netdev - Re: [PATCH net-next RFC 3/4] vhost: interrupt coalescing support

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <bf53ae2d-faca-a454-4df0-48296f8106b2@redhat.com>
Date:   Wed, 8 Mar 2017 11:25:21 +0800
From:   Jason Wang <jasowang@...hat.com>
To:     Willem de Bruijn <willemdebruijn.kernel@...il.com>
Cc:     Network Development <netdev@...r.kernel.org>,
        "Michael S. Tsirkin" <mst@...hat.com>,
        Willem de Bruijn <willemb@...gle.com>
Subject: Re: [PATCH net-next RFC 3/4] vhost: interrupt coalescing support



On 2017年03月07日 01:31, Willem de Bruijn wrote:
> On Mon, Mar 6, 2017 at 4:28 AM, Jason Wang <jasowang@...hat.com> wrote:
>>
>> On 2017年03月03日 22:39, Willem de Bruijn wrote:
>>> +void vhost_signal(struct vhost_dev *dev, struct vhost_virtqueue *vq);
>>> +static enum hrtimer_restart vhost_coalesce_timer(struct hrtimer *timer)
>>> +{
>>> +       struct vhost_virtqueue *vq =
>>> +               container_of(timer, struct vhost_virtqueue, ctimer);
>>> +
>>> +       if (mutex_trylock(&vq->mutex)) {
>>> +               vq->coalesce_frames = vq->max_coalesce_frames;
>>> +               vhost_signal(vq->dev, vq);
>>> +               mutex_unlock(&vq->mutex);
>>> +       }
>>> +
>>> +       /* TODO: restart if lock failed and not held by handle_tx */
>>> +       return HRTIMER_NORESTART;
>>> +}
>>> +
>>
>> Then we may lose an interrupt forever if no new tx request? I believe we
>> need e.g vhost_poll_queue() here.
> Absolutely, I need to fix this. The common case for failing to grab
> the lock is competition with handle_tx. With careful coding we can
> probably avoid scheduling another run with vhost_poll_queue in
> the common case.

Yes, probably add some checking after releasing the mutex_lock in 
handle_tx().

Thans

>
> Your patch v7 cancels the pending hrtimer at the start of handle_tx.
> I need to reintroduce that, and also only schedule a timer at the end
> of handle_tx, not immediately when vq->coalesce_frames becomes
> non-zero.