Merge tag 'drm-fixes-for-v4.12-rc1' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
"AMD, nouveau, one i915, and one EDID fix for v4.12-rc1

Some fixes that it would be good to have in rc1. It contains the i915
quiet fix that you reported.

It also has an amdgpu fixes pull, with lots of ongoing work on Vega10
which is new in this kernel and is preliminary support so may have a
fair bit of movement.

Otherwise a few non-Vega10 AMD fixes, one EDID fix and some nouveau
regression fixers"

* tag 'drm-fixes-for-v4.12-rc1' of git://people.freedesktop.org/~airlied/linux: (144 commits)
drm/i915: Make vblank evade warnings optional
drm/nouveau/therm: remove ineffective workarounds for alarm bugs
drm/nouveau/tmr: avoid processing completed alarms when adding a new one
drm/nouveau/tmr: fix corruption of the pending list when rescheduling an alarm
drm/nouveau/tmr: handle races with hw when updating the next alarm time
drm/nouveau/tmr: ack interrupt before processing alarms
drm/nouveau/core: fix static checker warning
drm/nouveau/fb/ram/gf100-: remove 0x10f200 read
drm/nouveau/kms/nv50: skip core channel cursor update on position-only changes
drm/nouveau/kms/nv50: fix source-rect-only plane updates
drm/nouveau/kms/nv50: remove pointless argument to window atomic_check_acquire()
drm/amd/powerplay: refine pwm1_enable callback functions for CI.
drm/amd/powerplay: refine pwm1_enable callback functions for vi.
drm/amd/powerplay: refine pwm1_enable callback functions for Vega10.
drm/amdgpu: refine amdgpu pwm1_enable sysfs interface.
drm/amdgpu: add amd fan ctrl mode enums.
drm/amd/powerplay: add more smu message on Vega10.
drm/amdgpu: fix dependency issue
drm/amd: fix init order of sched job
drm/amdgpu: add some additional vega10 pci ids
...

drivers/gpu/drm/amd/amdgpu/amdgpu.h

···
 extern int amdgpu_cntl_sb_buf_per_se;
 extern int amdgpu_param_buf_per_se;
 
+#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)
···
 	unsigned mc_arb_ramcfg;
 	unsigned gb_addr_config;
 	unsigned num_rbs;
+	unsigned gs_vgt_table_depth;
+	unsigned gs_prim_buffer_depth;
 
 	uint32_t tile_mode_array[32];
 	uint32_t macrotile_mode_array[16];
···
 struct amdgpu_cu_info {
 	uint32_t number; /* total active CU number */
 	uint32_t ao_cu_mask;
+	uint32_t wave_front_size;
 	uint32_t bitmap[4][4];
 };
 
···
 };
 
 enum {
-	PRIM = 0,
-	POS,
-	CNTL,
-	PARAM,
+	NGG_PRIM = 0,
+	NGG_POS,
+	NGG_CNTL,
+	NGG_PARAM,
 	NGG_BUF_MAX
 };
 
···
 	void			*owner;
 	uint64_t		fence_ctx; /* the fence_context this job uses */
 	bool                    vm_needs_flush;
+	bool			need_pipeline_sync;
 	unsigned		vm_id;
 	uint64_t		vm_pd_addr;
 	uint32_t		gds_base, gds_size;
···
 
 #define WREG32_FIELD_OFFSET(reg, offset, field, val)	\
 	WREG32(mm##reg + offset, (RREG32(mm##reg + offset) & ~REG_FIELD_MASK(reg, field)) | (val) << REG_FIELD_SHIFT(reg, field))
-
-#define WREG32_FIELD15(ip, idx, reg, field, val)	\
-	WREG32(SOC15_REG_OFFSET(ip, idx, mm##reg), (RREG32(SOC15_REG_OFFSET(ip, idx, mm##reg)) & ~REG_FIELD_MASK(reg, field)) | (val) << REG_FIELD_SHIFT(reg, field))
 
 /*
  * BIOS helpers.

drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c

···
 {
 	int i;
 
+	/*
+	 * VBIOS will check ASIC_INIT_COMPLETE bit to decide if
+	 * execute ASIC_Init posting via driver
+	 */
+	adev->bios_scratch[7] &= ~ATOM_S7_ASIC_INIT_COMPLETE_MASK;
+
 	for (i = 0; i < AMDGPU_BIOS_NUM_SCRATCH; i++)
 		WREG32(mmBIOS_SCRATCH_0 + i, adev->bios_scratch[i]);
 }

drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c

···
 #include "atomfirmware.h"
 #include "amdgpu_atomfirmware.h"
 #include "atom.h"
+#include "atombios.h"
 
 #define get_index_into_master_table(master_table, table_name) (offsetof(struct master_table, table_name) / sizeof(uint16_t))
 
···
 {
 	int i;
 
+	/*
+	 * VBIOS will check ASIC_INIT_COMPLETE bit to decide if
+	 * execute ASIC_Init posting via driver
+	 */
+	adev->bios_scratch[7] &= ~ATOM_S7_ASIC_INIT_COMPLETE_MASK;
+
 	for (i = 0; i < AMDGPU_BIOS_NUM_SCRATCH; i++)
 		WREG32(adev->bios_scratch_reg_offset + i, adev->bios_scratch[i]);
+}
+
+void amdgpu_atomfirmware_scratch_regs_engine_hung(struct amdgpu_device *adev,
+						  bool hung)
+{
+	u32 tmp = RREG32(adev->bios_scratch_reg_offset + 3);
+
+	if (hung)
+		tmp |= ATOM_S3_ASIC_GUI_ENGINE_HUNG;
+	else
+		tmp &= ~ATOM_S3_ASIC_GUI_ENGINE_HUNG;
+
+	WREG32(adev->bios_scratch_reg_offset + 3, tmp);
 }
 
 int amdgpu_atomfirmware_allocate_fb_scratch(struct amdgpu_device *adev)

drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h

···
 void amdgpu_atomfirmware_scratch_regs_init(struct amdgpu_device *adev);
 void amdgpu_atomfirmware_scratch_regs_save(struct amdgpu_device *adev);
 void amdgpu_atomfirmware_scratch_regs_restore(struct amdgpu_device *adev);
+void amdgpu_atomfirmware_scratch_regs_engine_hung(struct amdgpu_device *adev,
+						  bool hung);
 int amdgpu_atomfirmware_allocate_fb_scratch(struct amdgpu_device *adev);
 
 #endif

drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c

···
 	}
 
 out_cleanup:
+	/* Check error value now. The value can be overwritten when clean up.*/
+	if (r) {
+		DRM_ERROR("Error while benchmarking BO move.\n");
+	}
+
 	if (sobj) {
-		r = amdgpu_bo_reserve(sobj, false);
+		r = amdgpu_bo_reserve(sobj, true);
 		if (likely(r == 0)) {
 			amdgpu_bo_unpin(sobj);
 			amdgpu_bo_unreserve(sobj);
···
 		amdgpu_bo_unref(&sobj);
 	}
 	if (dobj) {
-		r = amdgpu_bo_reserve(dobj, false);
+		r = amdgpu_bo_reserve(dobj, true);
 		if (likely(r == 0)) {
 			amdgpu_bo_unpin(dobj);
 			amdgpu_bo_unreserve(dobj);
 		}
 		amdgpu_bo_unref(&dobj);
-	}
-
-	if (r) {
-		DRM_ERROR("Error while benchmarking BO move.\n");
 	}
 }
 

drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c

···
 	struct amdgpu_device *adev =					\
 		((struct amdgpu_cgs_device *)cgs_device)->adev
 
-static int amdgpu_cgs_gpu_mem_info(struct cgs_device *cgs_device, enum cgs_gpu_mem_type type,
-				   uint64_t *mc_start, uint64_t *mc_size,
-				   uint64_t *mem_size)
-{
-	CGS_FUNC_ADEV;
-	switch(type) {
-	case CGS_GPU_MEM_TYPE__VISIBLE_CONTIG_FB:
-	case CGS_GPU_MEM_TYPE__VISIBLE_FB:
-		*mc_start = 0;
-		*mc_size = adev->mc.visible_vram_size;
-		*mem_size = adev->mc.visible_vram_size - adev->vram_pin_size;
-		break;
-	case CGS_GPU_MEM_TYPE__INVISIBLE_CONTIG_FB:
-	case CGS_GPU_MEM_TYPE__INVISIBLE_FB:
-		*mc_start = adev->mc.visible_vram_size;
-		*mc_size = adev->mc.real_vram_size - adev->mc.visible_vram_size;
-		*mem_size = *mc_size;
-		break;
-	case CGS_GPU_MEM_TYPE__GART_CACHEABLE:
-	case CGS_GPU_MEM_TYPE__GART_WRITECOMBINE:
-		*mc_start = adev->mc.gtt_start;
-		*mc_size = adev->mc.gtt_size;
-		*mem_size = adev->mc.gtt_size - adev->gart_pin_size;
-		break;
-	default:
-		return -EINVAL;
-	}
-
-	return 0;
-}
-
-static int amdgpu_cgs_gmap_kmem(struct cgs_device *cgs_device, void *kmem,
-				uint64_t size,
-				uint64_t min_offset, uint64_t max_offset,
-				cgs_handle_t *kmem_handle, uint64_t *mcaddr)
-{
-	CGS_FUNC_ADEV;
-	int ret;
-	struct amdgpu_bo *bo;
-	struct page *kmem_page = vmalloc_to_page(kmem);
-	int npages = ALIGN(size, PAGE_SIZE) >> PAGE_SHIFT;
-
-	struct sg_table *sg = drm_prime_pages_to_sg(&kmem_page, npages);
-	ret = amdgpu_bo_create(adev, size, PAGE_SIZE, false,
-			       AMDGPU_GEM_DOMAIN_GTT, 0, sg, NULL, &bo);
-	if (ret)
-		return ret;
-	ret = amdgpu_bo_reserve(bo, false);
-	if (unlikely(ret != 0))
-		return ret;
-
-	/* pin buffer into GTT */
-	ret = amdgpu_bo_pin_restricted(bo, AMDGPU_GEM_DOMAIN_GTT,
-				       min_offset, max_offset, mcaddr);
-	amdgpu_bo_unreserve(bo);
-
-	*kmem_handle = (cgs_handle_t)bo;
-	return ret;
-}
-
-static int amdgpu_cgs_gunmap_kmem(struct cgs_device *cgs_device, cgs_handle_t kmem_handle)
-{
-	struct amdgpu_bo *obj = (struct amdgpu_bo *)kmem_handle;
-
-	if (obj) {
-		int r = amdgpu_bo_reserve(obj, false);
-		if (likely(r == 0)) {
-			amdgpu_bo_unpin(obj);
-			amdgpu_bo_unreserve(obj);
-		}
-		amdgpu_bo_unref(&obj);
-
-	}
-	return 0;
-}
-
 static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device,
 				    enum cgs_gpu_mem_type type,
 				    uint64_t size, uint64_t align,
···
 	struct amdgpu_bo *obj = (struct amdgpu_bo *)handle;
 
 	if (obj) {
-		int r = amdgpu_bo_reserve(obj, false);
+		int r = amdgpu_bo_reserve(obj, true);
 		if (likely(r == 0)) {
 			amdgpu_bo_kunmap(obj);
 			amdgpu_bo_unpin(obj);
···
 	min_offset = obj->placements[0].fpfn << PAGE_SHIFT;
 	max_offset = obj->placements[0].lpfn << PAGE_SHIFT;
 
-	r = amdgpu_bo_reserve(obj, false);
+	r = amdgpu_bo_reserve(obj, true);
 	if (unlikely(r != 0))
 		return r;
 	r = amdgpu_bo_pin_restricted(obj, obj->prefered_domains,
···
 {
 	int r;
 	struct amdgpu_bo *obj = (struct amdgpu_bo *)handle;
-	r = amdgpu_bo_reserve(obj, false);
+	r = amdgpu_bo_reserve(obj, true);
 	if (unlikely(r != 0))
 		return r;
 	r = amdgpu_bo_unpin(obj);
···
 {
 	int r;
 	struct amdgpu_bo *obj = (struct amdgpu_bo *)handle;
-	r = amdgpu_bo_reserve(obj, false);
+	r = amdgpu_bo_reserve(obj, true);
 	if (unlikely(r != 0))
 		return r;
 	r = amdgpu_bo_kmap(obj, map);
···
 {
 	int r;
 	struct amdgpu_bo *obj = (struct amdgpu_bo *)handle;
-	r = amdgpu_bo_reserve(obj, false);
+	r = amdgpu_bo_reserve(obj, true);
 	if (unlikely(r != 0))
 		return r;
 	amdgpu_bo_kunmap(obj);
···
 	WARN(1, "Invalid indirect register space");
 }
 
-static uint8_t amdgpu_cgs_read_pci_config_byte(struct cgs_device *cgs_device, unsigned addr)
-{
-	CGS_FUNC_ADEV;
-	uint8_t val;
-	int ret = pci_read_config_byte(adev->pdev, addr, &val);
-	if (WARN(ret, "pci_read_config_byte error"))
-		return 0;
-	return val;
-}
-
-static uint16_t amdgpu_cgs_read_pci_config_word(struct cgs_device *cgs_device, unsigned addr)
-{
-	CGS_FUNC_ADEV;
-	uint16_t val;
-	int ret = pci_read_config_word(adev->pdev, addr, &val);
-	if (WARN(ret, "pci_read_config_word error"))
-		return 0;
-	return val;
-}
-
-static uint32_t amdgpu_cgs_read_pci_config_dword(struct cgs_device *cgs_device,
-						 unsigned addr)
-{
-	CGS_FUNC_ADEV;
-	uint32_t val;
-	int ret = pci_read_config_dword(adev->pdev, addr, &val);
-	if (WARN(ret, "pci_read_config_dword error"))
-		return 0;
-	return val;
-}
-
-static void amdgpu_cgs_write_pci_config_byte(struct cgs_device *cgs_device, unsigned addr,
-					     uint8_t value)
-{
-	CGS_FUNC_ADEV;
-	int ret = pci_write_config_byte(adev->pdev, addr, value);
-	WARN(ret, "pci_write_config_byte error");
-}
-
-static void amdgpu_cgs_write_pci_config_word(struct cgs_device *cgs_device, unsigned addr,
-					     uint16_t value)
-{
-	CGS_FUNC_ADEV;
-	int ret = pci_write_config_word(adev->pdev, addr, value);
-	WARN(ret, "pci_write_config_word error");
-}
-
-static void amdgpu_cgs_write_pci_config_dword(struct cgs_device *cgs_device, unsigned addr,
-					      uint32_t value)
-{
-	CGS_FUNC_ADEV;
-	int ret = pci_write_config_dword(adev->pdev, addr, value);
-	WARN(ret, "pci_write_config_dword error");
-}
-
-
 static int amdgpu_cgs_get_pci_resource(struct cgs_device *cgs_device,
 				       enum cgs_resource_type resource_type,
 				       uint64_t size,
···
 
 	return amdgpu_atom_execute_table(
 		adev->mode_info.atom_context, table, args);
-}
-
-static int amdgpu_cgs_create_pm_request(struct cgs_device *cgs_device, cgs_handle_t *request)
-{
-	/* TODO */
-	return 0;
-}
-
-static int amdgpu_cgs_destroy_pm_request(struct cgs_device *cgs_device, cgs_handle_t request)
-{
-	/* TODO */
-	return 0;
-}
-
-static int amdgpu_cgs_set_pm_request(struct cgs_device *cgs_device, cgs_handle_t request,
-				     int active)
-{
-	/* TODO */
-	return 0;
-}
-
-static int amdgpu_cgs_pm_request_clock(struct cgs_device *cgs_device, cgs_handle_t request,
-				       enum cgs_clock clock, unsigned freq)
-{
-	/* TODO */
-	return 0;
-}
-
-static int amdgpu_cgs_pm_request_engine(struct cgs_device *cgs_device, cgs_handle_t request,
-					enum cgs_engine engine, int powered)
-{
-	/* TODO */
-	return 0;
-}
-
-
-
-static int amdgpu_cgs_pm_query_clock_limits(struct cgs_device *cgs_device,
-					    enum cgs_clock clock,
-					    struct cgs_clock_limits *limits)
-{
-	/* TODO */
-	return 0;
-}
-
-static int amdgpu_cgs_set_camera_voltages(struct cgs_device *cgs_device, uint32_t mask,
-					  const uint32_t *voltages)
-{
-	DRM_ERROR("not implemented");
-	return -EPERM;
 }
 
 struct cgs_irq_params {
···
 }
 
 static const struct cgs_ops amdgpu_cgs_ops = {
-	.gpu_mem_info = amdgpu_cgs_gpu_mem_info,
-	.gmap_kmem = amdgpu_cgs_gmap_kmem,
-	.gunmap_kmem = amdgpu_cgs_gunmap_kmem,
 	.alloc_gpu_mem = amdgpu_cgs_alloc_gpu_mem,
 	.free_gpu_mem = amdgpu_cgs_free_gpu_mem,
 	.gmap_gpu_mem = amdgpu_cgs_gmap_gpu_mem,
···
 	.write_register = amdgpu_cgs_write_register,
 	.read_ind_register = amdgpu_cgs_read_ind_register,
 	.write_ind_register = amdgpu_cgs_write_ind_register,
-	.read_pci_config_byte = amdgpu_cgs_read_pci_config_byte,
-	.read_pci_config_word = amdgpu_cgs_read_pci_config_word,
-	.read_pci_config_dword = amdgpu_cgs_read_pci_config_dword,
-	.write_pci_config_byte = amdgpu_cgs_write_pci_config_byte,
-	.write_pci_config_word = amdgpu_cgs_write_pci_config_word,
-	.write_pci_config_dword = amdgpu_cgs_write_pci_config_dword,
 	.get_pci_resource = amdgpu_cgs_get_pci_resource,
 	.atom_get_data_table = amdgpu_cgs_atom_get_data_table,
 	.atom_get_cmd_table_revs = amdgpu_cgs_atom_get_cmd_table_revs,
 	.atom_exec_cmd_table = amdgpu_cgs_atom_exec_cmd_table,
-	.create_pm_request = amdgpu_cgs_create_pm_request,
-	.destroy_pm_request = amdgpu_cgs_destroy_pm_request,
-	.set_pm_request = amdgpu_cgs_set_pm_request,
-	.pm_request_clock = amdgpu_cgs_pm_request_clock,
-	.pm_request_engine = amdgpu_cgs_pm_request_engine,
-	.pm_query_clock_limits = amdgpu_cgs_pm_query_clock_limits,
-	.set_camera_voltages = amdgpu_cgs_set_camera_voltages,
 	.get_firmware_info = amdgpu_cgs_get_firmware_info,
 	.rel_firmware = amdgpu_cgs_rel_firmware,
 	.set_powergating_state = amdgpu_cgs_set_powergating_state,

drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c

···
 	cs->out.handle = amdgpu_ctx_add_fence(p->ctx, ring, p->fence);
 	job->uf_sequence = cs->out.handle;
 	amdgpu_job_free_resources(job);
+	amdgpu_cs_parser_fini(p, 0, true);
 
 	trace_amdgpu_cs_ioctl(job);
 	amd_sched_entity_push_job(&job->base);
···
 		goto out;
 
 	r = amdgpu_cs_submit(&parser, cs);
+	if (r)
+		goto out;
 
+	return 0;
 out:
 	amdgpu_cs_parser_fini(&parser, r, reserved_buffers);
 	return r;

drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c

···
 
 	spin_lock(&ctx->ring_lock);
 
+	if (seq == ~0ull)
+		seq = ctx->rings[ring->idx].sequence - 1;
+
 	if (seq >= cring->sequence) {
 		spin_unlock(&ctx->ring_lock);
 		return ERR_PTR(-EINVAL);

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c

···
 #include "bif/bif_4_1_d.h"
 #include <linux/pci.h>
 #include <linux/firmware.h>
-#include "amdgpu_pm.h"
 
 static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev);
 static void amdgpu_debugfs_regs_cleanup(struct amdgpu_device *adev);
···
 	if (adev->vram_scratch.robj == NULL) {
 		return;
 	}
-	r = amdgpu_bo_reserve(adev->vram_scratch.robj, false);
+	r = amdgpu_bo_reserve(adev->vram_scratch.robj, true);
 	if (likely(r == 0)) {
 		amdgpu_bo_kunmap(adev->vram_scratch.robj);
 		amdgpu_bo_unpin(adev->vram_scratch.robj);
···
 	if (adev->doorbell.num_doorbells == 0)
 		return -EINVAL;
 
-	adev->doorbell.ptr = ioremap(adev->doorbell.base, adev->doorbell.num_doorbells * sizeof(u32));
-	if (adev->doorbell.ptr == NULL) {
+	adev->doorbell.ptr = ioremap(adev->doorbell.base,
+				     adev->doorbell.num_doorbells *
+				     sizeof(u32));
+	if (adev->doorbell.ptr == NULL)
 		return -ENOMEM;
-	}
-	DRM_INFO("doorbell mmio base: 0x%08X\n", (uint32_t)adev->doorbell.base);
-	DRM_INFO("doorbell mmio size: %u\n", (unsigned)adev->doorbell.size);
 
 	return 0;
 }
···
 		}
 	}
 
-	amdgpu_dpm_enable_uvd(adev, false);
-	amdgpu_dpm_enable_vce(adev, false);
-
 	return 0;
 }
 
···
 
 	/* mutex initialization are all done here so we
 	 * can recall function without having locking issues */
-	mutex_init(&adev->vm_manager.lock);
 	atomic_set(&adev->irq.ih.lock, 0);
 	mutex_init(&adev->firmware.mutex);
 	mutex_init(&adev->pm.mutex);
···
 
 	DRM_INFO("amdgpu: finishing device.\n");
 	adev->shutdown = true;
-	drm_crtc_force_disable_all(adev->ddev);
+	if (adev->mode_info.mode_config_initialized)
+		drm_crtc_force_disable_all(adev->ddev);
 	/* evict vram memory */
 	amdgpu_bo_evict_vram(adev);
 	amdgpu_ib_pool_fini(adev);
···
 
 		if (amdgpu_crtc->cursor_bo) {
 			struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
-			r = amdgpu_bo_reserve(aobj, false);
+			r = amdgpu_bo_reserve(aobj, true);
 			if (r == 0) {
 				amdgpu_bo_unpin(aobj);
 				amdgpu_bo_unreserve(aobj);
···
 		robj = gem_to_amdgpu_bo(rfb->obj);
 		/* don't unpin kernel fb objects */
 		if (!amdgpu_fbdev_robj_is_fb(adev, robj)) {
-			r = amdgpu_bo_reserve(robj, false);
+			r = amdgpu_bo_reserve(robj, true);
 			if (r == 0) {
 				amdgpu_bo_unpin(robj);
 				amdgpu_bo_unreserve(robj);
···
 	struct drm_connector *connector;
 	struct amdgpu_device *adev = dev->dev_private;
 	struct drm_crtc *crtc;
-	int r;
+	int r = 0;
 
 	if (dev->switch_power_state == DRM_SWITCH_POWER_OFF)
 		return 0;
···
 		pci_set_power_state(dev->pdev, PCI_D0);
 		pci_restore_state(dev->pdev);
 		r = pci_enable_device(dev->pdev);
-		if (r) {
-			if (fbcon)
-				console_unlock();
-			return r;
-		}
+		if (r)
+			goto unlock;
 	}
 	if (adev->is_atom_fw)
 		amdgpu_atomfirmware_scratch_regs_restore(adev);
···
 	r = amdgpu_resume(adev);
 	if (r) {
 		DRM_ERROR("amdgpu_resume failed (%d).\n", r);
-		return r;
+		goto unlock;
 	}
 	amdgpu_fence_driver_resume(adev);
 
···
 	}
 
 	r = amdgpu_late_init(adev);
-	if (r) {
-		if (fbcon)
-			console_unlock();
-		return r;
-	}
+	if (r)
+		goto unlock;
 
 	/* pin cursors */
 	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
···
 
 		if (amdgpu_crtc->cursor_bo) {
 			struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
-			r = amdgpu_bo_reserve(aobj, false);
+			r = amdgpu_bo_reserve(aobj, true);
 			if (r == 0) {
 				r = amdgpu_bo_pin(aobj,
 						  AMDGPU_GEM_DOMAIN_VRAM,
···
 	dev->dev->power.disable_depth--;
 #endif
 
-	if (fbcon) {
+	if (fbcon)
 		amdgpu_fbdev_set_suspend(adev, 0);
-		console_unlock();
-	}
 
-	return 0;
+unlock:
+	if (fbcon)
+		console_unlock();
+
+	return r;
 }
 
 static bool amdgpu_check_soft_reset(struct amdgpu_device *adev)
···
 	uint32_t domain;
 	int r;
 
-       if (!bo->shadow)
-               return 0;
+	if (!bo->shadow)
+		return 0;
 
-       r = amdgpu_bo_reserve(bo, false);
-       if (r)
-               return r;
-       domain = amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type);
-       /* if bo has been evicted, then no need to recover */
-       if (domain == AMDGPU_GEM_DOMAIN_VRAM) {
-               r = amdgpu_bo_restore_from_shadow(adev, ring, bo,
+	r = amdgpu_bo_reserve(bo, true);
+	if (r)
+		return r;
+	domain = amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type);
+	/* if bo has been evicted, then no need to recover */
+	if (domain == AMDGPU_GEM_DOMAIN_VRAM) {
+		r = amdgpu_bo_validate(bo->shadow);
+		if (r) {
+			DRM_ERROR("bo validate failed!\n");
+			goto err;
+		}
+
+		r = amdgpu_ttm_bind(&bo->shadow->tbo, &bo->shadow->tbo.mem);
+		if (r) {
+			DRM_ERROR("%p bind failed\n", bo->shadow);
+			goto err;
+		}
+
+		r = amdgpu_bo_restore_from_shadow(adev, ring, bo,
 						 NULL, fence, true);
-               if (r) {
-                       DRM_ERROR("recover page table failed!\n");
-                       goto err;
-               }
-       }
+		if (r) {
+			DRM_ERROR("recover page table failed!\n");
+			goto err;
+		}
+	}
 err:
-       amdgpu_bo_unreserve(bo);
-       return r;
+	amdgpu_bo_unreserve(bo);
+	return r;
 }
 
 /**
···
 	ring = adev->mman.buffer_funcs_ring;
 	mutex_lock(&adev->shadow_list_lock);
 	list_for_each_entry_safe(bo, tmp, &adev->shadow_list, shadow_list) {
+		next = NULL;
 		amdgpu_recover_vram_from_shadow(adev, ring, bo, &next);
 		if (fence) {
 			r = dma_fence_wait(fence, false);
···
 	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 		struct amdgpu_ring *ring = adev->rings[i];
 
-		if (!ring)
+		if (!ring || !ring->sched.thread)
 			continue;
 		kthread_park(ring->sched.thread);
 		amd_sched_hw_job_reset(&ring->sched);
···
 			DRM_INFO("recover vram bo from shadow\n");
 			mutex_lock(&adev->shadow_list_lock);
 			list_for_each_entry_safe(bo, tmp, &adev->shadow_list, shadow_list) {
+				next = NULL;
 				amdgpu_recover_vram_from_shadow(adev, ring, bo, &next);
 				if (fence) {
 					r = dma_fence_wait(fence, false);
···
 		}
 		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 			struct amdgpu_ring *ring = adev->rings[i];
-			if (!ring)
+
+			if (!ring || !ring->sched.thread)
 				continue;
 
 			amd_sched_job_recovery(&ring->sched);
···
 	} else {
 		dev_err(adev->dev, "asic resume failed (%d).\n", r);
 		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
-			if (adev->rings[i]) {
+			if (adev->rings[i] && adev->rings[i]->sched.thread) {
 				kthread_unpark(adev->rings[i]->sched.thread);
 			}
 		}

drivers/gpu/drm/amd/amdgpu/amdgpu_display.c

···
 	int r;
 
 	/* unpin of the old buffer */
-	r = amdgpu_bo_reserve(work->old_abo, false);
+	r = amdgpu_bo_reserve(work->old_abo, true);
 	if (likely(r == 0)) {
 		r = amdgpu_bo_unpin(work->old_abo);
 		if (unlikely(r != 0)) {
···
 	kfree(work);
 }
 
-
-static void amdgpu_flip_work_cleanup(struct amdgpu_flip_work *work)
-{
-	int i;
-
-	amdgpu_bo_unref(&work->old_abo);
-	dma_fence_put(work->excl);
-	for (i = 0; i < work->shared_count; ++i)
-		dma_fence_put(work->shared[i]);
-	kfree(work->shared);
-	kfree(work);
-}
-
-static void amdgpu_flip_cleanup_unreserve(struct amdgpu_flip_work *work,
-					  struct amdgpu_bo *new_abo)
-{
-	amdgpu_bo_unreserve(new_abo);
-	amdgpu_flip_work_cleanup(work);
-}
-
-static void amdgpu_flip_cleanup_unpin(struct amdgpu_flip_work *work,
-				      struct amdgpu_bo *new_abo)
-{
-	if (unlikely(amdgpu_bo_unpin(new_abo) != 0))
-		DRM_ERROR("failed to unpin new abo in error path\n");
-	amdgpu_flip_cleanup_unreserve(work, new_abo);
-}
-
-void amdgpu_crtc_cleanup_flip_ctx(struct amdgpu_flip_work *work,
-				  struct amdgpu_bo *new_abo)
-{
-	if (unlikely(amdgpu_bo_reserve(new_abo, false) != 0)) {
-		DRM_ERROR("failed to reserve new abo in error path\n");
-		amdgpu_flip_work_cleanup(work);
-		return;
-	}
-	amdgpu_flip_cleanup_unpin(work, new_abo);
-}
-
-int amdgpu_crtc_prepare_flip(struct drm_crtc *crtc,
-			     struct drm_framebuffer *fb,
-			     struct drm_pending_vblank_event *event,
-			     uint32_t page_flip_flags,
-			     uint32_t target,
-			     struct amdgpu_flip_work **work_p,
-			     struct amdgpu_bo **new_abo_p)
+int amdgpu_crtc_page_flip_target(struct drm_crtc *crtc,
+				 struct drm_framebuffer *fb,
+				 struct drm_pending_vblank_event *event,
+				 uint32_t page_flip_flags, uint32_t target,
+				 struct drm_modeset_acquire_ctx *ctx)
 {
 	struct drm_device *dev = crtc->dev;
 	struct amdgpu_device *adev = dev->dev_private;
···
 	unsigned long flags;
 	u64 tiling_flags;
 	u64 base;
-	int r;
+	int i, r;
 
 	work = kzalloc(sizeof *work, GFP_KERNEL);
 	if (work == NULL)
···
 		spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
 		r = -EBUSY;
 		goto pflip_cleanup;
-
 	}
-	spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
 
-	*work_p = work;
-	*new_abo_p = new_abo;
-
-	return 0;
-
-pflip_cleanup:
-	amdgpu_crtc_cleanup_flip_ctx(work, new_abo);
-	return r;
-
-unpin:
-	amdgpu_flip_cleanup_unpin(work, new_abo);
-	return r;
-
-unreserve:
-	amdgpu_flip_cleanup_unreserve(work, new_abo);
-	return r;
-
-cleanup:
-	amdgpu_flip_work_cleanup(work);
-	return r;
-
-}
-
-void amdgpu_crtc_submit_flip(struct drm_crtc *crtc,
-			     struct drm_framebuffer *fb,
-			     struct amdgpu_flip_work *work,
-			     struct amdgpu_bo *new_abo)
-{
-	unsigned long flags;
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-
-	spin_lock_irqsave(&crtc->dev->event_lock, flags);
 	amdgpu_crtc->pflip_status = AMDGPU_FLIP_PENDING;
 	amdgpu_crtc->pflip_works = work;
 
+
+	DRM_DEBUG_DRIVER("crtc:%d[%p], pflip_stat:AMDGPU_FLIP_PENDING, work: %p,\n",
+					 amdgpu_crtc->crtc_id, amdgpu_crtc, work);
 	/* update crtc fb */
 	crtc->primary->fb = fb;
 	spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
-
-	DRM_DEBUG_DRIVER(
-			"crtc:%d[%p], pflip_stat:AMDGPU_FLIP_PENDING, work: %p,\n",
-			amdgpu_crtc->crtc_id, amdgpu_crtc, work);
-
 	amdgpu_flip_work_func(&work->flip_work.work);
-}
-
-int amdgpu_crtc_page_flip_target(struct drm_crtc *crtc,
-				 struct drm_framebuffer *fb,
-				 struct drm_pending_vblank_event *event,
-				 uint32_t page_flip_flags,
-				 uint32_t target,
-				 struct drm_modeset_acquire_ctx *ctx)
-{
-	struct amdgpu_bo *new_abo;
-	struct amdgpu_flip_work *work;
-	int r;
-
-	r = amdgpu_crtc_prepare_flip(crtc,
-				     fb,
-				     event,
-				     page_flip_flags,
-				     target,
-				     &work,
-				     &new_abo);
-	if (r)
-		return r;
-
-	amdgpu_crtc_submit_flip(crtc, fb, work, new_abo);
-
 	return 0;
+
+pflip_cleanup:
+	if (unlikely(amdgpu_bo_reserve(new_abo, false) != 0)) {
+		DRM_ERROR("failed to reserve new abo in error path\n");
+		goto cleanup;
+	}
+unpin:
+	if (unlikely(amdgpu_bo_unpin(new_abo) != 0)) {
+		DRM_ERROR("failed to unpin new abo in error path\n");
+	}
+unreserve:
+	amdgpu_bo_unreserve(new_abo);
+
+cleanup:
+	amdgpu_bo_unref(&work->old_abo);
+	dma_fence_put(work->excl);
+	for (i = 0; i < work->shared_count; ++i)
+		dma_fence_put(work->shared[i]);
+	kfree(work->shared);
+	kfree(work);
+
+	return r;
 }
 
 int amdgpu_crtc_set_config(struct drm_mode_set *set,

drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c

···
  * - 3.11.0 - Add support for sensor query info (clocks, temp, etc).
  * - 3.12.0 - Add query for double offchip LDS buffers
  * - 3.13.0 - Add PRT support
+ * - 3.14.0 - Fix race in amdgpu_ctx_get_fence() and note new functionality
+ * - 3.15.0 - Export more gpu info for gfx9
  */
 #define KMS_DRIVER_MAJOR	3
-#define KMS_DRIVER_MINOR	13
+#define KMS_DRIVER_MINOR	15
 #define KMS_DRIVER_PATCHLEVEL	0
 
 int amdgpu_vram_limit = 0;
···
 	{0x1002, 0x6861, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT},
 	{0x1002, 0x6862, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT},
 	{0x1002, 0x6863, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT},
+	{0x1002, 0x6864, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT},
 	{0x1002, 0x6867, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT},
+	{0x1002, 0x6868, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT},
 	{0x1002, 0x686c, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT},
 	{0x1002, 0x687f, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT},
 	{0, 0, 0}

drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c

···
 	struct amdgpu_bo *abo = gem_to_amdgpu_bo(gobj);
 	int ret;
 
-	ret = amdgpu_bo_reserve(abo, false);
+	ret = amdgpu_bo_reserve(abo, true);
 	if (likely(ret == 0)) {
 		amdgpu_bo_kunmap(abo);
 		amdgpu_bo_unpin(abo);

drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c

···
 	if (adev->gart.robj == NULL) {
 		return;
 	}
-	r = amdgpu_bo_reserve(adev->gart.robj, false);
+	r = amdgpu_bo_reserve(adev->gart.robj, true);
 	if (likely(r == 0)) {
 		amdgpu_bo_kunmap(adev->gart.robj);
 		amdgpu_bo_unpin(adev->gart.robj);

drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c

···
 	return 0;
 }
 
+static int amdgpu_gem_vm_check(void *param, struct amdgpu_bo *bo)
+{
+	/* if anything is swapped out don't swap it in here,
+	   just abort and wait for the next CS */
+	if (!amdgpu_bo_gpu_accessible(bo))
+		return -ERESTARTSYS;
+
+	if (bo->shadow && !amdgpu_bo_gpu_accessible(bo->shadow))
+		return -ERESTARTSYS;
+
+	return 0;
+}
+
+static bool amdgpu_gem_vm_ready(struct amdgpu_device *adev,
+				struct amdgpu_vm *vm,
+				struct list_head *list)
+{
+	struct ttm_validate_buffer *entry;
+
+	list_for_each_entry(entry, list, head) {
+		struct amdgpu_bo *bo =
+			container_of(entry->bo, struct amdgpu_bo, tbo);
+		if (amdgpu_gem_vm_check(NULL, bo))
+			return false;
+	}
+
+	return !amdgpu_vm_validate_pt_bos(adev, vm, amdgpu_gem_vm_check, NULL);
+}
+
 void amdgpu_gem_object_close(struct drm_gem_object *obj,
 			     struct drm_file *file_priv)
 {
···
 	struct amdgpu_vm *vm = &fpriv->vm;
 
 	struct amdgpu_bo_list_entry vm_pd;
-	struct list_head list, duplicates;
+	struct list_head list;
 	struct ttm_validate_buffer tv;
 	struct ww_acquire_ctx ticket;
 	struct amdgpu_bo_va *bo_va;
-	struct dma_fence *fence = NULL;
 	int r;
 
 	INIT_LIST_HEAD(&list);
-	INIT_LIST_HEAD(&duplicates);
 
 	tv.bo = &bo->tbo;
 	tv.shared = true;
···
 
 	amdgpu_vm_get_pd_bo(vm, &list, &vm_pd);
 
-	r = ttm_eu_reserve_buffers(&ticket, &list, false, &duplicates);
+	r = ttm_eu_reserve_buffers(&ticket, &list, false, NULL);
 	if (r) {
 		dev_err(adev->dev, "leaking bo va because "
 			"we fail to reserve bo (%d)\n", r);
 		return;
 	}
 	bo_va = amdgpu_vm_bo_find(vm, bo);
-	if (bo_va) {
-		if (--bo_va->ref_count == 0) {
-			amdgpu_vm_bo_rmv(adev, bo_va);
+	if (bo_va && --bo_va->ref_count == 0) {
+		amdgpu_vm_bo_rmv(adev, bo_va);
+
+		if (amdgpu_gem_vm_ready(adev, vm, &list)) {
+			struct dma_fence *fence = NULL;
 
 			r = amdgpu_vm_clear_freed(adev, vm, &fence);
 			if (unlikely(r)) {
···
 	return r;
 }
 
-static int amdgpu_gem_va_check(void *param, struct amdgpu_bo *bo)
-{
-	/* if anything is swapped out don't swap it in here,
-	   just abort and wait for the next CS */
-	if (!amdgpu_bo_gpu_accessible(bo))
-		return -ERESTARTSYS;
-
-	if (bo->shadow && !amdgpu_bo_gpu_accessible(bo->shadow))
-		return -ERESTARTSYS;
-
-	return 0;
-}
-
 /**
  * amdgpu_gem_va_update_vm -update the bo_va in its VM
  *
···
 				    struct list_head *list,
 				    uint32_t operation)
 {
-	struct ttm_validate_buffer *entry;
 	int r = -ERESTARTSYS;
 
-	list_for_each_entry(entry, list, head) {
-		struct amdgpu_bo *bo =
-			container_of(entry->bo, struct amdgpu_bo, tbo);
-		if (amdgpu_gem_va_check(NULL, bo))
-			goto error;
-	}
-
-	r = amdgpu_vm_validate_pt_bos(adev, vm, amdgpu_gem_va_check,
-				      NULL);
-	if (r)
+	if (!amdgpu_gem_vm_ready(adev, vm, list))
 		goto error;
 
 	r = amdgpu_vm_update_directories(adev, vm);

drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c

···
 	return r;
 }
 
+void amdgpu_gtt_mgr_print(struct seq_file *m, struct ttm_mem_type_manager *man)
+{
+	struct amdgpu_device *adev = amdgpu_ttm_adev(man->bdev);
+	struct amdgpu_gtt_mgr *mgr = man->priv;
+
+	seq_printf(m, "man size:%llu pages, gtt available:%llu pages, usage:%lluMB\n",
+		   man->size, mgr->available, (u64)atomic64_read(&adev->gtt_usage) >> 20);
+
+}
 /**
  * amdgpu_gtt_mgr_new - allocate a new node
  *

drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c

···
 		dev_err(adev->dev, "scheduling IB failed (%d).\n", r);
 		return r;
 	}
+	if (ring->funcs->emit_pipeline_sync && job && job->need_pipeline_sync)
+		amdgpu_ring_emit_pipeline_sync(ring);
 
 	if (vm) {
 		r = amdgpu_vm_flush(ring, job);
···
 	if (r) {
 		dev_err(adev->dev, "failed to emit fence (%d)\n", r);
 		if (job && job->vm_id)
-			amdgpu_vm_reset_id(adev, job->vm_id);
+			amdgpu_vm_reset_id(adev, ring->funcs->vmhub,
+					   job->vm_id);
 		amdgpu_ring_undo(ring);
 		return r;
 	}

drivers/gpu/drm/amd/amdgpu/amdgpu_job.c

···
 	(*job)->vm = vm;
 	(*job)->ibs = (void *)&(*job)[1];
 	(*job)->num_ibs = num_ibs;
+	(*job)->need_pipeline_sync = false;
 
 	amdgpu_sync_create(&(*job)->sync);
 
···
 
 	struct dma_fence *fence = amdgpu_sync_get_fence(&job->sync);
 
-	if (fence == NULL && vm && !job->vm_id) {
+	while (fence == NULL && vm && !job->vm_id) {
 		struct amdgpu_ring *ring = job->ring;
 		int r;
 
···
 
 		fence = amdgpu_sync_get_fence(&job->sync);
 	}
+
+	if (amd_sched_dependency_optimized(fence, sched_job->s_entity))
+		job->need_pipeline_sync = true;
 
 	return fence;
 }

drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c

···
 			adev->gfx.config.double_offchip_lds_buf;
 
 		if (amdgpu_ngg) {
-			dev_info.prim_buf_gpu_addr = adev->gfx.ngg.buf[PRIM].gpu_addr;
-			dev_info.pos_buf_gpu_addr = adev->gfx.ngg.buf[POS].gpu_addr;
-			dev_info.cntl_sb_buf_gpu_addr = adev->gfx.ngg.buf[CNTL].gpu_addr;
-			dev_info.param_buf_gpu_addr = adev->gfx.ngg.buf[PARAM].gpu_addr;
+			dev_info.prim_buf_gpu_addr = adev->gfx.ngg.buf[NGG_PRIM].gpu_addr;
+			dev_info.prim_buf_size = adev->gfx.ngg.buf[NGG_PRIM].size;
+			dev_info.pos_buf_gpu_addr = adev->gfx.ngg.buf[NGG_POS].gpu_addr;
+			dev_info.pos_buf_size = adev->gfx.ngg.buf[NGG_POS].size;
+			dev_info.cntl_sb_buf_gpu_addr = adev->gfx.ngg.buf[NGG_CNTL].gpu_addr;
+			dev_info.cntl_sb_buf_size = adev->gfx.ngg.buf[NGG_CNTL].size;
+			dev_info.param_buf_gpu_addr = adev->gfx.ngg.buf[NGG_PARAM].gpu_addr;
+			dev_info.param_buf_size = adev->gfx.ngg.buf[NGG_PARAM].size;
 		}
+		dev_info.wave_front_size = adev->gfx.cu_info.wave_front_size;
+		dev_info.num_shader_visible_vgprs = adev->gfx.config.max_gprs;
+		dev_info.num_cu_per_sh = adev->gfx.config.max_cu_per_sh;
+		dev_info.num_tcc_blocks = adev->gfx.config.max_texture_channel_caches;
+		dev_info.gs_vgt_table_depth = adev->gfx.config.gs_vgt_table_depth;
+		dev_info.gs_prim_buffer_depth = adev->gfx.config.gs_prim_buffer_depth;
+		dev_info.max_gs_waves_per_vgt = adev->gfx.config.max_gs_threads;
 
 		return copy_to_user(out, &dev_info,
 				    min((size_t)size, sizeof(dev_info))) ? -EFAULT : 0;
···
 
 	if (amdgpu_sriov_vf(adev)) {
 		/* TODO: how to handle reserve failure */
-		BUG_ON(amdgpu_bo_reserve(adev->virt.csa_obj, false));
+		BUG_ON(amdgpu_bo_reserve(adev->virt.csa_obj, true));
 		amdgpu_vm_bo_rmv(adev, fpriv->vm.csa_bo_va);
 		fpriv->vm.csa_bo_va = NULL;
 		amdgpu_bo_unreserve(adev->virt.csa_obj);

drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h

···
 				 struct drm_pending_vblank_event *event,
 				 uint32_t page_flip_flags, uint32_t target,
 				 struct drm_modeset_acquire_ctx *ctx);
-void amdgpu_crtc_cleanup_flip_ctx(struct amdgpu_flip_work *work,
-				  struct amdgpu_bo *new_abo);
-int amdgpu_crtc_prepare_flip(struct drm_crtc *crtc,
-			     struct drm_framebuffer *fb,
-			     struct drm_pending_vblank_event *event,
-			     uint32_t page_flip_flags,
-			     uint32_t target,
-			     struct amdgpu_flip_work **work,
-			     struct amdgpu_bo **new_abo);
-
-void amdgpu_crtc_submit_flip(struct drm_crtc *crtc,
-			     struct drm_framebuffer *fb,
-			     struct amdgpu_flip_work *work,
-			     struct amdgpu_bo *new_abo);
-
 extern const struct drm_mode_config_funcs amdgpu_mode_funcs;
 
 #endif

drivers/gpu/drm/amd/amdgpu/amdgpu_object.c

···
 	if (*bo == NULL)
 		return;
 
-	if (likely(amdgpu_bo_reserve(*bo, false) == 0)) {
+	if (likely(amdgpu_bo_reserve(*bo, true) == 0)) {
 		if (cpu_addr)
 			amdgpu_bo_kunmap(*bo);
 
···
 		amdgpu_bo_fence(bo, *fence, true);
 
 err:
+	return r;
+}
+
+int amdgpu_bo_validate(struct amdgpu_bo *bo)
+{
+	uint32_t domain;
+	int r;
+
+	if (bo->pin_count)
+		return 0;
+
+	domain = bo->prefered_domains;
+
+retry:
+	amdgpu_ttm_placement_from_domain(bo, domain);
+	r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false);
+	if (unlikely(r == -ENOMEM) && domain != bo->allowed_domains) {
+		domain = bo->allowed_domains;
+		goto retry;
+	}
+
 	return r;
 }
 

drivers/gpu/drm/amd/amdgpu/amdgpu_object.h

···
 			       struct amdgpu_bo *bo,
 			       struct reservation_object *resv,
 			       struct dma_fence **fence, bool direct);
+int amdgpu_bo_validate(struct amdgpu_bo *bo);
 int amdgpu_bo_restore_from_shadow(struct amdgpu_device *adev,
 				  struct amdgpu_ring *ring,
 				  struct amdgpu_bo *bo,

drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c

···
 
 	pwm_mode = amdgpu_dpm_get_fan_control_mode(adev);
 
-	/* never 0 (full-speed), fuse or smc-controlled always */
-	return sprintf(buf, "%i\n", pwm_mode == FDO_PWM_MODE_STATIC ? 1 : 2);
+	return sprintf(buf, "%i\n", pwm_mode);
 }
 
 static ssize_t amdgpu_hwmon_set_pwm1_enable(struct device *dev,
···
 	if (err)
 		return err;
 
-	switch (value) {
-	case 1: /* manual, percent-based */
-		amdgpu_dpm_set_fan_control_mode(adev, FDO_PWM_MODE_STATIC);
-		break;
-	default: /* disable */
-		amdgpu_dpm_set_fan_control_mode(adev, 0);
-		break;
-	}
+	amdgpu_dpm_set_fan_control_mode(adev, value);
 
 	return count;
 }

drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c

···
 	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
 	int ret = 0;
 
-	ret = amdgpu_bo_reserve(bo, false);
+	ret = amdgpu_bo_reserve(bo, true);
 	if (unlikely(ret != 0))
 		return;
 

drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c

···
 		psp->bootloader_load_sos = psp_v3_1_bootloader_load_sos;
 		psp->prep_cmd_buf = psp_v3_1_prep_cmd_buf;
 		psp->ring_init = psp_v3_1_ring_init;
+		psp->ring_create = psp_v3_1_ring_create;
+		psp->ring_destroy = psp_v3_1_ring_destroy;
 		psp->cmd_submit = psp_v3_1_cmd_submit;
 		psp->compare_sram_data = psp_v3_1_compare_sram_data;
 		psp->smu_reload_quirk = psp_v3_1_smu_reload_quirk;
···
 static int psp_tmr_init(struct psp_context *psp)
 {
 	int ret;
-	struct psp_gfx_cmd_resp *cmd;
-
-	cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL);
-	if (!cmd)
-		return -ENOMEM;
 
 	/*
 	 * Allocate 3M memory aligned to 1M from Frame Buffer (local
···
 	ret = amdgpu_bo_create_kernel(psp->adev, 0x300000, 0x100000,
 				      AMDGPU_GEM_DOMAIN_VRAM,
 				      &psp->tmr_bo, &psp->tmr_mc_addr, &psp->tmr_buf);
-	if (ret)
-		goto failed;
+
+	return ret;
+}
+
+static int psp_tmr_load(struct psp_context *psp)
+{
+	int ret;
+	struct psp_gfx_cmd_resp *cmd;
+
+	cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL);
+	if (!cmd)
+		return -ENOMEM;
 
 	psp_prep_tmr_cmd_buf(cmd, psp->tmr_mc_addr, 0x300000);
 
 	ret = psp_cmd_submit_buf(psp, NULL, cmd,
 				 psp->fence_buf_mc_addr, 1);
 	if (ret)
-		goto failed_mem;
+		goto failed;
 
 	kfree(cmd);
 
 	return 0;
 
-failed_mem:
-	amdgpu_bo_free_kernel(&psp->tmr_bo, &psp->tmr_mc_addr, &psp->tmr_buf);
 failed:
 	kfree(cmd);
 	return ret;
···
 	cmd->cmd.cmd_load_ta.cmd_buf_len = shared_size;
 }
 
-static int psp_asd_load(struct psp_context *psp)
+static int psp_asd_init(struct psp_context *psp)
 {
 	int ret;
-	struct amdgpu_bo *asd_bo, *asd_shared_bo;
-	uint64_t asd_mc_addr, asd_shared_mc_addr;
-	void *asd_buf, *asd_shared_buf;
-	struct psp_gfx_cmd_resp *cmd;
-
-	cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL);
-	if (!cmd)
-		return -ENOMEM;
 
 	/*
 	 * Allocate 16k memory aligned to 4k from Frame Buffer (local
 	 * physical) for shared ASD <-> Driver
 	 */
-	ret = amdgpu_bo_create_kernel(psp->adev, PSP_ASD_SHARED_MEM_SIZE, PAGE_SIZE,
-				      AMDGPU_GEM_DOMAIN_VRAM,
-				      &asd_shared_bo, &asd_shared_mc_addr, &asd_buf);
-	if (ret)
-		goto failed;
+	ret = amdgpu_bo_create_kernel(psp->adev, PSP_ASD_SHARED_MEM_SIZE,
+				      PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM,
+				      &psp->asd_shared_bo,
+				      &psp->asd_shared_mc_addr,
+				      &psp->asd_shared_buf);
 
-	/*
-	 * Allocate 256k memory aligned to 4k from Frame Buffer (local
-	 * physical) for ASD firmware
-	 */
-	ret = amdgpu_bo_create_kernel(psp->adev, PSP_ASD_BIN_SIZE, PAGE_SIZE,
-				      AMDGPU_GEM_DOMAIN_VRAM,
-				      &asd_bo, &asd_mc_addr, &asd_buf);
-	if (ret)
-		goto failed_mem;
-
-	memcpy(asd_buf, psp->asd_start_addr, psp->asd_ucode_size);
-
-	psp_prep_asd_cmd_buf(cmd, asd_mc_addr, asd_shared_mc_addr,
-			     psp->asd_ucode_size, PSP_ASD_SHARED_MEM_SIZE);
-
-	ret = psp_cmd_submit_buf(psp, NULL, cmd,
-				 psp->fence_buf_mc_addr, 2);
-	if (ret)
-		goto failed_mem1;
-
-	amdgpu_bo_free_kernel(&asd_bo, &asd_mc_addr, &asd_buf);
-	amdgpu_bo_free_kernel(&asd_shared_bo, &asd_shared_mc_addr, &asd_shared_buf);
-	kfree(cmd);
-
-	return 0;
-
-failed_mem1:
-	amdgpu_bo_free_kernel(&asd_bo, &asd_mc_addr, &asd_buf);
-failed_mem:
-	amdgpu_bo_free_kernel(&asd_shared_bo, &asd_shared_mc_addr, &asd_shared_buf);
-failed:
-	kfree(cmd);
 	return ret;
 }
 
-static int psp_load_fw(struct amdgpu_device *adev)
+static int psp_asd_load(struct psp_context *psp)
 {
 	int ret;
 	struct psp_gfx_cmd_resp *cmd;
-	int i;
-	struct amdgpu_firmware_info *ucode;
-	struct psp_context *psp = &adev->psp;
 
 	cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL);
 	if (!cmd)
 		return -ENOMEM;
 
+	memset(psp->fw_pri_buf, 0, PSP_1_MEG);
+	memcpy(psp->fw_pri_buf, psp->asd_start_addr, psp->asd_ucode_size);
+
+	psp_prep_asd_cmd_buf(cmd, psp->fw_pri_mc_addr, psp->asd_shared_mc_addr,
+			     psp->asd_ucode_size, PSP_ASD_SHARED_MEM_SIZE);
+
+	ret = psp_cmd_submit_buf(psp, NULL, cmd,
+				 psp->fence_buf_mc_addr, 2);
+
+	kfree(cmd);
+
+	return ret;
+}
+
+static int psp_hw_start(struct psp_context *psp)
+{
+	int ret;
+
 	ret = psp_bootloader_load_sysdrv(psp);
 	if (ret)
-		goto failed;
+		return ret;
 
 	ret = psp_bootloader_load_sos(psp);
 	if (ret)
-		goto failed;
+		return ret;
 
-	ret = psp_ring_init(psp, PSP_RING_TYPE__KM);
+	ret = psp_ring_create(psp, PSP_RING_TYPE__KM);
 	if (ret)
-		goto failed;
+		return ret;
 
-	ret = amdgpu_bo_create_kernel(adev, PSP_FENCE_BUFFER_SIZE, PAGE_SIZE,
-				      AMDGPU_GEM_DOMAIN_VRAM,
-				      &psp->fence_buf_bo,
-				      &psp->fence_buf_mc_addr,
-				      &psp->fence_buf);
+	ret = psp_tmr_load(psp);
 	if (ret)
-		goto failed;
-
-	memset(psp->fence_buf, 0, PSP_FENCE_BUFFER_SIZE);
-
-	ret = psp_tmr_init(psp);
-	if (ret)
-		goto failed_mem;
+		return ret;
 
 	ret = psp_asd_load(psp);
 	if (ret)
-		goto failed_mem;
+		return ret;
+
+	return 0;
+}
+
+static int psp_np_fw_load(struct psp_context *psp)
+{
+	int i, ret;
+	struct amdgpu_firmware_info *ucode;
+	struct amdgpu_device* adev = psp->adev;
 
 	for (i = 0; i < adev->firmware.max_ucodes; i++) {
 		ucode = &adev->firmware.ucode[i];
···
 		if (ucode->ucode_id == AMDGPU_UCODE_ID_SMC &&
 		    psp_smu_reload_quirk(psp))
 			continue;
+		if (amdgpu_sriov_vf(adev) &&
+		   (ucode->ucode_id == AMDGPU_UCODE_ID_SDMA0
+		    || ucode->ucode_id == AMDGPU_UCODE_ID_SDMA1
+		    || ucode->ucode_id == AMDGPU_UCODE_ID_RLC_G))
+			/*skip ucode loading in SRIOV VF */
+			continue;
 
-		ret = psp_prep_cmd_buf(ucode, cmd);
+		ret = psp_prep_cmd_buf(ucode, psp->cmd);
 		if (ret)
-			goto failed_mem;
+			return ret;
 
-		ret = psp_cmd_submit_buf(psp, ucode, cmd,
+		ret = psp_cmd_submit_buf(psp, ucode, psp->cmd,
 					 psp->fence_buf_mc_addr, i + 3);
 		if (ret)
-			goto failed_mem;
+			return ret;
 
 #if 0
 		/* check if firmware loaded sucessfully */
···
 #endif
 	}
 
-	amdgpu_bo_free_kernel(&psp->fence_buf_bo,
-			      &psp->fence_buf_mc_addr, &psp->fence_buf);
+	return 0;
+}
+
+static int psp_load_fw(struct amdgpu_device *adev)
+{
+	int ret;
+	struct psp_context *psp = &adev->psp;
+	struct psp_gfx_cmd_resp *cmd;
+
+	cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL);
+	if (!cmd)
+		return -ENOMEM;
+
+	psp->cmd = cmd;
+
+	ret = amdgpu_bo_create_kernel(adev, PSP_1_MEG, PSP_1_MEG,
+				      AMDGPU_GEM_DOMAIN_GTT,
+				      &psp->fw_pri_bo,
+				      &psp->fw_pri_mc_addr,
+				      &psp->fw_pri_buf);
+	if (ret)
+		goto failed;
+
+	ret = amdgpu_bo_create_kernel(adev, PSP_FENCE_BUFFER_SIZE, PAGE_SIZE,
+				      AMDGPU_GEM_DOMAIN_VRAM,
+				      &psp->fence_buf_bo,
+				      &psp->fence_buf_mc_addr,
+				      &psp->fence_buf);
+	if (ret)
+		goto failed_mem1;
+
+	memset(psp->fence_buf, 0, PSP_FENCE_BUFFER_SIZE);
+
+	ret = psp_ring_init(psp, PSP_RING_TYPE__KM);
+	if (ret)
+		goto failed_mem1;
+
+	ret = psp_tmr_init(psp);
+	if (ret)
+		goto failed_mem;
+
+	ret = psp_asd_init(psp);
+	if (ret)
+		goto failed_mem;
+
+	ret = psp_hw_start(psp);
+	if (ret)
+		goto failed_mem;
+
+	ret = psp_np_fw_load(psp);
+	if (ret)
+		goto failed_mem;
+
 	kfree(cmd);
 
 	return 0;
···
 failed_mem:
 	amdgpu_bo_free_kernel(&psp->fence_buf_bo,
 			      &psp->fence_buf_mc_addr, &psp->fence_buf);
+failed_mem1:
+	amdgpu_bo_free_kernel(&psp->fw_pri_bo,
+			      &psp->fw_pri_mc_addr, &psp->fw_pri_buf);
 failed:
 	kfree(cmd);
 	return ret;
···
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct psp_context *psp = &adev->psp;
 
-	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP)
-		amdgpu_ucode_fini_bo(adev);
+	if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP)
+		return 0;
+
+	amdgpu_ucode_fini_bo(adev);
+
+	psp_ring_destroy(psp, PSP_RING_TYPE__KM);
 
 	if (psp->tmr_buf)
 		amdgpu_bo_free_kernel(&psp->tmr_bo, &psp->tmr_mc_addr, &psp->tmr_buf);
+
+	if (psp->fw_pri_buf)
+		amdgpu_bo_free_kernel(&psp->fw_pri_bo,
+				      &psp->fw_pri_mc_addr, &psp->fw_pri_buf);
+
+	if (psp->fence_buf_bo)
+		amdgpu_bo_free_kernel(&psp->fence_buf_bo,
+				      &psp->fence_buf_mc_addr, &psp->fence_buf);
 
 	return 0;
 }
···
 {
 	int ret;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct psp_context *psp = &adev->psp;
 
 	if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP)
 		return 0;
 
+	DRM_INFO("PSP is resuming...\n");
+
 	mutex_lock(&adev->firmware.mutex);
 
-	ret = psp_load_fw(adev);
+	ret = psp_hw_start(psp);
 	if (ret)
-		DRM_ERROR("PSP resume failed\n");
+		goto failed;
+
+	ret = psp_np_fw_load(psp);
+	if (ret)
+		goto failed;
 
 	mutex_unlock(&adev->firmware.mutex);
 
+	return 0;
+
+failed:
+	DRM_ERROR("PSP resume failed\n");
+	mutex_unlock(&adev->firmware.mutex);
 	return ret;
 }
 

drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h

···
 
 #define PSP_FENCE_BUFFER_SIZE	0x1000
 #define PSP_CMD_BUFFER_SIZE	0x1000
-#define PSP_ASD_BIN_SIZE	0x40000
 #define PSP_ASD_SHARED_MEM_SIZE	0x4000
+#define PSP_1_MEG		0x100000
 
 enum psp_ring_type
 {
···
 {
 	struct amdgpu_device            *adev;
 	struct psp_ring                 km_ring;
+	struct psp_gfx_cmd_resp		*cmd;
 
 	int (*init_microcode)(struct psp_context *psp);
 	int (*bootloader_load_sysdrv)(struct psp_context *psp);
···
 	int (*prep_cmd_buf)(struct amdgpu_firmware_info *ucode,
 			    struct psp_gfx_cmd_resp *cmd);
 	int (*ring_init)(struct psp_context *psp, enum psp_ring_type ring_type);
+	int (*ring_create)(struct psp_context *psp, enum psp_ring_type ring_type);
+	int (*ring_destroy)(struct psp_context *psp,
+			    enum psp_ring_type ring_type);
 	int (*cmd_submit)(struct psp_context *psp, struct amdgpu_firmware_info *ucode,
 			  uint64_t cmd_buf_mc_addr, uint64_t fence_mc_addr, int index);
 	bool (*compare_sram_data)(struct psp_context *psp,
 				  struct amdgpu_firmware_info *ucode,
 				  enum AMDGPU_UCODE_ID ucode_type);
 	bool (*smu_reload_quirk)(struct psp_context *psp);
+
+	/* fence buffer */
+	struct amdgpu_bo 		*fw_pri_bo;
+	uint64_t 			fw_pri_mc_addr;
+	void				*fw_pri_buf;
 
 	/* sos firmware */
 	const struct firmware		*sos_fw;
···
 	uint64_t 			tmr_mc_addr;
 	void				*tmr_buf;
 
-	/* asd firmware */
+	/* asd firmware and buffer */
 	const struct firmware		*asd_fw;
 	uint32_t			asd_fw_version;
 	uint32_t			asd_feature_version;
 	uint32_t			asd_ucode_size;
 	uint8_t				*asd_start_addr;
+	struct amdgpu_bo 		*asd_shared_bo;
+	uint64_t 			asd_shared_mc_addr;
+	void				*asd_shared_buf;
 
 	/* fence buffer */
 	struct amdgpu_bo 		*fence_buf_bo;
···
 
 #define psp_prep_cmd_buf(ucode, type) (psp)->prep_cmd_buf((ucode), (type))
 #define psp_ring_init(psp, type) (psp)->ring_init((psp), (type))
+#define psp_ring_create(psp, type) (psp)->ring_create((psp), (type))
+#define psp_ring_destroy(psp, type) ((psp)->ring_destroy((psp), (type)))
 #define psp_cmd_submit(psp, ucode, cmd_mc, fence_mc, index) \
 		(psp)->cmd_submit((psp), (ucode), (cmd_mc), (fence_mc), (index))
 #define psp_compare_sram_data(psp, ucode, type) \

drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h

···
 	uint32_t		align_mask;
 	u32			nop;
 	bool			support_64bit_ptrs;
+	unsigned		vmhub;
 
 	/* ring read/write ptr handling */
 	u64 (*get_rptr)(struct amdgpu_ring *ring);
···
 	unsigned		cond_exe_offs;
 	u64			cond_exe_gpu_addr;
 	volatile u32		*cond_exe_cpu_addr;
+	unsigned		vm_inv_eng;
 #if defined(CONFIG_DEBUG_FS)
 	struct dentry *ent;
 #endif

drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c

···
 		return -EINVAL;
 	}
 
-	r = amdgpu_bo_reserve(sa_manager->bo, false);
+	r = amdgpu_bo_reserve(sa_manager->bo, true);
 	if (!r) {
 		amdgpu_bo_kunmap(sa_manager->bo);
 		amdgpu_bo_unpin(sa_manager->bo);

drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h

···
 
 
 TRACE_EVENT(amdgpu_vm_grab_id,
-	    TP_PROTO(struct amdgpu_vm *vm, int ring, struct amdgpu_job *job),
+	    TP_PROTO(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
+		     struct amdgpu_job *job),
 	    TP_ARGS(vm, ring, job),
 	    TP_STRUCT__entry(
 			     __field(struct amdgpu_vm *, vm)
 			     __field(u32, ring)
-			     __field(u32, vmid)
+			     __field(u32, vm_id)
+			     __field(u32, vm_hub)
 			     __field(u64, pd_addr)
 			     __field(u32, needs_flush)
 			     ),
 
 	    TP_fast_assign(
 			   __entry->vm = vm;
-			   __entry->ring = ring;
-			   __entry->vmid = job->vm_id;
+			   __entry->ring = ring->idx;
+			   __entry->vm_id = job->vm_id;
+			   __entry->vm_hub = ring->funcs->vmhub,
 			   __entry->pd_addr = job->vm_pd_addr;
 			   __entry->needs_flush = job->vm_needs_flush;
 			   ),
-	    TP_printk("vm=%p, ring=%u, id=%u, pd_addr=%010Lx needs_flush=%u",
-		      __entry->vm, __entry->ring, __entry->vmid,
-		      __entry->pd_addr, __entry->needs_flush)
+	    TP_printk("vm=%p, ring=%u, id=%u, hub=%u, pd_addr=%010Lx needs_flush=%u",
+		      __entry->vm, __entry->ring, __entry->vm_id,
+		      __entry->vm_hub, __entry->pd_addr, __entry->needs_flush)
 );
 
 TRACE_EVENT(amdgpu_vm_bo_map,
···
 );
 
 TRACE_EVENT(amdgpu_vm_flush,
-	    TP_PROTO(uint64_t pd_addr, unsigned ring, unsigned id),
-	    TP_ARGS(pd_addr, ring, id),
+	    TP_PROTO(struct amdgpu_ring *ring, unsigned vm_id,
+		     uint64_t pd_addr),
+	    TP_ARGS(ring, vm_id, pd_addr),
 	    TP_STRUCT__entry(
-			     __field(u64, pd_addr)
 			     __field(u32, ring)
-			     __field(u32, id)
+			     __field(u32, vm_id)
+			     __field(u32, vm_hub)
+			     __field(u64, pd_addr)
 			     ),
 
 	    TP_fast_assign(
+			   __entry->ring = ring->idx;
+			   __entry->vm_id = vm_id;
+			   __entry->vm_hub = ring->funcs->vmhub;
 			   __entry->pd_addr = pd_addr;
-			   __entry->ring = ring;
-			   __entry->id = id;
 			   ),
-	    TP_printk("ring=%u, id=%u, pd_addr=%010Lx",
-		      __entry->ring, __entry->id, __entry->pd_addr)
+	    TP_printk("ring=%u, id=%u, hub=%u, pd_addr=%010Lx",
+		      __entry->ring, __entry->vm_id,
+		      __entry->vm_hub,__entry->pd_addr)
 );
 
 TRACE_EVENT(amdgpu_bo_list_set,

drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

···
 	abo = container_of(bo, struct amdgpu_bo, tbo);
 	switch (bo->mem.mem_type) {
 	case TTM_PL_VRAM:
-		if (adev->mman.buffer_funcs_ring->ready == false) {
+		if (adev->mman.buffer_funcs &&
+		    adev->mman.buffer_funcs_ring &&
+		    adev->mman.buffer_funcs_ring->ready == false) {
 			amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_CPU);
 		} else {
 			amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_GTT);
···
 {
 	struct amdgpu_ttm_tt *gtt, *tmp;
 	struct ttm_mem_reg bo_mem;
-	uint32_t flags;
+	uint64_t flags;
 	int r;
 
 	bo_mem.mem_type = TTM_PL_TT;
···
 static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo,
 					    const struct ttm_place *place)
 {
-	if (bo->mem.mem_type == TTM_PL_VRAM &&
-	    bo->mem.start == AMDGPU_BO_INVALID_OFFSET) {
-		unsigned long num_pages = bo->mem.num_pages;
-		struct drm_mm_node *node = bo->mem.mm_node;
+	unsigned long num_pages = bo->mem.num_pages;
+	struct drm_mm_node *node = bo->mem.mm_node;
 
+	if (bo->mem.start != AMDGPU_BO_INVALID_OFFSET)
+		return ttm_bo_eviction_valuable(bo, place);
+
+	switch (bo->mem.mem_type) {
+	case TTM_PL_TT:
+		return true;
+
+	case TTM_PL_VRAM:
 		/* Check each drm MM node individually */
 		while (num_pages) {
 			if (place->fpfn < (node->start + node->size) &&
···
 			num_pages -= node->size;
 			++node;
 		}
+		break;
 
-		return false;
+	default:
+		break;
 	}
 
 	return ttm_bo_eviction_valuable(bo, place);
···
 		return;
 	amdgpu_ttm_debugfs_fini(adev);
 	if (adev->stollen_vga_memory) {
-		r = amdgpu_bo_reserve(adev->stollen_vga_memory, false);
+		r = amdgpu_bo_reserve(adev->stollen_vga_memory, true);
 		if (r == 0) {
 			amdgpu_bo_unpin(adev->stollen_vga_memory);
 			amdgpu_bo_unreserve(adev->stollen_vga_memory);
···
 
 #if defined(CONFIG_DEBUG_FS)
 
+extern void amdgpu_gtt_mgr_print(struct seq_file *m, struct ttm_mem_type_manager
+				 *man);
 static int amdgpu_mm_dump_table(struct seq_file *m, void *data)
 {
 	struct drm_info_node *node = (struct drm_info_node *)m->private;
···
 	spin_lock(&glob->lru_lock);
 	drm_mm_print(mm, &p);
 	spin_unlock(&glob->lru_lock);
-	if (ttm_pl == TTM_PL_VRAM)
+	switch (ttm_pl) {
+	case TTM_PL_VRAM:
 		seq_printf(m, "man size:%llu pages, ram usage:%lluMB, vis usage:%lluMB\n",
 			   adev->mman.bdev.man[ttm_pl].size,
 			   (u64)atomic64_read(&adev->vram_usage) >> 20,
 			   (u64)atomic64_read(&adev->vram_vis_usage) >> 20);
+		break;
+	case TTM_PL_TT:
+		amdgpu_gtt_mgr_print(m, &adev->mman.bdev.man[TTM_PL_TT]);
+		break;
+	}
 	return 0;
 }
 

drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c

···
 	 * if SMU loaded firmware, it needn't add SMC, UVD, and VCE
 	 * ucode info here
 	 */
-	if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP)
-		adev->firmware.max_ucodes = AMDGPU_UCODE_ID_MAXIMUM - 4;
-	else
+	if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
+		if (amdgpu_sriov_vf(adev))
+			adev->firmware.max_ucodes = AMDGPU_UCODE_ID_MAXIMUM - 3;
+		else
+			adev->firmware.max_ucodes = AMDGPU_UCODE_ID_MAXIMUM - 4;
+	} else {
 		adev->firmware.max_ucodes = AMDGPU_UCODE_ID_MAXIMUM;
+	}
 
 	for (i = 0; i < adev->firmware.max_ucodes; i++) {
 		ucode = &adev->firmware.ucode[i];

drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c

···
 	struct amdgpu_device *adev = ring->adev;
 	uint32_t rptr = amdgpu_ring_get_rptr(ring);
 	unsigned i;
-	int r;
+	int r, timeout = adev->usec_timeout;
 
-	/* TODO: remove it if VCE can work for sriov */
+	/* workaround VCE ring test slow issue for sriov*/
 	if (amdgpu_sriov_vf(adev))
-		return 0;
+		timeout *= 10;
 
 	r = amdgpu_ring_alloc(ring, 16);
 	if (r) {
···
 	amdgpu_ring_write(ring, VCE_CMD_END);
 	amdgpu_ring_commit(ring);
 
-	for (i = 0; i < adev->usec_timeout; i++) {
+	for (i = 0; i < timeout; i++) {
 		if (amdgpu_ring_get_rptr(ring) != rptr)
 			break;
 		DRM_UDELAY(1);
 	}
 
-	if (i < adev->usec_timeout) {
+	if (i < timeout) {
 		DRM_INFO("ring test on %d succeeded in %d usecs\n",
 			 ring->idx, i);
 	} else {
···
 {
 	struct dma_fence *fence = NULL;
 	long r;
-
-	/* TODO: remove it if VCE can work for sriov */
-	if (amdgpu_sriov_vf(ring->adev))
-		return 0;
 
 	/* skip vce ring1/2 ib test for now, since it's not reliable */
 	if (ring != &ring->adev->vce.ring[0])

drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c

···
 
 	return 0;
 }
+
+/**
+ * amdgpu_virt_alloc_mm_table() - alloc memory for mm table
+ * @amdgpu:	amdgpu device.
+ * MM table is used by UVD and VCE for its initialization
+ * Return: Zero if allocate success.
+ */
+int amdgpu_virt_alloc_mm_table(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (!amdgpu_sriov_vf(adev) || adev->virt.mm_table.gpu_addr)
+		return 0;
+
+	r = amdgpu_bo_create_kernel(adev, PAGE_SIZE, PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_VRAM,
+				    &adev->virt.mm_table.bo,
+				    &adev->virt.mm_table.gpu_addr,
+				    (void *)&adev->virt.mm_table.cpu_addr);
+	if (r) {
+		DRM_ERROR("failed to alloc mm table and error = %d.\n", r);
+		return r;
+	}
+
+	memset((void *)adev->virt.mm_table.cpu_addr, 0, PAGE_SIZE);
+	DRM_INFO("MM table gpu addr = 0x%llx, cpu addr = %p.\n",
+		 adev->virt.mm_table.gpu_addr,
+		 adev->virt.mm_table.cpu_addr);
+	return 0;
+}
+
+/**
+ * amdgpu_virt_free_mm_table() - free mm table memory
+ * @amdgpu:	amdgpu device.
+ * Free MM table memory
+ */
+void amdgpu_virt_free_mm_table(struct amdgpu_device *adev)
+{
+	if (!amdgpu_sriov_vf(adev) || !adev->virt.mm_table.gpu_addr)
+		return;
+
+	amdgpu_bo_free_kernel(&adev->virt.mm_table.bo,
+			      &adev->virt.mm_table.gpu_addr,
+			      (void *)&adev->virt.mm_table.cpu_addr);
+	adev->virt.mm_table.gpu_addr = 0;
+}

drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h

···
 int amdgpu_virt_release_full_gpu(struct amdgpu_device *adev, bool init);
 int amdgpu_virt_reset_gpu(struct amdgpu_device *adev);
 int amdgpu_sriov_gpu_reset(struct amdgpu_device *adev, bool voluntary);
+int amdgpu_virt_alloc_mm_table(struct amdgpu_device *adev);
+void amdgpu_virt_free_mm_table(struct amdgpu_device *adev);
 
 #endif

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c

···
 		      struct amdgpu_job *job)
 {
 	struct amdgpu_device *adev = ring->adev;
+	unsigned vmhub = ring->funcs->vmhub;
+	struct amdgpu_vm_id_manager *id_mgr = &adev->vm_manager.id_mgr[vmhub];
 	uint64_t fence_context = adev->fence_context + ring->idx;
 	struct dma_fence *updates = sync->last_vm_update;
 	struct amdgpu_vm_id *id, *idle;
···
 	unsigned i;
 	int r = 0;
 
-	fences = kmalloc_array(sizeof(void *), adev->vm_manager.num_ids,
-			       GFP_KERNEL);
+	fences = kmalloc_array(sizeof(void *), id_mgr->num_ids, GFP_KERNEL);
 	if (!fences)
 		return -ENOMEM;
 
-	mutex_lock(&adev->vm_manager.lock);
+	mutex_lock(&id_mgr->lock);
 
 	/* Check if we have an idle VMID */
 	i = 0;
-	list_for_each_entry(idle, &adev->vm_manager.ids_lru, list) {
+	list_for_each_entry(idle, &id_mgr->ids_lru, list) {
 		fences[i] = amdgpu_sync_peek_fence(&idle->active, ring);
 		if (!fences[i])
 			break;
···
 	}
 
 	/* If we can't find a idle VMID to use, wait till one becomes available */
-	if (&idle->list == &adev->vm_manager.ids_lru) {
+	if (&idle->list == &id_mgr->ids_lru) {
 		u64 fence_context = adev->vm_manager.fence_context + ring->idx;
 		unsigned seqno = ++adev->vm_manager.seqno[ring->idx];
 		struct dma_fence_array *array;
···
 		if (r)
 			goto error;
 
-		mutex_unlock(&adev->vm_manager.lock);
+		mutex_unlock(&id_mgr->lock);
 		return 0;
 
 	}
 	kfree(fences);
 
-	job->vm_needs_flush = true;
+	job->vm_needs_flush = false;
 	/* Check if we can use a VMID already assigned to this VM */
-	i = ring->idx;
-	do {
+	list_for_each_entry_reverse(id, &id_mgr->ids_lru, list) {
 		struct dma_fence *flushed;
-
-		id = vm->ids[i++];
-		if (i == AMDGPU_MAX_RINGS)
-			i = 0;
+		bool needs_flush = false;
 
 		/* Check all the prerequisites to using this VMID */
-		if (!id)
-			continue;
 		if (amdgpu_vm_had_gpu_reset(adev, id))
 			continue;
 
···
 		if (job->vm_pd_addr != id->pd_gpu_addr)
 			continue;
 
-		if (!id->last_flush)
-			continue;
-
-		if (id->last_flush->context != fence_context &&
-		    !dma_fence_is_signaled(id->last_flush))
-			continue;
+		if (!id->last_flush ||
+		    (id->last_flush->context != fence_context &&
+		     !dma_fence_is_signaled(id->last_flush)))
+			needs_flush = true;
 
 		flushed  = id->flushed_updates;
-		if (updates &&
-		    (!flushed || dma_fence_is_later(updates, flushed)))
+		if (updates && (!flushed || dma_fence_is_later(updates, flushed)))
+			needs_flush = true;
+
+		/* Concurrent flushes are only possible starting with Vega10 */
+		if (adev->asic_type < CHIP_VEGA10 && needs_flush)
 			continue;
 
 		/* Good we can use this VMID. Remember this submission as
···
 		if (r)
 			goto error;
 
-		list_move_tail(&id->list, &adev->vm_manager.ids_lru);
-		vm->ids[ring->idx] = id;
+		if (updates && (!flushed || dma_fence_is_later(updates, flushed))) {
+			dma_fence_put(id->flushed_updates);
+			id->flushed_updates = dma_fence_get(updates);
+		}
 
-		job->vm_id = id - adev->vm_manager.ids;
-		job->vm_needs_flush = false;
-		trace_amdgpu_vm_grab_id(vm, ring->idx, job);
+		if (needs_flush)
+			goto needs_flush;
+		else
+			goto no_flush_needed;
 
-		mutex_unlock(&adev->vm_manager.lock);
-		return 0;
-
-	} while (i != ring->idx);
+	};
 
 	/* Still no ID to use? Then use the idle one found earlier */
 	id = idle;
···
 	if (r)
 		goto error;
 
+	id->pd_gpu_addr = job->vm_pd_addr;
+	dma_fence_put(id->flushed_updates);
+	id->flushed_updates = dma_fence_get(updates);
+	id->current_gpu_reset_count = atomic_read(&adev->gpu_reset_counter);
+	atomic64_set(&id->owner, vm->client_id);
+
+needs_flush:
+	job->vm_needs_flush = true;
 	dma_fence_put(id->last_flush);
 	id->last_flush = NULL;
 
-	dma_fence_put(id->flushed_updates);
-	id->flushed_updates = dma_fence_get(updates);
+no_flush_needed:
+	list_move_tail(&id->list, &id_mgr->ids_lru);
 
-	id->pd_gpu_addr = job->vm_pd_addr;
-	id->current_gpu_reset_count = atomic_read(&adev->gpu_reset_counter);
-	list_move_tail(&id->list, &adev->vm_manager.ids_lru);
-	atomic64_set(&id->owner, vm->client_id);
-	vm->ids[ring->idx] = id;
-
-	job->vm_id = id - adev->vm_manager.ids;
-	trace_amdgpu_vm_grab_id(vm, ring->idx, job);
+	job->vm_id = id - id_mgr->ids;
+	trace_amdgpu_vm_grab_id(vm, ring, job);
 
 error:
-	mutex_unlock(&adev->vm_manager.lock);
+	mutex_unlock(&id_mgr->lock);
 	return r;
 }
 
···
 int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job)
 {
 	struct amdgpu_device *adev = ring->adev;
-	struct amdgpu_vm_id *id = &adev->vm_manager.ids[job->vm_id];
+	unsigned vmhub = ring->funcs->vmhub;
+	struct amdgpu_vm_id_manager *id_mgr = &adev->vm_manager.id_mgr[vmhub];
+	struct amdgpu_vm_id *id = &id_mgr->ids[job->vm_id];
 	bool gds_switch_needed = ring->funcs->emit_gds_switch && (
 		id->gds_base != job->gds_base ||
 		id->gds_size != job->gds_size ||
···
 	if (ring->funcs->init_cond_exec)
 		patch_offset = amdgpu_ring_init_cond_exec(ring);
 
-	if (ring->funcs->emit_pipeline_sync)
+	if (ring->funcs->emit_pipeline_sync && !job->need_pipeline_sync)
 		amdgpu_ring_emit_pipeline_sync(ring);
 
 	if (ring->funcs->emit_vm_flush && vm_flush_needed) {
 		u64 pd_addr = amdgpu_vm_adjust_mc_addr(adev, job->vm_pd_addr);
 		struct dma_fence *fence;
 
-		trace_amdgpu_vm_flush(pd_addr, ring->idx, job->vm_id);
+		trace_amdgpu_vm_flush(ring, job->vm_id, pd_addr);
 		amdgpu_ring_emit_vm_flush(ring, job->vm_id, pd_addr);
 
 		r = amdgpu_fence_emit(ring, &fence);
 		if (r)
 			return r;
 
-		mutex_lock(&adev->vm_manager.lock);
+		mutex_lock(&id_mgr->lock);
 		dma_fence_put(id->last_flush);
 		id->last_flush = fence;
-		mutex_unlock(&adev->vm_manager.lock);
+		mutex_unlock(&id_mgr->lock);
 	}
 
 	if (gds_switch_needed) {
···
  *
  * Reset saved GDW, GWS and OA to force switch on next flush.
  */
-void amdgpu_vm_reset_id(struct amdgpu_device *adev, unsigned vm_id)
+void amdgpu_vm_reset_id(struct amdgpu_device *adev, unsigned vmhub,
+			unsigned vmid)
 {
-	struct amdgpu_vm_id *id = &adev->vm_manager.ids[vm_id];
+	struct amdgpu_vm_id_manager *id_mgr = &adev->vm_manager.id_mgr[vmhub];
+	struct amdgpu_vm_id *id = &id_mgr->ids[vmid];
 
 	id->gds_base = 0;
 	id->gds_size = 0;
···
 	flags &= ~AMDGPU_PTE_MTYPE_MASK;
 	flags |= (mapping->flags & AMDGPU_PTE_MTYPE_MASK);
 
+	if ((mapping->flags & AMDGPU_PTE_PRT) &&
+	    (adev->asic_type >= CHIP_VEGA10)) {
+		flags |= AMDGPU_PTE_PRT;
+		flags &= ~AMDGPU_PTE_VALID;
+	}
+
 	trace_amdgpu_vm_bo_update(mapping);
 
 	pfn = mapping->offset >> PAGE_SHIFT;
···
 			struct amdgpu_bo_va_mapping, list);
 		list_del(&mapping->list);
 
-		r = amdgpu_vm_bo_split_mapping(adev, NULL, 0, NULL, vm, mapping,
-					       0, 0, &f);
+		r = amdgpu_vm_bo_update_mapping(adev, NULL, 0, NULL, vm,
+						mapping->start, mapping->last,
+						0, 0, &f);
 		amdgpu_vm_free_mapping(adev, vm, mapping, f);
 		if (r) {
 			dma_fence_put(f);
···
 	unsigned ring_instance;
 	struct amdgpu_ring *ring;
 	struct amd_sched_rq *rq;
-	int i, r;
+	int r;
 
-	for (i = 0; i < AMDGPU_MAX_RINGS; ++i)
-		vm->ids[i] = NULL;
 	vm->va = RB_ROOT;
 	vm->client_id = atomic64_inc_return(&adev->vm_manager.client_counter);
 	spin_lock_init(&vm->status_lock);
···
  */
 void amdgpu_vm_manager_init(struct amdgpu_device *adev)
 {
-	unsigned i;
+	unsigned i, j;
 
-	INIT_LIST_HEAD(&adev->vm_manager.ids_lru);
+	for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) {
+		struct amdgpu_vm_id_manager *id_mgr =
+			&adev->vm_manager.id_mgr[i];
 
-	/* skip over VMID 0, since it is the system VM */
-	for (i = 1; i < adev->vm_manager.num_ids; ++i) {
-		amdgpu_vm_reset_id(adev, i);
-		amdgpu_sync_create(&adev->vm_manager.ids[i].active);
-		list_add_tail(&adev->vm_manager.ids[i].list,
-			      &adev->vm_manager.ids_lru);
+		mutex_init(&id_mgr->lock);
+		INIT_LIST_HEAD(&id_mgr->ids_lru);
+
+		/* skip over VMID 0, since it is the system VM */
+		for (j = 1; j < id_mgr->num_ids; ++j) {
+			amdgpu_vm_reset_id(adev, i, j);
+			amdgpu_sync_create(&id_mgr->ids[i].active);
+			list_add_tail(&id_mgr->ids[j].list, &id_mgr->ids_lru);
+		}
 	}
 
 	adev->vm_manager.fence_context =
 		dma_fence_context_alloc(AMDGPU_MAX_RINGS);
 	for (i = 0; i < AMDGPU_MAX_RINGS; ++i)
 		adev->vm_manager.seqno[i] = 0;
+
 
 	atomic_set(&adev->vm_manager.vm_pte_next_ring, 0);
 	atomic64_set(&adev->vm_manager.client_counter, 0);
···
  */
 void amdgpu_vm_manager_fini(struct amdgpu_device *adev)
 {
-	unsigned i;
+	unsigned i, j;
 
-	for (i = 0; i < AMDGPU_NUM_VM; ++i) {
-		struct amdgpu_vm_id *id = &adev->vm_manager.ids[i];
+	for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) {
+		struct amdgpu_vm_id_manager *id_mgr =
+			&adev->vm_manager.id_mgr[i];
 
-		amdgpu_sync_free(&adev->vm_manager.ids[i].active);
-		dma_fence_put(id->flushed_updates);
-		dma_fence_put(id->last_flush);
+		mutex_destroy(&id_mgr->lock);
+		for (j = 0; j < AMDGPU_NUM_VM; ++j) {
+			struct amdgpu_vm_id *id = &id_mgr->ids[j];
+
+			amdgpu_sync_free(&id->active);
+			dma_fence_put(id->flushed_updates);
+			dma_fence_put(id->last_flush);
+		}
 	}
 }

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h

···
 
 #define AMDGPU_PTE_FRAG(x)	((x & 0x1fULL) << 7)
 
-#define AMDGPU_PTE_PRT		(1ULL << 63)
+/* TILED for VEGA10, reserved for older ASICs  */
+#define AMDGPU_PTE_PRT		(1ULL << 51)
 
 /* VEGA10 only */
 #define AMDGPU_PTE_MTYPE(a)    ((uint64_t)a << 57)
···
 	struct dma_fence	*last_dir_update;
 	uint64_t		last_eviction_counter;
 
-	/* for id and flush management per ring */
-	struct amdgpu_vm_id	*ids[AMDGPU_MAX_RINGS];
-
 	/* protecting freed */
 	spinlock_t		freed_lock;
 
···
 	uint32_t		oa_size;
 };
 
+struct amdgpu_vm_id_manager {
+	struct mutex		lock;
+	unsigned		num_ids;
+	struct list_head	ids_lru;
+	struct amdgpu_vm_id	ids[AMDGPU_NUM_VM];
+};
+
 struct amdgpu_vm_manager {
 	/* Handling of VMIDs */
-	struct mutex				lock;
-	unsigned				num_ids;
-	struct list_head			ids_lru;
-	struct amdgpu_vm_id			ids[AMDGPU_NUM_VM];
+	struct amdgpu_vm_id_manager		id_mgr[AMDGPU_MAX_VMHUBS];
 
 	/* Handling of VM fences */
 	u64					fence_context;
···
 		      struct amdgpu_sync *sync, struct dma_fence *fence,
 		      struct amdgpu_job *job);
 int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job);
-void amdgpu_vm_reset_id(struct amdgpu_device *adev, unsigned vm_id);
+void amdgpu_vm_reset_id(struct amdgpu_device *adev, unsigned vmhub,
+			unsigned vmid);
 int amdgpu_vm_update_directories(struct amdgpu_device *adev,
 				 struct amdgpu_vm *vm);
 int amdgpu_vm_clear_freed(struct amdgpu_device *adev,

drivers/gpu/drm/amd/amdgpu/ci_dpm.c

···
 
 static void ci_dpm_set_fan_control_mode(struct amdgpu_device *adev, u32 mode)
 {
-	if (mode) {
-		/* stop auto-manage */
+	switch (mode) {
+	case AMD_FAN_CTRL_NONE:
 		if (adev->pm.dpm.fan.ucode_fan_control)
 			ci_fan_ctrl_stop_smc_fan_control(adev);
-		ci_fan_ctrl_set_static_mode(adev, mode);
-	} else {
-		/* restart auto-manage */
+		ci_dpm_set_fan_speed_percent(adev, 100);
+		break;
+	case AMD_FAN_CTRL_MANUAL:
+		if (adev->pm.dpm.fan.ucode_fan_control)
+			ci_fan_ctrl_stop_smc_fan_control(adev);
+		break;
+	case AMD_FAN_CTRL_AUTO:
 		if (adev->pm.dpm.fan.ucode_fan_control)
 			ci_thermal_start_smc_fan_control(adev);
-		else
-			ci_fan_ctrl_set_default_mode(adev);
+		break;
+	default:
+		break;
 	}
 }
 
 static u32 ci_dpm_get_fan_control_mode(struct amdgpu_device *adev)
 {
 	struct ci_power_info *pi = ci_get_pi(adev);
-	u32 tmp;
 
 	if (pi->fan_is_controlled_by_smc)
-		return 0;
-
-	tmp = RREG32_SMC(ixCG_FDO_CTRL2) & CG_FDO_CTRL2__FDO_PWM_MODE_MASK;
-	return (tmp >> CG_FDO_CTRL2__FDO_PWM_MODE__SHIFT);
+		return AMD_FAN_CTRL_AUTO;
+	else
+		return AMD_FAN_CTRL_MANUAL;
 }
 
 #if 0
···
 						      memory_clock,
 						      &memory_level->MinVddcPhases);
 
