Re: [PATCH for-7.2] vhost: enable vrings in vhost_dev_start() for vhost-user devices

[Michael S. Tsirkin]

On Wed, Nov 30, 2022 at 04:03:28PM -0500, Stefan Hajnoczi wrote:
> On Fri, Nov 25, 2022 at 09:12:43AM +0100, Stefano Garzarella wrote:
> > On Thu, Nov 24, 2022 at 01:36:29PM -0500, Stefan Hajnoczi wrote:
> > > On Wed, Nov 23, 2022 at 02:16:30PM +0100, Stefano Garzarella wrote:
> > > > Commit 02b61f38d3 ("hw/virtio: incorporate backend features in 
> > > > features")
> > > > properly negotiates VHOST_USER_F_PROTOCOL_FEATURES with the vhost-user
> > > > backend, but we forgot to enable vrings as specified in
> > > > docs/interop/vhost-user.rst:
> > > > 
> > > >     If ``VHOST_USER_F_PROTOCOL_FEATURES`` has not been negotiated, the
> > > >     ring starts directly in the enabled state.
> > > > 
> > > >     If ``VHOST_USER_F_PROTOCOL_FEATURES`` has been negotiated, the ring 
> > > > is
> > > >     initialized in a disabled state and is enabled by
> > > >     ``VHOST_USER_SET_VRING_ENABLE`` with parameter 1.
> > > > 
> > > > Some vhost-user front-ends already did this by calling
> > > > vhost_ops.vhost_set_vring_enable() directly:
> > > > - backends/cryptodev-vhost.c
> > > > - hw/net/virtio-net.c
> > > > - hw/virtio/vhost-user-gpio.c
> > > > 
> > > > But most didn't do that, so we would leave the vrings disabled and some
> > > > backends would not work. We observed this issue with the rust version of
> > > > virtiofsd [1], which uses the event loop [2] provided by the
> > > > vhost-user-backend crate where requests are not processed if vring is
> > > > not enabled.
> > > > 
> > > > Let's fix this issue by enabling the vrings in vhost_dev_start() for
> > > > vhost-user front-ends that don't already do this directly. Same thing
> > > > also in vhost_dev_stop() where we disable vrings.
> > > > 
> > > > [1] https://gitlab.com/virtio-fs/virtiofsd
> > > > [2] 
> > > > https://github.com/rust-vmm/vhost/blob/240fc2966/crates/vhost-user-backend/src/event_loop.rs#L217
> > > > 
> > > > Fixes: 02b61f38d3 ("hw/virtio: incorporate backend features in 
> > > > features")
> > > > Reported-by: German Maglione <gmaglione@redhat.com>
> > > > Tested-by: German Maglione <gmaglione@redhat.com>
> > > > Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
> > > > ---
> > > >  include/hw/virtio/vhost.h      |  6 +++--
> > > >  backends/cryptodev-vhost.c     |  4 ++--
> > > >  backends/vhost-user.c          |  4 ++--
> > > >  hw/block/vhost-user-blk.c      |  4 ++--
> > > >  hw/net/vhost_net.c             |  8 +++----
> > > >  hw/scsi/vhost-scsi-common.c    |  4 ++--
> > > >  hw/virtio/vhost-user-fs.c      |  4 ++--
> > > >  hw/virtio/vhost-user-gpio.c    |  4 ++--
> > > >  hw/virtio/vhost-user-i2c.c     |  4 ++--
> > > >  hw/virtio/vhost-user-rng.c     |  4 ++--
> > > >  hw/virtio/vhost-vsock-common.c |  4 ++--
> > > >  hw/virtio/vhost.c              | 44 ++++++++++++++++++++++++++++++----
> > > >  hw/virtio/trace-events         |  4 ++--
> > > >  13 files changed, 67 insertions(+), 31 deletions(-)
> > > > 
> > > > diff --git a/include/hw/virtio/vhost.h b/include/hw/virtio/vhost.h
> > > > index 353252ac3e..67a6807fac 100644
> > > > --- a/include/hw/virtio/vhost.h
> > > > +++ b/include/hw/virtio/vhost.h
> > > > @@ -184,24 +184,26 @@ static inline bool vhost_dev_is_started(struct 
> > > > vhost_dev *hdev)
> > > >   * vhost_dev_start() - start the vhost device
> > > >   * @hdev: common vhost_dev structure
> > > >   * @vdev: the VirtIODevice structure
> > > > + * @vrings: true to have vrings enabled in this call
> > > >   *
> > > >   * Starts the vhost device. From this point VirtIO feature negotiation
> > > >   * can start and the device can start processing VirtIO transactions.
