From 38d4e4638e85ae52bec2c33869b2131e24d49229 Mon Sep 17 00:00:00 2001 From: Jingwen Chen Date: Fri, 22 Oct 2021 11:30:01 +0800 Subject: drm/amd/amdgpu: fix bad job hw_fence use after free in advance tdr [Why] In advance tdr mode, the real bad job will be resubmitted twice, while in drm_sched_resubmit_jobs_ext, there's a dma_fence_put, so the bad job is put one more time than other jobs. [How] Adding dma_fence_get before resbumit job in amdgpu_device_recheck_guilty_jobs and put the fence for normal jobs Signed-off-by: Jingwen Chen Reviewed-by: Andrey Grodzovsky Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 6e40cc1bc6dc..7c3f6ee14853 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4850,6 +4850,9 @@ static void amdgpu_device_recheck_guilty_jobs( /* clear job's guilty and depend the folowing step to decide the real one */ drm_sched_reset_karma(s_job); + /* for the real bad job, it will be resubmitted twice, adding a dma_fence_get + * to make sure fence is balanced */ + dma_fence_get(s_job->s_fence->parent); drm_sched_resubmit_jobs_ext(&ring->sched, 1); ret = dma_fence_wait_timeout(s_job->s_fence->parent, false, ring->sched.timeout); @@ -4885,6 +4888,7 @@ retry: /* got the hw fence, signal finished fence */ atomic_dec(ring->sched.score); + dma_fence_put(s_job->s_fence->parent); dma_fence_get(&s_job->s_fence->finished); dma_fence_signal(&s_job->s_fence->finished); dma_fence_put(&s_job->s_fence->finished); -- cgit From 93cec184788b0cf3926bc1f7b47fed74ba87990c Mon Sep 17 00:00:00 2001 From: James Zhu Date: Tue, 2 Nov 2021 21:33:50 -0400 Subject: drm/amdgpu: remove duplicated kfd_resume_iommu Remove duplicated kfd_resume_iommu which already runs in mdgpu_amdkfd_device_init. Tested-By: Ken Moffat Reviewed-by: Alex Deucher Signed-off-by: James Zhu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 7c3f6ee14853..6c38dd9f41be 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2398,10 +2398,6 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev) if (!adev->gmc.xgmi.pending_reset) amdgpu_amdkfd_device_init(adev); - r = amdgpu_amdkfd_resume_iommu(adev); - if (r) - goto init_failed; - amdgpu_fru_get_product_info(adev); init_failed: -- cgit From 2d32ffd6e9e5f28fab3f52ea4044e3c14418cfb7 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Wed, 3 Nov 2021 23:17:50 -0400 Subject: drm/amdgpu: fix SI handling in amdgpu_device_asic_has_dc_support() Properly handle SI DC support when CONFIG_DRM_AMD_DC_SI is not set. Fixes: f7f12b25823c0d ("drm/amdgpu: default to true in amdgpu_device_asic_has_dc_support") Reviewed-by: Evan Quan Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 6c38dd9f41be..b45f4751adf8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3167,11 +3167,21 @@ bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type) { switch (asic_type) { #if defined(CONFIG_DRM_AMD_DC) -#if defined(CONFIG_DRM_AMD_DC_SI) case CHIP_TAHITI: case CHIP_PITCAIRN: case CHIP_VERDE: case CHIP_OLAND: + /* + * We have systems in the wild with these ASICs that require + * LVDS and VGA support which is not supported with DC. + * + * Fallback to the non-DC driver here by default so as not to + * cause regressions. + */ +#if defined(CONFIG_DRM_AMD_DC_SI) + return amdgpu_dc > 0; +#else + return false; #endif case CHIP_BONAIRE: case CHIP_KAVERI: -- cgit From 9f4f2c1a35248f56b2a9c1c004e0aaff3609b15d Mon Sep 17 00:00:00 2001 From: shaoyunl Date: Fri, 5 Nov 2021 12:34:14 -0400 Subject: drm/amd/amdgpu: fix the kfd pre_reset sequence in sriov The KFD pre_reset should be called before reset been executed, it will hold the lock to prevent other rocm process to sent the packlage to hiq during host execute the real reset on the HW Signed-off-by: shaoyunl Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index b45f4751adf8..5625f7736e37 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4293,8 +4293,6 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev, if (r) return r; - amdgpu_amdkfd_pre_reset(adev); - /* Resume IP prior to SMC */ r = amdgpu_device_ip_reinit_early_sriov(adev); if (r) @@ -5030,8 +5028,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, cancel_delayed_work_sync(&tmp_adev->delayed_init_work); - if (!amdgpu_sriov_vf(tmp_adev)) - amdgpu_amdkfd_pre_reset(tmp_adev); + amdgpu_amdkfd_pre_reset(tmp_adev); /* * Mark these ASICs to be reseted as untracked first -- cgit From 6ee27ee27ba8b2e725886951ba2d2d87f113bece Mon Sep 17 00:00:00 2001 From: Evan Quan Date: Fri, 5 Nov 2021 15:25:30 +0800 Subject: drm/amd/pm: avoid duplicate powergate/ungate setting Just bail out if the target IP block is already in the desired powergate/ungate state. This can avoid some duplicate settings which sometimes may cause unexpected issues. Link: https://lore.kernel.org/all/YV81vidWQLWvATMM@zn.tnic/ Bug: https://bugzilla.kernel.org/show_bug.cgi?id=214921 Bug: https://bugzilla.kernel.org/show_bug.cgi?id=215025 Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1789 Fixes: bf756fb833cb ("drm/amdgpu: add missing cleanups for Polaris12 UVD/VCE on suspend") Signed-off-by: Evan Quan Tested-by: Borislav Petkov Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 5625f7736e37..188accb71249 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3509,6 +3509,9 @@ int amdgpu_device_init(struct amdgpu_device *adev, adev->rmmio_size = pci_resource_len(adev->pdev, 2); } + for (i = 0; i < AMD_IP_BLOCK_TYPE_NUM; i++) + atomic_set(&adev->pm.pwr_state[i], POWER_STATE_UNKNOWN); + adev->rmmio = ioremap(adev->rmmio_base, adev->rmmio_size); if (adev->rmmio == NULL) { return -ENOMEM; -- cgit