zink: force-add usage when adding last-ref tracking

this fixes desync+crash when:
1. usage is added for bs A
2. tracking is added for bs B
3. tracking is removed for bs B
4. context is destroyed
5. usage A is now dangling and will crash if accessed

as seen in glmark2

cc: mesa-stable

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15429>

diff --git a/src/gallium/drivers/zink/zink_context.c b/src/gallium/drivers/zink/zink_context.c
index 9c07ce6..c969016 100644 (file)
--- a/src/gallium/drivers/zink/zink_context.c
+++ b/src/gallium/drivers/zink/zink_context.c
@@ -73,8 +73,18 @@ debug_describe_zink_buffer_view(char *buf, const struct zink_buffer_view *ptr)
 ALWAYS_INLINE static void
 check_resource_for_batch_ref(struct zink_context *ctx, struct zink_resource *res)
 {
-   if (!zink_resource_has_binds(res))
-      zink_batch_reference_resource(&ctx->batch, res);
+   if (!zink_resource_has_binds(res)) {
+      /* avoid desync between usage and tracking:
+       * - if usage exists, it must be removed before the context is destroyed
+       * - having usage does not imply having tracking
+       * - if tracking will be added here, also reapply usage to avoid dangling usage once tracking is removed
+       * TODO: somehow fix this for perf because it's an extra hash lookup
+       */
+      if (res->obj->bo->reads || res->obj->bo->writes)
+         zink_batch_reference_resource_rw(&ctx->batch, res, !!res->obj->bo->writes);
+      else
+         zink_batch_reference_resource(&ctx->batch, res);
+   }
 }
 
 static void