drm/nouveau: remove fence wait code from deferred client work handler

Fences attached to deferred client work items now originate from channels
belonging to the client, meaning we can be certain they've been signalled
before we destroy a client.

This closes a race that could happen if the dma_fence_wait_timeout() call
didn't succeed.  When the fence was later signalled, a use-after-free was
possible.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

diff --git a/drivers/gpu/drm/nouveau/nouveau_drm.c b/drivers/gpu/drm/nouveau/nouveau_drm.c
index 6caece4..64b8fd0 100644 (file)
--- a/drivers/gpu/drm/nouveau/nouveau_drm.c
+++ b/drivers/gpu/drm/nouveau/nouveau_drm.c
@@ -113,24 +113,22 @@ nouveau_name(struct drm_device *dev)
 }
 
 static inline bool
-nouveau_cli_work_ready(struct dma_fence *fence, bool wait)
+nouveau_cli_work_ready(struct dma_fence *fence)
 {
-       if (!dma_fence_is_signaled(fence)) {
-               if (!wait)
-                       return false;
-               WARN_ON(dma_fence_wait_timeout(fence, false, 2 * HZ) <= 0);
-       }
+       if (!dma_fence_is_signaled(fence))
+               return false;
        dma_fence_put(fence);
        return true;
 }
 
 static void
-nouveau_cli_work_flush(struct nouveau_cli *cli, bool wait)
+nouveau_cli_work(struct work_struct *w)
 {
+       struct nouveau_cli *cli = container_of(w, typeof(*cli), work);
        struct nouveau_cli_work *work, *wtmp;
        mutex_lock(&cli->lock);
        list_for_each_entry_safe(work, wtmp, &cli->worker, head) {
-               if (!work->fence || nouveau_cli_work_ready(work->fence, wait)) {
+               if (!work->fence || nouveau_cli_work_ready(work->fence)) {
                        list_del(&work->head);
                        work->func(work);
                }
@@ -158,17 +156,17 @@ nouveau_cli_work_queue(struct nouveau_cli *cli, struct dma_fence *fence,
        mutex_unlock(&cli->lock);
 }
 
-static void
-nouveau_cli_work(struct work_struct *w)
-{
-       struct nouveau_cli *cli = container_of(w, typeof(*cli), work);
-       nouveau_cli_work_flush(cli, false);
-}
-
 static void
 nouveau_cli_fini(struct nouveau_cli *cli)
 {
-       nouveau_cli_work_flush(cli, true);
+       /* All our channels are dead now, which means all the fences they
+        * own are signalled, and all callback functions have been called.
+        *
+        * So, after flushing the workqueue, there should be nothing left.
+        */
+       flush_work(&cli->work);
+       WARN_ON(!list_empty(&cli->worker));
+
        usif_client_fini(cli);
        nouveau_vmm_fini(&cli->vmm);
        nvif_mmu_fini(&cli->mmu);