Skip to content
jetson-customize-pcie logo

Jetson Customize Pcie

jetson-customize-pcie

Per-controller PCIe enable / disable / lanes / link-speed for a Jetson Thor or Orin custom carrier via ODMDATA + kernel-DT overlay. Do NOT use for UPHY lane allocation or endpoint-mode bring-up.

nvidia/skills1Kinstalls3.5Kstars

SKILL.md

Full skill instructions

Customize PCIe (per-controller status / lanes / speed)

Overview

PCIe on Tegra264 (Thor, [email protected]) and Tegra234 (Orin, [email protected]) is split across multiple controllers that share the UPHY lane pool with USB3 / MGBE / UFS. Each controller's runtime behavior is determined by two surfaces, both required:

SurfaceTargetAuthoritative for
ODMDATA pcie@N_status=… (+ pcie@N_max-link-speed, pcie@N_pcie-mode, pcie@N_clk-scheme, pcie-cN-endpoint-enable)/​pcie/​pcie@N in BPMP DTBUPHY lane power, refclk gating, controller-side power rails
Kernel-DT overlay on &pcieN/​bus@0/​pcie@<addr> in kernel DTBKernel probe, lane width, link speed, RC/​EP mode

Skipping the kernel overlay on a disable lets the kernel probe a powered-down PHY (link timeouts in dmesg). Skipping the ODMDATA token on a disable leaves BPMP holding the PHY hot.

Agentic, not table-driven — no controller table, no questions.json. Every controller, lane width, schematic-routed receptacle, and authoritative DT node address is discovered at runtime from the docs + DTB + carrier pinmap.

The output is a kernel-DT overlay commit only. Per-controller fragment@N blocks are appended to the composite custom overlay .dts per ../​../​references/​bsp-customization-kernel-dtb.md and committed to the bsp_sources/ hardware repo. /​jetson-build-source compiles the composite to .dtbo and owns its Makefile + flash-conf registration.

This skill MUST NOT edit ODMDATA="...". All ODMDATA tokens (pcie@N_status=…, pcie@N_max-link-speed, pcie@N_pcie-mode, pcie@N_clk-scheme, pcie-cN-endpoint-enable, plus the uphyX-config-N surface tokens and UPHY_CONFIG="" clear) are emitted by /​jetson-customize-uphy in a single atomic commit on the carrier flash-conf fork. The allocation table this skill consumes from the UPHY sidecar already tells the operator which controllers are okay / disabled / per-lane sized; this skill only translates that table into kernel-DT overlay fragments and verifies that the overlay agrees with the ODMDATA already committed by customize-uphy (consistency check in Step 8 — disagreement is reported, not silently fixed).

When to invoke

  • The user says "configure PCIe", "enable PCIe controller", "set PCIe num-lanes", "change PCIe link speed", or asks to flip a pcie@N_status token.
  • A specific PCIe slot or M.2 receptacle doesn't enumerate after flash, OR the link trains at the wrong width / speed.
  • jetson-customize-uphy ran and re-allocated lanes across PCIe controllers (e.g. switched from uphy0-config-7 to uphy0-config-6 enabling PCIe C3); the per-controller side now needs to be brought up.
  • jetson-customize-mgbe reports the QSFP path is wired but the kernel doesn't probe its PCIe-side companion (rare; XFI configurations).

Prerequisites:

  • Active profile with reference_devkit: + custom_carrier: blocks.
  • <source.root_path>/​Linux_for_Tegra/​.git exists (/​jetson-init-source).
  • /​jetson-derive-carrier has run — carrier flash-conf fork is in the overlay tracker.
  • /​jetson-customize-uphy has run — its JSON sidecar at <workspace>/​target-platform/<profile-stem>.jetson-customize-uphy.json drives the per-controller enable decision.
  • Source-of-truth docs registered or supplied at prompt: Adaptation Guide, Module Design Guide, SoC TRM.
  • When custom_carrier: is present, both documents.custom_carrier_schematic AND documents.custom_carrier_pinmux_xls are REQUIRED. Refuse the run if either is missing — routing on a custom carrier cannot be guessed. Reference-devkit-only profiles skip this check.
  • dtc on PATH.

Procedure (summary)

Full step-by-step walkthrough lives in references/​procedure.md. High-level flow:

  1. Resolve active target + open source-of-truth documents (incl. <carrier-pinmap>, <ref-dtb>, <uphy-state>). Refuse if <uphy-state> is missing.
  2. Diff PCIe topology — devkit vs custom carrier — by decompiling <ref-dtb> and grepping the schematic for PEX<N>_* net labels.
  3. AskUserQuestion (multiSelect) — which controllers to customize.
  4. Per-controller verification: pinmap + schematic + pin_verifier.py for PE<N>_CLKREQ_L, PE<N>_RST_L, optional PE<N>_WAKE_L.
  5. Auto-derive per-controller plan (enable from <uphy-state>, lanes / speed from Adaptation Guide, mode hard-pinned to "rc") → mandatory confirm-or-customize gate.
  6. Append per-controller fragment@N blocks (marker /​* custom-bsp: pcie:pcie@<addr> */) to the composite custom overlay .dts in bsp_sources/. Pre-flight dtc + fdtoverlay. