Public > -----Original Message----- > From: Limonciello, Mario <[email protected]> > Sent: Monday, September 28, 2026 1:15 PM > To: Jason Gunthorpe <[email protected]>; Deucher, Alexander > <[email protected]> > Cc: Mario Limonciello <[email protected]>; Joerg Roedel > <[email protected]>; [email protected]; Suthikulpanit, Suravee > <[email protected]>; Hegde, Vasant > <[email protected]>; Will Deacon <[email protected]>; Robin Murphy > <[email protected]>; open list:AMD IOMMU (AMD-VI) > <[email protected]>; Jatin Kataria <[email protected]>; Boqun Feng > <[email protected]>; Derek J . Clark <[email protected]> > Subject: Re: [PATCH v2 1/2] iommu/amd: Add PerfOpt IOMMU performance > optimization support > > > > On 9/28/26 12:11, Jason Gunthorpe wrote: > > On Mon, Sep 28, 2026 at 10:32:41AM -0500, Mario Limonciello wrote: > > > >> Actually; no. amdgpu has migrated away from PCI IDs and probes based > >> on the "ATI vendor" + "display class" instead. As part of the probe > >> sequence there is something called an IP discovery table pull from > >> the platform that identifies all IP blocks in the hardware. > > > > Oh, so that big table is not processing pci ids.. OK > > > >> This IP discovery table will indicate graphics IP and that is mapped > >> to relevant driver code. So for example GC 11.5.1 is what you find > >> in a Strix Halo APU. If a future APU had exactly the same graphics > >> IP but a different display IP it would still be GC 11.5.1. > > > > Move this stuff to drivers/firmware and iommu and GPU can access it > > together? > > > > "IP discovery table" sure sounds like a platform firmware datablob to > > me at least? > > > > Jason > > Yes; it comes from platform firmware. > > Alex, what do you think of doing this?
It's not platform specific, it's AMD GPU specific. It also exists on dGPUs which will work in any platform. I guess we could put some sort of data table in there for this sort of thing but as it stands today, I don't really see it as being much different from PCI IDs with respect to the IOMMU. I think IVRS tables would be a better fit for this since they are supposed to describe the behavior of the IOMMU with respect to the devices present on the system. Alex
