Merge branch 'drm-next-4.20' of git://people.freedesktop.org/~agd5f/linux into drm-next

More new features and fixes for 4.20:
- Add dynamic powergating support for VCN on picasso
- Scheduler cleanup
- Vega20 support for KFD
- DC cleanups and bug fixes

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180927184348.2696-1-alexander.deucher@amd.com

55 files changed, 1011 insertions, 421 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 6cb35e3dab30..c05b39438663 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -146,6 +146,7 @@ extern int amdgpu_cik_support;
 #define AMDGPU_DEFAULT_GTT_SIZE_MB		3072ULL /* 3GB by default */
 #define AMDGPU_WAIT_IDLE_TIMEOUT_IN_MS	        3000
 #define AMDGPU_MAX_USEC_TIMEOUT			100000	/* 100 ms */
+#define AMDGPU_FENCE_JIFFIES_TIMEOUT		(HZ / 2)
 /* AMDGPU_IB_POOL_SIZE must be a power of 2 */
 #define AMDGPU_IB_POOL_SIZE			16
 #define AMDGPU_DEBUGFS_MAX_COMPONENTS		32
@@ -408,16 +409,25 @@ typedef enum _AMDGPU_DOORBELL64_ASSIGNMENT
 	AMDGPU_DOORBELL64_GFX_RING0               = 0x8b,
 
 	/*
-	 * Other graphics doorbells can be allocated here: from 0x8c to 0xef
+	 * Other graphics doorbells can be allocated here: from 0x8c to 0xdf
 	 * Graphics voltage island aperture 1
-	 * default non-graphics QWORD index is 0xF0 - 0xFF inclusive
+	 * default non-graphics QWORD index is 0xe0 - 0xFF inclusive
 	 */
 
-	/* sDMA engines */
-	AMDGPU_DOORBELL64_sDMA_ENGINE0            = 0xF0,
-	AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE0     = 0xF1,
-	AMDGPU_DOORBELL64_sDMA_ENGINE1            = 0xF2,
-	AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE1     = 0xF3,
+	/* sDMA engines  reserved from 0xe0 -oxef  */
+	AMDGPU_DOORBELL64_sDMA_ENGINE0            = 0xE0,
+	AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE0     = 0xE1,
+	AMDGPU_DOORBELL64_sDMA_ENGINE1            = 0xE8,
+	AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE1     = 0xE9,
+
+	/* For vega10 sriov, the sdma doorbell must be fixed as follow
+	 * to keep the same setting with host driver, or it will
+	 * happen conflicts
+	 */
+	AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE0            = 0xF0,
+	AMDGPU_VEGA10_DOORBELL64_sDMA_HI_PRI_ENGINE0     = 0xF1,
+	AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE1            = 0xF2,
+	AMDGPU_VEGA10_DOORBELL64_sDMA_HI_PRI_ENGINE1     = 0xF3,
 
 	/* Interrupt handler */
 	AMDGPU_DOORBELL64_IH                      = 0xF4,  /* For legacy interrupt ring buffer */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 0f9947edb12a..c31a8849e9f8 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -76,6 +76,7 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev)
 		kfd2kgd = amdgpu_amdkfd_gfx_8_0_get_functions();
 		break;
 	case CHIP_VEGA10:
+	case CHIP_VEGA20:
 	case CHIP_RAVEN:
 		kfd2kgd = amdgpu_amdkfd_gfx_9_0_get_functions();
 		break;
@@ -123,7 +124,7 @@ static void amdgpu_doorbell_get_kfd_info(struct amdgpu_device *adev,
 
 void amdgpu_amdkfd_device_init(struct amdgpu_device *adev)
 {
-	int i;
+	int i, n;
 	int last_valid_bit;
 	if (adev->kfd) {
 		struct kgd2kfd_shared_resources gpu_resources = {
@@ -162,7 +163,15 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev)
 				&gpu_resources.doorbell_physical_address,
 				&gpu_resources.doorbell_aperture_size,
 				&gpu_resources.doorbell_start_offset);
-		if (adev->asic_type >= CHIP_VEGA10) {
+
+		if (adev->asic_type < CHIP_VEGA10) {
+			kgd2kfd->device_init(adev->kfd, &gpu_resources);
+			return;
+		}
+
+		n = (adev->asic_type < CHIP_VEGA20) ? 2 : 8;
+
+		for (i = 0; i < n; i += 2) {
 			/* On SOC15 the BIF is involved in routing
 			 * doorbells using the low 12 bits of the
 			 * address. Communicate the assignments to
@@ -170,20 +179,31 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev)
 			 * process in case of 64-bit doorbells so we
 			 * can use each doorbell assignment twice.
 			 */
-			gpu_resources.sdma_doorbell[0][0] =
-				AMDGPU_DOORBELL64_sDMA_ENGINE0;
-			gpu_resources.sdma_doorbell[0][1] =
-				AMDGPU_DOORBELL64_sDMA_ENGINE0 + 0x200;
-			gpu_resources.sdma_doorbell[1][0] =
-				AMDGPU_DOORBELL64_sDMA_ENGINE1;
-			gpu_resources.sdma_doorbell[1][1] =
-				AMDGPU_DOORBELL64_sDMA_ENGINE1 + 0x200;
-			/* Doorbells 0x0f0-0ff and 0x2f0-2ff are reserved for
-			 * SDMA, IH and VCN. So don't use them for the CP.
-			 */
-			gpu_resources.reserved_doorbell_mask = 0x1f0;
-			gpu_resources.reserved_doorbell_val  = 0x0f0;
+			if (adev->asic_type == CHIP_VEGA10) {
+				gpu_resources.sdma_doorbell[0][i] =
+					AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE0 + (i >> 1);
+				gpu_resources.sdma_doorbell[0][i+1] =
+					AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE0 + 0x200 + (i >> 1);
+				gpu_resources.sdma_doorbell[1][i] =
+					AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE1 + (i >> 1);
+				gpu_resources.sdma_doorbell[1][i+1] =
+					AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE1 + 0x200 + (i >> 1);
+			} else {
+				gpu_resources.sdma_doorbell[0][i] =
+					AMDGPU_DOORBELL64_sDMA_ENGINE0 + (i >> 1);
+				gpu_resources.sdma_doorbell[0][i+1] =
+					AMDGPU_DOORBELL64_sDMA_ENGINE0 + 0x200 + (i >> 1);
+				gpu_resources.sdma_doorbell[1][i] =
+					AMDGPU_DOORBELL64_sDMA_ENGINE1 + (i >> 1);
+				gpu_resources.sdma_doorbell[1][i+1] =
+					AMDGPU_DOORBELL64_sDMA_ENGINE1 + 0x200 + (i >> 1);
+			}
 		}
+		/* Doorbells 0x0e0-0ff and 0x2e0-2ff are reserved for
+		 * SDMA, IH and VCN. So don't use them for the CP.
+		 */
+		gpu_resources.reserved_doorbell_mask = 0x1e0;
+		gpu_resources.reserved_doorbell_val  = 0x0e0;
 
 		kgd2kfd->device_init(adev->kfd, &gpu_resources);
 	}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index 056fc6ef6c63..8e0d4f7196b4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -174,7 +174,7 @@ void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev,
 				struct amdgpu_vm *vm);
 void amdgpu_amdkfd_gpuvm_destroy_process_vm(struct kgd_dev *kgd, void *vm);
 void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *vm);
-uint32_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm);
+uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm);
 int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 		struct kgd_dev *kgd, uint64_t va, uint64_t size,
 		void *vm, struct kgd_mem **mem,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
index b2e45c8e2e0d..244d9834a381 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
@@ -142,7 +142,7 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type);
 static void set_scratch_backing_va(struct kgd_dev *kgd,
 					uint64_t va, uint32_t vmid);
 static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-		uint32_t page_table_base);
+		uint64_t page_table_base);
 static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid);
 static int invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid);
 static uint32_t read_vmid_from_vmfault_reg(struct kgd_dev *kgd);
@@ -874,7 +874,7 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type)
 }
 
 static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-			uint32_t page_table_base)
+			uint64_t page_table_base)
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 
@@ -882,7 +882,8 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
 		pr_err("trying to set page table base for wrong VMID\n");
 		return;
 	}
-	WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vmid - 8, page_table_base);
+	WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vmid - 8,
+		lower_32_bits(page_table_base));
 }
 
 static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
index ea7c18ce7754..9f149914ad6c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
@@ -45,8 +45,6 @@ enum hqd_dequeue_request_type {
 	RESET_WAVES
 };
 
-struct vi_sdma_mqd;
-
 /*
  * Register access functions
  */
@@ -100,7 +98,7 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type);
 static void set_scratch_backing_va(struct kgd_dev *kgd,
 					uint64_t va, uint32_t vmid);
 static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-		uint32_t page_table_base);
+		uint64_t page_table_base);
 static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid);
 static int invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid);
 
@@ -282,7 +280,8 @@ static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)
 
 	lock_srbm(kgd, mec, pipe, 0, 0);
 
-	WREG32(mmCPC_INT_CNTL, CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK);
+	WREG32(mmCPC_INT_CNTL, CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK |
+			CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK);
 
 	unlock_srbm(kgd);
 
@@ -834,7 +833,7 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type)
 }
 
 static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-		uint32_t page_table_base)
+		uint64_t page_table_base)
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 
@@ -842,7 +841,8 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
 		pr_err("trying to set page table base for wrong VMID\n");
 		return;
 	}
-	WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vmid - 8, page_table_base);
+	WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vmid - 8,
+			lower_32_bits(page_table_base));
 }
 
 static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
index c9176537550b..42cb4c4e0929 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -138,7 +138,7 @@ static bool get_atc_vmid_pasid_mapping_valid(struct kgd_dev *kgd,
 static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd,
 		uint8_t vmid);
 static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-		uint32_t page_table_base);
+		uint64_t page_table_base);
 static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type);
 static void set_scratch_backing_va(struct kgd_dev *kgd,
 					uint64_t va, uint32_t vmid);
@@ -1013,11 +1013,10 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type)
 }
 
 static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-		uint32_t page_table_base)
+		uint64_t page_table_base)
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-	uint64_t base = (uint64_t)page_table_base << PAGE_SHIFT |
-		AMDGPU_PTE_VALID;
+	uint64_t base = page_table_base | AMDGPU_PTE_VALID;
 
 	if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) {
 		pr_err("trying to set page table base for wrong VMID %u\n",
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index 6ee9dc476c86..df0a059565f9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -1131,11 +1131,15 @@ void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *vm)
 	amdgpu_vm_release_compute(adev, avm);
 }
 
-uint32_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm)
+uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm)
 {
 	struct amdgpu_vm *avm = (struct amdgpu_vm *)vm;
+	struct amdgpu_bo *pd = avm->root.base.bo;
+	struct amdgpu_device *adev = amdgpu_ttm_adev(pd->tbo.bdev);
 
-	return avm->pd_phys_addr >> AMDGPU_GPU_PAGE_SHIFT;
+	if (adev->asic_type < CHIP_VEGA10)
+		return avm->pd_phys_addr >> AMDGPU_GPU_PAGE_SHIFT;
+	return avm->pd_phys_addr;
 }
 
 int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
index 176f28777f5e..5448cf27654e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
@@ -196,6 +196,19 @@ int amdgpu_fence_emit_polling(struct amdgpu_ring *ring, uint32_t *s)
 }
 
 /**
+ * amdgpu_fence_schedule_fallback - schedule fallback check
+ *
+ * @ring: pointer to struct amdgpu_ring
+ *
+ * Start a timer as fallback to our interrupts.
+ */
+static void amdgpu_fence_schedule_fallback(struct amdgpu_ring *ring)
+{
+	mod_timer(&ring->fence_drv.fallback_timer,
+		  jiffies + AMDGPU_FENCE_JIFFIES_TIMEOUT);
+}
+
+/**
  * amdgpu_fence_process - check for fence activity
  *
  * @ring: pointer to struct amdgpu_ring
@@ -203,8 +216,10 @@ int amdgpu_fence_emit_polling(struct amdgpu_ring *ring, uint32_t *s)
  * Checks the current fence value and calculates the last
  * signalled fence value. Wakes the fence queue if the
  * sequence number has increased.
+ *
+ * Returns true if fence was processed
  */
-void amdgpu_fence_process(struct amdgpu_ring *ring)
+bool amdgpu_fence_process(struct amdgpu_ring *ring)
 {
 	struct amdgpu_fence_driver *drv = &ring->fence_drv;
 	uint32_t seq, last_seq;
@@ -216,8 +231,12 @@ void amdgpu_fence_process(struct amdgpu_ring *ring)
 
 	} while (atomic_cmpxchg(&drv->last_seq, last_seq, seq) != last_seq);
 
+	if (del_timer(&ring->fence_drv.fallback_timer) &&
+	    seq != ring->fence_drv.sync_seq)
+		amdgpu_fence_schedule_fallback(ring);
+
 	if (unlikely(seq == last_seq))
-		return;
+		return false;
 
 	last_seq &= drv->num_fences_mask;
 	seq &= drv->num_fences_mask;
@@ -244,6 +263,24 @@ void amdgpu_fence_process(struct amdgpu_ring *ring)
 
 		dma_fence_put(fence);
 	} while (last_seq != seq);
+
+	return true;
+}
+
+/**
+ * amdgpu_fence_fallback - fallback for hardware interrupts
+ *
+ * @work: delayed work item
+ *
+ * Checks for fence activity.
+ */
+static void amdgpu_fence_fallback(struct timer_list *t)
+{
+	struct amdgpu_ring *ring = from_timer(ring, t,
+					      fence_drv.fallback_timer);
+
+	if (amdgpu_fence_process(ring))
+		DRM_WARN("Fence fallback timer expired on ring %s\n", ring->name);
 }
 
 /**
@@ -393,6 +430,8 @@ int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring,
 	atomic_set(&ring->fence_drv.last_seq, 0);
 	ring->fence_drv.initialized = false;
 
