Merge branch 'drm-next-5.2' of git://people.freedesktop.org/~agd5f/linux into drm-next

amdgpu:
- Switch to HMM for userptr (reverted until HMM fixes land)
- New experimental SMU 11 replacement for powerplay for vega20 (not enabled by default)
- Initial RAS support for vega20
- BACO support for vega12
- BACO fixes for vega20
- Rework IH handling for page fault and retry interrupts
- Cleanly split CPU and GPU paths for GPUVM updates
- Powerplay fixes
- XGMI fixes
- Rework how DC interacts with atomic for planes
- Clean up and simplify DC/Powerplay interfaces
- Misc cleanups and bug fixes

amdkfd:
- Switch to HMM for userptr (reverted until HMM fixes land)
- Add initial RAS support
- MQD fixes

ttm:
- Unify DRM_FILE_PAGE_OFFSET handling
- Account for kernel allocations in kernel zone only
- Misc cleanups

Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190402170820.22197-1-alexander.deucher@amd.com

drivers/gpu/drm/Kconfig

@@ -231,8 +231,6 @@ 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

@@ -57,7 +57,6 @@ obj-$(CONFIG_DRM_TTM)	+= ttm/
 obj-$(CONFIG_DRM_SCHED)	+= scheduler/
 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

@@ -53,7 +53,8 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \
 	amdgpu_ucode.o amdgpu_bo_list.o amdgpu_ctx.o amdgpu_sync.o \
 	amdgpu_gtt_mgr.o amdgpu_vram_mgr.o amdgpu_virt.o amdgpu_atomfirmware.o \
 	amdgpu_vf_error.o amdgpu_sched.o amdgpu_debugfs.o amdgpu_ids.o \
-	amdgpu_gmc.o amdgpu_xgmi.o amdgpu_csa.o
+	amdgpu_gmc.o amdgpu_xgmi.o amdgpu_csa.o amdgpu_ras.o amdgpu_vm_cpu.o \
+	amdgpu_vm_sdma.o
 
 # add asic specific block
 amdgpu-$(CONFIG_DRM_AMDGPU_CIK)+= cik.o cik_ih.o kv_smc.o kv_dpm.o \

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

@@ -83,6 +83,7 @@
 #include "amdgpu_gem.h"
 #include "amdgpu_doorbell.h"
 #include "amdgpu_amdkfd.h"
+#include "amdgpu_smu.h"
 
 #define MAX_GPU_INSTANCE		16
 
@@ -156,6 +157,8 @@ extern int amdgpu_emu_mode;
 extern uint amdgpu_smu_memory_pool_size;
 extern uint amdgpu_dc_feature_mask;
 extern struct amdgpu_mgpu_info mgpu_info;
+extern int amdgpu_ras_enable;
+extern uint amdgpu_ras_mask;
 
 #ifdef CONFIG_DRM_AMDGPU_SI
 extern int amdgpu_si_support;
@@ -702,7 +705,6 @@ enum amd_hw_ip_block_type {
 struct amd_powerplay {
 	void *pp_handle;
 	const struct amd_pm_funcs *pp_funcs;
-	uint32_t pp_feature;
 };
 
 #define AMDGPU_RESET_MAGIC_NUM 64
@@ -842,6 +844,9 @@ struct amdgpu_device {
 	struct amd_powerplay		powerplay;
 	bool				pp_force_state_enabled;
 
+	/* smu */
+	struct smu_context		smu;
+
 	/* dpm */
 	struct amdgpu_pm		pm;
 	u32				cg_flags;
@@ -922,6 +927,8 @@ struct amdgpu_device {
 
 	int asic_reset_res;
 	struct work_struct		xgmi_reset_work;
+
+	bool                            in_baco_reset;
 };
 
 static inline struct amdgpu_device *amdgpu_ttm_adev(struct ttm_bo_device *bdev)

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

@@ -640,4 +640,8 @@ int kgd2kfd_post_reset(struct kfd_dev *kfd)
 void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry)
 {
 }
+
+void kgd2kfd_set_sram_ecc_flag(struct kfd_dev *kfd)
+{
+}
 #endif

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

@@ -230,5 +230,6 @@ int kgd2kfd_quiesce_mm(struct mm_struct *mm);
 int kgd2kfd_resume_mm(struct mm_struct *mm);
 int kgd2kfd_schedule_evict_and_restore_process(struct mm_struct *mm,
 					       struct dma_fence *fence);
+void kgd2kfd_set_sram_ecc_flag(struct kfd_dev *kfd);
 
 #endif /* AMDGPU_AMDKFD_H_INCLUDED */

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

@@ -410,15 +410,7 @@ static int add_bo_to_vm(struct amdgpu_device *adev, struct kgd_mem *mem,
 	if (p_bo_va_entry)
 		*p_bo_va_entry = bo_va_entry;
 
-	/* Allocate new page tables if needed and validate
-	 * them.
-	 */
-	ret = amdgpu_vm_alloc_pts(adev, vm, va, amdgpu_bo_size(bo));
-	if (ret) {
-		pr_err("Failed to allocate pts, err=%d\n", ret);
-		goto err_alloc_pts;
-	}
-
+	/* Allocate validate page tables if needed */
 	ret = vm_validate_pt_pd_bos(vm);
 	if (ret) {
 		pr_err("validate_pt_pd_bos() failed\n");
@@ -741,13 +733,7 @@ static int update_gpuvm_pte(struct amdgpu_device *adev,
 		struct amdgpu_sync *sync)
 {
 	int ret;
-	struct amdgpu_vm *vm;
-	struct amdgpu_bo_va *bo_va;
-	struct amdgpu_bo *bo;
-
-	bo_va = entry->bo_va;
-	vm = bo_va->base.vm;
-	bo = bo_va->base.bo;
+	struct amdgpu_bo_va *bo_va = entry->bo_va;
 
 	/* Update the page tables  */
 	ret = amdgpu_vm_bo_update(adev, bo_va, false);
@@ -906,7 +892,8 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info,
 		pr_err("validate_pt_pd_bos() failed\n");
 		goto validate_pd_fail;
 	}
-	amdgpu_bo_sync_wait(vm->root.base.bo, AMDGPU_FENCE_OWNER_KFD, false);
+	ret = amdgpu_bo_sync_wait(vm->root.base.bo,
+				  AMDGPU_FENCE_OWNER_KFD, false);
 	if (ret)
 		goto wait_pd_fail;
 	amdgpu_bo_fence(vm->root.base.bo,

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

@@ -28,8 +28,6 @@
 #include "atom.h"
 #include "atombios.h"
 
-#define get_index_into_master_table(master_table, table_name) (offsetof(struct master_table, table_name) / sizeof(uint16_t))
-
 bool amdgpu_atomfirmware_gpu_supports_virtualization(struct amdgpu_device *adev)
 {
 	int index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1,
@@ -238,10 +236,71 @@ int amdgpu_atomfirmware_get_vram_type(struct amdgpu_device *adev)
 	return 0;
 }
 
+/*
+ * Return true if vbios enabled ecc by default, if umc info table is available
+ * or false if ecc is not enabled or umc info table is not available
+ */
+bool amdgpu_atomfirmware_mem_ecc_supported(struct amdgpu_device *adev)
+{
+	struct amdgpu_mode_info *mode_info = &adev->mode_info;
+	int index;
+	u16 data_offset, size;
+	union umc_info *umc_info;
+	u8 frev, crev;
+	bool ecc_default_enabled = false;
+
+	index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1,
+			umc_info);
+
+	if (amdgpu_atom_parse_data_header(mode_info->atom_context,
+				index, &size, &frev, &crev, &data_offset)) {
+		/* support umc_info 3.1+ */
+		if ((frev == 3 && crev >= 1) || (frev > 3)) {
+			umc_info = (union umc_info *)
+				(mode_info->atom_context->bios + data_offset);
+			ecc_default_enabled =
+				(le32_to_cpu(umc_info->v31.umc_config) &
+				 UMC_CONFIG__DEFAULT_MEM_ECC_ENABLE) ? true : false;
+		}
+	}
+
+	return ecc_default_enabled;
+}
+
 union firmware_info {
 	struct atom_firmware_info_v3_1 v31;
 };
 
+/*
+ * Return true if vbios supports sram ecc or false if not
+ */
+bool amdgpu_atomfirmware_sram_ecc_supported(struct amdgpu_device *adev)
+{
+	struct amdgpu_mode_info *mode_info = &adev->mode_info;
+	int index;
+	u16 data_offset, size;
+	union firmware_info *firmware_info;
+	u8 frev, crev;
+	bool sram_ecc_supported = false;
+
+	index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1,
+			firmwareinfo);
+
+	if (amdgpu_atom_parse_data_header(adev->mode_info.atom_context,
+				index, &size, &frev, &crev, &data_offset)) {
+		/* support firmware_info 3.1 + */
+		if ((frev == 3 && crev >=1) || (frev > 3)) {
+			firmware_info = (union firmware_info *)
+				(mode_info->atom_context->bios + data_offset);
+			sram_ecc_supported =
+				(le32_to_cpu(firmware_info->v31.firmware_capability) &
+				 ATOM_FIRMWARE_CAP_SRAM_ECC) ? true : false;
+		}
+	}
+
+	return sram_ecc_supported;
+}
+
 union smu_info {
 	struct atom_smu_info_v3_1 v31;
 };
@@ -346,11 +405,11 @@ int amdgpu_atomfirmware_get_gfx_info(struct amdgpu_device *adev)
 			(mode_info->atom_context->bios + data_offset);
 		switch (crev) {
 		case 4:
-			adev->gfx.config.max_shader_engines = gfx_info->v24.gc_num_se;
-			adev->gfx.config.max_cu_per_sh = gfx_info->v24.gc_num_cu_per_sh;
-			adev->gfx.config.max_sh_per_se = gfx_info->v24.gc_num_sh_per_se;
-			adev->gfx.config.max_backends_per_se = gfx_info->v24.gc_num_rb_per_se;
-			adev->gfx.config.max_texture_channel_caches = gfx_info->v24.gc_num_tccs;
+			adev->gfx.config.max_shader_engines = gfx_info->v24.max_shader_engines;
+			adev->gfx.config.max_cu_per_sh = gfx_info->v24.max_cu_per_sh;
+			adev->gfx.config.max_sh_per_se = gfx_info->v24.max_sh_per_se;
+			adev->gfx.config.max_backends_per_se = gfx_info->v24.max_backends_per_se;
+			adev->gfx.config.max_texture_channel_caches = gfx_info->v24.max_texture_channel_caches;
 			adev->gfx.config.max_gprs = le16_to_cpu(gfx_info->v24.gc_num_gprs);
 			adev->gfx.config.max_gs_threads = gfx_info->v24.gc_num_max_gs_thds;
 			adev->gfx.config.gs_vgt_table_depth = gfx_info->v24.gc_gs_table_depth;

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

@@ -24,6 +24,8 @@
 #ifndef __AMDGPU_ATOMFIRMWARE_H__
 #define __AMDGPU_ATOMFIRMWARE_H__
 
+#define get_index_into_master_table(master_table, table_name) (offsetof(struct master_table, table_name) / sizeof(uint16_t))
+
 bool amdgpu_atomfirmware_gpu_supports_virtualization(struct amdgpu_device *adev);
 void amdgpu_atomfirmware_scratch_regs_init(struct amdgpu_device *adev);
 int amdgpu_atomfirmware_allocate_fb_scratch(struct amdgpu_device *adev);
@@ -31,5 +33,7 @@ int amdgpu_atomfirmware_get_vram_width(struct amdgpu_device *adev);
 int amdgpu_atomfirmware_get_vram_type(struct amdgpu_device *adev);
 int amdgpu_atomfirmware_get_clock_info(struct amdgpu_device *adev);
 int amdgpu_atomfirmware_get_gfx_info(struct amdgpu_device *adev);
+bool amdgpu_atomfirmware_mem_ecc_supported(struct amdgpu_device *adev);
+bool amdgpu_atomfirmware_sram_ecc_supported(struct amdgpu_device *adev);
 
 #endif

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

@@ -92,15 +92,6 @@ int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		return -ENOMEM;
 	}
 
-	r = amdgpu_vm_alloc_pts(adev, (*bo_va)->base.vm, csa_addr,
-				size);
-	if (r) {
-		DRM_ERROR("failed to allocate pts for static CSA, err=%d\n", r);
-		amdgpu_vm_bo_rmv(adev, *bo_va);
-		ttm_eu_backoff_reservation(&ticket, &list);
-		return r;
-	}
-
 	r = amdgpu_vm_bo_map(adev, *bo_va, csa_addr, 0, size,
 			     AMDGPU_PTE_READABLE | AMDGPU_PTE_WRITEABLE |
 			     AMDGPU_PTE_EXECUTABLE);

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

@@ -26,6 +26,7 @@
 #include <drm/drm_auth.h>
 #include "amdgpu.h"
 #include "amdgpu_sched.h"
+#include "amdgpu_ras.h"
 
 #define to_amdgpu_ctx_entity(e)	\
 	container_of((e), struct amdgpu_ctx_entity, entity)
@@ -344,6 +345,7 @@ static int amdgpu_ctx_query2(struct amdgpu_device *adev,
 {
 	struct amdgpu_ctx *ctx;
 	struct amdgpu_ctx_mgr *mgr;
+	uint32_t ras_counter;
 
 	if (!fpriv)
 		return -EINVAL;
@@ -368,6 +370,21 @@ static int amdgpu_ctx_query2(struct amdgpu_device *adev,
 	if (atomic_read(&ctx->guilty))
 		out->state.flags |= AMDGPU_CTX_QUERY2_FLAGS_GUILTY;
 
+	/*query ue count*/
+	ras_counter = amdgpu_ras_query_error_count(adev, false);
+	/*ras counter is monotonic increasing*/
+	if (ras_counter != ctx->ras_counter_ue) {
+		out->state.flags |= AMDGPU_CTX_QUERY2_FLAGS_RAS_UE;
+		ctx->ras_counter_ue = ras_counter;
+	}
+
+	/*query ce count*/
+	ras_counter = amdgpu_ras_query_error_count(adev, true);
+	if (ras_counter != ctx->ras_counter_ce) {
+		out->state.flags |= AMDGPU_CTX_QUERY2_FLAGS_RAS_CE;
+		ctx->ras_counter_ce = ras_counter;
+	}
+
 	mutex_unlock(&mgr->lock);
 	return 0;
 }
@@ -541,32 +558,26 @@ void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr)
 	idr_init(&mgr->ctx_handles);
 }
 
-void amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr)
+long amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr, long timeout)
 {
 	unsigned num_entities = amdgput_ctx_total_num_entities();
 	struct amdgpu_ctx *ctx;
 	struct idr *idp;
 	uint32_t id, i;
-	long max_wait = MAX_WAIT_SCHED_ENTITY_Q_EMPTY;
 
 	idp = &mgr->ctx_handles;
 
 	mutex_lock(&mgr->lock);
 	idr_for_each_entry(idp, ctx, id) {
-
-		if (!ctx->adev) {
-			mutex_unlock(&mgr->lock);
-			return;
-		}
-
 		for (i = 0; i < num_entities; i++) {
 			struct drm_sched_entity *entity;
 
 			entity = &ctx->entities[0][i].entity;
-			max_wait = drm_sched_entity_flush(entity, max_wait);
+			timeout = drm_sched_entity_flush(entity, timeout);
 		}
 	}
 	mutex_unlock(&mgr->lock);
+	return timeout;
 }
 
 void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr)
@@ -579,10 +590,6 @@ void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr)
 	idp = &mgr->ctx_handles;
 
 	idr_for_each_entry(idp, ctx, id) {
-
-		if (!ctx->adev)
-			return;
-
 		if (kref_read(&ctx->refcount) != 1) {
 			DRM_ERROR("ctx %p is still alive\n", ctx);
 			continue;

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

@@ -49,6 +49,8 @@ struct amdgpu_ctx {
 	enum drm_sched_priority		override_priority;
 	struct mutex			lock;
 	atomic_t			guilty;
+	uint32_t			ras_counter_ce;
+	uint32_t			ras_counter_ue;
 };
 
 struct amdgpu_ctx_mgr {
@@ -82,7 +84,7 @@ int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx,
 
 void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr);
 void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr);
-void amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr);
+long amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr, long timeout);
 void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr);
 
 #endif

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

@@ -568,10 +568,9 @@ static ssize_t amdgpu_debugfs_sensor_read(struct file *f, char __user *buf,
 	idx = *pos >> 2;
 
 	valuesize = sizeof(values);
-	if (adev->powerplay.pp_funcs && adev->powerplay.pp_funcs->read_sensor)
-		r = amdgpu_dpm_read_sensor(adev, idx, &values[0], &valuesize);
-	else
-		return -EINVAL;
+	r = amdgpu_dpm_read_sensor(adev, idx, &values[0], &valuesize);
+	if (r)
+		return r;
 
 	if (size > valuesize)
 		return -EINVAL;

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

@@ -60,6 +60,7 @@
 #include "amdgpu_pm.h"
 
 #include "amdgpu_xgmi.h"
+#include "amdgpu_ras.h"
 
 MODULE_FIRMWARE("amdgpu/vega10_gpu_info.bin");
 MODULE_FIRMWARE("amdgpu/vega12_gpu_info.bin");
@@ -1506,7 +1507,9 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev)
 			return -EAGAIN;
 	}
 
-	adev->powerplay.pp_feature = amdgpu_pp_feature_mask;
+	adev->pm.pp_feature = amdgpu_pp_feature_mask;
+	if (amdgpu_sriov_vf(adev))
+		adev->pm.pp_feature &= ~PP_GFXOFF_MASK;
 
 	for (i = 0; i < adev->num_ip_blocks; i++) {
 		if ((amdgpu_ip_block_mask & (1 << i)) == 0) {
@@ -1638,6 +1641,10 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
 {
 	int i, r;
 
+	r = amdgpu_ras_init(adev);
+	if (r)
+		return r;
+
 	for (i = 0; i < adev->num_ip_blocks; i++) {
 		if (!adev->ip_blocks[i].status.valid)
 			continue;
@@ -1681,6 +1688,13 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
 		}
 	}
 
+	r = amdgpu_ib_pool_init(adev);
+	if (r) {
+		dev_err(adev->dev, "IB initialization failed (%d).\n", r);
+		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_IB_INIT_FAIL, 0, r);
+		goto init_failed;
+	}
+
 	r = amdgpu_ucode_create_bo(adev); /* create ucode bo when sw_init complete*/
 	if (r)
 		goto init_failed;
@@ -1869,6 +1883,8 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev)
 {
 	int i, r;
 
+	amdgpu_ras_pre_fini(adev);
+
 	if (adev->gmc.xgmi.num_physical_nodes > 1)
 		amdgpu_xgmi_remove_device(adev);
 
@@ -1917,6 +1933,7 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev)
 			amdgpu_free_static_csa(&adev->virt.csa_obj);
 			amdgpu_device_wb_fini(adev);
 			amdgpu_device_vram_scratch_fini(adev);
+			amdgpu_ib_pool_fini(adev);
 		}
 
 		r = adev->ip_blocks[i].version->funcs->sw_fini((void *)adev);
@@ -1937,6 +1954,8 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev)
 		adev->ip_blocks[i].status.late_initialized = false;
 	}
 
+	amdgpu_ras_fini(adev);
+
 	if (amdgpu_sriov_vf(adev))
 		if (amdgpu_virt_release_full_gpu(adev, false))
 			DRM_ERROR("failed to release exclusive mode on fini\n");
@@ -1999,6 +2018,10 @@ static void amdgpu_device_ip_late_init_func_handler(struct work_struct *work)
 	r = amdgpu_device_enable_mgpu_fan_boost();
 	if (r)
 		DRM_ERROR("enable mgpu fan boost failed (%d).\n", r);
+
+	/*set to low pstate by default */
+	amdgpu_xgmi_set_pstate(adev, 0);
+
 }
 
 static void amdgpu_device_delay_enable_gfx_off(struct work_struct *work)
@@ -2369,7 +2392,7 @@ static void amdgpu_device_xgmi_reset_func(struct work_struct *__work)
 
 	adev->asic_reset_res =  amdgpu_asic_reset(adev);
 	if (adev->asic_reset_res)
-		DRM_WARN("ASIC reset failed with err r, %d for drm dev, %s",
+		DRM_WARN("ASIC reset failed with error, %d for drm dev, %s",
 			 adev->asic_reset_res, adev->ddev->unique);
 }
 
@@ -2642,13 +2665,6 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	/* Get a log2 for easy divisions. */
 	adev->mm_stats.log2_max_MBps = ilog2(max(1u, max_MBps));
 
-	r = amdgpu_ib_pool_init(adev);
-	if (r) {
-		dev_err(adev->dev, "IB initialization failed (%d).\n", r);
-		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_IB_INIT_FAIL, 0, r);
-		goto failed;
-	}
-
 	amdgpu_fbdev_init(adev);
 
 	r = amdgpu_pm_sysfs_init(adev);
@@ -2694,6 +2710,9 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 		goto failed;
 	}
 
+	/* must succeed. */
+	amdgpu_ras_post_init(adev);
+
 	return 0;
 
 failed:
@@ -2726,7 +2745,6 @@ void amdgpu_device_fini(struct amdgpu_device *adev)
 		else
 			drm_atomic_helper_shutdown(adev->ddev);
 	}
-	amdgpu_ib_pool_fini(adev);
 	amdgpu_fence_driver_fini(adev);
 	amdgpu_pm_sysfs_fini(adev);
 	amdgpu_fbdev_fini(adev);
@@ -3219,6 +3237,8 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev,
 	if (r)
 		return r;
 
+	amdgpu_amdkfd_pre_reset(adev);
+
 	/* Resume IP prior to SMC */
 	r = amdgpu_device_ip_reinit_early_sriov(adev);
 	if (r)
@@ -3238,6 +3258,7 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev,
 
 	amdgpu_irq_gpu_reset_resume_helper(adev);
 	r = amdgpu_ib_ring_tests(adev);
+	amdgpu_amdkfd_post_reset(adev);
 
 error:
 	amdgpu_virt_init_data_exchange(adev);
@@ -3370,7 +3391,7 @@ static int amdgpu_do_asic_reset(struct amdgpu_hive_info *hive,
 				r = amdgpu_asic_reset(tmp_adev);
 
 			if (r) {
-				DRM_ERROR("ASIC reset failed with err r, %d for drm dev, %s",
+				DRM_ERROR("ASIC reset failed with error, %d for drm dev, %s",
 					 r, tmp_adev->ddev->unique);
 				break;
 			}
@@ -3387,6 +3408,11 @@ static int amdgpu_do_asic_reset(struct amdgpu_hive_info *hive,
 						break;
 				}
 			}
+
+			list_for_each_entry(tmp_adev, device_list_handle,
+					gmc.xgmi.head) {
+				amdgpu_ras_reserve_bad_pages(tmp_adev);
+			}
 		}
 	}
 

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

@@ -904,3 +904,19 @@ amdgpu_get_vce_clock_state(void *handle, u32 idx)
 
 	return NULL;
 }
+
+int amdgpu_dpm_get_sclk(struct amdgpu_device *adev, bool low)
+{
+	if (is_support_sw_smu(adev))
+		return smu_get_sclk(&adev->smu, low);
+	else
+		return (adev)->powerplay.pp_funcs->get_sclk((adev)->powerplay.pp_handle, (low));
+}
+
+int amdgpu_dpm_get_mclk(struct amdgpu_device *adev, bool low)
+{
+	if (is_support_sw_smu(adev))
+		return smu_get_mclk(&adev->smu, low);
+	else
+		return (adev)->powerplay.pp_funcs->get_mclk((adev)->powerplay.pp_handle, (low));
+}

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

@@ -260,9 +260,6 @@ enum amdgpu_pcie_gen {
 #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)->powerplay.pp_funcs->read_sensor((adev)->powerplay.pp_handle, (idx), (value), (size)))
-
 #define amdgpu_dpm_set_fan_control_mode(adev, m) \
 		((adev)->powerplay.pp_funcs->set_fan_control_mode((adev)->powerplay.pp_handle, (m)))
 
@@ -281,18 +278,18 @@ enum amdgpu_pcie_gen {
 #define amdgpu_dpm_set_fan_speed_rpm(adev, s) \
 		((adev)->powerplay.pp_funcs->set_fan_speed_rpm)((adev)->powerplay.pp_handle, (s))
 
-#define amdgpu_dpm_get_sclk(adev, l) \
-		((adev)->powerplay.pp_funcs->get_sclk((adev)->powerplay.pp_handle, (l)))
-
-#define amdgpu_dpm_get_mclk(adev, l)  \
-		((adev)->powerplay.pp_funcs->get_mclk((adev)->powerplay.pp_handle, (l)))
-
 #define amdgpu_dpm_force_performance_level(adev, l) \
 		((adev)->powerplay.pp_funcs->force_performance_level((adev)->powerplay.pp_handle, (l)))
 
 #define amdgpu_dpm_get_current_power_state(adev) \
 		((adev)->powerplay.pp_funcs->get_current_power_state((adev)->powerplay.pp_handle))
 
+#define amdgpu_smu_get_current_power_state(adev) \
+		((adev)->smu.ppt_funcs->get_current_power_state(&((adev)->smu)))
+
+#define amdgpu_smu_set_power_state(adev) \
+		((adev)->smu.ppt_funcs->set_power_state(&((adev)->smu)))
+
 #define amdgpu_dpm_get_pp_num_states(adev, data) \
 		((adev)->powerplay.pp_funcs->get_pp_num_states((adev)->powerplay.pp_handle, data))
 
@@ -448,6 +445,9 @@ struct amdgpu_pm {
 	uint32_t                smu_prv_buffer_size;
 	struct amdgpu_bo        *smu_prv_buffer;
 	bool ac_power;
+	/* powerplay feature */
+	uint32_t pp_feature;
+
 };
 
 #define R600_SSTU_DFLT                               0
@@ -486,6 +486,8 @@ void amdgpu_dpm_print_ps_status(struct amdgpu_device *adev,
 u32 amdgpu_dpm_get_vblank_time(struct amdgpu_device *adev);
 u32 amdgpu_dpm_get_vrefresh(struct amdgpu_device *adev);
 void amdgpu_dpm_get_active_displays(struct amdgpu_device *adev);
+int amdgpu_dpm_read_sensor(struct amdgpu_device *adev, enum amd_pp_sensors sensor,
+			   void *data, uint32_t *size);
 
 bool amdgpu_is_internal_thermal_sensor(enum amdgpu_int_thermal_type sensor);
 
@@ -504,4 +506,8 @@ enum amdgpu_pcie_gen amdgpu_get_pcie_gen_support(struct amdgpu_device *adev,
 struct amd_vce_state*
 amdgpu_get_vce_clock_state(void *handle, u32 idx);
 
+extern int amdgpu_dpm_get_sclk(struct amdgpu_device *adev, bool low);
+
+extern int amdgpu_dpm_get_mclk(struct amdgpu_device *adev, bool low);
+
 #endif

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

@@ -74,9 +74,10 @@
  * - 3.28.0 - Add AMDGPU_CHUNK_ID_SCHEDULED_DEPENDENCIES
  * - 3.29.0 - Add AMDGPU_IB_FLAG_RESET_GDS_MAX_WAVE_ID
  * - 3.30.0 - Add AMDGPU_SCHED_OP_CONTEXT_PRIORITY_OVERRIDE.
+ * - 3.31.0 - Add support for per-flip tiling attribute changes with DC
  */
 #define KMS_DRIVER_MAJOR	3
-#define KMS_DRIVER_MINOR	30
+#define KMS_DRIVER_MINOR	31
 #define KMS_DRIVER_PATCHLEVEL	0
 
 int amdgpu_vram_limit = 0;
@@ -117,8 +118,8 @@ uint amdgpu_pg_mask = 0xffffffff;
 uint amdgpu_sdma_phase_quantum = 32;
 char *amdgpu_disable_cu = NULL;
 char *amdgpu_virtual_display = NULL;
-/* OverDrive(bit 14),gfxoff(bit 15),stutter mode(bit 17) disabled by default*/
-uint amdgpu_pp_feature_mask = 0xfffd3fff;
+/* OverDrive(bit 14) disabled by default*/
+uint amdgpu_pp_feature_mask = 0xffffbfff;
 int amdgpu_ngg = 0;
 int amdgpu_prim_buf_per_se = 0;
 int amdgpu_pos_buf_per_se = 0;
@@ -136,6 +137,8 @@ uint amdgpu_dc_feature_mask = 0;
 struct amdgpu_mgpu_info mgpu_info = {
 	.mutex = __MUTEX_INITIALIZER(mgpu_info.mutex),
 };
+int amdgpu_ras_enable = -1;
+uint amdgpu_ras_mask = 0xffffffff;
 
 /**
  * DOC: vramlimit (int)
@@ -494,6 +497,21 @@ module_param_named(gpu_recovery, amdgpu_gpu_recovery, int, 0444);
 MODULE_PARM_DESC(emu_mode, "Emulation mode, (1 = enable, 0 = disable)");
 module_param_named(emu_mode, amdgpu_emu_mode, int, 0444);
 
+/**
+ * DOC: ras_enable (int)
+ * Enable RAS features on the GPU (0 = disable, 1 = enable, -1 = auto (default))
+ */
+MODULE_PARM_DESC(ras_enable, "Enable RAS features on the GPU (0 = disable, 1 = enable, -1 = auto (default))");
+module_param_named(ras_enable, amdgpu_ras_enable, int, 0444);
+
+/**
+ * DOC: ras_mask (uint)
+ * Mask of RAS features to enable (default 0xffffffff), only valid when ras_enable == 1
+ * See the flags in drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
+ */
+MODULE_PARM_DESC(ras_mask, "Mask of RAS features to enable (default 0xffffffff), only valid when ras_enable == 1");
+module_param_named(ras_mask, amdgpu_ras_mask, uint, 0444);
+
 /**
  * DOC: si_support (int)
  * Set SI support driver. This parameter works after set config CONFIG_DRM_AMDGPU_SI. For SI asic, when radeon driver is enabled,
@@ -1159,13 +1177,14 @@ static int amdgpu_flush(struct file *f, fl_owner_t id)
 {
 	struct drm_file *file_priv = f->private_data;
 	struct amdgpu_fpriv *fpriv = file_priv->driver_priv;
+	long timeout = MAX_WAIT_SCHED_ENTITY_Q_EMPTY;
 
-	amdgpu_ctx_mgr_entity_flush(&fpriv->ctx_mgr);
+	timeout = amdgpu_ctx_mgr_entity_flush(&fpriv->ctx_mgr, timeout);
+	timeout = amdgpu_vm_wait_idle(&fpriv->vm, timeout);
 
-	return 0;
+	return timeout >= 0 ? 0 : timeout;
 }
 
-
 static const struct file_operations amdgpu_driver_kms_fops = {
 	.owner = THIS_MODULE,
 	.open = drm_open,

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

@@ -627,11 +627,6 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 
 	switch (args->operation) {
 	case AMDGPU_VA_OP_MAP:
-		r = amdgpu_vm_alloc_pts(adev, bo_va->base.vm, args->va_address,
-					args->map_size);
-		if (r)
-			goto error_backoff;
-
 		va_flags = amdgpu_gmc_get_pte_flags(adev, args->flags);
 		r = amdgpu_vm_bo_map(adev, bo_va, args->va_address,
 				     args->offset_in_bo, args->map_size,
@@ -647,11 +642,6 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 						args->map_size);
 		break;
 	case AMDGPU_VA_OP_REPLACE:
-		r = amdgpu_vm_alloc_pts(adev, bo_va->base.vm, args->va_address,
-					args->map_size);
-		if (r)
-			goto error_backoff;
-
 		va_flags = amdgpu_gmc_get_pte_flags(adev, args->flags);
 		r = amdgpu_vm_bo_replace_map(adev, bo_va, args->va_address,
 					     args->offset_in_bo, args->map_size,
@@ -745,17 +735,25 @@ int amdgpu_mode_dumb_create(struct drm_file *file_priv,
 	struct amdgpu_device *adev = dev->dev_private;
 	struct drm_gem_object *gobj;
 	uint32_t handle;
+	u64 flags = AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
 	u32 domain;
 	int r;
 
+	/*
+	 * The buffer returned from this function should be cleared, but
+	 * it can only be done if the ring is enabled or we'll fail to
+	 * create the buffer.
+	 */
+	if (adev->mman.buffer_funcs_enabled)
+		flags |= AMDGPU_GEM_CREATE_VRAM_CLEARED;
+
 	args->pitch = amdgpu_align_pitch(adev, args->width,
 					 DIV_ROUND_UP(args->bpp, 8), 0);
 	args->size = (u64)args->pitch * args->height;
 	args->size = ALIGN(args->size, PAGE_SIZE);
 	domain = amdgpu_bo_get_preferred_pin_domain(adev,
 				amdgpu_display_supported_domains(adev));
-	r = amdgpu_gem_object_create(adev, args->size, 0, domain,
-				     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
+	r = amdgpu_gem_object_create(adev, args->size, 0, domain, flags,
 				     ttm_bo_type_device, NULL, &gobj);
 	if (r)
 		return -ENOMEM;

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

@@ -390,7 +390,7 @@ void amdgpu_gfx_compute_mqd_sw_fini(struct amdgpu_device *adev)
 
 void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable)
 {
-	if (!(adev->powerplay.pp_feature & PP_GFXOFF_MASK))
+	if (!(adev->pm.pp_feature & PP_GFXOFF_MASK))
 		return;
 
 	if (!adev->powerplay.pp_funcs || !adev->powerplay.pp_funcs->set_powergating_by_smu)

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

@@ -258,6 +258,9 @@ struct amdgpu_gfx {
 	/* pipe reservation */
 	struct mutex			pipe_reserve_mutex;
 	DECLARE_BITMAP			(pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
+
+	/*ras */
+	struct ras_common_if		*ras_if;
 };
 
 #define amdgpu_gfx_get_gpu_clock_counter(adev) (adev)->gfx.funcs->get_gpu_clock_counter((adev))

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

@@ -79,6 +79,33 @@ uint64_t amdgpu_gmc_pd_addr(struct amdgpu_bo *bo)
 	return pd_addr;
 }
 
+/**
+ * amdgpu_gmc_set_pte_pde - update the page tables using CPU
+ *
+ * @adev: amdgpu_device pointer
+ * @cpu_pt_addr: cpu address of the page table
+ * @gpu_page_idx: entry in the page table to update
+ * @addr: dst addr to write into pte/pde
+ * @flags: access flags
+ *
+ * Update the page tables using CPU.
+ */
+int amdgpu_gmc_set_pte_pde(struct amdgpu_device *adev, void *cpu_pt_addr,
+				uint32_t gpu_page_idx, uint64_t addr,
+				uint64_t flags)
+{
+	void __iomem *ptr = (void *)cpu_pt_addr;
+	uint64_t value;
+
+	/*
+	 * The following is for PTE only. GART does not have PDEs.
+	*/
+	value = addr & 0x0000FFFFFFFFF000ULL;
+	value |= flags;
+	writeq(value, ptr + (gpu_page_idx * 8));
+	return 0;
+}
+
 /**
  * amdgpu_gmc_agp_addr - return the address in the AGP address space
  *
@@ -213,3 +240,58 @@ void amdgpu_gmc_agp_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc)
 	dev_info(adev->dev, "AGP: %lluM 0x%016llX - 0x%016llX\n",
 			mc->agp_size >> 20, mc->agp_start, mc->agp_end);
 }
+
+/**
+ * amdgpu_gmc_filter_faults - filter VM faults
+ *
+ * @adev: amdgpu device structure
+ * @addr: address of the VM fault
+ * @pasid: PASID of the process causing the fault
+ * @timestamp: timestamp of the fault
+ *
+ * Returns:
+ * True if the fault was filtered and should not be processed further.
+ * False if the fault is a new one and needs to be handled.
+ */
+bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev, uint64_t addr,
+			      uint16_t pasid, uint64_t timestamp)
+{
+	struct amdgpu_gmc *gmc = &adev->gmc;
+
+	uint64_t stamp, key = addr << 4 | pasid;
+	struct amdgpu_gmc_fault *fault;
+	uint32_t hash;
+
+	/* If we don't have space left in the ring buffer return immediately */
+	stamp = max(timestamp, AMDGPU_GMC_FAULT_TIMEOUT + 1) -
+		AMDGPU_GMC_FAULT_TIMEOUT;
+	if (gmc->fault_ring[gmc->last_fault].timestamp >= stamp)
+		return true;
+
+	/* Try to find the fault in the hash */
+	hash = hash_64(key, AMDGPU_GMC_FAULT_HASH_ORDER);
+	fault = &gmc->fault_ring[gmc->fault_hash[hash].idx];
+	while (fault->timestamp >= stamp) {
+		uint64_t tmp;
+
+		if (fault->key == key)
+			return true;
+
+		tmp = fault->timestamp;
+		fault = &gmc->fault_ring[fault->next];
+
+		/* Check if the entry was reused */
+		if (fault->timestamp >= tmp)
+			break;
+	}
+
+	/* Add the fault to the ring */
+	fault = &gmc->fault_ring[gmc->last_fault];
+	fault->key = key;
+	fault->timestamp = timestamp;
+
+	/* And update the hash */
+	fault->next = gmc->fault_hash[hash].idx;
+	gmc->fault_hash[hash].idx = gmc->last_fault++;
+	return false;
+}

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

@@ -43,8 +43,34 @@
  */
 #define AMDGPU_GMC_HOLE_MASK	0x0000ffffffffffffULL
 
+/*
+ * Ring size as power of two for the log of recent faults.
+ */
+#define AMDGPU_GMC_FAULT_RING_ORDER	8
+#define AMDGPU_GMC_FAULT_RING_SIZE	(1 << AMDGPU_GMC_FAULT_RING_ORDER)
+
+/*
+ * Hash size as power of two for the log of recent faults
+ */
+#define AMDGPU_GMC_FAULT_HASH_ORDER	8
+#define AMDGPU_GMC_FAULT_HASH_SIZE	(1 << AMDGPU_GMC_FAULT_HASH_ORDER)
+
+/*
+ * Number of IH timestamp ticks until a fault is considered handled
+ */
+#define AMDGPU_GMC_FAULT_TIMEOUT	5000ULL
+
 struct firmware;
 
+/*
+ * GMC page fault information
+ */
+struct amdgpu_gmc_fault {
+	uint64_t	timestamp;
+	uint64_t	next:AMDGPU_GMC_FAULT_RING_ORDER;
+	uint64_t	key:52;
+};
+
 /*
  * VMHUB structures, functions & helpers
  */
@@ -71,12 +97,6 @@ struct amdgpu_gmc_funcs {
 	/* Change the VMID -> PASID mapping */
 	void (*emit_pasid_mapping)(struct amdgpu_ring *ring, unsigned vmid,
 				   unsigned pasid);
-	/* write pte/pde updates using the cpu */
-	int (*set_pte_pde)(struct amdgpu_device *adev,
-			   void *cpu_pt_addr, /* cpu addr of page table */
-			   uint32_t gpu_page_idx, /* pte/pde to update */
-			   uint64_t addr, /* addr to write into pte/pde */
-			   uint64_t flags); /* access flags */
 	/* enable/disable PRT support */
 	void (*set_prt)(struct amdgpu_device *adev, bool enable);
 	/* set pte flags based per asic */
@@ -147,15 +167,22 @@ struct amdgpu_gmc {
 	struct kfd_vm_fault_info *vm_fault_info;
 	atomic_t		vm_fault_info_updated;
 
+	struct amdgpu_gmc_fault	fault_ring[AMDGPU_GMC_FAULT_RING_SIZE];
+	struct {
+		uint64_t	idx:AMDGPU_GMC_FAULT_RING_ORDER;
+	} fault_hash[AMDGPU_GMC_FAULT_HASH_SIZE];
+	uint64_t		last_fault:AMDGPU_GMC_FAULT_RING_ORDER;
+
 	const struct amdgpu_gmc_funcs	*gmc_funcs;
 
 	struct amdgpu_xgmi xgmi;
+	struct amdgpu_irq_src	ecc_irq;
+	struct ras_common_if    *ras_if;
 };
 
 #define amdgpu_gmc_flush_gpu_tlb(adev, vmid, type) (adev)->gmc.gmc_funcs->flush_gpu_tlb((adev), (vmid), (type))
 #define amdgpu_gmc_emit_flush_gpu_tlb(r, vmid, addr) (r)->adev->gmc.gmc_funcs->emit_flush_gpu_tlb((r), (vmid), (addr))
 #define amdgpu_gmc_emit_pasid_mapping(r, vmid, pasid) (r)->adev->gmc.gmc_funcs->emit_pasid_mapping((r), (vmid), (pasid))
-#define amdgpu_gmc_set_pte_pde(adev, pt, idx, addr, flags) (adev)->gmc.gmc_funcs->set_pte_pde((adev), (pt), (idx), (addr), (flags))
 #define amdgpu_gmc_get_vm_pde(adev, level, dst, flags) (adev)->gmc.gmc_funcs->get_vm_pde((adev), (level), (dst), (flags))
 #define amdgpu_gmc_get_pte_flags(adev, flags) (adev)->gmc.gmc_funcs->get_vm_pte_flags((adev),(flags))
 
@@ -189,6 +216,9 @@ static inline uint64_t amdgpu_gmc_sign_extend(uint64_t addr)
 
 void amdgpu_gmc_get_pde_for_bo(struct amdgpu_bo *bo, int level,
 			       uint64_t *addr, uint64_t *flags);
+int amdgpu_gmc_set_pte_pde(struct amdgpu_device *adev, void *cpu_pt_addr,
+				uint32_t gpu_page_idx, uint64_t addr,
+				uint64_t flags);
 uint64_t amdgpu_gmc_pd_addr(struct amdgpu_bo *bo);
 uint64_t amdgpu_gmc_agp_addr(struct ttm_buffer_object *bo);
 void amdgpu_gmc_vram_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc,
@@ -197,5 +227,7 @@ void amdgpu_gmc_gart_location(struct amdgpu_device *adev,
 			      struct amdgpu_gmc *mc);
 void amdgpu_gmc_agp_location(struct amdgpu_device *adev,
 			     struct amdgpu_gmc *mc);
+bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev, uint64_t addr,
+			      uint16_t pasid, uint64_t timestamp);
 
 #endif

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

@@ -36,6 +36,47 @@ struct amdgpu_gtt_node {
 	struct ttm_buffer_object *tbo;
 };
 
+/**
+ * DOC: mem_info_gtt_total
+ *
+ * The amdgpu driver provides a sysfs API for reporting current total size of
+ * the GTT.
+ * The file mem_info_gtt_total is used for this, and returns the total size of
+ * the GTT block, in bytes
+ */
+static ssize_t amdgpu_mem_info_gtt_total_show(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = ddev->dev_private;
+
+	return snprintf(buf, PAGE_SIZE, "%llu\n",
+			(adev->mman.bdev.man[TTM_PL_TT].size) * PAGE_SIZE);
+}
+
+/**
+ * DOC: mem_info_gtt_used
+ *
+ * The amdgpu driver provides a sysfs API for reporting current total amount of
+ * used GTT.
+ * The file mem_info_gtt_used is used for this, and returns the current used
+ * size of the GTT block, in bytes
+ */
+static ssize_t amdgpu_mem_info_gtt_used_show(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = ddev->dev_private;
+
+	return snprintf(buf, PAGE_SIZE, "%llu\n",
+			amdgpu_gtt_mgr_usage(&adev->mman.bdev.man[TTM_PL_TT]));
+}
+
+static DEVICE_ATTR(mem_info_gtt_total, S_IRUGO,
+	           amdgpu_mem_info_gtt_total_show, NULL);
+static DEVICE_ATTR(mem_info_gtt_used, S_IRUGO,
+	           amdgpu_mem_info_gtt_used_show, NULL);
+
 /**
  * amdgpu_gtt_mgr_init - init GTT manager and DRM MM
  *
@@ -50,6 +91,7 @@ static int amdgpu_gtt_mgr_init(struct ttm_mem_type_manager *man,
 	struct amdgpu_device *adev = amdgpu_ttm_adev(man->bdev);
 	struct amdgpu_gtt_mgr *mgr;
 	uint64_t start, size;
+	int ret;
 
 	mgr = kzalloc(sizeof(*mgr), GFP_KERNEL);
 	if (!mgr)
@@ -61,6 +103,18 @@ static int amdgpu_gtt_mgr_init(struct ttm_mem_type_manager *man,
 	spin_lock_init(&mgr->lock);
 	atomic64_set(&mgr->available, p_size);
 	man->priv = mgr;
+
+	ret = device_create_file(adev->dev, &dev_attr_mem_info_gtt_total);
+	if (ret) {
+		DRM_ERROR("Failed to create device file mem_info_gtt_total\n");
+		return ret;
+	}
+	ret = device_create_file(adev->dev, &dev_attr_mem_info_gtt_used);
+	if (ret) {
+		DRM_ERROR("Failed to create device file mem_info_gtt_used\n");
+		return ret;
+	}
+
 	return 0;
 }
 
@@ -74,12 +128,17 @@ static int amdgpu_gtt_mgr_init(struct ttm_mem_type_manager *man,
  */
 static int amdgpu_gtt_mgr_fini(struct ttm_mem_type_manager *man)
 {
+	struct amdgpu_device *adev = amdgpu_ttm_adev(man->bdev);
 	struct amdgpu_gtt_mgr *mgr = man->priv;
 	spin_lock(&mgr->lock);
 	drm_mm_takedown(&mgr->mm);
 	spin_unlock(&mgr->lock);
 	kfree(mgr);
 	man->priv = NULL;
+
+	device_remove_file(adev->dev, &dev_attr_mem_info_gtt_total);
+	device_remove_file(adev->dev, &dev_attr_mem_info_gtt_used);
+
 	return 0;
 }
 

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

@@ -142,6 +142,7 @@ void amdgpu_ih_ring_fini(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih)
  */
 int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih)
 {
+	unsigned int count = AMDGPU_IH_MAX_NUM_IVS;
 	u32 wptr;
 
 	if (!ih->enabled || adev->shutdown)
@@ -159,7 +160,7 @@ int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih)
 	/* Order reading of wptr vs. reading of IH ring data */
 	rmb();
 
-	while (ih->rptr != wptr) {
+	while (ih->rptr != wptr && --count) {
 		amdgpu_irq_dispatch(adev, ih);
 		ih->rptr &= ih->ptr_mask;
 	}

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

@@ -24,6 +24,9 @@
 #ifndef __AMDGPU_IH_H__
 #define __AMDGPU_IH_H__
 
+/* Maximum number of IVs processed at once */
+#define AMDGPU_IH_MAX_NUM_IVS	32
+
 struct amdgpu_device;
 struct amdgpu_iv_entry;
 

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

@@ -39,6 +39,7 @@
 #include "amdgpu_amdkfd.h"
 #include "amdgpu_gem.h"
 #include "amdgpu_display.h"
+#include "amdgpu_ras.h"
 
 static void amdgpu_unregister_gpu_instance(struct amdgpu_device *adev)
 {
@@ -296,6 +297,17 @@ static int amdgpu_firmware_info(struct drm_amdgpu_info_firmware *fw_info,
 		fw_info->ver = adev->pm.fw_version;
 		fw_info->feature = 0;
 		break;
+	case AMDGPU_INFO_FW_TA:
+		if (query_fw->index > 1)
+			return -EINVAL;
+		if (query_fw->index == 0) {
+			fw_info->ver = adev->psp.ta_fw_version;
+			fw_info->feature = adev->psp.ta_xgmi_ucode_version;
+		} else {
+			fw_info->ver = adev->psp.ta_fw_version;
+			fw_info->feature = adev->psp.ta_ras_ucode_version;
+		}
+		break;
 	case AMDGPU_INFO_FW_SDMA:
 		if (query_fw->index >= adev->sdma.num_instances)
 			return -EINVAL;
@@ -909,6 +921,18 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file
 	case AMDGPU_INFO_VRAM_LOST_COUNTER:
 		ui32 = atomic_read(&adev->vram_lost_counter);
 		return copy_to_user(out, &ui32, min(size, 4u)) ? -EFAULT : 0;
+	case AMDGPU_INFO_RAS_ENABLED_FEATURES: {
+		struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+		uint64_t ras_mask;
+
+		if (!ras)
+			return -EINVAL;
+		ras_mask = (uint64_t)ras->supported << 32 | ras->features;
+
+		return copy_to_user(out, &ras_mask,
+				min_t(u64, size, sizeof(ras_mask))) ?
+			-EFAULT : 0;
+	}
 	default:
 		DRM_DEBUG_KMS("Invalid request %d\n", info->query);
 		return -EINVAL;
@@ -1328,6 +1352,16 @@ static int amdgpu_debugfs_firmware_info(struct seq_file *m, void *data)
 	seq_printf(m, "ASD feature version: %u, firmware version: 0x%08x\n",
 		   fw_info.feature, fw_info.ver);
 
+	query_fw.fw_type = AMDGPU_INFO_FW_TA;
+	for (i = 0; i < 2; i++) {
+		query_fw.index = i;
+		ret = amdgpu_firmware_info(&fw_info, &query_fw, adev);
+		if (ret)
+			continue;
+		seq_printf(m, "TA %s feature version: %u, firmware version: 0x%08x\n",
+				i ? "RAS" : "XGMI", fw_info.feature, fw_info.ver);
+	}
+
 	/* SMC */
 	query_fw.fw_type = AMDGPU_INFO_FW_SMC;
 	ret = amdgpu_firmware_info(&fw_info, &query_fw, adev);

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

@@ -72,6 +72,8 @@ struct amdgpu_bo_va {
 
 	/* If the mappings are cleared or filled */
 	bool				cleared;
+
+	bool				is_xgmi;
 };
 
 struct amdgpu_bo {

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

@@ -28,6 +28,7 @@
 #include "amdgpu_pm.h"
 #include "amdgpu_dpm.h"
 #include "amdgpu_display.h"
+#include "amdgpu_smu.h"
 #include "atom.h"
 #include <linux/power_supply.h>
 #include <linux/hwmon.h>
@@ -80,6 +81,27 @@ void amdgpu_pm_acpi_event_handler(struct amdgpu_device *adev)
 	}
 }
 
+int amdgpu_dpm_read_sensor(struct amdgpu_device *adev, enum amd_pp_sensors sensor,
+			   void *data, uint32_t *size)
+{
+	int ret = 0;
+
+	if (!data || !size)
+		return -EINVAL;
+
+	if (is_support_sw_smu(adev))
+		ret = smu_read_sensor(&adev->smu, sensor, data, size);
+	else {
+		if (adev->powerplay.pp_funcs && adev->powerplay.pp_funcs->read_sensor)
+			ret = adev->powerplay.pp_funcs->read_sensor((adev)->powerplay.pp_handle,
+								    sensor, data, size);
+		else
+			ret = -EINVAL;
+	}
+
+	return ret;
+}
+
 /**
  * DOC: power_dpm_state
  *
@@ -122,7 +144,9 @@ 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->powerplay.pp_funcs->get_current_power_state)
+	if (adev->smu.ppt_funcs->get_current_power_state)
+		pm = amdgpu_smu_get_current_power_state(adev);
+	else if (adev->powerplay.pp_funcs->get_current_power_state)
 		pm = amdgpu_dpm_get_current_power_state(adev);
 	else
 		pm = adev->pm.dpm.user_state;
@@ -240,7 +264,9 @@ static ssize_t amdgpu_get_dpm_forced_performance_level(struct device *dev,
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return snprintf(buf, PAGE_SIZE, "off\n");
 
-	if (adev->powerplay.pp_funcs->get_performance_level)
+	if (is_support_sw_smu(adev))
+		level = smu_get_performance_level(&adev->smu);
+	else if (adev->powerplay.pp_funcs->get_performance_level)
 		level = amdgpu_dpm_get_performance_level(adev);
 	else
 		level = adev->pm.dpm.forced_level;
@@ -273,7 +299,9 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev,
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
 
-	if (adev->powerplay.pp_funcs->get_performance_level)
+	if (is_support_sw_smu(adev))
+		current_level = smu_get_performance_level(&adev->smu);
+	else if (adev->powerplay.pp_funcs->get_performance_level)
 		current_level = amdgpu_dpm_get_performance_level(adev);
 
 	if (strncmp("low", buf, strlen("low")) == 0) {
@@ -302,7 +330,20 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev,
 	if (current_level == level)
 		return count;
 
-	if (adev->powerplay.pp_funcs->force_performance_level) {
+	if (is_support_sw_smu(adev)) {
+		mutex_lock(&adev->pm.mutex);
+		if (adev->pm.dpm.thermal_active) {
+			count = -EINVAL;
+			mutex_unlock(&adev->pm.mutex);
+			goto fail;
+		}
+		ret = smu_force_performance_level(&adev->smu, level);
+		if (ret)
+			count = -EINVAL;
+		else
+			adev->pm.dpm.forced_level = level;
+		mutex_unlock(&adev->pm.mutex);
+	} else if (adev->powerplay.pp_funcs->force_performance_level) {
 		mutex_lock(&adev->pm.mutex);
 		if (adev->pm.dpm.thermal_active) {
 			count = -EINVAL;
@@ -328,9 +369,13 @@ static ssize_t amdgpu_get_pp_num_states(struct device *dev,
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
 	struct pp_states_info data;
-	int i, buf_len;
+	int i, buf_len, ret;
 
-	if (adev->powerplay.pp_funcs->get_pp_num_states)
+	if (is_support_sw_smu(adev)) {
+		ret = smu_get_power_num_states(&adev->smu, &data);
+		if (ret)
+			return ret;
+	} else 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);
@@ -351,23 +396,29 @@ static ssize_t amdgpu_get_pp_cur_state(struct device *dev,
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
 	struct pp_states_info data;
+	struct smu_context *smu = &adev->smu;
 	enum amd_pm_state_type pm = 0;
-	int i = 0;
+	int i = 0, ret = 0;
 
-	if (adev->powerplay.pp_funcs->get_current_power_state
+	if (is_support_sw_smu(adev)) {
+		pm = smu_get_current_power_state(smu);
+		ret = smu_get_power_num_states(smu, &data);
+		if (ret)
+			return ret;
+	} else 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);
-
-		for (i = 0; i < data.nums; i++) {
-			if (pm == data.states[i])
-				break;
-		}
-
-		if (i == data.nums)
-			i = -EINVAL;
 	}
 
+	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);
 }
 
@@ -397,6 +448,8 @@ static ssize_t amdgpu_set_pp_force_state(struct device *dev,
 
 	if (strlen(buf) == 1)
 		adev->pp_force_state_enabled = false;
+	else if (is_support_sw_smu(adev))
+		adev->pp_force_state_enabled = false;
 	else if (adev->powerplay.pp_funcs->dispatch_tasks &&
 			adev->powerplay.pp_funcs->get_pp_num_states) {
 		struct pp_states_info data;
@@ -442,7 +495,12 @@ static ssize_t amdgpu_get_pp_table(struct device *dev,
 	char *table = NULL;
 	int size;
 
-	if (adev->powerplay.pp_funcs->get_pp_table)
+	if (is_support_sw_smu(adev)) {
+		size = smu_sys_get_pp_table(&adev->smu, (void **)&table);
+		if (size < 0)
+			return size;
+	}
+	else if (adev->powerplay.pp_funcs->get_pp_table)
 		size = amdgpu_dpm_get_pp_table(adev, &table);
 	else
 		return 0;
@@ -462,8 +520,13 @@ 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;
+	int ret = 0;
 
-	if (adev->powerplay.pp_funcs->set_pp_table)
+	if (is_support_sw_smu(adev)) {
+		ret = smu_sys_set_pp_table(&adev->smu, (void *)buf, count);
+		if (ret)
+			return ret;
+	} else if (adev->powerplay.pp_funcs->set_pp_table)
 		amdgpu_dpm_set_pp_table(adev, buf, count);
 
 	return count;
@@ -586,19 +649,29 @@ static ssize_t amdgpu_set_pp_od_clk_voltage(struct device *dev,
 			tmp_str++;
 	}
 
-	if (adev->powerplay.pp_funcs->odn_edit_dpm_table)
-		ret = amdgpu_dpm_odn_edit_dpm_table(adev, type,
+	if (is_support_sw_smu(adev)) {
+		ret = smu_od_edit_dpm_table(&adev->smu, type,
+					    parameter, parameter_size);
+
+		if (ret)
+			return -EINVAL;
+	} else {
+		if (adev->powerplay.pp_funcs->odn_edit_dpm_table)
+			ret = amdgpu_dpm_odn_edit_dpm_table(adev, type,
 						parameter, parameter_size);
 
-	if (ret)
-		return -EINVAL;
-
-	if (type == PP_OD_COMMIT_DPM_TABLE) {
-		if (adev->powerplay.pp_funcs->dispatch_tasks) {
-			amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL);
-			return count;
-		} else {
+		if (ret)
 			return -EINVAL;
+
+		if (type == PP_OD_COMMIT_DPM_TABLE) {
+			if (adev->powerplay.pp_funcs->dispatch_tasks) {
+				amdgpu_dpm_dispatch_task(adev,
+						AMD_PP_TASK_READJUST_POWER_STATE,
+						NULL);
+				return count;
+			} else {
+				return -EINVAL;
+			}
 		}
 	}
 
@@ -613,7 +686,13 @@ static ssize_t amdgpu_get_pp_od_clk_voltage(struct device *dev,
 	struct amdgpu_device *adev = ddev->dev_private;
 	uint32_t size = 0;
 
-	if (adev->powerplay.pp_funcs->print_clock_levels) {
+	if (is_support_sw_smu(adev)) {
+		size = smu_print_clk_levels(&adev->smu, OD_SCLK, buf);
+		size += smu_print_clk_levels(&adev->smu, OD_MCLK, buf+size);
+		size += smu_print_clk_levels(&adev->smu, OD_VDDC_CURVE, buf+size);
+		size += smu_print_clk_levels(&adev->smu, OD_RANGE, buf+size);
+		return size;
+	} else if (adev->powerplay.pp_funcs->print_clock_levels) {
 		size = amdgpu_dpm_print_clock_levels(adev, OD_SCLK, buf);
 		size += amdgpu_dpm_print_clock_levels(adev, OD_MCLK, buf+size);
 		size += amdgpu_dpm_print_clock_levels(adev, OD_VDDC_CURVE, buf+size);
@@ -711,7 +790,9 @@ 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;
 
-	if (adev->powerplay.pp_funcs->print_clock_levels)
+	if (is_support_sw_smu(adev))
+		return smu_print_clk_levels(&adev->smu, PP_SCLK, buf);
+	else 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");
@@ -767,7 +848,9 @@ static ssize_t amdgpu_set_pp_dpm_sclk(struct device *dev,
 	if (ret)
 		return ret;
 
-	if (adev->powerplay.pp_funcs->force_clock_level)
+	if (is_support_sw_smu(adev))
+		ret = smu_force_clk_levels(&adev->smu, PP_SCLK, mask);
+	else if (adev->powerplay.pp_funcs->force_clock_level)
 		ret = amdgpu_dpm_force_clock_level(adev, PP_SCLK, mask);
 
 	if (ret)
@@ -783,7 +866,9 @@ 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;
 
-	if (adev->powerplay.pp_funcs->print_clock_levels)
+	if (is_support_sw_smu(adev))
+		return smu_print_clk_levels(&adev->smu, PP_MCLK, buf);
+	else 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");
@@ -803,7 +888,9 @@ static ssize_t amdgpu_set_pp_dpm_mclk(struct device *dev,
 	if (ret)
 		return ret;
 
-	if (adev->powerplay.pp_funcs->force_clock_level)
+	if (is_support_sw_smu(adev))
+		ret = smu_force_clk_levels(&adev->smu, PP_MCLK, mask);
+	else if (adev->powerplay.pp_funcs->force_clock_level)
 		ret = amdgpu_dpm_force_clock_level(adev, PP_MCLK, mask);
 
 	if (ret)
@@ -819,7 +906,9 @@ static ssize_t amdgpu_get_pp_dpm_socclk(struct device *dev,
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
 
-	if (adev->powerplay.pp_funcs->print_clock_levels)
+	if (is_support_sw_smu(adev))
+		return smu_print_clk_levels(&adev->smu, PP_SOCCLK, buf);
+	else if (adev->powerplay.pp_funcs->print_clock_levels)
 		return amdgpu_dpm_print_clock_levels(adev, PP_SOCCLK, buf);
 	else
 		return snprintf(buf, PAGE_SIZE, "\n");
@@ -839,7 +928,9 @@ static ssize_t amdgpu_set_pp_dpm_socclk(struct device *dev,
 	if (ret)
 		return ret;
 
-	if (adev->powerplay.pp_funcs->force_clock_level)
+	if (is_support_sw_smu(adev))
+		ret = smu_force_clk_levels(&adev->smu, PP_SOCCLK, mask);
+	else if (adev->powerplay.pp_funcs->force_clock_level)
 		ret = amdgpu_dpm_force_clock_level(adev, PP_SOCCLK, mask);
 
 	if (ret)
@@ -855,7 +946,9 @@ static ssize_t amdgpu_get_pp_dpm_fclk(struct device *dev,
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
 
-	if (adev->powerplay.pp_funcs->print_clock_levels)
+	if (is_support_sw_smu(adev))
+		return smu_print_clk_levels(&adev->smu, PP_FCLK, buf);
+	else if (adev->powerplay.pp_funcs->print_clock_levels)
 		return amdgpu_dpm_print_clock_levels(adev, PP_FCLK, buf);
 	else
 		return snprintf(buf, PAGE_SIZE, "\n");
@@ -875,7 +968,9 @@ static ssize_t amdgpu_set_pp_dpm_fclk(struct device *dev,
 	if (ret)
 		return ret;
 
-	if (adev->powerplay.pp_funcs->force_clock_level)
+	if (is_support_sw_smu(adev))
+		ret = smu_force_clk_levels(&adev->smu, PP_FCLK, mask);
+	else if (adev->powerplay.pp_funcs->force_clock_level)
 		ret = amdgpu_dpm_force_clock_level(adev, PP_FCLK, mask);
 
 	if (ret)
@@ -891,7 +986,9 @@ static ssize_t amdgpu_get_pp_dpm_dcefclk(struct device *dev,
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
 
-	if (adev->powerplay.pp_funcs->print_clock_levels)
+	if (is_support_sw_smu(adev))
+		return smu_print_clk_levels(&adev->smu, PP_DCEFCLK, buf);
+	else if (adev->powerplay.pp_funcs->print_clock_levels)
 		return amdgpu_dpm_print_clock_levels(adev, PP_DCEFCLK, buf);
 	else
 		return snprintf(buf, PAGE_SIZE, "\n");
@@ -911,7 +1008,9 @@ static ssize_t amdgpu_set_pp_dpm_dcefclk(struct device *dev,
 	if (ret)
 		return ret;
 
-	if (adev->powerplay.pp_funcs->force_clock_level)
+	if (is_support_sw_smu(adev))
+		ret = smu_force_clk_levels(&adev->smu, PP_DCEFCLK, mask);
+	else if (adev->powerplay.pp_funcs->force_clock_level)
 		ret = amdgpu_dpm_force_clock_level(adev, PP_DCEFCLK, mask);
 
 	if (ret)
@@ -927,7 +1026,9 @@ 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;
 
-	if (adev->powerplay.pp_funcs->print_clock_levels)
+	if (is_support_sw_smu(adev))
+		return smu_print_clk_levels(&adev->smu, PP_PCIE, buf);
+	else 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");
@@ -947,7 +1048,9 @@ static ssize_t amdgpu_set_pp_dpm_pcie(struct device *dev,
 	if (ret)
 		return ret;
 
-	if (adev->powerplay.pp_funcs->force_clock_level)
+	if (is_support_sw_smu(adev))
+		ret = smu_force_clk_levels(&adev->smu, PP_PCIE, mask);
+	else if (adev->powerplay.pp_funcs->force_clock_level)
 		ret = amdgpu_dpm_force_clock_level(adev, PP_PCIE, mask);
 
 	if (ret)
@@ -964,7 +1067,9 @@ static ssize_t amdgpu_get_pp_sclk_od(struct device *dev,
 	struct amdgpu_device *adev = ddev->dev_private;
 	uint32_t value = 0;
 
-	if (adev->powerplay.pp_funcs->get_sclk_od)
+	if (is_support_sw_smu(adev))
+		value = smu_get_od_percentage(&(adev->smu), OD_SCLK);
+	else if (adev->powerplay.pp_funcs->get_sclk_od)
 		value = amdgpu_dpm_get_sclk_od(adev);
 
 	return snprintf(buf, PAGE_SIZE, "%d\n", value);
@@ -986,14 +1091,19 @@ 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->powerplay.pp_funcs->dispatch_tasks) {
-		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL);
+	if (is_support_sw_smu(adev)) {
+		value = smu_set_od_percentage(&(adev->smu), OD_SCLK, (uint32_t)value);
 	} else {
-		adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps;
-		amdgpu_pm_compute_clocks(adev);
+		if (adev->powerplay.pp_funcs->set_sclk_od)
+			amdgpu_dpm_set_sclk_od(adev, (uint32_t)value);
+
+		if (adev->powerplay.pp_funcs->dispatch_tasks) {
+			amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL);
+		} else {
+			adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps;
+			amdgpu_pm_compute_clocks(adev);
+		}
 	}
 
 fail:
@@ -1008,7 +1118,9 @@ static ssize_t amdgpu_get_pp_mclk_od(struct device *dev,
 	struct amdgpu_device *adev = ddev->dev_private;
 	uint32_t value = 0;
 
-	if (adev->powerplay.pp_funcs->get_mclk_od)
+	if (is_support_sw_smu(adev))
+		value = smu_get_od_percentage(&(adev->smu), OD_MCLK);
+	else if (adev->powerplay.pp_funcs->get_mclk_od)
 		value = amdgpu_dpm_get_mclk_od(adev);
 
 	return snprintf(buf, PAGE_SIZE, "%d\n", value);
@@ -1030,14 +1142,19 @@ 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->powerplay.pp_funcs->dispatch_tasks) {
-		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL);
+	if (is_support_sw_smu(adev)) {
+		value = smu_set_od_percentage(&(adev->smu), OD_MCLK, (uint32_t)value);
 	} else {
-		adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps;
-		amdgpu_pm_compute_clocks(adev);
+		if (adev->powerplay.pp_funcs->set_mclk_od)
+			amdgpu_dpm_set_mclk_od(adev, (uint32_t)value);
+
+		if (adev->powerplay.pp_funcs->dispatch_tasks) {
+			amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL);
+		} else {
+			adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps;
+			amdgpu_pm_compute_clocks(adev);
+		}
 	}
 
 fail:
@@ -1071,7 +1188,9 @@ static ssize_t amdgpu_get_pp_power_profile_mode(struct device *dev,
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
 
-	if (adev->powerplay.pp_funcs->get_power_profile_mode)
+	if (is_support_sw_smu(adev))
+		return smu_get_power_profile_mode(&adev->smu, buf);
+	else if (adev->powerplay.pp_funcs->get_power_profile_mode)
 		return amdgpu_dpm_get_power_profile_mode(adev, buf);
 
 	return snprintf(buf, PAGE_SIZE, "\n");
@@ -1121,9 +1240,10 @@ static ssize_t amdgpu_set_pp_power_profile_mode(struct device *dev,
 		}
 	}
 	parameter[parameter_size] = profile_mode;
-	if (adev->powerplay.pp_funcs->set_power_profile_mode)
+	if (is_support_sw_smu(adev))
+		ret = smu_set_power_profile_mode(&adev->smu, parameter, parameter_size);
+	else if (adev->powerplay.pp_funcs->set_power_profile_mode)
 		ret = amdgpu_dpm_set_power_profile_mode(adev, parameter, parameter_size);
-
 	if (!ret)
 		return count;
 fail:
@@ -1146,14 +1266,10 @@ static ssize_t amdgpu_get_busy_percent(struct device *dev,
 	struct amdgpu_device *adev = ddev->dev_private;
 	int r, value, size = sizeof(value);
 
-	/* sanity check PP is enabled */
-	if (!(adev->powerplay.pp_funcs &&
-	      adev->powerplay.pp_funcs->read_sensor))
-		return -EINVAL;
-
 	/* read the IP busy sensor */
 	r = amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_GPU_LOAD,
 				   (void *)&value, &size);
+
 	if (r)
 		return r;
 
@@ -1247,11 +1363,6 @@ static ssize_t amdgpu_hwmon_show_temp(struct device *dev,
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
 
-	/* sanity check PP is enabled */
-	if (!(adev->powerplay.pp_funcs &&
-	      adev->powerplay.pp_funcs->read_sensor))
-		return -EINVAL;
-
 	/* get the temperature */
 	r = amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_GPU_TEMP,
 				   (void *)&temp, &size);
@@ -1283,11 +1394,14 @@ static ssize_t amdgpu_hwmon_get_pwm1_enable(struct device *dev,
 {
 	struct amdgpu_device *adev = dev_get_drvdata(dev);
 	u32 pwm_mode = 0;
+	if (is_support_sw_smu(adev)) {
+		pwm_mode = smu_get_fan_control_mode(&adev->smu);
+	} else {
+		if (!adev->powerplay.pp_funcs->get_fan_control_mode)
+			return -EINVAL;
 
-	if (!adev->powerplay.pp_funcs->get_fan_control_mode)
-		return -EINVAL;
-
-	pwm_mode = amdgpu_dpm_get_fan_control_mode(adev);
+		pwm_mode = amdgpu_dpm_get_fan_control_mode(adev);
+	}
 
 	return sprintf(buf, "%i\n", pwm_mode);
 }
@@ -1306,14 +1420,22 @@ static ssize_t amdgpu_hwmon_set_pwm1_enable(struct device *dev,
 	     (adev->ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
 
-	if (!adev->powerplay.pp_funcs->set_fan_control_mode)
-		return -EINVAL;
+	if (is_support_sw_smu(adev)) {
+		err = kstrtoint(buf, 10, &value);
+		if (err)
+			return err;
 
-	err = kstrtoint(buf, 10, &value);
-	if (err)
-		return err;
+		smu_set_fan_control_mode(&adev->smu, value);
+	} else {
+		if (!adev->powerplay.pp_funcs->set_fan_control_mode)
+			return -EINVAL;
 
-	amdgpu_dpm_set_fan_control_mode(adev, value);
+		err = kstrtoint(buf, 10, &value);
+		if (err)
+			return err;
+
+		amdgpu_dpm_set_fan_control_mode(adev, value);
+	}
 
 	return count;
 }
@@ -1345,8 +1467,10 @@ static ssize_t amdgpu_hwmon_set_pwm1(struct device *dev,
 	if  ((adev->flags & AMD_IS_PX) &&
 	     (adev->ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
-
-	pwm_mode = amdgpu_dpm_get_fan_control_mode(adev);
+	if (is_support_sw_smu(adev))
+		pwm_mode = smu_get_fan_control_mode(&adev->smu);
+	else
+		pwm_mode = amdgpu_dpm_get_fan_control_mode(adev);
 	if (pwm_mode != AMD_FAN_CTRL_MANUAL) {
 		pr_info("manual fan speed control should be enabled first\n");
 		return -EINVAL;
@@ -1358,7 +1482,11 @@ static ssize_t amdgpu_hwmon_set_pwm1(struct device *dev,
 
 	value = (value * 100) / 255;
 
-	if (adev->powerplay.pp_funcs->set_fan_speed_percent) {
+	if (is_support_sw_smu(adev)) {
+		err = smu_set_fan_speed_percent(&adev->smu, value);
+		if (err)
+			return err;
+	} else if (adev->powerplay.pp_funcs->set_fan_speed_percent) {
 		err = amdgpu_dpm_set_fan_speed_percent(adev, value);
 		if (err)
 			return err;
@@ -1380,7 +1508,11 @@ static ssize_t amdgpu_hwmon_get_pwm1(struct device *dev,
 	     (adev->ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
 
-	if (adev->powerplay.pp_funcs->get_fan_speed_percent) {
+	if (is_support_sw_smu(adev)) {
+		err = smu_get_fan_speed_percent(&adev->smu, &speed);
+		if (err)
+			return err;
+	} else if (adev->powerplay.pp_funcs->get_fan_speed_percent) {
 		err = amdgpu_dpm_get_fan_speed_percent(adev, &speed);
 		if (err)
 			return err;
@@ -1404,7 +1536,11 @@ static ssize_t amdgpu_hwmon_get_fan1_input(struct device *dev,
 	     (adev->ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
 
-	if (adev->powerplay.pp_funcs->get_fan_speed_rpm) {
+	if (is_support_sw_smu(adev)) {
+		err = smu_get_current_rpm(&adev->smu, &speed);
+		if (err)
+			return err;
+	} else if (adev->powerplay.pp_funcs->get_fan_speed_rpm) {
 		err = amdgpu_dpm_get_fan_speed_rpm(adev, &speed);
 		if (err)
 			return err;
@@ -1422,9 +1558,6 @@ static ssize_t amdgpu_hwmon_get_fan1_min(struct device *dev,
 	u32 size = sizeof(min_rpm);
 	int r;
 
-	if (!adev->powerplay.pp_funcs->read_sensor)
-		return -EINVAL;
-
 	r = amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_MIN_FAN_RPM,
 				   (void *)&min_rpm, &size);
 	if (r)
@@ -1442,9 +1575,6 @@ static ssize_t amdgpu_hwmon_get_fan1_max(struct device *dev,
 	u32 size = sizeof(max_rpm);
 	int r;
 
-	if (!adev->powerplay.pp_funcs->read_sensor)
-		return -EINVAL;
-
 	r = amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_MAX_FAN_RPM,
 				   (void *)&max_rpm, &size);
 	if (r)
@@ -1466,7 +1596,11 @@ static ssize_t amdgpu_hwmon_get_fan1_target(struct device *dev,
 	     (adev->ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
 
-	if (adev->powerplay.pp_funcs->get_fan_speed_rpm) {
+	if (is_support_sw_smu(adev)) {
+		err = smu_get_current_rpm(&adev->smu, &rpm);
+		if (err)
+			return err;
+	} else if (adev->powerplay.pp_funcs->get_fan_speed_rpm) {
 		err = amdgpu_dpm_get_fan_speed_rpm(adev, &rpm);
 		if (err)
 			return err;
@@ -1484,7 +1618,11 @@ static ssize_t amdgpu_hwmon_set_fan1_target(struct device *dev,
 	u32 value;
 	u32 pwm_mode;
 
-	pwm_mode = amdgpu_dpm_get_fan_control_mode(adev);
+	if (is_support_sw_smu(adev))
+		pwm_mode = smu_get_fan_control_mode(&adev->smu);
+	else
+		pwm_mode = amdgpu_dpm_get_fan_control_mode(adev);
+
 	if (pwm_mode != AMD_FAN_CTRL_MANUAL)
 		return -ENODATA;
 
@@ -1497,7 +1635,11 @@ static ssize_t amdgpu_hwmon_set_fan1_target(struct device *dev,
 	if (err)
 		return err;
 
-	if (adev->powerplay.pp_funcs->set_fan_speed_rpm) {
+	if (is_support_sw_smu(adev)) {
+		err = smu_set_fan_speed_rpm(&adev->smu, value);
+		if (err)
+			return err;
+	} else if (adev->powerplay.pp_funcs->set_fan_speed_rpm) {
 		err = amdgpu_dpm_set_fan_speed_rpm(adev, value);
 		if (err)
 			return err;
@@ -1513,11 +1655,14 @@ static ssize_t amdgpu_hwmon_get_fan1_enable(struct device *dev,
 	struct amdgpu_device *adev = dev_get_drvdata(dev);
 	u32 pwm_mode = 0;
 
-	if (!adev->powerplay.pp_funcs->get_fan_control_mode)
-		return -EINVAL;
-
-	pwm_mode = amdgpu_dpm_get_fan_control_mode(adev);
+	if (is_support_sw_smu(adev)) {
+		pwm_mode = smu_get_fan_control_mode(&adev->smu);
+	} else {
+		if (!adev->powerplay.pp_funcs->get_fan_control_mode)
+			return -EINVAL;
 
+		pwm_mode = amdgpu_dpm_get_fan_control_mode(adev);
+	}
 	return sprintf(buf, "%i\n", pwm_mode == AMD_FAN_CTRL_AUTO ? 0 : 1);
 }
 
@@ -1536,8 +1681,6 @@ static ssize_t amdgpu_hwmon_set_fan1_enable(struct device *dev,
 	     (adev->ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
 
-	if (!adev->powerplay.pp_funcs->set_fan_control_mode)
-		return -EINVAL;
 
 	err = kstrtoint(buf, 10, &value);
 	if (err)
@@ -1550,7 +1693,13 @@ static ssize_t amdgpu_hwmon_set_fan1_enable(struct device *dev,
 	else
 		return -EINVAL;
 
-	amdgpu_dpm_set_fan_control_mode(adev, pwm_mode);
+	if (is_support_sw_smu(adev)) {
+		smu_set_fan_control_mode(&adev->smu, pwm_mode);
+	} else {
+		if (!adev->powerplay.pp_funcs->set_fan_control_mode)
+			return -EINVAL;
+		amdgpu_dpm_set_fan_control_mode(adev, pwm_mode);
+	}
 
 	return count;
 }
@@ -1569,11 +1718,6 @@ static ssize_t amdgpu_hwmon_show_vddgfx(struct device *dev,
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
 
-	/* sanity check PP is enabled */
-	if (!(adev->powerplay.pp_funcs &&
-	      adev->powerplay.pp_funcs->read_sensor))
-	      return -EINVAL;
-
 	/* get the voltage */
 	r = amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_VDDGFX,
 				   (void *)&vddgfx, &size);
@@ -1608,11 +1752,6 @@ static ssize_t amdgpu_hwmon_show_vddnb(struct device *dev,
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
 
-	/* sanity check PP is enabled */
-	if (!(adev->powerplay.pp_funcs &&
-	      adev->powerplay.pp_funcs->read_sensor))
-	      return -EINVAL;
-
 	/* get the voltage */
 	r = amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_VDDNB,
 				   (void *)&vddnb, &size);
@@ -1644,11 +1783,6 @@ static ssize_t amdgpu_hwmon_show_power_avg(struct device *dev,
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
 
-	/* sanity check PP is enabled */
-	if (!(adev->powerplay.pp_funcs &&
-	      adev->powerplay.pp_funcs->read_sensor))
-	      return -EINVAL;
-
 	/* get the voltage */
 	r = amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_GPU_POWER,
 				   (void *)&query, &size);
@@ -1675,7 +1809,10 @@ static ssize_t amdgpu_hwmon_show_power_cap_max(struct device *dev,
 	struct amdgpu_device *adev = dev_get_drvdata(dev);
 	uint32_t limit = 0;
 
-	if (adev->powerplay.pp_funcs && adev->powerplay.pp_funcs->get_power_limit) {
+	if (is_support_sw_smu(adev)) {
+		smu_get_power_limit(&adev->smu, &limit, true);
+		return snprintf(buf, PAGE_SIZE, "%u\n", limit * 1000000);
+	} else if (adev->powerplay.pp_funcs && adev->powerplay.pp_funcs->get_power_limit) {
 		adev->powerplay.pp_funcs->get_power_limit(adev->powerplay.pp_handle, &limit, true);
 		return snprintf(buf, PAGE_SIZE, "%u\n", limit * 1000000);
 	} else {
@@ -1690,7 +1827,10 @@ static ssize_t amdgpu_hwmon_show_power_cap(struct device *dev,
 	struct amdgpu_device *adev = dev_get_drvdata(dev);
 	uint32_t limit = 0;
 
-	if (adev->powerplay.pp_funcs && adev->powerplay.pp_funcs->get_power_limit) {
+	if (is_support_sw_smu(adev)) {
+		smu_get_power_limit(&adev->smu, &limit, false);
+		return snprintf(buf, PAGE_SIZE, "%u\n", limit * 1000000);
+	} else if (adev->powerplay.pp_funcs && adev->powerplay.pp_funcs->get_power_limit) {
 		adev->powerplay.pp_funcs->get_power_limit(adev->powerplay.pp_handle, &limit, false);
 		return snprintf(buf, PAGE_SIZE, "%u\n", limit * 1000000);
 	} else {
@@ -1713,7 +1853,9 @@ static ssize_t amdgpu_hwmon_set_power_cap(struct device *dev,
 		return err;
 
 	value = value / 1000000; /* convert to Watt */
-	if (adev->powerplay.pp_funcs && adev->powerplay.pp_funcs->set_power_limit) {
+	if (is_support_sw_smu(adev)) {
+		adev->smu.funcs->set_power_limit(&adev->smu, value);
+	} else if (adev->powerplay.pp_funcs && adev->powerplay.pp_funcs->set_power_limit) {
 		err = adev->powerplay.pp_funcs->set_power_limit(adev->powerplay.pp_handle, value);
 		if (err)
 			return err;
@@ -1967,18 +2109,20 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj,
 	     attr == &sensor_dev_attr_fan1_enable.dev_attr.attr))
 		return 0;
 
-	/* mask fan attributes if we have no bindings for this asic to expose */
-	if ((!adev->powerplay.pp_funcs->get_fan_speed_percent &&
-	     attr == &sensor_dev_attr_pwm1.dev_attr.attr) || /* can't query fan */
-	    (!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 (!is_support_sw_smu(adev)) {
+		/* mask fan attributes if we have no bindings for this asic to expose */
+		if ((!adev->powerplay.pp_funcs->get_fan_speed_percent &&
+		     attr == &sensor_dev_attr_pwm1.dev_attr.attr) || /* can't query fan */
+		    (!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->powerplay.pp_funcs->set_fan_speed_percent &&
-	     attr == &sensor_dev_attr_pwm1.dev_attr.attr) || /* can't manage fan */
-	    (!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;
+		if ((!adev->powerplay.pp_funcs->set_fan_speed_percent &&
+		     attr == &sensor_dev_attr_pwm1.dev_attr.attr) || /* can't manage fan */
+		    (!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;
+	}
 
 	if ((adev->flags & AMD_IS_APU) &&
 	    (attr == &sensor_dev_attr_power1_average.dev_attr.attr ||
@@ -1987,20 +2131,22 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj,
 	     attr == &sensor_dev_attr_power1_cap.dev_attr.attr))
 		return 0;
 
-	/* hide max/min values if we can't both query and manage the fan */
-	if ((!adev->powerplay.pp_funcs->set_fan_speed_percent &&
-	     !adev->powerplay.pp_funcs->get_fan_speed_percent) &&
-	     (!adev->powerplay.pp_funcs->set_fan_speed_rpm &&
-	     !adev->powerplay.pp_funcs->get_fan_speed_rpm) &&
-	    (attr == &sensor_dev_attr_pwm1_max.dev_attr.attr ||
-	     attr == &sensor_dev_attr_pwm1_min.dev_attr.attr))
-		return 0;
+	if (!is_support_sw_smu(adev)) {
+		/* hide max/min values if we can't both query and manage the fan */
+		if ((!adev->powerplay.pp_funcs->set_fan_speed_percent &&
+		     !adev->powerplay.pp_funcs->get_fan_speed_percent) &&
+		     (!adev->powerplay.pp_funcs->set_fan_speed_rpm &&
+		     !adev->powerplay.pp_funcs->get_fan_speed_rpm) &&
+		    (attr == &sensor_dev_attr_pwm1_max.dev_attr.attr ||
+		     attr == &sensor_dev_attr_pwm1_min.dev_attr.attr))
+			return 0;
 
-	if ((!adev->powerplay.pp_funcs->set_fan_speed_rpm &&
-	     !adev->powerplay.pp_funcs->get_fan_speed_rpm) &&
-	    (attr == &sensor_dev_attr_fan1_max.dev_attr.attr ||
-	     attr == &sensor_dev_attr_fan1_min.dev_attr.attr))
-		return 0;
+		if ((!adev->powerplay.pp_funcs->set_fan_speed_rpm &&
+		     !adev->powerplay.pp_funcs->get_fan_speed_rpm) &&
+		    (attr == &sensor_dev_attr_fan1_max.dev_attr.attr ||
+		     attr == &sensor_dev_attr_fan1_min.dev_attr.attr))
+			return 0;
+	}
 
 	/* only APUs have vddnb */
 	if (!(adev->flags & AMD_IS_APU) &&
@@ -2039,9 +2185,7 @@ void amdgpu_dpm_thermal_work_handler(struct work_struct *work)
 	if (!adev->pm.dpm_enabled)
 		return;
 
-	if (adev->powerplay.pp_funcs &&
-	    adev->powerplay.pp_funcs->read_sensor &&
-	    !amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_GPU_TEMP,
+	if (!amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_GPU_TEMP,
 				    (void *)&temp, &size)) {
 		if (temp < adev->pm.dpm.thermal.min_temp)
 			/* switch back the user state */
@@ -2267,7 +2411,13 @@ 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->powerplay.pp_funcs->set_powergating_by_smu) {
+	int ret = 0;
+	if (is_support_sw_smu(adev)) {
+	    ret = smu_dpm_set_power_gate(&adev->smu, AMD_IP_BLOCK_TYPE_UVD, enable);
+	    if (ret)
+		DRM_ERROR("[SW SMU]: dpm enable uvd failed, state = %s, ret = %d. \n",
+			  enable ? "true" : "false", ret);
+	} else if (adev->powerplay.pp_funcs->set_powergating_by_smu) {
 		/* enable/disable UVD */
 		mutex_lock(&adev->pm.mutex);
 		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_UVD, !enable);
@@ -2288,7 +2438,13 @@ void amdgpu_dpm_enable_uvd(struct amdgpu_device *adev, bool enable)
 
 void amdgpu_dpm_enable_vce(struct amdgpu_device *adev, bool enable)
 {
-	if (adev->powerplay.pp_funcs->set_powergating_by_smu) {
+	int ret = 0;
+	if (is_support_sw_smu(adev)) {
+	    ret = smu_dpm_set_power_gate(&adev->smu, AMD_IP_BLOCK_TYPE_VCE, enable);
+	    if (ret)
+		DRM_ERROR("[SW SMU]: dpm enable vce failed, state = %s, ret = %d. \n",
+			  enable ? "true" : "false", ret);
+	} else if (adev->powerplay.pp_funcs->set_powergating_by_smu) {
 		/* enable/disable VCE */
 		mutex_lock(&adev->pm.mutex);
 		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_VCE, !enable);
@@ -2413,7 +2569,8 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev)
 				"pp_power_profile_mode\n");
 		return ret;
 	}
-	if (hwmgr->od_enabled) {
+	if ((is_support_sw_smu(adev) && adev->smu.od_enabled) ||
+	    (!is_support_sw_smu(adev) && hwmgr->od_enabled)) {
 		ret = device_create_file(adev->dev,
 				&dev_attr_pp_od_clk_voltage);
 		if (ret) {
@@ -2489,7 +2646,8 @@ void amdgpu_pm_sysfs_fini(struct amdgpu_device *adev)
 	device_remove_file(adev->dev, &dev_attr_pp_mclk_od);
 	device_remove_file(adev->dev,
 			&dev_attr_pp_power_profile_mode);
-	if (hwmgr->od_enabled)
+	if ((is_support_sw_smu(adev) && adev->smu.od_enabled) ||
+	    (!is_support_sw_smu(adev) && hwmgr->od_enabled))
 		device_remove_file(adev->dev,
 				&dev_attr_pp_od_clk_voltage);
 	device_remove_file(adev->dev, &dev_attr_gpu_busy_percent);
@@ -2516,28 +2674,38 @@ void amdgpu_pm_compute_clocks(struct amdgpu_device *adev)
 			amdgpu_fence_wait_empty(ring);
 	}
 
-	if (adev->powerplay.pp_funcs->dispatch_tasks) {
-		if (!amdgpu_device_has_dc_support(adev)) {
+	if (is_support_sw_smu(adev)) {
+		struct smu_context *smu = &adev->smu;
+		struct smu_dpm_context *smu_dpm = &adev->smu.smu_dpm;
+		mutex_lock(&(smu->mutex));
+		smu_handle_task(&adev->smu,
+				smu_dpm->dpm_level,
+				AMD_PP_TASK_DISPLAY_CONFIG_CHANGE);
+		mutex_unlock(&(smu->mutex));
+	} else {
+		if (adev->powerplay.pp_funcs->dispatch_tasks) {
+			if (!amdgpu_device_has_dc_support(adev)) {
+				mutex_lock(&adev->pm.mutex);
+				amdgpu_dpm_get_active_displays(adev);
+				adev->pm.pm_display_cfg.num_display = adev->pm.dpm.new_active_crtc_count;
+				adev->pm.pm_display_cfg.vrefresh = amdgpu_dpm_get_vrefresh(adev);
+				adev->pm.pm_display_cfg.min_vblank_time = amdgpu_dpm_get_vblank_time(adev);
+				/* we have issues with mclk switching with refresh rates over 120 hz on the non-DC code. */
+				if (adev->pm.pm_display_cfg.vrefresh > 120)
+					adev->pm.pm_display_cfg.min_vblank_time = 0;
+				if (adev->powerplay.pp_funcs->display_configuration_change)
+					adev->powerplay.pp_funcs->display_configuration_change(
+									adev->powerplay.pp_handle,
+									&adev->pm.pm_display_cfg);
+				mutex_unlock(&adev->pm.mutex);
+			}
+			amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_DISPLAY_CONFIG_CHANGE, NULL);
+		} else {
 			mutex_lock(&adev->pm.mutex);
 			amdgpu_dpm_get_active_displays(adev);
-			adev->pm.pm_display_cfg.num_display = adev->pm.dpm.new_active_crtc_count;
-			adev->pm.pm_display_cfg.vrefresh = amdgpu_dpm_get_vrefresh(adev);
-			adev->pm.pm_display_cfg.min_vblank_time = amdgpu_dpm_get_vblank_time(adev);
-			/* we have issues with mclk switching with refresh rates over 120 hz on the non-DC code. */
-			if (adev->pm.pm_display_cfg.vrefresh > 120)
-				adev->pm.pm_display_cfg.min_vblank_time = 0;
-			if (adev->powerplay.pp_funcs->display_configuration_change)
-				adev->powerplay.pp_funcs->display_configuration_change(
-								adev->powerplay.pp_handle,
-								&adev->pm.pm_display_cfg);
+			amdgpu_dpm_change_power_state_locked(adev);
 			mutex_unlock(&adev->pm.mutex);
 		}
-		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_DISPLAY_CONFIG_CHANGE, NULL);
-	} else {
-		mutex_lock(&adev->pm.mutex);
-		amdgpu_dpm_get_active_displays(adev);
-		amdgpu_dpm_change_power_state_locked(adev);
-		mutex_unlock(&adev->pm.mutex);
 	}
 }
 
@@ -2553,11 +2721,6 @@ static int amdgpu_debugfs_pm_info_pp(struct seq_file *m, struct amdgpu_device *a
 	uint32_t query = 0;
 	int size;
 
-	/* sanity check PP is enabled */
-	if (!(adev->powerplay.pp_funcs &&
-	      adev->powerplay.pp_funcs->read_sensor))
-	      return -EINVAL;
-
 	/* GPU Clocks */
 	size = sizeof(value);
 	seq_printf(m, "GFX Clocks and Power:\n");
@@ -2649,7 +2812,7 @@ static int amdgpu_debugfs_pm_info(struct seq_file *m, void *data)
 	if  ((adev->flags & AMD_IS_PX) &&
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON)) {
 		seq_printf(m, "PX asic powered off\n");
-	} else if (adev->powerplay.pp_funcs->debugfs_print_current_performance_level) {
+	} else if (!is_support_sw_smu(adev) && adev->powerplay.pp_funcs->debugfs_print_current_performance_level) {
 		mutex_lock(&adev->pm.mutex);
 		if (adev->powerplay.pp_funcs->debugfs_print_current_performance_level)
 			adev->powerplay.pp_funcs->debugfs_print_current_performance_level(adev, m);

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

@@ -120,6 +120,7 @@ psp_cmd_submit_buf(struct psp_context *psp,
 {
 	int ret;
 	int index;
+	int timeout = 2000;
 
 	memset(psp->cmd_buf_mem, 0, PSP_CMD_BUFFER_SIZE);
 
@@ -133,8 +134,11 @@ psp_cmd_submit_buf(struct psp_context *psp,
 		return ret;
 	}
 
-	while (*((unsigned int *)psp->fence_buf) != index)
+	while (*((unsigned int *)psp->fence_buf) != index) {
+		if (--timeout == 0)
+			break;
 		msleep(1);
+	}
 
 	/* In some cases, psp response status is not 0 even there is no
 	 * problem while the command is submitted. Some version of PSP FW
@@ -143,12 +147,14 @@ psp_cmd_submit_buf(struct psp_context *psp,
 	 * during psp initialization to avoid breaking hw_init and it doesn't
 	 * return -EINVAL.
 	 */
-	if (psp->cmd_buf_mem->resp.status) {
+	if (psp->cmd_buf_mem->resp.status || !timeout) {
 		if (ucode)
 			DRM_WARN("failed to load ucode id (%d) ",
 				  ucode->ucode_id);
 		DRM_WARN("psp command failed and response status is (%d)\n",
 			  psp->cmd_buf_mem->resp.status);
+		if (!timeout)
+			return -EINVAL;
 	}
 
 	/* get xGMI session id from response buffer */
@@ -466,6 +472,206 @@ static int psp_xgmi_initialize(struct psp_context *psp)
 	return ret;
 }
 
