* [PATCH net] vsock/virtio: enable VQs early on probe
@ 2022-03-22 10:38 Stefano Garzarella
2022-03-22 13:36 ` Michael S. Tsirkin
0 siblings, 1 reply; 5+ messages in thread
From: Stefano Garzarella @ 2022-03-22 10:38 UTC (permalink / raw)
To: netdev
Cc: virtualization, Stefano Garzarella, Asias He, David S. Miller,
Jakub Kicinski, Stefan Hajnoczi, kvm, linux-kernel, Paolo Abeni,
Michael S. Tsirkin
virtio spec requires drivers to set DRIVER_OK before using VQs.
This is set automatically after probe returns, but virtio-vsock
driver uses VQs in the probe function to fill rx and event VQs
with new buffers.
Let's fix this, calling virtio_device_ready() before using VQs
in the probe function.
Fixes: 0ea9e1d3a9e3 ("VSOCK: Introduce virtio_transport.ko")
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
---
net/vmw_vsock/virtio_transport.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c
index 5afc194a58bb..b1962f8cd502 100644
--- a/net/vmw_vsock/virtio_transport.c
+++ b/net/vmw_vsock/virtio_transport.c
@@ -622,6 +622,8 @@ static int virtio_vsock_probe(struct virtio_device *vdev)
INIT_WORK(&vsock->event_work, virtio_transport_event_work);
INIT_WORK(&vsock->send_pkt_work, virtio_transport_send_pkt_work);
+ virtio_device_ready(vdev);
+
mutex_lock(&vsock->tx_lock);
vsock->tx_run = true;
mutex_unlock(&vsock->tx_lock);
--
2.35.1
^ permalink raw reply related [flat|nested] 5+ messages in thread* Re: [PATCH net] vsock/virtio: enable VQs early on probe 2022-03-22 10:38 [PATCH net] vsock/virtio: enable VQs early on probe Stefano Garzarella @ 2022-03-22 13:36 ` Michael S. Tsirkin 2022-03-22 14:05 ` Stefano Garzarella 0 siblings, 1 reply; 5+ messages in thread From: Michael S. Tsirkin @ 2022-03-22 13:36 UTC (permalink / raw) To: Stefano Garzarella Cc: netdev, virtualization, Asias He, David S. Miller, Jakub Kicinski, Stefan Hajnoczi, kvm, linux-kernel, Paolo Abeni On Tue, Mar 22, 2022 at 11:38:23AM +0100, Stefano Garzarella wrote: > virtio spec requires drivers to set DRIVER_OK before using VQs. > This is set automatically after probe returns, but virtio-vsock > driver uses VQs in the probe function to fill rx and event VQs > with new buffers. So this is a spec violation. absolutely. > Let's fix this, calling virtio_device_ready() before using VQs > in the probe function. > > Fixes: 0ea9e1d3a9e3 ("VSOCK: Introduce virtio_transport.ko") > Signed-off-by: Stefano Garzarella <sgarzare@redhat.com> > --- > net/vmw_vsock/virtio_transport.c | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c > index 5afc194a58bb..b1962f8cd502 100644 > --- a/net/vmw_vsock/virtio_transport.c > +++ b/net/vmw_vsock/virtio_transport.c > @@ -622,6 +622,8 @@ static int virtio_vsock_probe(struct virtio_device *vdev) > INIT_WORK(&vsock->event_work, virtio_transport_event_work); > INIT_WORK(&vsock->send_pkt_work, virtio_transport_send_pkt_work); > > + virtio_device_ready(vdev); > + > mutex_lock(&vsock->tx_lock); > vsock->tx_run = true; > mutex_unlock(&vsock->tx_lock); Here's the whole code snippet: mutex_lock(&vsock->tx_lock); vsock->tx_run = true; mutex_unlock(&vsock->tx_lock); mutex_lock(&vsock->rx_lock); virtio_vsock_rx_fill(vsock); vsock->rx_run = true; mutex_unlock(&vsock->rx_lock); mutex_lock(&vsock->event_lock); virtio_vsock_event_fill(vsock); vsock->event_run = true; mutex_unlock(&vsock->event_lock); if (virtio_has_feature(vdev, VIRTIO_VSOCK_F_SEQPACKET)) vsock->seqpacket_allow = true; vdev->priv = vsock; rcu_assign_pointer(the_virtio_vsock, vsock); mutex_unlock(&the_virtio_vsock_mutex); I worry that this is not the only problem here: seqpacket_allow and setting of vdev->priv at least after device is active look suspicious. E.g.: static void virtio_vsock_event_done(struct virtqueue *vq) { struct virtio_vsock *vsock = vq->vdev->priv; if (!vsock) return; queue_work(virtio_vsock_workqueue, &vsock->event_work); } looks like it will miss events now they will be reported earlier. One might say that since vq has been kicked it might send interrupts earlier too so not a new problem, but there's a chance device actually waits until DRIVER_OK to start operating. > -- > 2.35.1 ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH net] vsock/virtio: enable VQs early on probe 2022-03-22 13:36 ` Michael S. Tsirkin @ 2022-03-22 14:05 ` Stefano Garzarella 2022-03-22 14:09 ` Michael S. Tsirkin 0 siblings, 1 reply; 5+ messages in thread From: Stefano Garzarella @ 2022-03-22 14:05 UTC (permalink / raw) To: Michael S. Tsirkin Cc: netdev, virtualization, David S. Miller, Jakub Kicinski, Stefan Hajnoczi, kvm, linux-kernel, Paolo Abeni On Tue, Mar 22, 2022 at 09:36:14AM -0400, Michael S. Tsirkin wrote: >On Tue, Mar 22, 2022 at 11:38:23AM +0100, Stefano Garzarella wrote: >> virtio spec requires drivers to set DRIVER_OK before using VQs. >> This is set automatically after probe returns, but virtio-vsock >> driver uses VQs in the probe function to fill rx and event VQs >> with new buffers. > > >So this is a spec violation. absolutely. > >> Let's fix this, calling virtio_device_ready() before using VQs >> in the probe function. >> >> Fixes: 0ea9e1d3a9e3 ("VSOCK: Introduce virtio_transport.ko") >> Signed-off-by: Stefano Garzarella <sgarzare@redhat.com> >> --- >> net/vmw_vsock/virtio_transport.c | 2 ++ >> 1 file changed, 2 insertions(+) >> >> diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c >> index 5afc194a58bb..b1962f8cd502 100644 >> --- a/net/vmw_vsock/virtio_transport.c >> +++ b/net/vmw_vsock/virtio_transport.c >> @@ -622,6 +622,8 @@ static int virtio_vsock_probe(struct virtio_device *vdev) >> INIT_WORK(&vsock->event_work, virtio_transport_event_work); >> INIT_WORK(&vsock->send_pkt_work, virtio_transport_send_pkt_work); >> >> + virtio_device_ready(vdev); >> + >> mutex_lock(&vsock->tx_lock); >> vsock->tx_run = true; >> mutex_unlock(&vsock->tx_lock); > >Here's the whole code snippet: > > > mutex_lock(&vsock->tx_lock); > vsock->tx_run = true; > mutex_unlock(&vsock->tx_lock); > > mutex_lock(&vsock->rx_lock); > virtio_vsock_rx_fill(vsock); > vsock->rx_run = true; > mutex_unlock(&vsock->rx_lock); > > mutex_lock(&vsock->event_lock); > virtio_vsock_event_fill(vsock); > vsock->event_run = true; > mutex_unlock(&vsock->event_lock); > > if (virtio_has_feature(vdev, VIRTIO_VSOCK_F_SEQPACKET)) > vsock->seqpacket_allow = true; > > vdev->priv = vsock; > rcu_assign_pointer(the_virtio_vsock, vsock); > > mutex_unlock(&the_virtio_vsock_mutex); > > >I worry that this is not the only problem here: >seqpacket_allow and setting of vdev->priv at least after >device is active look suspicious. Right, so if you agree I'll move these before virtio_device_ready(). >E.g.: > >static void virtio_vsock_event_done(struct virtqueue *vq) >{ > struct virtio_vsock *vsock = vq->vdev->priv; > > if (!vsock) > return; > queue_work(virtio_vsock_workqueue, &vsock->event_work); >} > >looks like it will miss events now they will be reported earlier. >One might say that since vq has been kicked it might send >interrupts earlier too so not a new problem, but >there's a chance device actually waits until DRIVER_OK >to start operating. Yes I see, should I break into 2 patches (one where I move the code already present and this one)? Maybe a single patch is fine since it's the complete solution. Thank you for the detailed explanation, Stefano ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH net] vsock/virtio: enable VQs early on probe 2022-03-22 14:05 ` Stefano Garzarella @ 2022-03-22 14:09 ` Michael S. Tsirkin 2022-03-22 14:36 ` Stefano Garzarella 0 