video: tegra: host: dla: manage operation desc

- pin mapped operation descriptor buffers during task submission
- get operation descriptors handle from user and pass its IOVA to engine
- pin API returns IOVA for given mem handle
- unpin operation descriptors buffers in task cleanup

Jira DLA-93

Change-Id: I78fb22301ab472685c3bae7c424d75140b814887
Signed-off-by: Shridhar Rasal <srasal@nvidia.com>
Reviewed-on: http://git-master/r/1213761
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Prashant Gaikwad <pgaikwad@nvidia.com>
GVS: Gerrit_Virtual_Submit

drivers/video/tegra/host/nvdla/nvdla.h

@@ -22,6 +22,7 @@
 #define __NVHOST_NVDLA_H__
 
 #include <linux/nvhost_nvdla_ioctl.h>
+#include "nvhost_buffer.h"
 
 /**
  * Method ID and Method data THI registers
@@ -30,6 +31,7 @@
 #define NV_DLA_THI_METHOD_DATA	0x00000044      /* RW-4R */
 
 #define MAX_NUM_ACTION_LIST	1
+#define MAX_HANDLE_PER_OP_DESC	1
 
 /* TODO: Below should come from firmware interface */
 #define ACTION_OPCODE_TERMINATE		0x00
@@ -78,6 +80,7 @@ struct nvdla_task_fence {
  * struct nvdla_task:	structure for task info
  *
  * @queue		Queue in which task submitted
+ * @buffers		nvhost buffers for priv/task
  * @sp			pointer to syncpt
  * @prefences		pointer to prefences
  * @postfences		pointer to post fences
@@ -88,9 +91,12 @@ struct nvdla_task_fence {
  * @task_desc_pa	DLA task desc PA
  * @buf_size		Total size of task dma alloc
  * @timeout		max timeout to wait for task completion
+ * @op_handle		pointer to handle list of operation descriptor
+ *
  */
 struct nvdla_task {
 	struct nvhost_queue *queue;
+	struct nvhost_buffers *buffers;
 	struct nvhost_syncpt *sp;
 	struct nvdla_task_fence *prefences;
 	struct nvdla_task_fence *postfences;
@@ -101,6 +107,7 @@ struct nvdla_task {
 	dma_addr_t task_desc_pa;
 	size_t buf_size;
 	int timeout;
+	u32 *op_handle;
 };
 
 extern const struct file_operations tegra_nvdla_ctrl_ops;
@@ -184,6 +191,7 @@ void nvdla_task_get(struct nvdla_task *task);
  * nvdla_task_alloc()	allocate task for a give queue
  *
  * @queue		Pointer to nvhost queue
+ * @buffers		Pointer to nvhost buffers
  * @user_task		Pointer to user task passed from UMD
  *
  * Return		allocated task in success, otherwise pointer to err
@@ -192,6 +200,7 @@ void nvdla_task_get(struct nvdla_task *task);
  * parameter detais
  */
 struct nvdla_task *nvdla_task_alloc(struct nvhost_queue *queue,
+			struct nvhost_buffers *buffers,
 			struct nvdla_ctrl_ioctl_submit_task user_task);
 
 /**

drivers/video/tegra/host/nvdla/nvdla_ioctl.c

@@ -184,6 +184,7 @@ static int nvdla_ctrl_submit(struct nvdla_private *priv, void *arg)
 	struct nvdla_ctrl_ioctl_submit_task *local_tasks;
 	struct platform_device *pdev;
 	struct nvhost_queue *queue;
+	struct nvhost_buffers *buffers;
 	u32 num_tasks;
 	struct nvdla_task *task;
 	int err = 0, i = 0;
@@ -196,6 +197,8 @@ static int nvdla_ctrl_submit(struct nvdla_private *priv, void *arg)
 	if (!queue)
 		return -EINVAL;
 
+	buffers = priv->buffers;
+
 	user_tasks = (struct nvdla_ctrl_ioctl_submit_task __user *)
 			(uintptr_t)args->tasks;
 	num_tasks = args->num_tasks;
@@ -222,7 +225,7 @@ static int nvdla_ctrl_submit(struct nvdla_private *priv, void *arg)
 		nvdla_dbg_info(pdev, "submit [%d]th task", i + 1);
 
 		/* allocate per task and update fields */
-		task = nvdla_task_alloc(queue, local_tasks[i]);
+		task = nvdla_task_alloc(queue, buffers, local_tasks[i]);
 		if (IS_ERR(task)) {
 			err = PTR_ERR(task);
 			goto fail_to_task_alloc;

drivers/video/tegra/host/nvdla/nvdla_queue.c

@@ -119,6 +119,9 @@ static void task_free(struct kref *ref)
 		task->task_desc = NULL;
 	}
 
+	/* free operation descriptor handle */
+	kfree(task->op_handle);
+
 	/* finally free task */
 	kfree(task);
 }
@@ -153,6 +156,10 @@ static void nvdla_queue_update(void *priv, int nr_completed)
 			/* give syncpoint reference */
 			nvhost_syncpt_put_ref(task->sp, queue->syncpt_id);
 
+			/* unpin submit ref */
+			nvhost_buffer_submit_unpin(task->buffers,
+				task->op_handle, MAX_HANDLE_PER_OP_DESC);
+
 			/* update takslist */
 			list_del(&task->list);
 
@@ -192,9 +199,11 @@ dma_addr_t get_semaphore_pa(struct platform_device *pdev)
 #endif
 
 struct nvdla_task *nvdla_task_alloc(struct nvhost_queue *queue,
+			struct nvhost_buffers *buffers,
 			struct nvdla_ctrl_ioctl_submit_task user_task)
 {
 	struct platform_device *pdev = queue->pool->pdev;
+	u32 num_operations = user_task.num_operations;
 	u32 num_postfences = user_task.num_postfences;
 	u32 num_prefences = user_task.num_prefences;
 	struct dla_action_semaphore *postaction;
@@ -210,9 +219,11 @@ struct nvdla_task *nvdla_task_alloc(struct nvhost_queue *queue,
 	size_t preactionlist_size;
 	uint16_t postactionl_of;
 	uint16_t preactionl_of;
+	dma_addr_t op_dma_addr;
 	dma_addr_t buffer_pa;
 	size_t task_size;
 	size_t buf_size;
+	size_t op_size;
 	u32 *buffer_va;
 	void *mem;
 	int err;
@@ -400,10 +411,37 @@ struct nvdla_task *nvdla_task_alloc(struct nvhost_queue *queue,
 		postaction->address = get_semaphore_pa(pdev);
 	}
 
+	if (num_operations) {
+		task->buffers = buffers;
+		task->op_handle = kcalloc(MAX_HANDLE_PER_OP_DESC, sizeof(u32),
+				GFP_KERNEL);
+		if (!task->op_handle) {
+			err = -ENOMEM;
+			goto fail_to_alloc_opdesc;
+		}
+
+		if (copy_from_user(task->op_handle,
+				(void __user *)user_task.operation_desc,
+				(MAX_HANDLE_PER_OP_DESC * sizeof(u32)))) {
+			err = -EFAULT;
+			goto fail_to_cpy_buffer;
+		}
+
+		err = nvhost_buffer_submit_pin(buffers, task->op_handle,
+			MAX_HANDLE_PER_OP_DESC, &op_dma_addr, &op_size);
+		if (!err) {
+			task_desc->operation_desc = op_dma_addr;
+			task_desc->num_operations = num_operations;
+		}
+	}
+
 	nvdla_dbg_info(pdev, "task[%p] initialized", task);
 
 	return task;
 
+fail_to_cpy_buffer:
+	kfree(task->op_handle);
+fail_to_alloc_opdesc:
 fail_to_dma_alloc:
 	kfree(task);
 fail_to_alloc_task: