fix(memory): compare mapped module identity against a mapping, not stat

[not-matthias]

TLDR: This is a regression from the gen2 macro runner, which uses overlayfs.

Stops memory-mode artifact extraction from rejecting unchanged system libraries on nested overlayfs, which left libc frames unresolved in memory flamegraphs.

• names_mapped_file compared the perf MMAP2 (dev, ino) with stat(path). On nested overlayfs, stat can return a per-layer pseudo device while mappings carry the overlay's own device, so every system library was treated as changed and shipped no symbols or unwind data.
• The current identity now comes from a read-only mapping of the file, read back from /proc/self/maps, which the kernel derives the same way as MMAP2.
• Replaced files still get a different inode and are still rejected.

Evidence

• Failing report: 28 changed since it was mapped lines for libc, ld.so and others, recorded minor 56 vs stat 58, same inode. Only 4 modules were saved.
• New root-only test accepts_an_unchanged_file_in_a_nested_overlay builds a nested overlay, asserts stat and the mapping disagree, and checks the file is accepted. It passes in a privileged container. The other module_artifacts tests pass too.

Review notes

• The nested overlay test is #[ignore] because it needs root to mount.
• Not yet re-recorded end to end: whether libc symbols resolve and whether truncated stacks get deeper.

Fixes COD-3746

[codspeed]

Merging this PR will not alter performance

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

✅ 31 untouched benchmarks
⏩ 6 skipped benchmarks¹

---

_{Comparing cod-3746-investigate-unresolved-symbols-and-truncated-memory-call (8c603c2) with main (bcad8db)²}

Footnotes

1. 6 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

2. No successful run was found on main (4421efa) during the generation of this report, so bcad8db was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[greptile-apps]

Confidence Score: 4/5

[Medium risk] Refactors module artifact parsing to bind identity checks to file contents.

The PR is not safe to merge until executor tests are separated by their actual modes so memtrack cannot overlap Valgrind.

Summary

The PR changes memory-mode module extraction to compare mapping identity using a mapped file and parse artifacts from those same bytes. Since the prior review, it also adds uretprobe return-address restoration and moves executor integration tests into a Docker-backed CI job.

• The new test sharder does not reliably separate memory tests from simulation tests, which must not overlap.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  A[Parametrized executor tests] --> B[Generated positional test names]
  B --> C[Name-based mode classifier]
  C --> D[Walltime shard]
  C --> E[Simulation shard]
  D --> F[Concurrent container runs]
  E --> F
  F --> G[Memtrack probes can overlap Valgrind]

Loading

_{Reviews (5) · Last reviewed commit: "fix(memory): compare mapped module ident..."}

[GuillaumeLagrange]

good catch, lgtm