drm/amdgpu: Add queue id support to the user queue wait IOCTL

Add queue id support to the user queue wait IOCTL
drm_amdgpu_userq_wait structure.

This is required to retrieve the wait user queue and maintain
the fence driver references in it so that the user queue in
the same context releases their reference to the fence drivers
at some point before queue destruction.

Otherwise, we would gather those references until we
don't have any more space left and crash.

v2: Modify the UAPI comment as per the mesa and libdrm UAPI comment.

Libdrm MR: https://gitlab.freedesktop.org/mesa/drm/-/merge_requests/408
Mesa MR: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34493

Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Suggested-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c
index 0a3032e..ca19836 100644 (file)
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c
@@ -91,7 +91,6 @@ int amdgpu_userq_fence_driver_alloc(struct amdgpu_device *adev,
        spin_lock_init(&fence_drv->fence_list_lock);
 
        fence_drv->adev = adev;
-       fence_drv->fence_drv_xa_ptr = &userq->fence_drv_xa;
        fence_drv->context = dma_fence_context_alloc(1);
        get_task_comm(fence_drv->timeline_name, current);
 
@@ -611,6 +610,9 @@ int amdgpu_userq_wait_ioctl(struct drm_device *dev, void *data,
        u32 num_syncobj, num_read_bo_handles, num_write_bo_handles;
        struct drm_amdgpu_userq_fence_info *fence_info = NULL;
        struct drm_amdgpu_userq_wait *wait_info = data;
+       struct amdgpu_fpriv *fpriv = filp->driver_priv;
+       struct amdgpu_userq_mgr *userq_mgr = &fpriv->userq_mgr;
+       struct amdgpu_usermode_queue *waitq;
        struct drm_gem_object **gobj_write;
        struct drm_gem_object **gobj_read;
        struct dma_fence **fences = NULL;
@@ -860,6 +862,10 @@ int amdgpu_userq_wait_ioctl(struct drm_device *dev, void *data,
                        fences[num_fences++] = fence;
                }
 
+               waitq = idr_find(&userq_mgr->userq_idr, wait_info->waitq_id);
+               if (!waitq)
+                       goto free_fences;
+
                for (i = 0, cnt = 0; i < num_fences; i++) {
                        struct amdgpu_userq_fence_driver *fence_drv;
                        struct amdgpu_userq_fence *userq_fence;
@@ -888,14 +894,12 @@ int amdgpu_userq_wait_ioctl(struct drm_device *dev, void *data,
                         * Otherwise, we would gather those references until we don't
                         * have any more space left and crash.
                         */
-                       if (fence_drv->fence_drv_xa_ptr) {
-                               r = xa_alloc(fence_drv->fence_drv_xa_ptr, &index, fence_drv,
-                                            xa_limit_32b, GFP_KERNEL);
-                               if (r)
-                                       goto free_fences;
+                       r = xa_alloc(&waitq->fence_drv_xa, &index, fence_drv,
+                                    xa_limit_32b, GFP_KERNEL);
+                       if (r)
+                               goto free_fences;
 
-                               amdgpu_userq_fence_driver_get(fence_drv);
-                       }
+                       amdgpu_userq_fence_driver_get(fence_drv);
 
                        /* Store drm syncobj's gpu va address and value */
                        fence_info[cnt].va = fence_drv->va;

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.h
index 7bbae23..2af4e0c 100644 (file)
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.h
@@ -55,7 +55,6 @@ struct amdgpu_userq_fence_driver {
        spinlock_t fence_list_lock;
        struct list_head fences;
        struct amdgpu_device *adev;
-       struct xarray *fence_drv_xa_ptr;
        char timeline_name[TASK_COMM_LEN];
 };
 

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userqueue.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userqueue.c
index 4e02d6c..8f6510b 100644 (file)
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userqueue.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userqueue.c
@@ -51,7 +51,6 @@ amdgpu_userqueue_cleanup(struct amdgpu_userq_mgr *uq_mgr,
        }
 
        uq_funcs->mqd_destroy(uq_mgr, queue);
-       queue->fence_drv->fence_drv_xa_ptr = NULL;
        amdgpu_userq_fence_driver_free(queue);
        idr_remove(&uq_mgr->userq_idr, queue_id);
        kfree(queue);

diff --git a/include/uapi/drm/amdgpu_drm.h b/include/uapi/drm/amdgpu_drm.h
index 284ac25..1fd9647 100644 (file)
--- a/include/uapi/drm/amdgpu_drm.h
+++ b/include/uapi/drm/amdgpu_drm.h
@@ -515,6 +515,12 @@ struct drm_amdgpu_userq_fence_info {
 };
 
 struct drm_amdgpu_userq_wait {
+       /**
+        * @waitq_id: Queue handle used by the userq wait IOCTL to retrieve the
+        * wait queue and maintain the fence driver references in it.
+        */
+       __u32   waitq_id;
+       __u32   pad;
        /**
         * @syncobj_handles: The list of syncobj handles submitted by the user queue
         * job to get the va/value pairs.