[libvirt] [PATCH] qemu: Explain why mdevs are assumed to be PCI Express

[Erik Skultety]

On Mon, Jun 12, 2017 at 06:00:13PM +0800, Andrea Bolognani wrote:
> Signed-off-by: Andrea Bolognani <abologna at redhat.com>
> ---
>  src/qemu/qemu_domain_address.c | 5 +++++
>  1 file changed, 5 insertions(+)
>
> diff --git a/src/qemu/qemu_domain_address.c b/src/qemu/qemu_domain_address.c
> index 2106b34..b5b863f 100644
> --- a/src/qemu/qemu_domain_address.c
> +++ b/src/qemu/qemu_domain_address.c
> @@ -645,6 +645,11 @@ qemuDomainDeviceCalculatePCIConnectFlags(virDomainDeviceDefPtr dev,
>              return pcieFlags;
>          }
>
> +        /* mdevs don't have corresponding files in /sys that we can poke to
> +         * try and figure out whether they are legacy PCI or PCI Express, so
> +         * the logic below would never work; instead, we just go ahead and
> +         * assume they're PCI Express. This is a very reasonable assumption,
> +         * as all current mdev-capable devices are indeed PCI Express */
>          if (hostdev->source.subsys.type == VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_MDEV)
>              return pcieFlags;
>
> --
> 2.7.5
>

ACK

Erik