Lukas Wunner
2022-Feb-28 22:32 UTC
[Nouveau] [PATCH v5 3/7] PCI: Drop the `is_thunderbolt` attribute from PCI core
On Mon, Feb 28, 2022 at 04:13:44PM -0600, Bjorn Helgaas wrote:> On Mon, Feb 28, 2022 at 03:33:13PM +0000, Limonciello, Mario wrote: > > > On Fri, Feb 25, 2022 at 11:42:24AM -0600, Bjorn Helgaas wrote: > > > > That would just leave the "PCI_VSEC_ID_INTEL_TBT implies external- > > > facing" > > > > assumption above. Not having a Thunderbolt spec, I have no idea how > > > > you deal with that. > > > > > > You can download the spec here:[...]> > > Inside the archive there is also the DVSEC spec with name "USB4 DVSEC > > > Version 1.0.pdf". > > > > The spec has Host_Router_indication (bits 18-19) as meaning external facing. > > I'll respin the patch 3 for using that. > > Thanks, please include the spec citation when you do. And probably > the URL, because it's not at all obvious how the casual reader would > get from "is_thunderbolt" to a recent add-on to the USB4 spec.PCI_VSEC_ID_INTEL_TBT is not mentioned at all in the USB4 spec, hence there's no connection between "is_thunderbolt" and the USB4 spec. It's a proprietary VSEC used by Intel and the only way to recognize pre-USB4 Thunderbolt devices that I know of. Its ID is also different from the DVSEC IDs given in the above-mentioned spec. Thanks, Lukas
Limonciello, Mario
2022-Feb-28 22:36 UTC
[Nouveau] [PATCH v5 3/7] PCI: Drop the `is_thunderbolt` attribute from PCI core
[AMD Official Use Only]> -----Original Message----- > From: Lukas Wunner <lukas at wunner.de> > Sent: Monday, February 28, 2022 16:33 > To: Bjorn Helgaas <helgaas at kernel.org> > Cc: Limonciello, Mario <Mario.Limonciello at amd.com>; Mika Westerberg > <mika.westerberg at linux.intel.com>; Michael Jamet > <michael.jamet at intel.com>; open list:PCI SUBSYSTEM <linux- > pci at vger.kernel.org>; open list:THUNDERBOLT DRIVER <linux- > usb at vger.kernel.org>; Yehezkel Bernat <YehezkelShB at gmail.com>; open > list:DRM DRIVERS <dri-devel at lists.freedesktop.org>; open list:X86 > PLATFORM DRIVERS <platform-driver-x86 at vger.kernel.org>; Andreas > Noever <andreas.noever at gmail.com>; open list:RADEON and AMDGPU > DRM DRIVERS <amd-gfx at lists.freedesktop.org>; open list:DRM DRIVER FOR > NVIDIA GEFORCE/QUADRO GPUS <nouveau at lists.freedesktop.org>; Bjorn > Helgaas <bhelgaas at google.com>; Deucher, Alexander > <Alexander.Deucher at amd.com> > Subject: Re: [PATCH v5 3/7] PCI: Drop the `is_thunderbolt` attribute from PCI > core > > On Mon, Feb 28, 2022 at 04:13:44PM -0600, Bjorn Helgaas wrote: > > On Mon, Feb 28, 2022 at 03:33:13PM +0000, Limonciello, Mario wrote: > > > > On Fri, Feb 25, 2022 at 11:42:24AM -0600, Bjorn Helgaas wrote: > > > > > That would just leave the "PCI_VSEC_ID_INTEL_TBT implies external- > > > > facing" > > > > > assumption above. Not having a Thunderbolt spec, I have no idea > how > > > > > you deal with that. > > > > > > > > You can download the spec here: > [...] > > > > Inside the archive there is also the DVSEC spec with name "USB4 DVSEC > > > > Version 1.0.pdf". > > > > > > The spec has Host_Router_indication (bits 18-19) as meaning external > facing. > > > I'll respin the patch 3 for using that. > > > > Thanks, please include the spec citation when you do. And probably > > the URL, because it's not at all obvious how the casual reader would > > get from "is_thunderbolt" to a recent add-on to the USB4 spec. > > PCI_VSEC_ID_INTEL_TBT is not mentioned at all in the USB4 spec, > hence there's no connection between "is_thunderbolt" and the USB4 spec. > > It's a proprietary VSEC used by Intel and the only way to recognize > pre-USB4 Thunderbolt devices that I know of. Its ID is also > different from the DVSEC IDs given in the above-mentioned spec. > > Thanks,The USB4 DVSEC spec makes comments about DVSEC_ID of 0x8086 and also DVSEC VENDOR_ID of 0x8086. Is that not also present on the Intel TBT3 controllers? My interpretation of this (and Mika's comment) was that rather than looking at the Intel VSEC we should look at the USB4 DVSEC to detect the Intel TBT3 controllers.
