2020-07-28 10:49:51

by Kai-Heng Feng

[permalink] [raw]
Subject: [PATCH] PCI: Mark AMD Navi10 GPU rev 0x00 ATS as broken

We are seeing AMD Radeon Pro W5700 doesn't work when IOMMU is enabled:
[ 3.375841] iommu ivhd0: AMD-Vi: Event logged [IOTLB_INV_TIMEOUT device=63:00.0 address=0x42b5b01a0]
[ 3.375845] iommu ivhd0: AMD-Vi: Event logged [IOTLB_INV_TIMEOUT device=63:00.0 address=0x42b5b01c0]

The error also makes graphics driver fail to probe the device.

It appears to be the same issue as commit 5e89cd303e3a ("PCI: Mark AMD
Navi14 GPU rev 0xc5 ATS as broken") addresses, and indeed the same ATS
quirk can workaround the issue.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=208725
Cc: Alex Deucher <[email protected]>
Signed-off-by: Kai-Heng Feng <[email protected]>
---
drivers/pci/quirks.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
index 812bfc32ecb8..052efeb9f053 100644
--- a/drivers/pci/quirks.c
+++ b/drivers/pci/quirks.c
@@ -5192,7 +5192,8 @@ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_SERVERWORKS, 0x0422, quirk_no_ext_tags);
*/
static void quirk_amd_harvest_no_ats(struct pci_dev *pdev)
{
- if (pdev->device == 0x7340 && pdev->revision != 0xc5)
+ if ((pdev->device == 0x7312 && pdev->revision != 0x00) ||
+ (pdev->device == 0x7340 && pdev->revision != 0xc5))
return;

pci_info(pdev, "disabling ATS\n");
@@ -5203,6 +5204,8 @@ static void quirk_amd_harvest_no_ats(struct pci_dev *pdev)
DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x98e4, quirk_amd_harvest_no_ats);
/* AMD Iceland dGPU */
DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x6900, quirk_amd_harvest_no_ats);
+/* AMD Navi10 dGPU */
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x7312, quirk_amd_harvest_no_ats);
/* AMD Navi14 dGPU */
DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x7340, quirk_amd_harvest_no_ats);
#endif /* CONFIG_PCI_ATS */
--
2.17.1


2020-07-28 14:38:24

by Deucher, Alexander

[permalink] [raw]
Subject: RE: [PATCH] PCI: Mark AMD Navi10 GPU rev 0x00 ATS as broken

[AMD Public Use]

> -----Original Message-----
> From: Kai-Heng Feng <[email protected]>
> Sent: Tuesday, July 28, 2020 6:46 AM
> To: [email protected]
> Cc: Kai-Heng Feng <[email protected]>; Deucher, Alexander
> <[email protected]>; open list:PCI SUBSYSTEM <linux-
> [email protected]>; open list <[email protected]>
> Subject: [PATCH] PCI: Mark AMD Navi10 GPU rev 0x00 ATS as broken
>
> We are seeing AMD Radeon Pro W5700 doesn't work when IOMMU is
> enabled:
> [ 3.375841] iommu ivhd0: AMD-Vi: Event logged [IOTLB_INV_TIMEOUT
> device=63:00.0 address=0x42b5b01a0]
> [ 3.375845] iommu ivhd0: AMD-Vi: Event logged [IOTLB_INV_TIMEOUT
> device=63:00.0 address=0x42b5b01c0]
>
> The error also makes graphics driver fail to probe the device.
>
> It appears to be the same issue as commit 5e89cd303e3a ("PCI: Mark AMD
> Navi14 GPU rev 0xc5 ATS as broken") addresses, and indeed the same ATS
> quirk can workaround the issue.
>
> Bugzilla:
> https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fbugz
> illa.kernel.org%2Fshow_bug.cgi%3Fid%3D208725&amp;data=02%7C01%7Cal
> exander.deucher%40amd.com%7Cbb49d8e71c29459d631a08d832e36d56%7
> C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637315299664339358&
> amp;sdata=SUAXEIoIJfgTm54FmgwUCMUI%2Bk2qWNcvSpvpU09Ak5k%3D&
> amp;reserved=0
> Cc: Alex Deucher <[email protected]>
> Signed-off-by: Kai-Heng Feng <[email protected]>

This was fixed in the vbios, but apparently that didn't make it out to everyone.
Acked-by: Alex Deucher <[email protected]>

> ---
> drivers/pci/quirks.c | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c index
> 812bfc32ecb8..052efeb9f053 100644
> --- a/drivers/pci/quirks.c
> +++ b/drivers/pci/quirks.c
> @@ -5192,7 +5192,8 @@
> DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_SERVERWORKS, 0x0422,
> quirk_no_ext_tags);
> */
> static void quirk_amd_harvest_no_ats(struct pci_dev *pdev) {
> - if (pdev->device == 0x7340 && pdev->revision != 0xc5)
> + if ((pdev->device == 0x7312 && pdev->revision != 0x00) ||
> + (pdev->device == 0x7340 && pdev->revision != 0xc5))
> return;
>
> pci_info(pdev, "disabling ATS\n");
> @@ -5203,6 +5204,8 @@ static void quirk_amd_harvest_no_ats(struct
> pci_dev *pdev) DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x98e4,
> quirk_amd_harvest_no_ats);
> /* AMD Iceland dGPU */
> DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x6900,
> quirk_amd_harvest_no_ats);
> +/* AMD Navi10 dGPU */
> +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x7312,
> +quirk_amd_harvest_no_ats);
> /* AMD Navi14 dGPU */
> DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x7340,
> quirk_amd_harvest_no_ats); #endif /* CONFIG_PCI_ATS */
> --
> 2.17.1

2020-07-29 19:59:17

by Bjorn Helgaas

[permalink] [raw]
Subject: Re: [PATCH] PCI: Mark AMD Navi10 GPU rev 0x00 ATS as broken

On Tue, Jul 28, 2020 at 06:45:53PM +0800, Kai-Heng Feng wrote:
> We are seeing AMD Radeon Pro W5700 doesn't work when IOMMU is enabled:
> [ 3.375841] iommu ivhd0: AMD-Vi: Event logged [IOTLB_INV_TIMEOUT device=63:00.0 address=0x42b5b01a0]
> [ 3.375845] iommu ivhd0: AMD-Vi: Event logged [IOTLB_INV_TIMEOUT device=63:00.0 address=0x42b5b01c0]
>
> The error also makes graphics driver fail to probe the device.
>
> It appears to be the same issue as commit 5e89cd303e3a ("PCI: Mark AMD
> Navi14 GPU rev 0xc5 ATS as broken") addresses, and indeed the same ATS
> quirk can workaround the issue.
>
> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=208725
> Cc: Alex Deucher <[email protected]>
> Signed-off-by: Kai-Heng Feng <[email protected]>

Applied with Alex's ack to pci/virtualization for v5.9, thanks!

I also added a stable tag since we did that for 5e89cd303e3a. Let me
know if you *don't* want that.

> ---
> drivers/pci/quirks.c | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
> index 812bfc32ecb8..052efeb9f053 100644
> --- a/drivers/pci/quirks.c
> +++ b/drivers/pci/quirks.c
> @@ -5192,7 +5192,8 @@ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_SERVERWORKS, 0x0422, quirk_no_ext_tags);
> */
> static void quirk_amd_harvest_no_ats(struct pci_dev *pdev)
> {
> - if (pdev->device == 0x7340 && pdev->revision != 0xc5)
> + if ((pdev->device == 0x7312 && pdev->revision != 0x00) ||
> + (pdev->device == 0x7340 && pdev->revision != 0xc5))
> return;
>
> pci_info(pdev, "disabling ATS\n");
> @@ -5203,6 +5204,8 @@ static void quirk_amd_harvest_no_ats(struct pci_dev *pdev)
> DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x98e4, quirk_amd_harvest_no_ats);
> /* AMD Iceland dGPU */
> DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x6900, quirk_amd_harvest_no_ats);
> +/* AMD Navi10 dGPU */
> +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x7312, quirk_amd_harvest_no_ats);
> /* AMD Navi14 dGPU */
> DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x7340, quirk_amd_harvest_no_ats);
> #endif /* CONFIG_PCI_ATS */
> --
> 2.17.1
>