fix(pcie): isolate CUDA graph channel lifecycle

[voipmonitor]

Summary

• bind recycled target and draft CUDA graph streams to independent PCIe channels
• checkpoint and roll back channels created by disposable graph captures
• coordinate IPC teardown across ranks and make rollback fail-safe
• keep all MoE execution-plan changes out of this PR

Why

Target and speculative graph managers can reuse CUDA stream identities while retaining different live graph resources. Sharing or prematurely recycling a PCIe channel across those captures can leave stale IPC state or destroy a channel still owned by a replayable graph.

This PR contains only the PCIe lifecycle half of #58 and is based directly on master.

Validation

• tests/comm/test_pcie_dcp_a2a.py
• tests/comm/test_pcie_oneshot.py
• 41 passed
• Ruff lint and format checks passed
• split integration tree is byte-identical to the PCIe files in Fix PCIe graph lifecycle and preserve safe MoE overlap #58

[coderabbitai]

Warning

Review limit reached

@voipmonitor, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 34 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: c67a965f-c879-4a55-819b-49d20ef3d8de

📥 Commits

Reviewing files that changed from the base of the PR and between ec2cd4d and ed988b2.

📒 Files selected for processing (4)

• sparkinfer/comm/pcie/pcie_dcp_a2a.py
• sparkinfer/comm/pcie/pcie_oneshot.py
• tests/comm/test_pcie_dcp_a2a.py
• tests/comm/test_pcie_oneshot.py

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/pcie-graph-lifecycle-20260721

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[voipmonitor]

Manual fallback review completed because the automated review was rate-limited. The channel ownership, capture isolation, rollback ordering, and coordinated IPC teardown paths were inspected against the current master; no actionable defect was found. Focused result: 41 passed; Ruff clean.

@coderabbitai review

[coderabbitai]

@voipmonitor Thanks for the manual fallback review and validation details. I’ll review the current changes with focus on the PCIe channel lifecycle, CUDA graph capture isolation, rollback safety, and coordinated IPC teardown.

✅ Action performed

Review finished.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.