[PATCH 2/3] drm/amdgpu: enable mode1 reset

Zhang, Hawking Hawking.Zhang at amd.com
Mon Jul 13 05:34:02 UTC 2020


[AMD Public Use]

V2 looks good to me expect for the following that can be improved

+	if (need_emergency_restart && amdgpu_ras_get_context(adev)->reboot) {

The check of reboot could be centralized to the new function amdgpu_ras_need_emergency_restart you added in amdgpu_ras.c. And it would be better you can make amdgpu_ras_need_emergency_restart related as a separated patch so that it helps to isolate ras recovery related issue from general gpu reset issue. 

Regards,
Hawking
-----Original Message-----
From: Sheng, Wenhui <Wenhui.Sheng at amd.com> 
Sent: Friday, July 10, 2020 22:17
To: amd-gfx at lists.freedesktop.org
Cc: Zhang, Hawking <Hawking.Zhang at amd.com>; Sheng, Wenhui <Wenhui.Sheng at amd.com>; Gao, Likun <Likun.Gao at amd.com>
Subject: [PATCH 2/3] drm/amdgpu: enable mode1 reset

For sienna cichlid, add mode1 reset path for sGPU.
And fix some corner cases about mode1 mode reuse.

v2: hiding MP0/MP1 mode1 reset under AMD_RESET_METHOD_MODE1

Signed-off-by: Likun Gao <Likun.Gao at amd.com>
Signed-off-by: Wenhui Sheng <Wenhui.Sheng at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 23 +++++++++++-----------
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c    | 11 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h    |  1 +
 drivers/gpu/drm/amd/amdgpu/nv.c            | 19 +++++++++++-------
 drivers/gpu/drm/amd/powerplay/navi10_ppt.c |  2 +-
 5 files changed, 37 insertions(+), 19 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index fec6cf3f0c8a..6ce6e6bb8b50 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -4274,18 +4274,19 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 	struct amdgpu_hive_info *hive = NULL;
 	struct amdgpu_device *tmp_adev = NULL;
 	int i, r = 0;
-	bool in_ras_intr = amdgpu_ras_intr_triggered();
-	bool use_baco =
-		(amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) ?
-		true : false;
+	bool need_emergency_restart = false;
 	bool audio_suspended = false;
 
+	/**
+	 * Special case: RAS triggered and full reset isn't supported
+	 */
+	need_emergency_restart = amdgpu_ras_need_emergency_restart(adev);
+
 	/*
 	 * Flush RAM to disk so that after reboot
 	 * the user can read log and see why the system rebooted.
 	 */
-	if (in_ras_intr && !use_baco && amdgpu_ras_get_context(adev)->reboot) {
-
+	if (need_emergency_restart && amdgpu_ras_get_context(adev)->reboot) {
 		DRM_WARN("Emergency reboot.");
 
 		ksys_sync_helper();
@@ -4293,7 +4294,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 	}
 
 	dev_info(adev->dev, "GPU %s begin!\n",
-		(in_ras_intr && !use_baco) ? "jobs stop":"reset");
+		need_emergency_restart ? "jobs stop":"reset");
 
 	/*
 	 * Here we trylock to avoid chain of resets executing from @@ -4365,7 +4366,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 		amdgpu_fbdev_set_suspend(tmp_adev, 1);
 
 		/* disable ras on ALL IPs */
-		if (!(in_ras_intr && !use_baco) &&
+		if (!need_emergency_restart &&
 		      amdgpu_device_ip_need_full_reset(tmp_adev))
 			amdgpu_ras_suspend(tmp_adev);
 
@@ -4377,12 +4378,12 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 
 			drm_sched_stop(&ring->sched, job ? &job->base : NULL);
 
-			if (in_ras_intr && !use_baco)
+			if (need_emergency_restart)
 				amdgpu_job_stop_all_jobs_on_sched(&ring->sched);
 		}
 	}
 
-	if (in_ras_intr && !use_baco)
+	if (need_emergency_restart)
 		goto skip_sched_resume;
 
 	/*
@@ -4459,7 +4460,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 skip_sched_resume:
 	list_for_each_entry(tmp_adev, device_list_handle, gmc.xgmi.head) {
 		/*unlock kfd: SRIOV would do it separately */
-		if (!(in_ras_intr && !use_baco) && !amdgpu_sriov_vf(tmp_adev))
+		if (!need_emergency_restart && !amdgpu_sriov_vf(tmp_adev))
 	                amdgpu_amdkfd_post_reset(tmp_adev);
 		if (audio_suspended)
 			amdgpu_device_resume_display_audio(tmp_adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
index 3a3fa8567c94..6f06e1214622 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
@@ -2135,3 +2135,14 @@ void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev)
 		amdgpu_ras_reset_gpu(adev);
 	}
 }
+
+bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev) {
+	if (adev->asic_type == CHIP_VEGA20 &&
+	    adev->pm.fw_version <= 0x283400) {
+		return !(amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) &&
+				amdgpu_ras_intr_triggered();
+	}
+
+	return false;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
index e7df5d8429f8..b2667342cf67 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
@@ -633,4 +633,5 @@ void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev);
 
 void amdgpu_ras_set_error_query_ready(struct amdgpu_device *adev, bool ready);
 
+bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/nv.c b/drivers/gpu/drm/amd/amdgpu/nv.c index 356849136d1d..9f1240bd0310 100644
--- a/drivers/gpu/drm/amd/amdgpu/nv.c
+++ b/drivers/gpu/drm/amd/amdgpu/nv.c
@@ -265,17 +265,21 @@ static int nv_asic_mode1_reset(struct amdgpu_device *adev)
 
 	amdgpu_atombios_scratch_regs_engine_hung(adev, true);
 
-	dev_info(adev->dev, "GPU mode1 reset\n");
-
 	/* disable BM */
 	pci_clear_master(adev->pdev);
 
 	pci_save_state(adev->pdev);
 
-	ret = psp_gpu_reset(adev);
+	if (amdgpu_dpm_is_mode1_reset_supported(adev)) {
+		dev_info(adev->dev, "GPU smu mode1 reset\n");
+		ret = amdgpu_dpm_mode1_reset(adev);
+	} else {
+		dev_info(adev->dev, "GPU psp mode1 reset\n");
+		ret = psp_gpu_reset(adev);
+	}
+
 	if (ret)
 		dev_err(adev->dev, "GPU mode1 reset failed\n");
-
 	pci_restore_state(adev->pdev);
 
 	/* wait for asic to come out of reset */ @@ -307,7 +311,7 @@ nv_asic_reset_method(struct amdgpu_device *adev)  {
 	struct smu_context *smu = &adev->smu;
 
-	if (!amdgpu_sriov_vf(adev) && smu_baco_is_support(smu))
+	if (smu_baco_is_support(smu))
 		return AMD_RESET_METHOD_BACO;
 	else
 		return AMD_RESET_METHOD_MODE1;
@@ -319,15 +323,16 @@ static int nv_asic_reset(struct amdgpu_device *adev)
 	struct smu_context *smu = &adev->smu;
 
 	if (nv_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) {
+		dev_info(adev->dev, "GPU BACO reset\n");
+
 		ret = smu_baco_enter(smu);
 		if (ret)
 			return ret;
 		ret = smu_baco_exit(smu);
 		if (ret)
 			return ret;
-	} else {
+	} else
 		ret = nv_asic_mode1_reset(adev);
-	}
 
 	return ret;
 }
diff --git a/drivers/gpu/drm/amd/powerplay/navi10_ppt.c b/drivers/gpu/drm/amd/powerplay/navi10_ppt.c
index 4f7d064e16e4..014815bcae93 100644
--- a/drivers/gpu/drm/amd/powerplay/navi10_ppt.c
+++ b/drivers/gpu/drm/amd/powerplay/navi10_ppt.c
@@ -2039,7 +2039,7 @@ static bool navi10_is_baco_supported(struct smu_context *smu)
 	struct amdgpu_device *adev = smu->adev;
 	uint32_t val;
 
-	if (!smu_v11_0_baco_is_support(smu))
+	if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
 		return false;
 
 	val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0);
--
2.17.1


More information about the amd-gfx mailing list