An issue was discovered in PyTorch v2.5 and v2.7.1. Omission of profiler.stop() can cause torch.profiler.profile (PythonTracer) to crash or hang during finalization, leading to a Denial of Service (DoS).
The product does not properly acquire or release a lock on a resource, leading to unexpected resource state changes and behaviors.
Locking is a type of synchronization behavior that ensures that multiple independently-operating processes or threads do not interfere with each other when accessing the same resource. All processes/threads are expected to follow the same steps for locking. If these steps are not followed precisely - or if no locking is done at all - then another process/thread could modify the shared resource in a way that is not visible or predictable to the original process. This can lead to data or memory corruption, denial of service, etc.