tjh.dev / kernel
Linux kernel mirror (for testing) git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel os linux
1
fork

Configure Feed

Select the types of activity you want to include in your feed.

drm/amd: Check that VPE has reached DPM0 in idle handler

[Why]
Newer VPE microcode has functionality that will decrease DPM level
only when a workload has run for 2 or more seconds. If VPE is turned
off before this DPM decrease and the PMFW doesn't reset it when
power gating VPE, the SOC can get stuck with a higher DPM level.

This can happen from amdgpu's ring buffer test because it's a short
quick workload for VPE and VPE is turned off after 1s.

[How]
In idle handler besides checking fences are drained check PMFW version
to determine if it will reset DPM when power gating VPE. If PMFW will
not do this, then check VPE DPM level. If it is not DPM0 reschedule
delayed work again until it is.

v2: squash in return fix (Alex)

Cc: Peyton.Lee@amd.com
Reported-by: Sultan Alsawaf <sultan@kerneltoast.com>
Reviewed-by: Sultan Alsawaf <sultan@kerneltoast.com>
Tested-by: Sultan Alsawaf <sultan@kerneltoast.com>
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4615
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 3ac635367eb589bee8edcc722f812a89970e14b7)
Cc: stable@vger.kernel.org

authored by

Mario Limonciello and committed by
Alex Deucher ba10f8d9 dcb6fa37

+30 -4
+30 -4
drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c
··· 322 322 return 0; 323 323 } 324 324 325 + static bool vpe_need_dpm0_at_power_down(struct amdgpu_device *adev) 326 + { 327 + switch (amdgpu_ip_version(adev, VPE_HWIP, 0)) { 328 + case IP_VERSION(6, 1, 1): 329 + return adev->pm.fw_version < 0x0a640500; 330 + default: 331 + return false; 332 + } 333 + } 334 + 335 + static int vpe_get_dpm_level(struct amdgpu_device *adev) 336 + { 337 + struct amdgpu_vpe *vpe = &adev->vpe; 338 + 339 + if (!adev->pm.dpm_enabled) 340 + return 0; 341 + 342 + return RREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_request_lv)); 343 + } 344 + 325 345 static void vpe_idle_work_handler(struct work_struct *work) 326 346 { 327 347 struct amdgpu_device *adev = ··· 349 329 unsigned int fences = 0; 350 330 351 331 fences += amdgpu_fence_count_emitted(&adev->vpe.ring); 332 + if (fences) 333 + goto reschedule; 352 334 353 - if (fences == 0) 354 - amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE, AMD_PG_STATE_GATE); 355 - else 356 - schedule_delayed_work(&adev->vpe.idle_work, VPE_IDLE_TIMEOUT); 335 + if (vpe_need_dpm0_at_power_down(adev) && vpe_get_dpm_level(adev) != 0) 336 + goto reschedule; 337 + 338 + amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE, AMD_PG_STATE_GATE); 339 + return; 340 + 341 + reschedule: 342 + schedule_delayed_work(&adev->vpe.idle_work, VPE_IDLE_TIMEOUT); 357 343 } 358 344 359 345 static int vpe_common_init(struct amdgpu_vpe *vpe)