1 files changed, 301 insertions, 143 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 58df107e3beb..ab8f970b2849 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -32,6 +32,9 @@
 #include <linux/slab.h>
 #include <linux/iommu.h>
 #include <linux/pci.h>
+#include <linux/devcoredump.h>
+#include <generated/utsrelease.h>
+#include <linux/pci-p2pdma.h>
 
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_probe_helper.h>
@@ -1942,35 +1945,6 @@ static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev)
 	}
 
 	switch (adev->asic_type) {
-#ifdef CONFIG_DRM_AMDGPU_SI
-	case CHIP_VERDE:
-	case CHIP_TAHITI:
-	case CHIP_PITCAIRN:
-	case CHIP_OLAND:
-	case CHIP_HAINAN:
-#endif
-#ifdef CONFIG_DRM_AMDGPU_CIK
-	case CHIP_BONAIRE:
-	case CHIP_HAWAII:
-	case CHIP_KAVERI:
-	case CHIP_KABINI:
-	case CHIP_MULLINS:
-#endif
-	case CHIP_TOPAZ:
-	case CHIP_TONGA:
-	case CHIP_FIJI:
-	case CHIP_POLARIS10:
-	case CHIP_POLARIS11:
-	case CHIP_POLARIS12:
-	case CHIP_VEGAM:
-	case CHIP_CARRIZO:
-	case CHIP_STONEY:
-	case CHIP_VEGA20:
-	case CHIP_ALDEBARAN:
-	case CHIP_SIENNA_CICHLID:
-	case CHIP_NAVY_FLOUNDER:
-	case CHIP_DIMGREY_CAVEFISH:
-	case CHIP_BEIGE_GOBY:
 	default:
 		return 0;
 	case CHIP_VEGA10:
@@ -2391,8 +2365,16 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
 		}
 		adev->ip_blocks[i].status.sw = true;
 
-		/* need to do gmc hw init early so we can allocate gpu mem */
-		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) {
+		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON) {
+			/* need to do common hw init early so everything is set up for gmc */
+			r = adev->ip_blocks[i].version->funcs->hw_init((void *)adev);
+			if (r) {
+				DRM_ERROR("hw_init %d failed %d\n", i, r);
+				goto init_failed;
+			}
+			adev->ip_blocks[i].status.hw = true;
+		} else if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) {
+			/* need to do gmc hw init early so we can allocate gpu mem */
 			/* Try to reserve bad pages early */
 			if (amdgpu_sriov_vf(adev))
 				amdgpu_virt_exchange_data(adev);
@@ -2477,17 +2459,21 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
 	 */
 	if (adev->gmc.xgmi.num_physical_nodes > 1) {
 		if (amdgpu_xgmi_add_device(adev) == 0) {
-			struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev);
+			if (!amdgpu_sriov_vf(adev)) {
+				struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev);
 
-			if (!hive->reset_domain ||
-			    !amdgpu_reset_get_reset_domain(hive->reset_domain)) {
-				r = -ENOENT;
-				goto init_failed;
-			}
+				if (!hive->reset_domain ||
+				    !amdgpu_reset_get_reset_domain(hive->reset_domain)) {
+					r = -ENOENT;
+					amdgpu_put_xgmi_hive(hive);
+					goto init_failed;
+				}
 
-			/* Drop the early temporary reset domain we created for device */
-			amdgpu_reset_put_reset_domain(adev->reset_domain);
-			adev->reset_domain = hive->reset_domain;
+				/* Drop the early temporary reset domain we created for device */
+				amdgpu_reset_put_reset_domain(adev->reset_domain);
+				adev->reset_domain = hive->reset_domain;
+				amdgpu_put_xgmi_hive(hive);
+			}
 		}
 	}
 
@@ -3076,8 +3062,8 @@ static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev)
 	int i, r;
 
 	static enum amd_ip_block_type ip_order[] = {
-		AMD_IP_BLOCK_TYPE_GMC,
 		AMD_IP_BLOCK_TYPE_COMMON,
+		AMD_IP_BLOCK_TYPE_GMC,
 		AMD_IP_BLOCK_TYPE_PSP,
 		AMD_IP_BLOCK_TYPE_IH,
 	};
@@ -3168,7 +3154,8 @@ static int amdgpu_device_ip_resume_phase1(struct amdgpu_device *adev)
 			continue;
 		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON ||
 		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC ||
-		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH) {
+		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH ||
+		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP && amdgpu_sriov_vf(adev))) {
 
 			r = adev->ip_blocks[i].version->funcs->resume(adev);
 			if (r) {
@@ -3316,38 +3303,12 @@ bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type)
 	case CHIP_MULLINS:
 		/*
 		 * We have systems in the wild with these ASICs that require
-		 * LVDS and VGA support which is not supported with DC.
+		 * VGA support which is not supported with DC.
 		 *
 		 * Fallback to the non-DC driver here by default so as not to
 		 * cause regressions.
 		 */
 		return amdgpu_dc > 0;
-	case CHIP_HAWAII:
-	case CHIP_CARRIZO:
-	case CHIP_STONEY:
-	case CHIP_POLARIS10:
-	case CHIP_POLARIS11:
-	case CHIP_POLARIS12:
-	case CHIP_VEGAM:
-	case CHIP_TONGA:
-	case CHIP_FIJI:
-	case CHIP_VEGA10:
-	case CHIP_VEGA12:
-	case CHIP_VEGA20:
-#if defined(CONFIG_DRM_AMD_DC_DCN)
-	case CHIP_RAVEN:
-	case CHIP_NAVI10:
-	case CHIP_NAVI14:
-	case CHIP_NAVI12:
-	case CHIP_RENOIR:
-	case CHIP_CYAN_SKILLFISH:
-	case CHIP_SIENNA_CICHLID:
-	case CHIP_NAVY_FLOUNDER:
-	case CHIP_DIMGREY_CAVEFISH:
-	case CHIP_BEIGE_GOBY:
-	case CHIP_VANGOGH:
-	case CHIP_YELLOW_CARP:
-#endif
 	default:
 		return amdgpu_dc != 0;
 #else
@@ -3369,7 +3330,7 @@ bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type)
  */
 bool amdgpu_device_has_dc_support(struct amdgpu_device *adev)
 {
-	if (amdgpu_sriov_vf(adev) || 
+	if (amdgpu_sriov_vf(adev) ||
 	    adev->enable_virtual_display ||
 	    (adev->harvest_ip_mask & AMD_HARVEST_IP_DMU_MASK))
 		return false;
@@ -3551,6 +3512,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	adev->gmc.gart_size = 512 * 1024 * 1024;
 	adev->accel_working = false;
 	adev->num_rings = 0;
+	RCU_INIT_POINTER(adev->gang_submit, dma_fence_get_stub());
 	adev->mman.buffer_funcs = NULL;
 	adev->mman.buffer_funcs_ring = NULL;
 	adev->vm_manager.vm_pte_funcs = NULL;
@@ -3629,6 +3591,8 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	INIT_WORK(&adev->xgmi_reset_work, amdgpu_device_xgmi_reset_func);
 
 	adev->gfx.gfx_off_req_count = 1;
+	adev->gfx.gfx_off_residency = 0;
+	adev->gfx.gfx_off_entrycount = 0;
 	adev->pm.ac_power = power_supply_is_system_supplied() > 0;
 
 	atomic_set(&adev->throttling_logging_enabled, 1);
@@ -3667,14 +3631,6 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	if (amdgpu_mcbp)
 		DRM_INFO("MCBP is enabled\n");
 
-	if (adev->asic_type >= CHIP_NAVI10) {
-		if (amdgpu_mes || amdgpu_mes_kiq)
-			adev->enable_mes = true;
-
-		if (amdgpu_mes_kiq)
-			adev->enable_mes_kiq = true;
-	}
-
 	/*
 	 * Reset domain needs to be present early, before XGMI hive discovered
 	 * (if any) and intitialized to use reset sem and in_gpu reset flag
@@ -4025,8 +3981,7 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev)
 
 	amdgpu_gart_dummy_page_fini(adev);
 
-	if (drm_dev_is_unplugged(adev_to_drm(adev)))
-		amdgpu_device_unmap_mmio(adev);
+	amdgpu_device_unmap_mmio(adev);
 
 }
 
@@ -4039,6 +3994,7 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev)
 	release_firmware(adev->firmware.gpu_info_fw);
 	adev->firmware.gpu_info_fw = NULL;
 	adev->accel_working = false;
+	dma_fence_put(rcu_dereference_protected(adev->gang_submit, true));
 
 	amdgpu_reset_fini(adev);
 
@@ -4114,12 +4070,20 @@ static void amdgpu_device_evict_resources(struct amdgpu_device *adev)
 int amdgpu_device_suspend(struct drm_device *dev, bool fbcon)
 {
 	struct amdgpu_device *adev = drm_to_adev(dev);
+	int r = 0;
 
 	if (dev->switch_power_state == DRM_SWITCH_POWER_OFF)
 		return 0;
 
 	adev->in_suspend = true;
 
+	if (amdgpu_sriov_vf(adev)) {
+		amdgpu_virt_fini_data_exchange(adev);
+		r = amdgpu_virt_request_full_gpu(adev, false);
+		if (r)
+			return r;
+	}
+
 	if (amdgpu_acpi_smart_shift_update(dev, AMDGPU_SS_DEV_D3))
 		DRM_WARN("smart shift update failed\n");
 
@@ -4143,6 +4107,9 @@ int amdgpu_device_suspend(struct drm_device *dev, bool fbcon)
 
 	amdgpu_device_ip_suspend_phase2(adev);
 
+	if (amdgpu_sriov_vf(adev))
+		amdgpu_virt_release_full_gpu(adev, false);
+
 	return 0;
 }
 
@@ -4161,6 +4128,12 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon)
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	int r = 0;
 
+	if (amdgpu_sriov_vf(adev)) {
+		r = amdgpu_virt_request_full_gpu(adev, true);
+		if (r)
+			return r;
+	}
+
 	if (dev->switch_power_state == DRM_SWITCH_POWER_OFF)
 		return 0;
 
@@ -4175,6 +4148,13 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon)
 	}
 
 	r = amdgpu_device_ip_resume(adev);
+
+	/* no matter what r is, always need to properly release full GPU */
+	if (amdgpu_sriov_vf(adev)) {
+		amdgpu_virt_init_data_exchange(adev);
+		amdgpu_virt_release_full_gpu(adev, true);
+	}
+
 	if (r) {
 		dev_err(adev->dev, "amdgpu_device_ip_resume failed (%d).\n", r);
 		return r;
@@ -4473,8 +4453,6 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev,
 retry:
 	amdgpu_amdkfd_pre_reset(adev);
 
-	amdgpu_amdkfd_pre_reset(adev);
-
 	if (from_hypervisor)
 		r = amdgpu_virt_request_full_gpu(adev, true);
 	else
@@ -4569,14 +4547,15 @@ bool amdgpu_device_has_job_running(struct amdgpu_device *adev)
  */
 bool amdgpu_device_should_recover_gpu(struct amdgpu_device *adev)
 {
-	if (!amdgpu_device_ip_check_soft_reset(adev)) {
-		dev_info(adev->dev, "Timeout, but no hardware hang detected.\n");
-		return false;
-	}
 
 	if (amdgpu_gpu_recovery == 0)
 		goto disabled;
 
+	if (!amdgpu_device_ip_check_soft_reset(adev)) {
+		dev_info(adev->dev,"Timeout, but no hardware hang detected.\n");
+		return false;
+	}
+
 	if (amdgpu_sriov_vf(adev))
 		return true;
 
@@ -4666,6 +4645,8 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 		amdgpu_virt_fini_data_exchange(adev);
 	}
 
+	amdgpu_fence_driver_isr_toggle(adev, true);
+
 	/* block all schedulers and reset given job's ring */
 	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 		struct amdgpu_ring *ring = adev->rings[i];
@@ -4681,6 +4662,8 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 		amdgpu_fence_driver_force_completion(ring);
 	}
 
+	amdgpu_fence_driver_isr_toggle(adev, false);
+
 	if (job && job->vm)
 		drm_sched_increase_karma(&job->base);
 
@@ -4697,7 +4680,7 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 		if (!need_full_reset)
 			need_full_reset = amdgpu_device_ip_need_full_reset(adev);
 
-		if (!need_full_reset) {
+		if (!need_full_reset && amdgpu_gpu_recovery) {
 			amdgpu_device_ip_pre_soft_reset(adev);
 			r = amdgpu_device_ip_soft_reset(adev);
 			amdgpu_device_ip_post_soft_reset(adev);
@@ -4721,31 +4704,86 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 
 static int amdgpu_reset_reg_dumps(struct amdgpu_device *adev)
 {
-	uint32_t reg_value;
 	int i;
 
 	lockdep_assert_held(&adev->reset_domain->sem);
-	dump_stack();
 
 	for (i = 0; i < adev->num_regs; i++) {
-		reg_value = RREG32(adev->reset_dump_reg_list[i]);
-		trace_amdgpu_reset_reg_dumps(adev->reset_dump_reg_list[i], reg_value);
+		adev->reset_dump_reg_value[i] = RREG32(adev->reset_dump_reg_list[i]);
+		trace_amdgpu_reset_reg_dumps(adev->reset_dump_reg_list[i],
+					     adev->reset_dump_reg_value[i]);
 	}
 
 	return 0;
 }
 
+#ifdef CONFIG_DEV_COREDUMP
+static ssize_t amdgpu_devcoredump_read(char *buffer, loff_t offset,
+		size_t count, void *data, size_t datalen)
+{
+	struct drm_printer p;
+	struct amdgpu_device *adev = data;
+	struct drm_print_iterator iter;
+	int i;
+
+	iter.data = buffer;
+	iter.offset = 0;
+	iter.start = offset;
+	iter.remain = count;
+
+	p = drm_coredump_printer(&iter);
+
+	drm_printf(&p, "**** AMDGPU Device Coredump ****\n");
+	drm_printf(&p, "kernel: " UTS_RELEASE "\n");
+	drm_printf(&p, "module: " KBUILD_MODNAME "\n");
+	drm_printf(&p, "time: %lld.%09ld\n", adev->reset_time.tv_sec, adev->reset_time.tv_nsec);
+	if (adev->reset_task_info.pid)
+		drm_printf(&p, "process_name: %s PID: %d\n",
+			   adev->reset_task_info.process_name,
+			   adev->reset_task_info.pid);
+
+	if (adev->reset_vram_lost)
+		drm_printf(&p, "VRAM is lost due to GPU reset!\n");
+	if (adev->num_regs) {
+		drm_printf(&p, "AMDGPU register dumps:\nOffset:     Value:\n");
+
+		for (i = 0; i < adev->num_regs; i++)
+			drm_printf(&p, "0x%08x: 0x%08x\n",
+				   adev->reset_dump_reg_list[i],
+				   adev->reset_dump_reg_value[i]);
+	}
+
+	return count - iter.remain;
+}
+
+static void amdgpu_devcoredump_free(void *data)
+{
+}
+
+static void amdgpu_reset_capture_coredumpm(struct amdgpu_device *adev)
+{
+	struct drm_device *dev = adev_to_drm(adev);
+
+	ktime_get_ts64(&adev->reset_time);
+	dev_coredumpm(dev->dev, THIS_MODULE, adev, 0, GFP_KERNEL,
+		      amdgpu_devcoredump_read, amdgpu_devcoredump_free);
+}
+#endif
+
 int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 			 struct amdgpu_reset_context *reset_context)
 {
 	struct amdgpu_device *tmp_adev = NULL;
 	bool need_full_reset, skip_hw_reset, vram_lost = false;
 	int r = 0;
+	bool gpu_reset_for_dev_remove = 0;
 
 	/* Try reset handler method first */
 	tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device,
 				    reset_list);
 	amdgpu_reset_reg_dumps(tmp_adev);
+
+	reset_context->reset_device_list = device_list_handle;
 	r = amdgpu_reset_perform_reset(tmp_adev, reset_context);
 	/* If reset handler not implemented, continue; otherwise return */
 	if (r == -ENOSYS)
@@ -4758,6 +4796,10 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 		test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
 	skip_hw_reset = test_bit(AMDGPU_SKIP_HW_RESET, &reset_context->flags);
 
+	gpu_reset_for_dev_remove =
+		test_bit(AMDGPU_RESET_FOR_DEVICE_REMOVE, &reset_context->flags) &&
+			test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
+
 	/*
 	 * ASIC reset has to be done on all XGMI hive nodes ASAP
 	 * to allow proper links negotiation in FW (within 1 sec)
@@ -4802,6 +4844,18 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 		amdgpu_ras_intr_cleared();
 	}
 
+	/* Since the mode1 reset affects base ip blocks, the
+	 * phase1 ip blocks need to be resumed. Otherwise there
+	 * will be a BIOS signature error and the psp bootloader
+	 * can't load kdb on the next amdgpu install.
+	 */
+	if (gpu_reset_for_dev_remove) {
+		list_for_each_entry(tmp_adev, device_list_handle, reset_list)
+			amdgpu_device_ip_resume_phase1(tmp_adev);
+
+		goto end;
+	}
+
 	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
 		if (need_full_reset) {
 			/* post card */
@@ -4819,6 +4873,15 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 					goto out;
 
 				vram_lost = amdgpu_device_check_vram_lost(tmp_adev);
+#ifdef CONFIG_DEV_COREDUMP
+				tmp_adev->reset_vram_lost = vram_lost;
+				memset(&tmp_adev->reset_task_info, 0,
+						sizeof(tmp_adev->reset_task_info));
+				if (reset_context->job && reset_context->job->vm)
+					tmp_adev->reset_task_info =
+						reset_context->job->vm->task_info;
+				amdgpu_reset_capture_coredumpm(tmp_adev);
+#endif
 				if (vram_lost) {
 					DRM_INFO("VRAM is lost due to GPU reset!\n");
 					amdgpu_inc_vram_lost(tmp_adev);
@@ -5004,18 +5067,35 @@ 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);
 
+		if (!s_job->s_fence->parent) {
+			DRM_WARN("Failed to get a HW fence for job!");
+			continue;
+		}
+
 		ret = dma_fence_wait_timeout(s_job->s_fence->parent, false, ring->sched.timeout);
 		if (ret == 0) { /* timeout */
 			DRM_ERROR("Found the real bad job! ring:%s, job_id:%llx\n",
 						ring->sched.name, s_job->id);
 
+
+			amdgpu_fence_driver_isr_toggle(adev, true);
+
+			/* Clear this failed job from fence array */
+			amdgpu_fence_driver_clear_job_fences(ring);
+
+			amdgpu_fence_driver_isr_toggle(adev, false);
+
+			/* Since the job won't signal and we go for
+			 * another resubmit drop this parent pointer
+			 */
+			dma_fence_put(s_job->s_fence->parent);
+			s_job->s_fence->parent = NULL;
+
 			/* set guilty */
 			drm_sched_increase_karma(s_job);
+			amdgpu_reset_prepare_hwcontext(adev, reset_context);
 retry:
 			/* do hw reset */
 			if (amdgpu_sriov_vf(adev)) {
@@ -5042,7 +5122,6 @@ 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);
@@ -5055,8 +5134,29 @@ retry:
 	}
 }
 
+static inline void amdgpu_device_stop_pending_resets(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+
+#if defined(CONFIG_DEBUG_FS)
+	if (!amdgpu_sriov_vf(adev))
+		cancel_work(&adev->reset_work);
+#endif
+
+	if (adev->kfd.dev)
+		cancel_work(&adev->kfd.reset_work);
+
+	if (amdgpu_sriov_vf(adev))
+		cancel_work(&adev->virt.flr_work);
+
+	if (con && adev->ras_enabled)
+		cancel_work(&con->recovery_work);
+
+}
+
+
 /**
- * amdgpu_device_gpu_recover_imp - reset the asic and recover scheduler
+ * amdgpu_device_gpu_recover - reset the asic and recover scheduler
  *
  * @adev: amdgpu_device pointer
  * @job: which job trigger hang
@@ -5066,8 +5166,9 @@ retry:
  * Returns 0 for success or an error on failure.
  */
 
-int amdgpu_device_gpu_recover_imp(struct amdgpu_device *adev,
-			      struct amdgpu_job *job)
+int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
+			      struct amdgpu_job *job,
+			      struct amdgpu_reset_context *reset_context)
 {
 	struct list_head device_list, *device_list_handle =  NULL;
 	bool job_signaled = false;
@@ -5077,9 +5178,11 @@ int amdgpu_device_gpu_recover_imp(struct amdgpu_device *adev,
 	bool need_emergency_restart = false;
 	bool audio_suspended = false;
 	int tmp_vram_lost_counter;
-	struct amdgpu_reset_context reset_context;
+	bool gpu_reset_for_dev_remove = false;
 
-	memset(&reset_context, 0, sizeof(reset_context));
+	gpu_reset_for_dev_remove =
+			test_bit(AMDGPU_RESET_FOR_DEVICE_REMOVE, &reset_context->flags) &&
+				test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
 
 	/*
 	 * Special case: RAS triggered and full reset isn't supported
@@ -5105,11 +5208,8 @@ int amdgpu_device_gpu_recover_imp(struct amdgpu_device *adev,
 	if (hive)
 		mutex_lock(&hive->hive_lock);
 
-	reset_context.method = AMD_RESET_METHOD_NONE;
-	reset_context.reset_req_dev = adev;
-	reset_context.job = job;
-	reset_context.hive = hive;
-	clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
+	reset_context->job = job;
+	reset_context->hive = hive;
 
 	/*
 	 * Build list of devices to reset.
@@ -5118,8 +5218,11 @@ int amdgpu_device_gpu_recover_imp(struct amdgpu_device *adev,
 	 */
 	INIT_LIST_HEAD(&device_list);
 	if (!amdgpu_sriov_vf(adev) && (adev->gmc.xgmi.num_physical_nodes > 1)) {
-		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head)
+		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
 			list_add_tail(&tmp_adev->reset_list, &device_list);
+			if (gpu_reset_for_dev_remove && adev->shutdown)
+				tmp_adev->shutdown = true;
+		}
 		if (!list_is_first(&adev->reset_list, &device_list))
 			list_rotate_to_front(&adev->reset_list, &device_list);
 		device_list_handle = &device_list;
@@ -5194,8 +5297,7 @@ int amdgpu_device_gpu_recover_imp(struct amdgpu_device *adev,
 	 *
 	 * job->base holds a reference to parent fence
 	 */
-	if (job && job->base.s_fence->parent &&
-	    dma_fence_is_signaled(job->base.s_fence->parent)) {
+	if (job && dma_fence_is_signaled(&job->hw_fence)) {
 		job_signaled = true;
 		dev_info(adev->dev, "Guilty job already signaled, skipping HW reset");
 		goto skip_hw_reset;
@@ -5203,13 +5305,23 @@ int amdgpu_device_gpu_recover_imp(struct amdgpu_device *adev,
 
