Skip to contentSkip to conversation

] drm/amdgpu/gfx9: add gfxoff quirk

PATCH In review Daniel Gomez 3 participants 8 messages Mar 10, 2021 – Mar 12, 2021

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ] drm/amdgpu/gfx9: add gfxoff quirk

Disabling GFXOFF via the quirk list fixes a hardware lockup in Ryzen V1605B, RAVEN 0x1002:0x15DD rev 0x83.

---

This patch is a continuation of the work here:
https://lkml.org/lkml/2021/2/3/122 where a hardware lockup was discussed and a dma_fence deadlock was provoke as a side effect. To reproduce the issue please refer to the above link.

The hardware lockup was introduced in 5.6-rc1 for our particular revision as it wasn't part of the new blacklist. Before that, in kernel v5.5, this hardware was working fine without any hardware lock because the GFXOFF was actually disabled by the if condition for the CHIP_RAVEN case. So this patch, adds the 'Radeon Vega Mobile Series [1002:15dd] (rev 83)' to the blacklist to disable the GFXOFF.

But besides the fix, I'd like to ask from where this revision comes from. Is it an ASIC revision or is it hardcoded in the VBIOS from our vendor? From what I can see, it comes from the ASIC and I wonder if somehow we can get an APU in the future, 'not blacklisted', with the same problem. Then, should this table only filter for the vendor and device and not the revision? Do you know if there are any revisions for the 1002:15dd validated, tested and functional?

Logs:

[   27.708348] [drm] initializing kernel modesetting (RAVEN
0x1002:0x15DD 0x1002:0x15DD 0x83).
[   27.789156] amdgpu: ATOM BIOS: 113-RAVEN-115

Thanks in advance,
Daniel

 drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +2 -0
index 65db88bb6cbc..319d4b99aec8 100644
@@ -1243,6 +1243,8 @@ static const struct amdgpu_gfxoff_quirk amdgpu_gfxoff_quirk_list[] = {
 	{ 0x1002, 0x15dd, 0x103c, 0x83e7, 0xd3 },
 	/* GFXOFF is unstable on C6 parts with a VBIOS 113-RAVEN-114 */
 	{ 0x1002, 0x15dd, 0x1002, 0x15dd, 0xc6 },
+	/* GFXOFF provokes a hw lockup on 83 parts with a VBIOS 113-RAVEN-115 */
+	{ 0x1002, 0x15dd, 0x1002, 0x15dd, 0x83 },
 	{ 0, 0, 0, 0, 0 },
 };
 

--
2.30.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#1 lore raw
Quoted 18 lines · Daniel Gomez

Disabling GFXOFF via the quirk list fixes a hardware lockup in Ryzen V1605B, RAVEN 0x1002:0x15DD rev 0x83.

---

This patch is a continuation of the work here:
https://lkml.org/lkml/2021/2/3/122 where a hardware lockup was discussed and a dma_fence deadlock was provoke as a side effect. To reproduce the issue please refer to the above link.

The hardware lockup was introduced in 5.6-rc1 for our particular revision as it wasn't part of the new blacklist. Before that, in kernel v5.5, this hardware was working fine without any hardware lock because the GFXOFF was actually disabled by the if condition for the CHIP_RAVEN case. So this patch, adds the 'Radeon Vega Mobile Series [1002:15dd] (rev 83)' to the blacklist to disable the GFXOFF.

Quoted 6 lines

But besides the fix, I'd like to ask from where this revision comes from. Is it an ASIC revision or is it hardcoded in the VBIOS from our vendor? From what I can see, it comes from the ASIC and I wonder if somehow we can get an APU in the future, 'not blacklisted', with the same problem. Then, should this table only filter for the vendor and device and not the revision? Do you know if there are any revisions for the 1002:15dd validated, tested and functional?

The pci revision id (RID) is used to specify the specific SKU within a family. GFXOFF is supposed to be working on all raven variants. It was tested and functional on all reference platforms and any OEM platforms that launched with Linux support. There are a lot of dependencies on sbios in the early raven variants (0x15dd), so it's likely more of a specific platform issue, but there is not a good way to detect this so we use the DID/SSID/RID as a proxy. The newer raven variants (0x15d8) have much better GFXOFF support since they all shipped with newer firmware and sbios.

Alex

Quoted 28 lines

Logs:

[   27.708348] [drm] initializing kernel modesetting (RAVEN
0x1002:0x15DD 0x1002:0x15DD 0x83).
[   27.789156] amdgpu: ATOM BIOS: 113-RAVEN-115

Thanks in advance,
Daniel

 drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +2 -1
index 65db88bb6cbc..319d4b99aec8 100644
@@ -1243,6 +1243,8 @@ static const struct amdgpu_gfxoff_quirk amdgpu_gfxoff_quirk_list[] = {
        { 0x1002, 0x15dd, 0x103c, 0x83e7, 0xd3 },
        /* GFXOFF is unstable on C6 parts with a VBIOS 113-RAVEN-114 */
        { 0x1002, 0x15dd, 0x1002, 0x15dd, 0xc6 },
+       /* GFXOFF provokes a hw lockup on 83 parts with a VBIOS 113-RAVEN-115 */
+       { 0x1002, 0x15dd, 0x1002, 0x15dd, 0x83 },
        { 0, 0, 0, 0, 0 },
 };
 
--

2.30.1

Quoted 4 lines

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#2 lore raw
Quoted 27 lines · Alex Deucher
Quoted 18 lines · Daniel Gomez

Disabling GFXOFF via the quirk list fixes a hardware lockup in Ryzen V1605B, RAVEN 0x1002:0x15DD rev 0x83.

---

This patch is a continuation of the work here:
https://lkml.org/lkml/2021/2/3/122 where a hardware lockup was discussed and a dma_fence deadlock was provoke as a side effect. To reproduce the issue please refer to the above link.

The hardware lockup was introduced in 5.6-rc1 for our particular revision as it wasn't part of the new blacklist. Before that, in kernel v5.5, this hardware was working fine without any hardware lock because the GFXOFF was actually disabled by the if condition for the CHIP_RAVEN case. So this patch, adds the 'Radeon Vega Mobile Series [1002:15dd] (rev 83)' to the blacklist to disable the GFXOFF.

Quoted 6 lines

But besides the fix, I'd like to ask from where this revision comes from. Is it an ASIC revision or is it hardcoded in the VBIOS from our vendor? From what I can see, it comes from the ASIC and I wonder if somehow we can get an APU in the future, 'not blacklisted', with the same problem. Then, should this table only filter for the vendor and device and not the revision? Do you know if there are any revisions for the 1002:15dd validated, tested and functional?

Quoted 9 lines

The pci revision id (RID) is used to specify the specific SKU within a family. GFXOFF is supposed to be working on all raven variants. It was tested and functional on all reference platforms and any OEM platforms that launched with Linux support. There are a lot of dependencies on sbios in the early raven variants (0x15dd), so it's likely more of a specific platform issue, but there is not a good way to detect this so we use the DID/SSID/RID as a proxy. The newer raven variants (0x15d8) have much better GFXOFF support since they all shipped with newer firmware and sbios.

We took one of the first reference platform boards to design our custom board based on the V1605B and I assume it has one of the early 'unstable' raven variants with RID 0x83. Also, as OEM we are in control of the bios (provided by insyde) but I wasn't sure about the RID so, thanks for the clarification. Is there anything we can do with the bios to have the GFXOFF enabled and 'stable' for this particular revision? Otherwise we'd need to add the 0x83 RID to the table. Also, there is an extra ']' in the patch subject. Sorry
for that. Would you need a new patch in case you accept it with the ']' removed?

Good to hear that the newer raven versions have better GFXOFF support.

Daniel

Quoted 29 lines

Alex

Quoted 25 lines

Logs:

[   27.708348] [drm] initializing kernel modesetting (RAVEN
0x1002:0x15DD 0x1002:0x15DD 0x83).
[   27.789156] amdgpu: ATOM BIOS: 113-RAVEN-115

Thanks in advance,
Daniel

 drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +2 -0
index 65db88bb6cbc..319d4b99aec8 100644
@@ -1243,6 +1243,8 @@ static const struct amdgpu_gfxoff_quirk amdgpu_gfxoff_quirk_list[] = {
        { 0x1002, 0x15dd, 0x103c, 0x83e7, 0xd3 },
        /* GFXOFF is unstable on C6 parts with a VBIOS 113-RAVEN-114 */
        { 0x1002, 0x15dd, 0x1002, 0x15dd, 0xc6 },
+       /* GFXOFF provokes a hw lockup on 83 parts with a VBIOS 113-RAVEN-115 */
+       { 0x1002, 0x15dd, 0x1002, 0x15dd, 0x83 },
        { 0, 0, 0, 0, 0 },
 };
 
Quoted 7 lines
Quoted 7 lines

--
2.30.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#3 lore raw
Quoted 49 lines · Daniel Gomez
Quoted 27 lines · Alex Deucher
Quoted 18 lines · Daniel Gomez

Disabling GFXOFF via the quirk list fixes a hardware lockup in Ryzen V1605B, RAVEN 0x1002:0x15DD rev 0x83.

---

This patch is a continuation of the work here:
https://lkml.org/lkml/2021/2/3/122 where a hardware lockup was discussed and a dma_fence deadlock was provoke as a side effect. To reproduce the issue please refer to the above link.

The hardware lockup was introduced in 5.6-rc1 for our particular revision as it wasn't part of the new blacklist. Before that, in kernel v5.5, this hardware was working fine without any hardware lock because the GFXOFF was actually disabled by the if condition for the CHIP_RAVEN case. So this patch, adds the 'Radeon Vega Mobile Series [1002:15dd] (rev 83)' to the blacklist to disable the GFXOFF.

Quoted 6 lines

But besides the fix, I'd like to ask from where this revision comes from. Is it an ASIC revision or is it hardcoded in the VBIOS from our vendor? From what I can see, it comes from the ASIC and I wonder if somehow we can get an APU in the future, 'not blacklisted', with the same problem. Then, should this table only filter for the vendor and device and not the revision? Do you know if there are any revisions for the 1002:15dd validated, tested and functional?

Quoted 9 lines

The pci revision id (RID) is used to specify the specific SKU within a family. GFXOFF is supposed to be working on all raven variants. It was tested and functional on all reference platforms and any OEM platforms that launched with Linux support. There are a lot of dependencies on sbios in the early raven variants (0x15dd), so it's likely more of a specific platform issue, but there is not a good way to detect this so we use the DID/SSID/RID as a proxy. The newer raven variants (0x15d8) have much better GFXOFF support since they all shipped with newer firmware and sbios.

We took one of the first reference platform boards to design our custom board based on the V1605B and I assume it has one of the early 'unstable' raven variants with RID 0x83. Also, as OEM we are in control of the bios (provided by insyde) but I wasn't sure about the RID so, thanks for the clarification. Is there anything we can do with the bios to have the GFXOFF enabled and 'stable' for this particular revision? Otherwise we'd need to add the 0x83 RID to the table. Also, there is an extra ']' in the patch subject. Sorry
for that. Would you need a new patch in case you accept it with the ']' removed?

Quoted 1 line

Good to hear that the newer raven versions have better GFXOFF support.

Adding Alex Desnoyer to the loop as he is the electronic/hardware and bios responsible so, he can
provide more information about this.

I've now done a test on the reference platform (dibbler) with the latest bios available
and the hw lockup can be also reproduced with the same steps.

For reference, I'm using mainline kernel 5.12-rc2.

[    5.938544] [drm] initializing kernel modesetting (RAVEN
0x1002:0x15DD 0x1002:0x15DD 0xC1).
[    5.939942] amdgpu: ATOM BIOS: 113-RAVEN-11

As in the previous cases, the clocks go to 100% of usage when the hang occurs.

However, when the gpu hangs, dmesg output displays the following:

