Validate NPY shape sizes before allocation - #6443

Category:

Bug fix

Description:

Reject invalid and overflowing NumPy array shapes before an allocation or file read can use a wrapped element or byte count.

Addresses:
When parsing numpy (.npy) files, the ParseHeaderContents function reads arbitrary int64_t shape values from the file header. The volume() function multiplies all shape dimensions together without overflow checking. The result is used in HeaderData::nbytes() which computes type_info->size() * size() (both size_t). A crafted .npy file with shape dimensions whose product overflows int64_t (e.g., shape=(2^32, 2^32)) would cause volume() to return a small or negative value, leading to an undersized memory allocation in Tensor::Resize(), followed by a buffer overflow when reading file data into the undersized buffer.

Additional information:

Affected modules and functionalities:

• NumPy .npy header parsing and HeaderData size calculation

Key points relevant for the review:

• Reject negative and out-of-range dimensions.
• Check both element-count and byte-count multiplication.
• Errors report the requested multiplication terms and the maximum size_t limit.

Tests:

• Existing tests apply
• New tests added
  • Python tests
  • GTests
    • NumpyLoaderTest.RejectsInvalidOrOverflowingShape
  • Benchmark
  • Other
• N/A

Checklist

Documentation

• Existing documentation applies
• Documentation updated
  • Docstring
  • Doxygen
  • RST
  • Jupyter
  • Other
• N/A

DALI team only

Requirements

• Implements new requirements
• Affects existing requirements
• N/A

REQ IDs: N/A

JIRA TASK: N/A