siblings, 1 reply; 5+ messages in thread From: Michael S. Tsirkin @ 2022-03-22 14:09 UTC (permalink / raw) To: Stefano Garzarella Cc: netdev, virtualization, David S. Miller, Jakub Kicinski, Stefan Hajnoczi, kvm, linux-kernel, Paolo Abeni On Tue, Mar 22, 2022 at 03:05:00PM +0100, Stefano Garzarella wrote: > On Tue, Mar 22, 2022 at 09:36:14AM -0400, Michael S. Tsirkin wrote: > > On Tue, Mar 22, 2022 at 11:38:23AM +0100, Stefano Garzarella wrote: > > > virtio spec requires drivers to set DRIVER_OK before using VQs. > > > This is set automatically after probe returns, but virtio-vsock > > > driver uses VQs in the probe function to fill rx and event VQs > > > with new buffers. > > > > > > So this is a spec violation. absolutely. > > > > > Let's fix this, calling virtio_device_ready() before using VQs > > > in the probe function. > > > > > > Fixes: 0ea9e1d3a9e3 ("VSOCK: Introduce virtio_transport.ko") > > > Signed-off-by: Stefano Garzarella <sgarzare@redhat.com> > > > --- > > > net/vmw_vsock/virtio_transport.c | 2 ++ > > > 1 file changed, 2 insertions(+) > > > > > > diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c > > > index 5afc194a58bb..b1962f8cd502 100644 > > > --- a/net/vmw_vsock/virtio_transport.c > > > +++ b/net/vmw_vsock/virtio_transport.c > > > @@ -622,6 +622,8 @@ static int virtio_vsock_probe(struct virtio_device *vdev) > > > INIT_WORK(&vsock->event_work, virtio_transport_event_work); > > > INIT_WORK(&vsock->send_pkt_work, virtio_transport_send_pkt_work); > > > > > > + virtio_device_ready(vdev); > > > + > > > mutex_lock(&vsock->tx_lock); > > > vsock->tx_run = true; > > > mutex_unlock(&vsock->tx_lock); > > > > Here's the whole code snippet: > > > > > > mutex_lock(&vsock->tx_lock); > > vsock->tx_run = true; > > mutex_unlock(&vsock->tx_lock); > > > > mutex_lock(&vsock->rx_lock); > > virtio_vsock_rx_fill(vsock); > > vsock->rx_run = true; > > mutex_unlock(&vsock->rx_lock); > > > > mutex_lock(&vsock->event_lock); > > virtio_vsock_event_fill(vsock); > > vsock->event_run = true; > > mutex_unlock(&vsock->event_lock); > > > > if (virtio_has_feature(vdev, VIRTIO_VSOCK_F_SEQPACKET)) > > vsock->seqpacket_allow = true; > > > > vdev->priv = vsock; > > rcu_assign_pointer(the_virtio_vsock, vsock); > > > > mutex_unlock(&the_virtio_vsock_mutex); > > > > > > I worry that this is not the only problem here: > > seqpacket_allow and setting of vdev->priv at least after > > device is active look suspicious. > > Right, so if you agree I'll move these before virtio_device_ready(). > > > E.g.: > > > > static void virtio_vsock_event_done(struct virtqueue *vq) > > { > > struct virtio_vsock *vsock = vq->vdev->priv; > > > > if (!vsock) > > return; > > queue_work(virtio_vsock_workqueue, &vsock->event_work); > > } > > > > looks like it will miss events now they will be reported earlier. > > One might say that since vq has been kicked it might send > > interrupts earlier too so not a new problem, but > > there's a chance device actually waits until DRIVER_OK > > to start operating. > > Yes I see, should I break into 2 patches (one where I move the code already > present and this one)? > > Maybe a single patch is fine since it's the complete solution. > > Thank you for the detailed explanation, > Stefano Two I think since movement can be backported to before the hardening effort. -- MST ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH net] vsock/virtio: enable VQs early on probe 2022-03-22 14:09 ` Michael S. Tsirkin @ 2022-03-22 14:36 ` Stefano Garzarella 0 siblings, 0 replies; 5+ messages in thread From: Stefano Garzarella @ 2022-03-22 14:36 UTC (permalink / raw) To: Michael S. Tsirkin Cc: netdev, virtualization, David S. Miller, Jakub Kicinski, Stefan Hajnoczi, kvm, linux-kernel, Paolo Abeni On