From b7dfbd2e601f3fee545bc158feceba4f340fe7cf Mon Sep 17 00:00:00 2001 From: Tushar Patel Date: Thu, 17 Mar 2022 15:31:22 -0400 Subject: drm/amdkfd: Fix Incorrect VMIDs passed to HWS Compute-only GPUs have more than 8 VMIDs allocated to KFD. Fix this by passing correct number of VMIDs to HWS v2: squash in warning fix (Alex) Signed-off-by: Tushar Patel Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index bb1c025d9001..b03663f42cc9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -680,7 +680,7 @@ MODULE_PARM_DESC(sched_policy, * Maximum number of processes that HWS can schedule concurrently. The maximum is the * number of VMIDs assigned to the HWS, which is also the default. */ -int hws_max_conc_proc = 8; +int hws_max_conc_proc = -1; module_param(hws_max_conc_proc, int, 0444); MODULE_PARM_DESC(hws_max_conc_proc, "Max # processes HWS can execute concurrently when sched_policy=0 (0 = no concurrency, #VMIDs for KFD = Maximum(default))"); -- cgit v1.2.3 From 887f75cfd0da44c19dda93b2ff9e70ca8792cdc1 Mon Sep 17 00:00:00 2001 From: Kai-Heng Feng Date: Thu, 7 Apr 2022 20:12:28 +0800 Subject: drm/amdgpu: Ensure HDA function is suspended before ASIC reset DP/HDMI audio on AMD PRO VII stops working after S3: [ 149.450391] amdgpu 0000:63:00.0: amdgpu: MODE1 reset [ 149.450395] amdgpu 0000:63:00.0: amdgpu: GPU mode1 reset [ 149.450494] amdgpu 0000:63:00.0: amdgpu: GPU psp mode1 reset [ 149.983693] snd_hda_intel 0000:63:00.1: refused to change power state from D0 to D3hot [ 150.003439] amdgpu 0000:63:00.0: refused to change power state from D0 to D3hot ... [ 155.432975] snd_hda_intel 0000:63:00.1: CORB reset timeout#2, CORBRP = 65535 The offending commit is daf8de0874ab5b ("drm/amdgpu: always reset the asic in suspend (v2)"). Commit 34452ac3038a7 ("drm/amdgpu: don't use BACO for reset in S3 ") doesn't help, so the issue is something different. Assuming that to make HDA resume to D0 fully realized, it needs to be successfully put to D3 first. And this guesswork proves working, by moving amdgpu_asic_reset() to noirq callback, so it's called after HDA function is in D3. Fixes: daf8de0874ab5b ("drm/amdgpu: always reset the asic in suspend (v2)") Signed-off-by: Kai-Heng Feng Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 18 ++++++++++++------ 1 file changed, 12 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index b03663f42cc9..29e9419a914b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -2323,18 +2323,23 @@ static int amdgpu_pmops_suspend(struct device *dev) { struct drm_device *drm_dev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(drm_dev); - int r; if (amdgpu_acpi_is_s0ix_active(adev)) adev->in_s0ix = true; else adev->in_s3 = true; - r = amdgpu_device_suspend(drm_dev, true); - if (r) - return r; + return amdgpu_device_suspend(drm_dev, true); +} + +static int amdgpu_pmops_suspend_noirq(struct device *dev) +{ + struct drm_device *drm_dev = dev_get_drvdata(dev); + struct amdgpu_device *adev = drm_to_adev(drm_dev); + if (!adev->in_s0ix) - r = amdgpu_asic_reset(adev); - return r; + return amdgpu_asic_reset(adev); + + return 0; } static int amdgpu_pmops_resume(struct device *dev) @@ -2575,6 +2580,7 @@ static const struct dev_pm_ops amdgpu_pm_ops = { .prepare = amdgpu_pmops_prepare, .complete = amdgpu_pmops_complete, .suspend = amdgpu_pmops_suspend, + .suspend_noirq = amdgpu_pmops_suspend_noirq, .resume = amdgpu_pmops_resume, .freeze = amdgpu_pmops_freeze, .thaw = amdgpu_pmops_thaw, -- cgit v1.2.3 From f95af4a9236695caed24fe6401256bb974e8f2a7 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 28 Dec 2021 17:26:24 -0500 Subject: drm/amdgpu: don't runtime suspend if there are displays attached (v3) We normally runtime suspend when there are displays attached if they are in the DPMS off state, however, if something wakes the GPU we send a hotplug event on resume (in case any displays were connected while the GPU was in suspend) which can cause userspace to light up the displays again soon after they were turned off. Prior to commit 087451f372bf76 ("drm/amdgpu: use generic fb helpers instead of setting up AMD own's."), the driver took a runtime pm reference when the fbdev emulation was enabled because we didn't implement proper shadowing support for vram access when the device was off so the device never runtime suspended when there was a console bound. Once that commit landed, we now utilize the core fb helper implementation which properly handles the emulation, so runtime pm now suspends in cases where it did not before. Ultimately, we need to sort out why runtime suspend in not working in this case for some users, but this should restore similar behavior to before. v2: move check into runtime_suspend v3: wake ups -> wakeups in comment, retain pm_runtime behavior in runtime_idle callback Fixes: 087451f372bf76 ("drm/amdgpu: use generic fb helpers instead of setting up AMD own's.") Link: https://lore.kernel.org/r/20220403132322.51c90903@darkstar.example.org/ Tested-by: Michele Ballabio Reviewed-by: Evan Quan Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 105 +++++++++++++++++++++----------- 1 file changed, 70 insertions(+), 35 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 29e9419a914b..7fd0277b2805 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -2395,6 +2395,71 @@ static int amdgpu_pmops_restore(struct device *dev) return amdgpu_device_resume(drm_dev, true); } +static int amdgpu_runtime_idle_check_display(struct device *dev) +{ + struct pci_dev *pdev = to_pci_dev(dev); + struct drm_device *drm_dev = pci_get_drvdata(pdev); + struct amdgpu_device *adev = drm_to_adev(drm_dev); + + if (adev->mode_info.num_crtc) { + struct drm_connector *list_connector; + struct drm_connector_list_iter iter; + int ret = 0; + + /* XXX: Return busy if any displays are connected to avoid + * possible display wakeups after runtime resume due to + * hotplug events in case any displays were connected while + * the GPU was in suspend. Remove this once that is fixed. + */ + mutex_lock(&drm_dev->mode_config.mutex); + drm_connector_list_iter_begin(drm_dev, &iter); + drm_for_each_connector_iter(list_connector, &iter) { + if (list_connector->status == connector_status_connected) { + ret = -EBUSY; + break; + } + } + drm_connector_list_iter_end(&iter); + mutex_unlock(&drm_dev->mode_config.mutex); + + if (ret) + return ret; + + if (amdgpu_device_has_dc_support(adev)) { + struct drm_crtc *crtc; + + drm_for_each_crtc(crtc, drm_dev) { + drm_modeset_lock(&crtc->mutex, NULL); + if (crtc->state->active) + ret = -EBUSY; + drm_modeset_unlock(&crtc->mutex); + if (ret < 0) + break; + } + } else { + mutex_lock(&drm_dev->mode_config.mutex); + drm_modeset_lock(&drm_dev->mode_config.connection_mutex, NULL); + + drm_connector_list_iter_begin(drm_dev, &iter); + drm_for_each_connector_iter(list_connector, &iter) { + if (list_connector->dpms == DRM_MODE_DPMS_ON) { + ret = -EBUSY; + break; + } + } + + drm_connector_list_iter_end(&iter); + + drm_modeset_unlock(&drm_dev->mode_config.connection_mutex); + mutex_unlock(&drm_dev->mode_config.mutex); + } + if (ret) + return ret; + } + + return 0; +} + static int amdgpu_pmops_runtime_suspend(struct device *dev) { struct pci_dev *pdev = to_pci_dev(dev); @@ -2407,6 +2472,10 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev) return -EBUSY; } + ret = amdgpu_runtime_idle_check_display(dev); + if (ret) + return ret; + /* wait for all rings to drain before suspending */ for (i = 0; i < AMDGPU_MAX_RINGS; i++) { struct amdgpu_ring *ring = adev->rings[i]; @@ -2516,41 +2585,7 @@ static int amdgpu_pmops_runtime_idle(struct device *dev) return -EBUSY; } - if (amdgpu_device_has_dc_support(adev)) { - struct drm_crtc *crtc; - - drm_for_each_crtc(crtc, drm_dev) { - drm_modeset_lock(&crtc->mutex, NULL); - if (crtc->state->active) - ret = -EBUSY; - drm_modeset_unlock(&crtc->mutex); - if (ret < 0) - break; - } - - } else { - struct drm_connector *list_connector; - struct drm_connector_list_iter iter; - - mutex_lock(&drm_dev->mode_config.mutex); - drm_modeset_lock(&drm_dev->mode_config.connection_mutex, NULL); - - drm_connector_list_iter_begin(drm_dev, &iter); - drm_for_each_connector_iter(list_connector, &iter) { - if (list_connector->dpms == DRM_MODE_DPMS_ON) { - ret = -EBUSY; - break; - } - } - - drm_connector_list_iter_end(&iter); - - drm_modeset_unlock(&drm_dev->mode_config.connection_mutex); - mutex_unlock(&drm_dev->mode_config.mutex); - } - - if (ret == -EBUSY) - DRM_DEBUG_DRIVER("failing to power off - crtc active\n"); + ret = amdgpu_runtime_idle_check_display(dev); pm_runtime_mark_last_busy(dev); pm_runtime_autosuspend(dev); -- cgit v1.2.3 From 7123d39dc24dcd21ff23d75f46f926b15269b9da Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Tue, 17 May 2022 12:00:37 -0500 Subject: drm/amd: Don't reset dGPUs if the system is going to s2idle An A+A configuration on ASUS ROG Strix G513QY proves that the ASIC reset for handling aborted suspend can't work with s2idle. This functionality was introduced in commit daf8de0874ab5b ("drm/amdgpu: always reset the asic in suspend (v2)"). A few other commits have gone on top of the ASIC reset, but this still doesn't work on the A+A configuration in s2idle. Avoid doing the reset on dGPUs specifically when using s2idle. Fixes: daf8de0874ab5b ("drm/amdgpu: always reset the asic in suspend (v2)") Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2008 Reviewed-by: Alex Deucher Signed-off-by: Mario Limonciello Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 ++ drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c | 14 ++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 2 +- 3 files changed, 17 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index cdf0818088b3..7606e3b6361e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1342,9 +1342,11 @@ static inline int amdgpu_acpi_smart_shift_update(struct drm_device *dev, #if defined(CONFIG_ACPI) && defined(CONFIG_SUSPEND) bool amdgpu_acpi_is_s3_active(struct amdgpu_device *adev); +bool amdgpu_acpi_should_gpu_reset(struct amdgpu_device *adev); bool amdgpu_acpi_is_s0ix_active(struct amdgpu_device *adev); #else static inline bool amdgpu_acpi_is_s0ix_active(struct amdgpu_device *adev) { return false; } +static inline bool amdgpu_acpi_should_gpu_reset(struct amdgpu_device *adev) { return false; } static inline bool amdgpu_acpi_is_s3_active(struct amdgpu_device *adev) { return false; } #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c index 0e12315fa0cb..98ac53ee6bb5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c @@ -1045,6 +1045,20 @@ bool amdgpu_acpi_is_s3_active(struct amdgpu_device *adev) (pm_suspend_target_state == PM_SUSPEND_MEM); } +/** + * amdgpu_acpi_should_gpu_reset + * + * @adev: amdgpu_device_pointer + * + * returns true if should reset GPU, false if not + */ +bool amdgpu_acpi_should_gpu_reset(struct amdgpu_device *adev) +{ + if (adev->flags & AMD_IS_APU) + return false; + return pm_suspend_target_state != PM_SUSPEND_TO_IDLE; +} + /** * amdgpu_acpi_is_s0ix_active * diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 7fd0277b2805..46ef57b07c15 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -2336,7 +2336,7 @@ static int amdgpu_pmops_suspend_noirq(struct device *dev) struct drm_device *drm_dev = dev_get_drvdata(dev); struct amdgpu_device *adev = drm_to_adev(drm_dev); - if (!adev->in_s0ix) + if (amdgpu_acpi_should_gpu_reset(adev)) return amdgpu_asic_reset(adev); return 0; -- cgit v1.2.3