Re: [PATCH v4 2/2] PCI: Enable NO_BUS_RESET quirk for Nvidia GPUs
From: Bjorn Helgaas
Date: Tue May 04 2021 - 22:13:15 EST
On Mon, May 03, 2021 at 09:07:11PM -0500, Shanker R Donthineni wrote:
> On 5/3/21 5:42 PM, Bjorn Helgaas wrote:
> > Obviously _RST only works for built-in devices, since there's no AML
> > for plug-in devices, right? So if there's a plug-in card with this
> > GPU, neither SBR nor _RST will work?
> These are not plug-in PCIe GPU cards, will exist on upcoming server
> baseboards. ACPI-reset should wok for plug-in devices as well as long
> as firmware has _RST method defined in ACPI-device associated with
> the PCIe hot-plug slot.
Maybe I'm missing something, but I don't see how _RST can work for
plug-in devices. _RST is part of the system firmware, and that
firmware knows nothing about what will be plugged into the slot. So
if system firmware supplies _RST that knows how to reset the Nvidia
GPU, it's not going to do the right thing if you plug in an NVMe
device instead.
Can you elaborate on how _RST would work for plug-in devices? My only
point here is that IF this GPU is ever on a plug-in card, neither _RST
nor SBR would work, so we'd have to use whatever other reset methods
*do* work (I guess only FLR?)
> I've verified PCIe plug-in feature using SYSFS interface.
>
> 1) Remove device using sysfs interface
> root@test:/sys/bus/pci# echo 1 > devices/0005:01:00.0/remove
> root@test:/sys/bus/pci# lspci -s 0005:01:00.0
>
> 2) Rescan PCI bus using sysfs interface
> root@test:/sys/bus/pci# echo 1 > devices/0005:00:00.0/rescan
> root@test:/sys/bus/pci# lspci -s 0005:01:00.0
> 0005:01:00.0 3D controller: NVIDIA Corporation Device 2341 (rev a1)
>
> 3) List current reset methods
> root@jetson:/sys/bus/pci# cat devices/0005:01:00.0/reset_method
> acpi,flr
>
> Example AML code:
> // Device definition for slot/devfn
> Device(GPU0) {
> Name(_ADR,0x00000000)
> Method (_RST, 0)
> {
> printf("Entering ACPI _RST method")
> // RESET code
> printf("Exiting ACPI _RST method")
> }
> }
>
> 4) Issue device reset from the userspace
> root@test:/sys/bus/pci# echo 1 > devices/0005:01:00.0/reset
>
> dmesg:
> [ 6156.426303] ACPI Debug: "Entering PCI9 _RST method"
> [ 6156.427007] ACPI Debug: "Exiting PCI9 _RST method"
>
> > I'm wondering if we should log something to dmesg in
> > quirk_no_bus_reset(), quirk_no_pm_reset(), quirk_no_flr(), etc., just
> > so we have a hint about the fact that resets won't work quite as
> > expected on these devices.
> Yes, it would be very useful to know what PCI quirks were applied
> during boot. Should I create a separate patch for adding pci_info()
> or include as part of this patch?
Don't include it as part of this patch. It's a separate logical
change so should be a separate patch. We can worry about that later.
> --- a/drivers/pci/quirks.c
> +++ b/drivers/pci/quirks.c
> @@ -3556,6 +3556,7 @@ DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_MELLANOX, PCI_ANY_ID,
> static void quirk_no_bus_reset(struct pci_dev *dev)
> {
> dev->dev_flags |= PCI_DEV_FLAGS_NO_BUS_RESET;
> +pci_info(dev, "Applied NO_BUS_RESET quirk\n");
> }
>
> /*
> @@ -3598,6 +3599,7 @@ static void quirk_no_pm_reset(struct pci_dev *dev)
> */
> if (!pci_is_root_bus(dev->bus))
> dev->dev_flags |= PCI_DEV_FLAGS_NO_PM_RESET;
> +pci_info(dev, "Applied NO_PM_RESET quirk\n");
> }
>
> /*
> @@ -5138,6 +5140,7 @@ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_INTEL, 0x443, quirk_intel_qat_vf_cap);
> static void quirk_no_flr(struct pci_dev *dev)
> {
> dev->dev_flags |= PCI_DEV_FLAGS_NO_FLR_RESET;
> +pci_info(dev, "Applied NO_FLR_RESET quirk\n");
> }
>
>