+	timer_setup(&ring->fence_drv.fallback_timer, amdgpu_fence_fallback, 0);
+
 	ring->fence_drv.num_fences_mask = num_hw_submission * 2 - 1;
 	spin_lock_init(&ring->fence_drv.lock);
 	ring->fence_drv.fences = kcalloc(num_hw_submission * 2, sizeof(void *),
@@ -468,6 +507,7 @@ void amdgpu_fence_driver_fini(struct amdgpu_device *adev)
 		amdgpu_irq_put(adev, ring->fence_drv.irq_src,
 			       ring->fence_drv.irq_type);
 		drm_sched_fini(&ring->sched);
+		del_timer_sync(&ring->fence_drv.fallback_timer);
 		for (j = 0; j <= ring->fence_drv.num_fences_mask; ++j)
 			dma_fence_put(ring->fence_drv.fences[j]);
 		kfree(ring->fence_drv.fences);
@@ -561,6 +601,27 @@ static const char *amdgpu_fence_get_timeline_name(struct dma_fence *f)
 }
 
 /**
+ * amdgpu_fence_enable_signaling - enable signalling on fence
+ * @fence: fence
+ *
+ * This function is called with fence_queue lock held, and adds a callback
+ * to fence_queue that checks if this fence is signaled, and if so it
+ * signals the fence and removes itself.
+ */
+static bool amdgpu_fence_enable_signaling(struct dma_fence *f)
+{
+	struct amdgpu_fence *fence = to_amdgpu_fence(f);
+	struct amdgpu_ring *ring = fence->ring;
+
+	if (!timer_pending(&ring->fence_drv.fallback_timer))
+		amdgpu_fence_schedule_fallback(ring);
+
+	DMA_FENCE_TRACE(&fence->base, "armed on ring %i!\n", ring->idx);
+
+	return true;
+}
+
+/**
  * amdgpu_fence_free - free up the fence memory
  *
  * @rcu: RCU callback head
@@ -590,6 +651,7 @@ static void amdgpu_fence_release(struct dma_fence *f)
 static const struct dma_fence_ops amdgpu_fence_ops = {
 	.get_driver_name = amdgpu_fence_get_driver_name,
 	.get_timeline_name = amdgpu_fence_get_timeline_name,
+	.enable_signaling = amdgpu_fence_enable_signaling,
 	.release = amdgpu_fence_release,
 };
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c
index 4ed86218cef3..8af67f649660 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c
@@ -24,46 +24,21 @@
 #include <drm/drmP.h>
 #include "amdgpu.h"
 #include "amdgpu_ih.h"
-#include "amdgpu_amdkfd.h"
-
-/**
- * amdgpu_ih_ring_alloc - allocate memory for the IH ring
- *
- * @adev: amdgpu_device pointer
- *
- * Allocate a ring buffer for the interrupt controller.
- * Returns 0 for success, errors for failure.
- */
-static int amdgpu_ih_ring_alloc(struct amdgpu_device *adev)
-{
-	int r;
-
-	/* Allocate ring buffer */
-	if (adev->irq.ih.ring_obj == NULL) {
-		r = amdgpu_bo_create_kernel(adev, adev->irq.ih.ring_size,
-					    PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT,
-					    &adev->irq.ih.ring_obj,
-					    &adev->irq.ih.gpu_addr,
-					    (void **)&adev->irq.ih.ring);
-		if (r) {
-			DRM_ERROR("amdgpu: failed to create ih ring buffer (%d).\n", r);
-			return r;
-		}
-	}
-	return 0;
-}
 
 /**
  * amdgpu_ih_ring_init - initialize the IH state
  *
  * @adev: amdgpu_device pointer
+ * @ih: ih ring to initialize
+ * @ring_size: ring size to allocate
+ * @use_bus_addr: true when we can use dma_alloc_coherent
  *
  * Initializes the IH state and allocates a buffer
  * for the IH ring buffer.
  * Returns 0 for success, errors for failure.
  */
-int amdgpu_ih_ring_init(struct amdgpu_device *adev, unsigned ring_size,
-			bool use_bus_addr)
+int amdgpu_ih_ring_init(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih,
+			unsigned ring_size, bool use_bus_addr)
 {
 	u32 rb_bufsz;
 	int r;
@@ -71,70 +46,76 @@ int amdgpu_ih_ring_init(struct amdgpu_device *adev, unsigned ring_size,
 	/* Align ring size */
 	rb_bufsz = order_base_2(ring_size / 4);
 	ring_size = (1 << rb_bufsz) * 4;
-	adev->irq.ih.ring_size = ring_size;
-	adev->irq.ih.ptr_mask = adev->irq.ih.ring_size - 1;
-	adev->irq.ih.rptr = 0;
-	adev->irq.ih.use_bus_addr = use_bus_addr;
-
-	if (adev->irq.ih.use_bus_addr) {
-		if (!adev->irq.ih.ring) {
-			/* add 8 bytes for the rptr/wptr shadows and
-			 * add them to the end of the ring allocation.
-			 */
-			adev->irq.ih.ring = pci_alloc_consistent(adev->pdev,
-								 adev->irq.ih.ring_size + 8,
-								 &adev->irq.ih.rb_dma_addr);
-			if (adev->irq.ih.ring == NULL)
-				return -ENOMEM;
-			memset((void *)adev->irq.ih.ring, 0, adev->irq.ih.ring_size + 8);
-			adev->irq.ih.wptr_offs = (adev->irq.ih.ring_size / 4) + 0;
-			adev->irq.ih.rptr_offs = (adev->irq.ih.ring_size / 4) + 1;
-		}
-		return 0;
+	ih->ring_size = ring_size;
+	ih->ptr_mask = ih->ring_size - 1;
+	ih->rptr = 0;
+	ih->use_bus_addr = use_bus_addr;
+
+	if (use_bus_addr) {
+		if (ih->ring)
+			return 0;
+
+		/* add 8 bytes for the rptr/wptr shadows and
+		 * add them to the end of the ring allocation.
+		 */
+		ih->ring = dma_alloc_coherent(adev->dev, ih->ring_size + 8,
+					      &ih->rb_dma_addr, GFP_KERNEL);
+		if (ih->ring == NULL)
+			return -ENOMEM;
+
+		memset((void *)ih->ring, 0, ih->ring_size + 8);
+		ih->wptr_offs = (ih->ring_size / 4) + 0;
+		ih->rptr_offs = (ih->ring_size / 4) + 1;
 	} else {
-		r = amdgpu_device_wb_get(adev, &adev->irq.ih.wptr_offs);
+		r = amdgpu_device_wb_get(adev, &ih->wptr_offs);
+		if (r)
+			return r;
+
+		r = amdgpu_device_wb_get(adev, &ih->rptr_offs);
 		if (r) {
-			dev_err(adev->dev, "(%d) ih wptr_offs wb alloc failed\n", r);
+			amdgpu_device_wb_free(adev, ih->wptr_offs);
 			return r;
 		}
 
-		r = amdgpu_device_wb_get(adev, &adev->irq.ih.rptr_offs);
+		r = amdgpu_bo_create_kernel(adev, ih->ring_size, PAGE_SIZE,
+					    AMDGPU_GEM_DOMAIN_GTT,
+					    &ih->ring_obj, &ih->gpu_addr,
+					    (void **)&ih->ring);
 		if (r) {
-			amdgpu_device_wb_free(adev, adev->irq.ih.wptr_offs);
-			dev_err(adev->dev, "(%d) ih rptr_offs wb alloc failed\n", r);
+			amdgpu_device_wb_free(adev, ih->rptr_offs);
+			amdgpu_device_wb_free(adev, ih->wptr_offs);
 			return r;
 		}
-
-		return amdgpu_ih_ring_alloc(adev);
 	}
+	return 0;
 }
 
 /**
  * amdgpu_ih_ring_fini - tear down the IH state
  *
  * @adev: amdgpu_device pointer
+ * @ih: ih ring to tear down
  *
  * Tears down the IH state and frees buffer
  * used for the IH ring buffer.
  */
-void amdgpu_ih_ring_fini(struct amdgpu_device *adev)
+void amdgpu_ih_ring_fini(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih)
 {
-	if (adev->irq.ih.use_bus_addr) {
-		if (adev->irq.ih.ring) {
-			/* add 8 bytes for the rptr/wptr shadows and
-			 * add them to the end of the ring allocation.
-			 */
-			pci_free_consistent(adev->pdev, adev->irq.ih.ring_size + 8,
-					    (void *)adev->irq.ih.ring,
-					    adev->irq.ih.rb_dma_addr);
-			adev->irq.ih.ring = NULL;
-		}
+	if (ih->use_bus_addr) {
+		if (!ih->ring)
+			return;
+
+		/* add 8 bytes for the rptr/wptr shadows and
+		 * add them to the end of the ring allocation.
+		 */
+		dma_free_coherent(adev->dev, ih->ring_size + 8,
+				  (void *)ih->ring, ih->rb_dma_addr);
+		ih->ring = NULL;
 	} else {
-		amdgpu_bo_free_kernel(&adev->irq.ih.ring_obj,
-				      &adev->irq.ih.gpu_addr,
-				      (void **)&adev->irq.ih.ring);
-		amdgpu_device_wb_free(adev, adev->irq.ih.wptr_offs);
-		amdgpu_device_wb_free(adev, adev->irq.ih.rptr_offs);
+		amdgpu_bo_free_kernel(&ih->ring_obj, &ih->gpu_addr,
+				      (void **)&ih->ring);
+		amdgpu_device_wb_free(adev, ih->wptr_offs);
+		amdgpu_device_wb_free(adev, ih->rptr_offs);
 	}
 }
 
@@ -142,56 +123,43 @@ void amdgpu_ih_ring_fini(struct amdgpu_device *adev)
  * amdgpu_ih_process - interrupt handler
  *
  * @adev: amdgpu_device pointer
+ * @ih: ih ring to process
  *
  * Interrupt hander (VI), walk the IH ring.
  * Returns irq process return code.
  */
-int amdgpu_ih_process(struct amdgpu_device *adev)
+int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih,
+		      void (*callback)(struct amdgpu_device *adev,
+				       struct amdgpu_ih_ring *ih))
 {
-	struct amdgpu_iv_entry entry;
 	u32 wptr;
 
-	if (!adev->irq.ih.enabled || adev->shutdown)
+	if (!ih->enabled || adev->shutdown)
 		return IRQ_NONE;
 
 	wptr = amdgpu_ih_get_wptr(adev);
 
 restart_ih:
 	/* is somebody else already processing irqs? */
-	if (atomic_xchg(&adev->irq.ih.lock, 1))
+	if (atomic_xchg(&ih->lock, 1))
 		return IRQ_NONE;
 
-	DRM_DEBUG("%s: rptr %d, wptr %d\n", __func__, adev->irq.ih.rptr, wptr);
+	DRM_DEBUG("%s: rptr %d, wptr %d\n", __func__, ih->rptr, wptr);
 
 	/* Order reading of wptr vs. reading of IH ring data */
 	rmb();
 
-	while (adev->irq.ih.rptr != wptr) {
-		u32 ring_index = adev->irq.ih.rptr >> 2;
-
-		/* Prescreening of high-frequency interrupts */
-		if (!amdgpu_ih_prescreen_iv(adev)) {
-			adev->irq.ih.rptr &= adev->irq.ih.ptr_mask;
-			continue;
-		}
-
-		/* Before dispatching irq to IP blocks, send it to amdkfd */
-		amdgpu_amdkfd_interrupt(adev,
-				(const void *) &adev->irq.ih.ring[ring_index]);
-
-		entry.iv_entry = (const uint32_t *)
-			&adev->irq.ih.ring[ring_index];
-		amdgpu_ih_decode_iv(adev, &entry);
-		adev->irq.ih.rptr &= adev->irq.ih.ptr_mask;
-
-		amdgpu_irq_dispatch(adev, &entry);
+	while (ih->rptr != wptr) {
+		callback(adev, ih);
+		ih->rptr &= ih->ptr_mask;
 	}
+
 	amdgpu_ih_set_rptr(adev);
-	atomic_set(&adev->irq.ih.lock, 0);
+	atomic_set(&ih->lock, 0);
 
 	/* make sure wptr hasn't changed while processing */
 	wptr = amdgpu_ih_get_wptr(adev);
-	if (wptr != adev->irq.ih.rptr)
+	if (wptr != ih->rptr)
 		goto restart_ih;
 
 	return IRQ_HANDLED;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h
index 0d5b3f5201d2..9ce8c93ec19b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h
@@ -24,12 +24,8 @@
 #ifndef __AMDGPU_IH_H__
 #define __AMDGPU_IH_H__
 
-#include "soc15_ih_clientid.h"
-
 struct amdgpu_device;
-
-#define AMDGPU_IH_CLIENTID_LEGACY 0
-#define AMDGPU_IH_CLIENTID_MAX SOC15_IH_CLIENTID_MAX
+struct amdgpu_iv_entry;
 
 /*
  * R6xx+ IH ring
@@ -51,22 +47,6 @@ struct amdgpu_ih_ring {
 	dma_addr_t		rb_dma_addr; /* only used when use_bus_addr = true */
 };
 
-#define AMDGPU_IH_SRC_DATA_MAX_SIZE_DW 4
-
-struct amdgpu_iv_entry {
-	unsigned client_id;
-	unsigned src_id;
-	unsigned ring_id;
-	unsigned vmid;
-	unsigned vmid_src;
-	uint64_t timestamp;
-	unsigned timestamp_src;
-	unsigned pasid;
-	unsigned pasid_src;
-	unsigned src_data[AMDGPU_IH_SRC_DATA_MAX_SIZE_DW];
-	const uint32_t *iv_entry;
-};
-
 /* provided by the ih block */
 struct amdgpu_ih_funcs {
 	/* ring read/write ptr handling, called from interrupt context */
@@ -82,9 +62,11 @@ struct amdgpu_ih_funcs {
 #define amdgpu_ih_decode_iv(adev, iv) (adev)->irq.ih_funcs->decode_iv((adev), (iv))
 #define amdgpu_ih_set_rptr(adev) (adev)->irq.ih_funcs->set_rptr((adev))
 
-int amdgpu_ih_ring_init(struct amdgpu_device *adev, unsigned ring_size,
-			bool use_bus_addr);
-void amdgpu_ih_ring_fini(struct amdgpu_device *adev);
-int amdgpu_ih_process(struct amdgpu_device *adev);
+int amdgpu_ih_ring_init(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih,
+			unsigned ring_size, bool use_bus_addr);
+void amdgpu_ih_ring_fini(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih);
+int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih,
+		      void (*callback)(struct amdgpu_device *adev,
+				       struct amdgpu_ih_ring *ih));
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
index b927e8798534..52c17f6219a7 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
@@ -51,6 +51,7 @@
 #include "atom.h"
 #include "amdgpu_connectors.h"
 #include "amdgpu_trace.h"
+#include "amdgpu_amdkfd.h"
 
 #include <linux/pm_runtime.h>
 
@@ -123,7 +124,7 @@ void amdgpu_irq_disable_all(struct amdgpu_device *adev)
 	int r;
 
 	spin_lock_irqsave(&adev->irq.lock, irqflags);
-	for (i = 0; i < AMDGPU_IH_CLIENTID_MAX; ++i) {
+	for (i = 0; i < AMDGPU_IRQ_CLIENTID_MAX; ++i) {
 		if (!adev->irq.client[i].sources)
 			continue;
 
@@ -147,6 +148,34 @@ void amdgpu_irq_disable_all(struct amdgpu_device *adev)
 }
 
 /**
+ * amdgpu_irq_callback - callback from the IH ring
+ *
+ * @adev: amdgpu device pointer
+ * @ih: amdgpu ih ring
+ *
+ * Callback from IH ring processing to handle the entry at the current position
+ * and advance the read pointer.
+ */
+static void amdgpu_irq_callback(struct amdgpu_device *adev,
+				struct amdgpu_ih_ring *ih)
+{
+	u32 ring_index = ih->rptr >> 2;
+	struct amdgpu_iv_entry entry;
+
+	/* Prescreening of high-frequency interrupts */
+	if (!amdgpu_ih_prescreen_iv(adev))
+		return;
+
+	/* Before dispatching irq to IP blocks, send it to amdkfd */
+	amdgpu_amdkfd_interrupt(adev, (const void *) &ih->ring[ring_index]);
+
+	entry.iv_entry = (const uint32_t *)&ih->ring[ring_index];
+	amdgpu_ih_decode_iv(adev, &entry);
+
+	amdgpu_irq_dispatch(adev, &entry);
+}
+
+/**
  * amdgpu_irq_handler - IRQ handler
  *
  * @irq: IRQ number (unused)
@@ -163,7 +192,7 @@ irqreturn_t amdgpu_irq_handler(int irq, void *arg)
 	struct amdgpu_device *adev = dev->dev_private;
 	irqreturn_t ret;
 
-	ret = amdgpu_ih_process(adev);
+	ret = amdgpu_ih_process(adev, &adev->irq.ih, amdgpu_irq_callback);
 	if (ret == IRQ_HANDLED)
 		pm_runtime_mark_last_busy(dev->dev);
 	return ret;
@@ -273,7 +302,7 @@ void amdgpu_irq_fini(struct amdgpu_device *adev)
 		cancel_work_sync(&adev->reset_work);
 	}
 
-	for (i = 0; i < AMDGPU_IH_CLIENTID_MAX; ++i) {
+	for (i = 0; i < AMDGPU_IRQ_CLIENTID_MAX; ++i) {
 		if (!adev->irq.client[i].sources)
 			continue;
 
@@ -313,7 +342,7 @@ int amdgpu_irq_add_id(struct amdgpu_device *adev,
 		      unsigned client_id, unsigned src_id,
 		      struct amdgpu_irq_src *source)
 {
-	if (client_id >= AMDGPU_IH_CLIENTID_MAX)
+	if (client_id >= AMDGPU_IRQ_CLIENTID_MAX)
 		return -EINVAL;
 
 	if (src_id >= AMDGPU_MAX_IRQ_SRC_ID)
@@ -367,7 +396,7 @@ void amdgpu_irq_dispatch(struct amdgpu_device *adev,
 
 	trace_amdgpu_iv(entry);
 
-	if (client_id >= AMDGPU_IH_CLIENTID_MAX) {
+	if (client_id >= AMDGPU_IRQ_CLIENTID_MAX) {
 		DRM_DEBUG("Invalid client_id in IV: %d\n", client_id);
 		return;
 	}
@@ -440,7 +469,7 @@ void amdgpu_irq_gpu_reset_resume_helper(struct amdgpu_device *adev)
 {
 	int i, j, k;
 
-	for (i = 0; i < AMDGPU_IH_CLIENTID_MAX; ++i) {
+	for (i = 0; i < AMDGPU_IRQ_CLIENTID_MAX; ++i) {
 		if (!adev->irq.client[i].sources)
 			continue;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
index 3375ad778edc..f6ce171cb8aa 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
@@ -25,19 +25,38 @@
 #define __AMDGPU_IRQ_H__
 
 #include <linux/irqdomain.h>
+#include "soc15_ih_clientid.h"
 #include "amdgpu_ih.h"
 
-#define AMDGPU_MAX_IRQ_SRC_ID	0x100
+#define AMDGPU_MAX_IRQ_SRC_ID		0x100
 #define AMDGPU_MAX_IRQ_CLIENT_ID	0x100
 
+#define AMDGPU_IRQ_CLIENTID_LEGACY	0
+#define AMDGPU_IRQ_CLIENTID_MAX		SOC15_IH_CLIENTID_MAX
+
+#define AMDGPU_IRQ_SRC_DATA_MAX_SIZE_DW	4
+
 struct amdgpu_device;
-struct amdgpu_iv_entry;
 
 enum amdgpu_interrupt_state {
 	AMDGPU_IRQ_STATE_DISABLE,
 	AMDGPU_IRQ_STATE_ENABLE,
 };
 
+struct amdgpu_iv_entry {
+	unsigned client_id;
+	unsigned src_id;
+	unsigned ring_id;
+	unsigned vmid;
+	unsigned vmid_src;
+	uint64_t timestamp;
+	unsigned timestamp_src;
+	unsigned pasid;
+	unsigned pasid_src;
+	unsigned src_data[AMDGPU_IRQ_SRC_DATA_MAX_SIZE_DW];
+	const uint32_t *iv_entry;
+};
+
 struct amdgpu_irq_src {
 	unsigned				num_types;
 	atomic_t				*enabled_types;
@@ -63,7 +82,7 @@ struct amdgpu_irq {
 	bool				installed;
 	spinlock_t			lock;
 	/* interrupt sources */
-	struct amdgpu_irq_client	client[AMDGPU_IH_CLIENTID_MAX];
+	struct amdgpu_irq_client	client[AMDGPU_IRQ_CLIENTID_MAX];
 
 	/* status, etc. */
 	bool				msi_enabled; /* msi enabled */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c
index 8c334fc808c2..18d989e0e362 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c
@@ -1976,6 +1976,7 @@ void amdgpu_pm_compute_clocks(struct amdgpu_device *adev)
 static int amdgpu_debugfs_pm_info_pp(struct seq_file *m, struct amdgpu_device *adev)
 {
 	uint32_t value;
+	uint64_t value64;
 	uint32_t query = 0;
 	int size;
 
@@ -2014,6 +2015,10 @@ static int amdgpu_debugfs_pm_info_pp(struct seq_file *m, struct amdgpu_device *a
 		seq_printf(m, "GPU Load: %u %%\n", value);
 	seq_printf(m, "\n");
 
+	/* SMC feature mask */
+	if (!amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_ENABLED_SMC_FEATURES_MASK, (void *)&value64, &size))
+		seq_printf(m, "SMC Feature Mask: 0x%016llx\n", value64);
+
 	/* UVD clocks */
 	if (!amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_UVD_POWER, (void *)&value, &size)) {
 		if (!value) {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
index 44fc665e4577..4caa301ce454 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
@@ -77,6 +77,7 @@ struct amdgpu_fence_driver {
 	bool				initialized;
 	struct amdgpu_irq_src		*irq_src;
 	unsigned			irq_type;
+	struct timer_list		fallback_timer;
 	unsigned			num_fences_mask;
 	spinlock_t			lock;
 	struct dma_fence		**fences;
@@ -96,7 +97,7 @@ void amdgpu_fence_driver_resume(struct amdgpu_device *adev);
 int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **fence,
 		      unsigned flags);
 int amdgpu_fence_emit_polling(struct amdgpu_ring *ring, uint32_t *s);
-void amdgpu_fence_process(struct amdgpu_ring *ring);
+bool amdgpu_fence_process(struct amdgpu_ring *ring);
 int amdgpu_fence_wait_empty(struct amdgpu_ring *ring);
 signed long amdgpu_fence_wait_polling(struct amdgpu_ring *ring,
 				      uint32_t wait_seq,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h
index 2e87414422f9..e9bf70e2ac51 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h
@@ -103,7 +103,7 @@ TRACE_EVENT(amdgpu_iv,
 			   __entry->src_data[2] = iv->src_data[2];
 			   __entry->src_data[3] = iv->src_data[3];
 			   ),
-	    TP_printk("client_id:%u src_id:%u ring:%u vmid:%u timestamp: %llu pasid:%u src_data: %08x %08x %08x %08x\n",
+	    TP_printk("client_id:%u src_id:%u ring:%u vmid:%u timestamp: %llu pasid:%u src_data: %08x %08x %08x %08x",
 		      __entry->client_id, __entry->src_id,
 		      __entry->ring_id, __entry->vmid,
 		      __entry->timestamp, __entry->pasid,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
index a73674f9a0f5..2a2eb0143f48 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
@@ -36,6 +36,7 @@
 #include "soc15_common.h"
 
 #include "vcn/vcn_1_0_offset.h"
+#include "vcn/vcn_1_0_sh_mask.h"
 
 /* 1 second timeout */
 #define VCN_IDLE_TIMEOUT	msecs_to_jiffies(1000)
@@ -212,18 +213,158 @@ int amdgpu_vcn_resume(struct amdgpu_device *adev)
 	return 0;
 }
 
+static int amdgpu_vcn_pause_dpg_mode(struct amdgpu_device *adev,
+				     struct dpg_pause_state *new_state)
+{
+	int ret_code;
+	uint32_t reg_data = 0;
+	uint32_t reg_data2 = 0;
+	struct amdgpu_ring *ring;
+
+	/* pause/unpause if state is changed */
+	if (adev->vcn.pause_state.fw_based != new_state->fw_based) {
+		DRM_DEBUG("dpg pause state changed %d:%d -> %d:%d",
+			adev->vcn.pause_state.fw_based, adev->vcn.pause_state.jpeg,
+			new_state->fw_based, new_state->jpeg);
+
+		reg_data = RREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE) &
+			(~UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK);
+
+		if (new_state->fw_based == VCN_DPG_STATE__PAUSE) {
+			ret_code = 0;
+
+			if (!(reg_data & UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK))
+				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
+						   UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF,
+						   UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code);
+
+			if (!ret_code) {
+				/* pause DPG non-jpeg */
+				reg_data |= UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK;
+				WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data);
+				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_DPG_PAUSE,
+						   UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK,
+						   UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK, ret_code);
+
+				/* Restore */
+				ring = &adev->vcn.ring_enc[0];
+				WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO, ring->gpu_addr);
+				WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
+				WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE, ring->ring_size / 4);
+				WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR, lower_32_bits(ring->wptr));
+				WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR, lower_32_bits(ring->wptr));
+
+				ring = &adev->vcn.ring_enc[1];
+				WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO2, ring->gpu_addr);
+				WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr));
+				WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE2, ring->ring_size / 4);
+				WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr));
+				WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr));
+
+				ring = &adev->vcn.ring_dec;
+				WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR,
+					     lower_32_bits(ring->wptr) | 0x80000000);
+				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
+						   UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON,
+						   UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code);
+			}
+		} else {
+			/* unpause dpg non-jpeg, no need to wait */
+			reg_data &= ~UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK;
+			WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data);
+		}
+		adev->vcn.pause_state.fw_based = new_state->fw_based;
+	}
+
+	/* pause/unpause if state is changed */
+	if (adev->vcn.pause_state.jpeg != new_state->jpeg) {
+		DRM_DEBUG("dpg pause state changed %d:%d -> %d:%d",
+			adev->vcn.pause_state.fw_based, adev->vcn.pause_state.jpeg,
+			new_state->fw_based, new_state->jpeg);
+
+		reg_data = RREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE) &
+			(~UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK);
+
+		if (new_state->jpeg == VCN_DPG_STATE__PAUSE) {
+			ret_code = 0;
+
+			if (!(reg_data & UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK))
+				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
+						   UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF,
+						   UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code);
+
+			if (!ret_code) {
+				/* Make sure JPRG Snoop is disabled before sending the pause */
+				reg_data2 = RREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS);
+				reg_data2 |= UVD_POWER_STATUS__JRBC_SNOOP_DIS_MASK;
+				WREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS, reg_data2);
+
+				/* pause DPG jpeg */
+				reg_data |= UVD_DPG_PAUSE__JPEG_PAUSE_DPG_REQ_MASK;
+				WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data);
+				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_DPG_PAUSE,
+							UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK,
+							UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK, ret_code);
+
+				/* Restore */
+				ring = &adev->vcn.ring_jpeg;
+				WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_VMID, 0);
+				WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL, 0x00000001L | 0x00000002L);
+				WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_LOW,
+					     lower_32_bits(ring->gpu_addr));
+				WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_HIGH,
+					     upper_32_bits(ring->gpu_addr));
+				WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_RPTR, ring->wptr);
+				WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR, ring->wptr);
+				WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL, 0x00000002L);
+
+				ring = &adev->vcn.ring_dec;
+				WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR,
+					     lower_32_bits(ring->wptr) | 0x80000000);
+				SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS,
+						   UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON,
+						   UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code);
+			}
+		} else {
+			/* unpause dpg jpeg, no need to wait */
+			reg_data &= ~UVD_DPG_PAUSE__JPEG_PAUSE_DPG_REQ_MASK;
+			WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data);
+		}
+		adev->vcn.pause_state.jpeg = new_state->jpeg;
+	}
+
+	return 0;
+}
+
 static void amdgpu_vcn_idle_work_handler(struct work_struct *work)
 {
 	struct amdgpu_device *adev =
 		container_of(work, struct amdgpu_device, vcn.idle_work.work);
-	unsigned fences = amdgpu_fence_count_emitted(&adev->vcn.ring_dec);
-	unsigned i;
+	unsigned int fences = 0;
+	unsigned int i;
 
 	for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
 		fences += amdgpu_fence_count_emitted(&adev->vcn.ring_enc[i]);
 	}
 
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)	{
+		struct dpg_pause_state new_state;
+
+		if (fences)
+			new_state.fw_based = VCN_DPG_STATE__PAUSE;
+		else
+			new_state.fw_based = VCN_DPG_STATE__UNPAUSE;
+
+		if (amdgpu_fence_count_emitted(&adev->vcn.ring_jpeg))
+			new_state.jpeg = VCN_DPG_STATE__PAUSE;
+		else
+			new_state.jpeg = VCN_DPG_STATE__UNPAUSE;
+
+		amdgpu_vcn_pause_dpg_mode(adev, &new_state);
+	}
+
 	fences += amdgpu_fence_count_emitted(&adev->vcn.ring_jpeg);
+	fences += amdgpu_fence_count_emitted(&adev->vcn.ring_dec);
 
 	if (fences == 0) {
 		amdgpu_gfx_off_ctrl(adev, true);
@@ -250,6 +391,22 @@ void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring)
 			amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN,
 							       AMD_PG_STATE_UNGATE);
 	}
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)	{
+		struct dpg_pause_state new_state;
+
+		if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC)
+			new_state.fw_based = VCN_DPG_STATE__PAUSE;
+		else
+			new_state.fw_based = adev->vcn.pause_state.fw_based;
+
+		if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_JPEG)
+			new_state.jpeg = VCN_DPG_STATE__PAUSE;
+		else
+			new_state.jpeg = adev->vcn.pause_state.jpeg;
+
+		amdgpu_vcn_pause_dpg_mode(adev, &new_state);
+	}
 }
 
 void amdgpu_vcn_ring_end_use(struct amdgpu_ring *ring)
@@ -264,7 +421,7 @@ int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring)
 	unsigned i;
 	int r;
 
-	WREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_CONTEXT_ID), 0xCAFEDEAD);
+	WREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0xCAFEDEAD);
 	r = amdgpu_ring_alloc(ring, 3);
 	if (r) {
 		DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n",
@@ -272,11 +429,11 @@ int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring)
 		return r;
 	}
 	amdgpu_ring_write(ring,
-		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_CONTEXT_ID), 0));
+		PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0));
 	amdgpu_ring_write(ring, 0xDEADBEEF);
 	amdgpu_ring_commit(ring);
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_CONTEXT_ID));
+		tmp = RREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9));
 		if (tmp == 0xDEADBEEF)
 			break;
 		DRM_UDELAY(1);
@@ -616,7 +773,7 @@ int amdgpu_vcn_jpeg_ring_test_ring(struct amdgpu_ring *ring)
 	unsigned i;
 	int r;
 
-	WREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_CONTEXT_ID), 0xCAFEDEAD);
+	WREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0xCAFEDEAD);
 	r = amdgpu_ring_alloc(ring, 3);
 
 	if (r) {
@@ -626,12 +783,12 @@ int amdgpu_vcn_jpeg_ring_test_ring(struct amdgpu_ring *ring)
 	}
 
 	amdgpu_ring_write(ring,
-		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_CONTEXT_ID), 0, 0, 0));
+		PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0, 0, 0));
 	amdgpu_ring_write(ring, 0xDEADBEEF);
 	amdgpu_ring_commit(ring);
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_CONTEXT_ID));
+		tmp = RREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9));
 		if (tmp == 0xDEADBEEF)
 			break;
 		DRM_UDELAY(1);
@@ -665,7 +822,7 @@ static int amdgpu_vcn_jpeg_set_reg(struct amdgpu_ring *ring, uint32_t handle,
 
 	ib = &job->ibs[0];
 
-	ib->ptr[0] = PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JPEG_PITCH), 0, 0, PACKETJ_TYPE0);
+	ib->ptr[0] = PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0, 0, PACKETJ_TYPE0);
 	ib->ptr[1] = 0xDEADBEEF;
 	for (i = 2; i < 16; i += 2) {
 		ib->ptr[i] = PACKETJ(0, 0, 0, PACKETJ_TYPE6);
@@ -714,7 +871,7 @@ int amdgpu_vcn_jpeg_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = 0;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_JPEG_PITCH));
+		tmp = RREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9));
 		if (tmp == 0xDEADBEEF)
 			break;
 		DRM_UDELAY(1);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
index 0b0b8638d73f..0b88a4672da5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
@@ -56,6 +56,16 @@ enum engine_status_constants {
 	UVD_STATUS__RBC_BUSY = 0x1,
 };
 
+enum internal_dpg_state {
+	VCN_DPG_STATE__UNPAUSE = 0,
+	VCN_DPG_STATE__PAUSE,
+};
+
+struct dpg_pause_state {
+	enum internal_dpg_state fw_based;
+	enum internal_dpg_state jpeg;
+};
+
 struct amdgpu_vcn {
 	struct amdgpu_bo	*vcpu_bo;
 	void			*cpu_addr;
@@ -69,6 +79,8 @@ struct amdgpu_vcn {
 	struct amdgpu_ring	ring_jpeg;
 	struct amdgpu_irq_src	irq;
 	unsigned		num_enc_rings;
+	enum amd_powergating_state cur_state;
+	struct dpg_pause_state pause_state;
 };
 
 int amdgpu_vcn_sw_init(struct amdgpu_device *adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c
index d2469453dca2..79220a91abe3 100644
--- a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c
+++ b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c
@@ -6277,12 +6277,12 @@ static int ci_dpm_sw_init(void *handle)
 	int ret;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	ret = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 230,
+	ret = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 230,
 				&adev->pm.dpm.thermal.irq);
 	if (ret)
 		return ret;
 
-	ret = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 231,
+	ret = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 231,
 				&adev->pm.dpm.thermal.irq);
 	if (ret)
 		return ret;
diff --git a/drivers/gpu/drm/amd/amdgpu/cik_ih.c b/drivers/gpu/drm/amd/amdgpu/cik_ih.c
index 44d10c2172f6..b5775c6a857b 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/cik_ih.c
@@ -276,7 +276,7 @@ static void cik_ih_decode_iv(struct amdgpu_device *adev,
 	dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]);
 	dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]);
 
-	entry->client_id = AMDGPU_IH_CLIENTID_LEGACY;
+	entry->client_id = AMDGPU_IRQ_CLIENTID_LEGACY;
 	entry->src_id = dw[0] & 0xff;
 	entry->src_data[0] = dw[1] & 0xfffffff;
 	entry->ring_id = dw[2] & 0xff;
@@ -318,7 +318,7 @@ static int cik_ih_sw_init(void *handle)
 	int r;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_ih_ring_init(adev, 64 * 1024, false);
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, false);
 	if (r)
 		return r;
 
@@ -332,7 +332,7 @@ static int cik_ih_sw_fini(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	amdgpu_irq_fini(adev);
-	amdgpu_ih_ring_fini(adev);
+	amdgpu_ih_ring_fini(adev, &adev->irq.ih);
 	amdgpu_irq_remove_domain(adev);
 
 	return 0;
@@ -468,8 +468,7 @@ static const struct amdgpu_ih_funcs cik_ih_funcs = {
 
 static void cik_ih_set_interrupt_funcs(struct amdgpu_device *adev)
 {
-	if (adev->irq.ih_funcs == NULL)
-		adev->irq.ih_funcs = &cik_ih_funcs;
+	adev->irq.ih_funcs = &cik_ih_funcs;
 }
 
 const struct amdgpu_ip_block_version cik_ih_ip_block =
diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
index 154b1499b07e..b918c8886b75 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
+++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
@@ -970,19 +970,19 @@ static int cik_sdma_sw_init(void *handle)
 	}
 
 	/* SDMA trap event */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 224,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 224,
 			      &adev->sdma.trap_irq);
 	if (r)
 		return r;
 
 	/* SDMA Privileged inst */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 241,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 241,
 			      &adev->sdma.illegal_inst_irq);
 	if (r)
 		return r;
 
 	/* SDMA Privileged inst */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 247,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 247,
 			      &adev->sdma.illegal_inst_irq);
 	if (r)
 		return r;
@@ -1370,10 +1370,8 @@ static const struct amdgpu_buffer_funcs cik_sdma_buffer_funcs = {
 
 static void cik_sdma_set_buffer_funcs(struct amdgpu_device *adev)
 {
-	if (adev->mman.buffer_funcs == NULL) {
-		adev->mman.buffer_funcs = &cik_sdma_buffer_funcs;
-		adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring;
-	}
+	adev->mman.buffer_funcs = &cik_sdma_buffer_funcs;
+	adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring;
 }
 
 static const struct amdgpu_vm_pte_funcs cik_sdma_vm_pte_funcs = {
@@ -1389,15 +1387,13 @@ static void cik_sdma_set_vm_pte_funcs(struct amdgpu_device *adev)
 	struct drm_gpu_scheduler *sched;
 	unsigned i;
 
-	if (adev->vm_manager.vm_pte_funcs == NULL) {
-		adev->vm_manager.vm_pte_funcs = &cik_sdma_vm_pte_funcs;
-		for (i = 0; i < adev->sdma.num_instances; i++) {
-			sched = &adev->sdma.instance[i].ring.sched;
-			adev->vm_manager.vm_pte_rqs[i] =
-				&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
-		}
-		adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
+	adev->vm_manager.vm_pte_funcs = &cik_sdma_vm_pte_funcs;
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		sched = &adev->sdma.instance[i].ring.sched;
+		adev->vm_manager.vm_pte_rqs[i] =
+			&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
 	}
+	adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
 }
 
 const struct amdgpu_ip_block_version cik_sdma_ip_block =
diff --git a/drivers/gpu/drm/amd/amdgpu/cz_ih.c b/drivers/gpu/drm/amd/amdgpu/cz_ih.c
index 960c29e17da6..df5ac4d85a00 100644
--- a/drivers/gpu/drm/amd/amdgpu/cz_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/cz_ih.c
@@ -255,7 +255,7 @@ static void cz_ih_decode_iv(struct amdgpu_device *adev,
 	dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]);
 	dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]);
 
-	entry->client_id = AMDGPU_IH_CLIENTID_LEGACY;
+	entry->client_id = AMDGPU_IRQ_CLIENTID_LEGACY;
 	entry->src_id = dw[0] & 0xff;
 	entry->src_data[0] = dw[1] & 0xfffffff;
 	entry->ring_id = dw[2] & 0xff;
@@ -297,7 +297,7 @@ static int cz_ih_sw_init(void *handle)
 	int r;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_ih_ring_init(adev, 64 * 1024, false);
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, false);
 	if (r)
 		return r;
 
@@ -311,7 +311,7 @@ static int cz_ih_sw_fini(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	amdgpu_irq_fini(adev);
-	amdgpu_ih_ring_fini(adev);
+	amdgpu_ih_ring_fini(adev, &adev->irq.ih);
 	amdgpu_irq_remove_domain(adev);
 
 	return 0;
@@ -449,8 +449,7 @@ static const struct amdgpu_ih_funcs cz_ih_funcs = {
 
 static void cz_ih_set_interrupt_funcs(struct amdgpu_device *adev)
 {
-	if (adev->irq.ih_funcs == NULL)
-		adev->irq.ih_funcs = &cz_ih_funcs;
+	adev->irq.ih_funcs = &cz_ih_funcs;
 }
 
 const struct amdgpu_ip_block_version cz_ih_ip_block =
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c
index 3916aa6cc4ec..4cfecdce29a3 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c
@@ -2746,19 +2746,19 @@ static int dce_v10_0_sw_init(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i + 1, &adev->crtc_irq);
+		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + 1, &adev->crtc_irq);
 		if (r)
 			return r;
 	}
 
 	for (i = VISLANDS30_IV_SRCID_D1_GRPH_PFLIP; i < 20; i += 2) {
-		r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i, &adev->pageflip_irq);
+		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i, &adev->pageflip_irq);
 		if (r)
 			return r;
 	}
 
 	/* HPD hotplug */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_HOTPLUG_DETECT_A, &adev->hpd_irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_HOTPLUG_DETECT_A, &adev->hpd_irq);
 	if (r)
 		return r;
 
@@ -3570,8 +3570,7 @@ static const struct amdgpu_display_funcs dce_v10_0_display_funcs = {
 
 static void dce_v10_0_set_display_funcs(struct amdgpu_device *adev)
 {
-	if (adev->mode_info.funcs == NULL)
-		adev->mode_info.funcs = &dce_v10_0_display_funcs;
+	adev->mode_info.funcs = &dce_v10_0_display_funcs;
 }
 
 static const struct amdgpu_irq_src_funcs dce_v10_0_crtc_irq_funcs = {
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c
index 4ffb612a4e53..7c868916d90f 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c
@@ -2867,19 +2867,19 @@ static int dce_v11_0_sw_init(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i + 1, &adev->crtc_irq);
+		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + 1, &adev->crtc_irq);
 		if (r)
 			return r;
 	}
 
 	for (i = VISLANDS30_IV_SRCID_D1_GRPH_PFLIP; i < 20; i += 2) {
-		r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i, &adev->pageflip_irq);
+		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i, &adev->pageflip_irq);
 		if (r)
 			return r;
 	}
 
 	/* HPD hotplug */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_HOTPLUG_DETECT_A, &adev->hpd_irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_HOTPLUG_DETECT_A, &adev->hpd_irq);
 	if (r)
 		return r;
 
@@ -3702,8 +3702,7 @@ static const struct amdgpu_display_funcs dce_v11_0_display_funcs = {
 
 static void dce_v11_0_set_display_funcs(struct amdgpu_device *adev)
 {
-	if (adev->mode_info.funcs == NULL)
-		adev->mode_info.funcs = &dce_v11_0_display_funcs;
+	adev->mode_info.funcs = &dce_v11_0_display_funcs;
 }
 
 static const struct amdgpu_irq_src_funcs dce_v11_0_crtc_irq_funcs = {
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c
index 480c5348a14f..17eaaba36017 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c
@@ -2616,19 +2616,19 @@ static int dce_v6_0_sw_init(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i + 1, &adev->crtc_irq);
+		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + 1, &adev->crtc_irq);
 		if (r)
 			return r;
 	}
 
 	for (i = 8; i < 20; i += 2) {
-		r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i, &adev->pageflip_irq);
+		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i, &adev->pageflip_irq);
 		if (r)
 			return r;
 	}
 
 	/* HPD hotplug */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 42, &adev->hpd_irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 42, &adev->hpd_irq);
 	if (r)
 		return r;
 
@@ -3376,8 +3376,7 @@ static const struct amdgpu_display_funcs dce_v6_0_display_funcs = {
 
 static void dce_v6_0_set_display_funcs(struct amdgpu_device *adev)
 {
-	if (adev->mode_info.funcs == NULL)
-		adev->mode_info.funcs = &dce_v6_0_display_funcs;
+	adev->mode_info.funcs = &dce_v6_0_display_funcs;
 }
 
 static const struct amdgpu_irq_src_funcs dce_v6_0_crtc_irq_funcs = {
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c
index 797196476c94..8c0576978d36 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c
@@ -2643,19 +2643,19 @@ static int dce_v8_0_sw_init(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i + 1, &adev->crtc_irq);
+		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + 1, &adev->crtc_irq);
 		if (r)
 			return r;
 	}
 
 	for (i = 8; i < 20; i += 2) {
-		r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i, &adev->pageflip_irq);
+		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i, &adev->pageflip_irq);
 		if (r)
 			return r;
 	}
 
 	/* HPD hotplug */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 42, &adev->hpd_irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 42, &adev->hpd_irq);
 	if (r)
 		return r;
 
@@ -3458,8 +3458,7 @@ static const struct amdgpu_display_funcs dce_v8_0_display_funcs = {
 
 static void dce_v8_0_set_display_funcs(struct amdgpu_device *adev)
 {
-	if (adev->mode_info.funcs == NULL)
-		adev->mode_info.funcs = &dce_v8_0_display_funcs;
+	adev->mode_info.funcs = &dce_v8_0_display_funcs;
 }
 
 static const struct amdgpu_irq_src_funcs dce_v8_0_crtc_irq_funcs = {
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_virtual.c b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c
index 15257634a53a..fdace004544d 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_virtual.c
+++ b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c
@@ -372,7 +372,7 @@ static int dce_virtual_sw_init(void *handle)
 	int r, i;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SMU_DISP_TIMER2_TRIGGER, &adev->crtc_irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SMU_DISP_TIMER2_TRIGGER, &adev->crtc_irq);
 	if (r)
 		return r;
 
@@ -649,8 +649,7 @@ static const struct amdgpu_display_funcs dce_virtual_display_funcs = {
 
 static void dce_virtual_set_display_funcs(struct amdgpu_device *adev)
 {
-	if (adev->mode_info.funcs == NULL)
-		adev->mode_info.funcs = &dce_virtual_display_funcs;
+	adev->mode_info.funcs = &dce_virtual_display_funcs;
 }
 
 static int dce_virtual_pageflip(struct amdgpu_device *adev,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c
index de184a886057..d76eb27945dc 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c
@@ -1552,7 +1552,7 @@ static void gfx_v6_0_config_init(struct amdgpu_device *adev)
 	adev->gfx.config.double_offchip_lds_buf = 0;
 }
 
-static void gfx_v6_0_gpu_init(struct amdgpu_device *adev)
+static void gfx_v6_0_constants_init(struct amdgpu_device *adev)
 {
 	u32 gb_addr_config = 0;
 	u32 mc_shared_chmap, mc_arb_ramcfg;
@@ -3094,15 +3094,15 @@ static int gfx_v6_0_sw_init(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	int i, r;
 
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 181, &adev->gfx.eop_irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 181, &adev->gfx.eop_irq);
 	if (r)
 		return r;
 
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 184, &adev->gfx.priv_reg_irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 184, &adev->gfx.priv_reg_irq);
 	if (r)
 		return r;
 
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 185, &adev->gfx.priv_inst_irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 185, &adev->gfx.priv_inst_irq);
 	if (r)
 		return r;
 
@@ -3175,7 +3175,7 @@ static int gfx_v6_0_hw_init(void *handle)
 	int r;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	gfx_v6_0_gpu_init(adev);
+	gfx_v6_0_constants_init(adev);
 
 	r = gfx_v6_0_rlc_resume(adev);
 	if (r)
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
index fc39ebbc9d9f..0e72bc09939a 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
@@ -1886,14 +1886,14 @@ static void gfx_v7_0_config_init(struct amdgpu_device *adev)
 }
 
 /**
- * gfx_v7_0_gpu_init - setup the 3D engine
+ * gfx_v7_0_constants_init - setup the 3D engine
  *
  * @adev: amdgpu_device pointer
  *
- * Configures the 3D engine and tiling configuration
- * registers so that the 3D engine is usable.
+ * init the gfx constants such as the 3D engine, tiling configuration
+ * registers, maximum number of quad pipes, render backends...
  */
-static void gfx_v7_0_gpu_init(struct amdgpu_device *adev)
+static void gfx_v7_0_constants_init(struct amdgpu_device *adev)
 {
 	u32 sh_mem_cfg, sh_static_mem_cfg, sh_mem_base;
 	u32 tmp;
@@ -4516,18 +4516,18 @@ static int gfx_v7_0_sw_init(void *handle)
 	adev->gfx.mec.num_queue_per_pipe = 8;
 
 	/* EOP Event */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 181, &adev->gfx.eop_irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 181, &adev->gfx.eop_irq);
 	if (r)
 		return r;
 
 	/* Privileged reg */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 184,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 184,
 			      &adev->gfx.priv_reg_irq);
 	if (r)
 		return r;
 
 	/* Privileged inst */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 185,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 185,
 			      &adev->gfx.priv_inst_irq);
 	if (r)
 		return r;
@@ -4624,7 +4624,7 @@ static int gfx_v7_0_hw_init(void *handle)
 	int r;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	gfx_v7_0_gpu_init(adev);
+	gfx_v7_0_constants_init(adev);
 
 	/* init rlc */
 	r = gfx_v7_0_rlc_resume(adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
index 470dc80f4fe7..2aeef2bb93a4 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
@@ -2049,35 +2049,35 @@ static int gfx_v8_0_sw_init(void *handle)
 	adev->gfx.mec.num_queue_per_pipe = 8;
 
 	/* KIQ event */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_INT_IB2, &adev->gfx.kiq.irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_INT_IB2, &adev->gfx.kiq.irq);
 	if (r)
 		return r;
 
 	/* EOP Event */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_END_OF_PIPE, &adev->gfx.eop_irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_END_OF_PIPE, &adev->gfx.eop_irq);
 	if (r)
 		return r;
 
 	/* Privileged reg */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_PRIV_REG_FAULT,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_PRIV_REG_FAULT,
 			      &adev->gfx.priv_reg_irq);
 	if (r)
 		return r;
 
 	/* Privileged inst */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_PRIV_INSTR_FAULT,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_PRIV_INSTR_FAULT,
 			      &adev->gfx.priv_inst_irq);
 	if (r)
 		return r;
 
 	/* Add CP EDC/ECC irq  */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_ECC_ERROR,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_ECC_ERROR,
 			      &adev->gfx.cp_ecc_error_irq);
 	if (r)
 		return r;
 
 	/* SQ interrupts. */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SQ_INTERRUPT_MSG,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SQ_INTERRUPT_MSG,
 			      &adev->gfx.sq_irq);
 	if (r) {
 		DRM_ERROR("amdgpu_irq_add() for SQ failed: %d\n", r);
@@ -3835,7 +3835,7 @@ static void gfx_v8_0_config_init(struct amdgpu_device *adev)
 	}
 }
 
-static void gfx_v8_0_gpu_init(struct amdgpu_device *adev)
+static void gfx_v8_0_constants_init(struct amdgpu_device *adev)
 {
 	u32 tmp, sh_static_mem_cfg;
 	int i;
@@ -4208,31 +4208,11 @@ static int gfx_v8_0_rlc_load_microcode(struct amdgpu_device *adev)
 static int gfx_v8_0_rlc_resume(struct amdgpu_device *adev)
 {
 	int r;
-	u32 tmp;
 
 	gfx_v8_0_rlc_stop(adev);
-
-	/* disable CG */
-	tmp = RREG32(mmRLC_CGCG_CGLS_CTRL);
-	tmp &= ~(RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK |
-		 RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK);
-	WREG32(mmRLC_CGCG_CGLS_CTRL, tmp);
-	if (adev->asic_type == CHIP_POLARIS11 ||
-	    adev->asic_type == CHIP_POLARIS10 ||
-	    adev->asic_type == CHIP_POLARIS12 ||
-	    adev->asic_type == CHIP_VEGAM) {
-		tmp = RREG32(mmRLC_CGCG_CGLS_CTRL_3D);
-		tmp &= ~0x3;
-		WREG32(mmRLC_CGCG_CGLS_CTRL_3D, tmp);
-	}
-
-	/* disable PG */
-	WREG32(mmRLC_PG_CNTL, 0);
-
 	gfx_v8_0_rlc_reset(adev);
 	gfx_v8_0_init_pg(adev);
 
-
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
 		/* legacy rlc firmware loading */
 		r = gfx_v8_0_rlc_load_microcode(adev);
@@ -5039,7 +5019,7 @@ static int gfx_v8_0_hw_init(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	gfx_v8_0_init_golden_registers(adev);
-	gfx_v8_0_gpu_init(adev);
+	gfx_v8_0_constants_init(adev);
 
 	r = gfx_v8_0_rlc_resume(adev);
 	if (r)
@@ -5080,6 +5060,55 @@ static int gfx_v8_0_kcq_disable(struct amdgpu_device *adev)
 	return r;
 }
 
+static bool gfx_v8_0_is_idle(void *handle)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+	if (REG_GET_FIELD(RREG32(mmGRBM_STATUS), GRBM_STATUS, GUI_ACTIVE)
+		|| RREG32(mmGRBM_STATUS2) != 0x8)
+		return false;
+	else
+		return true;
+}
+
+static bool gfx_v8_0_rlc_is_idle(void *handle)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+	if (RREG32(mmGRBM_STATUS2) != 0x8)
+		return false;
+	else
+		return true;
+}
+
+static int gfx_v8_0_wait_for_rlc_idle(void *handle)
+{
+	unsigned int i;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (gfx_v8_0_rlc_is_idle(handle))
+			return 0;
+
+		udelay(1);
+	}
+	return -ETIMEDOUT;
+}
+
+static int gfx_v8_0_wait_for_idle(void *handle)
+{
+	unsigned int i;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (gfx_v8_0_is_idle(handle))
+			return 0;
+
+		udelay(1);
+	}
+	return -ETIMEDOUT;
+}
+
 static int gfx_v8_0_hw_fini(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
@@ -5098,9 +5127,16 @@ static int gfx_v8_0_hw_fini(void *handle)
 		pr_debug("For SRIOV client, shouldn't do anything.\n");
 		return 0;
 	}
-	gfx_v8_0_cp_enable(adev, false);
-	gfx_v8_0_rlc_stop(adev);
-
+	adev->gfx.rlc.funcs->enter_safe_mode(adev);
+	if (!gfx_v8_0_wait_for_idle(adev))
+		gfx_v8_0_cp_enable(adev, false);
+	else
+		pr_err("cp is busy, skip halt cp\n");
+	if (!gfx_v8_0_wait_for_rlc_idle(adev))
+		gfx_v8_0_rlc_stop(adev);
+	else
+		pr_err("rlc is busy, skip halt rlc\n");
+	adev->gfx.rlc.funcs->exit_safe_mode(adev);
 	return 0;
 }
 
@@ -5121,30 +5157,6 @@ static int gfx_v8_0_resume(void *handle)
 	return r;
 }
 
-static bool gfx_v8_0_is_idle(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	if (REG_GET_FIELD(RREG32(mmGRBM_STATUS), GRBM_STATUS, GUI_ACTIVE))
-		return false;
-	else
-		return true;
-}
-
-static int gfx_v8_0_wait_for_idle(void *handle)
-{
-	unsigned i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	for (i = 0; i < adev->usec_timeout; i++) {
-		if (gfx_v8_0_is_idle(handle))
-			return 0;
-
-		udelay(1);
-	}
-	return -ETIMEDOUT;
-}
-
 static bool gfx_v8_0_check_soft_reset(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index f369d9603435..261bb051b14d 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
@@ -1847,7 +1847,7 @@ static void gfx_v9_0_init_compute_vmid(struct amdgpu_device *adev)
 	mutex_unlock(&adev->srbm_mutex);
 }
 
-static void gfx_v9_0_gpu_init(struct amdgpu_device *adev)
+static void gfx_v9_0_constants_init(struct amdgpu_device *adev)
 {
 	u32 tmp;
 	int i;
@@ -3235,7 +3235,7 @@ static int gfx_v9_0_hw_init(void *handle)
 
 	gfx_v9_0_init_golden_registers(adev);
 
-	gfx_v9_0_gpu_init(adev);
+	gfx_v9_0_constants_init(adev);
 
 	r = gfx_v9_0_csb_vram_pin(adev);
 	if (r)
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c
index 65f58ebcf835..ceb7847b504f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c
@@ -82,7 +82,8 @@ static void gfxhub_v1_0_init_system_aperture_regs(struct amdgpu_device *adev)
 		 * to get rid of the VM fault and hardware hang.
 		 */
 		WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
-			     (max(adev->gmc.vram_end, adev->gmc.agp_end) >> 18) + 0x1);
+			     max((adev->gmc.vram_end >> 18) + 0x1,
+				 adev->gmc.agp_end >> 18));
 	else
 		WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
 			     max(adev->gmc.vram_end, adev->gmc.agp_end) >> 18);
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
index 4411463ca719..e1c2b4e9c7b2 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
@@ -859,11 +859,11 @@ static int gmc_v6_0_sw_init(void *handle)
 		adev->gmc.vram_type = gmc_v6_0_convert_vram_type(tmp);
 	}
 
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 146, &adev->gmc.vm_fault);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 146, &adev->gmc.vm_fault);
 	if (r)
 		return r;
 
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 147, &adev->gmc.vm_fault);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 147, &adev->gmc.vm_fault);
 	if (r)
 		return r;
 
@@ -1180,8 +1180,7 @@ static const struct amdgpu_irq_src_funcs gmc_v6_0_irq_funcs = {
 
 static void gmc_v6_0_set_gmc_funcs(struct amdgpu_device *adev)
 {
-	if (adev->gmc.gmc_funcs == NULL)
-		adev->gmc.gmc_funcs = &gmc_v6_0_gmc_funcs;
+	adev->gmc.gmc_funcs = &gmc_v6_0_gmc_funcs;
 }
 
 static void gmc_v6_0_set_irq_funcs(struct amdgpu_device *adev)
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
index ae776ce9a415..910c4ce19cb3 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
@@ -991,11 +991,11 @@ static int gmc_v7_0_sw_init(void *handle)
 		adev->gmc.vram_type = gmc_v7_0_convert_vram_type(tmp);
 	}
 
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_GFX_PAGE_INV_FAULT, &adev->gmc.vm_fault);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_GFX_PAGE_INV_FAULT, &adev->gmc.vm_fault);
 	if (r)
 		return r;
 
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_GFX_MEM_PROT_FAULT, &adev->gmc.vm_fault);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_GFX_MEM_PROT_FAULT, &adev->gmc.vm_fault);
 	if (r)
 		return r;
 
@@ -1388,8 +1388,7 @@ static const struct amdgpu_irq_src_funcs gmc_v7_0_irq_funcs = {
 
 static void gmc_v7_0_set_gmc_funcs(struct amdgpu_device *adev)
 {
-	if (adev->gmc.gmc_funcs == NULL)
-		adev->gmc.gmc_funcs = &gmc_v7_0_gmc_funcs;
+	adev->gmc.gmc_funcs = &gmc_v7_0_gmc_funcs;
 }
 
 static void gmc_v7_0_set_irq_funcs(struct amdgpu_device *adev)
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
index 53ae49b8bde8..1d3265c97b70 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
@@ -1095,11 +1095,11 @@ static int gmc_v8_0_sw_init(void *handle)
 		adev->gmc.vram_type = gmc_v8_0_convert_vram_type(tmp);
 	}
 
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_GFX_PAGE_INV_FAULT, &adev->gmc.vm_fault);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_GFX_PAGE_INV_FAULT, &adev->gmc.vm_fault);
 	if (r)
 		return r;
 
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_GFX_MEM_PROT_FAULT, &adev->gmc.vm_fault);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_GFX_MEM_PROT_FAULT, &adev->gmc.vm_fault);
 	if (r)
 		return r;
 
@@ -1733,8 +1733,7 @@ static const struct amdgpu_irq_src_funcs gmc_v8_0_irq_funcs = {
 
 static void gmc_v8_0_set_gmc_funcs(struct amdgpu_device *adev)
 {
-	if (adev->gmc.gmc_funcs == NULL)
-		adev->gmc.gmc_funcs = &gmc_v8_0_gmc_funcs;
+	adev->gmc.gmc_funcs = &gmc_v8_0_gmc_funcs;
 }
 
 static void gmc_v8_0_set_irq_funcs(struct amdgpu_device *adev)
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index aad3c7c5fb3a..f35d7a554ad5 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
@@ -593,8 +593,7 @@ static const struct amdgpu_gmc_funcs gmc_v9_0_gmc_funcs = {
 
 static void gmc_v9_0_set_gmc_funcs(struct amdgpu_device *adev)
 {
-	if (adev->gmc.gmc_funcs == NULL)
-		adev->gmc.gmc_funcs = &gmc_v9_0_gmc_funcs;
+	adev->gmc.gmc_funcs = &gmc_v9_0_gmc_funcs;
 }
 
 static int gmc_v9_0_early_init(void *handle)
diff --git a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c
index 842c4b677b4d..cf0fc61aebe6 100644
--- a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c
@@ -255,7 +255,7 @@ static void iceland_ih_decode_iv(struct amdgpu_device *adev,
 	dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]);
 	dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]);
 
-	entry->client_id = AMDGPU_IH_CLIENTID_LEGACY;
+	entry->client_id = AMDGPU_IRQ_CLIENTID_LEGACY;
 	entry->src_id = dw[0] & 0xff;
 	entry->src_data[0] = dw[1] & 0xfffffff;
 	entry->ring_id = dw[2] & 0xff;
@@ -297,7 +297,7 @@ static int iceland_ih_sw_init(void *handle)
 	int r;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_ih_ring_init(adev, 64 * 1024, false);
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, false);
 	if (r)
 		return r;
 
@@ -311,7 +311,7 @@ static int iceland_ih_sw_fini(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	amdgpu_irq_fini(adev);
-	amdgpu_ih_ring_fini(adev);
+	amdgpu_ih_ring_fini(adev, &adev->irq.ih);
 	amdgpu_irq_remove_domain(adev);
 
 	return 0;
@@ -447,8 +447,7 @@ static const struct amdgpu_ih_funcs iceland_ih_funcs = {
 
 static void iceland_ih_set_interrupt_funcs(struct amdgpu_device *adev)
 {
-	if (adev->irq.ih_funcs == NULL)
-		adev->irq.ih_funcs = &iceland_ih_funcs;
+	adev->irq.ih_funcs = &iceland_ih_funcs;
 }
 
 const struct amdgpu_ip_block_version iceland_ih_ip_block =
diff --git a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c
index cb79a93c2eb7..d0e478f43443 100644
--- a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c
+++ b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c
@@ -2995,12 +2995,12 @@ static int kv_dpm_sw_init(void *handle)
 	int ret;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	ret = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 230,
+	ret = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 230,
 				&adev->pm.dpm.thermal.irq);
 	if (ret)
 		return ret;
 
-	ret = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 231,
+	ret = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 231,
 				&adev->pm.dpm.thermal.irq);
 	if (ret)
 		return ret;
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c
index 80698b5ffa4a..14649f8475f3 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c
@@ -100,7 +100,8 @@ static void mmhub_v1_0_init_system_aperture_regs(struct amdgpu_device *adev)
 		 * to get rid of the VM fault and hardware hang.
 		 */
 		WREG32_SOC15(MMHUB, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
-			     (max(adev->gmc.vram_end, adev->gmc.agp_end) >> 18) + 0x1);
+			     max((adev->gmc.vram_end >> 18) + 0x1,
+				 adev->gmc.agp_end >> 18));
 	else
 		WREG32_SOC15(MMHUB, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
 			     max(adev->gmc.vram_end, adev->gmc.agp_end) >> 18);
diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c
index 842567b53df5..64e875d528dd 100644
--- a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c
+++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c
@@ -580,11 +580,11 @@ int xgpu_vi_mailbox_add_irq_id(struct amdgpu_device *adev)
 {
 	int r;
 
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 135, &adev->virt.rcv_irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 135, &adev->virt.rcv_irq);
 	if (r)
 		return r;
 
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 138, &adev->virt.ack_irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 138, &adev->virt.ack_irq);
 	if (r) {
 		amdgpu_irq_put(adev, &adev->virt.rcv_irq, 0);
 		return r;
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
index c403bdf8ad70..cd781abc4953 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
@@ -898,19 +898,19 @@ static int sdma_v2_4_sw_init(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	/* SDMA trap event */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SDMA_TRAP,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SDMA_TRAP,
 			      &adev->sdma.trap_irq);
 	if (r)
 		return r;
 
 	/* SDMA Privileged inst */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 241,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 241,
 			      &adev->sdma.illegal_inst_irq);
 	if (r)
 		return r;
 
 	/* SDMA Privileged inst */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SDMA_SRBM_WRITE,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SDMA_SRBM_WRITE,
 			      &adev->sdma.illegal_inst_irq);
 	if (r)
 		return r;
@@ -1296,10 +1296,8 @@ static const struct amdgpu_buffer_funcs sdma_v2_4_buffer_funcs = {
 
 static void sdma_v2_4_set_buffer_funcs(struct amdgpu_device *adev)
 {
-	if (adev->mman.buffer_funcs == NULL) {
-		adev->mman.buffer_funcs = &sdma_v2_4_buffer_funcs;
-		adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring;
-	}
+	adev->mman.buffer_funcs = &sdma_v2_4_buffer_funcs;
+	adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring;
 }
 
 static const struct amdgpu_vm_pte_funcs sdma_v2_4_vm_pte_funcs = {
@@ -1315,15 +1313,13 @@ static void sdma_v2_4_set_vm_pte_funcs(struct amdgpu_device *adev)
 	struct drm_gpu_scheduler *sched;
 	unsigned i;
 
-	if (adev->vm_manager.vm_pte_funcs == NULL) {
-		adev->vm_manager.vm_pte_funcs = &sdma_v2_4_vm_pte_funcs;
-		for (i = 0; i < adev->sdma.num_instances; i++) {
-			sched = &adev->sdma.instance[i].ring.sched;
-			adev->vm_manager.vm_pte_rqs[i] =
-				&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
-		}
-		adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
+	adev->vm_manager.vm_pte_funcs = &sdma_v2_4_vm_pte_funcs;
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		sched = &adev->sdma.instance[i].ring.sched;
+		adev->vm_manager.vm_pte_rqs[i] =
+			&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
 	}
+	adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
 }
 
 const struct amdgpu_ip_block_version sdma_v2_4_ip_block =
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
index 2677d6a1bf42..6d5c8ac64874 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
@@ -1177,19 +1177,19 @@ static int sdma_v3_0_sw_init(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	/* SDMA trap event */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SDMA_TRAP,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SDMA_TRAP,
 			      &adev->sdma.trap_irq);
 	if (r)
 		return r;
 
 	/* SDMA Privileged inst */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 241,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 241,
 			      &adev->sdma.illegal_inst_irq);
 	if (r)
 		return r;
 
 	/* SDMA Privileged inst */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SDMA_SRBM_WRITE,
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SDMA_SRBM_WRITE,
 			      &adev->sdma.illegal_inst_irq);
 	if (r)
 		return r;
@@ -1736,10 +1736,8 @@ static const struct amdgpu_buffer_funcs sdma_v3_0_buffer_funcs = {
 
 static void sdma_v3_0_set_buffer_funcs(struct amdgpu_device *adev)
 {
-	if (adev->mman.buffer_funcs == NULL) {
-		adev->mman.buffer_funcs = &sdma_v3_0_buffer_funcs;
-		adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring;
-	}
+	adev->mman.buffer_funcs = &sdma_v3_0_buffer_funcs;
+	adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring;
 }
 
 static const struct amdgpu_vm_pte_funcs sdma_v3_0_vm_pte_funcs = {
@@ -1755,15 +1753,13 @@ static void sdma_v3_0_set_vm_pte_funcs(struct amdgpu_device *adev)
 	struct drm_gpu_scheduler *sched;
 	unsigned i;
 
-	if (adev->vm_manager.vm_pte_funcs == NULL) {
-		adev->vm_manager.vm_pte_funcs = &sdma_v3_0_vm_pte_funcs;
-		for (i = 0; i < adev->sdma.num_instances; i++) {
-			sched = &adev->sdma.instance[i].ring.sched;
-			adev->vm_manager.vm_pte_rqs[i] =
-				&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
-		}
-		adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
+	adev->vm_manager.vm_pte_funcs = &sdma_v3_0_vm_pte_funcs;
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		sched = &adev->sdma.instance[i].ring.sched;
+		adev->vm_manager.vm_pte_rqs[i] =
+			&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
 	}
+	adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
 }
 
 const struct amdgpu_ip_block_version sdma_v3_0_ip_block =
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
index 2ea1f0d8f5be..a3e2ed15fff2 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
@@ -1320,9 +1320,15 @@ static int sdma_v4_0_sw_init(void *handle)
 		DRM_INFO("use_doorbell being set to: [%s]\n",
 				ring->use_doorbell?"true":"false");
 
-		ring->doorbell_index = (i == 0) ?
-			(AMDGPU_DOORBELL64_sDMA_ENGINE0 << 1) //get DWORD offset
-			: (AMDGPU_DOORBELL64_sDMA_ENGINE1 << 1); // get DWORD offset
+		if (adev->asic_type == CHIP_VEGA10)
+			ring->doorbell_index = (i == 0) ?
+				(AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE0 << 1) //get DWORD offset
+				: (AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE1 << 1); // get DWORD offset
+		else
+			ring->doorbell_index = (i == 0) ?
+				(AMDGPU_DOORBELL64_sDMA_ENGINE0 << 1) //get DWORD offset
+				: (AMDGPU_DOORBELL64_sDMA_ENGINE1 << 1); // get DWORD offset
+
 
 		sprintf(ring->name, "sdma%d", i);
 		r = amdgpu_ring_init(adev, ring, 1024,
@@ -1801,10 +1807,8 @@ static const struct amdgpu_buffer_funcs sdma_v4_0_buffer_funcs = {
 
 static void sdma_v4_0_set_buffer_funcs(struct amdgpu_device *adev)
 {
-	if (adev->mman.buffer_funcs == NULL) {
-		adev->mman.buffer_funcs = &sdma_v4_0_buffer_funcs;
-		adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring;
-	}
+	adev->mman.buffer_funcs = &sdma_v4_0_buffer_funcs;
+	adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring;
 }
 
 static const struct amdgpu_vm_pte_funcs sdma_v4_0_vm_pte_funcs = {
@@ -1820,15 +1824,13 @@ static void sdma_v4_0_set_vm_pte_funcs(struct amdgpu_device *adev)
 	struct drm_gpu_scheduler *sched;
 	unsigned i;
 
-	if (adev->vm_manager.vm_pte_funcs == NULL) {
-		adev->vm_manager.vm_pte_funcs = &sdma_v4_0_vm_pte_funcs;
-		for (i = 0; i < adev->sdma.num_instances; i++) {
-			sched = &adev->sdma.instance[i].ring.sched;
-			adev->vm_manager.vm_pte_rqs[i] =
-				&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
-		}
-		adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
+	adev->vm_manager.vm_pte_funcs = &sdma_v4_0_vm_pte_funcs;
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		sched = &adev->sdma.instance[i].ring.sched;
+		adev->vm_manager.vm_pte_rqs[i] =
+			&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
 	}
+	adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
 }
 
 const struct amdgpu_ip_block_version sdma_v4_0_ip_block = {
diff --git a/drivers/gpu/drm/amd/amdgpu/si_dma.c b/drivers/gpu/drm/amd/amdgpu/si_dma.c
index fafaf259b17b..d4ceaf440f26 100644
--- a/drivers/gpu/drm/amd/amdgpu/si_dma.c
+++ b/drivers/gpu/drm/amd/amdgpu/si_dma.c
@@ -502,12 +502,12 @@ static int si_dma_sw_init(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	/* DMA0 trap event */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 224, &adev->sdma.trap_irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 224, &adev->sdma.trap_irq);
 	if (r)
 		return r;
 
 	/* DMA1 trap event */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 244, &adev->sdma.trap_irq_1);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 244, &adev->sdma.trap_irq_1);
 	if (r)
 		return r;
 
@@ -863,10 +863,8 @@ static const struct amdgpu_buffer_funcs si_dma_buffer_funcs = {
 
 static void si_dma_set_buffer_funcs(struct amdgpu_device *adev)
 {
-	if (adev->mman.buffer_funcs == NULL) {
-		adev->mman.buffer_funcs = &si_dma_buffer_funcs;
-		adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring;
-	}
+	adev->mman.buffer_funcs = &si_dma_buffer_funcs;
+	adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring;
 }
 
 static const struct amdgpu_vm_pte_funcs si_dma_vm_pte_funcs = {
@@ -882,15 +880,13 @@ static void si_dma_set_vm_pte_funcs(struct amdgpu_device *adev)
 	struct drm_gpu_scheduler *sched;
 	unsigned i;
 
-	if (adev->vm_manager.vm_pte_funcs == NULL) {
-		adev->vm_manager.vm_pte_funcs = &si_dma_vm_pte_funcs;
-		for (i = 0; i < adev->sdma.num_instances; i++) {
-			sched = &adev->sdma.instance[i].ring.sched;
-			adev->vm_manager.vm_pte_rqs[i] =
-				&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
-		}
-		adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
+	adev->vm_manager.vm_pte_funcs = &si_dma_vm_pte_funcs;
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		sched = &adev->sdma.instance[i].ring.sched;
+		adev->vm_manager.vm_pte_rqs[i] =
+			&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
 	}
+	adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
 }
 
 const struct amdgpu_ip_block_version si_dma_ip_block =
diff --git a/drivers/gpu/drm/amd/amdgpu/si_dpm.c b/drivers/gpu/drm/amd/amdgpu/si_dpm.c
index 1de96995e690..da58040fdbdc 100644
--- a/drivers/gpu/drm/amd/amdgpu/si_dpm.c
+++ b/drivers/gpu/drm/amd/amdgpu/si_dpm.c
@@ -7687,11 +7687,11 @@ static int si_dpm_sw_init(void *handle)
 	int ret;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	ret = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 230, &adev->pm.dpm.thermal.irq);
+	ret = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 230, &adev->pm.dpm.thermal.irq);
 	if (ret)
 		return ret;
 
-	ret = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 231, &adev->pm.dpm.thermal.irq);
+	ret = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 231, &adev->pm.dpm.thermal.irq);
 	if (ret)
 		return ret;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/si_ih.c b/drivers/gpu/drm/amd/amdgpu/si_ih.c
index 60dad63098a2..b3d7d9f83202 100644
--- a/drivers/gpu/drm/amd/amdgpu/si_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/si_ih.c
@@ -142,7 +142,7 @@ static void si_ih_decode_iv(struct amdgpu_device *adev,
 	dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]);
 	dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]);
 
-	entry->client_id = AMDGPU_IH_CLIENTID_LEGACY;
+	entry->client_id = AMDGPU_IRQ_CLIENTID_LEGACY;
 	entry->src_id = dw[0] & 0xff;
 	entry->src_data[0] = dw[1] & 0xfffffff;
 	entry->ring_id = dw[2] & 0xff;
@@ -170,7 +170,7 @@ static int si_ih_sw_init(void *handle)
 	int r;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_ih_ring_init(adev, 64 * 1024, false);
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, false);
 	if (r)
 		return r;
 
@@ -182,7 +182,7 @@ static int si_ih_sw_fini(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	amdgpu_irq_fini(adev);
-	amdgpu_ih_ring_fini(adev);
+	amdgpu_ih_ring_fini(adev, &adev->irq.ih);
 
 	return 0;
 }
@@ -308,8 +308,7 @@ static const struct amdgpu_ih_funcs si_ih_funcs = {
 
 static void si_ih_set_interrupt_funcs(struct amdgpu_device *adev)
 {
-	if (adev->irq.ih_funcs == NULL)
-		adev->irq.ih_funcs = &si_ih_funcs;
+	adev->irq.ih_funcs = &si_ih_funcs;
 }
 
 const struct amdgpu_ip_block_version si_ih_ip_block =
diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c
index 138c4810a3de..fb26039beeba 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc15.c
+++ b/drivers/gpu/drm/amd/amdgpu/soc15.c
@@ -739,7 +739,8 @@ static int soc15_common_early_init(void *handle)
 
 			adev->pg_flags = AMD_PG_SUPPORT_SDMA |
 				AMD_PG_SUPPORT_MMHUB |
-				AMD_PG_SUPPORT_VCN;
+				AMD_PG_SUPPORT_VCN |
+				AMD_PG_SUPPORT_VCN_DPG;
 		} else {
 			adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG |
 				AMD_CG_SUPPORT_GFX_MGLS |
diff --git a/drivers/gpu/drm/amd/amdgpu/soc15_common.h b/drivers/gpu/drm/amd/amdgpu/soc15_common.h
index f5d602540673..958b10a57073 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc15_common.h
+++ b/drivers/gpu/drm/amd/amdgpu/soc15_common.h
@@ -57,13 +57,33 @@
 			loop--;					\
 			if (!loop) {				\
 				DRM_ERROR("Register(%d) [%s] failed to reach value 0x%08x != 0x%08x\n", \
-							inst, #reg, expected_value, (tmp_ & (mask))); \
+					  inst, #reg, (unsigned)expected_value, (unsigned)(tmp_ & (mask))); \
 				ret = -ETIMEDOUT;		\
 				break;				\
 			}					\
 		}						\
 	} while (0)
 
+#define RREG32_SOC15_DPG_MODE(ip, inst, reg, mask, sram_sel) 	\
+		({ WREG32_SOC15(ip, inst, mmUVD_DPG_LMA_MASK, mask); \
+			WREG32_SOC15(ip, inst, mmUVD_DPG_LMA_CTL,	\
+				UVD_DPG_LMA_CTL__MASK_EN_MASK |				\
+				((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg) \
+				<< UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT) | \
+				(sram_sel << UVD_DPG_LMA_CTL__SRAM_SEL__SHIFT));	\
+			RREG32_SOC15(ip, inst, mmUVD_DPG_LMA_DATA); })
+
+#define WREG32_SOC15_DPG_MODE(ip, inst, reg, value, mask, sram_sel)	\
+	do {							\
+		WREG32_SOC15(ip, inst, mmUVD_DPG_LMA_DATA, value);	\
+		WREG32_SOC15(ip, inst, mmUVD_DPG_LMA_MASK, mask);		\
+		WREG32_SOC15(ip, inst, mmUVD_DPG_LMA_CTL,	\
+			UVD_DPG_LMA_CTL__READ_WRITE_MASK |	\
+			((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg) \
+			<< UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT) |	\
+			(sram_sel << UVD_DPG_LMA_CTL__SRAM_SEL__SHIFT)); \
+	} while (0)
+
 #endif
 
 
diff --git a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c
index 52853d8a8fdd..3abffd06b5c7 100644
--- a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c
@@ -266,7 +266,7 @@ static void tonga_ih_decode_iv(struct amdgpu_device *adev,
 	dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]);
 	dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]);
 
-	entry->client_id = AMDGPU_IH_CLIENTID_LEGACY;
+	entry->client_id = AMDGPU_IRQ_CLIENTID_LEGACY;
 	entry->src_id = dw[0] & 0xff;
 	entry->src_data[0] = dw[1] & 0xfffffff;
 	entry->ring_id = dw[2] & 0xff;
@@ -317,7 +317,7 @@ static int tonga_ih_sw_init(void *handle)
 	int r;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_ih_ring_init(adev, 64 * 1024, true);
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, true);
 	if (r)
 		return r;
 
@@ -334,7 +334,7 @@ static int tonga_ih_sw_fini(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	amdgpu_irq_fini(adev);
-	amdgpu_ih_ring_fini(adev);
+	amdgpu_ih_ring_fini(adev, &adev->irq.ih);
 	amdgpu_irq_remove_domain(adev);
 
 	return 0;
@@ -513,8 +513,7 @@ static const struct amdgpu_ih_funcs tonga_ih_funcs = {
 
 static void tonga_ih_set_interrupt_funcs(struct amdgpu_device *adev)
 {
-	if (adev->irq.ih_funcs == NULL)
-		adev->irq.ih_funcs = &tonga_ih_funcs;
+	adev->irq.ih_funcs = &tonga_ih_funcs;
 }
 
 const struct amdgpu_ip_block_version tonga_ih_ip_block =
diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c
index 8a926d1df939..1fc17bf39fed 100644
--- a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c
@@ -108,7 +108,7 @@ static int uvd_v4_2_sw_init(void *handle)
 	int r;
 
 	/* UVD TRAP */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 124, &adev->uvd.inst->irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 124, &adev->uvd.inst->irq);
 	if (r)
 		return r;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c
index 50248059412e..fde6ad5ac9ab 100644
--- a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c
@@ -105,7 +105,7 @@ static int uvd_v5_0_sw_init(void *handle)
 	int r;
 
 	/* UVD TRAP */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_UVD_SYSTEM_MESSAGE, &adev->uvd.inst->irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_UVD_SYSTEM_MESSAGE, &adev->uvd.inst->irq);
 	if (r)
 		return r;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c
index 6ae82cc2e55e..8ef4a5392112 100644
--- a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c
@@ -393,14 +393,14 @@ static int uvd_v6_0_sw_init(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	/* UVD TRAP */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_UVD_SYSTEM_MESSAGE, &adev->uvd.inst->irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_UVD_SYSTEM_MESSAGE, &adev->uvd.inst->irq);
 	if (r)
 		return r;
 
 	/* UVD ENC TRAP */
 	if (uvd_v6_0_enc_support(adev)) {
 		for (i = 0; i < adev->uvd.num_enc_rings; ++i) {
-			r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i + VISLANDS30_IV_SRCID_UVD_ENC_GEN_PURP, &adev->uvd.inst->irq);
+			r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + VISLANDS30_IV_SRCID_UVD_ENC_GEN_PURP, &adev->uvd.inst->irq);
 			if (r)
 				return r;
 		}
diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c
index 7eaa54ba016b..ea28828360d3 100644
--- a/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c
@@ -417,7 +417,7 @@ static int vce_v2_0_sw_init(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	/* VCE */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 167, &adev->vce.irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 167, &adev->vce.irq);
 	if (r)
 		return r;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c
index c8390f9adfd6..6dbd39730070 100644
--- a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c
@@ -423,7 +423,7 @@ static int vce_v3_0_sw_init(void *handle)
 	int r, i;
 
 	/* VCE */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_VCE_TRAP, &adev->vce.irq);
+	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_VCE_TRAP, &adev->vce.irq);
 	if (r)
 		return r;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
index 2664bb2c47c3..63d7f97e81b7 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
@@ -198,7 +198,8 @@ static int vcn_v1_0_hw_init(void *handle)
 
 done:
 	if (!r)
-		DRM_INFO("VCN decode and encode initialized successfully.\n");
+		DRM_INFO("VCN decode and encode initialized successfully(under %s).\n",
+			(adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)?"DPG Mode":"SPG Mode");
 
 	return r;
 }
@@ -266,13 +267,13 @@ static int vcn_v1_0_resume(void *handle)
 }
 
 /**
- * vcn_v1_0_mc_resume - memory controller programming
+ * vcn_v1_0_mc_resume_spg_mode - memory controller programming
  *
  * @adev: amdgpu_device pointer
  *
  * Let the VCN memory controller know it's offsets
  */
-static void vcn_v1_0_mc_resume(struct amdgpu_device *adev)
+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 offset;
@@ -319,6 +320,65 @@ static void vcn_v1_0_mc_resume(struct amdgpu_device *adev)
 			adev->gfx.config.gb_addr_config);
 }
 
