Re: [PULL v4 42/83] virtio-rng-pci: Allow setting nvectors, so we can use MSI-X

[Michael S. Tsirkin]

On Mon, Jan 09, 2023 at 10:20:45AM +0000, Dr. David Alan Gilbert wrote:
> * Michael S. Tsirkin (mst@redhat.com) wrote:
> > From: David Daney <david.daney@fungible.com>
> > 
> > Most other virtio-pci devices allow MSI-X, let's have it for rng too.
> > 
> > Signed-off-by: David Daney <david.daney@fungible.com>
> > Reviewed-by: Marcin Nowakowski <marcin.nowakowski@fungible.com>
> > Signed-off-by: Philippe Mathieu-Daudé <philmd@fungible.com>
> > Message-Id: <20221014160947.66105-1-philmd@fungible.com>
> > Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
> > Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
> > Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> 
> This breaks migration compatibility 7.1->7.2 :
> 
> (qemu) qemu: get_pci_config_device: Bad config data: i=0x34 read: 84 device: 
> 98 cmask: ff wmask: 0 w1cmask:0
> qemu: Failed to load PCIDevice:config
> qemu: Failed to load virtio-rng:virtio
> qemu: error while loading state for instance 0x0 of device 
> '0000:00:03.0/virtio-rng'
> qemu: load of migration failed: Invalid argument
> 
> because the destination is configured with msi-x but the source isn't.
> 
> The fix is in theory simple:
> diff --git a/hw/core/machine.c b/hw/core/machine.c
> index f589b92909..45459d1cef 100644
> --- a/hw/core/machine.c
> +++ b/hw/core/machine.c
> @@ -45,6 +45,7 @@ const size_t hw_compat_7_2_len = 
> G_N_ELEMENTS(hw_compat_7_2);
>  
>  GlobalProperty hw_compat_7_1[] = {
>      { "virtio-device", "queue_reset", "false" },
> +    { "virtio-rng-pci", "vectors", "0" },
>  };
>  const size_t hw_compat_7_1_len = G_N_ELEMENTS(hw_compat_7_1);
> 
> the gotcha is that will break 7.2->7.2-fixed.
> 
> (I guess you can also work around it by explicitly passing vectors=0 to
> the virtio-rng on the cli)
> 
> Does anyone have preferences as to whether that should be fixed in the
> 7.2 world or left as is?
> 
> This is:
> https://bugzilla.redhat.com/show_bug.cgi?id=2155749
> 
> Dave

I think that yes, it should be fixed in 7.2.


> > ---
> >  hw/virtio/virtio-rng-pci.c | 14 ++++++++++++++
> >  1 file changed, 14 insertions(+)
> > 
> > diff --git a/hw/virtio/virtio-rng-pci.c b/hw/virtio/virtio-rng-pci.c
> > index 151ece6f94..6e76f8b57b 100644
> > --- a/hw/virtio/virtio-rng-pci.c
> > +++ b/hw/virtio/virtio-rng-pci.c
> > @@ -13,6 +13,7 @@
> >  
> >  #include "hw/virtio/virtio-pci.h"
> >  #include "hw/virtio/virtio-rng.h"
> > +#include "hw/qdev-properties.h"
> >  #include "qapi/error.h"
> >  #include "qemu/module.h"
> >  #include "qom/object.h"
> > @@ -31,11 +32,23 @@ struct VirtIORngPCI {
> >      VirtIORNG vdev;
> >  };
> >  
> > +static Property virtio_rng_properties[] = {
> > +    DEFINE_PROP_BIT("ioeventfd", VirtIOPCIProxy, flags,
> > +                    VIRTIO_PCI_FLAG_USE_IOEVENTFD_BIT, true),
> > +    DEFINE_PROP_UINT32("vectors", VirtIOPCIProxy, nvectors,
> > +                       DEV_NVECTORS_UNSPECIFIED),
> > +    DEFINE_PROP_END_OF_LIST(),
> > +};
> > +
> >  static void virtio_rng_pci_realize(VirtIOPCIProxy *vpci_dev, Error **errp)
> >  {
> >      VirtIORngPCI *vrng = VIRTIO_RNG_PCI(vpci_dev);
> >      DeviceState *vdev = DEVICE(&vrng->vdev);
> >  
> > +    if (vpci_dev->nvectors == DEV_NVECTORS_UNSPECIFIED) {
> > +        vpci_dev->nvectors = 2;
> > +    }
> > +
> >      if (!qdev_realize(vdev, BUS(&vpci_dev->bus), errp)) {
> >          return;
> >      }
> > @@ -54,6 +67,7 @@ static void virtio_rng_pci_class_init(ObjectClass *klass, 
> > void *data)
> >      pcidev_k->device_id = PCI_DEVICE_ID_VIRTIO_RNG;
> >      pcidev_k->revision = VIRTIO_PCI_ABI_VERSION;
> >      pcidev_k->class_id = PCI_CLASS_OTHERS;
> > +    device_class_set_props(dc, virtio_rng_properties);
> >  }
> >  
> >  static void virtio_rng_initfn(Object *obj)
> > -- 
> > MST
> > 
> > 
> -- 
> Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK