Deadlock in mutually recursive `tf.function` objects

Summary

TensorFlow is an open source platform for machine learning. In affected versions the code behind `tf.function` API can be made to deadlock when two `tf.function` decorated Python functions are mutually recursive. This occurs due to using a non-reentrant `Lock` Python object. Loading any model which contains mutually recursive functions is vulnerable. An attacker can cause denial of service by causing users to load such models and calling a recursive `tf.function`, although this is not a frequent scenario. The fix will be included in TensorFlow 2.7.0. We will also cherrypick this commit on TensorFlow 2.6.1, TensorFlow 2.5.2, and TensorFlow 2.4.4, as these are also affected and still in supported range.

Timeline

1. Nov 5, 2021, 10:10 PM UTCCVE published
2. Aug 4, 2024, 3:08 AM UTCRecord last updated

Affected versions

• tensorflow tensorflow

  From 2.6.0 · fixed in 2.6.1; From 2.5.0 · fixed in 2.5.2; fixed in 2.4.4

• tensorflow-cpuPyPI

  fixed in 2.4.4; From 2.5.0 · fixed in 2.5.2; From 2.6.0 · fixed in 2.6.1; From 2.7.0rc0 · fixed in 2.7.0

• tensorflowPyPI

  fixed in 2.4.4; From 2.5.0 · fixed in 2.5.2; From 2.6.0 · fixed in 2.6.1; From 2.7.0rc0 · fixed in 2.7.0

• tensorflowPyPI

  From 2.6.0 · fixed in 2.6.1

• tensorflowPyPI

  From 2.5.0 · fixed in 2.5.2

• tensorflowPyPI

  fixed in 2.4.4

• tensorflow-cpuPyPI

  From 2.6.0 · fixed in 2.6.1

• tensorflow-cpuPyPI

  From 2.5.0 · fixed in 2.5.2

• tensorflow-cpuPyPI

  fixed in 2.4.4

• tensorflow-gpuPyPI

  From 2.6.0 · fixed in 2.6.1

• tensorflow-gpuPyPI

  From 2.5.0 · fixed in 2.5.2

• tensorflow-gpuPyPI

  fixed in 2.4.4

• tensorflow-gpuPyPI

  fixed in 2.4.4; From 2.5.0 · fixed in 2.5.2; From 2.6.0 · fixed in 2.6.1; From 2.7.0rc0 · fixed in 2.7.0

References 2