Impact
Under certain scenarios, TensorFlow can fail to specialize a type during shape inference:
void InferenceContext::PreInputInit(
const OpDef& op_def, const std::vector<const Tensor*>& input_tensors,
const std::vector<ShapeHandle>& input_tensors_as_shapes) {
const auto ret = full_type::SpecializeType(attrs_, op_def);
DCHECK(ret.status().ok()) << "while instantiating types: " << ret.status();
ret_types_ = ret.ValueOrDie();
// ...
}However, DCHECK is a no-op in production builds and an assertion failure in debug builds. In the first case execution proceeds to the ValueOrDie line. This results in an assertion failure as ret contains an error Status, not a value. In the second case we also get a crash due to the assertion failure.
Patches
We have patched the issue in GitHub commit cb164786dc891ea11d3a900e90367c339305dc7b.
The fix will be included in TensorFlow 2.8.0. We will also cherrypick this commit on TensorFlow 2.7.1, and TensorFlow 2.6.3, as these are also affected and still in supported range.
For more information
Please consult our security guide for more information regarding the security model and how to contact us with issues and questions.
References
Impact
Under certain scenarios, TensorFlow can fail to specialize a type during shape inference:
However,
DCHECKis a no-op in production builds and an assertion failure in debug builds. In the first case execution proceeds to theValueOrDieline. This results in an assertion failure asretcontains an errorStatus, not a value. In the second case we also get a crash due to the assertion failure.Patches
We have patched the issue in GitHub commit cb164786dc891ea11d3a900e90367c339305dc7b.
The fix will be included in TensorFlow 2.8.0. We will also cherrypick this commit on TensorFlow 2.7.1, and TensorFlow 2.6.3, as these are also affected and still in supported range.
For more information
Please consult our security guide for more information regarding the security model and how to contact us with issues and questions.
References