Fix two memory-safety defects in file-backed weights cache

[JakeStevens]

Summary:
Fixes two defects in the XNNPACK file-backed packed-weight cache (XNNWeightsCache)

1. Use-after-munmap via offset_to_addr (SIGBUS). offset_to_addr returns the raw pointer stored in packed_data_ptrs_[offset], which on the file-backed path is an address inside a MAP_SHARED region. reset_for_fresh_write() munmaps every region but left the offset->pointer slots populated, so a subsequent offset_to_addr handed XNNPACK a pointer into unmapped memory — dereferencing it faults (SIGBUS/SIGSEGV, or an ASAN report). Fix: null each file-backed slot before erasing its metadata entry, so offset_to_addr returns nullptr (the miss contract look_up_or_insert already expects) instead of a dangling pointer.

2. Mid-trailer-truncated cache file wrongly accepted. In load_packed_cache, the parse loop guard i < entry_count && cursor + 4 <= end terminated silently when fewer than 4 bytes remained, with no post-loop check that all entry_count entries were consumed. A file whose footer claimed entry_count=N but whose index region held only N-1 complete entries (plus <4 trailing bytes) was accepted as a partial cache — the existing truncated-entry-header check only fires when >=4 bytes remain. The next save_packed_index would then rewrite a trailer covering only the partial subset, permanently dropping the missing entries (the exact failure the code's own comment warns about). Fix: after the loop, require that exactly entry_count entries were read and that the cursor consumed the whole index region; otherwise roll back (munmap, clear state) and return false, matching the in-loop corruption branches.

A minimal test seam (friend class XNNWeightsCacheTestPeer) is added to the runtime header so the unit test can drive the private fresh-write path directly. Changes are mirrored across the fbcode/ and xplat/ copies.

Differential Revision: D109886666

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/20562

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit fa23731 with merge base 825bd30 ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[meta-codesync]

@JakeStevens has exported this pull request. If you are a Meta employee, you can view the originating Diff in D109886666.

[github-actions]

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.