Re: [PATCH] pci-ids.rst: Add Red Hat pci-id for AMD IOMMU device

["Michael S. Tsirkin" <mst@redhat.com>]

On Sun, May 11, 2025 at 06:20:41PM +0300, Yan Vugenfirer wrote:
> On Sun, May 11, 2025 at 4:27 PM Michael S. Tsirkin <mst@redhat.com> wrote:
> >
> > On Wed, Mar 12, 2025 at 02:43:52PM +0200, Yan Vugenfirer wrote:
> > >
> > >
> > > On Tue, Mar 11, 2025 at 4:02 AM Suthikulpanit, Suravee <
> > > suravee.suthikulpanit@amd.com> wrote:
> > >
> > >
> > >
> > >     On 3/9/2025 8:44 PM, Michael S. Tsirkin wrote:
> > >     > On Tue, Mar 04, 2025 at 06:37:47PM +0000, Suravee Suthikulpanit wrote:
> > >     >> The QEMU-emulated AMD IOMMU PCI device is implemented based on the AMD I
> > >     /O
> > >     >> Virtualization Technology (IOMMU) Specification [1]. The PCI id for this
> > >     >> device is platform-specific.
> > >     >>
> > >     >> Currently, the QEMU-emulated AMD IOMMU device is using AMD vendor id and
> > >     >> undefined device id.
> > >     >>
> > >     >> Therefore, change the vendor id to Red Hat and request a new
> > >     QEMU-specific
> > >     >> device id.
> > >     >>
> > >     >> [1] https://www.amd.com/content/dam/amd/en/documents/processor-tech-docs
> > >     /specifications/48882_IOMMU.pdf
> > >     >>
> > >     >> Cc: Gerd Hoffmann <kraxel@redhat.com>
> > >     >> Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>
> > >     >
> > >     > Will the existing drivers bind with the device then?
> > >
> > >     Existing Windows would not recognize the device ID.
> > >
> > >     Actually, Linux and Windows does not depend on the PCI vendor / device
> > >     ids to probe devices and initialize AMD IOMMU. Instead, it depends on
> > >     the ACPI IVRS table.
> > >
> > >     Checking on a real system w/ AMD IOMMU enabled booting Windows Server
> > >     2022, there is no AMD IOMMU device showing in the Device Manger.
> > >
> > >     In this case, I believe Windows is not fully initializing the
> > >     QEMU-emulated AMD IOMMU. So Windows would not remove the IOMMU PCIe from
> > >     the list of OS visible devices and therefore expose the PNPID to the
> > >     device manager. And since the device ID is zero, it appears as an "Other
> > >     devices->PCI Device (with warning sign).
> > >
> > >     Therefore, it we have two options:
> > >
> > >     1. Fake the device ID to 0x1419, which is current appear in the
> > >     machine.inf as an entry in section [AMD_SYS.NTamd64]:
> > >
> > >     %IommuDevice_Desc% = NO_DRV,PCI\VEN_1022&DEV_1419
> > >
> > > Considering that this is a "null driver" (no actual driver is loaded for the
> > > PCIe endpoint according to machine.inf), I recommend using this PNP ID.
> > >
> >
> > So this is the status quo really or did I miss anything?
> 
> There was a patch fixing it:
> https://patchwork.kernel.org/project/qemu-devel/patch/20250325021140.5676-1-suravee.suthikulpanit@amd.com/

Ah, right. Thanks.

> >
> > >
> > >     2. Figure out why Windows does not recognize the device.
> > >
> > > The answer is simple: the PCIe endpoint's PNP ID is unknown to Windows. So
> > > technically device is recognized (it is shown in Device Manager after all), but
> > > there are no compatible drivers for it. And in anycase, machine.inf specifies
> > > "null driver" for AMD PCIe endpoint IOMMU device. The device will get a
> > > friendly name in Device Manager and considered to be "installed" by Windows, by
> > > there is no actual driver associated with the device.
> > >
> > > Best regards,
> > > Yan.
> > >
> > >
> > >
> > >     Anyhow, we should still assign some PCI ID value (instead of zero).
> > >
> > >     Thanks,
> > >     Suravee
> > >
> > >
> > >
> > >
> >