[5.9][RemoteMirror] Fix demangle node corruption caused by excessively eager clearing of the NodeFactory. #64769
Conversation
…y eager clearing of the NodeFactory. We clear the NodeFactory to prevent unbounded buildup of allocated memory, but this is done too eagerly. In particular, normalizeReflectionName can end up clearing the factory while the calling code is still using nodes that were allocated from it. To keep peak memory usage low while avoiding this problem, we introduce a checkpoint mechanism in NodeFactory. A checkpoint can be pushed and then subsequently popped. When a checkpoint is popped, only the nodes allocated since the checkpoint was pushed are invalidated and the memory reclaimed. This allows us to quickly clear short-lived nodes like those created in normalizeReflectionName, while preserving longer-lived nodes used in code calling it. Uses of clearNodeFactory are replaced with this checkpoint mechanism. rdar://106547092 (cherry picked from commit fbc3f69)
|
Explanation: The demangler uses an internal bump allocator which can be cleared between uses to prevent memory buildup. Remote Mirror sometimes clears the allocator too eagerly, when some nodes are still in use, resulting in corrupted node structures and crashes. This fixes the problem by adding a checkpoint system to the bump allocator, where it can be partically cleared back to the checkpoint. This allows code to use the demangler for temporary local tasks and then clear the memory used without affecting the caller's use of the demangler. |
|
@swift-ci please test |
Cherry-pick #64674 to
release/5.9.We clear the NodeFactory to prevent unbounded buildup of allocated memory, but this is done too eagerly. In particular, normalizeReflectionName can end up clearing the factory while the calling code is still using nodes that were allocated from it.
To keep peak memory usage low while avoiding this problem, we introduce a checkpoint mechanism in NodeFactory. A checkpoint can be pushed and then subsequently popped. When a checkpoint is popped, only the nodes allocated since the checkpoint was pushed are invalidated and the memory reclaimed. This allows us to quickly clear short-lived nodes like those created in normalizeReflectionName, while preserving longer-lived nodes used in code calling it. Uses of clearNodeFactory are replaced with this checkpoint mechanism.
rdar://106547092