[PATCH v3 1/2] drm/amdgpu/mes: refactor the amdgpu_mes_alloc/free_proc|gang()

- Unify amdgpu_mes_alloc/free_proc|gang_ctx_index to provide
  centralized RS64mem bitmap management for both KGD and KFD.
- Retrieve the bitmap bit for userq contex index based on a
  per process granularity.

Signed-off-by: Prike Liang <[email protected]>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c    | 28 +++++++++++-----------
 drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h    |  8 +++----
 drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c  |  6 +++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h  |  3 ++-
 drivers/gpu/drm/amd/amdgpu/mes_userqueue.c | 26 ++++++++++++++------
 5 files changed, 45 insertions(+), 26 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
index 9af7c97af5be..5998e05ecd79 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
@@ -1040,13 +1040,13 @@ int amdgpu_mes_rs64mem_setup_bitmaps(struct amdgpu_mes *mes)
  * amdgpu_mes_alloc_proc_ctx_index - allocate a process context slot
  *
  * @mes: MES instance
- * @queue: Usermode queue receiving the allocated process context index
+ * @index: the allocated process context index
  *
  * Returns 0 on success, -ENOSPC if all slots are used, or
  * -EOPNOTSUPP if RS64 local memory is unavailable.
  */
 int amdgpu_mes_alloc_proc_ctx_index(struct amdgpu_mes *mes,
-				    struct amdgpu_usermode_queue *queue)
+				    uint32_t *index)
 {
 	unsigned long bit;
 
@@ -1061,7 +1061,7 @@ int amdgpu_mes_alloc_proc_ctx_index(struct amdgpu_mes *mes,
 		return -ENOSPC;
 	}
 	set_bit(bit, mes->proc_ctx_bitmap);
-	queue->proc_ctx_array_index = (uint32_t)bit;
+	*index = (uint32_t)bit;
 	amdgpu_mes_unlock(mes);
 
 	return 0;
@@ -1071,18 +1071,18 @@ int amdgpu_mes_alloc_proc_ctx_index(struct amdgpu_mes *mes,
   * amdgpu_mes_free_proc_ctx_index - free a process context slot
   *
   * @mes: MES instance
-  * @queue: Usermode queue whose process context index is released
+  * @index: process context index is released
   */
 void amdgpu_mes_free_proc_ctx_index(struct amdgpu_mes *mes,
-				    struct amdgpu_usermode_queue *queue)
+				    uint32_t index)
 {
 	if (!mes->use_rs64mem || !mes->proc_ctx_bitmap)
 		return;
-	if (queue->proc_ctx_array_index >= mes->proc_ctx_array_size)
+	if (index >= mes->proc_ctx_array_size)
 		return;
 
 	amdgpu_mes_lock(mes);
-	clear_bit(queue->proc_ctx_array_index, mes->proc_ctx_bitmap);
+	clear_bit(index, mes->proc_ctx_bitmap);
 	amdgpu_mes_unlock(mes);
 }
 
@@ -1090,13 +1090,13 @@ void amdgpu_mes_free_proc_ctx_index(struct amdgpu_mes *mes,
   * amdgpu_mes_alloc_gang_ctx_index - allocate a gang context slot
   *
   * @mes: MES instance
-  * @queue: Usermode queue receiving the allocated gang context index
+  * @index: the allocated gang context index
   *
   * Returns 0 on success, -ENOSPC if all slots are used, or
   * -EOPNOTSUPP if RS64 local memory is unavailable.
   */
 int amdgpu_mes_alloc_gang_ctx_index(struct amdgpu_mes *mes,
-				    struct amdgpu_usermode_queue *queue)
+				    uint32_t *index)
 {
 	unsigned long bit;
 
@@ -1111,7 +1111,7 @@ int amdgpu_mes_alloc_gang_ctx_index(struct amdgpu_mes *mes,
 		return -ENOSPC;
 	}
 	set_bit(bit, mes->gang_ctx_bitmap);
-	queue->gang_ctx_array_index = bit;
+	*index = bit;
 	amdgpu_mes_unlock(mes);
 
 	return 0;
@@ -1121,18 +1121,18 @@ int amdgpu_mes_alloc_gang_ctx_index(struct amdgpu_mes *mes,
   * amdgpu_mes_free_gang_ctx_index - free a gang context slot
   *
   * @mes: MES instance
-  * @queue: Usermode queue whose gang context index is released
+  * @index: gang context index is released
   */
 void amdgpu_mes_free_gang_ctx_index(struct amdgpu_mes *mes,
-				    struct amdgpu_usermode_queue *queue)
+				    uint32_t index)
 {
 	if (!mes->use_rs64mem || !mes->gang_ctx_bitmap)
 		return;
-	if (queue->gang_ctx_array_index >= mes->gang_ctx_array_size)
+	if (index >= mes->gang_ctx_array_size)
 		return;
 
 	amdgpu_mes_lock(mes);
-	clear_bit(queue->gang_ctx_array_index, mes->gang_ctx_bitmap);
+	clear_bit(index, mes->gang_ctx_bitmap);
 	amdgpu_mes_unlock(mes);
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
index c67db2d6e122..977c057dcce8 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
@@ -636,11 +636,11 @@ int amdgpu_mes_rs64mem_init(struct amdgpu_mes *mes);
 void amdgpu_mes_rs64mem_fini(struct amdgpu_mes *mes);
 int amdgpu_mes_rs64mem_setup_bitmaps(struct amdgpu_mes *mes);
 int amdgpu_mes_alloc_proc_ctx_index(struct amdgpu_mes *mes,
-				    struct amdgpu_usermode_queue *queue);
+				    uint32_t *index);
 void amdgpu_mes_free_proc_ctx_index(struct amdgpu_mes *mes,
-				    struct amdgpu_usermode_queue *queue);
+				    uint32_t index);
 int amdgpu_mes_alloc_gang_ctx_index(struct amdgpu_mes *mes,
-				    struct amdgpu_usermode_queue *queue);
+				    uint32_t *index);
 void amdgpu_mes_free_gang_ctx_index(struct amdgpu_mes *mes,
-				    struct amdgpu_usermode_queue *queue);
+				    uint32_t index);
 #endif /* __AMDGPU_MES_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
index 571413c9a611..5dd4fe258cdb 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
@@ -1244,6 +1244,7 @@ int amdgpu_userq_mgr_init(struct amdgpu_userq_mgr *userq_mgr, struct drm_file *f
 	xa_init_flags(&userq_mgr->userq_xa, XA_FLAGS_ALLOC);
 	userq_mgr->adev = adev;
 	userq_mgr->file = file_priv;
+	userq_mgr->proc_ctx_allocated = false;
 	mutex_init(&userq_mgr->proc_ctx_lock);
 
 	INIT_DELAYED_WORK(&userq_mgr->resume_work, amdgpu_userq_restore_worker);
@@ -1272,6 +1273,7 @@ void amdgpu_userq_mgr_cancel_resume(struct amdgpu_userq_mgr *userq_mgr)
 
 void amdgpu_userq_mgr_fini(struct amdgpu_userq_mgr *userq_mgr)
 {
+	struct amdgpu_mes *mes = &userq_mgr->adev->mes;
 	struct amdgpu_usermode_queue *queue;
 	unsigned long queue_id = 0;
 
@@ -1298,6 +1300,10 @@ void amdgpu_userq_mgr_fini(struct amdgpu_userq_mgr *userq_mgr)
 	 */
 	cancel_work_sync(&userq_mgr->reset_work);
 
+	if (userq_mgr->proc_ctx_allocated) {
+		amdgpu_mes_free_proc_ctx_index(mes, userq_mgr->proc_ctx_array_index);
+		userq_mgr->proc_ctx_allocated = false;
+	}
 	amdgpu_bo_free_kernel(&userq_mgr->proc_ctx_obj.obj,
 			      &userq_mgr->proc_ctx_obj.gpu_addr,
 			      &userq_mgr->proc_ctx_obj.cpu_ptr);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h
index 402dc2816f04..4dcf6151de6a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h
@@ -101,7 +101,6 @@ struct amdgpu_usermode_queue {
 		u64 va_array[6];
 	} userq_vas;
 
-	uint32_t	proc_ctx_array_index;
 	uint32_t	gang_ctx_array_index;
 };
 
@@ -133,6 +132,8 @@ struct amdgpu_userq_mgr {
 	struct mutex			proc_ctx_lock;
 	struct amdgpu_userq_obj		proc_ctx_obj;
 
+	bool				proc_ctx_allocated;
+	uint32_t			proc_ctx_array_index;
 	/**
 	 * @reset_work:
 	 *
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_userqueue.c b/drivers/gpu/drm/amd/amdgpu/mes_userqueue.c
index a914198a4d62..ae403bb89441 100644
--- a/drivers/gpu/drm/amd/amdgpu/mes_userqueue.c
+++ b/drivers/gpu/drm/amd/amdgpu/mes_userqueue.c
@@ -143,10 +143,24 @@ static int mes_userq_map(struct amdgpu_usermode_queue *queue)
 	queue_input.doorbell_offset = userq_props->doorbell_index;
 	queue_input.page_table_base_addr = amdgpu_gmc_pd_addr(queue->vm->root.bo);
 	queue_input.wptr_mc_addr = queue->wptr_obj.gpu_addr;
+
 	if (mes->use_rs64mem) {
-		amdgpu_mes_alloc_proc_ctx_index(mes, queue);
-		queue_input.process_context_array_index = queue->proc_ctx_array_index;
-		amdgpu_mes_alloc_gang_ctx_index(mes, queue);
+		 if (!uq_mgr->proc_ctx_allocated) {
+			r = amdgpu_mes_alloc_proc_ctx_index(mes, &uq_mgr->proc_ctx_array_index);
+			if (r) {
+				DRM_ERROR("Failed to allocate userq process index err:%d\n", r);
+				return r;
+			}
+			uq_mgr->proc_ctx_allocated = true;
+		}
+
+		r = amdgpu_mes_alloc_gang_ctx_index(mes, &queue->gang_ctx_array_index);
+		if (r) {
+			DRM_ERROR("Failed to allocate userq gang index err:%d\n", r);
+			amdgpu_mes_free_gang_ctx_index(mes, queue->gang_ctx_array_index);
+			return r;
+		}
+		queue_input.process_context_array_index = uq_mgr->proc_ctx_array_index;
 		queue_input.gang_context_array_index = queue->gang_ctx_array_index;
 	}
 	amdgpu_mes_lock(&adev->mes);
@@ -179,10 +193,8 @@ static int mes_userq_unmap(struct amdgpu_usermode_queue *queue)
 	amdgpu_mes_lock(&adev->mes);
 	r = adev->mes.funcs->remove_hw_queue(&adev->mes, &queue_input);
 	amdgpu_mes_unlock(&adev->mes);
-	if (mes->use_rs64mem) {
-		amdgpu_mes_free_proc_ctx_index(mes, queue);
-		amdgpu_mes_free_gang_ctx_index(mes, queue);
-	}
+	if (mes->use_rs64mem)
+		amdgpu_mes_free_gang_ctx_index(mes, queue->gang_ctx_array_index);
 	if (r)
 		DRM_ERROR("Failed to unmap queue in HW, err (%d)\n", r);
 	return r;
-- 
2.34.1