> > > >   *
> > > >   * Return: 0 on success, < 0 on error.
> > > >   */
> > > > -int vhost_dev_start(struct vhost_dev *hdev, VirtIODevice *vdev);
> > > > +int vhost_dev_start(struct vhost_dev *hdev, VirtIODevice *vdev, bool 
> > > > vrings);
> > > > 
> > > >  /**
> > > >   * vhost_dev_stop() - stop the vhost device
> > > >   * @hdev: common vhost_dev structure
> > > >   * @vdev: the VirtIODevice structure
> > > > + * @vrings: true to have vrings disabled in this call
> > > >   *
> > > >   * Stop the vhost device. After the device is stopped the notifiers
> > > >   * can be disabled (@vhost_dev_disable_notifiers) and the device can
> > > >   * be torn down (@vhost_dev_cleanup).
> > > >   */
> > > > -void vhost_dev_stop(struct vhost_dev *hdev, VirtIODevice *vdev);
> > > > +void vhost_dev_stop(struct vhost_dev *hdev, VirtIODevice *vdev, bool 
> > > > vrings);
> > > > 
> > > >  /**
> > > >   * DOC: vhost device configuration handling
> > > > diff --git a/backends/cryptodev-vhost.c b/backends/cryptodev-vhost.c
> > > > index bc13e466b4..572f87b3be 100644
> > > > --- a/backends/cryptodev-vhost.c
> > > > +++ b/backends/cryptodev-vhost.c
> > > > @@ -94,7 +94,7 @@ cryptodev_vhost_start_one(CryptoDevBackendVhost 
> > > > *crypto,
> > > >          goto fail_notifiers;
> > > >      }
> > > > 
> > > > -    r = vhost_dev_start(&crypto->dev, dev);
> > > > +    r = vhost_dev_start(&crypto->dev, dev, false);
> > > >      if (r < 0) {
> > > >          goto fail_start;
> > > >      }
> > > > @@ -111,7 +111,7 @@ static void
> > > >  cryptodev_vhost_stop_one(CryptoDevBackendVhost *crypto,
> > > >                                   VirtIODevice *dev)
> > > >  {
> > > > -    vhost_dev_stop(&crypto->dev, dev);
> > > > +    vhost_dev_stop(&crypto->dev, dev, false);
> > > >      vhost_dev_disable_notifiers(&crypto->dev, dev);
> > > >  }
> > > > 
> > > > diff --git a/backends/vhost-user.c b/backends/vhost-user.c
> > > > index 5dedb2d987..7bfcaef976 100644
> > > > --- a/backends/vhost-user.c
> > > > +++ b/backends/vhost-user.c
> > > > @@ -85,7 +85,7 @@ vhost_user_backend_start(VhostUserBackend *b)
> > > >      }
> > > > 
> > > >      b->dev.acked_features = b->vdev->guest_features;
> > > > -    ret = vhost_dev_start(&b->dev, b->vdev);
> > > > +    ret = vhost_dev_start(&b->dev, b->vdev, true);
> > > >      if (ret < 0) {
> > > >          error_report("Error start vhost dev");
> > > >          goto err_guest_notifiers;
> > > > @@ -120,7 +120,7 @@ vhost_user_backend_stop(VhostUserBackend *b)
> > > >          return;
> > > >      }
> > > > 
> > > > -    vhost_dev_stop(&b->dev, b->vdev);
> > > > +    vhost_dev_stop(&b->dev, b->vdev, true);
> > > > 
> > > >      if (k->set_guest_notifiers) {
> > > >          ret = k->set_guest_notifiers(qbus->parent,
> > > > diff --git a/hw/block/vhost-user-blk.c b/hw/block/vhost-user-blk.c
> > > > index 0d5190accf..1177064631 100644
> > > > --- a/hw/block/vhost-user-blk.c
> > > > +++ b/hw/block/vhost-user-blk.c
> > > > @@ -178,7 +178,7 @@ static int vhost_user_blk_start(VirtIODevice *vdev, 
> > > > Error **errp)
> > > >      }
> > > > 
> > > >      s->dev.vq_index_end = s->dev.nvqs;
> > > > -    ret = vhost_dev_start(&s->dev, vdev);
> > > > +    ret = vhost_dev_start(&s->dev, vdev, true);
> > > >      if (ret < 0) {
> > > >          error_setg_errno(errp, -ret, "Error starting vhost");
> > > >          goto err_guest_notifiers;
> > > > @@ -213,7 +213,7 @@ static void vhost_user_blk_stop(VirtIODevice *vdev)
> > > >          return;
> > > >      }
> > > > 
> > > > -    vhost_dev_stop(&s->dev, vdev);
> > > > +    vhost_dev_stop(&s->dev, vdev, true);
> > > > 
> > > >      ret = k->set_guest_notifiers(qbus->parent, s->dev.nvqs, false);
> > > >      if (ret < 0) {
> > > > diff --git a/hw/net/vhost_net.c b/hw/net/vhost_net.c
> > > > index 26e4930676..043058ff43 100644
> > > > --- a/hw/net/vhost_net.c
> > > > +++ b/hw/net/vhost_net.c
> > > > @@ -259,7 +259,7 @@ static int vhost_net_start_one(struct vhost_net 
> > > > *net,
> > > >          goto fail_notifiers;
> > > >      }
> > > > 
> > > > -    r = vhost_dev_start(&net->dev, dev);
> > > > +    r = vhost_dev_start(&net->dev, dev, false);
> > > >      if (r < 0) {
> > > >          goto fail_start;
> > > >      }
> > > > @@ -308,7 +308,7 @@ fail:
> > > >      if (net->nc->info->poll) {
> > > >          net->nc->info->poll(net->nc, true);
> > > >      }
> > > > -    vhost_dev_stop(&net->dev, dev);
> > > > +    vhost_dev_stop(&net->dev, dev, false);
> > > >  fail_start:
> > > >      vhost_dev_disable_notifiers(&net->dev, dev);
> > > >  fail_notifiers:
> > > > @@ -329,7 +329,7 @@ static void vhost_net_stop_one(struct vhost_net 
> > > > *net,
> > > >      if (net->nc->info->poll) {
> > > >          net->nc->info->poll(net->nc, true);
> > > >      }
> > > > -    vhost_dev_stop(&net->dev, dev);
> > > > +    vhost_dev_stop(&net->dev, dev, false);
> > > >      if (net->nc->info->stop) {
> > > >          net->nc->info->stop(net->nc);
> > > >      }
> > > > @@ -606,7 +606,7 @@ err_start:
> > > >          assert(r >= 0);
> > > >      }
> > > > 
> > > > -    vhost_dev_stop(&net->dev, vdev);
> > > > +    vhost_dev_stop(&net->dev, vdev, false);
> > > > 
> > > >      return r;
> > > >  }
> > > > diff --git a/hw/scsi/vhost-scsi-common.c b/hw/scsi/vhost-scsi-common.c
> > > > index 767f827e55..18ea5dcfa1 100644
> > > > --- a/hw/scsi/vhost-scsi-common.c
> > > > +++ b/hw/scsi/vhost-scsi-common.c
> > > > @@ -68,7 +68,7 @@ int vhost_scsi_common_start(VHostSCSICommon *vsc)
> > > >          goto err_guest_notifiers;
> > > >      }
> > > > 
> > > > -    ret = vhost_dev_start(&vsc->dev, vdev);
> > > > +    ret = vhost_dev_start(&vsc->dev, vdev, true);
> > > >      if (ret < 0) {
> > > >          error_report("Error start vhost dev");
> > > >          goto err_guest_notifiers;
> > > > @@ -101,7 +101,7 @@ void vhost_scsi_common_stop(VHostSCSICommon *vsc)
> > > >      VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus);
> > > >      int ret = 0;
