Merge drm/drm-next into drm-misc-next

Backmerging to get v6.9-rc2 changes into drm-misc-next.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>

57 files changed, 1454 insertions, 681 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile
index fa26a4e3a99d..4536c8ad0e11 100644
--- a/drivers/gpu/drm/amd/amdgpu/Makefile
+++ b/drivers/gpu/drm/amd/amdgpu/Makefile
@@ -98,7 +98,7 @@ amdgpu-y += \
 	vega20_reg_init.o nbio_v7_4.o nbio_v2_3.o nv.o arct_reg_init.o mxgpu_nv.o \
 	nbio_v7_2.o hdp_v4_0.o hdp_v5_0.o aldebaran_reg_init.o aldebaran.o soc21.o \
 	sienna_cichlid.o smu_v13_0_10.o nbio_v4_3.o hdp_v6_0.o nbio_v7_7.o hdp_v5_2.o lsdma_v6_0.o \
-	nbio_v7_9.o aqua_vanjaram.o nbio_v7_11.o lsdma_v7_0.o hdp_v7_0.o
+	nbio_v7_9.o aqua_vanjaram.o nbio_v7_11.o lsdma_v7_0.o hdp_v7_0.o nbif_v6_3_1.o
 
 # add DF block
 amdgpu-y += \
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index f5f2945711be..35dd6effa9a3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -146,7 +146,7 @@ int amdgpu_amdkfd_drm_client_create(struct amdgpu_device *adev)
 {
 	int ret;
 
-	if (!adev->kfd.init_complete)
+	if (!adev->kfd.init_complete || adev->kfd.client.dev)
 		return 0;
 
 	ret = drm_client_init(&adev->ddev, &adev->kfd.client, "kfd",
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index 14dc9d2d8d53..df58a6a1a67e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -2869,14 +2869,16 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence __rcu *
 
 	mutex_lock(&process_info->lock);
 
-	drm_exec_init(&exec, 0, 0);
+	drm_exec_init(&exec, DRM_EXEC_IGNORE_DUPLICATES, 0);
 	drm_exec_until_all_locked(&exec) {
 		list_for_each_entry(peer_vm, &process_info->vm_list_head,
 				    vm_list_node) {
 			ret = amdgpu_vm_lock_pd(peer_vm, &exec, 2);
 			drm_exec_retry_on_contention(&exec);
-			if (unlikely(ret))
+			if (unlikely(ret)) {
+				pr_err("Locking VM PD failed, ret: %d\n", ret);
 				goto ttm_reserve_fail;
+			}
 		}
 
 		/* Reserve all BOs and page tables/directory. Add all BOs from
@@ -2889,8 +2891,10 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence __rcu *
 			gobj = &mem->bo->tbo.base;
 			ret = drm_exec_prepare_obj(&exec, gobj, 1);
 			drm_exec_retry_on_contention(&exec);
-			if (unlikely(ret))
+			if (unlikely(ret)) {
+				pr_err("drm_exec_prepare_obj failed, ret: %d\n", ret);
 				goto ttm_reserve_fail;
+			}
 		}
 	}
 
@@ -2950,8 +2954,10 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence __rcu *
 	 * validations above would invalidate DMABuf imports again.
 	 */
 	ret = process_validate_vms(process_info, &exec.ticket);
-	if (ret)
+	if (ret) {
+		pr_debug("Validating VMs failed, ret: %d\n", ret);
 		goto validate_map_fail;
+	}
 
 	/* Update mappings not managed by KFD */
 	list_for_each_entry(peer_vm, &process_info->vm_list_head,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
index 1afbb2e932c6..f5d0fa207a88 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
@@ -1782,9 +1782,14 @@ static int amdgpu_debugfs_vm_info_show(struct seq_file *m, void *unused)
 	list_for_each_entry(file, &dev->filelist, lhead) {
 		struct amdgpu_fpriv *fpriv = file->driver_priv;
 		struct amdgpu_vm *vm = &fpriv->vm;
+		struct amdgpu_task_info *ti;
+
+		ti = amdgpu_vm_get_task_info_vm(vm);
+		if (ti) {
+			seq_printf(m, "pid:%d\tProcess:%s ----------\n", ti->pid, ti->process_name);
+			amdgpu_vm_put_task_info(ti);
+		}
 
-		seq_printf(m, "pid:%d\tProcess:%s ----------\n",
-				vm->task_info.pid, vm->task_info.process_name);
 		r = amdgpu_bo_reserve(vm->root.bo, true);
 		if (r)
 			break;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index e68bd6f8a6a4..aa16d51dd842 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -4040,10 +4040,8 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	 * early on during init and before calling to RREG32.
 	 */
 	adev->reset_domain = amdgpu_reset_create_reset_domain(SINGLE_DEVICE, "amdgpu-reset-dev");
-	if (!adev->reset_domain) {
-		r = -ENOMEM;
-		goto unmap_memory;
-	}
+	if (!adev->reset_domain)
+		return -ENOMEM;
 
 	/* detect hw virtualization here */
 	amdgpu_detect_virtualization(adev);
@@ -4053,20 +4051,20 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	r = amdgpu_device_get_job_timeout_settings(adev);
 	if (r) {
 		dev_err(adev->dev, "invalid lockup_timeout parameter syntax\n");
-		goto unmap_memory;
+		return r;
 	}
 
+	amdgpu_device_set_mcbp(adev);
+
 	/* early init functions */
 	r = amdgpu_device_ip_early_init(adev);
 	if (r)
-		goto unmap_memory;
-
-	amdgpu_device_set_mcbp(adev);
+		return r;
 
 	/* Get rid of things like offb */
 	r = drm_aperture_remove_conflicting_pci_framebuffers(adev->pdev, &amdgpu_kms_driver);
 	if (r)
-		goto unmap_memory;
+		return r;
 
 	/* Enable TMZ based on IP_VERSION */
 	amdgpu_gmc_tmz_set(adev);
@@ -4076,7 +4074,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	if (adev->gmc.xgmi.supported) {
 		r = adev->gfxhub.funcs->get_xgmi_info(adev);
 		if (r)
-			goto unmap_memory;
+			return r;
 	}
 
 	/* enable PCIE atomic ops */
@@ -4345,8 +4343,6 @@ release_ras_con:
 failed:
 	amdgpu_vf_error_trans_all(adev);
 
-unmap_memory:
-	iounmap(adev->rmmio);
 	return r;
 }
 
@@ -4543,6 +4539,8 @@ int amdgpu_device_prepare(struct drm_device *dev)
 	if (r)
 		goto unprepare;
 
+	flush_delayed_work(&adev->gfx.gfx_off_delay_work);
+
 	for (i = 0; i < adev->num_ip_blocks; i++) {
 		if (!adev->ip_blocks[i].status.valid)
 			continue;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c
index 502333725b49..fdd36fb027ab 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c
@@ -61,6 +61,7 @@
 #include "nbio_v4_3.h"
 #include "nbio_v7_2.h"
 #include "nbio_v7_7.h"
+#include "nbif_v6_3_1.h"
 #include "hdp_v5_0.h"
 #include "hdp_v5_2.h"
 #include "hdp_v6_0.h"
@@ -1319,6 +1320,15 @@ static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev)
 				}
 			}
 
+			if (le16_to_cpu(ip->hw_id) == VPE_HWID) {
+				if (adev->vpe.num_instances < AMDGPU_MAX_VPE_INSTANCES)
+					adev->vpe.num_instances++;
+				else
+					dev_err(adev->dev, "Too many VPE instances: %d vs %d\n",
+						adev->vpe.num_instances + 1,
+						AMDGPU_MAX_VPE_INSTANCES);
+			}
+
 			if (le16_to_cpu(ip->hw_id) == UMC_HWID) {
 				adev->gmc.num_umc++;
 				adev->umc.node_inst_num++;
@@ -1936,6 +1946,7 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev)
 		case IP_VERSION(3, 2, 0):
 		case IP_VERSION(3, 2, 1):
 		case IP_VERSION(3, 5, 0):
+		case IP_VERSION(3, 5, 1):
 			if (amdgpu_sriov_vf(adev))
 				amdgpu_discovery_set_sriov_display(adev);
 			else
@@ -2212,6 +2223,7 @@ static int amdgpu_discovery_set_vpe_ip_blocks(struct amdgpu_device *adev)
 {
 	switch (amdgpu_ip_version(adev, VPE_HWIP, 0)) {
 	case IP_VERSION(6, 1, 0):
+	case IP_VERSION(6, 1, 1):
 		amdgpu_device_ip_block_add(adev, &vpe_v6_1_ip_block);
 		break;
 	default:
@@ -2225,6 +2237,7 @@ static int amdgpu_discovery_set_umsch_mm_ip_blocks(struct amdgpu_device *adev)
 {
 	switch (amdgpu_ip_version(adev, VCN_HWIP, 0)) {
 	case IP_VERSION(4, 0, 5):
+	case IP_VERSION(4, 0, 6):
 		if (amdgpu_umsch_mm & 0x1) {
 			amdgpu_device_ip_block_add(adev, &umsch_mm_v4_0_ip_block);
 			adev->enable_umsch_mm = true;
@@ -2558,6 +2571,10 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		adev->nbio.funcs = &nbio_v7_7_funcs;
 		adev->nbio.hdp_flush_reg = &nbio_v7_7_hdp_flush_reg;
 		break;
+	case IP_VERSION(6, 3, 1):
+		adev->nbio.funcs = &nbif_v6_3_1_funcs;
+		adev->nbio.hdp_flush_reg = &nbif_v6_3_1_hdp_flush_reg;
+		break;
 	default:
 		break;
 	}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
index 6acffedf648c..80b9642f2bc4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
@@ -2479,8 +2479,11 @@ static void amdgpu_drv_delayed_reset_work_handler(struct work_struct *work)
 	}
 	for (i = 0; i < mgpu_info.num_dgpu; i++) {
 		adev = mgpu_info.gpu_ins[i].adev;
-		if (!adev->kfd.init_complete)
+		if (!adev->kfd.init_complete) {
+			kgd2kfd_init_zone_device(adev);
 			amdgpu_amdkfd_device_init(adev);
+			amdgpu_amdkfd_drm_client_create(adev);
+		}
 		amdgpu_ttm_set_buffer_funcs_status(adev, true);
 	}
 }
@@ -2693,7 +2696,7 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev)
 	}
 
 	adev->in_runpm = true;
-	if (amdgpu_device_supports_px(drm_dev))
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX)
 		drm_dev->switch_power_state = DRM_SWITCH_POWER_CHANGING;
 
 	/*
@@ -2703,7 +2706,7 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev)
 	 * platforms.
 	 * TODO: this may be also needed for PX capable platform.
 	 */
-	if (amdgpu_device_supports_boco(drm_dev))
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO)
 		adev->mp1_state = PP_MP1_STATE_UNLOAD;
 
 	ret = amdgpu_device_prepare(drm_dev);
@@ -2712,15 +2715,15 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev)
 	ret = amdgpu_device_suspend(drm_dev, false);
 	if (ret) {
 		adev->in_runpm = false;
-		if (amdgpu_device_supports_boco(drm_dev))
+		if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO)
 			adev->mp1_state = PP_MP1_STATE_NONE;
 		return ret;
 	}
 
-	if (amdgpu_device_supports_boco(drm_dev))
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO)
 		adev->mp1_state = PP_MP1_STATE_NONE;
 
-	if (amdgpu_device_supports_px(drm_dev)) {
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX) {
 		/* Only need to handle PCI state in the driver for ATPX
 		 * PCI core handles it for _PR3.
 		 */
@@ -2729,9 +2732,9 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev)
 		pci_ignore_hotplug(pdev);
 		pci_set_power_state(pdev, PCI_D3cold);
 		drm_dev->switch_power_state = DRM_SWITCH_POWER_DYNAMIC_OFF;
-	} else if (amdgpu_device_supports_boco(drm_dev)) {
+	} else if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO) {
 		/* nothing to do */
-	} else if (amdgpu_device_supports_baco(drm_dev)) {
+	} else if (adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) {
 		amdgpu_device_baco_enter(drm_dev);
 	}
 
@@ -2754,7 +2757,7 @@ static int amdgpu_pmops_runtime_resume(struct device *dev)
 	if (!pci_device_is_present(adev->pdev))
 		adev->no_hw_access = true;
 
-	if (amdgpu_device_supports_px(drm_dev)) {
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX) {
 		drm_dev->switch_power_state = DRM_SWITCH_POWER_CHANGING;
 
 		/* Only need to handle PCI state in the driver for ATPX
@@ -2766,22 +2769,22 @@ static int amdgpu_pmops_runtime_resume(struct device *dev)
 		if (ret)
 			return ret;
 		pci_set_master(pdev);
-	} else if (amdgpu_device_supports_boco(drm_dev)) {
+	} else if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO) {
 		/* Only need to handle PCI state in the driver for ATPX
 		 * PCI core handles it for _PR3.
 		 */
 		pci_set_master(pdev);
-	} else if (amdgpu_device_supports_baco(drm_dev)) {
+	} else if (adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) {
 		amdgpu_device_baco_exit(drm_dev);
 	}
 	ret = amdgpu_device_resume(drm_dev, false);
 	if (ret) {
-		if (amdgpu_device_supports_px(drm_dev))
+		if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX)
 			pci_disable_device(pdev);
 		return ret;
 	}
 
-	if (amdgpu_device_supports_px(drm_dev))
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX)
 		drm_dev->switch_power_state = DRM_SWITCH_POWER_ON;
 	adev->in_runpm = false;
 	return 0;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
index 22aeee8adb71..67c234bcf89f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
@@ -208,9 +208,15 @@ static int amdgpu_gem_object_open(struct drm_gem_object *obj,
 	if (!WARN_ON(!vm->process_info->eviction_fence)) {
 		r = amdgpu_amdkfd_bo_validate_and_fence(abo, AMDGPU_GEM_DOMAIN_GTT,
 							&vm->process_info->eviction_fence->base);
-		if (r)
-			dev_warn(adev->dev, "%d: validate_and_fence failed: %d\n",
-				 vm->task_info.pid, r);
+		if (r) {
+			struct amdgpu_task_info *ti = amdgpu_vm_get_task_info_vm(vm);
+
+			dev_warn(adev->dev, "validate_and_fence failed: %d\n", r);
+			if (ti) {
+				dev_warn(adev->dev, "pid %d\n", ti->pid);
+				amdgpu_vm_put_task_info(ti);
+			}
+		}
 	}
 	mutex_unlock(&vm->process_info->lock);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index f8b48fd93108..55d5508987ff 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -687,7 +687,7 @@ int amdgpu_gfx_enable_kgq(struct amdgpu_device *adev, int xcc_id)
 	r = amdgpu_ring_test_helper(kiq_ring);
 	spin_unlock(&kiq->ring_lock);
 	if (r)
-		DRM_ERROR("KCQ enable failed\n");
+		DRM_ERROR("KGQ enable failed\n");
 
 	return r;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c
index 55b65fc04b65..431ec72655ec 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c
@@ -129,13 +129,25 @@ static const struct mmu_interval_notifier_ops amdgpu_hmm_hsa_ops = {
  */
 int amdgpu_hmm_register(struct amdgpu_bo *bo, unsigned long addr)
 {
+	int r;
+
 	if (bo->kfd_bo)
-		return mmu_interval_notifier_insert(&bo->notifier, current->mm,
+		r = mmu_interval_notifier_insert(&bo->notifier, current->mm,
 						    addr, amdgpu_bo_size(bo),
 						    &amdgpu_hmm_hsa_ops);
-	return mmu_interval_notifier_insert(&bo->notifier, current->mm, addr,
-					    amdgpu_bo_size(bo),
-					    &amdgpu_hmm_gfx_ops);
+	else
+		r = mmu_interval_notifier_insert(&bo->notifier, current->mm, addr,
+							amdgpu_bo_size(bo),
+							&amdgpu_hmm_gfx_ops);
+	if (r)
+		/*
+		 * Make sure amdgpu_hmm_unregister() doesn't call
+		 * mmu_interval_notifier_remove() when the notifier isn't properly
+		 * initialized.
+		 */
+		bo->notifier.mm = NULL;
+
+	return r;
 }
 
 /**
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
index 6aa3b1d845ab..8b512dc28df8 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
@@ -131,7 +131,6 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned int num_ibs,
 	struct amdgpu_ib *ib = &ibs[0];
 	struct dma_fence *tmp = NULL;
 	bool need_ctx_switch;
-	unsigned int patch_offset = ~0;
 	struct amdgpu_vm *vm;
 	uint64_t fence_ctx;
 	uint32_t status = 0, alloc_size;
@@ -139,10 +138,11 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned int num_ibs,
 	bool secure, init_shadow;
 	u64 shadow_va, csa_va, gds_va;
 	int vmid = AMDGPU_JOB_GET_VMID(job);
+	bool need_pipe_sync = false;
+	unsigned int cond_exec;
 
 	unsigned int i;
 	int r = 0;
-	bool need_pipe_sync = false;
 
 	if (num_ibs == 0)
 		return -EINVAL;
@@ -228,7 +228,8 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned int num_ibs,
 					    init_shadow, vmid);
 
 	if (ring->funcs->init_cond_exec)
-		patch_offset = amdgpu_ring_init_cond_exec(ring);
+		cond_exec = amdgpu_ring_init_cond_exec(ring,
+						       ring->cond_exe_gpu_addr);
 
 	amdgpu_device_flush_hdp(adev, ring);
 
@@ -278,16 +279,9 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned int num_ibs,
 				       fence_flags | AMDGPU_FENCE_FLAG_64BIT);
 	}
 
-	if (ring->funcs->emit_gfx_shadow) {
+	if (ring->funcs->emit_gfx_shadow && ring->funcs->init_cond_exec) {
 		amdgpu_ring_emit_gfx_shadow(ring, 0, 0, 0, false, 0);
-
-		if (ring->funcs->init_cond_exec) {
-			unsigned int ce_offset = ~0;
-
-			ce_offset = amdgpu_ring_init_cond_exec(ring);
-			if (ce_offset != ~0 && ring->funcs->patch_cond_exec)
-				amdgpu_ring_patch_cond_exec(ring, ce_offset);
-		}
+		amdgpu_ring_init_cond_exec(ring, ring->cond_exe_gpu_addr);
 	}
 
 	r = amdgpu_fence_emit(ring, f, job, fence_flags);
@@ -302,8 +296,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned int num_ibs,
 	if (ring->funcs->insert_end)
 		ring->funcs->insert_end(ring);
 
-	if (patch_offset != ~0 && ring->funcs->patch_cond_exec)
-		amdgpu_ring_patch_cond_exec(ring, patch_offset);
+	amdgpu_ring_patch_cond_exec(ring, cond_exec);
 
 	ring->current_ctx = fence_ctx;
 	if (vm && ring->funcs->emit_switch_buffer)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
index 71a5cf37b472..4b3000c21ef2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
@@ -35,7 +35,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 {
 	struct amdgpu_ring *ring = to_amdgpu_ring(s_job->sched);
 	struct amdgpu_job *job = to_amdgpu_job(s_job);
-	struct amdgpu_task_info ti;
+	struct amdgpu_task_info *ti;
 	struct amdgpu_device *adev = ring->adev;
 	int idx;
 	int r;
@@ -48,7 +48,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 		return DRM_GPU_SCHED_STAT_ENODEV;
 	}
 
-	memset(&ti, 0, sizeof(struct amdgpu_task_info));
+
 	adev->job_hang = true;
 
 	if (amdgpu_gpu_recovery &&
@@ -58,12 +58,16 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 		goto exit;
 	}
 
-	amdgpu_vm_get_task_info(ring->adev, job->pasid, &ti);
 	DRM_ERROR("ring %s timeout, signaled seq=%u, emitted seq=%u\n",
-		  job->base.sched->name, atomic_read(&ring->fence_drv.last_seq),
-		  ring->fence_drv.sync_seq);
-	DRM_ERROR("Process information: process %s pid %d thread %s pid %d\n",
-		  ti.process_name, ti.tgid, ti.task_name, ti.pid);
+		   job->base.sched->name, atomic_read(&ring->fence_drv.last_seq),
+		   ring->fence_drv.sync_seq);
+
+	ti = amdgpu_vm_get_task_info_pasid(ring->adev, job->pasid);
+	if (ti) {
+		DRM_ERROR("Process information: process %s pid %d thread %s pid %d\n",
+			  ti->process_name, ti->tgid, ti->task_name, ti->pid);
+		amdgpu_vm_put_task_info(ti);
+	}
 
 	dma_fence_set_error(&s_job->s_fence->finished, -ETIME);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
index 3c2b1413058b..94b310fdb719 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
@@ -1830,6 +1830,10 @@ static int psp_hdcp_initialize(struct psp_context *psp)
 	if (amdgpu_sriov_vf(psp->adev))
 		return 0;
 
