Segmentation Fault in torch._C._nn.thnn_conv2d Due to Invalid Padding Argument

[Gwihwan-Go]

🐛 Describe the bug

A segmentation fault occurs when invoking torch._C._nn.thnn_conv2d with an empty list for the padding argument, alongside other parameters that are potentially mismatched or improperly specified.

Error log

Segmentation fault (core dumped)

Minimized Repro

import torch
torch._C._nn.thnn_conv2d(torch.rand([1]), bias=torch.rand([3]), kernel_size=[5], padding=[], stride=[9], weight=torch.rand([9]))

Versions

PyTorch version: 2.3.0.dev20240301+cu118
Is debug build: False
CUDA used to build PyTorch: 11.8
ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.3 LTS (x86_64)
GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Clang version: Could not collect
CMake version: Could not collect
Libc version: glibc-2.35

cc @albanD @mruberry @jbschlosser @walterddr @mikaylagawarecki @malfet

[albanD]

Thanks for the report. We would definitely accept a PR adding the appropriate error checking to avoid the segfault!

[Martim03]

Hi, I managed to replicate the issue and I would be willing to solve it if you don't mind! :)

[albanD]

Sounds good. Feel free to send a PR and link this issue as fixed in the PR description

[Martim03]

Hi again, upon reviewing the code, I've noticed that a Segmentation Fault could also occur if the function is invoked with an invalid kernel_size or stride. Would you like me to include error checking for those cases as well?