On Thu, Sep 28, 2017 at 12:39:23PM +0300, Marcel Apfelbaum wrote:
> On 27/09/2017 22:56, Eduardo Habkost wrote:
> > Changes v1 -> v2:
> > * Use "Conventional PCI" instead of "legacy PCI"
> > * Suggested-by: Alex Williamson <alex.william...@redhat.com>
> > * Mark base-xhci as hybrid too
> > * Reported-by: Marcel Apfelbaum <mar...@redhat.com>
> > * Mark pcie-pci-bridge as PCI Express only
> > * New Conventional PCI devices: sungem, sunhme
> >
> > v1 was a reimplementation of one portion of an old RFC:
> > Subject: [RFC v2 00/20] qmp: Report bus information on
> > 'query-machines'
> >
> > This series marks each PCI device type as Conventional PCI, PCI
> > Express, or "hybrid". It uses two new QOM interface names to do
> > that: INTERFACE_CONVENTIONAL_PCI_DEVICE
> > ("conventional-pci-device") and INTERFACE_PCIE_DEVICE
> > ("pci-express-device"). Conventional PCI devices will implement
> > only the former; PCIe-only devices will implement only the
> > latter; hybrid devices will implement both interfaces.
> >
> > With this, management software will then be able to use
> > qom-list-types to find out which PCI devices are
> > conventional/express/hybrid.
> >
> > In the future, the new interface names can be used in the
> > bus/slot querying commands, to indicate which types of devices
> > are accepted on each slot.
> >
> > The last patch in the series adds an assertion to the PCI device
> > class code, to ensure we won't forget to add the corresponding
> > interface names to new PCI device classes.
> >
>
> Hi Eduardo,
>
> Now we need to get rid of the "is_express" field,
> otherwise we have the same information on two places.
Agreed, it's on my TODO-list. is_express is completely
inconsistent on hybrid devices (some of them have is_express=0,
some of them have is_express=1).
--
Eduardo