+	/* bypass hdcp initialization if dmu is harvested */
+	if (!amdgpu_device_has_display_hardware(psp->adev))
+		return 0;
+
 	if (!psp->hdcp_context.context.bin_desc.size_bytes ||
 	    !psp->hdcp_context.context.bin_desc.start_addr) {
 		dev_info(psp->adev->dev, "HDCP: optional hdcp ta ucode is not available\n");
@@ -1862,6 +1866,9 @@ int psp_hdcp_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 	if (amdgpu_sriov_vf(psp->adev))
 		return 0;
 
+	if (!psp->hdcp_context.context.initialized)
+		return 0;
+
 	return psp_ta_invoke(psp, ta_cmd_id, &psp->hdcp_context.context);
 }
 
@@ -1897,6 +1904,10 @@ static int psp_dtm_initialize(struct psp_context *psp)
 	if (amdgpu_sriov_vf(psp->adev))
 		return 0;
 
+	/* bypass dtm initialization if dmu is harvested */
+	if (!amdgpu_device_has_display_hardware(psp->adev))
+		return 0;
+
 	if (!psp->dtm_context.context.bin_desc.size_bytes ||
 	    !psp->dtm_context.context.bin_desc.start_addr) {
 		dev_info(psp->adev->dev, "DTM: optional dtm ta ucode is not available\n");
@@ -1929,6 +1940,9 @@ int psp_dtm_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 	if (amdgpu_sriov_vf(psp->adev))
 		return 0;
 
+	if (!psp->dtm_context.context.initialized)
+		return 0;
+
 	return psp_ta_invoke(psp, ta_cmd_id, &psp->dtm_context.context);
 }
 
@@ -2063,6 +2077,10 @@ static int psp_securedisplay_initialize(struct psp_context *psp)
 	if (amdgpu_sriov_vf(psp->adev))
 		return 0;
 
+	/* bypass securedisplay initialization if dmu is harvested */
+	if (!amdgpu_device_has_display_hardware(psp->adev))
+		return 0;
+
 	if (!psp->securedisplay_context.context.bin_desc.size_bytes ||
 	    !psp->securedisplay_context.context.bin_desc.start_addr) {
 		dev_info(psp->adev->dev, "SECUREDISPLAY: securedisplay ta ucode is not available\n");
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c
index 4baa300121d8..147100c27c2d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c
@@ -196,6 +196,13 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,
 			   coredump->reset_task_info.process_name,
 			   coredump->reset_task_info.pid);
 
+	if (coredump->ring) {
+		drm_printf(&p, "\nRing timed out details\n");
+		drm_printf(&p, "IP Type: %d Ring Name: %s\n",
+			   coredump->ring->funcs->type,
+			   coredump->ring->name);
+	}
+
 	if (coredump->reset_vram_lost)
 		drm_printf(&p, "VRAM is lost due to GPU reset!\n");
 	if (coredump->adev->reset_info.num_regs) {
@@ -220,6 +227,8 @@ void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost,
 {
 	struct amdgpu_coredump_info *coredump;
 	struct drm_device *dev = adev_to_drm(adev);
+	struct amdgpu_job *job = reset_context->job;
+	struct drm_sched_job *s_job;
 
 	coredump = kzalloc(sizeof(*coredump), GFP_NOWAIT);
 
@@ -230,8 +239,21 @@ void amdgpu_coredump(struct amdgpu_device *adev, bool vram_lost,
 
 	coredump->reset_vram_lost = vram_lost;
 
-	if (reset_context->job && reset_context->job->vm)
-		coredump->reset_task_info = reset_context->job->vm->task_info;
+	if (reset_context->job && reset_context->job->vm) {
+		struct amdgpu_task_info *ti;
+		struct amdgpu_vm *vm = reset_context->job->vm;
+
+		ti = amdgpu_vm_get_task_info_vm(vm);
+		if (ti) {
+			coredump->reset_task_info = *ti;
+			amdgpu_vm_put_task_info(ti);
+		}
+	}
+
+	if (job) {
+		s_job = &job->base;
+		coredump->ring = to_amdgpu_ring(s_job->sched);
+	}
 
 	coredump->adev = adev;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
index 19899f6b9b2b..60522963aaca 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
@@ -97,6 +97,7 @@ struct amdgpu_coredump_info {
 	struct amdgpu_task_info         reset_task_info;
 	struct timespec64               reset_time;
 	bool                            reset_vram_lost;
+	struct amdgpu_ring			*ring;
 };
 #endif
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c
index 5505d646f43a..06f0a6534a94 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c
@@ -524,46 +524,58 @@ static ssize_t amdgpu_debugfs_mqd_read(struct file *f, char __user *buf,
 {
 	struct amdgpu_ring *ring = file_inode(f)->i_private;
 	volatile u32 *mqd;
-	int r;
+	u32 *kbuf;
+	int r, i;
 	uint32_t value, result;
 
 	if (*pos & 3 || size & 3)
 		return -EINVAL;
 
-	result = 0;
+	kbuf = kmalloc(ring->mqd_size, GFP_KERNEL);
+	if (!kbuf)
+		return -ENOMEM;
 
 	r = amdgpu_bo_reserve(ring->mqd_obj, false);
 	if (unlikely(r != 0))
-		return r;
+		goto err_free;
 
 	r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&mqd);
-	if (r) {
-		amdgpu_bo_unreserve(ring->mqd_obj);
-		return r;
-	}
+	if (r)
+		goto err_unreserve;
 
+	/*
+	 * Copy to local buffer to avoid put_user(), which might fault
+	 * and acquire mmap_sem, under reservation_ww_class_mutex.
+	 */
+	for (i = 0; i < ring->mqd_size/sizeof(u32); i++)
+		kbuf[i] = mqd[i];
+
+	amdgpu_bo_kunmap(ring->mqd_obj);
+	amdgpu_bo_unreserve(ring->mqd_obj);
+
+	result = 0;
 	while (size) {
 		if (*pos >= ring->mqd_size)
-			goto done;
+			break;
 
-		value = mqd[*pos/4];
+		value = kbuf[*pos/4];
 		r = put_user(value, (uint32_t *)buf);
 		if (r)
-			goto done;
+			goto err_free;
 		buf += 4;
 		result += 4;
 		size -= 4;
 		*pos += 4;
 	}
 
-done:
-	amdgpu_bo_kunmap(ring->mqd_obj);
-	mqd = NULL;
-	amdgpu_bo_unreserve(ring->mqd_obj);
-	if (r)
-		return r;
-
+	kfree(kbuf);
 	return result;
+
+err_unreserve:
+	amdgpu_bo_unreserve(ring->mqd_obj);
+err_free:
+	kfree(kbuf);
+	return r;
 }
 
 static const struct file_operations amdgpu_debugfs_mqd_fops = {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
index fe1a61eb6e4c..582053f1cd56 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
@@ -209,8 +209,7 @@ struct amdgpu_ring_funcs {
 	void (*insert_end)(struct amdgpu_ring *ring);
 	/* pad the indirect buffer to the necessary number of dw */
 	void (*pad_ib)(struct amdgpu_ring *ring, struct amdgpu_ib *ib);
-	unsigned (*init_cond_exec)(struct amdgpu_ring *ring);
-	void (*patch_cond_exec)(struct amdgpu_ring *ring, unsigned offset);
+	unsigned (*init_cond_exec)(struct amdgpu_ring *ring, uint64_t addr);
 	/* note usage for clock and power gating */
 	void (*begin_use)(struct amdgpu_ring *ring);
 	void (*end_use)(struct amdgpu_ring *ring);
@@ -286,6 +285,9 @@ struct amdgpu_ring {
 	unsigned		cond_exe_offs;
 	u64			cond_exe_gpu_addr;
 	volatile u32		*cond_exe_cpu_addr;
+	unsigned int		set_q_mode_offs;
+	volatile u32		*set_q_mode_ptr;
+	u64			set_q_mode_token;
 	unsigned		vm_hub;
 	unsigned		vm_inv_eng;
 	struct dma_fence	*vmid_wait;
@@ -327,8 +329,7 @@ struct amdgpu_ring {
 #define amdgpu_ring_emit_reg_write_reg_wait(r, d0, d1, v, m) (r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m))
 #define amdgpu_ring_emit_frame_cntl(r, b, s) (r)->funcs->emit_frame_cntl((r), (b), (s))
 #define amdgpu_ring_pad_ib(r, ib) ((r)->funcs->pad_ib((r), (ib)))
-#define amdgpu_ring_init_cond_exec(r) (r)->funcs->init_cond_exec((r))
-#define amdgpu_ring_patch_cond_exec(r,o) (r)->funcs->patch_cond_exec((r),(o))
+#define amdgpu_ring_init_cond_exec(r, a) (r)->funcs->init_cond_exec((r), (a))
 #define amdgpu_ring_preempt_ib(r) (r)->funcs->preempt_ib(r)
 #define amdgpu_ring_patch_cntl(r, o) ((r)->funcs->patch_cntl((r), (o)))
 #define amdgpu_ring_patch_ce(r, o) ((r)->funcs->patch_ce((r), (o)))
@@ -411,6 +412,30 @@ static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring,
 	ring->count_dw -= count_dw;
 }
 
+/**
+ * amdgpu_ring_patch_cond_exec - patch dw count of conditional execute
+ * @ring: amdgpu_ring structure
+ * @offset: offset returned by amdgpu_ring_init_cond_exec
+ *
+ * Calculate the dw count and patch it into a cond_exec command.
+ */
+static inline void amdgpu_ring_patch_cond_exec(struct amdgpu_ring *ring,
+					       unsigned int offset)
+{
+	unsigned cur;
+
+	if (!ring->funcs->init_cond_exec)
+		return;
+
+	WARN_ON(offset > ring->buf_mask);
+	WARN_ON(ring->ring[offset] != 0);
+
+	cur = (ring->wptr - 1) & ring->buf_mask;
+	if (cur < offset)
+		cur += ring->ring_size >> 2;
+	ring->ring[offset] = cur - offset;
+}
+
 #define amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset)			\
 	(ring->is_mes_queue && ring->mes_ctx ?				\
 	 (ring->mes_ctx->meta_data_gpu_addr + offset) : 0)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 8722beba494e..fc418e670fda 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -864,6 +864,7 @@ static void amdgpu_ttm_gart_bind(struct amdgpu_device *adev,
 		amdgpu_gart_bind(adev, gtt->offset, ttm->num_pages,
 				 gtt->ttm.dma_address, flags);
 	}
+	gtt->bound = true;
 }
 
 /*
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c
index ab820cf52668..0df97c3e3a70 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c
@@ -189,10 +189,13 @@ static void setup_vpe_queue(struct amdgpu_device *adev,
 	mqd->rptr_val = 0;
 	mqd->unmapped = 1;
 
+	if (adev->vpe.collaborate_mode)
+		memcpy(++mqd, test->mqd_data_cpu_addr, sizeof(struct MQD_INFO));
+
 	qinfo->mqd_addr = test->mqd_data_gpu_addr;
 	qinfo->csa_addr = test->ctx_data_gpu_addr +
 		offsetof(struct umsch_mm_test_ctx_data, vpe_ctx_csa);
-	qinfo->doorbell_offset_0 = (adev->doorbell_index.vpe_ring + 1) << 1;
+	qinfo->doorbell_offset_0 = 0;
 	qinfo->doorbell_offset_1 = 0;
 }
 
@@ -287,7 +290,10 @@ static int submit_vpe_queue(struct amdgpu_device *adev, struct umsch_mm_test *te
 	ring[5] = 0;
 
 	mqd->wptr_val = (6 << 2);
-	// WDOORBELL32(adev->umsch_mm.agdb_index[CONTEXT_PRIORITY_LEVEL_NORMAL], mqd->wptr_val);
+	if (adev->vpe.collaborate_mode)
+		(++mqd)->wptr_val = (6 << 2);
+
+	WDOORBELL32(adev->umsch_mm.agdb_index[CONTEXT_PRIORITY_LEVEL_NORMAL], mqd->wptr_val);
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		if (*fence == test_pattern)
@@ -571,6 +577,7 @@ int amdgpu_umsch_mm_init_microcode(struct amdgpu_umsch_mm *umsch)
 
 	switch (amdgpu_ip_version(adev, VCN_HWIP, 0)) {
 	case IP_VERSION(4, 0, 5):
+	case IP_VERSION(4, 0, 6):
 		fw_name = "amdgpu/umsch_mm_4_0_0.bin";
 		break;
 	default:
@@ -750,6 +757,7 @@ static int umsch_mm_early_init(void *handle)
 
 	switch (amdgpu_ip_version(adev, VCN_HWIP, 0)) {
 	case IP_VERSION(4, 0, 5):
+	case IP_VERSION(4, 0, 6):
 		umsch_mm_v4_0_set_funcs(&adev->umsch_mm);
 		break;
 	default:
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h
index 8258a43a6236..5014b5af95fd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h
@@ -33,13 +33,6 @@ enum UMSCH_SWIP_ENGINE_TYPE {
 	UMSCH_SWIP_ENGINE_TYPE_MAX
 };
 
-enum UMSCH_SWIP_AFFINITY_TYPE {
-	UMSCH_SWIP_AFFINITY_TYPE_ANY = 0,
-	UMSCH_SWIP_AFFINITY_TYPE_VCN0 = 1,
-	UMSCH_SWIP_AFFINITY_TYPE_VCN1 = 2,
-	UMSCH_SWIP_AFFINITY_TYPE_MAX
-};
-
 enum UMSCH_CONTEXT_PRIORITY_LEVEL {
 	CONTEXT_PRIORITY_LEVEL_IDLE = 0,
 	CONTEXT_PRIORITY_LEVEL_NORMAL = 1,
@@ -51,13 +44,15 @@ enum UMSCH_CONTEXT_PRIORITY_LEVEL {
 struct umsch_mm_set_resource_input {
 	uint32_t vmid_mask_mm_vcn;
 	uint32_t vmid_mask_mm_vpe;
+	uint32_t collaboration_mask_vpe;
 	uint32_t logging_vmid;
 	uint32_t engine_mask;
 	union {
 		struct {
 			uint32_t disable_reset : 1;
 			uint32_t disable_umsch_mm_log : 1;
-			uint32_t reserved : 30;
+			uint32_t use_rs64mem_for_proc_ctx_csa : 1;
+			uint32_t reserved : 29;
 		};
 		uint32_t uint32_all;
 	};
@@ -78,15 +73,18 @@ struct umsch_mm_add_queue_input {
 	uint32_t doorbell_offset_1;
 	enum UMSCH_SWIP_ENGINE_TYPE engine_type;
 	uint32_t affinity;
-	enum UMSCH_SWIP_AFFINITY_TYPE affinity_type;
 	uint64_t mqd_addr;
 	uint64_t h_context;
 	uint64_t h_queue;
 	uint32_t vm_context_cntl;
 
+	uint32_t process_csa_array_index;
+	uint32_t context_csa_array_index;
+
 	struct {
 		uint32_t is_context_suspended : 1;
-		uint32_t reserved : 31;
+		uint32_t collaboration_mode : 1;
+		uint32_t reserved : 30;
 	};
 };
 
@@ -94,6 +92,7 @@ struct umsch_mm_remove_queue_input {
 	uint32_t doorbell_offset_0;
 	uint32_t doorbell_offset_1;
 	uint64_t context_csa_addr;
+	uint32_t context_csa_array_index;
 };
 
 struct MQD_INFO {
@@ -103,6 +102,7 @@ struct MQD_INFO {
 	uint32_t wptr_val;
 	uint32_t rptr_val;
 	uint32_t unmapped;
+	uint32_t vmid;
 };
 
 struct amdgpu_umsch_mm;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
index b2535023764f..9c514a606a2f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
@@ -60,6 +60,7 @@
 #define FIRMWARE_VCN4_0_4		"amdgpu/vcn_4_0_4.bin"
 #define FIRMWARE_VCN4_0_5		"amdgpu/vcn_4_0_5.bin"
 #define FIRMWARE_VCN4_0_6		"amdgpu/vcn_4_0_6.bin"
+#define FIRMWARE_VCN4_0_6_1		"amdgpu/vcn_4_0_6_1.bin"
 #define FIRMWARE_VCN5_0_0		"amdgpu/vcn_5_0_0.bin"
 
 MODULE_FIRMWARE(FIRMWARE_RAVEN);
@@ -85,6 +86,7 @@ MODULE_FIRMWARE(FIRMWARE_VCN4_0_3);
 MODULE_FIRMWARE(FIRMWARE_VCN4_0_4);
 MODULE_FIRMWARE(FIRMWARE_VCN4_0_5);
 MODULE_FIRMWARE(FIRMWARE_VCN4_0_6);
+MODULE_FIRMWARE(FIRMWARE_VCN4_0_6_1);
 MODULE_FIRMWARE(FIRMWARE_VCN5_0_0);
 
 static void amdgpu_vcn_idle_work_handler(struct work_struct *work);
@@ -93,14 +95,22 @@ int amdgpu_vcn_early_init(struct amdgpu_device *adev)
 {
 	char ucode_prefix[30];
 	char fw_name[40];
-	int r;
+	int r, i;
 
-	amdgpu_ucode_ip_version_decode(adev, UVD_HWIP, ucode_prefix, sizeof(ucode_prefix));
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s.bin", ucode_prefix);
-	r = amdgpu_ucode_request(adev, &adev->vcn.fw, fw_name);
-	if (r)
-		amdgpu_ucode_release(&adev->vcn.fw);
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		amdgpu_ucode_ip_version_decode(adev, UVD_HWIP, ucode_prefix, sizeof(ucode_prefix));
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s.bin", ucode_prefix);
+		if (amdgpu_ip_version(adev, UVD_HWIP, 0) ==  IP_VERSION(4, 0, 6) &&
+			i == 1) {
+			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_%d.bin", ucode_prefix, i);
+		}
 
+		r = amdgpu_ucode_request(adev, &adev->vcn.fw[i], fw_name);
+		if (r) {
+			amdgpu_ucode_release(&adev->vcn.fw[i]);
+			return r;
+		}
+	}
 	return r;
 }
 
@@ -141,7 +151,7 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev)
 		}
 	}
 
-	hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
+	hdr = (const struct common_firmware_header *)adev->vcn.fw[0]->data;
 	adev->vcn.fw_version = le32_to_cpu(hdr->ucode_version);
 
 	/* Bit 20-23, it is encode major and non-zero for new naming convention.
@@ -256,9 +266,10 @@ int amdgpu_vcn_sw_fini(struct amdgpu_device *adev)
 
 		for (i = 0; i < adev->vcn.num_enc_rings; ++i)
 			amdgpu_ring_fini(&adev->vcn.inst[j].ring_enc[i]);
+
+		amdgpu_ucode_release(&adev->vcn.fw[j]);
 	}
 
-	amdgpu_ucode_release(&adev->vcn.fw);
 	mutex_destroy(&adev->vcn.vcn1_jpeg1_workaround);
 	mutex_destroy(&adev->vcn.vcn_pg_lock);
 
@@ -354,11 +365,12 @@ int amdgpu_vcn_resume(struct amdgpu_device *adev)
 			const struct common_firmware_header *hdr;
 			unsigned int offset;
 
-			hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
+			hdr = (const struct common_firmware_header *)adev->vcn.fw[i]->data;
 			if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
 				offset = le32_to_cpu(hdr->ucode_array_offset_bytes);
 				if (drm_dev_enter(adev_to_drm(adev), &idx)) {
-					memcpy_toio(adev->vcn.inst[i].cpu_addr, adev->vcn.fw->data + offset,
+					memcpy_toio(adev->vcn.inst[i].cpu_addr,
+						    adev->vcn.fw[i]->data + offset,
 						    le32_to_cpu(hdr->ucode_size_bytes));
 					drm_dev_exit(idx);
 				}
@@ -1043,11 +1055,11 @@ void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev)
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 		const struct common_firmware_header *hdr;
 
-		hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
-
 		for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
 			if (adev->vcn.harvest_config & (1 << i))
 				continue;
+
+			hdr = (const struct common_firmware_header *)adev->vcn.fw[i]->data;
 			/* currently only support 2 FW instances */
 			if (i >= 2) {
 				dev_info(adev->dev, "More then 2 VCN FW instances!\n");
@@ -1055,7 +1067,7 @@ void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev)
 			}
 			idx = AMDGPU_UCODE_ID_VCN + i;
 			adev->firmware.ucode[idx].ucode_id = idx;
-			adev->firmware.ucode[idx].fw = adev->vcn.fw;
+			adev->firmware.ucode[idx].fw = adev->vcn.fw[i];
 			adev->firmware.fw_size +=
 				ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
index 1985f71b4373..a418393d89ec 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
@@ -306,7 +306,7 @@ struct amdgpu_vcn_ras {
 struct amdgpu_vcn {
 	unsigned		fw_version;
 	struct delayed_work	idle_work;
-	const struct firmware	*fw;	/* VCN firmware */
+	const struct firmware	*fw[AMDGPU_MAX_VCN_INSTANCES];	/* VCN firmware */
 	unsigned		num_enc_rings;
 	enum amd_powergating_state cur_state;
 	bool			indirect_sram;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index ed4a8c5d26d7..4299ce386322 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -513,8 +513,14 @@ int amdgpu_vm_validate(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		bo = bo_base->bo;
 
 		if (dma_resv_locking_ctx(bo->tbo.base.resv) != ticket) {
-			pr_warn_ratelimited("Evicted user BO is not reserved in pid %d\n",
-					    vm->task_info.pid);
+			struct amdgpu_task_info *ti = amdgpu_vm_get_task_info_vm(vm);
+
+			pr_warn_ratelimited("Evicted user BO is not reserved\n");
+			if (ti) {
+				pr_warn_ratelimited("pid %d\n", ti->pid);
+				amdgpu_vm_put_task_info(ti);
+			}
+
 			return -EINVAL;
 		}
 
@@ -652,7 +658,7 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 	bool vm_flush_needed = job->vm_needs_flush;
 	struct dma_fence *fence = NULL;
 	bool pasid_mapping_needed = false;
-	unsigned patch_offset = 0;
+	unsigned int patch;
 	int r;
 
 	if (amdgpu_vmid_had_gpu_reset(adev, id)) {
@@ -679,7 +685,8 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 
 	amdgpu_ring_ib_begin(ring);
 	if (ring->funcs->init_cond_exec)
-		patch_offset = amdgpu_ring_init_cond_exec(ring);
+		patch = amdgpu_ring_init_cond_exec(ring,
+						   ring->cond_exe_gpu_addr);
 
 	if (need_pipe_sync)
 		amdgpu_ring_emit_pipeline_sync(ring);
@@ -727,8 +734,7 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 	}
 	dma_fence_put(fence);
 
-	if (ring->funcs->patch_cond_exec)
-		amdgpu_ring_patch_cond_exec(ring, patch_offset);
+	amdgpu_ring_patch_cond_exec(ring, patch);
 
 	/* the double SWITCH_BUFFER here *cannot* be skipped by COND_EXEC */
 	if (ring->funcs->emit_switch_buffer) {
@@ -1385,10 +1391,6 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
 			struct amdgpu_bo_va_mapping, list);
 		list_del(&mapping->list);
 
-		if (vm->pte_support_ats &&
-		    mapping->start < AMDGPU_GMC_HOLE_START)
-			init_pte_value = AMDGPU_PTE_DEFAULT_ATC;
-
 		r = amdgpu_vm_update_range(adev, vm, false, false, true, false,
 					   resv, mapping->start, mapping->last,
 					   init_pte_value, 0, 0, NULL, NULL,
@@ -2225,6 +2227,108 @@ long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout)
 	return dma_fence_wait_timeout(vm->last_unlocked, true, timeout);
 }
 
+static void amdgpu_vm_destroy_task_info(struct kref *kref)
+{
+	struct amdgpu_task_info *ti = container_of(kref, struct amdgpu_task_info, refcount);
+
+	kfree(ti);
+}
+
+static inline struct amdgpu_vm *
+amdgpu_vm_get_vm_from_pasid(struct amdgpu_device *adev, u32 pasid)
+{
+	struct amdgpu_vm *vm;
+	unsigned long flags;
+
+	xa_lock_irqsave(&adev->vm_manager.pasids, flags);
+	vm = xa_load(&adev->vm_manager.pasids, pasid);
+	xa_unlock_irqrestore(&adev->vm_manager.pasids, flags);
+
+	return vm;
+}
+
+/**
+ * amdgpu_vm_put_task_info - reference down the vm task_info ptr
+ *
+ * @task_info: task_info struct under discussion.
+ *
+ * frees the vm task_info ptr at the last put
+ */
+void amdgpu_vm_put_task_info(struct amdgpu_task_info *task_info)
+{
+	kref_put(&task_info->refcount, amdgpu_vm_destroy_task_info);
+}
+
+/**
+ * amdgpu_vm_get_task_info_vm - Extracts task info for a vm.
+ *
+ * @vm: VM to get info from
+ *
+ * Returns the reference counted task_info structure, which must be
+ * referenced down with amdgpu_vm_put_task_info.
+ */
+struct amdgpu_task_info *
+amdgpu_vm_get_task_info_vm(struct amdgpu_vm *vm)
+{
+	struct amdgpu_task_info *ti = NULL;
+
+	if (vm) {
+		ti = vm->task_info;
+		kref_get(&vm->task_info->refcount);
+	}
+
+	return ti;
+}
+
+/**
+ * amdgpu_vm_get_task_info_pasid - Extracts task info for a PASID.
+ *
+ * @adev: drm device pointer
+ * @pasid: PASID identifier for VM
+ *
+ * Returns the reference counted task_info structure, which must be
+ * referenced down with amdgpu_vm_put_task_info.
+ */
+struct amdgpu_task_info *
+amdgpu_vm_get_task_info_pasid(struct amdgpu_device *adev, u32 pasid)
+{
+	return amdgpu_vm_get_task_info_vm(
+			amdgpu_vm_get_vm_from_pasid(adev, pasid));
+}
+
+static int amdgpu_vm_create_task_info(struct amdgpu_vm *vm)
+{
+	vm->task_info = kzalloc(sizeof(struct amdgpu_task_info), GFP_KERNEL);
+	if (!vm->task_info)
+		return -ENOMEM;
+
+	kref_init(&vm->task_info->refcount);
+	return 0;
+}
+
+/**
+ * amdgpu_vm_set_task_info - Sets VMs task info.
+ *
+ * @vm: vm for which to set the info
+ */
+void amdgpu_vm_set_task_info(struct amdgpu_vm *vm)
+{
+	if (!vm->task_info)
+		return;
+
+	if (vm->task_info->pid == current->pid)
+		return;
+
+	vm->task_info->pid = current->pid;
+	get_task_comm(vm->task_info->task_name, current);
+
+	if (current->group_leader->mm != current->mm)
+		return;
+
+	vm->task_info->tgid = current->group_leader->pid;
+	get_task_comm(vm->task_info->process_name, current->group_leader);
+}
+
 /**
  * amdgpu_vm_init - initialize a vm instance
  *
@@ -2264,7 +2368,6 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	if (r)
 		return r;
 
-	vm->pte_support_ats = false;
 	vm->is_compute_context = false;
 
 	vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode &
@@ -2311,6 +2414,10 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	if (r)
 		goto error_free_root;
 
+	r = amdgpu_vm_create_task_info(vm);
+	if (r)
+		DRM_DEBUG("Failed to create task info for VM\n");
+
 	amdgpu_bo_unreserve(vm->root.bo);
 	amdgpu_bo_unref(&root_bo);
 
@@ -2350,30 +2457,12 @@ error_free_delayed:
  */
 int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 {
-	bool pte_support_ats = (adev->asic_type == CHIP_RAVEN);
 	int r;
 
 	r = amdgpu_bo_reserve(vm->root.bo, true);
 	if (r)
 		return r;
 
-	/* Check if PD needs to be reinitialized and do it before
-	 * changing any other state, in case it fails.
-	 */
-	if (pte_support_ats != vm->pte_support_ats) {
-		/* Sanity checks */
-		if (!amdgpu_vm_pt_is_root_clean(adev, vm)) {
-			r = -EINVAL;
-			goto unreserve_bo;
-		}
-
-		vm->pte_support_ats = pte_support_ats;
-		r = amdgpu_vm_pt_clear(adev, vm, to_amdgpu_bo_vm(vm->root.bo),
-				       false);
-		if (r)
-			goto unreserve_bo;
-	}
-
 	/* Update VM state */
 	vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode &
 				    AMDGPU_VM_USE_CPU_FOR_COMPUTE);
@@ -2450,6 +2539,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 
 	root = amdgpu_bo_ref(vm->root.bo);
 	amdgpu_bo_reserve(root, true);
+	amdgpu_vm_put_task_info(vm->task_info);
 	amdgpu_vm_set_pasid(adev, vm, 0);
 	dma_fence_wait(vm->last_unlocked, false);
 	dma_fence_put(vm->last_unlocked);
@@ -2607,48 +2697,6 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 }
 
 /**
- * amdgpu_vm_get_task_info - Extracts task info for a PASID.
- *
- * @adev: drm device pointer
- * @pasid: PASID identifier for VM
- * @task_info: task_info to fill.
- */
-void amdgpu_vm_get_task_info(struct amdgpu_device *adev, u32 pasid,
-			 struct amdgpu_task_info *task_info)
-{
-	struct amdgpu_vm *vm;
-	unsigned long flags;
-
-	xa_lock_irqsave(&adev->vm_manager.pasids, flags);
-
-	vm = xa_load(&adev->vm_manager.pasids, pasid);
-	if (vm)
-		*task_info = vm->task_info;
-
-	xa_unlock_irqrestore(&adev->vm_manager.pasids, flags);
-}
-
-/**
- * amdgpu_vm_set_task_info - Sets VMs task info.
- *
- * @vm: vm for which to set the info
- */
-void amdgpu_vm_set_task_info(struct amdgpu_vm *vm)
-{
-	if (vm->task_info.pid)
-		return;
-
-	vm->task_info.pid = current->pid;
-	get_task_comm(vm->task_info.task_name, current);
-
-	if (current->group_leader->mm != current->mm)
-		return;
-
-	vm->task_info.tgid = current->group_leader->pid;
-	get_task_comm(vm->task_info.process_name, current->group_leader);
-}
-
-/**
  * amdgpu_vm_handle_fault - graceful handling of VM faults.
  * @adev: amdgpu device pointer
  * @pasid: PASID of the VM
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
index 42f6ddec50c1..047ec1930d12 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
@@ -203,10 +203,11 @@ struct amdgpu_vm_pte_funcs {
 };
 
 struct amdgpu_task_info {
-	char	process_name[TASK_COMM_LEN];
-	char	task_name[TASK_COMM_LEN];
-	pid_t	pid;
-	pid_t	tgid;
+	char		process_name[TASK_COMM_LEN];
+	char		task_name[TASK_COMM_LEN];
+	pid_t		pid;
+	pid_t		tgid;
+	struct kref	refcount;
 };
 
 /**
@@ -357,9 +358,6 @@ struct amdgpu_vm {
 	/* Functions to use for VM table updates */
 	const struct amdgpu_vm_update_funcs	*update_funcs;
 
-	/* Flag to indicate ATS support from PTE for GFX9 */
-	bool			pte_support_ats;
-
 	/* Up to 128 pending retry page faults */
 	DECLARE_KFIFO(faults, u64, 128);
 
@@ -373,7 +371,7 @@ struct amdgpu_vm {
 	uint64_t		pd_phys_addr;
 
 	/* Some basic info about the task */
-	struct amdgpu_task_info task_info;
+	struct amdgpu_task_info *task_info;
 
 	/* Store positions of group of BOs */
 	struct ttm_lru_bulk_move lru_bulk_move;
@@ -514,8 +512,14 @@ bool amdgpu_vm_need_pipeline_sync(struct amdgpu_ring *ring,
 				  struct amdgpu_job *job);
 void amdgpu_vm_check_compute_bug(struct amdgpu_device *adev);
 
-void amdgpu_vm_get_task_info(struct amdgpu_device *adev, u32 pasid,
-			     struct amdgpu_task_info *task_info);
+struct amdgpu_task_info *
+amdgpu_vm_get_task_info_pasid(struct amdgpu_device *adev, u32 pasid);
+
+struct amdgpu_task_info *
+amdgpu_vm_get_task_info_vm(struct amdgpu_vm *vm);
+
+void amdgpu_vm_put_task_info(struct amdgpu_task_info *task_info);
+
 bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid,
 			    u32 vmid, u32 node_id, uint64_t addr,
 			    bool write_fault);
@@ -533,8 +537,6 @@ int amdgpu_vm_pt_create(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			int level, bool immediate, struct amdgpu_bo_vm **vmbo,
 			int32_t xcp_id);
 void amdgpu_vm_pt_free_root(struct amdgpu_device *adev, struct amdgpu_vm *vm);
-bool amdgpu_vm_pt_is_root_clean(struct amdgpu_device *adev,
-				struct amdgpu_vm *vm);
 
 int amdgpu_vm_pde_update(struct amdgpu_vm_update_params *params,
 			 struct amdgpu_vm_bo_base *entry);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c
index a160265ddc07..124389a6bf48 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c
@@ -90,22 +90,6 @@ static unsigned int amdgpu_vm_pt_num_entries(struct amdgpu_device *adev,
 }
 
 /**
- * amdgpu_vm_pt_num_ats_entries - return the number of ATS entries in the root PD
- *
- * @adev: amdgpu_device pointer
- *
- * Returns:
- * The number of entries in the root page directory which needs the ATS setting.
- */
-static unsigned int amdgpu_vm_pt_num_ats_entries(struct amdgpu_device *adev)
-{
-	unsigned int shift;
-
-	shift = amdgpu_vm_pt_level_shift(adev, adev->vm_manager.root_level);
-	return AMDGPU_GMC_HOLE_START >> (shift + AMDGPU_GPU_PAGE_SHIFT);
-}
-
-/**
  * amdgpu_vm_pt_entries_mask - the mask to get the entry number of a PD/PT
  *
  * @adev: amdgpu_device pointer
@@ -379,7 +363,7 @@ int amdgpu_vm_pt_clear(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	struct ttm_operation_ctx ctx = { true, false };
 	struct amdgpu_vm_update_params params;
 	struct amdgpu_bo *ancestor = &vmbo->bo;
-	unsigned int entries, ats_entries;
+	unsigned int entries;
 	struct amdgpu_bo *bo = &vmbo->bo;
 	uint64_t addr;
 	int r, idx;
@@ -394,27 +378,6 @@ int amdgpu_vm_pt_clear(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	}
 
 	entries = amdgpu_bo_size(bo) / 8;
-	if (!vm->pte_support_ats) {
-		ats_entries = 0;
-
-	} else if (!bo->parent) {
-		ats_entries = amdgpu_vm_pt_num_ats_entries(adev);
-		ats_entries = min(ats_entries, entries);
-		entries -= ats_entries;
-
-	} else {
-		struct amdgpu_vm_bo_base *pt;
-
-		pt = ancestor->vm_bo;
-		ats_entries = amdgpu_vm_pt_num_ats_entries(adev);
-		if ((pt - to_amdgpu_bo_vm(vm->root.bo)->entries) >=
-		    ats_entries) {
-			ats_entries = 0;
-		} else {
-			ats_entries = entries;
-			entries = 0;
-		}
-	}
 
 	r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 	if (r)
@@ -445,44 +408,24 @@ int amdgpu_vm_pt_clear(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		goto exit;
 
 	addr = 0;
-	if (ats_entries) {
-		uint64_t value = 0, flags;
 
-		flags = AMDGPU_PTE_DEFAULT_ATC;
+	uint64_t value = 0, flags = 0;
+	if (adev->asic_type >= CHIP_VEGA10) {
 		if (level != AMDGPU_VM_PTB) {
 			/* Handle leaf PDEs as PTEs */
 			flags |= AMDGPU_PDE_PTE;
-			amdgpu_gmc_get_vm_pde(adev, level, &value, &flags);
+			amdgpu_gmc_get_vm_pde(adev, level,
+					      &value, &flags);
+		} else {
+			/* Workaround for fault priority problem on GMC9 */
+			flags = AMDGPU_PTE_EXECUTABLE;
 		}
-
-		r = vm->update_funcs->update(&params, vmbo, addr, 0,
-					     ats_entries, value, flags);
-		if (r)
-			goto exit;
-
-		addr += ats_entries * 8;
 	}
 
-	if (entries) {
-		uint64_t value = 0, flags = 0;
-
-		if (adev->asic_type >= CHIP_VEGA10) {
-			if (level != AMDGPU_VM_PTB) {
-				/* Handle leaf PDEs as PTEs */
-				flags |= AMDGPU_PDE_PTE;
-				amdgpu_gmc_get_vm_pde(adev, level,
-						      &value, &flags);
-			} else {
-				/* Workaround for fault priority problem on GMC9 */
-				flags = AMDGPU_PTE_EXECUTABLE;
-			}
-		}
-
-		r = vm->update_funcs->update(&params, vmbo, addr, 0, entries,
-					     value, flags);
-		if (r)
-			goto exit;
-	}
+	r = vm->update_funcs->update(&params, vmbo, addr, 0, entries,
+				     value, flags);
+	if (r)
+		goto exit;
 
 	r = vm->update_funcs->commit(&params, NULL);
 exit:
@@ -728,33 +671,6 @@ void amdgpu_vm_pt_free_root(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 }
 
 /**
- * amdgpu_vm_pt_is_root_clean - check if a root PD is clean
- *
- * @adev: amdgpu_device pointer
- * @vm: the VM to check
- *
- * Check all entries of the root PD, if any subsequent PDs are allocated,
- * it means there are page table creating and filling, and is no a clean
- * VM
- *
- * Returns:
- *	0 if this VM is clean
- */
-bool amdgpu_vm_pt_is_root_clean(struct amdgpu_device *adev,
-				struct amdgpu_vm *vm)
-{
-	enum amdgpu_vm_level root = adev->vm_manager.root_level;
-	unsigned int entries = amdgpu_vm_pt_num_entries(adev, root);
-	unsigned int i = 0;
-
-	for (i = 0; i < entries; i++) {
-		if (to_amdgpu_bo_vm(vm->root.bo)->entries[i].bo)
-			return false;
-	}
-	return true;
-}
-
-/**
  * amdgpu_vm_pde_update - update a single level in the hierarchy
  *
  * @params: parameters for the update
@@ -1027,7 +943,7 @@ int amdgpu_vm_ptes_update(struct amdgpu_vm_update_params *params,
 			trace_amdgpu_vm_update_ptes(params, frag_start, upd_end,
 						    min(nptes, 32u), dst, incr,
 						    upd_flags,
-						    vm->task_info.tgid,
+						    vm->task_info ? vm->task_info->tgid : 0,
 						    vm->immediate.fence_context);
 			amdgpu_vm_pte_update_flags(params, to_amdgpu_bo_vm(pt),
 						   cursor.level, pe_start, dst,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c
index b9a15d51eb5c..6695481f870f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c
@@ -297,6 +297,10 @@ static int vpe_early_init(void *handle)
 	case IP_VERSION(6, 1, 0):
 		vpe_v6_1_set_funcs(vpe);
 		break;
+	case IP_VERSION(6, 1, 1):
+		vpe_v6_1_set_funcs(vpe);
+		vpe->collaborate_mode = true;
+		break;
 	default:
 		return -EINVAL;
 	}
@@ -304,6 +308,8 @@ static int vpe_early_init(void *handle)
 	vpe_set_ring_funcs(adev);
 	vpe_set_regs(vpe);
 
+	dev_info(adev->dev, "VPE: collaborate mode %s", vpe->collaborate_mode ? "true" : "false");
+
 	return 0;
 }
 
@@ -390,6 +396,12 @@ static int vpe_hw_init(void *handle)
 	struct amdgpu_vpe *vpe = &adev->vpe;
 	int ret;
 
+	/* Power on VPE */
+	ret = amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE,
+						     AMD_PG_STATE_UNGATE);
+	if (ret)
+		return ret;
+
 	ret = vpe_load_microcode(vpe);
 	if (ret)
 		return ret;
@@ -457,6 +469,18 @@ static uint64_t vpe_get_csa_mc_addr(struct amdgpu_ring *ring, uint32_t vmid)
 	return csa_mc_addr;
 }
 
+static void vpe_ring_emit_pred_exec(struct amdgpu_ring *ring,
+				    uint32_t device_select,
+				    uint32_t exec_count)
+{
+	if (!ring->adev->vpe.collaborate_mode)
+		return;
+
+	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_PRED_EXE, 0) |
+				(device_select << 16));
+	amdgpu_ring_write(ring, exec_count & 0x1fff);
+}
+
 static void vpe_ring_emit_ib(struct amdgpu_ring *ring,
 			     struct amdgpu_job *job,
 			     struct amdgpu_ib *ib,
@@ -505,6 +529,8 @@ static void vpe_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
 	uint32_t seq = ring->fence_drv.sync_seq;
 	uint64_t addr = ring->fence_drv.gpu_addr;
 
+	vpe_ring_emit_pred_exec(ring, 0, 6);
+
 	/* wait for idle */
 	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_POLL_REGMEM,
 				VPE_POLL_REGMEM_SUBOP_REGMEM) |
@@ -520,6 +546,8 @@ static void vpe_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
 
 static void vpe_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, uint32_t val)
 {
+	vpe_ring_emit_pred_exec(ring, 0, 3);
+
 	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_REG_WRITE, 0));
 	amdgpu_ring_write(ring,	reg << 2);
 	amdgpu_ring_write(ring, val);
@@ -528,6 +556,8 @@ static void vpe_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, uint32_t
 static void vpe_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
 				   uint32_t val, uint32_t mask)
 {
+	vpe_ring_emit_pred_exec(ring, 0, 6);
+
 	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_POLL_REGMEM,
 				VPE_POLL_REGMEM_SUBOP_REGMEM) |
 				VPE_CMD_POLL_REGMEM_HEADER_FUNC(3) | /* equal */
@@ -546,34 +576,21 @@ static void vpe_ring_emit_vm_flush(struct amdgpu_ring *ring, unsigned int vmid,
 	amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
 }
 
-static unsigned int vpe_ring_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned int vpe_ring_init_cond_exec(struct amdgpu_ring *ring,
+					    uint64_t addr)
 {
 	unsigned int ret;
 
 	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_COND_EXE, 0));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
 	amdgpu_ring_write(ring, 1);
-	ret = ring->wptr & ring->buf_mask;/* this is the offset we need patch later */
-	amdgpu_ring_write(ring, 0x55aa55aa);/* insert dummy here and patch it later */
+	ret = ring->wptr & ring->buf_mask;
+	amdgpu_ring_write(ring, 0);
 
 	return ret;
 }
 
-static void vpe_ring_patch_cond_exec(struct amdgpu_ring *ring, unsigned int offset)
-{
-	unsigned int cur;
-
-	WARN_ON_ONCE(offset > ring->buf_mask);
-	WARN_ON_ONCE(ring->ring[offset] != 0x55aa55aa);
-
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (cur > offset)
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->buf_mask + 1) - offset + cur;
-}
-
 static int vpe_ring_preempt_ib(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
@@ -695,16 +712,22 @@ static void vpe_ring_set_wptr(struct amdgpu_ring *ring)
 			upper_32_bits(ring->wptr << 2));
 		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, ring->wptr << 2);
 		WDOORBELL64(ring->doorbell_index, ring->wptr << 2);
+		if (vpe->collaborate_mode)
+			WDOORBELL64(ring->doorbell_index + 4, ring->wptr << 2);
 	} else {
-		dev_dbg(adev->dev, "Not using doorbell, \
-			regVPEC_QUEUE0_RB_WPTR == 0x%08x, \
-			regVPEC_QUEUE0_RB_WPTR_HI == 0x%08x\n",
-			lower_32_bits(ring->wptr << 2),
-			upper_32_bits(ring->wptr << 2));
-		WREG32(vpe_get_reg_offset(vpe, ring->me, vpe->regs.queue0_rb_wptr_lo),
-		       lower_32_bits(ring->wptr << 2));
-		WREG32(vpe_get_reg_offset(vpe, ring->me, vpe->regs.queue0_rb_wptr_hi),
-		       upper_32_bits(ring->wptr << 2));
+		int i;
+
+		for (i = 0; i < vpe->num_instances; i++) {
+			dev_dbg(adev->dev, "Not using doorbell, \
+				regVPEC_QUEUE0_RB_WPTR == 0x%08x, \
+				regVPEC_QUEUE0_RB_WPTR_HI == 0x%08x\n",
+				lower_32_bits(ring->wptr << 2),
+				upper_32_bits(ring->wptr << 2));
+			WREG32(vpe_get_reg_offset(vpe, i, vpe->regs.queue0_rb_wptr_lo),
+			       lower_32_bits(ring->wptr << 2));
+			WREG32(vpe_get_reg_offset(vpe, i, vpe->regs.queue0_rb_wptr_hi),
+			       upper_32_bits(ring->wptr << 2));
+		}
 	}
 }
 
@@ -864,7 +887,6 @@ static const struct amdgpu_ring_funcs vpe_ring_funcs = {
 	.test_ring = vpe_ring_test_ring,
 	.test_ib = vpe_ring_test_ib,
 	.init_cond_exec = vpe_ring_init_cond_exec,
-	.patch_cond_exec = vpe_ring_patch_cond_exec,
 	.preempt_ib = vpe_ring_preempt_ib,
 	.begin_use = vpe_ring_begin_use,
 	.end_use = vpe_ring_end_use,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h
index 1153ddaea64d..231d86d0953e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h
@@ -27,6 +27,8 @@
 #include "amdgpu_irq.h"
 #include "vpe_6_1_fw_if.h"
 
+#define AMDGPU_MAX_VPE_INSTANCES 2
+
 struct amdgpu_vpe;
 
 struct vpe_funcs {
@@ -74,6 +76,9 @@ struct amdgpu_vpe {
 	uint32_t			*cmdbuf_cpu_addr;
 	struct delayed_work		idle_work;
 	bool				context_started;
+
+	uint32_t			num_instances;
+	bool				collaborate_mode;
 };
 
 int amdgpu_vpe_psp_update_sram(struct amdgpu_device *adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index 691fa40e4e01..f90905ef32c7 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
@@ -3657,6 +3657,9 @@ static void gfx_v10_0_init_spm_golden_registers(struct amdgpu_device *adev)
 
 static void gfx_v10_0_init_golden_registers(struct amdgpu_device *adev)
 {
+	if (amdgpu_sriov_vf(adev))
+		return;
+
 	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 		soc15_program_register_sequence(adev,
@@ -4982,7 +4985,8 @@ static void gfx_v10_0_constants_init(struct amdgpu_device *adev)
 	u32 tmp;
 	int i;
 
-	WREG32_FIELD15(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff);
+	if (!amdgpu_sriov_vf(adev))
+		WREG32_FIELD15(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff);
 
 	gfx_v10_0_setup_rb(adev);
 	gfx_v10_0_get_cu_info(adev, &adev->gfx.cu_info);
@@ -7163,7 +7167,7 @@ static int gfx_v10_0_hw_init(void *handle)
 	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 3, 0))
 		gfx_v10_3_program_pbb_mode(adev);
 
-	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 3, 0))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 3, 0) && !amdgpu_sriov_vf(adev))
 		gfx_v10_3_set_power_brake_sequence(adev);
 
 	return r;
@@ -8542,34 +8546,23 @@ static void gfx_v10_0_ring_emit_cntxcntl(struct amdgpu_ring *ring,
 	amdgpu_ring_write(ring, 0);
 }
 
-static unsigned int gfx_v10_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned int gfx_v10_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring,
+						       uint64_t addr)
 {
 	unsigned int ret;
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, 0); /* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	/* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, 0);
 	ret = ring->wptr & ring->buf_mask;
-	amdgpu_ring_write(ring, 0x55aa55aa); /* patch dummy value later */
+	/* patch dummy value later */
+	amdgpu_ring_write(ring, 0);
 
 	return ret;
 }
 
-static void gfx_v10_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigned int offset)
-{
-	unsigned int cur;
-
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
-
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (likely(cur > offset))
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->buf_mask + 1) - offset + cur;
-}
-
 static int gfx_v10_0_ring_preempt_ib(struct amdgpu_ring *ring)
 {
 	int i, r = 0;
@@ -9224,7 +9217,6 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {
 	.emit_switch_buffer = gfx_v10_0_ring_emit_sb,
 	.emit_cntxcntl = gfx_v10_0_ring_emit_cntxcntl,
 	.init_cond_exec = gfx_v10_0_ring_emit_init_cond_exec,
-	.patch_cond_exec = gfx_v10_0_ring_emit_patch_cond_exec,
 	.preempt_ib = gfx_v10_0_ring_preempt_ib,
 	.emit_frame_cntl = gfx_v10_0_ring_emit_frame_cntl,
 	.emit_wreg = gfx_v10_0_ring_emit_wreg,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
index 0d90d60a21d6..1770e496c1b7 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
@@ -5461,6 +5461,11 @@ static void gfx_v11_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 		amdgpu_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0));
 		amdgpu_ring_write(ring, 0x0);
 	}
+
+	/* Make sure that we can't skip the SET_Q_MODE packets when the VM
+	 * changed in any way.
+	 */
+	ring->set_q_mode_ptr = NULL;
 }
 
 static void gfx_v11_0_ring_emit_fence_kiq(struct amdgpu_ring *ring, u64 addr,
@@ -5510,16 +5515,81 @@ static void gfx_v11_0_ring_emit_cntxcntl(struct amdgpu_ring *ring,
 	amdgpu_ring_write(ring, 0);
 }
 
+static unsigned gfx_v11_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring,
+						   uint64_t addr)
+{
+	unsigned ret;
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	/* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, 0);
+	ret = ring->wptr & ring->buf_mask;
+	/* patch dummy value later */
+	amdgpu_ring_write(ring, 0);
+
+	return ret;
+}
+
 static void gfx_v11_0_ring_emit_gfx_shadow(struct amdgpu_ring *ring,
 					   u64 shadow_va, u64 csa_va,
 					   u64 gds_va, bool init_shadow,
 					   int vmid)
 {
 	struct amdgpu_device *adev = ring->adev;
+	unsigned int offs, end;
 
-	if (!adev->gfx.cp_gfx_shadow)
+	if (!adev->gfx.cp_gfx_shadow || !ring->ring_obj)
 		return;
 
+	/*
+	 * The logic here isn't easy to understand because we need to keep state
+	 * accross multiple executions of the function as well as between the
+	 * CPU and GPU. The general idea is that the newly written GPU command
+	 * has a condition on the previous one and only executed if really
+	 * necessary.
+	 */
+
+	/*
+	 * The dw in the NOP controls if the next SET_Q_MODE packet should be
+	 * executed or not. Reserve 64bits just to be on the save side.
+	 */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_NOP, 1));
+	offs = ring->wptr & ring->buf_mask;
+
+	/*
+	 * We start with skipping the prefix SET_Q_MODE and always executing
+	 * the postfix SET_Q_MODE packet. This is changed below with a
+	 * WRITE_DATA command when the postfix executed.
+	 */
+	amdgpu_ring_write(ring, shadow_va ? 1 : 0);
+	amdgpu_ring_write(ring, 0);
+
+	if (ring->set_q_mode_offs) {
+		uint64_t addr;
+
+		addr = amdgpu_bo_gpu_offset(ring->ring_obj);
+		addr += ring->set_q_mode_offs << 2;
+		end = gfx_v11_0_ring_emit_init_cond_exec(ring, addr);
+	}
+
+	/*
+	 * When the postfix SET_Q_MODE packet executes we need to make sure that the
+	 * next prefix SET_Q_MODE packet executes as well.
+	 */
+	if (!shadow_va) {
+		uint64_t addr;
+
+		addr = amdgpu_bo_gpu_offset(ring->ring_obj);
+		addr += offs << 2;
+		amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+		amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM);
+		amdgpu_ring_write(ring, lower_32_bits(addr));
+		amdgpu_ring_write(ring, upper_32_bits(addr));
+		amdgpu_ring_write(ring, 0x1);
+	}
+
 	amdgpu_ring_write(ring, PACKET3(PACKET3_SET_Q_PREEMPTION_MODE, 7));
 	amdgpu_ring_write(ring, lower_32_bits(shadow_va));
 	amdgpu_ring_write(ring, upper_32_bits(shadow_va));
@@ -5531,33 +5601,26 @@ static void gfx_v11_0_ring_emit_gfx_shadow(struct amdgpu_ring *ring,
 			  PACKET3_SET_Q_PREEMPTION_MODE_IB_VMID(vmid) : 0);
 	amdgpu_ring_write(ring, init_shadow ?
 			  PACKET3_SET_Q_PREEMPTION_MODE_INIT_SHADOW_MEM : 0);
-}
 
-static unsigned gfx_v11_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring)
-{
-	unsigned ret;
+	if (ring->set_q_mode_offs)
+		amdgpu_ring_patch_cond_exec(ring, end);
 
-	amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, 0); /* discard following DWs if *cond_exec_gpu_addr==0 */
-	ret = ring->wptr & ring->buf_mask;
-	amdgpu_ring_write(ring, 0x55aa55aa); /* patch dummy value later */
+	if (shadow_va) {
+		uint64_t token = shadow_va ^ csa_va ^ gds_va ^ vmid;
 
-	return ret;
-}
+		/*
+		 * If the tokens match try to skip the last postfix SET_Q_MODE
+		 * packet to avoid saving/restoring the state all the time.
+		 */
+		if (ring->set_q_mode_ptr && ring->set_q_mode_token == token)
+			*ring->set_q_mode_ptr = 0;
 
-static void gfx_v11_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigned offset)
-{
-	unsigned cur;
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
+		ring->set_q_mode_token = token;
+	} else {
+		ring->set_q_mode_ptr = &ring->ring[ring->set_q_mode_offs];
+	}
 
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (likely(cur > offset))
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->buf_mask + 1) - offset + cur;
+	ring->set_q_mode_offs = offs;
 }
 
 static int gfx_v11_0_ring_preempt_ib(struct amdgpu_ring *ring)
@@ -6124,7 +6187,7 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_gfx = {
 	.emit_frame_size = /* totally 247 maximum if 16 IBs */
 		5 + /* update_spm_vmid */
 		5 + /* COND_EXEC */
-		9 + /* SET_Q_PREEMPTION_MODE */
+		22 + /* SET_Q_PREEMPTION_MODE */
 		7 + /* PIPELINE_SYNC */
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
@@ -6137,6 +6200,7 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_gfx = {
 		31 + /*	DE_META */
 		3 + /* CNTX_CTRL */
 		5 + /* HDP_INVL */
+		22 + /* SET_Q_PREEMPTION_MODE */
 		8 + 8 + /* FENCE x2 */
 		8, /* gfx_v11_0_emit_mem_sync */
 	.emit_ib_size =	4, /* gfx_v11_0_ring_emit_ib_gfx */
@@ -6153,7 +6217,6 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_gfx = {
 	.emit_cntxcntl = gfx_v11_0_ring_emit_cntxcntl,
 	.emit_gfx_shadow = gfx_v11_0_ring_emit_gfx_shadow,
 	.init_cond_exec = gfx_v11_0_ring_emit_init_cond_exec,
-	.patch_cond_exec = gfx_v11_0_ring_emit_patch_cond_exec,
 	.preempt_ib = gfx_v11_0_ring_preempt_ib,
 	.emit_frame_cntl = gfx_v11_0_ring_emit_frame_cntl,
 	.emit_wreg = gfx_v11_0_ring_emit_wreg,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
index b97ea62212b6..202ddda57f98 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
@@ -6326,33 +6326,22 @@ static void gfx_v8_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags)
 	amdgpu_ring_write(ring, 0);
 }
 
-static unsigned gfx_v8_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned gfx_v8_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring,
+						  uint64_t addr)
 {
 	unsigned ret;
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, 0); /* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	/* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, 0);
 	ret = ring->wptr & ring->buf_mask;
-	amdgpu_ring_write(ring, 0x55aa55aa); /* patch dummy value later */
+	/* patch dummy value later */
+	amdgpu_ring_write(ring, 0);
 	return ret;
 }
 
-static void gfx_v8_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigned offset)
-{
-	unsigned cur;
-
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
-
-	cur = (ring->wptr & ring->buf_mask) - 1;
-	if (likely(cur > offset))
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->ring_size >> 2) - offset + cur;
-}
-
 static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg,
 				    uint32_t reg_val_offs)
 {
@@ -6932,7 +6921,6 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = {
 	.emit_switch_buffer = gfx_v8_ring_emit_sb,
 	.emit_cntxcntl = gfx_v8_ring_emit_cntxcntl,
 	.init_cond_exec = gfx_v8_0_ring_emit_init_cond_exec,
-	.patch_cond_exec = gfx_v8_0_ring_emit_patch_cond_exec,
 	.emit_wreg = gfx_v8_0_ring_emit_wreg,
 	.soft_recovery = gfx_v8_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v8_0_emit_mem_sync,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index 7669f82aa1da..6f97a6d0e6d0 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
@@ -2080,7 +2080,7 @@ static int gfx_v9_0_sw_init(void *handle)
 		ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
 
 		/* disable scheduler on the real ring */
-		ring->no_scheduler = true;
+		ring->no_scheduler = adev->gfx.mcbp;
 		ring->vm_hub = AMDGPU_GFXHUB(0);
 		r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq,
 				     AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP,
@@ -2090,7 +2090,7 @@ static int gfx_v9_0_sw_init(void *handle)
 	}
 
 	/* set up the software rings */
-	if (adev->gfx.num_gfx_rings) {
+	if (adev->gfx.mcbp && adev->gfx.num_gfx_rings) {
 		for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++) {
 			ring = &adev->gfx.sw_gfx_ring[i];
 			ring->ring_obj = NULL;
@@ -2180,7 +2180,7 @@ static int gfx_v9_0_sw_fini(void *handle)
 	int i;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	if (adev->gfx.num_gfx_rings) {
+	if (adev->gfx.mcbp && adev->gfx.num_gfx_rings) {
 		for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++)
 			amdgpu_ring_fini(&adev->gfx.sw_gfx_ring[i]);
 		amdgpu_ring_mux_fini(&adev->gfx.muxer);
@@ -5610,31 +5610,21 @@ static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags)
 	amdgpu_ring_write(ring, 0);
 }
 
-static unsigned gfx_v9_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned gfx_v9_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring,
+						  uint64_t addr)
 {
 	unsigned ret;
 	amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, 0); /* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	/* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, 0);
 	ret = ring->wptr & ring->buf_mask;
-	amdgpu_ring_write(ring, 0x55aa55aa); /* patch dummy value later */
+	/* patch dummy value later */
+	amdgpu_ring_write(ring, 0);
 	return ret;
 }
 
-static void gfx_v9_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigned offset)
-{
-	unsigned cur;
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
-
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (likely(cur > offset))
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->ring_size>>2) - offset + cur;
-}
-
 static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg,
 				    uint32_t reg_val_offs)
 {
@@ -5909,11 +5899,14 @@ static int gfx_v9_0_eop_irq(struct amdgpu_device *adev,
 
 	switch (me_id) {
 	case 0:
-		if (adev->gfx.num_gfx_rings &&
-		    !amdgpu_mcbp_handle_trailing_fence_irq(&adev->gfx.muxer)) {
-			/* Fence signals are handled on the software rings*/
-			for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++)
-				amdgpu_fence_process(&adev->gfx.sw_gfx_ring[i]);
+		if (adev->gfx.num_gfx_rings) {
+			if (!adev->gfx.mcbp) {
+				amdgpu_fence_process(&adev->gfx.gfx_ring[0]);
+			} else if (!amdgpu_mcbp_handle_trailing_fence_irq(&adev->gfx.muxer)) {
+				/* Fence signals are handled on the software rings*/
+				for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++)
+					amdgpu_fence_process(&adev->gfx.sw_gfx_ring[i]);
+			}
 		}
 		break;
 	case 1:
@@ -6908,7 +6901,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = {
 	.emit_switch_buffer = gfx_v9_ring_emit_sb,
 	.emit_cntxcntl = gfx_v9_ring_emit_cntxcntl,
 	.init_cond_exec = gfx_v9_0_ring_emit_init_cond_exec,
-	.patch_cond_exec = gfx_v9_0_ring_emit_patch_cond_exec,
 	.preempt_ib = gfx_v9_0_ring_preempt_ib,
 	.emit_frame_cntl = gfx_v9_0_ring_emit_frame_cntl,
 	.emit_wreg = gfx_v9_0_ring_emit_wreg,
@@ -6963,7 +6955,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_sw_ring_funcs_gfx = {
 	.emit_switch_buffer = gfx_v9_ring_emit_sb,
 	.emit_cntxcntl = gfx_v9_ring_emit_cntxcntl,
 	.init_cond_exec = gfx_v9_0_ring_emit_init_cond_exec,
-	.patch_cond_exec = gfx_v9_0_ring_emit_patch_cond_exec,
 	.emit_frame_cntl = gfx_v9_0_ring_emit_frame_cntl,
 	.emit_wreg = gfx_v9_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v9_0_ring_emit_reg_wait,
@@ -7050,7 +7041,7 @@ static void gfx_v9_0_set_ring_funcs(struct amdgpu_device *adev)
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 		adev->gfx.gfx_ring[i].funcs = &gfx_v9_0_ring_funcs_gfx;
 
-	if (adev->gfx.num_gfx_rings) {
+	if (adev->gfx.mcbp && adev->gfx.num_gfx_rings) {
 		for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++)
 			adev->gfx.sw_gfx_ring[i].funcs = &gfx_v9_0_sw_ring_funcs_gfx;
 	}
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c
index cd0e8a321e46..17509f32f61a 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c
@@ -155,6 +155,9 @@ static void gfxhub_v2_1_init_system_aperture_regs(struct amdgpu_device *adev)
 {
 	uint64_t value;
 
+	if (amdgpu_sriov_vf(adev))
+		return;
+
 	/* Program the AGP BAR */
 	WREG32_SOC15(GC, 0, mmGCMC_VM_AGP_BASE, 0);
 	WREG32_SOC15(GC, 0, mmGCMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
index db89d13bd80d..d933e19e0cf5 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
@@ -105,7 +105,7 @@ static int gmc_v10_0_process_interrupt(struct amdgpu_device *adev,
 	struct amdgpu_vmhub *hub = &adev->vmhub[vmhub_index];
 	bool retry_fault = !!(entry->src_data[1] & 0x80);
 	bool write_fault = !!(entry->src_data[1] & 0x20);
-	struct amdgpu_task_info task_info;
+	struct amdgpu_task_info *task_info;
 	uint32_t status = 0;
 	u64 addr;
 
@@ -157,18 +157,22 @@ static int gmc_v10_0_process_interrupt(struct amdgpu_device *adev,
 	if (!printk_ratelimit())
 		return 0;
 
-	memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-	amdgpu_vm_get_task_info(adev, entry->pasid, &task_info);
-
 	dev_err(adev->dev,
-		"[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u, for process %s pid %d thread %s pid %d)\n",
+		"[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u)\n",
 		entry->vmid_src ? "mmhub" : "gfxhub",
-		entry->src_id, entry->ring_id, entry->vmid,
-		entry->pasid, task_info.process_name, task_info.tgid,
-		task_info.task_name, task_info.pid);
+		entry->src_id, entry->ring_id, entry->vmid, entry->pasid);
+	task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+	if (task_info) {
+		dev_err(adev->dev,
+			" in process %s pid %d thread %s pid %d\n",
+			task_info->process_name, task_info->tgid,
+			task_info->task_name, task_info->pid);
+		amdgpu_vm_put_task_info(task_info);
+	}
+
 	dev_err(adev->dev, "  in page starting at address 0x%016llx from client 0x%x (%s)\n",
-		addr, entry->client_id,
-		soc15_ih_clientid_name[entry->client_id]);
+			addr, entry->client_id,
+			soc15_ih_clientid_name[entry->client_id]);
 
 	if (!amdgpu_sriov_vf(adev))
 		hub->vmhub_funcs->print_l2_protection_fault_status(adev,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
index a3812f0036a0..527dc917e049 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
@@ -126,19 +126,24 @@ static int gmc_v11_0_process_interrupt(struct amdgpu_device *adev,
 	}
 
 	if (printk_ratelimit()) {
-		struct amdgpu_task_info task_info;
-
-		memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-		amdgpu_vm_get_task_info(adev, entry->pasid, &task_info);
+		struct amdgpu_task_info *task_info;
 
 		dev_err(adev->dev,
-			"[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u, for process %s pid %d thread %s pid %d)\n",
+			"[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u)\n",
 			entry->vmid_src ? "mmhub" : "gfxhub",
-			entry->src_id, entry->ring_id, entry->vmid,
-			entry->pasid, task_info.process_name, task_info.tgid,
-			task_info.task_name, task_info.pid);
+			entry->src_id, entry->ring_id, entry->vmid, entry->pasid);
+		task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+		if (task_info) {
+			dev_err(adev->dev,
+				" in process %s pid %d thread %s pid %d)\n",
+				task_info->process_name, task_info->tgid,
+				task_info->task_name, task_info->pid);
+			amdgpu_vm_put_task_info(task_info);
+		}
+
 		dev_err(adev->dev, "  in page starting at address 0x%016llx from client %d\n",
-			addr, entry->client_id);
+				addr, entry->client_id);
+
 		if (!amdgpu_sriov_vf(adev))
 			hub->vmhub_funcs->print_l2_protection_fault_status(adev, status);
 	}
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
index 969a9e867170..d20e5f20ee31 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
@@ -1445,18 +1445,24 @@ static int gmc_v8_0_process_interrupt(struct amdgpu_device *adev,
 		gmc_v8_0_set_fault_enable_default(adev, false);
 
 	if (printk_ratelimit()) {
-		struct amdgpu_task_info task_info;
+		struct amdgpu_task_info *task_info;
 
-		memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-		amdgpu_vm_get_task_info(adev, entry->pasid, &task_info);
+		dev_err(adev->dev, "GPU fault detected: %d 0x%08x\n",
+			entry->src_id, entry->src_data[0]);
+
+		task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+		if (task_info) {
+			dev_err(adev->dev, " for process %s pid %d thread %s pid %d\n",
+				task_info->process_name, task_info->tgid,
+				task_info->task_name, task_info->pid);
+			amdgpu_vm_put_task_info(task_info);
+		}
 
-		dev_err(adev->dev, "GPU fault detected: %d 0x%08x for process %s pid %d thread %s pid %d\n",
-			entry->src_id, entry->src_data[0], task_info.process_name,
-			task_info.tgid, task_info.task_name, task_info.pid);
 		dev_err(adev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x%08X\n",
-			addr);
+				addr);
 		dev_err(adev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n",
 			status);
+
 		gmc_v8_0_vm_decode_fault(adev, status, addr, mc_client,
 					 entry->pasid);
 	}
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index 1439e62e9378..47b63a4ce68b 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
@@ -549,7 +549,7 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev,
 	bool retry_fault = !!(entry->src_data[1] & 0x80);
 	bool write_fault = !!(entry->src_data[1] & 0x20);
 	uint32_t status = 0, cid = 0, rw = 0;
-	struct amdgpu_task_info task_info;
+	struct amdgpu_task_info *task_info;
 	struct amdgpu_vmhub *hub;
 	const char *mmhub_cid;
 	const char *hub_name;
@@ -626,15 +626,20 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev,
 	if (!printk_ratelimit())
 		return 0;
 
-	memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-	amdgpu_vm_get_task_info(adev, entry->pasid, &task_info);
-
 	dev_err(adev->dev,
-		"[%s] %s page fault (src_id:%u ring:%u vmid:%u pasid:%u, for process %s pid %d thread %s pid %d)\n",
-		hub_name, retry_fault ? "retry" : "no-retry",
-		entry->src_id, entry->ring_id, entry->vmid,
-		entry->pasid, task_info.process_name, task_info.tgid,
-		task_info.task_name, task_info.pid);
+		"[%s] %s page fault (src_id:%u ring:%u vmid:%u pasid:%u)\n", hub_name,
+		retry_fault ? "retry" : "no-retry",
+		entry->src_id, entry->ring_id, entry->vmid, entry->pasid);
+
+	task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+	if (task_info) {
+		dev_err(adev->dev,
+			" for process %s pid %d thread %s pid %d)\n",
+			task_info->process_name, task_info->tgid,
+			task_info->task_name, task_info->pid);
+		amdgpu_vm_put_task_info(task_info);
+	}
+
 	dev_err(adev->dev, "  in page starting at address 0x%016llx from IH client 0x%x (%s)\n",
 		addr, entry->client_id,
 		soc15_ih_clientid_name[entry->client_id]);
diff --git a/drivers/gpu/drm/amd/amdgpu/ih_v7_0.c b/drivers/gpu/drm/amd/amdgpu/ih_v7_0.c
index 16fe428c0722..7aed96fa10a9 100644
--- a/drivers/gpu/drm/amd/amdgpu/ih_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/ih_v7_0.c
@@ -418,6 +418,12 @@ static u32 ih_v7_0_get_wptr(struct amdgpu_device *adev,
 	tmp = RREG32_NO_KIQ(ih_regs->ih_rb_cntl);
 	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
 	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+
+	/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+	 * can be detected.
+	 */
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
 out:
 	return (wptr & ih->ptr_mask);
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_3.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_3.c
index b3961968c10c..238ea40c2450 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_3.c
@@ -99,16 +99,15 @@ mmhub_v3_3_print_l2_protection_fault_status(struct amdgpu_device *adev,
 	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(3, 3, 0):
 	case IP_VERSION(3, 3, 1):
-		mmhub_cid = mmhub_client_ids_v3_3[cid][rw];
+		mmhub_cid = cid < ARRAY_SIZE(mmhub_client_ids_v3_3) ?
+			    mmhub_client_ids_v3_3[cid][rw] :
+			    cid == 0x140 ? "UMSCH" : NULL;
 		break;
 	default:
 		mmhub_cid = NULL;
 		break;
 	}
 
-	if (!mmhub_cid && cid == 0x140)
-		mmhub_cid = "UMSCH";
-
 	dev_err(adev->dev, "\t Faulty UTCL2 client ID: %s (0x%x)\n",
 		mmhub_cid ? mmhub_cid : "unknown", cid);
 	dev_err(adev->dev, "\t MORE_FAULTS: 0x%lx\n",
diff --git a/drivers/gpu/drm/amd/amdgpu/nbif_v6_3_1.c b/drivers/gpu/drm/amd/amdgpu/nbif_v6_3_1.c
new file mode 100644
index 000000000000..96ed00ac81ac
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/nbif_v6_3_1.c
@@ -0,0 +1,495 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "amdgpu_atombios.h"
+#include "nbif_v6_3_1.h"
+
+#include "nbif/nbif_6_3_1_offset.h"
+#include "nbif/nbif_6_3_1_sh_mask.h"
+#include "pcie/pcie_6_1_0_offset.h"
+#include "pcie/pcie_6_1_0_sh_mask.h"
+#include <uapi/linux/kfd_ioctl.h>
+
+static void nbif_v6_3_1_remap_hdp_registers(struct amdgpu_device *adev)
+{
+	WREG32_SOC15(NBIO, 0, regBIF_BX0_REMAP_HDP_MEM_FLUSH_CNTL,
+		adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL);
+	WREG32_SOC15(NBIO, 0, regBIF_BX0_REMAP_HDP_REG_FLUSH_CNTL,
+		adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_REG_FLUSH_CNTL);
+}
+
+static u32 nbif_v6_3_1_get_rev_id(struct amdgpu_device *adev)
+{
+	u32 tmp = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_DEV0_EPF0_STRAP0);
+
+	tmp &= RCC_STRAP0_RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0_MASK;
+	tmp >>= RCC_STRAP0_RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0__SHIFT;
+
+	return tmp;
+}
+
+static void nbif_v6_3_1_mc_access_enable(struct amdgpu_device *adev, bool enable)
+{
+	if (enable)
+		WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_FB_EN,
+			     BIF_BX0_BIF_FB_EN__FB_READ_EN_MASK |
+			     BIF_BX0_BIF_FB_EN__FB_WRITE_EN_MASK);
+	else
+		WREG32_SOC15(NBIO, 0, regBIF_BX0_BIF_FB_EN, 0);
+}
+
+static u32 nbif_v6_3_1_get_memsize(struct amdgpu_device *adev)
+{
+	return RREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF0_RCC_CONFIG_MEMSIZE);
+}
+
+static void nbif_v6_3_1_sdma_doorbell_range(struct amdgpu_device *adev,
+					    int instance, bool use_doorbell,
+					    int doorbell_index,
+					    int doorbell_size)
+{
+	if (instance == 0) {
+		u32 doorbell_range = RREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_2_CTRL);
+
+		if (use_doorbell) {
+			doorbell_range = REG_SET_FIELD(doorbell_range,
+						       GDC_S2A0_S2A_DOORBELL_ENTRY_2_CTRL,
+						       S2A_DOORBELL_PORT2_ENABLE,
+						       0x1);
+			doorbell_range = REG_SET_FIELD(doorbell_range,
+						       GDC_S2A0_S2A_DOORBELL_ENTRY_2_CTRL,
+						       S2A_DOORBELL_PORT2_AWID,
+						       0xe);
+			doorbell_range = REG_SET_FIELD(doorbell_range,
+						       GDC_S2A0_S2A_DOORBELL_ENTRY_2_CTRL,
+						       S2A_DOORBELL_PORT2_RANGE_OFFSET,
+						       doorbell_index);
+			doorbell_range = REG_SET_FIELD(doorbell_range,
+						       GDC_S2A0_S2A_DOORBELL_ENTRY_2_CTRL,
+						       S2A_DOORBELL_PORT2_RANGE_SIZE,
+						       doorbell_size);
+			doorbell_range = REG_SET_FIELD(doorbell_range,
+						       GDC_S2A0_S2A_DOORBELL_ENTRY_2_CTRL,
+						       S2A_DOORBELL_PORT2_AWADDR_31_28_VALUE,
+						       0x3);
+		} else
+			doorbell_range = REG_SET_FIELD(doorbell_range,
+						       GDC_S2A0_S2A_DOORBELL_ENTRY_2_CTRL,
+						       S2A_DOORBELL_PORT2_RANGE_SIZE,
+						       0);
+
+		WREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_2_CTRL, doorbell_range);
+	}
+}
+
+static void nbif_v6_3_1_vcn_doorbell_range(struct amdgpu_device *adev,
+					   bool use_doorbell, int doorbell_index,
+					   int instance)
+{
+	u32 doorbell_range;
+
+	if (instance)
+		doorbell_range = RREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_5_CTRL);
+	else
+		doorbell_range = RREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_4_CTRL);
+
+	if (use_doorbell) {
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC_S2A0_S2A_DOORBELL_ENTRY_4_CTRL,
+					       S2A_DOORBELL_PORT4_ENABLE,
+					       0x1);
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC_S2A0_S2A_DOORBELL_ENTRY_4_CTRL,
+					       S2A_DOORBELL_PORT4_AWID,
+					       instance ? 0x7 : 0x4);
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC_S2A0_S2A_DOORBELL_ENTRY_4_CTRL,
+					       S2A_DOORBELL_PORT4_RANGE_OFFSET,
+					       doorbell_index);
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC_S2A0_S2A_DOORBELL_ENTRY_4_CTRL,
+					       S2A_DOORBELL_PORT4_RANGE_SIZE,
+					       8);
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC_S2A0_S2A_DOORBELL_ENTRY_4_CTRL,
+					       S2A_DOORBELL_PORT4_AWADDR_31_28_VALUE,
+					       instance ? 0x7 : 0x4);
+	} else
+		doorbell_range = REG_SET_FIELD(doorbell_range,
+					       GDC_S2A0_S2A_DOORBELL_ENTRY_4_CTRL,
+					       S2A_DOORBELL_PORT4_RANGE_SIZE,
+					       0);
+
+	if (instance)
+		WREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_5_CTRL, doorbell_range);
+	else
+		WREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_4_CTRL, doorbell_range);
+}
+
+static void nbif_v6_3_1_gc_doorbell_init(struct amdgpu_device *adev)
+{
+	WREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_0_CTRL, 0x30000007);
+	WREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_3_CTRL, 0x3000000d);
+}
+
+static void nbif_v6_3_1_enable_doorbell_aperture(struct amdgpu_device *adev,
+						 bool enable)
+{
+	WREG32_FIELD15_PREREG(NBIO, 0, RCC_DEV0_EPF0_RCC_DOORBELL_APER_EN,
+			BIF_DOORBELL_APER_EN, enable ? 1 : 0);
+}
+
+static void
+nbif_v6_3_1_enable_doorbell_selfring_aperture(struct amdgpu_device *adev,
+					      bool enable)
+{
+	u32 tmp = 0;
+
+	if (enable) {
+		tmp = REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL,
+				    DOORBELL_SELFRING_GPA_APER_EN, 1) |
+		      REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL,
+				    DOORBELL_SELFRING_GPA_APER_MODE, 1) |
+		      REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL,
+				    DOORBELL_SELFRING_GPA_APER_SIZE, 0);
+
+		WREG32_SOC15(NBIO, 0, regBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_BASE_LOW,
+			     lower_32_bits(adev->doorbell.base));
+		WREG32_SOC15(NBIO, 0, regBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_BASE_HIGH,
+			     upper_32_bits(adev->doorbell.base));
+	}
+
+	WREG32_SOC15(NBIO, 0, regBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, tmp);
+}
+
+static void nbif_v6_3_1_ih_doorbell_range(struct amdgpu_device *adev,
+					  bool use_doorbell, int doorbell_index)
+{
+	u32 ih_doorbell_range = RREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_1_CTRL);
+
+	if (use_doorbell) {
+		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
+						  GDC_S2A0_S2A_DOORBELL_ENTRY_1_CTRL,
+						  S2A_DOORBELL_PORT1_ENABLE,
+						  0x1);
+		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
+						  GDC_S2A0_S2A_DOORBELL_ENTRY_1_CTRL,
+						  S2A_DOORBELL_PORT1_AWID,
+						  0x0);
+		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
+						  GDC_S2A0_S2A_DOORBELL_ENTRY_1_CTRL,
+						  S2A_DOORBELL_PORT1_RANGE_OFFSET,
+						  doorbell_index);
+		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
+						  GDC_S2A0_S2A_DOORBELL_ENTRY_1_CTRL,
+						  S2A_DOORBELL_PORT1_RANGE_SIZE,
+						  2);
+		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
+						  GDC_S2A0_S2A_DOORBELL_ENTRY_1_CTRL,
+						  S2A_DOORBELL_PORT1_AWADDR_31_28_VALUE,
+						  0x0);
+	} else
+		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
+						  GDC_S2A0_S2A_DOORBELL_ENTRY_1_CTRL,
+						  S2A_DOORBELL_PORT1_RANGE_SIZE,
+						  0);
+
+	WREG32_SOC15(NBIO, 0, regGDC_S2A0_S2A_DOORBELL_ENTRY_1_CTRL, ih_doorbell_range);
+}
+
+static void nbif_v6_3_1_ih_control(struct amdgpu_device *adev)
+{
+	u32 interrupt_cntl;
+
+	/* setup interrupt control */
+	WREG32_SOC15(NBIO, 0, regBIF_BX0_INTERRUPT_CNTL2, adev->dummy_page_addr >> 8);
+
+	interrupt_cntl = RREG32_SOC15(NBIO, 0, regBIF_BX0_INTERRUPT_CNTL);
+	/*
+	 * BIF_BX0_INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=0 - dummy read disabled with msi, enabled without msi
+	 * BIF_BX0_INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=1 - dummy read controlled by IH_DUMMY_RD_EN
+	 */
+	interrupt_cntl = REG_SET_FIELD(interrupt_cntl, BIF_BX0_INTERRUPT_CNTL,
+				       IH_DUMMY_RD_OVERRIDE, 0);
+
+	/* BIF_BX0_INTERRUPT_CNTL__IH_REQ_NONSNOOP_EN_MASK=1 if ring is in non-cacheable memory, e.g., vram */
+	interrupt_cntl = REG_SET_FIELD(interrupt_cntl, BIF_BX0_INTERRUPT_CNTL,
+				       IH_REQ_NONSNOOP_EN, 0);
+
+	WREG32_SOC15(NBIO, 0, regBIF_BX0_INTERRUPT_CNTL, interrupt_cntl);
+}
+
+static void
+nbif_v6_3_1_update_medium_grain_clock_gating(struct amdgpu_device *adev,
+					     bool enable)
+{
+}
+
+static void
+nbif_v6_3_1_update_medium_grain_light_sleep(struct amdgpu_device *adev,
+					    bool enable)
+{
+}
+
+static void
+nbif_v6_3_1_get_clockgating_state(struct amdgpu_device *adev,
+				  u64 *flags)
+{
+}
+
+static u32 nbif_v6_3_1_get_hdp_flush_req_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_GPU_HDP_FLUSH_REQ);
+}
+
+static u32 nbif_v6_3_1_get_hdp_flush_done_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_GPU_HDP_FLUSH_DONE);
+}
+
+static u32 nbif_v6_3_1_get_pcie_index_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_RSMU_INDEX);
+}
+
+static u32 nbif_v6_3_1_get_pcie_data_offset(struct amdgpu_device *adev)
+{
+	return SOC15_REG_OFFSET(NBIO, 0, regBIF_BX_PF0_RSMU_DATA);
+}
+
+const struct nbio_hdp_flush_reg nbif_v6_3_1_hdp_flush_reg = {
+	.ref_and_mask_cp0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP0_MASK,
+	.ref_and_mask_cp1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP1_MASK,
+	.ref_and_mask_cp2 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP2_MASK,
+	.ref_and_mask_cp3 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP3_MASK,
+	.ref_and_mask_cp4 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP4_MASK,
+	.ref_and_mask_cp5 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP5_MASK,
+	.ref_and_mask_cp6 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP6_MASK,
+	.ref_and_mask_cp7 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP7_MASK,
+	.ref_and_mask_cp8 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP8_MASK,
+	.ref_and_mask_cp9 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP9_MASK,
+	.ref_and_mask_sdma0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA0_MASK,
+	.ref_and_mask_sdma1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA1_MASK,
+};
+
+static void nbif_v6_3_1_init_registers(struct amdgpu_device *adev)
+{
+	uint32_t data;
+
+	data = RREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF2_STRAP2);
+	data &= ~RCC_DEV0_EPF2_STRAP2__STRAP_NO_SOFT_RESET_DEV0_F2_MASK;
+	WREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF2_STRAP2, data);
+}
+
+static u32 nbif_v6_3_1_get_rom_offset(struct amdgpu_device *adev)
+{
+	u32 data, rom_offset;
+
+	data = RREG32_SOC15(NBIO, 0, regREGS_ROM_OFFSET_CTRL);
+	rom_offset = REG_GET_FIELD(data, REGS_ROM_OFFSET_CTRL, ROM_OFFSET);
+
+	return rom_offset;
+}
+
+#ifdef CONFIG_PCIEASPM
+static void nbif_v6_3_1_program_ltr(struct amdgpu_device *adev)
+{
+	uint32_t def, data;
+
+	def = RREG32_SOC15(NBIO, 0, regRCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL);
+	data = 0x35EB;
+	data &= ~RCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL__LTR_PRIV_MSG_DIS_IN_PM_NON_D0_MASK;
+	data &= ~RCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL__LTR_PRIV_RST_LTR_IN_DL_DOWN_MASK;
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regRCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL, data);
+
+	def = data = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP2);
+	data &= ~RCC_STRAP0_RCC_BIF_STRAP2__STRAP_LTR_IN_ASPML1_DIS_MASK;
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP2, data);
+
+	def = data = RREG32_SOC15(NBIO, 0, regBIF_CFG_DEV0_EPF0_DEVICE_CNTL2);
+	if (adev->pdev->ltr_path)
+		data |= BIF_CFG_DEV0_EPF0_DEVICE_CNTL2__LTR_EN_MASK;
+	else
+		data &= ~BIF_CFG_DEV0_EPF0_DEVICE_CNTL2__LTR_EN_MASK;
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regBIF_CFG_DEV0_EPF0_DEVICE_CNTL2, data);
+}
+#endif
+
+static void nbif_v6_3_1_program_aspm(struct amdgpu_device *adev)
+{
+#ifdef CONFIG_PCIEASPM
+	uint32_t def, data;
+
+	def = data = RREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL);
+	data &= ~PCIE_LC_CNTL__LC_L1_INACTIVITY_MASK;
+	data &= ~PCIE_LC_CNTL__LC_L0S_INACTIVITY_MASK;
+	data |= PCIE_LC_CNTL__LC_PMI_TO_L1_DIS_MASK;
+	if (def != data)
+		WREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL, data);
+
+	def = data = RREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL7);
+	data |= PCIE_LC_CNTL7__LC_NBIF_ASPM_INPUT_EN_MASK;
+	if (def != data)
+		WREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL7, data);
+
+	def = data = RREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL3);
+	data |= PCIE_LC_CNTL3__LC_DSC_DONT_ENTER_L23_AFTER_PME_ACK_MASK;
+	if (def != data)
+		WREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL3, data);
+
+	def = data = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP3);
+	data &= ~RCC_STRAP0_RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER_MASK;
+	data &= ~RCC_STRAP0_RCC_BIF_STRAP3__STRAP_VLINK_PM_L1_ENTRY_TIMER_MASK;
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP3, data);
+
+	def = data = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP5);
+	data &= ~RCC_STRAP0_RCC_BIF_STRAP5__STRAP_VLINK_LDN_ENTRY_TIMER_MASK;
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP5, data);
+
+	def = data = RREG32_SOC15(NBIO, 0, regBIF_CFG_DEV0_EPF0_DEVICE_CNTL2);
+	data &= ~BIF_CFG_DEV0_EPF0_DEVICE_CNTL2__LTR_EN_MASK;
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regBIF_CFG_DEV0_EPF0_DEVICE_CNTL2, data);
+
+	WREG32_SOC15(NBIO, 0, regBIF_CFG_DEV0_EPF0_PCIE_LTR_CAP, 0x10011001);
+
+#if 0
+	/* regPSWUSP0_PCIE_LC_CNTL2 should be replace by PCIE_LC_CNTL2 or someone else ? */
+	def = data = RREG32_SOC15(NBIO, 0, regPSWUSP0_PCIE_LC_CNTL2);
+	data |= PSWUSP0_PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L1_MASK |
+		PSWUSP0_PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L23_MASK;
+	data &= ~PSWUSP0_PCIE_LC_CNTL2__LC_RCV_L0_TO_RCV_L0S_DIS_MASK;
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regPSWUSP0_PCIE_LC_CNTL2, data);
+#endif
+	def = data = RREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL4);
+	data |= PCIE_LC_CNTL4__LC_L1_POWERDOWN_MASK;
+	if (def != data)
+		WREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL4, data);
+
+	def = data = RREG32_SOC15(PCIE, 0, regPCIE_LC_RXRECOVER_RXSTANDBY_CNTL);
+	data |= PCIE_LC_RXRECOVER_RXSTANDBY_CNTL__LC_RX_L0S_STANDBY_EN_MASK;
+	if (def != data)
+		WREG32_SOC15(PCIE, 0, regPCIE_LC_RXRECOVER_RXSTANDBY_CNTL, data);
+
+	nbif_v6_3_1_program_ltr(adev);
+
+	def = data = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP3);
+	data |= 0x5DE0 << RCC_STRAP0_RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER__SHIFT;
+	data |= 0x0010 << RCC_STRAP0_RCC_BIF_STRAP3__STRAP_VLINK_PM_L1_ENTRY_TIMER__SHIFT;
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP3, data);
+
+	def = data = RREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP5);
+	data |= 0x0010 << RCC_STRAP0_RCC_BIF_STRAP5__STRAP_VLINK_LDN_ENTRY_TIMER__SHIFT;
+	if (def != data)
+		WREG32_SOC15(NBIO, 0, regRCC_STRAP0_RCC_BIF_STRAP5, data);
+
+	def = data = RREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL);
+	data |= 0x0 << PCIE_LC_CNTL__LC_L0S_INACTIVITY__SHIFT;
+	data |= 0x9 << PCIE_LC_CNTL__LC_L1_INACTIVITY__SHIFT;
+	data &= ~PCIE_LC_CNTL__LC_PMI_TO_L1_DIS_MASK;
+	if (def != data)
+		WREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL, data);
+
+	def = data = RREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL3);
+	data &= ~PCIE_LC_CNTL3__LC_DSC_DONT_ENTER_L23_AFTER_PME_ACK_MASK;
+	if (def != data)
+		WREG32_SOC15(PCIE, 0, regPCIE_LC_CNTL3, data);
+#endif
+}
+
+const struct amdgpu_nbio_funcs nbif_v6_3_1_funcs = {
+	.get_hdp_flush_req_offset = nbif_v6_3_1_get_hdp_flush_req_offset,
+	.get_hdp_flush_done_offset = nbif_v6_3_1_get_hdp_flush_done_offset,
+	.get_pcie_index_offset = nbif_v6_3_1_get_pcie_index_offset,
+	.get_pcie_data_offset = nbif_v6_3_1_get_pcie_data_offset,
+	.get_rev_id = nbif_v6_3_1_get_rev_id,
+	.mc_access_enable = nbif_v6_3_1_mc_access_enable,
+	.get_memsize = nbif_v6_3_1_get_memsize,
+	.sdma_doorbell_range = nbif_v6_3_1_sdma_doorbell_range,
+	.vcn_doorbell_range = nbif_v6_3_1_vcn_doorbell_range,
+	.gc_doorbell_init = nbif_v6_3_1_gc_doorbell_init,
+	.enable_doorbell_aperture = nbif_v6_3_1_enable_doorbell_aperture,
+	.enable_doorbell_selfring_aperture = nbif_v6_3_1_enable_doorbell_selfring_aperture,
+	.ih_doorbell_range = nbif_v6_3_1_ih_doorbell_range,
+	.update_medium_grain_clock_gating = nbif_v6_3_1_update_medium_grain_clock_gating,
+	.update_medium_grain_light_sleep = nbif_v6_3_1_update_medium_grain_light_sleep,
+	.get_clockgating_state = nbif_v6_3_1_get_clockgating_state,
+	.ih_control = nbif_v6_3_1_ih_control,
+	.init_registers = nbif_v6_3_1_init_registers,
+	.remap_hdp_registers = nbif_v6_3_1_remap_hdp_registers,
+	.get_rom_offset = nbif_v6_3_1_get_rom_offset,
+	.program_aspm = nbif_v6_3_1_program_aspm,
+};
+
+
+static void nbif_v6_3_1_sriov_ih_doorbell_range(struct amdgpu_device *adev,
+						bool use_doorbell, int doorbell_index)
+{
+}
+
+static void nbif_v6_3_1_sriov_sdma_doorbell_range(struct amdgpu_device *adev,
+						  int instance, bool use_doorbell,
+						  int doorbell_index,
+						  int doorbell_size)
+{
+}
+
+static void nbif_v6_3_1_sriov_vcn_doorbell_range(struct amdgpu_device *adev,
+						 bool use_doorbell,
+						 int doorbell_index, int instance)
+{
+}
+
+static void nbif_v6_3_1_sriov_gc_doorbell_init(struct amdgpu_device *adev)
+{
+}
+
+const struct amdgpu_nbio_funcs nbif_v6_3_1_sriov_funcs = {
+	.get_hdp_flush_req_offset = nbif_v6_3_1_get_hdp_flush_req_offset,
+	.get_hdp_flush_done_offset = nbif_v6_3_1_get_hdp_flush_done_offset,
+	.get_pcie_index_offset = nbif_v6_3_1_get_pcie_index_offset,
+	.get_pcie_data_offset = nbif_v6_3_1_get_pcie_data_offset,
+	.get_rev_id = nbif_v6_3_1_get_rev_id,
+	.mc_access_enable = nbif_v6_3_1_mc_access_enable,
+	.get_memsize = nbif_v6_3_1_get_memsize,
+	.sdma_doorbell_range = nbif_v6_3_1_sriov_sdma_doorbell_range,
+	.vcn_doorbell_range = nbif_v6_3_1_sriov_vcn_doorbell_range,
+	.gc_doorbell_init = nbif_v6_3_1_sriov_gc_doorbell_init,
+	.enable_doorbell_aperture = nbif_v6_3_1_enable_doorbell_aperture,
+	.enable_doorbell_selfring_aperture = nbif_v6_3_1_enable_doorbell_selfring_aperture,
+	.ih_doorbell_range = nbif_v6_3_1_sriov_ih_doorbell_range,
+	.update_medium_grain_clock_gating = nbif_v6_3_1_update_medium_grain_clock_gating,
+	.update_medium_grain_light_sleep = nbif_v6_3_1_update_medium_grain_light_sleep,
+	.get_clockgating_state = nbif_v6_3_1_get_clockgating_state,
+	.ih_control = nbif_v6_3_1_ih_control,
+	.init_registers = nbif_v6_3_1_init_registers,
+	.remap_hdp_registers = nbif_v6_3_1_remap_hdp_registers,
+	.get_rom_offset = nbif_v6_3_1_get_rom_offset,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/nbif_v6_3_1.h b/drivers/gpu/drm/amd/amdgpu/nbif_v6_3_1.h
new file mode 100644
index 000000000000..b7f2e0d88905
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/nbif_v6_3_1.h
@@ -0,0 +1,33 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __NBIO_V6_3_1_H__
+#define __NBIO_V6_3_1_H__
+
+#include "soc15_common.h"
+
+extern const struct nbio_hdp_flush_reg nbif_v6_3_1_hdp_flush_reg;
+extern const struct amdgpu_nbio_funcs nbif_v6_3_1_funcs;
+extern const struct amdgpu_nbio_funcs nbif_v6_3_1_sriov_funcs;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
index 3d68dd5523c6..43775cb67ff5 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
@@ -2104,7 +2104,7 @@ static int sdma_v4_0_print_iv_entry(struct amdgpu_device *adev,
 					      struct amdgpu_iv_entry *entry)
 {
 	int instance;
-	struct amdgpu_task_info task_info;
+	struct amdgpu_task_info *task_info;
 	u64 addr;
 
 	instance = sdma_v4_0_irq_id_to_seq(entry->client_id);
@@ -2116,15 +2116,20 @@ static int sdma_v4_0_print_iv_entry(struct amdgpu_device *adev,
 	addr = (u64)entry->src_data[0] << 12;
 	addr |= ((u64)entry->src_data[1] & 0xf) << 44;
 
-	memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-	amdgpu_vm_get_task_info(adev, entry->pasid, &task_info);
-
 	dev_dbg_ratelimited(adev->dev,
-		   "[sdma%d] address:0x%016llx src_id:%u ring:%u vmid:%u "
-		   "pasid:%u, for process %s pid %d thread %s pid %d\n",
-		   instance, addr, entry->src_id, entry->ring_id, entry->vmid,
-		   entry->pasid, task_info.process_name, task_info.tgid,
-		   task_info.task_name, task_info.pid);
+			   "[sdma%d] address:0x%016llx src_id:%u ring:%u vmid:%u pasid:%u\n",
+			   instance, addr, entry->src_id, entry->ring_id, entry->vmid,
+			   entry->pasid);
+
+	task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+	if (task_info) {
+		dev_dbg_ratelimited(adev->dev,
+				    " for process %s pid %d thread %s pid %d\n",
+				    task_info->process_name, task_info->tgid,
+				    task_info->task_name, task_info->pid);
+		amdgpu_vm_put_task_info(task_info);
+	}
+
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index fec5a3d1c4bc..34237a1b1f2e 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -431,16 +431,11 @@ static void sdma_v4_4_2_inst_gfx_stop(struct amdgpu_device *adev,
 	struct amdgpu_ring *sdma[AMDGPU_MAX_SDMA_INSTANCES];
 	u32 doorbell_offset, doorbell;
 	u32 rb_cntl, ib_cntl;
-	int i, unset = 0;
+	int i;
 
 	for_each_inst(i, inst_mask) {
 		sdma[i] = &adev->sdma.instance[i].ring;
 
-		if ((adev->mman.buffer_funcs_ring == sdma[i]) && unset != 1) {
-			amdgpu_ttm_set_buffer_funcs_status(adev, false);
-			unset = 1;
-		}
-
 		rb_cntl = RREG32_SDMA(i, regSDMA_GFX_RB_CNTL);
 		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA_GFX_RB_CNTL, RB_ENABLE, 0);
 		WREG32_SDMA(i, regSDMA_GFX_RB_CNTL, rb_cntl);
@@ -487,20 +482,10 @@ static void sdma_v4_4_2_inst_rlc_stop(struct amdgpu_device *adev,
 static void sdma_v4_4_2_inst_page_stop(struct amdgpu_device *adev,
 				       uint32_t inst_mask)
 {
-	struct amdgpu_ring *sdma[AMDGPU_MAX_SDMA_INSTANCES];
 	u32 rb_cntl, ib_cntl;
 	int i;
-	bool unset = false;
 
 	for_each_inst(i, inst_mask) {
-		sdma[i] = &adev->sdma.instance[i].page;
-
-		if ((adev->mman.buffer_funcs_ring == sdma[i]) &&
-			(!unset)) {
-			amdgpu_ttm_set_buffer_funcs_status(adev, false);
-			unset = true;
-		}
-
 		rb_cntl = RREG32_SDMA(i, regSDMA_PAGE_RB_CNTL);
 		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA_PAGE_RB_CNTL,
 					RB_ENABLE, 0);
@@ -612,7 +597,7 @@ static uint32_t sdma_v4_4_2_rb_cntl(struct amdgpu_ring *ring, uint32_t rb_cntl)
 	/* Set ring buffer size in dwords */
 	uint32_t rb_bufsz = order_base_2(ring->ring_size / 4);
 
-	barrier(); /* work around https://bugs.llvm.org/show_bug.cgi?id=42576 */
+	barrier(); /* work around https://llvm.org/pr42576 */
 	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA_GFX_RB_CNTL, RB_SIZE, rb_bufsz);
 #ifdef __BIG_ENDIAN
 	rb_cntl = REG_SET_FIELD(rb_cntl, SDMA_GFX_RB_CNTL, RB_SWAP_ENABLE, 1);
@@ -950,13 +935,7 @@ static int sdma_v4_4_2_inst_start(struct amdgpu_device *adev,
 			r = amdgpu_ring_test_helper(page);
 			if (r)
 				return r;
-
-			if (adev->mman.buffer_funcs_ring == page)
-				amdgpu_ttm_set_buffer_funcs_status(adev, true);
 		}
-
-		if (adev->mman.buffer_funcs_ring == ring)
-			amdgpu_ttm_set_buffer_funcs_status(adev, true);
 	}
 
 	return r;
@@ -1644,7 +1623,7 @@ static int sdma_v4_4_2_print_iv_entry(struct amdgpu_device *adev,
 					      struct amdgpu_iv_entry *entry)
 {
 	int instance;
-	struct amdgpu_task_info task_info;
+	struct amdgpu_task_info *task_info;
 	u64 addr;
 
 	instance = sdma_v4_4_2_irq_id_to_seq(entry->client_id);
@@ -1656,15 +1635,19 @@ static int sdma_v4_4_2_print_iv_entry(struct amdgpu_device *adev,
 	addr = (u64)entry->src_data[0] << 12;
 	addr |= ((u64)entry->src_data[1] & 0xf) << 44;
 
-	memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-	amdgpu_vm_get_task_info(adev, entry->pasid, &task_info);
-
 	dev_dbg_ratelimited(adev->dev,
-		   "[sdma%d] address:0x%016llx src_id:%u ring:%u vmid:%u "
-		   "pasid:%u, for process %s pid %d thread %s pid %d\n",
-		   instance, addr, entry->src_id, entry->ring_id, entry->vmid,
-		   entry->pasid, task_info.process_name, task_info.tgid,
-		   task_info.task_name, task_info.pid);
+			    "[sdma%d] address:0x%016llx src_id:%u ring:%u vmid:%u pasid:%u\n",
+			    instance, addr, entry->src_id, entry->ring_id, entry->vmid,
+			    entry->pasid);
+
+	task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+	if (task_info) {
+		dev_dbg_ratelimited(adev->dev, " for process %s pid %d thread %s pid %d\n",
+				    task_info->process_name, task_info->tgid,
+				    task_info->task_name, task_info->pid);
+		amdgpu_vm_put_task_info(task_info);
+	}
+
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c
index 3c485e5a531a..883e8a1b8a40 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c
@@ -249,35 +249,23 @@ static int sdma_v5_0_init_microcode(struct amdgpu_device *adev)
 	return ret;
 }
 
-static unsigned sdma_v5_0_ring_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned sdma_v5_0_ring_init_cond_exec(struct amdgpu_ring *ring,
+					      uint64_t addr)
 {
 	unsigned ret;
 
 	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_COND_EXE));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
 	amdgpu_ring_write(ring, 1);
-	ret = ring->wptr & ring->buf_mask;/* this is the offset we need patch later */
-	amdgpu_ring_write(ring, 0x55aa55aa);/* insert dummy here and patch it later */
+	/* this is the offset we need patch later */
+	ret = ring->wptr & ring->buf_mask;
+	/* insert dummy here and patch it later */
+	amdgpu_ring_write(ring, 0);
 
 	return ret;
 }
 
-static void sdma_v5_0_ring_patch_cond_exec(struct amdgpu_ring *ring,
-					   unsigned offset)
-{
-	unsigned cur;
-
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
-
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (cur > offset)
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->buf_mask + 1) - offset + cur;
-}
-
 /**
  * sdma_v5_0_ring_get_rptr - get the current read pointer
  *
@@ -1780,7 +1768,6 @@ static const struct amdgpu_ring_funcs sdma_v5_0_ring_funcs = {
 	.emit_reg_wait = sdma_v5_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = sdma_v5_0_ring_emit_reg_write_reg_wait,
 	.init_cond_exec = sdma_v5_0_ring_init_cond_exec,
-	.patch_cond_exec = sdma_v5_0_ring_patch_cond_exec,
 	.preempt_ib = sdma_v5_0_ring_preempt_ib,
 };
 
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
index 0058f3f7cf6e..42f4bd250def 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
@@ -89,35 +89,23 @@ static u32 sdma_v5_2_get_reg_offset(struct amdgpu_device *adev, u32 instance, u3
 	return base + internal_offset;
 }
 
-static unsigned sdma_v5_2_ring_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned sdma_v5_2_ring_init_cond_exec(struct amdgpu_ring *ring,
+					      uint64_t addr)
 {
 	unsigned ret;
 
 	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_COND_EXE));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
 	amdgpu_ring_write(ring, 1);
-	ret = ring->wptr & ring->buf_mask;/* this is the offset we need patch later */
-	amdgpu_ring_write(ring, 0x55aa55aa);/* insert dummy here and patch it later */
+	/* this is the offset we need patch later */
+	ret = ring->wptr & ring->buf_mask;
+	/* insert dummy here and patch it later */
+	amdgpu_ring_write(ring, 0);
 
 	return ret;
 }
 
-static void sdma_v5_2_ring_patch_cond_exec(struct amdgpu_ring *ring,
-					   unsigned offset)
-{
-	unsigned cur;
-
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
-
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (cur > offset)
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->buf_mask + 1) - offset + cur;
-}
-
 /**
  * sdma_v5_2_ring_get_rptr - get the current read pointer
  *
@@ -1722,7 +1710,6 @@ static const struct amdgpu_ring_funcs sdma_v5_2_ring_funcs = {
 	.emit_reg_wait = sdma_v5_2_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = sdma_v5_2_ring_emit_reg_write_reg_wait,
 	.init_cond_exec = sdma_v5_2_ring_init_cond_exec,
-	.patch_cond_exec = sdma_v5_2_ring_patch_cond_exec,
 	.preempt_ib = sdma_v5_2_ring_preempt_ib,
 };
 
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
index 4874ded45653..361835a61f2e 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
@@ -80,35 +80,23 @@ static u32 sdma_v6_0_get_reg_offset(struct amdgpu_device *adev, u32 instance, u3
 	return base + internal_offset;
 }
 
-static unsigned sdma_v6_0_ring_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned sdma_v6_0_ring_init_cond_exec(struct amdgpu_ring *ring,
+					      uint64_t addr)
 {
 	unsigned ret;
 
 	amdgpu_ring_write(ring, SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_COND_EXE));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
 	amdgpu_ring_write(ring, 1);
-	ret = ring->wptr & ring->buf_mask;/* this is the offset we need patch later */
-	amdgpu_ring_write(ring, 0x55aa55aa);/* insert dummy here and patch it later */
+	/* this is the offset we need patch later */
+	ret = ring->wptr & ring->buf_mask;
+	/* insert dummy here and patch it later */
+	amdgpu_ring_write(ring, 0);
 
 	return ret;
 }
 
-static void sdma_v6_0_ring_patch_cond_exec(struct amdgpu_ring *ring,
-					   unsigned offset)
-{
-	unsigned cur;
-
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
-
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (cur > offset)
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->buf_mask + 1) - offset + cur;
-}
-
 /**
  * sdma_v6_0_ring_get_rptr - get the current read pointer
  *
@@ -1542,7 +1530,6 @@ static const struct amdgpu_ring_funcs sdma_v6_0_ring_funcs = {
 	.emit_reg_wait = sdma_v6_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = sdma_v6_0_ring_emit_reg_write_reg_wait,
 	.init_cond_exec = sdma_v6_0_ring_init_cond_exec,
-	.patch_cond_exec = sdma_v6_0_ring_patch_cond_exec,
 	.preempt_ib = sdma_v6_0_ring_preempt_ib,
 };
 
diff --git a/drivers/gpu/drm/amd/amdgpu/soc21.c b/drivers/gpu/drm/amd/amdgpu/soc21.c
index b92fd6760fa9..581a3bd11481 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc21.c
+++ b/drivers/gpu/drm/amd/amdgpu/soc21.c
@@ -388,6 +388,7 @@ soc21_asic_reset_method(struct amdgpu_device *adev)
 	case IP_VERSION(13, 0, 4):
 	case IP_VERSION(13, 0, 11):
 	case IP_VERSION(14, 0, 0):
+	case IP_VERSION(14, 0, 1):
 		return AMD_RESET_METHOD_MODE2;
 	default:
 		if (amdgpu_dpm_is_baco_supported(adev))
diff --git a/drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c b/drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c
index 8e7b763cfdb7..84368cf1e175 100644
--- a/drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/umsch_mm_v4_0.c
@@ -60,7 +60,7 @@ static int umsch_mm_v4_0_load_microcode(struct amdgpu_umsch_mm *umsch)
 
 	umsch->cmd_buf_curr_ptr = umsch->cmd_buf_ptr;
 
-	if (amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(4, 0, 5)) {
+	if (amdgpu_ip_version(adev, VCN_HWIP, 0) >= IP_VERSION(4, 0, 5)) {
 		WREG32_SOC15(VCN, 0, regUVD_IPX_DLDO_CONFIG,
 			1 << UVD_IPX_DLDO_CONFIG__ONO0_PWR_CONFIG__SHIFT);
 		SOC15_WAIT_ON_RREG(VCN, 0, regUVD_IPX_DLDO_STATUS,
@@ -248,7 +248,7 @@ static int umsch_mm_v4_0_ring_stop(struct amdgpu_umsch_mm *umsch)
 	data = REG_SET_FIELD(data, VCN_UMSCH_RB_DB_CTRL, EN, 0);
 	WREG32_SOC15(VCN, 0, regVCN_UMSCH_RB_DB_CTRL, data);
 
-	if (amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(4, 0, 5)) {
+	if (amdgpu_ip_version(adev, VCN_HWIP, 0) >= IP_VERSION(4, 0, 5)) {
 		WREG32_SOC15(VCN, 0, regUVD_IPX_DLDO_CONFIG,
 			2 << UVD_IPX_DLDO_CONFIG__ONO0_PWR_CONFIG__SHIFT);
 		SOC15_WAIT_ON_RREG(VCN, 0, regUVD_IPX_DLDO_STATUS,
@@ -271,6 +271,8 @@ static int umsch_mm_v4_0_set_hw_resources(struct amdgpu_umsch_mm *umsch)
 
 	set_hw_resources.vmid_mask_mm_vcn = umsch->vmid_mask_mm_vcn;
 	set_hw_resources.vmid_mask_mm_vpe = umsch->vmid_mask_mm_vpe;
+	set_hw_resources.collaboration_mask_vpe =
+		adev->vpe.collaborate_mode ? 0x3 : 0x0;
 	set_hw_resources.engine_mask = umsch->engine_mask;
 
 	set_hw_resources.vcn0_hqd_mask[0] = umsch->vcn0_hqd_mask;
@@ -346,6 +348,7 @@ static int umsch_mm_v4_0_add_queue(struct amdgpu_umsch_mm *umsch,
 	add_queue.h_queue = input_ptr->h_queue;
 	add_queue.vm_context_cntl = input_ptr->vm_context_cntl;
 	add_queue.is_context_suspended = input_ptr->is_context_suspended;
+	add_queue.collaboration_mode = adev->vpe.collaborate_mode ? 1 : 0;
 
 	add_queue.api_status.api_completion_fence_addr = umsch->ring.fence_drv.gpu_addr;
 	add_queue.api_status.api_completion_fence_value = ++umsch->ring.fence_drv.sync_seq;
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
index 25ba27151ac0..aaceecd558cf 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
@@ -304,7 +304,7 @@ static int vcn_v1_0_resume(void *handle)
  */
 static void vcn_v1_0_mc_resume_spg_mode(struct amdgpu_device *adev)
 {
-	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw[0]->size + 4);
 	uint32_t offset;
 
 	/* cache window 0: fw */
@@ -371,7 +371,7 @@ static void vcn_v1_0_mc_resume_spg_mode(struct amdgpu_device *adev)
 
 static void vcn_v1_0_mc_resume_dpg_mode(struct amdgpu_device *adev)
 {
-	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw[0]->size + 4);
 	uint32_t offset;
 
 	/* cache window 0: fw */
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c
index 18794394c5a0..e357d8cf0c01 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c
@@ -330,7 +330,7 @@ static int vcn_v2_0_resume(void *handle)
  */
 static void vcn_v2_0_mc_resume(struct amdgpu_device *adev)
 {
-	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw[0]->size + 4);
 	uint32_t offset;
 
 	if (amdgpu_sriov_vf(adev))
@@ -386,7 +386,7 @@ static void vcn_v2_0_mc_resume(struct amdgpu_device *adev)
 
 static void vcn_v2_0_mc_resume_dpg_mode(struct amdgpu_device *adev, bool indirect)
 {
-	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw[0]->size + 4);
 	uint32_t offset;
 
 	/* cache window 0: fw */
@@ -1878,7 +1878,7 @@ static int vcn_v2_0_start_sriov(struct amdgpu_device *adev)
 
 		init_table += header->vcn_table_offset;
 
-		size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+		size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw[0]->size + 4);
 
 		MMSCH_V2_0_INSERT_DIRECT_RD_MOD_WT(
 			SOC15_REG_OFFSET(UVD, i, mmUVD_STATUS),
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c
index aba403d71806..1cd8a94b0fbc 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c
@@ -414,13 +414,15 @@ static int vcn_v2_5_resume(void *handle)
  */
 static void vcn_v2_5_mc_resume(struct amdgpu_device *adev)
 {
-	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+	uint32_t size;
 	uint32_t offset;
 	int i;
 
 	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
 		if (adev->vcn.harvest_config & (1 << i))
 			continue;
+
+		size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw[i]->size + 4);
 		/* cache window 0: fw */
 		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 			WREG32_SOC15(VCN, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
@@ -469,7 +471,7 @@ static void vcn_v2_5_mc_resume(struct amdgpu_device *adev)
 
 static void vcn_v2_5_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_idx, bool indirect)
 {
-	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw[inst_idx]->size + 4);
 	uint32_t offset;
 
 	/* cache window 0: fw */
@@ -1240,7 +1242,7 @@ static int vcn_v2_5_sriov_start(struct amdgpu_device *adev)
 			SOC15_REG_OFFSET(VCN, i, mmUVD_STATUS),
 			~UVD_STATUS__UVD_BUSY, UVD_STATUS__UVD_BUSY);
 
-		size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+		size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw[i]->size + 4);
 		/* mc resume*/
 		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 			MMSCH_V1_0_INSERT_DIRECT_WT(
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
index e02af4de521c..8f82fb887e9c 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
@@ -449,7 +449,7 @@ static int vcn_v3_0_resume(void *handle)
  */
 static void vcn_v3_0_mc_resume(struct amdgpu_device *adev, int inst)
 {
-	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw[inst]->size + 4);
 	uint32_t offset;
 
 	/* cache window 0: fw */
@@ -499,7 +499,7 @@ static void vcn_v3_0_mc_resume(struct amdgpu_device *adev, int inst)
 
 static void vcn_v3_0_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_idx, bool indirect)
 {
-	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+	uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw[inst_idx]->size + 4);
 	uint32_t offset;
 
 	/* cache window 0: fw */
@@ -1332,7 +1332,7 @@ static int vcn_v3_0_start_sriov(struct amdgpu_device *adev)
 			mmUVD_STATUS),
 			~UVD_STATUS__UVD_BUSY, UVD_STATUS__UVD_BUSY);
 
-		cache_size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+		cache_size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw[i]->size + 4);
 
 		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 			MMSCH_V3_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i,
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c
index 8ab01ae919d2..832d15f7b5f6 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c
@@ -382,7 +382,7 @@ static void vcn_v4_0_mc_resume(struct amdgpu_device *adev, int inst)
 	uint32_t offset, size;
 	const struct common_firmware_header *hdr;
 
-	hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
+	hdr = (const struct common_firmware_header *)adev->vcn.fw[inst]->data;
 	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
 
 	/* cache window 0: fw */
@@ -442,7 +442,7 @@ static void vcn_v4_0_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_idx
 {
 	uint32_t offset, size;
 	const struct common_firmware_header *hdr;
-	hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
+	hdr = (const struct common_firmware_header *)adev->vcn.fw[inst_idx]->data;
 	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
 
 	/* cache window 0: fw */
@@ -1289,7 +1289,7 @@ static int vcn_v4_0_start_sriov(struct amdgpu_device *adev)
 			regUVD_STATUS),
 			~UVD_STATUS__UVD_BUSY, UVD_STATUS__UVD_BUSY);
 
-		cache_size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+		cache_size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw[i]->size + 4);
 
 		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 			MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, i,
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c
index 810bbfccd6f2..203fa988322b 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c
@@ -332,7 +332,7 @@ static void vcn_v4_0_3_mc_resume(struct amdgpu_device *adev, int inst_idx)
 	uint32_t offset, size, vcn_inst;
 	const struct common_firmware_header *hdr;
 
-	hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
+	hdr = (const struct common_firmware_header *)adev->vcn.fw[inst_idx]->data;
 	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
 
 	vcn_inst = GET_INST(VCN, inst_idx);
@@ -407,7 +407,7 @@ static void vcn_v4_0_3_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_i
 	uint32_t offset, size;
 	const struct common_firmware_header *hdr;
 
-	hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
+	hdr = (const struct common_firmware_header *)adev->vcn.fw[inst_idx]->data;
 	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
 
 	/* cache window 0: fw */
@@ -894,7 +894,7 @@ static int vcn_v4_0_3_start_sriov(struct amdgpu_device *adev)
 		MMSCH_V4_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCN, 0, regUVD_STATUS),
 			~UVD_STATUS__UVD_BUSY, UVD_STATUS__UVD_BUSY);
 
-		cache_size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4);
+		cache_size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw[i]->size + 4);
 
 		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 			MMSCH_V4_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCN, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c
index 0468955338b7..501e53e69f2a 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0_5.c
@@ -45,7 +45,7 @@
 #define mmUVD_DPG_LMA_DATA_BASE_IDX					regUVD_DPG_LMA_DATA_BASE_IDX
 
 #define VCN_VID_SOC_ADDRESS_2_0						0x1fb00
-#define VCN1_VID_SOC_ADDRESS_3_0					0x48300
+#define VCN1_VID_SOC_ADDRESS_3_0					(0x48300 + 0x38000)
 
 #define VCN_HARVEST_MMSCH							0
 
@@ -329,7 +329,7 @@ static void vcn_v4_0_5_mc_resume(struct amdgpu_device *adev, int inst)
 	uint32_t offset, size;
 	const struct common_firmware_header *hdr;
 
-	hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
+	hdr = (const struct common_firmware_header *)adev->vcn.fw[inst]->data;
 	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
 
 	/* cache window 0: fw */
@@ -390,7 +390,7 @@ static void vcn_v4_0_5_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_i
 	uint32_t offset, size;
 	const struct common_firmware_header *hdr;
 
-	hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
+	hdr = (const struct common_firmware_header *)adev->vcn.fw[inst_idx]->data;
 	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
 
 	/* cache window 0: fw */
@@ -486,7 +486,8 @@ static void vcn_v4_0_5_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_i
 
 	/* VCN global tiling registers */
 	WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET(
-		VCN, 0, regUVD_GFX10_ADDR_CONFIG), adev->gfx.config.gb_addr_config, 0, indirect);
+		VCN, inst_idx, regUVD_GFX10_ADDR_CONFIG),
+		adev->gfx.config.gb_addr_config, 0, indirect);
 }
 
 /**
@@ -911,7 +912,6 @@ static int vcn_v4_0_5_start_dpg_mode(struct amdgpu_device *adev, int inst_idx, b
 		VCN, inst_idx, regUVD_MASTINT_EN),
 		UVD_MASTINT_EN__VCPU_EN_MASK, 0, indirect);
 
-
 	if (indirect)
 		amdgpu_vcn_psp_update_sram(adev, inst_idx, 0);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.c
index d6ee9958ba5f..bc60c554eb32 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v5_0_0.c
@@ -290,7 +290,7 @@ static void vcn_v5_0_0_mc_resume(struct amdgpu_device *adev, int inst)
 	uint32_t offset, size;
 	const struct common_firmware_header *hdr;
 
-	hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
+	hdr = (const struct common_firmware_header *)adev->vcn.fw[inst]->data;
 	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
 
 	/* cache window 0: fw */
@@ -351,7 +351,7 @@ static void vcn_v5_0_0_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_i
 	uint32_t offset, size;
 	const struct common_firmware_header *hdr;
 
-	hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
+	hdr = (const struct common_firmware_header *)adev->vcn.fw[inst_idx]->data;
 	size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
 
 	/* cache window 0: fw */
diff --git a/drivers/gpu/drm/amd/amdgpu/vpe_6_1_fw_if.h b/drivers/gpu/drm/amd/amdgpu/vpe_6_1_fw_if.h
index 9b550deb48d3..47534dbbd137 100644
--- a/drivers/gpu/drm/amd/amdgpu/vpe_6_1_fw_if.h
+++ b/drivers/gpu/drm/amd/amdgpu/vpe_6_1_fw_if.h
@@ -40,7 +40,8 @@ enum VPE_CMD_OPCODE {
     VPE_CMD_OPCODE_POLL_REGMEM  = 0x8,
     VPE_CMD_OPCODE_COND_EXE     = 0x9,
     VPE_CMD_OPCODE_ATOMIC       = 0xA,
-    VPE_CMD_OPCODE_PLANE_FILL   = 0xB,
+    VPE_CMD_OPCODE_PRED_EXE     = 0xB,
+    VPE_CMD_OPCODE_COLLAB_SYNC  = 0xC,
     VPE_CMD_OPCODE_TIMESTAMP    = 0xD
 };
 
diff --git a/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.c b/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.c
index d20060a51e05..769eb8f7bb3c 100644
--- a/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/vpe_v6_1.c
@@ -33,14 +33,38 @@
 #include "vpe/vpe_6_1_0_sh_mask.h"
 
 MODULE_FIRMWARE("amdgpu/vpe_6_1_0.bin");
+MODULE_FIRMWARE("amdgpu/vpe_6_1_1.bin");
 
 #define VPE_THREAD1_UCODE_OFFSET	0x8000
 
+#define regVPEC_COLLABORATE_CNTL                                                0x0013
+#define regVPEC_COLLABORATE_CNTL_BASE_IDX                                       0
+#define VPEC_COLLABORATE_CNTL__COLLABORATE_MODE_EN__SHIFT                       0x0
+#define VPEC_COLLABORATE_CNTL__COLLABORATE_MODE_EN_MASK                         0x00000001L
+
+#define regVPEC_COLLABORATE_CFG                                                 0x0014
+#define regVPEC_COLLABORATE_CFG_BASE_IDX                                        0
+#define VPEC_COLLABORATE_CFG__MASTER_ID__SHIFT                                  0x0
+#define VPEC_COLLABORATE_CFG__MASTER_EN__SHIFT                                  0x3
+#define VPEC_COLLABORATE_CFG__SLAVE0_ID__SHIFT                                  0x4
+#define VPEC_COLLABORATE_CFG__SLAVE0_EN__SHIFT                                  0x7
+#define VPEC_COLLABORATE_CFG__MASTER_ID_MASK                                    0x00000007L
+#define VPEC_COLLABORATE_CFG__MASTER_EN_MASK                                    0x00000008L
+#define VPEC_COLLABORATE_CFG__SLAVE0_ID_MASK                                    0x00000070L
+#define VPEC_COLLABORATE_CFG__SLAVE0_EN_MASK                                    0x00000080L
+
+#define regVPEC_CNTL_6_1_1                                                      0x0016
+#define regVPEC_CNTL_6_1_1_BASE_IDX                                             0
+#define regVPEC_QUEUE_RESET_REQ_6_1_1                                           0x002c
+#define regVPEC_QUEUE_RESET_REQ_6_1_1_BASE_IDX                                  0
+#define regVPEC_PUB_DUMMY2_6_1_1                                                0x004c
+#define regVPEC_PUB_DUMMY2_6_1_1_BASE_IDX                                       0
+
 static uint32_t vpe_v6_1_get_reg_offset(struct amdgpu_vpe *vpe, uint32_t inst, uint32_t offset)
 {
 	uint32_t base;
 
-	base = vpe->ring.adev->reg_offset[VPE_HWIP][0][0];
+	base = vpe->ring.adev->reg_offset[VPE_HWIP][inst][0];
 
 	return base + offset;
 }
@@ -48,12 +72,14 @@ static uint32_t vpe_v6_1_get_reg_offset(struct amdgpu_vpe *vpe, uint32_t inst, u
 static void vpe_v6_1_halt(struct amdgpu_vpe *vpe, bool halt)
 {
 	struct amdgpu_device *adev = vpe->ring.adev;
-	uint32_t f32_cntl;
+	uint32_t i, f32_cntl;
 
-	f32_cntl = RREG32(vpe_get_reg_offset(vpe, 0, regVPEC_F32_CNTL));
-	f32_cntl = REG_SET_FIELD(f32_cntl, VPEC_F32_CNTL, HALT, halt ? 1 : 0);
-	f32_cntl = REG_SET_FIELD(f32_cntl, VPEC_F32_CNTL, TH1_RESET, halt ? 1 : 0);
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_F32_CNTL), f32_cntl);
+	for (i = 0; i < vpe->num_instances; i++) {
+		f32_cntl = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_F32_CNTL));
+		f32_cntl = REG_SET_FIELD(f32_cntl, VPEC_F32_CNTL, HALT, halt ? 1 : 0);
+		f32_cntl = REG_SET_FIELD(f32_cntl, VPEC_F32_CNTL, TH1_RESET, halt ? 1 : 0);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_F32_CNTL), f32_cntl);
+	}
 }
 
 static int vpe_v6_1_irq_init(struct amdgpu_vpe *vpe)
@@ -70,20 +96,58 @@ static int vpe_v6_1_irq_init(struct amdgpu_vpe *vpe)
 	return 0;
 }
 
+static void vpe_v6_1_set_collaborate_mode(struct amdgpu_vpe *vpe, bool enable)
+{
+	struct amdgpu_device *adev = vpe->ring.adev;
+	uint32_t vpe_colla_cntl, vpe_colla_cfg, i;
+
+	if (!vpe->collaborate_mode)
+		return;
+
+	for (i = 0; i < vpe->num_instances; i++) {
+		vpe_colla_cntl = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_COLLABORATE_CNTL));
+		vpe_colla_cntl = REG_SET_FIELD(vpe_colla_cntl, VPEC_COLLABORATE_CNTL,
+					       COLLABORATE_MODE_EN, enable ? 1 : 0);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_COLLABORATE_CNTL), vpe_colla_cntl);
+
+		vpe_colla_cfg = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_COLLABORATE_CFG));
+		vpe_colla_cfg = REG_SET_FIELD(vpe_colla_cfg, VPEC_COLLABORATE_CFG, MASTER_ID, 0);
+		vpe_colla_cfg = REG_SET_FIELD(vpe_colla_cfg, VPEC_COLLABORATE_CFG, MASTER_EN, enable ? 1 : 0);
+		vpe_colla_cfg = REG_SET_FIELD(vpe_colla_cfg, VPEC_COLLABORATE_CFG, SLAVE0_ID, 1);
+		vpe_colla_cfg = REG_SET_FIELD(vpe_colla_cfg, VPEC_COLLABORATE_CFG, SLAVE0_EN, enable ? 1 : 0);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_COLLABORATE_CFG), vpe_colla_cfg);
+	}
+}
+
 static int vpe_v6_1_load_microcode(struct amdgpu_vpe *vpe)
 {
 	struct amdgpu_device *adev = vpe->ring.adev;
 	const struct vpe_firmware_header_v1_0 *vpe_hdr;
 	const __le32 *data;
 	uint32_t ucode_offset[2], ucode_size[2];
-	uint32_t i, size_dw;
+	uint32_t i, j, size_dw;
 	uint32_t ret;
 
-	// disable UMSCH_INT_ENABLE
-	ret = RREG32(vpe_get_reg_offset(vpe, 0, regVPEC_CNTL));
-	ret = REG_SET_FIELD(ret, VPEC_CNTL, UMSCH_INT_ENABLE, 0);
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_CNTL), ret);
+	/* disable UMSCH_INT_ENABLE */
+	for (j = 0; j < vpe->num_instances; j++) {
+
+		if (amdgpu_ip_version(adev, VPE_HWIP, 0) == IP_VERSION(6, 1, 1))
+			ret = RREG32(vpe_get_reg_offset(vpe, j, regVPEC_CNTL_6_1_1));
+		else
+			ret = RREG32(vpe_get_reg_offset(vpe, j, regVPEC_CNTL));
+
+		ret = REG_SET_FIELD(ret, VPEC_CNTL, UMSCH_INT_ENABLE, 0);
 
+		if (amdgpu_ip_version(adev, VPE_HWIP, 0) == IP_VERSION(6, 1, 1))
+			WREG32(vpe_get_reg_offset(vpe, j, regVPEC_CNTL_6_1_1), ret);
+		else
+			WREG32(vpe_get_reg_offset(vpe, j, regVPEC_CNTL), ret);
+	}
+
+	/*
+	 * For VPE 6.1.1, still only need to add master's offset, and psp will apply it to slave as well.
+	 * Here use instance 0 as master.
+	 */
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 		uint32_t f32_offset, f32_cntl;
 
@@ -96,8 +160,7 @@ static int vpe_v6_1_load_microcode(struct amdgpu_vpe *vpe)
 		adev->vpe.cmdbuf_cpu_addr[1] = f32_cntl;
 
 		amdgpu_vpe_psp_update_sram(adev);
-
-		/* Config DPM */
+		vpe_v6_1_set_collaborate_mode(vpe, true);
 		amdgpu_vpe_configure_dpm(vpe);
 
 		return 0;
@@ -114,25 +177,26 @@ static int vpe_v6_1_load_microcode(struct amdgpu_vpe *vpe)
 
 	vpe_v6_1_halt(vpe, true);
 
-	for (i = 0; i < 2; i++) {
-		if (i > 0)
-			WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_UCODE_ADDR), VPE_THREAD1_UCODE_OFFSET);
-		else
-			WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_UCODE_ADDR), 0);
-
-		data = (const __le32 *)(adev->vpe.fw->data + ucode_offset[i]);
-		size_dw = ucode_size[i] / sizeof(__le32);
-
-		while (size_dw--) {
-			if (amdgpu_emu_mode && size_dw % 500 == 0)
-				msleep(1);
-			WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_UCODE_DATA), le32_to_cpup(data++));
+	for (j = 0; j < vpe->num_instances; j++) {
+		for (i = 0; i < 2; i++) {
+			if (i > 0)
+				WREG32(vpe_get_reg_offset(vpe, j, regVPEC_UCODE_ADDR), VPE_THREAD1_UCODE_OFFSET);
+			else
+				WREG32(vpe_get_reg_offset(vpe, j, regVPEC_UCODE_ADDR), 0);
+
+			data = (const __le32 *)(adev->vpe.fw->data + ucode_offset[i]);
+			size_dw = ucode_size[i] / sizeof(__le32);
+
+			while (size_dw--) {
+				if (amdgpu_emu_mode && size_dw % 500 == 0)
+					msleep(1);
+				WREG32(vpe_get_reg_offset(vpe, j, regVPEC_UCODE_DATA), le32_to_cpup(data++));
+			}
 		}
-
 	}
 
 	vpe_v6_1_halt(vpe, false);
-	/* Config DPM */
+	vpe_v6_1_set_collaborate_mode(vpe, true);
 	amdgpu_vpe_configure_dpm(vpe);
 
 	return 0;
@@ -142,68 +206,68 @@ static int vpe_v6_1_ring_start(struct amdgpu_vpe *vpe)
 {
 	struct amdgpu_ring *ring = &vpe->ring;
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t rb_bufsz, rb_cntl;
-	uint32_t ib_cntl;
 	uint32_t doorbell, doorbell_offset;
+	uint32_t rb_bufsz, rb_cntl;
+	uint32_t ib_cntl, i;
 	int ret;
 
-	rb_bufsz = order_base_2(ring->ring_size / 4);
-	rb_cntl = RREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_RB_CNTL));
-	rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RB_SIZE, rb_bufsz);
-	rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RB_PRIV, 1);
-	rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RB_VMID, 0);
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_RB_CNTL), rb_cntl);
-
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_RB_RPTR), 0);
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_RB_RPTR_HI), 0);
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_RB_WPTR), 0);
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_RB_WPTR_HI), 0);
-
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_RB_RPTR_ADDR_LO),
-	       lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC);
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_RB_RPTR_ADDR_HI),
-	       upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF);
-
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_RB_BASE), ring->gpu_addr >> 8);
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_RB_BASE_HI), ring->gpu_addr >> 40);
-
-	ring->wptr = 0;
-
-	/* before programing wptr to a less value, need set minor_ptr_update first */
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_MINOR_PTR_UPDATE), 1);
-
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_RB_WPTR), lower_32_bits(ring->wptr) << 2);
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_RB_WPTR_HI), upper_32_bits(ring->wptr) << 2);
-
-	/* set minor_ptr_update to 0 after wptr programed */
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_MINOR_PTR_UPDATE), 0);
-
-	doorbell = RREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_DOORBELL));
-	doorbell_offset = RREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_DOORBELL_OFFSET));
-
-	doorbell = REG_SET_FIELD(doorbell, VPEC_QUEUE0_DOORBELL, ENABLE, ring->use_doorbell ? 1 : 0);
-	doorbell_offset = REG_SET_FIELD(doorbell_offset, VPEC_QUEUE0_DOORBELL_OFFSET, OFFSET, ring->doorbell_index);
-
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_DOORBELL), doorbell);
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_DOORBELL_OFFSET), doorbell_offset);
-
-	adev->nbio.funcs->vpe_doorbell_range(adev, 0, ring->use_doorbell, ring->doorbell_index, 2);
-
-	rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1);
-	rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RB_ENABLE, 1);
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_RB_CNTL), rb_cntl);
-
-	ib_cntl = RREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_IB_CNTL));
-	ib_cntl = REG_SET_FIELD(ib_cntl, VPEC_QUEUE0_IB_CNTL, IB_ENABLE, 1);
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE0_IB_CNTL), ib_cntl);
-
-	ring->sched.ready = true;
+	for (i = 0; i < vpe->num_instances; i++) {
+		/* Set ring buffer size in dwords */
+		rb_bufsz = order_base_2(ring->ring_size / 4);
+		rb_cntl = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_CNTL));
+		rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RB_SIZE, rb_bufsz);
+		rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RB_PRIV, 1);
+		rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RB_VMID, 0);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_CNTL), rb_cntl);
+
+		/* Initialize the ring buffer's read and write pointers */
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_RPTR), 0);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_RPTR_HI), 0);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_WPTR), 0);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_WPTR_HI), 0);
+
+		/* set the wb address whether it's enabled or not */
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_RPTR_ADDR_LO),
+			lower_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFC);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_RPTR_ADDR_HI),
+			upper_32_bits(ring->rptr_gpu_addr) & 0xFFFFFFFF);
+
+		rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1);
+
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_BASE), ring->gpu_addr >> 8);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_BASE_HI), ring->gpu_addr >> 40);
+
+		ring->wptr = 0;
+
+		/* before programing wptr to a less value, need set minor_ptr_update first */
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_MINOR_PTR_UPDATE), 1);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_WPTR), lower_32_bits(ring->wptr) << 2);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_WPTR_HI), upper_32_bits(ring->wptr) << 2);
+		/* set minor_ptr_update to 0 after wptr programed */
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_MINOR_PTR_UPDATE), 0);
+
+		doorbell_offset = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_DOORBELL_OFFSET));
+		doorbell_offset = REG_SET_FIELD(doorbell_offset, VPEC_QUEUE0_DOORBELL_OFFSET, OFFSET, ring->doorbell_index + i*4);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_DOORBELL_OFFSET), doorbell_offset);
+
+		doorbell = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_DOORBELL));
+		doorbell = REG_SET_FIELD(doorbell, VPEC_QUEUE0_DOORBELL, ENABLE, ring->use_doorbell ? 1 : 0);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_DOORBELL), doorbell);
+
+		adev->nbio.funcs->vpe_doorbell_range(adev, i, ring->use_doorbell, ring->doorbell_index + i*4, 4);
+
+		rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1);
+		rb_cntl = REG_SET_FIELD(rb_cntl, VPEC_QUEUE0_RB_CNTL, RB_ENABLE, 1);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_RB_CNTL), rb_cntl);
+
+		ib_cntl = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_IB_CNTL));
+		ib_cntl = REG_SET_FIELD(ib_cntl, VPEC_QUEUE0_IB_CNTL, IB_ENABLE, 1);
+		WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE0_IB_CNTL), ib_cntl);
+	}
 
 	ret = amdgpu_ring_test_helper(ring);
-	if (ret) {
-		ring->sched.ready = false;
+	if (ret)
 		return ret;
-	}
 
 	return 0;
 }
@@ -211,17 +275,30 @@ static int vpe_v6_1_ring_start(struct amdgpu_vpe *vpe)
 static int vpe_v_6_1_ring_stop(struct amdgpu_vpe *vpe)
 {
 	struct amdgpu_device *adev = vpe->ring.adev;
-	uint32_t queue_reset;
+	uint32_t queue_reset, i;
 	int ret;
 
-	queue_reset = RREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE_RESET_REQ));
-	queue_reset = REG_SET_FIELD(queue_reset, VPEC_QUEUE_RESET_REQ, QUEUE0_RESET, 1);
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_QUEUE_RESET_REQ), queue_reset);
+	for (i = 0; i < vpe->num_instances; i++) {
+		if (amdgpu_ip_version(adev, VPE_HWIP, 0) == IP_VERSION(6, 1, 1))
+			queue_reset = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE_RESET_REQ_6_1_1));
+		else
+			queue_reset = RREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE_RESET_REQ));
+
+		queue_reset = REG_SET_FIELD(queue_reset, VPEC_QUEUE_RESET_REQ, QUEUE0_RESET, 1);
+
+		if (amdgpu_ip_version(adev, VPE_HWIP, 0) == IP_VERSION(6, 1, 1)) {
+			WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE_RESET_REQ_6_1_1), queue_reset);
+			ret = SOC15_WAIT_ON_RREG(VPE, i, regVPEC_QUEUE_RESET_REQ_6_1_1, 0,
+						 VPEC_QUEUE_RESET_REQ__QUEUE0_RESET_MASK);
+		} else {
+			WREG32(vpe_get_reg_offset(vpe, i, regVPEC_QUEUE_RESET_REQ), queue_reset);
+			ret = SOC15_WAIT_ON_RREG(VPE, i, regVPEC_QUEUE_RESET_REQ, 0,
+						 VPEC_QUEUE_RESET_REQ__QUEUE0_RESET_MASK);
+		}
 
-	ret = SOC15_WAIT_ON_RREG(VPE, 0, regVPEC_QUEUE_RESET_REQ, 0,
-				 VPEC_QUEUE_RESET_REQ__QUEUE0_RESET_MASK);
-	if (ret)
-		dev_err(adev->dev, "VPE queue reset failed\n");
+		if (ret)
+			dev_err(adev->dev, "VPE queue reset failed\n");
+	}
 
 	vpe->ring.sched.ready = false;
 
@@ -236,10 +313,18 @@ static int vpe_v6_1_set_trap_irq_state(struct amdgpu_device *adev,
 	struct amdgpu_vpe *vpe = &adev->vpe;
 	uint32_t vpe_cntl;
 
-	vpe_cntl = RREG32(vpe_get_reg_offset(vpe, 0, regVPEC_CNTL));
+	if (amdgpu_ip_version(adev, VPE_HWIP, 0) == IP_VERSION(6, 1, 1))
+		vpe_cntl = RREG32(vpe_get_reg_offset(vpe, 0, regVPEC_CNTL_6_1_1));
+	else
+		vpe_cntl = RREG32(vpe_get_reg_offset(vpe, 0, regVPEC_CNTL));
+
 	vpe_cntl = REG_SET_FIELD(vpe_cntl, VPEC_CNTL, TRAP_ENABLE,
 				 state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
-	WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_CNTL), vpe_cntl);
+
+	if (amdgpu_ip_version(adev, VPE_HWIP, 0) == IP_VERSION(6, 1, 1))
+		WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_CNTL_6_1_1), vpe_cntl);
+	else
+		WREG32(vpe_get_reg_offset(vpe, 0, regVPEC_CNTL), vpe_cntl);
 
 	return 0;
 }
@@ -264,13 +349,19 @@ static int vpe_v6_1_process_trap_irq(struct amdgpu_device *adev,
 
 static int vpe_v6_1_set_regs(struct amdgpu_vpe *vpe)
 {
+	struct amdgpu_device *adev = container_of(vpe, struct amdgpu_device, vpe);
+
 	vpe->regs.queue0_rb_rptr_lo = regVPEC_QUEUE0_RB_RPTR;
 	vpe->regs.queue0_rb_rptr_hi = regVPEC_QUEUE0_RB_RPTR_HI;
 	vpe->regs.queue0_rb_wptr_lo = regVPEC_QUEUE0_RB_WPTR;
 	vpe->regs.queue0_rb_wptr_hi = regVPEC_QUEUE0_RB_WPTR_HI;
 	vpe->regs.queue0_preempt = regVPEC_QUEUE0_PREEMPT;
 
-	vpe->regs.dpm_enable = regVPEC_PUB_DUMMY2;
+	if (amdgpu_ip_version(adev, VPE_HWIP, 0) == IP_VERSION(6, 1, 1))
+		vpe->regs.dpm_enable = regVPEC_PUB_DUMMY2_6_1_1;
+	else
+		vpe->regs.dpm_enable = regVPEC_PUB_DUMMY2;
+
 	vpe->regs.dpm_pratio = regVPEC_QUEUE6_DUMMY4;
 	vpe->regs.dpm_request_interval = regVPEC_QUEUE5_DUMMY3;
 	vpe->regs.dpm_decision_threshold = regVPEC_QUEUE5_DUMMY4;