drm/amdkfd: Implement doorbell allocation for SOC15

Allocate doorbells according to the doorbell routing information on
SOC15 ASICs (Vega10 and later). On older ASICs we continue to use the
queue_id as the doorbell ID to maintain compatibility with the Thunk.

Signed-off-by: Shaoyun Liu <Shaoyun.Liu@amd.com>
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

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

@@ -295,6 +295,13 @@ static int kfd_ioctl_create_queue(struct file *filep, struct kfd_process *p,
 	args->doorbell_offset = KFD_MMAP_TYPE_DOORBELL;
 	args->doorbell_offset |= KFD_MMAP_GPU_ID(args->gpu_id);
 	args->doorbell_offset <<= PAGE_SHIFT;
+	if (KFD_IS_SOC15(dev->device_info->asic_family))
+		/* On SOC15 ASICs, doorbell allocation must be
+		 * per-device, and independent from the per-process
+		 * queue_id. Return the doorbell offset within the
+		 * doorbell aperture to user mode.
+		 */
+		args->doorbell_offset |= q_properties.doorbell_off;
 
 	mutex_unlock(&p->mutex);
 

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

@@ -110,6 +110,57 @@ void program_sh_mem_settings(struct device_queue_manager *dqm,
 						qpd->sh_mem_bases);
 }
 
+static int allocate_doorbell(struct qcm_process_device *qpd, struct queue *q)
+{
+	struct kfd_dev *dev = qpd->dqm->dev;
+
+	if (!KFD_IS_SOC15(dev->device_info->asic_family)) {
+		/* On pre-SOC15 chips we need to use the queue ID to
+		 * preserve the user mode ABI.
+		 */
+		q->doorbell_id = q->properties.queue_id;
+	} else if (q->properties.type == KFD_QUEUE_TYPE_SDMA) {
+		/* For SDMA queues on SOC15, use static doorbell
+		 * assignments based on the engine and queue.
+		 */
+		q->doorbell_id = dev->shared_resources.sdma_doorbell
+			[q->properties.sdma_engine_id]
+			[q->properties.sdma_queue_id];
+	} else {
+		/* For CP queues on SOC15 reserve a free doorbell ID */
+		unsigned int found;
+
+		found = find_first_zero_bit(qpd->doorbell_bitmap,
+					    KFD_MAX_NUM_OF_QUEUES_PER_PROCESS);
+		if (found >= KFD_MAX_NUM_OF_QUEUES_PER_PROCESS) {
+			pr_debug("No doorbells available");
+			return -EBUSY;
+		}
+		set_bit(found, qpd->doorbell_bitmap);
+		q->doorbell_id = found;
+	}
+
+	q->properties.doorbell_off =
+		kfd_doorbell_id_to_offset(dev, q->process,
+					  q->doorbell_id);
+
+	return 0;
+}
+
+static void deallocate_doorbell(struct qcm_process_device *qpd,
+				struct queue *q)
+{
+	unsigned int old;
+	struct kfd_dev *dev = qpd->dqm->dev;
+
+	if (!KFD_IS_SOC15(dev->device_info->asic_family) ||
+	    q->properties.type == KFD_QUEUE_TYPE_SDMA)
+		return;
+
+	old = test_and_clear_bit(q->doorbell_id, qpd->doorbell_bitmap);
+	WARN_ON(!old);
+}
+
 static int allocate_vmid(struct device_queue_manager *dqm,
 			struct qcm_process_device *qpd,
 			struct queue *q)
@@ -301,10 +352,14 @@ static int create_compute_queue_nocpsch(struct device_queue_manager *dqm,
 	if (retval)
 		return retval;
 
+	retval = allocate_doorbell(qpd, q);
+	if (retval)
+		goto out_deallocate_hqd;
+
 	retval = mqd->init_mqd(mqd, &q->mqd, &q->mqd_mem_obj,
 				&q->gart_mqd_addr, &q->properties);
 	if (retval)
-		goto out_deallocate_hqd;
+		goto out_deallocate_doorbell;
 
 	pr_debug("Loading mqd to hqd on pipe %d, queue %d\n",
 			q->pipe, q->queue);
@@ -324,6 +379,8 @@ static int create_compute_queue_nocpsch(struct device_queue_manager *dqm,
 
 out_uninit_mqd:
 	mqd->uninit_mqd(mqd, q->mqd, q->mqd_mem_obj);
+out_deallocate_doorbell:
+	deallocate_doorbell(qpd, q);
 out_deallocate_hqd:
 	deallocate_hqd(dqm, q);
 
@@ -357,6 +414,8 @@ static int destroy_queue_nocpsch_locked(struct device_queue_manager *dqm,
 	}
 	dqm->total_queue_count--;
 
+	deallocate_doorbell(qpd, q);
+
 	retval = mqd->destroy_mqd(mqd, q->mqd,
 				KFD_PREEMPT_TYPE_WAVEFRONT_RESET,
 				KFD_UNMAP_LATENCY_MS,
@@ -861,6 +920,10 @@ static int create_sdma_queue_nocpsch(struct device_queue_manager *dqm,
 	q->properties.sdma_queue_id = q->sdma_id / CIK_SDMA_QUEUES_PER_ENGINE;
 	q->properties.sdma_engine_id = q->sdma_id % CIK_SDMA_QUEUES_PER_ENGINE;
 
+	retval = allocate_doorbell(qpd, q);
+	if (retval)
+		goto out_deallocate_sdma_queue;
+
 	pr_debug("SDMA id is:    %d\n", q->sdma_id);
 	pr_debug("SDMA queue id: %d\n", q->properties.sdma_queue_id);
 	pr_debug("SDMA engine id: %d\n", q->properties.sdma_engine_id);
@@ -869,7 +932,7 @@ static int create_sdma_queue_nocpsch(struct device_queue_manager *dqm,
 	retval = mqd->init_mqd(mqd, &q->mqd, &q->mqd_mem_obj,
 				&q->gart_mqd_addr, &q->properties);
 	if (retval)
-		goto out_deallocate_sdma_queue;
+		goto out_deallocate_doorbell;
 
 	retval = mqd->load_mqd(mqd, q->mqd, 0, 0, &q->properties, NULL);
 	if (retval)
@@ -879,6 +942,8 @@ static int create_sdma_queue_nocpsch(struct device_queue_manager *dqm,
 
 out_uninit_mqd:
 	mqd->uninit_mqd(mqd, q->mqd, q->mqd_mem_obj);
+out_deallocate_doorbell:
+	deallocate_doorbell(qpd, q);
 out_deallocate_sdma_queue:
 	deallocate_sdma_queue(dqm, q->sdma_id);
 
@@ -1070,12 +1135,17 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q,
 		q->properties.sdma_engine_id =
 			q->sdma_id % CIK_SDMA_QUEUES_PER_ENGINE;
 	}
+
+	retval = allocate_doorbell(qpd, q);
+	if (retval)
+		goto out_deallocate_sdma_queue;
+
 	mqd = dqm->ops.get_mqd_manager(dqm,
 			get_mqd_type_from_queue_type(q->properties.type));
 
 	if (!mqd) {
 		retval = -ENOMEM;
-		goto out_deallocate_sdma_queue;
+		goto out_deallocate_doorbell;
 	}
 	/*
 	 * Eviction state logic: we only mark active queues as evicted
@@ -1093,7 +1163,7 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q,
 	retval = mqd->init_mqd(mqd, &q->mqd, &q->mqd_mem_obj,
 				&q->gart_mqd_addr, &q->properties);
 	if (retval)
-		goto out_deallocate_sdma_queue;
+		goto out_deallocate_doorbell;
 
 	list_add(&q->list, &qpd->queues_list);
 	qpd->queue_count++;
@@ -1117,6 +1187,8 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q,
 	mutex_unlock(&dqm->lock);
 	return retval;
 
+out_deallocate_doorbell:
+	deallocate_doorbell(qpd, q);
 out_deallocate_sdma_queue:
 	if (q->properties.type == KFD_QUEUE_TYPE_SDMA)
 		deallocate_sdma_queue(dqm, q->sdma_id);
@@ -1257,6 +1329,8 @@ static int destroy_queue_cpsch(struct device_queue_manager *dqm,
 		goto failed;
 	}
 
+	deallocate_doorbell(qpd, q);
+
 	if (q->properties.type == KFD_QUEUE_TYPE_SDMA) {
 		dqm->sdma_queue_count--;
 		deallocate_sdma_queue(dqm, q->sdma_id);

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

@@ -49,7 +49,7 @@ static unsigned int max_doorbell_slices;
  */
 
 /* # of doorbell bytes allocated for each process. */
-static size_t kfd_doorbell_process_slice(struct kfd_dev *kfd)
+size_t kfd_doorbell_process_slice(struct kfd_dev *kfd)
 {
 	return roundup(kfd->device_info->doorbell_size *
 			KFD_MAX_NUM_OF_QUEUES_PER_PROCESS,
@@ -214,13 +214,9 @@ void write_kernel_doorbell(void __iomem *db, u32 value)
 	}
 }
 
-/*
- * queue_ids are in the range [0,MAX_PROCESS_QUEUES) and are mapped 1:1
- * to doorbells with the process's doorbell page
- */
-unsigned int kfd_queue_id_to_doorbell(struct kfd_dev *kfd,
+unsigned int kfd_doorbell_id_to_offset(struct kfd_dev *kfd,
 					struct kfd_process *process,
-					unsigned int queue_id)
+					unsigned int doorbell_id)
 {
 	/*
 	 * doorbell_id_offset accounts for doorbells taken by KGD.
@@ -231,7 +227,7 @@ unsigned int kfd_queue_id_to_doorbell(struct kfd_dev *kfd,
 	return kfd->doorbell_id_offset +
 		process->doorbell_index
 		* kfd_doorbell_process_slice(kfd) / sizeof(u32) +
-		queue_id * kfd->device_info->doorbell_size / sizeof(u32);
+		doorbell_id * kfd->device_info->doorbell_size / sizeof(u32);
 }
 
 uint64_t kfd_get_number_elems(struct kfd_dev *kfd)

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

@@ -169,6 +169,8 @@ enum cache_policy {
 	cache_policy_noncoherent
 };
 
+#define KFD_IS_SOC15(chip) ((chip) >= CHIP_VEGA10)
+
 struct kfd_event_interrupt_class {
 	bool (*interrupt_isr)(struct kfd_dev *dev,
 				const uint32_t *ih_ring_entry);
@@ -449,6 +451,7 @@ struct queue {
 	uint32_t queue;
 
 	unsigned int sdma_id;
+	unsigned int doorbell_id;
 
 	struct kfd_process	*process;
 	struct kfd_dev		*device;
@@ -523,6 +526,9 @@ struct qcm_process_device {
 	/* IB memory */
 	uint64_t ib_base;
 	void *ib_kaddr;
+
+	/* doorbell resources per process per device */
+	unsigned long *doorbell_bitmap;
 };
 
 /* KFD Memory Eviction */
@@ -747,6 +753,7 @@ unsigned int kfd_pasid_alloc(void);
 void kfd_pasid_free(unsigned int pasid);
 
 /* Doorbells */
+size_t kfd_doorbell_process_slice(struct kfd_dev *kfd);
 int kfd_doorbell_init(struct kfd_dev *kfd);
 void kfd_doorbell_fini(struct kfd_dev *kfd);
 int kfd_doorbell_mmap(struct kfd_dev *dev, struct kfd_process *process,
@@ -756,9 +763,9 @@ void __iomem *kfd_get_kernel_doorbell(struct kfd_dev *kfd,
 void kfd_release_kernel_doorbell(struct kfd_dev *kfd, u32 __iomem *db_addr);
 u32 read_kernel_doorbell(u32 __iomem *db);
 void write_kernel_doorbell(void __iomem *db, u32 value);
-unsigned int kfd_queue_id_to_doorbell(struct kfd_dev *kfd,
+unsigned int kfd_doorbell_id_to_offset(struct kfd_dev *kfd,
 					struct kfd_process *process,
-					unsigned int queue_id);
+					unsigned int doorbell_id);
 phys_addr_t kfd_get_process_doorbells(struct kfd_dev *dev,
 					struct kfd_process *process);
 int kfd_alloc_process_doorbells(struct kfd_process *process);

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

@@ -332,6 +332,7 @@ static void kfd_process_destroy_pdds(struct kfd_process *p)
 			free_pages((unsigned long)pdd->qpd.cwsr_kaddr,
 				get_order(KFD_CWSR_TBA_TMA_SIZE));
 
+		kfree(pdd->qpd.doorbell_bitmap);
 		idr_destroy(&pdd->alloc_idr);
 
 		kfree(pdd);
@@ -586,6 +587,31 @@ err_alloc_process:
 	return ERR_PTR(err);
 }
 
+static int init_doorbell_bitmap(struct qcm_process_device *qpd,
+			struct kfd_dev *dev)
+{
+	unsigned int i;
+
+	if (!KFD_IS_SOC15(dev->device_info->asic_family))
+		return 0;
+
+	qpd->doorbell_bitmap =
+		kzalloc(DIV_ROUND_UP(KFD_MAX_NUM_OF_QUEUES_PER_PROCESS,
+				     BITS_PER_BYTE), GFP_KERNEL);
+	if (!qpd->doorbell_bitmap)
+		return -ENOMEM;
+
+	/* Mask out any reserved doorbells */
+	for (i = 0; i < KFD_MAX_NUM_OF_QUEUES_PER_PROCESS; i++)
+		if ((dev->shared_resources.reserved_doorbell_mask & i) ==
+		    dev->shared_resources.reserved_doorbell_val) {
+			set_bit(i, qpd->doorbell_bitmap);
+			pr_debug("reserved doorbell 0x%03x\n", i);
+		}
+
+	return 0;
+}
+
 struct kfd_process_device *kfd_get_process_device_data(struct kfd_dev *dev,
 							struct kfd_process *p)
 {
@@ -607,6 +633,12 @@ struct kfd_process_device *kfd_create_process_device_data(struct kfd_dev *dev,
 	if (!pdd)
 		return NULL;
 
+	if (init_doorbell_bitmap(&pdd->qpd, dev)) {
+		pr_err("Failed to init doorbell for process\n");
+		kfree(pdd);
+		return NULL;
+	}
+
 	pdd->dev = dev;
 	INIT_LIST_HEAD(&pdd->qpd.queues_list);
 	INIT_LIST_HEAD(&pdd->qpd.priv_queue_list);

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

@@ -119,9 +119,6 @@ static int create_cp_queue(struct process_queue_manager *pqm,
 	/* Doorbell initialized in user space*/
 	q_properties->doorbell_ptr = NULL;
 
-	q_properties->doorbell_off =
-			kfd_queue_id_to_doorbell(dev, pqm->process, qid);
-
 	/* let DQM handle it*/
 	q_properties->vmid = 0;
 	q_properties->queue_id = qid;
@@ -248,6 +245,15 @@ int pqm_create_queue(struct process_queue_manager *pqm,
 		goto err_create_queue;
 	}
 
+	if (q)
+		/* Return the doorbell offset within the doorbell page
+		 * to the caller so it can be passed up to user mode
+		 * (in bytes).
+		 */
+		properties->doorbell_off =
+			(q->properties.doorbell_off * sizeof(uint32_t)) &
+			(kfd_doorbell_process_slice(dev) - 1);
+
 	pr_debug("PQM After DQM create queue\n");
 
 	list_add(&pqn->process_queue_list, &pqm->queues);