskiy [this message]
2021-05-21 16:43 ` Paolo Bonzini
2021-05-21 17:51 ` Vladimir Sementsov-Ogievskiy
2021-05-27 8:20 ` Stefan Hajnoczi
2021-05-27 19:04 ` Paolo Bonzini
2021-05-18 10:07 ` [PATCH v2 2/7] block-copy: improve documentation of BlockCopyTask and BlockCopyState types and functions Emanuele Giuseppe Esposito
2021-05-20 15:00 ` Vladimir Sementsov-Ogievskiy
2021-05-20 15:15 ` Emanuele Giuseppe Esposito
2021-05-18 10:07 ` [PATCH v2 3/7] block-copy: move progress_set_remaining in block_copy_task_end Emanuele Giuseppe Esposito
2021-05-20 15:03 ` Vladimir Sementsov-Ogievskiy
2021-05-18 10:07 ` [PATCH v2 4/7] block-copy: add a CoMutex to the BlockCopyTask list Emanuele Giuseppe Esposito
2021-05-20 15:19 ` Vladimir Sementsov-Ogievskiy
2021-05-25 10:07 ` Emanuele Giuseppe Esposito
2021-05-25 10:25 ` Vladimir Sementsov-Ogievskiy
2021-05-26 14:58 ` Paolo Bonzini
2021-05-26 16:13 ` Vladimir Sementsov-Ogievskiy
2021-05-27 9:07 ` Stefan Hajnoczi
2021-05-18 10:07 ` [PATCH v2 5/7] block-copy: add QemuMutex lock for BlockCopyCallState list Emanuele Giuseppe Esposito
2021-05-20 15:30 ` Vladimir Sementsov-Ogievskiy
2021-05-21 15:01 ` Paolo Bonzini
2021-05-25 10:58 ` Emanuele Giuseppe Esposito
2021-05-26 14:49 ` Paolo Bonzini
2021-05-28 10:53 ` Paolo Bonzini
2021-05-18 10:07 ` [PATCH v2 6/7] block-copy: atomic .cancelled and .finished fields in BlockCopyCallState Emanuele Giuseppe Esposito
2021-05-20 15:34 ` Vladimir Sementsov-Ogievskiy
2021-05-21 15:02 ` Paolo Bonzini
2021-05-21 15:53 ` Vladimir Sementsov-Ogievskiy
2021-05-21 15:56 ` Vladimir Sementsov-Ogievskiy
2021-05-21 16:47 ` Paolo Bonzini
2021-05-18 10:07 ` [PATCH v2 7/7] block-copy: protect BlockCopyState .method fields Emanuele Giuseppe Esposito
2021-05-21 17:10 ` Vladimir Sementsov-Ogievskiy
2021-05-25 10:18 ` Emanuele Giuseppe Esposito
2021-05-25 11:00 ` Vladimir Sementsov-Ogievskiy
2021-05-26 14:48 ` Paolo Bonzini
2021-05-26 17:18 ` Vladimir Sementsov-Ogievskiy
2021-05-28 10:24 ` Paolo Bonzini
2021-05-28 11:01 ` Paolo Bonzini
2021-05-28 11:52 ` Vladimir Sementsov-Ogievskiy
2021-05-28 12:44 ` Vladimir Sementsov-Ogievskiy
2021-05-20 13:47 ` [PATCH v2 0/7] block-copy: protect block-copy internal structures Vladimir Sementsov-Ogievskiy
2021-05-20 14:33 ` Emanuele Giuseppe Esposito
2021-05-27 10:31 ` Stefan Hajnoczi
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=5f93d053-0370-a179-0c81-27434430fd43@virtuozzo.com \
--to=vsementsov@virtuozzo.com \
--cc=eesposit@redhat.com \
--cc=jsnow@redhat.com \
--cc=kwolf@redhat.com \
--cc=mreitz@redhat.com \
--cc=pbonzini@redhat.com \
--cc=qemu-block@nongnu.org \
--cc=qemu-devel@nongnu.org \
--cc=stefanha@redhat.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;
as well as URLs for NNTP newsgroup(s).