[ 1568.279847] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx
timeout, signaled seq=188, emitted seq=191
[ 1568.434084] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* Process
information: process Xorg pid 311 thread Xorg:cs0 pid 312
[ 1568.279847] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx
timeout, signaled seq=188, emitted seq=191
[ 1568.434084] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* Process
information: process Xorg pid 311 thread Xorg:cs0 pid 312
[ 1568.507000] amdgpu 0000:01:00.0: amdgpu: GPU reset begin!
[ 1628.491882] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1628.491882] rcu:     3-...!: (665 ticks this GP)
idle=f9a/1/0x4000000000000000 softirq=188533/188533 fqs=15
[ 1628.491882] rcu: rcu_sched kthread timer wakeup didn't happen for
58497 jiffies! g726761 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402
[ 1628.491882] rcu:     Possible timer handling issue on cpu=2
timer-softirq=55225
[ 1628.491882] rcu: rcu_sched kthread starved for 58500 jiffies!
g726761 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402 ->cpu=2
[ 1628.491882] rcu:     Unless rcu_sched kthread gets sufficient CPU
time, OOM is now expected behavior.
[ 1628.491882] rcu: RCU grace-period kthread stack dump:
[ 1628.491882] rcu: Stack dump where RCU GP kthread last ran:
[ 1808.518445] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1808.518445] rcu:     3-...!: (2643 ticks this GP)
idle=f9a/1/0x4000000000000000 softirq=188533/188533 fqs=15
[ 1808.518445] rcu: rcu_sched kthread starved for 238526 jiffies!
g726761 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x0 ->cpu=2
[ 1808.518445] rcu:     Unless rcu_sched kthread gets sufficient CPU
time, OOM is now expected behavior.
[ 1808.518445] rcu: RCU grace-period kthread stack dump:
[ 1808.518445] rcu: Stack dump where RCU GP kthread last ran:
Quoted 32 lines

Daniel

Quoted 29 lines

Alex

Quoted 25 lines

Logs:

[   27.708348] [drm] initializing kernel modesetting (RAVEN
0x1002:0x15DD 0x1002:0x15DD 0x83).
[   27.789156] amdgpu: ATOM BIOS: 113-RAVEN-115

Thanks in advance,
Daniel

 drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +2 -0
index 65db88bb6cbc..319d4b99aec8 100644
@@ -1243,6 +1243,8 @@ static const struct amdgpu_gfxoff_quirk amdgpu_gfxoff_quirk_list[] = {
        { 0x1002, 0x15dd, 0x103c, 0x83e7, 0xd3 },
        /* GFXOFF is unstable on C6 parts with a VBIOS 113-RAVEN-114 */
        { 0x1002, 0x15dd, 0x1002, 0x15dd, 0xc6 },
+       /* GFXOFF provokes a hw lockup on 83 parts with a VBIOS 113-RAVEN-115 */
+       { 0x1002, 0x15dd, 0x1002, 0x15dd, 0x83 },
        { 0, 0, 0, 0, 0 },
 };
 
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

--
2.30.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#4 lore raw
Quoted 51 lines · Daniel Gomez
Quoted 49 lines · Daniel Gomez
Quoted 27 lines · Alex Deucher
Quoted 18 lines · Daniel Gomez

Disabling GFXOFF via the quirk list fixes a hardware lockup in Ryzen V1605B, RAVEN 0x1002:0x15DD rev 0x83.

---

This patch is a continuation of the work here:
https://lkml.org/lkml/2021/2/3/122 where a hardware lockup was discussed and a dma_fence deadlock was provoke as a side effect. To reproduce the issue please refer to the above link.

The hardware lockup was introduced in 5.6-rc1 for our particular revision as it wasn't part of the new blacklist. Before that, in kernel v5.5, this hardware was working fine without any hardware lock because the GFXOFF was actually disabled by the if condition for the CHIP_RAVEN case. So this patch, adds the 'Radeon Vega Mobile Series [1002:15dd] (rev 83)' to the blacklist to disable the GFXOFF.

Quoted 6 lines

But besides the fix, I'd like to ask from where this revision comes from. Is it an ASIC revision or is it hardcoded in the VBIOS from our vendor? From what I can see, it comes from the ASIC and I wonder if somehow we can get an APU in the future, 'not blacklisted', with the same problem. Then, should this table only filter for the vendor and device and not the revision? Do you know if there are any revisions for the 1002:15dd validated, tested and functional?

Quoted 9 lines

The pci revision id (RID) is used to specify the specific SKU within a family. GFXOFF is supposed to be working on all raven variants. It was tested and functional on all reference platforms and any OEM platforms that launched with Linux support. There are a lot of dependencies on sbios in the early raven variants (0x15dd), so it's likely more of a specific platform issue, but there is not a good way to detect this so we use the DID/SSID/RID as a proxy. The newer raven variants (0x15d8) have much better GFXOFF support since they all shipped with newer firmware and sbios.

We took one of the first reference platform boards to design our custom board based on the V1605B and I assume it has one of the early 'unstable' raven variants with RID 0x83. Also, as OEM we are in control of the bios (provided by insyde) but I wasn't sure about the RID so, thanks for the clarification. Is there anything we can do with the bios to have the GFXOFF enabled and 'stable' for this particular revision? Otherwise we'd need to add the 0x83 RID to the table. Also, there is an extra ']' in the patch subject. Sorry
for that. Would you need a new patch in case you accept it with the ']' removed?

Quoted 5 lines
Quoted 1 line

Good to hear that the newer raven versions have better GFXOFF support.

Adding Alex Desnoyer to the loop as he is the electronic/hardware and bios responsible so, he can
provide more information about this.

Hello everyone,

We, Qtechnology, are the OEM of the hardware platform where we originally discovered the bug. Our platform is based on the AMD Dibbler V-1000 reference design, with the latest Insyde BIOS release available for the (now unsupported) Dibbler platform. We have the Insyde BIOS source code internally, so we can make some modifications as needed.

The last test that Daniel and myself performed was on a standard Dibbler PCB rev.B1 motherboard (NOT our platform), and using the corresponding latest AMD released BIOS "RDB1109GA". As Daniel wrote, the hardware lockup can be reproduced on the Dibbler, even if it has a different RID that our V1605B APU.

We also have a Neousys Technology POC-515 embedded computer (V-1000, V1605B) in our office. The Neousys PC also uses Insyde BIOS. This computer is also locking-up in the test.
https://www.neousys-tech.com/en/product/application/rugged-embedded/poc-500-amd-ryzen-ultra-compact-embedded-computer

Digging into the BIOS source code, the only reference to GFXOFF is in the SMU and PSP firmware release notes, where some bug fixes have been mentioned for previous SMU/PSP releases. After a quick "git grep -i gfx | grep -i off", there seems to be no mention of GFXOFF in the Insyde UEFI (inluding AMD PI) code base. I would appreciate any information regarding BIOS modification needed to make the GFXOFF feature stable. As you (Alex Deucher) mentionned, it should be functional on all AMD Raven reference platforms.

Regards,

Alexandre Desnoyers

Quoted 86 lines

I've now done a test on the reference platform (dibbler) with the latest bios available
and the hw lockup can be also reproduced with the same steps.

For reference, I'm using mainline kernel 5.12-rc2.

[    5.938544] [drm] initializing kernel modesetting (RAVEN
0x1002:0x15DD 0x1002:0x15DD 0xC1).
[    5.939942] amdgpu: ATOM BIOS: 113-RAVEN-11

As in the previous cases, the clocks go to 100% of usage when the hang occurs.

However, when the gpu hangs, dmesg output displays the following:

[ 1568.279847] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx
timeout, signaled seq=188, emitted seq=191
[ 1568.434084] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* Process
information: process Xorg pid 311 thread Xorg:cs0 pid 312
[ 1568.279847] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx
timeout, signaled seq=188, emitted seq=191
[ 1568.434084] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* Process
information: process Xorg pid 311 thread Xorg:cs0 pid 312
[ 1568.507000] amdgpu 0000:01:00.0: amdgpu: GPU reset begin!
[ 1628.491882] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1628.491882] rcu:     3-...!: (665 ticks this GP)
idle=f9a/1/0x4000000000000000 softirq=188533/188533 fqs=15
[ 1628.491882] rcu: rcu_sched kthread timer wakeup didn't happen for
58497 jiffies! g726761 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402
[ 1628.491882] rcu:     Possible timer handling issue on cpu=2
timer-softirq=55225
[ 1628.491882] rcu: rcu_sched kthread starved for 58500 jiffies!
g726761 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402 ->cpu=2
[ 1628.491882] rcu:     Unless rcu_sched kthread gets sufficient CPU
time, OOM is now expected behavior.
[ 1628.491882] rcu: RCU grace-period kthread stack dump:
[ 1628.491882] rcu: Stack dump where RCU GP kthread last ran:
[ 1808.518445] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1808.518445] rcu:     3-...!: (2643 ticks this GP)
idle=f9a/1/0x4000000000000000 softirq=188533/188533 fqs=15
[ 1808.518445] rcu: rcu_sched kthread starved for 238526 jiffies!
g726761 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x0 ->cpu=2
[ 1808.518445] rcu:     Unless rcu_sched kthread gets sufficient CPU
time, OOM is now expected behavior.
[ 1808.518445] rcu: RCU grace-period kthread stack dump:
[ 1808.518445] rcu: Stack dump where RCU GP kthread last ran:
Quoted 39 lines

Daniel

Quoted 36 lines

Alex

Quoted 32 lines

Logs:

[   27.708348] [drm] initializing kernel modesetting (RAVEN
0x1002:0x15DD 0x1002:0x15DD 0x83).
[   27.789156] amdgpu: ATOM BIOS: 113-RAVEN-115

Thanks in advance,
Daniel

 drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +2 -1
index 65db88bb6cbc..319d4b99aec8 100644
@@ -1243,6 +1243,8 @@ static const struct amdgpu_gfxoff_quirk amdgpu_gfxoff_quirk_list[] = {
        { 0x1002, 0x15dd, 0x103c, 0x83e7, 0xd3 },
        /* GFXOFF is unstable on C6 parts with a VBIOS 113-RAVEN-114 */
        { 0x1002, 0x15dd, 0x1002, 0x15dd, 0xc6 },
+       /* GFXOFF provokes a hw lockup on 83 parts with a VBIOS 113-RAVEN-115 */
+       { 0x1002, 0x15dd, 0x1002, 0x15dd, 0x83 },
        { 0, 0, 0, 0, 0 },
 };
 
--

2.30.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#5 lore raw
Quoted 80 lines · Alexandre Desnoyers
Quoted 51 lines · Daniel Gomez
Quoted 49 lines · Daniel Gomez
Quoted 27 lines · Alex Deucher
Quoted 18 lines · Daniel Gomez

Disabling GFXOFF via the quirk list fixes a hardware lockup in Ryzen V1605B, RAVEN 0x1002:0x15DD rev 0x83.

---

This patch is a continuation of the work here:
https://lkml.org/lkml/2021/2/3/122 where a hardware lockup was discussed and a dma_fence deadlock was provoke as a side effect. To reproduce the issue please refer to the above link.

The hardware lockup was introduced in 5.6-rc1 for our particular revision as it wasn't part of the new blacklist. Before that, in kernel v5.5, this hardware was working fine without any hardware lock because the GFXOFF was actually disabled by the if condition for the CHIP_RAVEN case. So this patch, adds the 'Radeon Vega Mobile Series [1002:15dd] (rev 83)' to the blacklist to disable the GFXOFF.

Quoted 6 lines

But besides the fix, I'd like to ask from where this revision comes from. Is it an ASIC revision or is it hardcoded in the VBIOS from our vendor? From what I can see, it comes from the ASIC and I wonder if somehow we can get an APU in the future, 'not blacklisted', with the same problem. Then, should this table only filter for the vendor and device and not the revision? Do you know if there are any revisions for the 1002:15dd validated, tested and functional?

Quoted 9 lines

The pci revision id (RID) is used to specify the specific SKU within a family. GFXOFF is supposed to be working on all raven variants. It was tested and functional on all reference platforms and any OEM platforms that launched with Linux support. There are a lot of dependencies on sbios in the early raven variants (0x15dd), so it's likely more of a specific platform issue, but there is not a good way to detect this so we use the DID/SSID/RID as a proxy. The newer raven variants (0x15d8) have much better GFXOFF support since they all shipped with newer firmware and sbios.

We took one of the first reference platform boards to design our custom board based on the V1605B and I assume it has one of the early 'unstable' raven variants with RID 0x83. Also, as OEM we are in control of the bios (provided by insyde) but I wasn't sure about the RID so, thanks for the clarification. Is there anything we can do with the bios to have the GFXOFF enabled and 'stable' for this particular revision? Otherwise we'd need to add the 0x83 RID to the table. Also, there is an extra ']' in the patch subject. Sorry
for that. Would you need a new patch in case you accept it with the ']' removed?

Quoted 5 lines
Quoted 1 line

Good to hear that the newer raven versions have better GFXOFF support.

Adding Alex Desnoyer to the loop as he is the electronic/hardware and bios responsible so, he can
provide more information about this.

Hello everyone,

We, Qtechnology, are the OEM of the hardware platform where we originally discovered the bug. Our platform is based on the AMD Dibbler V-1000 reference design, with the latest Insyde BIOS release available for the (now unsupported) Dibbler platform. We have the Insyde BIOS source code internally, so we can make some modifications as needed.

The last test that Daniel and myself performed was on a standard Dibbler PCB rev.B1 motherboard (NOT our platform), and using the corresponding latest AMD released BIOS "RDB1109GA". As Daniel wrote, the hardware lockup can be reproduced on the Dibbler, even if it has a different RID that our V1605B APU.

We also have a Neousys Technology POC-515 embedded computer (V-1000, V1605B) in our office. The Neousys PC also uses Insyde BIOS. This computer is also locking-up in the test.
https://www.neousys-tech.com/en/product/application/rugged-embedded/poc-500-amd-ryzen-ultra-compact-embedded-computer

Quoted 9 lines

Digging into the BIOS source code, the only reference to GFXOFF is in the SMU and PSP firmware release notes, where some bug fixes have been mentioned for previous SMU/PSP releases. After a quick "git grep -i gfx | grep -i off", there seems to be no mention of GFXOFF in the Insyde UEFI (inluding AMD PI) code base. I would appreciate any information regarding BIOS modification needed to make the GFXOFF feature stable. As you (Alex Deucher) mentionned, it should be functional on all AMD Raven reference platforms.

It's handled by the firmwares carried by the sbios. I'm not sure what versions off hand. Probably want to make sure you have the latest ones. Do you have an AMD partner contact? It might be best to bring this up with them.

Regarding the issues you are seeing is this a general issue with all workloads that use the GFX shader cores? Or just specific workloads? If it's just compute workloads, you might try this patch. It may fix the issue for you.

Alex

Quoted 85 lines

Regards,

Alexandre Desnoyers

Quoted 79 lines

I've now done a test on the reference platform (dibbler) with the latest bios available
and the hw lockup can be also reproduced with the same steps.

For reference, I'm using mainline kernel 5.12-rc2.

[    5.938544] [drm] initializing kernel modesetting (RAVEN
0x1002:0x15DD 0x1002:0x15DD 0xC1).
[    5.939942] amdgpu: ATOM BIOS: 113-RAVEN-11

As in the previous cases, the clocks go to 100% of usage when the hang occurs.

However, when the gpu hangs, dmesg output displays the following:

[ 1568.279847] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx
timeout, signaled seq=188, emitted seq=191
[ 1568.434084] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* Process
information: process Xorg pid 311 thread Xorg:cs0 pid 312
[ 1568.279847] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx
timeout, signaled seq=188, emitted seq=191
[ 1568.434084] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* Process
information: process Xorg pid 311 thread Xorg:cs0 pid 312
[ 1568.507000] amdgpu 0000:01:00.0: amdgpu: GPU reset begin!
[ 1628.491882] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1628.491882] rcu:     3-...!: (665 ticks this GP)
idle=f9a/1/0x4000000000000000 softirq=188533/188533 fqs=15
[ 1628.491882] rcu: rcu_sched kthread timer wakeup didn't happen for
58497 jiffies! g726761 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402
[ 1628.491882] rcu:     Possible timer handling issue on cpu=2
timer-softirq=55225
[ 1628.491882] rcu: rcu_sched kthread starved for 58500 jiffies!
g726761 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402 ->cpu=2
[ 1628.491882] rcu:     Unless rcu_sched kthread gets sufficient CPU
time, OOM is now expected behavior.
[ 1628.491882] rcu: RCU grace-period kthread stack dump:
[ 1628.491882] rcu: Stack dump where RCU GP kthread last ran:
[ 1808.518445] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1808.518445] rcu:     3-...!: (2643 ticks this GP)
idle=f9a/1/0x4000000000000000 softirq=188533/188533 fqs=15
[ 1808.518445] rcu: rcu_sched kthread starved for 238526 jiffies!
g726761 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x0 ->cpu=2
[ 1808.518445] rcu:     Unless rcu_sched kthread gets sufficient CPU
time, OOM is now expected behavior.
[ 1808.518445] rcu: RCU grace-period kthread stack dump:
[ 1808.518445] rcu: Stack dump where RCU GP kthread last ran:
Quoted 32 lines

Daniel

Quoted 29 lines

Alex

Quoted 25 lines

Logs:

[   27.708348] [drm] initializing kernel modesetting (RAVEN
0x1002:0x15DD 0x1002:0x15DD 0x83).
[   27.789156] amdgpu: ATOM BIOS: 113-RAVEN-115

Thanks in advance,
Daniel

 drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +2 -0
index 65db88bb6cbc..319d4b99aec8 100644
@@ -1243,6 +1243,8 @@ static const struct amdgpu_gfxoff_quirk amdgpu_gfxoff_quirk_list[] = {
        { 0x1002, 0x15dd, 0x103c, 0x83e7, 0xd3 },
        /* GFXOFF is unstable on C6 parts with a VBIOS 113-RAVEN-114 */
        { 0x1002, 0x15dd, 0x1002, 0x15dd, 0xc6 },
+       /* GFXOFF provokes a hw lockup on 83 parts with a VBIOS 113-RAVEN-115 */
+       { 0x1002, 0x15dd, 0x1002, 0x15dd, 0x83 },
        { 0, 0, 0, 0, 0 },
 };
 
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

--
2.30.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

From 9fc1ec9a1adda2a333a4eca120a0f56199ea80de Mon Sep 17 00:00:00 2001
From: Alex Deucher <[email protected]>
Date: Thu, 11 Mar 2021 11:02:51 -0500
Subject: [PATCH] drm/amdgpu: disable gfxoff for ROCm workloads on Raven1

Causes hangs.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +3 -1
index fa27262b5c2a..894215ce346a 100644
@@ -642,7 +642,9 @@ void amdgpu_amdkfd_set_compute_idle(struct kgd_dev *kgd, bool idle)
 	/* Temp workaround to fix the soft hang observed in certain compute
 	 * applications if GFXOFF is enabled.
 	 */
-	if (adev->asic_type == CHIP_SIENNA_CICHLID) {
+	if ((adev->asic_type == CHIP_SIENNA_CICHLID) ||
+	    ((adev->asic_type == CHIP_RAVEN) &&
+	     (adev->apu_flags &AMD_APU_IS_RAVEN))) {
 		pr_debug("GFXOFF is %s\n", idle ? "enabled" : "disabled");
 		amdgpu_gfx_off_ctrl(adev, idle);
 	}
Signature · 1 line
2.29.2

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#6 lore raw
Quoted 92 lines · Alex Deucher
Quoted 80 lines · Alexandre Desnoyers
Quoted 51 lines · Daniel Gomez
Quoted 49 lines · Daniel Gomez
Quoted 27 lines · Alex Deucher
Quoted 18 lines · Daniel Gomez

Disabling GFXOFF via the quirk list fixes a hardware lockup in Ryzen V1605B, RAVEN 0x1002:0x15DD rev 0x83.

---

This patch is a continuation of the work here:
https://lkml.org/lkml/2021/2/3/122 where a hardware lockup was discussed and a dma_fence deadlock was provoke as a side effect. To reproduce the issue please refer to the above link.

The hardware lockup was introduced in 5.6-rc1 for our particular revision as it wasn't part of the new blacklist. Before that, in kernel v5.5, this hardware was working fine without any hardware lock because the GFXOFF was actually disabled by the if condition for the CHIP_RAVEN case. So this patch, adds the 'Radeon Vega Mobile Series [1002:15dd] (rev 83)' to the blacklist to disable the GFXOFF.

Quoted 6 lines

But besides the fix, I'd like to ask from where this revision comes from. Is it an ASIC revision or is it hardcoded in the VBIOS from our vendor? From what I can see, it comes from the ASIC and I wonder if somehow we can get an APU in the future, 'not blacklisted', with the same problem. Then, should this table only filter for the vendor and device and not the revision? Do you know if there are any revisions for the 1002:15dd validated, tested and functional?

Quoted 9 lines

The pci revision id (RID) is used to specify the specific SKU within a family. GFXOFF is supposed to be working on all raven variants. It was tested and functional on all reference platforms and any OEM platforms that launched with Linux support. There are a lot of dependencies on sbios in the early raven variants (0x15dd), so it's likely more of a specific platform issue, but there is not a good way to detect this so we use the DID/SSID/RID as a proxy. The newer raven variants (0x15d8) have much better GFXOFF support since they all shipped with newer firmware and sbios.

We took one of the first reference platform boards to design our custom board based on the V1605B and I assume it has one of the early 'unstable' raven variants with RID 0x83. Also, as OEM we are in control of the bios (provided by insyde) but I wasn't sure about the RID so, thanks for the clarification. Is there anything we can do with the bios to have the GFXOFF enabled and 'stable' for this particular revision? Otherwise we'd need to add the 0x83 RID to the table. Also, there is an extra ']' in the patch subject. Sorry
for that. Would you need a new patch in case you accept it with the ']' removed?

Quoted 5 lines
Quoted 1 line

Good to hear that the newer raven versions have better GFXOFF support.

Adding Alex Desnoyer to the loop as he is the electronic/hardware and bios responsible so, he can
provide more information about this.

Hello everyone,

We, Qtechnology, are the OEM of the hardware platform where we originally discovered the bug. Our platform is based on the AMD Dibbler V-1000 reference design, with the latest Insyde BIOS release available for the (now unsupported) Dibbler platform. We have the Insyde BIOS source code internally, so we can make some modifications as needed.

The last test that Daniel and myself performed was on a standard Dibbler PCB rev.B1 motherboard (NOT our platform), and using the corresponding latest AMD released BIOS "RDB1109GA". As Daniel wrote, the hardware lockup can be reproduced on the Dibbler, even if it has a different RID that our V1605B APU.

We also have a Neousys Technology POC-515 embedded computer (V-1000, V1605B) in our office. The Neousys PC also uses Insyde BIOS. This computer is also locking-up in the test.
https://www.neousys-tech.com/en/product/application/rugged-embedded/poc-500-amd-ryzen-ultra-compact-embedded-computer

Quoted 9 lines

Digging into the BIOS source code, the only reference to GFXOFF is in the SMU and PSP firmware release notes, where some bug fixes have been mentioned for previous SMU/PSP releases. After a quick "git grep -i gfx | grep -i off", there seems to be no mention of GFXOFF in the Insyde UEFI (inluding AMD PI) code base. I would appreciate any information regarding BIOS modification needed to make the GFXOFF feature stable. As you (Alex Deucher) mentionned, it should be functional on all AMD Raven reference platforms.

Quoted 4 lines

It's handled by the firmwares carried by the sbios. I'm not sure what versions off hand. Probably want to make sure you have the latest ones. Do you have an AMD partner contact? It might be best to bring this up with them.

I'm sure we were using the latest but let us double-check with our AMD partner and insyde just in case.

Quoted 5 lines

Regarding the issues you are seeing is this a general issue with all workloads that use the GFX shader cores? Or just specific workloads? If it's just compute workloads, you might try this patch. It may fix the issue for you.

Thanks Alex for the patch. I think it's kind of a general issue with all the workloads but the way we've been able to reproduce it was with the MatrixMultiplication test (from AMD) and clinfo. With the patch, I'm still able to reproduce the problem in our custom board. I'll check it tomorrow on the dribbler.

Quoted 89 lines

Alex

Quoted 85 lines

Regards,

Alexandre Desnoyers

Quoted 79 lines

I've now done a test on the reference platform (dibbler) with the latest bios available
and the hw lockup can be also reproduced with the same steps.

For reference, I'm using mainline kernel 5.12-rc2.

[    5.938544] [drm] initializing kernel modesetting (RAVEN
0x1002:0x15DD 0x1002:0x15DD 0xC1).
[    5.939942] amdgpu: ATOM BIOS: 113-RAVEN-11

As in the previous cases, the clocks go to 100% of usage when the hang occurs.

However, when the gpu hangs, dmesg output displays the following:

[ 1568.279847] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx
timeout, signaled seq=188, emitted seq=191
[ 1568.434084] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* Process
information: process Xorg pid 311 thread Xorg:cs0 pid 312
[ 1568.279847] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx
timeout, signaled seq=188, emitted seq=191
[ 1568.434084] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* Process
information: process Xorg pid 311 thread Xorg:cs0 pid 312
[ 1568.507000] amdgpu 0000:01:00.0: amdgpu: GPU reset begin!
[ 1628.491882] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1628.491882] rcu:     3-...!: (665 ticks this GP)
idle=f9a/1/0x4000000000000000 softirq=188533/188533 fqs=15
[ 1628.491882] rcu: rcu_sched kthread timer wakeup didn't happen for
58497 jiffies! g726761 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402
[ 1628.491882] rcu:     Possible timer handling issue on cpu=2
timer-softirq=55225
[ 1628.491882] rcu: rcu_sched kthread starved for 58500 jiffies!
g726761 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402 ->cpu=2
[ 1628.491882] rcu:     Unless rcu_sched kthread gets sufficient CPU
time, OOM is now expected behavior.
[ 1628.491882] rcu: RCU grace-period kthread stack dump:
[ 1628.491882] rcu: Stack dump where RCU GP kthread last ran:
[ 1808.518445] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1808.518445] rcu:     3-...!: (2643 ticks this GP)
idle=f9a/1/0x4000000000000000 softirq=188533/188533 fqs=15
[ 1808.518445] rcu: rcu_sched kthread starved for 238526 jiffies!
g726761 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x0 ->cpu=2
[ 1808.518445] rcu:     Unless rcu_sched kthread gets sufficient CPU
time, OOM is now expected behavior.
[ 1808.518445] rcu: RCU grace-period kthread stack dump:
[ 1808.518445] rcu: Stack dump where RCU GP kthread last ran:
Quoted 32 lines

Daniel

Quoted 29 lines

Alex

Quoted 25 lines

Logs:

[   27.708348] [drm] initializing kernel modesetting (RAVEN
0x1002:0x15DD 0x1002:0x15DD 0x83).
[   27.789156] amdgpu: ATOM BIOS: 113-RAVEN-115

Thanks in advance,
Daniel

 drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +2 -0
index 65db88bb6cbc..319d4b99aec8 100644
@@ -1243,6 +1243,8 @@ static const struct amdgpu_gfxoff_quirk amdgpu_gfxoff_quirk_list[] = {
        { 0x1002, 0x15dd, 0x103c, 0x83e7, 0xd3 },
        /* GFXOFF is unstable on C6 parts with a VBIOS 113-RAVEN-114 */
        { 0x1002, 0x15dd, 0x1002, 0x15dd, 0xc6 },
+       /* GFXOFF provokes a hw lockup on 83 parts with a VBIOS 113-RAVEN-115 */
+       { 0x1002, 0x15dd, 0x1002, 0x15dd, 0x83 },
        { 0, 0, 0, 0, 0 },
 };
 
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

--
2.30.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#7 lore raw
Quoted 133 lines · Daniel Gomez
Quoted 2 lines · Alex Deucher

On Thu, Mar 11, 2021 at 10:02 AM Alexandre Desnoyers <[email protected]>

Quoted 6 lines · wrote:
Quoted 6 lines · Daniel Gomez
Quoted 4 lines · Daniel Gomez
Quoted 2 lines

On Wed, 10 Mar 2021 at 18:06, Alex Deucher <[email protected]>

Quoted 2 lines · wrote:
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

On Wed, Mar 10, 2021 at 11:37 AM Daniel Gomez <[email protected]>

Quoted 9 lines · wrote:
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines

Disabling GFXOFF via the quirk list fixes a hardware lockup in Ryzen V1605B, RAVEN 0x1002:0x15DD rev 0x83.

---

This patch is a continuation of the work here:
https://lkml.org/lkml/2021/2/3/122 where a hardware lockup

was discussed and

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

a dma_fence deadlock was provoke as a side effect. To

reproduce the issue

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

please refer to the above link.

The hardware lockup was introduced in 5.6-rc1 for our

particular revision as it

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

wasn't part of the new blacklist. Before that, in kernel v5.5,

this hardware was

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

working fine without any hardware lock because the GFXOFF was

actually disabled

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

by the if condition for the CHIP_RAVEN case. So this patch,

adds the 'Radeon

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

Vega Mobile Series [1002:15dd] (rev 83)' to the blacklist to

disable the GFXOFF.

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

But besides the fix, I'd like to ask from where this revision

comes from. Is it

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

an ASIC revision or is it hardcoded in the VBIOS from our

vendor? From what I

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

can see, it comes from the ASIC and I wonder if somehow we can

get an APU in the

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

future, 'not blacklisted', with the same problem. Then, should

this table only

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

filter for the vendor and device and not the revision? Do you

know if there are

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

any revisions for the 1002:15dd validated, tested and

functional?

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

The pci revision id (RID) is used to specify the specific SKU

within a

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

family. GFXOFF is supposed to be working on all raven

variants. It

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

was tested and functional on all reference platforms and any OEM platforms that launched with Linux support. There are a lot of dependencies on sbios in the early raven variants (0x15dd), so

it's

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

likely more of a specific platform issue, but there is not a

good way

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

to detect this so we use the DID/SSID/RID as a proxy. The newer

raven

Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 2 lines

variants (0x15d8) have much better GFXOFF support since they all shipped with newer firmware and sbios.

We took one of the first reference platform boards to design our custom board based on the V1605B and I assume it has one of the

early 'unstable'

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

raven variants with RID 0x83. Also, as OEM we are in control of

the bios

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

(provided by insyde) but I wasn't sure about the RID so, thanks

for the

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

clarification. Is there anything we can do with the bios to have

the GFXOFF

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

enabled and 'stable' for this particular revision? Otherwise we'd

need to add

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

the 0x83 RID to the table. Also, there is an extra ']' in the patch subject. Sorry
for that. Would you need a new patch in case you accept it with

the ']' removed?

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

Good to hear that the newer raven versions have better GFXOFF

support.

Quoted 20 lines
Quoted 20 lines
Quoted 4 lines

Adding Alex Desnoyer to the loop as he is the electronic/hardware and bios responsible so, he can
provide more information about this.

Hello everyone,

We, Qtechnology, are the OEM of the hardware platform where we originally discovered the bug. Our platform is based on the AMD Dibbler V-1000 reference design, with the latest Insyde BIOS release available for the (now unsupported) Dibbler platform. We have the Insyde BIOS source code internally, so we can make some modifications as needed.

The last test that Daniel and myself performed was on a standard Dibbler PCB rev.B1 motherboard (NOT our platform), and using the corresponding latest AMD released BIOS "RDB1109GA". As Daniel wrote, the hardware lockup can be reproduced on the Dibbler, even if it has a different RID that our V1605B APU.

Quoted 4 lines
Quoted 4 lines

We also have a Neousys Technology POC-515 embedded computer (V-1000, V1605B) in our office. The Neousys PC also uses Insyde BIOS. This computer is also locking-up in the test.

https://www.neousys-tech.com/en/product/application/rugged-embedded/poc-500-amd-ryzen-ultra-compact-embedded-computer

Quoted 12 lines
Quoted 11 lines

Digging into the BIOS source code, the only reference to GFXOFF is in the SMU and PSP firmware release notes, where some bug fixes have been mentioned for previous SMU/PSP releases. After a quick "git grep -i gfx | grep -i off", there seems to be no mention of GFXOFF in the Insyde UEFI (inluding AMD PI) code base. I would appreciate any information regarding BIOS modification needed to make the GFXOFF feature stable. As you (Alex Deucher) mentionned, it should be functional on all AMD Raven reference platforms.

