drm/amd: Interface change to support 64 bit page_table_base

amdgpu_gpuvm_get_process_page_dir should return the page table address
in the format expected by the pm4_map_process packet for all ASIC
generations.

Signed-off-by: Shaoyun Liu <Shaoyun.Liu@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>

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

@@ -174,7 +174,7 @@ void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev,
 				struct amdgpu_vm *vm);
 void amdgpu_amdkfd_gpuvm_destroy_process_vm(struct kgd_dev *kgd, void *vm);
 void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *vm);
-uint32_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm);
+uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm);
 int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 		struct kgd_dev *kgd, uint64_t va, uint64_t size,
 		void *vm, struct kgd_mem **mem,

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

@@ -142,7 +142,7 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type);
 static void set_scratch_backing_va(struct kgd_dev *kgd,
 					uint64_t va, uint32_t vmid);
 static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-		uint32_t page_table_base);
+		uint64_t page_table_base);
 static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid);
 static int invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid);
 static uint32_t read_vmid_from_vmfault_reg(struct kgd_dev *kgd);
@@ -874,7 +874,7 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type)
 }
 
 static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-			uint32_t page_table_base)
+			uint64_t page_table_base)
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 
@@ -882,7 +882,8 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
 		pr_err("trying to set page table base for wrong VMID\n");
 		return;
 	}
-	WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vmid - 8, page_table_base);
+	WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vmid - 8,
+		lower_32_bits(page_table_base));
 }
 
 static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid)

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

@@ -98,7 +98,7 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type);
 static void set_scratch_backing_va(struct kgd_dev *kgd,
 					uint64_t va, uint32_t vmid);
 static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-		uint32_t page_table_base);
+		uint64_t page_table_base);
 static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid);
 static int invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid);
 
@@ -833,7 +833,7 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type)
 }
 
 static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-		uint32_t page_table_base)
+		uint64_t page_table_base)
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
 
@@ -841,7 +841,8 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
 		pr_err("trying to set page table base for wrong VMID\n");
 		return;
 	}
-	WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vmid - 8, page_table_base);
+	WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vmid - 8,
+			lower_32_bits(page_table_base));
 }
 
 static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid)

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c

@@ -138,7 +138,7 @@ static bool get_atc_vmid_pasid_mapping_valid(struct kgd_dev *kgd,
 static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd,
 		uint8_t vmid);
 static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-		uint32_t page_table_base);
+		uint64_t page_table_base);
 static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type);
 static void set_scratch_backing_va(struct kgd_dev *kgd,
 					uint64_t va, uint32_t vmid);
@@ -1013,11 +1013,10 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type)
 }
 
 static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
-		uint32_t page_table_base)
+		uint64_t page_table_base)
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
-	uint64_t base = (uint64_t)page_table_base << PAGE_SHIFT |
-		AMDGPU_PTE_VALID;
+	uint64_t base = page_table_base | AMDGPU_PTE_VALID;
 
 	if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) {
 		pr_err("trying to set page table base for wrong VMID %u\n",

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

@@ -1131,11 +1131,15 @@ void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *vm)
 	amdgpu_vm_release_compute(adev, avm);
 }
 
-uint32_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm)
+uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm)
 {
 	struct amdgpu_vm *avm = (struct amdgpu_vm *)vm;
+	struct amdgpu_bo *pd = avm->root.base.bo;
+	struct amdgpu_device *adev = amdgpu_ttm_adev(pd->tbo.bdev);
 
-	return avm->pd_phys_addr >> AMDGPU_GPU_PAGE_SHIFT;
+	if (adev->asic_type < CHIP_VEGA10)
+		return avm->pd_phys_addr >> AMDGPU_GPU_PAGE_SHIFT;
+	return avm->pd_phys_addr;
 }
 
 int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(

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

@@ -656,7 +656,7 @@ static int restore_process_queues_nocpsch(struct device_queue_manager *dqm,
 	struct queue *q;
 	struct mqd_manager *mqd_mgr;
 	struct kfd_process_device *pdd;
-	uint32_t pd_base;
+	uint64_t pd_base;
 	int retval = 0;
 
 	pdd = qpd_to_pdd(qpd);
@@ -676,7 +676,7 @@ static int restore_process_queues_nocpsch(struct device_queue_manager *dqm,
 
 	/* Update PD Base in QPD */
 	qpd->page_table_base = pd_base;
-	pr_debug("Updated PD address to 0x%08x\n", pd_base);
+	pr_debug("Updated PD address to 0x%llx\n", pd_base);
 
 	if (!list_empty(&qpd->queues_list)) {
 		dqm->dev->kfd2kgd->set_vm_context_page_table_base(
@@ -717,7 +717,7 @@ static int restore_process_queues_cpsch(struct device_queue_manager *dqm,
 {
 	struct queue *q;
 	struct kfd_process_device *pdd;
-	uint32_t pd_base;
+	uint64_t pd_base;
 	int retval = 0;
 
 	pdd = qpd_to_pdd(qpd);
@@ -737,7 +737,7 @@ static int restore_process_queues_cpsch(struct device_queue_manager *dqm,
 
 	/* Update PD Base in QPD */
 	qpd->page_table_base = pd_base;
-	pr_debug("Updated PD address to 0x%08x\n", pd_base);
+	pr_debug("Updated PD address to 0x%llx\n", pd_base);
 
 	/* activate all active queues on the qpd */
 	list_for_each_entry(q, &qpd->queues_list, list) {
@@ -761,7 +761,7 @@ static int register_process(struct device_queue_manager *dqm,
 {
 	struct device_process_node *n;
 	struct kfd_process_device *pdd;
-	uint32_t pd_base;
+	uint64_t pd_base;
 	int retval;
 
 	n = kzalloc(sizeof(*n), GFP_KERNEL);
@@ -779,6 +779,7 @@ static int register_process(struct device_queue_manager *dqm,
 
 	/* Update PD Base in QPD */
 	qpd->page_table_base = pd_base;
+	pr_debug("Updated PD address to 0x%llx\n", pd_base);
 
 	retval = dqm->asic_ops.update_qpd(dqm, qpd);
 

drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_v9.c

@@ -71,8 +71,7 @@ static int pm_map_process_v9(struct packet_manager *pm,
 		uint32_t *buffer, struct qcm_process_device *qpd)
 {
 	struct pm4_mes_map_process *packet;
-	uint64_t vm_page_table_base_addr =
-		(uint64_t)(qpd->page_table_base) << 12;
+	uint64_t vm_page_table_base_addr = qpd->page_table_base;
 
 	packet = (struct pm4_mes_map_process *)buffer;
 	memset(buffer, 0, sizeof(struct pm4_mes_map_process));

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

@@ -507,11 +507,11 @@ struct qcm_process_device {
 	 * All the memory management data should be here too
 	 */
 	uint64_t gds_context_area;
+	uint64_t page_table_base;
 	uint32_t sh_mem_config;
 	uint32_t sh_mem_bases;
 	uint32_t sh_mem_ape1_base;
 	uint32_t sh_mem_ape1_limit;
-	uint32_t page_table_base;
 	uint32_t gds_size;
 	uint32_t num_gws;
 	uint32_t num_oac;

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

@@ -409,9 +409,9 @@ struct kfd2kgd_calls {
 			struct dma_fence **ef);
 	void (*destroy_process_vm)(struct kgd_dev *kgd, void *vm);
 	void (*release_process_vm)(struct kgd_dev *kgd, void *vm);
-	uint32_t (*get_process_page_dir)(void *vm);
+	uint64_t (*get_process_page_dir)(void *vm);
 	void (*set_vm_context_page_table_base)(struct kgd_dev *kgd,
-			uint32_t vmid, uint32_t page_table_base);
+			uint32_t vmid, uint64_t page_table_base);
 	int (*alloc_memory_of_gpu)(struct kgd_dev *kgd, uint64_t va,
 			uint64_t size, void *vm,
 			struct kgd_mem **mem, uint64_t *offset,