LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
From: Jason Wang <jasowang@redhat.com>
To: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Vincent Whitchurch <vincent.whitchurch@axis.com>,
	kernel@axis.com, kvm <kvm@vger.kernel.org>,
	virtualization <virtualization@lists.linux-foundation.org>,
	netdev <netdev@vger.kernel.org>,
	linux-kernel <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] vhost: add support for mandatory barriers
Date: Tue, 24 Aug 2021 10:56:26 +0800	[thread overview]
Message-ID: <CACGkMEvR6GVfgSCDvFWvHJ3UryN4GOMDQhWMSAAqVHsbfAfPiA@mail.gmail.com> (raw)
In-Reply-To: <20210823171609-mutt-send-email-mst@kernel.org>

On Tue, Aug 24, 2021 at 5:20 AM Michael S. Tsirkin <mst@redhat.com> wrote:
>
> On Mon, Aug 23, 2021 at 10:14:37AM +0200, Vincent Whitchurch wrote:
> > vhost always uses SMP-conditional barriers, but these may not be
> > sufficient when vhost is used to communicate between heterogeneous
> > processors in an AMP configuration, especially since they're NOPs on
> > !SMP builds.
> >
> > To solve this, use the virtio_*() barrier functions and ask them for
> > non-weak barriers if requested by userspace.
> >
> > Signed-off-by: Vincent Whitchurch <vincent.whitchurch@axis.com>
>
> I am inclined to say let's (ab)use VIRTIO_F_ORDER_PLATFORM for this.
> Jason what do you think?

Yes, it looks fine to me.

>
> Also is the use of DMA variants really the intended thing here? Could
> you point me at some examples please?

Yes, we need to know which setup we need.