+// ras begin
+static void psp_prep_ras_ta_load_cmd_buf(struct psp_gfx_cmd_resp *cmd,
+		uint64_t ras_ta_mc, uint64_t ras_mc_shared,
+		uint32_t ras_ta_size, uint32_t shared_size)
+{
+	cmd->cmd_id = GFX_CMD_ID_LOAD_TA;
+	cmd->cmd.cmd_load_ta.app_phy_addr_lo = lower_32_bits(ras_ta_mc);
+	cmd->cmd.cmd_load_ta.app_phy_addr_hi = upper_32_bits(ras_ta_mc);
+	cmd->cmd.cmd_load_ta.app_len = ras_ta_size;
+
+	cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_lo = lower_32_bits(ras_mc_shared);
+	cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_hi = upper_32_bits(ras_mc_shared);
+	cmd->cmd.cmd_load_ta.cmd_buf_len = shared_size;
+}
+
+static int psp_ras_init_shared_buf(struct psp_context *psp)
+{
+	int ret;
+
+	/*
+	 * Allocate 16k memory aligned to 4k from Frame Buffer (local
+	 * physical) for ras ta <-> Driver
+	 */
+	ret = amdgpu_bo_create_kernel(psp->adev, PSP_RAS_SHARED_MEM_SIZE,
+			PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM,
+			&psp->ras.ras_shared_bo,
+			&psp->ras.ras_shared_mc_addr,
+			&psp->ras.ras_shared_buf);
+
+	return ret;
+}
+
+static int psp_ras_load(struct psp_context *psp)
+{
+	int ret;
+	struct psp_gfx_cmd_resp *cmd;
+
+	/*
+	 * TODO: bypass the loading in sriov for now
+	 */
+	if (amdgpu_sriov_vf(psp->adev))
+		return 0;
+
+	cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL);
+	if (!cmd)
+		return -ENOMEM;
+
+	memset(psp->fw_pri_buf, 0, PSP_1_MEG);
+	memcpy(psp->fw_pri_buf, psp->ta_ras_start_addr, psp->ta_ras_ucode_size);
+
+	psp_prep_ras_ta_load_cmd_buf(cmd, psp->fw_pri_mc_addr,
+			psp->ras.ras_shared_mc_addr,
+			psp->ta_ras_ucode_size, PSP_RAS_SHARED_MEM_SIZE);
+
+	ret = psp_cmd_submit_buf(psp, NULL, cmd,
+			psp->fence_buf_mc_addr);
+
+	if (!ret) {
+		psp->ras.ras_initialized = 1;
+		psp->ras.session_id = cmd->resp.session_id;
+	}
+
+	kfree(cmd);
+
+	return ret;
+}
+
+static void psp_prep_ras_ta_unload_cmd_buf(struct psp_gfx_cmd_resp *cmd,
+						uint32_t ras_session_id)
+{
+	cmd->cmd_id = GFX_CMD_ID_UNLOAD_TA;
+	cmd->cmd.cmd_unload_ta.session_id = ras_session_id;
+}
+
+static int psp_ras_unload(struct psp_context *psp)
+{
+	int ret;
+	struct psp_gfx_cmd_resp *cmd;
+
+	/*
+	 * TODO: bypass the unloading in sriov for now
+	 */
+	if (amdgpu_sriov_vf(psp->adev))
+		return 0;
+
+	cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL);
+	if (!cmd)
+		return -ENOMEM;
+
+	psp_prep_ras_ta_unload_cmd_buf(cmd, psp->ras.session_id);
+
+	ret = psp_cmd_submit_buf(psp, NULL, cmd,
+			psp->fence_buf_mc_addr);
+
+	kfree(cmd);
+
+	return ret;
+}
+
+static void psp_prep_ras_ta_invoke_cmd_buf(struct psp_gfx_cmd_resp *cmd,
+		uint32_t ta_cmd_id,
+		uint32_t ras_session_id)
+{
+	cmd->cmd_id = GFX_CMD_ID_INVOKE_CMD;
+	cmd->cmd.cmd_invoke_cmd.session_id = ras_session_id;
+	cmd->cmd.cmd_invoke_cmd.ta_cmd_id = ta_cmd_id;
+	/* Note: cmd_invoke_cmd.buf is not used for now */
+}
+
+int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
+{
+	int ret;
+	struct psp_gfx_cmd_resp *cmd;
+
+	/*
+	 * TODO: bypass the loading in sriov for now
+	 */
+	if (amdgpu_sriov_vf(psp->adev))
+		return 0;
+
+	cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL);
+	if (!cmd)
+		return -ENOMEM;
+
+	psp_prep_ras_ta_invoke_cmd_buf(cmd, ta_cmd_id,
+			psp->ras.session_id);
+
+	ret = psp_cmd_submit_buf(psp, NULL, cmd,
+			psp->fence_buf_mc_addr);
+
+	kfree(cmd);
+
+	return ret;
+}
+
+int psp_ras_enable_features(struct psp_context *psp,
+		union ta_ras_cmd_input *info, bool enable)
+{
+	struct ta_ras_shared_memory *ras_cmd;
+	int ret;
+
+	if (!psp->ras.ras_initialized)
+		return -EINVAL;
+
+	ras_cmd = (struct ta_ras_shared_memory *)psp->ras.ras_shared_buf;
+	memset(ras_cmd, 0, sizeof(struct ta_ras_shared_memory));
+
+	if (enable)
+		ras_cmd->cmd_id = TA_RAS_COMMAND__ENABLE_FEATURES;
+	else
+		ras_cmd->cmd_id = TA_RAS_COMMAND__DISABLE_FEATURES;
+
+	ras_cmd->ras_in_message = *info;
+
+	ret = psp_ras_invoke(psp, ras_cmd->cmd_id);
+	if (ret)
+		return -EINVAL;
+
+	return ras_cmd->ras_status;
+}
+
+static int psp_ras_terminate(struct psp_context *psp)
+{
+	int ret;
+
+	if (!psp->ras.ras_initialized)
+		return 0;
+
+	ret = psp_ras_unload(psp);
+	if (ret)
+		return ret;
+
+	psp->ras.ras_initialized = 0;
+
+	/* free ras shared memory */
+	amdgpu_bo_free_kernel(&psp->ras.ras_shared_bo,
+			&psp->ras.ras_shared_mc_addr,
+			&psp->ras.ras_shared_buf);
+
+	return 0;
+}
+
+static int psp_ras_initialize(struct psp_context *psp)
+{
+	int ret;
+
+	if (!psp->ras.ras_initialized) {
+		ret = psp_ras_init_shared_buf(psp);
+		if (ret)
+			return ret;
+	}
+
+	ret = psp_ras_load(psp);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+// ras end
+
 static int psp_hw_start(struct psp_context *psp)
 {
 	struct amdgpu_device *adev = psp->adev;
@@ -473,25 +679,35 @@ static int psp_hw_start(struct psp_context *psp)
 
 	if (!amdgpu_sriov_vf(adev) || !adev->in_gpu_reset) {
 		ret = psp_bootloader_load_sysdrv(psp);
-		if (ret)
+		if (ret) {
+			DRM_ERROR("PSP load sysdrv failed!\n");
 			return ret;
+		}
 
 		ret = psp_bootloader_load_sos(psp);
-		if (ret)
+		if (ret) {
+			DRM_ERROR("PSP load sos failed!\n");
 			return ret;
+		}
 	}
 
 	ret = psp_ring_create(psp, PSP_RING_TYPE__KM);
-	if (ret)
+	if (ret) {
+		DRM_ERROR("PSP create ring failed!\n");
 		return ret;
+	}
 
 	ret = psp_tmr_load(psp);
-	if (ret)
+	if (ret) {
+		DRM_ERROR("PSP load tmr failed!\n");
 		return ret;
+	}
 
 	ret = psp_asd_load(psp);
-	if (ret)
+	if (ret) {
+		DRM_ERROR("PSP load asd failed!\n");
 		return ret;
+	}
 
 	if (adev->gmc.xgmi.num_physical_nodes > 1) {
 		ret = psp_xgmi_initialize(psp);
@@ -502,6 +718,15 @@ static int psp_hw_start(struct psp_context *psp)
 			dev_err(psp->adev->dev,
 				"XGMI: Failed to initialize XGMI session\n");
 	}
+
+
+	if (psp->adev->psp.ta_fw) {
+		ret = psp_ras_initialize(psp);
+		if (ret)
+			dev_err(psp->adev->dev,
+					"RAS: Failed to initialize RAS\n");
+	}
+
 	return 0;
 }
 
@@ -665,53 +890,52 @@ static int psp_load_fw(struct amdgpu_device *adev)
 					&psp->fence_buf_mc_addr,
 					&psp->fence_buf);
 	if (ret)
-		goto failed_mem2;
+		goto failed;
 
 	ret = amdgpu_bo_create_kernel(adev, PSP_CMD_BUFFER_SIZE, PAGE_SIZE,
 				      AMDGPU_GEM_DOMAIN_VRAM,
 				      &psp->cmd_buf_bo, &psp->cmd_buf_mc_addr,
 				      (void **)&psp->cmd_buf_mem);
 	if (ret)
-		goto failed_mem1;
+		goto failed;
 
 	memset(psp->fence_buf, 0, PSP_FENCE_BUFFER_SIZE);
 
 	ret = psp_ring_init(psp, PSP_RING_TYPE__KM);
-	if (ret)
-		goto failed_mem;
+	if (ret) {
+		DRM_ERROR("PSP ring init failed!\n");
+		goto failed;
+	}
 
 	ret = psp_tmr_init(psp);
-	if (ret)
-		goto failed_mem;
+	if (ret) {
+		DRM_ERROR("PSP tmr init failed!\n");
+		goto failed;
+	}
 
 	ret = psp_asd_init(psp);
-	if (ret)
-		goto failed_mem;
+	if (ret) {
+		DRM_ERROR("PSP asd init failed!\n");
+		goto failed;
+	}
 
 skip_memalloc:
 	ret = psp_hw_start(psp);
 	if (ret)
-		goto failed_mem;
+		goto failed;
 
 	ret = psp_np_fw_load(psp);
 	if (ret)
-		goto failed_mem;
+		goto failed;
 
 	return 0;
 
-failed_mem:
-	amdgpu_bo_free_kernel(&psp->cmd_buf_bo,
-			      &psp->cmd_buf_mc_addr,
-			      (void **)&psp->cmd_buf_mem);
-failed_mem1:
-	amdgpu_bo_free_kernel(&psp->fence_buf_bo,
-			      &psp->fence_buf_mc_addr, &psp->fence_buf);
-failed_mem2:
-	amdgpu_bo_free_kernel(&psp->fw_pri_bo,
-			      &psp->fw_pri_mc_addr, &psp->fw_pri_buf);
 failed:
-	kfree(psp->cmd);
-	psp->cmd = NULL;
+	/*
+	 * all cleanup jobs (xgmi terminate, ras terminate,
+	 * ring destroy, cmd/fence/fw buffers destory,
+	 * psp->cmd destory) are delayed to psp_hw_fini
+	 */
 	return ret;
 }
 
@@ -753,6 +977,9 @@ static int psp_hw_fini(void *handle)
 	    psp->xgmi_context.initialized == 1)
                 psp_xgmi_terminate(psp);
 
+	if (psp->adev->psp.ta_fw)
+		psp_ras_terminate(psp);
+
 	psp_ring_destroy(psp, PSP_RING_TYPE__KM);
 
 	amdgpu_bo_free_kernel(&psp->tmr_bo, &psp->tmr_mc_addr, &psp->tmr_buf);
@@ -786,6 +1013,14 @@ static int psp_suspend(void *handle)
 		}
 	}
 
+	if (psp->adev->psp.ta_fw) {
+		ret = psp_ras_terminate(psp);
+		if (ret) {
+			DRM_ERROR("Failed to terminate ras ta\n");
+			return ret;
+		}
+	}
+
 	ret = psp_ring_stop(psp, PSP_RING_TYPE__KM);
 	if (ret) {
 		DRM_ERROR("PSP ring stop failed\n");

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

@@ -28,11 +28,13 @@
 #include "amdgpu.h"
 #include "psp_gfx_if.h"
 #include "ta_xgmi_if.h"
+#include "ta_ras_if.h"
 
 #define PSP_FENCE_BUFFER_SIZE	0x1000
 #define PSP_CMD_BUFFER_SIZE	0x1000
 #define PSP_ASD_SHARED_MEM_SIZE 0x4000
 #define PSP_XGMI_SHARED_MEM_SIZE 0x4000
+#define PSP_RAS_SHARED_MEM_SIZE 0x4000
 #define PSP_1_MEG		0x100000
 #define PSP_TMR_SIZE	0x400000
 
@@ -88,6 +90,9 @@ struct psp_funcs
 	int (*xgmi_set_topology_info)(struct psp_context *psp, int number_devices,
 				      struct psp_xgmi_topology_info *topology);
 	bool (*support_vmr_ring)(struct psp_context *psp);
+	int (*ras_trigger_error)(struct psp_context *psp,
+			struct ta_ras_trigger_error_input *info);
+	int (*ras_cure_posion)(struct psp_context *psp, uint64_t *mode_ptr);
 };
 
 struct psp_xgmi_context {
@@ -98,6 +103,16 @@ struct psp_xgmi_context {
 	void                            *xgmi_shared_buf;
 };
 
+struct psp_ras_context {
+	/*ras fw*/
+	bool			ras_initialized;
+	uint32_t		session_id;
+	struct amdgpu_bo	*ras_shared_bo;
+	uint64_t		ras_shared_mc_addr;
+	void			*ras_shared_buf;
+	struct amdgpu_ras	*ras;
+};
+
 struct psp_context
 {
 	struct amdgpu_device            *adev;
@@ -150,10 +165,15 @@ struct psp_context
 
 	/* xgmi ta firmware and buffer */
 	const struct firmware		*ta_fw;
+	uint32_t			ta_fw_version;
 	uint32_t			ta_xgmi_ucode_version;
 	uint32_t			ta_xgmi_ucode_size;
 	uint8_t				*ta_xgmi_start_addr;
+	uint32_t			ta_ras_ucode_version;
+	uint32_t			ta_ras_ucode_size;
+	uint8_t				*ta_ras_start_addr;
 	struct psp_xgmi_context		xgmi_context;
+	struct psp_ras_context		ras;
 };
 
 struct amdgpu_psp_funcs {
@@ -207,6 +227,13 @@ struct psp_xgmi_topology_info {
 
 #define amdgpu_psp_check_fw_loading_status(adev, i) (adev)->firmware.funcs->check_fw_loading_status((adev), (i))
 
+#define psp_ras_trigger_error(psp, info) \
+	((psp)->funcs->ras_trigger_error ? \
+	(psp)->funcs->ras_trigger_error((psp), (info)) : -EINVAL)
+#define psp_ras_cure_posion(psp, addr) \
+	((psp)->funcs->ras_cure_posion ? \
+	(psp)->funcs->ras_cure_posion(psp, (addr)) : -EINVAL)
+
 extern const struct amd_ip_funcs psp_ip_funcs;
 
 extern const struct amdgpu_ip_block_version psp_v3_1_ip_block;
@@ -217,6 +244,11 @@ extern const struct amdgpu_ip_block_version psp_v10_0_ip_block;
 
 int psp_gpu_reset(struct amdgpu_device *adev);
 int psp_xgmi_invoke(struct psp_context *psp, uint32_t ta_cmd_id);
+
+int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id);
+int psp_ras_enable_features(struct psp_context *psp,
+		union ta_ras_cmd_input *info, bool enable);
+
 extern const struct amdgpu_ip_block_version psp_v11_0_ip_block;
 
 #endif

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

@@ -0,0 +1,291 @@
+/*
+ * Copyright 2018 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 _AMDGPU_RAS_H
+#define _AMDGPU_RAS_H
+
+#include <linux/debugfs.h>
+#include <linux/list.h>
+#include "amdgpu.h"
+#include "amdgpu_psp.h"
+#include "ta_ras_if.h"
+
+enum amdgpu_ras_block {
+	AMDGPU_RAS_BLOCK__UMC = 0,
+	AMDGPU_RAS_BLOCK__SDMA,
+	AMDGPU_RAS_BLOCK__GFX,
+	AMDGPU_RAS_BLOCK__MMHUB,
+	AMDGPU_RAS_BLOCK__ATHUB,
+	AMDGPU_RAS_BLOCK__PCIE_BIF,
+	AMDGPU_RAS_BLOCK__HDP,
+	AMDGPU_RAS_BLOCK__XGMI_WAFL,
+	AMDGPU_RAS_BLOCK__DF,
+	AMDGPU_RAS_BLOCK__SMN,
+	AMDGPU_RAS_BLOCK__SEM,
+	AMDGPU_RAS_BLOCK__MP0,
+	AMDGPU_RAS_BLOCK__MP1,
+	AMDGPU_RAS_BLOCK__FUSE,
+
+	AMDGPU_RAS_BLOCK__LAST
+};
+
+#define AMDGPU_RAS_BLOCK_COUNT	AMDGPU_RAS_BLOCK__LAST
+#define AMDGPU_RAS_BLOCK_MASK	((1ULL << AMDGPU_RAS_BLOCK_COUNT) - 1)
+
+enum amdgpu_ras_error_type {
+	AMDGPU_RAS_ERROR__NONE							= 0,
+	AMDGPU_RAS_ERROR__PARITY						= 1,
+	AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE					= 2,
+	AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE					= 4,
+	AMDGPU_RAS_ERROR__POISON						= 8,
+};
+
+enum amdgpu_ras_ret {
+	AMDGPU_RAS_SUCCESS = 0,
+	AMDGPU_RAS_FAIL,
+	AMDGPU_RAS_UE,
+	AMDGPU_RAS_CE,
+	AMDGPU_RAS_PT,
+};
+
+struct ras_common_if {
+	enum amdgpu_ras_block block;
+	enum amdgpu_ras_error_type type;
+	uint32_t sub_block_index;
+	/* block name */
+	char name[32];
+};
+
+typedef int (*ras_ih_cb)(struct amdgpu_device *adev,
+		struct amdgpu_iv_entry *entry);
+
+struct amdgpu_ras {
+	/* ras infrastructure */
+	/* for ras itself. */
+	uint32_t hw_supported;
+	/* for IP to check its ras ability. */
+	uint32_t supported;
+	uint32_t features;
+	struct list_head head;
+	/* debugfs */
+	struct dentry *dir;
+	/* debugfs ctrl */
+	struct dentry *ent;
+	/* sysfs */
+	struct device_attribute features_attr;
+	/* block array */
+	struct ras_manager *objs;
+
+	/* gpu recovery */
+	struct work_struct recovery_work;
+	atomic_t in_recovery;
+	struct amdgpu_device *adev;
+	/* error handler data */
+	struct ras_err_handler_data *eh_data;
+	struct mutex recovery_lock;
+
+	uint32_t flags;
+};
+
+/* interfaces for IP */
+
+struct ras_fs_if {
+	struct ras_common_if head;
+	char sysfs_name[32];
+	char debugfs_name[32];
+};
+
+struct ras_query_if {
+	struct ras_common_if head;
+	unsigned long ue_count;
+	unsigned long ce_count;
+};
+
+struct ras_inject_if {
+	struct ras_common_if head;
+	uint64_t address;
+	uint64_t value;
+};
+
+struct ras_cure_if {
+	struct ras_common_if head;
+	uint64_t address;
+};
+
+struct ras_ih_if {
+	struct ras_common_if head;
+	ras_ih_cb cb;
+};
+
+struct ras_dispatch_if {
+	struct ras_common_if head;
+	struct amdgpu_iv_entry *entry;
+};
+
+struct ras_debug_if {
+	union {
+		struct ras_common_if head;
+		struct ras_inject_if inject;
+	};
+	int op;
+};
+/* work flow
+ * vbios
+ * 1: ras feature enable (enabled by default)
+ * psp
+ * 2: ras framework init (in ip_init)
+ * IP
+ * 3: IH add
+ * 4: debugfs/sysfs create
+ * 5: query/inject
+ * 6: debugfs/sysfs remove
+ * 7: IH remove
+ * 8: feature disable
+ */
+
+#define amdgpu_ras_get_context(adev)		((adev)->psp.ras.ras)
+#define amdgpu_ras_set_context(adev, ras_con)	((adev)->psp.ras.ras = (ras_con))
+
+/* check if ras is supported on block, say, sdma, gfx */
+static inline int amdgpu_ras_is_supported(struct amdgpu_device *adev,
+		unsigned int block)
+{
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+
+	return ras && (ras->supported & (1 << block));
+}
+
+int amdgpu_ras_query_error_count(struct amdgpu_device *adev,
+		bool is_ce);
+
+/* error handling functions */
+int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev,
+		unsigned long *bps, int pages);
+
+int amdgpu_ras_reserve_bad_pages(struct amdgpu_device *adev);
+
+static inline int amdgpu_ras_reset_gpu(struct amdgpu_device *adev,
+		bool is_baco)
+{
+	/* remove me when gpu reset works on vega20 A1. */
+#if 0
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+
+	if (atomic_cmpxchg(&ras->in_recovery, 0, 1) == 0)
+		schedule_work(&ras->recovery_work);
+#endif
+	return 0;
+}
+
+static inline enum ta_ras_block
+amdgpu_ras_block_to_ta(enum amdgpu_ras_block block) {
+	switch (block) {
+	case AMDGPU_RAS_BLOCK__UMC:
+		return TA_RAS_BLOCK__UMC;
+	case AMDGPU_RAS_BLOCK__SDMA:
+		return TA_RAS_BLOCK__SDMA;
+	case AMDGPU_RAS_BLOCK__GFX:
+		return TA_RAS_BLOCK__GFX;
+	case AMDGPU_RAS_BLOCK__MMHUB:
+		return TA_RAS_BLOCK__MMHUB;
+	case AMDGPU_RAS_BLOCK__ATHUB:
+		return TA_RAS_BLOCK__ATHUB;
+	case AMDGPU_RAS_BLOCK__PCIE_BIF:
+		return TA_RAS_BLOCK__PCIE_BIF;
+	case AMDGPU_RAS_BLOCK__HDP:
+		return TA_RAS_BLOCK__HDP;
+	case AMDGPU_RAS_BLOCK__XGMI_WAFL:
+		return TA_RAS_BLOCK__XGMI_WAFL;
+	case AMDGPU_RAS_BLOCK__DF:
+		return TA_RAS_BLOCK__DF;
+	case AMDGPU_RAS_BLOCK__SMN:
+		return TA_RAS_BLOCK__SMN;
+	case AMDGPU_RAS_BLOCK__SEM:
+		return TA_RAS_BLOCK__SEM;
+	case AMDGPU_RAS_BLOCK__MP0:
+		return TA_RAS_BLOCK__MP0;
+	case AMDGPU_RAS_BLOCK__MP1:
+		return TA_RAS_BLOCK__MP1;
+	case AMDGPU_RAS_BLOCK__FUSE:
+		return TA_RAS_BLOCK__FUSE;
+	default:
+		WARN_ONCE(1, "RAS ERROR: unexpected block id %d\n", block);
+		return TA_RAS_BLOCK__UMC;
+	}
+}
+
+static inline enum ta_ras_error_type
+amdgpu_ras_error_to_ta(enum amdgpu_ras_error_type error) {
+	switch (error) {
+	case AMDGPU_RAS_ERROR__NONE:
+		return TA_RAS_ERROR__NONE;
+	case AMDGPU_RAS_ERROR__PARITY:
+		return TA_RAS_ERROR__PARITY;
+	case AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE:
+		return TA_RAS_ERROR__SINGLE_CORRECTABLE;
+	case AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE:
+		return TA_RAS_ERROR__MULTI_UNCORRECTABLE;
+	case AMDGPU_RAS_ERROR__POISON:
+		return TA_RAS_ERROR__POISON;
+	default:
+		WARN_ONCE(1, "RAS ERROR: unexpected error type %d\n", error);
+		return TA_RAS_ERROR__NONE;
+	}
+}
+
+/* called in ip_init and ip_fini */
+int amdgpu_ras_init(struct amdgpu_device *adev);
+void amdgpu_ras_post_init(struct amdgpu_device *adev);
+int amdgpu_ras_fini(struct amdgpu_device *adev);
+int amdgpu_ras_pre_fini(struct amdgpu_device *adev);
+
+int amdgpu_ras_feature_enable(struct amdgpu_device *adev,
+		struct ras_common_if *head, bool enable);
+
+int amdgpu_ras_sysfs_create(struct amdgpu_device *adev,
+		struct ras_fs_if *head);
+
+int amdgpu_ras_sysfs_remove(struct amdgpu_device *adev,
+		struct ras_common_if *head);
+
+int amdgpu_ras_debugfs_create(struct amdgpu_device *adev,
+		struct ras_fs_if *head);
+
+int amdgpu_ras_debugfs_remove(struct amdgpu_device *adev,
+		struct ras_common_if *head);
+
+int amdgpu_ras_error_query(struct amdgpu_device *adev,
+		struct ras_query_if *info);
+
+int amdgpu_ras_error_inject(struct amdgpu_device *adev,
+		struct ras_inject_if *info);
+
+int amdgpu_ras_interrupt_add_handler(struct amdgpu_device *adev,
+		struct ras_ih_if *info);
+
+int amdgpu_ras_interrupt_remove_handler(struct amdgpu_device *adev,
+		struct ras_ih_if *info);
+
+int amdgpu_ras_interrupt_dispatch(struct amdgpu_device *adev,
+		struct ras_dispatch_if *info);
+#endif

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

@@ -248,6 +248,8 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 	 */
 	if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ)
 		sched_hw_submission = max(sched_hw_submission, 256);
+	else if (ring == &adev->sdma.instance[0].page)
+		sched_hw_submission = 256;
 
 	if (ring->adev == NULL) {
 		if (adev->num_rings >= AMDGPU_MAX_RINGS)

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

@@ -30,6 +30,8 @@
 enum amdgpu_sdma_irq {
 	AMDGPU_SDMA_IRQ_TRAP0 = 0,
 	AMDGPU_SDMA_IRQ_TRAP1,
+	AMDGPU_SDMA_IRQ_ECC0,
+	AMDGPU_SDMA_IRQ_ECC1,
 
 	AMDGPU_SDMA_IRQ_LAST
 };
@@ -49,9 +51,11 @@ struct amdgpu_sdma {
 	struct amdgpu_sdma_instance instance[AMDGPU_MAX_SDMA_INSTANCES];
 	struct amdgpu_irq_src	trap_irq;
 	struct amdgpu_irq_src	illegal_inst_irq;
+	struct amdgpu_irq_src	ecc_irq;
 	int			num_instances;
 	uint32_t                    srbm_soft_reset;
 	bool			has_page_queue;
+	struct ras_common_if	*ras_if;
 };
 
 /*

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

@@ -50,8 +50,6 @@
 #include "amdgpu_sdma.h"
 #include "bif/bif_4_1_d.h"
 
-#define DRM_FILE_PAGE_OFFSET (0x100000000ULL >> PAGE_SHIFT)
-
 static int amdgpu_map_buffer(struct ttm_buffer_object *bo,
 			     struct ttm_mem_reg *mem, unsigned num_pages,
 			     uint64_t offset, unsigned window,
@@ -1424,6 +1422,13 @@ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo,
 	struct dma_fence *f;
 	int i;
 
+	/* Don't evict VM page tables while they are busy, otherwise we can't
+	 * cleanly handle page faults.
+	 */
+	if (bo->type == ttm_bo_type_kernel &&
+	    !reservation_object_test_signaled_rcu(bo->resv, true))
+		return false;
+
 	/* If bo is a KFD BO, check if the bo belongs to the current process.
 	 * If true, then return false as any KFD process needs all its BOs to
 	 * be resident to run successfully
@@ -1671,7 +1676,6 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
 	r = ttm_bo_device_init(&adev->mman.bdev,
 			       &amdgpu_bo_driver,
 			       adev->ddev->anon_inode->i_mapping,
-			       DRM_FILE_PAGE_OFFSET,
 			       adev->need_dma32);
 	if (r) {
 		DRM_ERROR("failed initializing buffer object driver(%d).\n", r);
@@ -1877,14 +1881,9 @@ void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable)
 
 int amdgpu_mmap(struct file *filp, struct vm_area_struct *vma)
 {
-	struct drm_file *file_priv;
-	struct amdgpu_device *adev;
+	struct drm_file *file_priv = filp->private_data;
+	struct amdgpu_device *adev = file_priv->minor->dev->dev_private;
 
-	if (unlikely(vma->vm_pgoff < DRM_FILE_PAGE_OFFSET))
-		return -EINVAL;
-
-	file_priv = filp->private_data;
-	adev = file_priv->minor->dev->dev_private;
 	if (adev == NULL)
 		return -EINVAL;
 

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

@@ -30,7 +30,6 @@
 #include <drm/gpu_scheduler.h>
 #include <drm/drm_file.h>
 #include <drm/ttm/ttm_bo_driver.h>
-#include <linux/chash.h>
 
 #include "amdgpu_sync.h"
 #include "amdgpu_ring.h"
@@ -140,7 +139,6 @@ struct amdgpu_vm_bo_base {
 
 struct amdgpu_vm_pt {
 	struct amdgpu_vm_bo_base	base;
-	bool				huge;
 
 	/* array of page tables, one for each directory entry */
 	struct amdgpu_vm_pt		*entries;
@@ -167,11 +165,6 @@ struct amdgpu_vm_pte_funcs {
 			    uint32_t incr, uint64_t flags);
 };
 
-#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_task_info {
 	char	process_name[TASK_COMM_LEN];
 	char	task_name[TASK_COMM_LEN];
@@ -179,11 +172,52 @@ struct amdgpu_task_info {
 	pid_t	tgid;
 };
 
-#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;
+/**
+ * struct amdgpu_vm_update_params
+ *
+ * Encapsulate some VM table update parameters to reduce
+ * the number of function parameters
+ *
+ */
+struct amdgpu_vm_update_params {
+
+	/**
+	 * @adev: amdgpu device we do this update for
+	 */
+	struct amdgpu_device *adev;
+
+	/**
+	 * @vm: optional amdgpu_vm we do this update for
+	 */
+	struct amdgpu_vm *vm;
+
+	/**
+	 * @pages_addr:
+	 *
+	 * DMA addresses to use for mapping
+	 */
+	dma_addr_t *pages_addr;
+
+	/**
+	 * @job: job to used for hw submission
+	 */
+	struct amdgpu_job *job;
+
+	/**
+	 * @num_dw_left: number of dw left for the IB
+	 */
+	unsigned int num_dw_left;
+};
+
+struct amdgpu_vm_update_funcs {
+	int (*map_table)(struct amdgpu_bo *bo);
+	int (*prepare)(struct amdgpu_vm_update_params *p, void * owner,
+		       struct dma_fence *exclusive);
+	int (*update)(struct amdgpu_vm_update_params *p,
+		      struct amdgpu_bo *bo, uint64_t pe, uint64_t addr,
+		      unsigned count, uint32_t incr, uint64_t flags);
+	int (*commit)(struct amdgpu_vm_update_params *p,
+		      struct dma_fence **fence);
 };
 
 struct amdgpu_vm {
@@ -221,7 +255,10 @@ struct amdgpu_vm {
 	struct amdgpu_vmid	*reserved_vmid[AMDGPU_MAX_VMHUBS];
 
 	/* Flag to indicate if VM tables are updated by CPU or GPU (SDMA) */
-	bool                    use_cpu_for_update;
+	bool					use_cpu_for_update;
+
+	/* Functions to use for VM table updates */
+	const struct amdgpu_vm_update_funcs	*update_funcs;
 
 	/* Flag to indicate ATS support from PTE for GFX9 */
 	bool			pte_support_ats;
@@ -245,7 +282,6 @@ struct amdgpu_vm {
 	struct ttm_lru_bulk_move lru_bulk_move;
 	/* mark whether can do the bulk move */
 	bool			bulk_moveable;
-	struct amdgpu_retryfault_hashtable *fault_hash;
 };
 
 struct amdgpu_vm_manager {
@@ -283,14 +319,23 @@ struct amdgpu_vm_manager {
 	 */
 	struct idr				pasid_idr;
 	spinlock_t				pasid_lock;
+
+	/* counter of mapped memory through xgmi */
+	uint32_t				xgmi_map_counter;
+	struct mutex				lock_pstate;
 };
 
 #define amdgpu_vm_copy_pte(adev, ib, pe, src, count) ((adev)->vm_manager.vm_pte_funcs->copy_pte((ib), (pe), (src), (count)))
 #define amdgpu_vm_write_pte(adev, ib, pe, value, count, incr) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pe), (value), (count), (incr)))
 #define amdgpu_vm_set_pte_pde(adev, ib, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->set_pte_pde((ib), (pe), (addr), (count), (incr), (flags)))
 
+extern const struct amdgpu_vm_update_funcs amdgpu_vm_cpu_funcs;
+extern const struct amdgpu_vm_update_funcs amdgpu_vm_sdma_funcs;
+
 void amdgpu_vm_manager_init(struct amdgpu_device *adev);
 void amdgpu_vm_manager_fini(struct amdgpu_device *adev);
+
+long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout);
 int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		   int vm_context, unsigned int pasid);
 int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm, unsigned int pasid);
@@ -303,9 +348,6 @@ 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);
-int amdgpu_vm_alloc_pts(struct amdgpu_device *adev,
-			struct amdgpu_vm *vm,
-			uint64_t saddr, uint64_t size);
 int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_pipe_sync);
 int amdgpu_vm_update_directories(struct amdgpu_device *adev,
 				 struct amdgpu_vm *vm);
@@ -319,6 +361,7 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev,
 			bool clear);
 void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
 			     struct amdgpu_bo *bo, bool evicted);
+uint64_t amdgpu_vm_map_gart(const dma_addr_t *pages_addr, uint64_t addr);
 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,
@@ -358,11 +401,6 @@ void amdgpu_vm_set_task_info(struct amdgpu_vm *vm);
 
 void amdgpu_vm_move_to_lru_tail(struct amdgpu_device *adev,
 				struct amdgpu_vm *vm);
-
-int amdgpu_vm_add_fault(struct amdgpu_retryfault_hashtable *fault_hash, u64 key);
-
-void amdgpu_vm_clear_fault(struct amdgpu_retryfault_hashtable *fault_hash, u64 key);
-
 void amdgpu_vm_del_from_lru_notify(struct ttm_buffer_object *bo);
 
 #endif

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

@@ -0,0 +1,127 @@
+/*
+ * Copyright 2019 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "amdgpu_vm.h"
+#include "amdgpu_object.h"
+#include "amdgpu_trace.h"
+
+/**
+ * amdgpu_vm_cpu_map_table - make sure new PDs/PTs are kmapped
+ *
+ * @table: newly allocated or validated PD/PT
+ */
+static int amdgpu_vm_cpu_map_table(struct amdgpu_bo *table)
+{
+	return amdgpu_bo_kmap(table, NULL);
+}
+
+/**
+ * amdgpu_vm_cpu_prepare - prepare page table update with the CPU
+ *
+ * @p: see amdgpu_vm_update_params definition
+ * @owner: owner we need to sync to
+ * @exclusive: exclusive move fence we need to sync to
+ *
+ * Returns:
+ * Negativ errno, 0 for success.
+ */
+static int amdgpu_vm_cpu_prepare(struct amdgpu_vm_update_params *p, void *owner,
+				 struct dma_fence *exclusive)
+{
+	int r;
+
+	/* Wait for PT BOs to be idle. PTs share the same resv. object
+	 * as the root PD BO
+	 */
+	r = amdgpu_bo_sync_wait(p->vm->root.base.bo, owner, true);
+	if (unlikely(r))
+		return r;
+
+	/* Wait for any BO move to be completed */
+	if (exclusive) {
+		r = dma_fence_wait(exclusive, true);
+		if (unlikely(r))
+			return r;
+	}
+
+	return 0;
+}
+
+/**
+ * amdgpu_vm_cpu_update - helper to update page tables via CPU
+ *
+ * @p: see amdgpu_vm_update_params definition
+ * @bo: PD/PT to update
+ * @pe: kmap addr of the page entry
+ * @addr: dst addr to write into pe
+ * @count: number of page entries to update
+ * @incr: increase next addr by incr bytes
+ * @flags: hw access flags
+ *
+ * Write count number of PT/PD entries directly.
+ */
+static int amdgpu_vm_cpu_update(struct amdgpu_vm_update_params *p,
+				struct amdgpu_bo *bo, uint64_t pe,
+				uint64_t addr, unsigned count, uint32_t incr,
+				uint64_t flags)
+{
+	unsigned int i;
+	uint64_t value;
+
+	pe += (unsigned long)amdgpu_bo_kptr(bo);
+
+	trace_amdgpu_vm_set_ptes(pe, addr, count, incr, flags);
+
+	for (i = 0; i < count; i++) {
+		value = p->pages_addr ?
+			amdgpu_vm_map_gart(p->pages_addr, addr) :
+			addr;
+		amdgpu_gmc_set_pte_pde(p->adev, (void *)(uintptr_t)pe,
+				       i, value, flags);
+		addr += incr;
+	}
+	return 0;
+}
+
+/**
+ * amdgpu_vm_cpu_commit - commit page table update to the HW
+ *
+ * @p: see amdgpu_vm_update_params definition
+ * @fence: unused
+ *
+ * Make sure that the hardware sees the page table updates.
+ */
+static int amdgpu_vm_cpu_commit(struct amdgpu_vm_update_params *p,
+				struct dma_fence **fence)
+{
+	/* Flush HDP */
+	mb();
+	amdgpu_asic_flush_hdp(p->adev, NULL);
+	return 0;
+}
+
+const struct amdgpu_vm_update_funcs amdgpu_vm_cpu_funcs = {
+	.map_table = amdgpu_vm_cpu_map_table,
+	.prepare = amdgpu_vm_cpu_prepare,
+	.update = amdgpu_vm_cpu_update,
+	.commit = amdgpu_vm_cpu_commit
+};

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

@@ -0,0 +1,270 @@
+/*
+ * Copyright 2019 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "amdgpu_vm.h"
+#include "amdgpu_job.h"
+#include "amdgpu_object.h"
+#include "amdgpu_trace.h"
+
+#define AMDGPU_VM_SDMA_MIN_NUM_DW	256u
+#define AMDGPU_VM_SDMA_MAX_NUM_DW	(16u * 1024u)
+
+/**
+ * amdgpu_vm_sdma_map_table - make sure new PDs/PTs are GTT mapped
+ *
+ * @table: newly allocated or validated PD/PT
+ */
+static int amdgpu_vm_sdma_map_table(struct amdgpu_bo *table)
+{
+	int r;
+
+	r = amdgpu_ttm_alloc_gart(&table->tbo);
+	if (r)
+		return r;
+
+	if (table->shadow)
+		r = amdgpu_ttm_alloc_gart(&table->shadow->tbo);
+
+	return r;
+}
+
+/**
+ * amdgpu_vm_sdma_prepare - prepare SDMA command submission
+ *
+ * @p: see amdgpu_vm_update_params definition
+ * @owner: owner we need to sync to
+ * @exclusive: exclusive move fence we need to sync to
+ *
+ * Returns:
+ * Negativ errno, 0 for success.
+ */
+static int amdgpu_vm_sdma_prepare(struct amdgpu_vm_update_params *p,
+				  void *owner, struct dma_fence *exclusive)
+{
+	struct amdgpu_bo *root = p->vm->root.base.bo;
+	unsigned int ndw = AMDGPU_VM_SDMA_MIN_NUM_DW;
+	int r;
+
+	r = amdgpu_job_alloc_with_ib(p->adev, ndw * 4, &p->job);
+	if (r)
+		return r;
+
+	r = amdgpu_sync_fence(p->adev, &p->job->sync, exclusive, false);
+	if (r)
+		return r;
+
+	r = amdgpu_sync_resv(p->adev, &p->job->sync, root->tbo.resv,
+			     owner, false);
+	if (r)
+		return r;
+
+	p->num_dw_left = ndw;
+	return 0;
+}
+
+/**
+ * amdgpu_vm_sdma_commit - commit SDMA command submission
+ *
+ * @p: see amdgpu_vm_update_params definition
+ * @fence: resulting fence
+ *
+ * Returns:
+ * Negativ errno, 0 for success.
+ */
+static int amdgpu_vm_sdma_commit(struct amdgpu_vm_update_params *p,
+				 struct dma_fence **fence)
+{
+	struct amdgpu_bo *root = p->vm->root.base.bo;
+	struct amdgpu_ib *ib = p->job->ibs;
+	struct amdgpu_ring *ring;
+	struct dma_fence *f;
+	int r;
+
+	ring = container_of(p->vm->entity.rq->sched, struct amdgpu_ring, sched);
+
+	WARN_ON(ib->length_dw == 0);
+	amdgpu_ring_pad_ib(ring, ib);
+	WARN_ON(ib->length_dw > p->num_dw_left);
+	r = amdgpu_job_submit(p->job, &p->vm->entity,
+			      AMDGPU_FENCE_OWNER_VM, &f);
+	if (r)
+		goto error;
+
+	amdgpu_bo_fence(root, f, true);
+	if (fence)
+		swap(*fence, f);
+	dma_fence_put(f);
+	return 0;
+
+error:
+	amdgpu_job_free(p->job);
+	return r;
+}
+
+
+/**
+ * amdgpu_vm_sdma_copy_ptes - copy the PTEs from mapping
+ *
+ * @p: see amdgpu_vm_update_params definition
+ * @bo: PD/PT to update
+ * @pe: addr of the page entry
+ * @count: number of page entries to copy
+ *
+ * Traces the parameters and calls the DMA function to copy the PTEs.
+ */
+static void amdgpu_vm_sdma_copy_ptes(struct amdgpu_vm_update_params *p,
+				     struct amdgpu_bo *bo, uint64_t pe,
+				     unsigned count)
+{
+	struct amdgpu_ib *ib = p->job->ibs;
+	uint64_t src = ib->gpu_addr;
+
+	src += p->num_dw_left * 4;
+
+	pe += amdgpu_bo_gpu_offset(bo);
+	trace_amdgpu_vm_copy_ptes(pe, src, count);
+
+	amdgpu_vm_copy_pte(p->adev, ib, pe, src, count);
+}
+
+/**
+ * amdgpu_vm_sdma_set_ptes - helper to call the right asic function
+ *
+ * @p: see amdgpu_vm_update_params definition
+ * @bo: PD/PT to update
+ * @pe: addr of the page entry
+ * @addr: dst addr to write into pe
+ * @count: number of page entries to update
+ * @incr: increase next addr by incr bytes
+ * @flags: hw access flags
+ *
+ * Traces the parameters and calls the right asic functions
+ * to setup the page table using the DMA.
+ */
+static void amdgpu_vm_sdma_set_ptes(struct amdgpu_vm_update_params *p,
+				    struct amdgpu_bo *bo, uint64_t pe,
+				    uint64_t addr, unsigned count,
+				    uint32_t incr, uint64_t flags)
+{
+	struct amdgpu_ib *ib = p->job->ibs;
+
+	pe += amdgpu_bo_gpu_offset(bo);
+	trace_amdgpu_vm_set_ptes(pe, addr, count, incr, flags);
+	if (count < 3) {
+		amdgpu_vm_write_pte(p->adev, ib, pe, addr | flags,
+				    count, incr);
+	} else {
+		amdgpu_vm_set_pte_pde(p->adev, ib, pe, addr,
+				      count, incr, flags);
+	}
+}
+
+/**
+ * amdgpu_vm_sdma_update - execute VM update
+ *
+ * @p: see amdgpu_vm_update_params definition
+ * @bo: PD/PT to update
+ * @pe: addr of the page entry
+ * @addr: dst addr to write into pe
+ * @count: number of page entries to update
+ * @incr: increase next addr by incr bytes
+ * @flags: hw access flags
+ *
+ * Reserve space in the IB, setup mapping buffer on demand and write commands to
+ * the IB.
+ */
+static int amdgpu_vm_sdma_update(struct amdgpu_vm_update_params *p,
+				 struct amdgpu_bo *bo, uint64_t pe,
+				 uint64_t addr, unsigned count, uint32_t incr,
+				 uint64_t flags)
+{
+	unsigned int i, ndw, nptes;
+	uint64_t *pte;
+	int r;
+
+	do {
+		ndw = p->num_dw_left;
+		ndw -= p->job->ibs->length_dw;
+
+		if (ndw < 32) {
+			r = amdgpu_vm_sdma_commit(p, NULL);
+			if (r)
+				return r;
+
+			/* estimate how many dw we need */
+			ndw = 32;
+			if (p->pages_addr)
+				ndw += count * 2;
+			ndw = max(ndw, AMDGPU_VM_SDMA_MIN_NUM_DW);
+			ndw = min(ndw, AMDGPU_VM_SDMA_MAX_NUM_DW);
+
+			r = amdgpu_job_alloc_with_ib(p->adev, ndw * 4, &p->job);
+			if (r)
+				return r;
+
+			p->num_dw_left = ndw;
+		}
+
+		if (!p->pages_addr) {
+			/* set page commands needed */
+			if (bo->shadow)
+				amdgpu_vm_sdma_set_ptes(p, bo->shadow, pe, addr,
+							count, incr, flags);
+			amdgpu_vm_sdma_set_ptes(p, bo, pe, addr, count,
+						incr, flags);
+			return 0;
+		}
+
+		/* copy commands needed */
+		ndw -= p->adev->vm_manager.vm_pte_funcs->copy_pte_num_dw *
+			(bo->shadow ? 2 : 1);
+
+		/* for padding */
+		ndw -= 7;
+
+		nptes = min(count, ndw / 2);
+
+		/* Put the PTEs at the end of the IB. */
+		p->num_dw_left -= nptes * 2;
+		pte = (uint64_t *)&(p->job->ibs->ptr[p->num_dw_left]);
+		for (i = 0; i < nptes; ++i, addr += incr) {
+			pte[i] = amdgpu_vm_map_gart(p->pages_addr, addr);
+			pte[i] |= flags;
+		}
+
+		if (bo->shadow)
+			amdgpu_vm_sdma_copy_ptes(p, bo->shadow, pe, nptes);
+		amdgpu_vm_sdma_copy_ptes(p, bo, pe, nptes);
+
+		pe += nptes * 8;
+		count -= nptes;
+	} while (count);
+
+	return 0;
+}
+
+const struct amdgpu_vm_update_funcs amdgpu_vm_sdma_funcs = {
+	.map_table = amdgpu_vm_sdma_map_table,
+	.prepare = amdgpu_vm_sdma_prepare,
+	.update = amdgpu_vm_sdma_update,
+	.commit = amdgpu_vm_sdma_commit
+};

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

@@ -32,6 +32,85 @@ struct amdgpu_vram_mgr {
 	atomic64_t vis_usage;
 };
 
+/**
+ * DOC: mem_info_vram_total
+ *
+ * The amdgpu driver provides a sysfs API for reporting current total VRAM
+ * available on the device
+ * The file mem_info_vram_total is used for this and returns the total
+ * amount of VRAM in bytes
+ */
+static ssize_t amdgpu_mem_info_vram_total_show(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = ddev->dev_private;
+
+	return snprintf(buf, PAGE_SIZE, "%llu\n", adev->gmc.real_vram_size);
+}
+
+/**
+ * DOC: mem_info_vis_vram_total
+ *
+ * The amdgpu driver provides a sysfs API for reporting current total
+ * visible VRAM available on the device
+ * The file mem_info_vis_vram_total is used for this and returns the total
+ * amount of visible VRAM in bytes
+ */
+static ssize_t amdgpu_mem_info_vis_vram_total_show(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = ddev->dev_private;
+
+	return snprintf(buf, PAGE_SIZE, "%llu\n", adev->gmc.visible_vram_size);
+}
+
+/**
+ * DOC: mem_info_vram_used
+ *
+ * The amdgpu driver provides a sysfs API for reporting current total VRAM
+ * available on the device
+ * The file mem_info_vram_used is used for this and returns the total
+ * amount of currently used VRAM in bytes
+ */
+static ssize_t amdgpu_mem_info_vram_used_show(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = ddev->dev_private;
+
+	return snprintf(buf, PAGE_SIZE, "%llu\n",
+		amdgpu_vram_mgr_usage(&adev->mman.bdev.man[TTM_PL_VRAM]));
+}
+
+/**
+ * DOC: mem_info_vis_vram_used
+ *
+ * The amdgpu driver provides a sysfs API for reporting current total of
+ * used visible VRAM
+ * The file mem_info_vis_vram_used is used for this and returns the total
+ * amount of currently used visible VRAM in bytes
+ */
+static ssize_t amdgpu_mem_info_vis_vram_used_show(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = ddev->dev_private;
+
+	return snprintf(buf, PAGE_SIZE, "%llu\n",
+		amdgpu_vram_mgr_vis_usage(&adev->mman.bdev.man[TTM_PL_VRAM]));
+}
+
+static DEVICE_ATTR(mem_info_vram_total, S_IRUGO,
+		   amdgpu_mem_info_vram_total_show, NULL);
+static DEVICE_ATTR(mem_info_vis_vram_total, S_IRUGO,
+		   amdgpu_mem_info_vis_vram_total_show,NULL);
+static DEVICE_ATTR(mem_info_vram_used, S_IRUGO,
+		   amdgpu_mem_info_vram_used_show, NULL);
+static DEVICE_ATTR(mem_info_vis_vram_used, S_IRUGO,
+		   amdgpu_mem_info_vis_vram_used_show, NULL);
+
 /**
  * amdgpu_vram_mgr_init - init VRAM manager and DRM MM
  *
@@ -43,7 +122,9 @@ struct amdgpu_vram_mgr {
 static int amdgpu_vram_mgr_init(struct ttm_mem_type_manager *man,
 				unsigned long p_size)
 {
+	struct amdgpu_device *adev = amdgpu_ttm_adev(man->bdev);
 	struct amdgpu_vram_mgr *mgr;
+	int ret;
 
 	mgr = kzalloc(sizeof(*mgr), GFP_KERNEL);
 	if (!mgr)
@@ -52,6 +133,29 @@ static int amdgpu_vram_mgr_init(struct ttm_mem_type_manager *man,
 	drm_mm_init(&mgr->mm, 0, p_size);
 	spin_lock_init(&mgr->lock);
 	man->priv = mgr;
+
+	/* Add the two VRAM-related sysfs files */
+	ret = device_create_file(adev->dev, &dev_attr_mem_info_vram_total);
+	if (ret) {
+		DRM_ERROR("Failed to create device file mem_info_vram_total\n");
+		return ret;
+	}
+	ret = device_create_file(adev->dev, &dev_attr_mem_info_vis_vram_total);
+	if (ret) {
+		DRM_ERROR("Failed to create device file mem_info_vis_vram_total\n");
+		return ret;
+	}
+	ret = device_create_file(adev->dev, &dev_attr_mem_info_vram_used);
+	if (ret) {
+		DRM_ERROR("Failed to create device file mem_info_vram_used\n");
+		return ret;
+	}
+	ret = device_create_file(adev->dev, &dev_attr_mem_info_vis_vram_used);
+	if (ret) {
+		DRM_ERROR("Failed to create device file mem_info_vis_vram_used\n");
+		return ret;
+	}
+
 	return 0;
 }
 
@@ -65,6 +169,7 @@ static int amdgpu_vram_mgr_init(struct ttm_mem_type_manager *man,
  */
 static int amdgpu_vram_mgr_fini(struct ttm_mem_type_manager *man)
 {
+	struct amdgpu_device *adev = amdgpu_ttm_adev(man->bdev);
 	struct amdgpu_vram_mgr *mgr = man->priv;
 
 	spin_lock(&mgr->lock);
@@ -72,6 +177,10 @@ static int amdgpu_vram_mgr_fini(struct ttm_mem_type_manager *man)
 	spin_unlock(&mgr->lock);
 	kfree(mgr);
 	man->priv = NULL;
+	device_remove_file(adev->dev, &dev_attr_mem_info_vram_total);
+	device_remove_file(adev->dev, &dev_attr_mem_info_vis_vram_total);
+	device_remove_file(adev->dev, &dev_attr_mem_info_vram_used);
+	device_remove_file(adev->dev, &dev_attr_mem_info_vis_vram_used);
 	return 0;
 }
 

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

@@ -34,12 +34,132 @@ static DEFINE_MUTEX(xgmi_mutex);
 static struct amdgpu_hive_info xgmi_hives[AMDGPU_MAX_XGMI_HIVE];
 static unsigned hive_count = 0;
 
-
 void *amdgpu_xgmi_hive_try_lock(struct amdgpu_hive_info *hive)
 {
 	return &hive->device_list;
 }
 
+static ssize_t amdgpu_xgmi_show_hive_id(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct amdgpu_hive_info *hive =
+			container_of(attr, struct amdgpu_hive_info, dev_attr);
+
+	return snprintf(buf, PAGE_SIZE, "%llu\n", hive->hive_id);
+}
+
+static int amdgpu_xgmi_sysfs_create(struct amdgpu_device *adev,
+				    struct amdgpu_hive_info *hive)
+{
+	int ret = 0;
+
+	if (WARN_ON(hive->kobj))
+		return -EINVAL;
+
+	hive->kobj = kobject_create_and_add("xgmi_hive_info", &adev->dev->kobj);
+	if (!hive->kobj) {
+		dev_err(adev->dev, "XGMI: Failed to allocate sysfs entry!\n");
+		return -EINVAL;
+	}
+
+	hive->dev_attr = (struct device_attribute) {
+		.attr = {
+			.name = "xgmi_hive_id",
+			.mode = S_IRUGO,
+
+		},
+		.show = amdgpu_xgmi_show_hive_id,
+	};
+
+	ret = sysfs_create_file(hive->kobj, &hive->dev_attr.attr);
+	if (ret) {
+		dev_err(adev->dev, "XGMI: Failed to create device file xgmi_hive_id\n");
+		kobject_del(hive->kobj);
+		kobject_put(hive->kobj);
+		hive->kobj = NULL;
+	}
+
+	return ret;
+}
+
+static void amdgpu_xgmi_sysfs_destroy(struct amdgpu_device *adev,
+				    struct amdgpu_hive_info *hive)
+{
+	sysfs_remove_file(hive->kobj, &hive->dev_attr.attr);
+	kobject_del(hive->kobj);
+	kobject_put(hive->kobj);
+	hive->kobj = NULL;
+}
+
+static ssize_t amdgpu_xgmi_show_device_id(struct device *dev,
+				     struct device_attribute *attr,
+				     char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = ddev->dev_private;
+
+	return snprintf(buf, PAGE_SIZE, "%llu\n", adev->gmc.xgmi.node_id);
+
+}
+
+
+static DEVICE_ATTR(xgmi_device_id, S_IRUGO, amdgpu_xgmi_show_device_id, NULL);
+
+
+static int amdgpu_xgmi_sysfs_add_dev_info(struct amdgpu_device *adev,
+					 struct amdgpu_hive_info *hive)
+{
+	int ret = 0;
+	char node[10] = { 0 };
+
+	/* Create xgmi device id file */
+	ret = device_create_file(adev->dev, &dev_attr_xgmi_device_id);
+	if (ret) {
+		dev_err(adev->dev, "XGMI: Failed to create device file xgmi_device_id\n");
+		return ret;
+	}
+
+	/* Create sysfs link to hive info folder on the first device */
+	if (adev != hive->adev) {
+		ret = sysfs_create_link(&adev->dev->kobj, hive->kobj,
+					"xgmi_hive_info");
+		if (ret) {
+			dev_err(adev->dev, "XGMI: Failed to create link to hive info");
+			goto remove_file;
+		}
+	}
+
+	sprintf(node, "node%d", hive->number_devices);
+	/* Create sysfs link form the hive folder to yourself */
+	ret = sysfs_create_link(hive->kobj, &adev->dev->kobj, node);
+	if (ret) {
+		dev_err(adev->dev, "XGMI: Failed to create link from hive info");
+		goto remove_link;
+	}
+
+	goto success;
+
+
+remove_link:
+	sysfs_remove_link(&adev->dev->kobj, adev->ddev->unique);
+
+remove_file:
+	device_remove_file(adev->dev, &dev_attr_xgmi_device_id);
+
+success:
+	return ret;
+}
+
+static void amdgpu_xgmi_sysfs_rem_dev_info(struct amdgpu_device *adev,
+					  struct amdgpu_hive_info *hive)
+{
+	device_remove_file(adev->dev, &dev_attr_xgmi_device_id);
+	sysfs_remove_link(&adev->dev->kobj, adev->ddev->unique);
+	sysfs_remove_link(hive->kobj, adev->ddev->unique);
+}
+
+
+
 struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev, int lock)
 {
 	int i;
@@ -66,18 +186,40 @@ struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev, int lo
 
 	/* initialize new hive if not exist */
 	tmp = &xgmi_hives[hive_count++];
+
+	if (amdgpu_xgmi_sysfs_create(adev, tmp)) {
+		mutex_unlock(&xgmi_mutex);
+		return NULL;
+	}
+
+	tmp->adev = adev;
 	tmp->hive_id = adev->gmc.xgmi.hive_id;
 	INIT_LIST_HEAD(&tmp->device_list);
 	mutex_init(&tmp->hive_lock);
 	mutex_init(&tmp->reset_lock);
+
 	if (lock)
 		mutex_lock(&tmp->hive_lock);
-
+	tmp->pstate = -1;
 	mutex_unlock(&xgmi_mutex);
 
 	return tmp;
 }
 
+int amdgpu_xgmi_set_pstate(struct amdgpu_device *adev, int pstate)
+{
+	int ret = 0;
+	struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev, 0);
+
+	if (!hive)
+		return 0;
+
+	if (hive->pstate == pstate)
+		return 0;
+	/* Todo : sent the message to SMU for pstate change */
+	return ret;
+}
+
 int amdgpu_xgmi_update_topology(struct amdgpu_hive_info *hive, struct amdgpu_device *adev)
 {
 	int ret = -EINVAL;
@@ -156,8 +298,17 @@ int amdgpu_xgmi_add_device(struct amdgpu_device *adev)
 			break;
 	}
 
-	dev_info(adev->dev, "XGMI: Add node %d, hive 0x%llx.\n",
-		 adev->gmc.xgmi.physical_node_id, adev->gmc.xgmi.hive_id);
+	if (!ret)
+		ret = amdgpu_xgmi_sysfs_add_dev_info(adev, hive);
+
+	if (!ret)
+		dev_info(adev->dev, "XGMI: Add node %d, hive 0x%llx.\n",
+			 adev->gmc.xgmi.physical_node_id, adev->gmc.xgmi.hive_id);
+	else
+		dev_err(adev->dev, "XGMI: Failed to add node %d, hive 0x%llx ret: %d\n",
+			adev->gmc.xgmi.physical_node_id, adev->gmc.xgmi.hive_id,
+			ret);
+
 
 	mutex_unlock(&hive->hive_lock);
 exit:
@@ -176,9 +327,11 @@ void amdgpu_xgmi_remove_device(struct amdgpu_device *adev)
 		return;
 
 	if (!(hive->number_devices--)) {
+		amdgpu_xgmi_sysfs_destroy(adev, hive);
 		mutex_destroy(&hive->hive_lock);
 		mutex_destroy(&hive->reset_lock);
 	} else {
+		amdgpu_xgmi_sysfs_rem_dev_info(adev, hive);
 		mutex_unlock(&hive->hive_lock);
 	}
 }

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

@@ -29,13 +29,25 @@ struct amdgpu_hive_info {
 	struct list_head	device_list;
 	struct psp_xgmi_topology_info	topology_info;
 	int number_devices;
-	struct mutex hive_lock,
-		     reset_lock;
+	struct mutex hive_lock, reset_lock;
+	struct kobject *kobj;
+	struct device_attribute dev_attr;
+	struct amdgpu_device *adev;
+	int pstate; /*0 -- low , 1 -- high , -1 unknown*/
 };
 
 struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev, int lock);
 int amdgpu_xgmi_update_topology(struct amdgpu_hive_info *hive, struct amdgpu_device *adev);
 int amdgpu_xgmi_add_device(struct amdgpu_device *adev);
 void amdgpu_xgmi_remove_device(struct amdgpu_device *adev);
+int amdgpu_xgmi_set_pstate(struct amdgpu_device *adev, int pstate);
+
+static inline bool amdgpu_xgmi_same_hive(struct amdgpu_device *adev,
+		struct amdgpu_device *bo_adev)
+{
+	return (adev != bo_adev &&
+		adev->gmc.xgmi.hive_id &&
+		adev->gmc.xgmi.hive_id == bo_adev->gmc.xgmi.hive_id);
+}
 
 #endif

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

@@ -782,6 +782,25 @@ static void gfx_v6_0_tiling_mode_table_init(struct amdgpu_device *adev)
 				BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
 				BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
 				MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2);
+		tilemode[18] =  MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
+				ARRAY_MODE(ARRAY_1D_TILED_THICK) |
+				PIPE_CONFIG(ADDR_SURF_P4_8x16);
+		tilemode[19] =  MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
+				ARRAY_MODE(ARRAY_2D_TILED_XTHICK) |
+				PIPE_CONFIG(ADDR_SURF_P4_8x16) |
+				BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+				BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+				MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+				NUM_BANKS(ADDR_SURF_16_BANK) |
+				TILE_SPLIT(split_equal_to_row_size);
+		tilemode[20] =  MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
+				ARRAY_MODE(ARRAY_2D_TILED_THICK) |
+				PIPE_CONFIG(ADDR_SURF_P4_8x16) |
+				BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+				BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+				MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+				NUM_BANKS(ADDR_SURF_16_BANK) |
+				TILE_SPLIT(split_equal_to_row_size);
 		tilemode[21] =  MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
 				ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
 				PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |

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

@@ -3236,6 +3236,7 @@ static void gfx_v8_0_tiling_mode_table_init(struct amdgpu_device *adev)
 		dev_warn(adev->dev,
 			 "Unknown chip type (%d) in function gfx_v8_0_tiling_mode_table_init() falling through to CHIP_CARRIZO\n",
 			 adev->asic_type);
+		/* fall through */
 
 	case CHIP_CARRIZO:
 		modearray[0] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |

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

@@ -40,6 +40,8 @@
 
 #include "ivsrcid/gfx/irqsrcs_gfx_9_0.h"
 
+#include "amdgpu_ras.h"
+
 #define GFX9_NUM_GFX_RINGS     1
 #define GFX9_MEC_HPD_SIZE 4096
 #define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L
@@ -576,6 +578,27 @@ static void gfx_v9_0_check_fw_write_wait(struct amdgpu_device *adev)
 	}
 }
 
+static void gfx_v9_0_check_if_need_gfxoff(struct amdgpu_device *adev)
+{
+	switch (adev->asic_type) {
+	case CHIP_VEGA10:
+	case CHIP_VEGA12:
+	case CHIP_VEGA20:
+		break;
+	case CHIP_RAVEN:
+		if (adev->rev_id >= 0x8 || adev->pdev->device == 0x15d8)
+			break;
+		if ((adev->gfx.rlc_fw_version < 531) ||
+		    (adev->gfx.rlc_fw_version == 53815) ||
+		    (adev->gfx.rlc_feature_version < 1) ||
+		    !adev->gfx.rlc.is_rlc_v2_1)
+			adev->pm.pp_feature &= ~PP_GFXOFF_MASK;
+		break;
+	default:
+		break;
+	}
+}
+
 static int gfx_v9_0_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
@@ -828,6 +851,7 @@ static int gfx_v9_0_init_microcode(struct amdgpu_device *adev)
 	}
 
 out:
+	gfx_v9_0_check_if_need_gfxoff(adev);
 	gfx_v9_0_check_fw_write_wait(adev);
 	if (err) {
 		dev_err(adev->dev,
@@ -1639,6 +1663,18 @@ static int gfx_v9_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	/* ECC error */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_9_0__SRCID__CP_ECC_ERROR,
+			      &adev->gfx.cp_ecc_error_irq);
+	if (r)
+		return r;
+
+	/* FUE error */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_9_0__SRCID__CP_FUE_ERROR,
+			      &adev->gfx.cp_ecc_error_irq);
+	if (r)
+		return r;
+
 	adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE;
 
 	gfx_v9_0_scratch_init(adev);
@@ -1731,6 +1767,20 @@ static int gfx_v9_0_sw_fini(void *handle)
 	int i;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX) &&
+			adev->gfx.ras_if) {
+		struct ras_common_if *ras_if = adev->gfx.ras_if;
+		struct ras_ih_if ih_info = {
+			.head = *ras_if,
+		};
+
+		amdgpu_ras_debugfs_remove(adev, ras_if);
+		amdgpu_ras_sysfs_remove(adev, ras_if);
+		amdgpu_ras_interrupt_remove_handler(adev,  &ih_info);
+		amdgpu_ras_feature_enable(adev, ras_if, 0);
+		kfree(ras_if);
+	}
+
 	amdgpu_bo_free_kernel(&adev->gds.oa_gfx_bo, NULL, NULL);
 	amdgpu_bo_free_kernel(&adev->gds.gws_gfx_bo, NULL, NULL);
 	amdgpu_bo_free_kernel(&adev->gds.gds_gfx_bo, NULL, NULL);
@@ -3305,6 +3355,7 @@ static int gfx_v9_0_hw_fini(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
+	amdgpu_irq_put(adev, &adev->gfx.cp_ecc_error_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
 
@@ -3494,6 +3545,80 @@ static int gfx_v9_0_early_init(void *handle)
 	return 0;
 }
 
+static int gfx_v9_0_process_ras_data_cb(struct amdgpu_device *adev,
+		struct amdgpu_iv_entry *entry);
+
+static int gfx_v9_0_ecc_late_init(void *handle)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct ras_common_if **ras_if = &adev->gfx.ras_if;
+	struct ras_ih_if ih_info = {
+		.cb = gfx_v9_0_process_ras_data_cb,
+	};
+	struct ras_fs_if fs_info = {
+		.sysfs_name = "gfx_err_count",
+		.debugfs_name = "gfx_err_inject",
+	};
+	struct ras_common_if ras_block = {
+		.block = AMDGPU_RAS_BLOCK__GFX,
+		.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
+		.sub_block_index = 0,
+		.name = "gfx",
+	};
+	int r;
+
+	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) {
+		amdgpu_ras_feature_enable(adev, &ras_block, 0);
+		return 0;
+	}
+
+	if (*ras_if)
+		goto resume;
+
+	*ras_if = kmalloc(sizeof(**ras_if), GFP_KERNEL);
+	if (!*ras_if)
+		return -ENOMEM;
+
+	**ras_if = ras_block;
+
+	r = amdgpu_ras_feature_enable(adev, *ras_if, 1);
+	if (r)
+		goto feature;
+
+	ih_info.head = **ras_if;
+	fs_info.head = **ras_if;
+
+	r = amdgpu_ras_interrupt_add_handler(adev, &ih_info);
+	if (r)
+		goto interrupt;
+
+	r = amdgpu_ras_debugfs_create(adev, &fs_info);
+	if (r)
+		goto debugfs;
+
+	r = amdgpu_ras_sysfs_create(adev, &fs_info);
+	if (r)
+		goto sysfs;
+resume:
+	r = amdgpu_irq_get(adev, &adev->gfx.cp_ecc_error_irq, 0);
+	if (r)
+		goto irq;
+
+	return 0;
+irq:
+	amdgpu_ras_sysfs_remove(adev, *ras_if);
+sysfs:
+	amdgpu_ras_debugfs_remove(adev, *ras_if);
+debugfs:
+	amdgpu_ras_interrupt_remove_handler(adev, &ih_info);
+interrupt:
+	amdgpu_ras_feature_enable(adev, *ras_if, 0);
+feature:
+	kfree(*ras_if);
+	*ras_if = NULL;
+	return -EINVAL;
+}
+
 static int gfx_v9_0_late_init(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
@@ -3507,6 +3632,10 @@ static int gfx_v9_0_late_init(void *handle)
 	if (r)
 		return r;
 
+	r = gfx_v9_0_ecc_late_init(handle);
+	if (r)
+		return r;
+
 	return 0;
 }
 
@@ -4543,6 +4672,45 @@ static int gfx_v9_0_set_priv_inst_fault_state(struct amdgpu_device *adev,
 	return 0;
 }
 
+#define ENABLE_ECC_ON_ME_PIPE(me, pipe)				\
+	WREG32_FIELD15(GC, 0, CP_ME##me##_PIPE##pipe##_INT_CNTL,\
+			CP_ECC_ERROR_INT_ENABLE, 1)
+
+#define DISABLE_ECC_ON_ME_PIPE(me, pipe)			\
+	WREG32_FIELD15(GC, 0, CP_ME##me##_PIPE##pipe##_INT_CNTL,\
+			CP_ECC_ERROR_INT_ENABLE, 0)
+
+static int gfx_v9_0_set_cp_ecc_error_state(struct amdgpu_device *adev,
+					      struct amdgpu_irq_src *source,
+					      unsigned type,
+					      enum amdgpu_interrupt_state state)
+{
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		WREG32_FIELD15(GC, 0, CP_INT_CNTL_RING0,
+				CP_ECC_ERROR_INT_ENABLE, 0);
+		DISABLE_ECC_ON_ME_PIPE(1, 0);
+		DISABLE_ECC_ON_ME_PIPE(1, 1);
+		DISABLE_ECC_ON_ME_PIPE(1, 2);
+		DISABLE_ECC_ON_ME_PIPE(1, 3);
+		break;
+
+	case AMDGPU_IRQ_STATE_ENABLE:
+		WREG32_FIELD15(GC, 0, CP_INT_CNTL_RING0,
+				CP_ECC_ERROR_INT_ENABLE, 1);
+		ENABLE_ECC_ON_ME_PIPE(1, 0);
+		ENABLE_ECC_ON_ME_PIPE(1, 1);
+		ENABLE_ECC_ON_ME_PIPE(1, 2);
+		ENABLE_ECC_ON_ME_PIPE(1, 3);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+
 static int gfx_v9_0_set_eop_interrupt_state(struct amdgpu_device *adev,
 					    struct amdgpu_irq_src *src,
 					    unsigned type,
@@ -4659,6 +4827,28 @@ static int gfx_v9_0_priv_inst_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int gfx_v9_0_process_ras_data_cb(struct amdgpu_device *adev,
+		struct amdgpu_iv_entry *entry)
+{
+	/* TODO ue will trigger an interrupt. */
+	kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
+	amdgpu_ras_reset_gpu(adev, 0);
+	return AMDGPU_RAS_UE;
+}
+
+static int gfx_v9_0_cp_ecc_error_irq(struct amdgpu_device *adev,
+				  struct amdgpu_irq_src *source,
+				  struct amdgpu_iv_entry *entry)
+{
+	struct ras_dispatch_if ih_data = {
+		.head = *adev->gfx.ras_if,
+		.entry = entry,
+	};
+	DRM_ERROR("CP ECC ERROR IRQ\n");
+	amdgpu_ras_interrupt_dispatch(adev, &ih_data);
+	return 0;
+}
+
 static const struct amd_ip_funcs gfx_v9_0_ip_funcs = {
 	.name = "gfx_v9_0",
 	.early_init = gfx_v9_0_early_init,
@@ -4820,6 +5010,12 @@ static const struct amdgpu_irq_src_funcs gfx_v9_0_priv_inst_irq_funcs = {
 	.process = gfx_v9_0_priv_inst_irq,
 };
 
+static const struct amdgpu_irq_src_funcs gfx_v9_0_cp_ecc_error_irq_funcs = {
+	.set = gfx_v9_0_set_cp_ecc_error_state,
+	.process = gfx_v9_0_cp_ecc_error_irq,
+};
+
+
 static void gfx_v9_0_set_irq_funcs(struct amdgpu_device *adev)
 {
 	adev->gfx.eop_irq.num_types = AMDGPU_CP_IRQ_LAST;
@@ -4830,6 +5026,9 @@ static void gfx_v9_0_set_irq_funcs(struct amdgpu_device *adev)
 
 	adev->gfx.priv_inst_irq.num_types = 1;
 	adev->gfx.priv_inst_irq.funcs = &gfx_v9_0_priv_inst_irq_funcs;
+
+	adev->gfx.cp_ecc_error_irq.num_types = 2; /*C5 ECC error and C9 FUE error*/
+	adev->gfx.cp_ecc_error_irq.funcs = &gfx_v9_0_cp_ecc_error_irq_funcs;
 }
 
 static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev)

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

@@ -143,7 +143,7 @@ static void gfxhub_v1_0_init_cache_regs(struct amdgpu_device *adev)
 	/* XXX for emulation, Refer to closed source code.*/
 	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, L2_PDE0_CACHE_TAG_GENERATION_MODE,
 			    0);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 1);
+	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0);
 	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
 	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0);
 	WREG32_SOC15(GC, 0, mmVM_L2_CNTL, tmp);
@@ -236,7 +236,7 @@ static void gfxhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
 				    block_size);
 		/* Send no-retry XNACK on fault to suppress VM fault storm. */
 		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
+				    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 1);
 		WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT1_CNTL, i, tmp);
 		WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, i*2, 0);
 		WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, i*2, 0);

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

@@ -225,7 +225,7 @@ static void gmc_v6_0_vram_gtt_location(struct amdgpu_device *adev,
 	u64 base = RREG32(mmMC_VM_FB_LOCATION) & 0xFFFF;
 	base <<= 24;
 
-	amdgpu_gmc_vram_location(adev, &adev->gmc, base);
+	amdgpu_gmc_vram_location(adev, mc, base);
 	amdgpu_gmc_gart_location(adev, mc);
 }
 
@@ -383,20 +383,6 @@ static uint64_t gmc_v6_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 	return pd_addr;
 }
 
-static int gmc_v6_0_set_pte_pde(struct amdgpu_device *adev, void *cpu_pt_addr,
-				uint32_t gpu_page_idx, uint64_t addr,
-				uint64_t flags)
-{
-	void __iomem *ptr = (void *)cpu_pt_addr;
-	uint64_t value;
-
-	value = addr & 0xFFFFFFFFFFFFF000ULL;
-	value |= flags;
-	writeq(value, ptr + (gpu_page_idx * 8));
-
-	return 0;
-}
-
 static uint64_t gmc_v6_0_get_vm_pte_flags(struct amdgpu_device *adev,
 					  uint32_t flags)
 {
@@ -1169,7 +1155,6 @@ static const struct amd_ip_funcs gmc_v6_0_ip_funcs = {
 static const struct amdgpu_gmc_funcs gmc_v6_0_gmc_funcs = {
 	.flush_gpu_tlb = gmc_v6_0_flush_gpu_tlb,
 	.emit_flush_gpu_tlb = gmc_v6_0_emit_flush_gpu_tlb,
-	.set_pte_pde = gmc_v6_0_set_pte_pde,
 	.set_prt = gmc_v6_0_set_prt,
 	.get_vm_pde = gmc_v6_0_get_vm_pde,
 	.get_vm_pte_flags = gmc_v6_0_get_vm_pte_flags

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

@@ -242,7 +242,7 @@ static void gmc_v7_0_vram_gtt_location(struct amdgpu_device *adev,
 	u64 base = RREG32(mmMC_VM_FB_LOCATION) & 0xFFFF;
 	base <<= 24;
 
-	amdgpu_gmc_vram_location(adev, &adev->gmc, base);
+	amdgpu_gmc_vram_location(adev, mc, base);
 	amdgpu_gmc_gart_location(adev, mc);
 }
 
@@ -460,31 +460,6 @@ static void gmc_v7_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
 	amdgpu_ring_emit_wreg(ring, mmIH_VMID_0_LUT + vmid, pasid);
 }
 
-/**
- * gmc_v7_0_set_pte_pde - update the page tables using MMIO
- *
- * @adev: amdgpu_device pointer
- * @cpu_pt_addr: cpu address of the page table
- * @gpu_page_idx: entry in the page table to update
- * @addr: dst addr to write into pte/pde
- * @flags: access flags
- *
- * Update the page tables using the CPU.
- */
-static int gmc_v7_0_set_pte_pde(struct amdgpu_device *adev, void *cpu_pt_addr,
-				 uint32_t gpu_page_idx, uint64_t addr,
-				 uint64_t flags)
-{
-	void __iomem *ptr = (void *)cpu_pt_addr;
-	uint64_t value;
-
-	value = addr & 0xFFFFFFFFFFFFF000ULL;
-	value |= flags;
-	writeq(value, ptr + (gpu_page_idx * 8));
-
-	return 0;
-}
-
 static uint64_t gmc_v7_0_get_vm_pte_flags(struct amdgpu_device *adev,
 					  uint32_t flags)
 {
@@ -1376,7 +1351,6 @@ static const struct amdgpu_gmc_funcs gmc_v7_0_gmc_funcs = {
 	.flush_gpu_tlb = gmc_v7_0_flush_gpu_tlb,
 	.emit_flush_gpu_tlb = gmc_v7_0_emit_flush_gpu_tlb,
 	.emit_pasid_mapping = gmc_v7_0_emit_pasid_mapping,
-	.set_pte_pde = gmc_v7_0_set_pte_pde,
 	.set_prt = gmc_v7_0_set_prt,
 	.get_vm_pte_flags = gmc_v7_0_get_vm_pte_flags,
 	.get_vm_pde = gmc_v7_0_get_vm_pde

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

@@ -433,7 +433,7 @@ static void gmc_v8_0_vram_gtt_location(struct amdgpu_device *adev,
 		base = RREG32(mmMC_VM_FB_LOCATION) & 0xFFFF;
 	base <<= 24;
 
-	amdgpu_gmc_vram_location(adev, &adev->gmc, base);
+	amdgpu_gmc_vram_location(adev, mc, base);
 	amdgpu_gmc_gart_location(adev, mc);
 }
 
@@ -662,50 +662,26 @@ static void gmc_v8_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
 	amdgpu_ring_emit_wreg(ring, mmIH_VMID_0_LUT + vmid, pasid);
 }
 
-/**
- * gmc_v8_0_set_pte_pde - update the page tables using MMIO
+/*
+ * PTE format on VI:
+ * 63:40 reserved
+ * 39:12 4k physical page base address
+ * 11:7 fragment
+ * 6 write
+ * 5 read
+ * 4 exe
+ * 3 reserved
+ * 2 snooped
+ * 1 system
+ * 0 valid
  *
- * @adev: amdgpu_device pointer
- * @cpu_pt_addr: cpu address of the page table
- * @gpu_page_idx: entry in the page table to update
- * @addr: dst addr to write into pte/pde
- * @flags: access flags
- *
- * Update the page tables using the CPU.
+ * PDE format on VI:
+ * 63:59 block fragment size
+ * 58:40 reserved
+ * 39:1 physical base address of PTE
+ * bits 5:1 must be 0.
+ * 0 valid
  */
-static int gmc_v8_0_set_pte_pde(struct amdgpu_device *adev, void *cpu_pt_addr,
-				uint32_t gpu_page_idx, uint64_t addr,
-				uint64_t flags)
-{
-	void __iomem *ptr = (void *)cpu_pt_addr;
-	uint64_t value;
-
-	/*
-	 * PTE format on VI:
-	 * 63:40 reserved
-	 * 39:12 4k physical page base address
-	 * 11:7 fragment
-	 * 6 write
-	 * 5 read
-	 * 4 exe
-	 * 3 reserved
-	 * 2 snooped
-	 * 1 system
-	 * 0 valid
-	 *
-	 * PDE format on VI:
-	 * 63:59 block fragment size
-	 * 58:40 reserved
-	 * 39:1 physical base address of PTE
-	 * bits 5:1 must be 0.
-	 * 0 valid
-	 */
-	value = addr & 0x000000FFFFFFF000ULL;
-	value |= flags;
-	writeq(value, ptr + (gpu_page_idx * 8));
-
-	return 0;
-}
 
 static uint64_t gmc_v8_0_get_vm_pte_flags(struct amdgpu_device *adev,
 					  uint32_t flags)
@@ -1743,7 +1719,6 @@ static const struct amdgpu_gmc_funcs gmc_v8_0_gmc_funcs = {
 	.flush_gpu_tlb = gmc_v8_0_flush_gpu_tlb,
 	.emit_flush_gpu_tlb = gmc_v8_0_emit_flush_gpu_tlb,
 	.emit_pasid_mapping = gmc_v8_0_emit_pasid_mapping,
-	.set_pte_pde = gmc_v8_0_set_pte_pde,
 	.set_prt = gmc_v8_0_set_prt,
 	.get_vm_pte_flags = gmc_v8_0_get_vm_pte_flags,
 	.get_vm_pde = gmc_v8_0_get_vm_pde

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

@@ -47,6 +47,8 @@
 
 #include "ivsrcid/vmc/irqsrcs_vmc_1_0.h"
 
+#include "amdgpu_ras.h"
+
 /* add these here since we already include dce12 headers and these are for DCN */
 #define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION                                                          0x055d
 #define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION_BASE_IDX                                                 2
@@ -84,121 +86,176 @@ static const struct soc15_reg_golden golden_settings_athub_1_0_0[] =
 	SOC15_REG_GOLDEN_VALUE(ATHUB, 0, mmRPB_ARB_CNTL2, 0x00ff00ff, 0x00080008)
 };
 
-/* Ecc related register addresses, (BASE + reg offset) */
-/* Universal Memory Controller caps (may be fused). */
-/* UMCCH:UmcLocalCap */
-#define UMCLOCALCAPS_ADDR0	(0x00014306 + 0x00000000)
-#define UMCLOCALCAPS_ADDR1	(0x00014306 + 0x00000800)
-#define UMCLOCALCAPS_ADDR2	(0x00014306 + 0x00001000)
-#define UMCLOCALCAPS_ADDR3	(0x00014306 + 0x00001800)
-#define UMCLOCALCAPS_ADDR4	(0x00054306 + 0x00000000)
-#define UMCLOCALCAPS_ADDR5	(0x00054306 + 0x00000800)
-#define UMCLOCALCAPS_ADDR6	(0x00054306 + 0x00001000)
-#define UMCLOCALCAPS_ADDR7	(0x00054306 + 0x00001800)
-#define UMCLOCALCAPS_ADDR8	(0x00094306 + 0x00000000)
-#define UMCLOCALCAPS_ADDR9	(0x00094306 + 0x00000800)
-#define UMCLOCALCAPS_ADDR10	(0x00094306 + 0x00001000)
-#define UMCLOCALCAPS_ADDR11	(0x00094306 + 0x00001800)
-#define UMCLOCALCAPS_ADDR12	(0x000d4306 + 0x00000000)
-#define UMCLOCALCAPS_ADDR13	(0x000d4306 + 0x00000800)
-#define UMCLOCALCAPS_ADDR14	(0x000d4306 + 0x00001000)
-#define UMCLOCALCAPS_ADDR15	(0x000d4306 + 0x00001800)
-
-/* Universal Memory Controller Channel config. */
-/* UMCCH:UMC_CONFIG */
-#define UMCCH_UMC_CONFIG_ADDR0	(0x00014040 + 0x00000000)
-#define UMCCH_UMC_CONFIG_ADDR1	(0x00014040 + 0x00000800)
-#define UMCCH_UMC_CONFIG_ADDR2	(0x00014040 + 0x00001000)
-#define UMCCH_UMC_CONFIG_ADDR3	(0x00014040 + 0x00001800)
-#define UMCCH_UMC_CONFIG_ADDR4	(0x00054040 + 0x00000000)
-#define UMCCH_UMC_CONFIG_ADDR5	(0x00054040 + 0x00000800)
-#define UMCCH_UMC_CONFIG_ADDR6	(0x00054040 + 0x00001000)
-#define UMCCH_UMC_CONFIG_ADDR7	(0x00054040 + 0x00001800)
-#define UMCCH_UMC_CONFIG_ADDR8	(0x00094040 + 0x00000000)
-#define UMCCH_UMC_CONFIG_ADDR9	(0x00094040 + 0x00000800)
-#define UMCCH_UMC_CONFIG_ADDR10	(0x00094040 + 0x00001000)
-#define UMCCH_UMC_CONFIG_ADDR11	(0x00094040 + 0x00001800)
-#define UMCCH_UMC_CONFIG_ADDR12	(0x000d4040 + 0x00000000)
-#define UMCCH_UMC_CONFIG_ADDR13	(0x000d4040 + 0x00000800)
-#define UMCCH_UMC_CONFIG_ADDR14	(0x000d4040 + 0x00001000)
-#define UMCCH_UMC_CONFIG_ADDR15	(0x000d4040 + 0x00001800)
-
-/* Universal Memory Controller Channel Ecc config. */
-/* UMCCH:EccCtrl */
-#define UMCCH_ECCCTRL_ADDR0	(0x00014053 + 0x00000000)
-#define UMCCH_ECCCTRL_ADDR1	(0x00014053 + 0x00000800)
-#define UMCCH_ECCCTRL_ADDR2	(0x00014053 + 0x00001000)
-#define UMCCH_ECCCTRL_ADDR3	(0x00014053 + 0x00001800)
-#define UMCCH_ECCCTRL_ADDR4	(0x00054053 + 0x00000000)
-#define UMCCH_ECCCTRL_ADDR5	(0x00054053 + 0x00000800)
-#define UMCCH_ECCCTRL_ADDR6	(0x00054053 + 0x00001000)
-#define UMCCH_ECCCTRL_ADDR7	(0x00054053 + 0x00001800)
-#define UMCCH_ECCCTRL_ADDR8	(0x00094053 + 0x00000000)
-#define UMCCH_ECCCTRL_ADDR9	(0x00094053 + 0x00000800)
-#define UMCCH_ECCCTRL_ADDR10	(0x00094053 + 0x00001000)
-#define UMCCH_ECCCTRL_ADDR11	(0x00094053 + 0x00001800)
-#define UMCCH_ECCCTRL_ADDR12	(0x000d4053 + 0x00000000)
-#define UMCCH_ECCCTRL_ADDR13	(0x000d4053 + 0x00000800)
-#define UMCCH_ECCCTRL_ADDR14	(0x000d4053 + 0x00001000)
-#define UMCCH_ECCCTRL_ADDR15	(0x000d4053 + 0x00001800)
-
-static const uint32_t ecc_umclocalcap_addrs[] = {
-	UMCLOCALCAPS_ADDR0,
-	UMCLOCALCAPS_ADDR1,
-	UMCLOCALCAPS_ADDR2,
-	UMCLOCALCAPS_ADDR3,
-	UMCLOCALCAPS_ADDR4,
-	UMCLOCALCAPS_ADDR5,
-	UMCLOCALCAPS_ADDR6,
-	UMCLOCALCAPS_ADDR7,
-	UMCLOCALCAPS_ADDR8,
-	UMCLOCALCAPS_ADDR9,
-	UMCLOCALCAPS_ADDR10,
-	UMCLOCALCAPS_ADDR11,
-	UMCLOCALCAPS_ADDR12,
-	UMCLOCALCAPS_ADDR13,
-	UMCLOCALCAPS_ADDR14,
-	UMCLOCALCAPS_ADDR15,
+static const uint32_t ecc_umc_mcumc_ctrl_addrs[] = {
+	(0x000143c0 + 0x00000000),
+	(0x000143c0 + 0x00000800),
+	(0x000143c0 + 0x00001000),
+	(0x000143c0 + 0x00001800),
+	(0x000543c0 + 0x00000000),
+	(0x000543c0 + 0x00000800),
+	(0x000543c0 + 0x00001000),
+	(0x000543c0 + 0x00001800),
+	(0x000943c0 + 0x00000000),
+	(0x000943c0 + 0x00000800),
+	(0x000943c0 + 0x00001000),
+	(0x000943c0 + 0x00001800),
+	(0x000d43c0 + 0x00000000),
+	(0x000d43c0 + 0x00000800),
+	(0x000d43c0 + 0x00001000),
+	(0x000d43c0 + 0x00001800),
+	(0x001143c0 + 0x00000000),
+	(0x001143c0 + 0x00000800),
+	(0x001143c0 + 0x00001000),
+	(0x001143c0 + 0x00001800),
+	(0x001543c0 + 0x00000000),
+	(0x001543c0 + 0x00000800),
+	(0x001543c0 + 0x00001000),
+	(0x001543c0 + 0x00001800),
+	(0x001943c0 + 0x00000000),
+	(0x001943c0 + 0x00000800),
+	(0x001943c0 + 0x00001000),
+	(0x001943c0 + 0x00001800),
+	(0x001d43c0 + 0x00000000),
+	(0x001d43c0 + 0x00000800),
+	(0x001d43c0 + 0x00001000),
+	(0x001d43c0 + 0x00001800),
 };
 
-static const uint32_t ecc_umcch_umc_config_addrs[] = {
-	UMCCH_UMC_CONFIG_ADDR0,
-	UMCCH_UMC_CONFIG_ADDR1,
-	UMCCH_UMC_CONFIG_ADDR2,
-	UMCCH_UMC_CONFIG_ADDR3,
-	UMCCH_UMC_CONFIG_ADDR4,
-	UMCCH_UMC_CONFIG_ADDR5,
-	UMCCH_UMC_CONFIG_ADDR6,
-	UMCCH_UMC_CONFIG_ADDR7,
-	UMCCH_UMC_CONFIG_ADDR8,
-	UMCCH_UMC_CONFIG_ADDR9,
-	UMCCH_UMC_CONFIG_ADDR10,
-	UMCCH_UMC_CONFIG_ADDR11,
-	UMCCH_UMC_CONFIG_ADDR12,
-	UMCCH_UMC_CONFIG_ADDR13,
-	UMCCH_UMC_CONFIG_ADDR14,
-	UMCCH_UMC_CONFIG_ADDR15,
+static const uint32_t ecc_umc_mcumc_ctrl_mask_addrs[] = {
+	(0x000143e0 + 0x00000000),
+	(0x000143e0 + 0x00000800),
+	(0x000143e0 + 0x00001000),
+	(0x000143e0 + 0x00001800),
+	(0x000543e0 + 0x00000000),
+	(0x000543e0 + 0x00000800),
+	(0x000543e0 + 0x00001000),
+	(0x000543e0 + 0x00001800),
+	(0x000943e0 + 0x00000000),
+	(0x000943e0 + 0x00000800),
+	(0x000943e0 + 0x00001000),
+	(0x000943e0 + 0x00001800),
+	(0x000d43e0 + 0x00000000),
+	(0x000d43e0 + 0x00000800),
+	(0x000d43e0 + 0x00001000),
+	(0x000d43e0 + 0x00001800),
+	(0x001143e0 + 0x00000000),
+	(0x001143e0 + 0x00000800),
+	(0x001143e0 + 0x00001000),
+	(0x001143e0 + 0x00001800),
+	(0x001543e0 + 0x00000000),
+	(0x001543e0 + 0x00000800),
+	(0x001543e0 + 0x00001000),
+	(0x001543e0 + 0x00001800),
+	(0x001943e0 + 0x00000000),
+	(0x001943e0 + 0x00000800),
+	(0x001943e0 + 0x00001000),
+	(0x001943e0 + 0x00001800),
+	(0x001d43e0 + 0x00000000),
+	(0x001d43e0 + 0x00000800),
+	(0x001d43e0 + 0x00001000),
+	(0x001d43e0 + 0x00001800),
 };
 
-static const uint32_t ecc_umcch_eccctrl_addrs[] = {
-	UMCCH_ECCCTRL_ADDR0,
-	UMCCH_ECCCTRL_ADDR1,
-	UMCCH_ECCCTRL_ADDR2,
-	UMCCH_ECCCTRL_ADDR3,
-	UMCCH_ECCCTRL_ADDR4,
-	UMCCH_ECCCTRL_ADDR5,
-	UMCCH_ECCCTRL_ADDR6,
-	UMCCH_ECCCTRL_ADDR7,
-	UMCCH_ECCCTRL_ADDR8,
-	UMCCH_ECCCTRL_ADDR9,
-	UMCCH_ECCCTRL_ADDR10,
-	UMCCH_ECCCTRL_ADDR11,
-	UMCCH_ECCCTRL_ADDR12,
-	UMCCH_ECCCTRL_ADDR13,
-	UMCCH_ECCCTRL_ADDR14,
-	UMCCH_ECCCTRL_ADDR15,
+static const uint32_t ecc_umc_mcumc_status_addrs[] = {
+	(0x000143c2 + 0x00000000),
+	(0x000143c2 + 0x00000800),
+	(0x000143c2 + 0x00001000),
+	(0x000143c2 + 0x00001800),
+	(0x000543c2 + 0x00000000),
+	(0x000543c2 + 0x00000800),
+	(0x000543c2 + 0x00001000),
+	(0x000543c2 + 0x00001800),
+	(0x000943c2 + 0x00000000),
+	(0x000943c2 + 0x00000800),
+	(0x000943c2 + 0x00001000),
+	(0x000943c2 + 0x00001800),
+	(0x000d43c2 + 0x00000000),
+	(0x000d43c2 + 0x00000800),
+	(0x000d43c2 + 0x00001000),
+	(0x000d43c2 + 0x00001800),
+	(0x001143c2 + 0x00000000),
+	(0x001143c2 + 0x00000800),
+	(0x001143c2 + 0x00001000),
+	(0x001143c2 + 0x00001800),
+	(0x001543c2 + 0x00000000),
+	(0x001543c2 + 0x00000800),
+	(0x001543c2 + 0x00001000),
+	(0x001543c2 + 0x00001800),
+	(0x001943c2 + 0x00000000),
+	(0x001943c2 + 0x00000800),
+	(0x001943c2 + 0x00001000),
+	(0x001943c2 + 0x00001800),
+	(0x001d43c2 + 0x00000000),
+	(0x001d43c2 + 0x00000800),
+	(0x001d43c2 + 0x00001000),
+	(0x001d43c2 + 0x00001800),
 };
 
+static int gmc_v9_0_ecc_interrupt_state(struct amdgpu_device *adev,
+		struct amdgpu_irq_src *src,
+		unsigned type,
+		enum amdgpu_interrupt_state state)
+{
+	u32 bits, i, tmp, reg;
+
+	bits = 0x7f;
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		for (i = 0; i < ARRAY_SIZE(ecc_umc_mcumc_ctrl_addrs); i++) {
+			reg = ecc_umc_mcumc_ctrl_addrs[i];
+			tmp = RREG32(reg);
+			tmp &= ~bits;
+			WREG32(reg, tmp);
+		}
+		for (i = 0; i < ARRAY_SIZE(ecc_umc_mcumc_ctrl_mask_addrs); i++) {
+			reg = ecc_umc_mcumc_ctrl_mask_addrs[i];
+			tmp = RREG32(reg);
+			tmp &= ~bits;
+			WREG32(reg, tmp);
+		}
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < ARRAY_SIZE(ecc_umc_mcumc_ctrl_addrs); i++) {
+			reg = ecc_umc_mcumc_ctrl_addrs[i];
+			tmp = RREG32(reg);
+			tmp |= bits;
+			WREG32(reg, tmp);
+		}
+		for (i = 0; i < ARRAY_SIZE(ecc_umc_mcumc_ctrl_mask_addrs); i++) {
+			reg = ecc_umc_mcumc_ctrl_mask_addrs[i];
+			tmp = RREG32(reg);
+			tmp |= bits;
+			WREG32(reg, tmp);
+		}
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int gmc_v9_0_process_ras_data_cb(struct amdgpu_device *adev,
+		struct amdgpu_iv_entry *entry)
+{
+	kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
+	amdgpu_ras_reset_gpu(adev, 0);
+	return AMDGPU_RAS_UE;
+}
+
+static int gmc_v9_0_process_ecc_irq(struct amdgpu_device *adev,
+		struct amdgpu_irq_src *source,
+		struct amdgpu_iv_entry *entry)
+{
+	struct ras_dispatch_if ih_data = {
+		.head = *adev->gmc.ras_if,
+		.entry = entry,
+	};
+	amdgpu_ras_interrupt_dispatch(adev, &ih_data);
+	return 0;
+}
+
 static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 					struct amdgpu_irq_src *src,
 					unsigned type,
@@ -244,62 +301,6 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 	return 0;
 }
 
-/**
- * vega10_ih_prescreen_iv - prescreen an interrupt vector
- *
- * @adev: amdgpu_device pointer
- *
- * Returns true if the interrupt vector should be further processed.
- */
-static bool gmc_v9_0_prescreen_iv(struct amdgpu_device *adev,
-				  struct amdgpu_iv_entry *entry,
-				  uint64_t addr)
-{
-	struct amdgpu_vm *vm;
-	u64 key;
-	int r;
-
-	/* No PASID, can't identify faulting process */
-	if (!entry->pasid)
-		return true;
-
-	/* Not a retry fault */
-	if (!(entry->src_data[1] & 0x80))
-		return true;
-
-	/* Track retry faults in per-VM fault FIFO. */
-	spin_lock(&adev->vm_manager.pasid_lock);
-	vm = idr_find(&adev->vm_manager.pasid_idr, entry->pasid);
-	if (!vm) {
-		/* VM not found, process it normally */
-		spin_unlock(&adev->vm_manager.pasid_lock);
-		return true;
-	}
-
-	key = AMDGPU_VM_FAULT(entry->pasid, addr);
-	r = amdgpu_vm_add_fault(vm->fault_hash, key);
-
-	/* Hash table is full or the fault is already being processed,
-	 * ignore further page faults
-	 */
-	if (r != 0) {
-		spin_unlock(&adev->vm_manager.pasid_lock);
-		return false;
-	}
-	/* 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_vm_clear_fault(vm->fault_hash, key);
-		spin_unlock(&adev->vm_manager.pasid_lock);
-		return false;
-	}
-
-	spin_unlock(&adev->vm_manager.pasid_lock);
-	/* It's the first fault for this address, process it normally */
-	return true;
-}
-
 static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev,
 				struct amdgpu_irq_src *source,
 				struct amdgpu_iv_entry *entry)
@@ -312,9 +313,11 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev,
 	addr = (u64)entry->src_data[0] << 12;
 	addr |= ((u64)entry->src_data[1] & 0xf) << 44;
 
-	if (!gmc_v9_0_prescreen_iv(adev, entry, addr))
+	if (retry_fault && amdgpu_gmc_filter_faults(adev, addr, entry->pasid,
+						    entry->timestamp))
 		return 1; /* This also prevents sending it to KFD */
 
+	/* If it's the first fault for this address, process it normally */
 	if (!amdgpu_sriov_vf(adev)) {
 		status = RREG32(hub->vm_l2_pro_fault_status);
 		WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1);
@@ -350,10 +353,19 @@ static const struct amdgpu_irq_src_funcs gmc_v9_0_irq_funcs = {
 	.process = gmc_v9_0_process_interrupt,
 };
 
+
+static const struct amdgpu_irq_src_funcs gmc_v9_0_ecc_funcs = {
+	.set = gmc_v9_0_ecc_interrupt_state,
+	.process = gmc_v9_0_process_ecc_irq,
+};
+
 static void gmc_v9_0_set_irq_funcs(struct amdgpu_device *adev)
 {
 	adev->gmc.vm_fault.num_types = 1;
 	adev->gmc.vm_fault.funcs = &gmc_v9_0_irq_funcs;
+
+	adev->gmc.ecc_irq.num_types = 1;
+	adev->gmc.ecc_irq.funcs = &gmc_v9_0_ecc_funcs;
 }
 
 static uint32_t gmc_v9_0_get_invalidate_req(unsigned int vmid,
@@ -466,64 +478,37 @@ static void gmc_v9_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
 	amdgpu_ring_emit_wreg(ring, reg, pasid);
 }
 
-/**
- * gmc_v9_0_set_pte_pde - update the page tables using MMIO
+/*
+ * PTE format on VEGA 10:
+ * 63:59 reserved
+ * 58:57 mtype
+ * 56 F
+ * 55 L
+ * 54 P
+ * 53 SW
+ * 52 T
+ * 50:48 reserved
+ * 47:12 4k physical page base address
+ * 11:7 fragment
+ * 6 write
+ * 5 read
+ * 4 exe
+ * 3 Z
+ * 2 snooped
+ * 1 system
+ * 0 valid
  *
- * @adev: amdgpu_device pointer
- * @cpu_pt_addr: cpu address of the page table
- * @gpu_page_idx: entry in the page table to update
- * @addr: dst addr to write into pte/pde
- * @flags: access flags
- *
- * Update the page tables using the CPU.
+ * PDE format on VEGA 10:
+ * 63:59 block fragment size
+ * 58:55 reserved
+ * 54 P
+ * 53:48 reserved
+ * 47:6 physical base address of PD or PTE
+ * 5:3 reserved
+ * 2 C
+ * 1 system
+ * 0 valid
  */
-static int gmc_v9_0_set_pte_pde(struct amdgpu_device *adev, void *cpu_pt_addr,
-				uint32_t gpu_page_idx, uint64_t addr,
-				uint64_t flags)
-{
-	void __iomem *ptr = (void *)cpu_pt_addr;
-	uint64_t value;
-
-	/*
-	 * PTE format on VEGA 10:
-	 * 63:59 reserved
-	 * 58:57 mtype
-	 * 56 F
-	 * 55 L
-	 * 54 P
-	 * 53 SW
-	 * 52 T
-	 * 50:48 reserved
-	 * 47:12 4k physical page base address
-	 * 11:7 fragment
-	 * 6 write
-	 * 5 read
-	 * 4 exe
-	 * 3 Z
-	 * 2 snooped
-	 * 1 system
-	 * 0 valid
-	 *
-	 * PDE format on VEGA 10:
-	 * 63:59 block fragment size
-	 * 58:55 reserved
-	 * 54 P
-	 * 53:48 reserved
-	 * 47:6 physical base address of PD or PTE
-	 * 5:3 reserved
-	 * 2 C
-	 * 1 system
-	 * 0 valid
-	 */
-
-	/*
-	 * The following is for PTE only. GART does not have PDEs.
-	*/
-	value = addr & 0x0000FFFFFFFFF000ULL;
-	value |= flags;
-	writeq(value, ptr + (gpu_page_idx * 8));
-	return 0;
-}
 
 static uint64_t gmc_v9_0_get_vm_pte_flags(struct amdgpu_device *adev,
 						uint32_t flags)
@@ -593,7 +578,6 @@ static const struct amdgpu_gmc_funcs gmc_v9_0_gmc_funcs = {
 	.flush_gpu_tlb = gmc_v9_0_flush_gpu_tlb,
 	.emit_flush_gpu_tlb = gmc_v9_0_emit_flush_gpu_tlb,
 	.emit_pasid_mapping = gmc_v9_0_emit_pasid_mapping,
-	.set_pte_pde = gmc_v9_0_set_pte_pde,
 	.get_vm_pte_flags = gmc_v9_0_get_vm_pte_flags,
 	.get_vm_pde = gmc_v9_0_get_vm_pde
 };
@@ -620,85 +604,6 @@ static int gmc_v9_0_early_init(void *handle)
 	return 0;
 }
 
-static int gmc_v9_0_ecc_available(struct amdgpu_device *adev)
-{
-	uint32_t reg_val;
-	uint32_t reg_addr;
-	uint32_t field_val;
-	size_t i;
-	uint32_t fv2;
-	size_t lost_sheep;
-
-	DRM_DEBUG("ecc: gmc_v9_0_ecc_available()\n");
-
-	lost_sheep = 0;
-	for (i = 0; i < ARRAY_SIZE(ecc_umclocalcap_addrs); ++i) {
-		reg_addr = ecc_umclocalcap_addrs[i];
-		DRM_DEBUG("ecc: "
-			  "UMCCH_UmcLocalCap[%zu]: reg_addr: 0x%08x\n",
-			  i, reg_addr);
-		reg_val = RREG32(reg_addr);
-		field_val = REG_GET_FIELD(reg_val, UMCCH0_0_UmcLocalCap,
-					  EccDis);
-		DRM_DEBUG("ecc: "
-			  "reg_val: 0x%08x, "
-			  "EccDis: 0x%08x, ",
-			  reg_val, field_val);
-		if (field_val) {
-			DRM_ERROR("ecc: UmcLocalCap:EccDis is set.\n");
-			++lost_sheep;
-		}
-	}
-
-	for (i = 0; i < ARRAY_SIZE(ecc_umcch_umc_config_addrs); ++i) {
-		reg_addr = ecc_umcch_umc_config_addrs[i];
-		DRM_DEBUG("ecc: "
-			  "UMCCH0_0_UMC_CONFIG[%zu]: reg_addr: 0x%08x",
-			  i, reg_addr);
-		reg_val = RREG32(reg_addr);
-		field_val = REG_GET_FIELD(reg_val, UMCCH0_0_UMC_CONFIG,
-					  DramReady);
-		DRM_DEBUG("ecc: "
-			  "reg_val: 0x%08x, "
-			  "DramReady: 0x%08x\n",
-			  reg_val, field_val);
-
-		if (!field_val) {
-			DRM_ERROR("ecc: UMC_CONFIG:DramReady is not set.\n");
-			++lost_sheep;
-		}
-	}
-
-	for (i = 0; i < ARRAY_SIZE(ecc_umcch_eccctrl_addrs); ++i) {
-		reg_addr = ecc_umcch_eccctrl_addrs[i];
-		DRM_DEBUG("ecc: "
-			  "UMCCH_EccCtrl[%zu]: reg_addr: 0x%08x, ",
-			  i, reg_addr);
-		reg_val = RREG32(reg_addr);
-		field_val = REG_GET_FIELD(reg_val, UMCCH0_0_EccCtrl,
-					  WrEccEn);
-		fv2 = REG_GET_FIELD(reg_val, UMCCH0_0_EccCtrl,
-				    RdEccEn);
-		DRM_DEBUG("ecc: "
-			  "reg_val: 0x%08x, "
-			  "WrEccEn: 0x%08x, "
-			  "RdEccEn: 0x%08x\n",
-			  reg_val, field_val, fv2);
-
-		if (!field_val) {
-			DRM_DEBUG("ecc: WrEccEn is not set\n");
-			++lost_sheep;
-		}
-		if (!fv2) {
-			DRM_DEBUG("ecc: RdEccEn is not set\n");
-			++lost_sheep;
-		}
-	}
-
-	DRM_DEBUG("ecc: lost_sheep: %zu\n", lost_sheep);
-	return lost_sheep == 0;
-}
-
 static bool gmc_v9_0_keep_stolen_memory(struct amdgpu_device *adev)
 {
 
@@ -751,10 +656,82 @@ static int gmc_v9_0_allocate_vm_inv_eng(struct amdgpu_device *adev)
 	return 0;
 }
 
+static int gmc_v9_0_ecc_late_init(void *handle)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct ras_common_if **ras_if = &adev->gmc.ras_if;
+	struct ras_ih_if ih_info = {
+		.cb = gmc_v9_0_process_ras_data_cb,
+	};
+	struct ras_fs_if fs_info = {
+		.sysfs_name = "umc_err_count",
+		.debugfs_name = "umc_err_inject",
+	};
+	struct ras_common_if ras_block = {
+		.block = AMDGPU_RAS_BLOCK__UMC,
+		.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
+		.sub_block_index = 0,
+		.name = "umc",
+	};
+	int r;
+
+	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC)) {
+		amdgpu_ras_feature_enable(adev, &ras_block, 0);
+		return 0;
+	}
+	/* handle resume path. */
+	if (*ras_if)
+		goto resume;
+
+	*ras_if = kmalloc(sizeof(**ras_if), GFP_KERNEL);
+	if (!*ras_if)
+		return -ENOMEM;
+
+	**ras_if = ras_block;
+
+	r = amdgpu_ras_feature_enable(adev, *ras_if, 1);
+	if (r)
+		goto feature;
+
+	ih_info.head = **ras_if;
+	fs_info.head = **ras_if;
+
+	r = amdgpu_ras_interrupt_add_handler(adev, &ih_info);
+	if (r)
+		goto interrupt;
+
+	r = amdgpu_ras_debugfs_create(adev, &fs_info);
+	if (r)
+		goto debugfs;
+
+	r = amdgpu_ras_sysfs_create(adev, &fs_info);
+	if (r)
+		goto sysfs;
+resume:
+	r = amdgpu_irq_get(adev, &adev->gmc.ecc_irq, 0);
+	if (r)
+		goto irq;
+
+	return 0;
+irq:
+	amdgpu_ras_sysfs_remove(adev, *ras_if);
+sysfs:
+	amdgpu_ras_debugfs_remove(adev, *ras_if);
+debugfs:
+	amdgpu_ras_interrupt_remove_handler(adev, &ih_info);
+interrupt:
+	amdgpu_ras_feature_enable(adev, *ras_if, 0);
+feature:
+	kfree(*ras_if);
+	*ras_if = NULL;
+	return -EINVAL;
+}
+
+
 static int gmc_v9_0_late_init(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int r;
+	bool r;
 
 	if (!gmc_v9_0_keep_stolen_memory(adev))
 		amdgpu_bo_late_init(adev);
@@ -762,20 +739,36 @@ static int gmc_v9_0_late_init(void *handle)
 	r = gmc_v9_0_allocate_vm_inv_eng(adev);
 	if (r)
 		return r;
+	/* Check if ecc is available */
+	if (!amdgpu_sriov_vf(adev)) {
+		switch (adev->asic_type) {
+		case CHIP_VEGA10:
+		case CHIP_VEGA20:
+			r = amdgpu_atomfirmware_mem_ecc_supported(adev);
+			if (!r) {
+				DRM_INFO("ECC is not present.\n");
+				if (adev->df_funcs->enable_ecc_force_par_wr_rmw)
+					adev->df_funcs->enable_ecc_force_par_wr_rmw(adev, false);
+			} else {
+				DRM_INFO("ECC is active.\n");
+			}
 
-	if (adev->asic_type == CHIP_VEGA10 && !amdgpu_sriov_vf(adev)) {
-		r = gmc_v9_0_ecc_available(adev);
-		if (r == 1) {
-			DRM_INFO("ECC is active.\n");
-		} else if (r == 0) {
-			DRM_INFO("ECC is not present.\n");
-			adev->df_funcs->enable_ecc_force_par_wr_rmw(adev, false);
-		} else {
-			DRM_ERROR("gmc_v9_0_ecc_available() failed. r: %d\n", r);
-			return r;
+			r = amdgpu_atomfirmware_sram_ecc_supported(adev);
+			if (!r) {
+				DRM_INFO("SRAM ECC is not present.\n");
+			} else {
+				DRM_INFO("SRAM ECC is active.\n");
+			}
+			break;
+		default:
+			break;
 		}
 	}
 
+	r = gmc_v9_0_ecc_late_init(handle);
+	if (r)
+		return r;
+
 	return amdgpu_irq_get(adev, &adev->gmc.vm_fault, 0);
 }
 
@@ -787,7 +780,7 @@ static void gmc_v9_0_vram_gtt_location(struct amdgpu_device *adev,
 		base = mmhub_v1_0_get_fb_location(adev);
 	/* add the xgmi offset of the physical node */
 	base += adev->gmc.xgmi.physical_node_id * adev->gmc.xgmi.node_segment_size;
-	amdgpu_gmc_vram_location(adev, &adev->gmc, base);
+	amdgpu_gmc_vram_location(adev, mc, base);
 	amdgpu_gmc_gart_location(adev, mc);
 	if (!amdgpu_sriov_vf(adev))
 		amdgpu_gmc_agp_location(adev, mc);
@@ -987,6 +980,12 @@ static int gmc_v9_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	/* interrupt sent to DF. */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_DF, 0,
+			&adev->gmc.ecc_irq);
+	if (r)
+		return r;
+
 	/* Set the internal MC address mask
 	 * This is the max address of the GPU's
 	 * internal address space.
@@ -1052,6 +1051,22 @@ static int gmc_v9_0_sw_fini(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC) &&
+			adev->gmc.ras_if) {
+		struct ras_common_if *ras_if = adev->gmc.ras_if;
+		struct ras_ih_if ih_info = {
+			.head = *ras_if,
+		};
+
+		/*remove fs first*/
+		amdgpu_ras_debugfs_remove(adev, ras_if);
+		amdgpu_ras_sysfs_remove(adev, ras_if);
+		/*remove the IH*/
+		amdgpu_ras_interrupt_remove_handler(adev, &ih_info);
+		amdgpu_ras_feature_enable(adev, ras_if, 0);
+		kfree(ras_if);
+	}
+
 	amdgpu_gem_force_release(adev);
 	amdgpu_vm_manager_fini(adev);
 
@@ -1198,6 +1213,7 @@ static int gmc_v9_0_hw_fini(void *handle)
 		return 0;
 	}
 
+	amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
 	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
 	gmc_v9_0_gart_disable(adev);
 

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

@@ -2824,7 +2824,7 @@ static int kv_dpm_init(struct amdgpu_device *adev)
 		pi->caps_tcp_ramping = true;
 	}
 
-	if (adev->powerplay.pp_feature & PP_SCLK_DEEP_SLEEP_MASK)
+	if (adev->pm.pp_feature & PP_SCLK_DEEP_SLEEP_MASK)
 		pi->caps_sclk_ds = true;
 	else
 		pi->caps_sclk_ds = false;

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

@@ -163,7 +163,7 @@ static void mmhub_v1_0_init_cache_regs(struct amdgpu_device *adev)
 	/* XXX for emulation, Refer to closed source code.*/
 	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, L2_PDE0_CACHE_TAG_GENERATION_MODE,
 			    0);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 1);
+	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0);
 	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
 	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0);
 	WREG32_SOC15(MMHUB, 0, mmVM_L2_CNTL, tmp);
@@ -255,7 +255,7 @@ static void mmhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
 				    block_size);
 		/* Send no-retry XNACK on fault to suppress VM fault storm. */
 		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
+				    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 1);
 		WREG32_SOC15_OFFSET(MMHUB, 0, mmVM_CONTEXT1_CNTL, i, tmp);
 		WREG32_SOC15_OFFSET(MMHUB, 0, mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, i*2, 0);
 		WREG32_SOC15_OFFSET(MMHUB, 0, mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, i*2, 0);

drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c

@@ -118,7 +118,8 @@ static void nbio_v6_1_ih_doorbell_range(struct amdgpu_device *adev,
 
 	if (use_doorbell) {
 		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, OFFSET, doorbell_index);
-		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, SIZE, 2);
+		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
+						  BIF_IH_DOORBELL_RANGE, SIZE, 6);
 	} else
 		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, SIZE, 0);
 

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

@@ -43,6 +43,7 @@ enum psp_gfx_crtl_cmd_id
     GFX_CTRL_CMD_ID_ENABLE_INT      = 0x00050000,   /* enable PSP-to-Gfx interrupt */
     GFX_CTRL_CMD_ID_DISABLE_INT     = 0x00060000,   /* disable PSP-to-Gfx interrupt */
     GFX_CTRL_CMD_ID_MODE1_RST       = 0x00070000,   /* trigger the Mode 1 reset */
+    GFX_CTRL_CMD_ID_GBR_IH_SET      = 0x00080000,   /* set Gbr IH_RB_CNTL registers */
     GFX_CTRL_CMD_ID_CONSUME_CMD     = 0x000A0000,   /* send interrupt to psp for updating write pointer of vf */
     GFX_CTRL_CMD_ID_DESTROY_GPCOM_RING = 0x000C0000, /* destroy GPCOM ring */
 

drivers/gpu/drm/amd/amdgpu/psp_v11_0.c

@@ -33,6 +33,9 @@
 #include "sdma0/sdma0_4_0_offset.h"
 #include "nbio/nbio_7_4_offset.h"
 
+#include "oss/osssys_4_0_offset.h"
+#include "oss/osssys_4_0_sh_mask.h"
+
 MODULE_FIRMWARE("amdgpu/vega20_sos.bin");
 MODULE_FIRMWARE("amdgpu/vega20_asd.bin");
 MODULE_FIRMWARE("amdgpu/vega20_ta.bin");
@@ -113,6 +116,13 @@ static int psp_v11_0_init_microcode(struct psp_context *psp)
 		adev->psp.ta_xgmi_ucode_size = le32_to_cpu(ta_hdr->ta_xgmi_size_bytes);
 		adev->psp.ta_xgmi_start_addr = (uint8_t *)ta_hdr +
 			le32_to_cpu(ta_hdr->header.ucode_array_offset_bytes);
+
+		adev->psp.ta_fw_version = le32_to_cpu(ta_hdr->header.ucode_version);
+
+		adev->psp.ta_ras_ucode_version = le32_to_cpu(ta_hdr->ta_ras_ucode_version);
+		adev->psp.ta_ras_ucode_size = le32_to_cpu(ta_hdr->ta_ras_size_bytes);
+		adev->psp.ta_ras_start_addr = (uint8_t *)adev->psp.ta_xgmi_start_addr +
+			le32_to_cpu(ta_hdr->ta_ras_offset_bytes);
 	}
 
 	return 0;
@@ -217,6 +227,37 @@ static int psp_v11_0_bootloader_load_sos(struct psp_context *psp)
 	return ret;
 }
 
+static void psp_v11_0_reroute_ih(struct psp_context *psp)
+{
+	struct amdgpu_device *adev = psp->adev;
+	uint32_t tmp;
+
+	/* Change IH ring for VMC */
+	tmp = REG_SET_FIELD(0, IH_CLIENT_CFG_DATA, CREDIT_RETURN_ADDR, 0x1244b);
+	tmp = REG_SET_FIELD(tmp, IH_CLIENT_CFG_DATA, CLIENT_TYPE, 1);
+	tmp = REG_SET_FIELD(tmp, IH_CLIENT_CFG_DATA, RING_ID, 1);
+
+	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_69, 3);
+	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_70, tmp);
+	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, GFX_CTRL_CMD_ID_GBR_IH_SET);
+
+	mdelay(20);
+	psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
+		     0x80000000, 0x8000FFFF, false);
+
+	/* Change IH ring for UMC */
+	tmp = REG_SET_FIELD(0, IH_CLIENT_CFG_DATA, CREDIT_RETURN_ADDR, 0x1216b);
+	tmp = REG_SET_FIELD(tmp, IH_CLIENT_CFG_DATA, RING_ID, 1);
+
+	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_69, 4);
+	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_70, tmp);
+	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, GFX_CTRL_CMD_ID_GBR_IH_SET);
+
+	mdelay(20);
+	psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
+		     0x80000000, 0x8000FFFF, false);
+}
+
 static int psp_v11_0_ring_init(struct psp_context *psp,
 			      enum psp_ring_type ring_type)
 {
@@ -224,6 +265,8 @@ static int psp_v11_0_ring_init(struct psp_context *psp,
 	struct psp_ring *ring;
 	struct amdgpu_device *adev = psp->adev;
 
+	psp_v11_0_reroute_ih(psp);
+
 	ring = &psp->km_ring;
 
 	ring->ring_type = ring_type;
@@ -679,6 +722,54 @@ static int psp_v11_0_xgmi_get_node_id(struct psp_context *psp, uint64_t *node_id
 	return 0;
 }
 
+static int psp_v11_0_ras_trigger_error(struct psp_context *psp,
+		struct ta_ras_trigger_error_input *info)
+{
+	struct ta_ras_shared_memory *ras_cmd;
+	int ret;
+
+	if (!psp->ras.ras_initialized)
+		return -EINVAL;
+
+	ras_cmd = (struct ta_ras_shared_memory *)psp->ras.ras_shared_buf;
+	memset(ras_cmd, 0, sizeof(struct ta_ras_shared_memory));
+
+	ras_cmd->cmd_id = TA_RAS_COMMAND__TRIGGER_ERROR;
+	ras_cmd->ras_in_message.trigger_error = *info;
+
+	ret = psp_ras_invoke(psp, ras_cmd->cmd_id);
+	if (ret)
+		return -EINVAL;
+
+	return ras_cmd->ras_status;
+}
+
+static int psp_v11_0_ras_cure_posion(struct psp_context *psp, uint64_t *mode_ptr)
+{
+#if 0
+	// not support yet.
+	struct ta_ras_shared_memory *ras_cmd;
+	int ret;
+
+	if (!psp->ras.ras_initialized)
+		return -EINVAL;
+
+	ras_cmd = (struct ta_ras_shared_memory *)psp->ras.ras_shared_buf;
+	memset(ras_cmd, 0, sizeof(struct ta_ras_shared_memory));
+
+	ras_cmd->cmd_id = TA_RAS_COMMAND__CURE_POISON;
+	ras_cmd->ras_in_message.cure_poison.mode_ptr = mode_ptr;
+
+	ret = psp_ras_invoke(psp, ras_cmd->cmd_id);
+	if (ret)
+		return -EINVAL;
+
+	return ras_cmd->ras_status;
+#else
+	return -EINVAL;
+#endif
+}
+
 static const struct psp_funcs psp_v11_0_funcs = {
 	.init_microcode = psp_v11_0_init_microcode,
 	.bootloader_load_sysdrv = psp_v11_0_bootloader_load_sysdrv,
@@ -695,6 +786,8 @@ static const struct psp_funcs psp_v11_0_funcs = {
 	.xgmi_get_hive_id = psp_v11_0_xgmi_get_hive_id,
 	.xgmi_get_node_id = psp_v11_0_xgmi_get_node_id,
 	.support_vmr_ring = psp_v11_0_support_vmr_ring,
+	.ras_trigger_error = psp_v11_0_ras_trigger_error,
+	.ras_cure_posion = psp_v11_0_ras_cure_posion,
 };
 
 void psp_v11_0_set_psp_funcs(struct psp_context *psp)

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

@@ -37,6 +37,9 @@
 #include "sdma0/sdma0_4_0_offset.h"
 #include "nbio/nbio_6_1_offset.h"
 
+#include "oss/osssys_4_0_offset.h"
+#include "oss/osssys_4_0_sh_mask.h"
+
 MODULE_FIRMWARE("amdgpu/vega10_sos.bin");
 MODULE_FIRMWARE("amdgpu/vega10_asd.bin");
 MODULE_FIRMWARE("amdgpu/vega12_sos.bin");
@@ -252,6 +255,37 @@ static int psp_v3_1_ring_init(struct psp_context *psp,
 	return 0;
 }
 
+static void psp_v3_1_reroute_ih(struct psp_context *psp)
+{
+	struct amdgpu_device *adev = psp->adev;
+	uint32_t tmp;
+
+	/* Change IH ring for VMC */
+	tmp = REG_SET_FIELD(0, IH_CLIENT_CFG_DATA, CREDIT_RETURN_ADDR, 0x1244b);
+	tmp = REG_SET_FIELD(tmp, IH_CLIENT_CFG_DATA, CLIENT_TYPE, 1);
+	tmp = REG_SET_FIELD(tmp, IH_CLIENT_CFG_DATA, RING_ID, 1);
+
+	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_69, 3);
+	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_70, tmp);
+	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, GFX_CTRL_CMD_ID_GBR_IH_SET);
+
+	mdelay(20);
+	psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
+		     0x80000000, 0x8000FFFF, false);
+
+	/* Change IH ring for UMC */
+	tmp = REG_SET_FIELD(0, IH_CLIENT_CFG_DATA, CREDIT_RETURN_ADDR, 0x1216b);
+	tmp = REG_SET_FIELD(tmp, IH_CLIENT_CFG_DATA, RING_ID, 1);
+
+	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_69, 4);
+	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_70, tmp);
+	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, GFX_CTRL_CMD_ID_GBR_IH_SET);
+
+	mdelay(20);
+	psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64),
+		     0x80000000, 0x8000FFFF, false);
+}
+
 static int psp_v3_1_ring_create(struct psp_context *psp,
 				enum psp_ring_type ring_type)
 {
@@ -260,6 +294,8 @@ static int psp_v3_1_ring_create(struct psp_context *psp,
 	struct psp_ring *ring = &psp->km_ring;
 	struct amdgpu_device *adev = psp->adev;
 
+	psp_v3_1_reroute_ih(psp);
+
 	/* Write low address of the ring to C2PMSG_69 */
 	psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr);
 	WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_69, psp_ring_reg);

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

@@ -41,6 +41,8 @@
 #include "ivsrcid/sdma0/irqsrcs_sdma0_4_0.h"
 #include "ivsrcid/sdma1/irqsrcs_sdma1_4_0.h"
 
+#include "amdgpu_ras.h"
+
 MODULE_FIRMWARE("amdgpu/vega10_sdma.bin");
 MODULE_FIRMWARE("amdgpu/vega10_sdma1.bin");
 MODULE_FIRMWARE("amdgpu/vega12_sdma.bin");
@@ -1493,6 +1495,87 @@ static int sdma_v4_0_early_init(void *handle)
 	return 0;
 }
 
+static int sdma_v4_0_process_ras_data_cb(struct amdgpu_device *adev,
+		struct amdgpu_iv_entry *entry);
+
+static int sdma_v4_0_late_init(void *handle)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct ras_common_if **ras_if = &adev->sdma.ras_if;
+	struct ras_ih_if ih_info = {
+		.cb = sdma_v4_0_process_ras_data_cb,
+	};
+	struct ras_fs_if fs_info = {
+		.sysfs_name = "sdma_err_count",
+		.debugfs_name = "sdma_err_inject",
+	};
+	struct ras_common_if ras_block = {
+		.block = AMDGPU_RAS_BLOCK__SDMA,
+		.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
+		.sub_block_index = 0,
+		.name = "sdma",
+	};
+	int r;
+
+	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__SDMA)) {
+		amdgpu_ras_feature_enable(adev, &ras_block, 0);
+		return 0;
+	}
+
+	/* handle resume path. */
+	if (*ras_if)
+		goto resume;
+
+	*ras_if = kmalloc(sizeof(**ras_if), GFP_KERNEL);
+	if (!*ras_if)
+		return -ENOMEM;
+
+	**ras_if = ras_block;
+
+	r = amdgpu_ras_feature_enable(adev, *ras_if, 1);
+	if (r)
+		goto feature;
+
+	ih_info.head = **ras_if;
+	fs_info.head = **ras_if;
+
+	r = amdgpu_ras_interrupt_add_handler(adev, &ih_info);
+	if (r)
+		goto interrupt;
+
+	r = amdgpu_ras_debugfs_create(adev, &fs_info);
+	if (r)
+		goto debugfs;
+
+	r = amdgpu_ras_sysfs_create(adev, &fs_info);
+	if (r)
+		goto sysfs;
+resume:
+	r = amdgpu_irq_get(adev, &adev->sdma.ecc_irq, AMDGPU_SDMA_IRQ_ECC0);
+	if (r)
+		goto irq;
+
+	r = amdgpu_irq_get(adev, &adev->sdma.ecc_irq, AMDGPU_SDMA_IRQ_ECC1);
+	if (r) {
+		amdgpu_irq_put(adev, &adev->sdma.ecc_irq, AMDGPU_SDMA_IRQ_ECC0);
+		goto irq;
+	}
+
+	return 0;
+irq:
+	amdgpu_ras_sysfs_remove(adev, *ras_if);
+sysfs:
+	amdgpu_ras_debugfs_remove(adev, *ras_if);
+debugfs:
+	amdgpu_ras_interrupt_remove_handler(adev, &ih_info);
+interrupt:
+	amdgpu_ras_feature_enable(adev, *ras_if, 0);
+feature:
+	kfree(*ras_if);
+	*ras_if = NULL;
+	return -EINVAL;
+}
+
 static int sdma_v4_0_sw_init(void *handle)
 {
 	struct amdgpu_ring *ring;
@@ -1511,6 +1594,18 @@ static int sdma_v4_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	/* SDMA SRAM ECC event */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_SDMA0, SDMA0_4_0__SRCID__SDMA_SRAM_ECC,
+			&adev->sdma.ecc_irq);
+	if (r)
+		return r;
+
+	/* SDMA SRAM ECC event */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_SDMA1, SDMA1_4_0__SRCID__SDMA_SRAM_ECC,
+			&adev->sdma.ecc_irq);
+	if (r)
+		return r;
+
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		ring = &adev->sdma.instance[i].ring;
 		ring->ring_obj = NULL;
@@ -1561,6 +1656,22 @@ static int sdma_v4_0_sw_fini(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	int i;
 
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__SDMA) &&
+			adev->sdma.ras_if) {
+		struct ras_common_if *ras_if = adev->sdma.ras_if;
+		struct ras_ih_if ih_info = {
+			.head = *ras_if,
+		};
+
+		/*remove fs first*/
+		amdgpu_ras_debugfs_remove(adev, ras_if);
+		amdgpu_ras_sysfs_remove(adev, ras_if);
+		/*remove the IH*/
+		amdgpu_ras_interrupt_remove_handler(adev, &ih_info);
+		amdgpu_ras_feature_enable(adev, ras_if, 0);
+		kfree(ras_if);
+	}
+
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		amdgpu_ring_fini(&adev->sdma.instance[i].ring);
 		if (adev->sdma.has_page_queue)
@@ -1598,6 +1709,9 @@ static int sdma_v4_0_hw_fini(void *handle)
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
+	amdgpu_irq_put(adev, &adev->sdma.ecc_irq, AMDGPU_SDMA_IRQ_ECC0);
+	amdgpu_irq_put(adev, &adev->sdma.ecc_irq, AMDGPU_SDMA_IRQ_ECC1);
+
 	sdma_v4_0_ctx_switch_enable(adev, false);
 	sdma_v4_0_enable(adev, false);
 
@@ -1714,6 +1828,52 @@ static int sdma_v4_0_process_trap_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int sdma_v4_0_process_ras_data_cb(struct amdgpu_device *adev,
+		struct amdgpu_iv_entry *entry)
+{
+	uint32_t instance, err_source;
+
+	switch (entry->client_id) {
+	case SOC15_IH_CLIENTID_SDMA0:
+		instance = 0;
+		break;
+	case SOC15_IH_CLIENTID_SDMA1:
+		instance = 1;
+		break;
+	default:
+		return 0;
+	}
+
+	switch (entry->src_id) {
+	case SDMA0_4_0__SRCID__SDMA_SRAM_ECC:
+		err_source = 0;
+		break;
+	case SDMA0_4_0__SRCID__SDMA_ECC:
+		err_source = 1;
+		break;
+	default:
+		return 0;
+	}
+
+	kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
+
+	amdgpu_ras_reset_gpu(adev, 0);
+
+	return AMDGPU_RAS_UE;
+}
+
+static int sdma_v4_0_process_ecc_irq(struct amdgpu_device *adev,
+				      struct amdgpu_irq_src *source,
+				      struct amdgpu_iv_entry *entry)
+{
+	struct ras_dispatch_if ih_data = {
+		.head = *adev->sdma.ras_if,
+		.entry = entry,
+	};
+	amdgpu_ras_interrupt_dispatch(adev, &ih_data);
+	return 0;
+}
+
 static int sdma_v4_0_process_illegal_inst_irq(struct amdgpu_device *adev,
 					      struct amdgpu_irq_src *source,
 					      struct amdgpu_iv_entry *entry)
@@ -1741,6 +1901,25 @@ static int sdma_v4_0_process_illegal_inst_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int sdma_v4_0_set_ecc_irq_state(struct amdgpu_device *adev,
+					struct amdgpu_irq_src *source,
+					unsigned type,
+					enum amdgpu_interrupt_state state)
+{
+	u32 sdma_edc_config;
+
+	u32 reg_offset = (type == AMDGPU_SDMA_IRQ_ECC0) ?
+		sdma_v4_0_get_reg_offset(adev, 0, mmSDMA0_EDC_CONFIG) :
+		sdma_v4_0_get_reg_offset(adev, 1, mmSDMA0_EDC_CONFIG);
+
+	sdma_edc_config = RREG32(reg_offset);
+	sdma_edc_config = REG_SET_FIELD(sdma_edc_config, SDMA0_EDC_CONFIG, ECC_INT_ENABLE,
+		       state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+	WREG32(reg_offset, sdma_edc_config);
+
+	return 0;
+}
+
 static void sdma_v4_0_update_medium_grain_clock_gating(
 		struct amdgpu_device *adev,
 		bool enable)
@@ -1906,7 +2085,7 @@ static void sdma_v4_0_get_clockgating_state(void *handle, u32 *flags)
 const struct amd_ip_funcs sdma_v4_0_ip_funcs = {
 	.name = "sdma_v4_0",
 	.early_init = sdma_v4_0_early_init,
-	.late_init = NULL,
+	.late_init = sdma_v4_0_late_init,
 	.sw_init = sdma_v4_0_sw_init,
 	.sw_fini = sdma_v4_0_sw_fini,
 	.hw_init = sdma_v4_0_hw_init,
@@ -2008,11 +2187,20 @@ static const struct amdgpu_irq_src_funcs sdma_v4_0_illegal_inst_irq_funcs = {
 	.process = sdma_v4_0_process_illegal_inst_irq,
 };
 
+static const struct amdgpu_irq_src_funcs sdma_v4_0_ecc_irq_funcs = {
+	.set = sdma_v4_0_set_ecc_irq_state,
+	.process = sdma_v4_0_process_ecc_irq,
+};
+
+
+
 static void sdma_v4_0_set_irq_funcs(struct amdgpu_device *adev)
 {
 	adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_LAST;
 	adev->sdma.trap_irq.funcs = &sdma_v4_0_trap_irq_funcs;
 	adev->sdma.illegal_inst_irq.funcs = &sdma_v4_0_illegal_inst_irq_funcs;
+	adev->sdma.ecc_irq.num_types = AMDGPU_SDMA_IRQ_LAST;
+	adev->sdma.ecc_irq.funcs = &sdma_v4_0_ecc_irq_funcs;
 }
 
 /**
@@ -2077,8 +2265,8 @@ static const struct amdgpu_buffer_funcs sdma_v4_0_buffer_funcs = {
 static void sdma_v4_0_set_buffer_funcs(struct amdgpu_device *adev)
 {
 	adev->mman.buffer_funcs = &sdma_v4_0_buffer_funcs;
-	if (adev->sdma.has_page_queue)
-		adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].page;
+	if (adev->sdma.has_page_queue && adev->sdma.num_instances > 1)
+		adev->mman.buffer_funcs_ring = &adev->sdma.instance[1].page;
 	else
 		adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring;
 }
@@ -2097,15 +2285,21 @@ static void sdma_v4_0_set_vm_pte_funcs(struct amdgpu_device *adev)
 	unsigned i;
 
 	adev->vm_manager.vm_pte_funcs = &sdma_v4_0_vm_pte_funcs;
-	for (i = 0; i < adev->sdma.num_instances; i++) {
-		if (adev->sdma.has_page_queue)
+	if (adev->sdma.has_page_queue && adev->sdma.num_instances > 1) {
+		for (i = 1; i < adev->sdma.num_instances; i++) {
 			sched = &adev->sdma.instance[i].page.sched;
-		else
+			adev->vm_manager.vm_pte_rqs[i - 1] =
+				&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
+		}
+		adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances - 1;
+	} else {
+		for (i = 0; i < adev->sdma.num_instances; i++) {
 			sched = &adev->sdma.instance[i].ring.sched;
-		adev->vm_manager.vm_pte_rqs[i] =
-			&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
+			adev->vm_manager.vm_pte_rqs[i] =
+				&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
+		}
+		adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
 	}
-	adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
 }
 
 const struct amdgpu_ip_block_version sdma_v4_0_ip_block = {

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

@@ -4098,14 +4098,13 @@ static int si_notify_smc_display_change(struct amdgpu_device *adev,
 
 static void si_program_response_times(struct amdgpu_device *adev)
 {
-	u32 voltage_response_time, backbias_response_time, acpi_delay_time, vbi_time_out;
+	u32 voltage_response_time, acpi_delay_time, vbi_time_out;
 	u32 vddc_dly, acpi_dly, vbi_dly;
 	u32 reference_clock;
 
 	si_write_smc_soft_register(adev, SI_SMC_SOFT_REGISTER_mvdd_chg_time, 1);
 
 	voltage_response_time = (u32)adev->pm.dpm.voltage_response_time;
-	backbias_response_time = (u32)adev->pm.dpm.backbias_response_time;
 
 	if (voltage_response_time == 0)
 		voltage_response_time = 1000;

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

@@ -63,6 +63,7 @@
 #include "vcn_v1_0.h"
 #include "dce_virtual.h"
 #include "mxgpu_ai.h"
+#include "amdgpu_smu.h"
 
 #define mmMP0_MISC_CGTT_CTRL0                                                                   0x01b9
 #define mmMP0_MISC_CGTT_CTRL0_BASE_IDX                                                          0
@@ -392,6 +393,7 @@ void soc15_program_register_sequence(struct amdgpu_device *adev,
 static int soc15_asic_mode1_reset(struct amdgpu_device *adev)
 {
 	u32 i;
+	int ret = 0;
 
 	amdgpu_atombios_scratch_regs_engine_hung(adev, true);
 
@@ -402,7 +404,9 @@ static int soc15_asic_mode1_reset(struct amdgpu_device *adev)
 
 	pci_save_state(adev->pdev);
 
-	psp_gpu_reset(adev);
+	ret = psp_gpu_reset(adev);
+	if (ret)
+		dev_err(adev->dev, "GPU mode1 reset failed\n");
 
 	pci_restore_state(adev->pdev);
 
@@ -417,7 +421,7 @@ static int soc15_asic_mode1_reset(struct amdgpu_device *adev)
 
 	amdgpu_atombios_scratch_regs_engine_hung(adev, false);
 
-	return 0;
+	return ret;
 }
 
 static int soc15_asic_get_baco_capability(struct amdgpu_device *adev, bool *cap)
@@ -451,6 +455,8 @@ static int soc15_asic_baco_reset(struct amdgpu_device *adev)
 
 	dev_info(adev->dev, "GPU BACO reset\n");
 
+	adev->in_baco_reset = 1;
+
 	return 0;
 }
 
@@ -461,6 +467,7 @@ static int soc15_asic_reset(struct amdgpu_device *adev)
 
 	switch (adev->asic_type) {
 	case CHIP_VEGA10:
+	case CHIP_VEGA12:
 		soc15_asic_get_baco_capability(adev, &baco_reset);
 		break;
 	default:
@@ -602,8 +609,12 @@ int soc15_set_ip_blocks(struct amdgpu_device *adev)
 		}
 		amdgpu_device_ip_block_add(adev, &gfx_v9_0_ip_block);
 		amdgpu_device_ip_block_add(adev, &sdma_v4_0_ip_block);
-		if (!amdgpu_sriov_vf(adev))
-			amdgpu_device_ip_block_add(adev, &pp_smu_ip_block);
+		if (!amdgpu_sriov_vf(adev)) {
+			if (is_support_sw_smu(adev))
+				amdgpu_device_ip_block_add(adev, &smu_v11_0_ip_block);
+			else
+				amdgpu_device_ip_block_add(adev, &pp_smu_ip_block);
+		}
 		if (adev->enable_virtual_display || amdgpu_sriov_vf(adev))
 			amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block);
 #if defined(CONFIG_DRM_AMD_DC)
@@ -927,7 +938,7 @@ static int soc15_common_early_init(void *handle)
 			adev->pg_flags = AMD_PG_SUPPORT_SDMA | AMD_PG_SUPPORT_VCN;
 		}
 
-		if (adev->powerplay.pp_feature & PP_GFXOFF_MASK)
+		if (adev->pm.pp_feature & PP_GFXOFF_MASK)
 			adev->pg_flags |= AMD_PG_SUPPORT_GFX_PG |
 				AMD_PG_SUPPORT_CP |
 				AMD_PG_SUPPORT_RLC_SMU_HS;

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

@@ -0,0 +1,108 @@
+/****************************************************************************\
+* 
+*  File Name      ta_ras_if.h
+*  Project        AMD PSP SW IP Module
+*
+*  Description    Interface to the RAS Trusted Application
+*
+*  Copyright 2019 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 _TA_RAS_IF_H
+#define _TA_RAS_IF_H
+
+/* Responses have bit 31 set */
+#define RSP_ID_MASK (1U << 31)
+#define RSP_ID(cmdId) (((uint32_t)(cmdId)) | RSP_ID_MASK)
+
+#define TA_NUM_BLOCK_MAX		14
+
+enum ras_command {
+	TA_RAS_COMMAND__ENABLE_FEATURES = 0,
+	TA_RAS_COMMAND__DISABLE_FEATURES,
+	TA_RAS_COMMAND__TRIGGER_ERROR,
+};
+
+enum ta_ras_status {
+	TA_RAS_STATUS__SUCCESS				= 0x00,
+	TA_RAS_STATUS__RESET_NEEDED			= 0x01,
+	TA_RAS_STATUS__ERROR_INVALID_PARAMETER		= 0x02,
+	TA_RAS_STATUS__ERROR_RAS_NOT_AVAILABLE		= 0x03,
+	TA_RAS_STATUS__ERROR_RAS_DUPLICATE_CMD		= 0x04,
+	TA_RAS_STATUS__ERROR_INJECTION_FAILED		= 0x05
+};
+
+enum ta_ras_block {
+	TA_RAS_BLOCK__UMC = 0,
+	TA_RAS_BLOCK__SDMA,
+	TA_RAS_BLOCK__GFX,
+	TA_RAS_BLOCK__MMHUB,
+	TA_RAS_BLOCK__ATHUB,
+	TA_RAS_BLOCK__PCIE_BIF,
+	TA_RAS_BLOCK__HDP,
+	TA_RAS_BLOCK__XGMI_WAFL,
+	TA_RAS_BLOCK__DF,
+	TA_RAS_BLOCK__SMN,
+	TA_RAS_BLOCK__SEM,
+	TA_RAS_BLOCK__MP0,
+	TA_RAS_BLOCK__MP1,
+	TA_RAS_BLOCK__FUSE = (TA_NUM_BLOCK_MAX - 1),
+};
+
+enum ta_ras_error_type {
+	TA_RAS_ERROR__NONE				= 0,
+	TA_RAS_ERROR__PARITY				= 1,
+	TA_RAS_ERROR__SINGLE_CORRECTABLE		= 2,
+	TA_RAS_ERROR__MULTI_UNCORRECTABLE		= 4,
+	TA_RAS_ERROR__POISON				= 8
+};
+
+struct ta_ras_enable_features_input {
+	enum ta_ras_block       block_id;
+	enum ta_ras_error_type  error_type;
+};
+
+struct ta_ras_disable_features_input {
+	enum ta_ras_block       block_id;
+	enum ta_ras_error_type  error_type;
+};
+
+struct ta_ras_trigger_error_input {
+	enum ta_ras_block		block_id;
+	enum ta_ras_error_type		inject_error_type;
+	uint32_t			sub_block_index;
+	uint64_t			address;
+	uint64_t			value;
+};
+
+union ta_ras_cmd_input {
+	struct ta_ras_enable_features_input	enable_features;
+	struct ta_ras_disable_features_input	disable_features;
+	struct ta_ras_trigger_error_input	trigger_error;
+};
+
+struct ta_ras_shared_memory {
+	uint32_t			cmd_id;
+	uint32_t			resp_id;
+	enum ta_ras_status		ras_status;
+	uint32_t			reserved;
+	union ta_ras_cmd_input		ras_in_message;
+};
+
+#endif // TL_RAS_IF_H_

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

@@ -136,6 +136,25 @@ static uint32_t vega10_ih_rb_cntl(struct amdgpu_ih_ring *ih, uint32_t ih_rb_cntl
 	return ih_rb_cntl;
 }
 
+static uint32_t vega10_ih_doorbell_rptr(struct amdgpu_ih_ring *ih)
+{
+	u32 ih_doorbell_rtpr = 0;
+
+	if (ih->use_doorbell) {
+		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+						 IH_DOORBELL_RPTR, OFFSET,
+						 ih->doorbell_index);
+		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+						 IH_DOORBELL_RPTR,
+						 ENABLE, 1);
+	} else {
+		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+						 IH_DOORBELL_RPTR,
+						 ENABLE, 0);
+	}
+	return ih_doorbell_rtpr;
+}
+
 /**
  * vega10_ih_irq_init - init and enable the interrupt ring
  *
@@ -150,8 +169,8 @@ static uint32_t vega10_ih_rb_cntl(struct amdgpu_ih_ring *ih, uint32_t ih_rb_cntl
 static int vega10_ih_irq_init(struct amdgpu_device *adev)
 {
 	struct amdgpu_ih_ring *ih;
+	u32 ih_rb_cntl;
 	int ret = 0;
-	u32 ih_rb_cntl, ih_doorbell_rtpr;
 	u32 tmp;
 
 	/* disable irqs */
@@ -177,23 +196,11 @@ static int vega10_ih_irq_init(struct amdgpu_device *adev)
 		     upper_32_bits(ih->wptr_addr) & 0xFFFF);
 
 	/* set rptr, wptr to 0 */
-	WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR, 0);
 	WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR, 0);
+	WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR, 0);
 
-	ih_doorbell_rtpr = RREG32_SOC15(OSSSYS, 0, mmIH_DOORBELL_RPTR);
-	if (adev->irq.ih.use_doorbell) {
-		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
-						 IH_DOORBELL_RPTR, OFFSET,
-						 adev->irq.ih.doorbell_index);
-		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
-						 IH_DOORBELL_RPTR,
-						 ENABLE, 1);
-	} else {
-		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
-						 IH_DOORBELL_RPTR,
-						 ENABLE, 0);
-	}
-	WREG32_SOC15(OSSSYS, 0, mmIH_DOORBELL_RPTR, ih_doorbell_rtpr);
+	WREG32_SOC15(OSSSYS, 0, mmIH_DOORBELL_RPTR,
+		     vega10_ih_doorbell_rptr(ih));
 
 	ih = &adev->irq.ih1;
 	if (ih->ring_size) {
@@ -203,11 +210,18 @@ static int vega10_ih_irq_init(struct amdgpu_device *adev)
 
 		ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING1);
 		ih_rb_cntl = vega10_ih_rb_cntl(ih, ih_rb_cntl);
+		ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+					   WPTR_OVERFLOW_ENABLE, 0);
+		ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+					   RB_FULL_DRAIN_ENABLE, 1);
 		WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING1, ih_rb_cntl);
 
 		/* set rptr, wptr to 0 */
-		WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR_RING1, 0);
 		WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_RING1, 0);
+		WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR_RING1, 0);
+
+		WREG32_SOC15(OSSSYS, 0, mmIH_DOORBELL_RPTR_RING1,
+			     vega10_ih_doorbell_rptr(ih));
 	}
 
 	ih = &adev->irq.ih2;
@@ -216,13 +230,16 @@ static int vega10_ih_irq_init(struct amdgpu_device *adev)
 		WREG32_SOC15(OSSSYS, 0, mmIH_RB_BASE_HI_RING2,
 			     (ih->gpu_addr >> 40) & 0xff);
 
-		ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING1);
+		ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING2);
 		ih_rb_cntl = vega10_ih_rb_cntl(ih, ih_rb_cntl);
 		WREG32_SOC15(OSSSYS, 0, mmIH_RB_CNTL_RING2, ih_rb_cntl);
 
 		/* set rptr, wptr to 0 */
-		WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR_RING2, 0);
 		WREG32_SOC15(OSSSYS, 0, mmIH_RB_WPTR_RING2, 0);
+		WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR_RING2, 0);
+
+		WREG32_SOC15(OSSSYS, 0, mmIH_DOORBELL_RPTR_RING2,
+			     vega10_ih_doorbell_rptr(ih));
 	}
 
 	tmp = RREG32_SOC15(OSSSYS, 0, mmIH_STORM_CLIENT_LIST_CNTL);
@@ -449,20 +466,23 @@ static int vega10_ih_sw_init(void *handle)
 	if (r)
 		return r;
 
-	if (adev->asic_type == CHIP_VEGA10) {
-		r = amdgpu_ih_ring_init(adev, &adev->irq.ih1, PAGE_SIZE, true);
-		if (r)
-			return r;
-
-		r = amdgpu_ih_ring_init(adev, &adev->irq.ih2, PAGE_SIZE, true);
-		if (r)
-			return r;
-	}
-
-	/* TODO add doorbell for IH1 & IH2 as well */
 	adev->irq.ih.use_doorbell = true;
 	adev->irq.ih.doorbell_index = adev->doorbell_index.ih << 1;
 
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih1, PAGE_SIZE, true);
+	if (r)
+		return r;
+
+	adev->irq.ih1.use_doorbell = true;
+	adev->irq.ih1.doorbell_index = (adev->doorbell_index.ih + 1) << 1;
+
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih2, PAGE_SIZE, true);
+	if (r)
+		return r;
+
+	adev->irq.ih2.use_doorbell = true;
+	adev->irq.ih2.doorbell_index = (adev->doorbell_index.ih + 2) << 1;
+
 	r = amdgpu_irq_init(adev);
 
 	return r;

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

@@ -466,6 +466,8 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd,
 	memset(&kfd->doorbell_available_index, 0,
 		sizeof(kfd->doorbell_available_index));
 
+	atomic_set(&kfd->sram_ecc_flag, 0);
+
 	return kfd;
 }
 
@@ -661,6 +663,9 @@ int kgd2kfd_post_reset(struct kfd_dev *kfd)
 		return ret;
 	count = atomic_dec_return(&kfd_locked);
 	WARN_ONCE(count != 0, "KFD reset ref. error");
+
+	atomic_set(&kfd->sram_ecc_flag, 0);
+
 	return 0;
 }
 
@@ -1024,6 +1029,12 @@ int kfd_gtt_sa_free(struct kfd_dev *kfd, struct kfd_mem_obj *mem_obj)
 	return 0;
 }
 
+void kgd2kfd_set_sram_ecc_flag(struct kfd_dev *kfd)
+{
+	if (kfd)
+		atomic_inc(&kfd->sram_ecc_flag);
+}
+
 #if defined(CONFIG_DEBUG_FS)
 
 /* This function will send a package to HIQ to hang the HWS

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

@@ -1011,25 +1011,41 @@ void kfd_signal_vm_fault_event(struct kfd_dev *dev, unsigned int pasid,
 void kfd_signal_reset_event(struct kfd_dev *dev)
 {
 	struct kfd_hsa_hw_exception_data hw_exception_data;
+	struct kfd_hsa_memory_exception_data memory_exception_data;
 	struct kfd_process *p;
 	struct kfd_event *ev;
 	unsigned int temp;
 	uint32_t id, idx;
+	int reset_cause = atomic_read(&dev->sram_ecc_flag) ?
+			KFD_HW_EXCEPTION_ECC :
+			KFD_HW_EXCEPTION_GPU_HANG;
 
 	/* Whole gpu reset caused by GPU hang and memory is lost */
 	memset(&hw_exception_data, 0, sizeof(hw_exception_data));
 	hw_exception_data.gpu_id = dev->id;
 	hw_exception_data.memory_lost = 1;
+	hw_exception_data.reset_cause = reset_cause;
+
+	memset(&memory_exception_data, 0, sizeof(memory_exception_data));
+	memory_exception_data.ErrorType = KFD_MEM_ERR_SRAM_ECC;
+	memory_exception_data.gpu_id = dev->id;
+	memory_exception_data.failure.imprecise = true;
 
 	idx = srcu_read_lock(&kfd_processes_srcu);
 	hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {
 		mutex_lock(&p->event_mutex);
 		id = KFD_FIRST_NONSIGNAL_EVENT_ID;
-		idr_for_each_entry_continue(&p->event_idr, ev, id)
+		idr_for_each_entry_continue(&p->event_idr, ev, id) {
 			if (ev->type == KFD_EVENT_TYPE_HW_EXCEPTION) {
 				ev->hw_exception_data = hw_exception_data;
 				set_event(ev);
 			}
+			if (ev->type == KFD_EVENT_TYPE_MEMORY &&
+			    reset_cause == KFD_HW_EXCEPTION_ECC) {
+				ev->memory_exception_data = memory_exception_data;
+				set_event(ev);
+			}
+		}
 		mutex_unlock(&p->event_mutex);
 	}
 	srcu_read_unlock(&kfd_processes_srcu, idx);

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

@@ -276,6 +276,9 @@ struct kfd_dev {
 	uint64_t hive_id;
 
 	bool pci_atomic_requested;
+
+	/* SRAM ECC flag */
+	atomic_t sram_ecc_flag;
 };
 
 enum kfd_mempool {

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

@@ -37,6 +37,7 @@
 #include "kfd_device_queue_manager.h"
 #include "kfd_iommu.h"
 #include "amdgpu_amdkfd.h"
+#include "amdgpu_ras.h"
 
 /* topology_device_list - Master list of all topology devices */
 static struct list_head topology_device_list;
@@ -1197,6 +1198,7 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
 	void *crat_image = NULL;
 	size_t image_size = 0;
 	int proximity_domain;
+	struct amdgpu_ras *ctx;
 
 	INIT_LIST_HEAD(&temp_topology_device_list);
 
@@ -1328,6 +1330,20 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
 		dev->node_props.capability |= HSA_CAP_ATS_PRESENT;
 	}
 
+	ctx = amdgpu_ras_get_context((struct amdgpu_device *)(dev->gpu->kgd));
+	if (ctx) {
+		/* kfd only concerns sram ecc on GFX/SDMA and HBM ecc on UMC */
+		dev->node_props.capability |=
+			(((ctx->features & BIT(AMDGPU_RAS_BLOCK__SDMA)) != 0) ||
+			 ((ctx->features & BIT(AMDGPU_RAS_BLOCK__GFX)) != 0)) ?
+			HSA_CAP_SRAM_EDCSUPPORTED : 0;
+		dev->node_props.capability |= ((ctx->features & BIT(AMDGPU_RAS_BLOCK__UMC)) != 0) ?
+			HSA_CAP_MEM_EDCSUPPORTED : 0;
+
+		dev->node_props.capability |= (ctx->features != 0) ?
+			HSA_CAP_RASEVENTNOTIFY : 0;
+	}
+
 	kfd_debug_print_topology();
 
 	if (!res)

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

@@ -48,6 +48,10 @@
 #define HSA_CAP_DOORBELL_TYPE_2_0		0x2
 #define HSA_CAP_AQL_QUEUE_DOUBLE_MAP		0x00004000
 
+#define HSA_CAP_SRAM_EDCSUPPORTED		0x00080000
+#define HSA_CAP_MEM_EDCSUPPORTED		0x00100000
+#define HSA_CAP_RASEVENTNOTIFY			0x00200000
+
 struct kfd_node_properties {
 	uint64_t hive_id;
 	uint32_t cpu_cores_count;

drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h

@@ -132,8 +132,6 @@ struct amdgpu_display_manager {
 	 */
 	struct drm_private_obj atomic_obj;
 
-	struct drm_modeset_lock atomic_obj_lock;
-
 	/**
 	 * @dc_lock:
 	 *
@@ -240,6 +238,10 @@ struct amdgpu_dm_connector {
 	struct mutex hpd_lock;
 
 	bool fake_enable;
+#ifdef CONFIG_DEBUG_FS
+	uint32_t debugfs_dpcd_address;
+	uint32_t debugfs_dpcd_size;
+#endif
 };
 
 #define to_amdgpu_dm_connector(x) container_of(x, struct amdgpu_dm_connector, base)

drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_color.c

@@ -126,46 +126,51 @@ int amdgpu_dm_set_regamma_lut(struct dm_crtc_state *crtc)
 		crtc->base.state->dev->dev_private;
 	struct drm_color_lut *lut;
 	uint32_t lut_size;
-	struct dc_gamma *gamma;
+	struct dc_gamma *gamma = NULL;
 	enum dc_transfer_func_type old_type = stream->out_transfer_func->type;
 
 	bool ret;
 
-	if (!blob) {
+	if (!blob && adev->asic_type <= CHIP_RAVEN) {
 		/* By default, use the SRGB predefined curve.*/
 		stream->out_transfer_func->type = TF_TYPE_PREDEFINED;
 		stream->out_transfer_func->tf = TRANSFER_FUNCTION_SRGB;
 		return 0;
 	}
 
-	lut = (struct drm_color_lut *)blob->data;
-	lut_size = blob->length / sizeof(struct drm_color_lut);
+	if (blob) {
+		lut = (struct drm_color_lut *)blob->data;
+		lut_size = blob->length / sizeof(struct drm_color_lut);
 
-	gamma = dc_create_gamma();
-	if (!gamma)
-		return -ENOMEM;
+		gamma = dc_create_gamma();
+		if (!gamma)
+			return -ENOMEM;
 
-	gamma->num_entries = lut_size;
-	if (gamma->num_entries == MAX_COLOR_LEGACY_LUT_ENTRIES)
-		gamma->type = GAMMA_RGB_256;
-	else if (gamma->num_entries == MAX_COLOR_LUT_ENTRIES)
-		gamma->type = GAMMA_CS_TFM_1D;
-	else {
-		/* Invalid lut size */
-		dc_gamma_release(&gamma);
-		return -EINVAL;
+		gamma->num_entries = lut_size;
+		if (gamma->num_entries == MAX_COLOR_LEGACY_LUT_ENTRIES)
+			gamma->type = GAMMA_RGB_256;
+		else if (gamma->num_entries == MAX_COLOR_LUT_ENTRIES)
+			gamma->type = GAMMA_CS_TFM_1D;
+		else {
+			/* Invalid lut size */
+			dc_gamma_release(&gamma);
+			return -EINVAL;
+		}
+
+		/* Convert drm_lut into dc_gamma */
+		__drm_lut_to_dc_gamma(lut, gamma, gamma->type == GAMMA_RGB_256);
 	}
 
-	/* Convert drm_lut into dc_gamma */
-	__drm_lut_to_dc_gamma(lut, gamma, gamma->type == GAMMA_RGB_256);
-
-	/* Call color module to translate into something DC understands. Namely
-	 * a transfer function.
+	/* predefined gamma ROM only exist for RAVEN and pre-RAVEN ASIC,
+	 * set canRomBeUsed accordingly
 	 */
 	stream->out_transfer_func->type = TF_TYPE_DISTRIBUTED_POINTS;
 	ret = mod_color_calculate_regamma_params(stream->out_transfer_func,
-						 gamma, true, adev->asic_type <= CHIP_RAVEN, NULL);
-	dc_gamma_release(&gamma);
+			gamma, true, adev->asic_type <= CHIP_RAVEN, NULL);
+
+	if (gamma)
+		dc_gamma_release(&gamma);
+
 	if (!ret) {
 		stream->out_transfer_func->type = old_type;
 		DRM_ERROR("Out of memory when calculating regamma params\n");

drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_debugfs.c

@@ -29,6 +29,7 @@
 #include "amdgpu.h"
 #include "amdgpu_dm.h"
 #include "amdgpu_dm_debugfs.h"
+#include "dm_helpers.h"
 
 /* function description
  * get/ set DP configuration: lane_count, link_rate, spread_spectrum
@@ -688,8 +689,131 @@ static int vrr_range_show(struct seq_file *m, void *data)
 
 	return 0;
 }
+
+/* function description
+ *
+ * generic SDP message access for testing
+ *
+ * debugfs sdp_message is located at /syskernel/debug/dri/0/DP-x
+ *
+ * SDP header
+ * Hb0 : Secondary-Data Packet ID
+ * Hb1 : Secondary-Data Packet type
+ * Hb2 : Secondary-Data-packet-specific header, Byte 0
+ * Hb3 : Secondary-Data-packet-specific header, Byte 1
+ *
+ * for using custom sdp message: input 4 bytes SDP header and 32 bytes raw data
+ */
+static ssize_t dp_sdp_message_debugfs_write(struct file *f, const char __user *buf,
+				 size_t size, loff_t *pos)
+{
+	int r;
+	uint8_t data[36];
+	struct amdgpu_dm_connector *connector = file_inode(f)->i_private;
+	struct dm_crtc_state *acrtc_state;
+	uint32_t write_size = 36;
+
+	if (connector->base.status != connector_status_connected)
+		return -ENODEV;
+
+	if (size == 0)
+		return 0;
+
+	acrtc_state = to_dm_crtc_state(connector->base.state->crtc->state);
+
+	r = copy_from_user(data, buf, write_size);
+
+	write_size -= r;
+
+	dc_stream_send_dp_sdp(acrtc_state->stream, data, write_size);
+
+	return write_size;
+}
+
 DEFINE_SHOW_ATTRIBUTE(vrr_range);
 
+static ssize_t dp_dpcd_address_write(struct file *f, const char __user *buf,
+				 size_t size, loff_t *pos)
+{
+	int r;
+	struct amdgpu_dm_connector *connector = file_inode(f)->i_private;
+
+	if (size < sizeof(connector->debugfs_dpcd_address))
+		return 0;
+
+	r = copy_from_user(&connector->debugfs_dpcd_address,
+			buf, sizeof(connector->debugfs_dpcd_address));
+
+	return size - r;
+}
+
+static ssize_t dp_dpcd_size_write(struct file *f, const char __user *buf,
+				 size_t size, loff_t *pos)
+{
+	int r;
+	struct amdgpu_dm_connector *connector = file_inode(f)->i_private;
+
+	if (size < sizeof(connector->debugfs_dpcd_size))
+		return 0;
+
+	r = copy_from_user(&connector->debugfs_dpcd_size,
+			buf, sizeof(connector->debugfs_dpcd_size));
+
+	if (connector->debugfs_dpcd_size > 256)
+		connector->debugfs_dpcd_size = 0;
+
+	return size - r;
+}
+
+static ssize_t dp_dpcd_data_write(struct file *f, const char __user *buf,
+				 size_t size, loff_t *pos)
+{
+	int r;
+	char *data;
+	struct amdgpu_dm_connector *connector = file_inode(f)->i_private;
+	struct dc_link *link = connector->dc_link;
+	uint32_t write_size = connector->debugfs_dpcd_size;
+
+	if (size < write_size)
+		return 0;
+
+	data = kzalloc(write_size, GFP_KERNEL);
+	if (!data)
+		return 0;
+
+	r = copy_from_user(data, buf, write_size);
+
+	dm_helpers_dp_write_dpcd(link->ctx, link,
+			connector->debugfs_dpcd_address, data, write_size - r);
+	kfree(data);
+	return write_size - r;
+}
+
+static ssize_t dp_dpcd_data_read(struct file *f, char __user *buf,
+				 size_t size, loff_t *pos)
+{
+	int r;
+	char *data;
+	struct amdgpu_dm_connector *connector = file_inode(f)->i_private;
+	struct dc_link *link = connector->dc_link;
+	uint32_t read_size = connector->debugfs_dpcd_size;
+
+	if (size < read_size)
+		return 0;
+
+	data = kzalloc(read_size, GFP_KERNEL);
+	if (!data)
+		return 0;
+
+	dm_helpers_dp_read_dpcd(link->ctx, link,
+			connector->debugfs_dpcd_address, data, read_size);
+
+	r = copy_to_user(buf, data, read_size);
+
+	kfree(data);
+	return read_size - r;
+}
+
 static const struct file_operations dp_link_settings_debugfs_fops = {
 	.owner = THIS_MODULE,
 	.read = dp_link_settings_read,
@@ -710,6 +834,31 @@ static const struct file_operations dp_phy_test_pattern_fops = {
 	.llseek = default_llseek
 };
 
+static const struct file_operations sdp_message_fops = {
+	.owner = THIS_MODULE,
+	.write = dp_sdp_message_debugfs_write,
+	.llseek = default_llseek
+};
+
+static const struct file_operations dp_dpcd_address_debugfs_fops = {
+	.owner = THIS_MODULE,
+	.write = dp_dpcd_address_write,
+	.llseek = default_llseek
+};
+
+static const struct file_operations dp_dpcd_size_debugfs_fops = {
+	.owner = THIS_MODULE,
+	.write = dp_dpcd_size_write,
+	.llseek = default_llseek
+};
+
+static const struct file_operations dp_dpcd_data_debugfs_fops = {
+	.owner = THIS_MODULE,
+	.read = dp_dpcd_data_read,
+	.write = dp_dpcd_data_write,
+	.llseek = default_llseek
+};
+
 static const struct {
 	char *name;
 	const struct file_operations *fops;
@@ -717,7 +866,11 @@ static const struct {
 		{"link_settings", &dp_link_settings_debugfs_fops},
 		{"phy_settings", &dp_phy_settings_debugfs_fop},
 		{"test_pattern", &dp_phy_test_pattern_fops},
-		{"vrr_range", &vrr_range_fops}
+		{"vrr_range", &vrr_range_fops},
+		{"sdp_message", &sdp_message_fops},
+		{"aux_dpcd_address", &dp_dpcd_address_debugfs_fops},
+		{"aux_dpcd_size", &dp_dpcd_size_debugfs_fops},
+		{"aux_dpcd_data", &dp_dpcd_data_debugfs_fops}
 };
 
 int connector_debugfs_init(struct amdgpu_dm_connector *connector)

drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c

@@ -264,7 +264,7 @@ bool dm_helpers_dp_mst_write_payload_allocation_table(
 }
 
 /*
- * poll pending down reply before clear payload allocation table
+ * poll pending down reply
  */
 void dm_helpers_dp_mst_poll_pending_down_reply(
 	struct dc_context *ctx,

drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c

@@ -84,6 +84,7 @@ static ssize_t dm_dp_aux_transfer(struct drm_dp_aux *aux,
 {
 	ssize_t result = 0;
 	struct aux_payload payload;
+	enum aux_channel_operation_result operation_result;
 
 	if (WARN_ON(msg->size > 16))
 		return -E2BIG;
@@ -97,13 +98,27 @@ static ssize_t dm_dp_aux_transfer(struct drm_dp_aux *aux,
 	payload.mot = (msg->request & DP_AUX_I2C_MOT) != 0;
 	payload.defer_delay = 0;
 
-	result = dc_link_aux_transfer(TO_DM_AUX(aux)->ddc_service, &payload);
+	result = dc_link_aux_transfer_raw(TO_DM_AUX(aux)->ddc_service, &payload,
+				      &operation_result);
 
 	if (payload.write)
 		result = msg->size;
 
-	if (result < 0) /* DC doesn't know about kernel error codes */
-		result = -EIO;
+	if (result < 0)
+		switch (operation_result) {
+		case AUX_CHANNEL_OPERATION_SUCCEEDED:
+			break;
+		case AUX_CHANNEL_OPERATION_FAILED_HPD_DISCON:
+		case AUX_CHANNEL_OPERATION_FAILED_REASON_UNKNOWN:
+			result = -EIO;
+			break;
+		case AUX_CHANNEL_OPERATION_FAILED_INVALID_REPLY:
+			result = -EBUSY;
+			break;
+		case AUX_CHANNEL_OPERATION_FAILED_TIMEOUT:
+			result = -ETIMEDOUT;
+			break;
+		}
 
 	return result;
 }

drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_pp_smu.c

@@ -33,6 +33,7 @@
 #include "amdgpu_dm_irq.h"
 #include "amdgpu_pm.h"
 #include "dm_pp_smu.h"
+#include "amdgpu_smu.h"
 
 
 bool dm_pp_apply_display_requirements(
@@ -40,6 +41,7 @@ bool dm_pp_apply_display_requirements(
 		const struct dm_pp_display_configuration *pp_display_cfg)
 {
 	struct amdgpu_device *adev = ctx->driver_context;
+	struct smu_context *smu = &adev->smu;
 	int i;
 
 	if (adev->pm.dpm_enabled) {
@@ -105,6 +107,9 @@ bool dm_pp_apply_display_requirements(
 			adev->powerplay.pp_funcs->display_configuration_change(
 				adev->powerplay.pp_handle,
 				&adev->pm.pm_display_cfg);
+		else
+			smu_display_configuration_change(smu,
+							 &adev->pm.pm_display_cfg);
 
 		amdgpu_pm_compute_clocks(adev);
 	}
@@ -308,6 +313,12 @@ bool dm_pp_get_clock_levels_by_type(
 		if (adev->powerplay.pp_funcs->get_clock_by_type(pp_handle,
 			dc_to_pp_clock_type(clk_type), &pp_clks)) {
 		/* Error in pplib. Provide default values. */
+			return true;
+		}
+	} else if (adev->smu.funcs && adev->smu.funcs->get_clock_by_type) {
+		if (smu_get_clock_by_type(&adev->smu,
+					  dc_to_pp_clock_type(clk_type),
+					  &pp_clks)) {
 			get_default_clock_levels(clk_type, dc_clks);
 			return true;
 		}
@@ -324,6 +335,13 @@ bool dm_pp_get_clock_levels_by_type(
 			validation_clks.memory_max_clock = 80000;
 			validation_clks.level = 0;
 		}
+	} else if (adev->smu.funcs && adev->smu.funcs->get_max_high_clocks) {
+		if (smu_get_max_high_clocks(&adev->smu, &validation_clks)) {
+			DRM_INFO("DM_PPLIB: Warning: using default validation clocks!\n");
+			validation_clks.engine_max_clock = 72000;
+			validation_clks.memory_max_clock = 80000;
+			validation_clks.level = 0;
+		}
 	}
 
 	DRM_INFO("DM_PPLIB: Validation clocks:\n");
@@ -374,14 +392,21 @@ bool dm_pp_get_clock_levels_by_type_with_latency(
 	void *pp_handle = adev->powerplay.pp_handle;
 	struct pp_clock_levels_with_latency pp_clks = { 0 };
 	const struct amd_pm_funcs *pp_funcs = adev->powerplay.pp_funcs;
+	int ret;
 
-	if (!pp_funcs || !pp_funcs->get_clock_by_type_with_latency)
-		return false;
+	if (pp_funcs && pp_funcs->get_clock_by_type_with_latency) {
+		ret = pp_funcs->get_clock_by_type_with_latency(pp_handle,
+						dc_to_pp_clock_type(clk_type),
+						&pp_clks);
+		if (ret)
+			return false;
+	} else if (adev->smu.ppt_funcs && adev->smu.ppt_funcs->get_clock_by_type_with_latency) {
+		if (smu_get_clock_by_type_with_latency(&adev->smu,
+						       dc_to_pp_clock_type(clk_type),
+						       &pp_clks))
+			return false;
+	}
 
-	if (pp_funcs->get_clock_by_type_with_latency(pp_handle,
-						     dc_to_pp_clock_type(clk_type),
-						     &pp_clks))
-		return false;
 
 	pp_to_dc_clock_levels_with_latency(&pp_clks, clk_level_info, clk_type);
 
@@ -397,14 +422,20 @@ bool dm_pp_get_clock_levels_by_type_with_voltage(
 	void *pp_handle = adev->powerplay.pp_handle;
 	struct pp_clock_levels_with_voltage pp_clk_info = {0};
 	const struct amd_pm_funcs *pp_funcs = adev->powerplay.pp_funcs;
+	int ret;
 
-	if (!pp_funcs || !pp_funcs->get_clock_by_type_with_voltage)
-		return false;
-
-	if (pp_funcs->get_clock_by_type_with_voltage(pp_handle,
-						     dc_to_pp_clock_type(clk_type),
-						     &pp_clk_info))
-		return false;
+	if (pp_funcs && pp_funcs->get_clock_by_type_with_voltage) {
+		ret = pp_funcs->get_clock_by_type_with_voltage(pp_handle,
+						dc_to_pp_clock_type(clk_type),
+						&pp_clk_info);
+		if (ret)
+			return false;
+	} else if (adev->smu.ppt_funcs && adev->smu.ppt_funcs->get_clock_by_type_with_voltage) {
+		if (smu_get_clock_by_type_with_voltage(&adev->smu,
+						       dc_to_pp_clock_type(clk_type),
+						       &pp_clk_info))
+			return false;
+	}
 
 	pp_to_dc_clock_levels_with_voltage(&pp_clk_info, clk_level_info, clk_type);
 
@@ -445,6 +476,10 @@ bool dm_pp_apply_clock_for_voltage_request(
 		ret = adev->powerplay.pp_funcs->display_clock_voltage_request(
 			adev->powerplay.pp_handle,
 			&pp_clock_request);
+	else if (adev->smu.funcs &&
+		 adev->smu.funcs->display_clock_voltage_request)
+		ret = smu_display_clock_voltage_request(&adev->smu,
+							&pp_clock_request);
 	if (ret)
 		return false;
 	return true;
@@ -462,6 +497,8 @@ bool dm_pp_get_static_clocks(
 		ret = adev->powerplay.pp_funcs->get_current_clocks(
 			adev->powerplay.pp_handle,
 			&pp_clk_info);
+	else if (adev->smu.funcs)
+		ret = smu_get_current_clocks(&adev->smu, &pp_clk_info);
 	if (ret)
 		return false;
 
@@ -472,27 +509,6 @@ bool dm_pp_get_static_clocks(
 	return true;
 }
 
-void pp_rv_set_display_requirement(struct pp_smu *pp,
-		struct pp_smu_display_requirement_rv *req)
-{
-	const struct dc_context *ctx = pp->dm;
-	struct amdgpu_device *adev = ctx->driver_context;
-	void *pp_handle = adev->powerplay.pp_handle;
-	const struct amd_pm_funcs *pp_funcs = adev->powerplay.pp_funcs;
-	struct pp_display_clock_request clock = {0};
-
-	if (!pp_funcs || !pp_funcs->display_clock_voltage_request)
-		return;
-
-	clock.clock_type = amd_pp_dcf_clock;
-	clock.clock_freq_in_khz = req->hard_min_dcefclk_mhz * 1000;
-	pp_funcs->display_clock_voltage_request(pp_handle, &clock);
-
-	clock.clock_type = amd_pp_f_clock;
-	clock.clock_freq_in_khz = req->hard_min_fclk_mhz * 1000;
-	pp_funcs->display_clock_voltage_request(pp_handle, &clock);
-}
-
 void pp_rv_set_wm_ranges(struct pp_smu *pp,
 		struct pp_smu_wm_range_sets *ranges)
 {
@@ -508,9 +524,6 @@ void pp_rv_set_wm_ranges(struct pp_smu *pp,
 	wm_with_clock_ranges.num_wm_dmif_sets = ranges->num_reader_wm_sets;
 	wm_with_clock_ranges.num_wm_mcif_sets = ranges->num_writer_wm_sets;
 
-	if (!pp_funcs || !pp_funcs->set_watermarks_for_clocks_ranges)
-		return;
-
 	for (i = 0; i < wm_with_clock_ranges.num_wm_dmif_sets; i++) {
 		if (ranges->reader_wm_sets[i].wm_inst > 3)
 			wm_dce_clocks[i].wm_set_id = WM_SET_A;
@@ -543,7 +556,13 @@ void pp_rv_set_wm_ranges(struct pp_smu *pp,
 				ranges->writer_wm_sets[i].min_drain_clk_mhz * 1000;
 	}
 
-	pp_funcs->set_watermarks_for_clocks_ranges(pp_handle, &wm_with_clock_ranges);
+	if (pp_funcs && pp_funcs->set_watermarks_for_clocks_ranges)
+		pp_funcs->set_watermarks_for_clocks_ranges(pp_handle,
+							   &wm_with_clock_ranges);
+	else if (adev->smu.funcs &&
+		 adev->smu.funcs->set_watermarks_for_clock_ranges)
+		smu_set_watermarks_for_clock_ranges(&adev->smu,
+						    &wm_with_clock_ranges);
 }
 
 void pp_rv_set_pme_wa_enable(struct pp_smu *pp)
@@ -553,10 +572,10 @@ void pp_rv_set_pme_wa_enable(struct pp_smu *pp)
 	void *pp_handle = adev->powerplay.pp_handle;
 	const struct amd_pm_funcs *pp_funcs = adev->powerplay.pp_funcs;
 
-	if (!pp_funcs || !pp_funcs->notify_smu_enable_pwe)
-		return;
-
-	pp_funcs->notify_smu_enable_pwe(pp_handle);
+	if (pp_funcs && pp_funcs->notify_smu_enable_pwe)
+		pp_funcs->notify_smu_enable_pwe(pp_handle);
+	else if (adev->smu.funcs)
+		smu_notify_smu_enable_pwe(&adev->smu);
 }
 
 void pp_rv_set_active_display_count(struct pp_smu *pp, int count)
@@ -611,17 +630,16 @@ void pp_rv_set_hard_min_fclk_by_freq(struct pp_smu *pp, int mhz)
 	pp_funcs->set_hard_min_fclk_by_freq(pp_handle, mhz);
 }
 
-void dm_pp_get_funcs_rv(
+void dm_pp_get_funcs(
 		struct dc_context *ctx,
-		struct pp_smu_funcs_rv *funcs)
+		struct pp_smu_funcs *funcs)
 {
-	funcs->pp_smu.dm = ctx;
-	funcs->set_display_requirement = pp_rv_set_display_requirement;
-	funcs->set_wm_ranges = pp_rv_set_wm_ranges;
-	funcs->set_pme_wa_enable = pp_rv_set_pme_wa_enable;
-	funcs->set_display_count = pp_rv_set_active_display_count;
-	funcs->set_min_deep_sleep_dcfclk = pp_rv_set_min_deep_sleep_dcfclk;
-	funcs->set_hard_min_dcfclk_by_freq = pp_rv_set_hard_min_dcefclk_by_freq;
-	funcs->set_hard_min_fclk_by_freq = pp_rv_set_hard_min_fclk_by_freq;
+	funcs->rv_funcs.pp_smu.dm = ctx;
+	funcs->rv_funcs.set_wm_ranges = pp_rv_set_wm_ranges;
+	funcs->rv_funcs.set_pme_wa_enable = pp_rv_set_pme_wa_enable;
+	funcs->rv_funcs.set_display_count = pp_rv_set_active_display_count;
+	funcs->rv_funcs.set_min_deep_sleep_dcfclk = pp_rv_set_min_deep_sleep_dcfclk;
+	funcs->rv_funcs.set_hard_min_dcfclk_by_freq = pp_rv_set_hard_min_dcefclk_by_freq;
+	funcs->rv_funcs.set_hard_min_fclk_by_freq = pp_rv_set_hard_min_fclk_by_freq;
 }
 

drivers/gpu/drm/amd/display/dc/basics/fixpt31_32.c

@@ -449,6 +449,11 @@ static inline unsigned int clamp_ux_dy(
 		return min_clamp;
 }
 
+unsigned int dc_fixpt_u4d19(struct fixed31_32 arg)
+{
+	return ux_dy(arg.value, 4, 19);
+}
+
 unsigned int dc_fixpt_u3d19(struct fixed31_32 arg)
 {
 	return ux_dy(arg.value, 3, 19);

drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c

@@ -247,6 +247,53 @@ static enum dcn_bw_defs tl_pixel_format_to_bw_defs(enum surface_pixel_format for
 	}
 }
 
+enum source_macro_tile_size swizzle_mode_to_macro_tile_size(enum swizzle_mode_values sw_mode)
+{
+	switch (sw_mode) {
+	/* for 4/8/16 high tiles */
+	case DC_SW_LINEAR:
+		return dm_4k_tile;
+	case DC_SW_4KB_S:
+	case DC_SW_4KB_S_X:
+		return dm_4k_tile;
+	case DC_SW_64KB_S:
+	case DC_SW_64KB_S_X:
+	case DC_SW_64KB_S_T:
+		return dm_64k_tile;
+	case DC_SW_VAR_S:
+	case DC_SW_VAR_S_X:
+		return dm_256k_tile;
+
+	/* For 64bpp 2 high tiles */
+	case DC_SW_4KB_D:
+	case DC_SW_4KB_D_X:
+		return dm_4k_tile;
+	case DC_SW_64KB_D:
+	case DC_SW_64KB_D_X:
+	case DC_SW_64KB_D_T:
+		return dm_64k_tile;
+	case DC_SW_VAR_D:
+	case DC_SW_VAR_D_X:
+		return dm_256k_tile;
+
+	case DC_SW_4KB_R:
+	case DC_SW_4KB_R_X:
+		return dm_4k_tile;
+	case DC_SW_64KB_R:
+	case DC_SW_64KB_R_X:
+		return dm_64k_tile;
+	case DC_SW_VAR_R:
+	case DC_SW_VAR_R_X:
+		return dm_256k_tile;
+
+	/* Unsupported swizzle modes for dcn */
+	case DC_SW_256B_S:
+	default:
+		ASSERT(0); /* Not supported */
+		return 0;
+	}
+}
+
 static void pipe_ctx_to_e2e_pipe_params (
 		const struct pipe_ctx *pipe,
 		struct _vcs_dpi_display_pipe_params_st *input)
@@ -287,46 +334,7 @@ static void pipe_ctx_to_e2e_pipe_params (
 	input->src.cur0_src_width      = 128; /* TODO: Cursor calcs, not curently stored */
 	input->src.cur0_bpp            = 32;
 
-	switch (pipe->plane_state->tiling_info.gfx9.swizzle) {
-	/* for 4/8/16 high tiles */
-	case DC_SW_LINEAR:
-		input->src.macro_tile_size = dm_4k_tile;
-		break;
-	case DC_SW_4KB_S:
-	case DC_SW_4KB_S_X:
-		input->src.macro_tile_size = dm_4k_tile;
-		break;
-	case DC_SW_64KB_S:
-	case DC_SW_64KB_S_X:
-	case DC_SW_64KB_S_T:
-		input->src.macro_tile_size = dm_64k_tile;
-		break;
-	case DC_SW_VAR_S:
-	case DC_SW_VAR_S_X:
-		input->src.macro_tile_size = dm_256k_tile;
-		break;
-
-	/* For 64bpp 2 high tiles */
-	case DC_SW_4KB_D:
-	case DC_SW_4KB_D_X:
-		input->src.macro_tile_size = dm_4k_tile;
-		break;
-	case DC_SW_64KB_D:
-	case DC_SW_64KB_D_X:
-	case DC_SW_64KB_D_T:
-		input->src.macro_tile_size = dm_64k_tile;
-		break;
-	case DC_SW_VAR_D:
-	case DC_SW_VAR_D_X:
-		input->src.macro_tile_size = dm_256k_tile;
-		break;
-
-	/* Unsupported swizzle modes for dcn */
-	case DC_SW_256B_S:
-	default:
-		ASSERT(0); /* Not supported */
-		break;
-	}
+	input->src.macro_tile_size = swizzle_mode_to_macro_tile_size(pipe->plane_state->tiling_info.gfx9.swizzle);
 
 	switch (pipe->plane_state->rotation) {
 	case ROTATION_ANGLE_0:
@@ -466,7 +474,7 @@ static void dcn_bw_calc_rq_dlg_ttu(
 	input.clks_cfg.dcfclk_mhz = v->dcfclk;
 	input.clks_cfg.dispclk_mhz = v->dispclk;
 	input.clks_cfg.dppclk_mhz = v->dppclk;
-	input.clks_cfg.refclk_mhz = dc->res_pool->ref_clock_inKhz / 1000.0;
+	input.clks_cfg.refclk_mhz = dc->res_pool->ref_clocks.dchub_ref_clock_inKhz / 1000.0;
 	input.clks_cfg.socclk_mhz = v->socclk;
 	input.clks_cfg.voltage = v->voltage_level;
 //	dc->dml.logger = pool->base.logger;
@@ -1141,7 +1149,7 @@ bool dcn_validate_bandwidth(
 						hsplit_pipe->pipe_dlg_param.vblank_end = pipe->pipe_dlg_param.vblank_end;
 					} else {
 						/* pipe not split previously needs split */
-						hsplit_pipe = find_idle_secondary_pipe(&context->res_ctx, pool);
+						hsplit_pipe = find_idle_secondary_pipe(&context->res_ctx, pool, pipe);
 						ASSERT(hsplit_pipe);
 						split_stream_across_pipes(
 							&context->res_ctx, pool,
@@ -1395,12 +1403,14 @@ void dcn_bw_update_from_pplib(struct dc *dc)
 
 void dcn_bw_notify_pplib_of_wm_ranges(struct dc *dc)
 {
-	struct pp_smu_funcs_rv *pp = dc->res_pool->pp_smu;
+	struct pp_smu_funcs_rv *pp = NULL;
 	struct pp_smu_wm_range_sets ranges = {0};
 	int min_fclk_khz, min_dcfclk_khz, socclk_khz;
 	const int overdrive = 5000000; /* 5 GHz to cover Overdrive */
 
-	if (!pp->set_wm_ranges)
+	if (dc->res_pool->pp_smu)
+		pp = &dc->res_pool->pp_smu->rv_funcs;
+	if (!pp || !pp->set_wm_ranges)
 		return;
 
 	kernel_fpu_begin();

drivers/gpu/drm/amd/display/dc/core/dc.c

@@ -524,6 +524,14 @@ void dc_link_set_preferred_link_settings(struct dc *dc,
 	struct dc_stream_state *link_stream;
 	struct dc_link_settings store_settings = *link_setting;
 
+	link->preferred_link_setting = store_settings;
+
+	/* Retrain with preferred link settings only relevant for
+	 * DP signal type
+	 */
+	if (!dc_is_dp_signal(link->connector_signal))
+		return;
+
 	for (i = 0; i < MAX_PIPES; i++) {
 		pipe = &dc->current_state->res_ctx.pipe_ctx[i];
 		if (pipe->stream && pipe->stream->link) {
@@ -538,7 +546,10 @@ void dc_link_set_preferred_link_settings(struct dc *dc,
 
 	link_stream = link->dc->current_state->res_ctx.pipe_ctx[i].stream;
 
-	link->preferred_link_setting = store_settings;
+	/* Cannot retrain link if backend is off */
+	if (link_stream->dpms_off)
+		return;
+
 	if (link_stream)
 		decide_link_settings(link_stream, &store_settings);
 
@@ -621,6 +632,8 @@ static bool construct(struct dc *dc,
 #endif
 
 	enum dce_version dc_version = DCE_VERSION_UNKNOWN;
+	memcpy(&dc->bb_overrides, &init_params->bb_overrides, sizeof(dc->bb_overrides));
+
 	dc_dceip = kzalloc(sizeof(*dc_dceip), GFP_KERNEL);
 	if (!dc_dceip) {
 		dm_error("%s: failed to create dceip\n", __func__);
@@ -722,11 +735,7 @@ static bool construct(struct dc *dc,
 		goto fail;
 	}
 
-	dc->res_pool = dc_create_resource_pool(
-			dc,
-			init_params->num_virtual_links,
-			dc_version,
-			init_params->asic_id);
+	dc->res_pool = dc_create_resource_pool(dc, init_params, dc_version);
 	if (!dc->res_pool)
 		goto fail;
 
@@ -969,7 +978,7 @@ static bool context_changed(
 	return false;
 }
 
-bool dc_validate_seamless_boot_timing(struct dc *dc,
+bool dc_validate_seamless_boot_timing(const struct dc *dc,
 				const struct dc_sink *sink,
 				struct dc_crtc_timing *crtc_timing)
 {
@@ -1060,7 +1069,13 @@ static enum dc_status dc_commit_state_no_check(struct dc *dc, struct dc_state *c
 	if (!dcb->funcs->is_accelerated_mode(dcb))
 		dc->hwss.enable_accelerated_mode(dc, context);
 
-	dc->hwss.prepare_bandwidth(dc, context);
+	for (i = 0; i < context->stream_count; i++) {
+		if (context->streams[i]->apply_seamless_boot_optimization)
+			dc->optimize_seamless_boot = true;
+	}
+
+	if (!dc->optimize_seamless_boot)
+		dc->hwss.prepare_bandwidth(dc, context);
 
 	/* re-program planes for existing stream, in case we need to
 	 * free up plane resource for later use
@@ -1135,12 +1150,15 @@ static enum dc_status dc_commit_state_no_check(struct dc *dc, struct dc_state *c
 
 	dc_enable_stereo(dc, context, dc_streams, context->stream_count);
 
-	/* pplib is notified if disp_num changed */
-	dc->hwss.optimize_bandwidth(dc, context);
+	if (!dc->optimize_seamless_boot)
+		/* pplib is notified if disp_num changed */
+		dc->hwss.optimize_bandwidth(dc, context);
 
 	for (i = 0; i < context->stream_count; i++)
 		context->streams[i]->mode_changed = false;
 
+	memset(&context->commit_hints, 0, sizeof(context->commit_hints));
+
 	dc_release_state(dc->current_state);
 
 	dc->current_state = context;
@@ -1177,7 +1195,7 @@ bool dc_post_update_surfaces_to_stream(struct dc *dc)
 	int i;
 	struct dc_state *context = dc->current_state;
 
-	if (dc->optimized_required == false)
+	if (!dc->optimized_required || dc->optimize_seamless_boot)
 		return true;
 
 	post_surface_trace(dc);
@@ -1661,6 +1679,7 @@ static void commit_planes_do_stream_update(struct dc *dc,
 				continue;
 
 			if (stream_update->dpms_off) {
+				dc->hwss.pipe_control_lock(dc, pipe_ctx, true);
 				if (*stream_update->dpms_off) {
 					core_link_disable_stream(pipe_ctx, KEEP_ACQUIRED_RESOURCE);
 					dc->hwss.optimize_bandwidth(dc, dc->current_state);
@@ -1668,6 +1687,7 @@ static void commit_planes_do_stream_update(struct dc *dc,
 					dc->hwss.prepare_bandwidth(dc, dc->current_state);
 					core_link_enable_stream(dc->current_state, pipe_ctx);
 				}
+				dc->hwss.pipe_control_lock(dc, pipe_ctx, false);
 			}
 
 			if (stream_update->abm_level && pipe_ctx->stream_res.abm) {
@@ -1695,7 +1715,16 @@ static void commit_planes_for_stream(struct dc *dc,
 	int i, j;
 	struct pipe_ctx *top_pipe_to_program = NULL;
 
-	if (update_type == UPDATE_TYPE_FULL) {
+	if (dc->optimize_seamless_boot && surface_count > 0) {
+		/* Optimize seamless boot flag keeps clocks and watermarks high until
+		 * first flip. After first flip, optimization is required to lower
+		 * bandwidth.
+		 */
+		dc->optimize_seamless_boot = false;
+		dc->optimized_required = true;
+	}
+
+	if (update_type == UPDATE_TYPE_FULL && !dc->optimize_seamless_boot) {
 		dc->hwss.prepare_bandwidth(dc, context);
 		context_clock_trace(dc, context);
 	}

drivers/gpu/drm/amd/display/dc/core/dc_link.c

@@ -640,7 +640,8 @@ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason)
 	bool same_dpcd = true;
 	enum dc_connection_type new_connection_type = dc_connection_none;
 	DC_LOGGER_INIT(link->ctx->logger);
-	if (link->connector_signal == SIGNAL_TYPE_VIRTUAL)
+
+	if (dc_is_virtual_signal(link->connector_signal))
 		return false;
 
 	if (false == dc_link_detect_sink(link, &new_connection_type)) {
@@ -720,9 +721,8 @@ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason)
 					same_dpcd = false;
 			}
 			/* Active dongle plug in without display or downstream unplug*/
-			if (link->type == dc_connection_active_dongle
-					&& link->dpcd_caps.sink_count.
-					bits.SINK_COUNT == 0) {
+			if (link->type == dc_connection_active_dongle &&
+				link->dpcd_caps.sink_count.bits.SINK_COUNT == 0) {
 				if (prev_sink != NULL) {
 					/* Downstream unplug */
 					dc_sink_release(prev_sink);
@@ -1172,8 +1172,6 @@ static bool construct(
 		goto create_fail;
 	}
 
-
-
 	/* TODO: #DAL3 Implement id to str function.*/
 	LINK_INFO("Connector[%d] description:"
 			"signal %d\n",
@@ -1207,7 +1205,7 @@ static bool construct(
 	link->link_enc = link->dc->res_pool->funcs->link_enc_create(
 								&enc_init_data);
 
-	if( link->link_enc == NULL) {
+	if (link->link_enc == NULL) {
 		DC_ERROR("Failed to create link encoder!\n");
 		goto link_enc_create_fail;
 	}
@@ -1466,14 +1464,14 @@ static enum dc_status enable_link_dp_mst(
 	if (link->cur_link_settings.lane_count != LANE_COUNT_UNKNOWN)
 		return DC_OK;
 
-	/* to make sure the pending down rep can be processed
-	 * before clear payload table
-	 */
-	dm_helpers_dp_mst_poll_pending_down_reply(link->ctx, link);
-
 	/* clear payload table */
 	dm_helpers_dp_mst_clear_payload_allocation_table(link->ctx, link);
 
+	/* to make sure the pending down rep can be processed
+	 * before enabling the link
+	 */
+	dm_helpers_dp_mst_poll_pending_down_reply(link->ctx, link);
+
 	/* set the sink to MST mode before enabling the link */
 	dp_enable_mst_on_sink(link, true);
 
@@ -1982,7 +1980,7 @@ static void enable_link_hdmi(struct pipe_ctx *pipe_ctx)
 			pipe_ctx->stream->signal,
 			stream->phy_pix_clk);
 
-	if (pipe_ctx->stream->signal == SIGNAL_TYPE_HDMI_TYPE_A)
+	if (dc_is_hdmi_signal(pipe_ctx->stream->signal))
 		dal_ddc_service_read_scdc_data(link->ddc);
 }
 
@@ -2074,11 +2072,28 @@ static void disable_link(struct dc_link *link, enum signal_type signal)
 	}
 }
 
+static uint32_t get_timing_pixel_clock_100hz(const struct dc_crtc_timing *timing)
+{
+
+	uint32_t pxl_clk = timing->pix_clk_100hz;
+
+	if (timing->pixel_encoding == PIXEL_ENCODING_YCBCR420)
+		pxl_clk /= 2;
+	else if (timing->pixel_encoding == PIXEL_ENCODING_YCBCR422)
+		pxl_clk = pxl_clk * 2 / 3;
+
+	if (timing->display_color_depth == COLOR_DEPTH_101010)
+		pxl_clk = pxl_clk * 10 / 8;
+	else if (timing->display_color_depth == COLOR_DEPTH_121212)
+		pxl_clk = pxl_clk * 12 / 8;
+
+	return pxl_clk;
+}
+
 static bool dp_active_dongle_validate_timing(
 		const struct dc_crtc_timing *timing,
 		const struct dpcd_caps *dpcd_caps)
 {
-	unsigned int required_pix_clk_100hz = timing->pix_clk_100hz;
 	const struct dc_dongle_caps *dongle_caps = &dpcd_caps->dongle_caps;
 
 	switch (dpcd_caps->dongle_type) {
@@ -2115,13 +2130,6 @@ static bool dp_active_dongle_validate_timing(
 		return false;
 	}
 
-
-	/* Check Color Depth and Pixel Clock */
-	if (timing->pixel_encoding == PIXEL_ENCODING_YCBCR420)
-		required_pix_clk_100hz /= 2;
-	else if (timing->pixel_encoding == PIXEL_ENCODING_YCBCR422)
-		required_pix_clk_100hz = required_pix_clk_100hz * 2 / 3;
-
 	switch (timing->display_color_depth) {
 	case COLOR_DEPTH_666:
 	case COLOR_DEPTH_888:
@@ -2130,14 +2138,11 @@ static bool dp_active_dongle_validate_timing(
 	case COLOR_DEPTH_101010:
 		if (dongle_caps->dp_hdmi_max_bpc < 10)
 			return false;
-		required_pix_clk_100hz = required_pix_clk_100hz * 10 / 8;
 		break;
 	case COLOR_DEPTH_121212:
 		if (dongle_caps->dp_hdmi_max_bpc < 12)
 			return false;
-		required_pix_clk_100hz = required_pix_clk_100hz * 12 / 8;
 		break;
-
 	case COLOR_DEPTH_141414:
 	case COLOR_DEPTH_161616:
 	default:
@@ -2145,7 +2150,7 @@ static bool dp_active_dongle_validate_timing(
 		return false;
 	}
 
-	if (required_pix_clk_100hz > (dongle_caps->dp_hdmi_max_pixel_clk * 10))
+	if (get_timing_pixel_clock_100hz(timing) > (dongle_caps->dp_hdmi_max_pixel_clk * 10))
 		return false;
 
 	return true;
@@ -2166,7 +2171,7 @@ enum dc_status dc_link_validate_mode_timing(
 		return DC_OK;
 
 	/* Passive Dongle */
-	if (0 != max_pix_clk && timing->pix_clk_100hz > max_pix_clk)
+	if (max_pix_clk != 0 && get_timing_pixel_clock_100hz(timing) > max_pix_clk)
 		return DC_EXCEED_DONGLE_CAP;
 
 	/* Active Dongle*/
@@ -2551,12 +2556,12 @@ void core_link_enable_stream(
 		struct dc_state *state,
 		struct pipe_ctx *pipe_ctx)
 {
-	struct dc  *core_dc = pipe_ctx->stream->ctx->dc;
+	struct dc *core_dc = pipe_ctx->stream->ctx->dc;
 	struct dc_stream_state *stream = pipe_ctx->stream;
 	enum dc_status status;
 	DC_LOGGER_INIT(pipe_ctx->stream->ctx->logger);
 
-	if (pipe_ctx->stream->signal != SIGNAL_TYPE_VIRTUAL) {
+	if (!dc_is_virtual_signal(pipe_ctx->stream->signal)) {
 		stream->link->link_enc->funcs->setup(
 			stream->link->link_enc,
 			pipe_ctx->stream->signal);
@@ -2570,9 +2575,10 @@ void core_link_enable_stream(
 		pipe_ctx->stream_res.stream_enc->funcs->dp_set_stream_attribute(
 			pipe_ctx->stream_res.stream_enc,
 			&stream->timing,
-			stream->output_color_space);
+			stream->output_color_space,
+			stream->link->dpcd_caps.dprx_feature.bits.SST_SPLIT_SDP_CAP);
 
-	if (dc_is_hdmi_signal(pipe_ctx->stream->signal))
+	if (dc_is_hdmi_tmds_signal(pipe_ctx->stream->signal))
 		pipe_ctx->stream_res.stream_enc->funcs->hdmi_set_stream_attribute(
 			pipe_ctx->stream_res.stream_enc,
 			&stream->timing,

drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c

@@ -573,12 +573,28 @@ bool dal_ddc_service_query_ddc_data(
 	return ret;
 }
 
-int dc_link_aux_transfer(struct ddc_service *ddc,
-		struct aux_payload *payload)
+/* dc_link_aux_transfer_raw() - Attempt to transfer
+ * the given aux payload.  This function does not perform
+ * retries or handle error states.  The reply is returned
+ * in the payload->reply and the result through
+ * *operation_result.  Returns the number of bytes transferred,
+ * or -1 on a failure.
+ */
+int dc_link_aux_transfer_raw(struct ddc_service *ddc,
+		struct aux_payload *payload,
+		enum aux_channel_operation_result *operation_result)
 {
-	return dce_aux_transfer(ddc, payload);
+	return dce_aux_transfer_raw(ddc, payload, operation_result);
 }
 
+/* dc_link_aux_transfer_with_retries() - Attempt to submit an
+ * aux payload, retrying on timeouts, defers, and busy states
+ * as outlined in the DP spec.  Returns true if the request
+ * was successful.
+ *
+ * Unless you want to implement your own retry semantics, this
+ * is probably the one you want.
+ */
 bool dc_link_aux_transfer_with_retries(struct ddc_service *ddc,
 		struct aux_payload *payload)
 {

drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c

@@ -93,12 +93,10 @@ static void dpcd_set_link_settings(
 	struct dc_link *link,
 	const struct link_training_settings *lt_settings)
 {
-	uint8_t rate = (uint8_t)
-	(lt_settings->link_settings.link_rate);
+	uint8_t rate;
 
 	union down_spread_ctrl downspread = { {0} };
 	union lane_count_set lane_count_set = { {0} };
-	uint8_t link_set_buffer[2];
 
 	downspread.raw = (uint8_t)
 	(lt_settings->link_settings.link_spread);
@@ -111,29 +109,42 @@ static void dpcd_set_link_settings(
 	lane_count_set.bits.POST_LT_ADJ_REQ_GRANTED =
 		link->dpcd_caps.max_ln_count.bits.POST_LT_ADJ_REQ_SUPPORTED;
 
-	link_set_buffer[0] = rate;
-	link_set_buffer[1] = lane_count_set.raw;
-
-	core_link_write_dpcd(link, DP_LINK_BW_SET,
-	link_set_buffer, 2);
 	core_link_write_dpcd(link, DP_DOWNSPREAD_CTRL,
 	&downspread.raw, sizeof(downspread));
 
+	core_link_write_dpcd(link, DP_LANE_COUNT_SET,
+	&lane_count_set.raw, 1);
+
 	if (link->dpcd_caps.dpcd_rev.raw >= DPCD_REV_14 &&
-		(link->dpcd_caps.link_rate_set >= 1 &&
-		link->dpcd_caps.link_rate_set <= 8)) {
+			lt_settings->link_settings.use_link_rate_set == true) {
+		rate = 0;
+		core_link_write_dpcd(link, DP_LINK_BW_SET, &rate, 1);
 		core_link_write_dpcd(link, DP_LINK_RATE_SET,
-		&link->dpcd_caps.link_rate_set, 1);
+				&lt_settings->link_settings.link_rate_set, 1);
+	} else {
+		rate = (uint8_t) (lt_settings->link_settings.link_rate);
+		core_link_write_dpcd(link, DP_LINK_BW_SET, &rate, 1);
 	}
 
-	DC_LOG_HW_LINK_TRAINING("%s\n %x rate = %x\n %x lane = %x\n %x spread = %x\n",
-		__func__,
-		DP_LINK_BW_SET,
-		lt_settings->link_settings.link_rate,
-		DP_LANE_COUNT_SET,
-		lt_settings->link_settings.lane_count,
-		DP_DOWNSPREAD_CTRL,
-		lt_settings->link_settings.link_spread);
+	if (rate) {
+		DC_LOG_HW_LINK_TRAINING("%s\n %x rate = %x\n %x lane = %x\n %x spread = %x\n",
+			__func__,
+			DP_LINK_BW_SET,
+			lt_settings->link_settings.link_rate,
+			DP_LANE_COUNT_SET,
+			lt_settings->link_settings.lane_count,
+			DP_DOWNSPREAD_CTRL,
+			lt_settings->link_settings.link_spread);
+	} else {
+		DC_LOG_HW_LINK_TRAINING("%s\n %x rate set = %x\n %x lane = %x\n %x spread = %x\n",
+			__func__,
+			DP_LINK_RATE_SET,
+			lt_settings->link_settings.link_rate_set,
+			DP_LANE_COUNT_SET,
+			lt_settings->link_settings.lane_count,
+			DP_DOWNSPREAD_CTRL,
+			lt_settings->link_settings.link_spread);
+	}
 
 }
 
@@ -952,6 +963,8 @@ enum link_training_result dc_link_dp_perform_link_training(
 
 	lt_settings.link_settings.link_rate = link_setting->link_rate;
 	lt_settings.link_settings.lane_count = link_setting->lane_count;
+	lt_settings.link_settings.use_link_rate_set = link_setting->use_link_rate_set;
+	lt_settings.link_settings.link_rate_set = link_setting->link_rate_set;
 
 	/*@todo[vdevulap] move SS to LS, should not be handled by displaypath*/
 
@@ -1075,7 +1088,7 @@ static struct dc_link_settings get_max_link_cap(struct dc_link *link)
 {
 	/* Set Default link settings */
 	struct dc_link_settings max_link_cap = {LANE_COUNT_FOUR, LINK_RATE_HIGH,
-			LINK_SPREAD_05_DOWNSPREAD_30KHZ};
+			LINK_SPREAD_05_DOWNSPREAD_30KHZ, false, 0};
 
 	/* Higher link settings based on feature supported */
 	if (link->link_enc->features.flags.bits.IS_HBR2_CAPABLE)
@@ -1629,17 +1642,102 @@ bool dp_validate_mode_timing(
 		return false;
 }
 
+static bool decide_dp_link_settings(struct dc_link *link, struct dc_link_settings *link_setting, uint32_t req_bw)
+{
+	struct dc_link_settings initial_link_setting = {
+		LANE_COUNT_ONE, LINK_RATE_LOW, LINK_SPREAD_DISABLED, false, 0};
+	struct dc_link_settings current_link_setting =
+			initial_link_setting;
+	uint32_t link_bw;
+
+	/* search for the minimum link setting that:
+	 * 1. is supported according to the link training result
+	 * 2. could support the b/w requested by the timing
+	 */
+	while (current_link_setting.link_rate <=
+			link->verified_link_cap.link_rate) {
+		link_bw = bandwidth_in_kbps_from_link_settings(
+				&current_link_setting);
+		if (req_bw <= link_bw) {
+			*link_setting = current_link_setting;
+			return true;
+		}
+
+		if (current_link_setting.lane_count <
+				link->verified_link_cap.lane_count) {
+			current_link_setting.lane_count =
+					increase_lane_count(
+							current_link_setting.lane_count);
+		} else {
+			current_link_setting.link_rate =
+					increase_link_rate(
+							current_link_setting.link_rate);
+			current_link_setting.lane_count =
+					initial_link_setting.lane_count;
+		}
+	}
+
+	return false;
+}
+
+static bool decide_edp_link_settings(struct dc_link *link, struct dc_link_settings *link_setting, uint32_t req_bw)
+{
+	struct dc_link_settings initial_link_setting;
+	struct dc_link_settings current_link_setting;
+	uint32_t link_bw;
+
+	if (link->dpcd_caps.dpcd_rev.raw < DPCD_REV_14 ||
+			link->dpcd_caps.edp_supported_link_rates_count == 0 ||
+			link->dc->config.optimize_edp_link_rate == false) {
+		*link_setting = link->verified_link_cap;
+		return true;
+	}
+
+	memset(&initial_link_setting, 0, sizeof(initial_link_setting));
+	initial_link_setting.lane_count = LANE_COUNT_ONE;
+	initial_link_setting.link_rate = link->dpcd_caps.edp_supported_link_rates[0];
+	initial_link_setting.link_spread = LINK_SPREAD_DISABLED;
+	initial_link_setting.use_link_rate_set = true;
+	initial_link_setting.link_rate_set = 0;
+	current_link_setting = initial_link_setting;
+
+	/* search for the minimum link setting that:
+	 * 1. is supported according to the link training result
+	 * 2. could support the b/w requested by the timing
+	 */
+	while (current_link_setting.link_rate <=
+			link->verified_link_cap.link_rate) {
+		link_bw = bandwidth_in_kbps_from_link_settings(
+				&current_link_setting);
+		if (req_bw <= link_bw) {
+			*link_setting = current_link_setting;
+			return true;
+		}
+
+		if (current_link_setting.lane_count <
+				link->verified_link_cap.lane_count) {
+			current_link_setting.lane_count =
+					increase_lane_count(
+							current_link_setting.lane_count);
+		} else {
+			if (current_link_setting.link_rate_set < link->dpcd_caps.edp_supported_link_rates_count) {
+				current_link_setting.link_rate_set++;
+				current_link_setting.link_rate =
+					link->dpcd_caps.edp_supported_link_rates[current_link_setting.link_rate_set];
+				current_link_setting.lane_count =
+									initial_link_setting.lane_count;
+			} else
+				break;
+		}
+	}
+	return false;
+}
+
 void decide_link_settings(struct dc_stream_state *stream,
 	struct dc_link_settings *link_setting)
 {
-
-	struct dc_link_settings initial_link_setting = {
-		LANE_COUNT_ONE, LINK_RATE_LOW, LINK_SPREAD_DISABLED};
-	struct dc_link_settings current_link_setting =
-			initial_link_setting;
 	struct dc_link *link;
 	uint32_t req_bw;
-	uint32_t link_bw;
 
 	req_bw = bandwidth_in_kbps_from_timing(&stream->timing);
 
@@ -1664,38 +1762,11 @@ void decide_link_settings(struct dc_stream_state *stream,
 		return;
 	}
 
-	/* EDP use the link cap setting */
 	if (link->connector_signal == SIGNAL_TYPE_EDP) {
-		*link_setting = link->verified_link_cap;
-		return;
-	}
-
-	/* search for the minimum link setting that:
-	 * 1. is supported according to the link training result
-	 * 2. could support the b/w requested by the timing
-	 */
-	while (current_link_setting.link_rate <=
-			link->verified_link_cap.link_rate) {
-		link_bw = bandwidth_in_kbps_from_link_settings(
-				&current_link_setting);
-		if (req_bw <= link_bw) {
-			*link_setting = current_link_setting;
+		if (decide_edp_link_settings(link, link_setting, req_bw))
 			return;
-		}
-
-		if (current_link_setting.lane_count <
-				link->verified_link_cap.lane_count) {
-			current_link_setting.lane_count =
-					increase_lane_count(
-							current_link_setting.lane_count);
-		} else {
-			current_link_setting.link_rate =
-					increase_link_rate(
-							current_link_setting.link_rate);
-			current_link_setting.lane_count =
-					initial_link_setting.lane_count;
-		}
-	}
+	} else if (decide_dp_link_settings(link, link_setting, req_bw))
+		return;
 
 	BREAK_TO_DEBUGGER();
 	ASSERT(link->verified_link_cap.lane_count != LANE_COUNT_UNKNOWN);
@@ -2155,11 +2226,7 @@ bool is_mst_supported(struct dc_link *link)
 
 bool is_dp_active_dongle(const struct dc_link *link)
 {
-	enum display_dongle_type dongle_type = link->dpcd_caps.dongle_type;
-
-	return (dongle_type == DISPLAY_DONGLE_DP_VGA_CONVERTER) ||
-			(dongle_type == DISPLAY_DONGLE_DP_DVI_CONVERTER) ||
-			(dongle_type == DISPLAY_DONGLE_DP_HDMI_CONVERTER);
+	return link->dpcd_caps.is_branch_dev;
 }
 
 static int translate_dpcd_max_bpc(enum dpcd_downstream_port_max_bpc bpc)
@@ -2193,6 +2260,9 @@ static void get_active_converter_info(
 		return;
 	}
 
+	/* DPCD 0x5 bit 0 = 1, it indicate it's branch device */
+	link->dpcd_caps.is_branch_dev = ds_port.fields.PORT_PRESENT;
+
 	switch (ds_port.fields.PORT_TYPE) {
 	case DOWNSTREAM_VGA:
 		link->dpcd_caps.dongle_type = DISPLAY_DONGLE_DP_VGA_CONVERTER;
@@ -2347,6 +2417,10 @@ static bool retrieve_link_cap(struct dc_link *link)
 {
 	uint8_t dpcd_data[DP_ADAPTER_CAP - DP_DPCD_REV + 1];
 
+	/*Only need to read 1 byte starting from DP_DPRX_FEATURE_ENUMERATION_LIST.
+	 */
+	uint8_t dpcd_dprx_data = '\0';
+
 	struct dp_device_vendor_id sink_id;
 	union down_stream_port_count down_strm_port_count;
 	union edp_configuration_cap edp_config_cap;
@@ -2383,7 +2457,10 @@ static bool retrieve_link_cap(struct dc_link *link)
 		aux_rd_interval.raw =
 			dpcd_data[DP_TRAINING_AUX_RD_INTERVAL];
 
-		if (aux_rd_interval.bits.EXT_RECIEVER_CAP_FIELD_PRESENT == 1) {
+		link->dpcd_caps.ext_receiver_cap_field_present =
+				aux_rd_interval.bits.EXT_RECEIVER_CAP_FIELD_PRESENT == 1 ? true:false;
+
+		if (aux_rd_interval.bits.EXT_RECEIVER_CAP_FIELD_PRESENT == 1) {
 			uint8_t ext_cap_data[16];
 
 			memset(ext_cap_data, '\0', sizeof(ext_cap_data));
@@ -2404,7 +2481,38 @@ static bool retrieve_link_cap(struct dc_link *link)
 	}
 
 	link->dpcd_caps.dpcd_rev.raw =
-		dpcd_data[DP_DPCD_REV - DP_DPCD_REV];
+			dpcd_data[DP_DPCD_REV - DP_DPCD_REV];
+
+	if (link->dpcd_caps.dpcd_rev.raw >= 0x14) {
+		for (i = 0; i < read_dpcd_retry_cnt; i++) {
+			status = core_link_read_dpcd(
+					link,
+					DP_DPRX_FEATURE_ENUMERATION_LIST,
+					&dpcd_dprx_data,
+					sizeof(dpcd_dprx_data));
+			if (status == DC_OK)
+				break;
+		}
+
+		link->dpcd_caps.dprx_feature.raw = dpcd_dprx_data;
+
+		if (status != DC_OK)
+			dm_error("%s: Read DPRX caps data failed.\n", __func__);
+	}
+
+	else {
+		link->dpcd_caps.dprx_feature.raw = 0;
+	}
+
+
+	/* Error condition checking...
+	 * It is impossible for Sink to report Max Lane Count = 0.
+	 * It is possible for Sink to report Max Link Rate = 0, if it is
+	 * an eDP device that is reporting specialized link rates in the
+	 * SUPPORTED_LINK_RATE table.
+	 */
+	if (dpcd_data[DP_MAX_LANE_COUNT - DP_DPCD_REV] == 0)
+		return false;
 
 	ds_port.byte = dpcd_data[DP_DOWNSTREAMPORT_PRESENT -
 				 DP_DPCD_REV];
@@ -2536,31 +2644,31 @@ enum dc_link_rate linkRateInKHzToLinkRateMultiplier(uint32_t link_rate_in_khz)
 
 void detect_edp_sink_caps(struct dc_link *link)
 {
-	uint8_t supported_link_rates[16] = {0};
+	uint8_t supported_link_rates[16];
 	uint32_t entry;
 	uint32_t link_rate_in_khz;
 	enum dc_link_rate link_rate = LINK_RATE_UNKNOWN;
 
 	retrieve_link_cap(link);
+	link->dpcd_caps.edp_supported_link_rates_count = 0;
+	memset(supported_link_rates, 0, sizeof(supported_link_rates));
 
-	if (link->dpcd_caps.dpcd_rev.raw >= DPCD_REV_14) {
+	if (link->dpcd_caps.dpcd_rev.raw >= DPCD_REV_14 &&
+			link->dc->config.optimize_edp_link_rate) {
 		// Read DPCD 00010h - 0001Fh 16 bytes at one shot
 		core_link_read_dpcd(link, DP_SUPPORTED_LINK_RATES,
 							supported_link_rates, sizeof(supported_link_rates));
 
-		link->dpcd_caps.link_rate_set = 0;
 		for (entry = 0; entry < 16; entry += 2) {
 			// DPCD register reports per-lane link rate = 16-bit link rate capability
-			// value X 200 kHz. Need multipler to find link rate in kHz.
+			// value X 200 kHz. Need multiplier to find link rate in kHz.
 			link_rate_in_khz = (supported_link_rates[entry+1] * 0x100 +
 										supported_link_rates[entry]) * 200;
 
 			if (link_rate_in_khz != 0) {
 				link_rate = linkRateInKHzToLinkRateMultiplier(link_rate_in_khz);
-				if (link->reported_link_cap.link_rate < link_rate) {
-					link->reported_link_cap.link_rate = link_rate;
-					link->dpcd_caps.link_rate_set = entry;
-				}
+				link->dpcd_caps.edp_supported_link_rates[link->dpcd_caps.edp_supported_link_rates_count] = link_rate;
+				link->dpcd_caps.edp_supported_link_rates_count++;
 			}
 		}
 	}
@@ -2601,6 +2709,7 @@ static void set_crtc_test_pattern(struct dc_link *link,
 	enum dc_color_depth color_depth = pipe_ctx->
 		stream->timing.display_color_depth;
 	struct bit_depth_reduction_params params;
+	struct output_pixel_processor *opp = pipe_ctx->stream_res.opp;
 
 	memset(&params, 0, sizeof(params));
 
@@ -2640,8 +2749,7 @@ static void set_crtc_test_pattern(struct dc_link *link,
 	{
 		/* disable bit depth reduction */
 		pipe_ctx->stream->bit_depth_params = params;
-		pipe_ctx->stream_res.opp->funcs->
-			opp_program_bit_depth_reduction(pipe_ctx->stream_res.opp, &params);
+		opp->funcs->opp_program_bit_depth_reduction(opp, &params);
 		if (pipe_ctx->stream_res.tg->funcs->set_test_pattern)
 			pipe_ctx->stream_res.tg->funcs->set_test_pattern(pipe_ctx->stream_res.tg,
 				controller_test_pattern, color_depth);
@@ -2650,11 +2758,9 @@ static void set_crtc_test_pattern(struct dc_link *link,
 	case DP_TEST_PATTERN_VIDEO_MODE:
 	{
 		/* restore bitdepth reduction */
-		resource_build_bit_depth_reduction_params(pipe_ctx->stream,
-					&params);
+		resource_build_bit_depth_reduction_params(pipe_ctx->stream, &params);
 		pipe_ctx->stream->bit_depth_params = params;
-		pipe_ctx->stream_res.opp->funcs->
-			opp_program_bit_depth_reduction(pipe_ctx->stream_res.opp, &params);
+		opp->funcs->opp_program_bit_depth_reduction(opp, &params);
 		if (pipe_ctx->stream_res.tg->funcs->set_test_pattern)
 			pipe_ctx->stream_res.tg->funcs->set_test_pattern(pipe_ctx->stream_res.tg,
 				CONTROLLER_DP_TEST_PATTERN_VIDEOMODE,

drivers/gpu/drm/amd/display/dc/core/dc_resource.c

@@ -31,6 +31,8 @@
 #include "opp.h"
 #include "timing_generator.h"
 #include "transform.h"
+#include "dccg.h"
+#include "dchubbub.h"
 #include "dpp.h"
 #include "core_types.h"
 #include "set_mode_types.h"
@@ -104,44 +106,43 @@ enum dce_version resource_parse_asic_id(struct hw_asic_id asic_id)
 	return dc_version;
 }
 
-struct resource_pool *dc_create_resource_pool(
-				struct dc  *dc,
-				int num_virtual_links,
-				enum dce_version dc_version,
-				struct hw_asic_id asic_id)
+struct resource_pool *dc_create_resource_pool(struct dc  *dc,
+					      const struct dc_init_data *init_data,
+					      enum dce_version dc_version)
 {
 	struct resource_pool *res_pool = NULL;
 
 	switch (dc_version) {
 	case DCE_VERSION_8_0:
 		res_pool = dce80_create_resource_pool(
-			num_virtual_links, dc);
+				init_data->num_virtual_links, dc);
 		break;
 	case DCE_VERSION_8_1:
 		res_pool = dce81_create_resource_pool(
-			num_virtual_links, dc);
+				init_data->num_virtual_links, dc);
 		break;
 	case DCE_VERSION_8_3:
 		res_pool = dce83_create_resource_pool(
-			num_virtual_links, dc);
+				init_data->num_virtual_links, dc);
 		break;
 	case DCE_VERSION_10_0:
 		res_pool = dce100_create_resource_pool(
-				num_virtual_links, dc);
+				init_data->num_virtual_links, dc);
 		break;
 	case DCE_VERSION_11_0:
 		res_pool = dce110_create_resource_pool(
-			num_virtual_links, dc, asic_id);
+				init_data->num_virtual_links, dc,
+				init_data->asic_id);
 		break;
 	case DCE_VERSION_11_2:
 	case DCE_VERSION_11_22:
 		res_pool = dce112_create_resource_pool(
-			num_virtual_links, dc);
+				init_data->num_virtual_links, dc);
 		break;
 	case DCE_VERSION_12_0:
 	case DCE_VERSION_12_1:
 		res_pool = dce120_create_resource_pool(
-			num_virtual_links, dc);
+				init_data->num_virtual_links, dc);
 		break;
 
 #if defined(CONFIG_DRM_AMD_DC_DCN1_0)
@@ -149,8 +150,7 @@ struct resource_pool *dc_create_resource_pool(
 #if defined(CONFIG_DRM_AMD_DC_DCN1_01)
 	case DCN_VERSION_1_01:
 #endif
-		res_pool = dcn10_create_resource_pool(
-				num_virtual_links, dc);
+		res_pool = dcn10_create_resource_pool(init_data, dc);
 		break;
 #endif
 
@@ -163,7 +163,28 @@ struct resource_pool *dc_create_resource_pool(
 
 		if (dc->ctx->dc_bios->funcs->get_firmware_info(
 				dc->ctx->dc_bios, &fw_info) == BP_RESULT_OK) {
-				res_pool->ref_clock_inKhz = fw_info.pll_info.crystal_frequency;
+				res_pool->ref_clocks.xtalin_clock_inKhz = fw_info.pll_info.crystal_frequency;
+
+				if (IS_FPGA_MAXIMUS_DC(dc->ctx->dce_environment)) {
+					// On FPGA these dividers are currently not configured by GDB
+					res_pool->ref_clocks.dccg_ref_clock_inKhz = res_pool->ref_clocks.xtalin_clock_inKhz;
+					res_pool->ref_clocks.dchub_ref_clock_inKhz = res_pool->ref_clocks.xtalin_clock_inKhz;
+				} else if (res_pool->dccg && res_pool->hubbub) {
+					// If DCCG reference frequency cannot be determined (usually means not set to xtalin) then this is a critical error
+					// as this value must be known for DCHUB programming
+					(res_pool->dccg->funcs->get_dccg_ref_freq)(res_pool->dccg,
+							fw_info.pll_info.crystal_frequency,
+							&res_pool->ref_clocks.dccg_ref_clock_inKhz);
+
+					// Similarly, if DCHUB reference frequency cannot be determined, then it is also a critical error
+					(res_pool->hubbub->funcs->get_dchub_ref_freq)(res_pool->hubbub,
+							res_pool->ref_clocks.dccg_ref_clock_inKhz,
+							&res_pool->ref_clocks.dchub_ref_clock_inKhz);
+				} else {
+					// Not all ASICs have DCCG sw component
+					res_pool->ref_clocks.dccg_ref_clock_inKhz = res_pool->ref_clocks.xtalin_clock_inKhz;
+					res_pool->ref_clocks.dchub_ref_clock_inKhz = res_pool->ref_clocks.xtalin_clock_inKhz;
+				}
 			} else
 				ASSERT_CRITICAL(false);
 	}
@@ -260,6 +281,7 @@ bool resource_construct(
 			pool->stream_enc_count++;
 		}
 	}
+
 	dc->caps.dynamic_audio = false;
 	if (pool->audio_count < pool->stream_enc_count) {
 		dc->caps.dynamic_audio = true;
@@ -1014,24 +1036,60 @@ enum dc_status resource_build_scaling_params_for_context(
 
 struct pipe_ctx *find_idle_secondary_pipe(
 		struct resource_context *res_ctx,
-		const struct resource_pool *pool)
+		const struct resource_pool *pool,
+		const struct pipe_ctx *primary_pipe)
 {
 	int i;
 	struct pipe_ctx *secondary_pipe = NULL;
 
 	/*
-	 * search backwards for the second pipe to keep pipe
-	 * assignment more consistent
+	 * We add a preferred pipe mapping to avoid the chance that
+	 * MPCCs already in use will need to be reassigned to other trees.
+	 * For example, if we went with the strict, assign backwards logic:
+	 *
+	 * (State 1)
+	 * Display A on, no surface, top pipe = 0
+	 * Display B on, no surface, top pipe = 1
+	 *
+	 * (State 2)
+	 * Display A on, no surface, top pipe = 0
+	 * Display B on, surface enable, top pipe = 1, bottom pipe = 5
+	 *
+	 * (State 3)
+	 * Display A on, surface enable, top pipe = 0, bottom pipe = 5
+	 * Display B on, surface enable, top pipe = 1, bottom pipe = 4
+	 *
+	 * The state 2->3 transition requires remapping MPCC 5 from display B
+	 * to display A.
+	 *
+	 * However, with the preferred pipe logic, state 2 would look like:
+	 *
+	 * (State 2)
+	 * Display A on, no surface, top pipe = 0
+	 * Display B on, surface enable, top pipe = 1, bottom pipe = 4
+	 *
+	 * This would then cause 2->3 to not require remapping any MPCCs.
 	 */
-
-	for (i = pool->pipe_count - 1; i >= 0; i--) {
-		if (res_ctx->pipe_ctx[i].stream == NULL) {
-			secondary_pipe = &res_ctx->pipe_ctx[i];
-			secondary_pipe->pipe_idx = i;
-			break;
+	if (primary_pipe) {
+		int preferred_pipe_idx = (pool->pipe_count - 1) - primary_pipe->pipe_idx;
+		if (res_ctx->pipe_ctx[preferred_pipe_idx].stream == NULL) {
+			secondary_pipe = &res_ctx->pipe_ctx[preferred_pipe_idx];
+			secondary_pipe->pipe_idx = preferred_pipe_idx;
 		}
 	}
 
+	/*
+	 * search backwards for the second pipe to keep pipe
+	 * assignment more consistent
+	 */
+	if (!secondary_pipe)
+		for (i = pool->pipe_count - 1; i >= 0; i--) {
+			if (res_ctx->pipe_ctx[i].stream == NULL) {
+				secondary_pipe = &res_ctx->pipe_ctx[i];
+				secondary_pipe->pipe_idx = i;
+				break;
+			}
+		}
 
 	return secondary_pipe;
 }
@@ -1214,6 +1272,9 @@ bool dc_add_plane_to_context(
 		free_pipe->clock_source = tail_pipe->clock_source;
 		free_pipe->top_pipe = tail_pipe;
 		tail_pipe->bottom_pipe = free_pipe;
+	} else if (free_pipe->bottom_pipe && free_pipe->bottom_pipe->plane_state == NULL) {
+		ASSERT(free_pipe->bottom_pipe->stream_res.opp != free_pipe->stream_res.opp);
+		free_pipe->bottom_pipe->plane_state = plane_state;
 	}
 
 	/* assign new surfaces*/
@@ -1224,6 +1285,40 @@ bool dc_add_plane_to_context(
 	return true;
 }
 
+struct pipe_ctx *dc_res_get_odm_bottom_pipe(struct pipe_ctx *pipe_ctx)
+{
+	struct pipe_ctx *bottom_pipe = pipe_ctx->bottom_pipe;
+
+	/* ODM should only be updated once per otg */
+	if (pipe_ctx->top_pipe)
+		return NULL;
+
+	while (bottom_pipe) {
+		if (bottom_pipe->stream_res.opp != pipe_ctx->stream_res.opp)
+			break;
+		bottom_pipe = bottom_pipe->bottom_pipe;
+	}
+
+	return bottom_pipe;
+}
+
+bool dc_res_is_odm_head_pipe(struct pipe_ctx *pipe_ctx)
+{
+	struct pipe_ctx *top_pipe = pipe_ctx->top_pipe;
+	bool result = false;
+
+	if (top_pipe && top_pipe->stream_res.opp == pipe_ctx->stream_res.opp)
+		return false;
+
+	while (top_pipe) {
+		if (!top_pipe->top_pipe && top_pipe->stream_res.opp != pipe_ctx->stream_res.opp)
+			result = true;
+		top_pipe = top_pipe->top_pipe;
+	}
+
+	return result;
+}
+
 bool dc_remove_plane_from_context(
 		const struct dc *dc,
 		struct dc_stream_state *stream,
@@ -1247,10 +1342,14 @@ bool dc_remove_plane_from_context(
 
 	/* release pipe for plane*/
 	for (i = pool->pipe_count - 1; i >= 0; i--) {
-		struct pipe_ctx *pipe_ctx;
+		struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i];
 
-		if (context->res_ctx.pipe_ctx[i].plane_state == plane_state) {
-			pipe_ctx = &context->res_ctx.pipe_ctx[i];
+		if (pipe_ctx->plane_state == plane_state) {
+			if (dc_res_is_odm_head_pipe(pipe_ctx)) {
+				pipe_ctx->plane_state = NULL;
+				pipe_ctx->bottom_pipe = NULL;
+				continue;
+			}
 
 			if (pipe_ctx->top_pipe)
 				pipe_ctx->top_pipe->bottom_pipe = pipe_ctx->bottom_pipe;
@@ -1268,8 +1367,9 @@ bool dc_remove_plane_from_context(
 			 */
 			if (!pipe_ctx->top_pipe) {
 				pipe_ctx->plane_state = NULL;
-				pipe_ctx->bottom_pipe = NULL;
-			} else  {
+				if (!dc_res_get_odm_bottom_pipe(pipe_ctx))
+					pipe_ctx->bottom_pipe = NULL;
+			} else {
 				memset(pipe_ctx, 0, sizeof(*pipe_ctx));
 			}
 		}
@@ -1674,6 +1774,9 @@ enum dc_status dc_remove_stream_from_ctx(
 	for (i = 0; i < MAX_PIPES; i++) {
 		if (new_ctx->res_ctx.pipe_ctx[i].stream == stream &&
 				!new_ctx->res_ctx.pipe_ctx[i].top_pipe) {
+			struct pipe_ctx *odm_pipe =
+					dc_res_get_odm_bottom_pipe(&new_ctx->res_ctx.pipe_ctx[i]);
+
 			del_pipe = &new_ctx->res_ctx.pipe_ctx[i];
 
 			ASSERT(del_pipe->stream_res.stream_enc);
@@ -1698,6 +1801,8 @@ enum dc_status dc_remove_stream_from_ctx(
 				dc->res_pool->funcs->remove_stream_from_ctx(dc, new_ctx, stream);
 
 			memset(del_pipe, 0, sizeof(*del_pipe));
+			if (odm_pipe)
+				memset(odm_pipe, 0, sizeof(*odm_pipe));
 
 			break;
 		}
@@ -1855,6 +1960,7 @@ enum dc_status resource_map_pool_resources(
 	struct dc_context *dc_ctx = dc->ctx;
 	struct pipe_ctx *pipe_ctx = NULL;
 	int pipe_idx = -1;
+	struct dc_bios *dcb = dc->ctx->dc_bios;
 
 	/* TODO Check if this is needed */
 	/*if (!resource_is_stream_unchanged(old_context, stream)) {
@@ -1869,6 +1975,13 @@ enum dc_status resource_map_pool_resources(
 
 	calculate_phy_pix_clks(stream);
 
+	/* TODO: Check Linux */
+	if (dc->config.allow_seamless_boot_optimization &&
+			!dcb->funcs->is_accelerated_mode(dcb)) {
+		if (dc_validate_seamless_boot_timing(dc, stream->sink, &stream->timing))
+			stream->apply_seamless_boot_optimization = true;
+	}
+
 	if (stream->apply_seamless_boot_optimization)
 		pipe_idx = acquire_resource_from_hw_enabled_state(
 				&context->res_ctx,
@@ -2315,6 +2428,21 @@ static void set_spd_info_packet(
 	*info_packet = stream->vrr_infopacket;
 }
 
+static void set_dp_sdp_info_packet(
+		struct dc_info_packet *info_packet,
+		struct dc_stream_state *stream)
+{
+	/* SPD info packet for custom sdp message */
+
+	/* Return if false. If true,
+	 * set the corresponding bit in the info packet
+	 */
+	if (!stream->dpsdp_infopacket.valid)
+		return;
+
+	*info_packet = stream->dpsdp_infopacket;
+}
+
 static void set_hdr_static_info_packet(
 		struct dc_info_packet *info_packet,
 		struct dc_stream_state *stream)
@@ -2411,6 +2539,7 @@ void resource_build_info_frame(struct pipe_ctx *pipe_ctx)
 	info->spd.valid = false;
 	info->hdrsmd.valid = false;
 	info->vsc.valid = false;
+	info->dpsdp.valid = false;
 
 	signal = pipe_ctx->stream->signal;
 
@@ -2430,6 +2559,8 @@ void resource_build_info_frame(struct pipe_ctx *pipe_ctx)
 		set_spd_info_packet(&info->spd, pipe_ctx->stream);
 
 		set_hdr_static_info_packet(&info->hdrsmd, pipe_ctx->stream);
+
+		set_dp_sdp_info_packet(&info->dpsdp, pipe_ctx->stream);
 	}
 
 	patch_gamut_packet_checksum(&info->gamut);
@@ -2657,10 +2788,11 @@ enum dc_status dc_validate_stream(struct dc *dc, struct dc_stream_state *stream)
 	if (!tg->funcs->validate_timing(tg, &stream->timing))
 		res = DC_FAIL_CONTROLLER_VALIDATE;
 
-	if (res == DC_OK)
+	if (res == DC_OK) {
 		if (!link->link_enc->funcs->validate_output_with_stream(
 						link->link_enc, stream))
 			res = DC_FAIL_ENC_VALIDATE;
+	}
 
 	/* TODO: validate audio ASIC caps, encoder */
 

drivers/gpu/drm/amd/display/dc/core/dc_stream.c

@@ -29,6 +29,9 @@
 #include "resource.h"
 #include "ipp.h"
 #include "timing_generator.h"
+#if defined(CONFIG_DRM_AMD_DC_DCN1_0)
+#include "dcn10/dcn10_hw_sequencer.h"
+#endif
 
 #define DC_LOGGER dc->ctx->logger
 
@@ -196,6 +199,34 @@ struct dc_stream_status *dc_stream_get_status(
 	return dc_stream_get_status_from_state(dc->current_state, stream);
 }
 
+static void delay_cursor_until_vupdate(struct pipe_ctx *pipe_ctx, struct dc *dc)
+{
+#if defined(CONFIG_DRM_AMD_DC_DCN1_0)
+	unsigned int vupdate_line;
+	unsigned int lines_to_vupdate, us_to_vupdate, vpos, nvpos;
+	struct dc_stream_state *stream = pipe_ctx->stream;
+	unsigned int us_per_line;
+
+	if (stream->ctx->asic_id.chip_family == FAMILY_RV &&
+			ASIC_REV_IS_RAVEN(stream->ctx->asic_id.hw_internal_rev)) {
+
+		vupdate_line = get_vupdate_offset_from_vsync(pipe_ctx);
+		dc_stream_get_crtc_position(dc, &stream, 1, &vpos, &nvpos);
+
+		if (vpos >= vupdate_line)
+			return;
+
+		us_per_line = stream->timing.h_total * 10000 / stream->timing.pix_clk_100hz;
+		lines_to_vupdate = vupdate_line - vpos;
+		us_to_vupdate = lines_to_vupdate * us_per_line;
+
+		/* 70 us is a conservative estimate of cursor update time*/
+		if (us_to_vupdate < 70)
+			udelay(us_to_vupdate);
+	}
+#endif
+}
+
 /**
  * dc_stream_set_cursor_attributes() - Update cursor attributes and set cursor surface address
  */
@@ -234,6 +265,8 @@ bool dc_stream_set_cursor_attributes(
 
 		if (!pipe_to_program) {
 			pipe_to_program = pipe_ctx;
+
+			delay_cursor_until_vupdate(pipe_ctx, core_dc);
 			core_dc->hwss.pipe_control_lock(core_dc, pipe_to_program, true);
 		}
 
@@ -283,6 +316,8 @@ bool dc_stream_set_cursor_position(
 
 		if (!pipe_to_program) {
 			pipe_to_program = pipe_ctx;
+
+			delay_cursor_until_vupdate(pipe_ctx, core_dc);
 			core_dc->hwss.pipe_control_lock(core_dc, pipe_to_program, true);
 		}
 
@@ -314,6 +349,68 @@ uint32_t dc_stream_get_vblank_counter(const struct dc_stream_state *stream)
 	return 0;
 }
 
+static void build_dp_sdp_info_frame(struct pipe_ctx *pipe_ctx,
+		const uint8_t  *custom_sdp_message,
+		unsigned int sdp_message_size)
+{
+	uint8_t i;
+	struct encoder_info_frame *info = &pipe_ctx->stream_res.encoder_info_frame;
+
+	/* set valid info */
+	info->dpsdp.valid = true;
+
+	/* set sdp message header */
+	info->dpsdp.hb0 = custom_sdp_message[0]; /* package id */
+	info->dpsdp.hb1 = custom_sdp_message[1]; /* package type */
+	info->dpsdp.hb2 = custom_sdp_message[2]; /* package specific byte 0 any data */
+	info->dpsdp.hb3 = custom_sdp_message[3]; /* package specific byte 0 any data */
+
+	/* set sdp message data */
+	for (i = 0; i < 32; i++)
+		info->dpsdp.sb[i] = (custom_sdp_message[i+4]);
+
+}
+
+static void invalid_dp_sdp_info_frame(struct pipe_ctx *pipe_ctx)
+{
+	struct encoder_info_frame *info = &pipe_ctx->stream_res.encoder_info_frame;
+
+	/* in-valid info */
+	info->dpsdp.valid = false;
+}
+
+bool dc_stream_send_dp_sdp(const struct dc_stream_state *stream,
+		const uint8_t *custom_sdp_message,
+		unsigned int sdp_message_size)
+{
+	int i;
+	struct dc  *core_dc;
+	struct resource_context *res_ctx;
+
+	if (stream == NULL) {
+		dm_error("DC: dc_stream is NULL!\n");
+		return false;
+	}
+
+	core_dc = stream->ctx->dc;
+	res_ctx = &core_dc->current_state->res_ctx;
+
+	for (i = 0; i < MAX_PIPES; i++) {
+		struct pipe_ctx *pipe_ctx = &res_ctx->pipe_ctx[i];
+
+		if (pipe_ctx->stream != stream)
+			continue;
+
+		build_dp_sdp_info_frame(pipe_ctx, custom_sdp_message, sdp_message_size);
+
+		core_dc->hwss.update_info_frame(pipe_ctx);
+
+		invalid_dp_sdp_info_frame(pipe_ctx);
+	}
+
+	return true;
+}
+
 bool dc_stream_get_scanoutpos(const struct dc_stream_state *stream,
 				  uint32_t *v_blank_start,
 				  uint32_t *v_blank_end,

drivers/gpu/drm/amd/display/dc/core/dc_surface.c

@@ -119,6 +119,19 @@ const struct dc_plane_status *dc_plane_get_status(
 	if (core_dc->current_state == NULL)
 		return NULL;
 
