[libvirt] [PATCH v4 08/14] qemu: Assign PCI addresses for mediated devices as well

Laine Stump laine at laine.org
Sun Mar 26 18:32:23 UTC 2017


On 03/22/2017 11:27 AM, Erik Skultety wrote:
> So far, the official support is for x86_64 arch guests so unless a
> different device API than vfio-pci is available let's only turn on
> support for PCI address assignment. Once a different device API is
> introduced, we can enable another address type easily.
> 
> Signed-off-by: Erik Skultety <eskultet at redhat.com>
> ---
>  src/qemu/qemu_domain.h         |  1 +
>  src/qemu/qemu_domain_address.c | 14 +++++++++++---
>  2 files changed, 12 insertions(+), 3 deletions(-)
> 
> diff --git a/src/qemu/qemu_domain.h b/src/qemu/qemu_domain.h
> index 1f266bffb5..4377560261 100644
> --- a/src/qemu/qemu_domain.h
> +++ b/src/qemu/qemu_domain.h
> @@ -34,6 +34,7 @@
>  # include "qemu_agent.h"
>  # include "qemu_conf.h"
>  # include "qemu_capabilities.h"
> +# include "virmdev.h"
>  # include "virchrdev.h"
>  # include "virobject.h"
>  # include "logging/log_manager.h"
> diff --git a/src/qemu/qemu_domain_address.c b/src/qemu/qemu_domain_address.c
> index 6d3a627868..22d8bf67d9 100644
> --- a/src/qemu/qemu_domain_address.c
> +++ b/src/qemu/qemu_domain_address.c
> @@ -619,7 +619,8 @@ qemuDomainDeviceCalculatePCIConnectFlags(virDomainDeviceDefPtr dev,
>          virPCIDeviceAddressPtr hostAddr = &hostdev->source.subsys.u.pci.addr;
>  
>          if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS ||
> -            hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI) {
> +            (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI &&
> +             hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_MDEV)) {
>              return 0;
>          }
>  
> @@ -643,6 +644,9 @@ qemuDomainDeviceCalculatePCIConnectFlags(virDomainDeviceDefPtr dev,
>              return pcieFlags;
>          }
>  
> +        if (hostdev->source.subsys.type == VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_MDEV)
> +            return pcieFlags;
> +


I think we talked about this a long time ago, and agreed this was the
correct thing to do, but I just want to make sure everyone
sees/understands/agrees - the bit right above this comment assumes that
all mediated devices are PCI Express devices and should be plugged into
a PCIe port (which would include adding a pcie-root-port and plugging it
into that rather than directly into pcie-root).

(Hopefully nobody will come up with a new mediated device that is
legacy-pci only...)


>          if (!(pciDev = virPCIDeviceNew(hostAddr->domain,
>                                         hostAddr->bus,
>                                         hostAddr->slot,
> @@ -1727,13 +1731,17 @@ qemuDomainAssignDevicePCISlots(virDomainDefPtr def,
>  
>      /* Host PCI devices */
>      for (i = 0; i < def->nhostdevs; i++) {
> +        virDomainHostdevSubsysPtr subsys = &def->hostdevs[i]->source.subsys;
>          if (!virDeviceInfoPCIAddressWanted(def->hostdevs[i]->info))
>              continue;
>          if (def->hostdevs[i]->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS)
>              continue;
> -        if (def->hostdevs[i]->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI &&
> -            def->hostdevs[i]->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_SCSI_HOST)
> +        if (subsys->type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI &&
> +            subsys->type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_SCSI_HOST &&
> +            !(subsys->type == VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_MDEV &&
> +              subsys->u.mdev.model == VIR_MDEV_MODEL_TYPE_VFIO_PCI)) {
>              continue;
> +        }
>  
>          if (qemuDomainPCIAddressReserveNextAddr(addrs,
>                                                  def->hostdevs[i]->info) < 0)
> 

ACK.




More information about the libvir-list mailing list