Quoted 4 lines

It's handled by the firmwares carried by the sbios. I'm not sure what versions off hand. Probably want to make sure you have the latest ones. Do you have an AMD partner contact? It might be best to bring this up with them.

I'm sure we were using the latest but let us double-check with our AMD partner and insyde just in case.

Quoted 5 lines

Regarding the issues you are seeing is this a general issue with all workloads that use the GFX shader cores? Or just specific workloads? If it's just compute workloads, you might try this patch. It may fix the issue for you.

Quoted 7 lines

Thanks Alex for the patch. I think it's kind of a general issue with all the
workloads but the way we've been able to reproduce it was with the MatrixMultiplication test (from AMD) and clinfo. With the patch, I'm still able to reproduce the problem in our custom board. I'll check it tomorrow on the dribbler.

Quick update on this:
I've tested the patch on the dribbler and the hang also occurs. What we are now trying to figure out if we actually have the latest bios available for the V1605B (we think we do) but this might take a while (already reported
to the AMD partner).
To add more info about our current bios:
Dibbler:

-  Version: RDB1109GA
-  AGESA version: RavenPi-FP5-AM4 1.1.0.9
-  VBIOS 0113
QT5222 (V1605B with Radeon Vega Gfx 8):
-  Version: 0.0.29
-  AGESA version: RavenPi-FP5-AM4 1.1.0.A
-  VBIOS 0115

Also, in the release notes for the latest amdgpu release available in the amd support page
says: "Disable the GFXOFF for OpenCL(ROCm 3.8) usage as below in the grub.
Workaround: add “amdgpu.ppfeaturemask=0xffff3fff”in the grub." which confirms the issue.
https://drivers.amd.com/relnotes/linux_driver_2020.40_release_notes.pdf

My final question would be if we can/should add the patch as there are other raven
devices already added with different revision 0x1002:0x15dd {0xc8,0xc6} or we
should wait forAMD-partner confirmation to know if the bios/raven is buggy or not.

Quoted 84 lines
Quoted 22 lines

Alex

Quoted 18 lines

Regards,

Alexandre Desnoyers

Quoted 12 lines

I've now done a test on the reference platform (dibbler) with the latest bios available
and the hw lockup can be also reproduced with the same steps.

For reference, I'm using mainline kernel 5.12-rc2.

[    5.938544] [drm] initializing kernel modesetting (RAVEN
0x1002:0x15DD 0x1002:0x15DD 0xC1).
[    5.939942] amdgpu: ATOM BIOS: 113-RAVEN-11

As in the previous cases, the clocks go to 100% of usage when the

hang occurs.

Quoted 48 lines
Quoted 48 lines
Quoted 48 lines

However, when the gpu hangs, dmesg output displays the following:

[ 1568.279847] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx
timeout, signaled seq=188, emitted seq=191
[ 1568.434084] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* Process
information: process Xorg pid 311 thread Xorg:cs0 pid 312
[ 1568.279847] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx
timeout, signaled seq=188, emitted seq=191
[ 1568.434084] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* Process
information: process Xorg pid 311 thread Xorg:cs0 pid 312
[ 1568.507000] amdgpu 0000:01:00.0: amdgpu: GPU reset begin!
[ 1628.491882] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1628.491882] rcu:     3-...!: (665 ticks this GP)
idle=f9a/1/0x4000000000000000 softirq=188533/188533 fqs=15
[ 1628.491882] rcu: rcu_sched kthread timer wakeup didn't happen for
58497 jiffies! g726761 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402
[ 1628.491882] rcu:     Possible timer handling issue on cpu=2
timer-softirq=55225
[ 1628.491882] rcu: rcu_sched kthread starved for 58500 jiffies!
g726761 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402 ->cpu=2
[ 1628.491882] rcu:     Unless rcu_sched kthread gets sufficient CPU
time, OOM is now expected behavior.
[ 1628.491882] rcu: RCU grace-period kthread stack dump:
[ 1628.491882] rcu: Stack dump where RCU GP kthread last ran:
[ 1808.518445] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1808.518445] rcu:     3-...!: (2643 ticks this GP)
idle=f9a/1/0x4000000000000000 softirq=188533/188533 fqs=15
[ 1808.518445] rcu: rcu_sched kthread starved for 238526 jiffies!
g726761 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x0 ->cpu=2
[ 1808.518445] rcu:     Unless rcu_sched kthread gets sufficient CPU
time, OOM is now expected behavior.
[ 1808.518445] rcu: RCU grace-period kthread stack dump:
[ 1808.518445] rcu: Stack dump where RCU GP kthread last ran:
Quoted 13 lines

Daniel

Quoted 10 lines

Alex

Quoted 6 lines

Logs:

[   27.708348] [drm] initializing kernel modesetting (RAVEN
0x1002:0x15DD 0x1002:0x15DD 0x83).
[   27.789156] amdgpu: ATOM BIOS: 113-RAVEN-115
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

Review

Quoted 7 lines

Thanks in advance,
Daniel

 drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 2 ++
 1 file changed, 2 insertions(+)
(quoted diff) +0 -0

b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c static const struct amdgpu_gfxoff_quirk

Quoted 4 lines

index 65db88bb6cbc..319d4b99aec8 100644

drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +0 -0
@@ -1243,6 +1243,8 @@ static const struct amdgpu_gfxoff_quirk

amdgpu_gfxoff_quirk_list[] = {

Quoted 18 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c static const struct amdgpu_gfxoff_quirk · line 1244

Quoted 2 lines
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +0 -0
        { 0x1002, 0x15dd, 0x103c, 0x83e7, 0xd3 },
        /* GFXOFF is unstable on C6 parts with a VBIOS

113-RAVEN-114 */

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c static const struct amdgpu_gfxoff_quirk · line 1246

Quoted 2 lines
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +1 -0
        { 0x1002, 0x15dd, 0x1002, 0x15dd, 0xc6 },
+       /* GFXOFF provokes a hw lockup on 83 parts with a

VBIOS 113-RAVEN-115 */

Quoted 11 lines
Quoted 11 lines
Quoted 11 lines
Quoted 11 lines
Quoted 11 lines
Quoted 11 lines
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +1 -1
+       { 0x1002, 0x15dd, 0x1002, 0x15dd, 0x83 },
        { 0, 0, 0, 0, 0 },
 };
 
--

2.30.1

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx