fix: add bounds checking for size() and stride() methods #375
Conversation
Summary of ChangesHello @guan404ming, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! This pull request significantly improves the safety and reliability of tensor dimension access by introducing explicit index bounds checking to the Highlights
🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console. Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request adds bounds checking to the size() and stride() methods of tvm::ffi::Tensor and tvm::ffi::TensorView to prevent out-of-bounds memory access, along with corresponding C++ tests. It also improves Python type hints in the test suite.
The changes are a good improvement for robustness. However, there's a significant discrepancy between the pull request title/description and the actual code changes. The title ("fix: integer overflow in GetDataSize") and description state that this PR fixes an integer overflow in GetDataSize, but the provided patches only contain bounds checking logic for size() and stride() methods. Please consider updating the title and description to accurately reflect the changes made in this PR.
I've left one comment regarding code duplication that could be addressed to improve maintainability.
| int64_t size(int64_t idx) const { | ||
| const TensorObj* ptr = get(); | ||
| return ptr->shape[idx >= 0 ? idx : (ptr->ndim + idx)]; | ||
| int64_t adjusted_idx = idx >= 0 ? idx : (ptr->ndim + idx); | ||
| if (adjusted_idx < 0 || adjusted_idx >= ptr->ndim) { | ||
| TVM_FFI_THROW(IndexError) << "Index " << idx << " out of bounds for tensor with " << ptr->ndim | ||
| << " dimensions"; | ||
| } | ||
| return ptr->shape[adjusted_idx]; | ||
| } |
There was a problem hiding this comment.
The index adjustment and bounds checking logic is duplicated in Tensor::size, Tensor::stride, TensorView::size, and TensorView::stride. To improve maintainability and reduce code duplication, consider extracting this logic into a common helper function.
For example, you could add a free helper function in a details namespace:
namespace tvm {
namespace ffi {
namespace details {
inline int64_t CheckAndAdjustTensorIndex(int64_t idx, int32_t ndim) {
int64_t adjusted_idx = idx >= 0 ? idx : (ndim + idx);
if (adjusted_idx < 0 || adjusted_idx >= ndim) {
TVM_FFI_THROW(IndexError) << "Index " << idx << " out of bounds for tensor with " << ndim
<< " dimensions";
}
return adjusted_idx;
}
} // namespace details
} // namespace ffi
} // namespace tvmThen Tensor::size and TensorView::size could be simplified to:
// Tensor::size
int64_t size(int64_t idx) const {
const TensorObj* ptr = get();
return ptr->shape[details::CheckAndAdjustTensorIndex(idx, ptr->ndim)];
}
// TensorView::size
int64_t size(int64_t idx) const {
return tensor_.shape[details::CheckAndAdjustTensorIndex(idx, tensor_.ndim)];
}A similar simplification would apply to the stride methods.
guan404ming
changed the title
|
Given that we already have conditionals in |
|
Thanks! |
2 participants
Why
size() and stride() methods could access memory out of bounds when given invalid indices
How