Re: xen-pci-passthrough PCI Express support? (Re: [Qemu-devel] [PATCH v2 4/5] pci: Add INTERFACE_CONVENTIONAL_PCI_DEVICE to Conventional PCI devices)

From: Stefano Stabellini <sstabellini@kernel.org>
To: Eduardo Habkost <ehabkost@redhat.com>
Cc: "Peter Maydell" <peter.maydell@linaro.org>,
	"Michael S. Tsirkin" <mst@redhat.com>,
	"Jason Wang" <jasowang@redhat.com>,
	"Mark Cave-Ayland" <mark.cave-ayland@ilande.co.uk>,
	qemu-devel@nongnu.org, "Gerd Hoffmann" <kraxel@redhat.com>,
	"Jan Beulich" <JBeulich@suse.com>,
	"David Gibson" <dgibson@redhat.com>,
	"Jiri Slaby" <jslaby@suse.cz>, "Hannes Reinecke" <hare@suse.com>,
	"Stefano Stabellini" <sstabellini@kernel.org>,
	"Alberto Garcia" <berto@igalia.com>,
	qemu-block@nongnu.org, "Alexander Graf" <agraf@suse.de>,
	"Hervé Poussineau" <hpoussin@reactos.org>,
	"Marcel Apfelbaum" <marcel@redhat.com>,
	xen-devel@lists.xenproject.org,
	"David Gibson" <david@gibson.dropbear.id.au>,
	"Artyom Tarasenko" <atar4qemu@gmail.com>,
	"Jiri Pirko" <jiri@resnulli.us>,
	"Anthony Perard" <anthony.perard@citrix.com>,
	"Alex Williamson" <alex.williamson@redhat.com>,
	qemu-arm@nongnu.org, "Laine Stump" <laine@redhat.com>
Subject: Re: xen-pci-passthrough PCI Express support? (Re: [Qemu-devel] [PATCH v2 4/5] pci: Add INTERFACE_CONVENTIONAL_PCI_DEVICE to Conventional PCI devices)
Date: Tue, 3 Oct 2017 10:12:11 -0700 (PDT)	[thread overview]
Message-ID: <alpine.DEB.2.10.1710031003331.17914__23053.219021722$1507050804$gmane$org@sstabellini-ThinkPad-X260> (raw)
In-Reply-To: <20171003001257.GA7087@localhost.localdomain>

On Mon, 2 Oct 2017, Eduardo Habkost wrote:
> On Thu, Sep 28, 2017 at 10:12:34AM -0300, Eduardo Habkost wrote:
> > On Thu, Sep 28, 2017 at 02:33:57AM -0600, Jan Beulich wrote:
> > > >>> On 27.09.17 at 21:56, <ehabkost@redhat.com> wrote:
> > > > --- a/hw/xen/xen_pt.c
> > > > +++ b/hw/xen/xen_pt.c
> > > > @@ -964,6 +964,10 @@ static const TypeInfo xen_pci_passthrough_info = {
> > > >      .instance_size = sizeof(XenPCIPassthroughState),
> > > >      .instance_finalize = xen_pci_passthrough_finalize,
> > > >      .class_init = xen_pci_passthrough_class_init,
> > > > +    .interfaces = (InterfaceInfo[]) {
> > > > +        { INTERFACE_CONVENTIONAL_PCI_DEVICE },
> > > > +        { },
> > > > +    },
> > > >  };
> > > 
> > > Passed through devices can be both PCI and PCIe, so following
> > > the description of the patch I don't think these can be statically
> > > given either property. Granted quite a bit of PCIe specific
> > > functionality may be missing in the Xen code ...
> > 
> > This is just static data about what the device type supports, not
> > about what a given device instance really is.  Deciding if the
> > device is PCIe or Conventional at runtime is out of the scope of
> > this series.
> > 
> > That said, if passed through PCI Express devices are really
> > supported, it looks like this should be marked as hybrid.
> 
> Can anybody confirm if PCI Express devices are really supported
> by xen-pci-passthrough?
> 
> I suggest we add only INTERFACE_CONVENTIONAL_PCI_DEVICE to the
> class info until we confirm that.
> 
> (In other words, apply this patch as-is, and add
> INTERFACE_PCIE_DEVICE later as a follow-up patch if appropriate.)

Yes, xen-pci-passthrough can expose PCI Express devices to the guest.  I
don't think the old xen-pci-passthrough code does anything special to
mark it as such internally in QEMU, but it can deal with the PCI Express
Capabilities and expose/emulate them for the guest's benefit.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel