linux.git/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c, branch v5.14 Linux kernel source tree drm/amdgpu: Cancel delayed work when GFXOFF is disabled 2021-08-20T17:35:42+00:00 Michel Dänzer mdaenzer@redhat.com 2021-08-17T08:23:25+00:00 32bc8f8373d2d6a681c96e4b25dca60d4d1c6016 schedule_delayed_work does not push back the work if it was already scheduled before, so amdgpu_device_delay_enable_gfx_off ran ~100 ms after the first time GFXOFF was disabled and re-enabled, even if GFXOFF was disabled and re-enabled again during those 100 ms. This resulted in frame drops / stutter with the upcoming mutter 41 release on Navi 14, due to constantly enabling GFXOFF in the HW and disabling it again (for getting the GPU clock counter). To fix this, call cancel_delayed_work_sync when the disable count transitions from 0 to 1, and only schedule the delayed work on the reverse transition, not if the disable count was already 0. This makes sure the delayed work doesn't run at unexpected times, and allows it to be lock-free. v2: * Use cancel_delayed_work_sync & mutex_trylock instead of mod_delayed_work. v3: * Make amdgpu_device_delay_enable_gfx_off lock-free (Christian König) v4: * Fix race condition between amdgpu_gfx_off_ctrl incrementing adev->gfx.gfx_off_req_count and amdgpu_device_delay_enable_gfx_off checking for it to be 0 (Evan Quan) Cc: stable@vger.kernel.org Reviewed-by: Evan Quan <evan.quan@amd.com> Reviewed-by: Lijo Lazar <lijo.lazar@amd.com> # v3 Acked-by: Christian König <christian.koenig@amd.com> # v3 Signed-off-by: Michel Dänzer <mdaenzer@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> 
schedule_delayed_work does not push back the work if it was already
scheduled before, so amdgpu_device_delay_enable_gfx_off ran ~100 ms
after the first time GFXOFF was disabled and re-enabled, even if GFXOFF
was disabled and re-enabled again during those 100 ms.

This resulted in frame drops / stutter with the upcoming mutter 41
release on Navi 14, due to constantly enabling GFXOFF in the HW and
disabling it again (for getting the GPU clock counter).

To fix this, call cancel_delayed_work_sync when the disable count
transitions from 0 to 1, and only schedule the delayed work on the
reverse transition, not if the disable count was already 0. This makes
sure the delayed work doesn't run at unexpected times, and allows it to
be lock-free.

v2:
* Use cancel_delayed_work_sync & mutex_trylock instead of
  mod_delayed_work.
v3:
* Make amdgpu_device_delay_enable_gfx_off lock-free (Christian König)
v4:
* Fix race condition between amdgpu_gfx_off_ctrl incrementing
  adev->gfx.gfx_off_req_count and amdgpu_device_delay_enable_gfx_off
  checking for it to be 0 (Evan Quan)

Cc: stable@vger.kernel.org
Reviewed-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com> # v3
Acked-by: Christian König <christian.koenig@amd.com> # v3
Signed-off-by: Michel Dänzer <mdaenzer@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amdgpu: Conditionally reset RAS counters on boot 2021-05-20T02:38:11+00:00 John Clements john.clements@amd.com 2021-05-17T08:36:26+00:00 8f6368a9c92645e72fdd55862aa09821cdb81b43 Only clear RAS error counters if perestent EDC harvesting is not supported Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Signed-off-by: John Clements <john.clements@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> 
Only clear RAS error counters if perestent EDC harvesting is not supported

Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amdgpu: split gfx callbacks into ras and non-ras ones 2021-04-09T20:51:22+00:00 Hawking Zhang Hawking.Zhang@amd.com 2021-03-19T08:59:09+00:00 719a9b332305b8c4b91805c4bedee27ce82ee916 gfx ras is only available in cerntain ip generations. Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com> Reviewed-by: Dennis Li <Dennis.Li@amd.com> Reviewed-by: John Clements <John.Clements@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> 
gfx ras is only available in cerntain ip generations.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Dennis Li <Dennis.Li@amd.com>
Reviewed-by: John Clements <John.Clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amd/pm: unify the interface for gfx state setting 2021-04-09T20:46:51+00:00 Evan Quan evan.quan@amd.com 2021-03-25T05:16:48+00:00 2d64d23e9596b1815fa1b536b3ac096afac10bd5 No need to have special handling for swSMU supported ASICs. Signed-off-by: Evan Quan <evan.quan@amd.com> Reviewed-by: Lijo Lazar <lijo.lazar@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> 
No need to have special handling for swSMU supported ASICs.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amdgpu: add the sched_score to amdgpu_ring_init 2021-04-09T20:44:56+00:00 Christian König christian.koenig@amd.com 2021-02-02T12:05:49+00:00 c107171b8d3241d872807c04917e7e8fb70e8b71 Allow separate ring to share the same scheduler score. No functional change. Signed-off-by: Christian König <christian.koenig@amd.com> Reviewed-and-Tested-by: Leo Liu <leo.liu@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> 
Allow separate ring to share the same scheduler score.

No functional change.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-and-Tested-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amdgpu: wrap kiq ring ops with kiq spinlock 2021-04-09T20:35:31+00:00 Nirmoy Das nirmoy.das@amd.com 2021-03-12T09:42:41+00:00 5a8cd98e6e335ad03493502b3479fcbadcb4889e KIQ ring is being operated by kfd as well as amdgpu. KFD is using kiq lock, we should the same from amdgpu side as well. Signed-off-by: Nirmoy Das <nirmoy.das@amd.com> Acked-by: Christian König <christian.koenig@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> 
KIQ ring is being operated by kfd as well as amdgpu.
KFD is using kiq lock, we should the same from amdgpu side
as well.

Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amdgpu: add codes to capture invalid hardware access when recovery 2021-04-09T20:34:53+00:00 Dennis Li Dennis.Li@amd.com 2021-03-10T09:20:45+00:00 56b53c0b5aa5de49747351b2ad323fd36089eb52 When recovery thread has begun GPU reset, there should be not other threads to access hardware, otherwise system randomly hang. v2 (chk): rewritten from scratch, use trylock and lockdep instead of hand wiring the logic. v3: add in_irq check v4: change to check in_task Signed-off-by: Dennis Li <Dennis.Li@amd.com> Signed-off-by: Christian König <christian.koenig@amd.com> Reviewed-by: Christian König <christian.koenig@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> 
When recovery thread has begun GPU reset, there should be not other
threads to access hardware, otherwise system randomly hang.

v2 (chk): rewritten from scratch, use trylock and lockdep instead of
hand wiring the logic.

v3: add in_irq check

v4: change to check in_task

Signed-off-by: Dennis Li <Dennis.Li@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amdgpu: harvest edc status when connected to host via xGMI 2021-03-24T03:00:41+00:00 Dennis Li Dennis.Li@amd.com 2021-02-04T05:32:05+00:00 761d86d37f86ebba77e59fa59ccef4dc2f38674f When connected to a host via xGMI, system fatal errors may trigger warm reset, driver has no change to query edc status before reset. Therefore in this case, driver should harvest previous error loging registers during boot, instead of only resetting them. v2: 1. IP's ras_manager object is created when its ras feature is enabled, so change to query edc status after amdgpu_ras_late_init called 2. change to enable watchdog timer after finishing gfx edc init Signed-off-by: Dennis Li <Dennis.Li@amd.com> Reivewed-by: Hawking Zhang <hawking.zhang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> 
When connected to a host via xGMI, system fatal errors may trigger
warm reset, driver has no change to query edc status before reset.
Therefore in this case, driver should harvest previous error loging
registers during boot, instead of only resetting them.

v2:
1. IP's ras_manager object is created when its ras feature is enabled,
so change to query edc status after amdgpu_ras_late_init called

2. change to enable watchdog timer after finishing gfx edc init

Signed-off-by: Dennis Li <Dennis.Li@amd.com>
Reivewed-by: Hawking Zhang <hawking.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amdgpu: enable only one high prio compute queue 2021-02-09T20:26:56+00:00 Nirmoy Das nirmoy.das@amd.com 2021-02-01T11:12:34+00:00 8c0225d79273968a65e73a4204fba023ae02714d For high priority compute to work properly we need to enable wave limiting on gfx pipe. Wave limiting is done through writing into mmSPI_WCL_PIPE_PERCENT_GFX register. Enable only one high priority compute queue to avoid race condition between multiple high priority compute queues writing that register simultaneously. Signed-off-by: Nirmoy Das <nirmoy.das@amd.com> Acked-by: Christian König <christian.koenig@amd.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> 
For high priority compute to work properly we need to enable
wave limiting on gfx pipe. Wave limiting is done through writing
into mmSPI_WCL_PIPE_PERCENT_GFX register. Enable only one high
priority compute queue to avoid race condition between multiple
high priority compute queues writing that register simultaneously.

Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amd/pm: add gfx_state_change_set() for rn gfx power switch (v2) 2020-11-13T22:29:45+00:00 Prike Liang Prike.Liang@amd.com 2020-10-20T07:58:30+00:00 8279bb4ec78568a34c4102e723df170b49fb7c33 The gfx_state_change_set() funtion can support set GFX power change status to D0/D3. v2: make sure to register callback (Alex) Signed-off-by: Prike Liang <Prike.Liang@amd.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Huang Rui <ray.huang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> 
The gfx_state_change_set() funtion can support set GFX power
change status to D0/D3.

v2: make sure to register callback (Alex)

Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>