Tue, Mar 22, 2022 at 10:09:06AM -0400, Michael S. Tsirkin wrote: >On Tue, Mar 22, 2022 at 03:05:00PM +0100, Stefano Garzarella wrote: >> On Tue, Mar 22, 2022 at 09:36:14AM -0400, Michael S. Tsirkin wrote: >> > On Tue, Mar 22, 2022 at 11:38:23AM +0100, Stefano Garzarella wrote: >> > > virtio spec requires drivers to set DRIVER_OK before using VQs. >> > > This is set automatically after probe returns, but virtio-vsock >> > > driver uses VQs in the probe function to fill rx and event VQs >> > > with new buffers. >> > >> > >> > So this is a spec violation. absolutely. >> > >> > > Let's fix this, calling virtio_device_ready() before using VQs >> > > in the probe function. >> > > >> > > Fixes: 0ea9e1d3a9e3 ("VSOCK: Introduce virtio_transport.ko") >> > > Signed-off-by: Stefano Garzarella <sgarzare@redhat.com> >> > > --- >> > > net/vmw_vsock/virtio_transport.c | 2 ++ >> > > 1 file changed, 2 insertions(+) >> > > >> > > diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c >> > > index 5afc194a58bb..b1962f8cd502 100644 >> > > --- a/net/vmw_vsock/virtio_transport.c >> > > +++ b/net/vmw_vsock/virtio_transport.c >> > > @@ -622,6 +622,8 @@ static int virtio_vsock_probe(struct virtio_device *vdev) >> > > INIT_WORK(&vsock->event_work, virtio_transport_event_work); >> > > INIT_WORK(&vsock->send_pkt_work, virtio_transport_send_pkt_work); >> > > >> > > + virtio_device_ready(vdev); >> > > + >> > > mutex_lock(&vsock->tx_lock); >> > > vsock->tx_run = true; >> > > mutex_unlock(&vsock->tx_lock); >> > >> > Here's the whole code snippet: >> > >> > >> > mutex_lock(&vsock->tx_lock); >> > vsock->tx_run = true; >> > mutex_unlock(&vsock->tx_lock); >> > >> > mutex_lock(&vsock->rx_lock); >> > virtio_vsock_rx_fill(vsock); >> > vsock->rx_run = true; >> > mutex_unlock(&vsock->rx_lock); >> > >> > mutex_lock(&vsock->event_lock); >> > virtio_vsock_event_fill(vsock); >> > vsock->event_run = true; >> > mutex_unlock(&vsock->event_lock); >> > >> > if (virtio_has_feature(vdev, VIRTIO_VSOCK_F_SEQPACKET)) >> > vsock->seqpacket_allow = true; >> > >> > vdev->priv = vsock; >> > rcu_assign_pointer(the_virtio_vsock, vsock); >> > >> > mutex_unlock(&the_virtio_vsock_mutex); >> > >> > >> > I worry that this is not the only problem here: >> > seqpacket_allow and setting of vdev->priv at least after >> > device is active look suspicious. >> >> Right, so if you agree I'll move these before virtio_device_ready(). >> >> > E.g.: >> > >> > static void virtio_vsock_event_done(struct virtqueue *vq) >> > { >> > struct virtio_vsock *vsock = vq->vdev->priv; >> > >> > if (!vsock) >> > return; >> > queue_work(virtio_vsock_workqueue, &vsock->event_work); >> > } >> > >> > looks like it will miss events now they will be reported earlier. >> > One might say that since vq has been kicked it might send >> > interrupts earlier too so not a new problem, but >> > there's a chance device actually waits until DRIVER_OK >> > to start operating. >> >> Yes I see, should I break into 2 patches (one where I move the code already >> present and this one)? >> >> Maybe a single patch is fine since it's the complete solution. >> >> Thank you for the detailed explanation, >> Stefano > >Two I think since movement can be backported to before the hardening >effort. Yep, maybe 3 since seqpacket was added later. Thanks, Stefano ^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2022-03-22 14:36 UTC | newest] Thread overview: 5+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2022-03-22 10:38 [PATCH net] vsock/virtio: enable VQs early on probe Stefano Garzarella 2022-03-22 13:36 ` Michael S. Tsirkin 2022-03-22 14:05 ` Stefano Garzarella 2022-03-22 14:09 ` Michael S. Tsirkin 2022-03-22 14:36 ` Stefano Garzarella
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).