From: Filipe Manana <fdmanana@gmail.com>
To: Nikolay Borisov <nborisov@suse.com>
Cc: linux-btrfs <linux-btrfs@vger.kernel.org>,
Filipe David Borba Manana <fdmanana@suse.com>
Subject: Re: [PATCH] btrfs: Simplify join_running_log_trans
Date: Mon, 20 May 2019 10:10:16 +0100 [thread overview]
Message-ID: <CAL3q7H5MHYhrvLnqHagkmLS_cey6Bc-nZh-21ivGjbSPjZ76+Q@mail.gmail.com> (raw)
In-Reply-To: <560224d9-28ec-033f-c7c0-f324576e5e04@suse.com>
On Mon, May 20, 2019 at 9:23 AM Nikolay Borisov <nborisov@suse.com> wrote:
>
>
>
> On 20.05.19 г. 11:11 ч., Nikolay Borisov wrote:
> > This patch removes stray smp_mb before root->log_root from join_running_log_trans
> > as well as the unlocked check for root->log_root. log_root is only set in
> > btrfs_add_log_tree, called from start_log_trans under root->log_mutex.
> > Furthermore, log_root is freed in btrfs_free_log, called from commit_fs_root,
> > which in turn is called from transaction's critical section (TRANS_COMMIT_DOING).
> > Those 2 locking invariants ensure join_running_log_trans don't see invalid
> > values of ->log_root.
> >
> > Additionally this results in around 26% improvement when deleting 500k files/dir.
> > All values are in seconds.
> >
> > With Patch (real) With patch (sys) Without patch (real) Without patch (sys)
> > 80 78 91 90
> > 63 62 93 91
> > 65 64 92 90
> > 67 65 93 90
> > 75 73 90 88
> > 75 73 91 89
> > 75 73 93 90
> > 74 73 89 87
> > 76 74 91 89
> > stddev 5.76146200581454 5.45690184791497 1.42400062421959 1.22474487139159
> > mean 72.2222222222222 70.5555555555556 91.4444444444444 89.3333333333333
> > median 75 73 91 90
> >
> With Patch (real) With patch (sys) Without patch (real) Without patch (sys)
> 80 78 91 90
> 63 62 93 91
> 65 64 92 90
> 67 65 93 90
> 75 73 90 88
> 75 73 91 89
> 75 73 93 90
> 74 73 89 87
> 76 74 91 89
> stddev 5.76146200581454 5.45690184791497 1.42400062421959 1.22474487139159
> mean 72.2222222222222 70.5555555555556 91.4444444444444 89.3333333333333
> median 75 73 91 90
Great.
How was that test done?
Simply deleting the files with nothing else running in parallel?
How does it behave if while the files are being deleted, there are
concurrent fsyncs on other files of the same subvolume, that is, while
the mutex is held?
Because that check without holding the mutex, is likely there for
performance reasons.
Thanks.
>
>
> Here's the perf data without being butchered.
>
> > Signed-off-by: Nikolay Borisov <nborisov@suse.com>
> > ---
> >
> > This passed full xfstest run and the performance results were obtained with the
> > following testcase:
> >
> > #!/bin/bash
> > for i in {1..10}; do
> > echo "Testun run : %i"
> > ./ltp/fsstress -z -d /media/scratch/ -f creat=100 -n 500000
> > sync
> > time rm -rf /media/scratch/*
> > sync
> > done
> >
> > fs/btrfs/tree-log.c | 4 ----
> > 1 file changed, 4 deletions(-)
> >
> > diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
> > index 6adcd8a2c5c7..61744d8af106 100644
> > --- a/fs/btrfs/tree-log.c
> > +++ b/fs/btrfs/tree-log.c
> > @@ -188,10 +188,6 @@ static int join_running_log_trans(struct btrfs_root *root)
> > {
> > int ret = -ENOENT;
> >
> > - smp_mb();
> > - if (!root->log_root)
> > - return -ENOENT;
> > -
> > mutex_lock(&root->log_mutex);
> > if (root->log_root) {
> > ret = 0;
> >
--
Filipe David Manana,
“Whether you think you can, or you think you can't — you're right.”
next prev parent reply other threads:[~2019-05-20 9:10 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-05-20 8:11 [PATCH] btrfs: Simplify join_running_log_trans Nikolay Borisov
2019-05-20 8:18 ` Nikolay Borisov
2019-05-20 9:10 ` Filipe Manana [this message]
2019-05-20 9:13 ` Nikolay Borisov
2019-05-20 14:42 ` Filipe Manana
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=CAL3q7H5MHYhrvLnqHagkmLS_cey6Bc-nZh-21ivGjbSPjZ76+Q@mail.gmail.com \
--to=fdmanana@gmail.com \
--cc=fdmanana@suse.com \
--cc=linux-btrfs@vger.kernel.org \
--cc=nborisov@suse.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.