Mika Westerberg
2022-Mar-01 07:04 UTC
[Nouveau] [PATCH v5 3/7] PCI: Drop the `is_thunderbolt` attribute from PCI core
Hi, On Mon, Feb 28, 2022 at 10:36:59PM +0000, Limonciello, Mario wrote:> [AMD Official Use Only] > > > -----Original Message----- > > From: Lukas Wunner <lukas at wunner.de> > > Sent: Monday, February 28, 2022 16:33 > > To: Bjorn Helgaas <helgaas at kernel.org> > > Cc: Limonciello, Mario <Mario.Limonciello at amd.com>; Mika Westerberg > > <mika.westerberg at linux.intel.com>; Michael Jamet > > <michael.jamet at intel.com>; open list:PCI SUBSYSTEM <linux- > > pci at vger.kernel.org>; open list:THUNDERBOLT DRIVER <linux- > > usb at vger.kernel.org>; Yehezkel Bernat <YehezkelShB at gmail.com>; open > > list:DRM DRIVERS <dri-devel at lists.freedesktop.org>; open list:X86 > > PLATFORM DRIVERS <platform-driver-x86 at vger.kernel.org>; Andreas > > Noever <andreas.noever at gmail.com>; open list:RADEON and AMDGPU > > DRM DRIVERS <amd-gfx at lists.freedesktop.org>; open list:DRM DRIVER FOR > > NVIDIA GEFORCE/QUADRO GPUS <nouveau at lists.freedesktop.org>; Bjorn > > Helgaas <bhelgaas at google.com>; Deucher, Alexander > > <Alexander.Deucher at amd.com> > > Subject: Re: [PATCH v5 3/7] PCI: Drop the `is_thunderbolt` attribute from PCI > > core > > > > On Mon, Feb 28, 2022 at 04:13:44PM -0600, Bjorn Helgaas wrote: > > > On Mon, Feb 28, 2022 at 03:33:13PM +0000, Limonciello, Mario wrote: > > > > > On Fri, Feb 25, 2022 at 11:42:24AM -0600, Bjorn Helgaas wrote: > > > > > > That would just leave the "PCI_VSEC_ID_INTEL_TBT implies external- > > > > > facing" > > > > > > assumption above. Not having a Thunderbolt spec, I have no idea > > how > > > > > > you deal with that. > > > > > > > > > > You can download the spec here: > > [...] > > > > > Inside the archive there is also the DVSEC spec with name "USB4 DVSEC > > > > > Version 1.0.pdf". > > > > > > > > The spec has Host_Router_indication (bits 18-19) as meaning external > > facing. > > > > I'll respin the patch 3 for using that. > > > > > > Thanks, please include the spec citation when you do. And probably > > > the URL, because it's not at all obvious how the casual reader would > > > get from "is_thunderbolt" to a recent add-on to the USB4 spec. > > > > PCI_VSEC_ID_INTEL_TBT is not mentioned at all in the USB4 spec, > > hence there's no connection between "is_thunderbolt" and the USB4 spec. > > > > It's a proprietary VSEC used by Intel and the only way to recognize > > pre-USB4 Thunderbolt devices that I know of. Its ID is also > > different from the DVSEC IDs given in the above-mentioned spec. > > > > Thanks, > > The USB4 DVSEC spec makes comments about DVSEC_ID of 0x8086 and also > DVSEC VENDOR_ID of 0x8086. Is that not also present on the Intel TBT3 controllers? > > My interpretation of this (and Mika's comment) was that rather than > looking at the Intel VSEC we should look at the USB4 DVSEC to detect > the Intel TBT3 controllers.For pre-USB4 controllers (TBT 1-3) we need to use the existing method (or a quirk based on device ID) as they don't have the USB4 DVSEC.