> > > > 
> > > > -    vhost_dev_stop(&vsc->dev, vdev);
> > > > +    vhost_dev_stop(&vsc->dev, vdev, true);
> > > > 
> > > >      if (k->set_guest_notifiers) {
> > > >          ret = k->set_guest_notifiers(qbus->parent, vsc->dev.nvqs, 
> > > > false);
> > > > diff --git a/hw/virtio/vhost-user-fs.c b/hw/virtio/vhost-user-fs.c
> > > > index dc4014cdef..d97b179e6f 100644
> > > > --- a/hw/virtio/vhost-user-fs.c
> > > > +++ b/hw/virtio/vhost-user-fs.c
> > > > @@ -76,7 +76,7 @@ static void vuf_start(VirtIODevice *vdev)
> > > >      }
> > > > 
> > > >      fs->vhost_dev.acked_features = vdev->guest_features;
> > > > -    ret = vhost_dev_start(&fs->vhost_dev, vdev);
> > > > +    ret = vhost_dev_start(&fs->vhost_dev, vdev, true);
> > > >      if (ret < 0) {
> > > >          error_report("Error starting vhost: %d", -ret);
> > > >          goto err_guest_notifiers;
> > > > @@ -110,7 +110,7 @@ static void vuf_stop(VirtIODevice *vdev)
> > > >          return;
> > > >      }
> > > > 
> > > > -    vhost_dev_stop(&fs->vhost_dev, vdev);
> > > > +    vhost_dev_stop(&fs->vhost_dev, vdev, true);
> > > > 
> > > >      ret = k->set_guest_notifiers(qbus->parent, fs->vhost_dev.nvqs, 
> > > > false);
> > > >      if (ret < 0) {
> > > > diff --git a/hw/virtio/vhost-user-gpio.c b/hw/virtio/vhost-user-gpio.c
> > > > index 5851cb3bc9..0b40ebd15a 100644
> > > > --- a/hw/virtio/vhost-user-gpio.c
> > > > +++ b/hw/virtio/vhost-user-gpio.c
> > > > @@ -81,7 +81,7 @@ static int vu_gpio_start(VirtIODevice *vdev)
> > > >       */
> > > >      vhost_ack_features(&gpio->vhost_dev, feature_bits, 
> > > > vdev->guest_features);
> > > > 
> > > > -    ret = vhost_dev_start(&gpio->vhost_dev, vdev);
> > > > +    ret = vhost_dev_start(&gpio->vhost_dev, vdev, false);
> > > >      if (ret < 0) {
> > > >          error_report("Error starting vhost-user-gpio: %d", ret);
> > > >          goto err_guest_notifiers;
> > > > @@ -139,7 +139,7 @@ static void vu_gpio_stop(VirtIODevice *vdev)
> > > >          return;
> > > >      }
> > > > 
> > > > -    vhost_dev_stop(vhost_dev, vdev);
> > > > +    vhost_dev_stop(vhost_dev, vdev, false);
> > > > 
> > > >      ret = k->set_guest_notifiers(qbus->parent, vhost_dev->nvqs, false);
> > > >      if (ret < 0) {
> > > > diff --git a/hw/virtio/vhost-user-i2c.c b/hw/virtio/vhost-user-i2c.c
> > > > index 1c9f3d20dc..dc5c828ba6 100644
> > > > --- a/hw/virtio/vhost-user-i2c.c
> > > > +++ b/hw/virtio/vhost-user-i2c.c
> > > > @@ -46,7 +46,7 @@ static void vu_i2c_start(VirtIODevice *vdev)
> > > > 
> > > >      i2c->vhost_dev.acked_features = vdev->guest_features;
> > > > 
> > > > -    ret = vhost_dev_start(&i2c->vhost_dev, vdev);
> > > > +    ret = vhost_dev_start(&i2c->vhost_dev, vdev, true);
> > > >      if (ret < 0) {
> > > >          error_report("Error starting vhost-user-i2c: %d", -ret);
> > > >          goto err_guest_notifiers;
> > > > @@ -80,7 +80,7 @@ static void vu_i2c_stop(VirtIODevice *vdev)
> > > >          return;
> > > >      }
> > > > 
> > > > -    vhost_dev_stop(&i2c->vhost_dev, vdev);
> > > > +    vhost_dev_stop(&i2c->vhost_dev, vdev, true);
> > > > 
> > > >      ret = k->set_guest_notifiers(qbus->parent, i2c->vhost_dev.nvqs, 
> > > > false);
> > > >      if (ret < 0) {
> > > > diff --git a/hw/virtio/vhost-user-rng.c b/hw/virtio/vhost-user-rng.c
> > > > index f9084cde58..201a39e220 100644
> > > > --- a/hw/virtio/vhost-user-rng.c
> > > > +++ b/hw/virtio/vhost-user-rng.c
> > > > @@ -47,7 +47,7 @@ static void vu_rng_start(VirtIODevice *vdev)
> > > >      }
> > > > 
> > > >      rng->vhost_dev.acked_features = vdev->guest_features;
> > > > -    ret = vhost_dev_start(&rng->vhost_dev, vdev);
> > > > +    ret = vhost_dev_start(&rng->vhost_dev, vdev, true);
> > > >      if (ret < 0) {
> > > >          error_report("Error starting vhost-user-rng: %d", -ret);
> > > >          goto err_guest_notifiers;
> > > > @@ -81,7 +81,7 @@ static void vu_rng_stop(VirtIODevice *vdev)
> > > >          return;
> > > >      }
> > > > 
> > > > -    vhost_dev_stop(&rng->vhost_dev, vdev);
> > > > +    vhost_dev_stop(&rng->vhost_dev, vdev, true);
> > > > 
> > > >      ret = k->set_guest_notifiers(qbus->parent, rng->vhost_dev.nvqs, 
> > > > false);
> > > >      if (ret < 0) {
> > > > diff --git a/hw/virtio/vhost-vsock-common.c 
> > > > b/hw/virtio/vhost-vsock-common.c
> > > > index a67a275de2..d21c72b401 100644
> > > > --- a/hw/virtio/vhost-vsock-common.c
> > > > +++ b/hw/virtio/vhost-vsock-common.c
> > > > @@ -70,7 +70,7 @@ int vhost_vsock_common_start(VirtIODevice *vdev)
> > > >      }
> > > > 
> > > >      vvc->vhost_dev.acked_features = vdev->guest_features;
> > > > -    ret = vhost_dev_start(&vvc->vhost_dev, vdev);
> > > > +    ret = vhost_dev_start(&vvc->vhost_dev, vdev, true);
> > > >      if (ret < 0) {
> > > >          error_report("Error starting vhost: %d", -ret);
> > > >          goto err_guest_notifiers;
> > > > @@ -105,7 +105,7 @@ void vhost_vsock_common_stop(VirtIODevice *vdev)
> > > >          return;
> > > >      }
> > > > 
> > > > -    vhost_dev_stop(&vvc->vhost_dev, vdev);
> > > > +    vhost_dev_stop(&vvc->vhost_dev, vdev, true);
> > > > 
> > > >      ret = k->set_guest_notifiers(qbus->parent, vvc->vhost_dev.nvqs, 
> > > > false);
> > > >      if (ret < 0) {
> > > > diff --git a/hw/virtio/vhost.c b/hw/virtio/vhost.c
> > > > index d1c4c20b8c..7fb008bc9e 100644
> > > > --- a/hw/virtio/vhost.c
> > > > +++ b/hw/virtio/vhost.c
> > > > @@ -1777,15 +1777,36 @@ int vhost_dev_get_inflight(struct vhost_dev 
> > > > *dev, uint16_t queue_size,
> > > >      return 0;
> > > >  }
> > > > 
> > > > +static int vhost_dev_set_vring_enable(struct vhost_dev *hdev, int 
> > > > enable)
> > > 
> > > There is a similarly-named vhost_set_vring_enable(NetClientState *nc,
> > > int enable) function which is actually part of vhost_net. Please rename
> > > it to vhost_net_set_vring_enable().
> > 
> > Should I rename it in this patch?
> > 
> > > It should probably call
> > > vhost_dev_set_vring_enable().
