Quoting Chris Wilson (2019-10-31 14:18:56) > My memory says, and my assumption in this code, is that the > the iterator is safe against insertions -- we won't get horribly lost if > the tree is rebalanced as we walk. Actually, the iterator is not perfect across rebalances. It won't matter here in the selftest, since we are the only accessor, the two other users deserve throught. In __active_retire, we have exclusive access to the tree as we are freeing the nodes. Safe. In i915_active_wait() [we can't take the mutex here due to shrinker inversions!], we walk the tree to kick signaling on the nodes. So the iterator is not perfect, but calling enable_signaling() is mostly an optimisation so that we don't have to wait upon the background flush. So I think we are safe to miss nodes, so long as the iterator itself is bounded (which it must be). -Chris _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
