Re: [PATCH] vhost: add support for mandatory barriers

From: Jason Wang
Date: Mon Aug 23 2021 - 22:57:14 EST

On Tue, Aug 24, 2021 at 5:20 AM Michael S. Tsirkin <mst@xxxxxxxxxx> 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@xxxxxxxx>
>
> 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
>