Skip to contentSkip to conversation

PCI: Mark AMD Navi14 GPU 0x7341 rev 0x00 ATS as broken

PATCH In review Evan Quan 2 participants 3 messages May 31, 2021 – Jun 1, 2021 PCI

lore
Share & tools
lore

Evan Quan

author
#0 lore raw

PATCH 1/1 PCI: Mark AMD Navi14 GPU 0x7341 rev 0x00 ATS as broken

Unexpected GPU hang was observed during runpm stress test on 0x7341 rev 0x00. Further debugging shows broken ATS is related. Thus as a followup of commit 5e89cd303e3a ("PCI:
Mark AMD Navi14 GPU rev 0xc5 ATS as broken"), we disable the ATS for the specific SKU also.

Change-Id I3d9d570bd473762e3bfbb251cf8abaf5af38ced9
---
 drivers/pci/quirks.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/pci/quirks.c +3 -1
index b7e19bbb901a..70803ad6d2ac 100644
@@ -5176,7 +5176,8 @@ 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 == 0x7312 && pdev->revision != 0x00) ||
-	    (pdev->device == 0x7340 && pdev->revision != 0xc5))
+	    (pdev->device == 0x7340 && pdev->revision != 0xc5) ||
+	    (pdev->device == 0x7341 && pdev->revision != 0x00))
 		return;
 
 	if (pdev->device == 0x15d8) {
@@ -5203,6 +5204,7 @@ DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x6900, quirk_amd_harvest_no_ats);
 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);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x7341, quirk_amd_harvest_no_ats);
 /* AMD Raven platform iGPU */
 DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x15d8, quirk_amd_harvest_no_ats);
 #endif /* CONFIG_PCI_ATS */
Signature · 1 line
2.29.0
#1 lore raw

Hi Evan,

Thank you for sending the patch over!

A small nitpick: the subject line could just say:

Add quirk for AMD Navi 14 to disable ATS support

Or something along these lines, as I am not sure how useful the ID and revision are in the subject, especially since the commit message explains in details what hardware is affected, etc.

Quoted 5 lines

Unexpected GPU hang was observed during runpm stress test on 0x7341 rev 0x00. Further debugging shows broken ATS is related. Thus as a followup of commit 5e89cd303e3a ("PCI:
Mark AMD Navi14 GPU rev 0xc5 ATS as broken"), we disable the ATS for the specific SKU also.

As this might be a candidate for a back-port to current stable and long-term kernels, does it have any "Fixes" tag we could include here for reference? If not, then it's OK.

Quoted 1 line

Change-Id: I3d9d570bd473762e3bfbb251cf8abaf5af38ced9

I assume this is from some code review service like Gerrit? We usually as people to drop these when sending patches to be included in the kernel (so when sending patches upstream).

Krzysztof

author
#2 lore raw

[AMD Official Use Only]

Thanks Krzysztof. Comment inline

Quoted 13 lines

-----Original Message-----
From: Krzysztof Wilczyński <[email protected]>
Sent: Monday, May 31, 2021 10:41 PM
To: Quan, Evan <[email protected]>
Cc: [email protected]; Deucher, Alexander <[email protected]>
Subject: Re: [PATCH] PCI: Mark AMD Navi14 GPU 0x7341 rev 0x00 ATS as broken

Hi Evan,

Thank you for sending the patch over!

Quoted 8 lines

A small nitpick: the subject line could just say:

Add quirk for AMD Navi 14 to disable ATS support

Or something along these lines, as I am not sure how useful the ID and revision are in the subject, especially since the commit message explains in details what hardware is affected, etc.

[Quan, Evan] Will update it in V2.

Quoted 9 lines
Quoted 5 lines

Unexpected GPU hang was observed during runpm stress test on 0x7341 rev 0x00. Further debugging shows broken ATS is related. Thus as a followup of commit 5e89cd303e3a ("PCI:
Mark AMD Navi14 GPU rev 0xc5 ATS as broken"), we disable the ATS for the specific SKU also.

As this might be a candidate for a back-port to current stable and long-term kernels, does it have any "Fixes" tag we could include here for reference? If not, then it's OK.

[Quan, Evan] We have an internal ticket/link for this. But that is not accessible outside AMD. So, I would say there is no "Fixes" tag available for this.

Quoted 8 lines
Quoted 2 lines

Change-Id: I3d9d570bd473762e3bfbb251cf8abaf5af38ced9

I assume this is from some code review service like Gerrit? We usually as people to drop these when sending patches to be included in the kernel (so when sending patches upstream).

Krzysztof

[Quan, Evan] Thanks! Will drop it in V2.

BR
Evan