The user queue object destroy requires ensuring its
VA keeps mapping prior to the queue being destroyed.
Otherwise, it seems a bug in the user space or VA
freed wrongly, and the kernel driver should report an
invalidated state to the user IOCTL request.
Signed-off-by: Prike Liang <prike.li...@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
index 6cdfeb224f6c..e5891674b4d0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
@@ -520,6 +520,13 @@ amdgpu_userq_destroy(struct drm_file *filp, int queue_id)
amdgpu_bo_unreserve(queue->db_obj.obj);
}
amdgpu_bo_unref(&queue->db_obj.obj);
+ /*
+ * At this point the userq obj va should be mapped,
+ * otherwise will return error to user.
+ */
+ if (!amdgpu_userq_buffer_vas_mapped(&fpriv->vm, queue))
+ queue->state = AMDGPU_USERQ_STATE_INVALID_VA;
+
r = amdgpu_userq_unmap_helper(uq_mgr, queue);
/*TODO: It requires a reset for userq hw unmap error*/
if (unlikely(r != AMDGPU_USERQ_STATE_UNMAPPED)) {
--
2.34.1