+	memory_level->EnabledForActivity = 1;
 	memory_level->EnabledForThrottle = 1;
 	memory_level->UpH = 0;
 	memory_level->DownH = 100;
···
 		if (ret)
 			return ret;
 	}
-
-	pi->smc_state_table.MemoryLevel[0].EnabledForActivity = 1;
 
 	if ((dpm_table->mclk_table.count >= 2) &&
 	    ((adev->pdev->device == 0x67B0) || (adev->pdev->device == 0x67B1))) {

drivers/gpu/drm/amd/amdgpu/dce_v10_0.c

···
 	if (!atomic && fb && fb != crtc->primary->fb) {
 		amdgpu_fb = to_amdgpu_framebuffer(fb);
 		abo = gem_to_amdgpu_bo(amdgpu_fb->obj);
-		r = amdgpu_bo_reserve(abo, false);
+		r = amdgpu_bo_reserve(abo, true);
 		if (unlikely(r != 0))
 			return r;
 		amdgpu_bo_unpin(abo);
···
 unpin:
 	if (amdgpu_crtc->cursor_bo) {
 		struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
-		ret = amdgpu_bo_reserve(aobj, false);
+		ret = amdgpu_bo_reserve(aobj, true);
 		if (likely(ret == 0)) {
 			amdgpu_bo_unpin(aobj);
 			amdgpu_bo_unreserve(aobj);
···
 
 		amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb);
 		abo = gem_to_amdgpu_bo(amdgpu_fb->obj);
-		r = amdgpu_bo_reserve(abo, false);
+		r = amdgpu_bo_reserve(abo, true);
 		if (unlikely(r))
 			DRM_ERROR("failed to reserve abo before unpin\n");
 		else {

drivers/gpu/drm/amd/amdgpu/dce_v11_0.c

···
 	if (!atomic && fb && fb != crtc->primary->fb) {
 		amdgpu_fb = to_amdgpu_framebuffer(fb);
 		abo = gem_to_amdgpu_bo(amdgpu_fb->obj);
-		r = amdgpu_bo_reserve(abo, false);
+		r = amdgpu_bo_reserve(abo, true);
 		if (unlikely(r != 0))
 			return r;
 		amdgpu_bo_unpin(abo);
···
 unpin:
 	if (amdgpu_crtc->cursor_bo) {
 		struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
-		ret = amdgpu_bo_reserve(aobj, false);
+		ret = amdgpu_bo_reserve(aobj, true);
 		if (likely(ret == 0)) {
 			amdgpu_bo_unpin(aobj);
 			amdgpu_bo_unreserve(aobj);
···
 
 		amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb);
 		abo = gem_to_amdgpu_bo(amdgpu_fb->obj);
-		r = amdgpu_bo_reserve(abo, false);
+		r = amdgpu_bo_reserve(abo, true);
 		if (unlikely(r))
 			DRM_ERROR("failed to reserve abo before unpin\n");
 		else {

drivers/gpu/drm/amd/amdgpu/dce_v6_0.c

···
 	u32 priority_a_mark = 0, priority_b_mark = 0;
 	u32 priority_a_cnt = PRIORITY_OFF;
 	u32 priority_b_cnt = PRIORITY_OFF;
-	u32 tmp, arb_control3;
+	u32 tmp, arb_control3, lb_vblank_lead_lines = 0;
 	fixed20_12 a, b, c;
 
 	if (amdgpu_crtc->base.enabled && num_heads && mode) {
···
 		c.full = dfixed_div(c, a);
 		priority_b_mark = dfixed_trunc(c);
 		priority_b_cnt |= priority_b_mark & PRIORITY_MARK_MASK;
+
+		lb_vblank_lead_lines = DIV_ROUND_UP(lb_size, mode->crtc_hdisplay);
 	}
 
 	/* select wm A */
···
 	/* save values for DPM */
 	amdgpu_crtc->line_time = line_time;
 	amdgpu_crtc->wm_high = latency_watermark_a;
+
+	/* Save number of lines the linebuffer leads before the scanout */
+	amdgpu_crtc->lb_vblank_lead_lines = lb_vblank_lead_lines;
 }
 
 /* watermark setup */
···
 	if (!atomic && fb && fb != crtc->primary->fb) {
 		amdgpu_fb = to_amdgpu_framebuffer(fb);
 		abo = gem_to_amdgpu_bo(amdgpu_fb->obj);
-		r = amdgpu_bo_reserve(abo, false);
+		r = amdgpu_bo_reserve(abo, true);
 		if (unlikely(r != 0))
 			return r;
 		amdgpu_bo_unpin(abo);
···
 unpin:
 	if (amdgpu_crtc->cursor_bo) {
 		struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
-		ret = amdgpu_bo_reserve(aobj, false);
+		ret = amdgpu_bo_reserve(aobj, true);
 		if (likely(ret == 0)) {
 			amdgpu_bo_unpin(aobj);
 			amdgpu_bo_unreserve(aobj);
···
 
 		amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb);
 		abo = gem_to_amdgpu_bo(amdgpu_fb->obj);
-		r = amdgpu_bo_reserve(abo, false);
+		r = amdgpu_bo_reserve(abo, true);
 		if (unlikely(r))
 			DRM_ERROR("failed to reserve abo before unpin\n");
 		else {

drivers/gpu/drm/amd/amdgpu/dce_v8_0.c

···
 	if (!atomic && fb && fb != crtc->primary->fb) {
 		amdgpu_fb = to_amdgpu_framebuffer(fb);
 		abo = gem_to_amdgpu_bo(amdgpu_fb->obj);
-		r = amdgpu_bo_reserve(abo, false);
+		r = amdgpu_bo_reserve(abo, true);
 		if (unlikely(r != 0))
 			return r;
 		amdgpu_bo_unpin(abo);
···
 unpin:
 	if (amdgpu_crtc->cursor_bo) {
 		struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
-		ret = amdgpu_bo_reserve(aobj, false);
+		ret = amdgpu_bo_reserve(aobj, true);
 		if (likely(ret == 0)) {
 			amdgpu_bo_unpin(aobj);
 			amdgpu_bo_unreserve(aobj);
···
 
 		amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb);
 		abo = gem_to_amdgpu_bo(amdgpu_fb->obj);
-		r = amdgpu_bo_reserve(abo, false);
+		r = amdgpu_bo_reserve(abo, true);
 		if (unlikely(r))
 			DRM_ERROR("failed to reserve abo before unpin\n");
 		else {

drivers/gpu/drm/amd/amdgpu/dce_virtual.c

···
 
 		amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb);
 		abo = gem_to_amdgpu_bo(amdgpu_fb->obj);
-		r = amdgpu_bo_reserve(abo, false);
+		r = amdgpu_bo_reserve(abo, true);
 		if (unlikely(r))
 			DRM_ERROR("failed to reserve abo before unpin\n");
 		else {

drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c

···
 
 static void gfx_v6_0_config_init(struct amdgpu_device *adev)
 {
-	adev->gfx.config.double_offchip_lds_buf = 1;
+	adev->gfx.config.double_offchip_lds_buf = 0;
 }
 
 static void gfx_v6_0_gpu_init(struct amdgpu_device *adev)
···
 	int r;
 
 	if (adev->gfx.rlc.save_restore_obj) {
-		r = amdgpu_bo_reserve(adev->gfx.rlc.save_restore_obj, false);
+		r = amdgpu_bo_reserve(adev->gfx.rlc.save_restore_obj, true);
 		if (unlikely(r != 0))
 			dev_warn(adev->dev, "(%d) reserve RLC sr bo failed\n", r);
 		amdgpu_bo_unpin(adev->gfx.rlc.save_restore_obj);
···
 	}
 
 	if (adev->gfx.rlc.clear_state_obj) {
-		r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, false);
+		r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, true);
 		if (unlikely(r != 0))
 			dev_warn(adev->dev, "(%d) reserve RLC c bo failed\n", r);
 		amdgpu_bo_unpin(adev->gfx.rlc.clear_state_obj);
···
 	}
 
 	if (adev->gfx.rlc.cp_table_obj) {
-		r = amdgpu_bo_reserve(adev->gfx.rlc.cp_table_obj, false);
+		r = amdgpu_bo_reserve(adev->gfx.rlc.cp_table_obj, true);
 		if (unlikely(r != 0))
 			dev_warn(adev->dev, "(%d) reserve RLC cp table bo failed\n", r);
 		amdgpu_bo_unpin(adev->gfx.rlc.cp_table_obj);
···
 		ring->me = 1;
 		ring->pipe = i;
 		ring->queue = i;
-		sprintf(ring->name, "comp %d.%d.%d", ring->me, ring->pipe, ring->queue);
+		sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
 		irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP + ring->pipe;
 		r = amdgpu_ring_init(adev, ring, 1024,
 				     &adev->gfx.eop_irq, irq_type);

drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c

···
 				   INDEX_STRIDE, 3);
 
 	mutex_lock(&adev->srbm_mutex);
-	for (i = 0; i < adev->vm_manager.num_ids; i++) {
+	for (i = 0; i < adev->vm_manager.id_mgr[0].num_ids; i++) {
 		if (i == 0)
 			sh_mem_base = 0;
 		else
···
 		struct amdgpu_ring *ring = &adev->gfx.compute_ring[i];
 
 		if (ring->mqd_obj) {
-			r = amdgpu_bo_reserve(ring->mqd_obj, false);
+			r = amdgpu_bo_reserve(ring->mqd_obj, true);
 			if (unlikely(r != 0))
 				dev_warn(adev->dev, "(%d) reserve MQD bo failed\n", r);
 
···
 	int r;
 
 	if (adev->gfx.mec.hpd_eop_obj) {
-		r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, false);
+		r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, true);
 		if (unlikely(r != 0))
 			dev_warn(adev->dev, "(%d) reserve HPD EOP bo failed\n", r);
 		amdgpu_bo_unpin(adev->gfx.mec.hpd_eop_obj);
···
 
 	/* save restore block */
 	if (adev->gfx.rlc.save_restore_obj) {
-		r = amdgpu_bo_reserve(adev->gfx.rlc.save_restore_obj, false);
+		r = amdgpu_bo_reserve(adev->gfx.rlc.save_restore_obj, true);
 		if (unlikely(r != 0))
 			dev_warn(adev->dev, "(%d) reserve RLC sr bo failed\n", r);
 		amdgpu_bo_unpin(adev->gfx.rlc.save_restore_obj);
···
 
 	/* clear state block */
 	if (adev->gfx.rlc.clear_state_obj) {
-		r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, false);
+		r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, true);
 		if (unlikely(r != 0))
 			dev_warn(adev->dev, "(%d) reserve RLC c bo failed\n", r);
 		amdgpu_bo_unpin(adev->gfx.rlc.clear_state_obj);
···
 
 	/* clear state block */
 	if (adev->gfx.rlc.cp_table_obj) {
-		r = amdgpu_bo_reserve(adev->gfx.rlc.cp_table_obj, false);
+		r = amdgpu_bo_reserve(adev->gfx.rlc.cp_table_obj, true);
 		if (unlikely(r != 0))
 			dev_warn(adev->dev, "(%d) reserve RLC cp table bo failed\n", r);
 		amdgpu_bo_unpin(adev->gfx.rlc.cp_table_obj);

drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c

···
 
 	/* clear state block */
 	if (adev->gfx.rlc.clear_state_obj) {
-		r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, false);
+		r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, true);
 		if (unlikely(r != 0))
 			dev_warn(adev->dev, "(%d) reserve RLC cbs bo failed\n", r);
 		amdgpu_bo_unpin(adev->gfx.rlc.clear_state_obj);
···
 
 	/* jump table block */
 	if (adev->gfx.rlc.cp_table_obj) {
-		r = amdgpu_bo_reserve(adev->gfx.rlc.cp_table_obj, false);
+		r = amdgpu_bo_reserve(adev->gfx.rlc.cp_table_obj, true);
 		if (unlikely(r != 0))
 			dev_warn(adev->dev, "(%d) reserve RLC cp table bo failed\n", r);
 		amdgpu_bo_unpin(adev->gfx.rlc.cp_table_obj);
···
 	int r;
 
 	if (adev->gfx.mec.hpd_eop_obj) {
-		r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, false);
+		r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, true);
 		if (unlikely(r != 0))
 			dev_warn(adev->dev, "(%d) reserve HPD EOP bo failed\n", r);
 		amdgpu_bo_unpin(adev->gfx.mec.hpd_eop_obj);
···
 
 	memset(hpd, 0, MEC_HPD_SIZE);
 
-	r = amdgpu_bo_reserve(kiq->eop_obj, false);
+	r = amdgpu_bo_reserve(kiq->eop_obj, true);
 	if (unlikely(r != 0))
 		dev_warn(adev->dev, "(%d) reserve kiq eop bo failed\n", r);
 	amdgpu_bo_kunmap(kiq->eop_obj);
···
 		case 0xca:
 		case 0xce:
 		case 0x88:
+		case 0xe6:
 			/* B6 */
 			adev->gfx.config.max_cu_per_sh = 6;
 			break;
···
 		adev->gfx.config.max_backends_per_se = 1;
 
 		switch (adev->pdev->revision) {
+		case 0x80:
+		case 0x81:
 		case 0xc0:
 		case 0xc1:
 		case 0xc2:
 		case 0xc4:
 		case 0xc8:
 		case 0xc9:
+		case 0xd6:
+		case 0xda:
+		case 0xe9:
+		case 0xea:
 			adev->gfx.config.max_cu_per_sh = 3;
 			break;
+		case 0x83:
 		case 0xd0:
 		case 0xd1:
 		case 0xd2:
+		case 0xd4:
+		case 0xdb:
+		case 0xe1:
+		case 0xe2:
 		default:
 			adev->gfx.config.max_cu_per_sh = 2;
 			break;
···
 	sh_static_mem_cfg = REG_SET_FIELD(sh_static_mem_cfg, SH_STATIC_MEM_CONFIG,
 				   INDEX_STRIDE, 3);
 	mutex_lock(&adev->srbm_mutex);
-	for (i = 0; i < adev->vm_manager.num_ids; i++) {
+	for (i = 0; i < adev->vm_manager.id_mgr[0].num_ids; i++) {
 		vi_srbm_select(adev, 0, 0, 0, i);
 		/* CP and shaders */
 		if (i == 0) {

drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c

···
 
 #define GFX9_NUM_GFX_RINGS     1
 #define GFX9_NUM_COMPUTE_RINGS 8
-#define GFX9_NUM_SE		4
 #define RLCG_UCODE_LOADING_START_ADDRESS 0x2000
 
 MODULE_FIRMWARE("amdgpu/vega10_ce.bin");
···
 	int r;
 
 	if (adev->gfx.mec.hpd_eop_obj) {
-		r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, false);
+		r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, true);
 		if (unlikely(r != 0))
 			dev_warn(adev->dev, "(%d) reserve HPD EOP bo failed\n", r);
 		amdgpu_bo_unpin(adev->gfx.mec.hpd_eop_obj);
···
 		adev->gfx.mec.hpd_eop_obj = NULL;
 	}
 	if (adev->gfx.mec.mec_fw_obj) {
-		r = amdgpu_bo_reserve(adev->gfx.mec.mec_fw_obj, false);
+		r = amdgpu_bo_reserve(adev->gfx.mec.mec_fw_obj, true);
 		if (unlikely(r != 0))
 			dev_warn(adev->dev, "(%d) reserve mec firmware bo failed\n", r);
 		amdgpu_bo_unpin(adev->gfx.mec.mec_fw_obj);
···
 
 	memset(hpd, 0, MEC_HPD_SIZE);
 
-	r = amdgpu_bo_reserve(kiq->eop_obj, false);
+	r = amdgpu_bo_reserve(kiq->eop_obj, true);
 	if (unlikely(r != 0))
 		dev_warn(adev->dev, "(%d) reserve kiq eop bo failed\n", r);
 	amdgpu_bo_kunmap(kiq->eop_obj);
···
 		ring->pipe = 1;
 	}
 
-	irq->data = ring;
 	ring->queue = 0;
 	ring->eop_gpu_addr = kiq->eop_gpu_addr;
 	sprintf(ring->name, "kiq %d.%d.%d", ring->me, ring->pipe, ring->queue);
···
 {
 	amdgpu_wb_free(ring->adev, ring->adev->virt.reg_val_offs);
 	amdgpu_ring_fini(ring);
-	irq->data = NULL;
 }
 
 /* create MQD for each compute queue */
···
 
 static uint32_t wave_read_ind(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t address)
 {
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_IND_INDEX),
+	WREG32_SOC15(GC, 0, mmSQ_IND_INDEX,
 		(wave << SQ_IND_INDEX__WAVE_ID__SHIFT) |
 		(simd << SQ_IND_INDEX__SIMD_ID__SHIFT) |
 		(address << SQ_IND_INDEX__INDEX__SHIFT) |
 		(SQ_IND_INDEX__FORCE_READ_MASK));
-	return RREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_IND_DATA));
+	return RREG32_SOC15(GC, 0, mmSQ_IND_DATA);
 }
 
 static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd,
 			   uint32_t wave, uint32_t thread,
 			   uint32_t regno, uint32_t num, uint32_t *out)
 {
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_IND_INDEX),
+	WREG32_SOC15(GC, 0, mmSQ_IND_INDEX,
 		(wave << SQ_IND_INDEX__WAVE_ID__SHIFT) |
 		(simd << SQ_IND_INDEX__SIMD_ID__SHIFT) |
 		(regno << SQ_IND_INDEX__INDEX__SHIFT) |
···
 		(SQ_IND_INDEX__FORCE_READ_MASK) |
 		(SQ_IND_INDEX__AUTO_INCR_MASK));
 	while (num--)
-		*(out++) = RREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_IND_DATA));
+		*(out++) = RREG32_SOC15(GC, 0, mmSQ_IND_DATA);
 }
 
 static void gfx_v9_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
···
 	switch (adev->asic_type) {
 	case CHIP_VEGA10:
 		adev->gfx.config.max_shader_engines = 4;
-		adev->gfx.config.max_tile_pipes = 8; //??
 		adev->gfx.config.max_cu_per_sh = 16;
 		adev->gfx.config.max_sh_per_se = 1;
 		adev->gfx.config.max_backends_per_se = 4;
···
 		adev->gfx.config.sc_prim_fifo_size_backend = 0x100;
 		adev->gfx.config.sc_hiz_tile_fifo_size = 0x30;
 		adev->gfx.config.sc_earlyz_tile_fifo_size = 0x4C0;
+		adev->gfx.config.gs_vgt_table_depth = 32;
+		adev->gfx.config.gs_prim_buffer_depth = 1792;
 		gb_addr_config = VEGA10_GB_ADDR_CONFIG_GOLDEN;
 		break;
 	default:
···
 					adev->gfx.config.gb_addr_config,
 					GB_ADDR_CONFIG,
 					NUM_PIPES);
+
+	adev->gfx.config.max_tile_pipes =
+		adev->gfx.config.gb_addr_config_fields.num_pipes;
+
 	adev->gfx.config.gb_addr_config_fields.num_banks = 1 <<
 			REG_GET_FIELD(
 					adev->gfx.config.gb_addr_config,
···
 	}
 	size_se = size_se ? size_se : default_size_se;
 
-	ngg_buf->size = size_se * GFX9_NUM_SE;
+	ngg_buf->size = size_se * adev->gfx.config.max_shader_engines;
 	r = amdgpu_bo_create_kernel(adev, ngg_buf->size,
 				    PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM,
 				    &ngg_buf->bo,
···
 	adev->gfx.ngg.gds_reserve_addr += adev->gds.mem.gfx_partition_size;
 
 	/* Primitive Buffer */
-	r = gfx_v9_0_ngg_create_buf(adev, &adev->gfx.ngg.buf[PRIM],
+	r = gfx_v9_0_ngg_create_buf(adev, &adev->gfx.ngg.buf[NGG_PRIM],
 				    amdgpu_prim_buf_per_se,
 				    64 * 1024);
 	if (r) {
···
 	}
 
 	/* Position Buffer */
-	r = gfx_v9_0_ngg_create_buf(adev, &adev->gfx.ngg.buf[POS],
+	r = gfx_v9_0_ngg_create_buf(adev, &adev->gfx.ngg.buf[NGG_POS],
 				    amdgpu_pos_buf_per_se,
 				    256 * 1024);
 	if (r) {
···
 	}
 
 	/* Control Sideband */
-	r = gfx_v9_0_ngg_create_buf(adev, &adev->gfx.ngg.buf[CNTL],
+	r = gfx_v9_0_ngg_create_buf(adev, &adev->gfx.ngg.buf[NGG_CNTL],
 				    amdgpu_cntl_sb_buf_per_se,
 				    256);
 	if (r) {
···
 	if (amdgpu_param_buf_per_se <= 0)
 		goto out;
 
-	r = gfx_v9_0_ngg_create_buf(adev, &adev->gfx.ngg.buf[PARAM],
+	r = gfx_v9_0_ngg_create_buf(adev, &adev->gfx.ngg.buf[NGG_PARAM],
 				    amdgpu_param_buf_per_se,
 				    512 * 1024);
 	if (r) {
···
 
 	/* Program buffer size */
 	data = 0;
-	size = adev->gfx.ngg.buf[PRIM].size / 256;
+	size = adev->gfx.ngg.buf[NGG_PRIM].size / 256;
 	data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, INDEX_BUF_SIZE, size);
 
-	size = adev->gfx.ngg.buf[POS].size / 256;
+	size = adev->gfx.ngg.buf[NGG_POS].size / 256;
 	data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, POS_BUF_SIZE, size);
 
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmWD_BUF_RESOURCE_1), data);
+	WREG32_SOC15(GC, 0, mmWD_BUF_RESOURCE_1, data);
 
 	data = 0;
-	size = adev->gfx.ngg.buf[CNTL].size / 256;
+	size = adev->gfx.ngg.buf[NGG_CNTL].size / 256;
 	data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, CNTL_SB_BUF_SIZE, size);
 
-	size = adev->gfx.ngg.buf[PARAM].size / 1024;
+	size = adev->gfx.ngg.buf[NGG_PARAM].size / 1024;
 	data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, PARAM_BUF_SIZE, size);
 
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmWD_BUF_RESOURCE_2), data);
+	WREG32_SOC15(GC, 0, mmWD_BUF_RESOURCE_2, data);
 
 	/* Program buffer base address */
-	base = lower_32_bits(adev->gfx.ngg.buf[PRIM].gpu_addr);
+	base = lower_32_bits(adev->gfx.ngg.buf[NGG_PRIM].gpu_addr);
 	data = REG_SET_FIELD(0, WD_INDEX_BUF_BASE, BASE, base);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmWD_INDEX_BUF_BASE), data);
+	WREG32_SOC15(GC, 0, mmWD_INDEX_BUF_BASE, data);
 
-	base = upper_32_bits(adev->gfx.ngg.buf[PRIM].gpu_addr);
+	base = upper_32_bits(adev->gfx.ngg.buf[NGG_PRIM].gpu_addr);
 	data = REG_SET_FIELD(0, WD_INDEX_BUF_BASE_HI, BASE_HI, base);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmWD_INDEX_BUF_BASE_HI), data);
+	WREG32_SOC15(GC, 0, mmWD_INDEX_BUF_BASE_HI, data);
 
-	base = lower_32_bits(adev->gfx.ngg.buf[POS].gpu_addr);
+	base = lower_32_bits(adev->gfx.ngg.buf[NGG_POS].gpu_addr);
 	data = REG_SET_FIELD(0, WD_POS_BUF_BASE, BASE, base);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmWD_POS_BUF_BASE), data);
+	WREG32_SOC15(GC, 0, mmWD_POS_BUF_BASE, data);
 
-	base = upper_32_bits(adev->gfx.ngg.buf[POS].gpu_addr);
+	base = upper_32_bits(adev->gfx.ngg.buf[NGG_POS].gpu_addr);
 	data = REG_SET_FIELD(0, WD_POS_BUF_BASE_HI, BASE_HI, base);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmWD_POS_BUF_BASE_HI), data);
+	WREG32_SOC15(GC, 0, mmWD_POS_BUF_BASE_HI, data);
 
-	base = lower_32_bits(adev->gfx.ngg.buf[CNTL].gpu_addr);
+	base = lower_32_bits(adev->gfx.ngg.buf[NGG_CNTL].gpu_addr);
 	data = REG_SET_FIELD(0, WD_CNTL_SB_BUF_BASE, BASE, base);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmWD_CNTL_SB_BUF_BASE), data);
+	WREG32_SOC15(GC, 0, mmWD_CNTL_SB_BUF_BASE, data);
 
-	base = upper_32_bits(adev->gfx.ngg.buf[CNTL].gpu_addr);
+	base = upper_32_bits(adev->gfx.ngg.buf[NGG_CNTL].gpu_addr);
 	data = REG_SET_FIELD(0, WD_CNTL_SB_BUF_BASE_HI, BASE_HI, base);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmWD_CNTL_SB_BUF_BASE_HI), data);
+	WREG32_SOC15(GC, 0, mmWD_CNTL_SB_BUF_BASE_HI, data);
 
 	/* Clear GDS reserved memory */
 	r = amdgpu_ring_alloc(ring, 17);
···
 		ring->pipe = i / 8;
 		ring->queue = i % 8;
 		ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr + (i * MEC_HPD_SIZE);
-		sprintf(ring->name, "comp %d.%d.%d", ring->me, ring->pipe, ring->queue);
+		sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
 		irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP + ring->pipe;
 		/* type-2 packets are deprecated on MEC, use type-3 instead */
 		r = amdgpu_ring_init(adev, ring, 1024,
···
 		data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_INDEX, sh_num);
 		data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_INDEX, se_num);
 	}
-	WREG32( SOC15_REG_OFFSET(GC, 0, mmGRBM_GFX_INDEX), data);
+	WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, data);
 }
 
 static u32 gfx_v9_0_create_bitmask(u32 bit_width)
···
 {
 	u32 data, mask;
 
-	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmCC_RB_BACKEND_DISABLE));
-	data |= RREG32(SOC15_REG_OFFSET(GC, 0, mmGC_USER_RB_BACKEND_DISABLE));
+	data = RREG32_SOC15(GC, 0, mmCC_RB_BACKEND_DISABLE);
+	data |= RREG32_SOC15(GC, 0, mmGC_USER_RB_BACKEND_DISABLE);
 
 	data &= CC_RB_BACKEND_DISABLE__BACKEND_DISABLE_MASK;
 	data >>= GC_USER_RB_BACKEND_DISABLE__BACKEND_DISABLE__SHIFT;
···
 	for (i = FIRST_COMPUTE_VMID; i < LAST_COMPUTE_VMID; i++) {
 		soc15_grbm_select(adev, 0, 0, 0, i);
 		/* CP and shaders */
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_CONFIG), sh_mem_config);
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_BASES), sh_mem_bases);
+		WREG32_SOC15(GC, 0, mmSH_MEM_CONFIG, sh_mem_config);
+		WREG32_SOC15(GC, 0, mmSH_MEM_BASES, sh_mem_bases);
 	}
 	soc15_grbm_select(adev, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
···
 		tmp = 0;
 		tmp = REG_SET_FIELD(tmp, SH_MEM_CONFIG, ALIGNMENT_MODE,
 				    SH_MEM_ALIGNMENT_MODE_UNALIGNED);
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_CONFIG), tmp);
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_BASES), 0);
+		WREG32_SOC15(GC, 0, mmSH_MEM_CONFIG, tmp);
+		WREG32_SOC15(GC, 0, mmSH_MEM_BASES, 0);
 	}
 	soc15_grbm_select(adev, 0, 0, 0, 0);
 
···
 	 */
 	gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
 
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmPA_SC_FIFO_SIZE),
+	WREG32_SOC15(GC, 0, mmPA_SC_FIFO_SIZE,
 		   (adev->gfx.config.sc_prim_fifo_size_frontend <<
 			PA_SC_FIFO_SIZE__SC_FRONTEND_PRIM_FIFO_SIZE__SHIFT) |
 		   (adev->gfx.config.sc_prim_fifo_size_backend <<
···
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
 			gfx_v9_0_select_se_sh(adev, i, j, 0xffffffff);
 			for (k = 0; k < adev->usec_timeout; k++) {
-				if (RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SERDES_CU_MASTER_BUSY)) == 0)
+				if (RREG32_SOC15(GC, 0, mmRLC_SERDES_CU_MASTER_BUSY) == 0)
 					break;
 				udelay(1);
 			}
···
 		RLC_SERDES_NONCU_MASTER_BUSY__TC0_MASTER_BUSY_MASK |
 		RLC_SERDES_NONCU_MASTER_BUSY__TC1_MASTER_BUSY_MASK;
 	for (k = 0; k < adev->usec_timeout; k++) {
-		if ((RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SERDES_NONCU_MASTER_BUSY)) & mask) == 0)
+		if ((RREG32_SOC15(GC, 0, mmRLC_SERDES_NONCU_MASTER_BUSY) & mask) == 0)
 			break;
 		udelay(1);
 	}
···
 static void gfx_v9_0_enable_gui_idle_interrupt(struct amdgpu_device *adev,
 					       bool enable)
 {
-	u32 tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_INT_CNTL_RING0));
+	u32 tmp = RREG32_SOC15(GC, 0, mmCP_INT_CNTL_RING0);
 
 	if (enable)
 		return;
···
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE, enable ? 1 : 0);
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE, enable ? 1 : 0);
 
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_INT_CNTL_RING0), tmp);
+	WREG32_SOC15(GC, 0, mmCP_INT_CNTL_RING0, tmp);
 }
 
 void gfx_v9_0_rlc_stop(struct amdgpu_device *adev)
 {
-	u32 tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CNTL));
+	u32 tmp = RREG32_SOC15(GC, 0, mmRLC_CNTL);
 
 	tmp = REG_SET_FIELD(tmp, RLC_CNTL, RLC_ENABLE_F32, 0);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CNTL), tmp);
+	WREG32_SOC15(GC, 0, mmRLC_CNTL, tmp);
 
 	gfx_v9_0_enable_gui_idle_interrupt(adev, false);
 
···
 
 #ifdef AMDGPU_RLC_DEBUG_RETRY
 	/* RLC_GPM_GENERAL_6 : RLC Ucode version */
-	rlc_ucode_ver = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_GENERAL_6));
+	rlc_ucode_ver = RREG32_SOC15(GC, 0, mmRLC_GPM_GENERAL_6);
 	if(rlc_ucode_ver == 0x108) {
 		DRM_INFO("Using rlc debug ucode. mmRLC_GPM_GENERAL_6 ==0x08%x / fw_ver == %i \n",
 				rlc_ucode_ver, adev->gfx.rlc_fw_version);
 		/* RLC_GPM_TIMER_INT_3 : Timer interval in RefCLK cycles,
 		 * default is 0x9C4 to create a 100us interval */
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_TIMER_INT_3), 0x9C4);
+		WREG32_SOC15(GC, 0, mmRLC_GPM_TIMER_INT_3, 0x9C4);
 		/* RLC_GPM_GENERAL_12 : Minimum gap between wptr and rptr
 		 * to disable the page fault retry interrupts, default is 
 		 * 0x100 (256) */
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_GENERAL_12), 0x100);
+		WREG32_SOC15(GC, 0, mmRLC_GPM_GENERAL_12, 0x100);
 	}
 #endif
 }
···
 			   le32_to_cpu(hdr->header.ucode_array_offset_bytes));
 	fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4;
 
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_UCODE_ADDR),
+	WREG32_SOC15(GC, 0, mmRLC_GPM_UCODE_ADDR,
 			RLCG_UCODE_LOADING_START_ADDRESS);
 	for (i = 0; i < fw_size; i++)
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_UCODE_DATA), le32_to_cpup(fw_data++));
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_UCODE_ADDR), adev->gfx.rlc_fw_version);
+		WREG32_SOC15(GC, 0, mmRLC_GPM_UCODE_DATA, le32_to_cpup(fw_data++));
+	WREG32_SOC15(GC, 0, mmRLC_GPM_UCODE_ADDR, adev->gfx.rlc_fw_version);
 
 	return 0;
 }
···
 	gfx_v9_0_rlc_stop(adev);
 
 	/* disable CG */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGCG_CGLS_CTRL), 0);
+	WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL, 0);
 
 	/* disable PG */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), 0);
+	WREG32_SOC15(GC, 0, mmRLC_PG_CNTL, 0);
 
 	gfx_v9_0_rlc_reset(adev);
 
···
 static void gfx_v9_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable)
 {
 	int i;
-	u32 tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_ME_CNTL));
+	u32 tmp = RREG32_SOC15(GC, 0, mmCP_ME_CNTL);
 
 	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_HALT, enable ? 0 : 1);
 	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_HALT, enable ? 0 : 1);
···
 		for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 			adev->gfx.gfx_ring[i].ready = false;
 	}
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_ME_CNTL), tmp);
+	WREG32_SOC15(GC, 0, mmCP_ME_CNTL, tmp);
 	udelay(50);
 }
 
···
 		(adev->gfx.pfp_fw->data +
 		 le32_to_cpu(pfp_hdr->header.ucode_array_offset_bytes));
 	fw_size = le32_to_cpu(pfp_hdr->header.ucode_size_bytes) / 4;
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_PFP_UCODE_ADDR), 0);
+	WREG32_SOC15(GC, 0, mmCP_PFP_UCODE_ADDR, 0);
 	for (i = 0; i < fw_size; i++)
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_PFP_UCODE_DATA), le32_to_cpup(fw_data++));
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_PFP_UCODE_ADDR), adev->gfx.pfp_fw_version);
+		WREG32_SOC15(GC, 0, mmCP_PFP_UCODE_DATA, le32_to_cpup(fw_data++));
+	WREG32_SOC15(GC, 0, mmCP_PFP_UCODE_ADDR, adev->gfx.pfp_fw_version);
 
 	/* CE */
 	fw_data = (const __le32 *)
 		(adev->gfx.ce_fw->data +
 		 le32_to_cpu(ce_hdr->header.ucode_array_offset_bytes));
 	fw_size = le32_to_cpu(ce_hdr->header.ucode_size_bytes) / 4;
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_CE_UCODE_ADDR), 0);
+	WREG32_SOC15(GC, 0, mmCP_CE_UCODE_ADDR, 0);
 	for (i = 0; i < fw_size; i++)
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_CE_UCODE_DATA), le32_to_cpup(fw_data++));
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_CE_UCODE_ADDR), adev->gfx.ce_fw_version);
+		WREG32_SOC15(GC, 0, mmCP_CE_UCODE_DATA, le32_to_cpup(fw_data++));
+	WREG32_SOC15(GC, 0, mmCP_CE_UCODE_ADDR, adev->gfx.ce_fw_version);
 
 	/* ME */
 	fw_data = (const __le32 *)
 		(adev->gfx.me_fw->data +
 		 le32_to_cpu(me_hdr->header.ucode_array_offset_bytes));
 	fw_size = le32_to_cpu(me_hdr->header.ucode_size_bytes) / 4;
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_ME_RAM_WADDR), 0);
+	WREG32_SOC15(GC, 0, mmCP_ME_RAM_WADDR, 0);
 	for (i = 0; i < fw_size; i++)
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_ME_RAM_DATA), le32_to_cpup(fw_data++));
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_ME_RAM_WADDR), adev->gfx.me_fw_version);
+		WREG32_SOC15(GC, 0, mmCP_ME_RAM_DATA, le32_to_cpup(fw_data++));
+	WREG32_SOC15(GC, 0, mmCP_ME_RAM_WADDR, adev->gfx.me_fw_version);
 
 	return 0;
 }
···
 	int r, i;
 
 	/* init the CP */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MAX_CONTEXT), adev->gfx.config.max_hw_contexts - 1);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_DEVICE_ID), 1);
+	WREG32_SOC15(GC, 0, mmCP_MAX_CONTEXT, adev->gfx.config.max_hw_contexts - 1);
+	WREG32_SOC15(GC, 0, mmCP_DEVICE_ID, 1);
 
 	gfx_v9_0_cp_gfx_enable(adev, true);
 
···
 	u64 rb_addr, rptr_addr, wptr_gpu_addr;
 
 	/* Set the write pointer delay */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB_WPTR_DELAY), 0);
+	WREG32_SOC15(GC, 0, mmCP_RB_WPTR_DELAY, 0);
 
 	/* set the RB to use vmid 0 */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB_VMID), 0);
+	WREG32_SOC15(GC, 0, mmCP_RB_VMID, 0);
 
 	/* Set ring buffer size */
 	ring = &adev->gfx.gfx_ring[0];
···
 #ifdef __BIG_ENDIAN
 	tmp = REG_SET_FIELD(tmp, CP_RB0_CNTL, BUF_SWAP, 1);
 #endif
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB0_CNTL), tmp);
+	WREG32_SOC15(GC, 0, mmCP_RB0_CNTL, tmp);
 
 	/* Initialize the ring buffer's write pointers */
 	ring->wptr = 0;
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB0_WPTR), lower_32_bits(ring->wptr));
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB0_WPTR_HI), upper_32_bits(ring->wptr));
+	WREG32_SOC15(GC, 0, mmCP_RB0_WPTR, lower_32_bits(ring->wptr));
+	WREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI, upper_32_bits(ring->wptr));
 
 	/* set the wb address wether it's enabled or not */
 	rptr_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB0_RPTR_ADDR), lower_32_bits(rptr_addr));
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB0_RPTR_ADDR_HI), upper_32_bits(rptr_addr) & CP_RB_RPTR_ADDR_HI__RB_RPTR_ADDR_HI_MASK);
+	WREG32_SOC15(GC, 0, mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr));
+	WREG32_SOC15(GC, 0, mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & CP_RB_RPTR_ADDR_HI__RB_RPTR_ADDR_HI_MASK);
 
 	wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB_WPTR_POLL_ADDR_LO), lower_32_bits(wptr_gpu_addr));
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB_WPTR_POLL_ADDR_HI), upper_32_bits(wptr_gpu_addr));
+	WREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_ADDR_LO, lower_32_bits(wptr_gpu_addr));
+	WREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_ADDR_HI, upper_32_bits(wptr_gpu_addr));
 
 	mdelay(1);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB0_CNTL), tmp);
+	WREG32_SOC15(GC, 0, mmCP_RB0_CNTL, tmp);
 
 	rb_addr = ring->gpu_addr >> 8;
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB0_BASE), rb_addr);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB0_BASE_HI), upper_32_bits(rb_addr));
+	WREG32_SOC15(GC, 0, mmCP_RB0_BASE, rb_addr);
+	WREG32_SOC15(GC, 0, mmCP_RB0_BASE_HI, upper_32_bits(rb_addr));
 
-	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB_DOORBELL_CONTROL));
+	tmp = RREG32_SOC15(GC, 0, mmCP_RB_DOORBELL_CONTROL);
 	if (ring->use_doorbell) {
 		tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL,
 				    DOORBELL_OFFSET, ring->doorbell_index);
···
 	} else {
 		tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL, DOORBELL_EN, 0);
 	}
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB_DOORBELL_CONTROL), tmp);
+	WREG32_SOC15(GC, 0, mmCP_RB_DOORBELL_CONTROL, tmp);
 
 	tmp = REG_SET_FIELD(0, CP_RB_DOORBELL_RANGE_LOWER,
 			DOORBELL_RANGE_LOWER, ring->doorbell_index);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB_DOORBELL_RANGE_LOWER), tmp);
+	WREG32_SOC15(GC, 0, mmCP_RB_DOORBELL_RANGE_LOWER, tmp);
 
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB_DOORBELL_RANGE_UPPER),
+	WREG32_SOC15(GC, 0, mmCP_RB_DOORBELL_RANGE_UPPER,
 		       CP_RB_DOORBELL_RANGE_UPPER__DOORBELL_RANGE_UPPER_MASK);
 
 
···
 	int i;
 
 	if (enable) {
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEC_CNTL), 0);
+		WREG32_SOC15(GC, 0, mmCP_MEC_CNTL, 0);
 	} else {
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEC_CNTL),
+		WREG32_SOC15(GC, 0, mmCP_MEC_CNTL,
 			(CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK));
 		for (i = 0; i < adev->gfx.num_compute_rings; i++)
 			adev->gfx.compute_ring[i].ready = false;
···
 	tmp = 0;
 	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, VMID, 0);
 	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, CACHE_POLICY, 0);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_CPC_IC_BASE_CNTL), tmp);
+	WREG32_SOC15(GC, 0, mmCP_CPC_IC_BASE_CNTL, tmp);
 
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_CPC_IC_BASE_LO),
+	WREG32_SOC15(GC, 0, mmCP_CPC_IC_BASE_LO,
 		adev->gfx.mec.mec_fw_gpu_addr & 0xFFFFF000);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_CPC_IC_BASE_HI),
+	WREG32_SOC15(GC, 0, mmCP_CPC_IC_BASE_HI,
 		upper_32_bits(adev->gfx.mec.mec_fw_gpu_addr));
  
 	/* MEC1 */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEC_ME1_UCODE_ADDR),
+	WREG32_SOC15(GC, 0, mmCP_MEC_ME1_UCODE_ADDR,
 			 mec_hdr->jt_offset);
 	for (i = 0; i < mec_hdr->jt_size; i++)
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEC_ME1_UCODE_DATA),
+		WREG32_SOC15(GC, 0, mmCP_MEC_ME1_UCODE_DATA,
 			le32_to_cpup(fw_data + mec_hdr->jt_offset + i));
 
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEC_ME1_UCODE_ADDR),
+	WREG32_SOC15(GC, 0, mmCP_MEC_ME1_UCODE_ADDR,
 			adev->gfx.mec_fw_version);
 	/* Todo : Loading MEC2 firmware is only necessary if MEC2 should run different microcode than MEC1. */
 
···
 		struct amdgpu_ring *ring = &adev->gfx.compute_ring[i];
 
 		if (ring->mqd_obj) {
-			r = amdgpu_bo_reserve(ring->mqd_obj, false);
+			r = amdgpu_bo_reserve(ring->mqd_obj, true);
 			if (unlikely(r != 0))
 				dev_warn(adev->dev, "(%d) reserve MQD bo failed\n", r);
 
···
 	struct amdgpu_device *adev = ring->adev;
 
 	/* tell RLC which is KIQ queue */
-	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CP_SCHEDULERS));
+	tmp = RREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS);
 	tmp &= 0xffffff00;
 	tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CP_SCHEDULERS), tmp);
+	WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS, tmp);
 	tmp |= 0x80;
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CP_SCHEDULERS), tmp);
+	WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS, tmp);
 }
 
 static void gfx_v9_0_kiq_enable(struct amdgpu_ring *ring)
···
 	mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr);
 
 	/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
-	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_EOP_CONTROL));
+	tmp = RREG32_SOC15(GC, 0, mmCP_HQD_EOP_CONTROL);
 	tmp = REG_SET_FIELD(tmp, CP_HQD_EOP_CONTROL, EOP_SIZE,
 			(order_base_2(MEC_HPD_SIZE / 4) - 1));
 
 	mqd->cp_hqd_eop_control = tmp;
 
 	/* enable doorbell? */
-	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL));
+	tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL);
 
 	if (ring->use_doorbell) {
 		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
···
 	mqd->cp_mqd_base_addr_hi = upper_32_bits(ring->mqd_gpu_addr);
 
 	/* set MQD vmid to 0 */
-	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MQD_CONTROL));
+	tmp = RREG32_SOC15(GC, 0, mmCP_MQD_CONTROL);
 	tmp = REG_SET_FIELD(tmp, CP_MQD_CONTROL, VMID, 0);
 	mqd->cp_mqd_control = tmp;
 
···
 	mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr);
 
 	/* set up the HQD, this is similar to CP_RB0_CNTL */
-	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_CONTROL));
+	tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_CONTROL);
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE,
 			    (order_base_2(ring->ring_size / 4) - 1));
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE,
···
 	tmp = 0;
 	/* enable the doorbell if requested */
 	if (ring->use_doorbell) {
-		tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL));
+		tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL);
 		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
 				DOORBELL_OFFSET, ring->doorbell_index);
 
···
 
 	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
 	ring->wptr = 0;
-	mqd->cp_hqd_pq_rptr = RREG32(mmCP_HQD_PQ_RPTR);
+	mqd->cp_hqd_pq_rptr = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_RPTR);
 
 	/* set the vmid for the queue */
 	mqd->cp_hqd_vmid = 0;
 
-	tmp = RREG32(mmCP_HQD_PERSISTENT_STATE);
+	tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PERSISTENT_STATE);
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PERSISTENT_STATE, PRELOAD_SIZE, 0x53);
 	mqd->cp_hqd_persistent_state = tmp;
+
+	/* set MIN_IB_AVAIL_SIZE */
+	tmp = RREG32_SOC15(GC, 0, mmCP_HQD_IB_CONTROL);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_IB_CONTROL, MIN_IB_AVAIL_SIZE, 3);
+	mqd->cp_hqd_ib_control = tmp;
 
 	/* activate the queue */
 	mqd->cp_hqd_active = 1;
···
 	/* disable wptr polling */
 	WREG32_FIELD15(GC, 0, CP_PQ_WPTR_POLL_CNTL, EN, 0);
 
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_EOP_BASE_ADDR),
+	WREG32_SOC15(GC, 0, mmCP_HQD_EOP_BASE_ADDR,
 	       mqd->cp_hqd_eop_base_addr_lo);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_EOP_BASE_ADDR_HI),
+	WREG32_SOC15(GC, 0, mmCP_HQD_EOP_BASE_ADDR_HI,
 	       mqd->cp_hqd_eop_base_addr_hi);
 
 	/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_EOP_CONTROL),
+	WREG32_SOC15(GC, 0, mmCP_HQD_EOP_CONTROL,
 	       mqd->cp_hqd_eop_control);
 
 	/* enable doorbell? */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL,
 	       mqd->cp_hqd_pq_doorbell_control);
 
 	/* disable the queue if it's active */
-	if (RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE)) & 1) {
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_DEQUEUE_REQUEST), 1);
+	if (RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1) {
+		WREG32_SOC15(GC, 0, mmCP_HQD_DEQUEUE_REQUEST, 1);
 		for (j = 0; j < adev->usec_timeout; j++) {
-			if (!(RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE)) & 1))
+			if (!(RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1))
 				break;
 			udelay(1);
 		}
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_DEQUEUE_REQUEST),
+		WREG32_SOC15(GC, 0, mmCP_HQD_DEQUEUE_REQUEST,
 		       mqd->cp_hqd_dequeue_request);
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_RPTR),
+		WREG32_SOC15(GC, 0, mmCP_HQD_PQ_RPTR,
 		       mqd->cp_hqd_pq_rptr);
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_LO),
+		WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_LO,
 		       mqd->cp_hqd_pq_wptr_lo);
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI),
+		WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_HI,
 		       mqd->cp_hqd_pq_wptr_hi);
 	}
 
 	/* set the pointer to the MQD */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR),
+	WREG32_SOC15(GC, 0, mmCP_MQD_BASE_ADDR,
 	       mqd->cp_mqd_base_addr_lo);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR_HI),
+	WREG32_SOC15(GC, 0, mmCP_MQD_BASE_ADDR_HI,
 	       mqd->cp_mqd_base_addr_hi);
 
 	/* set MQD vmid to 0 */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MQD_CONTROL),
+	WREG32_SOC15(GC, 0, mmCP_MQD_CONTROL,
 	       mqd->cp_mqd_control);
 
 	/* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_BASE),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_BASE,
 	       mqd->cp_hqd_pq_base_lo);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_BASE_HI),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_BASE_HI,
 	       mqd->cp_hqd_pq_base_hi);
 
 	/* set up the HQD, this is similar to CP_RB0_CNTL */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_CONTROL),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_CONTROL,
 	       mqd->cp_hqd_pq_control);
 