 retry:	/* Rest of adevs pre asic reset from XGMI hive. */
 	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
-		r = amdgpu_device_pre_asic_reset(tmp_adev, &reset_context);
+		if (gpu_reset_for_dev_remove) {
+			/* Workaroud for ASICs need to disable SMC first */
+			amdgpu_device_smu_fini_early(tmp_adev);
+		}
+		r = amdgpu_device_pre_asic_reset(tmp_adev, reset_context);
 		/*TODO Should we stop ?*/
 		if (r) {
 			dev_err(tmp_adev->dev, "GPU pre asic reset failed with err, %d for drm dev, %s ",
 				  r, adev_to_drm(tmp_adev)->unique);
 			tmp_adev->asic_reset_res = r;
 		}
+
+		/*
+		 * Drop all pending non scheduler resets. Scheduler resets
+		 * were already dropped during drm_sched_stop
+		 */
+		amdgpu_device_stop_pending_resets(tmp_adev);
 	}
 
 	tmp_vram_lost_counter = atomic_read(&((adev)->vram_lost_counter));
@@ -5224,9 +5336,15 @@ retry:	/* Rest of adevs pre asic reset from XGMI hive. */
 		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2))
 			amdgpu_ras_resume(adev);
 	} else {
-		r = amdgpu_do_asic_reset(device_list_handle, &reset_context);
-		if (r && r == -EAGAIN)
+		r = amdgpu_do_asic_reset(device_list_handle, reset_context);
+		if (r && r == -EAGAIN) {
+			set_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context->flags);
+			adev->asic_reset_res = 0;
 			goto retry;
+		}
+
+		if (!r && gpu_reset_for_dev_remove)
+			goto recover_end;
 	}
 
 skip_hw_reset:
@@ -5244,7 +5362,7 @@ skip_hw_reset:
 		if (amdgpu_gpu_recovery == 2 &&
 			!(tmp_vram_lost_counter < atomic_read(&adev->vram_lost_counter)))
 			amdgpu_device_recheck_guilty_jobs(
-				tmp_adev, device_list_handle, &reset_context);
+				tmp_adev, device_list_handle, reset_context);
 
 		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 			struct amdgpu_ring *ring = tmp_adev->rings[i];
@@ -5259,6 +5377,9 @@ skip_hw_reset:
 			drm_sched_start(&ring->sched, !tmp_adev->asic_reset_res);
 		}
 
+		if (adev->enable_mes)
+			amdgpu_mes_self_test(tmp_adev);
+
 		if (!drm_drv_uses_atomic_modeset(adev_to_drm(tmp_adev)) && !job_signaled) {
 			drm_helper_resume_force_mode(adev_to_drm(tmp_adev));
 		}
@@ -5297,6 +5418,7 @@ skip_sched_resume:
 		amdgpu_device_unset_mp1_state(tmp_adev);
 	}
 
+recover_end:
 	tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device,
 					    reset_list);
 	amdgpu_device_unlock_reset_domain(tmp_adev->reset_domain);
@@ -5308,38 +5430,9 @@ skip_sched_resume:
 
 	if (r)
 		dev_info(adev->dev, "GPU reset end with ret = %d\n", r);
-	return r;
-}
-
-struct amdgpu_recover_work_struct {
-	struct work_struct base;
-	struct amdgpu_device *adev;
-	struct amdgpu_job *job;
-	int ret;
-};
 
-static void amdgpu_device_queue_gpu_recover_work(struct work_struct *work)
-{
-	struct amdgpu_recover_work_struct *recover_work = container_of(work, struct amdgpu_recover_work_struct, base);
-
-	recover_work->ret = amdgpu_device_gpu_recover_imp(recover_work->adev, recover_work->job);
-}
-/*
- * Serialize gpu recover into reset domain single threaded wq
- */
-int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
-				    struct amdgpu_job *job)
-{
-	struct amdgpu_recover_work_struct work = {.adev = adev, .job = job};
-
-	INIT_WORK(&work.base, amdgpu_device_queue_gpu_recover_work);
-
-	if (!amdgpu_reset_domain_schedule(adev->reset_domain, &work.base))
-		return -EAGAIN;
-
-	flush_work(&work.base);
-
-	return work.ret;
+	atomic_set(&adev->reset_domain->reset_res, r);
+	return r;
 }
 
 /**
@@ -5490,6 +5583,37 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev)
 	}
 }
 
+/**
+ * amdgpu_device_is_peer_accessible - Check peer access through PCIe BAR
+ *
+ * @adev: amdgpu_device pointer
+ * @peer_adev: amdgpu_device pointer for peer device trying to access @adev
+ *
+ * Return true if @peer_adev can access (DMA) @adev through the PCIe
+ * BAR, i.e. @adev is "large BAR" and the BAR matches the DMA mask of
+ * @peer_adev.
+ */
+bool amdgpu_device_is_peer_accessible(struct amdgpu_device *adev,
+				      struct amdgpu_device *peer_adev)
+{
+#ifdef CONFIG_HSA_AMD_P2P
+	uint64_t address_mask = peer_adev->dev->dma_mask ?
+		~*peer_adev->dev->dma_mask : ~((1ULL << 32) - 1);
+	resource_size_t aper_limit =
+		adev->gmc.aper_base + adev->gmc.aper_size - 1;
+	bool p2p_access =
+		!adev->gmc.xgmi.connected_to_cpu &&
+		!(pci_p2pdma_distance(adev->pdev, peer_adev->dev, false) < 0);
+
+	return pcie_p2p && p2p_access && (adev->gmc.visible_vram_size &&
+		adev->gmc.real_vram_size == adev->gmc.visible_vram_size &&
+		!(adev->gmc.aper_base & address_mask ||
+		  aper_limit & address_mask));
+#else
+	return false;
+#endif
+}
+
 int amdgpu_device_baco_enter(struct drm_device *dev)
 {
 	struct amdgpu_device *adev = drm_to_adev(dev);
@@ -5653,6 +5777,7 @@ pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev)
 	reset_context.reset_req_dev = adev;
 	set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
 	set_bit(AMDGPU_SKIP_HW_RESET, &reset_context.flags);
+	set_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context.flags);
 
 	adev->no_hw_access = true;
 	r = amdgpu_device_pre_asic_reset(adev, &reset_context);
@@ -5862,3 +5987,36 @@ void amdgpu_device_pcie_port_wreg(struct amdgpu_device *adev,
 	(void)RREG32(data);
 	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
 }
+
+/**
+ * amdgpu_device_switch_gang - switch to a new gang
+ * @adev: amdgpu_device pointer
+ * @gang: the gang to switch to
+ *
+ * Try to switch to a new gang.
+ * Returns: NULL if we switched to the new gang or a reference to the current
+ * gang leader.
+ */
+struct dma_fence *amdgpu_device_switch_gang(struct amdgpu_device *adev,
+					    struct dma_fence *gang)
+{
+	struct dma_fence *old = NULL;
+
+	do {
+		dma_fence_put(old);
+		rcu_read_lock();
+		old = dma_fence_get_rcu_safe(&adev->gang_submit);
+		rcu_read_unlock();
+
+		if (old == gang)
+			break;
+
+		if (!dma_fence_is_signaled(old))
+			return old;
+
+	} while (cmpxchg((struct dma_fence __force **)&adev->gang_submit,
+			 old, gang) != old);
+
+	dma_fence_put(old);
+	return NULL;
+}