diff mbox series

[for-9.0,v2] vdpa-dev: Fix initialisation order to restore VDUSE compatibility

Message ID 20240315140331.59227-1-kwolf@redhat.com (mailing list archive)
State New, archived
Headers show
Series [for-9.0,v2] vdpa-dev: Fix initialisation order to restore VDUSE compatibility | expand

Commit Message

Kevin Wolf March 15, 2024, 2:03 p.m. UTC
VDUSE requires that virtqueues are first enabled before the DRIVER_OK
status flag is set; with the current API of the kernel module, it is
impossible to enable the opposite order in our block export code because
userspace is not notified when a virtqueue is enabled.

This requirement also mathces the normal initialisation order as done by
the generic vhost code in QEMU. However, commit 6c482547 accidentally
changed the order for vdpa-dev and broke access to VDUSE devices with
this.

This changes vdpa-dev to use the normal order again and use the standard
vhost callback .vhost_set_vring_enable for this. VDUSE devices can be
used with vdpa-dev again after this fix.

vhost_net intentionally avoided enabling the vrings for vdpa and does
this manually later while it does enable them for other vhost backends.
Reflect this in the vhost_net code and return early for vdpa, so that
the behaviour doesn't change for this device.

Cc: qemu-stable@nongnu.org
Fixes: 6c4825476a4351530bcac17abab72295b75ffe98
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
---
v2:
- Actually make use of the @enable parameter
- Change vhost_net to preserve the current behaviour

 hw/net/vhost_net.c     | 10 ++++++++++
 hw/virtio/vdpa-dev.c   |  5 +----
 hw/virtio/vhost-vdpa.c | 27 +++++++++++++++++++++++++--
 hw/virtio/vhost.c      |  8 +++++++-
 4 files changed, 43 insertions(+), 7 deletions(-)

Comments

Stefano Garzarella March 15, 2024, 3:07 p.m. UTC | #1
On Fri, Mar 15, 2024 at 03:03:31PM +0100, Kevin Wolf wrote:
>VDUSE requires that virtqueues are first enabled before the DRIVER_OK
>status flag is set; with the current API of the kernel module, it is
>impossible to enable the opposite order in our block export code because
>userspace is not notified when a virtqueue is enabled.
>
>This requirement also mathces the normal initialisation order as done by
>the generic vhost code in QEMU. However, commit 6c482547 accidentally
>changed the order for vdpa-dev and broke access to VDUSE devices with
>this.
>
>This changes vdpa-dev to use the normal order again and use the standard
>vhost callback .vhost_set_vring_enable for this. VDUSE devices can be
>used with vdpa-dev again after this fix.
>
>vhost_net intentionally avoided enabling the vrings for vdpa and does
>this manually later while it does enable them for other vhost backends.
>Reflect this in the vhost_net code and return early for vdpa, so that
>the behaviour doesn't change for this device.
>
>Cc: qemu-stable@nongnu.org
>Fixes: 6c4825476a4351530bcac17abab72295b75ffe98
>Signed-off-by: Kevin Wolf <kwolf@redhat.com>
>---
>v2:
>- Actually make use of the @enable parameter
>- Change vhost_net to preserve the current behaviour
>
> hw/net/vhost_net.c     | 10 ++++++++++
> hw/virtio/vdpa-dev.c   |  5 +----
> hw/virtio/vhost-vdpa.c | 27 +++++++++++++++++++++++++--
> hw/virtio/vhost.c      |  8 +++++++-
> 4 files changed, 43 insertions(+), 7 deletions(-)
>
>diff --git a/hw/net/vhost_net.c b/hw/net/vhost_net.c
>index e8e1661646..fd1a93701a 100644
>--- a/hw/net/vhost_net.c
>+++ b/hw/net/vhost_net.c
>@@ -541,6 +541,16 @@ int vhost_set_vring_enable(NetClientState *nc, int enable)
>     VHostNetState *net = get_vhost_net(nc);
>     const VhostOps *vhost_ops = net->dev.vhost_ops;
>
>+    /*
>+     * vhost-vdpa network devices need to enable dataplane virtqueues after
>+     * DRIVER_OK, so they can recover device state before starting dataplane.
>+     * Because of that, we don't enable virtqueues here and leave it to
>+     * net/vhost-vdpa.c.
>+     */
>+    if (nc->info->type == NET_CLIENT_DRIVER_VHOST_VDPA) {
>+        return 0;
>+    }
>+
>     nc->vring_enable = enable;
>
>     if (vhost_ops && vhost_ops->vhost_set_vring_enable) {
>diff --git a/hw/virtio/vdpa-dev.c b/hw/virtio/vdpa-dev.c
>index eb9ecea83b..13e87f06f6 100644
>--- a/hw/virtio/vdpa-dev.c
>+++ b/hw/virtio/vdpa-dev.c
>@@ -253,14 +253,11 @@ static int vhost_vdpa_device_start(VirtIODevice *vdev, Error **errp)
>
>     s->dev.acked_features = vdev->guest_features;
>
>-    ret = vhost_dev_start(&s->dev, vdev, false);
>+    ret = vhost_dev_start(&s->dev, vdev, true);
>     if (ret < 0) {
>         error_setg_errno(errp, -ret, "Error starting vhost");
>         goto err_guest_notifiers;
>     }
>-    for (i = 0; i < s->dev.nvqs; ++i) {
>-        vhost_vdpa_set_vring_ready(&s->vdpa, i);
>-    }
>     s->started = true;
>
>     /*
>diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c
>index ddae494ca8..401afac2f5 100644
>--- a/hw/virtio/vhost-vdpa.c
>+++ b/hw/virtio/vhost-vdpa.c
>@@ -886,12 +886,13 @@ static int vhost_vdpa_get_vq_index(struct vhost_dev *dev, int idx)
>     return idx;
> }
>
>-int vhost_vdpa_set_vring_ready(struct vhost_vdpa *v, unsigned idx)
>+static int vhost_vdpa_set_vring_enable_one(struct vhost_vdpa *v, unsigned idx,
>+                                           int enable)
> {
>     struct vhost_dev *dev = v->dev;
>     struct vhost_vring_state state = {
>         .index = idx,
>-        .num = 1,
>+        .num = enable,
>     };
>     int r = vhost_vdpa_call(dev, VHOST_VDPA_SET_VRING_ENABLE, &state);
>

After this line we now have:

       trace_vhost_vdpa_set_vring_ready(dev, idx, r);

Should we rename it or move it to the new function?

If we rename it, we should trace also `enable`.

>@@ -899,6 +900,27 @@ int vhost_vdpa_set_vring_ready(struct vhost_vdpa 
>*v, unsigned idx)
>     return r;
> }
>
>+static int vhost_vdpa_set_vring_enable(struct vhost_dev *dev, int enable)
>+{
>+    struct vhost_vdpa *v = dev->opaque;
>+    unsigned int i;
>+    int ret;
>+
>+    for (i = 0; i < dev->nvqs; ++i) {
>+        ret = vhost_vdpa_set_vring_enable_one(v, i, enable);
>+        if (ret < 0) {
>+            return ret;
>+        }
>+    }
>+
>+    return 0;
>+}
>+
>+int vhost_vdpa_set_vring_ready(struct vhost_vdpa *v, unsigned idx)
>+{
>+    return vhost_vdpa_set_vring_enable_one(v, idx, 1);
>+}
>+
> static int vhost_vdpa_set_config_call(struct vhost_dev *dev,
>                                        int fd)
> {
>@@ -1514,6 +1536,7 @@ const VhostOps vdpa_ops = {
>         .vhost_set_features = vhost_vdpa_set_features,
>         .vhost_reset_device = vhost_vdpa_reset_device,
>         .vhost_get_vq_index = vhost_vdpa_get_vq_index,
>+        .vhost_set_vring_enable = vhost_vdpa_set_vring_enable,
>         .vhost_get_config  = vhost_vdpa_get_config,
>         .vhost_set_config = vhost_vdpa_set_config,
>         .vhost_requires_shm_log = NULL,
>diff --git a/hw/virtio/vhost.c b/hw/virtio/vhost.c
>index 2c9ac79468..decfb85184 100644
>--- a/hw/virtio/vhost.c
>+++ b/hw/virtio/vhost.c
>@@ -1984,7 +1984,13 @@ static int vhost_dev_set_vring_enable(struct vhost_dev *hdev, int enable)
>     return hdev->vhost_ops->vhost_set_vring_enable(hdev, enable);
> }
>
>-/* Host notifiers must be enabled at this point. */
>+/*
>+ * Host notifiers must be enabled at this point.
>+ *
>+ * If @vring is true, this function will enable all vrings before starting the

Nit: @vrings

>+ * device. If it is false, the vring initialization is left to be done by the
>+ * caller.
>+ */
> int vhost_dev_start(struct vhost_dev *hdev, VirtIODevice *vdev, bool vrings)
> {
>     int i, r;
>-- 
>2.44.0
>

The rest LGTM, but while reviewing I was thinking whether it would be
better to split the patch into two: first patch to support
.vhost_set_vring_enable() for vhost-vdpa, second patch to fix
hw/virtio/vdpa-dev.c.

But I'm not sure this is a good idea since the second is strictly
dependent on the first.

Thanks,
Stefano
Kevin Wolf March 15, 2024, 3:46 p.m. UTC | #2
Am 15.03.2024 um 16:07 hat Stefano Garzarella geschrieben:
> On Fri, Mar 15, 2024 at 03:03:31PM +0100, Kevin Wolf wrote:
> > VDUSE requires that virtqueues are first enabled before the DRIVER_OK
> > status flag is set; with the current API of the kernel module, it is
> > impossible to enable the opposite order in our block export code because
> > userspace is not notified when a virtqueue is enabled.
> > 
> > This requirement also mathces the normal initialisation order as done by
> > the generic vhost code in QEMU. However, commit 6c482547 accidentally
> > changed the order for vdpa-dev and broke access to VDUSE devices with
> > this.
> > 
> > This changes vdpa-dev to use the normal order again and use the standard
> > vhost callback .vhost_set_vring_enable for this. VDUSE devices can be
> > used with vdpa-dev again after this fix.
> > 
> > vhost_net intentionally avoided enabling the vrings for vdpa and does
> > this manually later while it does enable them for other vhost backends.
> > Reflect this in the vhost_net code and return early for vdpa, so that
> > the behaviour doesn't change for this device.
> > 
> > Cc: qemu-stable@nongnu.org
> > Fixes: 6c4825476a4351530bcac17abab72295b75ffe98
> > Signed-off-by: Kevin Wolf <kwolf@redhat.com>
> > ---
> > v2:
> > - Actually make use of the @enable parameter
> > - Change vhost_net to preserve the current behaviour
> > 
> > hw/net/vhost_net.c     | 10 ++++++++++
> > hw/virtio/vdpa-dev.c   |  5 +----
> > hw/virtio/vhost-vdpa.c | 27 +++++++++++++++++++++++++--
> > hw/virtio/vhost.c      |  8 +++++++-
> > 4 files changed, 43 insertions(+), 7 deletions(-)
> > 
> > diff --git a/hw/net/vhost_net.c b/hw/net/vhost_net.c
> > index e8e1661646..fd1a93701a 100644
> > --- a/hw/net/vhost_net.c
> > +++ b/hw/net/vhost_net.c
> > @@ -541,6 +541,16 @@ int vhost_set_vring_enable(NetClientState *nc, int enable)
> >     VHostNetState *net = get_vhost_net(nc);
> >     const VhostOps *vhost_ops = net->dev.vhost_ops;
> > 
> > +    /*
> > +     * vhost-vdpa network devices need to enable dataplane virtqueues after
> > +     * DRIVER_OK, so they can recover device state before starting dataplane.
> > +     * Because of that, we don't enable virtqueues here and leave it to
> > +     * net/vhost-vdpa.c.
> > +     */
> > +    if (nc->info->type == NET_CLIENT_DRIVER_VHOST_VDPA) {
> > +        return 0;
> > +    }
> > +
> >     nc->vring_enable = enable;
> > 
> >     if (vhost_ops && vhost_ops->vhost_set_vring_enable) {
> > diff --git a/hw/virtio/vdpa-dev.c b/hw/virtio/vdpa-dev.c
> > index eb9ecea83b..13e87f06f6 100644
> > --- a/hw/virtio/vdpa-dev.c
> > +++ b/hw/virtio/vdpa-dev.c
> > @@ -253,14 +253,11 @@ static int vhost_vdpa_device_start(VirtIODevice *vdev, Error **errp)
> > 
> >     s->dev.acked_features = vdev->guest_features;
> > 
> > -    ret = vhost_dev_start(&s->dev, vdev, false);
> > +    ret = vhost_dev_start(&s->dev, vdev, true);
> >     if (ret < 0) {
> >         error_setg_errno(errp, -ret, "Error starting vhost");
> >         goto err_guest_notifiers;
> >     }
> > -    for (i = 0; i < s->dev.nvqs; ++i) {
> > -        vhost_vdpa_set_vring_ready(&s->vdpa, i);
> > -    }
> >     s->started = true;
> > 
> >     /*
> > diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c
> > index ddae494ca8..401afac2f5 100644
> > --- a/hw/virtio/vhost-vdpa.c
> > +++ b/hw/virtio/vhost-vdpa.c
> > @@ -886,12 +886,13 @@ static int vhost_vdpa_get_vq_index(struct vhost_dev *dev, int idx)
> >     return idx;
> > }
> > 
> > -int vhost_vdpa_set_vring_ready(struct vhost_vdpa *v, unsigned idx)
> > +static int vhost_vdpa_set_vring_enable_one(struct vhost_vdpa *v, unsigned idx,
> > +                                           int enable)
> > {
> >     struct vhost_dev *dev = v->dev;
> >     struct vhost_vring_state state = {
> >         .index = idx,
> > -        .num = 1,
> > +        .num = enable,
> >     };
> >     int r = vhost_vdpa_call(dev, VHOST_VDPA_SET_VRING_ENABLE, &state);
> > 
> 
> After this line we now have:
> 
>       trace_vhost_vdpa_set_vring_ready(dev, idx, r);
> 
> Should we rename it or move it to the new function?
> 
> If we rename it, we should trace also `enable`.

