Merge branch 'drm-next-4.15' of git://people.freedesktop.org/~agd5f/linux into drm-next

First feature pull for 4.15.  Highlights:
- Per VM BO support
- Lots of powerplay cleanups
- Powerplay support for CI
- pasid mgr for kfd
- interrupt infrastructure for recoverable page faults
- SR-IOV fixes
- initial GPU reset for vega10
- prime mmap support
- ttm page table debugging improvements
- lots of bug fixes

* 'drm-next-4.15' of git://people.freedesktop.org/~agd5f/linux: (232 commits)
  drm/amdgpu: clarify license in amdgpu_trace_points.c
  drm/amdgpu: Add gem_prime_mmap support
  drm/amd/powerplay: delete dead code in smumgr
  drm/amd/powerplay: delete SMUM_FIELD_MASK
  drm/amd/powerplay: delete SMUM_WAIT_INDIRECT_FIELD
  drm/amd/powerplay: delete SMUM_READ_FIELD
  drm/amd/powerplay: delete SMUM_SET_FIELD
  drm/amd/powerplay: delete SMUM_READ_VFPF_INDIRECT_FIELD
  drm/amd/powerplay: delete SMUM_WRITE_VFPF_INDIRECT_FIELD
  drm/amd/powerplay: delete SMUM_WRITE_FIELD
  drm/amd/powerplay: delete SMU_WRITE_INDIRECT_FIELD
  drm/amd/powerplay: move macros to hwmgr.h
  drm/amd/powerplay: move PHM_WAIT_VFPF_INDIRECT_FIELD to hwmgr.h
  drm/amd/powerplay: move SMUM_WAIT_VFPF_INDIRECT_FIELD_UNEQUAL to hwmgr.h
  drm/amd/powerplay: move SMUM_WAIT_INDIRECT_FIELD_UNEQUAL to hwmgr.h
  drm/amd/powerplay: add new helper functions in hwmgr.h
  drm/amd/powerplay: use SMU_IND_INDEX/DATA_11 pair
  drm/amd/powerplay: refine powerplay code.
  drm/amd/powerplay: delete dead code in hwmgr.h
  drm/amd/powerplay: refine interface in struct pp_smumgr_func
  ...

drivers/gpu/drm/Kconfig

@@ -184,6 +184,7 @@ config DRM_AMDGPU
 	select BACKLIGHT_CLASS_DEVICE
 	select BACKLIGHT_LCD_SUPPORT
 	select INTERVAL_TREE
+	select CHASH
 	help
 	  Choose this option if you have a recent AMD Radeon graphics card.
 
@@ -191,6 +192,8 @@ config DRM_AMDGPU
 
 source "drivers/gpu/drm/amd/amdgpu/Kconfig"
 
+source "drivers/gpu/drm/amd/lib/Kconfig"
+
 source "drivers/gpu/drm/nouveau/Kconfig"
 
 source "drivers/gpu/drm/i915/Kconfig"

drivers/gpu/drm/Makefile

@@ -50,6 +50,7 @@ obj-$(CONFIG_DRM_ARM)	+= arm/
 obj-$(CONFIG_DRM_TTM)	+= ttm/
 obj-$(CONFIG_DRM_TDFX)	+= tdfx/
 obj-$(CONFIG_DRM_R128)	+= r128/
+obj-y			+= amd/lib/
 obj-$(CONFIG_HSA_AMD) += amd/amdkfd/
 obj-$(CONFIG_DRM_RADEON)+= radeon/
 obj-$(CONFIG_DRM_AMDGPU)+= amd/amdgpu/

drivers/gpu/drm/amd/amdgpu/Makefile

@@ -133,5 +133,3 @@ include $(FULL_AMD_PATH)/powerplay/Makefile
 amdgpu-y += $(AMD_POWERPLAY_FILES)
 
 obj-$(CONFIG_DRM_AMDGPU)+= amdgpu.o
-
-CFLAGS_amdgpu_trace_points.o := -I$(src)

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

@@ -65,6 +65,7 @@
 #include "amdgpu_uvd.h"
 #include "amdgpu_vce.h"
 #include "amdgpu_vcn.h"
+#include "amdgpu_mn.h"
 
 #include "gpu_scheduler.h"
 #include "amdgpu_virt.h"
@@ -91,7 +92,7 @@ extern int amdgpu_dpm;
 extern int amdgpu_fw_load_type;
 extern int amdgpu_aspm;
 extern int amdgpu_runtime_pm;
-extern unsigned amdgpu_ip_block_mask;
+extern uint amdgpu_ip_block_mask;
 extern int amdgpu_bapm;
 extern int amdgpu_deep_color;
 extern int amdgpu_vm_size;
@@ -104,14 +105,14 @@ extern int amdgpu_sched_jobs;
 extern int amdgpu_sched_hw_submission;
 extern int amdgpu_no_evict;
 extern int amdgpu_direct_gma_size;
-extern unsigned amdgpu_pcie_gen_cap;
-extern unsigned amdgpu_pcie_lane_cap;
-extern unsigned amdgpu_cg_mask;
-extern unsigned amdgpu_pg_mask;
-extern unsigned amdgpu_sdma_phase_quantum;
+extern uint amdgpu_pcie_gen_cap;
+extern uint amdgpu_pcie_lane_cap;
+extern uint amdgpu_cg_mask;
+extern uint amdgpu_pg_mask;
+extern uint amdgpu_sdma_phase_quantum;
 extern char *amdgpu_disable_cu;
 extern char *amdgpu_virtual_display;
-extern unsigned amdgpu_pp_feature_mask;
+extern uint amdgpu_pp_feature_mask;
 extern int amdgpu_vram_page_split;
 extern int amdgpu_ngg;
 extern int amdgpu_prim_buf_per_se;
@@ -178,6 +179,7 @@ struct amdgpu_cs_parser;
 struct amdgpu_job;
 struct amdgpu_irq_src;
 struct amdgpu_fpriv;
+struct amdgpu_bo_va_mapping;
 
 enum amdgpu_cp_irq {
 	AMDGPU_CP_IRQ_GFX_EOP = 0,
@@ -292,14 +294,25 @@ struct amdgpu_buffer_funcs {
 
 /* provided by hw blocks that can write ptes, e.g., sdma */
 struct amdgpu_vm_pte_funcs {
+	/* number of dw to reserve per operation */
+	unsigned	copy_pte_num_dw;
+
 	/* copy pte entries from GART */
 	void (*copy_pte)(struct amdgpu_ib *ib,
 			 uint64_t pe, uint64_t src,
 			 unsigned count);
+
 	/* write pte one entry at a time with addr mapping */
 	void (*write_pte)(struct amdgpu_ib *ib, uint64_t pe,
 			  uint64_t value, unsigned count,
 			  uint32_t incr);
+
+	/* maximum nums of PTEs/PDEs in a single operation */
+	uint32_t	set_max_nums_pte_pde;
+
+	/* number of dw to reserve per operation */
+	unsigned	set_pte_pde_num_dw;
+
 	/* for linear pte/pde updates without addr mapping */
 	void (*set_pte_pde)(struct amdgpu_ib *ib,
 			    uint64_t pe,
@@ -332,6 +345,7 @@ struct amdgpu_gart_funcs {
 struct amdgpu_ih_funcs {
 	/* ring read/write ptr handling, called from interrupt context */
 	u32 (*get_wptr)(struct amdgpu_device *adev);
+	bool (*prescreen_iv)(struct amdgpu_device *adev);
 	void (*decode_iv)(struct amdgpu_device *adev,
 			  struct amdgpu_iv_entry *entry);
 	void (*set_rptr)(struct amdgpu_device *adev);
@@ -399,6 +413,7 @@ void amdgpu_gem_prime_unpin(struct drm_gem_object *obj);
 struct reservation_object *amdgpu_gem_prime_res_obj(struct drm_gem_object *);
 void *amdgpu_gem_prime_vmap(struct drm_gem_object *obj);
 void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr);
+int amdgpu_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma);
 int amdgpu_gem_debugfs_init(struct amdgpu_device *adev);
 
 /* sub-allocation manager, it has to be protected by another lock.
@@ -455,9 +470,10 @@ struct amdgpu_sa_bo {
  */
 void amdgpu_gem_force_release(struct amdgpu_device *adev);
 int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
-				int alignment, u32 initial_domain,
-				u64 flags, bool kernel,
-				struct drm_gem_object **obj);
+			     int alignment, u32 initial_domain,
+			     u64 flags, bool kernel,
+			     struct reservation_object *resv,
+			     struct drm_gem_object **obj);
 
 int amdgpu_mode_dumb_create(struct drm_file *file_priv,
 			    struct drm_device *dev,
@@ -731,8 +747,8 @@ struct amdgpu_ctx_mgr {
 struct amdgpu_ctx *amdgpu_ctx_get(struct amdgpu_fpriv *fpriv, uint32_t id);
 int amdgpu_ctx_put(struct amdgpu_ctx *ctx);
 
-uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring,
-			      struct dma_fence *fence);
+int amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring,
+			      struct dma_fence *fence, uint64_t *seq);
 struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx,
 				   struct amdgpu_ring *ring, uint64_t seq);
 
@@ -1014,7 +1030,6 @@ struct amdgpu_gfx {
 	/* reset mask */
 	uint32_t                        grbm_soft_reset;
 	uint32_t                        srbm_soft_reset;
-	bool                            in_reset;
 	/* s3/s4 mask */
 	bool                            in_suspend;
 	/* NGG */
@@ -1056,6 +1071,7 @@ struct amdgpu_cs_parser {
 	/* buffer objects */
 	struct ww_acquire_ctx		ticket;
 	struct amdgpu_bo_list		*bo_list;
+	struct amdgpu_mn		*mn;
 	struct amdgpu_bo_list_entry	vm_pd;
 	struct list_head		validated;
 	struct dma_fence		*fence;
@@ -1183,6 +1199,9 @@ struct amdgpu_firmware {
 
 	/* gpu info firmware data pointer */
 	const struct firmware *gpu_info_fw;
+
+	void *fw_buf_ptr;
+	uint64_t fw_buf_mc;
 };
 
 /*
@@ -1196,20 +1215,6 @@ void amdgpu_benchmark(struct amdgpu_device *adev, int test_number);
  */
 void amdgpu_test_moves(struct amdgpu_device *adev);
 
-/*
- * MMU Notifier
- */
-#if defined(CONFIG_MMU_NOTIFIER)
-int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr);
-void amdgpu_mn_unregister(struct amdgpu_bo *bo);
-#else
-static inline int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
-{
-	return -ENODEV;
-}
-static inline void amdgpu_mn_unregister(struct amdgpu_bo *bo) {}
-#endif
-
 /*
  * Debugfs
  */
@@ -1592,6 +1597,7 @@ struct amdgpu_device {
 
 	/* record last mm index being written through WREG32*/
 	unsigned long last_mm_index;
+	bool                            in_sriov_reset;
 };
 
 static inline struct amdgpu_device *amdgpu_ttm_adev(struct ttm_bo_device *bdev)
@@ -1759,6 +1765,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring)
 #define amdgpu_ring_init_cond_exec(r) (r)->funcs->init_cond_exec((r))
 #define amdgpu_ring_patch_cond_exec(r,o) (r)->funcs->patch_cond_exec((r),(o))
 #define amdgpu_ih_get_wptr(adev) (adev)->irq.ih_funcs->get_wptr((adev))
+#define amdgpu_ih_prescreen_iv(adev) (adev)->irq.ih_funcs->prescreen_iv((adev))
 #define amdgpu_ih_decode_iv(adev, iv) (adev)->irq.ih_funcs->decode_iv((adev), (iv))
 #define amdgpu_ih_set_rptr(adev) (adev)->irq.ih_funcs->set_rptr((adev))
 #define amdgpu_display_vblank_get_counter(adev, crtc) (adev)->mode_info.funcs->vblank_get_counter((adev), (crtc))
@@ -1791,18 +1798,6 @@ void amdgpu_cs_report_moved_bytes(struct amdgpu_device *adev, u64 num_bytes,
 				  u64 num_vis_bytes);
 void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain);
 bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo);
-int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages);
-int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr,
-				     uint32_t flags);
-bool amdgpu_ttm_tt_has_userptr(struct ttm_tt *ttm);
-struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm);
-bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start,
-				  unsigned long end);
-bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm,
-				       int *last_invalidated);
-bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm);
-uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm,
-				 struct ttm_mem_reg *mem);
 void amdgpu_vram_location(struct amdgpu_device *adev, struct amdgpu_mc *mc, u64 base);
 void amdgpu_gart_location(struct amdgpu_device *adev, struct amdgpu_mc *mc);
 void amdgpu_ttm_set_active_vram_size(struct amdgpu_device *adev, u64 size);
@@ -1885,10 +1880,9 @@ static inline int amdgpu_acpi_init(struct amdgpu_device *adev) { return 0; }
 static inline void amdgpu_acpi_fini(struct amdgpu_device *adev) { }
 #endif
 
-struct amdgpu_bo_va_mapping *
-amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
-		       uint64_t addr, struct amdgpu_bo **bo);
-int amdgpu_cs_sysvm_access_required(struct amdgpu_cs_parser *parser);
+int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
+			   uint64_t addr, struct amdgpu_bo **bo,
+			   struct amdgpu_bo_va_mapping **mapping);
 
 #include "amdgpu_object.h"
 #endif

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c

@@ -169,6 +169,8 @@ static const struct kfd2kgd_calls kfd2kgd = {
 	.get_vmem_size = get_vmem_size,
 	.get_gpu_clock_counter = get_gpu_clock_counter,
 	.get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz,
+	.alloc_pasid = amdgpu_vm_alloc_pasid,
+	.free_pasid = amdgpu_vm_free_pasid,
 	.program_sh_mem_settings = kgd_program_sh_mem_settings,
 	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
 	.init_pipeline = kgd_init_pipeline,

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c

@@ -128,6 +128,8 @@ static const struct kfd2kgd_calls kfd2kgd = {
 	.get_vmem_size = get_vmem_size,
 	.get_gpu_clock_counter = get_gpu_clock_counter,
 	.get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz,
+	.alloc_pasid = amdgpu_vm_alloc_pasid,
+	.free_pasid = amdgpu_vm_free_pasid,
 	.program_sh_mem_settings = kgd_program_sh_mem_settings,
 	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
 	.init_pipeline = kgd_init_pipeline,

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

@@ -45,7 +45,6 @@ struct amdgpu_cgs_device {
 static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device,
 				    enum cgs_gpu_mem_type type,
 				    uint64_t size, uint64_t align,
-				    uint64_t min_offset, uint64_t max_offset,
 				    cgs_handle_t *handle)
 {
 	CGS_FUNC_ADEV;
@@ -53,13 +52,6 @@ static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device,
 	int ret = 0;
 	uint32_t domain = 0;
 	struct amdgpu_bo *obj;
-	struct ttm_placement placement;
-	struct ttm_place place;
-
-	if (min_offset > max_offset) {
-		BUG_ON(1);
-		return -EINVAL;
-	}
 
 	/* fail if the alignment is not a power of 2 */
 	if (((align != 1) && (align & (align - 1)))
@@ -73,41 +65,19 @@ static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device,
 		flags = AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
 			AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 		domain = AMDGPU_GEM_DOMAIN_VRAM;
-		if (max_offset > adev->mc.real_vram_size)
-			return -EINVAL;
-		place.fpfn = min_offset >> PAGE_SHIFT;
-		place.lpfn = max_offset >> PAGE_SHIFT;
-		place.flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED |
-			TTM_PL_FLAG_VRAM;
 		break;
 	case CGS_GPU_MEM_TYPE__INVISIBLE_CONTIG_FB:
 	case CGS_GPU_MEM_TYPE__INVISIBLE_FB:
 		flags = AMDGPU_GEM_CREATE_NO_CPU_ACCESS |
 			AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 		domain = AMDGPU_GEM_DOMAIN_VRAM;
-		if (adev->mc.visible_vram_size < adev->mc.real_vram_size) {
-			place.fpfn =
-				max(min_offset, adev->mc.visible_vram_size) >> PAGE_SHIFT;
-			place.lpfn =
-				min(max_offset, adev->mc.real_vram_size) >> PAGE_SHIFT;
-			place.flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED |
-				TTM_PL_FLAG_VRAM;
-		}
-
 		break;
 	case CGS_GPU_MEM_TYPE__GART_CACHEABLE:
 		domain = AMDGPU_GEM_DOMAIN_GTT;
-		place.fpfn = min_offset >> PAGE_SHIFT;
-		place.lpfn = max_offset >> PAGE_SHIFT;
-		place.flags = TTM_PL_FLAG_CACHED | TTM_PL_FLAG_TT;
 		break;
 	case CGS_GPU_MEM_TYPE__GART_WRITECOMBINE:
 		flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC;
 		domain = AMDGPU_GEM_DOMAIN_GTT;
-		place.fpfn = min_offset >> PAGE_SHIFT;
-		place.lpfn = max_offset >> PAGE_SHIFT;
-		place.flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_TT |
-			TTM_PL_FLAG_UNCACHED;
 		break;
 	default:
 		return -EINVAL;
@@ -116,15 +86,8 @@ static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device,
 
 	*handle = 0;
 
-	placement.placement = &place;
-	placement.num_placement = 1;
-	placement.busy_placement = &place;
-	placement.num_busy_placement = 1;
-
-	ret = amdgpu_bo_create_restricted(adev, size, PAGE_SIZE,
-					  true, domain, flags,
-					  NULL, &placement, NULL,
-					  0, &obj);
+	ret = amdgpu_bo_create(adev, size, align, true, domain, flags,
+			       NULL, NULL, 0, &obj);
 	if (ret) {
 		DRM_ERROR("(%d) bo create failed\n", ret);
 		return ret;
@@ -155,19 +118,14 @@ static int amdgpu_cgs_gmap_gpu_mem(struct cgs_device *cgs_device, cgs_handle_t h
 				   uint64_t *mcaddr)
 {
 	int r;
-	u64 min_offset, max_offset;
 	struct amdgpu_bo *obj = (struct amdgpu_bo *)handle;
 
 	WARN_ON_ONCE(obj->placement.num_placement > 1);
 
-	min_offset = obj->placements[0].fpfn << PAGE_SHIFT;
-	max_offset = obj->placements[0].lpfn << PAGE_SHIFT;
-
 	r = amdgpu_bo_reserve(obj, true);
 	if (unlikely(r != 0))
 		return r;
-	r = amdgpu_bo_pin_restricted(obj, obj->preferred_domains,
-				     min_offset, max_offset, mcaddr);
+	r = amdgpu_bo_pin(obj, obj->preferred_domains, mcaddr);
 	amdgpu_bo_unreserve(obj);
 	return r;
 }
@@ -675,6 +633,85 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
 
 		if (!adev->pm.fw) {
 			switch (adev->asic_type) {
+			case CHIP_TAHITI:
+				strcpy(fw_name, "radeon/tahiti_smc.bin");
+				break;
+			case CHIP_PITCAIRN:
+				if ((adev->pdev->revision == 0x81) &&
+				    ((adev->pdev->device == 0x6810) ||
+				    (adev->pdev->device == 0x6811))) {
+					info->is_kicker = true;
+					strcpy(fw_name, "radeon/pitcairn_k_smc.bin");
+				} else {
+					strcpy(fw_name, "radeon/pitcairn_smc.bin");
+				}
+				break;
+			case CHIP_VERDE:
+				if (((adev->pdev->device == 0x6820) &&
+					((adev->pdev->revision == 0x81) ||
+					(adev->pdev->revision == 0x83))) ||
+				    ((adev->pdev->device == 0x6821) &&
+					((adev->pdev->revision == 0x83) ||
+					(adev->pdev->revision == 0x87))) ||
+				    ((adev->pdev->revision == 0x87) &&
+					((adev->pdev->device == 0x6823) ||
+					(adev->pdev->device == 0x682b)))) {
+					info->is_kicker = true;
+					strcpy(fw_name, "radeon/verde_k_smc.bin");
+				} else {
+					strcpy(fw_name, "radeon/verde_smc.bin");
+				}
+				break;
+			case CHIP_OLAND:
+				if (((adev->pdev->revision == 0x81) &&
+					((adev->pdev->device == 0x6600) ||
+					(adev->pdev->device == 0x6604) ||
+					(adev->pdev->device == 0x6605) ||
+					(adev->pdev->device == 0x6610))) ||
+				    ((adev->pdev->revision == 0x83) &&
+					(adev->pdev->device == 0x6610))) {
+					info->is_kicker = true;
+					strcpy(fw_name, "radeon/oland_k_smc.bin");
+				} else {
+					strcpy(fw_name, "radeon/oland_smc.bin");
+				}
+				break;
+			case CHIP_HAINAN:
+				if (((adev->pdev->revision == 0x81) &&
+					(adev->pdev->device == 0x6660)) ||
+				    ((adev->pdev->revision == 0x83) &&
+					((adev->pdev->device == 0x6660) ||
+					(adev->pdev->device == 0x6663) ||
+					(adev->pdev->device == 0x6665) ||
+					 (adev->pdev->device == 0x6667)))) {
+					info->is_kicker = true;
+					strcpy(fw_name, "radeon/hainan_k_smc.bin");
+				} else if ((adev->pdev->revision == 0xc3) &&
+					 (adev->pdev->device == 0x6665)) {
+					info->is_kicker = true;
+					strcpy(fw_name, "radeon/banks_k_2_smc.bin");
+				} else {
+					strcpy(fw_name, "radeon/hainan_smc.bin");
+				}
+				break;
+			case CHIP_BONAIRE:
+				if ((adev->pdev->revision == 0x80) ||
+					(adev->pdev->revision == 0x81) ||
+					(adev->pdev->device == 0x665f)) {
+					info->is_kicker = true;
+					strcpy(fw_name, "radeon/bonaire_k_smc.bin");
+				} else {
+					strcpy(fw_name, "radeon/bonaire_smc.bin");
+				}
+				break;
+			case CHIP_HAWAII:
+				if (adev->pdev->revision == 0x80) {
+					info->is_kicker = true;
+					strcpy(fw_name, "radeon/hawaii_k_smc.bin");
+				} else {
+					strcpy(fw_name, "radeon/hawaii_smc.bin");
+				}
+				break;
 			case CHIP_TOPAZ:
 				if (((adev->pdev->device == 0x6900) && (adev->pdev->revision == 0x81)) ||
 				    ((adev->pdev->device == 0x6900) && (adev->pdev->revision == 0x83)) ||
@@ -838,6 +875,9 @@ static int amdgpu_cgs_query_system_info(struct cgs_device *cgs_device,
 	case CGS_SYSTEM_INFO_PCIE_SUB_SYS_VENDOR_ID:
 		sys_info->value = adev->pdev->subsystem_vendor;
 		break;
+	case CGS_SYSTEM_INFO_PCIE_BUS_DEVFN:
+		sys_info->value = adev->pdev->devfn;
+		break;
 	default:
 		return -ENODEV;
 	}

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

@@ -346,10 +346,8 @@ static void amdgpu_connector_free_edid(struct drm_connector *connector)
 {
 	struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
 
-	if (amdgpu_connector->edid) {
-		kfree(amdgpu_connector->edid);
-		amdgpu_connector->edid = NULL;
-	}
+	kfree(amdgpu_connector->edid);
+	amdgpu_connector->edid = NULL;
 }
 
 static int amdgpu_connector_ddc_get_modes(struct drm_connector *connector)

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

@@ -473,11 +473,16 @@ static int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p,
 			return -EPERM;
 
 		/* Check if we have user pages and nobody bound the BO already */
-		if (lobj->user_pages && bo->tbo.ttm->state != tt_bound) {
-			size_t size = sizeof(struct page *);
-
-			size *= bo->tbo.ttm->num_pages;
-			memcpy(bo->tbo.ttm->pages, lobj->user_pages, size);
+		if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm) &&
+		    lobj->user_pages) {
+			amdgpu_ttm_placement_from_domain(bo,
+							 AMDGPU_GEM_DOMAIN_CPU);
+			r = ttm_bo_validate(&bo->tbo, &bo->placement, true,
+					    false);
+			if (r)
+				return r;
+			amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm,
+						     lobj->user_pages);
 			binding_userptr = true;
 		}
 
@@ -502,7 +507,6 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
 	struct amdgpu_bo_list_entry *e;
 	struct list_head duplicates;
-	bool need_mmap_lock = false;
 	unsigned i, tries = 10;
 	int r;
 
@@ -510,9 +514,9 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 
 	p->bo_list = amdgpu_bo_list_get(fpriv, cs->in.bo_list_handle);
 	if (p->bo_list) {
-		need_mmap_lock = p->bo_list->first_userptr !=
-			p->bo_list->num_entries;
 		amdgpu_bo_list_get_list(p->bo_list, &p->validated);
+		if (p->bo_list->first_userptr != p->bo_list->num_entries)
+			p->mn = amdgpu_mn_get(p->adev);
 	}
 
 	INIT_LIST_HEAD(&duplicates);
@@ -521,9 +525,6 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 	if (p->uf_entry.robj)
 		list_add(&p->uf_entry.tv.head, &p->validated);
 
-	if (need_mmap_lock)
-		down_read(&current->mm->mmap_sem);
-
 	while (1) {
 		struct list_head need_pages;
 		unsigned i;
@@ -543,23 +544,25 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 		INIT_LIST_HEAD(&need_pages);
 		for (i = p->bo_list->first_userptr;
 		     i < p->bo_list->num_entries; ++i) {
+			struct amdgpu_bo *bo;
 
 			e = &p->bo_list->array[i];
+			bo = e->robj;
 
-			if (amdgpu_ttm_tt_userptr_invalidated(e->robj->tbo.ttm,
+			if (amdgpu_ttm_tt_userptr_invalidated(bo->tbo.ttm,
 				 &e->user_invalidated) && e->user_pages) {
 
 				/* We acquired a page array, but somebody
 				 * invalidated it. Free it and try again
 				 */
 				release_pages(e->user_pages,
-					      e->robj->tbo.ttm->num_pages,
+					      bo->tbo.ttm->num_pages,
 					      false);
 				kvfree(e->user_pages);
 				e->user_pages = NULL;
 			}
 
-			if (e->robj->tbo.ttm->state != tt_bound &&
+			if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm) &&
 			    !e->user_pages) {
 				list_del(&e->tv.head);
 				list_add(&e->tv.head, &need_pages);
@@ -636,9 +639,6 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 
 	amdgpu_cs_report_moved_bytes(p->adev, p->bytes_moved,
 				     p->bytes_moved_vis);
-	fpriv->vm.last_eviction_counter =
-		atomic64_read(&p->adev->num_evictions);
-
 	if (p->bo_list) {
 		struct amdgpu_bo *gds = p->bo_list->gds_obj;
 		struct amdgpu_bo *gws = p->bo_list->gws_obj;
@@ -679,9 +679,6 @@ error_validate:
 
 error_free_pages:
 
-	if (need_mmap_lock)
-		up_read(&current->mm->mmap_sem);
-
 	if (p->bo_list) {
 		for (i = p->bo_list->first_userptr;
 		     i < p->bo_list->num_entries; ++i) {
@@ -728,11 +725,7 @@ static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser, int error,
 {
 	unsigned i;
 
-	if (!error)
-		ttm_eu_fence_buffer_objects(&parser->ticket,
-					    &parser->validated,
-					    parser->fence);
-	else if (backoff)
+	if (error && backoff)
 		ttm_eu_backoff_reservation(&parser->ticket,
 					   &parser->validated);
 
@@ -768,10 +761,6 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p)
 	if (r)
 		return r;
 
-	r = amdgpu_sync_fence(adev, &p->job->sync, vm->last_dir_update);
-	if (r)
-		return r;
-
 	r = amdgpu_vm_clear_freed(adev, vm, NULL);
 	if (r)
 		return r;
@@ -825,7 +814,13 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p)
 
 	}
 
-	r = amdgpu_vm_clear_moved(adev, vm, &p->job->sync);
+	r = amdgpu_vm_handle_moved(adev, vm);
+	if (r)
+		return r;
+
+	r = amdgpu_sync_fence(adev, &p->job->sync, vm->last_update);
+	if (r)
+		return r;
 
 	if (amdgpu_vm_debug && p->bo_list) {
 		/* Invalidate all BOs to test for userspace bugs */
@@ -835,7 +830,7 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p)
 			if (!bo)
 				continue;
 
-			amdgpu_vm_bo_invalidate(adev, bo);
+			amdgpu_vm_bo_invalidate(adev, bo, false);
 		}
 	}
 
@@ -860,7 +855,7 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev,
 	}
 
 	if (p->job->vm) {
-		p->job->vm_pd_addr = amdgpu_bo_gpu_offset(vm->root.bo);
+		p->job->vm_pd_addr = amdgpu_bo_gpu_offset(vm->root.base.bo);
 
 		r = amdgpu_bo_vm_update_pte(p);
 		if (r)
@@ -928,11 +923,11 @@ static int amdgpu_cs_ib_fill(struct amdgpu_device *adev,
 			uint64_t offset;
 			uint8_t *kptr;
 
-			m = amdgpu_cs_find_mapping(parser, chunk_ib->va_start,
-						   &aobj);
-			if (!aobj) {
+			r = amdgpu_cs_find_mapping(parser, chunk_ib->va_start,
+						   &aobj, &m);
+			if (r) {
 				DRM_ERROR("IB va_start is invalid\n");
-				return -EINVAL;
+				return r;
 			}
 
 			if ((chunk_ib->va_start + chunk_ib->ib_bytes) >
@@ -1133,14 +1128,31 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	struct amdgpu_ring *ring = p->job->ring;
 	struct amd_sched_entity *entity = &p->ctx->rings[ring->idx].entity;
 	struct amdgpu_job *job;
+	unsigned i;
+	uint64_t seq;
+
 	int r;
 
+	amdgpu_mn_lock(p->mn);
+	if (p->bo_list) {
+		for (i = p->bo_list->first_userptr;
+		     i < p->bo_list->num_entries; ++i) {
+			struct amdgpu_bo *bo = p->bo_list->array[i].robj;
+
+			if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm)) {
+				amdgpu_mn_unlock(p->mn);
+				return -ERESTARTSYS;
+			}
+		}
+	}
+
 	job = p->job;
 	p->job = NULL;
 
 	r = amd_sched_job_init(&job->base, &ring->sched, entity, p->filp);
 	if (r) {
 		amdgpu_job_free(job);
+		amdgpu_mn_unlock(p->mn);
 		return r;
 	}
 
@@ -1148,14 +1160,28 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	job->fence_ctx = entity->fence_context;
 	p->fence = dma_fence_get(&job->base.s_fence->finished);
 
+	r = amdgpu_ctx_add_fence(p->ctx, ring, p->fence, &seq);
+	if (r) {
+		dma_fence_put(p->fence);
+		dma_fence_put(&job->base.s_fence->finished);
+		amdgpu_job_free(job);
+		amdgpu_mn_unlock(p->mn);
+		return r;
+	}
+
 	amdgpu_cs_post_dependencies(p);
 
-	cs->out.handle = amdgpu_ctx_add_fence(p->ctx, ring, p->fence);
-	job->uf_sequence = cs->out.handle;
+	cs->out.handle = seq;
+	job->uf_sequence = seq;
+
 	amdgpu_job_free_resources(job);
 
 	trace_amdgpu_cs_ioctl(job);
 	amd_sched_entity_push_job(&job->base);
+
+	ttm_eu_fence_buffer_objects(&p->ticket, &p->validated, p->fence);
+	amdgpu_mn_unlock(p->mn);
+
 	return 0;
 }
 
@@ -1383,6 +1409,7 @@ static int amdgpu_cs_wait_any_fence(struct amdgpu_device *adev,
 			array[i] = fence;
 		} else { /* NULL, the fence has been already signaled */
 			r = 1;
+			first = i;
 			goto out;
 		}
 	}
@@ -1462,78 +1489,36 @@ err_free_fences:
  * virtual memory address. Returns allocation structure when found, NULL
  * otherwise.
  */
-struct amdgpu_bo_va_mapping *
-amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
-		       uint64_t addr, struct amdgpu_bo **bo)
+int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
+			   uint64_t addr, struct amdgpu_bo **bo,
+			   struct amdgpu_bo_va_mapping **map)
 {
+	struct amdgpu_fpriv *fpriv = parser->filp->driver_priv;
+	struct amdgpu_vm *vm = &fpriv->vm;
 	struct amdgpu_bo_va_mapping *mapping;
-	unsigned i;
-
-	if (!parser->bo_list)
-		return NULL;
+	int r;
 
 	addr /= AMDGPU_GPU_PAGE_SIZE;
 
-	for (i = 0; i < parser->bo_list->num_entries; i++) {
-		struct amdgpu_bo_list_entry *lobj;
-
-		lobj = &parser->bo_list->array[i];
-		if (!lobj->bo_va)
-			continue;
-
-		list_for_each_entry(mapping, &lobj->bo_va->valids, list) {
-			if (mapping->start > addr ||
-			    addr > mapping->last)
-				continue;
-
-			*bo = lobj->bo_va->base.bo;
-			return mapping;
-		}
-
-		list_for_each_entry(mapping, &lobj->bo_va->invalids, list) {
-			if (mapping->start > addr ||
-			    addr > mapping->last)
-				continue;
+	mapping = amdgpu_vm_bo_lookup_mapping(vm, addr);
+	if (!mapping || !mapping->bo_va || !mapping->bo_va->base.bo)
+		return -EINVAL;
 
-			*bo = lobj->bo_va->base.bo;
-			return mapping;
-		}
-	}
+	*bo = mapping->bo_va->base.bo;
+	*map = mapping;
 
-	return NULL;
-}
+	/* Double check that the BO is reserved by this CS */
+	if (READ_ONCE((*bo)->tbo.resv->lock.ctx) != &parser->ticket)
+		return -EINVAL;
 
-/**
- * amdgpu_cs_sysvm_access_required - make BOs accessible by the system VM
- *
- * @parser: command submission parser context
- *
- * Helper for UVD/VCE VM emulation, make sure BOs are accessible by the system VM.
- */
-int amdgpu_cs_sysvm_access_required(struct amdgpu_cs_parser *parser)
-{
-	unsigned i;
-	int r;
+	r = amdgpu_ttm_bind(&(*bo)->tbo, &(*bo)->tbo.mem);
+	if (unlikely(r))
+		return r;
 
-	if (!parser->bo_list)
+	if ((*bo)->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)
 		return 0;
 
-	for (i = 0; i < parser->bo_list->num_entries; i++) {
-		struct amdgpu_bo *bo = parser->bo_list->array[i].robj;
-
-		r = amdgpu_ttm_bind(&bo->tbo, &bo->tbo.mem);
-		if (unlikely(r))
-			return r;
-
-		if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)
-			continue;
-
-		bo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
-		amdgpu_ttm_placement_from_domain(bo, bo->allowed_domains);
-		r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false);
-		if (unlikely(r))
-			return r;
-	}
-
-	return 0;
+	(*bo)->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
+	amdgpu_ttm_placement_from_domain(*bo, (*bo)->allowed_domains);
+	return ttm_bo_validate(&(*bo)->tbo, &(*bo)->placement, false, false);
 }

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

@@ -246,8 +246,8 @@ int amdgpu_ctx_put(struct amdgpu_ctx *ctx)
 	return 0;
 }
 
-uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring,
-			      struct dma_fence *fence)
+int amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring,
+			      struct dma_fence *fence, uint64_t* handler)
 {
 	struct amdgpu_ctx_ring *cring = & ctx->rings[ring->idx];
 	uint64_t seq = cring->sequence;
@@ -258,9 +258,9 @@ uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring,
 	other = cring->fences[idx];
 	if (other) {
 		signed long r;
-		r = dma_fence_wait_timeout(other, false, MAX_SCHEDULE_TIMEOUT);
+		r = dma_fence_wait_timeout(other, true, MAX_SCHEDULE_TIMEOUT);
 		if (r < 0)
-			DRM_ERROR("Error (%ld) waiting for fence!\n", r);
+			return r;
 	}
 
 	dma_fence_get(fence);
@@ -271,8 +271,10 @@ uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring,
 	spin_unlock(&ctx->ring_lock);
 
 	dma_fence_put(other);
+	if (handler)
+		*handler = seq;
 
-	return seq;
+	return 0;
 }
 
 struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx,

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

@@ -65,6 +65,7 @@ MODULE_FIRMWARE("amdgpu/raven_gpu_info.bin");
 static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev);
 static void amdgpu_debugfs_regs_cleanup(struct amdgpu_device *adev);
 static int amdgpu_debugfs_test_ib_ring_init(struct amdgpu_device *adev);
+static int amdgpu_debugfs_vbios_dump_init(struct amdgpu_device *adev);
 
 static const char *amdgpu_asic_name[] = {
 	"TAHITI",
@@ -402,6 +403,15 @@ void amdgpu_pci_config_reset(struct amdgpu_device *adev)
  */
 static int amdgpu_doorbell_init(struct amdgpu_device *adev)
 {
+	/* No doorbell on SI hardware generation */
+	if (adev->asic_type < CHIP_BONAIRE) {
+		adev->doorbell.base = 0;
+		adev->doorbell.size = 0;
+		adev->doorbell.num_doorbells = 0;
+		adev->doorbell.ptr = NULL;
+		return 0;
+	}
+
 	/* doorbell bar mapping */
 	adev->doorbell.base = pci_resource_start(adev->pdev, 2);
 	adev->doorbell.size = pci_resource_len(adev->pdev, 2);
@@ -887,6 +897,20 @@ static uint32_t cail_ioreg_read(struct card_info *info, uint32_t reg)
 	return r;
 }
 
+static ssize_t amdgpu_atombios_get_vbios_version(struct device *dev,
+						 struct device_attribute *attr,
+						 char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = ddev->dev_private;
+	struct atom_context *ctx = adev->mode_info.atom_context;
+
+	return snprintf(buf, PAGE_SIZE, "%s\n", ctx->vbios_version);
+}
+
+static DEVICE_ATTR(vbios_version, 0444, amdgpu_atombios_get_vbios_version,
+		   NULL);
+
 /**
  * amdgpu_atombios_fini - free the driver info and callbacks for atombios
  *
@@ -906,6 +930,7 @@ static void amdgpu_atombios_fini(struct amdgpu_device *adev)
 	adev->mode_info.atom_context = NULL;
 	kfree(adev->mode_info.atom_card_info);
 	adev->mode_info.atom_card_info = NULL;
+	device_remove_file(adev->dev, &dev_attr_vbios_version);
 }
 
 /**
@@ -922,6 +947,7 @@ static int amdgpu_atombios_init(struct amdgpu_device *adev)
 {
 	struct card_info *atom_card_info =
 	    kzalloc(sizeof(struct card_info), GFP_KERNEL);
+	int ret;
 
 	if (!atom_card_info)
 		return -ENOMEM;
@@ -958,6 +984,13 @@ static int amdgpu_atombios_init(struct amdgpu_device *adev)
 		amdgpu_atombios_scratch_regs_init(adev);
 		amdgpu_atombios_allocate_fb_scratch(adev);
 	}
+
+	ret = device_create_file(adev->dev, &dev_attr_vbios_version);
+	if (ret) {
+		DRM_ERROR("Failed to create device file for VBIOS version\n");
+		return ret;
+	}
+
 	return 0;
 }
 
@@ -1757,10 +1790,8 @@ static int amdgpu_fini(struct amdgpu_device *adev)
 		adev->ip_blocks[i].status.late_initialized = false;
 	}
 
-	if (amdgpu_sriov_vf(adev)) {
-		amdgpu_bo_free_kernel(&adev->virt.csa_obj, &adev->virt.csa_vmid0_addr, NULL);
+	if (amdgpu_sriov_vf(adev))
 		amdgpu_virt_release_full_gpu(adev, false);
-	}
 
 	return 0;
 }
@@ -2051,9 +2082,8 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	DRM_INFO("register mmio base: 0x%08X\n", (uint32_t)adev->rmmio_base);
 	DRM_INFO("register mmio size: %u\n", (unsigned)adev->rmmio_size);
 
-	if (adev->asic_type >= CHIP_BONAIRE)
-		/* doorbell bar mapping */
-		amdgpu_doorbell_init(adev);
+	/* doorbell bar mapping */
+	amdgpu_doorbell_init(adev);
 
 	/* io port mapping */
 	for (i = 0; i < DEVICE_COUNT_RESOURCE; i++) {
@@ -2201,6 +2231,10 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	if (r)
 		DRM_ERROR("registering firmware debugfs failed (%d).\n", r);
 
+	r = amdgpu_debugfs_vbios_dump_init(adev);
+	if (r)
+		DRM_ERROR("Creating vbios dump debugfs failed (%d).\n", r);
+
 	if ((amdgpu_testing & 1)) {
 		if (adev->accel_working)
 			amdgpu_test_moves(adev);
@@ -2276,8 +2310,7 @@ void amdgpu_device_fini(struct amdgpu_device *adev)
 	adev->rio_mem = NULL;
 	iounmap(adev->rmmio);
 	adev->rmmio = NULL;
-	if (adev->asic_type >= CHIP_BONAIRE)
-		amdgpu_doorbell_fini(adev);
+	amdgpu_doorbell_fini(adev);
 	amdgpu_debugfs_regs_cleanup(adev);
 }
 
@@ -2546,7 +2579,8 @@ static bool amdgpu_need_full_reset(struct amdgpu_device *adev)
 		if ((adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) ||
 		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC) ||
 		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_ACP) ||
-		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE)) {
+		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE) ||
+		     adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP) {
 			if (adev->ip_blocks[i].status.hang) {
 				DRM_INFO("Some block need full reset!\n");
 				return true;
@@ -2654,7 +2688,7 @@ int amdgpu_sriov_gpu_reset(struct amdgpu_device *adev, struct amdgpu_job *job)
 
 	mutex_lock(&adev->virt.lock_reset);
 	atomic_inc(&adev->gpu_reset_counter);
-	adev->gfx.in_reset = true;
+	adev->in_sriov_reset = true;
 
 	/* block TTM */
 	resched = ttm_bo_lock_delayed_workqueue(&adev->mman.bdev);
@@ -2765,7 +2799,7 @@ give_up_reset:
 		dev_info(adev->dev, "GPU reset successed!\n");
 	}
 
-	adev->gfx.in_reset = false;
+	adev->in_sriov_reset = false;
 	mutex_unlock(&adev->virt.lock_reset);
 	return r;
 }
@@ -3463,10 +3497,7 @@ static ssize_t amdgpu_debugfs_sensor_read(struct file *f, char __user *buf,
 
 	valuesize = sizeof(values);
 	if (adev->powerplay.pp_funcs && adev->powerplay.pp_funcs->read_sensor)
-		r = adev->powerplay.pp_funcs->read_sensor(adev->powerplay.pp_handle, idx, &values[0], &valuesize);
-	else if (adev->pm.funcs && adev->pm.funcs->read_sensor)
-		r = adev->pm.funcs->read_sensor(adev, idx, &values[0],
-						&valuesize);
+		r = amdgpu_dpm_read_sensor(adev, idx, &values[0], &valuesize);
 	else
 		return -EINVAL;
 
@@ -3754,6 +3785,28 @@ int amdgpu_debugfs_init(struct drm_minor *minor)
 {
 	return 0;
 }
+
+static int amdgpu_debugfs_get_vbios_dump(struct seq_file *m, void *data)
+{
+	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	struct drm_device *dev = node->minor->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+
+	seq_write(m, adev->bios, adev->bios_size);
+	return 0;
+}
+
+static const struct drm_info_list amdgpu_vbios_dump_list[] = {
+		{"amdgpu_vbios",
+		 amdgpu_debugfs_get_vbios_dump,
+		 0, NULL},
+};
+
+static int amdgpu_debugfs_vbios_dump_init(struct amdgpu_device *adev)
+{
+	return amdgpu_debugfs_add_files(adev,
+					amdgpu_vbios_dump_list, 1);
+}
 #else
 static int amdgpu_debugfs_test_ib_ring_init(struct amdgpu_device *adev)
 {
@@ -3763,5 +3816,9 @@ static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev)
 {
 	return 0;
 }
+static int amdgpu_debugfs_vbios_dump_init(struct amdgpu_device *adev)
+{
+	return 0;
+}
 static void amdgpu_debugfs_regs_cleanup(struct amdgpu_device *adev) { }
 #endif

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

@@ -960,8 +960,10 @@ u8 amdgpu_encode_pci_lane_width(u32 lanes)
 }
 
 struct amd_vce_state*
-amdgpu_get_vce_clock_state(struct amdgpu_device *adev, unsigned idx)
+amdgpu_get_vce_clock_state(void *handle, u32 idx)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
 	if (idx < adev->pm.dpm.num_of_vce_states)
 		return &adev->pm.dpm.vce_states[idx];
 

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

@@ -241,177 +241,119 @@ enum amdgpu_pcie_gen {
 	AMDGPU_PCIE_GEN_INVALID = 0xffff
 };
 
-struct amdgpu_dpm_funcs {
-	int (*get_temperature)(struct amdgpu_device *adev);
-	int (*pre_set_power_state)(struct amdgpu_device *adev);
-	int (*set_power_state)(struct amdgpu_device *adev);
-	void (*post_set_power_state)(struct amdgpu_device *adev);
-	void (*display_configuration_changed)(struct amdgpu_device *adev);
-	u32 (*get_sclk)(struct amdgpu_device *adev, bool low);
-	u32 (*get_mclk)(struct amdgpu_device *adev, bool low);
-	void (*print_power_state)(struct amdgpu_device *adev, struct amdgpu_ps *ps);
-	void (*debugfs_print_current_performance_level)(struct amdgpu_device *adev, struct seq_file *m);
-	int (*force_performance_level)(struct amdgpu_device *adev, enum amd_dpm_forced_level level);
-	bool (*vblank_too_short)(struct amdgpu_device *adev);
-	void (*powergate_uvd)(struct amdgpu_device *adev, bool gate);
-	void (*powergate_vce)(struct amdgpu_device *adev, bool gate);
-	void (*enable_bapm)(struct amdgpu_device *adev, bool enable);
-	void (*set_fan_control_mode)(struct amdgpu_device *adev, u32 mode);
-	u32 (*get_fan_control_mode)(struct amdgpu_device *adev);
-	int (*set_fan_speed_percent)(struct amdgpu_device *adev, u32 speed);
-	int (*get_fan_speed_percent)(struct amdgpu_device *adev, u32 *speed);
-	int (*force_clock_level)(struct amdgpu_device *adev, enum pp_clock_type type, uint32_t mask);
-	int (*print_clock_levels)(struct amdgpu_device *adev, enum pp_clock_type type, char *buf);
-	int (*get_sclk_od)(struct amdgpu_device *adev);
-	int (*set_sclk_od)(struct amdgpu_device *adev, uint32_t value);
-	int (*get_mclk_od)(struct amdgpu_device *adev);
-	int (*set_mclk_od)(struct amdgpu_device *adev, uint32_t value);
-	int (*check_state_equal)(struct amdgpu_device *adev,
-				struct amdgpu_ps *cps,
-				struct amdgpu_ps *rps,
-				bool *equal);
-	int (*read_sensor)(struct amdgpu_device *adev, int idx, void *value,
-			   int *size);
-
-	struct amd_vce_state* (*get_vce_clock_state)(struct amdgpu_device *adev, unsigned idx);
-	int (*reset_power_profile_state)(struct amdgpu_device *adev,
-			struct amd_pp_profile *request);
-	int (*get_power_profile_state)(struct amdgpu_device *adev,
-			struct amd_pp_profile *query);
-	int (*set_power_profile_state)(struct amdgpu_device *adev,
-			struct amd_pp_profile *request);
-	int (*switch_power_profile)(struct amdgpu_device *adev,
-			enum amd_pp_profile_type type);
-};
+#define amdgpu_dpm_pre_set_power_state(adev) \
+		((adev)->powerplay.pp_funcs->pre_set_power_state((adev)->powerplay.pp_handle))
+
+#define amdgpu_dpm_set_power_state(adev) \
+		((adev)->powerplay.pp_funcs->set_power_state((adev)->powerplay.pp_handle))
+
+#define amdgpu_dpm_post_set_power_state(adev) \
+		((adev)->powerplay.pp_funcs->post_set_power_state((adev)->powerplay.pp_handle))
+
+#define amdgpu_dpm_display_configuration_changed(adev) \
+		((adev)->powerplay.pp_funcs->display_configuration_changed((adev)->powerplay.pp_handle))
 
-#define amdgpu_dpm_pre_set_power_state(adev) (adev)->pm.funcs->pre_set_power_state((adev))
-#define amdgpu_dpm_set_power_state(adev) (adev)->pm.funcs->set_power_state((adev))
-#define amdgpu_dpm_post_set_power_state(adev) (adev)->pm.funcs->post_set_power_state((adev))
-#define amdgpu_dpm_display_configuration_changed(adev) (adev)->pm.funcs->display_configuration_changed((adev))
-#define amdgpu_dpm_print_power_state(adev, ps) (adev)->pm.funcs->print_power_state((adev), (ps))
-#define amdgpu_dpm_vblank_too_short(adev) (adev)->pm.funcs->vblank_too_short((adev))
-#define amdgpu_dpm_enable_bapm(adev, e) (adev)->pm.funcs->enable_bapm((adev), (e))
+#define amdgpu_dpm_print_power_state(adev, ps) \
+		((adev)->powerplay.pp_funcs->print_power_state((adev)->powerplay.pp_handle, (ps)))
+
+#define amdgpu_dpm_vblank_too_short(adev) \
+		((adev)->powerplay.pp_funcs->vblank_too_short((adev)->powerplay.pp_handle))
+
+#define amdgpu_dpm_enable_bapm(adev, e) \
+		((adev)->powerplay.pp_funcs->enable_bapm((adev)->powerplay.pp_handle, (e)))
 
 #define amdgpu_dpm_read_sensor(adev, idx, value, size) \
-	((adev)->pp_enabled ? \
-		(adev)->powerplay.pp_funcs->read_sensor(adev->powerplay.pp_handle, (idx), (value), (size)) : \
-		(adev)->pm.funcs->read_sensor((adev), (idx), (value), (size)))
+		((adev)->powerplay.pp_funcs->read_sensor((adev)->powerplay.pp_handle, (idx), (value), (size)))
 
 #define amdgpu_dpm_get_temperature(adev) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->get_temperature((adev)->powerplay.pp_handle) : \
-	      (adev)->pm.funcs->get_temperature((adev)))
+		((adev)->powerplay.pp_funcs->get_temperature((adev)->powerplay.pp_handle))
 
 #define amdgpu_dpm_set_fan_control_mode(adev, m) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->set_fan_control_mode((adev)->powerplay.pp_handle, (m)) : \
-	      (adev)->pm.funcs->set_fan_control_mode((adev), (m)))
+		((adev)->powerplay.pp_funcs->set_fan_control_mode((adev)->powerplay.pp_handle, (m)))
 
 #define amdgpu_dpm_get_fan_control_mode(adev) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->get_fan_control_mode((adev)->powerplay.pp_handle) : \
-	      (adev)->pm.funcs->get_fan_control_mode((adev)))
+		((adev)->powerplay.pp_funcs->get_fan_control_mode((adev)->powerplay.pp_handle))
 
 #define amdgpu_dpm_set_fan_speed_percent(adev, s) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->set_fan_speed_percent((adev)->powerplay.pp_handle, (s)) : \
-	      (adev)->pm.funcs->set_fan_speed_percent((adev), (s)))
+		((adev)->powerplay.pp_funcs->set_fan_speed_percent((adev)->powerplay.pp_handle, (s)))
 
 #define amdgpu_dpm_get_fan_speed_percent(adev, s) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->get_fan_speed_percent((adev)->powerplay.pp_handle, (s)) : \
-	      (adev)->pm.funcs->get_fan_speed_percent((adev), (s)))
+		((adev)->powerplay.pp_funcs->get_fan_speed_percent((adev)->powerplay.pp_handle, (s)))
 
 #define amdgpu_dpm_get_fan_speed_rpm(adev, s) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->get_fan_speed_rpm((adev)->powerplay.pp_handle, (s)) : \
-	      -EINVAL)
+		((adev)->powerplay.pp_funcs->get_fan_speed_rpm)((adev)->powerplay.pp_handle, (s))
 
 #define amdgpu_dpm_get_sclk(adev, l) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->get_sclk((adev)->powerplay.pp_handle, (l)) : \
-		(adev)->pm.funcs->get_sclk((adev), (l)))
+		((adev)->powerplay.pp_funcs->get_sclk((adev)->powerplay.pp_handle, (l)))
 
 #define amdgpu_dpm_get_mclk(adev, l)  \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->get_mclk((adev)->powerplay.pp_handle, (l)) : \
-	      (adev)->pm.funcs->get_mclk((adev), (l)))
-
+		((adev)->powerplay.pp_funcs->get_mclk((adev)->powerplay.pp_handle, (l)))
 
 #define amdgpu_dpm_force_performance_level(adev, l) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->force_performance_level((adev)->powerplay.pp_handle, (l)) : \
-	      (adev)->pm.funcs->force_performance_level((adev), (l)))
+		((adev)->powerplay.pp_funcs->force_performance_level((adev)->powerplay.pp_handle, (l)))
 
 #define amdgpu_dpm_powergate_uvd(adev, g) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->powergate_uvd((adev)->powerplay.pp_handle, (g)) : \
-	      (adev)->pm.funcs->powergate_uvd((adev), (g)))
+		((adev)->powerplay.pp_funcs->powergate_uvd((adev)->powerplay.pp_handle, (g)))
 
 #define amdgpu_dpm_powergate_vce(adev, g) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->powergate_vce((adev)->powerplay.pp_handle, (g)) : \
-	      (adev)->pm.funcs->powergate_vce((adev), (g)))
+		((adev)->powerplay.pp_funcs->powergate_vce((adev)->powerplay.pp_handle, (g)))
 
 #define amdgpu_dpm_get_current_power_state(adev) \
-	(adev)->powerplay.pp_funcs->get_current_power_state((adev)->powerplay.pp_handle)
+		((adev)->powerplay.pp_funcs->get_current_power_state((adev)->powerplay.pp_handle))
 
 #define amdgpu_dpm_get_pp_num_states(adev, data) \
-	(adev)->powerplay.pp_funcs->get_pp_num_states((adev)->powerplay.pp_handle, data)
+		((adev)->powerplay.pp_funcs->get_pp_num_states((adev)->powerplay.pp_handle, data))
 
 #define amdgpu_dpm_get_pp_table(adev, table) \
-	(adev)->powerplay.pp_funcs->get_pp_table((adev)->powerplay.pp_handle, table)
+		((adev)->powerplay.pp_funcs->get_pp_table((adev)->powerplay.pp_handle, table))
 
 #define amdgpu_dpm_set_pp_table(adev, buf, size) \
-	(adev)->powerplay.pp_funcs->set_pp_table((adev)->powerplay.pp_handle, buf, size)
+		((adev)->powerplay.pp_funcs->set_pp_table((adev)->powerplay.pp_handle, buf, size))
 
 #define amdgpu_dpm_print_clock_levels(adev, type, buf) \
-	(adev)->powerplay.pp_funcs->print_clock_levels((adev)->powerplay.pp_handle, type, buf)
+		((adev)->powerplay.pp_funcs->print_clock_levels((adev)->powerplay.pp_handle, type, buf))
 
 #define amdgpu_dpm_force_clock_level(adev, type, level) \
-		(adev)->powerplay.pp_funcs->force_clock_level((adev)->powerplay.pp_handle, type, level)
+		((adev)->powerplay.pp_funcs->force_clock_level((adev)->powerplay.pp_handle, type, level))
 
 #define amdgpu_dpm_get_sclk_od(adev) \
-	(adev)->powerplay.pp_funcs->get_sclk_od((adev)->powerplay.pp_handle)
+		((adev)->powerplay.pp_funcs->get_sclk_od((adev)->powerplay.pp_handle))
 
 #define amdgpu_dpm_set_sclk_od(adev, value) \
-	(adev)->powerplay.pp_funcs->set_sclk_od((adev)->powerplay.pp_handle, value)
+		((adev)->powerplay.pp_funcs->set_sclk_od((adev)->powerplay.pp_handle, value))
 
 #define amdgpu_dpm_get_mclk_od(adev) \
-	((adev)->powerplay.pp_funcs->get_mclk_od((adev)->powerplay.pp_handle))
+		((adev)->powerplay.pp_funcs->get_mclk_od((adev)->powerplay.pp_handle))
 
 #define amdgpu_dpm_set_mclk_od(adev, value) \
-	((adev)->powerplay.pp_funcs->set_mclk_od((adev)->powerplay.pp_handle, value))
+		((adev)->powerplay.pp_funcs->set_mclk_od((adev)->powerplay.pp_handle, value))
 
-#define amdgpu_dpm_dispatch_task(adev, event_id, input, output)		\
-	(adev)->powerplay.pp_funcs->dispatch_tasks((adev)->powerplay.pp_handle, (event_id), (input), (output))
+#define amdgpu_dpm_dispatch_task(adev, task_id, input, output)		\
+		((adev)->powerplay.pp_funcs->dispatch_tasks)((adev)->powerplay.pp_handle, (task_id), (input), (output))
 
-#define amgdpu_dpm_check_state_equal(adev, cps, rps, equal) (adev)->pm.funcs->check_state_equal((adev), (cps),(rps),(equal))
+#define amdgpu_dpm_check_state_equal(adev, cps, rps, equal) \
+		((adev)->powerplay.pp_funcs->check_state_equal((adev)->powerplay.pp_handle, (cps), (rps), (equal)))
 
 #define amdgpu_dpm_get_vce_clock_state(adev, i)				\
-	((adev)->pp_enabled ?						\
-	 (adev)->powerplay.pp_funcs->get_vce_clock_state((adev)->powerplay.pp_handle, (i)) : \
-	 (adev)->pm.funcs->get_vce_clock_state((adev), (i)))
+		((adev)->powerplay.pp_funcs->get_vce_clock_state((adev)->powerplay.pp_handle, (i)))
 
-#define amdgpu_dpm_get_performance_level(adev) \
-	((adev)->pp_enabled ?						\
-	(adev)->powerplay.pp_funcs->get_performance_level((adev)->powerplay.pp_handle) : \
-	(adev)->pm.dpm.forced_level)
+#define amdgpu_dpm_get_performance_level(adev)				\
+		((adev)->powerplay.pp_funcs->get_performance_level((adev)->powerplay.pp_handle))
 
 #define amdgpu_dpm_reset_power_profile_state(adev, request) \
-	((adev)->powerplay.pp_funcs->reset_power_profile_state(\
+		((adev)->powerplay.pp_funcs->reset_power_profile_state(\
 			(adev)->powerplay.pp_handle, request))
 
 #define amdgpu_dpm_get_power_profile_state(adev, query) \
-	((adev)->powerplay.pp_funcs->get_power_profile_state(\
+		((adev)->powerplay.pp_funcs->get_power_profile_state(\
 			(adev)->powerplay.pp_handle, query))
 
 #define amdgpu_dpm_set_power_profile_state(adev, request) \
-	((adev)->powerplay.pp_funcs->set_power_profile_state(\
+		((adev)->powerplay.pp_funcs->set_power_profile_state(\
 			(adev)->powerplay.pp_handle, request))
 
 #define amdgpu_dpm_switch_power_profile(adev, type) \
-	((adev)->powerplay.pp_funcs->switch_power_profile(\
+		((adev)->powerplay.pp_funcs->switch_power_profile(\
 			(adev)->powerplay.pp_handle, type))
 
 struct amdgpu_dpm {
@@ -485,7 +427,6 @@ struct amdgpu_pm {
 	struct amdgpu_dpm       dpm;
 	const struct firmware	*fw;	/* SMC firmware */
 	uint32_t                fw_version;
-	const struct amdgpu_dpm_funcs *funcs;
 	uint32_t                pcie_gen_mask;
 	uint32_t                pcie_mlw_mask;
 	struct amd_pp_display_configuration pm_display_cfg;/* set by DAL */
@@ -551,6 +492,6 @@ u16 amdgpu_get_pcie_lane_support(struct amdgpu_device *adev,
 u8 amdgpu_encode_pci_lane_width(u32 lanes);
 
 struct amd_vce_state*
-amdgpu_get_vce_clock_state(struct amdgpu_device *adev, unsigned idx);
+amdgpu_get_vce_clock_state(void *handle, u32 idx);
 
 #endif

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

@@ -69,9 +69,10 @@
  * - 3.17.0 - Add AMDGPU_NUM_VRAM_CPU_PAGE_FAULTS.
  * - 3.18.0 - Export gpu always on cu bitmap
  * - 3.19.0 - Add support for UVD MJPEG decode
+ * - 3.20.0 - Add support for local BOs
  */
 #define KMS_DRIVER_MAJOR	3
-#define KMS_DRIVER_MINOR	19
+#define KMS_DRIVER_MINOR	20
 #define KMS_DRIVER_PATCHLEVEL	0
 
 int amdgpu_vram_limit = 0;
@@ -91,7 +92,7 @@ int amdgpu_dpm = -1;
 int amdgpu_fw_load_type = -1;
 int amdgpu_aspm = -1;
 int amdgpu_runtime_pm = -1;
-unsigned amdgpu_ip_block_mask = 0xffffffff;
+uint amdgpu_ip_block_mask = 0xffffffff;
 int amdgpu_bapm = -1;
 int amdgpu_deep_color = 0;
 int amdgpu_vm_size = -1;
@@ -106,14 +107,14 @@ int amdgpu_sched_jobs = 32;
 int amdgpu_sched_hw_submission = 2;
 int amdgpu_no_evict = 0;
 int amdgpu_direct_gma_size = 0;
-unsigned amdgpu_pcie_gen_cap = 0;
-unsigned amdgpu_pcie_lane_cap = 0;
-unsigned amdgpu_cg_mask = 0xffffffff;
-unsigned amdgpu_pg_mask = 0xffffffff;
-unsigned amdgpu_sdma_phase_quantum = 32;
+uint amdgpu_pcie_gen_cap = 0;
+uint amdgpu_pcie_lane_cap = 0;
+uint amdgpu_cg_mask = 0xffffffff;
+uint amdgpu_pg_mask = 0xffffffff;
+uint amdgpu_sdma_phase_quantum = 32;
 char *amdgpu_disable_cu = NULL;
 char *amdgpu_virtual_display = NULL;
-unsigned amdgpu_pp_feature_mask = 0xffffffff;
+uint amdgpu_pp_feature_mask = 0xffffffff;
 int amdgpu_ngg = 0;
 int amdgpu_prim_buf_per_se = 0;
 int amdgpu_pos_buf_per_se = 0;
@@ -608,6 +609,8 @@ amdgpu_pci_remove(struct pci_dev *pdev)
 
 	drm_dev_unregister(dev);
 	drm_dev_unref(dev);
+	pci_disable_device(pdev);
+	pci_set_drvdata(pdev, NULL);
 }
 
 static void
@@ -852,6 +855,7 @@ static struct drm_driver kms_driver = {
 	.gem_prime_import_sg_table = amdgpu_gem_prime_import_sg_table,
 	.gem_prime_vmap = amdgpu_gem_prime_vmap,
 	.gem_prime_vunmap = amdgpu_gem_prime_vunmap,
+	.gem_prime_mmap = amdgpu_gem_prime_mmap,
 
 	.name = DRIVER_NAME,
 	.desc = DRIVER_DESC,

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

@@ -149,7 +149,7 @@ static int amdgpufb_create_pinned_object(struct amdgpu_fbdev *rfbdev,
 				       AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
 				       AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS |
 				       AMDGPU_GEM_CREATE_VRAM_CLEARED,
-				       true, &gobj);
+				       true, NULL, &gobj);
 	if (ret) {
 		pr_err("failed to allocate framebuffer (%d)\n", aligned_size);
 		return -ENOMEM;
@@ -303,10 +303,10 @@ static int amdgpu_fbdev_destroy(struct drm_device *dev, struct amdgpu_fbdev *rfb
 	if (rfb->obj) {
 		amdgpufb_destroy_pinned_object(rfb->obj);
 		rfb->obj = NULL;
+		drm_framebuffer_unregister_private(&rfb->base);
+		drm_framebuffer_cleanup(&rfb->base);
 	}
 	drm_fb_helper_fini(&rfbdev->helper);
-	drm_framebuffer_unregister_private(&rfb->base);
-	drm_framebuffer_cleanup(&rfb->base);
 
 	return 0;
 }

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

@@ -44,11 +44,12 @@ void amdgpu_gem_object_free(struct drm_gem_object *gobj)
 }
 
 int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
-				int alignment, u32 initial_domain,
-				u64 flags, bool kernel,
-				struct drm_gem_object **obj)
+			     int alignment, u32 initial_domain,
+			     u64 flags, bool kernel,
+			     struct reservation_object *resv,
+			     struct drm_gem_object **obj)
 {
-	struct amdgpu_bo *robj;
+	struct amdgpu_bo *bo;
 	int r;
 
 	*obj = NULL;
@@ -59,7 +60,7 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
 
 retry:
 	r = amdgpu_bo_create(adev, size, alignment, kernel, initial_domain,
-			     flags, NULL, NULL, 0, &robj);
+			     flags, NULL, resv, 0, &bo);
 	if (r) {
 		if (r != -ERESTARTSYS) {
 			if (initial_domain == AMDGPU_GEM_DOMAIN_VRAM) {
@@ -71,7 +72,7 @@ retry:
 		}
 		return r;
 	}
-	*obj = &robj->gem_base;
+	*obj = &bo->gem_base;
 
 	return 0;
 }
@@ -112,7 +113,17 @@ int amdgpu_gem_object_open(struct drm_gem_object *obj,
 	struct amdgpu_fpriv *fpriv = file_priv->driver_priv;
 	struct amdgpu_vm *vm = &fpriv->vm;
 	struct amdgpu_bo_va *bo_va;
+	struct mm_struct *mm;
 	int r;
+
+	mm = amdgpu_ttm_tt_get_usermm(abo->tbo.ttm);
+	if (mm && mm != current->mm)
+		return -EPERM;
+
+	if (abo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID &&
+	    abo->tbo.resv != vm->root.base.bo->tbo.resv)
+		return -EPERM;
+
 	r = amdgpu_bo_reserve(abo, false);
 	if (r)
 		return r;
@@ -127,35 +138,6 @@ int amdgpu_gem_object_open(struct drm_gem_object *obj,
 	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)
 {
@@ -165,13 +147,14 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj,
 	struct amdgpu_vm *vm = &fpriv->vm;
 
 	struct amdgpu_bo_list_entry vm_pd;
-	struct list_head list;
+	struct list_head list, duplicates;
 	struct ttm_validate_buffer tv;
 	struct ww_acquire_ctx ticket;
 	struct amdgpu_bo_va *bo_va;
 	int r;
 
 	INIT_LIST_HEAD(&list);
+	INIT_LIST_HEAD(&duplicates);
 
 	tv.bo = &bo->tbo;
 	tv.shared = true;
@@ -179,7 +162,7 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj,
 
 	amdgpu_vm_get_pd_bo(vm, &list, &vm_pd);
 
-	r = ttm_eu_reserve_buffers(&ticket, &list, false, NULL);
+	r = ttm_eu_reserve_buffers(&ticket, &list, false, &duplicates);
 	if (r) {
 		dev_err(adev->dev, "leaking bo va because "
 			"we fail to reserve bo (%d)\n", r);
@@ -189,7 +172,7 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj,
 	if (bo_va && --bo_va->ref_count == 0) {
 		amdgpu_vm_bo_rmv(adev, bo_va);
 
-		if (amdgpu_gem_vm_ready(adev, vm, &list)) {
+		if (amdgpu_vm_ready(vm)) {
 			struct dma_fence *fence = NULL;
 
 			r = amdgpu_vm_clear_freed(adev, vm, &fence);
@@ -214,18 +197,22 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
 			    struct drm_file *filp)
 {
 	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
+	struct amdgpu_vm *vm = &fpriv->vm;
 	union drm_amdgpu_gem_create *args = data;
+	uint64_t flags = args->in.domain_flags;
 	uint64_t size = args->in.bo_size;
+	struct reservation_object *resv = NULL;
 	struct drm_gem_object *gobj;
 	uint32_t handle;
-	bool kernel = false;
 	int r;
 
 	/* reject invalid gem flags */
-	if (args->in.domain_flags & ~(AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
-				      AMDGPU_GEM_CREATE_NO_CPU_ACCESS |
-				      AMDGPU_GEM_CREATE_CPU_GTT_USWC |
-				      AMDGPU_GEM_CREATE_VRAM_CLEARED))
+	if (flags & ~(AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
+		      AMDGPU_GEM_CREATE_NO_CPU_ACCESS |
+		      AMDGPU_GEM_CREATE_CPU_GTT_USWC |
+		      AMDGPU_GEM_CREATE_VRAM_CLEARED |
+		      AMDGPU_GEM_CREATE_VM_ALWAYS_VALID))
 		return -EINVAL;
 
 	/* reject invalid gem domains */
@@ -240,7 +227,7 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
 	/* create a gem object to contain this object in */
 	if (args->in.domains & (AMDGPU_GEM_DOMAIN_GDS |
 	    AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA)) {
-		kernel = true;
+		flags |= AMDGPU_GEM_CREATE_NO_CPU_ACCESS;
 		if (args->in.domains == AMDGPU_GEM_DOMAIN_GDS)
 			size = size << AMDGPU_GDS_SHIFT;
 		else if (args->in.domains == AMDGPU_GEM_DOMAIN_GWS)
@@ -252,10 +239,25 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
 	}
 	size = roundup(size, PAGE_SIZE);
 
+	if (flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) {
+		r = amdgpu_bo_reserve(vm->root.base.bo, false);
+		if (r)
+			return r;
+
+		resv = vm->root.base.bo->tbo.resv;
+	}
+
 	r = amdgpu_gem_object_create(adev, size, args->in.alignment,
 				     (u32)(0xffffffff & args->in.domains),
-				     args->in.domain_flags,
-				     kernel, &gobj);
+				     flags, false, resv, &gobj);
+	if (flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) {
+		if (!r) {
+			struct amdgpu_bo *abo = gem_to_amdgpu_bo(gobj);
+
+			abo->parent = amdgpu_bo_ref(vm->root.base.bo);
+		}
+		amdgpu_bo_unreserve(vm->root.base.bo);
+	}
 	if (r)
 		return r;
 
@@ -297,9 +299,8 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 	}
 
 	/* create a gem object to contain this object in */
-	r = amdgpu_gem_object_create(adev, args->size, 0,
-				     AMDGPU_GEM_DOMAIN_CPU, 0,
-				     0, &gobj);
+	r = amdgpu_gem_object_create(adev, args->size, 0, AMDGPU_GEM_DOMAIN_CPU,
+				     0, 0, NULL, &gobj);
 	if (r)
 		return r;
 
@@ -317,8 +318,6 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 	}
 
 	if (args->flags & AMDGPU_GEM_USERPTR_VALIDATE) {
-		down_read(&current->mm->mmap_sem);
-
 		r = amdgpu_ttm_tt_get_user_pages(bo->tbo.ttm,
 						 bo->tbo.ttm->pages);
 		if (r)
@@ -333,8 +332,6 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 		amdgpu_bo_unreserve(bo);
 		if (r)
 			goto free_pages;
-
-		up_read(&current->mm->mmap_sem);
 	}
 
 	r = drm_gem_handle_create(filp, gobj, &handle);
@@ -511,10 +508,10 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
 				    struct list_head *list,
 				    uint32_t operation)
 {
-	int r = -ERESTARTSYS;
+	int r;
 
-	if (!amdgpu_gem_vm_ready(adev, vm, list))
-		goto error;
+	if (!amdgpu_vm_ready(vm))
+		return;
 
 	r = amdgpu_vm_update_directories(adev, vm);
 	if (r)
@@ -551,7 +548,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 	struct amdgpu_bo_list_entry vm_pd;
 	struct ttm_validate_buffer tv;
 	struct ww_acquire_ctx ticket;
-	struct list_head list;
+	struct list_head list, duplicates;
 	uint64_t va_flags;
 	int r = 0;
 
@@ -587,6 +584,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 	}
 
 	INIT_LIST_HEAD(&list);
+	INIT_LIST_HEAD(&duplicates);
 	if ((args->operation != AMDGPU_VA_OP_CLEAR) &&
 	    !(args->flags & AMDGPU_VM_PAGE_PRT)) {
 		gobj = drm_gem_object_lookup(filp, args->handle);
@@ -603,7 +601,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 
 	amdgpu_vm_get_pd_bo(&fpriv->vm, &list, &vm_pd);
 
-	r = ttm_eu_reserve_buffers(&ticket, &list, true, NULL);
+	r = ttm_eu_reserve_buffers(&ticket, &list, true, &duplicates);
 	if (r)
 		goto error_unref;
 
@@ -669,6 +667,7 @@ error_unref:
 int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *filp)
 {
+	struct amdgpu_device *adev = dev->dev_private;
 	struct drm_amdgpu_gem_op *args = data;
 	struct drm_gem_object *gobj;
 	struct amdgpu_bo *robj;
@@ -716,6 +715,9 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 		if (robj->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM)
 			robj->allowed_domains |= AMDGPU_GEM_DOMAIN_GTT;
 
+		if (robj->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID)
+			amdgpu_vm_bo_invalidate(adev, robj, true);
+
 		amdgpu_bo_unreserve(robj);
 		break;
 	default:
@@ -745,8 +747,7 @@ int amdgpu_mode_dumb_create(struct drm_file *file_priv,
 	r = amdgpu_gem_object_create(adev, args->size, 0,
 				     AMDGPU_GEM_DOMAIN_VRAM,
 				     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-				     ttm_bo_type_device,
-				     &gobj);
+				     false, NULL, &gobj);
 	if (r)
 		return -ENOMEM;
 

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

@@ -260,8 +260,13 @@ int amdgpu_gfx_compute_mqd_sw_init(struct amdgpu_device *adev,
 	/* create MQD for KIQ */
 	ring = &adev->gfx.kiq.ring;
 	if (!ring->mqd_obj) {
+		/* originaly the KIQ MQD is put in GTT domain, but for SRIOV VRAM domain is a must
+		 * otherwise hypervisor trigger SAVE_VF fail after driver unloaded which mean MQD
+		 * deallocated and gart_unbind, to strict diverage we decide to use VRAM domain for
+		 * KIQ MQD no matter SRIOV or Bare-metal
+		 */
 		r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE,
-					    AMDGPU_GEM_DOMAIN_GTT, &ring->mqd_obj,
+					    AMDGPU_GEM_DOMAIN_VRAM, &ring->mqd_obj,
 					    &ring->mqd_gpu_addr, &ring->mqd_ptr);
 		if (r) {
 			dev_warn(adev->dev, "failed to create ring mqd ob (%d)", r);

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

@@ -169,6 +169,12 @@ restart_ih:
 	while (adev->irq.ih.rptr != wptr) {
 		u32 ring_index = adev->irq.ih.rptr >> 2;
 
+		/* Prescreening of high-frequency interrupts */
+		if (!amdgpu_ih_prescreen_iv(adev)) {
+			adev->irq.ih.rptr &= adev->irq.ih.ptr_mask;
+			continue;
+		}
+
 		/* Before dispatching irq to IP blocks, send it to amdkfd */
 		amdgpu_amdkfd_interrupt(adev,
 				(const void *) &adev->irq.ih.ring[ring_index]);
@@ -190,3 +196,79 @@ restart_ih:
 
 	return IRQ_HANDLED;
 }
+
+/**
+ * amdgpu_ih_add_fault - Add a page fault record
+ *
+ * @adev: amdgpu device pointer
+ * @key: 64-bit encoding of PASID and address
+ *
+ * This should be called when a retry page fault interrupt is
+ * received. If this is a new page fault, it will be added to a hash
+ * table. The return value indicates whether this is a new fault, or
+ * a fault that was already known and is already being handled.
+ *
+ * If there are too many pending page faults, this will fail. Retry
+ * interrupts should be ignored in this case until there is enough
+ * free space.
+ *
+ * Returns 0 if the fault was added, 1 if the fault was already known,
+ * -ENOSPC if there are too many pending faults.
+ */
+int amdgpu_ih_add_fault(struct amdgpu_device *adev, u64 key)
+{
+	unsigned long flags;
+	int r = -ENOSPC;
+
+	if (WARN_ON_ONCE(!adev->irq.ih.faults))
+		/* Should be allocated in <IP>_ih_sw_init on GPUs that
+		 * support retry faults and require retry filtering.
+		 */
+		return r;
+
+	spin_lock_irqsave(&adev->irq.ih.faults->lock, flags);
+
+	/* Only let the hash table fill up to 50% for best performance */
+	if (adev->irq.ih.faults->count >= (1 << (AMDGPU_PAGEFAULT_HASH_BITS-1)))
+		goto unlock_out;
+
+	r = chash_table_copy_in(&adev->irq.ih.faults->hash, key, NULL);
+	if (!r)
+		adev->irq.ih.faults->count++;
+
+	/* chash_table_copy_in should never fail unless we're losing count */
+	WARN_ON_ONCE(r < 0);
+
+unlock_out:
+	spin_unlock_irqrestore(&adev->irq.ih.faults->lock, flags);
+	return r;
+}
+
+/**
+ * amdgpu_ih_clear_fault - Remove a page fault record
+ *
+ * @adev: amdgpu device pointer
+ * @key: 64-bit encoding of PASID and address
+ *
+ * This should be called when a page fault has been handled. Any
+ * future interrupt with this key will be processed as a new
+ * page fault.
+ */
+void amdgpu_ih_clear_fault(struct amdgpu_device *adev, u64 key)
+{
+	unsigned long flags;
+	int r;
+
+	if (!adev->irq.ih.faults)
+		return;
+
+	spin_lock_irqsave(&adev->irq.ih.faults->lock, flags);
+
+	r = chash_table_remove(&adev->irq.ih.faults->hash, key, NULL);
+	if (!WARN_ON_ONCE(r < 0)) {
+		adev->irq.ih.faults->count--;
+		WARN_ON_ONCE(adev->irq.ih.faults->count < 0);
+	}
+
+	spin_unlock_irqrestore(&adev->irq.ih.faults->lock, flags);
+}

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

@@ -24,6 +24,8 @@
 #ifndef __AMDGPU_IH_H__
 #define __AMDGPU_IH_H__
 
+#include <linux/chash.h>
+
 struct amdgpu_device;
  /*
   * vega10+ IH clients
@@ -69,6 +71,13 @@ enum amdgpu_ih_clientid
 
 #define AMDGPU_IH_CLIENTID_LEGACY 0
 
+#define AMDGPU_PAGEFAULT_HASH_BITS 8
+struct amdgpu_retryfault_hashtable {
+	DECLARE_CHASH_TABLE(hash, AMDGPU_PAGEFAULT_HASH_BITS, 8, 0);
+	spinlock_t	lock;
+	int		count;
+};
+
 /*
  * R6xx+ IH ring
  */
@@ -87,6 +96,7 @@ struct amdgpu_ih_ring {
 	bool			use_doorbell;
 	bool			use_bus_addr;
 	dma_addr_t		rb_dma_addr; /* only used when use_bus_addr = true */
+	struct amdgpu_retryfault_hashtable *faults;
 };
 
 #define AMDGPU_IH_SRC_DATA_MAX_SIZE_DW 4
@@ -109,5 +119,7 @@ int amdgpu_ih_ring_init(struct amdgpu_device *adev, unsigned ring_size,
 			bool use_bus_addr);
 void amdgpu_ih_ring_fini(struct amdgpu_device *adev);
 int amdgpu_ih_process(struct amdgpu_device *adev);
+int amdgpu_ih_add_fault(struct amdgpu_device *adev, u64 key);
+void amdgpu_ih_clear_fault(struct amdgpu_device *adev, u64 key);
 
 #endif

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

@@ -825,7 +825,7 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv)
 	}
 
 	r = amdgpu_vm_init(adev, &fpriv->vm,
-			   AMDGPU_VM_CONTEXT_GFX);
+			   AMDGPU_VM_CONTEXT_GFX, 0);
 	if (r) {
 		kfree(fpriv);
 		goto out_suspend;
@@ -841,8 +841,11 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv)
 
 	if (amdgpu_sriov_vf(adev)) {
 		r = amdgpu_map_static_csa(adev, &fpriv->vm, &fpriv->csa_va);
-		if (r)
+		if (r) {
+			amdgpu_vm_fini(adev, &fpriv->vm);
+			kfree(fpriv);
 			goto out_suspend;
+		}
 	}
 
 	mutex_init(&fpriv->bo_list_lock);

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

@@ -50,8 +50,10 @@ struct amdgpu_mn {
 	struct hlist_node	node;
 
 	/* objects protected by lock */
-	struct mutex		lock;
+	struct rw_semaphore	lock;
 	struct rb_root_cached	objects;
+	struct mutex		read_lock;
+	atomic_t		recursion;
 };
 
 struct amdgpu_mn_node {
@@ -74,7 +76,7 @@ static void amdgpu_mn_destroy(struct work_struct *work)
 	struct amdgpu_bo *bo, *next_bo;
 
 	mutex_lock(&adev->mn_lock);
-	mutex_lock(&rmn->lock);
+	down_write(&rmn->lock);
 	hash_del(&rmn->node);
 	rbtree_postorder_for_each_entry_safe(node, next_node,
 					     &rmn->objects.rb_root, it.rb) {
@@ -84,7 +86,7 @@ static void amdgpu_mn_destroy(struct work_struct *work)
 		}
 		kfree(node);
 	}
-	mutex_unlock(&rmn->lock);
+	up_write(&rmn->lock);
 	mutex_unlock(&adev->mn_lock);
 	mmu_notifier_unregister_no_release(&rmn->mn, rmn->mm);
 	kfree(rmn);
@@ -106,6 +108,53 @@ static void amdgpu_mn_release(struct mmu_notifier *mn,
 	schedule_work(&rmn->work);
 }
 
+
+/**
+ * amdgpu_mn_lock - take the write side lock for this mn
+ */
+void amdgpu_mn_lock(struct amdgpu_mn *mn)
+{
+	if (mn)
+		down_write(&mn->lock);
+}
+
+/**
+ * amdgpu_mn_unlock - drop the write side lock for this mn
+ */
+void amdgpu_mn_unlock(struct amdgpu_mn *mn)
+{
+	if (mn)
+		up_write(&mn->lock);
+}
+
+/**
+ * amdgpu_mn_read_lock - take the rmn read lock
+ *
+ * @rmn: our notifier
+ *
+ * Take the rmn read side lock.
+ */
+static void amdgpu_mn_read_lock(struct amdgpu_mn *rmn)
+{
+	mutex_lock(&rmn->read_lock);
+	if (atomic_inc_return(&rmn->recursion) == 1)
+		down_read_non_owner(&rmn->lock);
+	mutex_unlock(&rmn->read_lock);
+}
+
+/**
+ * amdgpu_mn_read_unlock - drop the rmn read lock
+ *
+ * @rmn: our notifier
+ *
+ * Drop the rmn read side lock.
+ */
+static void amdgpu_mn_read_unlock(struct amdgpu_mn *rmn)
+{
+	if (atomic_dec_return(&rmn->recursion) == 0)
+		up_read_non_owner(&rmn->lock);
+}
+
 /**
  * amdgpu_mn_invalidate_node - unmap all BOs of a node
  *
@@ -126,23 +175,12 @@ static void amdgpu_mn_invalidate_node(struct amdgpu_mn_node *node,
 		if (!amdgpu_ttm_tt_affect_userptr(bo->tbo.ttm, start, end))
 			continue;
 
-		r = amdgpu_bo_reserve(bo, true);
-		if (r) {
-			DRM_ERROR("(%ld) failed to reserve user bo\n", r);
-			continue;
-		}
-
 		r = reservation_object_wait_timeout_rcu(bo->tbo.resv,
 			true, false, MAX_SCHEDULE_TIMEOUT);
 		if (r <= 0)
 			DRM_ERROR("(%ld) failed to wait for user bo\n", r);
 
-		amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU);
-		r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false);
-		if (r)
-			DRM_ERROR("(%ld) failed to validate user bo\n", r);
-
-		amdgpu_bo_unreserve(bo);
+		amdgpu_ttm_tt_mark_user_pages(bo->tbo.ttm);
 	}
 }
 
@@ -168,7 +206,7 @@ static void amdgpu_mn_invalidate_range_start(struct mmu_notifier *mn,
 	/* notification is exclusive, but interval is inclusive */
 	end -= 1;
 
-	mutex_lock(&rmn->lock);
+	amdgpu_mn_read_lock(rmn);
 
 	it = interval_tree_iter_first(&rmn->objects, start, end);
 	while (it) {
@@ -179,13 +217,32 @@ static void amdgpu_mn_invalidate_range_start(struct mmu_notifier *mn,
 
 		amdgpu_mn_invalidate_node(node, start, end);
 	}
+}
 
-	mutex_unlock(&rmn->lock);
+/**
+ * amdgpu_mn_invalidate_range_end - callback to notify about mm change
+ *
+ * @mn: our notifier
+ * @mn: the mm this callback is about
+ * @start: start of updated range
+ * @end: end of updated range
+ *
+ * Release the lock again to allow new command submissions.
+ */
+static void amdgpu_mn_invalidate_range_end(struct mmu_notifier *mn,
+					   struct mm_struct *mm,
+					   unsigned long start,
+					   unsigned long end)
+{
+	struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn);
+
+	amdgpu_mn_read_unlock(rmn);
 }
 
 static const struct mmu_notifier_ops amdgpu_mn_ops = {
 	.release = amdgpu_mn_release,
 	.invalidate_range_start = amdgpu_mn_invalidate_range_start,
+	.invalidate_range_end = amdgpu_mn_invalidate_range_end,
 };
 
 /**
@@ -195,7 +252,7 @@ static const struct mmu_notifier_ops amdgpu_mn_ops = {
  *
  * Creates a notifier context for current->mm.
  */
-static struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev)
+struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev)
 {
 	struct mm_struct *mm = current->mm;
 	struct amdgpu_mn *rmn;
@@ -220,8 +277,10 @@ static struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev)
 	rmn->adev = adev;
 	rmn->mm = mm;
 	rmn->mn.ops = &amdgpu_mn_ops;
-	mutex_init(&rmn->lock);
+	init_rwsem(&rmn->lock);
 	rmn->objects = RB_ROOT_CACHED;
+	mutex_init(&rmn->read_lock);
+	atomic_set(&rmn->recursion, 0);
 
 	r = __mmu_notifier_register(&rmn->mn, mm);
 	if (r)
@@ -267,7 +326,7 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
 
 	INIT_LIST_HEAD(&bos);
 
-	mutex_lock(&rmn->lock);
+	down_write(&rmn->lock);
 
 	while ((it = interval_tree_iter_first(&rmn->objects, addr, end))) {
 		kfree(node);
@@ -281,7 +340,7 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
 	if (!node) {
 		node = kmalloc(sizeof(struct amdgpu_mn_node), GFP_KERNEL);
 		if (!node) {
-			mutex_unlock(&rmn->lock);
+			up_write(&rmn->lock);
 			return -ENOMEM;
 		}
 	}
@@ -296,7 +355,7 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
 
 	interval_tree_insert(&node->it, &rmn->objects);
 
-	mutex_unlock(&rmn->lock);
+	up_write(&rmn->lock);
 
 	return 0;
 }
@@ -322,7 +381,7 @@ void amdgpu_mn_unregister(struct amdgpu_bo *bo)
 		return;
 	}
 
-	mutex_lock(&rmn->lock);
+	down_write(&rmn->lock);
 
 	/* save the next list entry for later */
 	head = bo->mn_list.next;
@@ -337,6 +396,7 @@ void amdgpu_mn_unregister(struct amdgpu_bo *bo)
 		kfree(node);
 	}
 
-	mutex_unlock(&rmn->lock);
+	up_write(&rmn->lock);
 	mutex_unlock(&adev->mn_lock);
 }
+

drivers/gpu/drm/amd/powerplay/eventmgr/psm.h → drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h

@@ -1,5 +1,5 @@
 /*
- * Copyright 2015 Advanced Micro Devices, Inc.
+ * Copyright 2017 Advanced Micro Devices, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -19,20 +19,34 @@
  * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
  * OTHER DEALINGS IN THE SOFTWARE.
  *
+ * Authors: Christian König
  */
-#include "eventmgr.h"
-#include "eventinit.h"
-#include "eventmanagement.h"
-#include "eventmanager.h"
-#include "power_state.h"
-#include "hardwaremanager.h"
+#ifndef __AMDGPU_MN_H__
+#define __AMDGPU_MN_H__
 
-int psm_get_ui_state(struct pp_eventmgr *eventmgr, enum PP_StateUILabel ui_label, unsigned long *state_id);
-
-int psm_get_state_by_classification(struct pp_eventmgr *eventmgr, enum PP_StateClassificationFlag flag, unsigned long *state_id);
-
-int psm_set_states(struct pp_eventmgr *eventmgr, unsigned long *state_id);
+/*
+ * MMU Notifier
+ */
+struct amdgpu_mn;
 
-int psm_adjust_power_state_dynamic(struct pp_eventmgr *eventmgr, bool skip);
+#if defined(CONFIG_MMU_NOTIFIER)
+void amdgpu_mn_lock(struct amdgpu_mn *mn);
+void amdgpu_mn_unlock(struct amdgpu_mn *mn);
+struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev);
+int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr);
+void amdgpu_mn_unregister(struct amdgpu_bo *bo);
+#else
+static inline void amdgpu_mn_lock(struct amdgpu_mn *mn) {}
+static inline void amdgpu_mn_unlock(struct amdgpu_mn *mn) {}
+static inline struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev)
+{
+	return NULL;
+}
+static inline int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
+{
+	return -ENODEV;
+}
+static inline void amdgpu_mn_unregister(struct amdgpu_bo *bo) {}
+#endif
 
-int psm_adjust_power_state_static(struct pp_eventmgr *eventmgr, bool skip);
+#endif

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

@@ -64,11 +64,12 @@ bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo)
 	return false;
 }
 
-static void amdgpu_ttm_placement_init(struct amdgpu_device *adev,
-				      struct ttm_placement *placement,
-				      struct ttm_place *places,
-				      u32 domain, u64 flags)
+void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 {
+	struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev);
+	struct ttm_placement *placement = &abo->placement;
+	struct ttm_place *places = abo->placements;
+	u64 flags = abo->flags;
 	u32 c = 0;
 
 	if (domain & AMDGPU_GEM_DOMAIN_VRAM) {
@@ -151,27 +152,6 @@ static void amdgpu_ttm_placement_init(struct amdgpu_device *adev,
 	placement->busy_placement = places;
 }
 
-void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
-{
-	struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev);
-
-	amdgpu_ttm_placement_init(adev, &abo->placement, abo->placements,
-				  domain, abo->flags);
-}
-
-static void amdgpu_fill_placement_to_bo(struct amdgpu_bo *bo,
-					struct ttm_placement *placement)
-{
-	BUG_ON(placement->num_placement > (AMDGPU_GEM_DOMAIN_MAX + 1));
-
-	memcpy(bo->placements, placement->placement,
-	       placement->num_placement * sizeof(struct ttm_place));
-	bo->placement.num_placement = placement->num_placement;
-	bo->placement.num_busy_placement = placement->num_busy_placement;
-	bo->placement.placement = bo->placements;
-	bo->placement.busy_placement = bo->placements;
-}
-
 /**
  * amdgpu_bo_create_reserved - create reserved BO for kernel use
  *
@@ -303,14 +283,13 @@ void amdgpu_bo_free_kernel(struct amdgpu_bo **bo, u64 *gpu_addr,
 		*cpu_addr = NULL;
 }
 
-int amdgpu_bo_create_restricted(struct amdgpu_device *adev,
-				unsigned long size, int byte_align,
-				bool kernel, u32 domain, u64 flags,
-				struct sg_table *sg,
-				struct ttm_placement *placement,
-				struct reservation_object *resv,
-				uint64_t init_value,
-				struct amdgpu_bo **bo_ptr)
+static int amdgpu_bo_do_create(struct amdgpu_device *adev,
+			       unsigned long size, int byte_align,
+			       bool kernel, u32 domain, u64 flags,
+			       struct sg_table *sg,
+			       struct reservation_object *resv,
+			       uint64_t init_value,
+			       struct amdgpu_bo **bo_ptr)
 {
 	struct amdgpu_bo *bo;
 	enum ttm_bo_type type;
@@ -384,10 +363,11 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev,
 		bo->flags &= ~AMDGPU_GEM_CREATE_CPU_GTT_USWC;
 #endif
 
-	amdgpu_fill_placement_to_bo(bo, placement);
-	/* Kernel allocation are uninterruptible */
+	bo->tbo.bdev = &adev->mman.bdev;
+	amdgpu_ttm_placement_from_domain(bo, domain);
 
 	initial_bytes_moved = atomic64_read(&adev->num_bytes_moved);
+	/* Kernel allocation are uninterruptible */
 	r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, type,
 				 &bo->placement, page_align, !kernel, NULL,
 				 acc_size, sg, resv, &amdgpu_ttm_bo_destroy);
@@ -442,27 +422,17 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev,
 				   unsigned long size, int byte_align,
 				   struct amdgpu_bo *bo)
 {
-	struct ttm_placement placement = {0};
-	struct ttm_place placements[AMDGPU_GEM_DOMAIN_MAX + 1];
 	int r;
 
 	if (bo->shadow)
 		return 0;
 
-	memset(&placements, 0, sizeof(placements));
-	amdgpu_ttm_placement_init(adev, &placement, placements,
-				  AMDGPU_GEM_DOMAIN_GTT,
-				  AMDGPU_GEM_CREATE_CPU_GTT_USWC |
-				  AMDGPU_GEM_CREATE_SHADOW);
-
-	r = amdgpu_bo_create_restricted(adev, size, byte_align, true,
-					AMDGPU_GEM_DOMAIN_GTT,
-					AMDGPU_GEM_CREATE_CPU_GTT_USWC |
-					AMDGPU_GEM_CREATE_SHADOW,
-					NULL, &placement,
-					bo->tbo.resv,
-					0,
-					&bo->shadow);
+	r = amdgpu_bo_do_create(adev, size, byte_align, true,
+				AMDGPU_GEM_DOMAIN_GTT,
+				AMDGPU_GEM_CREATE_CPU_GTT_USWC |
+				AMDGPU_GEM_CREATE_SHADOW,
+				NULL, bo->tbo.resv, 0,
+				&bo->shadow);
 	if (!r) {
 		bo->shadow->parent = amdgpu_bo_ref(bo);
 		mutex_lock(&adev->shadow_list_lock);
@@ -484,18 +454,11 @@ int amdgpu_bo_create(struct amdgpu_device *adev,
 		     uint64_t init_value,
 		     struct amdgpu_bo **bo_ptr)
 {
-	struct ttm_placement placement = {0};
-	struct ttm_place placements[AMDGPU_GEM_DOMAIN_MAX + 1];
 	uint64_t parent_flags = flags & ~AMDGPU_GEM_CREATE_SHADOW;
 	int r;
 
-	memset(&placements, 0, sizeof(placements));
-	amdgpu_ttm_placement_init(adev, &placement, placements,
-				  domain, parent_flags);
-
-	r = amdgpu_bo_create_restricted(adev, size, byte_align, kernel, domain,
-					parent_flags, sg, &placement, resv,
-					init_value, bo_ptr);
+	r = amdgpu_bo_do_create(adev, size, byte_align, kernel, domain,
+				parent_flags, sg, resv, init_value, bo_ptr);
 	if (r)
 		return r;
 
@@ -672,7 +635,6 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
 	int r, i;
-	unsigned fpfn, lpfn;
 
 	if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm))
 		return -EPERM;
@@ -704,22 +666,16 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
 	}
 
 	bo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
+	/* force to pin into visible video ram */
+	if (!(bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS))
+		bo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
 	amdgpu_ttm_placement_from_domain(bo, domain);
 	for (i = 0; i < bo->placement.num_placement; i++) {
-		/* force to pin into visible video ram */
-		if ((bo->placements[i].flags & TTM_PL_FLAG_VRAM) &&
-		    !(bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) &&
-		    (!max_offset || max_offset >
-		     adev->mc.visible_vram_size)) {
-			if (WARN_ON_ONCE(min_offset >
-					 adev->mc.visible_vram_size))
-				return -EINVAL;
-			fpfn = min_offset >> PAGE_SHIFT;
-			lpfn = adev->mc.visible_vram_size >> PAGE_SHIFT;
-		} else {
-			fpfn = min_offset >> PAGE_SHIFT;
-			lpfn = max_offset >> PAGE_SHIFT;
-		}
+		unsigned fpfn, lpfn;
+
+		fpfn = min_offset >> PAGE_SHIFT;
+		lpfn = max_offset >> PAGE_SHIFT;
+
 		if (fpfn > bo->placements[i].fpfn)
 			bo->placements[i].fpfn = fpfn;
 		if (!bo->placements[i].lpfn ||
@@ -929,7 +885,7 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo,
 		return;
 
 	abo = container_of(bo, struct amdgpu_bo, tbo);
-	amdgpu_vm_bo_invalidate(adev, abo);
+	amdgpu_vm_bo_invalidate(adev, abo, evict);
 
 	amdgpu_bo_kunmap(abo);
 

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

@@ -35,6 +35,7 @@
 
 /* bo virtual addresses in a vm */
 struct amdgpu_bo_va_mapping {
+	struct amdgpu_bo_va		*bo_va;
 	struct list_head		list;
 	struct rb_node			rb;
 	uint64_t			start;
@@ -49,12 +50,17 @@ struct amdgpu_bo_va {
 	struct amdgpu_vm_bo_base	base;
 
 	/* protected by bo being reserved */
-	struct dma_fence	        *last_pt_update;
 	unsigned			ref_count;
 
+	/* all other members protected by the VM PD being reserved */
+	struct dma_fence	        *last_pt_update;
+
 	/* mappings for this bo_va */
 	struct list_head		invalids;
 	struct list_head		valids;
+
+	/* If the mappings are cleared or filled */
+	bool				cleared;
 };
 
 struct amdgpu_bo {
@@ -189,14 +195,6 @@ int amdgpu_bo_create(struct amdgpu_device *adev,
 			    struct reservation_object *resv,
 			    uint64_t init_value,
 			    struct amdgpu_bo **bo_ptr);
-int amdgpu_bo_create_restricted(struct amdgpu_device *adev,
-				unsigned long size, int byte_align,
-				bool kernel, u32 domain, u64 flags,
-				struct sg_table *sg,
-				struct ttm_placement *placement,
-			        struct reservation_object *resv,
-				uint64_t init_value,
-				struct amdgpu_bo **bo_ptr);
 int amdgpu_bo_create_reserved(struct amdgpu_device *adev,
 			      unsigned long size, int align,
 			      u32 domain, struct amdgpu_bo **bo_ptr,

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

@@ -74,7 +74,7 @@ void amdgpu_pm_acpi_event_handler(struct amdgpu_device *adev)
 			adev->pm.dpm.ac_power = true;
 		else
 			adev->pm.dpm.ac_power = false;
-		if (adev->pm.funcs->enable_bapm)
+		if (adev->powerplay.pp_funcs->enable_bapm)
 			amdgpu_dpm_enable_bapm(adev, adev->pm.dpm.ac_power);
 		mutex_unlock(&adev->pm.mutex);
 	}
@@ -88,9 +88,9 @@ static ssize_t amdgpu_get_dpm_state(struct device *dev,
 	struct amdgpu_device *adev = ddev->dev_private;
 	enum amd_pm_state_type pm;
 
-	if (adev->pp_enabled) {
+	if (adev->powerplay.pp_funcs->get_current_power_state)
 		pm = amdgpu_dpm_get_current_power_state(adev);
-	} else
+	else
 		pm = adev->pm.dpm.user_state;
 
 	return snprintf(buf, PAGE_SIZE, "%s\n",
@@ -119,7 +119,7 @@ static ssize_t amdgpu_set_dpm_state(struct device *dev,
 	}
 
 	if (adev->pp_enabled) {
-		amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_ENABLE_USER_STATE, &state, NULL);
+		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_ENABLE_USER_STATE, &state, NULL);
 	} else {
 		mutex_lock(&adev->pm.mutex);
 		adev->pm.dpm.user_state = state;
@@ -140,13 +140,17 @@ static ssize_t amdgpu_get_dpm_forced_performance_level(struct device *dev,
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
-	enum amd_dpm_forced_level level;
+	enum amd_dpm_forced_level level = 0xff;
 
 	if  ((adev->flags & AMD_IS_PX) &&
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return snprintf(buf, PAGE_SIZE, "off\n");
 
-	level = amdgpu_dpm_get_performance_level(adev);
+	if (adev->powerplay.pp_funcs->get_performance_level)
+		level = amdgpu_dpm_get_performance_level(adev);
+	else
+		level = adev->pm.dpm.forced_level;
+
 	return snprintf(buf, PAGE_SIZE, "%s\n",
 			(level == AMD_DPM_FORCED_LEVEL_AUTO) ? "auto" :
 			(level == AMD_DPM_FORCED_LEVEL_LOW) ? "low" :
@@ -167,7 +171,7 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev,
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
 	enum amd_dpm_forced_level level;
-	enum amd_dpm_forced_level current_level;
+	enum amd_dpm_forced_level current_level = 0xff;
 	int ret = 0;
 
 	/* Can't force performance level when the card is off */
@@ -175,7 +179,8 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev,
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
 
-	current_level = amdgpu_dpm_get_performance_level(adev);
+	if (adev->powerplay.pp_funcs->get_performance_level)
+		current_level = amdgpu_dpm_get_performance_level(adev);
 
 	if (strncmp("low", buf, strlen("low")) == 0) {
 		level = AMD_DPM_FORCED_LEVEL_LOW;
@@ -203,9 +208,7 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev,
 	if (current_level == level)
 		return count;
 
-	if (adev->pp_enabled)
-		amdgpu_dpm_force_performance_level(adev, level);
-	else {
+	if (adev->powerplay.pp_funcs->force_performance_level) {
 		mutex_lock(&adev->pm.mutex);
 		if (adev->pm.dpm.thermal_active) {
 			count = -EINVAL;
@@ -233,7 +236,7 @@ static ssize_t amdgpu_get_pp_num_states(struct device *dev,
 	struct pp_states_info data;
 	int i, buf_len;
 
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->get_pp_num_states)
 		amdgpu_dpm_get_pp_num_states(adev, &data);
 
 	buf_len = snprintf(buf, PAGE_SIZE, "states: %d\n", data.nums);
@@ -257,8 +260,8 @@ static ssize_t amdgpu_get_pp_cur_state(struct device *dev,
 	enum amd_pm_state_type pm = 0;
 	int i = 0;
 
-	if (adev->pp_enabled) {
-
+	if (adev->powerplay.pp_funcs->get_current_power_state
+		 && adev->powerplay.pp_funcs->get_pp_num_states) {
 		pm = amdgpu_dpm_get_current_power_state(adev);
 		amdgpu_dpm_get_pp_num_states(adev, &data);
 
@@ -280,25 +283,10 @@ static ssize_t amdgpu_get_pp_force_state(struct device *dev,
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
-	struct pp_states_info data;
-	enum amd_pm_state_type pm = 0;
-	int i;
-
-	if (adev->pp_force_state_enabled && adev->pp_enabled) {
-		pm = amdgpu_dpm_get_current_power_state(adev);
-		amdgpu_dpm_get_pp_num_states(adev, &data);
-
-		for (i = 0; i < data.nums; i++) {
-			if (pm == data.states[i])
-				break;
-		}
 
-		if (i == data.nums)
-			i = -EINVAL;
-
-		return snprintf(buf, PAGE_SIZE, "%d\n", i);
-
-	} else
+	if (adev->pp_force_state_enabled)
+		return amdgpu_get_pp_cur_state(dev, attr, buf);
+	else
 		return snprintf(buf, PAGE_SIZE, "\n");
 }
 
@@ -330,7 +318,7 @@ static ssize_t amdgpu_set_pp_force_state(struct device *dev,
 		if (state != POWER_STATE_TYPE_INTERNAL_BOOT &&
 		    state != POWER_STATE_TYPE_DEFAULT) {
 			amdgpu_dpm_dispatch_task(adev,
-					AMD_PP_EVENT_ENABLE_USER_STATE, &state, NULL);
+					AMD_PP_TASK_ENABLE_USER_STATE, &state, NULL);
 			adev->pp_force_state_enabled = true;
 		}
 	}
@@ -347,7 +335,7 @@ static ssize_t amdgpu_get_pp_table(struct device *dev,
 	char *table = NULL;
 	int size;
 
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->get_pp_table)
 		size = amdgpu_dpm_get_pp_table(adev, &table);
 	else
 		return 0;
@@ -368,7 +356,7 @@ static ssize_t amdgpu_set_pp_table(struct device *dev,
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
 
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->set_pp_table)
 		amdgpu_dpm_set_pp_table(adev, buf, count);
 
 	return count;
@@ -380,14 +368,11 @@ static ssize_t amdgpu_get_pp_dpm_sclk(struct device *dev,
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
-	ssize_t size = 0;
 
-	if (adev->pp_enabled)
-		size = amdgpu_dpm_print_clock_levels(adev, PP_SCLK, buf);
-	else if (adev->pm.funcs->print_clock_levels)
-		size = adev->pm.funcs->print_clock_levels(adev, PP_SCLK, buf);
-
-	return size;
+	if (adev->powerplay.pp_funcs->print_clock_levels)
+		return amdgpu_dpm_print_clock_levels(adev, PP_SCLK, buf);
+	else
+		return snprintf(buf, PAGE_SIZE, "\n");
 }
 
 static ssize_t amdgpu_set_pp_dpm_sclk(struct device *dev,
@@ -416,10 +401,9 @@ static ssize_t amdgpu_set_pp_dpm_sclk(struct device *dev,
 		mask |= 1 << level;
 	}
 
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->force_clock_level)
 		amdgpu_dpm_force_clock_level(adev, PP_SCLK, mask);
-	else if (adev->pm.funcs->force_clock_level)
-		adev->pm.funcs->force_clock_level(adev, PP_SCLK, mask);
+
 fail:
 	return count;
 }
@@ -430,14 +414,11 @@ static ssize_t amdgpu_get_pp_dpm_mclk(struct device *dev,
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
-	ssize_t size = 0;
-
-	if (adev->pp_enabled)
-		size = amdgpu_dpm_print_clock_levels(adev, PP_MCLK, buf);
-	else if (adev->pm.funcs->print_clock_levels)
-		size = adev->pm.funcs->print_clock_levels(adev, PP_MCLK, buf);
 
-	return size;
+	if (adev->powerplay.pp_funcs->print_clock_levels)
+		return amdgpu_dpm_print_clock_levels(adev, PP_MCLK, buf);
+	else
+		return snprintf(buf, PAGE_SIZE, "\n");
 }
 
 static ssize_t amdgpu_set_pp_dpm_mclk(struct device *dev,
@@ -465,11 +446,9 @@ static ssize_t amdgpu_set_pp_dpm_mclk(struct device *dev,
 		}
 		mask |= 1 << level;
 	}
-
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->force_clock_level)
 		amdgpu_dpm_force_clock_level(adev, PP_MCLK, mask);
-	else if (adev->pm.funcs->force_clock_level)
-		adev->pm.funcs->force_clock_level(adev, PP_MCLK, mask);
+
 fail:
 	return count;
 }
@@ -480,14 +459,11 @@ static ssize_t amdgpu_get_pp_dpm_pcie(struct device *dev,
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
-	ssize_t size = 0;
-
-	if (adev->pp_enabled)
-		size = amdgpu_dpm_print_clock_levels(adev, PP_PCIE, buf);
-	else if (adev->pm.funcs->print_clock_levels)
-		size = adev->pm.funcs->print_clock_levels(adev, PP_PCIE, buf);
 
-	return size;
+	if (adev->powerplay.pp_funcs->print_clock_levels)
+		return amdgpu_dpm_print_clock_levels(adev, PP_PCIE, buf);
+	else
+		return snprintf(buf, PAGE_SIZE, "\n");
 }
 
 static ssize_t amdgpu_set_pp_dpm_pcie(struct device *dev,
@@ -515,11 +491,9 @@ static ssize_t amdgpu_set_pp_dpm_pcie(struct device *dev,
 		}
 		mask |= 1 << level;
 	}
-
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->force_clock_level)
 		amdgpu_dpm_force_clock_level(adev, PP_PCIE, mask);
-	else if (adev->pm.funcs->force_clock_level)
-		adev->pm.funcs->force_clock_level(adev, PP_PCIE, mask);
+
 fail:
 	return count;
 }
@@ -532,10 +506,8 @@ static ssize_t amdgpu_get_pp_sclk_od(struct device *dev,
 	struct amdgpu_device *adev = ddev->dev_private;
 	uint32_t value = 0;
 
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->get_sclk_od)
 		value = amdgpu_dpm_get_sclk_od(adev);
-	else if (adev->pm.funcs->get_sclk_od)
-		value = adev->pm.funcs->get_sclk_od(adev);
 
 	return snprintf(buf, PAGE_SIZE, "%d\n", value);
 }
@@ -556,12 +528,12 @@ static ssize_t amdgpu_set_pp_sclk_od(struct device *dev,
 		count = -EINVAL;
 		goto fail;
 	}
+	if (adev->powerplay.pp_funcs->set_sclk_od)
+		amdgpu_dpm_set_sclk_od(adev, (uint32_t)value);
 
 	if (adev->pp_enabled) {
-		amdgpu_dpm_set_sclk_od(adev, (uint32_t)value);
-		amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_READJUST_POWER_STATE, NULL, NULL);
-	} else if (adev->pm.funcs->set_sclk_od) {
-		adev->pm.funcs->set_sclk_od(adev, (uint32_t)value);
+		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL, NULL);
+	} else {
 		adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps;
 		amdgpu_pm_compute_clocks(adev);
 	}
@@ -578,10 +550,8 @@ static ssize_t amdgpu_get_pp_mclk_od(struct device *dev,
 	struct amdgpu_device *adev = ddev->dev_private;
 	uint32_t value = 0;
 
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->get_mclk_od)
 		value = amdgpu_dpm_get_mclk_od(adev);
-	else if (adev->pm.funcs->get_mclk_od)
-		value = adev->pm.funcs->get_mclk_od(adev);
 
 	return snprintf(buf, PAGE_SIZE, "%d\n", value);
 }
@@ -602,12 +572,12 @@ static ssize_t amdgpu_set_pp_mclk_od(struct device *dev,
 		count = -EINVAL;
 		goto fail;
 	}
+	if (adev->powerplay.pp_funcs->set_mclk_od)
+		amdgpu_dpm_set_mclk_od(adev, (uint32_t)value);
 
 	if (adev->pp_enabled) {
-		amdgpu_dpm_set_mclk_od(adev, (uint32_t)value);
-		amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_READJUST_POWER_STATE, NULL, NULL);
-	} else if (adev->pm.funcs->set_mclk_od) {
-		adev->pm.funcs->set_mclk_od(adev, (uint32_t)value);
+		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL, NULL);
+	} else {
 		adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps;
 		amdgpu_pm_compute_clocks(adev);
 	}
@@ -621,14 +591,11 @@ static ssize_t amdgpu_get_pp_power_profile(struct device *dev,
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
-	int ret = 0;
+	int ret = 0xff;
 
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->get_power_profile_state)
 		ret = amdgpu_dpm_get_power_profile_state(
 				adev, query);
-	else if (adev->pm.funcs->get_power_profile_state)
-		ret = adev->pm.funcs->get_power_profile_state(
-				adev, query);
 
 	if (ret)
 		return ret;
@@ -675,15 +642,12 @@ static ssize_t amdgpu_set_pp_power_profile(struct device *dev,
 	char *sub_str, buf_cpy[128], *tmp_str;
 	const char delimiter[3] = {' ', '\n', '\0'};
 	long int value;
-	int ret = 0;
+	int ret = 0xff;
 
 	if (strncmp("reset", buf, strlen("reset")) == 0) {
-		if (adev->pp_enabled)
+		if (adev->powerplay.pp_funcs->reset_power_profile_state)
 			ret = amdgpu_dpm_reset_power_profile_state(
 					adev, request);
-		else if (adev->pm.funcs->reset_power_profile_state)
-			ret = adev->pm.funcs->reset_power_profile_state(
-					adev, request);
 		if (ret) {
 			count = -EINVAL;
 			goto fail;
@@ -692,12 +656,10 @@ static ssize_t amdgpu_set_pp_power_profile(struct device *dev,
 	}
 
 	if (strncmp("set", buf, strlen("set")) == 0) {
-		if (adev->pp_enabled)
+		if (adev->powerplay.pp_funcs->set_power_profile_state)
 			ret = amdgpu_dpm_set_power_profile_state(
 					adev, request);
-		else if (adev->pm.funcs->set_power_profile_state)
-			ret = adev->pm.funcs->set_power_profile_state(
-					adev, request);
+
 		if (ret) {
 			count = -EINVAL;
 			goto fail;
@@ -745,13 +707,8 @@ static ssize_t amdgpu_set_pp_power_profile(struct device *dev,
 
 		loop++;
 	}
-
-	if (adev->pp_enabled)
-		ret = amdgpu_dpm_set_power_profile_state(
-				adev, request);
-	else if (adev->pm.funcs->set_power_profile_state)
-		ret = adev->pm.funcs->set_power_profile_state(
-				adev, request);
+	if (adev->powerplay.pp_funcs->set_power_profile_state)
+		ret = amdgpu_dpm_set_power_profile_state(adev, request);
 
 	if (ret)
 		count = -EINVAL;
@@ -831,7 +788,7 @@ static ssize_t amdgpu_hwmon_show_temp(struct device *dev,
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
 
-	if (!adev->pp_enabled && !adev->pm.funcs->get_temperature)
+	if (!adev->powerplay.pp_funcs->get_temperature)
 		temp = 0;
 	else
 		temp = amdgpu_dpm_get_temperature(adev);
@@ -862,7 +819,7 @@ static ssize_t amdgpu_hwmon_get_pwm1_enable(struct device *dev,
 	struct amdgpu_device *adev = dev_get_drvdata(dev);
 	u32 pwm_mode = 0;
 
-	if (!adev->pp_enabled && !adev->pm.funcs->get_fan_control_mode)
+	if (!adev->powerplay.pp_funcs->get_fan_control_mode)
 		return -EINVAL;
 
 	pwm_mode = amdgpu_dpm_get_fan_control_mode(adev);
@@ -879,7 +836,7 @@ static ssize_t amdgpu_hwmon_set_pwm1_enable(struct device *dev,
 	int err;
 	int value;
 
-	if (!adev->pp_enabled && !adev->pm.funcs->set_fan_control_mode)
+	if (!adev->powerplay.pp_funcs->set_fan_control_mode)
 		return -EINVAL;
 
 	err = kstrtoint(buf, 10, &value);
@@ -919,9 +876,11 @@ static ssize_t amdgpu_hwmon_set_pwm1(struct device *dev,
 
 	value = (value * 100) / 255;
 
-	err = amdgpu_dpm_set_fan_speed_percent(adev, value);
-	if (err)
-		return err;
+	if (adev->powerplay.pp_funcs->set_fan_speed_percent) {
+		err = amdgpu_dpm_set_fan_speed_percent(adev, value);
+		if (err)
+			return err;
+	}
 
 	return count;
 }
@@ -932,11 +891,13 @@ static ssize_t amdgpu_hwmon_get_pwm1(struct device *dev,
 {
 	struct amdgpu_device *adev = dev_get_drvdata(dev);
 	int err;
-	u32 speed;
+	u32 speed = 0;
 
-	err = amdgpu_dpm_get_fan_speed_percent(adev, &speed);
-	if (err)
-		return err;
+	if (adev->powerplay.pp_funcs->get_fan_speed_percent) {
+		err = amdgpu_dpm_get_fan_speed_percent(adev, &speed);
+		if (err)
+			return err;
+	}
 
 	speed = (speed * 255) / 100;
 
@@ -949,11 +910,13 @@ static ssize_t amdgpu_hwmon_get_fan1_input(struct device *dev,
 {
 	struct amdgpu_device *adev = dev_get_drvdata(dev);
 	int err;
-	u32 speed;
+	u32 speed = 0;
 
-	err = amdgpu_dpm_get_fan_speed_rpm(adev, &speed);
-	if (err)
-		return err;
+	if (adev->powerplay.pp_funcs->get_fan_speed_rpm) {
+		err = amdgpu_dpm_get_fan_speed_rpm(adev, &speed);
+		if (err)
+			return err;
+	}
 
 	return sprintf(buf, "%i\n", speed);
 }
@@ -1008,21 +971,21 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj,
 		return 0;
 
 	/* mask fan attributes if we have no bindings for this asic to expose */
-	if ((!adev->pm.funcs->get_fan_speed_percent &&
+	if ((!adev->powerplay.pp_funcs->get_fan_speed_percent &&
 	     attr == &sensor_dev_attr_pwm1.dev_attr.attr) || /* can't query fan */
-	    (!adev->pm.funcs->get_fan_control_mode &&
+	    (!adev->powerplay.pp_funcs->get_fan_control_mode &&
 	     attr == &sensor_dev_attr_pwm1_enable.dev_attr.attr)) /* can't query state */
 		effective_mode &= ~S_IRUGO;
 
-	if ((!adev->pm.funcs->set_fan_speed_percent &&
+	if ((!adev->powerplay.pp_funcs->set_fan_speed_percent &&
 	     attr == &sensor_dev_attr_pwm1.dev_attr.attr) || /* can't manage fan */
-	    (!adev->pm.funcs->set_fan_control_mode &&
+	    (!adev->powerplay.pp_funcs->set_fan_control_mode &&
 	     attr == &sensor_dev_attr_pwm1_enable.dev_attr.attr)) /* can't manage state */
 		effective_mode &= ~S_IWUSR;
 
 	/* hide max/min values if we can't both query and manage the fan */
-	if ((!adev->pm.funcs->set_fan_speed_percent &&
-	     !adev->pm.funcs->get_fan_speed_percent) &&
+	if ((!adev->powerplay.pp_funcs->set_fan_speed_percent &&
+	     !adev->powerplay.pp_funcs->get_fan_speed_percent) &&
 	    (attr == &sensor_dev_attr_pwm1_max.dev_attr.attr ||
 	     attr == &sensor_dev_attr_pwm1_min.dev_attr.attr))
 		return 0;
@@ -1055,7 +1018,7 @@ void amdgpu_dpm_thermal_work_handler(struct work_struct *work)
 	if (!adev->pm.dpm_enabled)
 		return;
 
-	if (adev->pm.funcs->get_temperature) {
+	if (adev->powerplay.pp_funcs->get_temperature) {
 		int temp = amdgpu_dpm_get_temperature(adev);
 
 		if (temp < adev->pm.dpm.thermal.min_temp)
@@ -1087,7 +1050,7 @@ static struct amdgpu_ps *amdgpu_dpm_pick_power_state(struct amdgpu_device *adev,
 		true : false;
 
 	/* check if the vblank period is too short to adjust the mclk */
-	if (single_display && adev->pm.funcs->vblank_too_short) {
+	if (single_display && adev->powerplay.pp_funcs->vblank_too_short) {
 		if (amdgpu_dpm_vblank_too_short(adev))
 			single_display = false;
 	}
@@ -1216,7 +1179,7 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev)
 	struct amdgpu_ps *ps;
 	enum amd_pm_state_type dpm_state;
 	int ret;
-	bool equal;
+	bool equal = false;
 
 	/* if dpm init failed */
 	if (!adev->pm.dpm_enabled)
@@ -1236,7 +1199,7 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev)
 	else
 		return;
 
-	if (amdgpu_dpm == 1) {
+	if (amdgpu_dpm == 1 && adev->powerplay.pp_funcs->print_power_state) {
 		printk("switching from power state:\n");
 		amdgpu_dpm_print_power_state(adev, adev->pm.dpm.current_ps);
 		printk("switching to power state:\n");
@@ -1245,15 +1208,17 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev)
 
 	/* update whether vce is active */
 	ps->vce_active = adev->pm.dpm.vce_active;
-
-	amdgpu_dpm_display_configuration_changed(adev);
+	if (adev->powerplay.pp_funcs->display_configuration_changed)
+		amdgpu_dpm_display_configuration_changed(adev);
 
 	ret = amdgpu_dpm_pre_set_power_state(adev);
 	if (ret)
 		return;
 
-	if ((0 != amgdpu_dpm_check_state_equal(adev, adev->pm.dpm.current_ps, adev->pm.dpm.requested_ps, &equal)))
-		equal = false;
+	if (adev->powerplay.pp_funcs->check_state_equal) {
+		if (0 != amdgpu_dpm_check_state_equal(adev, adev->pm.dpm.current_ps, adev->pm.dpm.requested_ps, &equal))
+			equal = false;
+	}
 
 	if (equal)
 		return;
@@ -1264,7 +1229,7 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev)
 	adev->pm.dpm.current_active_crtcs = adev->pm.dpm.new_active_crtcs;
 	adev->pm.dpm.current_active_crtc_count = adev->pm.dpm.new_active_crtc_count;
 
-	if (adev->pm.funcs->force_performance_level) {
+	if (adev->powerplay.pp_funcs->force_performance_level) {
 		if (adev->pm.dpm.thermal_active) {
 			enum amd_dpm_forced_level level = adev->pm.dpm.forced_level;
 			/* force low perf level for thermal */
@@ -1280,7 +1245,7 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev)
 
 void amdgpu_dpm_enable_uvd(struct amdgpu_device *adev, bool enable)
 {
-	if (adev->pp_enabled || adev->pm.funcs->powergate_uvd) {
+	if (adev->powerplay.pp_funcs->powergate_uvd) {
 		/* enable/disable UVD */
 		mutex_lock(&adev->pm.mutex);
 		amdgpu_dpm_powergate_uvd(adev, !enable);
@@ -1302,7 +1267,7 @@ void amdgpu_dpm_enable_uvd(struct amdgpu_device *adev, bool enable)
 
 void amdgpu_dpm_enable_vce(struct amdgpu_device *adev, bool enable)
 {
-	if (adev->pp_enabled || adev->pm.funcs->powergate_vce) {
+	if (adev->powerplay.pp_funcs->powergate_vce) {
 		/* enable/disable VCE */
 		mutex_lock(&adev->pm.mutex);
 		amdgpu_dpm_powergate_vce(adev, !enable);
@@ -1337,8 +1302,7 @@ void amdgpu_pm_print_power_states(struct amdgpu_device *adev)
 {
 	int i;
 
-	if (adev->pp_enabled)
-		/* TO DO */
+	if (adev->powerplay.pp_funcs->print_power_state == NULL)
 		return;
 
 	for (i = 0; i < adev->pm.dpm.num_ps; i++)
@@ -1353,10 +1317,8 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev)
 	if (adev->pm.sysfs_initialized)
 		return 0;
 
-	if (!adev->pp_enabled) {
-		if (adev->pm.funcs->get_temperature == NULL)
-			return 0;
-	}
+	if (adev->powerplay.pp_funcs->get_temperature == NULL)
+		return 0;
 
 	adev->pm.int_hwmon_dev = hwmon_device_register_with_groups(adev->dev,
 								   DRIVER_NAME, adev,
@@ -1496,7 +1458,7 @@ void amdgpu_pm_compute_clocks(struct amdgpu_device *adev)
 	}
 
 	if (adev->pp_enabled) {
-		amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_DISPLAY_CONFIG_CHANGE, NULL, NULL);
+		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_DISPLAY_CONFIG_CHANGE, NULL, NULL);
 	} else {
 		mutex_lock(&adev->pm.mutex);
 		adev->pm.dpm.new_active_crtcs = 0;
@@ -1634,8 +1596,8 @@ static int amdgpu_debugfs_pm_info(struct seq_file *m, void *data)
 		return amdgpu_debugfs_pm_info_pp(m, adev);
 	} else {
 		mutex_lock(&adev->pm.mutex);
-		if (adev->pm.funcs->debugfs_print_current_performance_level)
-			adev->pm.funcs->debugfs_print_current_performance_level(adev, m);
+		if (adev->powerplay.pp_funcs->debugfs_print_current_performance_level)
+			adev->powerplay.pp_funcs->debugfs_print_current_performance_level(adev, m);
 		else
 			seq_printf(m, "Debugfs support not implemented for this asic\n");
 		mutex_unlock(&adev->pm.mutex);

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

@@ -87,17 +87,28 @@ static int amdgpu_pp_early_init(void *handle)
 	case CHIP_OLAND:
 	case CHIP_HAINAN:
 		amd_pp->ip_funcs = &si_dpm_ip_funcs;
+		amd_pp->pp_funcs = &si_dpm_funcs;
 	break;
 #endif
 #ifdef CONFIG_DRM_AMDGPU_CIK
 	case CHIP_BONAIRE:
 	case CHIP_HAWAII:
-		amd_pp->ip_funcs = &ci_dpm_ip_funcs;
+		if (amdgpu_dpm == -1) {
+			amd_pp->ip_funcs = &ci_dpm_ip_funcs;
+			amd_pp->pp_funcs = &ci_dpm_funcs;
+		} else {
+			adev->pp_enabled = true;
+			if (amdgpu_create_pp_handle(adev))
+				return -EINVAL;
+			amd_pp->ip_funcs = &pp_ip_funcs;
+			amd_pp->pp_funcs = &pp_dpm_funcs;
+		}
 		break;
 	case CHIP_KABINI:
 	case CHIP_MULLINS:
 	case CHIP_KAVERI:
 		amd_pp->ip_funcs = &kv_dpm_ip_funcs;
+		amd_pp->pp_funcs = &kv_dpm_funcs;
 		break;
 #endif
 	default:
@@ -128,7 +139,7 @@ static int amdgpu_pp_late_init(void *handle)
 
 	if (adev->pp_enabled && adev->pm.dpm_enabled) {
 		amdgpu_pm_sysfs_init(adev);
-		amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_COMPLETE_INIT, NULL, NULL);
+		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_COMPLETE_INIT, NULL, NULL);
 	}
 
 	return ret;

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

@@ -57,6 +57,40 @@ void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr)
 	ttm_bo_kunmap(&bo->dma_buf_vmap);
 }
 
+int amdgpu_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)
+{
+	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
+	unsigned asize = amdgpu_bo_size(bo);
+	int ret;
+
+	if (!vma->vm_file)
+		return -ENODEV;
+
+	if (adev == NULL)
+		return -ENODEV;
+
+	/* Check for valid size. */
+	if (asize < vma->vm_end - vma->vm_start)
+		return -EINVAL;
+
+	if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) ||
+	    (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)) {
+		return -EPERM;
+	}
+	vma->vm_pgoff += amdgpu_bo_mmap_offset(bo) >> PAGE_SHIFT;
+
+	/* prime mmap does not need to check access, so allow here */
+	ret = drm_vma_node_allow(&obj->vma_node, vma->vm_file->private_data);
+	if (ret)
+		return ret;
+
+	ret = ttm_bo_mmap(vma->vm_file, vma, &adev->mman.bdev);
+	drm_vma_node_revoke(&obj->vma_node, vma->vm_file->private_data);
+
+	return ret;
+}
+
 struct drm_gem_object *
 amdgpu_gem_prime_import_sg_table(struct drm_device *dev,
 				 struct dma_buf_attachment *attach,
@@ -136,7 +170,8 @@ struct dma_buf *amdgpu_gem_prime_export(struct drm_device *dev,
 {
 	struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj);
 
-	if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm))
+	if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) ||
+	    bo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID)
 		return ERR_PTR(-EPERM);
 
 	return drm_gem_prime_export(dev, gobj, flags);

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

@@ -57,21 +57,23 @@ static int psp_sw_init(void *handle)
 		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_stop = psp_v3_1_ring_stop;
 		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;
+		psp->mode1_reset = psp_v3_1_mode1_reset;
 		break;
 	case CHIP_RAVEN:
-#if 0
 		psp->init_microcode = psp_v10_0_init_microcode;
-#endif
 		psp->prep_cmd_buf = psp_v10_0_prep_cmd_buf;
 		psp->ring_init = psp_v10_0_ring_init;
 		psp->ring_create = psp_v10_0_ring_create;
+		psp->ring_stop = psp_v10_0_ring_stop;
 		psp->ring_destroy = psp_v10_0_ring_destroy;
 		psp->cmd_submit = psp_v10_0_cmd_submit;
 		psp->compare_sram_data = psp_v10_0_compare_sram_data;
+		psp->mode1_reset = psp_v10_0_mode1_reset;
 		break;
 	default:
 		return -EINVAL;
@@ -90,6 +92,12 @@ static int psp_sw_init(void *handle)
 
 static int psp_sw_fini(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+	release_firmware(adev->psp.sos_fw);
+	adev->psp.sos_fw = NULL;
+	release_firmware(adev->psp.asd_fw);
+	adev->psp.asd_fw = NULL;
 	return 0;
 }
 
@@ -253,15 +261,18 @@ static int psp_asd_load(struct psp_context *psp)
 
 static int psp_hw_start(struct psp_context *psp)
 {
+	struct amdgpu_device *adev = psp->adev;
 	int ret;
 
-	ret = psp_bootloader_load_sysdrv(psp);
-	if (ret)
-		return ret;
+	if (!amdgpu_sriov_vf(adev) || !adev->in_sriov_reset) {
+		ret = psp_bootloader_load_sysdrv(psp);
+		if (ret)
+			return ret;
 
-	ret = psp_bootloader_load_sos(psp);
-	if (ret)
-		return ret;
+		ret = psp_bootloader_load_sos(psp);
+		if (ret)
+			return ret;
+	}
 
 	ret = psp_ring_create(psp, PSP_RING_TYPE__KM);
 	if (ret)
@@ -453,6 +464,16 @@ static int psp_hw_fini(void *handle)
 
 static int psp_suspend(void *handle)
 {
+	int ret;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct psp_context *psp = &adev->psp;
+
+	ret = psp_ring_stop(psp, PSP_RING_TYPE__KM);
+	if (ret) {
+		DRM_ERROR("PSP ring stop failed\n");
+		return ret;
+	}
+
 	return 0;
 }
 
@@ -487,6 +508,22 @@ failed:
 	return ret;
 }
 
+static bool psp_check_reset(void* handle)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+	if (adev->flags & AMD_IS_APU)
+		return true;
+
+	return false;
+}
+
+static int psp_reset(void* handle)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	return psp_mode1_reset(&adev->psp);
+}
+
 static bool psp_check_fw_loading_status(struct amdgpu_device *adev,
 					enum AMDGPU_UCODE_ID ucode_type)
 {
@@ -530,8 +567,9 @@ const struct amd_ip_funcs psp_ip_funcs = {
 	.suspend = psp_suspend,
 	.resume = psp_resume,
 	.is_idle = NULL,
+	.check_soft_reset = psp_check_reset,
 	.wait_for_idle = NULL,
-	.soft_reset = NULL,
+	.soft_reset = psp_reset,
 	.set_clockgating_state = psp_set_clockgating_state,
 	.set_powergating_state = psp_set_powergating_state,
 };

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

@@ -66,6 +66,8 @@ struct psp_context
 			    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_stop)(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,
@@ -74,6 +76,7 @@ struct psp_context
 				  struct amdgpu_firmware_info *ucode,
 				  enum AMDGPU_UCODE_ID ucode_type);
 	bool (*smu_reload_quirk)(struct psp_context *psp);
+	int (*mode1_reset)(struct psp_context *psp);
 
 	/* fence buffer */
 	struct amdgpu_bo 		*fw_pri_bo;
@@ -123,6 +126,7 @@ struct amdgpu_psp_funcs {
 #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_stop(psp, type) (psp)->ring_stop((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))
@@ -136,6 +140,8 @@ struct amdgpu_psp_funcs {
 		((psp)->bootloader_load_sos ? (psp)->bootloader_load_sos((psp)) : 0)
 #define psp_smu_reload_quirk(psp) \
 		((psp)->smu_reload_quirk ? (psp)->smu_reload_quirk((psp)) : false)
+#define psp_mode1_reset(psp) \
+		((psp)->mode1_reset ? (psp)->mode1_reset((psp)) : false)
 
 extern const struct amd_ip_funcs psp_ip_funcs;
 

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

@@ -14,62 +14,6 @@
 #define AMDGPU_JOB_GET_TIMELINE_NAME(job) \
 	 job->base.s_fence->finished.ops->get_timeline_name(&job->base.s_fence->finished)
 
-TRACE_EVENT(amdgpu_ttm_tt_populate,
-	    TP_PROTO(struct amdgpu_device *adev, uint64_t dma_address, uint64_t phys_address),
-	    TP_ARGS(adev, dma_address, phys_address),
-	    TP_STRUCT__entry(
-				__field(uint16_t, domain)
-				__field(uint8_t, bus)
-				__field(uint8_t, slot)
-				__field(uint8_t, func)
-				__field(uint64_t, dma)
-				__field(uint64_t, phys)
-			    ),
-	    TP_fast_assign(
-			   __entry->domain = pci_domain_nr(adev->pdev->bus);
-			   __entry->bus = adev->pdev->bus->number;
-			   __entry->slot = PCI_SLOT(adev->pdev->devfn);
-			   __entry->func = PCI_FUNC(adev->pdev->devfn);
-			   __entry->dma = dma_address;
-			   __entry->phys = phys_address;
-			   ),
-	    TP_printk("%04x:%02x:%02x.%x: 0x%llx => 0x%llx",
-		      (unsigned)__entry->domain,
-		      (unsigned)__entry->bus,
-		      (unsigned)__entry->slot,
-		      (unsigned)__entry->func,
-		      (unsigned long long)__entry->dma,
-		      (unsigned long long)__entry->phys)
-);
-
-TRACE_EVENT(amdgpu_ttm_tt_unpopulate,
-	    TP_PROTO(struct amdgpu_device *adev, uint64_t dma_address, uint64_t phys_address),
-	    TP_ARGS(adev, dma_address, phys_address),
-	    TP_STRUCT__entry(
-				__field(uint16_t, domain)
-				__field(uint8_t, bus)
-				__field(uint8_t, slot)
-				__field(uint8_t, func)
-				__field(uint64_t, dma)
-				__field(uint64_t, phys)
-			    ),
-	    TP_fast_assign(
-			   __entry->domain = pci_domain_nr(adev->pdev->bus);
-			   __entry->bus = adev->pdev->bus->number;
-			   __entry->slot = PCI_SLOT(adev->pdev->devfn);
-			   __entry->func = PCI_FUNC(adev->pdev->devfn);
-			   __entry->dma = dma_address;
-			   __entry->phys = phys_address;
-			   ),
-	    TP_printk("%04x:%02x:%02x.%x: 0x%llx => 0x%llx",
-		      (unsigned)__entry->domain,
-		      (unsigned)__entry->bus,
-		      (unsigned)__entry->slot,
-		      (unsigned)__entry->func,
-		      (unsigned long long)__entry->dma,
-		      (unsigned long long)__entry->phys)
-);
-
 TRACE_EVENT(amdgpu_mm_rreg,
 	    TP_PROTO(unsigned did, uint32_t reg, uint32_t value),
 	    TP_ARGS(did, reg, value),
@@ -473,5 +417,5 @@ TRACE_EVENT(amdgpu_ttm_bo_move,
 
 /* This part must be outside protection */
 #undef TRACE_INCLUDE_PATH
-#define TRACE_INCLUDE_PATH .
+#define TRACE_INCLUDE_PATH ../../drivers/gpu/drm/amd/amdgpu
 #include <trace/define_trace.h>

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

@@ -1,4 +1,23 @@
 /* Copyright Red Hat Inc 2010.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
  * Author : Dave Airlie <airlied@redhat.com>
  */
 #include <drm/drmP.h>

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

@@ -42,6 +42,7 @@
 #include <linux/swap.h>
 #include <linux/pagemap.h>
 #include <linux/debugfs.h>
+#include <linux/iommu.h>
 #include "amdgpu.h"
 #include "amdgpu_trace.h"
 #include "bif/bif_4_1_d.h"
@@ -608,6 +609,7 @@ struct amdgpu_ttm_tt {
 	spinlock_t              guptasklock;
 	struct list_head        guptasks;
 	atomic_t		mmu_invalidations;
+	uint32_t		last_set_pages;
 	struct list_head        list;
 };
 
@@ -621,6 +623,8 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages)
 	if (!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY))
 		flags |= FOLL_WRITE;
 
+	down_read(&current->mm->mmap_sem);
+
 	if (gtt->userflags & AMDGPU_GEM_USERPTR_ANONONLY) {
 		/* check that we only use anonymous memory
 		   to prevent problems with writeback */
@@ -628,8 +632,10 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages)
 		struct vm_area_struct *vma;
 
 		vma = find_vma(gtt->usermm, gtt->userptr);
-		if (!vma || vma->vm_file || vma->vm_end < end)
+		if (!vma || vma->vm_file || vma->vm_end < end) {
+			up_read(&current->mm->mmap_sem);
 			return -EPERM;
+		}
 	}
 
 	do {
@@ -656,42 +662,44 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages)
 
 	} while (pinned < ttm->num_pages);
 
+	up_read(&current->mm->mmap_sem);
 	return 0;
 
 release_pages:
 	release_pages(pages, pinned, 0);
+	up_read(&current->mm->mmap_sem);
 	return r;
 }
 
-static void amdgpu_trace_dma_map(struct ttm_tt *ttm)
+void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages)
 {
-	struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev);
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
 	unsigned i;
 
-	if (unlikely(trace_amdgpu_ttm_tt_populate_enabled())) {
-		for (i = 0; i < ttm->num_pages; i++) {
-			trace_amdgpu_ttm_tt_populate(
-				adev,
-				gtt->ttm.dma_address[i],
-				page_to_phys(ttm->pages[i]));
-		}
+	gtt->last_set_pages = atomic_read(&gtt->mmu_invalidations);
+	for (i = 0; i < ttm->num_pages; ++i) {
+		if (ttm->pages[i])
+			put_page(ttm->pages[i]);
+
+		ttm->pages[i] = pages ? pages[i] : NULL;
 	}
 }
 
-static void amdgpu_trace_dma_unmap(struct ttm_tt *ttm)
+void amdgpu_ttm_tt_mark_user_pages(struct ttm_tt *ttm)
 {
-	struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev);
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
 	unsigned i;
 
-	if (unlikely(trace_amdgpu_ttm_tt_unpopulate_enabled())) {
-		for (i = 0; i < ttm->num_pages; i++) {
-			trace_amdgpu_ttm_tt_unpopulate(
-				adev,
-				gtt->ttm.dma_address[i],
-				page_to_phys(ttm->pages[i]));
-		}
+	for (i = 0; i < ttm->num_pages; ++i) {
+		struct page *page = ttm->pages[i];
+
+		if (!page)
+			continue;
+
+		if (!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY))
+			set_page_dirty(page);
+
+		mark_page_accessed(page);
 	}
 }
 
@@ -721,8 +729,6 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm)
 	drm_prime_sg_to_page_addr_arrays(ttm->sg, ttm->pages,
 					 gtt->ttm.dma_address, ttm->num_pages);
 
-	amdgpu_trace_dma_map(ttm);
-
 	return 0;
 
 release_sg:
@@ -734,7 +740,6 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_tt *ttm)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev);
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
-	struct sg_page_iter sg_iter;
 
 	int write = !(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY);
 	enum dma_data_direction direction = write ?
@@ -747,16 +752,7 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_tt *ttm)
 	/* free the sg table and pages again */
 	dma_unmap_sg(adev->dev, ttm->sg->sgl, ttm->sg->nents, direction);
 
-	for_each_sg_page(ttm->sg->sgl, &sg_iter, ttm->sg->nents, 0) {
-		struct page *page = sg_page_iter_page(&sg_iter);
-		if (!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY))
-			set_page_dirty(page);
-
-		mark_page_accessed(page);
-		put_page(page);
-	}
-
-	amdgpu_trace_dma_unmap(ttm);
+	amdgpu_ttm_tt_mark_user_pages(ttm);
 
 	sg_free_table(ttm->sg);
 }
@@ -818,7 +814,6 @@ int amdgpu_ttm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *bo_mem)
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);
 	struct ttm_tt *ttm = bo->ttm;
 	struct ttm_mem_reg tmp;
-
 	struct ttm_placement placement;
 	struct ttm_place placements;
 	int r;
@@ -834,7 +829,7 @@ int amdgpu_ttm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *bo_mem)
 	placement.busy_placement = &placements;
 	placements.fpfn = 0;
 	placements.lpfn = adev->mc.gart_size >> PAGE_SHIFT;
-	placements.flags = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT;
+	placements.flags = bo->mem.placement | TTM_PL_FLAG_TT;
 
 	r = ttm_bo_mem_space(bo, &placement, &tmp, true, false);
 	if (unlikely(r))
@@ -941,8 +936,6 @@ static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev);
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
-	unsigned i;
-	int r;
 	bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG);
 
 	if (ttm->state != tt_unpopulated)
@@ -962,52 +955,26 @@ static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm)
 		drm_prime_sg_to_page_addr_arrays(ttm->sg, ttm->pages,
 						 gtt->ttm.dma_address, ttm->num_pages);
 		ttm->state = tt_unbound;
-		r = 0;
-		goto trace_mappings;
+		return 0;
 	}
 
 #ifdef CONFIG_SWIOTLB
 	if (swiotlb_nr_tbl()) {
-		r = ttm_dma_populate(&gtt->ttm, adev->dev);
-		goto trace_mappings;
+		return ttm_dma_populate(&gtt->ttm, adev->dev);
 	}
 #endif
 
-	r = ttm_pool_populate(ttm);
-	if (r) {
-		return r;
-	}
-
-	for (i = 0; i < ttm->num_pages; i++) {
-		gtt->ttm.dma_address[i] = pci_map_page(adev->pdev, ttm->pages[i],
-						       0, PAGE_SIZE,
-						       PCI_DMA_BIDIRECTIONAL);
-		if (pci_dma_mapping_error(adev->pdev, gtt->ttm.dma_address[i])) {
-			while (i--) {
-				pci_unmap_page(adev->pdev, gtt->ttm.dma_address[i],
-					       PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
-				gtt->ttm.dma_address[i] = 0;
-			}
-			ttm_pool_unpopulate(ttm);
-			return -EFAULT;
-		}
-	}
-
-	r = 0;
-trace_mappings:
-	if (likely(!r))
-		amdgpu_trace_dma_map(ttm);
-	return r;
+	return ttm_populate_and_map_pages(adev->dev, &gtt->ttm);
 }
 
 static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm)
 {
 	struct amdgpu_device *adev;
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
-	unsigned i;
 	bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG);
 
 	if (gtt && gtt->userptr) {
+		amdgpu_ttm_tt_set_user_pages(ttm, NULL);
 		kfree(ttm->sg);
 		ttm->page_flags &= ~TTM_PAGE_FLAG_SG;
 		return;
@@ -1018,8 +985,6 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm)
 
 	adev = amdgpu_ttm_adev(ttm->bdev);
 
-	amdgpu_trace_dma_unmap(ttm);
-
 #ifdef CONFIG_SWIOTLB
 	if (swiotlb_nr_tbl()) {
 		ttm_dma_unpopulate(&gtt->ttm, adev->dev);
@@ -1027,14 +992,7 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm)
 	}
 #endif
 
-	for (i = 0; i < ttm->num_pages; i++) {
-		if (gtt->ttm.dma_address[i]) {
-			pci_unmap_page(adev->pdev, gtt->ttm.dma_address[i],
-				       PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
-		}
-	}
-
-	ttm_pool_unpopulate(ttm);
+	ttm_unmap_and_unpopulate_pages(adev->dev, &gtt->ttm);
 }
 
 int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr,
@@ -1051,6 +1009,7 @@ int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr,
 	spin_lock_init(&gtt->guptasklock);
 	INIT_LIST_HEAD(&gtt->guptasks);
 	atomic_set(&gtt->mmu_invalidations, 0);
+	gtt->last_set_pages = 0;
 
 	return 0;
 }
@@ -1103,6 +1062,16 @@ bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm,
 	return prev_invalidated != *last_invalidated;
 }
 
+bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm)
+{
+	struct amdgpu_ttm_tt *gtt = (void *)ttm;
+
+	if (gtt == NULL || !gtt->userptr)
+		return false;
+
+	return atomic_read(&gtt->mmu_invalidations) != gtt->last_set_pages;
+}
+
 bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm)
 {
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
@@ -1202,14 +1171,14 @@ static int amdgpu_ttm_access_memory(struct ttm_buffer_object *bo,
 		}
 
 		spin_lock_irqsave(&adev->mmio_idx_lock, flags);
-		WREG32(mmMM_INDEX, ((uint32_t)aligned_pos) | 0x80000000);
-		WREG32(mmMM_INDEX_HI, aligned_pos >> 31);
+		WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t)aligned_pos) | 0x80000000);
+		WREG32_NO_KIQ(mmMM_INDEX_HI, aligned_pos >> 31);
 		if (!write || mask != 0xffffffff)
-			value = RREG32(mmMM_DATA);
+			value = RREG32_NO_KIQ(mmMM_DATA);
 		if (write) {
 			value &= ~mask;
 			value |= (*(uint32_t *)buf << shift) & mask;
-			WREG32(mmMM_DATA, value);
+			WREG32_NO_KIQ(mmMM_DATA, value);
 		}
 		spin_unlock_irqrestore(&adev->mmio_idx_lock, flags);
 		if (!write) {
@@ -1557,8 +1526,8 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
 		       struct dma_fence **fence)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
-	/* max_bytes applies to SDMA_OP_PTEPDE as well as SDMA_OP_CONST_FILL*/
-	uint32_t max_bytes = adev->mman.buffer_funcs->fill_max_bytes;
+	uint32_t max_bytes = 8 *
+			adev->vm_manager.vm_pte_funcs->set_max_nums_pte_pde;
 	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
 
 	struct drm_mm_node *mm_node;
@@ -1590,8 +1559,8 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
 		++mm_node;
 	}
 
-	/* 10 double words for each SDMA_OP_PTEPDE cmd */
-	num_dw = num_loops * 10;
+	/* num of dwords for each SDMA_OP_PTEPDE cmd */
+	num_dw = num_loops * adev->vm_manager.vm_pte_funcs->set_pte_pde_num_dw;
 
 	/* for IB padding */
 	num_dw += 64;
@@ -1697,9 +1666,9 @@ static ssize_t amdgpu_ttm_vram_read(struct file *f, char __user *buf,
 			return result;
 
 		spin_lock_irqsave(&adev->mmio_idx_lock, flags);
-		WREG32(mmMM_INDEX, ((uint32_t)*pos) | 0x80000000);
-		WREG32(mmMM_INDEX_HI, *pos >> 31);
-		value = RREG32(mmMM_DATA);
+		WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t)*pos) | 0x80000000);
+		WREG32_NO_KIQ(mmMM_INDEX_HI, *pos >> 31);
+		value = RREG32_NO_KIQ(mmMM_DATA);
 		spin_unlock_irqrestore(&adev->mmio_idx_lock, flags);
 
 		r = put_user(value, (uint32_t *)buf);
@@ -1715,10 +1684,50 @@ static ssize_t amdgpu_ttm_vram_read(struct file *f, char __user *buf,
 	return result;
 }
 
+static ssize_t amdgpu_ttm_vram_write(struct file *f, const char __user *buf,
+				    size_t size, loff_t *pos)
+{
+	struct amdgpu_device *adev = file_inode(f)->i_private;
+	ssize_t result = 0;
+	int r;
+
+	if (size & 0x3 || *pos & 0x3)
+		return -EINVAL;
+
+	if (*pos >= adev->mc.mc_vram_size)
+		return -ENXIO;
+
+	while (size) {
+		unsigned long flags;
+		uint32_t value;
+
+		if (*pos >= adev->mc.mc_vram_size)
+			return result;
+
+		r = get_user(value, (uint32_t *)buf);
+		if (r)
+			return r;
+
+		spin_lock_irqsave(&adev->mmio_idx_lock, flags);
+		WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t)*pos) | 0x80000000);
+		WREG32_NO_KIQ(mmMM_INDEX_HI, *pos >> 31);
+		WREG32_NO_KIQ(mmMM_DATA, value);
+		spin_unlock_irqrestore(&adev->mmio_idx_lock, flags);
+
+		result += 4;
+		buf += 4;
+		*pos += 4;
+		size -= 4;
+	}
+
+	return result;
+}
+
 static const struct file_operations amdgpu_ttm_vram_fops = {
 	.owner = THIS_MODULE,
 	.read = amdgpu_ttm_vram_read,
-	.llseek = default_llseek
+	.write = amdgpu_ttm_vram_write,
+	.llseek = default_llseek,
 };
 
 #ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS
@@ -1770,6 +1779,53 @@ static const struct file_operations amdgpu_ttm_gtt_fops = {
 
 #endif
 
+static ssize_t amdgpu_iova_to_phys_read(struct file *f, char __user *buf,
+				   size_t size, loff_t *pos)
+{
+	struct amdgpu_device *adev = file_inode(f)->i_private;
+	int r;
+	uint64_t phys;
+	struct iommu_domain *dom;
+
+	// always return 8 bytes
+	if (size != 8)
+		return -EINVAL;
+
+	// only accept page addresses
+	if (*pos & 0xFFF)
+		return -EINVAL;
+
+	dom = iommu_get_domain_for_dev(adev->dev);
+	if (dom)
+		phys = iommu_iova_to_phys(dom, *pos);
+	else
+		phys = *pos;
+
+	r = copy_to_user(buf, &phys, 8);
+	if (r)
+		return -EFAULT;
+
+	return 8;
+}
+
+static const struct file_operations amdgpu_ttm_iova_fops = {
+	.owner = THIS_MODULE,
+	.read = amdgpu_iova_to_phys_read,
+	.llseek = default_llseek
+};
+
+static const struct {
+	char *name;
+	const struct file_operations *fops;
+	int domain;
+} ttm_debugfs_entries[] = {
+	{ "amdgpu_vram", &amdgpu_ttm_vram_fops, TTM_PL_VRAM },
+#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS
+	{ "amdgpu_gtt", &amdgpu_ttm_gtt_fops, TTM_PL_TT },
+#endif
+	{ "amdgpu_iova", &amdgpu_ttm_iova_fops, TTM_PL_SYSTEM },
+};
+
 #endif
 
 static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev)
@@ -1780,22 +1836,21 @@ static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev)
 	struct drm_minor *minor = adev->ddev->primary;
 	struct dentry *ent, *root = minor->debugfs_root;
 
-	ent = debugfs_create_file("amdgpu_vram", S_IFREG | S_IRUGO, root,
-				  adev, &amdgpu_ttm_vram_fops);
-	if (IS_ERR(ent))
-		return PTR_ERR(ent);
-	i_size_write(ent->d_inode, adev->mc.mc_vram_size);
-	adev->mman.vram = ent;
-
-#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS
-	ent = debugfs_create_file("amdgpu_gtt", S_IFREG | S_IRUGO, root,
-				  adev, &amdgpu_ttm_gtt_fops);
-	if (IS_ERR(ent))
-		return PTR_ERR(ent);
-	i_size_write(ent->d_inode, adev->mc.gart_size);
-	adev->mman.gtt = ent;
+	for (count = 0; count < ARRAY_SIZE(ttm_debugfs_entries); count++) {
+		ent = debugfs_create_file(
+				ttm_debugfs_entries[count].name,
+				S_IFREG | S_IRUGO, root,
+				adev,
+				ttm_debugfs_entries[count].fops);
+		if (IS_ERR(ent))
+			return PTR_ERR(ent);
+		if (ttm_debugfs_entries[count].domain == TTM_PL_VRAM)
+			i_size_write(ent->d_inode, adev->mc.mc_vram_size);
+		else if (ttm_debugfs_entries[count].domain == TTM_PL_TT)
+			i_size_write(ent->d_inode, adev->mc.gart_size);
+		adev->mman.debugfs_entries[count] = ent;
+	}
 
-#endif
 	count = ARRAY_SIZE(amdgpu_ttm_debugfs_list);
 
 #ifdef CONFIG_SWIOTLB
@@ -1805,7 +1860,6 @@ static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev)
 
 	return amdgpu_debugfs_add_files(adev, amdgpu_ttm_debugfs_list, count);
 #else
-
 	return 0;
 #endif
 }
@@ -1813,14 +1867,9 @@ static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev)
 static void amdgpu_ttm_debugfs_fini(struct amdgpu_device *adev)
 {
 #if defined(CONFIG_DEBUG_FS)
+	unsigned i;
 
-	debugfs_remove(adev->mman.vram);
-	adev->mman.vram = NULL;
-
-#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS
-	debugfs_remove(adev->mman.gtt);
-	adev->mman.gtt = NULL;
-#endif
-
+	for (i = 0; i < ARRAY_SIZE(ttm_debugfs_entries); i++)
+		debugfs_remove(adev->mman.debugfs_entries[i]);
 #endif
 }

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

@@ -24,6 +24,7 @@
 #ifndef __AMDGPU_TTM_H__
 #define __AMDGPU_TTM_H__
 
+#include "amdgpu.h"
 #include "gpu_scheduler.h"
 
 #define AMDGPU_PL_GDS		(TTM_PL_PRIV + 0)
@@ -45,8 +46,7 @@ struct amdgpu_mman {
 	bool				initialized;
 
 #if defined(CONFIG_DEBUG_FS)
-	struct dentry			*vram;
-	struct dentry			*gtt;
+	struct dentry			*debugfs_entries[8];
 #endif
 
 	/* buffer handling */
@@ -82,4 +82,20 @@ bool amdgpu_ttm_is_bound(struct ttm_tt *ttm);
 int amdgpu_ttm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *bo_mem);
 int amdgpu_ttm_recover_gart(struct amdgpu_device *adev);
 
+int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages);
+void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages);
+void amdgpu_ttm_tt_mark_user_pages(struct ttm_tt *ttm);
+int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr,
+				     uint32_t flags);
+bool amdgpu_ttm_tt_has_userptr(struct ttm_tt *ttm);
+struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm);
+bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start,
+				  unsigned long end);
+bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm,
+				       int *last_invalidated);
+bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm);
+bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm);
+uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm,
+				 struct ttm_mem_reg *mem);
+
 #endif

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

@@ -270,12 +270,8 @@ amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type)
 		else
 			return AMDGPU_FW_LOAD_SMU;
 	case CHIP_VEGA10:
-		if (!load_type)
-			return AMDGPU_FW_LOAD_DIRECT;
-		else
-			return AMDGPU_FW_LOAD_PSP;
 	case CHIP_RAVEN:
-		if (load_type != 2)
+		if (!load_type)
 			return AMDGPU_FW_LOAD_DIRECT;
 		else
 			return AMDGPU_FW_LOAD_PSP;
@@ -364,8 +360,6 @@ static int amdgpu_ucode_patch_jt(struct amdgpu_firmware_info *ucode,
 int amdgpu_ucode_init_bo(struct amdgpu_device *adev)
 {
 	struct amdgpu_bo **bo = &adev->firmware.fw_buf;
-	uint64_t fw_mc_addr;
-	void *fw_buf_ptr = NULL;
 	uint64_t fw_offset = 0;
 	int i, err;
 	struct amdgpu_firmware_info *ucode = NULL;
@@ -376,37 +370,39 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev)
 		return 0;
 	}
 
-	err = amdgpu_bo_create(adev, adev->firmware.fw_size, PAGE_SIZE, true,
-				amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT,
-				AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS,
-				NULL, NULL, 0, bo);
-	if (err) {
-		dev_err(adev->dev, "(%d) Firmware buffer allocate failed\n", err);
-		goto failed;
-	}
+	if (!amdgpu_sriov_vf(adev) || !adev->in_sriov_reset) {
+		err = amdgpu_bo_create(adev, adev->firmware.fw_size, PAGE_SIZE, true,
+					amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT,
+					AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS,
+					NULL, NULL, 0, bo);
+		if (err) {
+			dev_err(adev->dev, "(%d) Firmware buffer allocate failed\n", err);
+			goto failed;
+		}
 
-	err = amdgpu_bo_reserve(*bo, false);
-	if (err) {
-		dev_err(adev->dev, "(%d) Firmware buffer reserve failed\n", err);
-		goto failed_reserve;
-	}
+		err = amdgpu_bo_reserve(*bo, false);
+		if (err) {
+			dev_err(adev->dev, "(%d) Firmware buffer reserve failed\n", err);
+			goto failed_reserve;
+		}
 
-	err = amdgpu_bo_pin(*bo, amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT,
-				&fw_mc_addr);
-	if (err) {
-		dev_err(adev->dev, "(%d) Firmware buffer pin failed\n", err);
-		goto failed_pin;
-	}
+		err = amdgpu_bo_pin(*bo, amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT,
+					&adev->firmware.fw_buf_mc);
+		if (err) {
+			dev_err(adev->dev, "(%d) Firmware buffer pin failed\n", err);
+			goto failed_pin;
+		}
 
-	err = amdgpu_bo_kmap(*bo, &fw_buf_ptr);
-	if (err) {
-		dev_err(adev->dev, "(%d) Firmware buffer kmap failed\n", err);
-		goto failed_kmap;
-	}
+		err = amdgpu_bo_kmap(*bo, &adev->firmware.fw_buf_ptr);
+		if (err) {
+			dev_err(adev->dev, "(%d) Firmware buffer kmap failed\n", err);
+			goto failed_kmap;
+		}
 
-	amdgpu_bo_unreserve(*bo);
+		amdgpu_bo_unreserve(*bo);
+	}
 
-	memset(fw_buf_ptr, 0, adev->firmware.fw_size);
+	memset(adev->firmware.fw_buf_ptr, 0, adev->firmware.fw_size);
 
 	/*
 	 * if SMU loaded firmware, it needn't add SMC, UVD, and VCE
@@ -425,14 +421,14 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev)
 		ucode = &adev->firmware.ucode[i];
 		if (ucode->fw) {
 			header = (const struct common_firmware_header *)ucode->fw->data;
-			amdgpu_ucode_init_single_fw(adev, ucode, fw_mc_addr + fw_offset,
-						    (void *)((uint8_t *)fw_buf_ptr + fw_offset));
+			amdgpu_ucode_init_single_fw(adev, ucode, adev->firmware.fw_buf_mc + fw_offset,
+						    adev->firmware.fw_buf_ptr + fw_offset);
 			if (i == AMDGPU_UCODE_ID_CP_MEC1 &&
 			    adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
 				const struct gfx_firmware_header_v1_0 *cp_hdr;
 				cp_hdr = (const struct gfx_firmware_header_v1_0 *)ucode->fw->data;
-				amdgpu_ucode_patch_jt(ucode, fw_mc_addr + fw_offset,
-						    fw_buf_ptr + fw_offset);
+				amdgpu_ucode_patch_jt(ucode,  adev->firmware.fw_buf_mc + fw_offset,
+						    adev->firmware.fw_buf_ptr + fw_offset);
 				fw_offset += ALIGN(le32_to_cpu(cp_hdr->jt_size) << 2, PAGE_SIZE);
 			}
 			fw_offset += ALIGN(ucode->ucode_size, PAGE_SIZE);

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

@@ -269,6 +269,7 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev)
 
 int amdgpu_uvd_sw_fini(struct amdgpu_device *adev)
 {
+	int i;
 	kfree(adev->uvd.saved_bo);
 
 	amd_sched_entity_fini(&adev->uvd.ring.sched, &adev->uvd.entity);
@@ -279,6 +280,9 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev)
 
 	amdgpu_ring_fini(&adev->uvd.ring);
 
+	for (i = 0; i < AMDGPU_MAX_UVD_ENC_RINGS; ++i)
+		amdgpu_ring_fini(&adev->uvd.ring_enc[i]);
+
 	release_firmware(adev->uvd.fw);
 
 	return 0;
@@ -410,10 +414,10 @@ static int amdgpu_uvd_cs_pass1(struct amdgpu_uvd_cs_ctx *ctx)
 	uint64_t addr = amdgpu_uvd_get_addr_from_ctx(ctx);
 	int r = 0;
 
-	mapping = amdgpu_cs_find_mapping(ctx->parser, addr, &bo);
-	if (mapping == NULL) {
+	r = amdgpu_cs_find_mapping(ctx->parser, addr, &bo, &mapping);
+	if (r) {
 		DRM_ERROR("Can't find BO for addr 0x%08Lx\n", addr);
-		return -EINVAL;
+		return r;
 	}
 
 	if (!ctx->parser->adev->uvd.address_64_bit) {
@@ -737,10 +741,10 @@ static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx)
 	uint64_t addr = amdgpu_uvd_get_addr_from_ctx(ctx);
 	int r;
 
-	mapping = amdgpu_cs_find_mapping(ctx->parser, addr, &bo);
-	if (mapping == NULL) {
+	r = amdgpu_cs_find_mapping(ctx->parser, addr, &bo, &mapping);
+	if (r) {
 		DRM_ERROR("Can't find BO for addr 0x%08Lx\n", addr);
-		return -EINVAL;
+		return r;
 	}
 
 	start = amdgpu_bo_gpu_offset(bo);
@@ -917,10 +921,6 @@ int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, uint32_t ib_idx)
 		return -EINVAL;
 	}
 
-	r = amdgpu_cs_sysvm_access_required(parser);
-	if (r)
-		return r;
-
 	ctx.parser = parser;
 	ctx.buf_sizes = buf_sizes;
 	ctx.ib_idx = ib_idx;

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

@@ -559,6 +559,7 @@ static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, uint32_t ib_idx,
 	struct amdgpu_bo_va_mapping *mapping;
 	struct amdgpu_bo *bo;
 	uint64_t addr;
+	int r;
 
 	if (index == 0xffffffff)
 		index = 0;
@@ -567,11 +568,11 @@ static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, uint32_t ib_idx,
 	       ((uint64_t)amdgpu_get_ib_value(p, ib_idx, hi)) << 32;
 	addr += ((uint64_t)size) * ((uint64_t)index);
 
-	mapping = amdgpu_cs_find_mapping(p, addr, &bo);
-	if (mapping == NULL) {
+	r = amdgpu_cs_find_mapping(p, addr, &bo, &mapping);
+	if (r) {
 		DRM_ERROR("Can't find BO for addr 0x%010Lx %d %d %d %d\n",
 			  addr, lo, hi, size, index);
-		return -EINVAL;
+		return r;
 	}
 
 	if ((addr + (uint64_t)size) >
@@ -652,10 +653,6 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx)
 	p->job->vm = NULL;
 	ib->gpu_addr = amdgpu_sa_bo_gpu_addr(ib->sa_bo);
 
-	r = amdgpu_cs_sysvm_access_required(p);
-	if (r)
-		return r;
-
 	while (idx < ib->length_dw) {
 		uint32_t len = amdgpu_get_ib_value(p, ib_idx, idx);
 		uint32_t cmd = amdgpu_get_ib_value(p, ib_idx, idx + 1);

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

@@ -27,11 +27,58 @@
  */
 #include <linux/dma-fence-array.h>
 #include <linux/interval_tree_generic.h>
+#include <linux/idr.h>
 #include <drm/drmP.h>
 #include <drm/amdgpu_drm.h>
 #include "amdgpu.h"
 #include "amdgpu_trace.h"
 
+/*
+ * PASID manager
+ *
+ * PASIDs are global address space identifiers that can be shared
+ * between the GPU, an IOMMU and the driver. VMs on different devices
+ * may use the same PASID if they share the same address
+ * space. Therefore PASIDs are allocated using a global IDA. VMs are
+ * looked up from the PASID per amdgpu_device.
+ */
+static DEFINE_IDA(amdgpu_vm_pasid_ida);
+
+/**
+ * amdgpu_vm_alloc_pasid - Allocate a PASID
+ * @bits: Maximum width of the PASID in bits, must be at least 1
+ *
+ * Allocates a PASID of the given width while keeping smaller PASIDs
+ * available if possible.
+ *
+ * Returns a positive integer on success. Returns %-EINVAL if bits==0.
+ * Returns %-ENOSPC if no PASID was available. Returns %-ENOMEM on
+ * memory allocation failure.
+ */
+int amdgpu_vm_alloc_pasid(unsigned int bits)
+{
+	int pasid = -EINVAL;
+
+	for (bits = min(bits, 31U); bits > 0; bits--) {
+		pasid = ida_simple_get(&amdgpu_vm_pasid_ida,
+				       1U << (bits - 1), 1U << bits,
+				       GFP_KERNEL);
+		if (pasid != -ENOSPC)
+			break;
+	}
+
+	return pasid;
+}
+
+/**
+ * amdgpu_vm_free_pasid - Free a PASID
+ * @pasid: PASID to free
+ */
+void amdgpu_vm_free_pasid(unsigned int pasid)
+{
+	ida_simple_remove(&amdgpu_vm_pasid_ida, pasid);
+}
+
 /*
  * GPUVM
  * GPUVM is similar to the legacy gart on older asics, however
@@ -140,7 +187,7 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
 			 struct list_head *validated,
 			 struct amdgpu_bo_list_entry *entry)
 {
-	entry->robj = vm->root.bo;
+	entry->robj = vm->root.base.bo;
 	entry->priority = 0;
 	entry->tv.bo = &entry->robj->tbo;
 	entry->tv.shared = true;
@@ -149,86 +196,80 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
 }
 
 /**
- * amdgpu_vm_validate_layer - validate a single page table level
+ * amdgpu_vm_validate_pt_bos - validate the page table BOs
  *
- * @parent: parent page table level
+ * @adev: amdgpu device pointer
+ * @vm: vm providing the BOs
  * @validate: callback to do the validation
  * @param: parameter for the validation callback
  *
  * Validate the page table BOs on command submission if neccessary.
  */
-static int amdgpu_vm_validate_level(struct amdgpu_vm_pt *parent,
-				    int (*validate)(void *, struct amdgpu_bo *),
-				    void *param, bool use_cpu_for_update,
-				    struct ttm_bo_global *glob)
+int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+			      int (*validate)(void *p, struct amdgpu_bo *bo),
+			      void *param)
 {
-	unsigned i;
+	struct ttm_bo_global *glob = adev->mman.bdev.glob;
 	int r;
 
-	if (use_cpu_for_update) {
-		r = amdgpu_bo_kmap(parent->bo, NULL);
-		if (r)
-			return r;
-	}
-
-	if (!parent->entries)
-		return 0;
+	spin_lock(&vm->status_lock);
+	while (!list_empty(&vm->evicted)) {
+		struct amdgpu_vm_bo_base *bo_base;
+		struct amdgpu_bo *bo;
 
-	for (i = 0; i <= parent->last_entry_used; ++i) {
-		struct amdgpu_vm_pt *entry = &parent->entries[i];
+		bo_base = list_first_entry(&vm->evicted,
+					   struct amdgpu_vm_bo_base,
+					   vm_status);
+		spin_unlock(&vm->status_lock);
 
-		if (!entry->bo)
-			continue;
+		bo = bo_base->bo;
+		BUG_ON(!bo);
+		if (bo->parent) {
+			r = validate(param, bo);
+			if (r)
+				return r;
 
-		r = validate(param, entry->bo);
-		if (r)
-			return r;
+			spin_lock(&glob->lru_lock);
+			ttm_bo_move_to_lru_tail(&bo->tbo);
+			if (bo->shadow)
+				ttm_bo_move_to_lru_tail(&bo->shadow->tbo);
+			spin_unlock(&glob->lru_lock);
+		}
 
-		spin_lock(&glob->lru_lock);
-		ttm_bo_move_to_lru_tail(&entry->bo->tbo);
-		if (entry->bo->shadow)
-			ttm_bo_move_to_lru_tail(&entry->bo->shadow->tbo);
-		spin_unlock(&glob->lru_lock);
+		if (bo->tbo.type == ttm_bo_type_kernel &&
+		    vm->use_cpu_for_update) {
+			r = amdgpu_bo_kmap(bo, NULL);
+			if (r)
+				return r;
+		}
 
-		/*
-		 * Recurse into the sub directory. This is harmless because we
-		 * have only a maximum of 5 layers.
-		 */
-		r = amdgpu_vm_validate_level(entry, validate, param,
-					     use_cpu_for_update, glob);
-		if (r)
-			return r;
+		spin_lock(&vm->status_lock);
+		if (bo->tbo.type != ttm_bo_type_kernel)
+			list_move(&bo_base->vm_status, &vm->moved);
+		else
+			list_move(&bo_base->vm_status, &vm->relocated);
 	}
+	spin_unlock(&vm->status_lock);
 
-	return r;
+	return 0;
 }
 
 /**
- * amdgpu_vm_validate_pt_bos - validate the page table BOs
+ * amdgpu_vm_ready - check VM is ready for updates
  *
- * @adev: amdgpu device pointer
- * @vm: vm providing the BOs
- * @validate: callback to do the validation
- * @param: parameter for the validation callback
+ * @vm: VM to check
  *
- * Validate the page table BOs on command submission if neccessary.
+ * Check if all VM PDs/PTs are ready for updates
  */
-int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			      int (*validate)(void *p, struct amdgpu_bo *bo),
-			      void *param)
+bool amdgpu_vm_ready(struct amdgpu_vm *vm)
 {
-	uint64_t num_evictions;
+	bool ready;
 
-	/* We only need to validate the page tables
-	 * if they aren't already valid.
-	 */
-	num_evictions = atomic64_read(&adev->num_evictions);
-	if (num_evictions == vm->last_eviction_counter)
-		return 0;
+	spin_lock(&vm->status_lock);
+	ready = list_empty(&vm->evicted);
+	spin_unlock(&vm->status_lock);
 
-	return amdgpu_vm_validate_level(&vm->root, validate, param,
-					vm->use_cpu_for_update,
-					adev->mman.bdev.glob);
+	return ready;
 }
 
 /**
@@ -294,11 +335,11 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev,
 
 	/* walk over the address space and allocate the page tables */
 	for (pt_idx = from; pt_idx <= to; ++pt_idx) {
-		struct reservation_object *resv = vm->root.bo->tbo.resv;
+		struct reservation_object *resv = vm->root.base.bo->tbo.resv;
 		struct amdgpu_vm_pt *entry = &parent->entries[pt_idx];
 		struct amdgpu_bo *pt;
 
-		if (!entry->bo) {
+		if (!entry->base.bo) {
 			r = amdgpu_bo_create(adev,
 					     amdgpu_vm_bo_size(adev, level),
 					     AMDGPU_GPU_PAGE_SIZE, true,
@@ -319,9 +360,14 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev,
 			/* Keep a reference to the root directory to avoid
 			* freeing them up in the wrong order.
 			*/
-			pt->parent = amdgpu_bo_ref(vm->root.bo);
-
-			entry->bo = pt;
+			pt->parent = amdgpu_bo_ref(parent->base.bo);
+
+			entry->base.vm = vm;
+			entry->base.bo = pt;
+			list_add_tail(&entry->base.bo_list, &pt->va);
+			spin_lock(&vm->status_lock);
+			list_add(&entry->base.vm_status, &vm->relocated);
+			spin_unlock(&vm->status_lock);
 			entry->addr = 0;
 		}
 
@@ -988,7 +1034,7 @@ static int amdgpu_vm_wait_pd(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	int r;
 
 	amdgpu_sync_create(&sync);
-	amdgpu_sync_resv(adev, &sync, vm->root.bo->tbo.resv, owner);
+	amdgpu_sync_resv(adev, &sync, vm->root.base.bo->tbo.resv, owner);
 	r = amdgpu_sync_wait(&sync, true);
 	amdgpu_sync_free(&sync);
 
@@ -1007,18 +1053,17 @@ static int amdgpu_vm_wait_pd(struct amdgpu_device *adev, struct amdgpu_vm *vm,
  */
 static int amdgpu_vm_update_level(struct amdgpu_device *adev,
 				  struct amdgpu_vm *vm,
-				  struct amdgpu_vm_pt *parent,
-				  unsigned level)
+				  struct amdgpu_vm_pt *parent)
 {
 	struct amdgpu_bo *shadow;
 	struct amdgpu_ring *ring = NULL;
 	uint64_t pd_addr, shadow_addr = 0;
-	uint32_t incr = amdgpu_vm_bo_size(adev, level + 1);
 	uint64_t last_pde = ~0, last_pt = ~0, last_shadow = ~0;
 	unsigned count = 0, pt_idx, ndw = 0;
 	struct amdgpu_job *job;
 	struct amdgpu_pte_update_params params;
 	struct dma_fence *fence = NULL;
+	uint32_t incr;
 
 	int r;
 
@@ -1027,10 +1072,10 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev,
 
 	memset(&params, 0, sizeof(params));
 	params.adev = adev;
-	shadow = parent->bo->shadow;
+	shadow = parent->base.bo->shadow;
 
 	if (vm->use_cpu_for_update) {
-		pd_addr = (unsigned long)amdgpu_bo_kptr(parent->bo);
+		pd_addr = (unsigned long)amdgpu_bo_kptr(parent->base.bo);
 		r = amdgpu_vm_wait_pd(adev, vm, AMDGPU_FENCE_OWNER_VM);
 		if (unlikely(r))
 			return r;
@@ -1046,7 +1091,7 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev,
 		/* assume the worst case */
 		ndw += parent->last_entry_used * 6;
 
-		pd_addr = amdgpu_bo_gpu_offset(parent->bo);
+		pd_addr = amdgpu_bo_gpu_offset(parent->base.bo);
 
 		if (shadow) {
 			shadow_addr = amdgpu_bo_gpu_offset(shadow);
@@ -1066,12 +1111,17 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev,
 
 	/* walk over the address space and update the directory */
 	for (pt_idx = 0; pt_idx <= parent->last_entry_used; ++pt_idx) {
-		struct amdgpu_bo *bo = parent->entries[pt_idx].bo;
+		struct amdgpu_vm_pt *entry = &parent->entries[pt_idx];
+		struct amdgpu_bo *bo = entry->base.bo;
 		uint64_t pde, pt;
 
 		if (bo == NULL)
 			continue;
 
+		spin_lock(&vm->status_lock);
+		list_del_init(&entry->base.vm_status);
+		spin_unlock(&vm->status_lock);
+
 		pt = amdgpu_bo_gpu_offset(bo);
 		pt = amdgpu_gart_get_vm_pde(adev, pt);
 		/* Don't update huge pages here */
@@ -1082,6 +1132,7 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev,
 		parent->entries[pt_idx].addr = pt | AMDGPU_PTE_VALID;
 
 		pde = pd_addr + pt_idx * 8;
+		incr = amdgpu_bo_size(bo);
 		if (((last_pde + 8 * count) != pde) ||
 		    ((last_pt + incr * count) != pt) ||
 		    (count == AMDGPU_VM_MAX_UPDATE_SIZE)) {
@@ -1109,7 +1160,7 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev,
 	}
 
 	if (count) {
-		if (vm->root.bo->shadow)
+		if (vm->root.base.bo->shadow)
 			params.func(&params, last_shadow, last_pt,
 				    count, incr, AMDGPU_PTE_VALID);
 
@@ -1122,7 +1173,8 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev,
 			amdgpu_job_free(job);
 		} else {
 			amdgpu_ring_pad_ib(ring, params.ib);
-			amdgpu_sync_resv(adev, &job->sync, parent->bo->tbo.resv,
+			amdgpu_sync_resv(adev, &job->sync,
+					 parent->base.bo->tbo.resv,
 					 AMDGPU_FENCE_OWNER_VM);
 			if (shadow)
 				amdgpu_sync_resv(adev, &job->sync,
@@ -1135,26 +1187,11 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev,
 			if (r)
 				goto error_free;
 
-			amdgpu_bo_fence(parent->bo, fence, true);
-			dma_fence_put(vm->last_dir_update);
-			vm->last_dir_update = dma_fence_get(fence);
-			dma_fence_put(fence);
+			amdgpu_bo_fence(parent->base.bo, fence, true);
+			dma_fence_put(vm->last_update);
+			vm->last_update = fence;
 		}
 	}
-	/*
-	 * Recurse into the subdirectories. This recursion is harmless because
-	 * we only have a maximum of 5 layers.
-	 */
-	for (pt_idx = 0; pt_idx <= parent->last_entry_used; ++pt_idx) {
-		struct amdgpu_vm_pt *entry = &parent->entries[pt_idx];
-
-		if (!entry->bo)
-			continue;
-
-		r = amdgpu_vm_update_level(adev, vm, entry, level + 1);
-		if (r)
-			return r;
-	}
 
 	return 0;
 
@@ -1170,7 +1207,8 @@ error_free:
  *
  * Mark all PD level as invalid after an error.
  */
-static void amdgpu_vm_invalidate_level(struct amdgpu_vm_pt *parent)
+static void amdgpu_vm_invalidate_level(struct amdgpu_vm *vm,
+				       struct amdgpu_vm_pt *parent)
 {
 	unsigned pt_idx;
 
@@ -1181,11 +1219,15 @@ static void amdgpu_vm_invalidate_level(struct amdgpu_vm_pt *parent)
 	for (pt_idx = 0; pt_idx <= parent->last_entry_used; ++pt_idx) {
 		struct amdgpu_vm_pt *entry = &parent->entries[pt_idx];
 
-		if (!entry->bo)
+		if (!entry->base.bo)
 			continue;
 
 		entry->addr = ~0ULL;
-		amdgpu_vm_invalidate_level(entry);
+		spin_lock(&vm->status_lock);
+		if (list_empty(&entry->base.vm_status))
+			list_add(&entry->base.vm_status, &vm->relocated);
+		spin_unlock(&vm->status_lock);
+		amdgpu_vm_invalidate_level(vm, entry);
 	}
 }
 
@@ -1203,9 +1245,38 @@ int amdgpu_vm_update_directories(struct amdgpu_device *adev,
 {
 	int r;
 
-	r = amdgpu_vm_update_level(adev, vm, &vm->root, 0);
-	if (r)
-		amdgpu_vm_invalidate_level(&vm->root);
+	spin_lock(&vm->status_lock);
+	while (!list_empty(&vm->relocated)) {
+		struct amdgpu_vm_bo_base *bo_base;
+		struct amdgpu_bo *bo;
+
+		bo_base = list_first_entry(&vm->relocated,
+					   struct amdgpu_vm_bo_base,
+					   vm_status);
+		spin_unlock(&vm->status_lock);
+
+		bo = bo_base->bo->parent;
+		if (bo) {
+			struct amdgpu_vm_bo_base *parent;
+			struct amdgpu_vm_pt *pt;
+
+			parent = list_first_entry(&bo->va,
+						  struct amdgpu_vm_bo_base,
+						  bo_list);
+			pt = container_of(parent, struct amdgpu_vm_pt, base);
+
+			r = amdgpu_vm_update_level(adev, vm, pt);
+			if (r) {
+				amdgpu_vm_invalidate_level(vm, &vm->root);
+				return r;
+			}
+			spin_lock(&vm->status_lock);
+		} else {
+			spin_lock(&vm->status_lock);
+			list_del_init(&bo_base->vm_status);
+		}
+	}
+	spin_unlock(&vm->status_lock);
 
 	if (vm->use_cpu_for_update) {
 		/* Flush HDP */
@@ -1236,7 +1307,7 @@ void amdgpu_vm_get_entry(struct amdgpu_pte_update_params *p, uint64_t addr,
 	*entry = &p->vm->root;
 	while ((*entry)->entries) {
 		idx = addr >> (p->adev->vm_manager.block_size * level--);
-		idx %= amdgpu_bo_size((*entry)->bo) / 8;
+		idx %= amdgpu_bo_size((*entry)->base.bo) / 8;
 		*parent = *entry;
 		*entry = &(*entry)->entries[idx];
 	}
@@ -1272,7 +1343,7 @@ static void amdgpu_vm_handle_huge_pages(struct amdgpu_pte_update_params *p,
 	    p->src ||
 	    !(flags & AMDGPU_PTE_VALID)) {
 
-		dst = amdgpu_bo_gpu_offset(entry->bo);
+		dst = amdgpu_bo_gpu_offset(entry->base.bo);
 		dst = amdgpu_gart_get_vm_pde(p->adev, dst);
 		flags = AMDGPU_PTE_VALID;
 	} else {
@@ -1298,18 +1369,18 @@ static void amdgpu_vm_handle_huge_pages(struct amdgpu_pte_update_params *p,
 		tmp = p->pages_addr;
 		p->pages_addr = NULL;
 
-		pd_addr = (unsigned long)amdgpu_bo_kptr(parent->bo);
+		pd_addr = (unsigned long)amdgpu_bo_kptr(parent->base.bo);
 		pde = pd_addr + (entry - parent->entries) * 8;
 		amdgpu_vm_cpu_set_ptes(p, pde, dst, 1, 0, flags);
 
 		p->pages_addr = tmp;
 	} else {
-		if (parent->bo->shadow) {
-			pd_addr = amdgpu_bo_gpu_offset(parent->bo->shadow);
+		if (parent->base.bo->shadow) {
+			pd_addr = amdgpu_bo_gpu_offset(parent->base.bo->shadow);
 			pde = pd_addr + (entry - parent->entries) * 8;
 			amdgpu_vm_do_set_ptes(p, pde, dst, 1, 0, flags);
 		}
-		pd_addr = amdgpu_bo_gpu_offset(parent->bo);
+		pd_addr = amdgpu_bo_gpu_offset(parent->base.bo);
 		pde = pd_addr + (entry - parent->entries) * 8;
 		amdgpu_vm_do_set_ptes(p, pde, dst, 1, 0, flags);
 	}
@@ -1360,7 +1431,7 @@ static int amdgpu_vm_update_ptes(struct amdgpu_pte_update_params *params,
 		if (entry->addr & AMDGPU_PDE_PTE)
 			continue;
 
-		pt = entry->bo;
+		pt = entry->base.bo;
 		if (use_cpu_update) {
 			pe_start = (unsigned long)amdgpu_bo_kptr(pt);
 		} else {
@@ -1396,8 +1467,6 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params	*params,
 				uint64_t start, uint64_t end,
 				uint64_t dst, uint64_t flags)
 {
-	int r;
-
 	/**
 	 * The MC L1 TLB supports variable sized pages, based on a fragment
 	 * field in the PTE. When this field is set to a non-zero value, page
@@ -1416,39 +1485,38 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params	*params,
 	 * Userspace can support this by aligning virtual base address and
 	 * allocation size to the fragment size.
 	 */
-	unsigned pages_per_frag = params->adev->vm_manager.fragment_size;
-	uint64_t frag_flags = AMDGPU_PTE_FRAG(pages_per_frag);
-	uint64_t frag_align = 1 << pages_per_frag;
-
-	uint64_t frag_start = ALIGN(start, frag_align);
-	uint64_t frag_end = end & ~(frag_align - 1);
+	unsigned max_frag = params->adev->vm_manager.fragment_size;
+	int r;
 
 	/* system pages are non continuously */
-	if (params->src || !(flags & AMDGPU_PTE_VALID) ||
-	    (frag_start >= frag_end))
+	if (params->src || !(flags & AMDGPU_PTE_VALID))
 		return amdgpu_vm_update_ptes(params, start, end, dst, flags);
 
-	/* handle the 4K area at the beginning */
-	if (start != frag_start) {
-		r = amdgpu_vm_update_ptes(params, start, frag_start,
-					  dst, flags);
+	while (start != end) {
+		uint64_t frag_flags, frag_end;
+		unsigned frag;
+
+		/* This intentionally wraps around if no bit is set */
+		frag = min((unsigned)ffs(start) - 1,
+			   (unsigned)fls64(end - start) - 1);
+		if (frag >= max_frag) {
+			frag_flags = AMDGPU_PTE_FRAG(max_frag);
+			frag_end = end & ~((1ULL << max_frag) - 1);
+		} else {
+			frag_flags = AMDGPU_PTE_FRAG(frag);
+			frag_end = start + (1 << frag);
+		}
+
+		r = amdgpu_vm_update_ptes(params, start, frag_end, dst,
+					  flags | frag_flags);
 		if (r)
 			return r;
-		dst += (frag_start - start) * AMDGPU_GPU_PAGE_SIZE;
-	}
 
-	/* handle the area in the middle */
-	r = amdgpu_vm_update_ptes(params, frag_start, frag_end, dst,
-				  flags | frag_flags);
-	if (r)
-		return r;
-
-	/* handle the 4K area at the end */
-	if (frag_end != end) {
-		dst += (frag_end - frag_start) * AMDGPU_GPU_PAGE_SIZE;
-		r = amdgpu_vm_update_ptes(params, frag_end, end, dst, flags);
+		dst += (frag_end - start) * AMDGPU_GPU_PAGE_SIZE;
+		start = frag_end;
 	}
-	return r;
+
+	return 0;
 }
 
 /**
@@ -1456,7 +1524,6 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params	*params,
  *
  * @adev: amdgpu_device pointer
  * @exclusive: fence we need to sync to
- * @src: address where to copy page table entries from
  * @pages_addr: DMA addresses to use for mapping
  * @vm: requested vm
  * @start: start of mapped range
@@ -1470,7 +1537,6 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params	*params,
  */
 static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 				       struct dma_fence *exclusive,
-				       uint64_t src,
 				       dma_addr_t *pages_addr,
 				       struct amdgpu_vm *vm,
 				       uint64_t start, uint64_t last,
@@ -1488,7 +1554,6 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 	memset(&params, 0, sizeof(params));
 	params.adev = adev;
 	params.vm = vm;
-	params.src = src;
 
 	/* sync to everything on unmapping */
 	if (!(flags & AMDGPU_PTE_VALID))
@@ -1517,10 +1582,12 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 	nptes = last - start + 1;
 
 	/*
-	 * reserve space for one command every (1 << BLOCK_SIZE)
+	 * reserve space for two commands every (1 << BLOCK_SIZE)
 	 *  entries or 2k dwords (whatever is smaller)
+         *
+         * The second command is for the shadow pagetables.
 	 */
-	ncmds = (nptes >> min(adev->vm_manager.block_size, 11u)) + 1;
+	ncmds = ((nptes >> min(adev->vm_manager.block_size, 11u)) + 1) * 2;
 
 	/* padding, etc. */
 	ndw = 64;
@@ -1528,15 +1595,9 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 	/* one PDE write for each huge page */
 	ndw += ((nptes >> adev->vm_manager.block_size) + 1) * 6;
 
-	if (src) {
-		/* only copy commands needed */
-		ndw += ncmds * 7;
-
-		params.func = amdgpu_vm_do_copy_ptes;
-
-	} else if (pages_addr) {
+	if (pages_addr) {
 		/* copy commands needed */
-		ndw += ncmds * 7;
+		ndw += ncmds * adev->vm_manager.vm_pte_funcs->copy_pte_num_dw;
 
 		/* and also PTEs */
 		ndw += nptes * 2;
@@ -1545,10 +1606,11 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 
 	} else {
 		/* set page commands needed */
-		ndw += ncmds * 10;
+		ndw += ncmds * adev->vm_manager.vm_pte_funcs->set_pte_pde_num_dw;
 
-		/* two extra commands for begin/end of fragment */
-		ndw += 2 * 10;
+		/* extra commands for begin/end fragments */
+		ndw += 2 * adev->vm_manager.vm_pte_funcs->set_pte_pde_num_dw
+				* adev->vm_manager.fragment_size;
 
 		params.func = amdgpu_vm_do_set_ptes;
 	}
@@ -1559,7 +1621,7 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 
 	params.ib = &job->ibs[0];
 
-	if (!src && pages_addr) {
+	if (pages_addr) {
 		uint64_t *pte;
 		unsigned i;
 
@@ -1580,12 +1642,12 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 	if (r)
 		goto error_free;
 
-	r = amdgpu_sync_resv(adev, &job->sync, vm->root.bo->tbo.resv,
+	r = amdgpu_sync_resv(adev, &job->sync, vm->root.base.bo->tbo.resv,
 			     owner);
 	if (r)
 		goto error_free;
 
-	r = reservation_object_reserve_shared(vm->root.bo->tbo.resv);
+	r = reservation_object_reserve_shared(vm->root.base.bo->tbo.resv);
 	if (r)
 		goto error_free;
 
@@ -1600,14 +1662,14 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 	if (r)
 		goto error_free;
 
-	amdgpu_bo_fence(vm->root.bo, f, true);
+	amdgpu_bo_fence(vm->root.base.bo, f, true);
 	dma_fence_put(*fence);
 	*fence = f;
 	return 0;
 
 error_free:
 	amdgpu_job_free(job);
-	amdgpu_vm_invalidate_level(&vm->root);
+	amdgpu_vm_invalidate_level(vm, &vm->root);
 	return r;
 }
 
@@ -1636,7 +1698,7 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev,
 				      struct drm_mm_node *nodes,
 				      struct dma_fence **fence)
 {
-	uint64_t pfn, src = 0, start = mapping->start;
+	uint64_t pfn, start = mapping->start;
 	int r;
 
 	/* normally,bo_va->flags only contians READABLE and WIRTEABLE bit go here
@@ -1691,8 +1753,7 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev,
 		addr += pfn << PAGE_SHIFT;
 
 		last = min((uint64_t)mapping->last, start + max_entries - 1);
-		r = amdgpu_vm_bo_update_mapping(adev, exclusive,
-						src, pages_addr, vm,
+		r = amdgpu_vm_bo_update_mapping(adev, exclusive, pages_addr, vm,
 						start, last, flags, addr,
 						fence);
 		if (r)
@@ -1730,7 +1791,7 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev,
 	dma_addr_t *pages_addr = NULL;
 	struct ttm_mem_reg *mem;
 	struct drm_mm_node *nodes;
-	struct dma_fence *exclusive;
+	struct dma_fence *exclusive, **last_update;
 	uint64_t flags;
 	int r;
 
@@ -1756,38 +1817,43 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev,
 	else
 		flags = 0x0;
 
-	spin_lock(&vm->status_lock);
-	if (!list_empty(&bo_va->base.vm_status))
+	if (clear || (bo && bo->tbo.resv == vm->root.base.bo->tbo.resv))
+		last_update = &vm->last_update;
+	else
+		last_update = &bo_va->last_pt_update;
+
+	if (!clear && bo_va->base.moved) {
+		bo_va->base.moved = false;
 		list_splice_init(&bo_va->valids, &bo_va->invalids);
-	spin_unlock(&vm->status_lock);
+
+	} else if (bo_va->cleared != clear) {
+		list_splice_init(&bo_va->valids, &bo_va->invalids);
+	}
 
 	list_for_each_entry(mapping, &bo_va->invalids, list) {
 		r = amdgpu_vm_bo_split_mapping(adev, exclusive, pages_addr, vm,
 					       mapping, flags, nodes,
-					       &bo_va->last_pt_update);
+					       last_update);
 		if (r)
 			return r;
 	}
 
-	if (trace_amdgpu_vm_bo_mapping_enabled()) {
-		list_for_each_entry(mapping, &bo_va->valids, list)
-			trace_amdgpu_vm_bo_mapping(mapping);
-
-		list_for_each_entry(mapping, &bo_va->invalids, list)
-			trace_amdgpu_vm_bo_mapping(mapping);
+	if (vm->use_cpu_for_update) {
+		/* Flush HDP */
+		mb();
+		amdgpu_gart_flush_gpu_tlb(adev, 0);
 	}
 
 	spin_lock(&vm->status_lock);
-	list_splice_init(&bo_va->invalids, &bo_va->valids);
 	list_del_init(&bo_va->base.vm_status);
-	if (clear)
-		list_add(&bo_va->base.vm_status, &vm->cleared);
 	spin_unlock(&vm->status_lock);
 
-	if (vm->use_cpu_for_update) {
-		/* Flush HDP */
-		mb();
-		amdgpu_gart_flush_gpu_tlb(adev, 0);
+	list_splice_init(&bo_va->invalids, &bo_va->valids);
+	bo_va->cleared = clear;
+
+	if (trace_amdgpu_vm_bo_mapping_enabled()) {
+		list_for_each_entry(mapping, &bo_va->valids, list)
+			trace_amdgpu_vm_bo_mapping(mapping);
 	}
 
 	return 0;
@@ -1895,7 +1961,7 @@ static void amdgpu_vm_free_mapping(struct amdgpu_device *adev,
  */
 static void amdgpu_vm_prt_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 {
-	struct reservation_object *resv = vm->root.bo->tbo.resv;
+	struct reservation_object *resv = vm->root.base.bo->tbo.resv;
 	struct dma_fence *excl, **shared;
 	unsigned i, shared_count;
 	int r;
@@ -1953,7 +2019,7 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
 		if (vm->pte_support_ats)
 			init_pte_value = AMDGPU_PTE_SYSTEM;
 
-		r = amdgpu_vm_bo_update_mapping(adev, NULL, 0, NULL, vm,
+		r = amdgpu_vm_bo_update_mapping(adev, NULL, NULL, vm,
 						mapping->start, mapping->last,
 						init_pte_value, 0, &f);
 		amdgpu_vm_free_mapping(adev, vm, mapping, f);
@@ -1975,29 +2041,35 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
 }
 
 /**
- * amdgpu_vm_clear_moved - clear moved BOs in the PT
+ * amdgpu_vm_handle_moved - handle moved BOs in the PT
  *
  * @adev: amdgpu_device pointer
  * @vm: requested vm
+ * @sync: sync object to add fences to
  *
- * Make sure all moved BOs are cleared in the PT.
+ * Make sure all BOs which are moved are updated in the PTs.
  * Returns 0 for success.
  *
- * PTs have to be reserved and mutex must be locked!
+ * PTs have to be reserved!
  */
-int amdgpu_vm_clear_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			    struct amdgpu_sync *sync)
+int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
+			   struct amdgpu_vm *vm)
 {
-	struct amdgpu_bo_va *bo_va = NULL;
+	bool clear;
 	int r = 0;
 
 	spin_lock(&vm->status_lock);
 	while (!list_empty(&vm->moved)) {
+		struct amdgpu_bo_va *bo_va;
+
 		bo_va = list_first_entry(&vm->moved,
 			struct amdgpu_bo_va, base.vm_status);
 		spin_unlock(&vm->status_lock);
 
-		r = amdgpu_vm_bo_update(adev, bo_va, true);
+		/* Per VM BOs never need to bo cleared in the page tables */
+		clear = bo_va->base.bo->tbo.resv != vm->root.base.bo->tbo.resv;
+
+		r = amdgpu_vm_bo_update(adev, bo_va, clear);
 		if (r)
 			return r;
 
@@ -2005,9 +2077,6 @@ int amdgpu_vm_clear_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	}
 	spin_unlock(&vm->status_lock);
 
-	if (bo_va)
-		r = amdgpu_sync_fence(adev, sync, bo_va->last_pt_update);
-
 	return r;
 }
 
@@ -2049,6 +2118,39 @@ struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev,
 	return bo_va;
 }
 
+
+/**
+ * amdgpu_vm_bo_insert_mapping - insert a new mapping
+ *
+ * @adev: amdgpu_device pointer
+ * @bo_va: bo_va to store the address
+ * @mapping: the mapping to insert
+ *
+ * Insert a new mapping into all structures.
+ */
+static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev,
+				    struct amdgpu_bo_va *bo_va,
+				    struct amdgpu_bo_va_mapping *mapping)
+{
+	struct amdgpu_vm *vm = bo_va->base.vm;
+	struct amdgpu_bo *bo = bo_va->base.bo;
+
+	mapping->bo_va = bo_va;
+	list_add(&mapping->list, &bo_va->invalids);
+	amdgpu_vm_it_insert(mapping, &vm->va);
+
+	if (mapping->flags & AMDGPU_PTE_PRT)
+		amdgpu_vm_prt_get(adev);
+
+	if (bo && bo->tbo.resv == vm->root.base.bo->tbo.resv) {
+		spin_lock(&vm->status_lock);
+		if (list_empty(&bo_va->base.vm_status))
+			list_add(&bo_va->base.vm_status, &vm->moved);
+		spin_unlock(&vm->status_lock);
+	}
+	trace_amdgpu_vm_bo_map(bo_va, mapping);
+}
+
 /**
  * amdgpu_vm_bo_map - map bo inside a vm
  *
@@ -2100,17 +2202,12 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 	if (!mapping)
 		return -ENOMEM;
 
-	INIT_LIST_HEAD(&mapping->list);
 	mapping->start = saddr;
 	mapping->last = eaddr;
 	mapping->offset = offset;
 	mapping->flags = flags;
 
-	list_add(&mapping->list, &bo_va->invalids);
-	amdgpu_vm_it_insert(mapping, &vm->va);
-
-	if (flags & AMDGPU_PTE_PRT)
-		amdgpu_vm_prt_get(adev);
+	amdgpu_vm_bo_insert_map(adev, bo_va, mapping);
 
 	return 0;
 }
@@ -2137,7 +2234,6 @@ int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev,
 {
 	struct amdgpu_bo_va_mapping *mapping;
 	struct amdgpu_bo *bo = bo_va->base.bo;
-	struct amdgpu_vm *vm = bo_va->base.vm;
 	uint64_t eaddr;
 	int r;
 
@@ -2171,11 +2267,7 @@ int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev,
 	mapping->offset = offset;
 	mapping->flags = flags;
 
-	list_add(&mapping->list, &bo_va->invalids);
-	amdgpu_vm_it_insert(mapping, &vm->va);
-
-	if (flags & AMDGPU_PTE_PRT)
-		amdgpu_vm_prt_get(adev);
+	amdgpu_vm_bo_insert_map(adev, bo_va, mapping);
 
 	return 0;
 }
@@ -2221,6 +2313,7 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
 
 	list_del(&mapping->list);
 	amdgpu_vm_it_remove(mapping, &vm->va);
+	mapping->bo_va = NULL;
 	trace_amdgpu_vm_bo_unmap(bo_va, mapping);
 
 	if (valid)
@@ -2306,6 +2399,7 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,
 		if (tmp->last > eaddr)
 		    tmp->last = eaddr;
 
+		tmp->bo_va = NULL;
 		list_add(&tmp->list, &vm->freed);
 		trace_amdgpu_vm_bo_unmap(NULL, tmp);
 	}
@@ -2331,6 +2425,19 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,
 	return 0;
 }
 
+/**
+ * amdgpu_vm_bo_lookup_mapping - find mapping by address
+ *
+ * @vm: the requested VM
+ *
+ * Find a mapping by it's address.
+ */
+struct amdgpu_bo_va_mapping *amdgpu_vm_bo_lookup_mapping(struct amdgpu_vm *vm,
+							 uint64_t addr)
+{
+	return amdgpu_vm_it_iter_first(&vm->va, addr, addr);
+}
+
 /**
  * amdgpu_vm_bo_rmv - remove a bo to a specific vm
  *
@@ -2356,6 +2463,7 @@ void amdgpu_vm_bo_rmv(struct amdgpu_device *adev,
 	list_for_each_entry_safe(mapping, next, &bo_va->valids, list) {
 		list_del(&mapping->list);
 		amdgpu_vm_it_remove(mapping, &vm->va);
+		mapping->bo_va = NULL;
 		trace_amdgpu_vm_bo_unmap(bo_va, mapping);
 		list_add(&mapping->list, &vm->freed);
 	}
@@ -2380,15 +2488,36 @@ void amdgpu_vm_bo_rmv(struct amdgpu_device *adev,
  * Mark @bo as invalid.
  */
 void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
-			     struct amdgpu_bo *bo)
+			     struct amdgpu_bo *bo, bool evicted)
 {
 	struct amdgpu_vm_bo_base *bo_base;
 
 	list_for_each_entry(bo_base, &bo->va, bo_list) {
+		struct amdgpu_vm *vm = bo_base->vm;
+
+		bo_base->moved = true;
+		if (evicted && bo->tbo.resv == vm->root.base.bo->tbo.resv) {
+			spin_lock(&bo_base->vm->status_lock);
+			if (bo->tbo.type == ttm_bo_type_kernel)
+				list_move(&bo_base->vm_status, &vm->evicted);
+			else
+				list_move_tail(&bo_base->vm_status,
+					       &vm->evicted);
+			spin_unlock(&bo_base->vm->status_lock);
+			continue;
+		}
+
+		if (bo->tbo.type == ttm_bo_type_kernel) {
+			spin_lock(&bo_base->vm->status_lock);
+			if (list_empty(&bo_base->vm_status))
+				list_add(&bo_base->vm_status, &vm->relocated);
+			spin_unlock(&bo_base->vm->status_lock);
+			continue;
+		}
+
 		spin_lock(&bo_base->vm->status_lock);
 		if (list_empty(&bo_base->vm_status))
-			list_add(&bo_base->vm_status,
-				 &bo_base->vm->moved);
+			list_add(&bo_base->vm_status, &vm->moved);
 		spin_unlock(&bo_base->vm->status_lock);
 	}
 }
@@ -2458,7 +2587,7 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint64_t vm_size, uint32_
  * Init @vm fields.
  */
 int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-		   int vm_context)
+		   int vm_context, unsigned int pasid)
 {
 	const unsigned align = min(AMDGPU_VM_PTB_ALIGN_SIZE,
 		AMDGPU_VM_PTE_COUNT(adev) * 8);
@@ -2474,8 +2603,9 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++)
 		vm->reserved_vmid[i] = NULL;
 	spin_lock_init(&vm->status_lock);
+	INIT_LIST_HEAD(&vm->evicted);
+	INIT_LIST_HEAD(&vm->relocated);
 	INIT_LIST_HEAD(&vm->moved);
-	INIT_LIST_HEAD(&vm->cleared);
 	INIT_LIST_HEAD(&vm->freed);
 
 	/* create scheduler entity for page table updates */
@@ -2506,7 +2636,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			 vm->use_cpu_for_update ? "CPU" : "SDMA");
 	WARN_ONCE((vm->use_cpu_for_update & !amdgpu_vm_is_large_bar(adev)),
 		  "CPU update of VM recommended only for large BAR system\n");
-	vm->last_dir_update = NULL;
+	vm->last_update = NULL;
 
 	flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS |
 			AMDGPU_GEM_CREATE_VRAM_CLEARED;
@@ -2519,30 +2649,46 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	r = amdgpu_bo_create(adev, amdgpu_vm_bo_size(adev, 0), align, true,
 			     AMDGPU_GEM_DOMAIN_VRAM,
 			     flags,
-			     NULL, NULL, init_pde_value, &vm->root.bo);
+			     NULL, NULL, init_pde_value, &vm->root.base.bo);
 	if (r)
 		goto error_free_sched_entity;
 
-	r = amdgpu_bo_reserve(vm->root.bo, false);
-	if (r)
-		goto error_free_root;
-
-	vm->last_eviction_counter = atomic64_read(&adev->num_evictions);
+	vm->root.base.vm = vm;
+	list_add_tail(&vm->root.base.bo_list, &vm->root.base.bo->va);
+	INIT_LIST_HEAD(&vm->root.base.vm_status);
 
 	if (vm->use_cpu_for_update) {
-		r = amdgpu_bo_kmap(vm->root.bo, NULL);
+		r = amdgpu_bo_reserve(vm->root.base.bo, false);
 		if (r)
 			goto error_free_root;
+
+		r = amdgpu_bo_kmap(vm->root.base.bo, NULL);
+		amdgpu_bo_unreserve(vm->root.base.bo);
+		if (r)
+			goto error_free_root;
+	}
+
+	if (pasid) {
+		unsigned long flags;
+
+		spin_lock_irqsave(&adev->vm_manager.pasid_lock, flags);
+		r = idr_alloc(&adev->vm_manager.pasid_idr, vm, pasid, pasid + 1,
+			      GFP_ATOMIC);
+		spin_unlock_irqrestore(&adev->vm_manager.pasid_lock, flags);
+		if (r < 0)
+			goto error_free_root;
+
+		vm->pasid = pasid;
 	}
 
-	amdgpu_bo_unreserve(vm->root.bo);
+	INIT_KFIFO(vm->faults);
 
 	return 0;
 
 error_free_root:
-	amdgpu_bo_unref(&vm->root.bo->shadow);
-	amdgpu_bo_unref(&vm->root.bo);
-	vm->root.bo = NULL;
+	amdgpu_bo_unref(&vm->root.base.bo->shadow);
+	amdgpu_bo_unref(&vm->root.base.bo);
+	vm->root.base.bo = NULL;
 
 error_free_sched_entity:
 	amd_sched_entity_fini(&ring->sched, &vm->entity);
@@ -2561,9 +2707,11 @@ static void amdgpu_vm_free_levels(struct amdgpu_vm_pt *level)
 {
 	unsigned i;
 
-	if (level->bo) {
-		amdgpu_bo_unref(&level->bo->shadow);
-		amdgpu_bo_unref(&level->bo);
+	if (level->base.bo) {
+		list_del(&level->base.bo_list);
+		list_del(&level->base.vm_status);
+		amdgpu_bo_unref(&level->base.bo->shadow);
+		amdgpu_bo_unref(&level->base.bo);
 	}
 
 	if (level->entries)
@@ -2586,8 +2734,21 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 {
 	struct amdgpu_bo_va_mapping *mapping, *tmp;
 	bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt;
+	u64 fault;
 	int i;
 
+	/* Clear pending page faults from IH when the VM is destroyed */
+	while (kfifo_get(&vm->faults, &fault))
+		amdgpu_ih_clear_fault(adev, fault);
+
+	if (vm->pasid) {
+		unsigned long flags;
+
+		spin_lock_irqsave(&adev->vm_manager.pasid_lock, flags);
+		idr_remove(&adev->vm_manager.pasid_idr, vm->pasid);
+		spin_unlock_irqrestore(&adev->vm_manager.pasid_lock, flags);
+	}
+
 	amd_sched_entity_fini(vm->entity.sched, &vm->entity);
 
 	if (!RB_EMPTY_ROOT(&vm->va.rb_root)) {
@@ -2610,7 +2771,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 	}
 
 	amdgpu_vm_free_levels(&vm->root);
-	dma_fence_put(vm->last_dir_update);
+	dma_fence_put(vm->last_update);
 	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++)
 		amdgpu_vm_free_reserved_vmid(adev, vm, i);
 }
@@ -2668,6 +2829,8 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev)
 	adev->vm_manager.vm_update_mode = 0;
 #endif
 
+	idr_init(&adev->vm_manager.pasid_idr);
+	spin_lock_init(&adev->vm_manager.pasid_lock);
 }
 
 /**
@@ -2681,6 +2844,9 @@ void amdgpu_vm_manager_fini(struct amdgpu_device *adev)
 {
 	unsigned i, j;
 
+	WARN_ON(!idr_is_empty(&adev->vm_manager.pasid_idr));
+	idr_destroy(&adev->vm_manager.pasid_idr);
+
 	for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) {
 		struct amdgpu_vm_id_manager *id_mgr =
 			&adev->vm_manager.id_mgr[i];

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

@@ -25,6 +25,7 @@
 #define __AMDGPU_VM_H__
 
 #include <linux/rbtree.h>
+#include <linux/idr.h>
 
 #include "gpu_scheduler.h"
 #include "amdgpu_sync.h"
@@ -105,17 +106,24 @@ struct amdgpu_vm_bo_base {
 
 	/* protected by spinlock */
 	struct list_head		vm_status;
+
+	/* protected by the BO being reserved */
+	bool				moved;
 };
 
 struct amdgpu_vm_pt {
-	struct amdgpu_bo	*bo;
-	uint64_t		addr;
+	struct amdgpu_vm_bo_base	base;
+	uint64_t			addr;
 
 	/* array of page tables, one for each directory entry */
-	struct amdgpu_vm_pt	*entries;
-	unsigned		last_entry_used;
+	struct amdgpu_vm_pt		*entries;
+	unsigned			last_entry_used;
 };
 
+#define AMDGPU_VM_FAULT(pasid, addr) (((u64)(pasid) << 48) | (addr))
+#define AMDGPU_VM_FAULT_PASID(fault) ((u64)(fault) >> 48)
+#define AMDGPU_VM_FAULT_ADDR(fault)  ((u64)(fault) & 0xfffffffff000ULL)
+
 struct amdgpu_vm {
 	/* tree of virtual addresses mapped */
 	struct rb_root_cached	va;
@@ -123,19 +131,21 @@ struct amdgpu_vm {
 	/* protecting invalidated */
 	spinlock_t		status_lock;
 
+	/* BOs who needs a validation */
+	struct list_head	evicted;
+
+	/* PT BOs which relocated and their parent need an update */
+	struct list_head	relocated;
+
 	/* BOs moved, but not yet updated in the PT */
 	struct list_head	moved;
 
-	/* BOs cleared in the PT because of a move */
-	struct list_head	cleared;
-
 	/* BO mappings freed, but not yet updated in the PT */
 	struct list_head	freed;
 
 	/* contains the page directory */
 	struct amdgpu_vm_pt     root;
-	struct dma_fence	*last_dir_update;
-	uint64_t		last_eviction_counter;
+	struct dma_fence	*last_update;
 
 	/* protecting freed */
 	spinlock_t		freed_lock;
@@ -143,8 +153,9 @@ struct amdgpu_vm {
 	/* Scheduler entity for page table updates */
 	struct amd_sched_entity	entity;
 
-	/* client id */
+	/* client id and PASID (TODO: replace client_id with PASID) */
 	u64                     client_id;
+	unsigned int		pasid;
 	/* dedicated to vm */
 	struct amdgpu_vm_id	*reserved_vmid[AMDGPU_MAX_VMHUBS];
 
@@ -153,6 +164,9 @@ struct amdgpu_vm {
 
 	/* Flag to indicate ATS support from PTE for GFX9 */
 	bool			pte_support_ats;
+
+	/* Up to 128 pending page faults */
+	DECLARE_KFIFO(faults, u64, 128);
 };
 
 struct amdgpu_vm_id {
@@ -215,16 +229,25 @@ struct amdgpu_vm_manager {
 	 * BIT1[= 0] Compute updated by SDMA [= 1] by CPU
 	 */
 	int					vm_update_mode;
+
+	/* PASID to VM mapping, will be used in interrupt context to
+	 * look up VM of a page fault
+	 */
+	struct idr				pasid_idr;
+	spinlock_t				pasid_lock;
 };
 
+int amdgpu_vm_alloc_pasid(unsigned int bits);
+void amdgpu_vm_free_pasid(unsigned int pasid);
 void amdgpu_vm_manager_init(struct amdgpu_device *adev);
 void amdgpu_vm_manager_fini(struct amdgpu_device *adev);
 int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-		   int vm_context);
+		   int vm_context, unsigned int pasid);
 void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm);
 void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
 			 struct list_head *validated,
 			 struct amdgpu_bo_list_entry *entry);
+bool amdgpu_vm_ready(struct amdgpu_vm *vm);
 int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			      int (*callback)(void *p, struct amdgpu_bo *bo),
 			      void *param);
@@ -243,13 +266,13 @@ int amdgpu_vm_update_directories(struct amdgpu_device *adev,
 int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
 			  struct amdgpu_vm *vm,
 			  struct dma_fence **fence);
-int amdgpu_vm_clear_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			  struct amdgpu_sync *sync);
+int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
+			   struct amdgpu_vm *vm);
 int amdgpu_vm_bo_update(struct amdgpu_device *adev,
 			struct amdgpu_bo_va *bo_va,
 			bool clear);
 void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
-			     struct amdgpu_bo *bo);
+			     struct amdgpu_bo *bo, bool evicted);
 struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm,
 				       struct amdgpu_bo *bo);
 struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev,
@@ -269,6 +292,8 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
 int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,
 				struct amdgpu_vm *vm,
 				uint64_t saddr, uint64_t size);
+struct amdgpu_bo_va_mapping *amdgpu_vm_bo_lookup_mapping(struct amdgpu_vm *vm,
+							 uint64_t addr);
 void amdgpu_vm_bo_rmv(struct amdgpu_device *adev,
 		      struct amdgpu_bo_va *bo_va);
 void amdgpu_vm_set_fragment_size(struct amdgpu_device *adev,

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

@@ -1343,8 +1343,11 @@ struct atom_context *amdgpu_atom_parse(struct card_info *card, void *bios)
 		idx = 0x80;
 
 	str = CSTR(idx);
-	if (*str != '\0')
+	if (*str != '\0') {
 		pr_info("ATOM BIOS: %s\n", str);
+		strlcpy(ctx->vbios_version, str, sizeof(ctx->vbios_version));
+	}
+
 
 	return ctx;
 }

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

@@ -140,6 +140,7 @@ struct atom_context {
 	int io_mode;
 	uint32_t *scratch;
 	int scratch_size_bytes;
+	char vbios_version[20];
 };
 
 extern int amdgpu_atom_debug;

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

@@ -307,7 +307,6 @@ static int ci_set_power_limit(struct amdgpu_device *adev, u32 n);
 static int ci_set_overdrive_target_tdp(struct amdgpu_device *adev,
 				       u32 target_tdp);
 static int ci_update_uvd_dpm(struct amdgpu_device *adev, bool gate);
-static void ci_dpm_set_dpm_funcs(struct amdgpu_device *adev);
 static void ci_dpm_set_irq_funcs(struct amdgpu_device *adev);
 
 static PPSMC_Result amdgpu_ci_send_msg_to_smc_with_parameter(struct amdgpu_device *adev,
@@ -883,8 +882,9 @@ static int ci_power_control_set_level(struct amdgpu_device *adev)
 	return ret;
 }
 
-static void ci_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate)
+static void ci_dpm_powergate_uvd(void *handle, bool gate)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 
 	pi->uvd_power_gated = gate;
@@ -901,8 +901,9 @@ static void ci_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate)
 	}
 }
 
-static bool ci_dpm_vblank_too_short(struct amdgpu_device *adev)
+static bool ci_dpm_vblank_too_short(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	u32 vblank_time = amdgpu_dpm_get_vblank_time(adev);
 	u32 switch_limit = adev->mc.vram_type == AMDGPU_VRAM_TYPE_GDDR5 ? 450 : 300;
 
@@ -1210,11 +1211,12 @@ static int ci_fan_ctrl_stop_smc_fan_control(struct amdgpu_device *adev)
 	}
 }
 
-static int ci_dpm_get_fan_speed_percent(struct amdgpu_device *adev,
+static int ci_dpm_get_fan_speed_percent(void *handle,
 					u32 *speed)
 {
 	u32 duty, duty100;
 	u64 tmp64;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	if (adev->pm.no_fan)
 		return -ENOENT;
@@ -1237,12 +1239,13 @@ static int ci_dpm_get_fan_speed_percent(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int ci_dpm_set_fan_speed_percent(struct amdgpu_device *adev,
+static int ci_dpm_set_fan_speed_percent(void *handle,
 					u32 speed)
 {
 	u32 tmp;
 	u32 duty, duty100;
 	u64 tmp64;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 
 	if (adev->pm.no_fan)
@@ -1271,8 +1274,10 @@ static int ci_dpm_set_fan_speed_percent(struct amdgpu_device *adev,
 	return 0;
 }
 
-static void ci_dpm_set_fan_control_mode(struct amdgpu_device *adev, u32 mode)
+static void ci_dpm_set_fan_control_mode(void *handle, u32 mode)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
 	switch (mode) {
 	case AMD_FAN_CTRL_NONE:
 		if (adev->pm.dpm.fan.ucode_fan_control)
@@ -1292,8 +1297,9 @@ static void ci_dpm_set_fan_control_mode(struct amdgpu_device *adev, u32 mode)
 	}
 }
 
-static u32 ci_dpm_get_fan_control_mode(struct amdgpu_device *adev)
+static u32 ci_dpm_get_fan_control_mode(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 
 	if (pi->fan_is_controlled_by_smc)
@@ -4378,9 +4384,10 @@ static u32 ci_get_lowest_enabled_level(struct amdgpu_device *adev,
 }
 
 
-static int ci_dpm_force_performance_level(struct amdgpu_device *adev,
+static int ci_dpm_force_performance_level(void *handle,
 					  enum amd_dpm_forced_level level)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	u32 tmp, levels, i;
 	int ret;
@@ -5291,8 +5298,9 @@ static void ci_update_requested_ps(struct amdgpu_device *adev,
 	adev->pm.dpm.requested_ps = &pi->requested_rps;
 }
 
-static int ci_dpm_pre_set_power_state(struct amdgpu_device *adev)
+static int ci_dpm_pre_set_power_state(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct amdgpu_ps requested_ps = *adev->pm.dpm.requested_ps;
 	struct amdgpu_ps *new_ps = &requested_ps;
@@ -5304,8 +5312,9 @@ static int ci_dpm_pre_set_power_state(struct amdgpu_device *adev)
 	return 0;
 }
 
-static void ci_dpm_post_set_power_state(struct amdgpu_device *adev)
+static void ci_dpm_post_set_power_state(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct amdgpu_ps *new_ps = &pi->requested_rps;
 
@@ -5479,8 +5488,9 @@ static void ci_dpm_disable(struct amdgpu_device *adev)
 	ci_update_current_ps(adev, boot_ps);
 }
 
-static int ci_dpm_set_power_state(struct amdgpu_device *adev)
+static int ci_dpm_set_power_state(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct amdgpu_ps *new_ps = &pi->requested_rps;
 	struct amdgpu_ps *old_ps = &pi->current_rps;
@@ -5551,8 +5561,10 @@ static void ci_dpm_reset_asic(struct amdgpu_device *adev)
 }
 #endif
 
-static void ci_dpm_display_configuration_changed(struct amdgpu_device *adev)
+static void ci_dpm_display_configuration_changed(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
 	ci_program_display_gap(adev);
 }
 
@@ -6105,9 +6117,10 @@ static int ci_dpm_init(struct amdgpu_device *adev)
 }
 
 static void
-ci_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev,
+ci_dpm_debugfs_print_current_performance_level(void *handle,
 					       struct seq_file *m)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct amdgpu_ps *rps = &pi->current_rps;
 	u32 sclk = ci_get_average_sclk_freq(adev);
@@ -6131,12 +6144,13 @@ ci_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev,
 	seq_printf(m, "GPU load: %u %%\n", activity_percent);
 }
 
-static void ci_dpm_print_power_state(struct amdgpu_device *adev,
-				     struct amdgpu_ps *rps)
+static void ci_dpm_print_power_state(void *handle, void *current_ps)
 {
+	struct amdgpu_ps *rps = (struct amdgpu_ps *)current_ps;
 	struct ci_ps *ps = ci_get_ps(rps);
 	struct ci_pl *pl;
 	int i;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	amdgpu_dpm_print_class_info(rps->class, rps->class2);
 	amdgpu_dpm_print_cap_info(rps->caps);
@@ -6158,20 +6172,23 @@ static inline bool ci_are_power_levels_equal(const struct ci_pl *ci_cpl1,
 		  (ci_cpl1->pcie_lane == ci_cpl2->pcie_lane));
 }
 
-static int ci_check_state_equal(struct amdgpu_device *adev,
-				struct amdgpu_ps *cps,
-				struct amdgpu_ps *rps,
+static int ci_check_state_equal(void *handle,
+				void *current_ps,
+				void *request_ps,
 				bool *equal)
 {
 	struct ci_ps *ci_cps;
 	struct ci_ps *ci_rps;
 	int i;
+	struct amdgpu_ps *cps = (struct amdgpu_ps *)current_ps;
+	struct amdgpu_ps *rps = (struct amdgpu_ps *)request_ps;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	if (adev == NULL || cps == NULL || rps == NULL || equal == NULL)
 		return -EINVAL;
 
-	ci_cps = ci_get_ps(cps);
-	ci_rps = ci_get_ps(rps);
+	ci_cps = ci_get_ps((struct amdgpu_ps *)cps);
+	ci_rps = ci_get_ps((struct amdgpu_ps *)rps);
 
 	if (ci_cps == NULL) {
 		*equal = false;
@@ -6199,8 +6216,9 @@ static int ci_check_state_equal(struct amdgpu_device *adev,
 	return 0;
 }
 
-static u32 ci_dpm_get_sclk(struct amdgpu_device *adev, bool low)
+static u32 ci_dpm_get_sclk(void *handle, bool low)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct ci_ps *requested_state = ci_get_ps(&pi->requested_rps);
 
@@ -6210,8 +6228,9 @@ static u32 ci_dpm_get_sclk(struct amdgpu_device *adev, bool low)
 		return requested_state->performance_levels[requested_state->performance_level_count - 1].sclk;
 }
 
-static u32 ci_dpm_get_mclk(struct amdgpu_device *adev, bool low)
+static u32 ci_dpm_get_mclk(void *handle, bool low)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct ci_ps *requested_state = ci_get_ps(&pi->requested_rps);
 
@@ -6222,10 +6241,11 @@ static u32 ci_dpm_get_mclk(struct amdgpu_device *adev, bool low)
 }
 
 /* get temperature in millidegrees */
-static int ci_dpm_get_temp(struct amdgpu_device *adev)
+static int ci_dpm_get_temp(void *handle)
 {
 	u32 temp;
 	int actual_temp = 0;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	temp = (RREG32_SMC(ixCG_MULT_THERMAL_STATUS) & CG_MULT_THERMAL_STATUS__CTF_TEMP_MASK) >>
 		CG_MULT_THERMAL_STATUS__CTF_TEMP__SHIFT;
@@ -6261,7 +6281,6 @@ static int ci_dpm_early_init(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	ci_dpm_set_dpm_funcs(adev);
 	ci_dpm_set_irq_funcs(adev);
 
 	return 0;
@@ -6551,9 +6570,10 @@ static int ci_dpm_set_powergating_state(void *handle,
 	return 0;
 }
 
-static int ci_dpm_print_clock_levels(struct amdgpu_device *adev,
+static int ci_dpm_print_clock_levels(void *handle,
 		enum pp_clock_type type, char *buf)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct ci_single_dpm_table *sclk_table = &pi->dpm_table.sclk_table;
 	struct ci_single_dpm_table *mclk_table = &pi->dpm_table.mclk_table;
@@ -6618,9 +6638,10 @@ static int ci_dpm_print_clock_levels(struct amdgpu_device *adev,
 	return size;
 }
 
-static int ci_dpm_force_clock_level(struct amdgpu_device *adev,
+static int ci_dpm_force_clock_level(void *handle,
 		enum pp_clock_type type, uint32_t mask)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 
 	if (adev->pm.dpm.forced_level & (AMD_DPM_FORCED_LEVEL_AUTO |
@@ -6664,8 +6685,9 @@ static int ci_dpm_force_clock_level(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int ci_dpm_get_sclk_od(struct amdgpu_device *adev)
+static int ci_dpm_get_sclk_od(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct ci_single_dpm_table *sclk_table = &(pi->dpm_table.sclk_table);
 	struct ci_single_dpm_table *golden_sclk_table =
@@ -6680,8 +6702,9 @@ static int ci_dpm_get_sclk_od(struct amdgpu_device *adev)
 	return value;
 }
 
-static int ci_dpm_set_sclk_od(struct amdgpu_device *adev, uint32_t value)
+static int ci_dpm_set_sclk_od(void *handle, uint32_t value)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct ci_ps *ps = ci_get_ps(adev->pm.dpm.requested_ps);
 	struct ci_single_dpm_table *golden_sclk_table =
@@ -6698,8 +6721,9 @@ static int ci_dpm_set_sclk_od(struct amdgpu_device *adev, uint32_t value)
 	return 0;
 }
 
-static int ci_dpm_get_mclk_od(struct amdgpu_device *adev)
+static int ci_dpm_get_mclk_od(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct ci_single_dpm_table *mclk_table = &(pi->dpm_table.mclk_table);
 	struct ci_single_dpm_table *golden_mclk_table =
@@ -6714,8 +6738,9 @@ static int ci_dpm_get_mclk_od(struct amdgpu_device *adev)
 	return value;
 }
 
-static int ci_dpm_set_mclk_od(struct amdgpu_device *adev, uint32_t value)
+static int ci_dpm_set_mclk_od(void *handle, uint32_t value)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct ci_ps *ps = ci_get_ps(adev->pm.dpm.requested_ps);
 	struct ci_single_dpm_table *golden_mclk_table =
@@ -6732,9 +6757,10 @@ static int ci_dpm_set_mclk_od(struct amdgpu_device *adev, uint32_t value)
 	return 0;
 }
 
-static int ci_dpm_get_power_profile_state(struct amdgpu_device *adev,
+static int ci_dpm_get_power_profile_state(void *handle,
 		struct amd_pp_profile *query)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 
 	if (!pi || !query)
@@ -6851,9 +6877,10 @@ static int ci_set_power_profile_state(struct amdgpu_device *adev,
 	return result;
 }
 
-static int ci_dpm_set_power_profile_state(struct amdgpu_device *adev,
+static int ci_dpm_set_power_profile_state(void *handle,
 		struct amd_pp_profile *request)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	int ret = -1;
 
@@ -6906,9 +6933,10 @@ static int ci_dpm_set_power_profile_state(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int ci_dpm_reset_power_profile_state(struct amdgpu_device *adev,
+static int ci_dpm_reset_power_profile_state(void *handle,
 		struct amd_pp_profile *request)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 
 	if (!pi || !request)
@@ -6927,9 +6955,10 @@ static int ci_dpm_reset_power_profile_state(struct amdgpu_device *adev,
 		return -EINVAL;
 }
 
-static int ci_dpm_switch_power_profile(struct amdgpu_device *adev,
+static int ci_dpm_switch_power_profile(void *handle,
 		enum amd_pp_profile_type type)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct amd_pp_profile request = {0};
 
@@ -6944,11 +6973,12 @@ static int ci_dpm_switch_power_profile(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int ci_dpm_read_sensor(struct amdgpu_device *adev, int idx,
+static int ci_dpm_read_sensor(void *handle, int idx,
 			      void *value, int *size)
 {
 	u32 activity_percent = 50;
 	int ret;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	/* size must be at least 4 bytes for all sensors */
 	if (*size < 4)
@@ -7003,7 +7033,7 @@ const struct amd_ip_funcs ci_dpm_ip_funcs = {
 	.set_powergating_state = ci_dpm_set_powergating_state,
 };
 
-static const struct amdgpu_dpm_funcs ci_dpm_funcs = {
+const struct amd_pm_funcs ci_dpm_funcs = {
 	.get_temperature = &ci_dpm_get_temp,
 	.pre_set_power_state = &ci_dpm_pre_set_power_state,
 	.set_power_state = &ci_dpm_set_power_state,
@@ -7035,12 +7065,6 @@ static const struct amdgpu_dpm_funcs ci_dpm_funcs = {
 	.read_sensor = ci_dpm_read_sensor,
 };
 
-static void ci_dpm_set_dpm_funcs(struct amdgpu_device *adev)
-{
-	if (adev->pm.funcs == NULL)
-		adev->pm.funcs = &ci_dpm_funcs;
-}
-
 static const struct amdgpu_irq_src_funcs ci_dpm_irq_funcs = {
 	.set = ci_dpm_set_interrupt_state,
 	.process = ci_dpm_process_interrupt,

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

@@ -26,5 +26,6 @@
 
 extern const struct amd_ip_funcs ci_dpm_ip_funcs;
 extern const struct amd_ip_funcs kv_dpm_ip_funcs;
-
+extern const struct amd_pm_funcs ci_dpm_funcs;
+extern const struct amd_pm_funcs kv_dpm_funcs;
 #endif

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

@@ -228,6 +228,19 @@ static u32 cik_ih_get_wptr(struct amdgpu_device *adev)
  * [127:96] - reserved
  */
 
+/**
+ * cik_ih_prescreen_iv - prescreen an interrupt vector
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Returns true if the interrupt vector should be further processed.
+ */
+static bool cik_ih_prescreen_iv(struct amdgpu_device *adev)
+{
+	/* Process all interrupts */
+	return true;
+}
+
  /**
  * cik_ih_decode_iv - decode an interrupt vector
  *
@@ -433,6 +446,7 @@ static const struct amd_ip_funcs cik_ih_ip_funcs = {
 
 static const struct amdgpu_ih_funcs cik_ih_funcs = {
 	.get_wptr = cik_ih_get_wptr,
+	.prescreen_iv = cik_ih_prescreen_iv,
 	.decode_iv = cik_ih_decode_iv,
 	.set_rptr = cik_ih_set_rptr
 };

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

@@ -1387,8 +1387,13 @@ static void cik_sdma_set_buffer_funcs(struct amdgpu_device *adev)
 }
 
 static const struct amdgpu_vm_pte_funcs cik_sdma_vm_pte_funcs = {
+	.copy_pte_num_dw = 7,
 	.copy_pte = cik_sdma_vm_copy_pte,
+
 	.write_pte = cik_sdma_vm_write_pte,
+
+	.set_max_nums_pte_pde = 0x1fffff >> 3,
+	.set_pte_pde_num_dw = 10,
 	.set_pte_pde = cik_sdma_vm_set_pte_pde,
 };
 

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

@@ -207,6 +207,19 @@ static u32 cz_ih_get_wptr(struct amdgpu_device *adev)
 	return (wptr & adev->irq.ih.ptr_mask);
 }
 
+/**
+ * cz_ih_prescreen_iv - prescreen an interrupt vector
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Returns true if the interrupt vector should be further processed.
+ */
+static bool cz_ih_prescreen_iv(struct amdgpu_device *adev)
+{
+	/* Process all interrupts */
+	return true;
+}
+
 /**
  * cz_ih_decode_iv - decode an interrupt vector
  *
@@ -414,6 +427,7 @@ static const struct amd_ip_funcs cz_ih_ip_funcs = {
 
 static const struct amdgpu_ih_funcs cz_ih_funcs = {
 	.get_wptr = cz_ih_get_wptr,
+	.prescreen_iv = cz_ih_prescreen_iv,
 	.decode_iv = cz_ih_decode_iv,
 	.set_rptr = cz_ih_set_rptr
 };

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

@@ -125,24 +125,39 @@ MODULE_FIRMWARE("amdgpu/fiji_mec2.bin");
 MODULE_FIRMWARE("amdgpu/fiji_rlc.bin");
 
 MODULE_FIRMWARE("amdgpu/polaris11_ce.bin");
+MODULE_FIRMWARE("amdgpu/polaris11_ce_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris11_pfp.bin");
+MODULE_FIRMWARE("amdgpu/polaris11_pfp_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris11_me.bin");
+MODULE_FIRMWARE("amdgpu/polaris11_me_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris11_mec.bin");
+MODULE_FIRMWARE("amdgpu/polaris11_mec_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris11_mec2.bin");
+MODULE_FIRMWARE("amdgpu/polaris11_mec2_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris11_rlc.bin");
 
 MODULE_FIRMWARE("amdgpu/polaris10_ce.bin");
+MODULE_FIRMWARE("amdgpu/polaris10_ce_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris10_pfp.bin");
+MODULE_FIRMWARE("amdgpu/polaris10_pfp_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris10_me.bin");
+MODULE_FIRMWARE("amdgpu/polaris10_me_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris10_mec.bin");
+MODULE_FIRMWARE("amdgpu/polaris10_mec_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris10_mec2.bin");
+MODULE_FIRMWARE("amdgpu/polaris10_mec2_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris10_rlc.bin");
 
 MODULE_FIRMWARE("amdgpu/polaris12_ce.bin");
+MODULE_FIRMWARE("amdgpu/polaris12_ce_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris12_pfp.bin");
+MODULE_FIRMWARE("amdgpu/polaris12_pfp_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris12_me.bin");
+MODULE_FIRMWARE("amdgpu/polaris12_me_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris12_mec.bin");
+MODULE_FIRMWARE("amdgpu/polaris12_mec_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris12_mec2.bin");
+MODULE_FIRMWARE("amdgpu/polaris12_mec2_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris12_rlc.bin");
 
 static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] =
@@ -918,8 +933,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 		BUG();
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name);
-	err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev);
+	if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp_2.bin", chip_name);
+		err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev);
+		if (err == -ENOENT) {
+			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name);
+			err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev);
+		}
+	} else {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name);
+		err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev);
+	}
 	if (err)
 		goto out;
 	err = amdgpu_ucode_validate(adev->gfx.pfp_fw);
@@ -929,8 +953,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 	adev->gfx.pfp_fw_version = le32_to_cpu(cp_hdr->header.ucode_version);
 	adev->gfx.pfp_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
-	err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev);
+	if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me_2.bin", chip_name);
+		err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev);
+		if (err == -ENOENT) {
+			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
+			err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev);
+		}
+	} else {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
+		err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev);
+	}
 	if (err)
 		goto out;
 	err = amdgpu_ucode_validate(adev->gfx.me_fw);
@@ -941,8 +974,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 
 	adev->gfx.me_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
-	err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev);
+	if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce_2.bin", chip_name);
+		err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev);
+		if (err == -ENOENT) {
+			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
+			err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev);
+		}
+	} else {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
+		err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev);
+	}
 	if (err)
 		goto out;
 	err = amdgpu_ucode_validate(adev->gfx.ce_fw);
@@ -1012,8 +1054,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 	for (i = 0 ; i < (rlc_hdr->reg_list_size_bytes >> 2); i++)
 		adev->gfx.rlc.register_restore[i] = le32_to_cpu(tmp[i]);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
-	err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev);
+	if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec_2.bin", chip_name);
+		err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev);
+		if (err == -ENOENT) {
+			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
+			err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev);
+		}
+	} else {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
+		err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev);
+	}
 	if (err)
 		goto out;
 	err = amdgpu_ucode_validate(adev->gfx.mec_fw);
@@ -1025,8 +1076,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 
 	if ((adev->asic_type != CHIP_STONEY) &&
 	    (adev->asic_type != CHIP_TOPAZ)) {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name);
-		err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev);
+		if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
+			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2_2.bin", chip_name);
+			err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev);
+			if (err == -ENOENT) {
+				snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name);
+				err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev);
+			}
+		} else {
+			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name);
+			err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev);
+		}
 		if (!err) {
 			err = amdgpu_ucode_validate(adev->gfx.mec2_fw);
 			if (err)
@@ -2053,6 +2113,7 @@ static int gfx_v8_0_sw_fini(void *handle)
 	amdgpu_gfx_compute_mqd_sw_fini(adev);
 	amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring, &adev->gfx.kiq.irq);
 	amdgpu_gfx_kiq_fini(adev);
+	amdgpu_bo_free_kernel(&adev->virt.csa_obj, &adev->virt.csa_vmid0_addr, NULL);
 
 	gfx_v8_0_mec_fini(adev);
 	gfx_v8_0_rlc_fini(adev);
@@ -4577,12 +4638,10 @@ static int gfx_v8_0_mqd_init(struct amdgpu_ring *ring)
 	mqd->compute_static_thread_mgmt_se2 = 0xffffffff;
 	mqd->compute_static_thread_mgmt_se3 = 0xffffffff;
 	mqd->compute_misc_reserved = 0x00000003;
-	if (!(adev->flags & AMD_IS_APU)) {
-		mqd->dynamic_cu_mask_addr_lo = lower_32_bits(ring->mqd_gpu_addr
-					     + offsetof(struct vi_mqd_allocation, dynamic_cu_mask));
-		mqd->dynamic_cu_mask_addr_hi = upper_32_bits(ring->mqd_gpu_addr
-					     + offsetof(struct vi_mqd_allocation, dynamic_cu_mask));
-	}
+	mqd->dynamic_cu_mask_addr_lo = lower_32_bits(ring->mqd_gpu_addr
+						     + offsetof(struct vi_mqd_allocation, dynamic_cu_mask));
+	mqd->dynamic_cu_mask_addr_hi = upper_32_bits(ring->mqd_gpu_addr
+						     + offsetof(struct vi_mqd_allocation, dynamic_cu_mask));
 	eop_base_addr = ring->eop_gpu_addr >> 8;
 	mqd->cp_hqd_eop_base_addr_lo = eop_base_addr;
 	mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr);
@@ -4753,7 +4812,7 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
 
 	gfx_v8_0_kiq_setting(ring);
 
-	if (adev->gfx.in_reset) { /* for GPU_RESET case */
+	if (adev->in_sriov_reset) { /* for GPU_RESET case */
 		/* reset MQD to a clean status */
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
 			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct vi_mqd_allocation));
@@ -4790,7 +4849,7 @@ static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring)
 	struct vi_mqd *mqd = ring->mqd_ptr;
 	int mqd_idx = ring - &adev->gfx.compute_ring[0];
 
-	if (!adev->gfx.in_reset && !adev->gfx.in_suspend) {
+	if (!adev->in_sriov_reset && !adev->gfx.in_suspend) {
 		memset((void *)mqd, 0, sizeof(struct vi_mqd_allocation));
 		((struct vi_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
 		((struct vi_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
@@ -4802,7 +4861,7 @@ static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring)
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
 			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct vi_mqd_allocation));
-	} else if (adev->gfx.in_reset) { /* for GPU_RESET case */
+	} else if (adev->in_sriov_reset) { /* for GPU_RESET case */
 		/* reset MQD to a clean status */
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
 			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct vi_mqd_allocation));
@@ -4975,12 +5034,69 @@ static int gfx_v8_0_hw_init(void *handle)
 	return r;
 }
 
+static int gfx_v8_0_kcq_disable(struct amdgpu_ring *kiq_ring,struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = kiq_ring->adev;
+	uint32_t scratch, tmp = 0;
+	int r, i;
+
+	r = amdgpu_gfx_scratch_get(adev, &scratch);
+	if (r) {
+		DRM_ERROR("Failed to get scratch reg (%d).\n", r);
+		return r;
+	}
+	WREG32(scratch, 0xCAFEDEAD);
+
+	r = amdgpu_ring_alloc(kiq_ring, 10);
+	if (r) {
+		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
+		amdgpu_gfx_scratch_free(adev, scratch);
+		return r;
+	}
+
+	/* unmap queues */
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4));
+	amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
+						PACKET3_UNMAP_QUEUES_ACTION(1) | /* RESET_QUEUES */
+						PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) |
+						PACKET3_UNMAP_QUEUES_ENGINE_SEL(0) |
+						PACKET3_UNMAP_QUEUES_NUM_QUEUES(1));
+	amdgpu_ring_write(kiq_ring, PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index));
+	amdgpu_ring_write(kiq_ring, 0);
+	amdgpu_ring_write(kiq_ring, 0);
+	amdgpu_ring_write(kiq_ring, 0);
+	/* write to scratch for completion */
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
+	amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START));
+	amdgpu_ring_write(kiq_ring, 0xDEADBEEF);
+	amdgpu_ring_commit(kiq_ring);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32(scratch);
+		if (tmp == 0xDEADBEEF)
+			break;
+		DRM_UDELAY(1);
+	}
+	if (i >= adev->usec_timeout) {
+		DRM_ERROR("KCQ disabled failed (scratch(0x%04X)=0x%08X)\n", scratch, tmp);
+		r = -EINVAL;
+	}
+	amdgpu_gfx_scratch_free(adev, scratch);
+	return r;
+}
+
 static int gfx_v8_0_hw_fini(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	int i;
 
 	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
+
+	/* disable KCQ to avoid CPC touch memory not valid anymore */
+	for (i = 0; i < adev->gfx.num_compute_rings; i++)
+		gfx_v8_0_kcq_disable(&adev->gfx.kiq.ring, &adev->gfx.compute_ring[i]);
+
 	if (amdgpu_sriov_vf(adev)) {
 		pr_debug("For SRIOV client, shouldn't do anything.\n");
 		return 0;

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

@@ -66,38 +66,70 @@ MODULE_FIRMWARE("amdgpu/raven_rlc.bin");
 
 static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] =
 {
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID0), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID0)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID1), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID1)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID2), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID2)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID3), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID3)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID4), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID4)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID5), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID5)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID6), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID6)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID7), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID7)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID8), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID8)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID9), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID9)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID10), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID10)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_SIZE),
-	       	SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID11), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID11)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID12), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID12)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID13), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID13)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID14), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID14)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID15), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID15)}
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID0),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID0) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID1),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID1) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID2),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID2) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID3),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID3) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID4),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID4) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID5),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID5) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID6),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID6) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID7),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID7) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID8),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID8) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID9),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID9) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID10),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID10) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID11),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID11) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID12),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID12)},
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID13),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID13) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID14),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID14) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID15),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID15) }
 };
 
 static const u32 golden_settings_gc_9_0[] =
@@ -352,6 +384,25 @@ err1:
         return r;
 }
 
+
+static void gfx_v9_0_free_microcode(struct amdgpu_device *adev)
+{
+	release_firmware(adev->gfx.pfp_fw);
+	adev->gfx.pfp_fw = NULL;
+	release_firmware(adev->gfx.me_fw);
+	adev->gfx.me_fw = NULL;
+	release_firmware(adev->gfx.ce_fw);
+	adev->gfx.ce_fw = NULL;
+	release_firmware(adev->gfx.rlc_fw);
+	adev->gfx.rlc_fw = NULL;
+	release_firmware(adev->gfx.mec_fw);
+	adev->gfx.mec_fw = NULL;
+	release_firmware(adev->gfx.mec2_fw);
+	adev->gfx.mec2_fw = NULL;
+
+	kfree(adev->gfx.rlc.register_list_format);
+}
+
 static int gfx_v9_0_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
@@ -1120,30 +1171,22 @@ static int gfx_v9_0_ngg_en(struct amdgpu_device *adev)
 {
 	struct amdgpu_ring *ring = &adev->gfx.gfx_ring[0];
 	int r;
-	u32 data;
-	u32 size;
-	u32 base;
+	u32 data, base;
 
 	if (!amdgpu_ngg)
 		return 0;
 
 	/* Program buffer size */
-	data = 0;
-	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[NGG_POS].size / 256;
-	data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, POS_BUF_SIZE, size);
-
+	data = REG_SET_FIELD(0, WD_BUF_RESOURCE_1, INDEX_BUF_SIZE,
+			     adev->gfx.ngg.buf[NGG_PRIM].size >> 8);
+	data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, POS_BUF_SIZE,
+			     adev->gfx.ngg.buf[NGG_POS].size >> 8);
 	WREG32_SOC15(GC, 0, mmWD_BUF_RESOURCE_1, data);
 
-	data = 0;
-	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[NGG_PARAM].size / 1024;
-	data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, PARAM_BUF_SIZE, size);
-
+	data = REG_SET_FIELD(0, WD_BUF_RESOURCE_2, CNTL_SB_BUF_SIZE,
+			     adev->gfx.ngg.buf[NGG_CNTL].size >> 8);
+	data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, PARAM_BUF_SIZE,
+			     adev->gfx.ngg.buf[NGG_PARAM].size >> 10);
 	WREG32_SOC15(GC, 0, mmWD_BUF_RESOURCE_2, data);
 
 	/* Program buffer base address */
@@ -1306,7 +1349,10 @@ static int gfx_v9_0_sw_init(void *handle)
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
 		ring = &adev->gfx.gfx_ring[i];
 		ring->ring_obj = NULL;
-		sprintf(ring->name, "gfx");
+		if (!i)
+			sprintf(ring->name, "gfx");
+		else
+			sprintf(ring->name, "gfx_%d", i);
 		ring->use_doorbell = true;
 		ring->doorbell_index = AMDGPU_DOORBELL64_GFX_RING0 << 1;
 		r = amdgpu_ring_init(adev, ring, 1024,
@@ -1346,7 +1392,7 @@ static int gfx_v9_0_sw_init(void *handle)
 		return r;
 
 	/* create MQD for all compute queues as wel as KIQ for SRIOV case */
-	r = amdgpu_gfx_compute_mqd_sw_init(adev, sizeof(struct v9_mqd));
+	r = amdgpu_gfx_compute_mqd_sw_init(adev, sizeof(struct v9_mqd_allocation));
 	if (r)
 		return r;
 
@@ -1398,9 +1444,11 @@ static int gfx_v9_0_sw_fini(void *handle)
 	amdgpu_gfx_compute_mqd_sw_fini(adev);
 	amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring, &adev->gfx.kiq.irq);
 	amdgpu_gfx_kiq_fini(adev);
+	amdgpu_bo_free_kernel(&adev->virt.csa_obj, &adev->virt.csa_vmid0_addr, NULL);
 
 	gfx_v9_0_mec_fini(adev);
 	gfx_v9_0_ngg_fini(adev);
+	gfx_v9_0_free_microcode(adev);
 
 	return 0;
 }
@@ -1740,11 +1788,7 @@ static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev)
 
 static void gfx_v9_0_enable_save_restore_machine(struct amdgpu_device *adev)
 {
-	u32 tmp = 0;
-
-	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_CNTL));
-	tmp |= RLC_SRM_CNTL__SRM_ENABLE_MASK;
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_CNTL), tmp);
+	WREG32_FIELD15(GC, 0, RLC_SRM_CNTL, SRM_ENABLE, 1);
 }
 
 static void pwr_10_0_gfxip_control_over_cgpg(struct amdgpu_device *adev,
@@ -1822,16 +1866,11 @@ static void gfx_v9_0_enable_sck_slow_down_on_power_up(struct amdgpu_device *adev
 	uint32_t default_data = 0;
 
 	default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL));
-
-	if (enable == true) {
-		data |= RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PU_ENABLE_MASK;
-		if (default_data != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
-	} else {
-		data &= ~RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PU_ENABLE_MASK;
-		if(default_data != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
-	}
+	data = REG_SET_FIELD(data, RLC_PG_CNTL,
+			     SMU_CLK_SLOWDOWN_ON_PU_ENABLE,
+			     enable ? 1 : 0);
+	if (default_data != data)
+		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
 }
 
 static void gfx_v9_0_enable_sck_slow_down_on_power_down(struct amdgpu_device *adev,
@@ -1841,16 +1880,11 @@ static void gfx_v9_0_enable_sck_slow_down_on_power_down(struct amdgpu_device *ad
 	uint32_t default_data = 0;
 
 	default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL));
-
-	if (enable == true) {
-		data |= RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PD_ENABLE_MASK;
-		if(default_data != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
-	} else {
-		data &= ~RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PD_ENABLE_MASK;
-		if(default_data != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
-	}
+	data = REG_SET_FIELD(data, RLC_PG_CNTL,
+			     SMU_CLK_SLOWDOWN_ON_PD_ENABLE,
+			     enable ? 1 : 0);
+	if(default_data != data)
+		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
 }
 
 static void gfx_v9_0_enable_cp_power_gating(struct amdgpu_device *adev,
@@ -1860,16 +1894,11 @@ static void gfx_v9_0_enable_cp_power_gating(struct amdgpu_device *adev,
 	uint32_t default_data = 0;
 
 	default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL));
-
-	if (enable == true) {
-		data &= ~RLC_PG_CNTL__CP_PG_DISABLE_MASK;
-		if(default_data != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
-	} else {
-		data |= RLC_PG_CNTL__CP_PG_DISABLE_MASK;
-		if(default_data != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
-	}
+	data = REG_SET_FIELD(data, RLC_PG_CNTL,
+			     CP_PG_DISABLE,
+			     enable ? 0 : 1);
+	if(default_data != data)
+		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
 }
 
 static void gfx_v9_0_enable_gfx_cg_power_gating(struct amdgpu_device *adev,
@@ -1878,10 +1907,9 @@ static void gfx_v9_0_enable_gfx_cg_power_gating(struct amdgpu_device *adev,
 	uint32_t data, default_data;
 
 	default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL));
-	if (enable == true)
-		data |= RLC_PG_CNTL__GFX_POWER_GATING_ENABLE_MASK;
-	else
-		data &= ~RLC_PG_CNTL__GFX_POWER_GATING_ENABLE_MASK;
+	data = REG_SET_FIELD(data, RLC_PG_CNTL,
+			     GFX_POWER_GATING_ENABLE,
+			     enable ? 1 : 0);
 	if(default_data != data)
 		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
 }
@@ -1892,10 +1920,9 @@ static void gfx_v9_0_enable_gfx_pipeline_powergating(struct amdgpu_device *adev,
 	uint32_t data, default_data;
 
 	default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL));
-	if (enable == true)
-		data |= RLC_PG_CNTL__GFX_PIPELINE_PG_ENABLE_MASK;
-	else
-		data &= ~RLC_PG_CNTL__GFX_PIPELINE_PG_ENABLE_MASK;
+	data = REG_SET_FIELD(data, RLC_PG_CNTL,
+			     GFX_PIPELINE_PG_ENABLE,
+			     enable ? 1 : 0);
 	if(default_data != data)
 		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
 
@@ -1910,10 +1937,9 @@ static void gfx_v9_0_enable_gfx_static_mg_power_gating(struct amdgpu_device *ade
 	uint32_t data, default_data;
 
 	default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL));
-	if (enable == true)
-		data |= RLC_PG_CNTL__STATIC_PER_CU_PG_ENABLE_MASK;
-	else
-		data &= ~RLC_PG_CNTL__STATIC_PER_CU_PG_ENABLE_MASK;
+	data = REG_SET_FIELD(data, RLC_PG_CNTL,
+			     STATIC_PER_CU_PG_ENABLE,
+			     enable ? 1 : 0);
 	if(default_data != data)
 		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
 }
@@ -1924,10 +1950,9 @@ static void gfx_v9_0_enable_gfx_dynamic_mg_power_gating(struct amdgpu_device *ad
 	uint32_t data, default_data;
 
 	default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL));
-	if (enable == true)
-		data |= RLC_PG_CNTL__DYN_PER_CU_PG_ENABLE_MASK;
-	else
-		data &= ~RLC_PG_CNTL__DYN_PER_CU_PG_ENABLE_MASK;
+	data = REG_SET_FIELD(data, RLC_PG_CNTL,
+			     DYN_PER_CU_PG_ENABLE,
+			     enable ? 1 : 0);
 	if(default_data != data)
 		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
 }
@@ -1967,13 +1992,8 @@ static void gfx_v9_0_init_pg(struct amdgpu_device *adev)
 
 void gfx_v9_0_rlc_stop(struct amdgpu_device *adev)
 {
-	u32 tmp = RREG32_SOC15(GC, 0, mmRLC_CNTL);
-
-	tmp = REG_SET_FIELD(tmp, RLC_CNTL, RLC_ENABLE_F32, 0);
-	WREG32_SOC15(GC, 0, mmRLC_CNTL, tmp);
-
+	WREG32_FIELD15(GC, 0, RLC_CNTL, RLC_ENABLE_F32, 0);
 	gfx_v9_0_enable_gui_idle_interrupt(adev, false);
-
 	gfx_v9_0_wait_for_rlc_serdes(adev);
 }
 
@@ -2045,8 +2065,10 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev)
 {
 	int r;
 
-	if (amdgpu_sriov_vf(adev))
+	if (amdgpu_sriov_vf(adev)) {
+		gfx_v9_0_init_csb(adev);
 		return 0;
+	}
 
 	gfx_v9_0_rlc_stop(adev);
 
@@ -2463,6 +2485,13 @@ static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring)
 	mqd->compute_static_thread_mgmt_se3 = 0xffffffff;
 	mqd->compute_misc_reserved = 0x00000003;
 
+	mqd->dynamic_cu_mask_addr_lo =
+		lower_32_bits(ring->mqd_gpu_addr
+			      + offsetof(struct v9_mqd_allocation, dynamic_cu_mask));
+	mqd->dynamic_cu_mask_addr_hi =
+		upper_32_bits(ring->mqd_gpu_addr
+			      + offsetof(struct v9_mqd_allocation, dynamic_cu_mask));
+
 	eop_base_addr = ring->eop_gpu_addr >> 8;
 	mqd->cp_hqd_eop_base_addr_lo = eop_base_addr;
 	mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr);
@@ -2486,10 +2515,10 @@ static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring)
 				    DOORBELL_SOURCE, 0);
 		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
 				    DOORBELL_HIT, 0);
-	}
-	else
+	} else {
 		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
 					 DOORBELL_EN, 0);
+	}
 
 	mqd->cp_hqd_pq_doorbell_control = tmp;
 
@@ -2692,10 +2721,10 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring)
 
 	gfx_v9_0_kiq_setting(ring);
 
-	if (adev->gfx.in_reset) { /* for GPU_RESET case */
+	if (adev->in_sriov_reset) { /* for GPU_RESET case */
 		/* reset MQD to a clean status */
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
+			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation));
 
 		/* reset ring buffer */
 		ring->wptr = 0;
@@ -2707,7 +2736,9 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring)
 		soc15_grbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 	} else {
-		memset((void *)mqd, 0, sizeof(*mqd));
+		memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation));
+		((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
+		((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
 		mutex_lock(&adev->srbm_mutex);
 		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
 		gfx_v9_0_mqd_init(ring);
@@ -2716,7 +2747,7 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring)
 		mutex_unlock(&adev->srbm_mutex);
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation));
 	}
 
 	return 0;
@@ -2728,8 +2759,10 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring)
 	struct v9_mqd *mqd = ring->mqd_ptr;
 	int mqd_idx = ring - &adev->gfx.compute_ring[0];
 
-	if (!adev->gfx.in_reset && !adev->gfx.in_suspend) {
-		memset((void *)mqd, 0, sizeof(*mqd));
+	if (!adev->in_sriov_reset && !adev->gfx.in_suspend) {
+		memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation));
+		((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
+		((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
 		mutex_lock(&adev->srbm_mutex);
 		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
 		gfx_v9_0_mqd_init(ring);
@@ -2737,11 +2770,11 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring)
 		mutex_unlock(&adev->srbm_mutex);
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
-	} else if (adev->gfx.in_reset) { /* for GPU_RESET case */
+			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation));
+	} else if (adev->in_sriov_reset) { /* for GPU_RESET case */
 		/* reset MQD to a clean status */
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
+			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation));
 
 		/* reset ring buffer */
 		ring->wptr = 0;
@@ -2882,12 +2915,70 @@ static int gfx_v9_0_hw_init(void *handle)
 	return r;
 }
 
+static int gfx_v9_0_kcq_disable(struct amdgpu_ring *kiq_ring,struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = kiq_ring->adev;
+	uint32_t scratch, tmp = 0;
+	int r, i;
+
+	r = amdgpu_gfx_scratch_get(adev, &scratch);
+	if (r) {
+		DRM_ERROR("Failed to get scratch reg (%d).\n", r);
+		return r;
+	}
+	WREG32(scratch, 0xCAFEDEAD);
+
+	r = amdgpu_ring_alloc(kiq_ring, 10);
+	if (r) {
+		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
+		amdgpu_gfx_scratch_free(adev, scratch);
+		return r;
+	}
+
+	/* unmap queues */
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4));
+	amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
+						PACKET3_UNMAP_QUEUES_ACTION(1) | /* RESET_QUEUES */
+						PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) |
+						PACKET3_UNMAP_QUEUES_ENGINE_SEL(0) |
+						PACKET3_UNMAP_QUEUES_NUM_QUEUES(1));
+	amdgpu_ring_write(kiq_ring, PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index));
+	amdgpu_ring_write(kiq_ring, 0);
+	amdgpu_ring_write(kiq_ring, 0);
+	amdgpu_ring_write(kiq_ring, 0);
+	/* write to scratch for completion */
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
+	amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START));
+	amdgpu_ring_write(kiq_ring, 0xDEADBEEF);
+	amdgpu_ring_commit(kiq_ring);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32(scratch);
+		if (tmp == 0xDEADBEEF)
+			break;
+		DRM_UDELAY(1);
+	}
+	if (i >= adev->usec_timeout) {
+		DRM_ERROR("KCQ disabled failed (scratch(0x%04X)=0x%08X)\n", scratch, tmp);
+		r = -EINVAL;
+	}
+	amdgpu_gfx_scratch_free(adev, scratch);
+	return r;
+}
+
+
 static int gfx_v9_0_hw_fini(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	int i;
 
 	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
+
+	/* disable KCQ to avoid CPC touch memory not valid anymore */
+	for (i = 0; i < adev->gfx.num_compute_rings; i++)
+		gfx_v9_0_kcq_disable(&adev->gfx.kiq.ring, &adev->gfx.compute_ring[i]);
+
 	if (amdgpu_sriov_vf(adev)) {
 		pr_debug("For SRIOV client, shouldn't do anything.\n");
 		return 0;
@@ -2930,15 +3021,10 @@ static bool gfx_v9_0_is_idle(void *handle)
 static int gfx_v9_0_wait_for_idle(void *handle)
 {
 	unsigned i;
-	u32 tmp;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		/* read MC_STATUS */
-		tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS) &
-			GRBM_STATUS__GUI_ACTIVE_MASK;
-
-		if (!REG_GET_FIELD(tmp, GRBM_STATUS, GUI_ACTIVE))
+		if (gfx_v9_0_is_idle(handle))
 			return 0;
 		udelay(1);
 	}
@@ -3499,7 +3585,9 @@ static void gfx_v9_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 	u32 ref_and_mask, reg_mem_engine;
 	struct nbio_hdp_flush_reg *nbio_hf_reg;
 
-	if (ring->adev->asic_type == CHIP_VEGA10)
+	if (ring->adev->flags & AMD_IS_APU)
+		nbio_hf_reg = &nbio_v7_0_hdp_flush_reg;
+	else
 		nbio_hf_reg = &nbio_v6_1_hdp_flush_reg;
 
 	if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) {
@@ -3528,7 +3616,7 @@ static void gfx_v9_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 static void gfx_v9_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring)
 {
 	gfx_v9_0_write_data_to_reg(ring, 0, true,
-				   SOC15_REG_OFFSET(HDP, 0, mmHDP_DEBUG0), 1);
+				   SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 1);
 }
 
 static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
@@ -3757,6 +3845,12 @@ static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring)
 	amdgpu_ring_write_multiple(ring, (void *)&de_payload, sizeof(de_payload) >> 2);
 }
 
+static void gfx_v9_0_ring_emit_tmz(struct amdgpu_ring *ring, bool start)
+{
+	amdgpu_ring_write(ring, PACKET3(PACKET3_FRAME_CONTROL, 0));
+	amdgpu_ring_write(ring, FRAME_CMD(start ? 0 : 1)); /* frame_end */
+}
+
 static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags)
 {
 	uint32_t dw2 = 0;
@@ -3764,6 +3858,8 @@ static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags)
 	if (amdgpu_sriov_vf(ring->adev))
 		gfx_v9_0_ring_emit_ce_meta(ring);
 
+	gfx_v9_0_ring_emit_tmz(ring, true);
+
 	dw2 |= 0x80000000; /* set load_enable otherwise this package is just NOPs */
 	if (flags & AMDGPU_HAVE_CTX_SWITCH) {
 		/* set load_global_config & load_global_uconfig */
@@ -3814,12 +3910,6 @@ static void gfx_v9_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne
 		ring->ring[offset] = (ring->ring_size>>2) - offset + cur;
 }
 
-static void gfx_v9_0_ring_emit_tmz(struct amdgpu_ring *ring, bool start)
-{
-	amdgpu_ring_write(ring, PACKET3(PACKET3_FRAME_CONTROL, 0));
-	amdgpu_ring_write(ring, FRAME_CMD(start ? 0 : 1)); /* frame_end */
-}
-
 static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)
 {
 	struct amdgpu_device *adev = ring->adev;

drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c

@@ -319,6 +319,12 @@ void gfxhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev,
 			WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
 	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
 			EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	if (!value) {
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_NO_RETRY_FAULT, 1);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_RETRY_FAULT, 1);
+    }
 	WREG32_SOC15(GC, 0, mmVM_L2_PROTECTION_FAULT_CNTL, tmp);
 }
 

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

@@ -831,7 +831,7 @@ static int gmc_v6_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	amdgpu_vm_adjust_size(adev, 64, 4);
+	amdgpu_vm_adjust_size(adev, 64, 9);
 	adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18;
 
 	adev->mc.mc_mask = 0xffffffffffULL;
@@ -901,6 +901,8 @@ static int gmc_v6_0_sw_fini(void *handle)
 	gmc_v6_0_gart_fini(adev);
 	amdgpu_gem_force_release(adev);
 	amdgpu_bo_fini(adev);
+	release_firmware(adev->mc.fw);
+	adev->mc.fw = NULL;
 
 	return 0;
 }

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

@@ -970,7 +970,7 @@ static int gmc_v7_0_sw_init(void *handle)
 	 * Currently set to 4GB ((1 << 20) 4k pages).
 	 * Max GPUVM size for cayman and SI is 40 bits.
 	 */
-	amdgpu_vm_adjust_size(adev, 64, 4);
+	amdgpu_vm_adjust_size(adev, 64, 9);
 	adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18;
 
 	/* Set the internal MC address mask
@@ -1050,6 +1050,8 @@ static int gmc_v7_0_sw_fini(void *handle)
 	gmc_v7_0_gart_fini(adev);
 	amdgpu_gem_force_release(adev);
 	amdgpu_bo_fini(adev);
+	release_firmware(adev->mc.fw);
+	adev->mc.fw = NULL;
 
 	return 0;
 }

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

@@ -1067,7 +1067,7 @@ static int gmc_v8_0_sw_init(void *handle)
 	 * Currently set to 4GB ((1 << 20) 4k pages).
 	 * Max GPUVM size for cayman and SI is 40 bits.
 	 */
-	amdgpu_vm_adjust_size(adev, 64, 4);
+	amdgpu_vm_adjust_size(adev, 64, 9);
 	adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18;
 
 	/* Set the internal MC address mask
@@ -1147,6 +1147,8 @@ static int gmc_v8_0_sw_fini(void *handle)
 	gmc_v8_0_gart_fini(adev);
 	amdgpu_gem_force_release(adev);
 	amdgpu_bo_fini(adev);
+	release_firmware(adev->mc.fw);
+	adev->mc.fw = NULL;
 
 	return 0;
 }

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

@@ -32,6 +32,8 @@
 #include "vega10/DC/dce_12_0_offset.h"
 #include "vega10/DC/dce_12_0_sh_mask.h"
 #include "vega10/vega10_enum.h"
+#include "vega10/MMHUB/mmhub_1_0_offset.h"
+#include "vega10/ATHUB/athub_1_0_offset.h"
 
 #include "soc15_common.h"
 
@@ -71,13 +73,25 @@ static const u32 golden_settings_vega10_hdp[] =
 	0xf6e, 0x0fffffff, 0x00000000,
 };
 
+static const u32 golden_settings_mmhub_1_0_0[] =
+{
+	SOC15_REG_OFFSET(MMHUB, 0, mmDAGB1_WRCLI2), 0x00000007, 0xfe5fe0fa,
+	SOC15_REG_OFFSET(MMHUB, 0, mmMMEA1_DRAM_WR_CLI2GRP_MAP0), 0x00000030, 0x55555565
+};
+
+static const u32 golden_settings_athub_1_0_0[] =
+{
+	SOC15_REG_OFFSET(ATHUB, 0, mmRPB_ARB_CNTL), 0x0000ff00, 0x00000800,
+	SOC15_REG_OFFSET(ATHUB, 0, mmRPB_ARB_CNTL2), 0x00ff00ff, 0x00080008
+};
+
 static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 					struct amdgpu_irq_src *src,
 					unsigned type,
 					enum amdgpu_interrupt_state state)
 {
 	struct amdgpu_vmhub *hub;
-	u32 tmp, reg, bits, i;
+	u32 tmp, reg, bits, i, j;
 
 	bits = VM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
 		VM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
@@ -89,43 +103,26 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
-		/* MM HUB */
-		hub = &adev->vmhub[AMDGPU_MMHUB];
-		for (i = 0; i< 16; i++) {
-			reg = hub->vm_context0_cntl + i;
-			tmp = RREG32(reg);
-			tmp &= ~bits;
-			WREG32(reg, tmp);
-		}
-
-		/* GFX HUB */
-		hub = &adev->vmhub[AMDGPU_GFXHUB];
-		for (i = 0; i < 16; i++) {
-			reg = hub->vm_context0_cntl + i;
-			tmp = RREG32(reg);
-			tmp &= ~bits;
-			WREG32(reg, tmp);
+		for (j = 0; j < AMDGPU_MAX_VMHUBS; j++) {
+			hub = &adev->vmhub[j];
+			for (i = 0; i < 16; i++) {
+				reg = hub->vm_context0_cntl + i;
+				tmp = RREG32(reg);
+				tmp &= ~bits;
+				WREG32(reg, tmp);
+			}
 		}
 		break;
 	case AMDGPU_IRQ_STATE_ENABLE:
-		/* MM HUB */
-		hub = &adev->vmhub[AMDGPU_MMHUB];
-		for (i = 0; i< 16; i++) {
-			reg = hub->vm_context0_cntl + i;
-			tmp = RREG32(reg);
-			tmp |= bits;
-			WREG32(reg, tmp);
-		}
-
-		/* GFX HUB */
-		hub = &adev->vmhub[AMDGPU_GFXHUB];
-		for (i = 0; i < 16; i++) {
-			reg = hub->vm_context0_cntl + i;
-			tmp = RREG32(reg);
-			tmp |= bits;
-			WREG32(reg, tmp);
+		for (j = 0; j < AMDGPU_MAX_VMHUBS; j++) {
+			hub = &adev->vmhub[j];
+			for (i = 0; i < 16; i++) {
+				reg = hub->vm_context0_cntl + i;
+				tmp = RREG32(reg);
+				tmp |= bits;
+				WREG32(reg, tmp);
+			}
 		}
-		break;
 	default:
 		break;
 	}
@@ -682,8 +679,17 @@ static void gmc_v9_0_init_golden_registers(struct amdgpu_device *adev)
 {
 	switch (adev->asic_type) {
 	case CHIP_VEGA10:
+		amdgpu_program_register_sequence(adev,
+						golden_settings_mmhub_1_0_0,
+						(const u32)ARRAY_SIZE(golden_settings_mmhub_1_0_0));
+		amdgpu_program_register_sequence(adev,
+						golden_settings_athub_1_0_0,
+						(const u32)ARRAY_SIZE(golden_settings_athub_1_0_0));
 		break;
 	case CHIP_RAVEN:
+		amdgpu_program_register_sequence(adev,
+						golden_settings_athub_1_0_0,
+						(const u32)ARRAY_SIZE(golden_settings_athub_1_0_0));
 		break;
 	default:
 		break;
@@ -713,12 +719,6 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev)
 	if (r)
 		return r;
 
-	/* After HDP is initialized, flush HDP.*/
-	if (adev->flags & AMD_IS_APU)
-		nbio_v7_0_hdp_flush(adev);
-	else
-		nbio_v6_1_hdp_flush(adev);
-
 	switch (adev->asic_type) {
 	case CHIP_RAVEN:
 		mmhub_v1_0_initialize_power_gating(adev);
@@ -736,13 +736,16 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev)
 	if (r)
 		return r;
 
-	tmp = RREG32_SOC15(HDP, 0, mmHDP_MISC_CNTL);
-	tmp |= HDP_MISC_CNTL__FLUSH_INVALIDATE_CACHE_MASK;
-	WREG32_SOC15(HDP, 0, mmHDP_MISC_CNTL, tmp);
+	WREG32_FIELD15(HDP, 0, HDP_MISC_CNTL, FLUSH_INVALIDATE_CACHE, 1);
 
 	tmp = RREG32_SOC15(HDP, 0, mmHDP_HOST_PATH_CNTL);
 	WREG32_SOC15(HDP, 0, mmHDP_HOST_PATH_CNTL, tmp);
 
+	/* After HDP is initialized, flush HDP.*/
+	if (adev->flags & AMD_IS_APU)
+		nbio_v7_0_hdp_flush(adev);
+	else
+		nbio_v6_1_hdp_flush(adev);
 
 	if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS)
 		value = false;
@@ -751,7 +754,6 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev)
 
 	gfxhub_v1_0_set_fault_enable_default(adev, value);
 	mmhub_v1_0_set_fault_enable_default(adev, value);
-
 	gmc_v9_0_gart_flush_gpu_tlb(adev, 0);
 
 	DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
@@ -770,17 +772,11 @@ static int gmc_v9_0_hw_init(void *handle)
 	gmc_v9_0_init_golden_registers(adev);
 
 	if (adev->mode_info.num_crtc) {
-		u32 tmp;
-
 		/* Lockout access through VGA aperture*/
-		tmp = RREG32_SOC15(DCE, 0, mmVGA_HDP_CONTROL);
-		tmp = REG_SET_FIELD(tmp, VGA_HDP_CONTROL, VGA_MEMORY_DISABLE, 1);
-		WREG32_SOC15(DCE, 0, mmVGA_HDP_CONTROL, tmp);
+		WREG32_FIELD15(DCE, 0, VGA_HDP_CONTROL, VGA_MEMORY_DISABLE, 1);
 
 		/* disable VGA render */
-		tmp = RREG32_SOC15(DCE, 0, mmVGA_RENDER_CONTROL);
-		tmp = REG_SET_FIELD(tmp, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0);
-		WREG32_SOC15(DCE, 0, mmVGA_RENDER_CONTROL, tmp);
+		WREG32_FIELD15(DCE, 0, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0);
 	}
 
 	r = gmc_v9_0_gart_enable(adev);
@@ -822,9 +818,7 @@ static int gmc_v9_0_suspend(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	gmc_v9_0_hw_fini(adev);
-
-	return 0;
+	return gmc_v9_0_hw_fini(adev);
 }
 
 static int gmc_v9_0_resume(void *handle)

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

@@ -207,6 +207,19 @@ static u32 iceland_ih_get_wptr(struct amdgpu_device *adev)
 	return (wptr & adev->irq.ih.ptr_mask);
 }
 
+/**
+ * iceland_ih_prescreen_iv - prescreen an interrupt vector
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Returns true if the interrupt vector should be further processed.
+ */
+static bool iceland_ih_prescreen_iv(struct amdgpu_device *adev)
+{
+	/* Process all interrupts */
+	return true;
+}
+
 /**
  * iceland_ih_decode_iv - decode an interrupt vector
  *
@@ -412,6 +425,7 @@ static const struct amd_ip_funcs iceland_ih_ip_funcs = {
 
 static const struct amdgpu_ih_funcs iceland_ih_funcs = {
 	.get_wptr = iceland_ih_get_wptr,
+	.prescreen_iv = iceland_ih_prescreen_iv,
 	.decode_iv = iceland_ih_decode_iv,
 	.set_rptr = iceland_ih_set_rptr
 };

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

@@ -42,7 +42,6 @@
 #define KV_MINIMUM_ENGINE_CLOCK         800
 #define SMC_RAM_END                     0x40000
 
-static void kv_dpm_set_dpm_funcs(struct amdgpu_device *adev);
 static void kv_dpm_set_irq_funcs(struct amdgpu_device *adev);
 static int kv_enable_nb_dpm(struct amdgpu_device *adev,
 			    bool enable);
@@ -64,7 +63,7 @@ static int kv_set_thermal_temperature_range(struct amdgpu_device *adev,
 					    int min_temp, int max_temp);
 static int kv_init_fps_limits(struct amdgpu_device *adev);
 
-static void kv_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate);
+static void kv_dpm_powergate_uvd(void *handle, bool gate);
 static void kv_dpm_powergate_vce(struct amdgpu_device *adev, bool gate);
 static void kv_dpm_powergate_samu(struct amdgpu_device *adev, bool gate);
 static void kv_dpm_powergate_acp(struct amdgpu_device *adev, bool gate);
@@ -1245,8 +1244,9 @@ static void kv_update_requested_ps(struct amdgpu_device *adev,
 	adev->pm.dpm.requested_ps = &pi->requested_rps;
 }
 
-static void kv_dpm_enable_bapm(struct amdgpu_device *adev, bool enable)
+static void kv_dpm_enable_bapm(void *handle, bool enable)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 	int ret;
 
@@ -1672,8 +1672,9 @@ static int kv_update_acp_dpm(struct amdgpu_device *adev, bool gate)
 	return kv_enable_acp_dpm(adev, !gate);
 }
 
-static void kv_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate)
+static void kv_dpm_powergate_uvd(void *handle, bool gate)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 	int ret;
 
@@ -1868,10 +1869,11 @@ static int kv_enable_nb_dpm(struct amdgpu_device *adev,
 	return ret;
 }
 
-static int kv_dpm_force_performance_level(struct amdgpu_device *adev,
+static int kv_dpm_force_performance_level(void *handle,
 					  enum amd_dpm_forced_level level)
 {
 	int ret;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	if (level == AMD_DPM_FORCED_LEVEL_HIGH) {
 		ret = kv_force_dpm_highest(adev);
@@ -1892,8 +1894,9 @@ static int kv_dpm_force_performance_level(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int kv_dpm_pre_set_power_state(struct amdgpu_device *adev)
+static int kv_dpm_pre_set_power_state(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 	struct amdgpu_ps requested_ps = *adev->pm.dpm.requested_ps;
 	struct amdgpu_ps *new_ps = &requested_ps;
@@ -1907,8 +1910,9 @@ static int kv_dpm_pre_set_power_state(struct amdgpu_device *adev)
 	return 0;
 }
 
-static int kv_dpm_set_power_state(struct amdgpu_device *adev)
+static int kv_dpm_set_power_state(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 	struct amdgpu_ps *new_ps = &pi->requested_rps;
 	struct amdgpu_ps *old_ps = &pi->current_rps;
@@ -1981,8 +1985,9 @@ static int kv_dpm_set_power_state(struct amdgpu_device *adev)
 	return 0;
 }
 
-static void kv_dpm_post_set_power_state(struct amdgpu_device *adev)
+static void kv_dpm_post_set_power_state(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 	struct amdgpu_ps *new_ps = &pi->requested_rps;
 
@@ -2848,9 +2853,10 @@ static int kv_dpm_init(struct amdgpu_device *adev)
 }
 
 static void
-kv_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev,
+kv_dpm_debugfs_print_current_performance_level(void *handle,
 					       struct seq_file *m)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 	u32 current_index =
 		(RREG32_SMC(ixTARGET_AND_CURRENT_PROFILE_INDEX) &
@@ -2875,11 +2881,12 @@ kv_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev,
 }
 
 static void
-kv_dpm_print_power_state(struct amdgpu_device *adev,
-			 struct amdgpu_ps *rps)
+kv_dpm_print_power_state(void *handle, void *request_ps)
 {
 	int i;
+	struct amdgpu_ps *rps = (struct amdgpu_ps *)request_ps;
 	struct kv_ps *ps = kv_get_ps(rps);
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	amdgpu_dpm_print_class_info(rps->class, rps->class2);
 	amdgpu_dpm_print_cap_info(rps->caps);
@@ -2905,13 +2912,14 @@ static void kv_dpm_fini(struct amdgpu_device *adev)
 	amdgpu_free_extended_power_table(adev);
 }
 
-static void kv_dpm_display_configuration_changed(struct amdgpu_device *adev)
+static void kv_dpm_display_configuration_changed(void *handle)
 {
 
 }
 
-static u32 kv_dpm_get_sclk(struct amdgpu_device *adev, bool low)
+static u32 kv_dpm_get_sclk(void *handle, bool low)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 	struct kv_ps *requested_state = kv_get_ps(&pi->requested_rps);
 
@@ -2921,18 +2929,20 @@ static u32 kv_dpm_get_sclk(struct amdgpu_device *adev, bool low)
 		return requested_state->levels[requested_state->num_levels - 1].sclk;
 }
 
-static u32 kv_dpm_get_mclk(struct amdgpu_device *adev, bool low)
+static u32 kv_dpm_get_mclk(void *handle, bool low)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 
 	return pi->sys_info.bootup_uma_clk;
 }
 
 /* get temperature in millidegrees */
-static int kv_dpm_get_temp(struct amdgpu_device *adev)
+static int kv_dpm_get_temp(void *handle)
 {
 	u32 temp;
 	int actual_temp = 0;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	temp = RREG32_SMC(0xC0300E0C);
 
@@ -2950,7 +2960,6 @@ static int kv_dpm_early_init(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	kv_dpm_set_dpm_funcs(adev);
 	kv_dpm_set_irq_funcs(adev);
 
 	return 0;
@@ -3222,14 +3231,17 @@ static inline bool kv_are_power_levels_equal(const struct kv_pl *kv_cpl1,
 		  (kv_cpl1->force_nbp_state == kv_cpl2->force_nbp_state));
 }
 
-static int kv_check_state_equal(struct amdgpu_device *adev,
-				struct amdgpu_ps *cps,
-				struct amdgpu_ps *rps,
+static int kv_check_state_equal(void *handle,
+				void *current_ps,
+				void *request_ps,
 				bool *equal)
 {
 	struct kv_ps *kv_cps;
 	struct kv_ps *kv_rps;
 	int i;
+	struct amdgpu_ps *cps = (struct amdgpu_ps *)current_ps;
+	struct amdgpu_ps *rps = (struct amdgpu_ps *)request_ps;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	if (adev == NULL || cps == NULL || rps == NULL || equal == NULL)
 		return -EINVAL;
@@ -3262,9 +3274,10 @@ static int kv_check_state_equal(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int kv_dpm_read_sensor(struct amdgpu_device *adev, int idx,
+static int kv_dpm_read_sensor(void *handle, int idx,
 			      void *value, int *size)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 	uint32_t sclk;
 	u32 pl_index =
@@ -3312,7 +3325,7 @@ const struct amd_ip_funcs kv_dpm_ip_funcs = {
 	.set_powergating_state = kv_dpm_set_powergating_state,
 };
 
-static const struct amdgpu_dpm_funcs kv_dpm_funcs = {
+const struct amd_pm_funcs kv_dpm_funcs = {
 	.get_temperature = &kv_dpm_get_temp,
 	.pre_set_power_state = &kv_dpm_pre_set_power_state,
 	.set_power_state = &kv_dpm_set_power_state,
@@ -3330,12 +3343,6 @@ static const struct amdgpu_dpm_funcs kv_dpm_funcs = {
 	.read_sensor = &kv_dpm_read_sensor,
 };
 
-static void kv_dpm_set_dpm_funcs(struct amdgpu_device *adev)
-{
-	if (adev->pm.funcs == NULL)
-		adev->pm.funcs = &kv_dpm_funcs;
-}
-
 static const struct amdgpu_irq_src_funcs kv_dpm_irq_funcs = {
 	.set = kv_dpm_set_interrupt_state,
 	.process = kv_dpm_process_interrupt,

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

@@ -273,7 +273,7 @@ static const struct pctl_data pctl0_data[] = {
 	{0x135, 0x12a810},
 	{0x149, 0x7a82c}
 };
-#define PCTL0_DATA_LEN (sizeof(pctl0_data)/sizeof(pctl0_data[0]))
+#define PCTL0_DATA_LEN (ARRAY_SIZE(pctl0_data))
 
 #define PCTL0_RENG_EXEC_END_PTR 0x151
 #define PCTL0_STCTRL_REG_SAVE_RANGE0_BASE  0xa640
@@ -309,7 +309,7 @@ static const struct pctl_data pctl1_data[] = {
 	{0x1f0, 0x5000a7f6},
 	{0x1f1, 0x5000a7e4}
 };
-#define PCTL1_DATA_LEN (sizeof(pctl1_data)/sizeof(pctl1_data[0]))
+#define PCTL1_DATA_LEN (ARRAY_SIZE(pctl1_data))
 
 #define PCTL1_RENG_EXEC_END_PTR 0x1f1
 #define PCTL1_STCTRL_REG_SAVE_RANGE0_BASE  0xa000
@@ -561,6 +561,13 @@ void mmhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev, bool value)
 			WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
 	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
 			EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	if (!value) {
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_NO_RETRY_FAULT, 1);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_RETRY_FAULT, 1);
+    }
+
 	WREG32_SOC15(MMHUB, 0, mmVM_L2_PROTECTION_FAULT_CNTL, tmp);
 }
 

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

@@ -24,7 +24,7 @@
 #ifndef __MXGPU_AI_H__
 #define __MXGPU_AI_H__
 
-#define AI_MAILBOX_TIMEDOUT	5000
+#define AI_MAILBOX_TIMEDOUT	12000
 
 enum idh_request {
 	IDH_REQ_GPU_INIT_ACCESS = 1,

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

@@ -23,7 +23,7 @@
 #ifndef __MXGPU_VI_H__
 #define __MXGPU_VI_H__
 
-#define VI_MAILBOX_TIMEDOUT	5000
+#define VI_MAILBOX_TIMEDOUT	12000
 #define VI_MAILBOX_RESET_TIME	12
 
 /* VI mailbox messages request */

drivers/gpu/drm/amd/amdgpu/psp_v10_0.c

@@ -35,6 +35,8 @@
 #include "raven1/GC/gc_9_1_offset.h"
 #include "raven1/SDMA0/sdma0_4_1_offset.h"
 
+MODULE_FIRMWARE("amdgpu/raven_asd.bin");
+
 static int
 psp_v10_0_get_fw_type(struct amdgpu_firmware_info *ucode, enum psp_gfx_fw_type *type)
 {
@@ -136,15 +138,13 @@ int psp_v10_0_prep_cmd_buf(struct amdgpu_firmware_info *ucode, struct psp_gfx_cm
 {
 	int ret;
 	uint64_t fw_mem_mc_addr = ucode->mc_addr;
-	struct  common_firmware_header *header;
 
 	memset(cmd, 0, sizeof(struct psp_gfx_cmd_resp));
-	header = (struct common_firmware_header *)ucode->fw;
 
 	cmd->cmd_id = GFX_CMD_ID_LOAD_IP_FW;
 	cmd->cmd.cmd_load_ip_fw.fw_phy_addr_lo = lower_32_bits(fw_mem_mc_addr);
 	cmd->cmd.cmd_load_ip_fw.fw_phy_addr_hi = upper_32_bits(fw_mem_mc_addr);
-	cmd->cmd.cmd_load_ip_fw.fw_size = le32_to_cpu(header->ucode_size_bytes);
+	cmd->cmd.cmd_load_ip_fw.fw_size = ucode->ucode_size;
 
 	ret = psp_v10_0_get_fw_type(ucode, &cmd->cmd.cmd_load_ip_fw.fw_type);
 	if (ret)
@@ -209,7 +209,7 @@ int psp_v10_0_ring_create(struct psp_context *psp, enum psp_ring_type ring_type)
 	return ret;
 }
 
-int psp_v10_0_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type)
+int psp_v10_0_ring_stop(struct psp_context *psp, enum psp_ring_type ring_type)
 {
 	int ret = 0;
 	struct psp_ring *ring;
@@ -229,6 +229,19 @@ int psp_v10_0_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
 			   0x80000000, 0x80000000, false);
 
+	return ret;
+}
+
+int psp_v10_0_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type)
+{
+	int ret = 0;
+	struct psp_ring *ring = &psp->km_ring;
+	struct amdgpu_device *adev = psp->adev;
+
+	ret = psp_v10_0_ring_stop(psp, ring_type);
+	if (ret)
+		DRM_ERROR("Fail to stop psp ring\n");
+
 	amdgpu_bo_free_kernel(&adev->firmware.rbuf,
 			      &ring->ring_mem_mc_addr,
 			      (void **)&ring->ring_mem);
@@ -245,15 +258,20 @@ int psp_v10_0_cmd_submit(struct psp_context *psp,
 	struct psp_gfx_rb_frame * write_frame = psp->km_ring.ring_mem;
 	struct psp_ring *ring = &psp->km_ring;
 	struct amdgpu_device *adev = psp->adev;
+	uint32_t ring_size_dw = ring->ring_size / 4;
+	uint32_t rb_frame_size_dw = sizeof(struct psp_gfx_rb_frame) / 4;
 
 	/* KM (GPCOM) prepare write pointer */
 	psp_write_ptr_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67);
 
 	/* Update KM RB frame pointer to new frame */
-	if ((psp_write_ptr_reg % ring->ring_size) == 0)
+	if ((psp_write_ptr_reg % ring_size_dw) == 0)
 		write_frame = ring->ring_mem;
 	else
-		write_frame = ring->ring_mem + (psp_write_ptr_reg / (sizeof(struct psp_gfx_rb_frame) / 4));
+		write_frame = ring->ring_mem + (psp_write_ptr_reg / rb_frame_size_dw);
+
+	/* Initialize KM RB frame */
+	memset(write_frame, 0, sizeof(struct psp_gfx_rb_frame));
 
 	/* Update KM RB frame */
 	write_frame->cmd_buf_addr_hi = upper_32_bits(cmd_buf_mc_addr);
@@ -263,8 +281,7 @@ int psp_v10_0_cmd_submit(struct psp_context *psp,
 	write_frame->fence_value = index;
 
 	/* Update the write Pointer in DWORDs */
-	psp_write_ptr_reg += sizeof(struct psp_gfx_rb_frame) / 4;
-	psp_write_ptr_reg = (psp_write_ptr_reg >= ring->ring_size) ? 0 : psp_write_ptr_reg;
+	psp_write_ptr_reg = (psp_write_ptr_reg + rb_frame_size_dw) % ring_size_dw;
 	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67, psp_write_ptr_reg);
 
 	return 0;
@@ -390,3 +407,10 @@ bool psp_v10_0_compare_sram_data(struct psp_context *psp,
 
 	return true;
 }
+
+
+int psp_v10_0_mode1_reset(struct psp_context *psp)
+{
+	DRM_INFO("psp mode 1 reset not supported now! \n");
+	return -EINVAL;
+}

drivers/gpu/drm/amd/amdgpu/psp_v10_0.h

@@ -34,6 +34,8 @@ extern int psp_v10_0_ring_init(struct psp_context *psp,
 			      enum psp_ring_type ring_type);
 extern int psp_v10_0_ring_create(struct psp_context *psp,
 				 enum psp_ring_type ring_type);
+extern int psp_v10_0_ring_stop(struct psp_context *psp,
+				  enum psp_ring_type ring_type);
 extern int psp_v10_0_ring_destroy(struct psp_context *psp,
 				  enum psp_ring_type ring_type);
 extern int psp_v10_0_cmd_submit(struct psp_context *psp,
@@ -43,4 +45,6 @@ extern int psp_v10_0_cmd_submit(struct psp_context *psp,
 extern bool psp_v10_0_compare_sram_data(struct psp_context *psp,
 				       struct amdgpu_firmware_info *ucode,
 				       enum AMDGPU_UCODE_ID ucode_type);
+
+extern int psp_v10_0_mode1_reset(struct psp_context *psp);
 #endif

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

@@ -319,7 +319,7 @@ int psp_v3_1_ring_create(struct psp_context *psp, enum psp_ring_type ring_type)
 	return ret;
 }
 
-int psp_v3_1_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type)
+int psp_v3_1_ring_stop(struct psp_context *psp, enum psp_ring_type ring_type)
 {
 	int ret = 0;
 	struct psp_ring *ring;
@@ -339,6 +339,19 @@ int psp_v3_1_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type)
 	ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
 			   0x80000000, 0x80000000, 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 = &psp->km_ring;
+	struct amdgpu_device *adev = psp->adev;
+
+	ret = psp_v3_1_ring_stop(psp, ring_type);
+	if (ret)
+		DRM_ERROR("Fail to stop psp ring\n");
+
 	amdgpu_bo_free_kernel(&adev->firmware.rbuf,
 			      &ring->ring_mem_mc_addr,
 			      (void **)&ring->ring_mem);
@@ -517,3 +530,37 @@ bool psp_v3_1_smu_reload_quirk(struct psp_context *psp)
 	reg = RREG32_SOC15(NBIO, 0, mmPCIE_DATA2);
 	return (reg & MP1_FIRMWARE_FLAGS__INTERRUPTS_ENABLED_MASK) ? true : false;
 }
+
+int psp_v3_1_mode1_reset(struct psp_context *psp)
+{
+	int ret;
+	uint32_t offset;
+	struct amdgpu_device *adev = psp->adev;
+
+	offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64);
+
+	ret = psp_wait_for(psp, offset, 0x80000000, 0x8000FFFF, false);
+
+	if (ret) {
+		DRM_INFO("psp is not working correctly before mode1 reset!\n");
+		return -EINVAL;
+	}
+
+	/*send the mode 1 reset command*/
+	WREG32(offset, 0x70000);
+
+	mdelay(1000);
+
+	offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_33);
+
+	ret = psp_wait_for(psp, offset, 0x80000000, 0x80000000, false);
+
+	if (ret) {
+		DRM_INFO("psp mode 1 reset failed!\n");
+		return -EINVAL;
+	}
+
+	DRM_INFO("psp mode1 reset succeed \n");
+
+	return 0;
+}

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

@@ -41,6 +41,8 @@ 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_stop(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,
@@ -51,4 +53,5 @@ extern bool psp_v3_1_compare_sram_data(struct psp_context *psp,
 				       struct amdgpu_firmware_info *ucode,
 				       enum AMDGPU_UCODE_ID ucode_type);
 extern bool psp_v3_1_smu_reload_quirk(struct psp_context *psp);
+extern int psp_v3_1_mode1_reset(struct psp_context *psp);
 #endif

drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c

@@ -1324,8 +1324,13 @@ static void sdma_v2_4_set_buffer_funcs(struct amdgpu_device *adev)
 }
 
 static const struct amdgpu_vm_pte_funcs sdma_v2_4_vm_pte_funcs = {
+	.copy_pte_num_dw = 7,
 	.copy_pte = sdma_v2_4_vm_copy_pte,
+
 	.write_pte = sdma_v2_4_vm_write_pte,
+
+	.set_max_nums_pte_pde = 0x1fffff >> 3,
+	.set_pte_pde_num_dw = 10,
 	.set_pte_pde = sdma_v2_4_vm_set_pte_pde,
 };
 

drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c

@@ -379,8 +379,10 @@ static void sdma_v3_0_ring_set_wptr(struct amdgpu_ring *ring)
 	struct amdgpu_device *adev = ring->adev;
 
 	if (ring->use_doorbell) {
+		u32 *wb = (u32 *)&adev->wb.wb[ring->wptr_offs];
+
 		/* XXX check if swapping is necessary on BE */
-		adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr) << 2;
+		WRITE_ONCE(*wb, (lower_32_bits(ring->wptr) << 2));
 		WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr) << 2);
 	} else {
 		int me = (ring == &ring->adev->sdma.instance[0].ring) ? 0 : 1;
@@ -641,10 +643,11 @@ static void sdma_v3_0_enable(struct amdgpu_device *adev, bool enable)
 static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev)
 {
 	struct amdgpu_ring *ring;
-	u32 rb_cntl, ib_cntl;
+	u32 rb_cntl, ib_cntl, wptr_poll_cntl;
 	u32 rb_bufsz;
 	u32 wb_offset;
 	u32 doorbell;
+	u64 wptr_gpu_addr;
 	int i, j, r;
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
@@ -707,6 +710,20 @@ static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev)
 		}
 		WREG32(mmSDMA0_GFX_DOORBELL + sdma_offsets[i], doorbell);
 
+		/* setup the wptr shadow polling */
+		wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4);
+
+		WREG32(mmSDMA0_GFX_RB_WPTR_POLL_ADDR_LO + sdma_offsets[i],
+		       lower_32_bits(wptr_gpu_addr));
+		WREG32(mmSDMA0_GFX_RB_WPTR_POLL_ADDR_HI + sdma_offsets[i],
+		       upper_32_bits(wptr_gpu_addr));
+		wptr_poll_cntl = RREG32(mmSDMA0_GFX_RB_WPTR_POLL_CNTL + sdma_offsets[i]);
+		if (amdgpu_sriov_vf(adev))
+			wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl, SDMA0_GFX_RB_WPTR_POLL_CNTL, F32_POLL_ENABLE, 1);
+		else
+			wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl, SDMA0_GFX_RB_WPTR_POLL_CNTL, F32_POLL_ENABLE, 0);
+		WREG32(mmSDMA0_GFX_RB_WPTR_POLL_CNTL + sdma_offsets[i], wptr_poll_cntl);
+
 		/* enable DMA RB */
 		rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 1);
 		WREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i], rb_cntl);
@@ -1713,11 +1730,11 @@ static void sdma_v3_0_emit_fill_buffer(struct amdgpu_ib *ib,
 }
 
 static const struct amdgpu_buffer_funcs sdma_v3_0_buffer_funcs = {
-	.copy_max_bytes = 0x1fffff,
+	.copy_max_bytes = 0x3fffe0, /* not 0x3fffff due to HW limitation */
 	.copy_num_dw = 7,
 	.emit_copy_buffer = sdma_v3_0_emit_copy_buffer,
 
-	.fill_max_bytes = 0x1fffff,
+	.fill_max_bytes = 0x3fffe0, /* not 0x3fffff due to HW limitation */
 	.fill_num_dw = 5,
 	.emit_fill_buffer = sdma_v3_0_emit_fill_buffer,
 };
@@ -1731,8 +1748,14 @@ static void sdma_v3_0_set_buffer_funcs(struct amdgpu_device *adev)
 }
 
 static const struct amdgpu_vm_pte_funcs sdma_v3_0_vm_pte_funcs = {
+	.copy_pte_num_dw = 7,
 	.copy_pte = sdma_v3_0_vm_copy_pte,
+
 	.write_pte = sdma_v3_0_vm_write_pte,
+
+	/* not 0x3fffff due to HW limitation */
+	.set_max_nums_pte_pde = 0x3fffe0 >> 3,
+	.set_pte_pde_num_dw = 10,
 	.set_pte_pde = sdma_v3_0_vm_set_pte_pde,
 };
 

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

@@ -54,7 +54,7 @@ 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[] = {
-	SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CHICKEN_BITS), 0xfe931f07, 0x02831f07,
+	SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CHICKEN_BITS), 0xfe931f07, 0x02831d07,
 	SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CLK_CTRL), 0xff000ff0, 0x3f000100,
 	SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GFX_IB_CNTL), 0x800f0100, 0x00000100,
 	SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GFX_RB_WPTR_POLL_CNTL), 0xfffffff7, 0x00403000,
@@ -89,7 +89,7 @@ static const u32 golden_settings_sdma_vg10[] = {
 
 static const u32 golden_settings_sdma_4_1[] =
 {
-	SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CHICKEN_BITS), 0xfe931f07, 0x02831f07,
+	SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CHICKEN_BITS), 0xfe931f07, 0x02831d07,
 	SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CLK_CTRL), 0xffffffff, 0x3f000100,
 	SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GFX_IB_CNTL), 0x800f0111, 0x00000100,
 	SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GFX_RB_WPTR_POLL_CNTL), 0xfffffff7, 0x00403000,
@@ -398,7 +398,7 @@ static void sdma_v4_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring)
 {
 	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) |
 			  SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf));
-	amdgpu_ring_write(ring, SOC15_REG_OFFSET(HDP, 0, mmHDP_DEBUG0));
+	amdgpu_ring_write(ring, SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE));
 	amdgpu_ring_write(ring, 1);
 }
 
@@ -1264,6 +1264,11 @@ static int sdma_v4_0_sw_fini(void *handle)
 	for (i = 0; i < adev->sdma.num_instances; i++)
 		amdgpu_ring_fini(&adev->sdma.instance[i].ring);
 
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		release_firmware(adev->sdma.instance[i].fw);
+		adev->sdma.instance[i].fw = NULL;
+	}
+
 	return 0;
 }
 
@@ -1714,8 +1719,13 @@ static void sdma_v4_0_set_buffer_funcs(struct amdgpu_device *adev)
 }
 
 static const struct amdgpu_vm_pte_funcs sdma_v4_0_vm_pte_funcs = {
+	.copy_pte_num_dw = 7,
 	.copy_pte = sdma_v4_0_vm_copy_pte,
+
 	.write_pte = sdma_v4_0_vm_write_pte,
+
+	.set_max_nums_pte_pde = 0x400000 >> 3,
+	.set_pte_pde_num_dw = 10,
 	.set_pte_pde = sdma_v4_0_vm_set_pte_pde,
 };
 

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

@@ -887,8 +887,13 @@ static void si_dma_set_buffer_funcs(struct amdgpu_device *adev)
 }
 
 static const struct amdgpu_vm_pte_funcs si_dma_vm_pte_funcs = {
+	.copy_pte_num_dw = 5,
 	.copy_pte = si_dma_vm_copy_pte,
+
 	.write_pte = si_dma_vm_write_pte,
+
+	.set_max_nums_pte_pde = 0xffff8 >> 3,
+	.set_pte_pde_num_dw = 9,
 	.set_pte_pde = si_dma_vm_set_pte_pde,
 };
 

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

@@ -1847,7 +1847,6 @@ static int si_calculate_sclk_params(struct amdgpu_device *adev,
 
 static void si_thermal_start_smc_fan_control(struct amdgpu_device *adev);
 static void si_fan_ctrl_set_default_mode(struct amdgpu_device *adev);
-static void si_dpm_set_dpm_funcs(struct amdgpu_device *adev);
 static void si_dpm_set_irq_funcs(struct amdgpu_device *adev);
 
 static struct si_power_info *si_get_pi(struct amdgpu_device *adev)
@@ -3060,9 +3059,9 @@ static int si_get_vce_clock_voltage(struct amdgpu_device *adev,
 	return ret;
 }
 
-static bool si_dpm_vblank_too_short(struct amdgpu_device *adev)
+static bool si_dpm_vblank_too_short(void *handle)
 {
-
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	u32 vblank_time = amdgpu_dpm_get_vblank_time(adev);
 	/* we never hit the non-gddr5 limit so disable it */
 	u32 switch_limit = adev->mc.vram_type == AMDGPU_VRAM_TYPE_GDDR5 ? 450 : 0;
@@ -3871,9 +3870,10 @@ static int si_restrict_performance_levels_before_switch(struct amdgpu_device *ad
 		0 : -EINVAL;
 }
 
-static int si_dpm_force_performance_level(struct amdgpu_device *adev,
+static int si_dpm_force_performance_level(void *handle,
 				   enum amd_dpm_forced_level level)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct amdgpu_ps *rps = adev->pm.dpm.current_ps;
 	struct  si_ps *ps = si_get_ps(rps);
 	u32 levels = ps->performance_level_count;
@@ -6575,11 +6575,12 @@ static int si_fan_ctrl_stop_smc_fan_control(struct amdgpu_device *adev)
 	}
 }
 
-static int si_dpm_get_fan_speed_percent(struct amdgpu_device *adev,
+static int si_dpm_get_fan_speed_percent(void *handle,
 				      u32 *speed)
 {
 	u32 duty, duty100;
 	u64 tmp64;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	if (adev->pm.no_fan)
 		return -ENOENT;
@@ -6600,9 +6601,10 @@ static int si_dpm_get_fan_speed_percent(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int si_dpm_set_fan_speed_percent(struct amdgpu_device *adev,
+static int si_dpm_set_fan_speed_percent(void *handle,
 				      u32 speed)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct si_power_info *si_pi = si_get_pi(adev);
 	u32 tmp;
 	u32 duty, duty100;
@@ -6633,8 +6635,10 @@ static int si_dpm_set_fan_speed_percent(struct amdgpu_device *adev,
 	return 0;
 }
 
-static void si_dpm_set_fan_control_mode(struct amdgpu_device *adev, u32 mode)
+static void si_dpm_set_fan_control_mode(void *handle, u32 mode)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
 	if (mode) {
 		/* stop auto-manage */
 		if (adev->pm.dpm.fan.ucode_fan_control)
@@ -6649,8 +6653,9 @@ static void si_dpm_set_fan_control_mode(struct amdgpu_device *adev, u32 mode)
 	}
 }
 
-static u32 si_dpm_get_fan_control_mode(struct amdgpu_device *adev)
+static u32 si_dpm_get_fan_control_mode(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct si_power_info *si_pi = si_get_pi(adev);
 	u32 tmp;
 
@@ -6946,8 +6951,9 @@ static void si_dpm_disable(struct amdgpu_device *adev)
 	ni_update_current_ps(adev, boot_ps);
 }
 
-static int si_dpm_pre_set_power_state(struct amdgpu_device *adev)
+static int si_dpm_pre_set_power_state(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct evergreen_power_info *eg_pi = evergreen_get_pi(adev);
 	struct amdgpu_ps requested_ps = *adev->pm.dpm.requested_ps;
 	struct amdgpu_ps *new_ps = &requested_ps;
@@ -6984,8 +6990,9 @@ static int si_power_control_set_level(struct amdgpu_device *adev)
 	return 0;
 }
 
-static int si_dpm_set_power_state(struct amdgpu_device *adev)
+static int si_dpm_set_power_state(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct evergreen_power_info *eg_pi = evergreen_get_pi(adev);
 	struct amdgpu_ps *new_ps = &eg_pi->requested_rps;
 	struct amdgpu_ps *old_ps = &eg_pi->current_rps;
@@ -7086,8 +7093,9 @@ static int si_dpm_set_power_state(struct amdgpu_device *adev)
 	return 0;
 }
 
-static void si_dpm_post_set_power_state(struct amdgpu_device *adev)
+static void si_dpm_post_set_power_state(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct evergreen_power_info *eg_pi = evergreen_get_pi(adev);
 	struct amdgpu_ps *new_ps = &eg_pi->requested_rps;
 
@@ -7103,8 +7111,10 @@ void si_dpm_reset_asic(struct amdgpu_device *adev)
 }
 #endif
 
-static void si_dpm_display_configuration_changed(struct amdgpu_device *adev)
+static void si_dpm_display_configuration_changed(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
 	si_program_display_gap(adev);
 }
 
@@ -7486,9 +7496,10 @@ static void si_dpm_fini(struct amdgpu_device *adev)
 	amdgpu_free_extended_power_table(adev);
 }
 
-static void si_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev,
+static void si_dpm_debugfs_print_current_performance_level(void *handle,
 						    struct seq_file *m)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct evergreen_power_info *eg_pi = evergreen_get_pi(adev);
 	struct amdgpu_ps *rps = &eg_pi->current_rps;
 	struct  si_ps *ps = si_get_ps(rps);
@@ -7860,10 +7871,11 @@ static int si_dpm_set_powergating_state(void *handle,
 }
 
 /* get temperature in millidegrees */
-static int si_dpm_get_temp(struct amdgpu_device *adev)
+static int si_dpm_get_temp(void *handle)
 {
 	u32 temp;
 	int actual_temp = 0;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	temp = (RREG32(CG_MULT_THERMAL_STATUS) & CTF_TEMP_MASK) >>
 		CTF_TEMP_SHIFT;
@@ -7878,8 +7890,9 @@ static int si_dpm_get_temp(struct amdgpu_device *adev)
 	return actual_temp;
 }
 
-static u32 si_dpm_get_sclk(struct amdgpu_device *adev, bool low)
+static u32 si_dpm_get_sclk(void *handle, bool low)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct evergreen_power_info *eg_pi = evergreen_get_pi(adev);
 	struct  si_ps *requested_state = si_get_ps(&eg_pi->requested_rps);
 
@@ -7889,8 +7902,9 @@ static u32 si_dpm_get_sclk(struct amdgpu_device *adev, bool low)
 		return requested_state->performance_levels[requested_state->performance_level_count - 1].sclk;
 }
 
-static u32 si_dpm_get_mclk(struct amdgpu_device *adev, bool low)
+static u32 si_dpm_get_mclk(void *handle, bool low)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct evergreen_power_info *eg_pi = evergreen_get_pi(adev);
 	struct  si_ps *requested_state = si_get_ps(&eg_pi->requested_rps);
 
@@ -7900,9 +7914,11 @@ static u32 si_dpm_get_mclk(struct amdgpu_device *adev, bool low)
 		return requested_state->performance_levels[requested_state->performance_level_count - 1].mclk;
 }
 
-static void si_dpm_print_power_state(struct amdgpu_device *adev,
-				     struct amdgpu_ps *rps)
+static void si_dpm_print_power_state(void *handle,
+				     void *current_ps)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_ps *rps = (struct amdgpu_ps *)current_ps;
 	struct  si_ps *ps = si_get_ps(rps);
 	struct rv7xx_pl *pl;
 	int i;
@@ -7927,7 +7943,6 @@ static int si_dpm_early_init(void *handle)
 
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	si_dpm_set_dpm_funcs(adev);
 	si_dpm_set_irq_funcs(adev);
 	return 0;
 }
@@ -7942,20 +7957,23 @@ static inline bool si_are_power_levels_equal(const struct rv7xx_pl  *si_cpl1,
 		  (si_cpl1->vddci == si_cpl2->vddci));
 }
 
-static int si_check_state_equal(struct amdgpu_device *adev,
-				struct amdgpu_ps *cps,
-				struct amdgpu_ps *rps,
+static int si_check_state_equal(void *handle,
+				void *current_ps,
+				void *request_ps,
 				bool *equal)
 {
 	struct si_ps *si_cps;
 	struct si_ps *si_rps;
 	int i;
+	struct amdgpu_ps *cps = (struct amdgpu_ps *)current_ps;
+	struct amdgpu_ps *rps = (struct amdgpu_ps *)request_ps;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	if (adev == NULL || cps == NULL || rps == NULL || equal == NULL)
 		return -EINVAL;
 
-	si_cps = si_get_ps(cps);
-	si_rps = si_get_ps(rps);
+	si_cps = si_get_ps((struct amdgpu_ps *)cps);
+	si_rps = si_get_ps((struct amdgpu_ps *)rps);
 
 	if (si_cps == NULL) {
 		printk("si_cps is NULL\n");
@@ -7983,9 +8001,10 @@ static int si_check_state_equal(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int si_dpm_read_sensor(struct amdgpu_device *adev, int idx,
+static int si_dpm_read_sensor(void *handle, int idx,
 			      void *value, int *size)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct evergreen_power_info *eg_pi = evergreen_get_pi(adev);
 	struct amdgpu_ps *rps = &eg_pi->current_rps;
 	struct  si_ps *ps = si_get_ps(rps);
@@ -8041,7 +8060,7 @@ const struct amd_ip_funcs si_dpm_ip_funcs = {
 	.set_powergating_state = si_dpm_set_powergating_state,
 };
 
-static const struct amdgpu_dpm_funcs si_dpm_funcs = {
+const struct amd_pm_funcs si_dpm_funcs = {
 	.get_temperature = &si_dpm_get_temp,
 	.pre_set_power_state = &si_dpm_pre_set_power_state,
 	.set_power_state = &si_dpm_set_power_state,
@@ -8062,12 +8081,6 @@ static const struct amdgpu_dpm_funcs si_dpm_funcs = {
 	.read_sensor = &si_dpm_read_sensor,
 };
 
-static void si_dpm_set_dpm_funcs(struct amdgpu_device *adev)
-{
-	if (adev->pm.funcs == NULL)
-		adev->pm.funcs = &si_dpm_funcs;
-}
-
 static const struct amdgpu_irq_src_funcs si_dpm_irq_funcs = {
 	.set = si_dpm_set_interrupt_state,
 	.process = si_dpm_process_interrupt,

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

@@ -246,6 +246,7 @@ enum si_display_gap
 };
 
 extern const struct amd_ip_funcs si_dpm_ip_funcs;
+extern const struct amd_pm_funcs si_dpm_funcs;
 
 struct ni_leakage_coeffients
 {

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

@@ -118,6 +118,19 @@ static u32 si_ih_get_wptr(struct amdgpu_device *adev)
 	return (wptr & adev->irq.ih.ptr_mask);
 }
 
+/**
+ * si_ih_prescreen_iv - prescreen an interrupt vector
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Returns true if the interrupt vector should be further processed.
+ */
+static bool si_ih_prescreen_iv(struct amdgpu_device *adev)
+{
+	/* Process all interrupts */
+	return true;
+}
+
 static void si_ih_decode_iv(struct amdgpu_device *adev,
 			     struct amdgpu_iv_entry *entry)
 {
@@ -288,6 +301,7 @@ static const struct amd_ip_funcs si_ih_ip_funcs = {
 
 static const struct amdgpu_ih_funcs si_ih_funcs = {
 	.get_wptr = si_ih_get_wptr,
+	.prescreen_iv = si_ih_prescreen_iv,
 	.decode_iv = si_ih_decode_iv,
 	.set_rptr = si_ih_set_rptr
 };

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

@@ -407,18 +407,27 @@ static int soc15_read_register(struct amdgpu_device *adev, u32 se_num,
 	return -EINVAL;
 }
 
-static void soc15_gpu_pci_config_reset(struct amdgpu_device *adev)
+static int soc15_asic_reset(struct amdgpu_device *adev)
 {
 	u32 i;
 
-	dev_info(adev->dev, "GPU pci config reset\n");
+	amdgpu_atombios_scratch_regs_engine_hung(adev, true);
+
+	dev_info(adev->dev, "GPU reset\n");
 
 	/* disable BM */
 	pci_clear_master(adev->pdev);
-	/* reset */
-	amdgpu_pci_config_reset(adev);
 
-	udelay(100);
+	pci_save_state(adev->pdev);
+
+	for (i = 0; i < AMDGPU_MAX_IP_NUM; i++) {
+		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP){
+			adev->ip_blocks[i].version->funcs->soft_reset((void *)adev);
+			break;
+		}
+	}
+
+	pci_restore_state(adev->pdev);
 
 	/* wait for asic to come out of reset */
 	for (i = 0; i < adev->usec_timeout; i++) {
@@ -430,14 +439,6 @@ static void soc15_gpu_pci_config_reset(struct amdgpu_device *adev)
 		udelay(1);
 	}
 
-}
-
-static int soc15_asic_reset(struct amdgpu_device *adev)
-{
-	amdgpu_atombios_scratch_regs_engine_hung(adev, true);
-
-	soc15_gpu_pci_config_reset(adev);
-
 	amdgpu_atombios_scratch_regs_engine_hung(adev, false);
 
 	return 0;

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

@@ -218,6 +218,19 @@ static u32 tonga_ih_get_wptr(struct amdgpu_device *adev)
 	return (wptr & adev->irq.ih.ptr_mask);
 }
 
+/**
+ * tonga_ih_prescreen_iv - prescreen an interrupt vector
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Returns true if the interrupt vector should be further processed.
+ */
+static bool tonga_ih_prescreen_iv(struct amdgpu_device *adev)
+{
+	/* Process all interrupts */
+	return true;
+}
+
 /**
  * tonga_ih_decode_iv - decode an interrupt vector
  *
@@ -478,6 +491,7 @@ static const struct amd_ip_funcs tonga_ih_ip_funcs = {
 
 static const struct amdgpu_ih_funcs tonga_ih_funcs = {
 	.get_wptr = tonga_ih_get_wptr,
+	.prescreen_iv = tonga_ih_prescreen_iv,
 	.decode_iv = tonga_ih_decode_iv,
 	.set_rptr = tonga_ih_set_rptr
 };

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

@@ -1161,7 +1161,7 @@ static void uvd_v7_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
  */
 static void uvd_v7_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring)
 {
-	amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(HDP, 0, mmHDP_DEBUG0), 0));
+	amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 0));
 	amdgpu_ring_write(ring, 1);
 }
 

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

@@ -1011,10 +1011,6 @@ static int vce_v4_0_process_interrupt(struct amdgpu_device *adev,
 {
 	DRM_DEBUG("IH: VCE\n");
 
-	WREG32_P(SOC15_REG_OFFSET(VCE, 0, mmVCE_SYS_INT_STATUS),
-			VCE_SYS_INT_STATUS__VCE_SYS_INT_TRAP_INTERRUPT_INT_MASK,
-			~VCE_SYS_INT_STATUS__VCE_SYS_INT_TRAP_INTERRUPT_INT_MASK);
-
 	switch (entry->src_data[0]) {
 	case 0:
 	case 1:

drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c

@@ -812,7 +812,7 @@ static void vcn_v1_0_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64
  */
 static void vcn_v1_0_dec_ring_emit_hdp_invalidate(struct amdgpu_ring *ring)
 {
-	amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(HDP, 0, mmHDP_DEBUG0), 0));
+	amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 0));
 	amdgpu_ring_write(ring, 1);
 }
 

drivers/gpu/drm/amd/amdgpu/vega10_ih.c

@@ -219,13 +219,91 @@ static u32 vega10_ih_get_wptr(struct amdgpu_device *adev)
 			wptr, adev->irq.ih.rptr, tmp);
 		adev->irq.ih.rptr = tmp;
 
-		tmp = RREG32(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL));
+		tmp = RREG32_NO_KIQ(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL));
 		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
-		WREG32(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL), tmp);
+		WREG32_NO_KIQ(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL), tmp);
 	}
 	return (wptr & adev->irq.ih.ptr_mask);
 }
 
+/**
+ * vega10_ih_prescreen_iv - prescreen an interrupt vector
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Returns true if the interrupt vector should be further processed.
+ */
+static bool vega10_ih_prescreen_iv(struct amdgpu_device *adev)
+{
+	u32 ring_index = adev->irq.ih.rptr >> 2;
+	u32 dw0, dw3, dw4, dw5;
+	u16 pasid;
+	u64 addr, key;
+	struct amdgpu_vm *vm;
+	int r;
+
+	dw0 = le32_to_cpu(adev->irq.ih.ring[ring_index + 0]);
+	dw3 = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]);
+	dw4 = le32_to_cpu(adev->irq.ih.ring[ring_index + 4]);
+	dw5 = le32_to_cpu(adev->irq.ih.ring[ring_index + 5]);
+
+	/* Filter retry page faults, let only the first one pass. If
+	 * there are too many outstanding faults, ignore them until
+	 * some faults get cleared.
+	 */
+	switch (dw0 & 0xff) {
+	case AMDGPU_IH_CLIENTID_VMC:
+	case AMDGPU_IH_CLIENTID_UTCL2:
+		break;
+	default:
+		/* Not a VM fault */
+		return true;
+	}
+
+	/* Not a retry fault */
+	if (!(dw5 & 0x80))
+		return true;
+
+	pasid = dw3 & 0xffff;
+	/* No PASID, can't identify faulting process */
+	if (!pasid)
+		return true;
+
+	addr = ((u64)(dw5 & 0xf) << 44) | ((u64)dw4 << 12);
+	key = AMDGPU_VM_FAULT(pasid, addr);
+	r = amdgpu_ih_add_fault(adev, key);
+
+	/* Hash table is full or the fault is already being processed,
+	 * ignore further page faults
+	 */
+	if (r != 0)
+		goto ignore_iv;
+
+	/* Track retry faults in per-VM fault FIFO. */
+	spin_lock(&adev->vm_manager.pasid_lock);
+	vm = idr_find(&adev->vm_manager.pasid_idr, pasid);
+	spin_unlock(&adev->vm_manager.pasid_lock);
+	if (WARN_ON_ONCE(!vm)) {
+		/* VM not found, process it normally */
+		amdgpu_ih_clear_fault(adev, key);
+		return true;
+	}
+	/* No locking required with single writer and single reader */
+	r = kfifo_put(&vm->faults, key);
+	if (!r) {
+		/* FIFO is full. Ignore it until there is space */
+		amdgpu_ih_clear_fault(adev, key);
+		goto ignore_iv;
+	}
+
+	/* It's the first fault for this address, process it normally */
+	return true;
+
+ignore_iv:
+	adev->irq.ih.rptr += 32;
+	return false;
+}
+
 /**
  * vega10_ih_decode_iv - decode an interrupt vector
  *
@@ -310,6 +388,14 @@ static int vega10_ih_sw_init(void *handle)
 	adev->irq.ih.use_doorbell = true;
 	adev->irq.ih.doorbell_index = AMDGPU_DOORBELL64_IH << 1;
 
+	adev->irq.ih.faults = kmalloc(sizeof(*adev->irq.ih.faults), GFP_KERNEL);
+	if (!adev->irq.ih.faults)
+		return -ENOMEM;
+	INIT_CHASH_TABLE(adev->irq.ih.faults->hash,
+			 AMDGPU_PAGEFAULT_HASH_BITS, 8, 0);
+	spin_lock_init(&adev->irq.ih.faults->lock);
+	adev->irq.ih.faults->count = 0;
+
 	r = amdgpu_irq_init(adev);
 
 	return r;
@@ -322,6 +408,9 @@ static int vega10_ih_sw_fini(void *handle)
 	amdgpu_irq_fini(adev);
 	amdgpu_ih_ring_fini(adev);
 
+	kfree(adev->irq.ih.faults);
+	adev->irq.ih.faults = NULL;
+
 	return 0;
 }
 
@@ -410,6 +499,7 @@ const struct amd_ip_funcs vega10_ih_ip_funcs = {
 
 static const struct amdgpu_ih_funcs vega10_ih_funcs = {
 	.get_wptr = vega10_ih_get_wptr,
+	.prescreen_iv = vega10_ih_prescreen_iv,
 	.decode_iv = vega10_ih_decode_iv,
 	.set_rptr = vega10_ih_set_rptr
 };

drivers/gpu/drm/amd/amdkfd/kfd_device.c

@@ -168,13 +168,6 @@ static bool device_iommu_pasid_init(struct kfd_dev *kfd)
 	pasid_limit = min_t(unsigned int,
 			(unsigned int)(1 << kfd->device_info->max_pasid_bits),
 			iommu_info.max_pasids);
-	/*
-	 * last pasid is used for kernel queues doorbells
-	 * in the future the last pasid might be used for a kernel thread.
-	 */
-	pasid_limit = min_t(unsigned int,
-				pasid_limit,
-				kfd->doorbell_process_limit - 1);
 
 	err = amd_iommu_init_device(kfd->pdev, pasid_limit);
 	if (err < 0) {

drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c

@@ -24,16 +24,15 @@
 #include <linux/mman.h>
 #include <linux/slab.h>
 #include <linux/io.h>
+#include <linux/idr.h>
 
 /*
- * This extension supports a kernel level doorbells management for
- * the kernel queues.
- * Basically the last doorbells page is devoted to kernel queues
- * and that's assures that any user process won't get access to the
- * kernel doorbells page
+ * This extension supports a kernel level doorbells management for the
+ * kernel queues using the first doorbell page reserved for the kernel.
  */
 
-#define KERNEL_DOORBELL_PASID 1
+static DEFINE_IDA(doorbell_ida);
+static unsigned int max_doorbell_slices;
 #define KFD_SIZE_OF_DOORBELL_IN_BYTES 4
 
 /*
@@ -84,13 +83,16 @@ int kfd_doorbell_init(struct kfd_dev *kfd)
 			(doorbell_aperture_size - doorbell_start_offset) /
 						doorbell_process_allocation();
 	else
-		doorbell_process_limit = 0;
+		return -ENOSPC;
+
+	if (!max_doorbell_slices ||
+	    doorbell_process_limit < max_doorbell_slices)
+		max_doorbell_slices = doorbell_process_limit;
 
 	kfd->doorbell_base = kfd->shared_resources.doorbell_physical_address +
 				doorbell_start_offset;
 
 	kfd->doorbell_id_offset = doorbell_start_offset / sizeof(u32);
-	kfd->doorbell_process_limit = doorbell_process_limit - 1;
 
 	kfd->doorbell_kernel_ptr = ioremap(kfd->doorbell_base,
 						doorbell_process_allocation());
@@ -185,11 +187,10 @@ u32 __iomem *kfd_get_kernel_doorbell(struct kfd_dev *kfd,
 		return NULL;
 
 	/*
-	 * Calculating the kernel doorbell offset using "faked" kernel
-	 * pasid that allocated for kernel queues only
+	 * Calculating the kernel doorbell offset using the first
+	 * doorbell page.
 	 */
-	*doorbell_off = KERNEL_DOORBELL_PASID * (doorbell_process_allocation() /
-							sizeof(u32)) + inx;
+	*doorbell_off = kfd->doorbell_id_offset + inx;
 
 	pr_debug("Get kernel queue doorbell\n"
 			 "     doorbell offset   == 0x%08X\n"
@@ -228,11 +229,12 @@ unsigned int kfd_queue_id_to_doorbell(struct kfd_dev *kfd,
 {
 	/*
 	 * doorbell_id_offset accounts for doorbells taken by KGD.
-	 * pasid * doorbell_process_allocation/sizeof(u32) adjusts
-	 * to the process's doorbells
+	 * index * doorbell_process_allocation/sizeof(u32) adjusts to
+	 * the process's doorbells.
 	 */
 	return kfd->doorbell_id_offset +
-		process->pasid * (doorbell_process_allocation()/sizeof(u32)) +
+		process->doorbell_index
+		* doorbell_process_allocation() / sizeof(u32) +
 		queue_id;
 }
 
@@ -250,5 +252,21 @@ phys_addr_t kfd_get_process_doorbells(struct kfd_dev *dev,
 					struct kfd_process *process)
 {
 	return dev->doorbell_base +
-		process->pasid * doorbell_process_allocation();
+		process->doorbell_index * doorbell_process_allocation();
+}
+
+int kfd_alloc_process_doorbells(struct kfd_process *process)
+{
+	int r = ida_simple_get(&doorbell_ida, 1, max_doorbell_slices,
+				GFP_KERNEL);
+	if (r > 0)
+		process->doorbell_index = r;
+
+	return r;
+}
+
+void kfd_free_process_doorbells(struct kfd_process *process)
+{
+	if (process->doorbell_index)
+		ida_simple_remove(&doorbell_ida, process->doorbell_index);
 }

drivers/gpu/drm/amd/amdkfd/kfd_module.c

@@ -103,10 +103,6 @@ static int __init kfd_module_init(void)
 		return -1;
 	}
 
-	err = kfd_pasid_init();
-	if (err < 0)
-		return err;
-
 	err = kfd_chardev_init();
 	if (err < 0)
 		goto err_ioctl;
@@ -126,7 +122,6 @@ static int __init kfd_module_init(void)
 err_topology:
 	kfd_chardev_exit();
 err_ioctl:
-	kfd_pasid_exit();
 	return err;
 }
 
@@ -137,7 +132,6 @@ static void __exit kfd_module_exit(void)
 	kfd_process_destroy_wq();
 	kfd_topology_shutdown();
 	kfd_chardev_exit();
-	kfd_pasid_exit();
 	dev_info(kfd_device, "Removed module\n");
 }
 

drivers/gpu/drm/amd/amdkfd/kfd_pasid.c

@@ -20,78 +20,64 @@
  * OTHER DEALINGS IN THE SOFTWARE.
  */
 
-#include <linux/slab.h>
 #include <linux/types.h>
 #include "kfd_priv.h"
 
-static unsigned long *pasid_bitmap;
-static unsigned int pasid_limit;
-static DEFINE_MUTEX(pasid_mutex);
-
-int kfd_pasid_init(void)
-{
-	pasid_limit = KFD_MAX_NUM_OF_PROCESSES;
-
-	pasid_bitmap = kcalloc(BITS_TO_LONGS(pasid_limit), sizeof(long),
-				GFP_KERNEL);
-	if (!pasid_bitmap)
-		return -ENOMEM;
-
-	set_bit(0, pasid_bitmap); /* PASID 0 is reserved. */
-
-	return 0;
-}
-
-void kfd_pasid_exit(void)
-{
-	kfree(pasid_bitmap);
-}
+static unsigned int pasid_bits = 16;
+static const struct kfd2kgd_calls *kfd2kgd;
 
 bool kfd_set_pasid_limit(unsigned int new_limit)
 {
-	if (new_limit < pasid_limit) {
-		bool ok;
-
-		mutex_lock(&pasid_mutex);
-
-		/* ensure that no pasids >= new_limit are in-use */
-		ok = (find_next_bit(pasid_bitmap, pasid_limit, new_limit) ==
-								pasid_limit);
-		if (ok)
-			pasid_limit = new_limit;
-
-		mutex_unlock(&pasid_mutex);
-
-		return ok;
+	if (new_limit < 2)
+		return false;
+
+	if (new_limit < (1U << pasid_bits)) {
+		if (kfd2kgd)
+			/* We've already allocated user PASIDs, too late to
+			 * change the limit
+			 */
+			return false;
+
+		while (new_limit < (1U << pasid_bits))
+			pasid_bits--;
 	}
 
 	return true;
 }
 
-inline unsigned int kfd_get_pasid_limit(void)
+unsigned int kfd_get_pasid_limit(void)
 {
-	return pasid_limit;
+	return 1U << pasid_bits;
 }
 
 unsigned int kfd_pasid_alloc(void)
 {
-	unsigned int found;
-
-	mutex_lock(&pasid_mutex);
-
-	found = find_first_zero_bit(pasid_bitmap, pasid_limit);
-	if (found == pasid_limit)
-		found = 0;
-	else
-		set_bit(found, pasid_bitmap);
+	int r;
+
+	/* Find the first best KFD device for calling KGD */
+	if (!kfd2kgd) {
+		struct kfd_dev *dev = NULL;
+		unsigned int i = 0;
+
+		while ((dev = kfd_topology_enum_kfd_devices(i)) != NULL) {
+			if (dev && dev->kfd2kgd) {
+				kfd2kgd = dev->kfd2kgd;
+				break;
+			}
+			i++;
+		}
+
+		if (!kfd2kgd)
+			return false;
+	}
 
-	mutex_unlock(&pasid_mutex);
+	r = kfd2kgd->alloc_pasid(pasid_bits);
 
-	return found;
+	return r > 0 ? r : 0;
 }
 
 void kfd_pasid_free(unsigned int pasid)
 {
-	if (!WARN_ON(pasid == 0 || pasid >= pasid_limit))
-		clear_bit(pasid, pasid_bitmap);
+	if (kfd2kgd)
+		kfd2kgd->free_pasid(pasid);
 }

drivers/gpu/drm/amd/amdkfd/kfd_priv.h

@@ -157,9 +157,6 @@ struct kfd_dev {
 					 * to HW doorbell, GFX reserved some
 					 * at the start)
 					 */
-	size_t doorbell_process_limit;	/* Number of processes we have doorbell
-					 * space for.
-					 */
 	u32 __iomem *doorbell_kernel_ptr; /* This is a pointer for a doorbells
 					   * page used by kernel queue
 					   */
@@ -495,6 +492,7 @@ struct kfd_process {
 	struct rcu_head	rcu;
 
 	unsigned int pasid;
+	unsigned int doorbell_index;
 
 	/*
 	 * List of kfd_process_device structures,
@@ -583,6 +581,10 @@ void write_kernel_doorbell(u32 __iomem *db, u32 value);
 unsigned int kfd_queue_id_to_doorbell(struct kfd_dev *kfd,
 					struct kfd_process *process,
 					unsigned int queue_id);
+phys_addr_t kfd_get_process_doorbells(struct kfd_dev *dev,
+					struct kfd_process *process);
+int kfd_alloc_process_doorbells(struct kfd_process *process);
+void kfd_free_process_doorbells(struct kfd_process *process);
 
 /* GTT Sub-Allocator */
 
@@ -694,8 +696,6 @@ int pm_send_unmap_queue(struct packet_manager *pm, enum kfd_queue_type type,
 void pm_release_ib(struct packet_manager *pm);
 
 uint64_t kfd_get_number_elems(struct kfd_dev *kfd);
-phys_addr_t kfd_get_process_doorbells(struct kfd_dev *dev,
-					struct kfd_process *process);
 
 /* Events */
 extern const struct kfd_event_interrupt_class event_interrupt_class_cik;

drivers/gpu/drm/amd/amdkfd/kfd_process.c

@@ -183,6 +183,7 @@ static void kfd_process_wq_release(struct work_struct *work)
 	kfd_event_free_process(p);
 
 	kfd_pasid_free(p->pasid);
+	kfd_free_process_doorbells(p);
 
 	mutex_unlock(&p->mutex);
 
@@ -288,6 +289,9 @@ static struct kfd_process *create_process(const struct task_struct *thread)
 	if (process->pasid == 0)
 		goto err_alloc_pasid;
 
+	if (kfd_alloc_process_doorbells(process) < 0)
+		goto err_alloc_doorbells;
+
 	mutex_init(&process->mutex);
 
 	process->mm = thread->mm;
@@ -329,6 +333,8 @@ err_process_pqm_init:
 	mmu_notifier_unregister_no_release(&process->mmu_notifier, process->mm);
 err_mmu_notifier:
 	mutex_destroy(&process->mutex);
+	kfd_free_process_doorbells(process);
+err_alloc_doorbells:
 	kfd_pasid_free(process->pasid);
 err_alloc_pasid:
 	kfree(process->queues);

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

@@ -25,6 +25,8 @@
 
 #define AMD_MAX_USEC_TIMEOUT		200000  /* 200 ms */
 
+struct seq_file;
+
 /*
  * Supported ASIC types
  */
@@ -144,6 +146,12 @@ enum amd_fan_ctrl_mode {
 	AMD_FAN_CTRL_AUTO = 2,
 };
 
+enum pp_clock_type {
+	PP_SCLK,
+	PP_MCLK,
+	PP_PCIE,
+};
+
 /* CG flags */
 #define AMD_CG_SUPPORT_GFX_MGCG			(1 << 0)
 #define AMD_CG_SUPPORT_GFX_MGLS			(1 << 1)
@@ -249,4 +257,62 @@ struct amd_ip_funcs {
 	void (*get_clockgating_state)(void *handle, u32 *flags);
 };
 
+enum amd_pp_task;
+
+struct pp_states_info;
+
+struct amd_pm_funcs {
+	int (*get_temperature)(void *handle);
+	int (*pre_set_power_state)(void *handle);
+	int (*set_power_state)(void *handle);
+	void (*post_set_power_state)(void *handle);
+	void (*display_configuration_changed)(void *handle);
+	u32 (*get_sclk)(void *handle, bool low);
+	u32 (*get_mclk)(void *handle, bool low);
+	void (*print_power_state)(void *handle, void *ps);
+	void (*debugfs_print_current_performance_level)(void *handle, struct seq_file *m);
+	int (*force_performance_level)(void *handle, enum amd_dpm_forced_level level);
+	bool (*vblank_too_short)(void *handle);
+	void (*powergate_uvd)(void *handle, bool gate);
+	void (*powergate_vce)(void *handle, bool gate);
+	void (*enable_bapm)(void *handle, bool enable);
+	void (*set_fan_control_mode)(void *handle, u32 mode);
+	u32 (*get_fan_control_mode)(void *handle);
+	int (*set_fan_speed_percent)(void *handle, u32 speed);
+	int (*get_fan_speed_percent)(void *handle, u32 *speed);
+	int (*force_clock_level)(void *handle, enum pp_clock_type type, uint32_t mask);
+	int (*print_clock_levels)(void *handle, enum pp_clock_type type, char *buf);
+	int (*get_sclk_od)(void *handle);
+	int (*set_sclk_od)(void *handle, uint32_t value);
+	int (*get_mclk_od)(void *handle);
+	int (*set_mclk_od)(void *handle, uint32_t value);
+	int (*check_state_equal)(void *handle,
+				void  *cps,
+				void  *rps,
+				bool  *equal);
+	int (*read_sensor)(void *handle, int idx, void *value,
+			   int *size);
+
+	struct amd_vce_state* (*get_vce_clock_state)(void *handle, u32 idx);
+	int (*reset_power_profile_state)(void *handle,
+			struct amd_pp_profile *request);
+	int (*get_power_profile_state)(void *handle,
+			struct amd_pp_profile *query);
+	int (*set_power_profile_state)(void *handle,
+			struct amd_pp_profile *request);
+	int (*switch_power_profile)(void *handle,
+			enum amd_pp_profile_type type);
+	int (*load_firmware)(void *handle);
+	int (*wait_for_fw_loading_complete)(void *handle);
+	enum amd_dpm_forced_level (*get_performance_level)(void *handle);
+	enum amd_pm_state_type (*get_current_power_state)(void *handle);
+	int (*dispatch_tasks)(void *handle, enum amd_pp_task task_id,
+				   void *input, void *output);
+	int (*get_fan_speed_rpm)(void *handle, uint32_t *rpm);
+	int (*get_pp_num_states)(void *handle, struct pp_states_info *data);
+	int (*get_pp_table)(void *handle, char **table);
+	int (*set_pp_table)(void *handle, const char *buf, size_t size);
+};
+
+
 #endif /* __AMD_SHARED_H__ */

drivers/gpu/drm/amd/include/asic_reg/smu/smu_7_0_1_sh_mask.h

@@ -5454,5 +5454,7 @@
 #define ROM_SW_DATA_64__ROM_SW_DATA__SHIFT 0x0
 #define CURRENT_PG_STATUS__VCE_PG_STATUS_MASK 0x00000002
 #define CURRENT_PG_STATUS__UVD_PG_STATUS_MASK 0x00000004
+#define SMC_SYSCON_MISC_CNTL__pre_fetcher_en_MASK  0x1
+#define SMC_SYSCON_MISC_CNTL__pre_fetcher_en__SHIFT 0
 
 #endif /* SMU_7_0_1_SH_MASK_H */

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

@@ -1017,6 +1017,19 @@ struct atom_14nm_combphy_tmds_vs_set
   uint8_t margin_deemph_lane0__deemph_sel_val;         
 };
 
+struct atom_i2c_reg_info {
+  uint8_t ucI2cRegIndex;
+  uint8_t ucI2cRegVal;
+};
+
+struct atom_hdmi_retimer_redriver_set {
+  uint8_t HdmiSlvAddr;
+  uint8_t HdmiRegNum;
+  uint8_t Hdmi6GRegNum;
+  struct atom_i2c_reg_info HdmiRegSetting[9];        //For non 6G Hz use
+  struct atom_i2c_reg_info Hdmi6GhzRegSetting[3];    //For 6G Hz use.
+};
+
 struct atom_integrated_system_info_v1_11
 {
   struct  atom_common_table_header  table_header;
@@ -1052,7 +1065,11 @@ struct atom_integrated_system_info_v1_11
   struct atom_14nm_dpphy_dp_tuningset dp_tuningset;
   struct atom_14nm_dpphy_dp_tuningset dp_hbr3_tuningset;
   struct atom_camera_data  camera_info;
-  uint32_t  reserved[138];
+  struct atom_hdmi_retimer_redriver_set dp0_retimer_set;   //for DP0
+  struct atom_hdmi_retimer_redriver_set dp1_retimer_set;   //for DP1
+  struct atom_hdmi_retimer_redriver_set dp2_retimer_set;   //for DP2
+  struct atom_hdmi_retimer_redriver_set dp3_retimer_set;   //for DP3
+  uint32_t  reserved[108];
 };
 
 

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

@@ -100,6 +100,7 @@ enum cgs_system_info_id {
 	CGS_SYSTEM_INFO_GFX_SE_INFO,
 	CGS_SYSTEM_INFO_PCIE_SUB_SYS_ID,
 	CGS_SYSTEM_INFO_PCIE_SUB_SYS_VENDOR_ID,
+	CGS_SYSTEM_INFO_PCIE_BUS_DEVFN,
 	CGS_SYSTEM_INFO_ID_MAXIMUM,
 };
 
@@ -193,8 +194,6 @@ struct cgs_acpi_method_info {
  * @type:	memory type
  * @size:	size in bytes
  * @align:	alignment in bytes
- * @min_offset: minimum offset from start of heap
- * @max_offset: maximum offset from start of heap
  * @handle:	memory handle (output)
  *
  * The memory types CGS_GPU_MEM_TYPE_*_CONTIG_FB force contiguous
@@ -216,7 +215,6 @@ struct cgs_acpi_method_info {
  */
 typedef int (*cgs_alloc_gpu_mem_t)(struct cgs_device *cgs_device, enum cgs_gpu_mem_type type,
 				   uint64_t size, uint64_t align,
-				   uint64_t min_offset, uint64_t max_offset,
 				   cgs_handle_t *handle);
 
 /**
@@ -310,6 +308,22 @@ typedef uint32_t (*cgs_read_ind_register_t)(struct cgs_device *cgs_device, enum
 typedef void (*cgs_write_ind_register_t)(struct cgs_device *cgs_device, enum cgs_ind_reg space,
 					 unsigned index, uint32_t value);
 
+#define CGS_REG_FIELD_SHIFT(reg, field) reg##__##field##__SHIFT
+#define CGS_REG_FIELD_MASK(reg, field) reg##__##field##_MASK
+
+#define CGS_REG_SET_FIELD(orig_val, reg, field, field_val)			\
+	(((orig_val) & ~CGS_REG_FIELD_MASK(reg, field)) |			\
+	 (CGS_REG_FIELD_MASK(reg, field) & ((field_val) << CGS_REG_FIELD_SHIFT(reg, field))))
+
+#define CGS_REG_GET_FIELD(value, reg, field)				\
+	(((value) & CGS_REG_FIELD_MASK(reg, field)) >> CGS_REG_FIELD_SHIFT(reg, field))
+
+#define CGS_WREG32_FIELD(device, reg, field, val)	\
+	cgs_write_register(device, mm##reg, (cgs_read_register(device, mm##reg) & ~CGS_REG_FIELD_MASK(reg, field)) | (val) << CGS_REG_FIELD_SHIFT(reg, field))
+
+#define CGS_WREG32_FIELD_IND(device, space, reg, field, val)	\
+	cgs_write_ind_register(device, space, ix##reg, (cgs_read_ind_register(device, space, ix##reg) & ~CGS_REG_FIELD_MASK(reg, field)) | (val) << CGS_REG_FIELD_SHIFT(reg, field))
+
 /**
  * cgs_get_pci_resource() - provide access to a device resource (PCI BAR)
  * @cgs_device:	opaque device handle
@@ -463,8 +477,8 @@ struct cgs_device
 #define CGS_OS_CALL(func,dev,...) \
 	(((struct cgs_device *)dev)->os_ops->func(dev, ##__VA_ARGS__))
 
-#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_alloc_gpu_mem(dev,type,size,align,handle)	\
+	CGS_CALL(alloc_gpu_mem,dev,type,size,align,handle)
 #define cgs_free_gpu_mem(dev,handle)		\
 	CGS_CALL(free_gpu_mem,dev,handle)
 #define cgs_gmap_gpu_mem(dev,handle,mcaddr)	\

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

@@ -112,6 +112,9 @@ struct tile_config {
  *
  * @get_max_engine_clock_in_mhz: Retrieves maximum GPU clock in MHz
  *
+ * @alloc_pasid: Allocate a PASID
+ * @free_pasid: Free a PASID
+ *
  * @program_sh_mem_settings: A function that should initiate the memory
  * properties such as main aperture memory type (cache / non cached) and
  * secondary aperture base address, size and memory type.
@@ -160,6 +163,9 @@ struct kfd2kgd_calls {
 
 	uint32_t (*get_max_engine_clock_in_mhz)(struct kgd_dev *kgd);
 
+	int (*alloc_pasid)(unsigned int bits);
+	void (*free_pasid)(unsigned int pasid);
+
 	/* Register access functions */
 	void (*program_sh_mem_settings)(struct kgd_dev *kgd, uint32_t vmid,
 			uint32_t sh_mem_config,	uint32_t sh_mem_ape1_base,

drivers/gpu/drm/amd/include/linux/chash.h

@@ -0,0 +1,366 @@
+/*
+ * Copyright 2017 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef _LINUX_CHASH_H
+#define _LINUX_CHASH_H
+
+#include <linux/types.h>
+#include <linux/hash.h>
+#include <linux/bug.h>
+#include <asm/bitsperlong.h>
+
+#if BITS_PER_LONG == 32
+# define _CHASH_LONG_SHIFT 5
+#elif BITS_PER_LONG == 64
+# define _CHASH_LONG_SHIFT 6
+#else
+# error "Unexpected BITS_PER_LONG"
+#endif
+
+struct __chash_table {
+	u8 bits;
+	u8 key_size;
+	unsigned int value_size;
+	u32 size_mask;
+	unsigned long *occup_bitmap, *valid_bitmap;
+	union {
+		u32 *keys32;
+		u64 *keys64;
+	};
+	u8 *values;
+
+#ifdef CONFIG_CHASH_STATS
+	u64 hits, hits_steps, hits_time_ns;
+	u64 miss, miss_steps, miss_time_ns;
+	u64 relocs, reloc_dist;
+#endif
+};
+
+#define __CHASH_BITMAP_SIZE(bits)				\
+	(((1 << (bits)) + BITS_PER_LONG - 1) / BITS_PER_LONG)
+#define __CHASH_ARRAY_SIZE(bits, size)				\
+	((((size) << (bits)) + sizeof(long) - 1) / sizeof(long))
+
+#define __CHASH_DATA_SIZE(bits, key_size, value_size)	\
+	(__CHASH_BITMAP_SIZE(bits) * 2 +		\
+	 __CHASH_ARRAY_SIZE(bits, key_size) +		\
+	 __CHASH_ARRAY_SIZE(bits, value_size))
+
+#define STRUCT_CHASH_TABLE(bits, key_size, value_size)			\
+	struct {							\
+		struct __chash_table table;				\
+		unsigned long data					\
+			[__CHASH_DATA_SIZE(bits, key_size, value_size)];\
+	}
+
+/**
+ * struct chash_table - Dynamically allocated closed hash table
+ *
+ * Use this struct for dynamically allocated hash tables (using
+ * chash_table_alloc and chash_table_free), where the size is
+ * determined at runtime.
+ */
+struct chash_table {
+	struct __chash_table table;
+	unsigned long *data;
+};
+
+/**
+ * DECLARE_CHASH_TABLE - macro to declare a closed hash table
+ * @table: name of the declared hash table
+ * @bts: Table size will be 2^bits entries
+ * @key_sz: Size of hash keys in bytes, 4 or 8
+ * @val_sz: Size of data values in bytes, can be 0
+ *
+ * This declares the hash table variable with a static size.
+ *
+ * The closed hash table stores key-value pairs with low memory and
+ * lookup overhead. In operation it performs no dynamic memory
+ * management. The data being stored does not require any
+ * list_heads. The hash table performs best with small @val_sz and as
+ * long as some space (about 50%) is left free in the table. But the
+ * table can still work reasonably efficiently even when filled up to
+ * about 90%. If bigger data items need to be stored and looked up,
+ * store the pointer to it as value in the hash table.
+ *
+ * @val_sz may be 0. This can be useful when all the stored
+ * information is contained in the key itself and the fact that it is
+ * in the hash table (or not).
+ */
+#define DECLARE_CHASH_TABLE(table, bts, key_sz, val_sz)		\
+	STRUCT_CHASH_TABLE(bts, key_sz, val_sz) table
+
+#ifdef CONFIG_CHASH_STATS
+#define __CHASH_STATS_INIT(prefix),		\
+		prefix.hits = 0,		\
+		prefix.hits_steps = 0,		\
+		prefix.hits_time_ns = 0,	\
+		prefix.miss = 0,		\
+		prefix.miss_steps = 0,		\
+		prefix.miss_time_ns = 0,	\
+		prefix.relocs = 0,		\
+		prefix.reloc_dist = 0
+#else
+#define __CHASH_STATS_INIT(prefix)
+#endif
+
+#define __CHASH_TABLE_INIT(prefix, data, bts, key_sz, val_sz)	\
+	prefix.bits = (bts),					\
+		prefix.key_size = (key_sz),			\
+		prefix.value_size = (val_sz),			\
+		prefix.size_mask = ((1 << bts) - 1),		\
+		prefix.occup_bitmap = &data[0],			\
+		prefix.valid_bitmap = &data			\
+			[__CHASH_BITMAP_SIZE(bts)],		\
+		prefix.keys64 = (u64 *)&data			\
+			[__CHASH_BITMAP_SIZE(bts) * 2],		\
+		prefix.values = (u8 *)&data			\
+			[__CHASH_BITMAP_SIZE(bts) * 2 +		\
+			 __CHASH_ARRAY_SIZE(bts, key_sz)]	\
+		__CHASH_STATS_INIT(prefix)
+
+/**
+ * DEFINE_CHASH_TABLE - macro to define and initialize a closed hash table
+ * @tbl: name of the declared hash table
+ * @bts: Table size will be 2^bits entries
+ * @key_sz: Size of hash keys in bytes, 4 or 8
+ * @val_sz: Size of data values in bytes, can be 0
+ *
+ * Note: the macro can be used for global and local hash table variables.
+ */
+#define DEFINE_CHASH_TABLE(tbl, bts, key_sz, val_sz)			\
+	DECLARE_CHASH_TABLE(tbl, bts, key_sz, val_sz) = {		\
+		.table = {						\
+			__CHASH_TABLE_INIT(, (tbl).data, bts, key_sz, val_sz) \
+		},							\
+		.data = {0}						\
+	}
+
+/**
+ * INIT_CHASH_TABLE - Initialize a hash table declared by DECLARE_CHASH_TABLE
+ * @tbl: name of the declared hash table
+ * @bts: Table size will be 2^bits entries
+ * @key_sz: Size of hash keys in bytes, 4 or 8
+ * @val_sz: Size of data values in bytes, can be 0
+ */
+#define INIT_CHASH_TABLE(tbl, bts, key_sz, val_sz)			\
+	__CHASH_TABLE_INIT(((tbl).table), (tbl).data, bts, key_sz, val_sz)
+
+int chash_table_alloc(struct chash_table *table, u8 bits, u8 key_size,
+		      unsigned int value_size, gfp_t gfp_mask);
+void chash_table_free(struct chash_table *table);
+
+/**
+ * chash_table_dump_stats - Dump statistics of a closed hash table
+ * @tbl: Pointer to the table structure
+ *
+ * Dumps some performance statistics of the table gathered in operation
+ * in the kernel log using pr_debug. If CONFIG_DYNAMIC_DEBUG is enabled,
+ * user must turn on messages for chash.c (file chash.c +p).
+ */
+#ifdef CONFIG_CHASH_STATS
+#define chash_table_dump_stats(tbl) __chash_table_dump_stats(&(*tbl).table)
+
+void __chash_table_dump_stats(struct __chash_table *table);
+#else
+#define chash_table_dump_stats(tbl)
+#endif
+
+/**
+ * chash_table_reset_stats - Reset statistics of a closed hash table
+ * @tbl: Pointer to the table structure
+ */
+#ifdef CONFIG_CHASH_STATS
+#define chash_table_reset_stats(tbl) __chash_table_reset_stats(&(*tbl).table)
+
+static inline void __chash_table_reset_stats(struct __chash_table *table)
+{
+	(void)table __CHASH_STATS_INIT((*table));
+}
+#else
+#define chash_table_reset_stats(tbl)
+#endif
+
+/**
+ * chash_table_copy_in - Copy a new value into the hash table
+ * @tbl: Pointer to the table structure
+ * @key: Key of the entry to add or update
+ * @value: Pointer to value to copy, may be NULL
+ *
+ * If @key already has an entry, its value is replaced. Otherwise a
+ * new entry is added. If @value is NULL, the value is left unchanged
+ * or uninitialized. Returns 1 if an entry already existed, 0 if a new
+ * entry was added or %-ENOMEM if there was no free space in the
+ * table.
+ */
+#define chash_table_copy_in(tbl, key, value)			\
+	__chash_table_copy_in(&(*tbl).table, key, value)
+
+int __chash_table_copy_in(struct __chash_table *table, u64 key,
+			  const void *value);
+
+/**
+ * chash_table_copy_out - Copy a value out of the hash table
+ * @tbl: Pointer to the table structure
+ * @key: Key of the entry to find
+ * @value: Pointer to value to copy, may be NULL
+ *
+ * If @value is not NULL and the table has a non-0 value_size, the
+ * value at @key is copied to @value. Returns the slot index of the
+ * entry or %-EINVAL if @key was not found.
+ */
+#define chash_table_copy_out(tbl, key, value)			\
+	__chash_table_copy_out(&(*tbl).table, key, value, false)
+
+int __chash_table_copy_out(struct __chash_table *table, u64 key,
+			   void *value, bool remove);
+
+/**
+ * chash_table_remove - Remove an entry from the hash table
+ * @tbl: Pointer to the table structure
+ * @key: Key of the entry to find
+ * @value: Pointer to value to copy, may be NULL
+ *
+ * If @value is not NULL and the table has a non-0 value_size, the
+ * value at @key is copied to @value. The entry is removed from the
+ * table. Returns the slot index of the removed entry or %-EINVAL if
+ * @key was not found.
+ */
+#define chash_table_remove(tbl, key, value)			\
+	__chash_table_copy_out(&(*tbl).table, key, value, true)
+
+/*
+ * Low level iterator API used internally by the above functions.
+ */
+struct chash_iter {
+	struct __chash_table *table;
+	unsigned long mask;
+	int slot;
+};
+
+/**
+ * CHASH_ITER_INIT - Initialize a hash table iterator
+ * @tbl: Pointer to hash table to iterate over
+ * @s: Initial slot number
+ */
+#define CHASH_ITER_INIT(table, s) {			\
+		table,					\
+		1UL << ((s) & (BITS_PER_LONG - 1)),	\
+		s					\
+	}
+/**
+ * CHASH_ITER_SET - Set hash table iterator to new slot
+ * @iter: Iterator
+ * @s: Slot number
+ */
+#define CHASH_ITER_SET(iter, s)					\
+	(iter).mask = 1UL << ((s) & (BITS_PER_LONG - 1)),	\
+	(iter).slot = (s)
+/**
+ * CHASH_ITER_INC - Increment hash table iterator
+ * @table: Hash table to iterate over
+ *
+ * Wraps around at the end.
+ */
+#define CHASH_ITER_INC(iter) do {					\
+		(iter).mask = (iter).mask << 1 |			\
+			(iter).mask >> (BITS_PER_LONG - 1);		\
+		(iter).slot = ((iter).slot + 1) & (iter).table->size_mask; \
+	} while (0)
+
+static inline bool chash_iter_is_valid(const struct chash_iter iter)
+{
+	BUG_ON((unsigned)iter.slot >= (1 << iter.table->bits));
+	return !!(iter.table->valid_bitmap[iter.slot >> _CHASH_LONG_SHIFT] &
+		  iter.mask);
+}
+static inline bool chash_iter_is_empty(const struct chash_iter iter)
+{
+	BUG_ON((unsigned)iter.slot >= (1 << iter.table->bits));
+	return !(iter.table->occup_bitmap[iter.slot >> _CHASH_LONG_SHIFT] &
+		 iter.mask);
+}
+
+static inline void chash_iter_set_valid(const struct chash_iter iter)
+{
+	BUG_ON((unsigned)iter.slot >= (1 << iter.table->bits));
+	iter.table->valid_bitmap[iter.slot >> _CHASH_LONG_SHIFT] |= iter.mask;
+	iter.table->occup_bitmap[iter.slot >> _CHASH_LONG_SHIFT] |= iter.mask;
+}
+static inline void chash_iter_set_invalid(const struct chash_iter iter)
+{
+	BUG_ON((unsigned)iter.slot >= (1 << iter.table->bits));
+	iter.table->valid_bitmap[iter.slot >> _CHASH_LONG_SHIFT] &= ~iter.mask;
+}
+static inline void chash_iter_set_empty(const struct chash_iter iter)
+{
+	BUG_ON((unsigned)iter.slot >= (1 << iter.table->bits));
+	iter.table->occup_bitmap[iter.slot >> _CHASH_LONG_SHIFT] &= ~iter.mask;
+}
+
+static inline u32 chash_iter_key32(const struct chash_iter iter)
+{
+	BUG_ON(iter.table->key_size != 4);
+	BUG_ON((unsigned)iter.slot >= (1 << iter.table->bits));
+	return iter.table->keys32[iter.slot];
+}
+static inline u64 chash_iter_key64(const struct chash_iter iter)
+{
+	BUG_ON(iter.table->key_size != 8);
+	BUG_ON((unsigned)iter.slot >= (1 << iter.table->bits));
+	return iter.table->keys64[iter.slot];
+}
+static inline u64 chash_iter_key(const struct chash_iter iter)
+{
+	BUG_ON((unsigned)iter.slot >= (1 << iter.table->bits));
+	return (iter.table->key_size == 4) ?
+		iter.table->keys32[iter.slot] : iter.table->keys64[iter.slot];
+}
+
+static inline u32 chash_iter_hash32(const struct chash_iter iter)
+{
+	BUG_ON(iter.table->key_size != 4);
+	return hash_32(chash_iter_key32(iter), iter.table->bits);
+}
+
+static inline u32 chash_iter_hash64(const struct chash_iter iter)
+{
+	BUG_ON(iter.table->key_size != 8);
+	return hash_64(chash_iter_key64(iter), iter.table->bits);
+}
+
+static inline u32 chash_iter_hash(const struct chash_iter iter)
+{
+	return (iter.table->key_size == 4) ?
+		hash_32(chash_iter_key32(iter), iter.table->bits) :
+		hash_64(chash_iter_key64(iter), iter.table->bits);
+}
+
+static inline void *chash_iter_value(const struct chash_iter iter)
+{
+	BUG_ON((unsigned)iter.slot >= (1 << iter.table->bits));
+	return iter.table->values +
+		((unsigned long)iter.slot * iter.table->value_size);
+}
+
+#endif /* _LINUX_CHASH_H */

drivers/gpu/drm/amd/include/v9_structs.h

@@ -284,8 +284,8 @@ struct v9_mqd {
 	uint32_t gds_save_mask_hi;
 	uint32_t ctx_save_base_addr_lo;
 	uint32_t ctx_save_base_addr_hi;
-	uint32_t reserved_126;
-	uint32_t reserved_127;
+	uint32_t dynamic_cu_mask_addr_lo;
+	uint32_t dynamic_cu_mask_addr_hi;
 	uint32_t cp_mqd_base_addr_lo;
 	uint32_t cp_mqd_base_addr_hi;
 	uint32_t cp_hqd_active;
@@ -672,6 +672,14 @@ struct v9_mqd {
 	uint32_t reserved_511;
 };
 
+struct v9_mqd_allocation {
+	struct v9_mqd mqd;
+	uint32_t wptr_poll_mem;
+	uint32_t rptr_report_mem;
+	uint32_t dynamic_cu_mask;
+	uint32_t dynamic_rb_mask;
+};
+
 /* from vega10 all CSA format is shifted to chain ib compatible mode */
 struct v9_ce_ib_state {
     /* section of non chained ib part */

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

@@ -423,265 +423,6 @@ struct vi_mqd_allocation {
 	uint32_t dynamic_rb_mask;
 };
 
-struct cz_mqd {
-	uint32_t header;
-	uint32_t compute_dispatch_initiator;
-	uint32_t compute_dim_x;
-	uint32_t compute_dim_y;
-	uint32_t compute_dim_z;
-	uint32_t compute_start_x;
-	uint32_t compute_start_y;
-	uint32_t compute_start_z;
-	uint32_t compute_num_thread_x;
-	uint32_t compute_num_thread_y;
-	uint32_t compute_num_thread_z;
-	uint32_t compute_pipelinestat_enable;
-	uint32_t compute_perfcount_enable;
-	uint32_t compute_pgm_lo;
-	uint32_t compute_pgm_hi;
-	uint32_t compute_tba_lo;
-	uint32_t compute_tba_hi;
-	uint32_t compute_tma_lo;
-	uint32_t compute_tma_hi;
-	uint32_t compute_pgm_rsrc1;
-	uint32_t compute_pgm_rsrc2;
-	uint32_t compute_vmid;
-	uint32_t compute_resource_limits;
-	uint32_t compute_static_thread_mgmt_se0;
-	uint32_t compute_static_thread_mgmt_se1;
-	uint32_t compute_tmpring_size;
-	uint32_t compute_static_thread_mgmt_se2;
-	uint32_t compute_static_thread_mgmt_se3;
-	uint32_t compute_restart_x;
-	uint32_t compute_restart_y;
-	uint32_t compute_restart_z;
-	uint32_t compute_thread_trace_enable;
-	uint32_t compute_misc_reserved;
-	uint32_t compute_dispatch_id;
-	uint32_t compute_threadgroup_id;
-	uint32_t compute_relaunch;
-	uint32_t compute_wave_restore_addr_lo;
-	uint32_t compute_wave_restore_addr_hi;
-	uint32_t compute_wave_restore_control;
-	uint32_t reserved_39;
-	uint32_t reserved_40;
-	uint32_t reserved_41;
-	uint32_t reserved_42;
-	uint32_t reserved_43;
-	uint32_t reserved_44;
-	uint32_t reserved_45;
-	uint32_t reserved_46;
-	uint32_t reserved_47;
-	uint32_t reserved_48;
-	uint32_t reserved_49;
-	uint32_t reserved_50;
-	uint32_t reserved_51;
-	uint32_t reserved_52;
-	uint32_t reserved_53;
-	uint32_t reserved_54;
-	uint32_t reserved_55;
-	uint32_t reserved_56;
-	uint32_t reserved_57;
-	uint32_t reserved_58;
-	uint32_t reserved_59;
-	uint32_t reserved_60;
-	uint32_t reserved_61;
-	uint32_t reserved_62;
-	uint32_t reserved_63;
-	uint32_t reserved_64;
-	uint32_t compute_user_data_0;
-	uint32_t compute_user_data_1;
-	uint32_t compute_user_data_2;
-	uint32_t compute_user_data_3;
-	uint32_t compute_user_data_4;
-	uint32_t compute_user_data_5;
-	uint32_t compute_user_data_6;
-	uint32_t compute_user_data_7;
-	uint32_t compute_user_data_8;
-	uint32_t compute_user_data_9;
-	uint32_t compute_user_data_10;
-	uint32_t compute_user_data_11;
-	uint32_t compute_user_data_12;
-	uint32_t compute_user_data_13;
-	uint32_t compute_user_data_14;
-	uint32_t compute_user_data_15;
-	uint32_t cp_compute_csinvoc_count_lo;
-	uint32_t cp_compute_csinvoc_count_hi;
-	uint32_t reserved_83;
-	uint32_t reserved_84;
-	uint32_t reserved_85;
-	uint32_t cp_mqd_query_time_lo;
-	uint32_t cp_mqd_query_time_hi;
-	uint32_t cp_mqd_connect_start_time_lo;
-	uint32_t cp_mqd_connect_start_time_hi;
-	uint32_t cp_mqd_connect_end_time_lo;
-	uint32_t cp_mqd_connect_end_time_hi;
-	uint32_t cp_mqd_connect_end_wf_count;
-	uint32_t cp_mqd_connect_end_pq_rptr;
-	uint32_t cp_mqd_connect_end_pq_wptr;
-	uint32_t cp_mqd_connect_end_ib_rptr;
-	uint32_t reserved_96;
-	uint32_t reserved_97;
-	uint32_t cp_mqd_save_start_time_lo;
-	uint32_t cp_mqd_save_start_time_hi;
-	uint32_t cp_mqd_save_end_time_lo;
-	uint32_t cp_mqd_save_end_time_hi;
-	uint32_t cp_mqd_restore_start_time_lo;
-	uint32_t cp_mqd_restore_start_time_hi;
-	uint32_t cp_mqd_restore_end_time_lo;
-	uint32_t cp_mqd_restore_end_time_hi;
-	uint32_t reserved_106;
-	uint32_t reserved_107;
-	uint32_t gds_cs_ctxsw_cnt0;
-	uint32_t gds_cs_ctxsw_cnt1;
-	uint32_t gds_cs_ctxsw_cnt2;
-	uint32_t gds_cs_ctxsw_cnt3;
-	uint32_t reserved_112;
-	uint32_t reserved_113;
-	uint32_t cp_pq_exe_status_lo;
-	uint32_t cp_pq_exe_status_hi;
-	uint32_t cp_packet_id_lo;
-	uint32_t cp_packet_id_hi;
-	uint32_t cp_packet_exe_status_lo;
-	uint32_t cp_packet_exe_status_hi;
-	uint32_t gds_save_base_addr_lo;
-	uint32_t gds_save_base_addr_hi;
-	uint32_t gds_save_mask_lo;
-	uint32_t gds_save_mask_hi;
-	uint32_t ctx_save_base_addr_lo;
-	uint32_t ctx_save_base_addr_hi;
-	uint32_t reserved_126;
-	uint32_t reserved_127;
-	uint32_t cp_mqd_base_addr_lo;
-	uint32_t cp_mqd_base_addr_hi;
-	uint32_t cp_hqd_active;
-	uint32_t cp_hqd_vmid;
-	uint32_t cp_hqd_persistent_state;
-	uint32_t cp_hqd_pipe_priority;
-	uint32_t cp_hqd_queue_priority;
-	uint32_t cp_hqd_quantum;
-	uint32_t cp_hqd_pq_base_lo;
-	uint32_t cp_hqd_pq_base_hi;
-	uint32_t cp_hqd_pq_rptr;
-	uint32_t cp_hqd_pq_rptr_report_addr_lo;
-	uint32_t cp_hqd_pq_rptr_report_addr_hi;
-	uint32_t cp_hqd_pq_wptr_poll_addr_lo;
-	uint32_t cp_hqd_pq_wptr_poll_addr_hi;
-	uint32_t cp_hqd_pq_doorbell_control;
-	uint32_t cp_hqd_pq_wptr;
-	uint32_t cp_hqd_pq_control;
-	uint32_t cp_hqd_ib_base_addr_lo;
-	uint32_t cp_hqd_ib_base_addr_hi;
-	uint32_t cp_hqd_ib_rptr;
-	uint32_t cp_hqd_ib_control;
-	uint32_t cp_hqd_iq_timer;
-	uint32_t cp_hqd_iq_rptr;
-	uint32_t cp_hqd_dequeue_request;
-	uint32_t cp_hqd_dma_offload;
-	uint32_t cp_hqd_sema_cmd;
-	uint32_t cp_hqd_msg_type;
-	uint32_t cp_hqd_atomic0_preop_lo;
-	uint32_t cp_hqd_atomic0_preop_hi;
-	uint32_t cp_hqd_atomic1_preop_lo;
-	uint32_t cp_hqd_atomic1_preop_hi;
-	uint32_t cp_hqd_hq_status0;
-	uint32_t cp_hqd_hq_control0;
-	uint32_t cp_mqd_control;
-	uint32_t cp_hqd_hq_status1;
-	uint32_t cp_hqd_hq_control1;
-	uint32_t cp_hqd_eop_base_addr_lo;
-	uint32_t cp_hqd_eop_base_addr_hi;
-	uint32_t cp_hqd_eop_control;
-	uint32_t cp_hqd_eop_rptr;
-	uint32_t cp_hqd_eop_wptr;
-	uint32_t cp_hqd_eop_done_events;
-	uint32_t cp_hqd_ctx_save_base_addr_lo;
-	uint32_t cp_hqd_ctx_save_base_addr_hi;
-	uint32_t cp_hqd_ctx_save_control;
-	uint32_t cp_hqd_cntl_stack_offset;
-	uint32_t cp_hqd_cntl_stack_size;
-	uint32_t cp_hqd_wg_state_offset;
-	uint32_t cp_hqd_ctx_save_size;
-	uint32_t cp_hqd_gds_resource_state;
-	uint32_t cp_hqd_error;
-	uint32_t cp_hqd_eop_wptr_mem;
-	uint32_t cp_hqd_eop_dones;
-	uint32_t reserved_182;
-	uint32_t reserved_183;
-	uint32_t reserved_184;
-	uint32_t reserved_185;
-	uint32_t reserved_186;
-	uint32_t reserved_187;
-	uint32_t reserved_188;
-	uint32_t reserved_189;
-	uint32_t reserved_190;
-	uint32_t reserved_191;
-	uint32_t iqtimer_pkt_header;
-	uint32_t iqtimer_pkt_dw0;
-	uint32_t iqtimer_pkt_dw1;
-	uint32_t iqtimer_pkt_dw2;
-	uint32_t iqtimer_pkt_dw3;
-	uint32_t iqtimer_pkt_dw4;
-	uint32_t iqtimer_pkt_dw5;
-	uint32_t iqtimer_pkt_dw6;
-	uint32_t iqtimer_pkt_dw7;
-	uint32_t iqtimer_pkt_dw8;
-	uint32_t iqtimer_pkt_dw9;
-	uint32_t iqtimer_pkt_dw10;
-	uint32_t iqtimer_pkt_dw11;
-	uint32_t iqtimer_pkt_dw12;
-	uint32_t iqtimer_pkt_dw13;
-	uint32_t iqtimer_pkt_dw14;
-	uint32_t iqtimer_pkt_dw15;
-	uint32_t iqtimer_pkt_dw16;
-	uint32_t iqtimer_pkt_dw17;
-	uint32_t iqtimer_pkt_dw18;
-	uint32_t iqtimer_pkt_dw19;
-	uint32_t iqtimer_pkt_dw20;
-	uint32_t iqtimer_pkt_dw21;
-	uint32_t iqtimer_pkt_dw22;
-	uint32_t iqtimer_pkt_dw23;
-	uint32_t iqtimer_pkt_dw24;
-	uint32_t iqtimer_pkt_dw25;
-	uint32_t iqtimer_pkt_dw26;
-	uint32_t iqtimer_pkt_dw27;
-	uint32_t iqtimer_pkt_dw28;
-	uint32_t iqtimer_pkt_dw29;
-	uint32_t iqtimer_pkt_dw30;
-	uint32_t iqtimer_pkt_dw31;
-	uint32_t reserved_225;
-	uint32_t reserved_226;
-	uint32_t reserved_227;
-	uint32_t set_resources_header;
-	uint32_t set_resources_dw1;
-	uint32_t set_resources_dw2;
-	uint32_t set_resources_dw3;
-	uint32_t set_resources_dw4;
-	uint32_t set_resources_dw5;
-	uint32_t set_resources_dw6;
-	uint32_t set_resources_dw7;
-	uint32_t reserved_236;
-	uint32_t reserved_237;
-	uint32_t reserved_238;
-	uint32_t reserved_239;
-	uint32_t queue_doorbell_id0;
-	uint32_t queue_doorbell_id1;
-	uint32_t queue_doorbell_id2;
-	uint32_t queue_doorbell_id3;
-	uint32_t queue_doorbell_id4;
-	uint32_t queue_doorbell_id5;
-	uint32_t queue_doorbell_id6;
-	uint32_t queue_doorbell_id7;
-	uint32_t queue_doorbell_id8;
-	uint32_t queue_doorbell_id9;
-	uint32_t queue_doorbell_id10;
-	uint32_t queue_doorbell_id11;
-	uint32_t queue_doorbell_id12;
-	uint32_t queue_doorbell_id13;
-	uint32_t queue_doorbell_id14;
-	uint32_t queue_doorbell_id15;
-};
-
 struct vi_ce_ib_state {
 	uint32_t    ce_ib_completion_status;
 	uint32_t    ce_constegnine_count;

drivers/gpu/drm/amd/lib/Kconfig

@@ -0,0 +1,27 @@
+menu "AMD Library routines"
+
+#
+# Closed hash table
+#
+config CHASH
+	tristate "Closed hash table"
+	help
+	 Statically sized closed hash table implementation with low
+	 memory and CPU overhead.
+
+config CHASH_STATS
+	bool "Closed hash table performance statistics"
+	depends on CHASH
+	default n
+	help
+	 Enable collection of performance statistics for closed hash tables.
+
+config CHASH_SELFTEST
+	bool "Closed hash table self test"
+	depends on CHASH
+	default n
+	help
+	 Runs a selftest during module load. Several module parameters
+	 are available to modify the behaviour of the test.
+
+endmenu

drivers/gpu/drm/amd/lib/Makefile

@@ -0,0 +1,11 @@
+#
+# Makefile for AMD library routines, which are used by AMD driver
+# components.
+#
+# This is for common library routines that can be shared between AMD
+# driver components or later moved to kernel/lib for sharing with
+# other drivers.
+
+ccflags-y := -I$(src)/../include
+
+obj-$(CONFIG_CHASH) += chash.o

drivers/gpu/drm/amd/lib/chash.c

@@ -0,0 +1,638 @@
+/*
+ * Copyright 2017 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/types.h>
+#include <linux/hash.h>
+#include <linux/bug.h>
+#include <linux/slab.h>
+#include <linux/module.h>
+#include <linux/sched/clock.h>
+#include <asm/div64.h>
+#include <linux/chash.h>
+
+/**
+ * chash_table_alloc - Allocate closed hash table
+ * @table: Pointer to the table structure
+ * @bits: Table size will be 2^bits entries
+ * @key_size: Size of hash keys in bytes, 4 or 8
+ * @value_size: Size of data values in bytes, can be 0
+ */
+int chash_table_alloc(struct chash_table *table, u8 bits, u8 key_size,
+		      unsigned int value_size, gfp_t gfp_mask)
+{
+	if (bits > 31)
+		return -EINVAL;
+
+	if (key_size != 4 && key_size != 8)
+		return -EINVAL;
+
+	table->data = kcalloc(__CHASH_DATA_SIZE(bits, key_size, value_size),
+		       sizeof(long), gfp_mask);
+	if (!table->data)
+		return -ENOMEM;
+
+	__CHASH_TABLE_INIT(table->table, table->data,
+			   bits, key_size, value_size);
+
+	return 0;
+}
+EXPORT_SYMBOL(chash_table_alloc);
+
+/**
+ * chash_table_free - Free closed hash table
+ * @table: Pointer to the table structure
+ */
+void chash_table_free(struct chash_table *table)
+{
+	kfree(table->data);
+}
+EXPORT_SYMBOL(chash_table_free);
+
+#ifdef CONFIG_CHASH_STATS
+
+#define DIV_FRAC(nom, denom, quot, frac, frac_digits) do {		\
+		u64 __nom = (nom);					\
+		u64 __denom = (denom);					\
+		u64 __quot, __frac;					\
+		u32 __rem;						\
+									\
+		while (__denom >> 32) {					\
+			__nom   >>= 1;					\
+			__denom >>= 1;					\
+		}							\
+		__quot = __nom;						\
+		__rem  = do_div(__quot, __denom);			\
+		__frac = __rem * (frac_digits) + (__denom >> 1);	\
+		do_div(__frac, __denom);				\
+		(quot) = __quot;					\
+		(frac) = __frac;					\
+	} while (0)
+
+void __chash_table_dump_stats(struct __chash_table *table)
+{
+	struct chash_iter iter = CHASH_ITER_INIT(table, 0);
+	u32 filled = 0, empty = 0, tombstones = 0;
+	u64 quot1, quot2;
+	u32 frac1, frac2;
+
+	do {
+		if (chash_iter_is_valid(iter))
+			filled++;
+		else if (chash_iter_is_empty(iter))
+			empty++;
+		else
+			tombstones++;
+		CHASH_ITER_INC(iter);
+	} while (iter.slot);
+
+	pr_debug("chash: key size %u, value size %u\n",
+		 table->key_size, table->value_size);
+	pr_debug("  Slots total/filled/empty/tombstones: %u / %u / %u / %u\n",
+		 1 << table->bits, filled, empty, tombstones);
+	if (table->hits > 0) {
+		DIV_FRAC(table->hits_steps, table->hits, quot1, frac1, 1000);
+		DIV_FRAC(table->hits * 1000, table->hits_time_ns,
+			 quot2, frac2, 1000);
+	} else {
+		quot1 = quot2 = 0;
+		frac1 = frac2 = 0;
+	}
+	pr_debug("  Hits   (avg.cost, rate): %llu (%llu.%03u, %llu.%03u M/s)\n",
+		 table->hits, quot1, frac1, quot2, frac2);
+	if (table->miss > 0) {
+		DIV_FRAC(table->miss_steps, table->miss, quot1, frac1, 1000);
+		DIV_FRAC(table->miss * 1000, table->miss_time_ns,
+			 quot2, frac2, 1000);
+	} else {
+		quot1 = quot2 = 0;
+		frac1 = frac2 = 0;
+	}
+	pr_debug("  Misses (avg.cost, rate): %llu (%llu.%03u, %llu.%03u M/s)\n",
+		 table->miss, quot1, frac1, quot2, frac2);
+	if (table->hits + table->miss > 0) {
+		DIV_FRAC(table->hits_steps + table->miss_steps,
+			 table->hits + table->miss, quot1, frac1, 1000);
+		DIV_FRAC((table->hits + table->miss) * 1000,
+			 (table->hits_time_ns + table->miss_time_ns),
+			 quot2, frac2, 1000);
+	} else {
+		quot1 = quot2 = 0;
+		frac1 = frac2 = 0;
+	}
+	pr_debug("  Total  (avg.cost, rate): %llu (%llu.%03u, %llu.%03u M/s)\n",
+		 table->hits + table->miss, quot1, frac1, quot2, frac2);
+	if (table->relocs > 0) {
+		DIV_FRAC(table->hits + table->miss, table->relocs,
+			 quot1, frac1, 1000);
+		DIV_FRAC(table->reloc_dist, table->relocs, quot2, frac2, 1000);
+		pr_debug("  Relocations (freq, avg.dist): %llu (1:%llu.%03u, %llu.%03u)\n",
+			 table->relocs, quot1, frac1, quot2, frac2);
+	} else {
+		pr_debug("  No relocations\n");
+	}
+}
+EXPORT_SYMBOL(__chash_table_dump_stats);
+
+#undef DIV_FRAC
+#endif
+
+#define CHASH_INC(table, a) ((a) = ((a) + 1) & (table)->size_mask)
+#define CHASH_ADD(table, a, b) (((a) + (b)) & (table)->size_mask)
+#define CHASH_SUB(table, a, b) (((a) - (b)) & (table)->size_mask)
+#define CHASH_IN_RANGE(table, slot, first, last) \
+	(CHASH_SUB(table, slot, first) <= CHASH_SUB(table, last, first))
+
+/*#define CHASH_DEBUG Uncomment this to enable verbose debug output*/
+#ifdef CHASH_DEBUG
+static void chash_table_dump(struct __chash_table *table)
+{
+	struct chash_iter iter = CHASH_ITER_INIT(table, 0);
+
+	do {
+		if ((iter.slot & 3) == 0)
+			pr_debug("%04x: ", iter.slot);
+
+		if (chash_iter_is_valid(iter))
+			pr_debug("[%016llx] ", chash_iter_key(iter));
+		else if (chash_iter_is_empty(iter))
+			pr_debug("[    <empty>     ] ");
+		else
+			pr_debug("[  <tombstone>   ] ");
+
+		if ((iter.slot & 3) == 3)
+			pr_debug("\n");
+
+		CHASH_ITER_INC(iter);
+	} while (iter.slot);
+
+	if ((iter.slot & 3) != 0)
+		pr_debug("\n");
+}
+
+static int chash_table_check(struct __chash_table *table)
+{
+	u32 hash;
+	struct chash_iter iter = CHASH_ITER_INIT(table, 0);
+	struct chash_iter cur = CHASH_ITER_INIT(table, 0);
+
+	do {
+		if (!chash_iter_is_valid(iter)) {
+			CHASH_ITER_INC(iter);
+			continue;
+		}
+
+		hash = chash_iter_hash(iter);
+		CHASH_ITER_SET(cur, hash);
+		while (cur.slot != iter.slot) {
+			if (chash_iter_is_empty(cur)) {
+				pr_err("Path to element at %x with hash %x broken at slot %x\n",
+				       iter.slot, hash, cur.slot);
+				chash_table_dump(table);
+				return -EINVAL;
+			}
+			CHASH_ITER_INC(cur);
+		}
+
+		CHASH_ITER_INC(iter);
+	} while (iter.slot);
+
+	return 0;
+}
+#endif
+
+static void chash_iter_relocate(struct chash_iter dst, struct chash_iter src)
+{
+	BUG_ON(src.table == dst.table && src.slot == dst.slot);
+	BUG_ON(src.table->key_size != src.table->key_size);
+	BUG_ON(src.table->value_size != src.table->value_size);
+
+	if (dst.table->key_size == 4)
+		dst.table->keys32[dst.slot] = src.table->keys32[src.slot];
+	else
+		dst.table->keys64[dst.slot] = src.table->keys64[src.slot];
+
+	if (dst.table->value_size)
+		memcpy(chash_iter_value(dst), chash_iter_value(src),
+		       dst.table->value_size);
+
+	chash_iter_set_valid(dst);
+	chash_iter_set_invalid(src);
+
+#ifdef CONFIG_CHASH_STATS
+	if (src.table == dst.table) {
+		dst.table->relocs++;
+		dst.table->reloc_dist +=
+			CHASH_SUB(dst.table, src.slot, dst.slot);
+	}
+#endif
+}
+
+/**
+ * __chash_table_find - Helper for looking up a hash table entry
+ * @iter: Pointer to hash table iterator
+ * @key: Key of the entry to find
+ * @for_removal: set to true if the element will be removed soon
+ *
+ * Searches for an entry in the hash table with a given key. iter must
+ * be initialized by the caller to point to the home position of the
+ * hypothetical entry, i.e. it must be initialized with the hash table
+ * and the key's hash as the initial slot for the search.
+ *
+ * This function also does some local clean-up to speed up future
+ * look-ups by relocating entries to better slots and removing
+ * tombstones that are no longer needed.
+ *
+ * If @for_removal is true, the function avoids relocating the entry
+ * that is being returned.
+ *
+ * Returns 0 if the search is successful. In this case iter is updated
+ * to point to the found entry. Otherwise %-EINVAL is returned and the
+ * iter is updated to point to the first available slot for the given
+ * key. If the table is full, the slot is set to -1.
+ */
+static int chash_table_find(struct chash_iter *iter, u64 key,
+			    bool for_removal)
+{
+#ifdef CONFIG_CHASH_STATS
+	u64 ts1 = local_clock();
+#endif
+	u32 hash = iter->slot;
+	struct chash_iter first_redundant = CHASH_ITER_INIT(iter->table, -1);
+	int first_avail = (for_removal ? -2 : -1);
+
+	while (!chash_iter_is_valid(*iter) || chash_iter_key(*iter) != key) {
+		if (chash_iter_is_empty(*iter)) {
+			/* Found an empty slot, which ends the
+			 * search. Clean up any preceding tombstones
+			 * that are no longer needed because they lead
+			 * to no-where
+			 */
+			if ((int)first_redundant.slot < 0)
+				goto not_found;
+			while (first_redundant.slot != iter->slot) {
+				if (!chash_iter_is_valid(first_redundant))
+					chash_iter_set_empty(first_redundant);
+				CHASH_ITER_INC(first_redundant);
+			}
+#ifdef CHASH_DEBUG
+			chash_table_check(iter->table);
+#endif
+			goto not_found;
+		} else if (!chash_iter_is_valid(*iter)) {
+			/* Found a tombstone. Remember it as candidate
+			 * for relocating the entry we're looking for
+			 * or for adding a new entry with the given key
+			 */
+			if (first_avail == -1)
+				first_avail = iter->slot;
+			/* Or mark it as the start of a series of
+			 * potentially redundant tombstones
+			 */
+			else if (first_redundant.slot == -1)
+				CHASH_ITER_SET(first_redundant, iter->slot);
+		} else if (first_redundant.slot >= 0) {
+			/* Found a valid, occupied slot with a
+			 * preceding series of tombstones. Relocate it
+			 * to a better position that no longer depends
+			 * on those tombstones
+			 */
+			u32 cur_hash = chash_iter_hash(*iter);
+
+			if (!CHASH_IN_RANGE(iter->table, cur_hash,
+					    first_redundant.slot + 1,
+					    iter->slot)) {
+				/* This entry has a hash at or before
+				 * the first tombstone we found. We
+				 * can relocate it to that tombstone
+				 * and advance to the next tombstone
+				 */
+				chash_iter_relocate(first_redundant, *iter);
+				do {
+					CHASH_ITER_INC(first_redundant);
+				} while (chash_iter_is_valid(first_redundant));
+			} else if (cur_hash != iter->slot) {
+				/* Relocate entry to its home position
+				 * or as close as possible so it no
+				 * longer depends on any preceding
+				 * tombstones
+				 */
+				struct chash_iter new_iter =
+					CHASH_ITER_INIT(iter->table, cur_hash);
+
+				while (new_iter.slot != iter->slot &&
+				       chash_iter_is_valid(new_iter))
+					CHASH_ITER_INC(new_iter);
+
+				if (new_iter.slot != iter->slot)
+					chash_iter_relocate(new_iter, *iter);
+			}
+		}
+
+		CHASH_ITER_INC(*iter);
+		if (iter->slot == hash) {
+			iter->slot = -1;
+			goto not_found;
+		}
+	}
+
+#ifdef CONFIG_CHASH_STATS
+	iter->table->hits++;
+	iter->table->hits_steps += CHASH_SUB(iter->table, iter->slot, hash) + 1;
+#endif
+
+	if (first_avail >= 0) {
+		CHASH_ITER_SET(first_redundant, first_avail);
+		chash_iter_relocate(first_redundant, *iter);
+		iter->slot = first_redundant.slot;
+		iter->mask = first_redundant.mask;
+	}
+
+#ifdef CONFIG_CHASH_STATS
+	iter->table->hits_time_ns += local_clock() - ts1;
+#endif
+
+	return 0;
+
+not_found:
+#ifdef CONFIG_CHASH_STATS
+	iter->table->miss++;
+	iter->table->miss_steps += (iter->slot < 0) ?
+		(1 << iter->table->bits) :
+		CHASH_SUB(iter->table, iter->slot, hash) + 1;
+#endif
+
+	if (first_avail >= 0)
+		CHASH_ITER_SET(*iter, first_avail);
+
+#ifdef CONFIG_CHASH_STATS
+	iter->table->miss_time_ns += local_clock() - ts1;
+#endif
+
+	return -EINVAL;
+}
+
+int __chash_table_copy_in(struct __chash_table *table, u64 key,
+			  const void *value)
+{
+	u32 hash = (table->key_size == 4) ?
+		hash_32(key, table->bits) : hash_64(key, table->bits);
+	struct chash_iter iter = CHASH_ITER_INIT(table, hash);
+	int r = chash_table_find(&iter, key, false);
+
+	/* Found an existing entry */
+	if (!r) {
+		if (value && table->value_size)
+			memcpy(chash_iter_value(iter), value,
+			       table->value_size);
+		return 1;
+	}
+
+	/* Is there a place to add a new entry? */
+	if (iter.slot < 0) {
+		pr_err("Hash table overflow\n");
+		return -ENOMEM;
+	}
+
+	chash_iter_set_valid(iter);
+
+	if (table->key_size == 4)
+		table->keys32[iter.slot] = key;
+	else
+		table->keys64[iter.slot] = key;
+	if (value && table->value_size)
+		memcpy(chash_iter_value(iter), value, table->value_size);
+
+	return 0;
+}
+EXPORT_SYMBOL(__chash_table_copy_in);
+
+int __chash_table_copy_out(struct __chash_table *table, u64 key,
+			   void *value, bool remove)
+{
+	u32 hash = (table->key_size == 4) ?
+		hash_32(key, table->bits) : hash_64(key, table->bits);
+	struct chash_iter iter = CHASH_ITER_INIT(table, hash);
+	int r = chash_table_find(&iter, key, remove);
+
+	if (r < 0)
+		return r;
+
+	if (value && table->value_size)
+		memcpy(value, chash_iter_value(iter), table->value_size);
+
+	if (remove)
+		chash_iter_set_invalid(iter);
+
+	return iter.slot;
+}
+EXPORT_SYMBOL(__chash_table_copy_out);
+
+#ifdef CONFIG_CHASH_SELFTEST
+/**
+ * chash_self_test - Run a self-test of the hash table implementation
+ * @bits: Table size will be 2^bits entries
+ * @key_size: Size of hash keys in bytes, 4 or 8
+ * @min_fill: Minimum fill level during the test
+ * @max_fill: Maximum fill level during the test
+ * @iterations: Number of test iterations
+ *
+ * The test adds and removes entries from a hash table, cycling the
+ * fill level between min_fill and max_fill entries. Also tests lookup
+ * and value retrieval.
+ */
+static int __init chash_self_test(u8 bits, u8 key_size,
+				  int min_fill, int max_fill,
+				  u64 iterations)
+{
+	struct chash_table table;
+	int ret;
+	u64 add_count, rmv_count;
+	u64 value;
+
+	if (key_size == 4 && iterations > 0xffffffff)
+		return -EINVAL;
+	if (min_fill >= max_fill)
+		return -EINVAL;
+
+	ret = chash_table_alloc(&table, bits, key_size, sizeof(u64),
+				GFP_KERNEL);
+	if (ret) {
+		pr_err("chash_table_alloc failed: %d\n", ret);
+		return ret;
+	}
+
+	for (add_count = 0, rmv_count = 0; add_count < iterations;
+	     add_count++) {
+		/* When we hit the max_fill level, remove entries down
+		 * to min_fill
+		 */
+		if (add_count - rmv_count == max_fill) {
+			u64 find_count = rmv_count;
+
+			/* First try to find all entries that we're
+			 * about to remove, confirm their value, test
+			 * writing them back a second time.
+			 */
+			for (; add_count - find_count > min_fill;
+			     find_count++) {
+				ret = chash_table_copy_out(&table, find_count,
+							   &value);
+				if (ret < 0) {
+					pr_err("chash_table_copy_out failed: %d\n",
+					       ret);
+					goto out;
+				}
+				if (value != ~find_count) {
+					pr_err("Wrong value retrieved for key 0x%llx, expected 0x%llx got 0x%llx\n",
+					       find_count, ~find_count, value);
+#ifdef CHASH_DEBUG
+					chash_table_dump(&table.table);
+#endif
+					ret = -EFAULT;
+					goto out;
+				}
+				ret = chash_table_copy_in(&table, find_count,
+							  &value);
+				if (ret != 1) {
+					pr_err("copy_in second time returned %d, expected 1\n",
+					       ret);
+					ret = -EFAULT;
+					goto out;
+				}
+			}
+			/* Remove them until we hit min_fill level */
+			for (; add_count - rmv_count > min_fill; rmv_count++) {
+				ret = chash_table_remove(&table, rmv_count,
+							 NULL);
+				if (ret < 0) {
+					pr_err("chash_table_remove failed: %d\n",
+					       ret);
+					goto out;
+				}
+			}
+		}
+
+		/* Add a new value */
+		value = ~add_count;
+		ret = chash_table_copy_in(&table, add_count, &value);
+		if (ret != 0) {
+			pr_err("copy_in first time returned %d, expected 0\n",
+			       ret);
+			ret = -EFAULT;
+			goto out;
+		}
+	}
+
+	chash_table_dump_stats(&table);
+	chash_table_reset_stats(&table);
+
+out:
+	chash_table_free(&table);
+	return ret;
+}
+
+static unsigned int chash_test_bits = 10;
+MODULE_PARM_DESC(test_bits,
+		 "Selftest number of hash bits ([4..20], default=10)");
+module_param_named(test_bits, chash_test_bits, uint, 0444);
+
+static unsigned int chash_test_keysize = 8;
+MODULE_PARM_DESC(test_keysize, "Selftest keysize (4 or 8, default=8)");
+module_param_named(test_keysize, chash_test_keysize, uint, 0444);
+
+static unsigned int chash_test_minfill;
+MODULE_PARM_DESC(test_minfill, "Selftest minimum #entries (default=50%)");
+module_param_named(test_minfill, chash_test_minfill, uint, 0444);
+
+static unsigned int chash_test_maxfill;
+MODULE_PARM_DESC(test_maxfill, "Selftest maximum #entries (default=80%)");
+module_param_named(test_maxfill, chash_test_maxfill, uint, 0444);
+
+static unsigned long chash_test_iters;
+MODULE_PARM_DESC(test_iters, "Selftest iterations (default=1000 x #entries)");
+module_param_named(test_iters, chash_test_iters, ulong, 0444);
+
+static int __init chash_init(void)
+{
+	int ret;
+	u64 ts1_ns;
+
+	/* Skip self test on user errors */
+	if (chash_test_bits < 4 || chash_test_bits > 20) {
+		pr_err("chash: test_bits out of range [4..20].\n");
+		return 0;
+	}
+	if (chash_test_keysize != 4 && chash_test_keysize != 8) {
+		pr_err("chash: test_keysize invalid. Must be 4 or 8.\n");
+		return 0;
+	}
+
+	if (!chash_test_minfill)
+		chash_test_minfill = (1 << chash_test_bits) / 2;
+	if (!chash_test_maxfill)
+		chash_test_maxfill = (1 << chash_test_bits) * 4 / 5;
+	if (!chash_test_iters)
+		chash_test_iters = (1 << chash_test_bits) * 1000;
+
+	if (chash_test_minfill >= (1 << chash_test_bits)) {
+		pr_err("chash: test_minfill too big. Must be < table size.\n");
+		return 0;
+	}
+	if (chash_test_maxfill >= (1 << chash_test_bits)) {
+		pr_err("chash: test_maxfill too big. Must be < table size.\n");
+		return 0;
+	}
+	if (chash_test_minfill >= chash_test_maxfill) {
+		pr_err("chash: test_minfill must be < test_maxfill.\n");
+		return 0;
+	}
+	if (chash_test_keysize == 4 && chash_test_iters > 0xffffffff) {
+		pr_err("chash: test_iters must be < 4G for 4 byte keys.\n");
+		return 0;
+	}
+
+	ts1_ns = local_clock();
+	ret = chash_self_test(chash_test_bits, chash_test_keysize,
+			      chash_test_minfill, chash_test_maxfill,
+			      chash_test_iters);
+	if (!ret) {
+		u64 ts_delta_us = local_clock() - ts1_ns;
+		u64 iters_per_second = (u64)chash_test_iters * 1000000;
+
+		do_div(ts_delta_us, 1000);
+		do_div(iters_per_second, ts_delta_us);
+		pr_info("chash: self test took %llu us, %llu iterations/s\n",
+			ts_delta_us, iters_per_second);
+	} else {
+		pr_err("chash: self test failed: %d\n", ret);
+	}
+
+	return ret;
+}
+
+module_init(chash_init);
+
+#endif /* CONFIG_CHASH_SELFTEST */
+
+MODULE_DESCRIPTION("Closed hash table");
+MODULE_LICENSE("GPL and additional rights");

drivers/gpu/drm/amd/powerplay/Makefile

@@ -4,12 +4,11 @@ subdir-ccflags-y += \
 		-I$(FULL_AMD_PATH)/include/asic_reg  \
 		-I$(FULL_AMD_PATH)/include  \
 		-I$(FULL_AMD_PATH)/powerplay/smumgr\
-		-I$(FULL_AMD_PATH)/powerplay/hwmgr \
-		-I$(FULL_AMD_PATH)/powerplay/eventmgr
+		-I$(FULL_AMD_PATH)/powerplay/hwmgr
 
 AMD_PP_PATH = ../powerplay
 
-PP_LIBS = smumgr hwmgr eventmgr
+PP_LIBS = smumgr hwmgr
 
 AMD_POWERPLAY = $(addsuffix /Makefile,$(addprefix $(FULL_AMD_PATH)/powerplay/,$(PP_LIBS)))
 

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

@@ -29,22 +29,19 @@
 #include "amd_powerplay.h"
 #include "pp_instance.h"
 #include "power_state.h"
-#include "eventmanager.h"
-
 
 static inline int pp_check(struct pp_instance *handle)
 {
 	if (handle == NULL || handle->pp_valid != PP_VALID)
 		return -EINVAL;
 
-	if (handle->smu_mgr == NULL || handle->smu_mgr->smumgr_funcs == NULL)
+	if (handle->hwmgr == NULL || handle->hwmgr->smumgr_funcs == NULL)
 		return -EINVAL;
 
 	if (handle->pm_en == 0)
 		return PP_DPM_DISABLED;
 
-	if (handle->hwmgr == NULL || handle->hwmgr->hwmgr_func == NULL
-		|| handle->eventmgr == NULL)
+	if (handle->hwmgr->hwmgr_func == NULL)
 		return PP_DPM_DISABLED;
 
 	return 0;
@@ -55,46 +52,32 @@ static int pp_early_init(void *handle)
 	int ret;
 	struct pp_instance *pp_handle = (struct pp_instance *)handle;
 
-	ret = smum_early_init(pp_handle);
+	ret = hwmgr_early_init(pp_handle);
 	if (ret)
-		return ret;
+		return -EINVAL;
 
 	if ((pp_handle->pm_en == 0)
 		|| cgs_is_virtualization_enabled(pp_handle->device))
 		return PP_DPM_DISABLED;
 
-	ret = hwmgr_early_init(pp_handle);
-	if (ret) {
-		pp_handle->pm_en = 0;
-		return PP_DPM_DISABLED;
-	}
-
-	ret = eventmgr_early_init(pp_handle);
-	if (ret) {
-		kfree(pp_handle->hwmgr);
-		pp_handle->hwmgr = NULL;
-		pp_handle->pm_en = 0;
-		return PP_DPM_DISABLED;
-	}
-
 	return 0;
 }
 
 static int pp_sw_init(void *handle)
 {
-	struct pp_smumgr *smumgr;
+	struct pp_hwmgr *hwmgr;
 	int ret = 0;
 	struct pp_instance *pp_handle = (struct pp_instance *)handle;
 
 	ret = pp_check(pp_handle);
 
 	if (ret == 0 || ret == PP_DPM_DISABLED) {
-		smumgr = pp_handle->smu_mgr;
+		hwmgr = pp_handle->hwmgr;
 
-		if (smumgr->smumgr_funcs->smu_init == NULL)
+		if (hwmgr->smumgr_funcs->smu_init == NULL)
 			return -EINVAL;
 
-		ret = smumgr->smumgr_funcs->smu_init(smumgr);
+		ret = hwmgr->smumgr_funcs->smu_init(hwmgr);
 
 		pr_info("amdgpu: powerplay sw initialized\n");
 	}
@@ -103,40 +86,39 @@ static int pp_sw_init(void *handle)
 
 static int pp_sw_fini(void *handle)
 {
-	struct pp_smumgr *smumgr;
+	struct pp_hwmgr *hwmgr;
 	int ret = 0;
 	struct pp_instance *pp_handle = (struct pp_instance *)handle;
 
 	ret = pp_check(pp_handle);
 	if (ret == 0 || ret == PP_DPM_DISABLED) {
-		smumgr = pp_handle->smu_mgr;
+		hwmgr = pp_handle->hwmgr;
 
-		if (smumgr->smumgr_funcs->smu_fini == NULL)
+		if (hwmgr->smumgr_funcs->smu_fini == NULL)
 			return -EINVAL;
 
-		ret = smumgr->smumgr_funcs->smu_fini(smumgr);
+		ret = hwmgr->smumgr_funcs->smu_fini(pp_handle->hwmgr);
 	}
 	return ret;
 }
 
 static int pp_hw_init(void *handle)
 {
-	struct pp_smumgr *smumgr;
-	struct pp_eventmgr *eventmgr;
 	int ret = 0;
 	struct pp_instance *pp_handle = (struct pp_instance *)handle;
+	struct pp_hwmgr *hwmgr;
 
 	ret = pp_check(pp_handle);
 
 	if (ret == 0 || ret == PP_DPM_DISABLED) {
-		smumgr = pp_handle->smu_mgr;
+		hwmgr = pp_handle->hwmgr;
 
-		if (smumgr->smumgr_funcs->start_smu == NULL)
+		if (hwmgr->smumgr_funcs->start_smu == NULL)
 			return -EINVAL;
 
-		if(smumgr->smumgr_funcs->start_smu(smumgr)) {
+		if(hwmgr->smumgr_funcs->start_smu(pp_handle->hwmgr)) {
 			pr_err("smc start failed\n");
-			smumgr->smumgr_funcs->smu_fini(smumgr);
+			hwmgr->smumgr_funcs->smu_fini(pp_handle->hwmgr);
 			return -EINVAL;;
 		}
 		if (ret == PP_DPM_DISABLED)
@@ -146,38 +128,21 @@ static int pp_hw_init(void *handle)
 	ret = hwmgr_hw_init(pp_handle);
 	if (ret)
 		goto err;
-
-	eventmgr = pp_handle->eventmgr;
-	if (eventmgr->pp_eventmgr_init == NULL ||
-		eventmgr->pp_eventmgr_init(eventmgr))
-		goto err;
-
 	return 0;
 err:
 	pp_handle->pm_en = 0;
-	kfree(pp_handle->eventmgr);
-	kfree(pp_handle->hwmgr);
-	pp_handle->hwmgr = NULL;
-	pp_handle->eventmgr = NULL;
 	return PP_DPM_DISABLED;
 }
 
 static int pp_hw_fini(void *handle)
 {
-	struct pp_eventmgr *eventmgr;
 	struct pp_instance *pp_handle = (struct pp_instance *)handle;
 	int ret = 0;
 
 	ret = pp_check(pp_handle);
-
-	if (ret == 0) {
-		eventmgr = pp_handle->eventmgr;
-
-		if (eventmgr->pp_eventmgr_fini != NULL)
-			eventmgr->pp_eventmgr_fini(eventmgr);
-
+	if (ret == 0)
 		hwmgr_hw_fini(pp_handle);
-	}
+
 	return 0;
 }
 
@@ -244,8 +209,6 @@ static int pp_set_powergating_state(void *handle,
 
 static int pp_suspend(void *handle)
 {
-	struct pp_eventmgr *eventmgr;
-	struct pem_event_data event_data = { {0} };
 	struct pp_instance *pp_handle = (struct pp_instance *)handle;
 	int ret = 0;
 
@@ -256,17 +219,12 @@ static int pp_suspend(void *handle)
 	else if (ret != 0)
 		return ret;
 
-	eventmgr = pp_handle->eventmgr;
-	pem_handle_event(eventmgr, AMD_PP_EVENT_SUSPEND, &event_data);
-
-	return 0;
+	return hwmgr_hw_suspend(pp_handle);
 }
 
 static int pp_resume(void *handle)
 {
-	struct pp_eventmgr *eventmgr;
-	struct pem_event_data event_data = { {0} };
-	struct pp_smumgr *smumgr;
+	struct pp_hwmgr  *hwmgr;
 	int ret, ret1;
 	struct pp_instance *pp_handle = (struct pp_instance *)handle;
 
@@ -275,26 +233,22 @@ static int pp_resume(void *handle)
 	if (ret1 != 0 && ret1 != PP_DPM_DISABLED)
 		return ret1;
 
-	smumgr = pp_handle->smu_mgr;
+	hwmgr = pp_handle->hwmgr;
 
-	if (smumgr->smumgr_funcs->start_smu == NULL)
+	if (hwmgr->smumgr_funcs->start_smu == NULL)
 		return -EINVAL;
 
-	ret = smumgr->smumgr_funcs->start_smu(smumgr);
+	ret = hwmgr->smumgr_funcs->start_smu(pp_handle->hwmgr);
 	if (ret) {
 		pr_err("smc start failed\n");
-		smumgr->smumgr_funcs->smu_fini(smumgr);
+		hwmgr->smumgr_funcs->smu_fini(pp_handle->hwmgr);
 		return ret;
 	}
 
 	if (ret1 == PP_DPM_DISABLED)
 		return 0;
 
-	eventmgr = pp_handle->eventmgr;
-
-	pem_handle_event(eventmgr, AMD_PP_EVENT_RESUME, &event_data);
-
-	return 0;
+	return hwmgr_hw_resume(pp_handle);
 }
 
 const struct amd_ip_funcs pp_ip_funcs = {
@@ -324,6 +278,42 @@ static int pp_dpm_fw_loading_complete(void *handle)
 	return 0;
 }
 
+static void pp_dpm_en_umd_pstate(struct pp_hwmgr  *hwmgr,
+						enum amd_dpm_forced_level *level)
+{
+	uint32_t profile_mode_mask = AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD |
+					AMD_DPM_FORCED_LEVEL_PROFILE_MIN_SCLK |
+					AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK |
+					AMD_DPM_FORCED_LEVEL_PROFILE_PEAK;
+
+	if (!(hwmgr->dpm_level & profile_mode_mask)) {
+		/* enter umd pstate, save current level, disable gfx cg*/
+		if (*level & profile_mode_mask) {
+			hwmgr->saved_dpm_level = hwmgr->dpm_level;
+			hwmgr->en_umd_pstate = true;
+			cgs_set_clockgating_state(hwmgr->device,
+						AMD_IP_BLOCK_TYPE_GFX,
+						AMD_CG_STATE_UNGATE);
+			cgs_set_powergating_state(hwmgr->device,
+					AMD_IP_BLOCK_TYPE_GFX,
+					AMD_PG_STATE_UNGATE);
+		}
+	} else {
+		/* exit umd pstate, restore level, enable gfx cg*/
+		if (!(*level & profile_mode_mask)) {
+			if (*level == AMD_DPM_FORCED_LEVEL_PROFILE_EXIT)
+				*level = hwmgr->saved_dpm_level;
+			hwmgr->en_umd_pstate = false;
+			cgs_set_clockgating_state(hwmgr->device,
+					AMD_IP_BLOCK_TYPE_GFX,
+					AMD_CG_STATE_GATE);
+			cgs_set_powergating_state(hwmgr->device,
+					AMD_IP_BLOCK_TYPE_GFX,
+					AMD_PG_STATE_GATE);
+		}
+	}
+}
+
 static int pp_dpm_force_performance_level(void *handle,
 					enum amd_dpm_forced_level level)
 {
@@ -338,13 +328,22 @@ static int pp_dpm_force_performance_level(void *handle,
 
 	hwmgr = pp_handle->hwmgr;
 
+	if (level == hwmgr->dpm_level)
+		return 0;
+
 	if (hwmgr->hwmgr_func->force_dpm_level == NULL) {
 		pr_info("%s was not implemented.\n", __func__);
 		return 0;
 	}
 
 	mutex_lock(&pp_handle->pp_lock);
-	hwmgr->hwmgr_func->force_dpm_level(hwmgr, level);
+	pp_dpm_en_umd_pstate(hwmgr, &level);
+	hwmgr->request_dpm_level = level;
+	hwmgr_handle_task(pp_handle, AMD_PP_TASK_READJUST_POWER_STATE, NULL, NULL);
+	ret = hwmgr->hwmgr_func->force_dpm_level(hwmgr, level);
+	if (!ret)
+		hwmgr->dpm_level = hwmgr->request_dpm_level;
+
 	mutex_unlock(&pp_handle->pp_lock);
 	return 0;
 }
@@ -369,11 +368,12 @@ static enum amd_dpm_forced_level pp_dpm_get_performance_level(
 	return level;
 }
 
-static int pp_dpm_get_sclk(void *handle, bool low)
+static uint32_t pp_dpm_get_sclk(void *handle, bool low)
 {
 	struct pp_hwmgr  *hwmgr;
 	struct pp_instance *pp_handle = (struct pp_instance *)handle;
 	int ret = 0;
+	uint32_t clk = 0;
 
 	ret = pp_check(pp_handle);
 
@@ -387,16 +387,17 @@ static int pp_dpm_get_sclk(void *handle, bool low)
 		return 0;
 	}
 	mutex_lock(&pp_handle->pp_lock);
-	ret = hwmgr->hwmgr_func->get_sclk(hwmgr, low);
+	clk = hwmgr->hwmgr_func->get_sclk(hwmgr, low);
 	mutex_unlock(&pp_handle->pp_lock);
-	return ret;
+	return clk;
 }
 
-static int pp_dpm_get_mclk(void *handle, bool low)
+static uint32_t pp_dpm_get_mclk(void *handle, bool low)
 {
 	struct pp_hwmgr  *hwmgr;
 	struct pp_instance *pp_handle = (struct pp_instance *)handle;
 	int ret = 0;
+	uint32_t clk = 0;
 
 	ret = pp_check(pp_handle);
 
@@ -410,12 +411,12 @@ static int pp_dpm_get_mclk(void *handle, bool low)
 		return 0;
 	}
 	mutex_lock(&pp_handle->pp_lock);
-	ret = hwmgr->hwmgr_func->get_mclk(hwmgr, low);
+	clk = hwmgr->hwmgr_func->get_mclk(hwmgr, low);
 	mutex_unlock(&pp_handle->pp_lock);
-	return ret;
+	return clk;
 }
 
-static int pp_dpm_powergate_vce(void *handle, bool gate)
+static void pp_dpm_powergate_vce(void *handle, bool gate)
 {
 	struct pp_hwmgr  *hwmgr;
 	struct pp_instance *pp_handle = (struct pp_instance *)handle;
@@ -424,21 +425,20 @@ static int pp_dpm_powergate_vce(void *handle, bool gate)
 	ret = pp_check(pp_handle);
 
 	if (ret != 0)
-		return ret;
+		return;
 
 	hwmgr = pp_handle->hwmgr;
 
 	if (hwmgr->hwmgr_func->powergate_vce == NULL) {
 		pr_info("%s was not implemented.\n", __func__);
-		return 0;
+		return;
 	}
 	mutex_lock(&pp_handle->pp_lock);
-	ret = hwmgr->hwmgr_func->powergate_vce(hwmgr, gate);
+	hwmgr->hwmgr_func->powergate_vce(hwmgr, gate);
 	mutex_unlock(&pp_handle->pp_lock);
-	return ret;
 }
 
-static int pp_dpm_powergate_uvd(void *handle, bool gate)
+static void pp_dpm_powergate_uvd(void *handle, bool gate)
 {
 	struct pp_hwmgr  *hwmgr;
 	struct pp_instance *pp_handle = (struct pp_instance *)handle;
@@ -447,74 +447,34 @@ static int pp_dpm_powergate_uvd(void *handle, bool gate)
 	ret = pp_check(pp_handle);
 
 	if (ret != 0)
-		return ret;
+		return;
 
 	hwmgr = pp_handle->hwmgr;
 
 	if (hwmgr->hwmgr_func->powergate_uvd == NULL) {
 		pr_info("%s was not implemented.\n", __func__);
-		return 0;
+		return;
 	}
 	mutex_lock(&pp_handle->pp_lock);
-	ret = hwmgr->hwmgr_func->powergate_uvd(hwmgr, gate);
+	hwmgr->hwmgr_func->powergate_uvd(hwmgr, gate);
 	mutex_unlock(&pp_handle->pp_lock);
-	return ret;
 }
 
-static enum PP_StateUILabel power_state_convert(enum amd_pm_state_type  state)
-{
-	switch (state) {
-	case POWER_STATE_TYPE_BATTERY:
-		return PP_StateUILabel_Battery;
-	case POWER_STATE_TYPE_BALANCED:
-		return PP_StateUILabel_Balanced;
-	case POWER_STATE_TYPE_PERFORMANCE:
-		return PP_StateUILabel_Performance;
-	default:
-		return PP_StateUILabel_None;
-	}
-}
-
-static int pp_dpm_dispatch_tasks(void *handle, enum amd_pp_event event_id,
+static int pp_dpm_dispatch_tasks(void *handle, enum amd_pp_task task_id,
 		void *input, void *output)
 {
 	int ret = 0;
-	struct pem_event_data data = { {0} };
 	struct pp_instance *pp_handle = (struct pp_instance *)handle;
 
 	ret = pp_check(pp_handle);
 
 	if (ret != 0)
 		return ret;
-	mutex_lock(&pp_handle->pp_lock);
-	switch (event_id) {
-	case AMD_PP_EVENT_DISPLAY_CONFIG_CHANGE:
-		ret = pem_handle_event(pp_handle->eventmgr, event_id, &data);
-		break;
-	case AMD_PP_EVENT_ENABLE_USER_STATE:
-	{
-		enum amd_pm_state_type  ps;
-
-		if (input == NULL) {
-			ret = -EINVAL;
-			break;
-		}
-		ps = *(unsigned long *)input;
 
-		data.requested_ui_label = power_state_convert(ps);
-		ret = pem_handle_event(pp_handle->eventmgr, event_id, &data);
-		break;
-	}
-	case AMD_PP_EVENT_COMPLETE_INIT:
-		ret = pem_handle_event(pp_handle->eventmgr, event_id, &data);
-		break;
-	case AMD_PP_EVENT_READJUST_POWER_STATE:
-		ret = pem_handle_event(pp_handle->eventmgr, event_id, &data);
-		break;
-	default:
-		break;
-	}
+	mutex_lock(&pp_handle->pp_lock);
+	ret = hwmgr_handle_task(pp_handle, task_id, input, output);
 	mutex_unlock(&pp_handle->pp_lock);
+
 	return ret;
 }
 
@@ -562,7 +522,7 @@ static enum amd_pm_state_type pp_dpm_get_current_power_state(void *handle)
 	return pm_type;
 }
 
-static int pp_dpm_set_fan_control_mode(void *handle, uint32_t mode)
+static void pp_dpm_set_fan_control_mode(void *handle, uint32_t mode)
 {
 	struct pp_hwmgr  *hwmgr;
 	struct pp_instance *pp_handle = (struct pp_instance *)handle;
@@ -571,25 +531,25 @@ static int pp_dpm_set_fan_control_mode(void *handle, uint32_t mode)
 	ret = pp_check(pp_handle);
 
 	if (ret != 0)
-		return ret;
+		return;
 
 	hwmgr = pp_handle->hwmgr;
 
 	if (hwmgr->hwmgr_func->set_fan_control_mode == NULL) {
 		pr_info("%s was not implemented.\n", __func__);
-		return 0;
+		return;
 	}
 	mutex_lock(&pp_handle->pp_lock);
-	ret = hwmgr->hwmgr_func->set_fan_control_mode(hwmgr, mode);
+	hwmgr->hwmgr_func->set_fan_control_mode(hwmgr, mode);
 	mutex_unlock(&pp_handle->pp_lock);
-	return ret;
 }
 
-static int pp_dpm_get_fan_control_mode(void *handle)
+static uint32_t pp_dpm_get_fan_control_mode(void *handle)
 {
 	struct pp_hwmgr  *hwmgr;
 	struct pp_instance *pp_handle = (struct pp_instance *)handle;
 	int ret = 0;
+	uint32_t mode = 0;
 
 	ret = pp_check(pp_handle);
 
@@ -603,9 +563,9 @@ static int pp_dpm_get_fan_control_mode(void *handle)
 		return 0;
 	}
 	mutex_lock(&pp_handle->pp_lock);
-	ret = hwmgr->hwmgr_func->get_fan_control_mode(hwmgr);
+	mode = hwmgr->hwmgr_func->get_fan_control_mode(hwmgr);
 	mutex_unlock(&pp_handle->pp_lock);
-	return ret;
+	return mode;
 }
 
 static int pp_dpm_set_fan_speed_percent(void *handle, uint32_t percent)
@@ -1128,7 +1088,7 @@ static int pp_dpm_switch_power_profile(void *handle,
 	return 0;
 }
 
-const struct amd_powerplay_funcs pp_dpm_funcs = {
+const struct amd_pm_funcs pp_dpm_funcs = {
 	.get_temperature = pp_dpm_get_temperature,
 	.load_firmware = pp_dpm_load_fw,
 	.wait_for_fw_loading_complete = pp_dpm_fw_loading_complete,
@@ -1189,15 +1149,9 @@ int amd_powerplay_destroy(void *handle)
 {
 	struct pp_instance *instance = (struct pp_instance *)handle;
 
-	if (instance->pm_en) {
-		kfree(instance->eventmgr);
-		kfree(instance->hwmgr);
-		instance->hwmgr = NULL;
-		instance->eventmgr = NULL;
-	}
+	kfree(instance->hwmgr);
+	instance->hwmgr = NULL;
 
-	kfree(instance->smu_mgr);
-	instance->smu_mgr = NULL;
 	kfree(instance);
 	instance = NULL;
 	return 0;
@@ -1206,18 +1160,16 @@ int amd_powerplay_destroy(void *handle)
 int amd_powerplay_reset(void *handle)
 {
 	struct pp_instance *instance = (struct pp_instance *)handle;
-	struct pp_eventmgr *eventmgr;
-	struct pem_event_data event_data = { {0} };
 	int ret;
 
-	if (cgs_is_virtualization_enabled(instance->smu_mgr->device))
+	if (cgs_is_virtualization_enabled(instance->hwmgr->device))
 		return PP_DPM_DISABLED;
 
 	ret = pp_check(instance);
 	if (ret != 0)
 		return ret;
 
-	ret = pp_hw_fini(handle);
+	ret = pp_hw_fini(instance);
 	if (ret)
 		return ret;
 
@@ -1225,16 +1177,7 @@ int amd_powerplay_reset(void *handle)
 	if (ret)
 		return PP_DPM_DISABLED;
 
-	eventmgr = instance->eventmgr;
-
-	if (eventmgr->pp_eventmgr_init == NULL)
-		return PP_DPM_DISABLED;
-
-	ret = eventmgr->pp_eventmgr_init(eventmgr);
-	if (ret)
-		return ret;
-
-	return pem_handle_event(eventmgr, AMD_PP_EVENT_COMPLETE_INIT, &event_data);
+	return hwmgr_handle_task(instance, AMD_PP_TASK_COMPLETE_INIT, NULL, NULL);
 }
 
 /* export this function to DAL */

drivers/gpu/drm/amd/powerplay/eventmgr/Makefile

@@ -1,11 +0,0 @@
-#
-# Makefile for the 'event manager' sub-component of powerplay.
-# It provides the event management services for the driver.
-
-EVENT_MGR = eventmgr.o eventinit.o eventmanagement.o  \
-		eventactionchains.o eventsubchains.o eventtasks.o psm.o
-
-AMD_PP_EVENT = $(addprefix $(AMD_PP_PATH)/eventmgr/,$(EVENT_MGR))
-
-AMD_POWERPLAY_FILES += $(AMD_PP_EVENT)
-

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

@@ -1,291 +0,0 @@
-/*
- * Copyright 2015 Advanced Micro Devices, Inc.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
- * OTHER DEALINGS IN THE SOFTWARE.
- *
- */
-#include "eventmgr.h"
-#include "eventactionchains.h"
-#include "eventsubchains.h"
-
-static const pem_event_action * const initialize_event[] = {
-	block_adjust_power_state_tasks,
-	power_budget_tasks,
-	system_config_tasks,
-	setup_asic_tasks,
-	enable_dynamic_state_management_tasks,
-	get_2d_performance_state_tasks,
-	set_performance_state_tasks,
-	initialize_thermal_controller_tasks,
-	conditionally_force_3d_performance_state_tasks,
-	process_vbios_eventinfo_tasks,
-	broadcast_power_policy_tasks,
-	NULL
-};
-
-const struct action_chain initialize_action_chain = {
-	"Initialize",
-	initialize_event
-};
-
-static const pem_event_action * const uninitialize_event[] = {
-	ungate_all_display_phys_tasks,
-	uninitialize_display_phy_access_tasks,
-	disable_gfx_voltage_island_power_gating_tasks,
-	disable_gfx_clock_gating_tasks,
-	uninitialize_thermal_controller_tasks,
-	set_boot_state_tasks,
-	adjust_power_state_tasks,
-	disable_dynamic_state_management_tasks,
-	disable_clock_power_gatings_tasks,
-	cleanup_asic_tasks,
-	prepare_for_pnp_stop_tasks,
-	NULL
-};
-
-const struct action_chain uninitialize_action_chain = {
-	"Uninitialize",
-	uninitialize_event
-};
-
-static const pem_event_action * const power_source_change_event_pp_enabled[] = {
-	set_power_source_tasks,
-	set_power_saving_state_tasks,
-	adjust_power_state_tasks,
-	enable_disable_fps_tasks,
-	set_nbmcu_state_tasks,
-	broadcast_power_policy_tasks,
-	NULL
-};
-
-const struct action_chain power_source_change_action_chain_pp_enabled = {
-	"Power source change - PowerPlay enabled",
-	power_source_change_event_pp_enabled
-};
-
-static const pem_event_action * const power_source_change_event_pp_disabled[] = {
-	set_power_source_tasks,
-	set_nbmcu_state_tasks,
-	NULL
-};
-
-const struct action_chain power_source_changes_action_chain_pp_disabled = {
-	"Power source change - PowerPlay disabled",
-	power_source_change_event_pp_disabled
-};
-
-static const pem_event_action * const power_source_change_event_hardware_dc[] = {
-	set_power_source_tasks,
-	set_power_saving_state_tasks,
-	adjust_power_state_tasks,
-	enable_disable_fps_tasks,
-	reset_hardware_dc_notification_tasks,
-	set_nbmcu_state_tasks,
-	broadcast_power_policy_tasks,
-	NULL
-};
-
-const struct action_chain power_source_change_action_chain_hardware_dc = {
-	"Power source change - with Hardware DC switching",
-	power_source_change_event_hardware_dc
-};
-
-static const pem_event_action * const suspend_event[] = {
-	reset_display_phy_access_tasks,
-	unregister_interrupt_tasks,
-	disable_gfx_voltage_island_power_gating_tasks,
-	disable_gfx_clock_gating_tasks,
-	notify_smu_suspend_tasks,
-	disable_smc_firmware_ctf_tasks,
-	set_boot_state_tasks,
-	adjust_power_state_tasks,
-	disable_fps_tasks,
-	vari_bright_suspend_tasks,
-	reset_fan_speed_to_default_tasks,
-	power_down_asic_tasks,
-	disable_stutter_mode_tasks,
-	set_connected_standby_tasks,
-	block_hw_access_tasks,
-	NULL
-};
-
-const struct action_chain suspend_action_chain = {
-	"Suspend",
-	suspend_event
-};
-
-static const pem_event_action * const resume_event[] = {
-	unblock_hw_access_tasks,
-	resume_connected_standby_tasks,
-	notify_smu_resume_tasks,
-	reset_display_configCounter_tasks,
-	update_dal_configuration_tasks,
-	vari_bright_resume_tasks,
-	setup_asic_tasks,
-	enable_stutter_mode_tasks, /*must do this in boot state and before SMC is started */
-	enable_dynamic_state_management_tasks,
-	enable_disable_bapm_tasks,
-	initialize_thermal_controller_tasks,
-	get_2d_performance_state_tasks,
-	set_performance_state_tasks,
-	adjust_power_state_tasks,
-	enable_disable_fps_tasks,
-	notify_hw_power_source_tasks,
-	process_vbios_event_info_tasks,
-	enable_gfx_clock_gating_tasks,
-	enable_gfx_voltage_island_power_gating_tasks,
-	reset_clock_gating_tasks,
-	notify_smu_vpu_recovery_end_tasks,
-	disable_vpu_cap_tasks,
-	execute_escape_sequence_tasks,
-	NULL
-};
-
-
-const struct action_chain resume_action_chain = {
-	"resume",
-	resume_event
-};
-
-static const pem_event_action * const complete_init_event[] = {
-	unblock_adjust_power_state_tasks,
-	adjust_power_state_tasks,
-	enable_gfx_clock_gating_tasks,
-	enable_gfx_voltage_island_power_gating_tasks,
-	notify_power_state_change_tasks,
-	NULL
-};
-
-const struct action_chain complete_init_action_chain = {
-	"complete init",
-	complete_init_event
-};
-
-static const pem_event_action * const enable_gfx_clock_gating_event[] = {
-	enable_gfx_clock_gating_tasks,
-	NULL
-};
-
-const struct action_chain enable_gfx_clock_gating_action_chain = {
-	"enable gfx clock gate",
-	enable_gfx_clock_gating_event
-};
-
-static const pem_event_action * const disable_gfx_clock_gating_event[] = {
-	disable_gfx_clock_gating_tasks,
-	NULL
-};
-
-const struct action_chain disable_gfx_clock_gating_action_chain = {
-	"disable gfx clock gate",
-	disable_gfx_clock_gating_event
-};
-
-static const pem_event_action * const enable_cgpg_event[] = {
-	enable_cgpg_tasks,
-	NULL
-};
-
-const struct action_chain enable_cgpg_action_chain = {
-	"eable cg pg",
-	enable_cgpg_event
-};
-
-static const pem_event_action * const disable_cgpg_event[] = {
-	disable_cgpg_tasks,
-	NULL
-};
-
-const struct action_chain disable_cgpg_action_chain = {
-	"disable cg pg",
-	disable_cgpg_event
-};
-
-
-/* Enable user _2d performance and activate */
-
-static const pem_event_action * const enable_user_state_event[] = {
-	create_new_user_performance_state_tasks,
-	adjust_power_state_tasks,
-	NULL
-};
-
-const struct action_chain enable_user_state_action_chain = {
-	"Enable user state",
-	enable_user_state_event
-};
-
-static const pem_event_action * const enable_user_2d_performance_event[] = {
-	enable_user_2d_performance_tasks,
-	add_user_2d_performance_state_tasks,
-	set_performance_state_tasks,
-	adjust_power_state_tasks,
-	delete_user_2d_performance_state_tasks,
-	NULL
-};
-
-const struct action_chain enable_user_2d_performance_action_chain = {
-	"enable_user_2d_performance_event_activate",
-	enable_user_2d_performance_event
-};
-
-
-static const pem_event_action * const disable_user_2d_performance_event[] = {
-	disable_user_2d_performance_tasks,
-	delete_user_2d_performance_state_tasks,
-	NULL
-};
-
-const struct action_chain disable_user_2d_performance_action_chain = {
-	"disable_user_2d_performance_event",
-	disable_user_2d_performance_event
-};
-
-
-static const pem_event_action * const display_config_change_event[] = {
-	/* countDisplayConfigurationChangeEventTasks, */
-	unblock_adjust_power_state_tasks,
-	set_cpu_power_state,
-	notify_hw_power_source_tasks,
-	get_2d_performance_state_tasks,
-	set_performance_state_tasks,
-	/* updateDALConfigurationTasks,
-	variBrightDisplayConfigurationChangeTasks, */
-	adjust_power_state_tasks,
-	/*enableDisableFPSTasks,
-	setNBMCUStateTasks,
-	notifyPCIEDeviceReadyTasks,*/
-	NULL
-};
-
-const struct action_chain display_config_change_action_chain = {
-	"Display configuration change",
-	display_config_change_event
-};
-
-static const pem_event_action * const readjust_power_state_event[] = {
-	adjust_power_state_tasks,
-	NULL
-};
-
-const struct action_chain readjust_power_state_action_chain = {
-	"re-adjust power state",
-	readjust_power_state_event
-};
-

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

@@ -1,62 +0,0 @@
-/*
- * Copyright 2015 Advanced Micro Devices, Inc.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
- * OTHER DEALINGS IN THE SOFTWARE.
- *
- */
-#ifndef _EVENT_ACTION_CHAINS_H_
-#define _EVENT_ACTION_CHAINS_H_
-#include "eventmgr.h"
-
-extern const struct action_chain initialize_action_chain;
-
-extern const struct action_chain uninitialize_action_chain;
-
-extern const struct action_chain power_source_change_action_chain_pp_enabled;
-
-extern const struct action_chain power_source_changes_action_chain_pp_disabled;
-
-extern const struct action_chain power_source_change_action_chain_hardware_dc;
-
-extern const struct action_chain suspend_action_chain;
-
-extern const struct action_chain resume_action_chain;
-
-extern const struct action_chain complete_init_action_chain;
-
-extern const struct action_chain enable_gfx_clock_gating_action_chain;
-
-extern const struct action_chain disable_gfx_clock_gating_action_chain;
-
-extern const struct action_chain enable_cgpg_action_chain;
-
-extern const struct action_chain disable_cgpg_action_chain;
-
-extern const struct action_chain enable_user_2d_performance_action_chain;
-
-extern const struct action_chain disable_user_2d_performance_action_chain;
-
-extern const struct action_chain enable_user_state_action_chain;
-
-extern const struct action_chain readjust_power_state_action_chain;
-
-extern const struct action_chain display_config_change_action_chain;
-
-#endif /*_EVENT_ACTION_CHAINS_H_*/
-

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

@@ -1,195 +0,0 @@
-/*
- * Copyright 2015 Advanced Micro Devices, Inc.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
- * OTHER DEALINGS IN THE SOFTWARE.
- *
- */
-#include "eventmgr.h"
-#include "eventinit.h"
-#include "ppinterrupt.h"
-#include "hardwaremanager.h"
-
-void pem_init_feature_info(struct pp_eventmgr *eventmgr)
-{
-
-	/* PowerPlay info */
-	eventmgr->ui_state_info[PP_PowerSource_AC].default_ui_lable =
-					    PP_StateUILabel_Performance;
-
-	eventmgr->ui_state_info[PP_PowerSource_AC].current_ui_label =
-					    PP_StateUILabel_Performance;
-
-	eventmgr->ui_state_info[PP_PowerSource_DC].default_ui_lable =
-						  PP_StateUILabel_Battery;
-
-	eventmgr->ui_state_info[PP_PowerSource_DC].current_ui_label =
-						  PP_StateUILabel_Battery;
-
-	if (phm_cap_enabled(eventmgr->platform_descriptor->platformCaps, PHM_PlatformCaps_PowerPlaySupport)) {
-		eventmgr->features[PP_Feature_PowerPlay].supported = true;
-		eventmgr->features[PP_Feature_PowerPlay].version = PEM_CURRENT_POWERPLAY_FEATURE_VERSION;
-		eventmgr->features[PP_Feature_PowerPlay].enabled_default = true;
-		eventmgr->features[PP_Feature_PowerPlay].enabled = true;
-	} else {
-		eventmgr->features[PP_Feature_PowerPlay].supported = false;
-		eventmgr->features[PP_Feature_PowerPlay].enabled = false;
-		eventmgr->features[PP_Feature_PowerPlay].enabled_default = false;
-	}
-
-	eventmgr->features[PP_Feature_Force3DClock].supported = true;
-	eventmgr->features[PP_Feature_Force3DClock].enabled = false;
-	eventmgr->features[PP_Feature_Force3DClock].enabled_default = false;
-	eventmgr->features[PP_Feature_Force3DClock].version = 1;
-
-	/* over drive*/
-	eventmgr->features[PP_Feature_User2DPerformance].version = 4;
-	eventmgr->features[PP_Feature_User3DPerformance].version = 4;
-	eventmgr->features[PP_Feature_OverdriveTest].version = 4;
-
-	eventmgr->features[PP_Feature_OverDrive].version = 4;
-	eventmgr->features[PP_Feature_OverDrive].enabled = false;
-	eventmgr->features[PP_Feature_OverDrive].enabled_default = false;
-
-	eventmgr->features[PP_Feature_User2DPerformance].supported = false;
-	eventmgr->features[PP_Feature_User2DPerformance].enabled = false;
-	eventmgr->features[PP_Feature_User2DPerformance].enabled_default = false;
-
-	eventmgr->features[PP_Feature_User3DPerformance].supported = false;
-	eventmgr->features[PP_Feature_User3DPerformance].enabled = false;
-	eventmgr->features[PP_Feature_User3DPerformance].enabled_default = false;
-
-	eventmgr->features[PP_Feature_OverdriveTest].supported = false;
-	eventmgr->features[PP_Feature_OverdriveTest].enabled = false;
-	eventmgr->features[PP_Feature_OverdriveTest].enabled_default = false;
-
-	eventmgr->features[PP_Feature_OverDrive].supported = false;
-
-	eventmgr->features[PP_Feature_PowerBudgetWaiver].enabled_default = false;
-	eventmgr->features[PP_Feature_PowerBudgetWaiver].version = 1;
-	eventmgr->features[PP_Feature_PowerBudgetWaiver].supported = false;
-	eventmgr->features[PP_Feature_PowerBudgetWaiver].enabled = false;
-
-	/* Multi UVD States support */
-	eventmgr->features[PP_Feature_MultiUVDState].supported = false;
-	eventmgr->features[PP_Feature_MultiUVDState].enabled = false;
-	eventmgr->features[PP_Feature_MultiUVDState].enabled_default = false;
-
-	/* Dynamic UVD States support */
-	eventmgr->features[PP_Feature_DynamicUVDState].supported = false;
-	eventmgr->features[PP_Feature_DynamicUVDState].enabled = false;
-	eventmgr->features[PP_Feature_DynamicUVDState].enabled_default = false;
-
-	/* VCE DPM support */
-	eventmgr->features[PP_Feature_VCEDPM].supported = false;
-	eventmgr->features[PP_Feature_VCEDPM].enabled = false;
-	eventmgr->features[PP_Feature_VCEDPM].enabled_default = false;
-
-	/* ACP PowerGating support */
-	eventmgr->features[PP_Feature_ACP_POWERGATING].supported = false;
-	eventmgr->features[PP_Feature_ACP_POWERGATING].enabled = false;
-	eventmgr->features[PP_Feature_ACP_POWERGATING].enabled_default = false;
-
-	/* PPM support */
-	eventmgr->features[PP_Feature_PPM].version = 1;
-	eventmgr->features[PP_Feature_PPM].supported = false;
-	eventmgr->features[PP_Feature_PPM].enabled = false;
-
-	/* FFC support (enables fan and temp settings, Gemini needs temp settings) */
-	if (phm_cap_enabled(eventmgr->platform_descriptor->platformCaps, PHM_PlatformCaps_ODFuzzyFanControlSupport) ||
-	    phm_cap_enabled(eventmgr->platform_descriptor->platformCaps, PHM_PlatformCaps_GeminiRegulatorFanControlSupport)) {
-		eventmgr->features[PP_Feature_FFC].version = 1;
-		eventmgr->features[PP_Feature_FFC].supported = true;
-		eventmgr->features[PP_Feature_FFC].enabled = true;
-		eventmgr->features[PP_Feature_FFC].enabled_default = true;
-	} else {
-		eventmgr->features[PP_Feature_FFC].supported = false;
-		eventmgr->features[PP_Feature_FFC].enabled = false;
-		eventmgr->features[PP_Feature_FFC].enabled_default = false;
-	}
-
-	eventmgr->features[PP_Feature_VariBright].supported = false;
-	eventmgr->features[PP_Feature_VariBright].enabled = false;
-	eventmgr->features[PP_Feature_VariBright].enabled_default = false;
-
-	eventmgr->features[PP_Feature_BACO].supported = false;
-	eventmgr->features[PP_Feature_BACO].supported = false;
-	eventmgr->features[PP_Feature_BACO].enabled_default = false;
-
-	/* PowerDown feature support */
-	eventmgr->features[PP_Feature_PowerDown].supported = false;
-	eventmgr->features[PP_Feature_PowerDown].enabled = false;
-	eventmgr->features[PP_Feature_PowerDown].enabled_default = false;
-
-	eventmgr->features[PP_Feature_FPS].version = 1;
-	eventmgr->features[PP_Feature_FPS].supported = false;
-	eventmgr->features[PP_Feature_FPS].enabled_default = false;
-	eventmgr->features[PP_Feature_FPS].enabled = false;
-
-	eventmgr->features[PP_Feature_ViPG].version = 1;
-	eventmgr->features[PP_Feature_ViPG].supported = false;
-	eventmgr->features[PP_Feature_ViPG].enabled_default = false;
-	eventmgr->features[PP_Feature_ViPG].enabled = false;
-}
-
-static int thermal_interrupt_callback(void *private_data,
-				      unsigned src_id, const uint32_t *iv_entry)
-{
-	/* TO DO hanle PEM_Event_ThermalNotification (struct pp_eventmgr *)private_data*/
-	pr_info("current thermal is out of range \n");
-	return 0;
-}
-
-int pem_register_interrupts(struct pp_eventmgr *eventmgr)
-{
-	int result = 0;
-	struct pp_interrupt_registration_info info;
-
-	info.call_back = thermal_interrupt_callback;
-	info.context = eventmgr;
-
-	result = phm_register_thermal_interrupt(eventmgr->hwmgr, &info);
-
-	/* TODO:
-	 * 2. Register CTF event interrupt
-	 * 3. Register for vbios events interrupt
-	 * 4. Register External Throttle Interrupt
-	 * 5. Register Smc To Host Interrupt
-	 * */
-	return result;
-}
-
-
-int pem_unregister_interrupts(struct pp_eventmgr *eventmgr)
-{
-	return 0;
-}
-
-
-void pem_uninit_featureInfo(struct pp_eventmgr *eventmgr)
-{
-	eventmgr->features[PP_Feature_MultiUVDState].supported = false;
-	eventmgr->features[PP_Feature_VariBright].supported = false;
-	eventmgr->features[PP_Feature_PowerBudgetWaiver].supported = false;
-	eventmgr->features[PP_Feature_OverDrive].supported = false;
-	eventmgr->features[PP_Feature_OverdriveTest].supported = false;
-	eventmgr->features[PP_Feature_User3DPerformance].supported = false;
-	eventmgr->features[PP_Feature_User2DPerformance].supported = false;
-	eventmgr->features[PP_Feature_PowerPlay].supported = false;
-	eventmgr->features[PP_Feature_Force3DClock].supported = false;
-}

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

@@ -1,215 +0,0 @@
-/*
- * Copyright 2015 Advanced Micro Devices, Inc.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
- * OTHER DEALINGS IN THE SOFTWARE.
- *
- */
-#include "eventmanagement.h"
-#include "eventmgr.h"
-#include "eventactionchains.h"
-
-int pem_init_event_action_chains(struct pp_eventmgr *eventmgr)
-{
-	int i;
-
-	for (i = 0; i < AMD_PP_EVENT_MAX; i++)
-		eventmgr->event_chain[i] = NULL;
-
-	eventmgr->event_chain[AMD_PP_EVENT_SUSPEND] = pem_get_suspend_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_INITIALIZE] = pem_get_initialize_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_UNINITIALIZE] = pem_get_uninitialize_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_POWER_SOURCE_CHANGE] = pem_get_power_source_change_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_HIBERNATE] = pem_get_hibernate_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_RESUME] = pem_get_resume_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_THERMAL_NOTIFICATION] = pem_get_thermal_notification_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_VBIOS_NOTIFICATION] = pem_get_vbios_notification_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_ENTER_THERMAL_STATE] = pem_get_enter_thermal_state_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_EXIT_THERMAL_STATE] = pem_get_exit_thermal_state_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_ENABLE_POWER_PLAY] = pem_get_enable_powerplay_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_DISABLE_POWER_PLAY] = pem_get_disable_powerplay_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_ENABLE_OVER_DRIVE_TEST] = pem_get_enable_overdrive_test_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_DISABLE_OVER_DRIVE_TEST] = pem_get_disable_overdrive_test_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_ENABLE_GFX_CLOCK_GATING] = pem_get_enable_gfx_clock_gating_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_DISABLE_GFX_CLOCK_GATING] = pem_get_disable_gfx_clock_gating_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_ENABLE_CGPG] = pem_get_enable_cgpg_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_DISABLE_CGPG] = pem_get_disable_cgpg_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_COMPLETE_INIT] = pem_get_complete_init_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_SCREEN_ON] = pem_get_screen_on_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_SCREEN_OFF] = pem_get_screen_off_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_PRE_SUSPEND] = pem_get_pre_suspend_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_PRE_RESUME] = pem_get_pre_resume_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_ENABLE_USER_STATE] = pem_enable_user_state_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_READJUST_POWER_STATE] = pem_readjust_power_state_action_chain(eventmgr);
-	eventmgr->event_chain[AMD_PP_EVENT_DISPLAY_CONFIG_CHANGE] = pem_display_config_change_action_chain(eventmgr);
-	return 0;
-}
-
-int pem_excute_event_chain(struct pp_eventmgr *eventmgr, const struct action_chain *event_chain, struct pem_event_data *event_data)
-{
-	const pem_event_action * const *paction_chain;
-	const pem_event_action *psub_chain;
-	int tmp_result = 0;
-	int result = 0;
-
-	if (eventmgr == NULL || event_chain == NULL || event_data == NULL)
-		return -EINVAL;
-
-	for (paction_chain = event_chain->action_chain; NULL != *paction_chain; paction_chain++) {
-		if (0 != result)
-			return result;
-
-		for (psub_chain = *paction_chain; NULL != *psub_chain; psub_chain++) {
-			tmp_result = (*psub_chain)(eventmgr, event_data);
-			if (0 == result)
-				result = tmp_result;
-		}
-	}
-
-	return result;
-}
-
-const struct action_chain *pem_get_suspend_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return &suspend_action_chain;
-}
-
-const struct action_chain *pem_get_initialize_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return &initialize_action_chain;
-}
-
-const struct action_chain *pem_get_uninitialize_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return &uninitialize_action_chain;
-}
-
-const struct action_chain *pem_get_power_source_change_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return &power_source_change_action_chain_pp_enabled;  /* other case base on feature info*/
-}
-
-const struct action_chain *pem_get_resume_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return &resume_action_chain;
-}
-
-const struct action_chain *pem_get_hibernate_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return NULL;
-}
-
-const struct action_chain *pem_get_thermal_notification_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return NULL;
-}
-
-const struct action_chain *pem_get_vbios_notification_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return NULL;
-}
-
-const struct action_chain *pem_get_enter_thermal_state_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return NULL;
-}
-
-const struct action_chain *pem_get_exit_thermal_state_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return NULL;
-}
-
-const struct action_chain *pem_get_enable_powerplay_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return NULL;
-}
-
-const struct action_chain *pem_get_disable_powerplay_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return NULL;
-}
-
-const struct action_chain *pem_get_enable_overdrive_test_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return NULL;
-}
-
-const struct action_chain *pem_get_disable_overdrive_test_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return NULL;
-}
-
-const struct action_chain *pem_get_enable_gfx_clock_gating_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return &enable_gfx_clock_gating_action_chain;
-}
-
-const struct action_chain *pem_get_disable_gfx_clock_gating_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return &disable_gfx_clock_gating_action_chain;
-}
-
-const struct action_chain *pem_get_enable_cgpg_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return &enable_cgpg_action_chain;
-}
-
-const struct action_chain *pem_get_disable_cgpg_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return &disable_cgpg_action_chain;
-}
-
-const struct action_chain *pem_get_complete_init_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return &complete_init_action_chain;
-}
-
-const struct action_chain *pem_get_screen_on_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return NULL;
-}
-
-const struct action_chain *pem_get_screen_off_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return NULL;
-}
-
-const struct action_chain *pem_get_pre_suspend_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return NULL;
-}
-
-const struct action_chain *pem_get_pre_resume_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return NULL;
-}
-
-const struct action_chain *pem_enable_user_state_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return &enable_user_state_action_chain;
-}
-
-const struct action_chain *pem_readjust_power_state_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return &readjust_power_state_action_chain;
-}
-
-const struct action_chain *pem_display_config_change_action_chain(struct pp_eventmgr *eventmgr)
-{
-	return &display_config_change_action_chain;
-}

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

@@ -1,59 +0,0 @@
-/*
- * Copyright 2015 Advanced Micro Devices, Inc.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
- * OTHER DEALINGS IN THE SOFTWARE.
- *
- */
-#ifndef _EVENT_MANAGEMENT_H_
-#define _EVENT_MANAGEMENT_H_
-
-#include "eventmgr.h"
-
-int pem_init_event_action_chains(struct pp_eventmgr *eventmgr);
-int pem_excute_event_chain(struct pp_eventmgr *eventmgr, const struct action_chain *event_chain, struct pem_event_data *event_data);
-const struct action_chain *pem_get_suspend_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_initialize_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_uninitialize_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_power_source_change_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_resume_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_hibernate_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_thermal_notification_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_vbios_notification_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_enter_thermal_state_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_exit_thermal_state_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_enable_powerplay_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_disable_powerplay_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_enable_overdrive_test_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_disable_overdrive_test_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_enable_gfx_clock_gating_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_disable_gfx_clock_gating_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_enable_cgpg_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_disable_cgpg_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_complete_init_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_screen_on_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_screen_off_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_pre_suspend_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_get_pre_resume_action_chain(struct pp_eventmgr *eventmgr);
-
-extern const struct action_chain *pem_enable_user_state_action_chain(struct pp_eventmgr *eventmgr);
-extern const struct action_chain *pem_readjust_power_state_action_chain(struct pp_eventmgr *eventmgr);
-const struct action_chain *pem_display_config_change_action_chain(struct pp_eventmgr *eventmgr);
-
-
-#endif /* _EVENT_MANAGEMENT_H_ */

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

@@ -1,104 +0,0 @@
-/*
- * Copyright 2015 Advanced Micro Devices, Inc.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
- * OTHER DEALINGS IN THE SOFTWARE.
- *
- */
-#include <linux/types.h>
-#include <linux/kernel.h>
-#include <linux/slab.h>
-#include "eventmgr.h"
-#include "hwmgr.h"
-#include "eventinit.h"
-#include "eventmanagement.h"
-
-static int pem_init(struct pp_eventmgr *eventmgr)
-{
-	int result = 0;
-	struct pem_event_data event_data = { {0} };
-
-	/* Initialize PowerPlay feature info */
-	pem_init_feature_info(eventmgr);
-
-	/* Initialize event action chains */
-	pem_init_event_action_chains(eventmgr);
-
-	/* Call initialization event */
-	result = pem_handle_event(eventmgr, AMD_PP_EVENT_INITIALIZE, &event_data);
-
-	/* if (0 != result)
-		return result; */
-
-	/* Register interrupt callback functions */
-	result = pem_register_interrupts(eventmgr);
-	return 0;
-}
-
-static void pem_fini(struct pp_eventmgr *eventmgr)
-{
-	struct pem_event_data event_data = { {0} };
-
-	pem_uninit_featureInfo(eventmgr);
-	pem_unregister_interrupts(eventmgr);
-
-	pem_handle_event(eventmgr, AMD_PP_EVENT_UNINITIALIZE, &event_data);
-}
-
-int eventmgr_early_init(struct pp_instance *handle)
-{
-	struct pp_eventmgr *eventmgr;
-
-	if (handle == NULL)
-		return -EINVAL;
-
-	eventmgr = kzalloc(sizeof(struct pp_eventmgr), GFP_KERNEL);
-	if (eventmgr == NULL)
-		return -ENOMEM;
-
-	eventmgr->hwmgr = handle->hwmgr;
-	handle->eventmgr = eventmgr;
-
-	eventmgr->platform_descriptor = &(eventmgr->hwmgr->platform_descriptor);
-	eventmgr->pp_eventmgr_init = pem_init;
-	eventmgr->pp_eventmgr_fini = pem_fini;
-
-	return 0;
-}
-
-static int pem_handle_event_unlocked(struct pp_eventmgr *eventmgr, enum amd_pp_event event, struct pem_event_data *data)
-{
-	if (eventmgr == NULL || event >= AMD_PP_EVENT_MAX || data == NULL)
-		return -EINVAL;
-
-	return pem_excute_event_chain(eventmgr, eventmgr->event_chain[event], data);
-}
-
-int pem_handle_event(struct pp_eventmgr *eventmgr, enum amd_pp_event event, struct pem_event_data *event_data)
-{
-	int r = 0;
-
-	r = pem_handle_event_unlocked(eventmgr, event, event_data);
-
-	return r;
-}
-
-bool pem_is_hw_access_blocked(struct pp_eventmgr *eventmgr)
-{
-	return (eventmgr->block_adjust_power_state || phm_is_hw_access_blocked(eventmgr->hwmgr));
-}