qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Nir Soffer <nsoffer@redhat.com>
To: "Daniel P. Berrangé" <berrange@redhat.com>
Cc: Kevin Wolf <kwolf@redhat.com>, Nir Soffer <nirsof@gmail.com>,
	qemu-block <qemu-block@nongnu.org>,
	Markus Armbruster <armbru@redhat.com>,
	QEMU Developers <qemu-devel@nongnu.org>,
	Max Reitz <mreitz@redhat.com>
Subject: Re: [PATCH 2/2] block: file-posix: Replace posix_fallocate with fallocate
Date: Tue, 15 Sep 2020 11:55:18 +0300	[thread overview]
Message-ID: <CAMRbyysRun-nkgugLuxHcBv3HJ2QqVs2rw9HO3CjgZEE5ftGjQ@mail.gmail.com> (raw)
In-Reply-To: <20200914173201.GR1252186@redhat.com>

On Mon, Sep 14, 2020 at 8:32 PM Daniel P. Berrangé <berrange@redhat.com> wrote:
>
> On Mon, Aug 31, 2020 at 05:01:27PM +0300, Nir Soffer wrote:
> > If fallocate() is not supported, posix_fallocate() falls back to
> > inefficient allocation, writing one byte for every 4k bytes[1]. This is
> > very slow compared with writing zeros. In oVirt we measured ~400%
> > improvement in allocation time when replacing posix_fallocate() with
> > manually writing zeroes[2].
> >
> > We also know that posix_fallocated() does not work well when using OFD
> > locks[3]. We don't know the reason yet for this issue yet.
> >
> > Change preallocate_falloc() to use fallocate() instead of
> > posix_falloate(), and fall back to full preallocation if not supported.
> >
> > Here are quick test results with this change.
> >
> > Before (qemu-img-5.1.0-2.fc32.x86_64):
> >
> > $ time qemu-img create -f raw -o preallocation=falloc /tmp/nfs3/test.raw 6g
> > Formatting '/tmp/nfs3/test.raw', fmt=raw size=6442450944 preallocation=falloc
> >
> > real 0m42.100s
> > user 0m0.602s
> > sys 0m4.137s
> >
> > NFS stats:
> > calls      retrans    authrefrsh    write
> > 1571583    0          1572205       1571321
> >
> > After:
> >
> > $ time ./qemu-img create -f raw -o preallocation=falloc /tmp/nfs3/test.raw 6g
> > Formatting '/tmp/nfs3/test.raw', fmt=raw size=6442450944 preallocation=falloc
> >
> > real 0m15.551s
> > user 0m0.070s
> > sys 0m2.623s
> >
> > NFS stats:
> > calls      retrans    authrefrsh    write
> > 24620      0          24624         24567
> >
> > [1] https://code.woboq.org/userspace/glibc/sysdeps/posix/posix_fallocate.c.html#96
> > [2] https://bugzilla.redhat.com/1850267#c25
> > [3] https://bugzilla.redhat.com/1851097
>
> This bug appears to be private to RH employees only, so rather than link
> to it, please summarize any important facts in it for benefit of nonn-RH
> QEMU contributors.

Thanks, I missed that detail when linking to the bug. The bug is public now.

> > Signed-off-by: Nir Soffer <nsoffer@redhat.com>
> > ---
> >  block/file-posix.c                     | 32 +++++++++-----------------
> >  docs/system/qemu-block-drivers.rst.inc | 11 +++++----
> >  docs/tools/qemu-img.rst                | 11 +++++----
> >  qapi/block-core.json                   |  4 ++--
> >  4 files changed, 25 insertions(+), 33 deletions(-)
>
> Regards,
> Daniel
> --
> |: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
> |: https://libvirt.org         -o-            https://fstop138.berrange.com :|
> |: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
>



  reply	other threads:[~2020-09-15  8:57 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-08-31 14:01 [PATCH 0/2] Replace posix_fallocate() with falloate() Nir Soffer
2020-08-31 14:01 ` [PATCH 1/2] block: file-posix: Extract preallocate helpers Nir Soffer
2020-09-01 10:24   ` Alberto Garcia
2020-09-01 10:26   ` Alberto Garcia
2020-09-01 10:47     ` Nir Soffer
2020-08-31 14:01 ` [PATCH 2/2] block: file-posix: Replace posix_fallocate with fallocate Nir Soffer
2020-09-01 15:51   ` Alberto Garcia
2020-09-14 17:32   ` Daniel P. Berrangé
2020-09-15  8:55     ` Nir Soffer [this message]
2020-08-31 15:55 ` [PATCH 0/2] Replace posix_fallocate() with falloate() no-reply
2020-09-14 17:19 ` Nir Soffer

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=CAMRbyysRun-nkgugLuxHcBv3HJ2QqVs2rw9HO3CjgZEE5ftGjQ@mail.gmail.com \
    --to=nsoffer@redhat.com \
    --cc=armbru@redhat.com \
    --cc=berrange@redhat.com \
    --cc=kwolf@redhat.com \
    --cc=mreitz@redhat.com \
    --cc=nirsof@gmail.com \
    --cc=qemu-block@nongnu.org \
    --cc=qemu-devel@nongnu.org \
    /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).