I think renaming is better so that we cover all code paths that enable a
vring. I'll change this and send a v3.

> > @@ -899,6 +900,27 @@ int vhost_vdpa_set_vring_ready(struct vhost_vdpa
> > *v, unsigned idx)
> >     return r;
> > }
> > 
> > +static int vhost_vdpa_set_vring_enable(struct vhost_dev *dev, int enable)
> > +{
> > +    struct vhost_vdpa *v = dev->opaque;
> > +    unsigned int i;
> > +    int ret;
> > +
> > +    for (i = 0; i < dev->nvqs; ++i) {
> > +        ret = vhost_vdpa_set_vring_enable_one(v, i, enable);
> > +        if (ret < 0) {
> > +            return ret;
> > +        }
> > +    }
> > +
> > +    return 0;
> > +}
> > +
> > +int vhost_vdpa_set_vring_ready(struct vhost_vdpa *v, unsigned idx)
> > +{
> > +    return vhost_vdpa_set_vring_enable_one(v, idx, 1);
> > +}
> > +
> > static int vhost_vdpa_set_config_call(struct vhost_dev *dev,
> >                                        int fd)
> > {
> > @@ -1514,6 +1536,7 @@ const VhostOps vdpa_ops = {
> >         .vhost_set_features = vhost_vdpa_set_features,
> >         .vhost_reset_device = vhost_vdpa_reset_device,
> >         .vhost_get_vq_index = vhost_vdpa_get_vq_index,
> > +        .vhost_set_vring_enable = vhost_vdpa_set_vring_enable,
> >         .vhost_get_config  = vhost_vdpa_get_config,
> >         .vhost_set_config = vhost_vdpa_set_config,
> >         .vhost_requires_shm_log = NULL,
> > diff --git a/hw/virtio/vhost.c b/hw/virtio/vhost.c
> > index 2c9ac79468..decfb85184 100644
> > --- a/hw/virtio/vhost.c
> > +++ b/hw/virtio/vhost.c
> > @@ -1984,7 +1984,13 @@ static int vhost_dev_set_vring_enable(struct vhost_dev *hdev, int enable)
> >     return hdev->vhost_ops->vhost_set_vring_enable(hdev, enable);
> > }
> > 
> > -/* Host notifiers must be enabled at this point. */
> > +/*
> > + * Host notifiers must be enabled at this point.
> > + *
> > + * If @vring is true, this function will enable all vrings before starting the
> 
> Nit: @vrings

Thanks, will fix.

> > + * device. If it is false, the vring initialization is left to be done by the
> > + * caller.
> > + */
> > int vhost_dev_start(struct vhost_dev *hdev, VirtIODevice *vdev, bool vrings)
> > {
> >     int i, r;
> > -- 
> > 2.44.0
> > 
> 
> The rest LGTM, but while reviewing I was thinking whether it would be
> better to split the patch into two: first patch to support
> .vhost_set_vring_enable() for vhost-vdpa, second patch to fix
> hw/virtio/vdpa-dev.c.
> 
> But I'm not sure this is a good idea since the second is strictly
> dependent on the first.

And the second part to be split off is really small, three lines removed
and one changed. I don't think it's worth it.