> > 
> > Ehm, the idea of this patch was to touch as little as possible to avoid new
> > regressions.
> > 
> > Also, the semantics of vhost_dev_set_vring_enable() was meant to keep
> > vhost_dev_start()/vhost_dev_stop() simple, not to be exposed to frontends.
> > (maybe I should have written it, sorry about that).
> > 
> > However I agree that we should clean up vhost-net and also the other
> > frontends as Raphael also suggested, but honestly I'm scared to do that now
> > in this patch...
> > 
> > What I would have wanted to do, would be similar to what we do for
> > vhost-vdpa: call SET_VRING_ENABLE in the vhost_ops->vhost_dev_start()
> > callback of vhost-user.c.
> > Removing all the call to vhost_ops->vhost_set_vring_enable() in the
> > frontends, but I think it's too risky to do that now.
> > 
> > > 
> > > > +{
> > > > +    if (!hdev->vhost_ops->vhost_set_vring_enable) {
> > > > +        return 0;
> > > > +    }
> > > > +
> > > > +    /*
> > > > +     * For vhost-user devices, if VHOST_USER_F_PROTOCOL_FEATURES has 
> > > > not
> > > > +     * been negotiated, the rings start directly in the enabled state, 
> > > > and
> > > > +     * .vhost_set_vring_enable callback will fail since
> > > > +     * VHOST_USER_SET_VRING_ENABLE is not supported.
> > > > +     */
> > > > +    if (hdev->vhost_ops->backend_type == VHOST_BACKEND_TYPE_USER &&
> > > > +        !virtio_has_feature(hdev->backend_features,
> > > > +                            VHOST_USER_F_PROTOCOL_FEATURES)) {
> > > > +        return 0;
> > > > +    }
> > > 
> > > These semantics are the opposite of vhost_user_set_vring_enable():
> > > 
> > >  if (!virtio_has_feature(dev->features, VHOST_USER_F_PROTOCOL_FEATURES)) {
> > >      return -EINVAL;
> > >  }
> > > 
> > > Please make vhost_user_set_vring_enable() and
> > > vhost_dev_set_vring_enable() consistent. Code gets really confusing when
> > > layers have different semantics for the same operation.
> > 
> > It's the opposite precisely because we shouldn't let
> > vhost_dev_start()/vhost_dev_stop() fail if
> > vhost_ops->vhost_set_vring_enable() can't be called because it would fail.
> > 
> > If I do it this way, then I have to put the check inside
> > vhost_dev_start()/vhost_dev_stop(), and at this point I remove the function
> > that would be useless (just a wrapper of
> > hdev->vhost_ops->vhost_set_vring_enable).
> > Actually this was the first implementation I did, then I added the function
> > just to have vhost_dev_start()/vhost_dev_stop() cleaner and to avoid
> > duplicating the check.
> > 
> > > 
> > > > +
> > > > +    return hdev->vhost_ops->vhost_set_vring_enable(hdev, enable);
> > > > +}
> > > 
> > > The return value is hard to understand. An error return is only returned
> > > by vhost-user devices with VHOST_USER_F_PROTOCOL_FEATURES. There are
> > > other cases that seem like they should return an error but return
> > > success instead. For example, when called with enable=false on a
> > > non-VHOST_USER_F_PROTOCOL_FEATURES device (e.g. vhost-kernel or legacy
> > > vhost-user) we return success even though the vring wasn't disabled.
> > 
> > As I explained above, the idea was not to expose this function outside, but
> > to use it only in vhost_dev_start()/vhost_dev_stop(). So the return value is
> > 0 both when it has successes and when there is no need/way to enable/disable
> > the vrings.
> > 
> > Perhaps since it is confusing, I will remove the function and put the code
> > directly into vhost_dev_start()/vhost_dev_stop().
> > 
> > What do you think?
> 
> It's late now. We can merge it as-is.
> 
> I think this patch makes the vhost code even harder to understand and
> it's important to do the clean ups that have been discussed for 8.0.

Yes I agree. Only acked because we don't have better ideas on how to
fix vmstate issues without breaking CI.

> Will you work on the changes we discussed for 8.0?
> 
> Thanks,
> Stefan

-- 
MST