+	/* Find the current plane state and set its pending bit to false */
+	for (i = 0; i < core_dc->res_pool->pipe_count; i++) {
+		struct pipe_ctx *pipe_ctx =
+				&core_dc->current_state->res_ctx.pipe_ctx[i];
+
+		if (pipe_ctx->plane_state != plane_state)
+			continue;
+
+		pipe_ctx->plane_state->status.is_flip_pending = false;
+
+		break;
+	}
+
 	for (i = 0; i < core_dc->res_pool->pipe_count; i++) {
 		struct pipe_ctx *pipe_ctx =
 				&core_dc->current_state->res_ctx.pipe_ctx[i];

drivers/gpu/drm/amd/display/dc/dc.h

@@ -39,9 +39,10 @@
 #include "inc/hw/dmcu.h"
 #include "dml/display_mode_lib.h"
 
-#define DC_VER "3.2.17"
+#define DC_VER "3.2.24"
 
 #define MAX_SURFACES 3
+#define MAX_PLANES 6
 #define MAX_STREAMS 6
 #define MAX_SINKS_PER_LINK 4
 
@@ -53,6 +54,22 @@ struct dc_versions {
 	struct dmcu_version dmcu_version;
 };
 
+enum dc_plane_type {
+	DC_PLANE_TYPE_INVALID,
+	DC_PLANE_TYPE_DCE_RGB,
+	DC_PLANE_TYPE_DCE_UNDERLAY,
+	DC_PLANE_TYPE_DCN_UNIVERSAL,
+};
+
+struct dc_plane_cap {
+	enum dc_plane_type type;
+	uint32_t blends_with_above : 1;
+	uint32_t blends_with_below : 1;
+	uint32_t per_pixel_alpha : 1;
+	uint32_t supports_argb8888 : 1;
+	uint32_t supports_nv12 : 1;
+};
+
 struct dc_caps {
 	uint32_t max_streams;
 	uint32_t max_links;
@@ -73,6 +90,7 @@ struct dc_caps {
 	bool force_dp_tps4_for_cp2520;
 	bool disable_dp_clk_share;
 	bool psp_setup_panel_mode;
+	struct dc_plane_cap planes[MAX_PLANES];
 };
 
 struct dc_dcc_surface_param {
@@ -164,6 +182,8 @@ struct dc_config {
 	bool gpu_vm_support;
 	bool disable_disp_pll_sharing;
 	bool fbc_support;
+	bool optimize_edp_link_rate;
+	bool allow_seamless_boot_optimization;
 };
 
 enum visual_confirm {
@@ -203,6 +223,7 @@ struct dc_clocks {
 	int fclk_khz;
 	int phyclk_khz;
 	int dramclk_khz;
+	bool p_state_change_support;
 };
 
 struct dc_debug_options {
@@ -257,6 +278,7 @@ struct dc_debug_options {
 	bool skip_detection_link_training;
 	unsigned int force_odm_combine; //bit vector based on otg inst
 	unsigned int force_fclk_khz;
+	bool disable_tri_buf;
 };
 
 struct dc_debug_data {
@@ -265,6 +287,14 @@ struct dc_debug_data {
 	uint32_t auxErrorCount;
 };
 
+struct dc_bounding_box_overrides {
+	int sr_exit_time_ns;
+	int sr_enter_plus_exit_time_ns;
+	int urgent_latency_ns;
+	int percent_of_ideal_drambw;
+	int dram_clock_change_latency_ns;
+};
+
 struct dc_state;
 struct resource_pool;
 struct dce_hwseq;
@@ -274,6 +304,7 @@ struct dc {
 	struct dc_cap_funcs cap_funcs;
 	struct dc_config config;
 	struct dc_debug_options debug;
+	struct dc_bounding_box_overrides bb_overrides;
 	struct dc_context *ctx;
 
 	uint8_t link_count;
@@ -298,8 +329,12 @@ struct dc {
 	struct hw_sequencer_funcs hwss;
 	struct dce_hwseq *hwseq;
 
+	/* Require to optimize clocks and bandwidth for added/removed planes */
 	bool optimized_required;
 
+	/* Require to maintain clocks and bandwidth for UEFI enabled HW */
+	bool optimize_seamless_boot;
+
 	/* FBC compressor */
 	struct compressor *fbc_compressor;
 
@@ -327,6 +362,7 @@ struct dc_init_data {
 	struct hw_asic_id asic_id;
 	void *driver; /* ctx */
 	struct cgs_device *cgs_device;
+	struct dc_bounding_box_overrides bb_overrides;
 
 	int num_virtual_links;
 	/*
@@ -594,7 +630,7 @@ struct dc_validation_set {
 	uint8_t plane_count;
 };
 
-bool dc_validate_seamless_boot_timing(struct dc *dc,
+bool dc_validate_seamless_boot_timing(const struct dc *dc,
 				const struct dc_sink *sink,
 				struct dc_crtc_timing *crtc_timing);
 
@@ -645,9 +681,16 @@ struct dpcd_caps {
 	union dpcd_rev dpcd_rev;
 	union max_lane_count max_ln_count;
 	union max_down_spread max_down_spread;
+	union dprx_feature dprx_feature;
+
+	/* valid only for eDP v1.4 or higher*/
+	uint8_t edp_supported_link_rates_count;
+	enum dc_link_rate edp_supported_link_rates[8];
 
 	/* dongle type (DP converter, CV smart dongle) */
 	enum display_dongle_type dongle_type;
+	/* branch device or sink device */
+	bool is_branch_dev;
 	/* Dongle's downstream count. */
 	union sink_count sink_count;
 	/* If dongle_type == DISPLAY_DONGLE_DP_HDMI_CONVERTER,
@@ -663,11 +706,11 @@ struct dpcd_caps {
 	int8_t branch_dev_name[6];
 	int8_t branch_hw_revision;
 	int8_t branch_fw_revision[2];
-	uint8_t link_rate_set;
 
 	bool allow_invalid_MSA_timing_param;
 	bool panel_mode_edp;
 	bool dpcd_display_control_capable;
+	bool ext_receiver_cap_field_present;
 };
 
 #include "dc_link.h"

drivers/gpu/drm/amd/display/dc/dc_ddc_types.h

@@ -68,6 +68,8 @@ enum aux_transaction_reply {
 	AUX_TRANSACTION_REPLY_AUX_ACK = 0x00,
 	AUX_TRANSACTION_REPLY_AUX_NACK = 0x01,
 	AUX_TRANSACTION_REPLY_AUX_DEFER = 0x02,
+	AUX_TRANSACTION_REPLY_I2C_OVER_AUX_NACK = 0x04,
+	AUX_TRANSACTION_REPLY_I2C_OVER_AUX_DEFER = 0x08,
 
 	AUX_TRANSACTION_REPLY_I2C_ACK = 0x00,
 	AUX_TRANSACTION_REPLY_I2C_NACK = 0x10,

drivers/gpu/drm/amd/display/dc/dc_dp_types.h

@@ -94,6 +94,8 @@ struct dc_link_settings {
 	enum dc_lane_count lane_count;
 	enum dc_link_rate link_rate;
 	enum dc_link_spread link_spread;
+	bool use_link_rate_set;
+	uint8_t link_rate_set;
 };
 
 struct dc_lane_settings {
@@ -420,10 +422,24 @@ union edp_configuration_cap {
 	uint8_t raw;
 };
 
+union dprx_feature {
+	struct {
+		uint8_t GTC_CAP:1;                             // bit 0: DP 1.3+
+		uint8_t SST_SPLIT_SDP_CAP:1;                   // bit 1: DP 1.4
+		uint8_t AV_SYNC_CAP:1;                         // bit 2: DP 1.3+
+		uint8_t VSC_SDP_COLORIMETRY_SUPPORTED:1;       // bit 3: DP 1.3+
+		uint8_t VSC_EXT_VESA_SDP_SUPPORTED:1;          // bit 4: DP 1.4
+		uint8_t VSC_EXT_VESA_SDP_CHAINING_SUPPORTED:1; // bit 5: DP 1.4
+		uint8_t VSC_EXT_CEA_SDP_SUPPORTED:1;           // bit 6: DP 1.4
+		uint8_t VSC_EXT_CEA_SDP_CHAINING_SUPPORTED:1;  // bit 7: DP 1.4
+	} bits;
+	uint8_t raw;
+};
+
 union training_aux_rd_interval {
 	struct {
 		uint8_t TRAINIG_AUX_RD_INTERVAL:7;
-		uint8_t EXT_RECIEVER_CAP_FIELD_PRESENT:1;
+		uint8_t EXT_RECEIVER_CAP_FIELD_PRESENT:1;
 	} bits;
 	uint8_t raw;
 };

drivers/gpu/drm/amd/display/dc/dc_helper.c

@@ -51,20 +51,16 @@ static inline void set_reg_field_value_masks(
 	field_value_mask->mask = field_value_mask->mask | mask;
 }
 
-uint32_t generic_reg_update_ex(const struct dc_context *ctx,
-		uint32_t addr, uint32_t reg_val, int n,
+static void set_reg_field_values(struct dc_reg_value_masks *field_value_mask,
+		uint32_t addr, int n,
 		uint8_t shift1, uint32_t mask1, uint32_t field_value1,
-		...)
+		va_list ap)
 {
-	struct dc_reg_value_masks field_value_mask = {0};
 	uint32_t shift, mask, field_value;
 	int i = 1;
 
-	va_list ap;
-	va_start(ap, field_value1);
-
 	/* gather all bits value/mask getting updated in this register */
-	set_reg_field_value_masks(&field_value_mask,
+	set_reg_field_value_masks(field_value_mask,
 			field_value1, mask1, shift1);
 
 	while (i < n) {
@@ -72,10 +68,48 @@ uint32_t generic_reg_update_ex(const struct dc_context *ctx,
 		mask = va_arg(ap, uint32_t);
 		field_value = va_arg(ap, uint32_t);
 
-		set_reg_field_value_masks(&field_value_mask,
+		set_reg_field_value_masks(field_value_mask,
 				field_value, mask, shift);
 		i++;
 	}
+}
+
+uint32_t generic_reg_update_ex(const struct dc_context *ctx,
+		uint32_t addr, int n,
+		uint8_t shift1, uint32_t mask1, uint32_t field_value1,
+		...)
+{
+	struct dc_reg_value_masks field_value_mask = {0};
+	uint32_t reg_val;
+	va_list ap;
+
+	va_start(ap, field_value1);
+
+	set_reg_field_values(&field_value_mask, addr, n, shift1, mask1,
+			field_value1, ap);
+
+	va_end(ap);
+
+	/* mmio write directly */
+	reg_val = dm_read_reg(ctx, addr);
+	reg_val = (reg_val & ~field_value_mask.mask) | field_value_mask.value;
+	dm_write_reg(ctx, addr, reg_val);
+	return reg_val;
+}
+
+uint32_t generic_reg_set_ex(const struct dc_context *ctx,
+		uint32_t addr, uint32_t reg_val, int n,
+		uint8_t shift1, uint32_t mask1, uint32_t field_value1,
+		...)
+{
+	struct dc_reg_value_masks field_value_mask = {0};
+	va_list ap;
+
+	va_start(ap, field_value1);
+
+	set_reg_field_values(&field_value_mask, addr, n, shift1, mask1,
+			field_value1, ap);
+
 	va_end(ap);
 
 
@@ -85,6 +119,24 @@ uint32_t generic_reg_update_ex(const struct dc_context *ctx,
 	return reg_val;
 }
 
+uint32_t dm_read_reg_func(
+	const struct dc_context *ctx,
+	uint32_t address,
+	const char *func_name)
+{
+	uint32_t value;
+#ifdef DM_CHECK_ADDR_0
+	if (address == 0) {
+		DC_ERR("invalid register read; address = 0\n");
+		return 0;
+	}
+#endif
+	value = cgs_read_register(ctx->cgs_device, address);
+	trace_amdgpu_dc_rreg(&ctx->perf_trace->read_count, address, value);
+
+	return value;
+}
+
 uint32_t generic_reg_get(const struct dc_context *ctx, uint32_t addr,
 		uint8_t shift, uint32_t mask, uint32_t *field_value)
 {
@@ -235,7 +287,7 @@ uint32_t generic_reg_get(const struct dc_context *ctx,
 }
 */
 
-uint32_t generic_reg_wait(const struct dc_context *ctx,
+void generic_reg_wait(const struct dc_context *ctx,
 	uint32_t addr, uint32_t shift, uint32_t mask, uint32_t condition_value,
 	unsigned int delay_between_poll_us, unsigned int time_out_num_tries,
 	const char *func_name, int line)
@@ -265,7 +317,7 @@ uint32_t generic_reg_wait(const struct dc_context *ctx,
 				DC_LOG_DC("REG_WAIT taking a while: %dms in %s line:%d\n",
 						delay_between_poll_us * i / 1000,
 						func_name, line);
-			return reg_val;
+			return;
 		}
 	}
 
@@ -275,8 +327,6 @@ uint32_t generic_reg_wait(const struct dc_context *ctx,
 
 	if (!IS_FPGA_MAXIMUS_DC(ctx->dce_environment))
 		BREAK_TO_DEBUGGER();
-
-	return reg_val;
 }
 
 void generic_write_indirect_reg(const struct dc_context *ctx,

drivers/gpu/drm/amd/display/dc/dc_stream.h

@@ -80,6 +80,7 @@ struct dc_stream_state {
 	struct dc_info_packet vrr_infopacket;
 	struct dc_info_packet vsc_infopacket;
 	struct dc_info_packet vsp_infopacket;
+	struct dc_info_packet dpsdp_infopacket;
 
 	struct rect src; /* composition area */
 	struct rect dst; /* stream addressable area */
@@ -221,6 +222,13 @@ struct dc_stream_state *dc_get_stream_at_index(struct dc *dc, uint8_t i);
  */
 uint32_t dc_stream_get_vblank_counter(const struct dc_stream_state *stream);
 
+/*
+ * Send dp sdp message.
+ */
+bool dc_stream_send_dp_sdp(const struct dc_stream_state *stream,
+		const uint8_t *custom_sdp_message,
+		unsigned int sdp_message_size);
+
 /* TODO: Return parsed values rather than direct register read
  * This has a dependency on the caller (amdgpu_display_get_crtc_scanoutpos)
  * being refactored properly to be dce-specific

drivers/gpu/drm/amd/display/dc/dc_types.h

@@ -103,7 +103,7 @@ struct dc_context {
 };
 
 
-#define DC_MAX_EDID_BUFFER_SIZE 512
+#define DC_MAX_EDID_BUFFER_SIZE 1024
 #define EDID_BLOCK_SIZE 128
 #define MAX_SURFACE_NUM 4
 #define NUM_PIXEL_FORMATS 10
@@ -550,9 +550,9 @@ struct psr_config {
 	unsigned char psr_version;
 	unsigned int psr_rfb_setup_time;
 	bool psr_exit_link_training_required;
-
 	bool psr_frame_capture_indication_req;
 	unsigned int psr_sdp_transmit_line_num_deadline;
+	bool allow_smu_optimizations;
 };
 
 union dmcu_psr_level {
@@ -654,6 +654,7 @@ struct psr_context {
 	 * continuing powerd own
 	 */
 	unsigned int frame_delay;
+	bool allow_smu_optimizations;
 };
 
 struct colorspace_transform {

drivers/gpu/drm/amd/display/dc/dce/dce_aux.c

@@ -171,24 +171,24 @@ static void submit_channel_request(
 		 (request->action == I2CAUX_TRANSACTION_ACTION_I2C_WRITE_MOT)));
 	if (REG(AUXN_IMPCAL)) {
 		/* clear_aux_error */
-		REG_UPDATE_SEQ(AUXN_IMPCAL, AUXN_CALOUT_ERROR_AK,
-				1,
-				0);
+		REG_UPDATE_SEQ_2(AUXN_IMPCAL,
+				AUXN_CALOUT_ERROR_AK, 1,
+				AUXN_CALOUT_ERROR_AK, 0);
 
-		REG_UPDATE_SEQ(AUXP_IMPCAL, AUXP_CALOUT_ERROR_AK,
-				1,
-				0);
+		REG_UPDATE_SEQ_2(AUXP_IMPCAL,
+				AUXP_CALOUT_ERROR_AK, 1,
+				AUXP_CALOUT_ERROR_AK, 0);
 
 		/* force_default_calibrate */
-		REG_UPDATE_1BY1_2(AUXN_IMPCAL,
+		REG_UPDATE_SEQ_2(AUXN_IMPCAL,
 				AUXN_IMPCAL_ENABLE, 1,
 				AUXN_IMPCAL_OVERRIDE_ENABLE, 0);
 
 		/* bug? why AUXN update EN and OVERRIDE_EN 1 by 1 while AUX P toggles OVERRIDE? */
 
-		REG_UPDATE_SEQ(AUXP_IMPCAL, AUXP_IMPCAL_OVERRIDE_ENABLE,
-				1,
-				0);
+		REG_UPDATE_SEQ_2(AUXP_IMPCAL,
+				AUXP_IMPCAL_OVERRIDE_ENABLE, 1,
+				AUXP_IMPCAL_OVERRIDE_ENABLE, 0);
 	}
 	/* set the delay and the number of bytes to write */
 
@@ -267,7 +267,7 @@ static int read_channel_reply(struct dce_aux *engine, uint32_t size,
 	if (!bytes_replied)
 		return -1;
 
-	REG_UPDATE_1BY1_3(AUX_SW_DATA,
+	REG_UPDATE_SEQ_3(AUX_SW_DATA,
 			  AUX_SW_INDEX, 0,
 			  AUX_SW_AUTOINCREMENT_DISABLE, 1,
 			  AUX_SW_DATA_RW, 1);
@@ -317,9 +317,10 @@ static enum aux_channel_operation_result get_channel_status(
 	*returned_bytes = 0;
 
 	/* poll to make sure that SW_DONE is asserted */
-	value = REG_WAIT(AUX_SW_STATUS, AUX_SW_DONE, 1,
+	REG_WAIT(AUX_SW_STATUS, AUX_SW_DONE, 1,
 				10, aux110->timeout_period/10);
 
+	value = REG_READ(AUX_SW_STATUS);
 	/* in case HPD is LOW, exit AUX transaction */
 	if ((value & AUX_SW_STATUS__AUX_SW_HPD_DISCON_MASK))
 		return AUX_CHANNEL_OPERATION_FAILED_HPD_DISCON;
@@ -374,7 +375,6 @@ static bool acquire(
 	struct dce_aux *engine,
 	struct ddc *ddc)
 {
-
 	enum gpio_result result;
 
 	if (!is_engine_available(engine))
@@ -439,12 +439,12 @@ static enum i2caux_transaction_action i2caux_action_from_payload(struct aux_payl
 	return I2CAUX_TRANSACTION_ACTION_DP_READ;
 }
 
-int dce_aux_transfer(struct ddc_service *ddc,
-		struct aux_payload *payload)
+int dce_aux_transfer_raw(struct ddc_service *ddc,
+		struct aux_payload *payload,
+		enum aux_channel_operation_result *operation_result)
 {
 	struct ddc *ddc_pin = ddc->ddc_pin;
 	struct dce_aux *aux_engine;
-	enum aux_channel_operation_result operation_result;
 	struct aux_request_transaction_data aux_req;
 	struct aux_reply_transaction_data aux_rep;
 	uint8_t returned_bytes = 0;
@@ -455,7 +455,8 @@ int dce_aux_transfer(struct ddc_service *ddc,
 	memset(&aux_rep, 0, sizeof(aux_rep));
 
 	aux_engine = ddc->ctx->dc->res_pool->engines[ddc_pin->pin_data->en];
-	acquire(aux_engine, ddc_pin);
+	if (!acquire(aux_engine, ddc_pin))
+		return -1;
 
 	if (payload->i2c_over_aux)
 		aux_req.type = AUX_TRANSACTION_TYPE_I2C;
@@ -470,28 +471,26 @@ int dce_aux_transfer(struct ddc_service *ddc,
 	aux_req.data = payload->data;
 
 	submit_channel_request(aux_engine, &aux_req);
-	operation_result = get_channel_status(aux_engine, &returned_bytes);
+	*operation_result = get_channel_status(aux_engine, &returned_bytes);
 
-	switch (operation_result) {
-	case AUX_CHANNEL_OPERATION_SUCCEEDED:
-		res = read_channel_reply(aux_engine, payload->length,
-							payload->data, payload->reply,
-							&status);
-		break;
-	case AUX_CHANNEL_OPERATION_FAILED_HPD_DISCON:
-		res = 0;
-		break;
-	case AUX_CHANNEL_OPERATION_FAILED_REASON_UNKNOWN:
-	case AUX_CHANNEL_OPERATION_FAILED_INVALID_REPLY:
-	case AUX_CHANNEL_OPERATION_FAILED_TIMEOUT:
+	if (*operation_result == AUX_CHANNEL_OPERATION_SUCCEEDED) {
+		read_channel_reply(aux_engine, payload->length,
+					 payload->data, payload->reply,
+					 &status);
+		res = returned_bytes;
+	} else {
 		res = -1;
-		break;
 	}
+
 	release_engine(aux_engine);
 	return res;
 }
 
-#define AUX_RETRY_MAX 7
+#define AUX_MAX_RETRIES 7
+#define AUX_MAX_DEFER_RETRIES 7
+#define AUX_MAX_I2C_DEFER_RETRIES 7
+#define AUX_MAX_INVALID_REPLY_RETRIES 2
+#define AUX_MAX_TIMEOUT_RETRIES 3
 
 bool dce_aux_transfer_with_retries(struct ddc_service *ddc,
 		struct aux_payload *payload)
@@ -499,24 +498,85 @@ bool dce_aux_transfer_with_retries(struct ddc_service *ddc,
 	int i, ret = 0;
 	uint8_t reply;
 	bool payload_reply = true;
+	enum aux_channel_operation_result operation_result;
+	int aux_ack_retries = 0,
+		aux_defer_retries = 0,
+		aux_i2c_defer_retries = 0,
+		aux_timeout_retries = 0,
+		aux_invalid_reply_retries = 0;
 
 	if (!payload->reply) {
 		payload_reply = false;
 		payload->reply = &reply;
 	}
 
-	for (i = 0; i < AUX_RETRY_MAX; i++) {
-		ret = dce_aux_transfer(ddc, payload);
+	for (i = 0; i < AUX_MAX_RETRIES; i++) {
+		ret = dce_aux_transfer_raw(ddc, payload, &operation_result);
+		switch (operation_result) {
+		case AUX_CHANNEL_OPERATION_SUCCEEDED:
+			aux_timeout_retries = 0;
+			aux_invalid_reply_retries = 0;
 
-		if (ret >= 0) {
-			if (*payload->reply == 0) {
-				if (!payload_reply)
-					payload->reply = NULL;
-				return true;
+			switch (*payload->reply) {
+			case AUX_TRANSACTION_REPLY_AUX_ACK:
+				if (!payload->write && payload->length != ret) {
+					if (++aux_ack_retries >= AUX_MAX_RETRIES)
+						goto fail;
+					else
+						udelay(300);
+				} else
+					return true;
+			break;
+
+			case AUX_TRANSACTION_REPLY_AUX_DEFER:
+			case AUX_TRANSACTION_REPLY_I2C_OVER_AUX_NACK:
+			case AUX_TRANSACTION_REPLY_I2C_OVER_AUX_DEFER:
+				if (++aux_defer_retries >= AUX_MAX_DEFER_RETRIES)
+					goto fail;
+				break;
+
+			case AUX_TRANSACTION_REPLY_I2C_DEFER:
+				aux_defer_retries = 0;
+				if (++aux_i2c_defer_retries >= AUX_MAX_I2C_DEFER_RETRIES)
+					goto fail;
+				break;
+
+			case AUX_TRANSACTION_REPLY_AUX_NACK:
+			case AUX_TRANSACTION_REPLY_HPD_DISCON:
+			default:
+				goto fail;
 			}
-		}
+			break;
 
-		udelay(1000);
+		case AUX_CHANNEL_OPERATION_FAILED_INVALID_REPLY:
+			if (++aux_invalid_reply_retries >= AUX_MAX_INVALID_REPLY_RETRIES)
+				goto fail;
+			else
+				udelay(400);
+			break;
+
+		case AUX_CHANNEL_OPERATION_FAILED_TIMEOUT:
+			if (++aux_timeout_retries >= AUX_MAX_TIMEOUT_RETRIES)
+				goto fail;
+			else {
+				/*
+				 * DP 1.4, 2.8.2:  AUX Transaction Response/Reply Timeouts
+				 * According to the DP spec there should be 3 retries total
+				 * with a 400us wait inbetween each. Hardware already waits
+				 * for 550us therefore no wait is required here.
+				 */
+			}
+			break;
+
+		case AUX_CHANNEL_OPERATION_FAILED_HPD_DISCON:
+		case AUX_CHANNEL_OPERATION_FAILED_REASON_UNKNOWN:
+		default:
+			goto fail;
+		}
 	}
+
+fail:
+	if (!payload_reply)
+		payload->reply = NULL;
 	return false;
 }

drivers/gpu/drm/amd/display/dc/dce/dce_aux.h

@@ -123,8 +123,9 @@ bool dce110_aux_engine_acquire(
 	struct dce_aux *aux_engine,
 	struct ddc *ddc);
 
-int dce_aux_transfer(struct ddc_service *ddc,
-		struct aux_payload *cmd);
+int dce_aux_transfer_raw(struct ddc_service *ddc,
+		struct aux_payload *cmd,
+		enum aux_channel_operation_result *operation_result);
 
 bool dce_aux_transfer_with_retries(struct ddc_service *ddc,
 		struct aux_payload *cmd);

drivers/gpu/drm/amd/display/dc/dce/dce_clock_source.c

@@ -978,7 +978,7 @@ static bool dce110_clock_source_power_down(
 }
 
 static bool get_pixel_clk_frequency_100hz(
-		struct clock_source *clock_source,
+		const struct clock_source *clock_source,
 		unsigned int inst,
 		unsigned int *pixel_clk_khz)
 {

drivers/gpu/drm/amd/display/dc/dce/dce_dmcu.c

@@ -213,9 +213,6 @@ static bool dce_dmcu_setup_psr(struct dmcu *dmcu,
 	link->link_enc->funcs->psr_program_secondary_packet(link->link_enc,
 			psr_context->sdpTransmitLineNumDeadline);
 
-	if (psr_context->psr_level.bits.SKIP_SMU_NOTIFICATION)
-		REG_UPDATE(SMU_INTERRUPT_CONTROL, DC_SMU_INT_ENABLE, 1);
-
 	/* waitDMCUReadyForCmd */
 	REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0,
 					dmcu_wait_reg_ready_interval,
@@ -594,7 +591,7 @@ static bool dcn10_dmcu_setup_psr(struct dmcu *dmcu,
 	link->link_enc->funcs->psr_program_secondary_packet(link->link_enc,
 			psr_context->sdpTransmitLineNumDeadline);
 
-	if (psr_context->psr_level.bits.SKIP_SMU_NOTIFICATION)
+	if (psr_context->allow_smu_optimizations)
 		REG_UPDATE(SMU_INTERRUPT_CONTROL, DC_SMU_INT_ENABLE, 1);
 
 	/* waitDMCUReadyForCmd */
@@ -615,6 +612,7 @@ static bool dcn10_dmcu_setup_psr(struct dmcu *dmcu,
 			psr_context->psrFrameCaptureIndicationReq;
 	masterCmdData1.bits.aux_chan = psr_context->channel;
 	masterCmdData1.bits.aux_repeat = psr_context->aux_repeats;
+	masterCmdData1.bits.allow_smu_optimizations = psr_context->allow_smu_optimizations;
 	dm_write_reg(dmcu->ctx, REG(MASTER_COMM_DATA_REG1),
 					masterCmdData1.u32All);
 
@@ -635,6 +633,7 @@ static bool dcn10_dmcu_setup_psr(struct dmcu *dmcu,
 	dm_write_reg(dmcu->ctx, REG(MASTER_COMM_DATA_REG3),
 			masterCmdData3.u32All);
 
+
 	/* setDMCUParam_Cmd */
 	REG_UPDATE(MASTER_COMM_CMD_REG,
 			MASTER_COMM_CMD_REG_BYTE0, PSR_SET);

drivers/gpu/drm/amd/display/dc/dce/dce_dmcu.h

@@ -199,16 +199,16 @@ struct dce_dmcu {
  ******************************************************************/
 union dce_dmcu_psr_config_data_reg1 {
 	struct {
-		unsigned int timehyst_frames:8;    /*[7:0]*/
-		unsigned int hyst_lines:7;         /*[14:8]*/
-		unsigned int rfb_update_auto_en:1; /*[15:15]*/
-		unsigned int dp_port_num:3;        /*[18:16]*/
-		unsigned int dcp_sel:3;            /*[21:19]*/
-		unsigned int phy_type:1;           /*[22:22]*/
-		unsigned int frame_cap_ind:1;      /*[23:23]*/
-		unsigned int aux_chan:3;           /*[26:24]*/
-		unsigned int aux_repeat:4;         /*[30:27]*/
-		unsigned int reserved:1;           /*[31:31]*/
+		unsigned int timehyst_frames:8;                  /*[7:0]*/
+		unsigned int hyst_lines:7;                       /*[14:8]*/
+		unsigned int rfb_update_auto_en:1;               /*[15:15]*/
+		unsigned int dp_port_num:3;                      /*[18:16]*/
+		unsigned int dcp_sel:3;                          /*[21:19]*/
+		unsigned int phy_type:1;                         /*[22:22]*/
+		unsigned int frame_cap_ind:1;                    /*[23:23]*/
+		unsigned int aux_chan:3;                         /*[26:24]*/
+		unsigned int aux_repeat:4;                       /*[30:27]*/
+		unsigned int allow_smu_optimizations:1;         /*[31:31]*/
 	} bits;
 	unsigned int u32All;
 };
@@ -236,7 +236,7 @@ union dce_dmcu_psr_config_data_reg3 {
 	struct {
 		unsigned int psr_level:16;      /*[15:0]*/
 		unsigned int link_rate:4;       /*[19:16]*/
-		unsigned int reserved:12;       /*[31:20]*/
+		unsigned int reserved:12;        /*[31:20]*/
 	} bits;
 	unsigned int u32All;
 };

drivers/gpu/drm/amd/display/dc/dce/dce_i2c_hw.c

@@ -346,6 +346,16 @@ static void release_engine(
 
 }
 
+static bool is_engine_available(struct dce_i2c_hw *dce_i2c_hw)
+{
+	unsigned int arbitrate;
+
+	REG_GET(DC_I2C_ARBITRATION, DC_I2C_REG_RW_CNTL_STATUS, &arbitrate);
+	if (arbitrate == DC_I2C_REG_RW_CNTL_STATUS_DMCU_ONLY)
+		return false;
+	return true;
+}
+
 struct dce_i2c_hw *acquire_i2c_hw_engine(
 	struct resource_pool *pool,
 	struct ddc *ddc)
@@ -368,7 +378,7 @@ struct dce_i2c_hw *acquire_i2c_hw_engine(
 	if (!dce_i2c_hw)
 		return NULL;
 
-	if (pool->i2c_hw_buffer_in_use)
+	if (pool->i2c_hw_buffer_in_use || !is_engine_available(dce_i2c_hw))
 		return NULL;
 
 	do {

drivers/gpu/drm/amd/display/dc/dce/dce_i2c_hw.h

@@ -29,7 +29,8 @@
 enum dc_i2c_status {
 	DC_I2C_STATUS__DC_I2C_STATUS_IDLE,
 	DC_I2C_STATUS__DC_I2C_STATUS_USED_BY_SW,
-	DC_I2C_STATUS__DC_I2C_STATUS_USED_BY_HW
+	DC_I2C_STATUS__DC_I2C_STATUS_USED_BY_HW,
+	DC_I2C_REG_RW_CNTL_STATUS_DMCU_ONLY = 2,
 };
 
 enum dc_i2c_arbitration {
@@ -129,7 +130,8 @@ enum {
 	I2C_SF(DC_I2C_DATA, DC_I2C_DATA, mask_sh),\
 	I2C_SF(DC_I2C_DATA, DC_I2C_INDEX, mask_sh),\
 	I2C_SF(DC_I2C_DATA, DC_I2C_INDEX_WRITE, mask_sh),\
-	I2C_SF(MICROSECOND_TIME_BASE_DIV, XTAL_REF_DIV, mask_sh)
+	I2C_SF(MICROSECOND_TIME_BASE_DIV, XTAL_REF_DIV, mask_sh),\
+	I2C_SF(DC_I2C_ARBITRATION, DC_I2C_REG_RW_CNTL_STATUS, mask_sh)
 
 #define I2C_COMMON_MASK_SH_LIST_DCE110(mask_sh)\
 	I2C_COMMON_MASK_SH_LIST_DCE_COMMON_BASE(mask_sh),\
@@ -170,6 +172,7 @@ struct dce_i2c_shift {
 	uint8_t DC_I2C_INDEX;
 	uint8_t DC_I2C_INDEX_WRITE;
 	uint8_t XTAL_REF_DIV;
+	uint8_t DC_I2C_REG_RW_CNTL_STATUS;
 };
 
 struct dce_i2c_mask {
@@ -207,6 +210,7 @@ struct dce_i2c_mask {
 	uint32_t DC_I2C_INDEX;
 	uint32_t DC_I2C_INDEX_WRITE;
 	uint32_t XTAL_REF_DIV;
+	uint32_t DC_I2C_REG_RW_CNTL_STATUS;
 };
 
 struct dce_i2c_registers {

drivers/gpu/drm/amd/display/dc/dce/dce_stream_encoder.c

@@ -272,7 +272,8 @@ static void dce110_update_hdmi_info_packet(
 static void dce110_stream_encoder_dp_set_stream_attribute(
 	struct stream_encoder *enc,
 	struct dc_crtc_timing *crtc_timing,
-	enum dc_color_space output_color_space)
+	enum dc_color_space output_color_space,
+	uint32_t enable_sdp_splitting)
 {
 #if defined(CONFIG_DRM_AMD_DC_DCN1_0)
 	uint32_t h_active_start;
@@ -977,7 +978,7 @@ static void dce110_stream_encoder_dp_unblank(
 
 		uint64_t m_vid_l = n_vid;
 
-		m_vid_l *= param->pixel_clk_khz;
+		m_vid_l *= param->timing.pix_clk_100hz / 10;
 		m_vid_l = div_u64(m_vid_l,
 			param->link_settings.link_rate
 				* LINK_RATE_REF_FREQ_IN_KHZ);

drivers/gpu/drm/amd/display/dc/dce100/dce100_resource.c

@@ -378,6 +378,11 @@ static const struct resource_caps res_cap = {
 	.num_ddc = 6,
 };
 
+static const struct dc_plane_cap plane_cap = {
+	.type = DC_PLANE_TYPE_DCE_RGB,
+	.supports_argb8888 = true,
+};
+
 #define CTX  ctx
 #define REG(reg) mm ## reg
 
@@ -1023,6 +1028,9 @@ static bool construct(
 
 	dc->caps.max_planes =  pool->base.pipe_count;
 
+	for (i = 0; i < dc->caps.max_planes; ++i)
+		dc->caps.planes[i] = plane_cap;
+
 	if (!resource_construct(num_virtual_links, dc, &pool->base,
 			&res_create_funcs))
 		goto res_create_fail;

drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c

@@ -616,7 +616,7 @@ dce110_set_output_transfer_func(struct pipe_ctx *pipe_ctx,
 
 void dce110_update_info_frame(struct pipe_ctx *pipe_ctx)
 {
-	bool is_hdmi;
+	bool is_hdmi_tmds;
 	bool is_dp;
 
 	ASSERT(pipe_ctx->stream);
@@ -624,13 +624,13 @@ void dce110_update_info_frame(struct pipe_ctx *pipe_ctx)
 	if (pipe_ctx->stream_res.stream_enc == NULL)
 		return;  /* this is not root pipe */
 
-	is_hdmi = dc_is_hdmi_signal(pipe_ctx->stream->signal);
+	is_hdmi_tmds = dc_is_hdmi_tmds_signal(pipe_ctx->stream->signal);
 	is_dp = dc_is_dp_signal(pipe_ctx->stream->signal);
 
-	if (!is_hdmi && !is_dp)
+	if (!is_hdmi_tmds && !is_dp)
 		return;
 
-	if (is_hdmi)
+	if (is_hdmi_tmds)
 		pipe_ctx->stream_res.stream_enc->funcs->update_hdmi_info_packets(
 			pipe_ctx->stream_res.stream_enc,
 			&pipe_ctx->stream_res.encoder_info_frame);
@@ -935,13 +935,31 @@ void hwss_edp_backlight_control(
 		edp_receiver_ready_T9(link);
 }
 
+// Static helper function which calls the correct function
+// based on pp_smu version
+static void set_pme_wa_enable_by_version(struct dc *dc)
+{
+	struct pp_smu_funcs *pp_smu = NULL;
+
+	if (dc->res_pool->pp_smu)
+		pp_smu = dc->res_pool->pp_smu;
+
+	if (pp_smu) {
+		if (pp_smu->ctx.ver == PP_SMU_VER_RV && pp_smu->rv_funcs.set_pme_wa_enable)
+			pp_smu->rv_funcs.set_pme_wa_enable(&(pp_smu->ctx));
+	}
+}
+
 void dce110_enable_audio_stream(struct pipe_ctx *pipe_ctx)
 {
-	struct dc *core_dc = pipe_ctx->stream->ctx->dc;
 	/* notify audio driver for audio modes of monitor */
-	struct pp_smu_funcs_rv *pp_smu = core_dc->res_pool->pp_smu;
+	struct dc *core_dc = pipe_ctx->stream->ctx->dc;
+	struct pp_smu_funcs *pp_smu = NULL;
 	unsigned int i, num_audio = 1;
 
+	if (core_dc->res_pool->pp_smu)
+		pp_smu = core_dc->res_pool->pp_smu;
+
 	if (pipe_ctx->stream_res.audio) {
 		for (i = 0; i < MAX_PIPES; i++) {
 			/*current_state not updated yet*/
@@ -951,30 +969,31 @@ void dce110_enable_audio_stream(struct pipe_ctx *pipe_ctx)
 
 		pipe_ctx->stream_res.audio->funcs->az_enable(pipe_ctx->stream_res.audio);
 
-		if (num_audio >= 1 && pp_smu != NULL && pp_smu->set_pme_wa_enable != NULL)
+		if (num_audio >= 1 && pp_smu != NULL)
 			/*this is the first audio. apply the PME w/a in order to wake AZ from D3*/
-			pp_smu->set_pme_wa_enable(&pp_smu->pp_smu);
+			set_pme_wa_enable_by_version(core_dc);
 		/* un-mute audio */
 		/* TODO: audio should be per stream rather than per link */
 		pipe_ctx->stream_res.stream_enc->funcs->audio_mute_control(
-			pipe_ctx->stream_res.stream_enc, false);
+					pipe_ctx->stream_res.stream_enc, false);
 	}
 }
 
 void dce110_disable_audio_stream(struct pipe_ctx *pipe_ctx, int option)
 {
 	struct dc *dc = pipe_ctx->stream->ctx->dc;
+	struct pp_smu_funcs *pp_smu = NULL;
 
 	pipe_ctx->stream_res.stream_enc->funcs->audio_mute_control(
 			pipe_ctx->stream_res.stream_enc, true);
 	if (pipe_ctx->stream_res.audio) {
-		struct pp_smu_funcs_rv *pp_smu = dc->res_pool->pp_smu;
+		if (dc->res_pool->pp_smu)
+			pp_smu = dc->res_pool->pp_smu;
 
 		if (option != KEEP_ACQUIRED_RESOURCE ||
-				!dc->debug.az_endpoint_mute_only) {
+				!dc->debug.az_endpoint_mute_only)
 			/*only disalbe az_endpoint if power down or free*/
 			pipe_ctx->stream_res.audio->funcs->az_disable(pipe_ctx->stream_res.audio);
-		}
 
 		if (dc_is_dp_signal(pipe_ctx->stream->signal))
 			pipe_ctx->stream_res.stream_enc->funcs->dp_audio_disable(
@@ -989,9 +1008,9 @@ void dce110_disable_audio_stream(struct pipe_ctx *pipe_ctx, int option)
 			update_audio_usage(&dc->current_state->res_ctx, dc->res_pool, pipe_ctx->stream_res.audio, false);
 			pipe_ctx->stream_res.audio = NULL;
 		}
-		if (pp_smu != NULL && pp_smu->set_pme_wa_enable != NULL)
+		if (pp_smu != NULL)
 			/*this is the first audio. apply the PME w/a in order to wake AZ from D3*/
-			pp_smu->set_pme_wa_enable(&pp_smu->pp_smu);
+			set_pme_wa_enable_by_version(dc);
 
 		/* TODO: notify audio driver for if audio modes list changed
 		 * add audio mode list change flag */
@@ -1007,7 +1026,7 @@ void dce110_disable_stream(struct pipe_ctx *pipe_ctx, int option)
 	struct dc_link *link = stream->link;
 	struct dc *dc = pipe_ctx->stream->ctx->dc;
 
-	if (dc_is_hdmi_signal(pipe_ctx->stream->signal))
+	if (dc_is_hdmi_tmds_signal(pipe_ctx->stream->signal))
 		pipe_ctx->stream_res.stream_enc->funcs->stop_hdmi_info_packets(
 			pipe_ctx->stream_res.stream_enc);
 
@@ -1032,7 +1051,7 @@ void dce110_unblank_stream(struct pipe_ctx *pipe_ctx,
 	struct dc_link *link = stream->link;
 
 	/* only 3 items below are used by unblank */
-	params.pixel_clk_khz = pipe_ctx->stream->timing.pix_clk_100hz / 10;
+	params.timing = pipe_ctx->stream->timing;
 	params.link_settings.link_rate = link_settings->link_rate;
 
 	if (dc_is_dp_signal(pipe_ctx->stream->signal))
@@ -1349,7 +1368,7 @@ static enum dc_status apply_single_controller_ctx_to_hw(
 		pipe_ctx->stream_res.tg->funcs->set_static_screen_control(
 				pipe_ctx->stream_res.tg, event_triggers);
 
-	if (pipe_ctx->stream->signal != SIGNAL_TYPE_VIRTUAL)
+	if (!dc_is_virtual_signal(pipe_ctx->stream->signal))
 		pipe_ctx->stream_res.stream_enc->funcs->dig_connect_to_otg(
 			pipe_ctx->stream_res.stream_enc,
 			pipe_ctx->stream_res.tg->inst);
@@ -1358,7 +1377,7 @@ static enum dc_status apply_single_controller_ctx_to_hw(
 			pipe_ctx->stream_res.opp,
 			COLOR_SPACE_YCBCR601,
 			stream->timing.display_color_depth,
-			pipe_ctx->stream->signal);
+			stream->signal);
 
 	pipe_ctx->stream_res.opp->funcs->opp_program_fmt(
 		pipe_ctx->stream_res.opp,
@@ -1532,6 +1551,9 @@ void dce110_enable_accelerated_mode(struct dc *dc, struct dc_state *context)
 		}
 	}
 
+	if (dc->hwss.init_pipes)
+		dc->hwss.init_pipes(dc, context);
+
 	if (edp_link) {
 		/* this seems to cause blank screens on DCE8 */
 		if ((dc->ctx->dce_version == DCE_VERSION_8_0) ||
@@ -2612,7 +2634,7 @@ void dce110_set_cursor_position(struct pipe_ctx *pipe_ctx)
 	struct mem_input *mi = pipe_ctx->plane_res.mi;
 	struct dc_cursor_mi_param param = {
 		.pixel_clk_khz = pipe_ctx->stream->timing.pix_clk_100hz / 10,
-		.ref_clk_khz = pipe_ctx->stream->ctx->dc->res_pool->ref_clock_inKhz,
+		.ref_clk_khz = pipe_ctx->stream->ctx->dc->res_pool->ref_clocks.xtalin_clock_inKhz,
 		.viewport = pipe_ctx->plane_res.scl_data.viewport,
 		.h_scale_ratio = pipe_ctx->plane_res.scl_data.ratios.horz,
 		.v_scale_ratio = pipe_ctx->plane_res.scl_data.ratios.vert,

drivers/gpu/drm/amd/display/dc/dce110/dce110_resource.c

@@ -392,6 +392,21 @@ static const struct resource_caps stoney_resource_cap = {
 		.num_ddc = 3,
 };
 
+static const struct dc_plane_cap plane_cap = {
+		.type = DC_PLANE_TYPE_DCE_RGB,
+		.blends_with_below = true,
+		.blends_with_above = true,
+		.per_pixel_alpha = 1,
+		.supports_argb8888 = true,
+};
+
+static const struct dc_plane_cap underlay_plane_cap = {
+		.type = DC_PLANE_TYPE_DCE_UNDERLAY,
+		.blends_with_above = true,
+		.per_pixel_alpha = 1,
+		.supports_nv12 = true
+};
+
 #define CTX  ctx
 #define REG(reg) mm ## reg
 
@@ -1371,6 +1386,11 @@ static bool construct(
 
 	dc->caps.max_planes =  pool->base.pipe_count;
 
+	for (i = 0; i < pool->base.underlay_pipe_index; ++i)
+		dc->caps.planes[i] = plane_cap;
+
+	dc->caps.planes[pool->base.underlay_pipe_index] = underlay_plane_cap;
+
 	bw_calcs_init(dc->bw_dceip, dc->bw_vbios, dc->ctx->asic_id);
 
 	bw_calcs_data_update_from_pplib(dc);

drivers/gpu/drm/amd/display/dc/dce112/dce112_resource.c

@@ -397,6 +397,11 @@ static const struct resource_caps polaris_11_resource_cap = {
 		.num_ddc = 5,
 };
 
+static const struct dc_plane_cap plane_cap = {
+	.type = DC_PLANE_TYPE_DCE_RGB,
+	.supports_argb8888 = true,
+};
+
 #define CTX  ctx
 #define REG(reg) mm ## reg
 
@@ -887,7 +892,7 @@ enum dc_status resource_map_phy_clock_resources(
 		return DC_ERROR_UNEXPECTED;
 
 	if (dc_is_dp_signal(pipe_ctx->stream->signal)
-		|| pipe_ctx->stream->signal == SIGNAL_TYPE_VIRTUAL)
+		|| dc_is_virtual_signal(pipe_ctx->stream->signal))
 		pipe_ctx->clock_source =
 				dc->res_pool->dp_clock_source;
 	else
@@ -1310,6 +1315,9 @@ static bool construct(
 
 	dc->caps.max_planes =  pool->base.pipe_count;
 
+	for (i = 0; i < dc->caps.max_planes; ++i)
+		dc->caps.planes[i] = plane_cap;
+
 	/* Create hardware sequencer */
 	dce112_hw_sequencer_construct(dc);