 	/* set the wb address whether it's enabled or not */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_RPTR_REPORT_ADDR),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_RPTR_REPORT_ADDR,
 				mqd->cp_hqd_pq_rptr_report_addr_lo);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_RPTR_REPORT_ADDR_HI),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_RPTR_REPORT_ADDR_HI,
 				mqd->cp_hqd_pq_rptr_report_addr_hi);
 
 	/* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR,
 	       mqd->cp_hqd_pq_wptr_poll_addr_lo);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI,
 	       mqd->cp_hqd_pq_wptr_poll_addr_hi);
 
 	/* enable the doorbell if requested */
 	if (ring->use_doorbell) {
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEC_DOORBELL_RANGE_LOWER),
+		WREG32_SOC15(GC, 0, mmCP_MEC_DOORBELL_RANGE_LOWER,
 					(AMDGPU_DOORBELL64_KIQ *2) << 2);
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEC_DOORBELL_RANGE_UPPER),
+		WREG32_SOC15(GC, 0, mmCP_MEC_DOORBELL_RANGE_UPPER,
 					(AMDGPU_DOORBELL64_USERQUEUE_END * 2) << 2);
 	}
 
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL,
 	       mqd->cp_hqd_pq_doorbell_control);
 
 	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_LO),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_LO,
 	       mqd->cp_hqd_pq_wptr_lo);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_HI,
 	       mqd->cp_hqd_pq_wptr_hi);
 
 	/* set the vmid for the queue */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_VMID), mqd->cp_hqd_vmid);
+	WREG32_SOC15(GC, 0, mmCP_HQD_VMID, mqd->cp_hqd_vmid);
 
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PERSISTENT_STATE),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PERSISTENT_STATE,
 	       mqd->cp_hqd_persistent_state);
 
 	/* activate the queue */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE),
+	WREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE,
 	       mqd->cp_hqd_active);
 
 	if (ring->use_doorbell)
···
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	if (REG_GET_FIELD(RREG32(SOC15_REG_OFFSET(GC, 0, mmGRBM_STATUS)), 
+	if (REG_GET_FIELD(RREG32_SOC15(GC, 0, mmGRBM_STATUS),
 				GRBM_STATUS, GUI_ACTIVE))
 		return false;
 	else
···
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		/* read MC_STATUS */
-		tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmGRBM_STATUS)) & 
+		tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS) &
 			GRBM_STATUS__GUI_ACTIVE_MASK;
 
 		if (!REG_GET_FIELD(tmp, GRBM_STATUS, GUI_ACTIVE))
···
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	/* GRBM_STATUS */
-	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmGRBM_STATUS));
+	tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS);
 	if (tmp & (GRBM_STATUS__PA_BUSY_MASK | GRBM_STATUS__SC_BUSY_MASK |
 		   GRBM_STATUS__BCI_BUSY_MASK | GRBM_STATUS__SX_BUSY_MASK |
 		   GRBM_STATUS__TA_BUSY_MASK | GRBM_STATUS__VGT_BUSY_MASK |
···
 	}
 
 	/* GRBM_STATUS2 */
-	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmGRBM_STATUS2));
+	tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS2);
 	if (REG_GET_FIELD(tmp, GRBM_STATUS2, RLC_BUSY))
 		grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset,
 						GRBM_SOFT_RESET, SOFT_RESET_RLC, 1);
···
 		gfx_v9_0_cp_compute_enable(adev, false);
 
 		if (grbm_soft_reset) {
-			tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmGRBM_SOFT_RESET));
+			tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
 			tmp |= grbm_soft_reset;
 			dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp);
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmGRBM_SOFT_RESET), tmp);
-			tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmGRBM_SOFT_RESET));
+			WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
+			tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
 
 			udelay(50);
 
 			tmp &= ~grbm_soft_reset;
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmGRBM_SOFT_RESET), tmp);
-			tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmGRBM_SOFT_RESET));
+			WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
+			tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
 		}
 
 		/* Wait a little for things to settle down */
···
 	uint64_t clock;
 
 	mutex_lock(&adev->gfx.gpu_clock_mutex);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CAPTURE_GPU_CLOCK_COUNT), 1);
-	clock = (uint64_t)RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPU_CLOCK_COUNT_LSB)) |
-		((uint64_t)RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPU_CLOCK_COUNT_MSB)) << 32ULL);
+	WREG32_SOC15(GC, 0, mmRLC_CAPTURE_GPU_CLOCK_COUNT, 1);
+	clock = (uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_LSB) |
+		((uint64_t)RREG32_SOC15(GC, 0, mmRLC_GPU_CLOCK_COUNT_MSB) << 32ULL);
 	mutex_unlock(&adev->gfx.gpu_clock_mutex);
 	return clock;
 }
···
 		return;
 
 	/* if RLC is not enabled, do nothing */
-	rlc_setting = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CNTL));
+	rlc_setting = RREG32_SOC15(GC, 0, mmRLC_CNTL);
 	if (!(rlc_setting & RLC_CNTL__RLC_ENABLE_F32_MASK))
 		return;
 
···
 	     AMD_CG_SUPPORT_GFX_3D_CGCG)) {
 		data = RLC_SAFE_MODE__CMD_MASK;
 		data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT);
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SAFE_MODE), data);
+		WREG32_SOC15(GC, 0, mmRLC_SAFE_MODE, data);
 
 		/* wait for RLC_SAFE_MODE */
 		for (i = 0; i < adev->usec_timeout; i++) {
···
 		return;
 
 	/* if RLC is not enabled, do nothing */
-	rlc_setting = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CNTL));
+	rlc_setting = RREG32_SOC15(GC, 0, mmRLC_CNTL);
 	if (!(rlc_setting & RLC_CNTL__RLC_ENABLE_F32_MASK))
 		return;
 
···
 		 * mode.
 		 */
 		data = RLC_SAFE_MODE__CMD_MASK;
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SAFE_MODE), data);
+		WREG32_SOC15(GC, 0, mmRLC_SAFE_MODE, data);
 		adev->gfx.rlc.in_safe_mode = false;
 	}
 }
···
 	/* It is disabled by HW by default */
 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) {
 		/* 1 - RLC_CGTT_MGCG_OVERRIDE */
-		def = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE));
+		def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE);
 		data &= ~(RLC_CGTT_MGCG_OVERRIDE__CPF_CGTT_SCLK_OVERRIDE_MASK |
 			  RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK |
 			  RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK |
···
 		data |= RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK;
 
 		if (def != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE), data);
+			WREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE, data);
 
 		/* MGLS is a global flag to control all MGLS in GFX */
 		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGLS) {
 			/* 2 - RLC memory Light sleep */
 			if (adev->cg_flags & AMD_CG_SUPPORT_GFX_RLC_LS) {
-				def = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_MEM_SLP_CNTL));
+				def = data = RREG32_SOC15(GC, 0, mmRLC_MEM_SLP_CNTL);
 				data |= RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK;
 				if (def != data)
-					WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_MEM_SLP_CNTL), data);
+					WREG32_SOC15(GC, 0, mmRLC_MEM_SLP_CNTL, data);
 			}
 			/* 3 - CP memory Light sleep */
 			if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CP_LS) {
-				def = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEM_SLP_CNTL));
+				def = data = RREG32_SOC15(GC, 0, mmCP_MEM_SLP_CNTL);
 				data |= CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK;
 				if (def != data)
-					WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEM_SLP_CNTL), data);
+					WREG32_SOC15(GC, 0, mmCP_MEM_SLP_CNTL, data);
 			}
 		}
 	} else {
 		/* 1 - MGCG_OVERRIDE */
-		def = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE));
+		def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE);
 		data |= (RLC_CGTT_MGCG_OVERRIDE__CPF_CGTT_SCLK_OVERRIDE_MASK |
 			 RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK |
 			 RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK |
 			 RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK |
 			 RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGLS_OVERRIDE_MASK);
 		if (def != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE), data);
+			WREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE, data);
 
 		/* 2 - disable MGLS in RLC */
-		data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_MEM_SLP_CNTL));
+		data = RREG32_SOC15(GC, 0, mmRLC_MEM_SLP_CNTL);
 		if (data & RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK) {
 			data &= ~RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK;
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_MEM_SLP_CNTL), data);
+			WREG32_SOC15(GC, 0, mmRLC_MEM_SLP_CNTL, data);
 		}
 
 		/* 3 - disable MGLS in CP */
-		data = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEM_SLP_CNTL));
+		data = RREG32_SOC15(GC, 0, mmCP_MEM_SLP_CNTL);
 		if (data & CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK) {
 			data &= ~CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK;
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEM_SLP_CNTL), data);
+			WREG32_SOC15(GC, 0, mmCP_MEM_SLP_CNTL, data);
 		}
 	}
 }
···
 	/* Enable 3D CGCG/CGLS */
 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGCG)) {
 		/* write cmd to clear cgcg/cgls ov */
-		def = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE));
+		def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE);
 		/* unset CGCG override */
 		data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_GFX3D_CG_OVERRIDE_MASK;
 		/* update CGCG and CGLS override bits */
 		if (def != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE), data);
+			WREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE, data);
 		/* enable 3Dcgcg FSM(0x0020003f) */
-		def = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D));
+		def = RREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D);
 		data = (0x2000 << RLC_CGCG_CGLS_CTRL_3D__CGCG_GFX_IDLE_THRESHOLD__SHIFT) |
 			RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK;
 		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGLS)
 			data |= (0x000F << RLC_CGCG_CGLS_CTRL_3D__CGLS_REP_COMPANSAT_DELAY__SHIFT) |
 				RLC_CGCG_CGLS_CTRL_3D__CGLS_EN_MASK;
 		if (def != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D), data);
+			WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D, data);
 
 		/* set IDLE_POLL_COUNT(0x00900100) */
-		def = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB_WPTR_POLL_CNTL));
+		def = RREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_CNTL);
 		data = (0x0100 << CP_RB_WPTR_POLL_CNTL__POLL_FREQUENCY__SHIFT) |
 			(0x0090 << CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT__SHIFT);
 		if (def != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB_WPTR_POLL_CNTL), data);
+			WREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_CNTL, data);
 	} else {
 		/* Disable CGCG/CGLS */
-		def = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D));
+		def = data = RREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D);
 		/* disable cgcg, cgls should be disabled */
 		data &= ~(RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK |
 			  RLC_CGCG_CGLS_CTRL_3D__CGLS_EN_MASK);
 		/* disable cgcg and cgls in FSM */
 		if (def != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D), data);
+			WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D, data);
 	}
 
 	adev->gfx.rlc.funcs->exit_safe_mode(adev);
···
 	adev->gfx.rlc.funcs->enter_safe_mode(adev);
 
 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) {
-		def = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE));
+		def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE);
 		/* unset CGCG override */
 		data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_CGCG_OVERRIDE_MASK;
 		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS)
···
 			data |= RLC_CGTT_MGCG_OVERRIDE__GFXIP_CGLS_OVERRIDE_MASK;
 		/* update CGCG and CGLS override bits */
 		if (def != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE), data);
+			WREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE, data);
 
 		/* enable cgcg FSM(0x0020003F) */
-		def = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGCG_CGLS_CTRL));
+		def = RREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL);
 		data = (0x2000 << RLC_CGCG_CGLS_CTRL__CGCG_GFX_IDLE_THRESHOLD__SHIFT) |
 			RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK;
 		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS)
 			data |= (0x000F << RLC_CGCG_CGLS_CTRL__CGLS_REP_COMPANSAT_DELAY__SHIFT) |
 				RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK;
 		if (def != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGCG_CGLS_CTRL), data);
+			WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL, data);
 
 		/* set IDLE_POLL_COUNT(0x00900100) */
-		def = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB_WPTR_POLL_CNTL));
+		def = RREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_CNTL);
 		data = (0x0100 << CP_RB_WPTR_POLL_CNTL__POLL_FREQUENCY__SHIFT) |
 			(0x0090 << CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT__SHIFT);
 		if (def != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB_WPTR_POLL_CNTL), data);
+			WREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_CNTL, data);
 	} else {
-		def = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGCG_CGLS_CTRL));
+		def = data = RREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL);
 		/* reset CGCG/CGLS bits */
 		data &= ~(RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK | RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK);
 		/* disable cgcg and cgls in FSM */
 		if (def != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGCG_CGLS_CTRL), data);
+			WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL, data);
 	}
 
 	adev->gfx.rlc.funcs->exit_safe_mode(adev);
···
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
+	if (amdgpu_sriov_vf(adev))
+		return 0;
+
 	switch (adev->asic_type) {
 	case CHIP_VEGA10:
 		gfx_v9_0_update_gfx_clock_gating(adev,
···
 		*flags = 0;
 
 	/* AMD_CG_SUPPORT_GFX_MGCG */
-	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE));
+	data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE);
 	if (!(data & RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK))
 		*flags |= AMD_CG_SUPPORT_GFX_MGCG;
 
 	/* AMD_CG_SUPPORT_GFX_CGCG */
-	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGCG_CGLS_CTRL));
+	data = RREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL);
 	if (data & RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK)
 		*flags |= AMD_CG_SUPPORT_GFX_CGCG;
 
···
 		*flags |= AMD_CG_SUPPORT_GFX_CGLS;
 
 	/* AMD_CG_SUPPORT_GFX_RLC_LS */
-	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_MEM_SLP_CNTL));
+	data = RREG32_SOC15(GC, 0, mmRLC_MEM_SLP_CNTL);
 	if (data & RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK)
 		*flags |= AMD_CG_SUPPORT_GFX_RLC_LS | AMD_CG_SUPPORT_GFX_MGLS;
 
 	/* AMD_CG_SUPPORT_GFX_CP_LS */
-	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEM_SLP_CNTL));
+	data = RREG32_SOC15(GC, 0, mmCP_MEM_SLP_CNTL);
 	if (data & CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK)
 		*flags |= AMD_CG_SUPPORT_GFX_CP_LS | AMD_CG_SUPPORT_GFX_MGLS;
 
 	/* AMD_CG_SUPPORT_GFX_3D_CGCG */
-	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D));
+	data = RREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D);
 	if (data & RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK)
 		*flags |= AMD_CG_SUPPORT_GFX_3D_CGCG;
 
···
 	if (ring->use_doorbell) {
 		wptr = atomic64_read((atomic64_t *)&adev->wb.wb[ring->wptr_offs]);
 	} else {
-		wptr = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB0_WPTR));
-		wptr += (u64)RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB0_WPTR_HI)) << 32;
+		wptr = RREG32_SOC15(GC, 0, mmCP_RB0_WPTR);
+		wptr += (u64)RREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI) << 32;
 	}
 
 	return wptr;
···
 		atomic64_set((atomic64_t*)&adev->wb.wb[ring->wptr_offs], ring->wptr);
 		WDOORBELL64(ring->doorbell_index, ring->wptr);
 	} else {
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB0_WPTR), lower_32_bits(ring->wptr));
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_RB0_WPTR_HI), upper_32_bits(ring->wptr));
+		WREG32_SOC15(GC, 0, mmCP_RB0_WPTR, lower_32_bits(ring->wptr));
+		WREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI, upper_32_bits(ring->wptr));
 	}
 }
 
···
 static void gfx_v9_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 					unsigned vm_id, uint64_t pd_addr)
 {
+	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX);
 	uint32_t req = ring->adev->gart.gart_funcs->get_invalidate_req(vm_id);
-	unsigned eng = ring->idx;
-	unsigned i;
+	unsigned eng = ring->vm_inv_eng;
 
 	pd_addr = pd_addr | 0x1; /* valid bit */
 	/* now only use physical base address of PDE and valid */
 	BUG_ON(pd_addr & 0xFFFF00000000003EULL);
 
-	for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) {
-		struct amdgpu_vmhub *hub = &ring->adev->vmhub[i];
+	gfx_v9_0_write_data_to_reg(ring, usepfp, true,
+				   hub->ctx0_ptb_addr_lo32 + (2 * vm_id),
+				   lower_32_bits(pd_addr));
 
-		gfx_v9_0_write_data_to_reg(ring, usepfp, true,
-					   hub->ctx0_ptb_addr_lo32
-					   + (2 * vm_id),
-					   lower_32_bits(pd_addr));
+	gfx_v9_0_write_data_to_reg(ring, usepfp, true,
+				   hub->ctx0_ptb_addr_hi32 + (2 * vm_id),
+				   upper_32_bits(pd_addr));
 
-		gfx_v9_0_write_data_to_reg(ring, usepfp, true,
-					   hub->ctx0_ptb_addr_hi32
-					   + (2 * vm_id),
-					   upper_32_bits(pd_addr));
+	gfx_v9_0_write_data_to_reg(ring, usepfp, true,
+				   hub->vm_inv_eng0_req + eng, req);
 
-		gfx_v9_0_write_data_to_reg(ring, usepfp, true,
-					   hub->vm_inv_eng0_req + eng, req);
-
-		/* wait for the invalidate to complete */
-		gfx_v9_0_wait_reg_mem(ring, 0, 0, 0, hub->vm_inv_eng0_ack +
-				      eng, 0, 1 << vm_id, 1 << vm_id, 0x20);
-	}
+	/* wait for the invalidate to complete */
+	gfx_v9_0_wait_reg_mem(ring, 0, 0, 0, hub->vm_inv_eng0_ack +
+			      eng, 0, 1 << vm_id, 1 << vm_id, 0x20);
 
 	/* compute doesn't have PFP */
 	if (usepfp) {
···
 					    enum amdgpu_interrupt_state state)
 {
 	uint32_t tmp, target;
-	struct amdgpu_ring *ring = (struct amdgpu_ring *)src->data;
-
-	BUG_ON(!ring || (ring->funcs->type != AMDGPU_RING_TYPE_KIQ));
+	struct amdgpu_ring *ring = &(adev->gfx.kiq.ring);
 
 	if (ring->me == 1)
 		target = SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE0_INT_CNTL);
···
 	switch (type) {
 	case AMDGPU_CP_KIQ_IRQ_DRIVER0:
 		if (state == AMDGPU_IRQ_STATE_DISABLE) {
-			tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCPC_INT_CNTL));
+			tmp = RREG32_SOC15(GC, 0, mmCPC_INT_CNTL);
 			tmp = REG_SET_FIELD(tmp, CPC_INT_CNTL,
 						 GENERIC2_INT_ENABLE, 0);
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmCPC_INT_CNTL), tmp);
+			WREG32_SOC15(GC, 0, mmCPC_INT_CNTL, tmp);
 
 			tmp = RREG32(target);
 			tmp = REG_SET_FIELD(tmp, CP_ME2_PIPE0_INT_CNTL,
 						 GENERIC2_INT_ENABLE, 0);
 			WREG32(target, tmp);
 		} else {
-			tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCPC_INT_CNTL));
+			tmp = RREG32_SOC15(GC, 0, mmCPC_INT_CNTL);
 			tmp = REG_SET_FIELD(tmp, CPC_INT_CNTL,
 						 GENERIC2_INT_ENABLE, 1);
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmCPC_INT_CNTL), tmp);
+			WREG32_SOC15(GC, 0, mmCPC_INT_CNTL, tmp);
 
 			tmp = RREG32(target);
 			tmp = REG_SET_FIELD(tmp, CP_ME2_PIPE0_INT_CNTL,
···
 			    struct amdgpu_iv_entry *entry)
 {
 	u8 me_id, pipe_id, queue_id;
-	struct amdgpu_ring *ring = (struct amdgpu_ring *)source->data;
-
-	BUG_ON(!ring || (ring->funcs->type != AMDGPU_RING_TYPE_KIQ));
+	struct amdgpu_ring *ring = &(adev->gfx.kiq.ring);
 
 	me_id = (entry->ring_id & 0x0c) >> 2;
 	pipe_id = (entry->ring_id & 0x03) >> 0;
···
 	.align_mask = 0xff,
 	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
 	.support_64bit_ptrs = true,
+	.vmhub = AMDGPU_GFXHUB,
 	.get_rptr = gfx_v9_0_ring_get_rptr_gfx,
 	.get_wptr = gfx_v9_0_ring_get_wptr_gfx,
 	.set_wptr = gfx_v9_0_ring_set_wptr_gfx,
 	.emit_frame_size = /* totally 242 maximum if 16 IBs */
 		5 +  /* COND_EXEC */
 		7 +  /* PIPELINE_SYNC */
-		46 + /* VM_FLUSH */
+		24 + /* VM_FLUSH */
 		8 +  /* FENCE for VM_FLUSH */
 		20 + /* GDS switch */
 		4 + /* double SWITCH_BUFFER,
···
 	.align_mask = 0xff,
 	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
 	.support_64bit_ptrs = true,
+	.vmhub = AMDGPU_GFXHUB,
 	.get_rptr = gfx_v9_0_ring_get_rptr_compute,
 	.get_wptr = gfx_v9_0_ring_get_wptr_compute,
 	.set_wptr = gfx_v9_0_ring_set_wptr_compute,
···
 		7 + /* gfx_v9_0_ring_emit_hdp_flush */
 		5 + /* gfx_v9_0_ring_emit_hdp_invalidate */
 		7 + /* gfx_v9_0_ring_emit_pipeline_sync */
-		64 + /* gfx_v9_0_ring_emit_vm_flush */
+		24 + /* gfx_v9_0_ring_emit_vm_flush */
 		8 + 8 + 8, /* gfx_v9_0_ring_emit_fence x3 for user fence, vm fence */
 	.emit_ib_size =	4, /* gfx_v9_0_ring_emit_ib_compute */
 	.emit_ib = gfx_v9_0_ring_emit_ib_compute,
···
 	.align_mask = 0xff,
 	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
 	.support_64bit_ptrs = true,
+	.vmhub = AMDGPU_GFXHUB,
 	.get_rptr = gfx_v9_0_ring_get_rptr_compute,
 	.get_wptr = gfx_v9_0_ring_get_wptr_compute,
 	.set_wptr = gfx_v9_0_ring_set_wptr_compute,
···
 		7 + /* gfx_v9_0_ring_emit_hdp_flush */
 		5 + /* gfx_v9_0_ring_emit_hdp_invalidate */
 		7 + /* gfx_v9_0_ring_emit_pipeline_sync */
-		64 + /* gfx_v9_0_ring_emit_vm_flush */
+		24 + /* gfx_v9_0_ring_emit_vm_flush */
 		8 + 8 + 8, /* gfx_v9_0_ring_emit_fence_kiq x3 for user fence, vm fence */
 	.emit_ib_size =	4, /* gfx_v9_0_ring_emit_ib_compute */
 	.emit_ib = gfx_v9_0_ring_emit_ib_compute,
···
 static void gfx_v9_0_set_gds_init(struct amdgpu_device *adev)
 {
 	/* init asci gds info */
-	adev->gds.mem.total_size = RREG32(SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_SIZE));
+	adev->gds.mem.total_size = RREG32_SOC15(GC, 0, mmGDS_VMID0_SIZE);
 	adev->gds.gws.total_size = 64;
 	adev->gds.oa.total_size = 16;
 
···
 {
 	u32 data, mask;
 
-	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmCC_GC_SHADER_ARRAY_CONFIG));
-	data |= RREG32(SOC15_REG_OFFSET(GC, 0, mmGC_USER_SHADER_ARRAY_CONFIG));
+	data = RREG32_SOC15(GC, 0, mmCC_GC_SHADER_ARRAY_CONFIG);
+	data |= RREG32_SOC15(GC, 0, mmGC_USER_SHADER_ARRAY_CONFIG);
 
 	data &= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_CUS_MASK;
 	data >>= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_CUS__SHIFT;
···
 	eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr + (ring->queue * MEC_HPD_SIZE);
 	eop_gpu_addr >>= 8;
 
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_EOP_BASE_ADDR), lower_32_bits(eop_gpu_addr));
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_EOP_BASE_ADDR_HI), upper_32_bits(eop_gpu_addr));
+	WREG32_SOC15(GC, 0, mmCP_HQD_EOP_BASE_ADDR, lower_32_bits(eop_gpu_addr));
+	WREG32_SOC15(GC, 0, mmCP_HQD_EOP_BASE_ADDR_HI, upper_32_bits(eop_gpu_addr));
 	mqd->cp_hqd_eop_base_addr_lo = lower_32_bits(eop_gpu_addr);
 	mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_gpu_addr);
 
 	/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
-	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_EOP_CONTROL));
+	tmp = RREG32_SOC15(GC, 0, mmCP_HQD_EOP_CONTROL);
 	tmp = REG_SET_FIELD(tmp, CP_HQD_EOP_CONTROL, EOP_SIZE,
 				    (order_base_2(MEC_HPD_SIZE / 4) - 1));
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_EOP_CONTROL), tmp);
+	WREG32_SOC15(GC, 0, mmCP_HQD_EOP_CONTROL, tmp);
 
 	/* enable doorbell? */
-	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL));
+	tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL);
 	if (use_doorbell)
 		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1);
 	else
 		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 0);
 
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL), tmp);
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL, tmp);
 	mqd->cp_hqd_pq_doorbell_control = tmp;
 
 	/* disable the queue if it's active */
···
 	mqd->cp_hqd_pq_rptr = 0;
 	mqd->cp_hqd_pq_wptr_lo = 0;
 	mqd->cp_hqd_pq_wptr_hi = 0;
-	if (RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE)) & 1) {
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_DEQUEUE_REQUEST), 1);
+	if (RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1) {
+		WREG32_SOC15(GC, 0, mmCP_HQD_DEQUEUE_REQUEST, 1);
 		for (j = 0; j < adev->usec_timeout; j++) {
-			if (!(RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE)) & 1))
+			if (!(RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1))
 				break;
 			udelay(1);
 		}
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_DEQUEUE_REQUEST), mqd->cp_hqd_dequeue_request);
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_RPTR), mqd->cp_hqd_pq_rptr);
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_LO), mqd->cp_hqd_pq_wptr_lo);
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI), mqd->cp_hqd_pq_wptr_hi);
+		WREG32_SOC15(GC, 0, mmCP_HQD_DEQUEUE_REQUEST, mqd->cp_hqd_dequeue_request);
+		WREG32_SOC15(GC, 0, mmCP_HQD_PQ_RPTR, mqd->cp_hqd_pq_rptr);
+		WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_LO, mqd->cp_hqd_pq_wptr_lo);
+		WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_HI, mqd->cp_hqd_pq_wptr_hi);
 	}
 
 	/* set the pointer to the MQD */
 	mqd->cp_mqd_base_addr_lo = mqd_gpu_addr & 0xfffffffc;
 	mqd->cp_mqd_base_addr_hi = upper_32_bits(mqd_gpu_addr);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR), mqd->cp_mqd_base_addr_lo);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR_HI), mqd->cp_mqd_base_addr_hi);
+	WREG32_SOC15(GC, 0, mmCP_MQD_BASE_ADDR, mqd->cp_mqd_base_addr_lo);
+	WREG32_SOC15(GC, 0, mmCP_MQD_BASE_ADDR_HI, mqd->cp_mqd_base_addr_hi);
 
 	/* set MQD vmid to 0 */
-	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MQD_CONTROL));
+	tmp = RREG32_SOC15(GC, 0, mmCP_MQD_CONTROL);
 	tmp = REG_SET_FIELD(tmp, CP_MQD_CONTROL, VMID, 0);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MQD_CONTROL), tmp);
+	WREG32_SOC15(GC, 0, mmCP_MQD_CONTROL, tmp);
 	mqd->cp_mqd_control = tmp;
 
 	/* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */
 	hqd_gpu_addr = ring->gpu_addr >> 8;
 	mqd->cp_hqd_pq_base_lo = hqd_gpu_addr;
 	mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_BASE), mqd->cp_hqd_pq_base_lo);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_BASE_HI), mqd->cp_hqd_pq_base_hi);
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_BASE, mqd->cp_hqd_pq_base_lo);
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_BASE_HI, mqd->cp_hqd_pq_base_hi);
 
 	/* set up the HQD, this is similar to CP_RB0_CNTL */
-	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_CONTROL));
+	tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_CONTROL);
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE,
 		(order_base_2(ring->ring_size / 4) - 1));
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE,
···
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, ROQ_PQ_IB_FLIP, 0);
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1);
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_CONTROL), tmp);
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_CONTROL, tmp);
 	mqd->cp_hqd_pq_control = tmp;
 
 	/* set the wb address wether it's enabled or not */
···
 	mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc;
 	mqd->cp_hqd_pq_rptr_report_addr_hi =
 	upper_32_bits(wb_gpu_addr) & 0xffff;
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_RPTR_REPORT_ADDR),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_RPTR_REPORT_ADDR,
 		mqd->cp_hqd_pq_rptr_report_addr_lo);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_RPTR_REPORT_ADDR_HI),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_RPTR_REPORT_ADDR_HI,
 		mqd->cp_hqd_pq_rptr_report_addr_hi);
 
 	/* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */
 	wb_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4);
 	mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc;
 	mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff;
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR,
 		mqd->cp_hqd_pq_wptr_poll_addr_lo);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI,
 		mqd->cp_hqd_pq_wptr_poll_addr_hi);
 
 	/* enable the doorbell if requested */
 	if (use_doorbell) {
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEC_DOORBELL_RANGE_LOWER),
+		WREG32_SOC15(GC, 0, mmCP_MEC_DOORBELL_RANGE_LOWER,
 			(AMDGPU_DOORBELL64_KIQ * 2) << 2);
-		WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEC_DOORBELL_RANGE_UPPER),
+		WREG32_SOC15(GC, 0, mmCP_MEC_DOORBELL_RANGE_UPPER,
 			(AMDGPU_DOORBELL64_MEC_RING7 * 2) << 2);
-		tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL));
+		tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL);
 		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
 			DOORBELL_OFFSET, ring->doorbell_index);
 		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1);
···
 	} else {
 		mqd->cp_hqd_pq_doorbell_control = 0;
 	}
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL),
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL,
 		mqd->cp_hqd_pq_doorbell_control);
 
 	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_LO), mqd->cp_hqd_pq_wptr_lo);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI), mqd->cp_hqd_pq_wptr_hi);
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_LO, mqd->cp_hqd_pq_wptr_lo);
+	WREG32_SOC15(GC, 0, mmCP_HQD_PQ_WPTR_HI, mqd->cp_hqd_pq_wptr_hi);
 
 	/* set the vmid for the queue */
 	mqd->cp_hqd_vmid = 0;
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_VMID), mqd->cp_hqd_vmid);
+	WREG32_SOC15(GC, 0, mmCP_HQD_VMID, mqd->cp_hqd_vmid);
 
-	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PERSISTENT_STATE));
+	tmp = RREG32_SOC15(GC, 0, mmCP_HQD_PERSISTENT_STATE);
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PERSISTENT_STATE, PRELOAD_SIZE, 0x53);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PERSISTENT_STATE), tmp);
+	WREG32_SOC15(GC, 0, mmCP_HQD_PERSISTENT_STATE, tmp);
 	mqd->cp_hqd_persistent_state = tmp;
 
 	/* activate the queue */
 	mqd->cp_hqd_active = 1;
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE), mqd->cp_hqd_active);
+	WREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE, mqd->cp_hqd_active);
 
 	soc15_grbm_select(adev, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);

drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c

···
 	 * size equal to the 1024 or vram, whichever is larger.
 	 */
 	if (amdgpu_gart_size == -1)
-		adev->mc.gtt_size = max((1024ULL << 20), adev->mc.mc_vram_size);
+		adev->mc.gtt_size = max((AMDGPU_DEFAULT_GTT_SIZE_MB << 20),
+					adev->mc.mc_vram_size);
 	else
 		adev->mc.gtt_size = (uint64_t)amdgpu_gart_size << 20;
 
···
 	 * amdgpu graphics/compute will use VMIDs 1-7
 	 * amdkfd will use VMIDs 8-15
 	 */
-	adev->vm_manager.num_ids = AMDGPU_NUM_OF_VMIDS;
+	adev->vm_manager.id_mgr[0].num_ids = AMDGPU_NUM_OF_VMIDS;
 	adev->vm_manager.num_level = 1;
 	amdgpu_vm_manager_init(adev);
 

drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c

···
 	 * size equal to the 1024 or vram, whichever is larger.
 	 */
 	if (amdgpu_gart_size == -1)
-		adev->mc.gtt_size = max((1024ULL << 20), adev->mc.mc_vram_size);
+		adev->mc.gtt_size = max((AMDGPU_DEFAULT_GTT_SIZE_MB << 20),
+					adev->mc.mc_vram_size);
 	else
 		adev->mc.gtt_size = (uint64_t)amdgpu_gart_size << 20;
 
···
 	 * amdgpu graphics/compute will use VMIDs 1-7
 	 * amdkfd will use VMIDs 8-15
 	 */
-	adev->vm_manager.num_ids = AMDGPU_NUM_OF_VMIDS;
+	adev->vm_manager.id_mgr[0].num_ids = AMDGPU_NUM_OF_VMIDS;
 	adev->vm_manager.num_level = 1;
 	amdgpu_vm_manager_init(adev);
 

drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c

···
 	 * size equal to the 1024 or vram, whichever is larger.
 	 */
 	if (amdgpu_gart_size == -1)
-		adev->mc.gtt_size = max((1024ULL << 20), adev->mc.mc_vram_size);
+		adev->mc.gtt_size = max((AMDGPU_DEFAULT_GTT_SIZE_MB << 20),
+					adev->mc.mc_vram_size);
 	else
 		adev->mc.gtt_size = (uint64_t)amdgpu_gart_size << 20;
 
···
 	 * amdgpu graphics/compute will use VMIDs 1-7
 	 * amdkfd will use VMIDs 8-15
 	 */
-	adev->vm_manager.num_ids = AMDGPU_NUM_OF_VMIDS;
+	adev->vm_manager.id_mgr[0].num_ids = AMDGPU_NUM_OF_VMIDS;
 	adev->vm_manager.num_level = 1;
 	amdgpu_vm_manager_init(adev);
 

drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c

···
 static int gmc_v9_0_late_init(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	unsigned vm_inv_eng[AMDGPU_MAX_VMHUBS] = { 3, 3 };
+	unsigned i;
+
+	for(i = 0; i < adev->num_rings; ++i) {
+		struct amdgpu_ring *ring = adev->rings[i];
+		unsigned vmhub = ring->funcs->vmhub;
+
+		ring->vm_inv_eng = vm_inv_eng[vmhub]++;
+		dev_info(adev->dev, "ring %u(%s) uses VM inv eng %u on hub %u\n",
+			 ring->idx, ring->name, ring->vm_inv_eng,
+			 ring->funcs->vmhub);
+	}
+
+	/* Engine 17 is used for GART flushes */
+	for(i = 0; i < AMDGPU_MAX_VMHUBS; ++i)
+		BUG_ON(vm_inv_eng[i] > 17);
+
 	return amdgpu_irq_get(adev, &adev->mc.vm_fault, 0);
 }
 
···
 	 * size equal to the 1024 or vram, whichever is larger.
 	 */
 	if (amdgpu_gart_size == -1)
-		adev->mc.gtt_size = max((1024ULL << 20), adev->mc.mc_vram_size);
+		adev->mc.gtt_size = max((AMDGPU_DEFAULT_GTT_SIZE_MB << 20),
+					adev->mc.mc_vram_size);
 	else
 		adev->mc.gtt_size = (uint64_t)amdgpu_gart_size << 20;
 
···
 	 * amdgpu graphics/compute will use VMIDs 1-7
 	 * amdkfd will use VMIDs 8-15
 	 */
-	adev->vm_manager.num_ids = AMDGPU_NUM_OF_VMIDS;
+	adev->vm_manager.id_mgr[AMDGPU_GFXHUB].num_ids = AMDGPU_NUM_OF_VMIDS;
+	adev->vm_manager.id_mgr[AMDGPU_MMHUB].num_ids = AMDGPU_NUM_OF_VMIDS;
 
 	/* TODO: fix num_level for APU when updating vm size and block size */
 	if (adev->flags & AMD_IS_APU)

drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c

···
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
+	if (amdgpu_sriov_vf(adev))
+		return 0;
+
 	switch (adev->asic_type) {
 	case CHIP_VEGA10:
 		mmhub_v1_0_update_medium_grain_clock_gating(adev,

drivers/gpu/drm/amd/amdgpu/mmsch_v1_0.h

···
 	uint32_t reg_value;
 };
 
+static inline void mmsch_v1_0_insert_direct_wt(struct mmsch_v1_0_cmd_direct_write *direct_wt,
+					       uint32_t *init_table,
+					       uint32_t reg_offset,
+					       uint32_t value)
+{
+	direct_wt->cmd_header.reg_offset = reg_offset;
+	direct_wt->reg_value = value;
+	memcpy((void *)init_table, direct_wt, sizeof(struct mmsch_v1_0_cmd_direct_write));
+}
+
+static inline void mmsch_v1_0_insert_direct_rd_mod_wt(struct mmsch_v1_0_cmd_direct_read_modify_write *direct_rd_mod_wt,
+						      uint32_t *init_table,
+						      uint32_t reg_offset,
+						      uint32_t mask, uint32_t data)
+{
+	direct_rd_mod_wt->cmd_header.reg_offset = reg_offset;
+	direct_rd_mod_wt->mask_value = mask;
+	direct_rd_mod_wt->write_data = data;
+	memcpy((void *)init_table, direct_rd_mod_wt,
+	       sizeof(struct mmsch_v1_0_cmd_direct_read_modify_write));
+}
+
+static inline void mmsch_v1_0_insert_direct_poll(struct mmsch_v1_0_cmd_direct_polling *direct_poll,
+						 uint32_t *init_table,
+						 uint32_t reg_offset,
+						 uint32_t mask, uint32_t wait)
+{
+	direct_poll->cmd_header.reg_offset = reg_offset;
+	direct_poll->mask_value = mask;
+	direct_poll->wait_value = wait;
+	memcpy((void *)init_table, direct_poll, sizeof(struct mmsch_v1_0_cmd_direct_polling));
+}
+
+#define MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(reg, mask, data) { \
+	mmsch_v1_0_insert_direct_rd_mod_wt(&direct_rd_mod_wt, \
+					   init_table, (reg), \
+					   (mask), (data)); \
+	init_table += sizeof(struct mmsch_v1_0_cmd_direct_read_modify_write)/4; \
+	table_size += sizeof(struct mmsch_v1_0_cmd_direct_read_modify_write)/4; \
+}
+
+#define MMSCH_V1_0_INSERT_DIRECT_WT(reg, value) { \
+	mmsch_v1_0_insert_direct_wt(&direct_wt, \
+				    init_table, (reg), \
+				    (value)); \
+	init_table += sizeof(struct mmsch_v1_0_cmd_direct_write)/4; \
+	table_size += sizeof(struct mmsch_v1_0_cmd_direct_write)/4; \
+}
+
+#define MMSCH_V1_0_INSERT_DIRECT_POLL(reg, mask, wait) { \
+	mmsch_v1_0_insert_direct_poll(&direct_poll, \
+				      init_table, (reg), \
+				      (mask), (wait)); \
+	init_table += sizeof(struct mmsch_v1_0_cmd_direct_polling)/4; \
+	table_size += sizeof(struct mmsch_v1_0_cmd_direct_polling)/4; \
+}
+
 #endif

drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c

···
 	u32 reg;
 	u32 mask = REG_FIELD_MASK(MAILBOX_CONTROL, RCV_MSG_VALID);
 
-	reg = RREG32_NO_KIQ(mmMAILBOX_CONTROL);
-	if (!(reg & mask))
-		return -ENOENT;
+	/* workaround: host driver doesn't set VALID for CMPL now */
+	if (event != IDH_FLR_NOTIFICATION_CMPL) {
+		reg = RREG32_NO_KIQ(mmMAILBOX_CONTROL);
+		if (!(reg & mask))
+			return -ENOENT;
+	}
 
 	reg = RREG32_NO_KIQ(mmMAILBOX_MSGBUF_RCV_DW0);
 	if (reg != event)

drivers/gpu/drm/amd/amdgpu/psp_v3_1.c

···
 {
 	int ret;
 	uint32_t psp_gfxdrv_command_reg = 0;
-	struct amdgpu_bo *psp_sysdrv;
-	void *psp_sysdrv_virt = NULL;
-	uint64_t psp_sysdrv_mem;
 	struct amdgpu_device *adev = psp->adev;
-	uint32_t size, sol_reg;
+	uint32_t sol_reg;
 
 	/* Check sOS sign of life register to confirm sys driver and sOS
 	 * are already been loaded.
···
 	if (ret)
 		return ret;
 
-	/*
-	 * Create a 1 meg GART memory to store the psp sys driver
-	 * binary with a 1 meg aligned address
-	 */
-	size = (psp->sys_bin_size + (PSP_BOOTLOADER_1_MEG_ALIGNMENT - 1)) &
-		(~(PSP_BOOTLOADER_1_MEG_ALIGNMENT - 1));
-
-	ret = amdgpu_bo_create_kernel(adev, size, PSP_BOOTLOADER_1_MEG_ALIGNMENT,
-				      AMDGPU_GEM_DOMAIN_GTT,
-				      &psp_sysdrv,
-				      &psp_sysdrv_mem,
-				      &psp_sysdrv_virt);
-	if (ret)
-		return ret;
+	memset(psp->fw_pri_buf, 0, PSP_1_MEG);
 
 	/* Copy PSP System Driver binary to memory */
-	memcpy(psp_sysdrv_virt, psp->sys_start_addr, psp->sys_bin_size);
+	memcpy(psp->fw_pri_buf, psp->sys_start_addr, psp->sys_bin_size);
 
 	/* Provide the sys driver to bootrom */
 	WREG32(SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_36),
-	       (uint32_t)(psp_sysdrv_mem >> 20));
+	       (uint32_t)(psp->fw_pri_mc_addr >> 20));
 	psp_gfxdrv_command_reg = 1 << 16;
 	WREG32(SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35),
 	       psp_gfxdrv_command_reg);
···
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35),
 			   0x80000000, 0x80000000, false);
 
-	amdgpu_bo_free_kernel(&psp_sysdrv, &psp_sysdrv_mem, &psp_sysdrv_virt);
-
 	return ret;
 }
 
···
 {
 	int ret;
 	unsigned int psp_gfxdrv_command_reg = 0;
-	struct amdgpu_bo *psp_sos;
-	void *psp_sos_virt = NULL;
-	uint64_t psp_sos_mem;
 	struct amdgpu_device *adev = psp->adev;
-	uint32_t size, sol_reg;
+	uint32_t sol_reg;
 
 	/* Check sOS sign of life register to confirm sys driver and sOS
 	 * are already been loaded.
···
 	if (ret)
 		return ret;
 
-	size = (psp->sos_bin_size + (PSP_BOOTLOADER_1_MEG_ALIGNMENT - 1)) &
-		(~((uint64_t)PSP_BOOTLOADER_1_MEG_ALIGNMENT - 1));
-
-	ret = amdgpu_bo_create_kernel(adev, size, PSP_BOOTLOADER_1_MEG_ALIGNMENT,
-				      AMDGPU_GEM_DOMAIN_GTT,
-				      &psp_sos,
-				      &psp_sos_mem,
-				      &psp_sos_virt);
-	if (ret)
-		return ret;
+	memset(psp->fw_pri_buf, 0, PSP_1_MEG);
 
 	/* Copy Secure OS binary to PSP memory */
-	memcpy(psp_sos_virt, psp->sos_start_addr, psp->sos_bin_size);
+	memcpy(psp->fw_pri_buf, psp->sos_start_addr, psp->sos_bin_size);
 
 	/* Provide the PSP secure OS to bootrom */
 	WREG32(SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_36),
-	       (uint32_t)(psp_sos_mem >> 20));
+	       (uint32_t)(psp->fw_pri_mc_addr >> 20));
 	psp_gfxdrv_command_reg = 2 << 16;
 	WREG32(SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35),
 	       psp_gfxdrv_command_reg);
···
 			   RREG32(SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_81)),
 			   0, true);
 #endif
-
-	amdgpu_bo_free_kernel(&psp_sos, &psp_sos_mem, &psp_sos_virt);
 
 	return ret;
 }
···
 int psp_v3_1_ring_init(struct psp_context *psp, enum psp_ring_type ring_type)
 {
 	int ret = 0;
-	unsigned int psp_ring_reg = 0;
 	struct psp_ring *ring;
 	struct amdgpu_device *adev = psp->adev;
 
···
 		ring->ring_size = 0;
 		return ret;
 	}
+
+	return 0;
+}
+
+int psp_v3_1_ring_create(struct psp_context *psp, enum psp_ring_type ring_type)
+{
+	int ret = 0;
+	unsigned int psp_ring_reg = 0;
+	struct psp_ring *ring = &psp->km_ring;
+	struct amdgpu_device *adev = psp->adev;
 
 	/* Write low address of the ring to C2PMSG_69 */
 	psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr);
···
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
 			   0x80000000, 0x8000FFFF, false);
 
+	return ret;
+}
+
+int psp_v3_1_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type)
+{
+	int ret = 0;
+	struct psp_ring *ring;
+	unsigned int psp_ring_reg = 0;
+	struct amdgpu_device *adev = psp->adev;
+
+	ring = &psp->km_ring;
+
+	/* Write the ring destroy command to C2PMSG_64 */
+	psp_ring_reg = 3 << 16;
+	WREG32(SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64), psp_ring_reg);
+
+	/* there might be handshake issue with hardware which needs delay */
+	mdelay(20);
+
+	/* Wait for response flag (bit 31) in C2PMSG_64 */
+	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
+			   0x80000000, 0x80000000, false);
+
+	if (ring->ring_mem)
+		amdgpu_bo_free_kernel(&adev->firmware.rbuf,
+				      &ring->ring_mem_mc_addr,
+				      (void **)&ring->ring_mem);
 	return ret;
 }
 

drivers/gpu/drm/amd/amdgpu/psp_v3_1.h

···
 				 struct psp_gfx_cmd_resp *cmd);
 extern int psp_v3_1_ring_init(struct psp_context *psp,
 			      enum psp_ring_type ring_type);
+extern int psp_v3_1_ring_create(struct psp_context *psp,
+				enum psp_ring_type ring_type);
+extern int psp_v3_1_ring_destroy(struct psp_context *psp,
+				enum psp_ring_type ring_type);
 extern int psp_v3_1_cmd_submit(struct psp_context *psp,
 			       struct amdgpu_firmware_info *ucode,
 			       uint64_t cmd_buf_mc_addr, uint64_t fence_mc_addr,

drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c

···
 static void sdma_v4_0_set_vm_pte_funcs(struct amdgpu_device *adev);
 static void sdma_v4_0_set_irq_funcs(struct amdgpu_device *adev);
 
-static const u32 golden_settings_sdma_4[] =
-{
+static const u32 golden_settings_sdma_4[] = {
 	SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CHICKEN_BITS), 0xfe931f07, 0x02831f07,
 	SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CLK_CTRL), 0xff000ff0, 0x3f000100,
 	SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GFX_IB_CNTL), 0x800f0100, 0x00000100,
···
 	SOC15_REG_OFFSET(SDMA1, 0, mmSDMA1_UTCL1_PAGE), 0x000003ff, 0x000003c0
 };
 
-static const u32 golden_settings_sdma_vg10[] =
-{
+static const u32 golden_settings_sdma_vg10[] = {
 	SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GB_ADDR_CONFIG), 0x0018773f, 0x00104002,
 	SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GB_ADDR_CONFIG_READ), 0x0018773f, 0x00104002,
 	SOC15_REG_OFFSET(SDMA1, 0, mmSDMA1_GB_ADDR_CONFIG), 0x0018773f, 0x00104002,
···
 static u32 sdma_v4_0_get_reg_offset(u32 instance, u32 internal_offset)
 {
 	u32 base = 0;
+
 	switch (instance) {
-		case 0:
-			base = SDMA0_BASE.instance[0].segment[0];
-			break;
-		case 1:
-			base = SDMA1_BASE.instance[0].segment[0];
-			break;
-		default:
-			BUG();
-			break;
+	case 0:
+		base = SDMA0_BASE.instance[0].segment[0];
+		break;
+	case 1:
+		base = SDMA1_BASE.instance[0].segment[0];
+		break;
+	default:
+		BUG();
+		break;
 	}
 
 	return base + internal_offset;
···
 	case CHIP_VEGA10:
 		chip_name = "vega10";
 		break;
-	default: BUG();
+	default:
+		BUG();
 	}
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
···
 		if (adev->sdma.instance[i].feature_version >= 20)
 			adev->sdma.instance[i].burst_nop = true;
 		DRM_DEBUG("psp_load == '%s'\n",
-				adev->firmware.load_type == AMDGPU_FW_LOAD_PSP? "true": "false");
+				adev->firmware.load_type == AMDGPU_FW_LOAD_PSP ? "true" : "false");
 
 		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 			info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SDMA0 + i];
···
 	}
 out:
 	if (err) {
-		printk(KERN_ERR
-		       "sdma_v4_0: Failed to load firmware \"%s\"\n",
-		       fw_name);
+		DRM_ERROR("sdma_v4_0: Failed to load firmware \"%s\"\n", fw_name);
 		for (i = 0; i < adev->sdma.num_instances; i++) {
 			release_firmware(adev->sdma.instance[i].fw);
 			adev->sdma.instance[i].fw = NULL;
···
  */
 static uint64_t sdma_v4_0_ring_get_rptr(struct amdgpu_ring *ring)
 {
-	u64* rptr;
+	u64 *rptr;
 
 	/* XXX check if swapping is necessary on BE */
-	rptr =((u64*)&ring->adev->wb.wb[ring->rptr_offs]);
+	rptr = ((u64 *)&ring->adev->wb.wb[ring->rptr_offs]);
 
 	DRM_DEBUG("rptr before shift == 0x%016llx\n", *rptr);
 	return ((*rptr) >> 2);
···
 static uint64_t sdma_v4_0_ring_get_wptr(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	u64* wptr = NULL;
-	uint64_t local_wptr=0;
+	u64 *wptr = NULL;
+	uint64_t local_wptr = 0;
 
 	if (ring->use_doorbell) {
 		/* XXX check if swapping is necessary on BE */
-		wptr = ((u64*)&adev->wb.wb[ring->wptr_offs]);
+		wptr = ((u64 *)&adev->wb.wb[ring->wptr_offs]);
 		DRM_DEBUG("wptr/doorbell before shift == 0x%016llx\n", *wptr);
 		*wptr = (*wptr) >> 2;
 		DRM_DEBUG("wptr/doorbell after shift == 0x%016llx\n", *wptr);
 	} else {
 		u32 lowbit, highbit;
 		int me = (ring == &adev->sdma.instance[0].ring) ? 0 : 1;
-		wptr=&local_wptr;
+
+		wptr = &local_wptr;
 		lowbit = RREG32(sdma_v4_0_get_reg_offset(me, mmSDMA0_GFX_RB_WPTR)) >> 2;
 		highbit = RREG32(sdma_v4_0_get_reg_offset(me, mmSDMA0_GFX_RB_WPTR_HI)) >> 2;
 
···
 		WDOORBELL64(ring->doorbell_index, ring->wptr << 2);
 	} else {
 		int me = (ring == &ring->adev->sdma.instance[0].ring) ? 0 : 1;
+
 		DRM_DEBUG("Not using doorbell -- "
 				"mmSDMA%i_GFX_RB_WPTR == 0x%08x "
-				"mmSDMA%i_GFX_RB_WPTR_HI == 0x%08x \n",
-				me,
+				"mmSDMA%i_GFX_RB_WPTR_HI == 0x%08x\n",
 				me,
 				lower_32_bits(ring->wptr << 2),
+				me,
 				upper_32_bits(ring->wptr << 2));
 		WREG32(sdma_v4_0_get_reg_offset(me, mmSDMA0_GFX_RB_WPTR), lower_32_bits(ring->wptr << 2));
 		WREG32(sdma_v4_0_get_reg_offset(me, mmSDMA0_GFX_RB_WPTR_HI), upper_32_bits(ring->wptr << 2));
···
  * Schedule an IB in the DMA ring (VEGA10).
  */
 static void sdma_v4_0_ring_emit_ib(struct amdgpu_ring *ring,
-                                   struct amdgpu_ib *ib,
-                                   unsigned vm_id, bool ctx_switch)
+					struct amdgpu_ib *ib,
+					unsigned vm_id, bool ctx_switch)
 {
-        u32 vmid = vm_id & 0xf;
+	u32 vmid = vm_id & 0xf;
 
-        /* IB packet must end on a 8 DW boundary */
-        sdma_v4_0_ring_insert_nop(ring, (10 - (lower_32_bits(ring->wptr) & 7)) % 8);
+	/* IB packet must end on a 8 DW boundary */
+	sdma_v4_0_ring_insert_nop(ring, (10 - (lower_32_bits(ring->wptr) & 7)) % 8);
 
-        amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_INDIRECT) |
-                          SDMA_PKT_INDIRECT_HEADER_VMID(vmid));
-        /* base must be 32 byte aligned */
-        amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr) & 0xffffffe0);
-        amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
-        amdgpu_ring_write(ring, ib->length_dw);
-        amdgpu_ring_write(ring, 0);
-        amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_INDIRECT) |
+			  SDMA_PKT_INDIRECT_HEADER_VMID(vmid));
+	/* base must be 32 byte aligned */
+	amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr) & 0xffffffe0);
+	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
+	amdgpu_ring_write(ring, ib->length_dw);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, 0);
 
 }
 
···
 	u32 doorbell;
 	u32 doorbell_offset;
 	u32 temp;
-	int i,r;
+	int i, r;
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		ring = &adev->sdma.instance[i].ring;
···
 		doorbell = RREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_GFX_DOORBELL));
 		doorbell_offset = RREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_GFX_DOORBELL_OFFSET));
 
-		if (ring->use_doorbell){
+		if (ring->use_doorbell) {
 			doorbell = REG_SET_FIELD(doorbell, SDMA0_GFX_DOORBELL, ENABLE, 1);
 			doorbell_offset = REG_SET_FIELD(doorbell_offset, SDMA0_GFX_DOORBELL_OFFSET,
 					OFFSET, ring->doorbell_index);
···
 
 
 		for (j = 0; j < fw_size; j++)
-		{
 			WREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_UCODE_DATA), le32_to_cpup(fw_data++));
-		}
 
 		WREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_UCODE_ADDR), adev->sdma.instance[i].fw_version);
 	}
···
 	if (r)
 		return r;
 	r = sdma_v4_0_rlc_resume(adev);
-	if (r)
-		return r;
 
-	return 0;
+	return r;
 }
 
 /**
···
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		tmp = le32_to_cpu(adev->wb.wb[index]);
-		if (tmp == 0xDEADBEEF) {
+		if (tmp == 0xDEADBEEF)
 			break;
-		}
 		DRM_UDELAY(1);
 	}
 
···
 	if (r)
 		goto err1;
 
-        r = dma_fence_wait_timeout(f, false, timeout);
-        if (r == 0) {
-                DRM_ERROR("amdgpu: IB test timed out\n");
-                r = -ETIMEDOUT;
-                goto err1;
-        } else if (r < 0) {
-                DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
-                goto err1;
-        }
-        tmp = le32_to_cpu(adev->wb.wb[index]);
-        if (tmp == 0xDEADBEEF) {
-                DRM_INFO("ib test on ring %d succeeded\n", ring->idx);
-                r = 0;
-        } else {
-                DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp);
-                r = -EINVAL;
-        }
+	r = dma_fence_wait_timeout(f, false, timeout);
+	if (r == 0) {
+		DRM_ERROR("amdgpu: IB test timed out\n");
+		r = -ETIMEDOUT;
+		goto err1;
+	} else if (r < 0) {
+		DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
+		goto err1;
+	}
+	tmp = le32_to_cpu(adev->wb.wb[index]);
+	if (tmp == 0xDEADBEEF) {
+		DRM_INFO("ib test on ring %d succeeded\n", ring->idx);
+		r = 0;
+	} else {
+		DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp);
+		r = -EINVAL;
+	}
 err1:
-        amdgpu_ib_free(adev, &ib, NULL);
-        dma_fence_put(f);
+	amdgpu_ib_free(adev, &ib, NULL);
+	dma_fence_put(f);
 err0:
-        amdgpu_wb_free(adev, index);
-        return r;
+	amdgpu_wb_free(adev, index);
+	return r;
 }
 
 
···
 static void sdma_v4_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 					 unsigned vm_id, uint64_t pd_addr)
 {
+	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = ring->adev->gart.gart_funcs->get_invalidate_req(vm_id);
-	unsigned eng = ring->idx;
-	unsigned i;
+	unsigned eng = ring->vm_inv_eng;
 
 	pd_addr = pd_addr | 0x1; /* valid bit */
 	/* now only use physical base address of PDE and valid */
 	BUG_ON(pd_addr & 0xFFFF00000000003EULL);
 
-	for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) {
-		struct amdgpu_vmhub *hub = &ring->adev->vmhub[i];
+	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) |
+			  SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf));
+	amdgpu_ring_write(ring, hub->ctx0_ptb_addr_lo32 + vm_id * 2);
+	amdgpu_ring_write(ring, lower_32_bits(pd_addr));
 
-		amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) |
-				  SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf));
-		amdgpu_ring_write(ring, hub->ctx0_ptb_addr_lo32 + vm_id * 2);
-		amdgpu_ring_write(ring, lower_32_bits(pd_addr));
+	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) |
+			  SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf));
+	amdgpu_ring_write(ring, hub->ctx0_ptb_addr_hi32 + vm_id * 2);
+	amdgpu_ring_write(ring, upper_32_bits(pd_addr));
 
-		amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) |
-				  SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf));
-		amdgpu_ring_write(ring, hub->ctx0_ptb_addr_hi32 + vm_id * 2);
-		amdgpu_ring_write(ring, upper_32_bits(pd_addr));
+	/* flush TLB */
+	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) |
+			  SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf));
+	amdgpu_ring_write(ring, hub->vm_inv_eng0_req + eng);
+	amdgpu_ring_write(ring, req);
 
-		/* flush TLB */
-		amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) |
-				  SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf));
-		amdgpu_ring_write(ring, hub->vm_inv_eng0_req + eng);
-		amdgpu_ring_write(ring, req);
-
-		/* wait for flush */
-		amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_POLL_REGMEM) |
-				  SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(0) |
-				  SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3)); /* equal */
-		amdgpu_ring_write(ring, (hub->vm_inv_eng0_ack + eng) << 2);
-		amdgpu_ring_write(ring, 0);
-		amdgpu_ring_write(ring, 1 << vm_id); /* reference */
-		amdgpu_ring_write(ring, 1 << vm_id); /* mask */
-		amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) |
-				  SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10));
-	}
+	/* wait for flush */
+	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_POLL_REGMEM) |
+			  SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(0) |
+			  SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3)); /* equal */
+	amdgpu_ring_write(ring, (hub->vm_inv_eng0_ack + eng) << 2);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, 1 << vm_id); /* reference */
+	amdgpu_ring_write(ring, 1 << vm_id); /* mask */
+	amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) |
+			  SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10));
 }
 
 static int sdma_v4_0_early_init(void *handle)
···
 	sdma_v4_0_init_golden_registers(adev);
 
 	r = sdma_v4_0_start(adev);
-	if (r)
-		return r;
 
 	return r;
 }
···
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	u32 i;
+
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		u32 tmp = RREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_STATUS_REG));
+
 		if (!(tmp & SDMA0_STATUS_REG__IDLE_MASK))
-	   		 return false;
+			return false;
 	}
 
 	return true;
···
 static int sdma_v4_0_wait_for_idle(void *handle)
 {
 	unsigned i;
-	u32 sdma0,sdma1;
+	u32 sdma0, sdma1;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
 	for (i = 0; i < adev->usec_timeout; i++) {
 		sdma0 = RREG32(sdma_v4_0_get_reg_offset(0, mmSDMA0_STATUS_REG));
 		sdma1 = RREG32(sdma_v4_0_get_reg_offset(1, mmSDMA0_STATUS_REG));
···
 
 	u32 reg_offset = (type == AMDGPU_SDMA_IRQ_TRAP0) ?
 		sdma_v4_0_get_reg_offset(0, mmSDMA0_CNTL) :
-	       	sdma_v4_0_get_reg_offset(1, mmSDMA0_CNTL);
+		sdma_v4_0_get_reg_offset(1, mmSDMA0_CNTL);
 
 	sdma_cntl = RREG32(reg_offset);
 	sdma_cntl = REG_SET_FIELD(sdma_cntl, SDMA0_CNTL, TRAP_ENABLE,
···
 				  SDMA1_CLK_CTRL__SOFT_OVERRIDE2_MASK |
 				  SDMA1_CLK_CTRL__SOFT_OVERRIDE1_MASK |
 				  SDMA1_CLK_CTRL__SOFT_OVERRIDE0_MASK);
-			if(def != data)
+			if (def != data)
 				WREG32(SOC15_REG_OFFSET(SDMA1, 0, mmSDMA1_CLK_CTRL), data);
 		}
 	} else {
···
 
 		/* 1-not override: enable sdma1 mem light sleep */
 		if (adev->asic_type == CHIP_VEGA10) {
-			 def = data = RREG32(SOC15_REG_OFFSET(SDMA1, 0, mmSDMA1_POWER_CNTL));
-			 data |= SDMA1_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
-			 if (def != data)
-				 WREG32(SOC15_REG_OFFSET(SDMA1, 0, mmSDMA1_POWER_CNTL), data);
+			def = data = RREG32(SOC15_REG_OFFSET(SDMA1, 0, mmSDMA1_POWER_CNTL));
+			data |= SDMA1_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
+			if (def != data)
+				WREG32(SOC15_REG_OFFSET(SDMA1, 0, mmSDMA1_POWER_CNTL), data);
 		}
 	} else {
 		/* 0-override:disable sdma0 mem light sleep */
 		def = data = RREG32(SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_POWER_CNTL));
 		data &= ~SDMA0_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
 		if (def != data)
-		       WREG32(SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_POWER_CNTL), data);
+			WREG32(SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_POWER_CNTL), data);
 
 		/* 0-override:disable sdma1 mem light sleep */
 		if (adev->asic_type == CHIP_VEGA10) {
···
 	.align_mask = 0xf,
 	.nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP),
 	.support_64bit_ptrs = true,
+	.vmhub = AMDGPU_MMHUB,
 	.get_rptr = sdma_v4_0_ring_get_rptr,
 	.get_wptr = sdma_v4_0_ring_get_wptr,
 	.set_wptr = sdma_v4_0_ring_set_wptr,
···
 		6 + /* sdma_v4_0_ring_emit_hdp_flush */
 		3 + /* sdma_v4_0_ring_emit_hdp_invalidate */
 		6 + /* sdma_v4_0_ring_emit_pipeline_sync */
-		36 + /* sdma_v4_0_ring_emit_vm_flush */
+		18 + /* sdma_v4_0_ring_emit_vm_flush */
 		10 + 10 + 10, /* sdma_v4_0_ring_emit_fence x3 for user fence, vm fence */
 	.emit_ib_size = 7 + 6, /* sdma_v4_0_ring_emit_ib */
 	.emit_ib = sdma_v4_0_ring_emit_ib,
···
 	}
 }
 
-const struct amdgpu_ip_block_version sdma_v4_0_ip_block =
-{
+const struct amdgpu_ip_block_version sdma_v4_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_SDMA,
 	.major = 4,
 	.minor = 0,

drivers/gpu/drm/amd/amdgpu/soc15.c

···
 #include <linux/module.h>
 #include "drmP.h"
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
+#include "amdgpu_atomfirmware.h"
 #include "amdgpu_ih.h"
 #include "amdgpu_uvd.h"
 #include "amdgpu_vce.h"
···
 
 static int soc15_asic_reset(struct amdgpu_device *adev)
 {
-	amdgpu_atombios_scratch_regs_engine_hung(adev, true);
+	amdgpu_atomfirmware_scratch_regs_engine_hung(adev, true);
 
 	soc15_gpu_pci_config_reset(adev);
 
-	amdgpu_atombios_scratch_regs_engine_hung(adev, false);
+	amdgpu_atomfirmware_scratch_regs_engine_hung(adev, false);
 
 	return 0;
 }
···
 			amdgpu_ip_block_add(adev, &dce_virtual_ip_block);
 		amdgpu_ip_block_add(adev, &gfx_v9_0_ip_block);
 		amdgpu_ip_block_add(adev, &sdma_v4_0_ip_block);
-		if (!amdgpu_sriov_vf(adev))
-			amdgpu_ip_block_add(adev, &uvd_v7_0_ip_block);
+		amdgpu_ip_block_add(adev, &uvd_v7_0_ip_block);
 		amdgpu_ip_block_add(adev, &vce_v4_0_ip_block);
 		break;
 	default:

drivers/gpu/drm/amd/amdgpu/soc15_common.h

···
 	u32 index_offset;
 	u32 data_offset;
 };
-// Register Access Macro
+
+/* Register Access Macros */
 #define SOC15_REG_OFFSET(ip, inst, reg)       (0 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG0 + reg : \
                                                 (1 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG1 + reg : \
                                                     (2 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG2 + reg : \
                                                         (3 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG3 + reg : \
                                                             (ip##_BASE__INST##inst##_SEG4 + reg)))))
+
+#define WREG32_FIELD15(ip, idx, reg, field, val)	\
+	WREG32(SOC15_REG_OFFSET(ip, idx, mm##reg), (RREG32(SOC15_REG_OFFSET(ip, idx, mm##reg)) & ~REG_FIELD_MASK(reg, field)) | (val) << REG_FIELD_SHIFT(reg, field))
+
+#define RREG32_SOC15(ip, inst, reg) \
+	RREG32( (0 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG0 + reg : \
+		(1 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG1 + reg : \
+		(2 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG2 + reg : \
+		(3 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG3 + reg : \
+		(ip##_BASE__INST##inst##_SEG4 + reg))))))
+
+#define WREG32_SOC15(ip, inst, reg, value) \
+	WREG32( (0 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG0 + reg : \
+		(1 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG1 + reg : \
+		(2 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG2 + reg : \
+		(3 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG3 + reg : \
+		(ip##_BASE__INST##inst##_SEG4 + reg))))), value)
 
 #endif
 

drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c

···
 #include "amdgpu_uvd.h"
 #include "soc15d.h"
 #include "soc15_common.h"
+#include "mmsch_v1_0.h"
 
 #include "vega10/soc15ip.h"
 #include "vega10/UVD/uvd_7_0_offset.h"
 #include "vega10/UVD/uvd_7_0_sh_mask.h"
+#include "vega10/VCE/vce_4_0_offset.h"
+#include "vega10/VCE/vce_4_0_default.h"
+#include "vega10/VCE/vce_4_0_sh_mask.h"
 #include "vega10/NBIF/nbif_6_1_offset.h"
 #include "vega10/HDP/hdp_4_0_offset.h"
 #include "vega10/MMHUB/mmhub_1_0_offset.h"
···
 static void uvd_v7_0_set_irq_funcs(struct amdgpu_device *adev);
 static int uvd_v7_0_start(struct amdgpu_device *adev);
 static void uvd_v7_0_stop(struct amdgpu_device *adev);
+static int uvd_v7_0_sriov_start(struct amdgpu_device *adev);
 
 /**
  * uvd_v7_0_ring_get_rptr - get read pointer
···
 {
 	struct amdgpu_device *adev = ring->adev;
 
+	if (ring->use_doorbell)
+		return adev->wb.wb[ring->wptr_offs];
+
 	if (ring == &adev->uvd.ring_enc[0])
 		return RREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_RB_WPTR));
 	else
···
 static void uvd_v7_0_enc_ring_set_wptr(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->use_doorbell) {
+		/* XXX check if swapping is necessary on BE */
+		adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr);
+		WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr));
+		return;
+	}
 
 	if (ring == &adev->uvd.ring_enc[0])
 		WREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_RB_WPTR),
···
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	adev->uvd.num_enc_rings = 2;
+	if (amdgpu_sriov_vf(adev))
+		adev->uvd.num_enc_rings = 1;
+	else
+		adev->uvd.num_enc_rings = 2;
 	uvd_v7_0_set_ring_funcs(adev);
 	uvd_v7_0_set_enc_ring_funcs(adev);
 	uvd_v7_0_set_irq_funcs(adev);
···
 	r = amdgpu_uvd_resume(adev);
 	if (r)
 		return r;
-
-	ring = &adev->uvd.ring;
-	sprintf(ring->name, "uvd");
-	r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.irq, 0);
-	if (r)
-		return r;
-
-	for (i = 0; i < adev->uvd.num_enc_rings; ++i) {
-		ring = &adev->uvd.ring_enc[i];
-		sprintf(ring->name, "uvd_enc%d", i);
+	if (!amdgpu_sriov_vf(adev)) {
+		ring = &adev->uvd.ring;
+		sprintf(ring->name, "uvd");
 		r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.irq, 0);
 		if (r)
 			return r;
 	}
+
+
+	for (i = 0; i < adev->uvd.num_enc_rings; ++i) {
+		ring = &adev->uvd.ring_enc[i];
+		sprintf(ring->name, "uvd_enc%d", i);
+		if (amdgpu_sriov_vf(adev)) {
+			ring->use_doorbell = true;
+			ring->doorbell_index = AMDGPU_DOORBELL64_UVD_RING0_1 * 2;
+		}
+		r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.irq, 0);
+		if (r)
+			return r;
+	}
+
+	r = amdgpu_virt_alloc_mm_table(adev);
+	if (r)
+		return r;
 
 	return r;
 }
···
 {
 	int i, r;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+	amdgpu_virt_free_mm_table(adev);
 
 	r = amdgpu_uvd_suspend(adev);
 	if (r)
···
 	uint32_t tmp;
 	int i, r;
 
-	r = uvd_v7_0_start(adev);
+	if (amdgpu_sriov_vf(adev))
+		r = uvd_v7_0_sriov_start(adev);
+	else
+		r = uvd_v7_0_start(adev);
 	if (r)
 		goto done;
 
-	ring->ready = true;
-	r = amdgpu_ring_test_ring(ring);
-	if (r) {
-		ring->ready = false;
-		goto done;
+	if (!amdgpu_sriov_vf(adev)) {
+		ring->ready = true;
+		r = amdgpu_ring_test_ring(ring);
+		if (r) {
+			ring->ready = false;
+			goto done;
+		}
+
+		r = amdgpu_ring_alloc(ring, 10);
+		if (r) {
+			DRM_ERROR("amdgpu: ring failed to lock UVD ring (%d).\n", r);
+			goto done;
+		}
+
+		tmp = PACKET0(SOC15_REG_OFFSET(UVD, 0,
+			mmUVD_SEMA_WAIT_FAULT_TIMEOUT_CNTL), 0);
+		amdgpu_ring_write(ring, tmp);
+		amdgpu_ring_write(ring, 0xFFFFF);
+
+		tmp = PACKET0(SOC15_REG_OFFSET(UVD, 0,
+			mmUVD_SEMA_WAIT_INCOMPLETE_TIMEOUT_CNTL), 0);
+		amdgpu_ring_write(ring, tmp);
+		amdgpu_ring_write(ring, 0xFFFFF);
+
+		tmp = PACKET0(SOC15_REG_OFFSET(UVD, 0,
+			mmUVD_SEMA_SIGNAL_INCOMPLETE_TIMEOUT_CNTL), 0);
+		amdgpu_ring_write(ring, tmp);
+		amdgpu_ring_write(ring, 0xFFFFF);
+
+		/* Clear timeout status bits */
+		amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, 0,
+			mmUVD_SEMA_TIMEOUT_STATUS), 0));
+		amdgpu_ring_write(ring, 0x8);
+
+		amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, 0,
+			mmUVD_SEMA_CNTL), 0));
+		amdgpu_ring_write(ring, 3);
+
+		amdgpu_ring_commit(ring);
 	}
-
-	r = amdgpu_ring_alloc(ring, 10);
-	if (r) {
-		DRM_ERROR("amdgpu: ring failed to lock UVD ring (%d).\n", r);
-		goto done;
-	}
-
-	tmp = PACKET0(SOC15_REG_OFFSET(UVD, 0,
-		mmUVD_SEMA_WAIT_FAULT_TIMEOUT_CNTL), 0);
-	amdgpu_ring_write(ring, tmp);
-	amdgpu_ring_write(ring, 0xFFFFF);
-
-	tmp = PACKET0(SOC15_REG_OFFSET(UVD, 0,
-		mmUVD_SEMA_WAIT_INCOMPLETE_TIMEOUT_CNTL), 0);
-	amdgpu_ring_write(ring, tmp);
-	amdgpu_ring_write(ring, 0xFFFFF);
-
-	tmp = PACKET0(SOC15_REG_OFFSET(UVD, 0,
-		mmUVD_SEMA_SIGNAL_INCOMPLETE_TIMEOUT_CNTL), 0);
-	amdgpu_ring_write(ring, tmp);
-	amdgpu_ring_write(ring, 0xFFFFF);
-
-	/* Clear timeout status bits */
-	amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, 0,
-		mmUVD_SEMA_TIMEOUT_STATUS), 0));
-	amdgpu_ring_write(ring, 0x8);
-
-	amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, 0,
-		mmUVD_SEMA_CNTL), 0));
-	amdgpu_ring_write(ring, 3);
-
-	amdgpu_ring_commit(ring);
 
 	for (i = 0; i < adev->uvd.num_enc_rings; ++i) {
 		ring = &adev->uvd.ring_enc[i];
···
 			adev->gfx.config.gb_addr_config);
 
 	WREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_GP_SCRATCH4), adev->uvd.max_handles);
+}
+
+static int uvd_v7_0_mmsch_start(struct amdgpu_device *adev,
+				struct amdgpu_mm_table *table)
+{
+	uint32_t data = 0, loop;
+	uint64_t addr = table->gpu_addr;
+	struct mmsch_v1_0_init_header *header = (struct mmsch_v1_0_init_header *)table->cpu_addr;
+	uint32_t size;
+
+	size = header->header_size + header->vce_table_size + header->uvd_table_size;
+
+	/* 1, write to vce_mmsch_vf_ctx_addr_lo/hi register with GPU mc addr of memory descriptor location */
+	WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_CTX_ADDR_LO), lower_32_bits(addr));
+	WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_CTX_ADDR_HI), upper_32_bits(addr));
+
+	/* 2, update vmid of descriptor */
+	data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_VMID));
+	data &= ~VCE_MMSCH_VF_VMID__VF_CTX_VMID_MASK;
+	data |= (0 << VCE_MMSCH_VF_VMID__VF_CTX_VMID__SHIFT); /* use domain0 for MM scheduler */
+	WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_VMID), data);
+
+	/* 3, notify mmsch about the size of this descriptor */
+	WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_CTX_SIZE), size);
+
+	/* 4, set resp to zero */
+	WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_RESP), 0);
+
+	/* 5, kick off the initialization and wait until VCE_MMSCH_VF_MAILBOX_RESP becomes non-zero */
+	WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_HOST), 0x10000001);
+
+	data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_RESP));
+	loop = 1000;
+	while ((data & 0x10000002) != 0x10000002) {
+		udelay(10);
+		data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_RESP));
+		loop--;
+		if (!loop)
+			break;
+	}
+
+	if (!loop) {
+		dev_err(adev->dev, "failed to init MMSCH, mmVCE_MMSCH_VF_MAILBOX_RESP = %x\n", data);
+		return -EBUSY;
+	}
+
+	return 0;
+}
+
+static int uvd_v7_0_sriov_start(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	uint32_t offset, size, tmp;
+	uint32_t table_size = 0;
+	struct mmsch_v1_0_cmd_direct_write direct_wt = { {0} };
+	struct mmsch_v1_0_cmd_direct_read_modify_write direct_rd_mod_wt = { {0} };
+	struct mmsch_v1_0_cmd_direct_polling direct_poll = { {0} };
+	struct mmsch_v1_0_cmd_end end = { {0} };
+	uint32_t *init_table = adev->virt.mm_table.cpu_addr;
+	struct mmsch_v1_0_init_header *header = (struct mmsch_v1_0_init_header *)init_table;
+
+	direct_wt.cmd_header.command_type = MMSCH_COMMAND__DIRECT_REG_WRITE;
+	direct_rd_mod_wt.cmd_header.command_type = MMSCH_COMMAND__DIRECT_REG_READ_MODIFY_WRITE;
+	direct_poll.cmd_header.command_type = MMSCH_COMMAND__DIRECT_REG_POLLING;
+	end.cmd_header.command_type = MMSCH_COMMAND__END;
+
+	if (header->uvd_table_offset == 0 && header->uvd_table_size == 0) {
+		header->version = MMSCH_VERSION;
+		header->header_size = sizeof(struct mmsch_v1_0_init_header) >> 2;
+
+		if (header->vce_table_offset == 0 && header->vce_table_size == 0)
+			header->uvd_table_offset = header->header_size;
+		else
+			header->uvd_table_offset = header->vce_table_size + header->vce_table_offset;
+
+		init_table += header->uvd_table_offset;
+
+		ring = &adev->uvd.ring;
+		size = AMDGPU_GPU_PAGE_ALIGN(adev->uvd.fw->size + 4);
+
+		/* disable clock gating */
+		MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_POWER_STATUS),
+						   ~UVD_POWER_STATUS__UVD_PG_MODE_MASK, 0);
+		MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_STATUS),
+						   0xFFFFFFFF, 0x00000004);
+		/* mc resume*/
+		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+			MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW),
+						    lower_32_bits(adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].mc_addr));
+			MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH),
+						    upper_32_bits(adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].mc_addr));
+			offset = 0;
+		} else {
+			MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW),
+						    lower_32_bits(adev->uvd.gpu_addr));
+			MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH),
+						    upper_32_bits(adev->uvd.gpu_addr));
+			offset = size;
+		}
+
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0),
+					    AMDGPU_UVD_FIRMWARE_OFFSET >> 3);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CACHE_SIZE0), size);
+
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW),
+					    lower_32_bits(adev->uvd.gpu_addr + offset));
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH),
+					    upper_32_bits(adev->uvd.gpu_addr + offset));
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CACHE_OFFSET1), (1 << 21));
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CACHE_SIZE1), AMDGPU_UVD_HEAP_SIZE);
+
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW),
+					    lower_32_bits(adev->uvd.gpu_addr + offset + AMDGPU_UVD_HEAP_SIZE));
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH),
+					    upper_32_bits(adev->uvd.gpu_addr + offset + AMDGPU_UVD_HEAP_SIZE));
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CACHE_OFFSET2), (2 << 21));
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CACHE_SIZE2),
+					    AMDGPU_UVD_STACK_SIZE + (AMDGPU_UVD_SESSION_SIZE * 40));
+
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_UDEC_ADDR_CONFIG),
+					    adev->gfx.config.gb_addr_config);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_UDEC_DB_ADDR_CONFIG),
+					    adev->gfx.config.gb_addr_config);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_UDEC_DBW_ADDR_CONFIG),
+					    adev->gfx.config.gb_addr_config);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_GP_SCRATCH4), adev->uvd.max_handles);
+		/* mc resume end*/
+
+		/* disable clock gating */
+		MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_CGC_CTRL),
+						   ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK, 0);
+
+		/* disable interupt */
+		MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN),
+						   ~UVD_MASTINT_EN__VCPU_EN_MASK, 0);
+
+		/* stall UMC and register bus before resetting VCPU */
+		MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2),
+						   ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK,
+						   UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
+
+		/* put LMI, VCPU, RBC etc... into reset */
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET),
+					    (uint32_t)(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));
+
+		/* initialize UVD memory controller */
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL),
+					    (uint32_t)((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));
+
+		/* disable byte swapping */
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_SWAP_CNTL), 0);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MP_SWAP_CNTL), 0);
+
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MPC_SET_MUXA0), 0x40c2040);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MPC_SET_MUXA1), 0x0);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MPC_SET_MUXB0), 0x40c2040);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MPC_SET_MUXB1), 0x0);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MPC_SET_ALU), 0);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MPC_SET_MUX), 0x88);
+
+		/* take all subblocks out of reset, except VCPU */
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET),
+					    UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
+
+		/* enable VCPU clock */
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CNTL),
+					    UVD_VCPU_CNTL__CLK_EN_MASK);
+
+		/* enable UMC */
+		MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2),
+						   ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, 0);
+
+		/* boot up the VCPU */
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 0);
+
+		MMSCH_V1_0_INSERT_DIRECT_POLL(SOC15_REG_OFFSET(UVD, 0, mmUVD_STATUS), 0x02, 0x02);
+
+		/* enable master interrupt */
+		MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(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));
+
+		/* clear the bit 4 of UVD_STATUS */
+		MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_STATUS),
+						   ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT), 0);
+
+		/* force RBC into idle state */
+		size = order_base_2(ring->ring_size);
+		tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, size);
+		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);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_CNTL), tmp);
+
+		/* set the write pointer delay */
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_WPTR_CNTL), 0);
+
+		/* set the wb address */
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_RPTR_ADDR),
+					    (upper_32_bits(ring->gpu_addr) >> 2));
+
+		/* programm the RB_BASE for ring buffer */
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW),
+					    lower_32_bits(ring->gpu_addr));
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH),
+					    upper_32_bits(ring->gpu_addr));
+
+		ring->wptr = 0;
+		ring = &adev->uvd.ring_enc[0];
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_RB_BASE_LO), ring->gpu_addr);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_RB_BASE_HI), upper_32_bits(ring->gpu_addr));
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_RB_SIZE), ring->ring_size / 4);
+
+		/* add end packet */
+		memcpy((void *)init_table, &end, sizeof(struct mmsch_v1_0_cmd_end));
+		table_size += sizeof(struct mmsch_v1_0_cmd_end) / 4;
+		header->uvd_table_size = table_size;
+
+		return uvd_v7_0_mmsch_start(adev, &adev->virt.mm_table);
+	}
+	return -EINVAL; /* already initializaed ? */
 }
 
 /**
···
 static void uvd_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 					unsigned vm_id, uint64_t pd_addr)
 {
+	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = ring->adev->gart.gart_funcs->get_invalidate_req(vm_id);
 	uint32_t data0, data1, mask;
-	unsigned eng = ring->idx;
-	unsigned i;
+	unsigned eng = ring->vm_inv_eng;
 
 	pd_addr = pd_addr | 0x1; /* valid bit */
 	/* now only use physical base address of PDE and valid */
 	BUG_ON(pd_addr & 0xFFFF00000000003EULL);
 
-	for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) {
-		struct amdgpu_vmhub *hub = &ring->adev->vmhub[i];
+	data0 = (hub->ctx0_ptb_addr_hi32 + vm_id * 2) << 2;
+	data1 = upper_32_bits(pd_addr);
+	uvd_v7_0_vm_reg_write(ring, data0, data1);
 
-		data0 = (hub->ctx0_ptb_addr_hi32 + vm_id * 2) << 2;
-		data1 = upper_32_bits(pd_addr);
-		uvd_v7_0_vm_reg_write(ring, data0, data1);
+	data0 = (hub->ctx0_ptb_addr_lo32 + vm_id * 2) << 2;
+	data1 = lower_32_bits(pd_addr);
+	uvd_v7_0_vm_reg_write(ring, data0, data1);
 
-		data0 = (hub->ctx0_ptb_addr_lo32 + vm_id * 2) << 2;
-		data1 = lower_32_bits(pd_addr);
-		uvd_v7_0_vm_reg_write(ring, data0, data1);
+	data0 = (hub->ctx0_ptb_addr_lo32 + vm_id * 2) << 2;
+	data1 = lower_32_bits(pd_addr);
+	mask = 0xffffffff;
+	uvd_v7_0_vm_reg_wait(ring, data0, data1, mask);
 
-		data0 = (hub->ctx0_ptb_addr_lo32 + vm_id * 2) << 2;
-		data1 = lower_32_bits(pd_addr);
-		mask = 0xffffffff;
-		uvd_v7_0_vm_reg_wait(ring, data0, data1, mask);
+	/* flush TLB */
+	data0 = (hub->vm_inv_eng0_req + eng) << 2;
+	data1 = req;
+	uvd_v7_0_vm_reg_write(ring, data0, data1);
 
-		/* flush TLB */
-		data0 = (hub->vm_inv_eng0_req + eng) << 2;
-		data1 = req;
-		uvd_v7_0_vm_reg_write(ring, data0, data1);
-
-		/* wait for flush */
-		data0 = (hub->vm_inv_eng0_ack + eng) << 2;
-		data1 = 1 << vm_id;
-		mask =  1 << vm_id;
-		uvd_v7_0_vm_reg_wait(ring, data0, data1, mask);
-	}
+	/* wait for flush */
+	data0 = (hub->vm_inv_eng0_ack + eng) << 2;
+	data1 = 1 << vm_id;
+	mask =  1 << vm_id;
+	uvd_v7_0_vm_reg_wait(ring, data0, data1, mask);
 }
 
 static void uvd_v7_0_enc_ring_insert_end(struct amdgpu_ring *ring)
···
 static void uvd_v7_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring,
 			 unsigned int vm_id, uint64_t pd_addr)
 {
+	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = ring->adev->gart.gart_funcs->get_invalidate_req(vm_id);
-	unsigned eng = ring->idx;
-	unsigned i;
+	unsigned eng = ring->vm_inv_eng;
 
 	pd_addr = pd_addr | 0x1; /* valid bit */
 	/* now only use physical base address of PDE and valid */
 	BUG_ON(pd_addr & 0xFFFF00000000003EULL);
 
-	for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) {
-		struct amdgpu_vmhub *hub = &ring->adev->vmhub[i];
+	amdgpu_ring_write(ring, HEVC_ENC_CMD_REG_WRITE);
+	amdgpu_ring_write(ring,	(hub->ctx0_ptb_addr_hi32 + vm_id * 2) << 2);
+	amdgpu_ring_write(ring, upper_32_bits(pd_addr));
 
-		amdgpu_ring_write(ring, HEVC_ENC_CMD_REG_WRITE);
-		amdgpu_ring_write(ring,
-			(hub->ctx0_ptb_addr_hi32 + vm_id * 2) << 2);
-		amdgpu_ring_write(ring, upper_32_bits(pd_addr));
+	amdgpu_ring_write(ring, HEVC_ENC_CMD_REG_WRITE);
+	amdgpu_ring_write(ring,	(hub->ctx0_ptb_addr_lo32 + vm_id * 2) << 2);
+	amdgpu_ring_write(ring, lower_32_bits(pd_addr));
 
-		amdgpu_ring_write(ring, HEVC_ENC_CMD_REG_WRITE);
-		amdgpu_ring_write(ring,
-			(hub->ctx0_ptb_addr_lo32 + vm_id * 2) << 2);
-		amdgpu_ring_write(ring, lower_32_bits(pd_addr));
+	amdgpu_ring_write(ring, HEVC_ENC_CMD_REG_WAIT);
+	amdgpu_ring_write(ring,	(hub->ctx0_ptb_addr_lo32 + vm_id * 2) << 2);
+	amdgpu_ring_write(ring, 0xffffffff);
+	amdgpu_ring_write(ring, lower_32_bits(pd_addr));
 
-		amdgpu_ring_write(ring, HEVC_ENC_CMD_REG_WAIT);
-		amdgpu_ring_write(ring,
-			(hub->ctx0_ptb_addr_lo32 + vm_id * 2) << 2);
-		amdgpu_ring_write(ring, 0xffffffff);
-		amdgpu_ring_write(ring, lower_32_bits(pd_addr));
+	/* flush TLB */
+	amdgpu_ring_write(ring, HEVC_ENC_CMD_REG_WRITE);
+	amdgpu_ring_write(ring,	(hub->vm_inv_eng0_req + eng) << 2);
+	amdgpu_ring_write(ring, req);
 
-		/* flush TLB */
-		amdgpu_ring_write(ring, HEVC_ENC_CMD_REG_WRITE);
-		amdgpu_ring_write(ring,	(hub->vm_inv_eng0_req + eng) << 2);
-		amdgpu_ring_write(ring, req);
-
-		/* wait for flush */
-		amdgpu_ring_write(ring, HEVC_ENC_CMD_REG_WAIT);
-		amdgpu_ring_write(ring, (hub->vm_inv_eng0_ack + eng) << 2);
-		amdgpu_ring_write(ring, 1 << vm_id);
-		amdgpu_ring_write(ring, 1 << vm_id);
-	}
+	/* wait for flush */
+	amdgpu_ring_write(ring, HEVC_ENC_CMD_REG_WAIT);
+	amdgpu_ring_write(ring, (hub->vm_inv_eng0_ack + eng) << 2);
+	amdgpu_ring_write(ring, 1 << vm_id);
+	amdgpu_ring_write(ring, 1 << vm_id);
 }
 
 #if 0
···
 		amdgpu_fence_process(&adev->uvd.ring_enc[0]);
 		break;
 	case 120:
-		amdgpu_fence_process(&adev->uvd.ring_enc[1]);
+		if (!amdgpu_sriov_vf(adev))
+			amdgpu_fence_process(&adev->uvd.ring_enc[1]);
 		break;
 	default:
 		DRM_ERROR("Unhandled interrupt: %d %d\n",
···
 	.align_mask = 0xf,
 	.nop = PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_NO_OP), 0),
 	.support_64bit_ptrs = false,
+	.vmhub = AMDGPU_MMHUB,
 	.get_rptr = uvd_v7_0_ring_get_rptr,
 	.get_wptr = uvd_v7_0_ring_get_wptr,
 	.set_wptr = uvd_v7_0_ring_set_wptr,
 	.emit_frame_size =
 		2 + /* uvd_v7_0_ring_emit_hdp_flush */
 		2 + /* uvd_v7_0_ring_emit_hdp_invalidate */
-		34 * AMDGPU_MAX_VMHUBS + /* uvd_v7_0_ring_emit_vm_flush */
+		34 + /* uvd_v7_0_ring_emit_vm_flush */
 		14 + 14, /* uvd_v7_0_ring_emit_fence x2 vm fence */
 	.emit_ib_size = 8, /* uvd_v7_0_ring_emit_ib */
 	.emit_ib = uvd_v7_0_ring_emit_ib,
···
 	.align_mask = 0x3f,
 	.nop = HEVC_ENC_CMD_NO_OP,
 	.support_64bit_ptrs = false,
+	.vmhub = AMDGPU_MMHUB,
 	.get_rptr = uvd_v7_0_enc_ring_get_rptr,
 	.get_wptr = uvd_v7_0_enc_ring_get_wptr,
 	.set_wptr = uvd_v7_0_enc_ring_set_wptr,
 	.emit_frame_size =
-		17 * AMDGPU_MAX_VMHUBS + /* uvd_v7_0_enc_ring_emit_vm_flush */
+		17 + /* uvd_v7_0_enc_ring_emit_vm_flush */
 		5 + 5 + /* uvd_v7_0_enc_ring_emit_fence x2 vm fence */
 		1, /* uvd_v7_0_enc_ring_insert_end */
 	.emit_ib_size = 5, /* uvd_v7_0_enc_ring_emit_ib */

drivers/gpu/drm/amd/amdgpu/vce_v4_0.c

···
 static void vce_v4_0_set_ring_funcs(struct amdgpu_device *adev);
 static void vce_v4_0_set_irq_funcs(struct amdgpu_device *adev);
 
-static inline void mmsch_insert_direct_wt(struct mmsch_v1_0_cmd_direct_write *direct_wt,
-					  uint32_t *init_table,
-					  uint32_t reg_offset,
-					  uint32_t value)
-{
-	direct_wt->cmd_header.reg_offset = reg_offset;
-	direct_wt->reg_value = value;
-	memcpy((void *)init_table, direct_wt, sizeof(struct mmsch_v1_0_cmd_direct_write));
-}
-
-static inline void mmsch_insert_direct_rd_mod_wt(struct mmsch_v1_0_cmd_direct_read_modify_write *direct_rd_mod_wt,
-						 uint32_t *init_table,
-						 uint32_t reg_offset,
-						 uint32_t mask, uint32_t data)
-{
-	direct_rd_mod_wt->cmd_header.reg_offset = reg_offset;
-	direct_rd_mod_wt->mask_value = mask;
-	direct_rd_mod_wt->write_data = data;
-	memcpy((void *)init_table, direct_rd_mod_wt,
-	       sizeof(struct mmsch_v1_0_cmd_direct_read_modify_write));
-}
-
-static inline void mmsch_insert_direct_poll(struct mmsch_v1_0_cmd_direct_polling *direct_poll,
-					    uint32_t *init_table,
-					    uint32_t reg_offset,
-					    uint32_t mask, uint32_t wait)
-{
-	direct_poll->cmd_header.reg_offset = reg_offset;
-	direct_poll->mask_value = mask;
-	direct_poll->wait_value = wait;
-	memcpy((void *)init_table, direct_poll, sizeof(struct mmsch_v1_0_cmd_direct_polling));
-}
-
-#define INSERT_DIRECT_RD_MOD_WT(reg, mask, data) { \
-	mmsch_insert_direct_rd_mod_wt(&direct_rd_mod_wt, \
-				      init_table, (reg), \
-				      (mask), (data)); \
-	init_table += sizeof(struct mmsch_v1_0_cmd_direct_read_modify_write)/4; \
-	table_size += sizeof(struct mmsch_v1_0_cmd_direct_read_modify_write)/4; \
-}
-
-#define INSERT_DIRECT_WT(reg, value) { \
-	mmsch_insert_direct_wt(&direct_wt, \
-			       init_table, (reg), \
-			       (value)); \
-	init_table += sizeof(struct mmsch_v1_0_cmd_direct_write)/4; \
-	table_size += sizeof(struct mmsch_v1_0_cmd_direct_write)/4; \
-}
-
-#define INSERT_DIRECT_POLL(reg, mask, wait) { \
-	mmsch_insert_direct_poll(&direct_poll, \
-				 init_table, (reg), \
-				 (mask), (wait)); \
-	init_table += sizeof(struct mmsch_v1_0_cmd_direct_polling)/4; \
-	table_size += sizeof(struct mmsch_v1_0_cmd_direct_polling)/4; \
-}
-
 /**
  * vce_v4_0_ring_get_rptr - get read pointer
  *
···
 		init_table += header->vce_table_offset;
 
 		ring = &adev->vce.ring[0];
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_RPTR), ring->wptr);
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_WPTR), ring->wptr);
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_BASE_LO), lower_32_bits(ring->gpu_addr));
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_BASE_HI), upper_32_bits(ring->gpu_addr));
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_SIZE), ring->ring_size / 4);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_BASE_LO),
+					    lower_32_bits(ring->gpu_addr));
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_BASE_HI),
+					    upper_32_bits(ring->gpu_addr));
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_SIZE),
+					    ring->ring_size / 4);
 
 		/* BEGING OF MC_RESUME */
-		INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_CLOCK_GATING_A), ~(1 << 16), 0);
-		INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_CLOCK_GATING), ~0xFF9FF000, 0x1FF000);
-		INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_REG_CLOCK_GATING), ~0x3F, 0x3F);
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_CLOCK_GATING_B), 0x1FF);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_CTRL), 0x398000);
+		MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_CACHE_CTRL), ~0x1, 0);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_SWAP_CNTL), 0);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_SWAP_CNTL1), 0);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VM_CTRL), 0);
 
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_CTRL), 0x398000);
-		INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_CACHE_CTRL), ~0x1, 0);
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_SWAP_CNTL), 0);
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_SWAP_CNTL1), 0);
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VM_CTRL), 0);
-
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_40BIT_BAR0), adev->vce.gpu_addr >> 8);
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_40BIT_BAR1), adev->vce.gpu_addr >> 8);
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_40BIT_BAR2), adev->vce.gpu_addr >> 8);
+		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		    MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_40BIT_BAR0),
+						adev->firmware.ucode[AMDGPU_UCODE_ID_VCE].mc_addr >> 8);
+		    MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_40BIT_BAR1),
+						adev->firmware.ucode[AMDGPU_UCODE_ID_VCE].mc_addr >> 8);
+		    MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_40BIT_BAR2),
+						adev->firmware.ucode[AMDGPU_UCODE_ID_VCE].mc_addr >> 8);
+		} else {
+		    MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_40BIT_BAR0),
+						adev->vce.gpu_addr >> 8);
+		    MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_40BIT_BAR1),
+						adev->vce.gpu_addr >> 8);
+		    MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_40BIT_BAR2),
+						adev->vce.gpu_addr >> 8);
+		}
 
 		offset = AMDGPU_VCE_FIRMWARE_OFFSET;
 		size = VCE_V4_0_FW_SIZE;
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_OFFSET0), offset & 0x7FFFFFFF);
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_SIZE0), size);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_OFFSET0),
+					    offset & 0x7FFFFFFF);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_SIZE0), size);
 
 		offset += size;
 		size = VCE_V4_0_STACK_SIZE;
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_OFFSET1), offset & 0x7FFFFFFF);
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_SIZE1), size);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_OFFSET1),
+					    offset & 0x7FFFFFFF);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_SIZE1), size);
 
 		offset += size;
 		size = VCE_V4_0_DATA_SIZE;
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_OFFSET2), offset & 0x7FFFFFFF);
-		INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_SIZE2), size);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_OFFSET2),
+					    offset & 0x7FFFFFFF);
+		MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_SIZE2), size);
 
-		INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_CTRL2), ~0x100, 0);
-		INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_SYS_INT_EN),
-				0xffffffff, VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK);
+		MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_CTRL2), ~0x100, 0);
+		MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_SYS_INT_EN),
+						   0xffffffff, VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK);
 
 		/* end of MC_RESUME */
-		INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CNTL),
-				~0x200001, VCE_VCPU_CNTL__CLK_EN_MASK);
-		INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_SOFT_RESET),
-				~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, 0);
+		MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_STATUS),
+						   VCE_STATUS__JOB_BUSY_MASK, ~VCE_STATUS__JOB_BUSY_MASK);
+		MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CNTL),
+						   ~0x200001, VCE_VCPU_CNTL__CLK_EN_MASK);
+		MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_SOFT_RESET),
+						   ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, 0);
 
-		INSERT_DIRECT_POLL(SOC15_REG_OFFSET(VCE, 0, mmVCE_STATUS),
-				VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK,
-				VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK);
+		MMSCH_V1_0_INSERT_DIRECT_POLL(SOC15_REG_OFFSET(VCE, 0, mmVCE_STATUS),
+					      VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK,
+					      VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK);
 
 		/* clear BUSY flag */
-		INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_STATUS),
-				~VCE_STATUS__JOB_BUSY_MASK, 0);
+		MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_STATUS),
+						   ~VCE_STATUS__JOB_BUSY_MASK, 0);
 
 		/* add end packet */
 		memcpy((void *)init_table, &end, sizeof(struct mmsch_v1_0_cmd_end));
···
 			return r;
 	}
 
-	if (amdgpu_sriov_vf(adev)) {
-		r = amdgpu_bo_create_kernel(adev, PAGE_SIZE, PAGE_SIZE,
-					    AMDGPU_GEM_DOMAIN_VRAM,
-					    &adev->virt.mm_table.bo,
-					    &adev->virt.mm_table.gpu_addr,
-					    (void *)&adev->virt.mm_table.cpu_addr);
-		if (!r) {
-			memset((void *)adev->virt.mm_table.cpu_addr, 0, PAGE_SIZE);
-			printk("mm table gpu addr = 0x%llx, cpu addr = %p. \n",
-			       adev->virt.mm_table.gpu_addr,
-			       adev->virt.mm_table.cpu_addr);
-		}
+	r = amdgpu_virt_alloc_mm_table(adev);
+	if (r)
 		return r;
-	}
 
 	return r;
 }
···
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	/* free MM table */
-	if (amdgpu_sriov_vf(adev))
-		amdgpu_bo_free_kernel(&adev->virt.mm_table.bo,
-				      &adev->virt.mm_table.gpu_addr,
-				      (void *)&adev->virt.mm_table.cpu_addr);
+	amdgpu_virt_free_mm_table(adev);
 
 	r = amdgpu_vce_suspend(adev);
 	if (r)
···
 static void vce_v4_0_emit_vm_flush(struct amdgpu_ring *ring,
 			 unsigned int vm_id, uint64_t pd_addr)
 {
+	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = ring->adev->gart.gart_funcs->get_invalidate_req(vm_id);
-	unsigned eng = ring->idx;
-	unsigned i;
+	unsigned eng = ring->vm_inv_eng;
 
 	pd_addr = pd_addr | 0x1; /* valid bit */
 	/* now only use physical base address of PDE and valid */
 	BUG_ON(pd_addr & 0xFFFF00000000003EULL);
 
-	for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) {
-		struct amdgpu_vmhub *hub = &ring->adev->vmhub[i];
+	amdgpu_ring_write(ring, VCE_CMD_REG_WRITE);
+	amdgpu_ring_write(ring,	(hub->ctx0_ptb_addr_hi32 + vm_id * 2) << 2);
+	amdgpu_ring_write(ring, upper_32_bits(pd_addr));
 
-		amdgpu_ring_write(ring, VCE_CMD_REG_WRITE);
-		amdgpu_ring_write(ring,
-			(hub->ctx0_ptb_addr_hi32 + vm_id * 2) << 2);
-		amdgpu_ring_write(ring, upper_32_bits(pd_addr));
+	amdgpu_ring_write(ring, VCE_CMD_REG_WRITE);
+	amdgpu_ring_write(ring,	(hub->ctx0_ptb_addr_lo32 + vm_id * 2) << 2);
+	amdgpu_ring_write(ring, lower_32_bits(pd_addr));
 
-		amdgpu_ring_write(ring, VCE_CMD_REG_WRITE);
-		amdgpu_ring_write(ring,
-			(hub->ctx0_ptb_addr_lo32 + vm_id * 2) << 2);
-		amdgpu_ring_write(ring, lower_32_bits(pd_addr));
+	amdgpu_ring_write(ring, VCE_CMD_REG_WAIT);
+	amdgpu_ring_write(ring,	(hub->ctx0_ptb_addr_lo32 + vm_id * 2) << 2);
+	amdgpu_ring_write(ring, 0xffffffff);
+	amdgpu_ring_write(ring, lower_32_bits(pd_addr));
 
-		amdgpu_ring_write(ring, VCE_CMD_REG_WAIT);
-		amdgpu_ring_write(ring,
-			(hub->ctx0_ptb_addr_lo32 + vm_id * 2) << 2);
-		amdgpu_ring_write(ring, 0xffffffff);
-		amdgpu_ring_write(ring, lower_32_bits(pd_addr));
+	/* flush TLB */
+	amdgpu_ring_write(ring, VCE_CMD_REG_WRITE);
+	amdgpu_ring_write(ring,	(hub->vm_inv_eng0_req + eng) << 2);
+	amdgpu_ring_write(ring, req);
 
-		/* flush TLB */
-		amdgpu_ring_write(ring, VCE_CMD_REG_WRITE);
-		amdgpu_ring_write(ring,	(hub->vm_inv_eng0_req + eng) << 2);
-		amdgpu_ring_write(ring, req);
-
-		/* wait for flush */
-		amdgpu_ring_write(ring, VCE_CMD_REG_WAIT);
-		amdgpu_ring_write(ring, (hub->vm_inv_eng0_ack + eng) << 2);
-		amdgpu_ring_write(ring, 1 << vm_id);
-		amdgpu_ring_write(ring, 1 << vm_id);
-	}
+	/* wait for flush */
+	amdgpu_ring_write(ring, VCE_CMD_REG_WAIT);
+	amdgpu_ring_write(ring, (hub->vm_inv_eng0_ack + eng) << 2);
+	amdgpu_ring_write(ring, 1 << vm_id);
+	amdgpu_ring_write(ring, 1 << vm_id);
 }
 
 static int vce_v4_0_set_interrupt_state(struct amdgpu_device *adev,
···
 	.align_mask = 0x3f,
 	.nop = VCE_CMD_NO_OP,
 	.support_64bit_ptrs = false,
+	.vmhub = AMDGPU_MMHUB,
 	.get_rptr = vce_v4_0_ring_get_rptr,
 	.get_wptr = vce_v4_0_ring_get_wptr,
 	.set_wptr = vce_v4_0_ring_set_wptr,
 	.parse_cs = amdgpu_vce_ring_parse_cs_vm,
 	.emit_frame_size =
-		17 * AMDGPU_MAX_VMHUBS + /* vce_v4_0_emit_vm_flush */
+		17 + /* vce_v4_0_emit_vm_flush */
 		5 + 5 + /* amdgpu_vce_ring_emit_fence x2 vm fence */
 		1, /* vce_v4_0_ring_insert_end */
 	.emit_ib_size = 5, /* vce_v4_0_ring_emit_ib */

drivers/gpu/drm/amd/include/amd_shared.h

···
 	uint8_t down_hyst;
 };
 
+enum amd_fan_ctrl_mode {
+	AMD_FAN_CTRL_NONE = 0,
+	AMD_FAN_CTRL_MANUAL = 1,
+	AMD_FAN_CTRL_AUTO = 2,
+};
+
 /* CG flags */
 #define AMD_CG_SUPPORT_GFX_MGCG			(1 << 0)
 #define AMD_CG_SUPPORT_GFX_MGLS			(1 << 1)

drivers/gpu/drm/amd/include/cgs_common.h

···
 };
 
 /**
- * enum cgs_clock - Clocks controlled by the SMU
- */
-enum cgs_clock {
-	CGS_CLOCK__SCLK,
-	CGS_CLOCK__MCLK,
-	CGS_CLOCK__VCLK,
-	CGS_CLOCK__DCLK,
-	CGS_CLOCK__ECLK,
-	CGS_CLOCK__ACLK,
-	CGS_CLOCK__ICLK,
-	/* ... */
-};
-
-/**
  * enum cgs_engine - Engines that can be statically power-gated
  */
 enum cgs_engine {
···
 	CGS_ENGINE__ACP_DSP0,
 	CGS_ENGINE__ACP_DSP1,
 	CGS_ENGINE__ISP,
-	/* ... */
-};
-
-/**
- * enum cgs_voltage_planes - Voltage planes for external camera HW
- */
-enum cgs_voltage_planes {
-	CGS_VOLTAGE_PLANE__SENSOR0,
-	CGS_VOLTAGE_PLANE__SENSOR1,
 	/* ... */
 };
 
···
 	CGS_RESOURCE_TYPE_IO,
 	CGS_RESOURCE_TYPE_DOORBELL,
 	CGS_RESOURCE_TYPE_ROM,
-};
-
-/**
- * struct cgs_clock_limits - Clock limits
- *
- * Clocks are specified in 10KHz units.
- */
-struct cgs_clock_limits {
-	unsigned min;		/**< Minimum supported frequency */
-	unsigned max;		/**< Maxumim supported frequency */
-	unsigned sustainable;	/**< Thermally sustainable frequency */
 };
 
 /**
···
 	struct cgs_acpi_method_argument *poutput_argument;
 	uint32_t padding[9];
 };
-
-/**
- * cgs_gpu_mem_info() - Return information about memory heaps
- * @cgs_device: opaque device handle
- * @type:	memory type
- * @mc_start:	Start MC address of the heap (output)
- * @mc_size:	MC address space size (output)
- * @mem_size:	maximum amount of memory available for allocation (output)
- *
- * This function returns information about memory heaps. The type
- * parameter is used to select the memory heap. The mc_start and
- * mc_size for GART heaps may be bigger than the memory available for
- * allocation.
- *
- * mc_start and mc_size are undefined for non-contiguous FB memory
- * types, since buffers allocated with these types may or may not be
- * GART mapped.
- *
- * Return:  0 on success, -errno otherwise
- */
-typedef int (*cgs_gpu_mem_info_t)(struct cgs_device *cgs_device, enum cgs_gpu_mem_type type,
-				  uint64_t *mc_start, uint64_t *mc_size,
-				  uint64_t *mem_size);
-
-/**
- * cgs_gmap_kmem() - map kernel memory to GART aperture
- * @cgs_device:	opaque device handle
- * @kmem:	pointer to kernel memory
- * @size:	size to map
- * @min_offset: minimum offset from start of GART aperture
- * @max_offset: maximum offset from start of GART aperture
- * @kmem_handle: kernel memory handle (output)
- * @mcaddr:	MC address (output)
- *
- * Return:  0 on success, -errno otherwise
- */
-typedef int (*cgs_gmap_kmem_t)(struct cgs_device *cgs_device, void *kmem, uint64_t size,
-			       uint64_t min_offset, uint64_t max_offset,
-			       cgs_handle_t *kmem_handle, uint64_t *mcaddr);
-
-/**
- * cgs_gunmap_kmem() - unmap kernel memory
- * @cgs_device:	opaque device handle
- * @kmem_handle: kernel memory handle returned by gmap_kmem
- *
- * Return:  0 on success, -errno otherwise
- */
-typedef int (*cgs_gunmap_kmem_t)(struct cgs_device *cgs_device, cgs_handle_t kmem_handle);
 
 /**
  * cgs_alloc_gpu_mem() - Allocate GPU memory
···
 					 unsigned index, uint32_t value);
 
 /**
- * cgs_read_pci_config_byte() - Read byte from PCI configuration space
- * @cgs_device:	opaque device handle
- * @addr:	address
- *
- * Return:  Value read
- */
-typedef uint8_t (*cgs_read_pci_config_byte_t)(struct cgs_device *cgs_device, unsigned addr);
-
-/**
- * cgs_read_pci_config_word() - Read word from PCI configuration space
- * @cgs_device:	opaque device handle
- * @addr:	address, must be word-aligned
- *
- * Return:  Value read
- */
-typedef uint16_t (*cgs_read_pci_config_word_t)(struct cgs_device *cgs_device, unsigned addr);
-
-/**
- * cgs_read_pci_config_dword() - Read dword from PCI configuration space
- * @cgs_device:	opaque device handle
- * @addr:	address, must be dword-aligned
- *
- * Return:  Value read
- */
-typedef uint32_t (*cgs_read_pci_config_dword_t)(struct cgs_device *cgs_device,
-						unsigned addr);
-
-/**
- * cgs_write_pci_config_byte() - Write byte to PCI configuration space
- * @cgs_device:	opaque device handle
- * @addr:	address
- * @value:	value to write
- */
-typedef void (*cgs_write_pci_config_byte_t)(struct cgs_device *cgs_device, unsigned addr,
-					    uint8_t value);
-
-/**
- * cgs_write_pci_config_word() - Write byte to PCI configuration space
- * @cgs_device:	opaque device handle
- * @addr:	address, must be word-aligned
- * @value:	value to write
- */
-typedef void (*cgs_write_pci_config_word_t)(struct cgs_device *cgs_device, unsigned addr,
-					    uint16_t value);
-
-/**
- * cgs_write_pci_config_dword() - Write byte to PCI configuration space
- * @cgs_device:	opaque device handle
- * @addr:	address, must be dword-aligned
- * @value:	value to write
- */
-typedef void (*cgs_write_pci_config_dword_t)(struct cgs_device *cgs_device, unsigned addr,
-					     uint32_t value);
-
-
-/**
  * cgs_get_pci_resource() - provide access to a device resource (PCI BAR)
  * @cgs_device:	opaque device handle
  * @resource_type:	Type of Resource (MMIO, IO, ROM, FB, DOORBELL)
···
 					 unsigned table, void *args);
 
 /**
- * cgs_create_pm_request() - Create a power management request
- * @cgs_device:	opaque device handle
- * @request:	handle of created PM request (output)
- *
- * Return:  0 on success, -errno otherwise
- */
-typedef int (*cgs_create_pm_request_t)(struct cgs_device *cgs_device, cgs_handle_t *request);
-
-/**
- * cgs_destroy_pm_request() - Destroy a power management request
- * @cgs_device:	opaque device handle
- * @request:	handle of created PM request
- *
- * Return:  0 on success, -errno otherwise
- */
-typedef int (*cgs_destroy_pm_request_t)(struct cgs_device *cgs_device, cgs_handle_t request);
-
-/**
- * cgs_set_pm_request() - Activate or deactiveate a PM request
- * @cgs_device:	opaque device handle
- * @request:	PM request handle
- * @active:	0 = deactivate, non-0 = activate
- *
- * While a PM request is active, its minimum clock requests are taken
- * into account as the requested engines are powered up. When the
- * request is inactive, the engines may be powered down and clocks may
- * be lower, depending on other PM requests by other driver
- * components.
- *
- * Return:  0 on success, -errno otherwise
- */
-typedef int (*cgs_set_pm_request_t)(struct cgs_device *cgs_device, cgs_handle_t request,
-				    int active);
-
-/**
- * cgs_pm_request_clock() - Request a minimum frequency for a specific clock
- * @cgs_device:	opaque device handle
- * @request:	PM request handle
- * @clock:	which clock?
- * @freq:	requested min. frequency in 10KHz units (0 to clear request)
- *
- * Return:  0 on success, -errno otherwise
- */
-typedef int (*cgs_pm_request_clock_t)(struct cgs_device *cgs_device, cgs_handle_t request,
-				      enum cgs_clock clock, unsigned freq);
-
-/**
- * cgs_pm_request_engine() - Request an engine to be powered up
- * @cgs_device:	opaque device handle
- * @request:	PM request handle
- * @engine:	which engine?
- * @powered:	0 = powered down, non-0 = powered up
- *
- * Return:  0 on success, -errno otherwise
- */
-typedef int (*cgs_pm_request_engine_t)(struct cgs_device *cgs_device, cgs_handle_t request,
-				       enum cgs_engine engine, int powered);
-
-/**
- * cgs_pm_query_clock_limits() - Query clock frequency limits
- * @cgs_device:	opaque device handle
- * @clock:	which clock?
- * @limits:	clock limits
- *
- * Return:  0 on success, -errno otherwise
- */
-typedef int (*cgs_pm_query_clock_limits_t)(struct cgs_device *cgs_device,
-					   enum cgs_clock clock,
-					   struct cgs_clock_limits *limits);
-
-/**
- * cgs_set_camera_voltages() - Apply specific voltages to PMIC voltage planes
- * @cgs_device:	opaque device handle
- * @mask:	bitmask of voltages to change (1<<CGS_VOLTAGE_PLANE__xyz|...)
- * @voltages:	pointer to array of voltage values in 1mV units
- *
- * Return: 0 on success, -errno otherwise
- */
-typedef int (*cgs_set_camera_voltages_t)(struct cgs_device *cgs_device, uint32_t mask,
-					 const uint32_t *voltages);
-/**
  * cgs_get_firmware_info - Get the firmware information from core driver
  * @cgs_device: opaque device handle
  * @type: the firmware type
···
 
 struct cgs_ops {
 	/* memory management calls (similar to KFD interface) */
-	cgs_gpu_mem_info_t gpu_mem_info;
-	cgs_gmap_kmem_t gmap_kmem;
-	cgs_gunmap_kmem_t gunmap_kmem;
 	cgs_alloc_gpu_mem_t alloc_gpu_mem;
 	cgs_free_gpu_mem_t free_gpu_mem;
 	cgs_gmap_gpu_mem_t gmap_gpu_mem;
···
 	cgs_write_register_t write_register;
 	cgs_read_ind_register_t read_ind_register;
 	cgs_write_ind_register_t write_ind_register;
-	/* PCI configuration space access */
-	cgs_read_pci_config_byte_t read_pci_config_byte;
-	cgs_read_pci_config_word_t read_pci_config_word;
-	cgs_read_pci_config_dword_t read_pci_config_dword;
-	cgs_write_pci_config_byte_t write_pci_config_byte;
-	cgs_write_pci_config_word_t write_pci_config_word;
-	cgs_write_pci_config_dword_t write_pci_config_dword;
 	/* PCI resources */
 	cgs_get_pci_resource_t get_pci_resource;
 	/* ATOM BIOS */
 	cgs_atom_get_data_table_t atom_get_data_table;
 	cgs_atom_get_cmd_table_revs_t atom_get_cmd_table_revs;
 	cgs_atom_exec_cmd_table_t atom_exec_cmd_table;
-	/* Power management */
-	cgs_create_pm_request_t create_pm_request;
-	cgs_destroy_pm_request_t destroy_pm_request;
-	cgs_set_pm_request_t set_pm_request;
-	cgs_pm_request_clock_t pm_request_clock;
-	cgs_pm_request_engine_t pm_request_engine;
-	cgs_pm_query_clock_limits_t pm_query_clock_limits;
-	cgs_set_camera_voltages_t set_camera_voltages;
 	/* Firmware Info */
 	cgs_get_firmware_info get_firmware_info;
 	cgs_rel_firmware rel_firmware;
···
 #define CGS_OS_CALL(func,dev,...) \
 	(((struct cgs_device *)dev)->os_ops->func(dev, ##__VA_ARGS__))
 
-#define cgs_gpu_mem_info(dev,type,mc_start,mc_size,mem_size)		\
-	CGS_CALL(gpu_mem_info,dev,type,mc_start,mc_size,mem_size)
-#define cgs_gmap_kmem(dev,kmem,size,min_off,max_off,kmem_handle,mcaddr)	\
-	CGS_CALL(gmap_kmem,dev,kmem,size,min_off,max_off,kmem_handle,mcaddr)
-#define cgs_gunmap_kmem(dev,kmem_handle)	\
-	CGS_CALL(gunmap_kmem,dev,keme_handle)
 #define cgs_alloc_gpu_mem(dev,type,size,align,min_off,max_off,handle)	\
 	CGS_CALL(alloc_gpu_mem,dev,type,size,align,min_off,max_off,handle)
 #define cgs_free_gpu_mem(dev,handle)		\
···
 #define cgs_write_ind_register(dev,space,index,value)		\
 	CGS_CALL(write_ind_register,dev,space,index,value)
 
-#define cgs_read_pci_config_byte(dev,addr)	\
-	CGS_CALL(read_pci_config_byte,dev,addr)
-#define cgs_read_pci_config_word(dev,addr)	\
-	CGS_CALL(read_pci_config_word,dev,addr)
-#define cgs_read_pci_config_dword(dev,addr)		\
-	CGS_CALL(read_pci_config_dword,dev,addr)
-#define cgs_write_pci_config_byte(dev,addr,value)	\
-	CGS_CALL(write_pci_config_byte,dev,addr,value)
-#define cgs_write_pci_config_word(dev,addr,value)	\
-	CGS_CALL(write_pci_config_word,dev,addr,value)
-#define cgs_write_pci_config_dword(dev,addr,value)	\
-	CGS_CALL(write_pci_config_dword,dev,addr,value)
-
 #define cgs_atom_get_data_table(dev,table,size,frev,crev)	\
 	CGS_CALL(atom_get_data_table,dev,table,size,frev,crev)
 #define cgs_atom_get_cmd_table_revs(dev,table,frev,crev)	\
···
 #define cgs_atom_exec_cmd_table(dev,table,args)		\
 	CGS_CALL(atom_exec_cmd_table,dev,table,args)
 
-#define cgs_create_pm_request(dev,request)	\
-	CGS_CALL(create_pm_request,dev,request)
-#define cgs_destroy_pm_request(dev,request)		\
-	CGS_CALL(destroy_pm_request,dev,request)
-#define cgs_set_pm_request(dev,request,active)		\
-	CGS_CALL(set_pm_request,dev,request,active)
-#define cgs_pm_request_clock(dev,request,clock,freq)		\
-	CGS_CALL(pm_request_clock,dev,request,clock,freq)
-#define cgs_pm_request_engine(dev,request,engine,powered)	\
-	CGS_CALL(pm_request_engine,dev,request,engine,powered)
-#define cgs_pm_query_clock_limits(dev,clock,limits)		\
-	CGS_CALL(pm_query_clock_limits,dev,clock,limits)
-#define cgs_set_camera_voltages(dev,mask,voltages)	\
-	CGS_CALL(set_camera_voltages,dev,mask,voltages)
 #define cgs_get_firmware_info(dev, type, info)	\
 	CGS_CALL(get_firmware_info, dev, type, info)
 #define cgs_rel_firmware(dev, type)	\

drivers/gpu/drm/amd/powerplay/amd_powerplay.c

···
 
 	ret = pp_check(pp_handle);
 
-	if (ret != 0)
+	if (ret == PP_DPM_DISABLED)
+		return 0;
+	else if (ret != 0)
 		return ret;
 
 	eventmgr = pp_handle->eventmgr;

drivers/gpu/drm/amd/powerplay/eventmgr/eventsubchains.c

···
 };
 
 const pem_event_action disable_smc_firmware_ctf_tasks[] = {
-	/* PEM_Task_DisableSMCFirmwareCTF,*/
+	pem_task_disable_smc_firmware_ctf,
 	NULL
 };
 

drivers/gpu/drm/amd/powerplay/eventmgr/eventtasks.c

···
 	return 0;
 }
 
+int pem_task_disable_smc_firmware_ctf(struct pp_eventmgr *eventmgr, struct pem_event_data *event_data)
+{
+	return phm_disable_smc_firmware_ctf(eventmgr->hwmgr);
+}
+
 int pem_task_setup_asic(struct pp_eventmgr *eventmgr, struct pem_event_data *event_data)
 {
 	return phm_setup_asic(eventmgr->hwmgr);

drivers/gpu/drm/amd/powerplay/eventmgr/eventtasks.h

···
 /*thermal */
 int pem_task_initialize_thermal_controller(struct pp_eventmgr *eventmgr, struct pem_event_data *event_data);
 int pem_task_uninitialize_thermal_controller(struct pp_eventmgr *eventmgr, struct pem_event_data *event_data);
+int pem_task_disable_smc_firmware_ctf(struct pp_eventmgr *eventmgr, struct pem_event_data *event_data);
 
 #endif /* _EVENT_TASKS_H_ */

drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c

···
 
 	return hwmgr->hwmgr_func->get_max_high_clocks(hwmgr, clocks);
 }
+
+int phm_disable_smc_firmware_ctf(struct pp_hwmgr *hwmgr)
+{
+	PHM_FUNC_CHECK(hwmgr);
+
+	if (hwmgr->hwmgr_func->disable_smc_firmware_ctf == NULL)
+		return -EINVAL;
+
+	return hwmgr->hwmgr_func->disable_smc_firmware_ctf(hwmgr);
+}

drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.c

···
 			le32_to_cpu(profile->gb_vdroop_table_ckson_a2);
 	param->ulGbFuseTableCksoffM1 =
 			le32_to_cpu(profile->avfsgb_fuse_table_cksoff_m1);
-	param->usGbFuseTableCksoffM2 =
+	param->ulGbFuseTableCksoffM2 =
 			le16_to_cpu(profile->avfsgb_fuse_table_cksoff_m2);
 	param->ulGbFuseTableCksoffB =
 			le32_to_cpu(profile->avfsgb_fuse_table_cksoff_b);
 	param->ulGbFuseTableCksonM1 =
 			le32_to_cpu(profile->avfsgb_fuse_table_ckson_m1);
-	param->usGbFuseTableCksonM2 =
+	param->ulGbFuseTableCksonM2 =
 			le16_to_cpu(profile->avfsgb_fuse_table_ckson_m2);
 	param->ulGbFuseTableCksonB =
 			le32_to_cpu(profile->avfsgb_fuse_table_ckson_b);
-	param->usMaxVoltage025mv =
-			le16_to_cpu(profile->max_voltage_0_25mv);
-	param->ucEnableGbVdroopTableCksoff =
-			profile->enable_gb_vdroop_table_cksoff;
+
 	param->ucEnableGbVdroopTableCkson =
 			profile->enable_gb_vdroop_table_ckson;
-	param->ucEnableGbFuseTableCksoff =
-			profile->enable_gb_fuse_table_cksoff;
 	param->ucEnableGbFuseTableCkson =
 			profile->enable_gb_fuse_table_ckson;
 	param->usPsmAgeComfactor =
 			le16_to_cpu(profile->psm_age_comfactor);
-	param->ucEnableApplyAvfsCksoffVoltage =
-			profile->enable_apply_avfs_cksoff_voltage;
 
 	param->ulDispclk2GfxclkM1 =
 			le32_to_cpu(profile->dispclk2gfxclk_a);
-	param->usDispclk2GfxclkM2 =
+	param->ulDispclk2GfxclkM2 =
 			le16_to_cpu(profile->dispclk2gfxclk_b);
 	param->ulDispclk2GfxclkB =
 			le32_to_cpu(profile->dispclk2gfxclk_c);
 	param->ulDcefclk2GfxclkM1 =
 			le32_to_cpu(profile->dcefclk2gfxclk_a);
-	param->usDcefclk2GfxclkM2 =
+	param->ulDcefclk2GfxclkM2 =
 			le16_to_cpu(profile->dcefclk2gfxclk_b);
 	param->ulDcefclk2GfxclkB =
 			le32_to_cpu(profile->dcefclk2gfxclk_c);
 	param->ulPixelclk2GfxclkM1 =
 			le32_to_cpu(profile->pixclk2gfxclk_a);
-	param->usPixelclk2GfxclkM2 =
+	param->ulPixelclk2GfxclkM2 =
 			le16_to_cpu(profile->pixclk2gfxclk_b);
 	param->ulPixelclk2GfxclkB =
 			le32_to_cpu(profile->pixclk2gfxclk_c);
 	param->ulPhyclk2GfxclkM1 =
 			le32_to_cpu(profile->phyclk2gfxclk_a);
-	param->usPhyclk2GfxclkM2 =
+	param->ulPhyclk2GfxclkM2 =
 			le16_to_cpu(profile->phyclk2gfxclk_b);
 	param->ulPhyclk2GfxclkB =
 			le32_to_cpu(profile->phyclk2gfxclk_c);
···
 	param->ucVR1HotPolarity = info->vr1hot_polarity;
 	param->ucFwCtfGpio      = info->fw_ctf_gpio_bit;
 	param->ucFwCtfPolarity  = info->fw_ctf_polarity;
+
+	return 0;
+}
+
+int pp_atomfwctrl_get_vbios_bootup_values(struct pp_hwmgr *hwmgr,
+			struct pp_atomfwctrl_bios_boot_up_values *boot_values)
+{
+	struct atom_firmware_info_v3_1 *info = NULL;
+	uint16_t ix;
+
+	ix = GetIndexIntoMasterDataTable(firmwareinfo);
+	info = (struct atom_firmware_info_v3_1 *)
+		cgs_atom_get_data_table(hwmgr->device,
+				ix, NULL, NULL, NULL);
+
+	if (!info) {
+		pr_info("Error retrieving BIOS firmwareinfo!");
+		return -EINVAL;
+	}
+
+	boot_values->ulRevision = info->firmware_revision;
+	boot_values->ulGfxClk   = info->bootup_sclk_in10khz;
+	boot_values->ulUClk     = info->bootup_mclk_in10khz;
+	boot_values->ulSocClk   = 0;
+	boot_values->usVddc     = info->bootup_vddc_mv;
+	boot_values->usVddci    = info->bootup_vddci_mv;
+	boot_values->usMvddc    = info->bootup_mvddc_mv;
+	boot_values->usVddGfx   = info->bootup_vddgfx_mv;
 
 	return 0;
 }

drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.h

···
 struct pp_atomfwctrl_avfs_parameters {
 	uint32_t   ulMaxVddc;
 	uint32_t   ulMinVddc;
-	uint8_t    ucMaxVidStep;
+
 	uint32_t   ulMeanNsigmaAcontant0;
 	uint32_t   ulMeanNsigmaAcontant1;
 	uint32_t   ulMeanNsigmaAcontant2;
···
 	uint32_t   ulGbVdroopTableCksonA0;
 	uint32_t   ulGbVdroopTableCksonA1;
 	uint32_t   ulGbVdroopTableCksonA2;
+
 	uint32_t   ulGbFuseTableCksoffM1;
-	uint16_t   usGbFuseTableCksoffM2;
-	uint32_t   ulGbFuseTableCksoffB;\
+	uint32_t   ulGbFuseTableCksoffM2;
+	uint32_t   ulGbFuseTableCksoffB;
+
 	uint32_t   ulGbFuseTableCksonM1;
-	uint16_t   usGbFuseTableCksonM2;
+	uint32_t   ulGbFuseTableCksonM2;
 	uint32_t   ulGbFuseTableCksonB;
-	uint16_t   usMaxVoltage025mv;
-	uint8_t    ucEnableGbVdroopTableCksoff;
+
 	uint8_t    ucEnableGbVdroopTableCkson;
-	uint8_t    ucEnableGbFuseTableCksoff;
 	uint8_t    ucEnableGbFuseTableCkson;
 	uint16_t   usPsmAgeComfactor;
-	uint8_t    ucEnableApplyAvfsCksoffVoltage;
+
 	uint32_t   ulDispclk2GfxclkM1;
-	uint16_t   usDispclk2GfxclkM2;
+	uint32_t   ulDispclk2GfxclkM2;
 	uint32_t   ulDispclk2GfxclkB;
 	uint32_t   ulDcefclk2GfxclkM1;
-	uint16_t   usDcefclk2GfxclkM2;
+	uint32_t   ulDcefclk2GfxclkM2;
 	uint32_t   ulDcefclk2GfxclkB;
 	uint32_t   ulPixelclk2GfxclkM1;
-	uint16_t   usPixelclk2GfxclkM2;
+	uint32_t   ulPixelclk2GfxclkM2;
 	uint32_t   ulPixelclk2GfxclkB;
 	uint32_t   ulPhyclk2GfxclkM1;
-	uint16_t   usPhyclk2GfxclkM2;
+	uint32_t   ulPhyclk2GfxclkM2;
 	uint32_t   ulPhyclk2GfxclkB;
 };
 
···
 	uint8_t   ucFwCtfGpio;
 	uint8_t   ucFwCtfPolarity;
 };
+
+struct pp_atomfwctrl_bios_boot_up_values {
+	uint32_t   ulRevision;
+	uint32_t   ulGfxClk;
+	uint32_t   ulUClk;
+	uint32_t   ulSocClk;
+	uint16_t   usVddc;
+	uint16_t   usVddci;
+	uint16_t   usMvddc;
+	uint16_t   usVddGfx;
+};
+
 int pp_atomfwctrl_get_gpu_pll_dividers_vega10(struct pp_hwmgr *hwmgr,
 		uint32_t clock_type, uint32_t clock_value,
 		struct pp_atomfwctrl_clock_dividers_soc15 *dividers);
···
 		struct pp_atomfwctrl_avfs_parameters *param);
 int pp_atomfwctrl_get_gpio_information(struct pp_hwmgr *hwmgr,
 		struct pp_atomfwctrl_gpio_parameters *param);
+
+int pp_atomfwctrl_get_vbios_bootup_values(struct pp_hwmgr *hwmgr,
+			struct pp_atomfwctrl_bios_boot_up_values *boot_values);
 
 #endif
 

drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c

···
 
 static int smu7_set_fan_control_mode(struct pp_hwmgr *hwmgr, uint32_t mode)
 {
-	if (mode) {
-		/* stop auto-manage */
-		if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps,
-				PHM_PlatformCaps_MicrocodeFanControl))
-			smu7_fan_ctrl_stop_smc_fan_control(hwmgr);
-		smu7_fan_ctrl_set_static_mode(hwmgr, mode);
-	} else
-		/* restart auto-manage */
-		smu7_fan_ctrl_reset_fan_speed_to_default(hwmgr);
+	int result = 0;
 
-	return 0;
+	switch (mode) {
+	case AMD_FAN_CTRL_NONE:
+		result = smu7_fan_ctrl_set_fan_speed_percent(hwmgr, 100);
+		break;
+	case AMD_FAN_CTRL_MANUAL:
+		if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps,
+			PHM_PlatformCaps_MicrocodeFanControl))
+			result = smu7_fan_ctrl_stop_smc_fan_control(hwmgr);
+		break;
+	case AMD_FAN_CTRL_AUTO:
+		result = smu7_fan_ctrl_set_static_mode(hwmgr, mode);
+		if (!result)
+			result = smu7_fan_ctrl_start_smc_fan_control(hwmgr);
+		break;
+	default:
+		break;
+	}
+	return result;
 }
 
 static int smu7_get_fan_control_mode(struct pp_hwmgr *hwmgr)
 {
-	if (hwmgr->fan_ctrl_is_in_default_mode)
-		return hwmgr->fan_ctrl_default_mode;
-	else
-		return PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC,
-				CG_FDO_CTRL2, FDO_PWM_MODE);
+	return hwmgr->fan_ctrl_enabled ? AMD_FAN_CTRL_AUTO : AMD_FAN_CTRL_MANUAL;
 }
 
 static int smu7_get_sclk_od(struct pp_hwmgr *hwmgr)
···
 	return 0;
 }
 
-static int smu7_request_firmware(struct pp_hwmgr *hwmgr)
-{
-	int ret;
-	struct cgs_firmware_info info = {0};
-
-	ret = cgs_get_firmware_info(hwmgr->device,
-				    smu7_convert_fw_type_to_cgs(UCODE_ID_SMU),
-				    &info);
-	if (ret || !info.kptr)
-		return -EINVAL;
-
-	return 0;
-}
-
-static int smu7_release_firmware(struct pp_hwmgr *hwmgr)
-{
-	int ret;
-
-	ret = cgs_rel_firmware(hwmgr->device,
-			       smu7_convert_fw_type_to_cgs(UCODE_ID_SMU));
-	if (ret)
-		return -EINVAL;
-
-	return 0;
-}
-
 static void smu7_find_min_clock_masks(struct pp_hwmgr *hwmgr,
 		uint32_t *sclk_mask, uint32_t *mclk_mask,
 		uint32_t min_sclk, uint32_t min_mclk)
···
 	.get_clock_by_type = smu7_get_clock_by_type,
 	.read_sensor = smu7_read_sensor,
 	.dynamic_state_management_disable = smu7_disable_dpm_tasks,
-	.request_firmware = smu7_request_firmware,
-	.release_firmware = smu7_release_firmware,
 	.set_power_profile_state = smu7_set_power_profile_state,
 	.avfs_control = smu7_avfs_control,
+	.disable_smc_firmware_ctf = smu7_thermal_disable_alert,
 };
 
 uint8_t smu7_get_sleep_divider_id_from_clock(uint32_t clock,

drivers/gpu/drm/amd/powerplay/hwmgr/smu7_thermal.c

···
 */
 int smu7_fan_ctrl_set_static_mode(struct pp_hwmgr *hwmgr, uint32_t mode)
 {
-
 	if (hwmgr->fan_ctrl_is_in_default_mode) {
 		hwmgr->fan_ctrl_default_mode =
-				PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device,	CGS_IND_REG__SMC,
+				PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC,
 						CG_FDO_CTRL2, FDO_PWM_MODE);
 		hwmgr->tmin =
 				PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC,
···
 	return 0;
 }
 
-static int smu7_fan_ctrl_start_smc_fan_control(struct pp_hwmgr *hwmgr)
+int smu7_fan_ctrl_start_smc_fan_control(struct pp_hwmgr *hwmgr)
 {
 	int result;
 
···
 				PPSMC_MSG_SetFanTemperatureTarget,
 				hwmgr->thermal_controller.
 				advanceFanControlParameters.ucTargetTemperature);
+	hwmgr->fan_ctrl_enabled = true;
 
 	return result;
 }
···
 
 int smu7_fan_ctrl_stop_smc_fan_control(struct pp_hwmgr *hwmgr)
 {
+	hwmgr->fan_ctrl_enabled = false;
 	return smum_send_msg_to_smc(hwmgr->smumgr, PPSMC_StopFanControl);
 }
 
···
 	PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC,
 				CG_TACH_STATUS, TACH_PERIOD, tach_period);
 
-	return smu7_fan_ctrl_set_static_mode(hwmgr, FDO_PWM_MODE_STATIC);
+	return smu7_fan_ctrl_set_static_mode(hwmgr, FDO_PWM_MODE_STATIC_RPM);
 }
 
 /**

drivers/gpu/drm/amd/powerplay/hwmgr/smu7_thermal.h

···
 extern int smu7_fan_ctrl_stop_smc_fan_control(struct pp_hwmgr *hwmgr);
 extern int smu7_thermal_enable_alert(struct pp_hwmgr *hwmgr);
 extern int smu7_thermal_disable_alert(struct pp_hwmgr *hwmgr);
-
+extern int smu7_fan_ctrl_start_smc_fan_control(struct pp_hwmgr *hwmgr);
 #endif
 

drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c

···
 			hwmgr->feature_mask & PP_SOCCLK_DPM_MASK ? false : true;
 	data->registry_data.mclk_dpm_key_disabled =
 			hwmgr->feature_mask & PP_MCLK_DPM_MASK ? false : true;
+	data->registry_data.pcie_dpm_key_disabled =
+			hwmgr->feature_mask & PP_PCIE_DPM_MASK ? false : true;
 
 	data->registry_data.dcefclk_dpm_key_disabled =
 			hwmgr->feature_mask & PP_DCEFCLK_DPM_MASK ? false : true;
···
 		data->registry_data.enable_tdc_limit_feature = 1;
 	}
 
-	data->registry_data.pcie_dpm_key_disabled = 1;
+	data->registry_data.clock_stretcher_support =
+			hwmgr->feature_mask & PP_CLOCK_STRETCH_MASK ? false : true;
+
 	data->registry_data.disable_water_mark = 0;
 
 	data->registry_data.fan_control_support = 1;
···
 	int i;
 
 	for (i = 0; i < dep_table->count; i++) {
-		if (i == 0 || dpm_table->dpm_levels[dpm_table->count - 1].value !=
+		if (i == 0 || dpm_table->dpm_levels[dpm_table->count - 1].value <=
 				dep_table->entries[i].clk) {
 			dpm_table->dpm_levels[dpm_table->count].value =
 					dep_table->entries[i].clk;
···
 		else
 			pcie_table->lclk[i] =
 					bios_pcie_table->entries[i].pcie_sclk;
-
-		pcie_table->count++;
 	}
 
-	if (data->registry_data.pcieSpeedOverride)
-		pcie_table->pcie_gen[i] = data->registry_data.pcieSpeedOverride;
-	else
-		pcie_table->pcie_gen[i] =
-			bios_pcie_table->entries[bios_pcie_table->count - 1].gen_speed;
-
-	if (data->registry_data.pcieLaneOverride)
-		pcie_table->pcie_lane[i] = data->registry_data.pcieLaneOverride;
-	else
-		pcie_table->pcie_lane[i] =
-			bios_pcie_table->entries[bios_pcie_table->count - 1].lane_width;
-
-	if (data->registry_data.pcieClockOverride)
-		pcie_table->lclk[i] = data->registry_data.pcieClockOverride;
-	else
-		pcie_table->lclk[i] =
-			bios_pcie_table->entries[bios_pcie_table->count - 1].pcie_sclk;
-
-	pcie_table->count++;
+	pcie_table->count = NUM_LINK_LEVELS;
 
 	return 0;
 }
···
 	dpm_table = &(data->dpm_table.eclk_table);
 	for (i = 0; i < dep_mm_table->count; i++) {
 		if (i == 0 || dpm_table->dpm_levels
-				[dpm_table->count - 1].value !=
+				[dpm_table->count - 1].value <=
 						dep_mm_table->entries[i].eclk) {
 			dpm_table->dpm_levels[dpm_table->count].value =
 					dep_mm_table->entries[i].eclk;
···
 	dpm_table = &(data->dpm_table.vclk_table);
 	for (i = 0; i < dep_mm_table->count; i++) {
 		if (i == 0 || dpm_table->dpm_levels
-				[dpm_table->count - 1].value !=
+				[dpm_table->count - 1].value <=
 						dep_mm_table->entries[i].vclk) {
 			dpm_table->dpm_levels[dpm_table->count].value =
 					dep_mm_table->entries[i].vclk;
···
 	dpm_table = &(data->dpm_table.dclk_table);
 	for (i = 0; i < dep_mm_table->count; i++) {
 		if (i == 0 || dpm_table->dpm_levels
-				[dpm_table->count - 1].value !=
+				[dpm_table->count - 1].value <=
 						dep_mm_table->entries[i].dclk) {
 			dpm_table->dpm_levels[dpm_table->count].value =
 					dep_mm_table->entries[i].dclk;
···
 			(struct phm_ppt_v2_information *)(hwmgr->pptable);
 
 	data->smc_state_table.pp_table.UlvOffsetVid =
-			(uint8_t)(table_info->us_ulv_voltage_offset *
-					VOLTAGE_VID_OFFSET_SCALE2 /
-					VOLTAGE_VID_OFFSET_SCALE1);
+			(uint8_t)table_info->us_ulv_voltage_offset;
 
 	data->smc_state_table.pp_table.UlvSmnclkDid =
 			(uint8_t)(table_info->us_ulv_smnclk_did);
···
 	current_gfxclk_level->FbMult =
 			cpu_to_le32(dividers.ulPll_fb_mult);
 	/* Spread FB Multiplier bit: bit 0:8 int, bit 31:16 frac */
-	current_gfxclk_level->SsOn = dividers.ucPll_ss_enable;
+	if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps,
+				PHM_PlatformCaps_EngineSpreadSpectrumSupport))
+		current_gfxclk_level->SsOn = dividers.ucPll_ss_enable;
+	else
+		current_gfxclk_level->SsOn = 0;
 	current_gfxclk_level->SsFbMult =
 			cpu_to_le32(dividers.ulPll_ss_fbsmult);
 	current_gfxclk_level->SsSlewFrac =
···
 			table_info->vdd_dep_on_sclk;
 	uint32_t i;
 
-	for (i = 0; dep_table->count; i++) {
+	for (i = 0; i < dep_table->count; i++) {
 		pp_table->CksEnable[i] = dep_table->entries[i].cks_enable;
-		pp_table->CksVidOffset[i] = convert_to_vid(
-				dep_table->entries[i].cks_voffset);
+		pp_table->CksVidOffset[i] = (uint8_t)(dep_table->entries[i].cks_voffset
+				* VOLTAGE_VID_OFFSET_SCALE2 / VOLTAGE_VID_OFFSET_SCALE1);
 	}
 
 	return 0;
···
 		result = pp_atomfwctrl_get_avfs_information(hwmgr, &avfs_params);
 		if (!result) {
 			pp_table->MinVoltageVid = (uint8_t)
-					convert_to_vid((uint16_t)(avfs_params.ulMaxVddc));
-			pp_table->MaxVoltageVid = (uint8_t)
 					convert_to_vid((uint16_t)(avfs_params.ulMinVddc));
-			pp_table->BtcGbVdroopTableCksOn.a0 =
-					cpu_to_le32(avfs_params.ulGbVdroopTableCksonA0);
-			pp_table->BtcGbVdroopTableCksOn.a1 =
-					cpu_to_le32(avfs_params.ulGbVdroopTableCksonA1);
-			pp_table->BtcGbVdroopTableCksOn.a2 =
-					cpu_to_le32(avfs_params.ulGbVdroopTableCksonA2);
+			pp_table->MaxVoltageVid = (uint8_t)
+					convert_to_vid((uint16_t)(avfs_params.ulMaxVddc));
+
+			pp_table->AConstant[0] = cpu_to_le32(avfs_params.ulMeanNsigmaAcontant0);
+			pp_table->AConstant[1] = cpu_to_le32(avfs_params.ulMeanNsigmaAcontant1);
+			pp_table->AConstant[2] = cpu_to_le32(avfs_params.ulMeanNsigmaAcontant2);
+			pp_table->DC_tol_sigma = cpu_to_le16(avfs_params.usMeanNsigmaDcTolSigma);
+			pp_table->Platform_mean = cpu_to_le16(avfs_params.usMeanNsigmaPlatformMean);
+			pp_table->Platform_sigma = cpu_to_le16(avfs_params.usMeanNsigmaDcTolSigma);
+			pp_table->PSM_Age_CompFactor = cpu_to_le16(avfs_params.usPsmAgeComfactor);
 
 			pp_table->BtcGbVdroopTableCksOff.a0 =
 					cpu_to_le32(avfs_params.ulGbVdroopTableCksoffA0);
+			pp_table->BtcGbVdroopTableCksOff.a0_shift = 20;
 			pp_table->BtcGbVdroopTableCksOff.a1 =
 					cpu_to_le32(avfs_params.ulGbVdroopTableCksoffA1);
+			pp_table->BtcGbVdroopTableCksOff.a1_shift = 20;
 			pp_table->BtcGbVdroopTableCksOff.a2 =
 					cpu_to_le32(avfs_params.ulGbVdroopTableCksoffA2);
+			pp_table->BtcGbVdroopTableCksOff.a2_shift = 20;
+
+			pp_table->OverrideBtcGbCksOn = avfs_params.ucEnableGbVdroopTableCkson;
+			pp_table->BtcGbVdroopTableCksOn.a0 =
+					cpu_to_le32(avfs_params.ulGbVdroopTableCksonA0);
+			pp_table->BtcGbVdroopTableCksOn.a0_shift = 20;
+			pp_table->BtcGbVdroopTableCksOn.a1 =
+					cpu_to_le32(avfs_params.ulGbVdroopTableCksonA1);
+			pp_table->BtcGbVdroopTableCksOn.a1_shift = 20;
+			pp_table->BtcGbVdroopTableCksOn.a2 =
+					cpu_to_le32(avfs_params.ulGbVdroopTableCksonA2);
+			pp_table->BtcGbVdroopTableCksOn.a2_shift = 20;
 
 			pp_table->AvfsGbCksOn.m1 =
 					cpu_to_le32(avfs_params.ulGbFuseTableCksonM1);
 			pp_table->AvfsGbCksOn.m2 =
-					cpu_to_le16(avfs_params.usGbFuseTableCksonM2);
+					cpu_to_le16(avfs_params.ulGbFuseTableCksonM2);
 			pp_table->AvfsGbCksOn.b =
 					cpu_to_le32(avfs_params.ulGbFuseTableCksonB);
 			pp_table->AvfsGbCksOn.m1_shift = 24;
 			pp_table->AvfsGbCksOn.m2_shift = 12;
+			pp_table->AvfsGbCksOn.b_shift = 0;
 
+			pp_table->OverrideAvfsGbCksOn =
+					avfs_params.ucEnableGbFuseTableCkson;
 			pp_table->AvfsGbCksOff.m1 =
 					cpu_to_le32(avfs_params.ulGbFuseTableCksoffM1);
 			pp_table->AvfsGbCksOff.m2 =
-					cpu_to_le16(avfs_params.usGbFuseTableCksoffM2);
+					cpu_to_le16(avfs_params.ulGbFuseTableCksoffM2);
 			pp_table->AvfsGbCksOff.b =
 					cpu_to_le32(avfs_params.ulGbFuseTableCksoffB);
 			pp_table->AvfsGbCksOff.m1_shift = 24;
 			pp_table->AvfsGbCksOff.m2_shift = 12;
+			pp_table->AvfsGbCksOff.b_shift = 0;
 
-			pp_table->AConstant[0] =
-					cpu_to_le32(avfs_params.ulMeanNsigmaAcontant0);
-			pp_table->AConstant[1] =
-					cpu_to_le32(avfs_params.ulMeanNsigmaAcontant1);
-			pp_table->AConstant[2] =
-					cpu_to_le32(avfs_params.ulMeanNsigmaAcontant2);
-			pp_table->DC_tol_sigma =
-					cpu_to_le16(avfs_params.usMeanNsigmaDcTolSigma);
-			pp_table->Platform_mean =
-					cpu_to_le16(avfs_params.usMeanNsigmaPlatformMean);
-			pp_table->PSM_Age_CompFactor =
-					cpu_to_le16(avfs_params.usPsmAgeComfactor);
-			pp_table->Platform_sigma =
-					cpu_to_le16(avfs_params.usMeanNsigmaDcTolSigma);
-
-			for (i = 0; i < dep_table->count; i++)
-				pp_table->StaticVoltageOffsetVid[i] = (uint8_t)
-						(dep_table->entries[i].sclk_offset *
+			for (i = 0; i < dep_table->count; i++) {
+				if (dep_table->entries[i].sclk_offset == 0)
+					pp_table->StaticVoltageOffsetVid[i] = 248;
+				else
+					pp_table->StaticVoltageOffsetVid[i] =
+						(uint8_t)(dep_table->entries[i].sclk_offset *
 								VOLTAGE_VID_OFFSET_SCALE2 /
 								VOLTAGE_VID_OFFSET_SCALE1);
-
-			pp_table->OverrideBtcGbCksOn =
-					avfs_params.ucEnableGbVdroopTableCkson;
-			pp_table->OverrideAvfsGbCksOn =
-					avfs_params.ucEnableGbFuseTableCkson;
+			}
 
 			if ((PPREGKEY_VEGA10QUADRATICEQUATION_DFLT !=
 					data->disp_clk_quad_eqn_a) &&
···
 				pp_table->DisplayClock2Gfxclk[DSPCLK_DISPCLK].m1 =
 						(int32_t)data->disp_clk_quad_eqn_a;
 				pp_table->DisplayClock2Gfxclk[DSPCLK_DISPCLK].m2 =
-						(int16_t)data->disp_clk_quad_eqn_b;
+						(int32_t)data->disp_clk_quad_eqn_b;
 				pp_table->DisplayClock2Gfxclk[DSPCLK_DISPCLK].b =
 						(int32_t)data->disp_clk_quad_eqn_c;
 			} else {
 				pp_table->DisplayClock2Gfxclk[DSPCLK_DISPCLK].m1 =
 						(int32_t)avfs_params.ulDispclk2GfxclkM1;
 				pp_table->DisplayClock2Gfxclk[DSPCLK_DISPCLK].m2 =
-						(int16_t)avfs_params.usDispclk2GfxclkM2;
+						(int32_t)avfs_params.ulDispclk2GfxclkM2;
 				pp_table->DisplayClock2Gfxclk[DSPCLK_DISPCLK].b =
 						(int32_t)avfs_params.ulDispclk2GfxclkB;
 			}
 
 			pp_table->DisplayClock2Gfxclk[DSPCLK_DISPCLK].m1_shift = 24;
 			pp_table->DisplayClock2Gfxclk[DSPCLK_DISPCLK].m2_shift = 12;
+			pp_table->DisplayClock2Gfxclk[DSPCLK_DISPCLK].b_shift = 12;
 
 			if ((PPREGKEY_VEGA10QUADRATICEQUATION_DFLT !=
 					data->dcef_clk_quad_eqn_a) &&
···
 				pp_table->DisplayClock2Gfxclk[DSPCLK_DCEFCLK].m1 =
 						(int32_t)data->dcef_clk_quad_eqn_a;
 				pp_table->DisplayClock2Gfxclk[DSPCLK_DCEFCLK].m2 =
-						(int16_t)data->dcef_clk_quad_eqn_b;
+						(int32_t)data->dcef_clk_quad_eqn_b;
 				pp_table->DisplayClock2Gfxclk[DSPCLK_DCEFCLK].b =
 						(int32_t)data->dcef_clk_quad_eqn_c;
 			} else {
 				pp_table->DisplayClock2Gfxclk[DSPCLK_DCEFCLK].m1 =
 						(int32_t)avfs_params.ulDcefclk2GfxclkM1;
 				pp_table->DisplayClock2Gfxclk[DSPCLK_DCEFCLK].m2 =
-						(int16_t)avfs_params.usDcefclk2GfxclkM2;
+						(int32_t)avfs_params.ulDcefclk2GfxclkM2;
 				pp_table->DisplayClock2Gfxclk[DSPCLK_DCEFCLK].b =
 						(int32_t)avfs_params.ulDcefclk2GfxclkB;
 			}
 
 			pp_table->DisplayClock2Gfxclk[DSPCLK_DCEFCLK].m1_shift = 24;
 			pp_table->DisplayClock2Gfxclk[DSPCLK_DCEFCLK].m2_shift = 12;
+			pp_table->DisplayClock2Gfxclk[DSPCLK_DCEFCLK].b_shift = 12;
 
 			if ((PPREGKEY_VEGA10QUADRATICEQUATION_DFLT !=
 					data->pixel_clk_quad_eqn_a) &&
···
 				pp_table->DisplayClock2Gfxclk[DSPCLK_PIXCLK].m1 =
 						(int32_t)data->pixel_clk_quad_eqn_a;
 				pp_table->DisplayClock2Gfxclk[DSPCLK_PIXCLK].m2 =
-						(int16_t)data->pixel_clk_quad_eqn_b;
+						(int32_t)data->pixel_clk_quad_eqn_b;
 				pp_table->DisplayClock2Gfxclk[DSPCLK_PIXCLK].b =
 						(int32_t)data->pixel_clk_quad_eqn_c;
 			} else {
 				pp_table->DisplayClock2Gfxclk[DSPCLK_PIXCLK].m1 =
 						(int32_t)avfs_params.ulPixelclk2GfxclkM1;
 				pp_table->DisplayClock2Gfxclk[DSPCLK_PIXCLK].m2 =
-						(int16_t)avfs_params.usPixelclk2GfxclkM2;
+						(int32_t)avfs_params.ulPixelclk2GfxclkM2;
 				pp_table->DisplayClock2Gfxclk[DSPCLK_PIXCLK].b =
 						(int32_t)avfs_params.ulPixelclk2GfxclkB;
 			}
 
 			pp_table->DisplayClock2Gfxclk[DSPCLK_PIXCLK].m1_shift = 24;
 			pp_table->DisplayClock2Gfxclk[DSPCLK_PIXCLK].m2_shift = 12;
-
+			pp_table->DisplayClock2Gfxclk[DSPCLK_PIXCLK].b_shift = 12;
 			if ((PPREGKEY_VEGA10QUADRATICEQUATION_DFLT !=
 					data->phy_clk_quad_eqn_a) &&
 				(PPREGKEY_VEGA10QUADRATICEQUATION_DFLT !=
···
 				pp_table->DisplayClock2Gfxclk[DSPCLK_PHYCLK].m1 =
 						(int32_t)data->phy_clk_quad_eqn_a;
 				pp_table->DisplayClock2Gfxclk[DSPCLK_PHYCLK].m2 =
-						(int16_t)data->phy_clk_quad_eqn_b;
+						(int32_t)data->phy_clk_quad_eqn_b;
 				pp_table->DisplayClock2Gfxclk[DSPCLK_PHYCLK].b =
 						(int32_t)data->phy_clk_quad_eqn_c;
 			} else {
 				pp_table->DisplayClock2Gfxclk[DSPCLK_PHYCLK].m1 =
 						(int32_t)avfs_params.ulPhyclk2GfxclkM1;
 				pp_table->DisplayClock2Gfxclk[DSPCLK_PHYCLK].m2 =
-						(int16_t)avfs_params.usPhyclk2GfxclkM2;
+						(int32_t)avfs_params.ulPhyclk2GfxclkM2;
 				pp_table->DisplayClock2Gfxclk[DSPCLK_PHYCLK].b =
 						(int32_t)avfs_params.ulPhyclk2GfxclkB;
 			}
 
 			pp_table->DisplayClock2Gfxclk[DSPCLK_PHYCLK].m1_shift = 24;
 			pp_table->DisplayClock2Gfxclk[DSPCLK_PHYCLK].m2_shift = 12;
+			pp_table->DisplayClock2Gfxclk[DSPCLK_PHYCLK].b_shift = 12;
 		} else {
 			data->smu_features[GNLD_AVFS].supported = false;
 		}
···
 			(struct phm_ppt_v2_information *)(hwmgr->pptable);
 	PPTable_t *pp_table = &(data->smc_state_table.pp_table);
 	struct pp_atomfwctrl_voltage_table voltage_table;
+	struct pp_atomfwctrl_bios_boot_up_values boot_up_values;
 
 	result = vega10_setup_default_dpm_tables(hwmgr);
 	PP_ASSERT_WITH_CODE(!result,
···
 			(uint8_t)(table_info->uc_vce_dpm_voltage_mode);
 	pp_table->Mp0DpmVoltageMode =
 			(uint8_t)(table_info->uc_mp0_dpm_voltage_mode);
+
 	pp_table->DisplayDpmVoltageMode =
 			(uint8_t)(table_info->uc_dcef_dpm_voltage_mode);
 
···
 			"Failed to initialize UVD Level!",
 			return result);
 
-	if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps,
-			PHM_PlatformCaps_ClockStretcher)) {
+	if (data->registry_data.clock_stretcher_support) {
 		result = vega10_populate_clock_stretcher_table(hwmgr);
 		PP_ASSERT_WITH_CODE(!result,
 				"Failed to populate Clock Stretcher Table!",
 				return result);
+	}
+
+	result = pp_atomfwctrl_get_vbios_bootup_values(hwmgr, &boot_up_values);
+	if (!result) {
+		data->vbios_boot_state.vddc     = boot_up_values.usVddc;
+		data->vbios_boot_state.vddci    = boot_up_values.usVddci;
+		data->vbios_boot_state.mvddc    = boot_up_values.usMvddc;
+		data->vbios_boot_state.gfx_clock = boot_up_values.ulGfxClk;
+		data->vbios_boot_state.mem_clock = boot_up_values.ulUClk;
+		data->vbios_boot_state.soc_clock = boot_up_values.ulSocClk;
+		if (0 != boot_up_values.usVddc) {
+			smum_send_msg_to_smc_with_parameter(hwmgr->smumgr,
+						PPSMC_MSG_SetFloorSocVoltage,
+						(boot_up_values.usVddc * 4));
+			data->vbios_boot_state.bsoc_vddc_lock = true;
+		} else {
+			data->vbios_boot_state.bsoc_vddc_lock = false;
+		}
 	}
 
 	result = vega10_populate_avfs_parameters(hwmgr);
···
 	PP_ASSERT_WITH_CODE(!result,
 			"Failed to upload PPtable!", return result);
 
-	if (data->smu_features[GNLD_AVFS].supported) {
-		uint32_t features_enabled;
-		result = vega10_get_smc_features(hwmgr->smumgr, &features_enabled);
-		PP_ASSERT_WITH_CODE(!result,
-				"Failed to Retrieve Enabled Features!",
-				return result);
-		if (!(features_enabled & (1 << FEATURE_AVFS_BIT))) {
-			result = vega10_perform_btc(hwmgr->smumgr);
-			PP_ASSERT_WITH_CODE(!result,
-					"Failed to Perform BTC!",
+	result = vega10_avfs_enable(hwmgr, true);
+	PP_ASSERT_WITH_CODE(!result, "Attempt to enable AVFS feature Failed!",
 					return result);
-			result = vega10_avfs_enable(hwmgr, true);
-			PP_ASSERT_WITH_CODE(!result,
-					"Attempt to enable AVFS feature Failed!",
-					return result);
-			result = vega10_save_vft_table(hwmgr->smumgr,
-					(uint8_t *)&(data->smc_state_table.avfs_table));
-			PP_ASSERT_WITH_CODE(!result,
-					"Attempt to save VFT table Failed!",
-					return result);
-		} else {
-			data->smu_features[GNLD_AVFS].enabled = true;
-			result = vega10_restore_vft_table(hwmgr->smumgr,
-					(uint8_t *)&(data->smc_state_table.avfs_table));
-			PP_ASSERT_WITH_CODE(!result,
-					"Attempt to restore VFT table Failed!",
-					return result;);
-		}
-	}
 
 	return 0;
 }
···
 				"Enable THERMAL Feature Failed!",
 				return -1);
 		data->smu_features[GNLD_THERMAL].enabled = true;
+	}
+
+	return 0;
+}
+
+static int vega10_disable_thermal_protection(struct pp_hwmgr *hwmgr)
+{
+	struct vega10_hwmgr *data = (struct vega10_hwmgr *)(hwmgr->backend);
+
+	if (data->smu_features[GNLD_THERMAL].supported) {
+		if (!data->smu_features[GNLD_THERMAL].enabled)
+			pr_info("THERMAL Feature Already disabled!");
+
+		PP_ASSERT_WITH_CODE(
+				!vega10_enable_smc_features(hwmgr->smumgr,
+				false,
+				data->smu_features[GNLD_THERMAL].smu_feature_bitmap),
+				"disable THERMAL Feature Failed!",
+				return -1);
+		data->smu_features[GNLD_THERMAL].enabled = false;
 	}
 
 	return 0;
···
 	return 0;
 }
 
+static int vega10_stop_dpm(struct pp_hwmgr *hwmgr, uint32_t bitmap)
+{
+	struct vega10_hwmgr *data =
+			(struct vega10_hwmgr *)(hwmgr->backend);
+	uint32_t i, feature_mask = 0;
+
+
+	if(data->smu_features[GNLD_LED_DISPLAY].supported == true){
+		PP_ASSERT_WITH_CODE(!vega10_enable_smc_features(hwmgr->smumgr,
+				true, data->smu_features[GNLD_LED_DISPLAY].smu_feature_bitmap),
+		"Attempt to Enable LED DPM feature Failed!", return -EINVAL);
+		data->smu_features[GNLD_LED_DISPLAY].enabled = true;
+	}
+
+	for (i = 0; i < GNLD_DPM_MAX; i++) {
+		if (data->smu_features[i].smu_feature_bitmap & bitmap) {
+			if (data->smu_features[i].supported) {
+				if (data->smu_features[i].enabled) {
+					feature_mask |= data->smu_features[i].
+							smu_feature_bitmap;
+					data->smu_features[i].enabled = false;
+				}
+			}
+		}
+	}
+
+	vega10_enable_smc_features(hwmgr->smumgr, false, feature_mask);
+
+	return 0;
+}
+
 /**
  * @brief Tell SMC to enabled the supported DPMs.
  *
···
 		data->smu_features[GNLD_LED_DISPLAY].enabled = true;
 	}
 
+	if (data->vbios_boot_state.bsoc_vddc_lock) {
+		smum_send_msg_to_smc_with_parameter(hwmgr->smumgr,
+						PPSMC_MSG_SetFloorSocVoltage, 0);
+		data->vbios_boot_state.bsoc_vddc_lock = false;
+	}
+
 	if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps,
 			PHM_PlatformCaps_Falcon_QuickTransition)) {
 		if (data->smu_features[GNLD_ACDC].supported) {
···
 	PP_ASSERT_WITH_CODE(!tmp_result,
 			"Failed to configure telemetry!",
 			return tmp_result);
-
-	vega10_set_tools_address(hwmgr->smumgr);
 
 	smum_send_msg_to_smc_with_parameter(hwmgr->smumgr,
 			PPSMC_MSG_NumOfDisplays, 0);
···
 
 static int vega10_set_fan_control_mode(struct pp_hwmgr *hwmgr, uint32_t mode)
 {
-	if (mode) {
-		/* stop auto-manage */
-		if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps,
-				PHM_PlatformCaps_MicrocodeFanControl))
-			vega10_fan_ctrl_stop_smc_fan_control(hwmgr);
-		vega10_fan_ctrl_set_static_mode(hwmgr, mode);
-	} else
-		/* restart auto-manage */
-		vega10_fan_ctrl_reset_fan_speed_to_default(hwmgr);
+	int result = 0;
 
-	return 0;
+	switch (mode) {
+	case AMD_FAN_CTRL_NONE:
+		result = vega10_fan_ctrl_set_fan_speed_percent(hwmgr, 100);
+		break;
+	case AMD_FAN_CTRL_MANUAL:
+		if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps,
+			PHM_PlatformCaps_MicrocodeFanControl))
+			result = vega10_fan_ctrl_stop_smc_fan_control(hwmgr);
+		break;
+	case AMD_FAN_CTRL_AUTO:
+		result = vega10_fan_ctrl_set_static_mode(hwmgr, mode);
+		if (!result)
+			result = vega10_fan_ctrl_start_smc_fan_control(hwmgr);
+		break;
+	default:
+		break;
+	}
+	return result;
 }
 
 static int vega10_get_fan_control_mode(struct pp_hwmgr *hwmgr)
 {
-	uint32_t reg;
+	struct vega10_hwmgr *data = (struct vega10_hwmgr *)(hwmgr->backend);
 
-	if (hwmgr->fan_ctrl_is_in_default_mode) {
-		return hwmgr->fan_ctrl_default_mode;
-	} else {
-		reg = soc15_get_register_offset(THM_HWID, 0,
-			mmCG_FDO_CTRL2_BASE_IDX, mmCG_FDO_CTRL2);
-		return (cgs_read_register(hwmgr->device, reg) &
-				CG_FDO_CTRL2__FDO_PWM_MODE_MASK) >>
-				CG_FDO_CTRL2__FDO_PWM_MODE__SHIFT;
-	}
+	if (data->smu_features[GNLD_FAN_CONTROL].enabled == false)
+		return AMD_FAN_CTRL_MANUAL;
+	else
+		return AMD_FAN_CTRL_AUTO;
 }
 
 static int vega10_get_dal_power_level(struct pp_hwmgr *hwmgr,
···
 
 	switch (type) {
 	case PP_SCLK:
-		if (data->registry_data.sclk_dpm_key_disabled)
-			break;
-
 		for (i = 0; i < 32; i++) {
 			if (mask & (1 << i))
 				break;
 		}
+		data->smc_state_table.gfx_boot_level = i;
 
-		PP_ASSERT_WITH_CODE(!smum_send_msg_to_smc_with_parameter(
-				hwmgr->smumgr,
-				PPSMC_MSG_SetSoftMinGfxclkByIndex,
-				i),
-				"Failed to set soft min sclk index!",
-				return -1);
+		for (i = 31; i >= 0; i--) {
+			if (mask & (1 << i))
+				break;
+		}
+		data->smc_state_table.gfx_max_level = i;
+
+		PP_ASSERT_WITH_CODE(!vega10_upload_dpm_bootup_level(hwmgr),
+			"Failed to upload boot level to lowest!",
+			return -EINVAL);
+
+		PP_ASSERT_WITH_CODE(!vega10_upload_dpm_max_level(hwmgr),
+			"Failed to upload dpm max level to highest!",
+			return -EINVAL);
 		break;
 
 	case PP_MCLK:
-		if (data->registry_data.mclk_dpm_key_disabled)
-			break;
-
 		for (i = 0; i < 32; i++) {
 			if (mask & (1 << i))
 				break;
 		}
 
-		PP_ASSERT_WITH_CODE(!smum_send_msg_to_smc_with_parameter(
-				hwmgr->smumgr,
-				PPSMC_MSG_SetSoftMinUclkByIndex,
-				i),
-				"Failed to set soft min mclk index!",
-				return -1);
+		for (i = 0; i < 32; i++) {
+			if (mask & (1 << i))
+				break;
+		}
+		data->smc_state_table.mem_boot_level = i;
+
+		for (i = 31; i >= 0; i--) {
+			if (mask & (1 << i))
+				break;
+		}
+		data->smc_state_table.mem_max_level = i;
+
+		PP_ASSERT_WITH_CODE(!vega10_upload_dpm_bootup_level(hwmgr),
+			"Failed to upload boot level to lowest!",
+			return -EINVAL);
+
+		PP_ASSERT_WITH_CODE(!vega10_upload_dpm_max_level(hwmgr),
+			"Failed to upload dpm max level to highest!",
+			return -EINVAL);
+
 		break;
 
 	case PP_PCIE:
-		if (data->registry_data.pcie_dpm_key_disabled)
-			break;
-
-		for (i = 0; i < 32; i++) {
-			if (mask & (1 << i))
-				break;
-		}
-
-		PP_ASSERT_WITH_CODE(!smum_send_msg_to_smc_with_parameter(
-				hwmgr->smumgr,
-				PPSMC_MSG_SetMinLinkDpmByIndex,
-				i),
-				"Failed to set min pcie index!",
-				return -1);
-		break;
 	default:
 		break;
 	}
···
 	return is_update_required;
 }
 
+static int vega10_disable_dpm_tasks(struct pp_hwmgr *hwmgr)
+{
+	int tmp_result, result = 0;
+
+	tmp_result = (vega10_is_dpm_running(hwmgr)) ? 0 : -1;
+	PP_ASSERT_WITH_CODE(tmp_result == 0,
+			"DPM is not running right now, no need to disable DPM!",
+			return 0);
+
+	if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps,
+			PHM_PlatformCaps_ThermalController))
+		vega10_disable_thermal_protection(hwmgr);
+
+	tmp_result = vega10_disable_power_containment(hwmgr);
+	PP_ASSERT_WITH_CODE((tmp_result == 0),
+			"Failed to disable power containment!", result = tmp_result);
+
+	tmp_result = vega10_avfs_enable(hwmgr, false);
+	PP_ASSERT_WITH_CODE((tmp_result == 0),
+			"Failed to disable AVFS!", result = tmp_result);
+
+	tmp_result = vega10_stop_dpm(hwmgr, SMC_DPM_FEATURES);
+	PP_ASSERT_WITH_CODE((tmp_result == 0),
+			"Failed to stop DPM!", result = tmp_result);
+
+	return result;
+}
+
+static int vega10_power_off_asic(struct pp_hwmgr *hwmgr)
+{
+	struct vega10_hwmgr *data = (struct vega10_hwmgr *)(hwmgr->backend);
+	int result;
+
+	result = vega10_disable_dpm_tasks(hwmgr);
+	PP_ASSERT_WITH_CODE((0 == result),
+			"[disable_dpm_tasks] Failed to disable DPM!",
+			);
+	data->water_marks_bitmap &= ~(WaterMarksLoaded);
+
+	return result;
+}
+
+
 static const struct pp_hwmgr_func vega10_hwmgr_funcs = {
 	.backend_init = vega10_hwmgr_backend_init,
 	.backend_fini = vega10_hwmgr_backend_fini,
 	.asic_setup = vega10_setup_asic_task,
 	.dynamic_state_management_enable = vega10_enable_dpm_tasks,
+	.dynamic_state_management_disable = vega10_disable_dpm_tasks,
 	.get_num_of_pp_table_entries =
 			vega10_get_number_of_powerplay_table_entries,
 	.get_power_state_size = vega10_get_power_state_size,
···
 	.check_states_equal = vega10_check_states_equal,
 	.check_smc_update_required_for_display_configuration =
 			vega10_check_smc_update_required_for_display_configuration,
+	.power_off_asic = vega10_power_off_asic,
+	.disable_smc_firmware_ctf = vega10_thermal_disable_alert,
 };
 
 int vega10_hwmgr_init(struct pp_hwmgr *hwmgr)

drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.h

···
 };
 
 struct vega10_vbios_boot_state {
+	bool        bsoc_vddc_lock;
 	uint16_t    vddc;
 	uint16_t    vddci;
+	uint16_t    mvddc;
+	uint16_t    vdd_gfx;
 	uint32_t    gfx_clock;
 	uint32_t    mem_clock;
 	uint32_t    soc_clock;

drivers/gpu/drm/amd/powerplay/hwmgr/vega10_powertune.c

···
 	table->Tliquid1Limit = cpu_to_le16(tdp_table->usTemperatureLimitLiquid1);
 	table->Tliquid2Limit = cpu_to_le16(tdp_table->usTemperatureLimitLiquid2);
 	table->TplxLimit = cpu_to_le16(tdp_table->usTemperatureLimitPlx);
-	table->LoadLineResistance = cpu_to_le16(
-			hwmgr->platform_descriptor.LoadLineSlope);
+	table->LoadLineResistance =
+			hwmgr->platform_descriptor.LoadLineSlope * 256;
 	table->FitLimit = 0; /* Not used for Vega10 */
 
 	table->Liquid1_I2C_address = tdp_table->ucLiquid1_I2C_address;
···
 	}
 
 	return result;
+}
+
+int vega10_disable_power_containment(struct pp_hwmgr *hwmgr)
+{
+	struct vega10_hwmgr *data =
+			(struct vega10_hwmgr *)(hwmgr->backend);
+
+	if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps,
+			PHM_PlatformCaps_PowerContainment)) {
+		if (data->smu_features[GNLD_PPT].supported)
+			PP_ASSERT_WITH_CODE(!vega10_enable_smc_features(hwmgr->smumgr,
+					false, data->smu_features[GNLD_PPT].smu_feature_bitmap),
+					"Attempt to disable PPT feature Failed!",
+					data->smu_features[GNLD_PPT].supported = false);
+
+		if (data->smu_features[GNLD_TDC].supported)
+			PP_ASSERT_WITH_CODE(!vega10_enable_smc_features(hwmgr->smumgr,
+					false, data->smu_features[GNLD_TDC].smu_feature_bitmap),
+					"Attempt to disable PPT feature Failed!",
+					data->smu_features[GNLD_TDC].supported = false);
+	}
+
+	return 0;
 }
 
 static int vega10_set_overdrive_target_percentage(struct pp_hwmgr *hwmgr,

drivers/gpu/drm/amd/powerplay/hwmgr/vega10_powertune.h

···
 int vega10_enable_power_containment(struct pp_hwmgr *hwmgr);
 int vega10_set_power_limit(struct pp_hwmgr *hwmgr, uint32_t n);
 int vega10_power_control_set_level(struct pp_hwmgr *hwmgr);
+int vega10_disable_power_containment(struct pp_hwmgr *hwmgr);
 
 #endif  /* _VEGA10_POWERTUNE_H_ */
 

drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.c

···
 		tdp_table->ucPlx_I2C_address = power_tune_table->ucPlx_I2C_address;
 		tdp_table->ucPlx_I2C_Line = power_tune_table->ucPlx_I2C_LineSCL;
 		tdp_table->ucPlx_I2C_LineSDA = power_tune_table->ucPlx_I2C_LineSDA;
-		hwmgr->platform_descriptor.LoadLineSlope = power_tune_table->usLoadLineResistance;
+		hwmgr->platform_descriptor.LoadLineSlope = le16_to_cpu(power_tune_table->usLoadLineResistance);
 	} else {
 		power_tune_table_v2 = (ATOM_Vega10_PowerTune_Table_V2 *)table;
 		tdp_table->usMaximumPowerDeliveryLimit = le16_to_cpu(power_tune_table_v2->usSocketPowerLimit);
···
 		tdp_table->ucPlx_I2C_LineSDA = sda;
 
 		hwmgr->platform_descriptor.LoadLineSlope =
-					power_tune_table_v2->usLoadLineResistance;
+					le16_to_cpu(power_tune_table_v2->usLoadLineResistance);
 	}
 
 	*info_tdp_table = tdp_table;

drivers/gpu/drm/amd/powerplay/hwmgr/vega10_thermal.c

···
 
 	temp = cgs_read_register(hwmgr->device, reg);
 
-	temp = (temp & CG_MULT_THERMAL_STATUS__CTF_TEMP_MASK) >>
-			CG_MULT_THERMAL_STATUS__CTF_TEMP__SHIFT;
+	temp = (temp & CG_MULT_THERMAL_STATUS__ASIC_MAX_TEMP_MASK) >>
+			CG_MULT_THERMAL_STATUS__ASIC_MAX_TEMP__SHIFT;
 
-	/* Bit 9 means the reading is lower than the lowest usable value. */
-	if (temp & 0x200)
-		temp = VEGA10_THERMAL_MAXIMUM_TEMP_READING;
-	else
-		temp = temp & 0x1ff;
+	temp = temp & 0x1ff;
 
 	temp *= PP_TEMPERATURE_UNITS_PER_CENTIGRADES;
 
···
 			mmTHM_THERMAL_INT_CTRL_BASE_IDX, mmTHM_THERMAL_INT_CTRL);
 
 	val = cgs_read_register(hwmgr->device, reg);
-	val &= ~(THM_THERMAL_INT_CTRL__DIG_THERM_INTH_MASK);
-	val |= (high / PP_TEMPERATURE_UNITS_PER_CENTIGRADES) <<
-			THM_THERMAL_INT_CTRL__DIG_THERM_INTH__SHIFT;
-	val &= ~(THM_THERMAL_INT_CTRL__DIG_THERM_INTL_MASK);
-	val |= (low / PP_TEMPERATURE_UNITS_PER_CENTIGRADES) <<
-			THM_THERMAL_INT_CTRL__DIG_THERM_INTL__SHIFT;
+
+	val &= (~THM_THERMAL_INT_CTRL__MAX_IH_CREDIT_MASK);
+	val |=  (5 << THM_THERMAL_INT_CTRL__MAX_IH_CREDIT__SHIFT);
+
+	val &= (~THM_THERMAL_INT_CTRL__THERM_IH_HW_ENA_MASK);
+	val |= (1 << THM_THERMAL_INT_CTRL__THERM_IH_HW_ENA__SHIFT);
+
+	val &= (~THM_THERMAL_INT_CTRL__DIG_THERM_INTH_MASK);
+	val |= ((high / PP_TEMPERATURE_UNITS_PER_CENTIGRADES)
+			<< THM_THERMAL_INT_CTRL__DIG_THERM_INTH__SHIFT);
+
+	val &= (~THM_THERMAL_INT_CTRL__DIG_THERM_INTL_MASK);
+	val |= ((low / PP_TEMPERATURE_UNITS_PER_CENTIGRADES)
+			<< THM_THERMAL_INT_CTRL__DIG_THERM_INTL__SHIFT);
+
+	val = val & (~THM_THERMAL_INT_CTRL__THERM_TRIGGER_MASK_MASK);
+
 	cgs_write_register(hwmgr->device, reg, val);
 
 	reg = soc15_get_register_offset(THM_HWID, 0,
 			mmTHM_TCON_HTC_BASE_IDX, mmTHM_TCON_HTC);
-
-	val = cgs_read_register(hwmgr->device, reg);
-	val &= ~(THM_TCON_HTC__HTC_TMP_LMT_MASK);
-	val |= (high / PP_TEMPERATURE_UNITS_PER_CENTIGRADES) <<
-			THM_TCON_HTC__HTC_TMP_LMT__SHIFT;
-	cgs_write_register(hwmgr->device, reg, val);
 
 	return 0;
 }
···
 static int vega10_thermal_enable_alert(struct pp_hwmgr *hwmgr)
 {
 	struct vega10_hwmgr *data = (struct vega10_hwmgr *)(hwmgr->backend);
+	uint32_t val = 0;
+	uint32_t reg;
 
 	if (data->smu_features[GNLD_FW_CTF].supported) {
 		if (data->smu_features[GNLD_FW_CTF].enabled)
 			printk("[Thermal_EnableAlert] FW CTF Already Enabled!\n");
+
+		PP_ASSERT_WITH_CODE(!vega10_enable_smc_features(hwmgr->smumgr,
+				true,
+				data->smu_features[GNLD_FW_CTF].smu_feature_bitmap),
+				"Attempt to Enable FW CTF feature Failed!",
+				return -1);
+		data->smu_features[GNLD_FW_CTF].enabled = true;
 	}
 
-	PP_ASSERT_WITH_CODE(!vega10_enable_smc_features(hwmgr->smumgr,
-			true,
-			data->smu_features[GNLD_FW_CTF].smu_feature_bitmap),
-			"Attempt to Enable FW CTF feature Failed!",
-			return -1);
-	data->smu_features[GNLD_FW_CTF].enabled = true;
+	val |= (1 << THM_THERMAL_INT_ENA__THERM_INTH_CLR__SHIFT);
+	val |= (1 << THM_THERMAL_INT_ENA__THERM_INTL_CLR__SHIFT);
+	val |= (1 << THM_THERMAL_INT_ENA__THERM_TRIGGER_CLR__SHIFT);
+
+	reg = soc15_get_register_offset(THM_HWID, 0, mmTHM_THERMAL_INT_ENA_BASE_IDX, mmTHM_THERMAL_INT_ENA);
+	cgs_write_register(hwmgr->device, reg, val);
+
 	return 0;
 }
 
···
 * Disable thermal alerts on the RV770 thermal controller.
 * @param    hwmgr The address of the hardware manager.
 */
-static int vega10_thermal_disable_alert(struct pp_hwmgr *hwmgr)
+int vega10_thermal_disable_alert(struct pp_hwmgr *hwmgr)
 {
 	struct vega10_hwmgr *data = (struct vega10_hwmgr *)(hwmgr->backend);
+	uint32_t reg;
 
 	if (data->smu_features[GNLD_FW_CTF].supported) {
 		if (!data->smu_features[GNLD_FW_CTF].enabled)
 			printk("[Thermal_EnableAlert] FW CTF Already disabled!\n");
-	}
 
-	PP_ASSERT_WITH_CODE(!vega10_enable_smc_features(hwmgr->smumgr,
+
+		PP_ASSERT_WITH_CODE(!vega10_enable_smc_features(hwmgr->smumgr,
 			false,
 			data->smu_features[GNLD_FW_CTF].smu_feature_bitmap),
 			"Attempt to disable FW CTF feature Failed!",
 			return -1);
-	data->smu_features[GNLD_FW_CTF].enabled = false;
+		data->smu_features[GNLD_FW_CTF].enabled = false;
+	}
+
+	reg = soc15_get_register_offset(THM_HWID, 0, mmTHM_THERMAL_INT_ENA_BASE_IDX, mmTHM_THERMAL_INT_ENA);
+	cgs_write_register(hwmgr->device, reg, 0);
+
 	return 0;
 }
 
···
 			advanceFanControlParameters.ulMinFanSCLKAcousticLimit);
 	table->FanTargetTemperature = hwmgr->thermal_controller.
 			advanceFanControlParameters.usTMax;
+
+	smum_send_msg_to_smc_with_parameter(hwmgr->smumgr,
+				PPSMC_MSG_SetFanTemperatureTarget,
+				(uint32_t)table->FanTargetTemperature);
+
 	table->FanPwmMin = hwmgr->thermal_controller.
 			advanceFanControlParameters.usPWMMin * 255 / 100;
 	table->FanTargetGfxclk = (uint16_t)(hwmgr->thermal_controller.

drivers/gpu/drm/amd/powerplay/hwmgr/vega10_thermal.h

···
 		uint32_t *speed);
 extern int vega10_fan_ctrl_stop_smc_fan_control(struct pp_hwmgr *hwmgr);
 extern uint32_t smu7_get_xclk(struct pp_hwmgr *hwmgr);
+extern int vega10_thermal_disable_alert(struct pp_hwmgr *hwmgr);
+int vega10_fan_ctrl_start_smc_fan_control(struct pp_hwmgr *hwmgr);
 
 #endif
 

drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h

···
 		struct pp_display_clock_request *clock);
 
 extern int phm_get_max_high_clocks(struct pp_hwmgr *hwmgr, struct amd_pp_simple_clock_info *clocks);
-
+extern int phm_disable_smc_firmware_ctf(struct pp_hwmgr *hwmgr);
 #endif /* _HARDWARE_MANAGER_H_ */
 

drivers/gpu/drm/amd/powerplay/inc/hwmgr.h

···
 	int (*get_mclk_od)(struct pp_hwmgr *hwmgr);
 	int (*set_mclk_od)(struct pp_hwmgr *hwmgr, uint32_t value);
 	int (*read_sensor)(struct pp_hwmgr *hwmgr, int idx, void *value, int *size);
-	int (*request_firmware)(struct pp_hwmgr *hwmgr);
-	int (*release_firmware)(struct pp_hwmgr *hwmgr);
 	int (*set_power_profile_state)(struct pp_hwmgr *hwmgr,
 			struct amd_pp_profile *request);
 	int (*avfs_control)(struct pp_hwmgr *hwmgr, bool enable);
+	int (*disable_smc_firmware_ctf)(struct pp_hwmgr *hwmgr);
 };
 
 struct pp_table_func {
···
 	struct pp_thermal_controller_info thermal_controller;
 	bool fan_ctrl_is_in_default_mode;
 	uint32_t fan_ctrl_default_mode;
+	bool fan_ctrl_enabled;
 	uint32_t tmin;
 	struct phm_microcode_version_info microcode_version_info;
 	uint32_t ps_size;

drivers/gpu/drm/amd/powerplay/inc/smu9_driver_if.h

···
  * SMU TEAM: Always increment the interface version if
  * any structure is changed in this file
  */
-#define SMU9_DRIVER_IF_VERSION 0xB
+#define SMU9_DRIVER_IF_VERSION 0xD
 
 #define PPTABLE_V10_SMU_VERSION 1
 
···
 
   uint32_t     DpmLevelPowerDelta;
 
-  uint32_t     Reserved[19];
+  uint8_t      EnableBoostState;
+  uint8_t      AConstant_Shift;
+  uint8_t      DC_tol_sigma_Shift;
+  uint8_t      PSM_Age_CompFactor_Shift;
+
+  uint16_t     BoostStartTemperature;
+  uint16_t     BoostStopTemperature;
+
+  PllSetting_t GfxBoostState;
+
+  uint32_t     Reserved[14];
 
   /* Padding - ignore */
   uint32_t     MmHubPadding[7]; /* SMU internal use */
···
 #define DB_IR_SHIFT 25
 #define DB_PCC_SHIFT 26
 #define DB_EDC_SHIFT 27
+
+#define REMOVE_FMAX_MARGIN_BIT     0x0
+#define REMOVE_DCTOL_MARGIN_BIT    0x1
+#define REMOVE_PLATFORM_MARGIN_BIT 0x2
 
 #endif

drivers/gpu/drm/amd/powerplay/inc/vega10_ppsmc.h

···
 #define PPSMC_MSG_SetFanMinPwm                   0x52
 #define PPSMC_MSG_ConfigureGfxDidt               0x55
 #define PPSMC_MSG_NumOfDisplays                  0x56
-#define PPSMC_Message_Count                      0x57
+#define PPSMC_MSG_ReadSerialNumTop32             0x58
+#define PPSMC_MSG_ReadSerialNumBottom32          0x59
+#define PPSMC_Message_Count                      0x5A
+
 
 typedef int PPSMC_Msg;
 

drivers/gpu/drm/amd/powerplay/smumgr/vega10_smumgr.c

···
 	return false;
 }
 
-/**
-* Check if SMC has responded to previous message.
-*
-* @param    smumgr  the address of the powerplay hardware manager.
-* @return   TRUE    SMC has responded, FALSE otherwise.
-*/
+/*
+ * Check if SMC has responded to previous message.
+ *
+ * @param    smumgr  the address of the powerplay hardware manager.
+ * @return   TRUE    SMC has responded, FALSE otherwise.
+ */
 static uint32_t vega10_wait_for_response(struct pp_smumgr *smumgr)
 {
 	uint32_t reg;
 
 	if (!vega10_is_smc_ram_running(smumgr))
-		return -1;
+		return -EINVAL;
 
 	reg = soc15_get_register_offset(MP1_HWID, 0,
 			mmMP1_SMN_C2PMSG_90_BASE_IDX, mmMP1_SMN_C2PMSG_90);
···
 	return cgs_read_register(smumgr->device, reg);
 }
 
-/**
-* Send a message to the SMC, and do not wait for its response.
-*
-* @param    smumgr  the address of the powerplay hardware manager.
-* @param    msg the message to send.
-* @return   Always return 0.
-*/
+/*
+ * Send a message to the SMC, and do not wait for its response.
+ * @param    smumgr  the address of the powerplay hardware manager.
+ * @param    msg the message to send.
+ * @return   Always return 0.
+ */
 int vega10_send_msg_to_smc_without_waiting(struct pp_smumgr *smumgr,
 		uint16_t msg)
 {
 	uint32_t reg;
 
 	if (!vega10_is_smc_ram_running(smumgr))
-		return -1;
+		return -EINVAL;
 
 	reg = soc15_get_register_offset(MP1_HWID, 0,
 			mmMP1_SMN_C2PMSG_66_BASE_IDX, mmMP1_SMN_C2PMSG_66);
···
 	return 0;
 }
 
-/**
-* Send a message to the SMC, and wait for its response.
-*
-* @param    smumgr  the address of the powerplay hardware manager.
-* @param    msg the message to send.
-* @return   The response that came from the SMC.
-*/
+/*
+ * Send a message to the SMC, and wait for its response.
+ * @param    smumgr  the address of the powerplay hardware manager.
+ * @param    msg the message to send.
+ * @return   Always return 0.
+ */
 int vega10_send_msg_to_smc(struct pp_smumgr *smumgr, uint16_t msg)
 {
 	uint32_t reg;
 
 	if (!vega10_is_smc_ram_running(smumgr))
-		return -1;
+		return -EINVAL;
 
 	vega10_wait_for_response(smumgr);
 
···
 
 	vega10_send_msg_to_smc_without_waiting(smumgr, msg);
 
-	PP_ASSERT_WITH_CODE(vega10_wait_for_response(smumgr) == 1,
-			"Failed to send Message.",
-			return -1);
+	if (vega10_wait_for_response(smumgr) != 1)
+		pr_err("Failed to send message: 0x%x\n", msg);
 
 	return 0;
 }
 
-/**
+/*
  * Send a message to the SMC with parameter
  * @param    smumgr:  the address of the powerplay hardware manager.
  * @param    msg: the message to send.
  * @param    parameter: the parameter to send
- * @return   The response that came from the SMC.
+ * @return   Always return 0.
  */
 int vega10_send_msg_to_smc_with_parameter(struct pp_smumgr *smumgr,
 		uint16_t msg, uint32_t parameter)
···
 	uint32_t reg;
 
 	if (!vega10_is_smc_ram_running(smumgr))
-		return -1;
+		return -EINVAL;
 
 	vega10_wait_for_response(smumgr);
 
···
 
 	vega10_send_msg_to_smc_without_waiting(smumgr, msg);
 
-	PP_ASSERT_WITH_CODE(vega10_wait_for_response(smumgr) == 1,
-			"Failed to send Message.",
-			return -1);
+	if (vega10_wait_for_response(smumgr) != 1)
+		pr_err("Failed to send message: 0x%x\n", msg);
 
 	return 0;
 }
 
 
-/**
-* Send a message to the SMC with parameter, do not wait for response
-*
-* @param    smumgr:  the address of the powerplay hardware manager.
-* @param    msg: the message to send.
-* @param    parameter: the parameter to send
-* @return   The response that came from the SMC.
-*/
+/*
+ * Send a message to the SMC with parameter, do not wait for response
+ * @param    smumgr:  the address of the powerplay hardware manager.
+ * @param    msg: the message to send.
+ * @param    parameter: the parameter to send
+ * @return   The response that came from the SMC.
+ */
 int vega10_send_msg_to_smc_with_parameter_without_waiting(
 		struct pp_smumgr *smumgr, uint16_t msg, uint32_t parameter)
 {
···
 	return vega10_send_msg_to_smc_without_waiting(smumgr, msg);
 }
 
-/**
-* Retrieve an argument from SMC.
-*
-* @param    smumgr  the address of the powerplay hardware manager.
-* @param    arg     pointer to store the argument from SMC.
-* @return   Always return 0.
-*/
+/*
+ * Retrieve an argument from SMC.
+ * @param    smumgr  the address of the powerplay hardware manager.
+ * @param    arg     pointer to store the argument from SMC.
+ * @return   Always return 0.
+ */
 int vega10_read_arg_from_smc(struct pp_smumgr *smumgr, uint32_t *arg)
 {
 	uint32_t reg;
···
 	return 0;
 }
 
-/**
-* Copy table from SMC into driver FB
-* @param   smumgr    the address of the SMC manager
-* @param   table_id    the driver's table ID to copy from
-*/
+/*
+ * Copy table from SMC into driver FB
+ * @param   smumgr    the address of the SMC manager
+ * @param   table_id    the driver's table ID to copy from
+ */
 int vega10_copy_table_from_smc(struct pp_smumgr *smumgr,
 		uint8_t *table, int16_t table_id)
 {
···
 			(struct vega10_smumgr *)(smumgr->backend);
 
 	PP_ASSERT_WITH_CODE(table_id < MAX_SMU_TABLE,
-			"Invalid SMU Table ID!", return -1;);
+			"Invalid SMU Table ID!", return -EINVAL);
 	PP_ASSERT_WITH_CODE(priv->smu_tables.entry[table_id].version != 0,
-			"Invalid SMU Table version!", return -1;);
+			"Invalid SMU Table version!", return -EINVAL);
 	PP_ASSERT_WITH_CODE(priv->smu_tables.entry[table_id].size != 0,
-			"Invalid SMU Table Length!", return -1;);
+			"Invalid SMU Table Length!", return -EINVAL);
 	PP_ASSERT_WITH_CODE(vega10_send_msg_to_smc_with_parameter(smumgr,
 			PPSMC_MSG_SetDriverDramAddrHigh,
 			priv->smu_tables.entry[table_id].table_addr_high) == 0,
-			"[CopyTableFromSMC] Attempt to Set Dram Addr High Failed!", return -1;);
+			"[CopyTableFromSMC] Attempt to Set Dram Addr High Failed!", return -EINVAL);
 	PP_ASSERT_WITH_CODE(vega10_send_msg_to_smc_with_parameter(smumgr,
 			PPSMC_MSG_SetDriverDramAddrLow,
 			priv->smu_tables.entry[table_id].table_addr_low) == 0,
 			"[CopyTableFromSMC] Attempt to Set Dram Addr Low Failed!",
-			return -1;);
+			return -EINVAL);
 	PP_ASSERT_WITH_CODE(vega10_send_msg_to_smc_with_parameter(smumgr,
 			PPSMC_MSG_TransferTableSmu2Dram,
 			priv->smu_tables.entry[table_id].table_id) == 0,
 			"[CopyTableFromSMC] Attempt to Transfer Table From SMU Failed!",
-			return -1;);
+			return -EINVAL);
 
 	memcpy(table, priv->smu_tables.entry[table_id].table,
 			priv->smu_tables.entry[table_id].size);
···
 	return 0;
 }
 
-/**
-* Copy table from Driver FB into SMC
-* @param   smumgr    the address of the SMC manager
-* @param   table_id    the table to copy from
-*/
+/*
+ * Copy table from Driver FB into SMC
+ * @param   smumgr    the address of the SMC manager
+ * @param   table_id    the table to copy from
+ */
 int vega10_copy_table_to_smc(struct pp_smumgr *smumgr,
 		uint8_t *table, int16_t table_id)
 {
···
 			(struct vega10_smumgr *)(smumgr->backend);
 
 	PP_ASSERT_WITH_CODE(table_id < MAX_SMU_TABLE,
-			"Invalid SMU Table ID!", return -1;);
+			"Invalid SMU Table ID!", return -EINVAL);
 	PP_ASSERT_WITH_CODE(priv->smu_tables.entry[table_id].version != 0,
-			"Invalid SMU Table version!", return -1;);
+			"Invalid SMU Table version!", return -EINVAL);
 	PP_ASSERT_WITH_CODE(priv->smu_tables.entry[table_id].size != 0,
-			"Invalid SMU Table Length!", return -1;);
+			"Invalid SMU Table Length!", return -EINVAL);
 
 	memcpy(priv->smu_tables.entry[table_id].table, table,
 			priv->smu_tables.entry[table_id].size);
···
 			PPSMC_MSG_SetDriverDramAddrHigh,
 			priv->smu_tables.entry[table_id].table_addr_high) == 0,
 			"[CopyTableToSMC] Attempt to Set Dram Addr High Failed!",
-			return -1;);
+			return -EINVAL;);
 	PP_ASSERT_WITH_CODE(vega10_send_msg_to_smc_with_parameter(smumgr,
 			PPSMC_MSG_SetDriverDramAddrLow,
 			priv->smu_tables.entry[table_id].table_addr_low) == 0,
 			"[CopyTableToSMC] Attempt to Set Dram Addr Low Failed!",
-			return -1;);
+			return -EINVAL);
 	PP_ASSERT_WITH_CODE(vega10_send_msg_to_smc_with_parameter(smumgr,
 			PPSMC_MSG_TransferTableDram2Smu,
 			priv->smu_tables.entry[table_id].table_id) == 0,
 			"[CopyTableToSMC] Attempt to Transfer Table To SMU Failed!",
-			return -1;);
+			return -EINVAL);
 
-	return 0;
-}
-
-int vega10_perform_btc(struct pp_smumgr *smumgr)
-{
-	PP_ASSERT_WITH_CODE(!vega10_send_msg_to_smc_with_parameter(
-			smumgr, PPSMC_MSG_RunBtc, 0),
-			"Attempt to run DC BTC Failed!",
-			return -1);
 	return 0;
 }
 
···
 {
 	PP_ASSERT_WITH_CODE(avfs_table,
 			"No access to SMC AVFS Table",
-			return -1);
+			return -EINVAL);
 
 	return vega10_copy_table_from_smc(smumgr, avfs_table, AVFSTABLE);
 }
···
 {
 	PP_ASSERT_WITH_CODE(avfs_table,
 			"No access to SMC AVFS Table",
-			return -1);
+			return -EINVAL);
 
 	return vega10_copy_table_to_smc(smumgr, avfs_table, AVFSTABLE);
 }
···
 int vega10_get_smc_features(struct pp_smumgr *smumgr,
 		uint32_t *features_enabled)
 {
+	if (features_enabled == NULL)
+		return -EINVAL;
+
 	if (!vega10_send_msg_to_smc(smumgr,
 			PPSMC_MSG_GetEnabledSmuFeatures)) {
-		if (!vega10_read_arg_from_smc(smumgr, features_enabled))
-			return 0;
+		vega10_read_arg_from_smc(smumgr, features_enabled);
+		return 0;
 	}
 
-	return -1;
+	return -EINVAL;
 }
 
 int vega10_set_tools_address(struct pp_smumgr *smumgr)
···
 	PP_ASSERT_WITH_CODE(!vega10_send_msg_to_smc(smumgr,
 			PPSMC_MSG_GetDriverIfVersion),
 			"Attempt to get SMC IF Version Number Failed!",
-			return -1);
-	PP_ASSERT_WITH_CODE(!vega10_read_arg_from_smc(smumgr,
-			&smc_driver_if_version),
-			"Attempt to read SMC IF Version Number Failed!",
-			return -1);
+			return -EINVAL);
+	vega10_read_arg_from_smc(smumgr, &smc_driver_if_version);
 
-	if (smc_driver_if_version != SMU9_DRIVER_IF_VERSION)
-		return -1;
+	if (smc_driver_if_version != SMU9_DRIVER_IF_VERSION) {
+		pr_err("Your firmware(0x%x) doesn't match \
+			SMU9_DRIVER_IF_VERSION(0x%x). \
+			Please update your firmware!\n",
+			smc_driver_if_version, SMU9_DRIVER_IF_VERSION);
+		return -EINVAL;
+	}
 
 	return 0;
 }
 
-/**
-* Write a 32bit value to the SMC SRAM space.
-* ALL PARAMETERS ARE IN HOST BYTE ORDER.
-* @param    smumgr  the address of the powerplay hardware manager.
-* @param    smc_addr the address in the SMC RAM to access.
-* @param    value to write to the SMC SRAM.
-*/
 static int vega10_smu_init(struct pp_smumgr *smumgr)
 {
 	struct vega10_smumgr *priv;
···
 			kfree(smumgr->backend);
 			cgs_free_gpu_mem(smumgr->device,
 			(cgs_handle_t)handle);
-			return -1);
+			return -EINVAL);
 
 	priv->smu_tables.entry[PPTABLE].version = 0x01;
 	priv->smu_tables.entry[PPTABLE].size = sizeof(PPTable_t);
···
 			(cgs_handle_t)priv->smu_tables.entry[PPTABLE].handle);
 			cgs_free_gpu_mem(smumgr->device,
 			(cgs_handle_t)handle);
-			return -1);
+			return -EINVAL);
 
 	priv->smu_tables.entry[WMTABLE].version = 0x01;
 	priv->smu_tables.entry[WMTABLE].size = sizeof(Watermarks_t);
···
 			(cgs_handle_t)priv->smu_tables.entry[WMTABLE].handle);
 			cgs_free_gpu_mem(smumgr->device,
 			(cgs_handle_t)handle);
-			return -1);
+			return -EINVAL);
 
 	priv->smu_tables.entry[AVFSTABLE].version = 0x01;
 	priv->smu_tables.entry[AVFSTABLE].size = sizeof(AvfsTable_t);
···
 	priv->smu_tables.entry[AVFSTABLE].table = kaddr;
 	priv->smu_tables.entry[AVFSTABLE].handle = handle;
 
-	tools_size = 0;
+	tools_size = 0x19000;
 	if (tools_size) {
 		smu_allocate_memory(smumgr->device,
 				tools_size,
···
 					smu_lower_32_bits(mc_addr);
 			priv->smu_tables.entry[TOOLSTABLE].table = kaddr;
 			priv->smu_tables.entry[TOOLSTABLE].handle = handle;
+			vega10_set_tools_address(smumgr);
 		}
 	}
+
+	/* allocate space for AVFS Fuse table */
+	smu_allocate_memory(smumgr->device,
+			sizeof(AvfsFuseOverride_t),
+			CGS_GPU_MEM_TYPE__VISIBLE_CONTIG_FB,
+			PAGE_SIZE,
+			&mc_addr,
+			&kaddr,
+			&handle);
+
+	PP_ASSERT_WITH_CODE(kaddr,
+			"[vega10_smu_init] Out of memory for avfs fuse table.",
+			kfree(smumgr->backend);
+			cgs_free_gpu_mem(smumgr->device,
+			(cgs_handle_t)priv->smu_tables.entry[PPTABLE].handle);
+			cgs_free_gpu_mem(smumgr->device,
+			(cgs_handle_t)priv->smu_tables.entry[WMTABLE].handle);
+			cgs_free_gpu_mem(smumgr->device,
+			(cgs_handle_t)priv->smu_tables.entry[AVFSTABLE].handle);
+			cgs_free_gpu_mem(smumgr->device,
+			(cgs_handle_t)priv->smu_tables.entry[TOOLSTABLE].handle);
+			cgs_free_gpu_mem(smumgr->device,
+			(cgs_handle_t)handle);
+			return -EINVAL);
+
+	priv->smu_tables.entry[AVFSFUSETABLE].version = 0x01;
+	priv->smu_tables.entry[AVFSFUSETABLE].size = sizeof(AvfsFuseOverride_t);
+	priv->smu_tables.entry[AVFSFUSETABLE].table_id = TABLE_AVFS_FUSE_OVERRIDE;
+	priv->smu_tables.entry[AVFSFUSETABLE].table_addr_high =
+			smu_upper_32_bits(mc_addr);
+	priv->smu_tables.entry[AVFSFUSETABLE].table_addr_low =
+			smu_lower_32_bits(mc_addr);
+	priv->smu_tables.entry[AVFSFUSETABLE].table = kaddr;
+	priv->smu_tables.entry[AVFSFUSETABLE].handle = handle;
 
 	return 0;
 }
···
 		if (priv->smu_tables.entry[TOOLSTABLE].table)
 			cgs_free_gpu_mem(smumgr->device,
 					(cgs_handle_t)priv->smu_tables.entry[TOOLSTABLE].handle);
+		cgs_free_gpu_mem(smumgr->device,
+				(cgs_handle_t)priv->smu_tables.entry[AVFSFUSETABLE].handle);
 		kfree(smumgr->backend);
 		smumgr->backend = NULL;
 	}
···
 {
 	PP_ASSERT_WITH_CODE(!vega10_verify_smc_interface(smumgr),
 			"Failed to verify SMC interface!",
-			return -1);
+			return -EINVAL);
 	return 0;
 }
 

drivers/gpu/drm/amd/powerplay/smumgr/vega10_smumgr.h

···
 	WMTABLE,
 	AVFSTABLE,
 	TOOLSTABLE,
+	AVFSFUSETABLE,
 	MAX_SMU_TABLE,
 };
 
···
 		uint32_t *features_enabled);
 int vega10_save_vft_table(struct pp_smumgr *smumgr, uint8_t *avfs_table);
 int vega10_restore_vft_table(struct pp_smumgr *smumgr, uint8_t *avfs_table);
-int vega10_perform_btc(struct pp_smumgr *smumgr);
 
 int vega10_set_tools_address(struct pp_smumgr *smumgr);
 

drivers/gpu/drm/amd/scheduler/gpu_scheduler.c

···
 	dma_fence_put(f);
 }
 
+bool amd_sched_dependency_optimized(struct dma_fence* fence,
+				    struct amd_sched_entity *entity)
+{
+	struct amd_gpu_scheduler *sched = entity->sched;
+	struct amd_sched_fence *s_fence;
+
+	if (!fence || dma_fence_is_signaled(fence))
+		return false;
+	if (fence->context == entity->fence_context)
+		return true;
+	s_fence = to_amd_sched_fence(fence);
+	if (s_fence && s_fence->sched == sched)
+		return true;
+
+	return false;
+}
+
 static bool amd_sched_entity_add_dependency_cb(struct amd_sched_entity *entity)
 {
 	struct amd_gpu_scheduler *sched = entity->sched;
···
 
 	spin_lock(&sched->job_list_lock);
 	list_for_each_entry_reverse(s_job, &sched->ring_mirror_list, node) {
-		if (dma_fence_remove_callback(s_job->s_fence->parent, &s_job->s_fence->cb)) {
+		if (s_job->s_fence->parent &&
+		    dma_fence_remove_callback(s_job->s_fence->parent,
+					      &s_job->s_fence->cb)) {
 			dma_fence_put(s_job->s_fence->parent);
 			s_job->s_fence->parent = NULL;
 		}
···
 	job->sched = sched;
 	job->s_entity = entity;
 	job->s_fence = amd_sched_fence_create(entity, owner);
-	job->id = atomic64_inc_return(&sched->job_id_count);
 	if (!job->s_fence)
 		return -ENOMEM;
+	job->id = atomic64_inc_return(&sched->job_id_count);
 
 	INIT_WORK(&job->finish_work, amd_sched_job_finish);
 	INIT_LIST_HEAD(&job->node);

drivers/gpu/drm/amd/scheduler/gpu_scheduler.h

···
 		       void *owner);
 void amd_sched_hw_job_reset(struct amd_gpu_scheduler *sched);
 void amd_sched_job_recovery(struct amd_gpu_scheduler *sched);
+bool amd_sched_dependency_optimized(struct dma_fence* fence,
+				    struct amd_sched_entity *entity);
 #endif

drivers/gpu/drm/drm_edid.c

···
 #define EDID_QUIRK_FORCE_12BPC			(1 << 9)
 /* Force 6bpc */
 #define EDID_QUIRK_FORCE_6BPC			(1 << 10)
+/* Force 10bpc */
+#define EDID_QUIRK_FORCE_10BPC			(1 << 11)
 
 struct detailed_mode_closure {
 	struct drm_connector *connector;
···
 	/* Funai Electronics PM36B */
 	{ "FCM", 13600, EDID_QUIRK_PREFER_LARGE_75 |
 	  EDID_QUIRK_DETAILED_IN_CM },
+
+	/* LGD panel of HP zBook 17 G2, eDP 10 bpc, but reports unknown bpc */
+	{ "LGD", 764, EDID_QUIRK_FORCE_10BPC },
 
 	/* LG Philips LCD LP154W01-A5 */
 	{ "LPL", 0, EDID_QUIRK_DETAILED_USE_MAXIMUM_SIZE },
···
 
 	if (quirks & EDID_QUIRK_FORCE_8BPC)
 		connector->display_info.bpc = 8;
+
+	if (quirks & EDID_QUIRK_FORCE_10BPC)
+		connector->display_info.bpc = 10;
 
 	if (quirks & EDID_QUIRK_FORCE_12BPC)
 		connector->display_info.bpc = 12;

drivers/gpu/drm/i915/Kconfig.debug

···
           and also analyze the request dependency resolving timeline.
 
           If in doubt, say "N".
+
+config DRM_I915_DEBUG_VBLANK_EVADE
+	bool "Enable extra debug warnings for vblank evasion"
+	depends on DRM_I915
+	default n
+	help
+	  Choose this option to turn on extra debug warnings for the
+	  vblank evade mechanism. This gives a warning every time the
+	  the deadline allotted for the vblank evade critical section
+	  is exceeded, even if there isn't an actual risk of missing
+	  the vblank.
+
+	  If in doubt, say "N".

drivers/gpu/drm/i915/intel_sprite.c

···
 			  ktime_us_delta(end_vbl_time, crtc->debug.start_vbl_time),
 			  crtc->debug.min_vbl, crtc->debug.max_vbl,
 			  crtc->debug.scanline_start, scanline_end);
-	} else if (ktime_us_delta(end_vbl_time, crtc->debug.start_vbl_time) >
-		   VBLANK_EVASION_TIME_US)
+	}
+#ifdef CONFIG_DRM_I915_DEBUG_VBLANK_EVADE
+	else if (ktime_us_delta(end_vbl_time, crtc->debug.start_vbl_time) >
+		 VBLANK_EVASION_TIME_US)
 		DRM_WARN("Atomic update on pipe (%c) took %lld us, max time under evasion is %u us\n",
 			 pipe_name(pipe),
 			 ktime_us_delta(end_vbl_time, crtc->debug.start_vbl_time),
 			 VBLANK_EVASION_TIME_US);
+#endif
 }
 
 static void

drivers/gpu/drm/nouveau/nv50_display.c

···
 static int
 nv50_wndw_atomic_check_acquire(struct nv50_wndw *wndw,
 			       struct nv50_wndw_atom *asyw,
-			       struct nv50_head_atom *asyh,
-			       u32 pflip_flags)
+			       struct nv50_head_atom *asyh)
 {
 	struct nouveau_framebuffer *fb = nouveau_framebuffer(asyw->state.fb);
 	struct nouveau_drm *drm = nouveau_drm(wndw->plane.dev);
···
 	asyw->image.h = fb->base.height;
 	asyw->image.kind = (fb->nvbo->tile_flags & 0x0000ff00) >> 8;
 
-	asyw->interval = pflip_flags & DRM_MODE_PAGE_FLIP_ASYNC ? 0 : 1;
+	if (asyh->state.pageflip_flags & DRM_MODE_PAGE_FLIP_ASYNC)
+		asyw->interval = 0;
+	else
+		asyw->interval = 1;
 
 	if (asyw->image.kind) {
 		asyw->image.layout = 0;
···
 	struct nv50_head_atom *harm = NULL, *asyh = NULL;
 	bool varm = false, asyv = false, asym = false;
 	int ret;
-	u32 pflip_flags = 0;
 
 	NV_ATOMIC(drm, "%s atomic_check\n", plane->name);
 	if (asyw->state.crtc) {
···
 			return PTR_ERR(asyh);
 		asym = drm_atomic_crtc_needs_modeset(&asyh->state);
 		asyv = asyh->state.active;
-		pflip_flags = asyh->state.pageflip_flags;
 	}
 
 	if (armw->state.crtc) {
···
 		if (memcmp(&armw->point, &asyw->point, sizeof(asyw->point)))
 			asyw->set.point = true;
 
-		if (!varm || asym || armw->state.fb != asyw->state.fb) {
-			ret = nv50_wndw_atomic_check_acquire(
-					wndw, asyw, asyh, pflip_flags);
-			if (ret)
-				return ret;
-		}
+		ret = nv50_wndw_atomic_check_acquire(wndw, asyw, asyh);
+		if (ret)
+			return ret;
 	} else
 	if (varm) {
 		nv50_wndw_atomic_check_release(wndw, asyw, harm);
···
 nv50_curs_prepare(struct nv50_wndw *wndw, struct nv50_head_atom *asyh,
 		  struct nv50_wndw_atom *asyw)
 {
-	asyh->curs.handle = nv50_disp(wndw->plane.dev)->mast.base.vram.handle;
-	asyh->curs.offset = asyw->image.offset;
-	asyh->set.curs = asyh->curs.visible;
+	u32 handle = nv50_disp(wndw->plane.dev)->mast.base.vram.handle;
+	u32 offset = asyw->image.offset;
+	if (asyh->curs.handle != handle || asyh->curs.offset != offset) {
+		asyh->curs.handle = handle;
+		asyh->curs.offset = offset;
+		asyh->set.curs = asyh->curs.visible;
+	}
 }
 
 static void

drivers/gpu/drm/nouveau/nvkm/core/object.c

···
 	INIT_LIST_HEAD(&object->head);
 	INIT_LIST_HEAD(&object->tree);
 	RB_CLEAR_NODE(&object->node);
-	WARN_ON(oclass->engine && !object->engine);
+	WARN_ON(IS_ERR(object->engine));
 }
 
 int

drivers/gpu/drm/nouveau/nvkm/subdev/fb/ramgf100.c

···
 			return ret;
 	}
 
-	ram->ranks = (nvkm_rd32(device, 0x10f200) & 0x00000004) ? 2 : 1;
 	return 0;
 }
 

drivers/gpu/drm/nouveau/nvkm/subdev/therm/base.c

···
 		poll = false;
 	}
 
-	if (list_empty(&therm->alarm.head) && poll)
+	if (poll)
 		nvkm_timer_alarm(tmr, 1000000000ULL, &therm->alarm);
 	spin_unlock_irqrestore(&therm->lock, flags);
 

drivers/gpu/drm/nouveau/nvkm/subdev/therm/fan.c

···
 	spin_unlock_irqrestore(&fan->lock, flags);
 
 	/* schedule next fan update, if not at target speed already */
-	if (list_empty(&fan->alarm.head) && target != duty) {
+	if (target != duty) {
 		u16 bump_period = fan->bios.bump_period;
 		u16 slow_down_period = fan->bios.slow_down_period;
 		u64 delay;

drivers/gpu/drm/nouveau/nvkm/subdev/therm/fantog.c

···
 	duty = !nvkm_gpio_get(gpio, 0, DCB_GPIO_FAN, 0xff);
 	nvkm_gpio_set(gpio, 0, DCB_GPIO_FAN, 0xff, duty);
 
-	if (list_empty(&fan->alarm.head) && percent != (duty * 100)) {
+	if (percent != (duty * 100)) {
 		u64 next_change = (percent * fan->period_us) / 100;
 		if (!duty)
 			next_change = fan->period_us - next_change;

drivers/gpu/drm/nouveau/nvkm/subdev/therm/temp.c

···
 	spin_unlock_irqrestore(&therm->sensor.alarm_program_lock, flags);
 
 	/* schedule the next poll in one second */
-	if (therm->func->temp_get(therm) >= 0 && list_empty(&alarm->head))
+	if (therm->func->temp_get(therm) >= 0)
 		nvkm_timer_alarm(tmr, 1000000000ULL, alarm);
 }
 

drivers/gpu/drm/nouveau/nvkm/subdev/timer/base.c

···
 	unsigned long flags;
 	LIST_HEAD(exec);
 
-	/* move any due alarms off the pending list */
+	/* Process pending alarms. */
 	spin_lock_irqsave(&tmr->lock, flags);
 	list_for_each_entry_safe(alarm, atemp, &tmr->alarms, head) {
-		if (alarm->timestamp <= nvkm_timer_read(tmr))
-			list_move_tail(&alarm->head, &exec);
+		/* Have we hit the earliest alarm that hasn't gone off? */
+		if (alarm->timestamp > nvkm_timer_read(tmr)) {
+			/* Schedule it.  If we didn't race, we're done. */
+			tmr->func->alarm_init(tmr, alarm->timestamp);
+			if (alarm->timestamp > nvkm_timer_read(tmr))
+				break;
+		}
+
+		/* Move to completed list.  We'll drop the lock before
+		 * executing the callback so it can reschedule itself.
+		 */
+		list_move_tail(&alarm->head, &exec);
 	}
 
-	/* reschedule interrupt for next alarm time */
-	if (!list_empty(&tmr->alarms)) {
-		alarm = list_first_entry(&tmr->alarms, typeof(*alarm), head);
-		tmr->func->alarm_init(tmr, alarm->timestamp);
-	} else {
+	/* Shut down interrupt if no more pending alarms. */
+	if (list_empty(&tmr->alarms))
 		tmr->func->alarm_fini(tmr);
-	}
 	spin_unlock_irqrestore(&tmr->lock, flags);
 
-	/* execute any pending alarm handlers */
+	/* Execute completed callbacks. */
 	list_for_each_entry_safe(alarm, atemp, &exec, head) {
 		list_del_init(&alarm->head);
 		alarm->func(alarm);
···
 	struct nvkm_alarm *list;
 	unsigned long flags;
 
-	alarm->timestamp = nvkm_timer_read(tmr) + nsec;
-
-	/* append new alarm to list, in soonest-alarm-first order */
+	/* Remove alarm from pending list.
+	 *
+	 * This both protects against the corruption of the list,
+	 * and implements alarm rescheduling/cancellation.
+	 */
 	spin_lock_irqsave(&tmr->lock, flags);
-	if (!nsec) {
-		if (!list_empty(&alarm->head))
-			list_del(&alarm->head);
-	} else {
+	list_del_init(&alarm->head);
+
+	if (nsec) {
+		/* Insert into pending list, ordered earliest to latest. */
+		alarm->timestamp = nvkm_timer_read(tmr) + nsec;
 		list_for_each_entry(list, &tmr->alarms, head) {
 			if (list->timestamp > alarm->timestamp)
 				break;
 		}
+
 		list_add_tail(&alarm->head, &list->head);
+
+		/* Update HW if this is now the earliest alarm. */
+		list = list_first_entry(&tmr->alarms, typeof(*list), head);
+		if (list == alarm) {
+			tmr->func->alarm_init(tmr, alarm->timestamp);
+			/* This shouldn't happen if callers aren't stupid.
+			 *
+			 * Worst case scenario is that it'll take roughly
+			 * 4 seconds for the next alarm to trigger.
+			 */
+			WARN_ON(alarm->timestamp <= nvkm_timer_read(tmr));
+		}
 	}
 	spin_unlock_irqrestore(&tmr->lock, flags);
-
-	/* process pending alarms */
-	nvkm_timer_alarm_trigger(tmr);
 }
 
 void

drivers/gpu/drm/nouveau/nvkm/subdev/timer/nv04.c

···
 	u32 stat = nvkm_rd32(device, NV04_PTIMER_INTR_0);
 
 	if (stat & 0x00000001) {
-		nvkm_timer_alarm_trigger(tmr);
 		nvkm_wr32(device, NV04_PTIMER_INTR_0, 0x00000001);
+		nvkm_timer_alarm_trigger(tmr);
 		stat &= ~0x00000001;
 	}
 

drivers/gpu/drm/radeon/cik.c

···
 	a.full = dfixed_const(available_bandwidth);
 	b.full = dfixed_const(wm->num_heads);
 	a.full = dfixed_div(a, b);
+	tmp = div_u64((u64) dmif_size * (u64) wm->disp_clk, mc_latency + 512);
+	tmp = min(dfixed_trunc(a), tmp);
 
-	b.full = dfixed_const(mc_latency + 512);
-	c.full = dfixed_const(wm->disp_clk);
-	b.full = dfixed_div(b, c);
-
-	c.full = dfixed_const(dmif_size);
-	b.full = dfixed_div(c, b);
-
-	tmp = min(dfixed_trunc(a), dfixed_trunc(b));
-
-	b.full = dfixed_const(1000);
-	c.full = dfixed_const(wm->disp_clk);
-	b.full = dfixed_div(c, b);
-	c.full = dfixed_const(wm->bytes_per_pixel);
-	b.full = dfixed_mul(b, c);
-
-	lb_fill_bw = min(tmp, dfixed_trunc(b));
+	lb_fill_bw = min(tmp, wm->disp_clk * wm->bytes_per_pixel / 1000);
 
 	a.full = dfixed_const(max_src_lines_per_dst_line * wm->src_width * wm->bytes_per_pixel);
 	b.full = dfixed_const(1000);
···
 {
 	struct drm_display_mode *mode = &radeon_crtc->base.mode;
 	struct dce8_wm_params wm_low, wm_high;
-	u32 pixel_period;
+	u32 active_time;
 	u32 line_time = 0;
 	u32 latency_watermark_a = 0, latency_watermark_b = 0;
 	u32 tmp, wm_mask;
 
 	if (radeon_crtc->base.enabled && num_heads && mode) {
-		pixel_period = 1000000 / (u32)mode->clock;
-		line_time = min((u32)mode->crtc_htotal * pixel_period, (u32)65535);
+		active_time = 1000000UL * (u32)mode->crtc_hdisplay / (u32)mode->clock;
+		line_time = min((u32) (1000000UL * (u32)mode->crtc_htotal / (u32)mode->clock), (u32)65535);
 
 		/* watermark for high clocks */
 		if ((rdev->pm.pm_method == PM_METHOD_DPM) &&
···
 
 		wm_high.disp_clk = mode->clock;
 		wm_high.src_width = mode->crtc_hdisplay;
-		wm_high.active_time = mode->crtc_hdisplay * pixel_period;
+		wm_high.active_time = active_time;
 		wm_high.blank_time = line_time - wm_high.active_time;
 		wm_high.interlaced = false;
 		if (mode->flags & DRM_MODE_FLAG_INTERLACE)
···
 
 		wm_low.disp_clk = mode->clock;
 		wm_low.src_width = mode->crtc_hdisplay;
-		wm_low.active_time = mode->crtc_hdisplay * pixel_period;
+		wm_low.active_time = active_time;
 		wm_low.blank_time = line_time - wm_low.active_time;
 		wm_low.interlaced = false;
 		if (mode->flags & DRM_MODE_FLAG_INTERLACE)

drivers/gpu/drm/radeon/evergreen.c

···
 	b.full = dfixed_const(wm->num_heads);
 	a.full = dfixed_div(a, b);
 
-	b.full = dfixed_const(1000);
-	c.full = dfixed_const(wm->disp_clk);
-	b.full = dfixed_div(c, b);
-	c.full = dfixed_const(wm->bytes_per_pixel);
-	b.full = dfixed_mul(b, c);
-
-	lb_fill_bw = min(dfixed_trunc(a), dfixed_trunc(b));
+	lb_fill_bw = min(dfixed_trunc(a), wm->disp_clk * wm->bytes_per_pixel / 1000);
 
 	a.full = dfixed_const(max_src_lines_per_dst_line * wm->src_width * wm->bytes_per_pixel);
 	b.full = dfixed_const(1000);
···
 	struct drm_display_mode *mode = &radeon_crtc->base.mode;
 	struct evergreen_wm_params wm_low, wm_high;
 	u32 dram_channels;
-	u32 pixel_period;
+	u32 active_time;
 	u32 line_time = 0;
 	u32 latency_watermark_a = 0, latency_watermark_b = 0;
 	u32 priority_a_mark = 0, priority_b_mark = 0;
···
 	fixed20_12 a, b, c;
 
 	if (radeon_crtc->base.enabled && num_heads && mode) {
-		pixel_period = 1000000 / (u32)mode->clock;
-		line_time = min((u32)mode->crtc_htotal * pixel_period, (u32)65535);
+		active_time = 1000000UL * (u32)mode->crtc_hdisplay / (u32)mode->clock;
+		line_time = min((u32) (1000000UL * (u32)mode->crtc_htotal / (u32)mode->clock), (u32)65535);
 		priority_a_cnt = 0;
 		priority_b_cnt = 0;
 		dram_channels = evergreen_get_number_of_dram_channels(rdev);
···
 
 		wm_high.disp_clk = mode->clock;
 		wm_high.src_width = mode->crtc_hdisplay;
-		wm_high.active_time = mode->crtc_hdisplay * pixel_period;
+		wm_high.active_time = active_time;
 		wm_high.blank_time = line_time - wm_high.active_time;
 		wm_high.interlaced = false;
 		if (mode->flags & DRM_MODE_FLAG_INTERLACE)
···
 
 		wm_low.disp_clk = mode->clock;
 		wm_low.src_width = mode->crtc_hdisplay;
-		wm_low.active_time = mode->crtc_hdisplay * pixel_period;
+		wm_low.active_time = active_time;
 		wm_low.blank_time = line_time - wm_low.active_time;
 		wm_low.interlaced = false;
 		if (mode->flags & DRM_MODE_FLAG_INTERLACE)

drivers/gpu/drm/radeon/r420.c

···
 
 static void r420_cp_errata_init(struct radeon_device *rdev)
 {
+	int r;
 	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
 
 	/* RV410 and R420 can lock up if CP DMA to host memory happens
···
 	 * of the CP init, apparently.
 	 */
 	radeon_scratch_get(rdev, &rdev->config.r300.resync_scratch);
-	radeon_ring_lock(rdev, ring, 8);
+	r = radeon_ring_lock(rdev, ring, 8);
+	WARN_ON(r);
 	radeon_ring_write(ring, PACKET0(R300_CP_RESYNC_ADDR, 1));
 	radeon_ring_write(ring, rdev->config.r300.resync_scratch);
 	radeon_ring_write(ring, 0xDEADBEEF);
···
 
 static void r420_cp_errata_fini(struct radeon_device *rdev)
 {
+	int r;
 	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
 
 	/* Catch the RESYNC we dispatched all the way back,
 	 * at the very beginning of the CP init.
 	 */
-	radeon_ring_lock(rdev, ring, 8);
+	r = radeon_ring_lock(rdev, ring, 8);
+	WARN_ON(r);
 	radeon_ring_write(ring, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
 	radeon_ring_write(ring, R300_RB3D_DC_FINISH);
 	radeon_ring_unlock_commit(rdev, ring, false);

drivers/gpu/drm/radeon/radeon_cs.c

···
 		priority = (r->flags & RADEON_RELOC_PRIO_MASK) * 2
 			   + !!r->write_domain;
 
-		/* the first reloc of an UVD job is the msg and that must be in
-		   VRAM, also but everything into VRAM on AGP cards and older
-		   IGP chips to avoid image corruptions */
+		/* The first reloc of an UVD job is the msg and that must be in
+		 * VRAM, the second reloc is the DPB and for WMV that must be in
+		 * VRAM as well. Also put everything into VRAM on AGP cards and older
+		 * IGP chips to avoid image corruptions
+		 */
 		if (p->ring == R600_RING_TYPE_UVD_INDEX &&
-		    (i == 0 || pci_find_capability(p->rdev->ddev->pdev,
+		    (i <= 0 || pci_find_capability(p->rdev->ddev->pdev,
 						   PCI_CAP_ID_AGP) ||
 		     p->rdev->family == CHIP_RS780 ||
 		     p->rdev->family == CHIP_RS880)) {

drivers/gpu/drm/radeon/radeon_object.c

···
 	list_del_init(&bo->list);
 	mutex_unlock(&bo->rdev->gem.mutex);
 	radeon_bo_clear_surface_reg(bo);
-	WARN_ON(!list_empty(&bo->va));
+	WARN_ON_ONCE(!list_empty(&bo->va));
 	drm_gem_object_release(&bo->gem_base);
 	kfree(bo);
 }

drivers/gpu/drm/radeon/radeon_test.c

···
 			DRM_ERROR("Failed to lock ring A %d\n", ring->idx);
 			return r;
 		}
-		radeon_fence_emit(rdev, fence, ring->idx);
+		r = radeon_fence_emit(rdev, fence, ring->idx);
+		if (r) {
+			DRM_ERROR("Failed to emit fence\n");
+			radeon_ring_unlock_undo(rdev, ring);
+			return r;
+		}
 		radeon_ring_unlock_commit(rdev, ring, false);
 	}
 	return 0;

drivers/gpu/drm/radeon/radeon_uvd.c

···
 	}
 
 	/* TODO: is this still necessary on NI+ ? */
-	if ((cmd == 0 || cmd == 0x3) &&
+	if ((cmd == 0 || cmd == 1 || cmd == 0x3) &&
 	    (start >> 28) != (p->rdev->uvd.gpu_addr >> 28)) {
 		DRM_ERROR("msg/fb buffer %LX-%LX out of 256MB segment!\n",
 			  start, end);

drivers/gpu/drm/radeon/si.c

···
 	a.full = dfixed_const(available_bandwidth);
 	b.full = dfixed_const(wm->num_heads);
 	a.full = dfixed_div(a, b);
+	tmp = div_u64((u64) dmif_size * (u64) wm->disp_clk, mc_latency + 512);
+	tmp = min(dfixed_trunc(a), tmp);
 
-	b.full = dfixed_const(mc_latency + 512);
-	c.full = dfixed_const(wm->disp_clk);
-	b.full = dfixed_div(b, c);
-
-	c.full = dfixed_const(dmif_size);
-	b.full = dfixed_div(c, b);
-
-	tmp = min(dfixed_trunc(a), dfixed_trunc(b));
-
-	b.full = dfixed_const(1000);
-	c.full = dfixed_const(wm->disp_clk);
-	b.full = dfixed_div(c, b);
-	c.full = dfixed_const(wm->bytes_per_pixel);
-	b.full = dfixed_mul(b, c);
-
-	lb_fill_bw = min(tmp, dfixed_trunc(b));
+	lb_fill_bw = min(tmp, wm->disp_clk * wm->bytes_per_pixel / 1000);
 
 	a.full = dfixed_const(max_src_lines_per_dst_line * wm->src_width * wm->bytes_per_pixel);
 	b.full = dfixed_const(1000);
···
 	struct drm_display_mode *mode = &radeon_crtc->base.mode;
 	struct dce6_wm_params wm_low, wm_high;
 	u32 dram_channels;
-	u32 pixel_period;
+	u32 active_time;
 	u32 line_time = 0;
 	u32 latency_watermark_a = 0, latency_watermark_b = 0;
 	u32 priority_a_mark = 0, priority_b_mark = 0;
···
 	fixed20_12 a, b, c;
 
 	if (radeon_crtc->base.enabled && num_heads && mode) {
-		pixel_period = 1000000 / (u32)mode->clock;
-		line_time = min((u32)mode->crtc_htotal * pixel_period, (u32)65535);
+		active_time = 1000000UL * (u32)mode->crtc_hdisplay / (u32)mode->clock;
+		line_time = min((u32) (1000000UL * (u32)mode->crtc_htotal / (u32)mode->clock), (u32)65535);
 		priority_a_cnt = 0;
 		priority_b_cnt = 0;
 
···
 
 		wm_high.disp_clk = mode->clock;
 		wm_high.src_width = mode->crtc_hdisplay;
-		wm_high.active_time = mode->crtc_hdisplay * pixel_period;
+		wm_high.active_time = active_time;
 		wm_high.blank_time = line_time - wm_high.active_time;
 		wm_high.interlaced = false;
 		if (mode->flags & DRM_MODE_FLAG_INTERLACE)
···
 
 		wm_low.disp_clk = mode->clock;
 		wm_low.src_width = mode->crtc_hdisplay;
-		wm_low.active_time = mode->crtc_hdisplay * pixel_period;
+		wm_low.active_time = active_time;
 		wm_low.blank_time = line_time - wm_low.active_time;
 		wm_low.interlaced = false;
 		if (mode->flags & DRM_MODE_FLAG_INTERLACE)

drivers/gpu/drm/ttm/ttm_bo.c

···
 int ttm_bo_init_mm(struct ttm_bo_device *bdev, unsigned type,
 			unsigned long p_size)
 {
-	int ret = -EINVAL;
+	int ret;
 	struct ttm_mem_type_manager *man;
 	unsigned i;
 
···
 		return ret;
 	man->bdev = bdev;
 
-	ret = 0;
 	if (type != TTM_PL_SYSTEM) {
 		ret = (*man->func->init)(man, p_size);
 		if (ret)

include/uapi/drm/amdgpu_drm.h

···
 };
 
 struct drm_amdgpu_wait_cs_in {
-	/** Command submission handle */
+	/* Command submission handle
+         * handle equals 0 means none to wait for
+         * handle equals ~0ull means wait for the latest sequence number
+         */
 	__u64 handle;
 	/** Absolute timeout to wait */
 	__u64 timeout;
···
 	__u64 cntl_sb_buf_gpu_addr;
 	/* NGG Parameter Cache */
 	__u64 param_buf_gpu_addr;
+	__u32 prim_buf_size;
+	__u32 pos_buf_size;
+	__u32 cntl_sb_buf_size;
+	__u32 param_buf_size;
+	/* wavefront size*/
+	__u32 wave_front_size;
+	/* shader visible vgprs*/
+	__u32 num_shader_visible_vgprs;
+	/* CU per shader array*/
+	__u32 num_cu_per_sh;
+	/* number of tcc blocks*/
+	__u32 num_tcc_blocks;
+	/* gs vgt table depth*/
+	__u32 gs_vgt_table_depth;
+	/* gs primitive buffer depth*/
+	__u32 gs_prim_buffer_depth;
+	/* max gs wavefront per vgt*/
+	__u32 max_gs_waves_per_vgt;
+	__u32 _pad1;
 };
 
 struct drm_amdgpu_info_hw_ip {