feat(memtrack): capture allocation stacks in eBPF

[not-matthias]

Adds raw allocation-stack capture to memtrack and ships everything needed for off-box unwinding.

At every allocator entry the eBPF program copies the caller's user stack (chunked, budget-capped, FNV-1a-hashed) and walks frame pointers in-kernel. Definitions are deduplicated by hash in-kernel: one StackDefinition event (raw bytes, registers, FP chain) per distinct stack, referenced by a stack_hash on allocation events. Everything flows through the ordinary MemtrackArtifact event stream; unwinding and symbolication happen server-side later.

After a capture-enabled run, the runner decodes the artifact and reuses the perf walltime machinery to dump per-module symbols, unwind data, and debug info into the profile folder, plus a memory_metadata.json with per-pid process names.

Refs COD-3222

[codspeed]

Merging this PR will degrade performance by 33.59%

⚠️ 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.

❌ 3 (👁 3) regressed benchmarks
✅ 16 untouched benchmarks
🆕 14 new benchmarks
⏩ 4 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
🆕	WallTime	encode_events_realistic[1]	N/A	733.4 ms	N/A
🆕	WallTime	encode_events_realistic[2]	N/A	432.6 ms	N/A
🆕	WallTime	write_stack_events	N/A	186.5 ms	N/A
🆕	Simulation	encode_events_realistic[1]	N/A	1.3 s	N/A
🆕	Simulation	encode_events_realistic[2]	N/A	1.3 s	N/A
🆕	Simulation	write_stack_events	N/A	405.5 ms	N/A
🆕	Memory	encode_events_realistic[1]	N/A	121.1 MB	N/A
🆕	Memory	encode_events_realistic[16]	N/A	131.5 MB	N/A
🆕	Memory	encode_events_realistic[2]	N/A	121.1 MB	N/A
🆕	Memory	encode_events_realistic[4]	N/A	121.1 MB	N/A
🆕	Memory	encode_events_realistic[8]	N/A	123.8 MB	N/A
🆕	Memory	write_events[10000]	N/A	2.1 MB	N/A
🆕	Memory	write_events[100000]	N/A	9.6 MB	N/A
🆕	Memory	write_stack_events	N/A	65.6 MB	N/A
👁	WallTime	encode_events_realistic[16]	121.1 ms	203.1 ms	-40.34%
👁	WallTime	encode_events_realistic[4]	214.4 ms	288.1 ms	-25.57%
👁	WallTime	encode_events_realistic[8]	143.1 ms	216.9 ms	-34.03%

---

_{Comparing cod-3222-add-ebpf-based-dwarffp-unwinding (d0f3114) with main (6c953d6)}

Footnotes

1. 4 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. ↩

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge; no new actionable defect remains, and all previous findings were resolved, withdrawn, or explicitly accepted.

Summary

Adds eBPF allocation-stack capture and the supporting artifact, module-metadata, mapping, test, and CI infrastructure for later off-box unwinding.

• Captures deduplicated raw stack definitions, registers, and frame-pointer chains alongside allocation stack hashes.
• Extends memtrack artifacts and runner processing with mapping and module metadata needed for symbolication.
• Batches ring-buffer delivery to reduce channel-allocation overhead.
• Adds stack integration tests and dedicated memtrack performance benchmarks.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  A[Allocator uprobe] --> B[Capture registers and stack bytes]
  B --> C[Deduplicate by stack hash]
  C --> D[Stack-definition ring buffer]
  A --> E[Allocation event with stack hash]
  E --> F[Event ring buffer]
  D --> G[Batched userspace channel]
  F --> G
  G --> H[Memtrack artifact]
  H --> I[Runner extracts mappings and module metadata]
  I --> J[Off-box unwind and symbolication]

Loading

_{Reviews (4) · Last reviewed commit: "feat(runner): add experimental flag to c..."}

[GuillaumeLagrange]

OLGTM, quite a big piece of work