Interpreter crash from `tf.io.decode_raw`
### Impact
The implementation of `tf.io.decode_raw` produces incorrect results and crashes the Python interpreter when combining `fixed_length` and wider datatypes.
```python
import tensorflow as tf
tf.io.decode_raw(tf.constant(["1","2","3","4"]), tf.uint16, fixed_length=4)
```
The implementation of the padded version
https://github.com/tensorflow/tensorflow/blob/1d8903e5b167ed0432077a3db6e462daf781d1fe/tensorflow/core/kernels/decode_padded_raw_op.cc is buggy due to a confusion about pointer arithmetic rules.
First, the code computes
https://github.com/tensorflow/tensorflow/blob/1d8903e5b167ed0432077a3db6e462daf781d1fe/tensorflow/core/kernels/decode_padded_raw_op.cc#L61 the width of each output element by dividing the `fixed_length` value to the size of the type argument:
```cc
int width = fixed_length / sizeof(T);
```
The `fixed_length` argument is also used to determine the size needed for the output tensor
https://github.com/tensorflow/tensorflow/blob/1d8903e5b167ed0432077a3db6e462daf781d1fe/tensorflow/core/kernels/decode_padded_raw_op.cc#L63-L79:
```cc
TensorShape out_shape = input.shape();
out_shape.AddDim(width);
Tensor* output_tensor = nullptr;
OP_REQUIRES_OK(context, context->allocate_output("output", out_shape, &output_tensor));
auto out = output_tensor->flat_inner_dims<T>();
T* out_data = out.data();
memset(out_data, 0, fixed_length * flat_in.size());
```
This is followed by reencoding code
https://github.com/tensorflow/tensorflow/blob/1d8903e5b167ed0432077a3db6e462daf781d1fe/tensorflow/core/kernels/decode_padded_raw_op.cc#L85-L94:
```cc
for (int64 i = 0; i < flat_in.size(); ++i) {
const T* in_data = reinterpret_cast<const T*>(flat_in(i).data());
if (flat_in(i).size() > fixed_length) {
memcpy(out_data, in_data, fixed_length);
} else {
memcpy(out_data, in_data, flat_in(i).size());
}
out_data += fixed_length;
}
```
The erroneous code is the last line above: it is moving the `out_data` pointer by `fixed_length * sizeof(T)` bytes whereas it only copied at most `fixed_length` bytes from the input. This results in parts of the input not being decoded into the output.
Furthermore, because the pointer advance is far wider than desired, this quickly leads to writing to outside the bounds of the backing data. This OOB write leads to interpreter crash in the reproducer mentioned here, but more severe attacks can be mounted too, given that this gadget allows writing to periodically placed locations in memory.
### Patches
We have patched the issue in GitHub commit 698e01511f62a3c185754db78ebce0eee1f0184d
https://github.com/tensorflow/tensorflow/commit/698e01511f62a3c185754db78ebce0eee1f0184d.
The fix will be included in TensorFlow 2.5.0. We will also cherrypick this commit on TensorFlow 2.4.2, TensorFlow 2.3.3, TensorFlow 2.2.3 and TensorFlow 2.1.4, 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.