Re: [PATCH 6/9] virtio_pci: harden MSI-X interrupts

From: Michael S. Tsirkin
Date: Mon Sep 13 2021 - 02:50:24 EST


On Mon, Sep 13, 2021 at 02:37:42AM -0400, Michael S. Tsirkin wrote:
> On Mon, Sep 13, 2021 at 02:34:01PM +0800, Jason Wang wrote:
> > On Mon, Sep 13, 2021 at 2:28 PM Michael S. Tsirkin <mst@xxxxxxxxxx> wrote:
> > >
> > > On Mon, Sep 13, 2021 at 02:08:02PM +0800, Jason Wang wrote:
> > > > On Mon, Sep 13, 2021 at 2:04 PM Michael S. Tsirkin <mst@xxxxxxxxxx> wrote:
> > > > >
> > > > > On Mon, Sep 13, 2021 at 01:53:50PM +0800, Jason Wang wrote:
> > > > > > We used to synchronize pending MSI-X irq handlers via
> > > > > > synchronize_irq(), this may not work for the untrusted device which
> > > > > > may keep sending interrupts after reset which may lead unexpected
> > > > > > results. Similarly, we should not enable MSI-X interrupt until the
> > > > > > device is ready. So this patch fixes those two issues by:
> > > > > >
> > > > > > 1) switching to use disable_irq() to prevent the virtio interrupt
> > > > > > handlers to be called after the device is reset.
> > > > > > 2) using IRQF_NO_AUTOEN and enable the MSI-X irq during .ready()
> > > > > >
> > > > > > This can make sure the virtio interrupt handler won't be called before
> > > > > > virtio_device_ready() and after reset.
> > > > > >
> > > > > > Signed-off-by: Jason Wang <jasowang@xxxxxxxxxx>
> > > > >
> > > > > I don't get the threat model here. Isn't disabling irqs done by the
> > > > > hypervisor anyway? Is there a reason to trust disable_irq but not
> > > > > device reset?
> > > >
> > > > My understanding is that e.g in the case of SEV/TDX we don't trust the
> > > > hypervisor. So the hypervisor can keep sending interrupts even if the
> > > > device is reset. The guest can only trust its own software interrupt
> > > > management logic to avoid call virtio callback in this case.
> > > >
> > > > Thanks
> > >
> > > Hmm but I don't see how do these patches do this.
> > > They call disable_irq but can't the hypervisor keep
> > > sending interrupts after disable_irq, too?
> >
> > Yes, but since the irq is disabled, the vring or config callback won't
> > be called in this case.
> >
> > Thanks
>
> But doen't "irq is disabled" basically mean "we told the hypervisor
> to disable the irq"? What extractly prevents hypervisor from
> sending the irq even if guest thinks it disabled it?

More generally, can't we for example blow away the
indir_desc array that we use to keep the ctx pointers?
Won't that be enough?


And looking at all this closely, I suspect it is wise to just completely
disable hotplug/unplug for encrypted guests. Or maybe it's already
the case?


