[https://nvbugs/6244695][fix] Revert Pass IPC HMAC key through file descriptor

[chenfeiz0326]

…descriptor (#14378)"

This reverts commit 50ca49f.

Summary by CodeRabbit

Release Notes

• Refactor

  • Streamlined inter-process communication (IPC) authentication mechanism for internal proxy processes by simplifying key management and eliminating legacy file descriptor-based passing.

• Tests

  • Removed outdated tests for deprecated authentication method.

Description

Test Coverage

PR Checklist

Please review the following before submitting your PR:

• PR description clearly explains what and why. If using CodeRabbit's summary, please make sure it makes sense.

• PR Follows TRT-LLM CODING GUIDELINES to the best of your knowledge.

• Test cases are provided for new code paths (see test instructions)

• If PR introduces API changes, an appropriate PR label is added - either api-compatible or api-breaking. For api-breaking, include BREAKING in the PR title.

• Any new dependencies have been scanned for license and vulnerabilities

• CODEOWNERS updated if ownership changes

• Documentation updated as needed

• Update tava architecture diagram if there is a significant design change in PR.

• The reviewers assigned automatically/manually are appropriate for the PR.

• Please check this after reviewing the above items as appropriate for this PR.

GitHub Bot Help

To see a list of available CI bot commands, please comment /bot help.

[coderabbitai]

📝 Walkthrough

Walkthrough

Spawn-proxy IPC authentication migrates from file-descriptor-based passing with cached helpers to direct hex-encoded environment variable export. The core accessor function, launcher script generation, and MPI server integration all simplify by removing FD plumbing and cached state, and the test module covering the old FD mechanism is removed.

Changes

IPC HMAC Key Mechanism Refactor

Layer / File(s)	Summary
IPC HMAC key contract and accessor tensorrt_llm/executor/utils.py	Enum adds TLLM_SPAWN_PROXY_PROCESS_IPC_HMAC_KEY environment variable, removes TLLM_SPAWN_PROXY_PROCESS_IPC_HMAC_KEY_FD. Function get_spawn_proxy_process_ipc_hmac_key_env() now reads hex-encoded key directly from env, asserts presence, and converts to bytes; removes FD-based reading and cached normalization helpers.
HMAC key generation in launch script tensorrt_llm/llmapi/trtllm-llmapi-launch	Generates random 32-byte hex key via openssl rand and exports as TLLM_SPAWN_PROXY_PROCESS_IPC_HMAC_KEY. Rank-0 task execution and MGMN leader invocation remove the run_with_ipc_hmac_key wrapper, execute directly, and preserve exit-code capture.
Disaggregated MPI leader HMAC key integration tensorrt_llm/commands/serve.py	Imports LlmLauncherEnvs, directly sets TLLM_SPAWN_PROXY_PROCESS_IPC_HMAC_KEY in child-process environment dict, removes FD assertions, and launches proxy via simplified subprocess.Popen without file-descriptor passing. Signal handler cleanup logic is removed from the finally block.

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly Related PRs

• NVIDIA/TensorRT-LLM#14378: Prior work introducing the FD-based TLLM_SPAWN_PROXY_PROCESS_IPC_HMAC_KEY_FD mechanism that this PR replaces with direct environment variable passing.

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 75.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.
Description check	⚠️ Warning	The PR description lacks substantive explanation beyond the revert statement; required template sections (Description, Test Coverage) are empty with only placeholder comments shown.	Add a clear explanation of why the previous IPC HMAC key file descriptor approach is being reverted and document any test coverage or validation that verifies the revert works correctly.

✅ Passed checks (3 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Title check	✅ Passed	The title clearly identifies this as a revert of a previous commit addressing IPC HMAC key handling via file descriptors, which matches the changeset modifications.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@tensorrt_llm/executor/utils.py`:
- Line 26: The new constant TLLM_SPAWN_PROXY_PROCESS_IPC_HMAC_KEY exposes the
IPC HMAC secret via process environment—restore the previous non-environment
mechanism: remove usage of TLLM_SPAWN_PROXY_PROCESS_IPC_HMAC_KEY from any
subprocess env setup and instead pass the HMAC key over the narrowly-scoped
inherited-FD / dedicated IPC channel used by ZeroMqQueue (or via explicit file
descriptor/pipe passed to the child), update the code paths in ZeroMqQueue and
the launcher/spawn code that currently read/write this env var so they accept
the key from the inherited-FD/IPC handle, and delete the environment-constant
definition so the secret is never placed into os.environ or subprocess env.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 90405160-bfc2-4fc3-a9e1-5c178eafe732

📥 Commits

Reviewing files that changed from the base of the PR and between f20858c and 99d27ff.

📒 Files selected for processing (4)

• tensorrt_llm/commands/serve.py
• tensorrt_llm/executor/utils.py
• tensorrt_llm/llmapi/trtllm-llmapi-launch
• tests/unittest/executor/test_launcher_envs.py

💤 Files with no reviewable changes (1)

• tests/unittest/executor/test_launcher_envs.py

[coderabbitai]

⚠️ Potential issue | 🟠 Major | 🏗️ Heavy lift

Keep the IPC HMAC key off the environment.

ZeroMqQueue uses this key to authenticate pickle traffic before pickle.loads() in tensorrt_llm/executor/ipc.py:33-103,354-389. Switching the contract back to TLLM_SPAWN_PROXY_PROCESS_IPC_HMAC_KEY means the secret now propagates through process environments in both launcher paths instead of staying on a narrowly scoped channel like the previous inherited-FD mechanism. That weakens the only barrier guarding unauthorized IPC deserialization.

Also applies to: 39-43

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tensorrt_llm/executor/utils.py` at line 26, The new constant
TLLM_SPAWN_PROXY_PROCESS_IPC_HMAC_KEY exposes the IPC HMAC secret via process
environment—restore the previous non-environment mechanism: remove usage of
TLLM_SPAWN_PROXY_PROCESS_IPC_HMAC_KEY from any subprocess env setup and instead
pass the HMAC key over the narrowly-scoped inherited-FD / dedicated IPC channel
used by ZeroMqQueue (or via explicit file descriptor/pipe passed to the child),
update the code paths in ZeroMqQueue and the launcher/spawn code that currently
read/write this env var so they accept the key from the inherited-FD/IPC handle,
and delete the environment-constant definition so the secret is never placed
into os.environ or subprocess env.

[chenfeiz0326]

/bot run --post-merge --disable-fail-fast

[tensorrt-cicd]

PR_Github #51187 [ run ] triggered by Bot. Commit: 588868b Link to invocation

[tensorrt-cicd]

PR_Github #51187 [ run ] completed with state FAILURE. Commit: 588868b
/LLM/main/L0_MergeRequest_PR pipeline #40617 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

[chenfeiz0326]

/bot run --disable-fail-fast --stage-list "*PerfSanity*"

[tensorrt-cicd]

PR_Github #51230 [ run ] triggered by Bot. Commit: a003643 Link to invocation

[tensorrt-cicd]

PR_Github #51230 [ run ] completed with state FAILURE. Commit: a003643
/LLM/main/L0_MergeRequest_PR pipeline #40653 (Partly Tested) completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation