Deadlock in mutually recursive `tf.function` objects
Medium5.5CVE-2021-41213 · Published Nov 10, 2021 · updated Jul 8, 2026
### Impact The [code behind `tf.function` API](https://github.com/tensorflow/tensorflow/blob/8d72537c6abf5a44103b57b9c2e22c14f5f49698/tensorflow/python/eager/def_function.py#L542) can be made to deadlock when two `tf.function` decorated Python functions are mutually recursive: ```python import tensorflow as tf @tf.function() def fun1(num): if num == 1: return print(num) fun2(num-1) @tf.function() def fun2(num): if num == 0: return print(num) fun1(num-1) fun1(9) ``` 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. ### Patches We have patched the issue in GitHub commit [afac8158d43691661ad083f6dd9e56f327c1dcb7](https://github.com/tensorflow/tensorflow/commit/afac8158d43691661ad083f6dd9e56f327c1dcb7). 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 ra...
Affected versions
| Package | Affected | Fixed in |
|---|---|---|
| tensorflow PyPI | >= 2.6.0, < 2.6.1 | 2.6.1 |
| >= 2.5.0, < 2.5.2 | 2.5.2 | |
| < 2.4.4 | 2.4.4 |
Details and references
### Impact The [code behind `tf.function` API](https://github.com/tensorflow/tensorflow/blob/8d72537c6abf5a44103b57b9c2e22c14f5f49698/tensorflow/python/eager/def_function.py#L542) can be made to deadlock when two `tf.function` decorated Python functions are mutually recursive: ```python import tensorflow as tf @tf.function() def fun1(num): if num == 1: return print(num) fun2(num-1) @tf.function() def fun2(num): if num == 0: return print(num) fun1(num-1) fun1(9) ``` 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. ### Patches We have patched the issue in GitHub commit [afac8158d43691661ad083f6dd9e56f327c1dcb7](https://github.com/tensorflow/tensorflow/commit/afac8158d43691661ad083f6dd9e56f327c1dcb7). 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. ### For more information Please consult [our security guide](https://github.com/tensorflow/tensorflow/blob/master/SECURITY.md) for more information regarding the security model and how to contact us with issues and questions. ### Attribution This vulnerability has been reported by members of the Aivul Team from Qihoo 360.
- CVSS 3.1
- CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H
- Severity from
- GitHub (reviewed advisory)
- Weakness
- CWE-662, CWE-667
- Also known as
- BIT-tensorflow-2021-41213, CVE-2021-41213, PYSEC-2021-405, PYSEC-2021-622, PYSEC-2021-820
- github.com/tensorflow/tensorflow/security/advisories/GHSA-h67m-xg8f-fxcf
- nvd.nist.gov/vuln/detail/CVE-2021-41213
- github.com/tensorflow/tensorflow/commit/afac8158d43691661ad083f6dd9e56f327c1dcb7
- github.com/pypa/advisory-database/tree/main/vulns/tensorflow-cpu/PYSEC-2021-622.yaml
- github.com/pypa/advisory-database/tree/main/vulns/tensorflow-gpu/PYSEC-2021-820.yaml
- github.com/pypa/advisory-database/tree/main/vulns/tensorflow/PYSEC-2021-405.yaml
- github.com/tensorflow/tensorflow
More TensorFlow advisories
All TensorFlow| Date | Advisory | Severity | Fixed in |
|---|---|---|---|
| Nov 102021 | Incomplete validation in boosted trees code Critical9.3Nov 10, 2021 | Critical9.3 | 2.4.4+2 more |
| Nov 102021 | Crash in `tf.math.segment_*` operationsMedium5.5Nov 10, 2021 | Medium5.5 | 2.4.4+2 more |
| Nov 102021 | Crash in `max_pool3d` when size argument is 0 or negativeMedium5.5Nov 10, 2021 | Medium5.5 | 2.4.4+2 more |
| Nov 102021 | Crashes due to overflow and `CHECK`-fail in ops with large tensor shapesMedium5.5Nov 10, 2021 | Medium5.5 | 2.4.4+2 more |
| Nov 102021 | Overflow/crash in `tf.tile` when tiling tensor is largeMedium5.5Nov 10, 2021 | Medium5.5 | 2.4.4+2 more |
| Nov 102021 | Overflow/crash in `tf.image.resize` when size is largeMedium5.5Nov 10, 2021 | Medium5.5 | 2.4.4+2 more |