> > >
> > >
> > >
> > > > >
> > > > > Cc a bunch more people ...
> > > > >
> > > > >
> > > > > > ---
> > > > > > drivers/virtio/virtio_pci_common.c | 27 +++++++++++++++++++++------
> > > > > > drivers/virtio/virtio_pci_common.h | 6 ++++--
> > > > > > drivers/virtio/virtio_pci_legacy.c | 5 +++--
> > > > > > drivers/virtio/virtio_pci_modern.c | 6 ++++--
> > > > > > 4 files changed, 32 insertions(+), 12 deletions(-)
> > > > > >
> > > > > > diff --git a/drivers/virtio/virtio_pci_common.c b/drivers/virtio/virtio_pci_common.c
> > > > > > index b35bb2d57f62..0b9523e6dd39 100644
> > > > > > --- a/drivers/virtio/virtio_pci_common.c
> > > > > > +++ b/drivers/virtio/virtio_pci_common.c
> > > > > > @@ -24,8 +24,8 @@ MODULE_PARM_DESC(force_legacy,
> > > > > > "Force legacy mode for transitional virtio 1 devices");
> > > > > > #endif
> > > > > >
> > > > > > -/* wait for pending irq handlers */
> > > > > > -void vp_synchronize_vectors(struct virtio_device *vdev)
> > > > > > +/* disable irq handlers */
> > > > > > +void vp_disable_vectors(struct virtio_device *vdev)
> > > > > > {
> > > > > > struct virtio_pci_device *vp_dev = to_vp_device(vdev);
> > > > > > int i;
> > > > > > @@ -34,7 +34,20 @@ void vp_synchronize_vectors(struct virtio_device *vdev)
> > > > > > synchronize_irq(vp_dev->pci_dev->irq);
> > > > > >
> > > > > > for (i = 0; i < vp_dev->msix_vectors; ++i)
> > > > > > - synchronize_irq(pci_irq_vector(vp_dev->pci_dev, i));
> > > > > > + disable_irq(pci_irq_vector(vp_dev->pci_dev, i));
> > > > > > +}
> > > > > > +
> > > > > > +/* enable irq handlers */
> > > > > > +void vp_enable_vectors(struct virtio_device *vdev)
> > > > > > +{
> > > > > > + struct virtio_pci_device *vp_dev = to_vp_device(vdev);
> > > > > > + int i;
> > > > > > +
> > > > > > + if (vp_dev->intx_enabled)
> > > > > > + return;
> > > > > > +
> > > > > > + for (i = 0; i < vp_dev->msix_vectors; ++i)
> > > > > > + enable_irq(pci_irq_vector(vp_dev->pci_dev, i));
> > > > > > }
> > > > > >
> > > > > > /* the notify function used when creating a virt queue */
> > > > > > @@ -141,7 +154,8 @@ static int vp_request_msix_vectors(struct virtio_device *vdev, int nvectors,
> > > > > > snprintf(vp_dev->msix_names[v], sizeof *vp_dev->msix_names,
> > > > > > "%s-config", name);
> > > > > > err = request_irq(pci_irq_vector(vp_dev->pci_dev, v),
> > > > > > - vp_config_changed, 0, vp_dev->msix_names[v],
> > > > > > + vp_config_changed, IRQF_NO_AUTOEN,
> > > > > > + vp_dev->msix_names[v],
> > > > > > vp_dev);
> > > > > > if (err)
> > > > > > goto error;
> > > > > > @@ -160,7 +174,8 @@ static int vp_request_msix_vectors(struct virtio_device *vdev, int nvectors,
> > > > > > snprintf(vp_dev->msix_names[v], sizeof *vp_dev->msix_names,
> > > > > > "%s-virtqueues", name);
> > > > > > err = request_irq(pci_irq_vector(vp_dev->pci_dev, v),
> > > > > > - vp_vring_interrupt, 0, vp_dev->msix_names[v],
> > > > > > + vp_vring_interrupt, IRQF_NO_AUTOEN,
> > > > > > + vp_dev->msix_names[v],
> > > > > > vp_dev);
> > > > > > if (err)
> > > > > > goto error;
> > > > > > @@ -337,7 +352,7 @@ static int vp_find_vqs_msix(struct virtio_device *vdev, unsigned nvqs,
> > > > > > "%s-%s",
> > > > > > dev_name(&vp_dev->vdev.dev), names[i]);
> > > > > > err = request_irq(pci_irq_vector(vp_dev->pci_dev, msix_vec),
> > > > > > - vring_interrupt, 0,
> > > > > > + vring_interrupt, IRQF_NO_AUTOEN,
> > > > > > vp_dev->msix_names[msix_vec],
> > > > > > vqs[i]);
> > > > > > if (err)
> > > > > > diff --git a/drivers/virtio/virtio_pci_common.h b/drivers/virtio/virtio_pci_common.h
> > > > > > index beec047a8f8d..a235ce9ff6a5 100644
> > > > > > --- a/drivers/virtio/virtio_pci_common.h
> > > > > > +++ b/drivers/virtio/virtio_pci_common.h
> > > > > > @@ -102,8 +102,10 @@ static struct virtio_pci_device *to_vp_device(struct virtio_device *vdev)
> > > > > > return container_of(vdev, struct virtio_pci_device, vdev);
> > > > > > }
> > > > > >
> > > > > > -/* wait for pending irq handlers */
> > > > > > -void vp_synchronize_vectors(struct virtio_device *vdev);
> > > > > > +/* disable irq handlers */
> > > > > > +void vp_disable_vectors(struct virtio_device *vdev);
> > > > > > +/* enable irq handlers */
> > > > > > +void vp_enable_vectors(struct virtio_device *vdev);
> > > > > > /* the notify function used when creating a virt queue */
> > > > > > bool vp_notify(struct virtqueue *vq);
> > > > > > /* the config->del_vqs() implementation */
> > > > > > diff --git a/drivers/virtio/virtio_pci_legacy.c b/drivers/virtio/virtio_pci_legacy.c
> > > > > > index d62e9835aeec..bdf6bc667ab5 100644
> > > > > > --- a/drivers/virtio/virtio_pci_legacy.c
> > > > > > +++ b/drivers/virtio/virtio_pci_legacy.c
> > > > > > @@ -97,8 +97,8 @@ static void vp_reset(struct virtio_device *vdev)
> > > > > > /* Flush out the status write, and flush in device writes,
> > > > > > * including MSi-X interrupts, if any. */
> > > > > > ioread8(vp_dev->ioaddr + VIRTIO_PCI_STATUS);
> > > > > > - /* Flush pending VQ/configuration callbacks. */
> > > > > > - vp_synchronize_vectors(vdev);
> > > > > > + /* Disable VQ/configuration callbacks. */
> > > > > > + vp_disable_vectors(vdev);
> > > > > > }
> > > > > >
> > > > > > static u16 vp_config_vector(struct virtio_pci_device *vp_dev, u16 vector)
> > > > > > @@ -194,6 +194,7 @@ static void del_vq(struct virtio_pci_vq_info *info)
> > > > > > }
> > > > > >
> > > > > > static const struct virtio_config_ops virtio_pci_config_ops = {
> > > > > > + .ready = vp_enable_vectors,
> > > > > > .get = vp_get,
> > > > > > .set = vp_set,
> > > > > > .get_status = vp_get_status,
> > > > > > diff --git a/drivers/virtio/virtio_pci_modern.c b/drivers/virtio/virtio_pci_modern.c
> > > > > > index 30654d3a0b41..acf0f6b6381d 100644
> > > > > > --- a/drivers/virtio/virtio_pci_modern.c
> > > > > > +++ b/drivers/virtio/virtio_pci_modern.c
> > > > > > @@ -172,8 +172,8 @@ static void vp_reset(struct virtio_device *vdev)
> > > > > > */
> > > > > > while (vp_modern_get_status(mdev))
> > > > > > msleep(1);
> > > > > > - /* Flush pending VQ/configuration callbacks. */
> > > > > > - vp_synchronize_vectors(vdev);
> > > > > > + /* Disable VQ/configuration callbacks. */
> > > > > > + vp_disable_vectors(vdev);
> > > > > > }
> > > > > >
> > > > > > static u16 vp_config_vector(struct virtio_pci_device *vp_dev, u16 vector)
> > > > > > @@ -380,6 +380,7 @@ static bool vp_get_shm_region(struct virtio_device *vdev,
> > > > > > }
> > > > > >
> > > > > > static const struct virtio_config_ops virtio_pci_config_nodev_ops = {
> > > > > > + .ready = vp_enable_vectors,
> > > > > > .get = NULL,
> > > > > > .set = NULL,
> > > > > > .generation = vp_generation,
> > > > > > @@ -397,6 +398,7 @@ static const struct virtio_config_ops virtio_pci_config_nodev_ops = {
> > > > > > };
> > > > > >
> > > > > > static const struct virtio_config_ops virtio_pci_config_ops = {
> > > > > > + .ready = vp_enable_vectors,
> > > > > > .get = vp_get,
> > > > > > .set = vp_set,
> > > > > > .generation = vp_generation,
> > > > > > --
> > > > > > 2.25.1
> > > > >
> > >