Kevin
diff mbox series

Patch

diff --git a/hw/net/vhost_net.c b/hw/net/vhost_net.c
index e8e1661646..fd1a93701a 100644
--- a/hw/net/vhost_net.c
+++ b/hw/net/vhost_net.c
@@ -541,6 +541,16 @@  int vhost_set_vring_enable(NetClientState *nc, int enable)
     VHostNetState *net = get_vhost_net(nc);
     const VhostOps *vhost_ops = net->dev.vhost_ops;
 
+    /*
+     * vhost-vdpa network devices need to enable dataplane virtqueues after
+     * DRIVER_OK, so they can recover device state before starting dataplane.
+     * Because of that, we don't enable virtqueues here and leave it to
+     * net/vhost-vdpa.c.
+     */
+    if (nc->info->type == NET_CLIENT_DRIVER_VHOST_VDPA) {
+        return 0;
+    }
+
     nc->vring_enable = enable;
 
     if (vhost_ops && vhost_ops->vhost_set_vring_enable) {
diff --git a/hw/virtio/vdpa-dev.c b/hw/virtio/vdpa-dev.c
index eb9ecea83b..13e87f06f6 100644
--- a/hw/virtio/vdpa-dev.c
+++ b/hw/virtio/vdpa-dev.c
@@ -253,14 +253,11 @@  static int vhost_vdpa_device_start(VirtIODevice *vdev, Error **errp)
 
     s->dev.acked_features = vdev->guest_features;
 
-    ret = vhost_dev_start(&s->dev, vdev, false);
+    ret = vhost_dev_start(&s->dev, vdev, true);
     if (ret < 0) {
         error_setg_errno(errp, -ret, "Error starting vhost");
         goto err_guest_notifiers;
     }
-    for (i = 0; i < s->dev.nvqs; ++i) {
-        vhost_vdpa_set_vring_ready(&s->vdpa, i);
-    }
     s->started = true;
 
     /*
diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c
index ddae494ca8..401afac2f5 100644
--- a/hw/virtio/vhost-vdpa.c
+++ b/hw/virtio/vhost-vdpa.c
@@ -886,12 +886,13 @@  static int vhost_vdpa_get_vq_index(struct vhost_dev *dev, int idx)
     return idx;
 }
 
-int vhost_vdpa_set_vring_ready(struct vhost_vdpa *v, unsigned idx)
+static int vhost_vdpa_set_vring_enable_one(struct vhost_vdpa *v, unsigned idx,
+                                           int enable)
 {
     struct vhost_dev *dev = v->dev;
     struct vhost_vring_state state = {
         .index = idx,
-        .num = 1,
+        .num = enable,
     };
     int r = vhost_vdpa_call(dev, VHOST_VDPA_SET_VRING_ENABLE, &state);
 
@@ -899,6 +900,27 @@  int vhost_vdpa_set_vring_ready(struct vhost_vdpa *v, unsigned idx)
     return r;
 }
 
+static int vhost_vdpa_set_vring_enable(struct vhost_dev *dev, int enable)
+{
+    struct vhost_vdpa *v = dev->opaque;
+    unsigned int i;
+    int ret;
+
+    for (i = 0; i < dev->nvqs; ++i) {
+        ret = vhost_vdpa_set_vring_enable_one(v, i, enable);
+        if (ret < 0) {
+            return ret;
+        }
+    }
+
+    return 0;
+}
+
+int vhost_vdpa_set_vring_ready(struct vhost_vdpa *v, unsigned idx)
+{
+    return vhost_vdpa_set_vring_enable_one(v, idx, 1);
+}
+
 static int vhost_vdpa_set_config_call(struct vhost_dev *dev,
                                        int fd)
 {
@@ -1514,6 +1536,7 @@  const VhostOps vdpa_ops = {
         .vhost_set_features = vhost_vdpa_set_features,
         .vhost_reset_device = vhost_vdpa_reset_device,
         .vhost_get_vq_index = vhost_vdpa_get_vq_index,
+        .vhost_set_vring_enable = vhost_vdpa_set_vring_enable,
         .vhost_get_config  = vhost_vdpa_get_config,
         .vhost_set_config = vhost_vdpa_set_config,
         .vhost_requires_shm_log = NULL,
diff --git a/hw/virtio/vhost.c b/hw/virtio/vhost.c
index 2c9ac79468..decfb85184 100644
--- a/hw/virtio/vhost.c
+++ b/hw/virtio/vhost.c
@@ -1984,7 +1984,13 @@  static int vhost_dev_set_vring_enable(struct vhost_dev *hdev, int enable)
     return hdev->vhost_ops->vhost_set_vring_enable(hdev, enable);
 }
 
-/* Host notifiers must be enabled at this point. */
+/*
+ * Host notifiers must be enabled at this point.
+ *
+ * If @vring is true, this function will enable all vrings before starting the
+ * device. If it is false, the vring initialization is left to be done by the
+ * caller.
+ */
 int vhost_dev_start(struct vhost_dev *hdev, VirtIODevice *vdev, bool vrings)
 {
     int i, r;