+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 offset;
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
+			     (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_lo),
+			     0xFFFFFFFF, 0);
+		WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
+			     (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_hi),
+			     0xFFFFFFFF, 0);
+		WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0, 0,
+			     0xFFFFFFFF, 0);
+		offset = 0;
+	} else {
+		WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW,
+			lower_32_bits(adev->vcn.gpu_addr), 0xFFFFFFFF, 0);
+		WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH,
+			upper_32_bits(adev->vcn.gpu_addr), 0xFFFFFFFF, 0);
+		offset = size;
+		WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0,
+			     AMDGPU_UVD_FIRMWARE_OFFSET >> 3, 0xFFFFFFFF, 0);
+	}
+
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_SIZE0, size, 0xFFFFFFFF, 0);
+
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW,
+		     lower_32_bits(adev->vcn.gpu_addr + offset), 0xFFFFFFFF, 0);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH,
+		     upper_32_bits(adev->vcn.gpu_addr + offset), 0xFFFFFFFF, 0);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_OFFSET1, 0,
+			     0xFFFFFFFF, 0);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_HEAP_SIZE,
+			     0xFFFFFFFF, 0);
+
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW,
+		     lower_32_bits(adev->vcn.gpu_addr + offset + AMDGPU_VCN_HEAP_SIZE),
+			     0xFFFFFFFF, 0);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH,
+		     upper_32_bits(adev->vcn.gpu_addr + offset + AMDGPU_VCN_HEAP_SIZE),
+			     0xFFFFFFFF, 0);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_OFFSET2, 0, 0xFFFFFFFF, 0);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_SIZE2,
+			AMDGPU_VCN_STACK_SIZE + (AMDGPU_VCN_SESSION_SIZE * 40),
+			     0xFFFFFFFF, 0);
+
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_UDEC_ADDR_CONFIG,
+			adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_UDEC_DB_ADDR_CONFIG,
+			adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_UDEC_DBW_ADDR_CONFIG,
+			adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_JPEG_ADDR_CONFIG,
+			adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_JPEG_UV_ADDR_CONFIG,
+			adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0);
+}
+
 /**
  * vcn_v1_0_disable_clock_gating - disable VCN clock gating
  *
@@ -519,6 +579,62 @@ static void vcn_v1_0_enable_clock_gating(struct amdgpu_device *adev)
 	WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL, data);
 }
 
+static void vcn_v1_0_clock_gating_dpg_mode(struct amdgpu_device *adev, uint8_t sram_sel)
+{
+	uint32_t reg_data = 0;
+
+	/* disable JPEG CGC */
+	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
+		reg_data = 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	else
+		reg_data = 0 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	reg_data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
+	reg_data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmJPEG_CGC_CTRL, reg_data, 0xFFFFFFFF, sram_sel);
+
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmJPEG_CGC_GATE, 0, 0xFFFFFFFF, sram_sel);
+
+	/* enable sw clock gating control */
+	if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG)
+		reg_data = 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	else
+		reg_data = 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT;
+	reg_data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT;
+	reg_data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT;
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_CGC_CTRL, reg_data, 0xFFFFFFFF, sram_sel);
+
+	reg_data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK |
+		 UVD_CGC_CTRL__UDEC_CM_MODE_MASK |
+		 UVD_CGC_CTRL__UDEC_IT_MODE_MASK |
+		 UVD_CGC_CTRL__UDEC_DB_MODE_MASK |
+		 UVD_CGC_CTRL__UDEC_MP_MODE_MASK |
+		 UVD_CGC_CTRL__SYS_MODE_MASK |
+		 UVD_CGC_CTRL__UDEC_MODE_MASK |
+		 UVD_CGC_CTRL__MPEG2_MODE_MASK |
+		 UVD_CGC_CTRL__REGS_MODE_MASK |
+		 UVD_CGC_CTRL__RBC_MODE_MASK |
+		 UVD_CGC_CTRL__LMI_MC_MODE_MASK |
+		 UVD_CGC_CTRL__LMI_UMC_MODE_MASK |
+		 UVD_CGC_CTRL__IDCT_MODE_MASK |
+		 UVD_CGC_CTRL__MPRD_MODE_MASK |
+		 UVD_CGC_CTRL__MPC_MODE_MASK |
+		 UVD_CGC_CTRL__LBSI_MODE_MASK |
+		 UVD_CGC_CTRL__LRBBM_MODE_MASK |
+		 UVD_CGC_CTRL__WCB_MODE_MASK |
+		 UVD_CGC_CTRL__VCPU_MODE_MASK |
+		 UVD_CGC_CTRL__SCPU_MODE_MASK);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_CGC_CTRL, reg_data, 0xFFFFFFFF, sram_sel);
+
+	/* turn off clock gating */
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_CGC_GATE, 0, 0xFFFFFFFF, sram_sel);
+
+	/* turn on SUVD clock gating */
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SUVD_CGC_GATE, 1, 0xFFFFFFFF, sram_sel);
+
+	/* turn on sw mode in UVD_SUVD_CGC_CTRL */
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SUVD_CGC_CTRL, 0, 0xFFFFFFFF, sram_sel);
+}
+
 static void vcn_1_0_disable_static_power_gating(struct amdgpu_device *adev)
 {
 	uint32_t data = 0;
@@ -614,7 +730,7 @@ static void vcn_1_0_enable_static_power_gating(struct amdgpu_device *adev)
  *
  * Setup and start the VCN block
  */
-static int vcn_v1_0_start(struct amdgpu_device *adev)
+static int vcn_v1_0_start_spg_mode(struct amdgpu_device *adev)
 {
 	struct amdgpu_ring *ring = &adev->vcn.ring_dec;
 	uint32_t rb_bufsz, tmp;
@@ -628,7 +744,7 @@ static int vcn_v1_0_start(struct amdgpu_device *adev)
 	/* disable clock gating */
 	vcn_v1_0_disable_clock_gating(adev);
 
-	vcn_v1_0_mc_resume(adev);
+	vcn_v1_0_mc_resume_spg_mode(adev);
 
 	/* disable interupt */
 	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN), 0,
@@ -799,6 +915,170 @@ static int vcn_v1_0_start(struct amdgpu_device *adev)
 	return 0;
 }
 
+static int vcn_v1_0_start_dpg_mode(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring = &adev->vcn.ring_dec;
+	uint32_t rb_bufsz, tmp, reg_data;
+	uint32_t lmi_swap_cntl;
+
+	/* disable byte swapping */
+	lmi_swap_cntl = 0;
+
+	vcn_1_0_enable_static_power_gating(adev);
+
+	/* enable dynamic power gating mode */
+	reg_data = RREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS);
+	reg_data |= UVD_POWER_STATUS__UVD_PG_MODE_MASK;
+	reg_data |= UVD_POWER_STATUS__UVD_PG_EN_MASK;
+	WREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS, reg_data);
+
+	/* enable clock gating */
+	vcn_v1_0_clock_gating_dpg_mode(adev, 0);
+
+	/* enable VCPU clock */
+	reg_data = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT);
+	reg_data |= UVD_VCPU_CNTL__CLK_EN_MASK;
+	reg_data |= UVD_VCPU_CNTL__MIF_WR_LOW_THRESHOLD_BP_MASK;
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CNTL, reg_data, 0xFFFFFFFF, 0);
+
+	/* disable interupt */
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MASTINT_EN,
+			0, UVD_MASTINT_EN__VCPU_EN_MASK, 0);
+
+	/* stall UMC and register bus before resetting VCPU */
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_CTRL2,
+			UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, 0);
+
+	/* put LMI, VCPU, RBC etc... into reset */
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SOFT_RESET,
+		UVD_SOFT_RESET__LMI_SOFT_RESET_MASK |
+		UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK |
+		UVD_SOFT_RESET__LBSI_SOFT_RESET_MASK |
+		UVD_SOFT_RESET__RBC_SOFT_RESET_MASK |
+		UVD_SOFT_RESET__CSM_SOFT_RESET_MASK |
+		UVD_SOFT_RESET__CXW_SOFT_RESET_MASK |
+		UVD_SOFT_RESET__TAP_SOFT_RESET_MASK |
+		UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK,
+		0xFFFFFFFF, 0);
+
+	/* initialize VCN memory controller */
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_CTRL,
+		(0x40 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) |
+		UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
+		UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
+		UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK |
+		UVD_LMI_CTRL__REQ_MODE_MASK |
+		0x00100000L, 0xFFFFFFFF, 0);
+
+#ifdef __BIG_ENDIAN
+	/* swap (8 in 32) RB and IB */
+	lmi_swap_cntl = 0xa;
+#endif
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_SWAP_CNTL, lmi_swap_cntl, 0xFFFFFFFF, 0);
+
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_SET_MUXA0, 0x40c2040, 0xFFFFFFFF, 0);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_SET_MUXA1, 0x0, 0xFFFFFFFF, 0);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_SET_MUXB0, 0x40c2040, 0xFFFFFFFF, 0);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_SET_MUXB1, 0x0, 0xFFFFFFFF, 0);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_SET_ALU, 0, 0xFFFFFFFF, 0);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_SET_MUX, 0x88, 0xFFFFFFFF, 0);
+
+	vcn_v1_0_mc_resume_dpg_mode(adev);
+
+	/* take all subblocks out of reset, except VCPU */
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SOFT_RESET,
+			UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK, 0xFFFFFFFF, 0);
+
+	/* enable VCPU clock */
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CNTL,
+			UVD_VCPU_CNTL__CLK_EN_MASK, 0xFFFFFFFF, 0);
+
+	/* enable UMC */
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_CTRL2,
+			0, UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, 0);
+
+	/* boot up the VCPU */
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SOFT_RESET, 0, 0xFFFFFFFF, 0);
+
+	/* enable master interrupt */
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MASTINT_EN,
+			(UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK),
+			(UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK), 0);
+
+	vcn_v1_0_clock_gating_dpg_mode(adev, 1);
+	/* setup mmUVD_LMI_CTRL */
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_CTRL,
+			(UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
+				UVD_LMI_CTRL__CRC_RESET_MASK |
+				UVD_LMI_CTRL__MASK_MC_URGENT_MASK |
+				UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
+				UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK |
+				(8 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) |
+				0x00100000L), 0xFFFFFFFF, 1);
+
+	tmp = adev->gfx.config.gb_addr_config;
+	/* setup VCN global tiling registers */
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_JPEG_ADDR_CONFIG, tmp, 0xFFFFFFFF, 1);
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_JPEG_UV_ADDR_CONFIG, tmp, 0xFFFFFFFF, 1);
+
+	/* enable System Interrupt for JRBC */
+	WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SYS_INT_EN,
+									UVD_SYS_INT_EN__UVD_JRBC_EN_MASK, 0xFFFFFFFF, 1);
+
+	/* force RBC into idle state */
+	rb_bufsz = order_base_2(ring->ring_size);
+	tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz);
+	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1);
+	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1);
+	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_WPTR_POLL_EN, 0);
+	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1);
+	tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1);
+	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_CNTL, tmp);
+
+	/* set the write pointer delay */
+	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR_CNTL, 0);
+
+	/* set the wb address */
+	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR_ADDR,
+								(upper_32_bits(ring->gpu_addr) >> 2));
+
+	/* programm the RB_BASE for ring buffer */
+	WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW,
+								lower_32_bits(ring->gpu_addr));
+	WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH,
+								upper_32_bits(ring->gpu_addr));
+
+	/* Initialize the ring buffer's read and write pointers */
+	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR, 0);
+
+	ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR);
+	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR,
+								lower_32_bits(ring->wptr));
+
+	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_CNTL), 0,
+			~UVD_RBC_RB_CNTL__RB_NO_FETCH_MASK);
+
+	/* initialize wptr */
+	ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR);
+
+	/* copy patch commands to the jpeg ring */
+	vcn_v1_0_jpeg_ring_set_patch_ring(ring,
+		(ring->wptr + ring->max_dw * amdgpu_sched_hw_submission));
+
+	return 0;
+}
+
+static int vcn_v1_0_start(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
+		r = vcn_v1_0_start_dpg_mode(adev);
+	else
+		r = vcn_v1_0_start_spg_mode(adev);
+	return r;
+}
+
 /**
  * vcn_v1_0_stop - stop VCN block
  *
@@ -806,7 +1086,7 @@ static int vcn_v1_0_start(struct amdgpu_device *adev)
  *
  * stop the VCN block
  */
-static int vcn_v1_0_stop(struct amdgpu_device *adev)
+static int vcn_v1_0_stop_spg_mode(struct amdgpu_device *adev)
 {
 	/* force RBC into idle state */
 	WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_CNTL, 0x11010101);
@@ -836,6 +1116,33 @@ static int vcn_v1_0_stop(struct amdgpu_device *adev)
 	return 0;
 }
 
+static int vcn_v1_0_stop_dpg_mode(struct amdgpu_device *adev)
+{
+	int ret_code;
+
+	/* Wait for power status to be 1 */
+	SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, 0x1,
+			UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code);
+
+	/* disable dynamic power gating mode */
+	WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_POWER_STATUS), 0,
+			~UVD_POWER_STATUS__UVD_PG_MODE_MASK);
+
+	return 0;
+}
+
+static int vcn_v1_0_stop(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)
+		r = vcn_v1_0_stop_dpg_mode(adev);
+	else
+		r = vcn_v1_0_stop_spg_mode(adev);
+
+	return r;
+}
+
 static bool vcn_v1_0_is_idle(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
@@ -1633,12 +1940,20 @@ static int vcn_v1_0_set_powergating_state(void *handle,
 	 * revisit this when there is a cleaner line between
 	 * the smc and the hw blocks
 	 */
+	int ret;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
+	if(state == adev->vcn.cur_state)
+		return 0;
+
 	if (state == AMD_PG_STATE_GATE)
-		return vcn_v1_0_stop(adev);
+		ret = vcn_v1_0_stop(adev);
 	else
-		return vcn_v1_0_start(adev);
+		ret = vcn_v1_0_start(adev);
+
+	if(!ret)
+		adev->vcn.cur_state = state;
+	return ret;
 }
 
 static const struct amd_ip_funcs vcn_v1_0_ip_funcs = {
diff --git a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c
index acbe5a770207..a99f71797aa3 100644
--- a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c
@@ -380,7 +380,7 @@ static int vega10_ih_sw_init(void *handle)
 	int r;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_ih_ring_init(adev, 256 * 1024, true);
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 256 * 1024, true);
 	if (r)
 		return r;
 
@@ -397,7 +397,7 @@ static int vega10_ih_sw_fini(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	amdgpu_irq_fini(adev);
-	amdgpu_ih_ring_fini(adev);
+	amdgpu_ih_ring_fini(adev, &adev->irq.ih);
 
 	return 0;
 }
@@ -494,8 +494,7 @@ static const struct amdgpu_ih_funcs vega10_ih_funcs = {
 
 static void vega10_ih_set_interrupt_funcs(struct amdgpu_device *adev)
 {
-	if (adev->irq.ih_funcs == NULL)
-		adev->irq.ih_funcs = &vega10_ih_funcs;
+	adev->irq.ih_funcs = &vega10_ih_funcs;
 }
 
 const struct amdgpu_ip_block_version vega10_ih_ip_block =