From: "Michael S. Tsirkin" <mst@redhat.com>
To: Bart Van Assche <bvanassche@acm.org>
Cc: Asias He <asias@redhat.com>, kvm-devel <kvm@vger.kernel.org>
Subject: Re: [PATCH 2/2] vhost: Avoid that vhost_work_flush() returns early
Date: Wed, 14 Aug 2013 14:46:40 +0300 [thread overview]
Message-ID: <20130814114640.GG5430@redhat.com> (raw)
In-Reply-To: <520B2BC0.3050807@acm.org>
On Wed, Aug 14, 2013 at 09:03:28AM +0200, Bart Van Assche wrote:
> If two or more items are queued on dev->work_list before
> vhost_worker() starts processing these then the value of
> work->done_seq will be set to the sequence number of a work item
> that has not yet been processed. Avoid this by letting
> vhost_worker() count the number of items that have already been
> processed.
>
> Signed-off-by: Bart Van Assche <bvanassche@acm.org>
> Cc: Michael S. Tsirkin <mst@redhat.com>
> Cc: Asias He <asias@redhat.com>
I'm confused by this explanation.
done_seq is set here:
if (work) {
work->done_seq = seq;
if (work->flushing)
wake_up_all(&work->done);
}
and work is set here:
if (!list_empty(&dev->work_list)) {
work = list_first_entry(&dev->work_list,
struct vhost_work, node);
list_del_init(&work->node);
seq = work->queue_seq;
}
this work is processed on the next line:
if (work) {
__set_current_state(TASK_RUNNING);
work->fn(work);
if (need_resched())
schedule();
}
so how do we end up with a sequence of a work item
that isn't processed?
> ---
> drivers/vhost/vhost.c | 5 ++---
> 1 file changed, 2 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
> index e7ffc10..11d668a 100644
> --- a/drivers/vhost/vhost.c
> +++ b/drivers/vhost/vhost.c
> @@ -202,7 +202,7 @@ static int vhost_worker(void *data)
> {
> struct vhost_dev *dev = data;
> struct vhost_work *work;
> - unsigned seq;
> + unsigned seq = 0;
> mm_segment_t oldfs = get_fs();
>
> set_fs(USER_DS);
> @@ -216,14 +216,13 @@ static int vhost_worker(void *data)
> work = list_first_entry(&dev->work_list,
> struct vhost_work, node);
> list_del_init(&work->node);
> - seq = work->queue_seq;
> spin_unlock_irq(&dev->work_lock);
>
> __set_current_state(TASK_RUNNING);
> work->fn(work);
>
> spin_lock_irq(&dev->work_lock);
> - work->done_seq = seq;
> + work->done_seq = ++seq;
> if (work->flushing)
> wake_up_all(&work->done);
> }
> --
> 1.7.10.4
next prev parent reply other threads:[~2013-08-14 11:45 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-08-14 7:01 [PATCH 0/2] vhost_worker fixes Bart Van Assche
2013-08-14 7:02 ` [PATCH 1/2] vhost: Reduce vhost_work_flush() wakeup latency Bart Van Assche
2013-08-14 11:37 ` Michael S. Tsirkin
2013-08-14 15:22 ` Bart Van Assche
2013-08-14 17:58 ` Michael S. Tsirkin
2013-08-15 1:30 ` Asias He
2013-08-15 6:13 ` Bart Van Assche
2013-08-15 7:35 ` Michael S. Tsirkin
2013-08-14 7:03 ` [PATCH 2/2] vhost: Avoid that vhost_work_flush() returns early Bart Van Assche
2013-08-14 11:46 ` Michael S. Tsirkin [this message]
2013-08-14 15:19 ` Bart Van Assche
2013-08-14 11:39 ` [PATCH 0/2] vhost_worker fixes Michael S. Tsirkin
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=20130814114640.GG5430@redhat.com \
--to=mst@redhat.com \
--cc=asias@redhat.com \
--cc=bvanassche@acm.org \
--cc=kvm@vger.kernel.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