Thanks

>
>
> > ---
> >  drivers/vhost/vhost.c      | 23 ++++++++++++++---------
> >  drivers/vhost/vhost.h      |  2 ++
> >  include/uapi/linux/vhost.h |  2 ++
> >  3 files changed, 18 insertions(+), 9 deletions(-)
> >
> > diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
> > index b9e853e6094d..f7172e1bc395 100644
> > --- a/drivers/vhost/vhost.c
> > +++ b/drivers/vhost/vhost.c
> > @@ -500,6 +500,7 @@ void vhost_dev_init(struct vhost_dev *dev,
> >               vq->indirect = NULL;
> >               vq->heads = NULL;
> >               vq->dev = dev;
> > +             vq->weak_barriers = true;
> >               mutex_init(&vq->mutex);
> >               vhost_vq_reset(dev, vq);
> >               if (vq->handle_kick)
> > @@ -1801,6 +1802,10 @@ long vhost_dev_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *argp)
> >               if (ctx)
> >                       eventfd_ctx_put(ctx);
> >               break;
> > +     case VHOST_SET_STRONG_BARRIERS:
> > +             for (i = 0; i < d->nvqs; ++i)
> > +                     d->vqs[i]->weak_barriers = false;
> > +             break;
> >       default:
> >               r = -ENOIOCTLCMD;
> >               break;
> > @@ -1927,7 +1932,7 @@ int vhost_log_write(struct vhost_virtqueue *vq, struct vhost_log *log,
> >       int i, r;
> >
> >       /* Make sure data written is seen before log. */
> > -     smp_wmb();
> > +     virtio_wmb(vq->weak_barriers);
> >
> >       if (vq->iotlb) {
> >               for (i = 0; i < count; i++) {
> > @@ -1964,7 +1969,7 @@ static int vhost_update_used_flags(struct vhost_virtqueue *vq)
> >               return -EFAULT;
> >       if (unlikely(vq->log_used)) {
> >               /* Make sure the flag is seen before log. */
> > -             smp_wmb();
> > +             virtio_wmb(vq->weak_barriers);
> >               /* Log used flag write. */
> >               used = &vq->used->flags;
> >               log_used(vq, (used - (void __user *)vq->used),
> > @@ -1982,7 +1987,7 @@ static int vhost_update_avail_event(struct vhost_virtqueue *vq, u16 avail_event)
> >       if (unlikely(vq->log_used)) {
> >               void __user *used;
> >               /* Make sure the event is seen before log. */
> > -             smp_wmb();
> > +             virtio_wmb(vq->weak_barriers);
> >               /* Log avail event write */
> >               used = vhost_avail_event(vq);
> >               log_used(vq, (used - (void __user *)vq->used),
> > @@ -2228,7 +2233,7 @@ int vhost_get_vq_desc(struct vhost_virtqueue *vq,
> >               /* Only get avail ring entries after they have been
> >                * exposed by guest.
> >                */
> > -             smp_rmb();
> > +             virtio_rmb(vq->weak_barriers);
> >       }
> >
> >       /* Grab the next descriptor number they're advertising, and increment
> > @@ -2367,7 +2372,7 @@ static int __vhost_add_used_n(struct vhost_virtqueue *vq,
> >       }
> >       if (unlikely(vq->log_used)) {
> >               /* Make sure data is seen before log. */
> > -             smp_wmb();
> > +             virtio_wmb(vq->weak_barriers);
> >               /* Log used ring entry write. */
> >               log_used(vq, ((void __user *)used - (void __user *)vq->used),
> >                        count * sizeof *used);
> > @@ -2402,14 +2407,14 @@ int vhost_add_used_n(struct vhost_virtqueue *vq, struct vring_used_elem *heads,
> >       r = __vhost_add_used_n(vq, heads, count);
> >
> >       /* Make sure buffer is written before we update index. */
> > -     smp_wmb();
> > +     virtio_wmb(vq->weak_barriers);
> >       if (vhost_put_used_idx(vq)) {
> >               vq_err(vq, "Failed to increment used idx");
> >               return -EFAULT;
> >       }
> >       if (unlikely(vq->log_used)) {
> >               /* Make sure used idx is seen before log. */
> > -             smp_wmb();
> > +             virtio_wmb(vq->weak_barriers);
> >               /* Log used index update. */
> >               log_used(vq, offsetof(struct vring_used, idx),
> >                        sizeof vq->used->idx);
> > @@ -2428,7 +2433,7 @@ static bool vhost_notify(struct vhost_dev *dev, struct vhost_virtqueue *vq)
> >       /* Flush out used index updates. This is paired
> >        * with the barrier that the Guest executes when enabling
> >        * interrupts. */
> > -     smp_mb();
> > +     virtio_mb(vq->weak_barriers);
> >
> >       if (vhost_has_feature(vq, VIRTIO_F_NOTIFY_ON_EMPTY) &&
> >           unlikely(vq->avail_idx == vq->last_avail_idx))
> > @@ -2530,7 +2535,7 @@ bool vhost_enable_notify(struct vhost_dev *dev, struct vhost_virtqueue *vq)
> >       }
> >       /* They could have slipped one in as we were doing that: make
> >        * sure it's written, then check again. */
> > -     smp_mb();
> > +     virtio_mb(vq->weak_barriers);
> >       r = vhost_get_avail_idx(vq, &avail_idx);
> >       if (r) {
> >               vq_err(vq, "Failed to check avail idx at %p: %d\n",
> > diff --git a/drivers/vhost/vhost.h b/drivers/vhost/vhost.h
> > index 638bb640d6b4..5bd20d0db457 100644
> > --- a/drivers/vhost/vhost.h
> > +++ b/drivers/vhost/vhost.h
> > @@ -108,6 +108,8 @@ struct vhost_virtqueue {
> >       bool log_used;
> >       u64 log_addr;
> >
> > +     bool weak_barriers;
> > +
> >       struct iovec iov[UIO_MAXIOV];
> >       struct iovec iotlb_iov[64];
> >       struct iovec *indirect;
> > diff --git a/include/uapi/linux/vhost.h b/include/uapi/linux/vhost.h
> > index c998860d7bbc..4b8656307f51 100644
> > --- a/include/uapi/linux/vhost.h
> > +++ b/include/uapi/linux/vhost.h
> > @@ -97,6 +97,8 @@
> >  #define VHOST_SET_BACKEND_FEATURES _IOW(VHOST_VIRTIO, 0x25, __u64)
> >  #define VHOST_GET_BACKEND_FEATURES _IOR(VHOST_VIRTIO, 0x26, __u64)
> >
> > +#define VHOST_SET_STRONG_BARRIERS _IO(VHOST_VIRTIO, 0x27)
> > +
> >  /* VHOST_NET specific defines */
> >
> >  /* Attach virtio net ring to a raw socket, or tap device.
> > --
> > 2.28.0
>


  reply	other threads:[~2021-08-24  2:56 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-08-23  8:14 Vincent Whitchurch
2021-08-23 21:19 ` Michael S. Tsirkin
2021-08-24  2:56   ` Jason Wang [this message]
2021-08-24  7:23   ` Vincent Whitchurch

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=CACGkMEvR6GVfgSCDvFWvHJ3UryN4GOMDQhWMSAAqVHsbfAfPiA@mail.gmail.com \
    --to=jasowang@redhat.com \
    --cc=kernel@axis.com \
    --cc=kvm@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mst@redhat.com \
    --cc=netdev@vger.kernel.org \
    --cc=vincent.whitchurch@axis.com \
    --cc=virtualization@lists.linux-foundation.org \
    --subject='Re: [PATCH] vhost: add support for mandatory barriers' \
    /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

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).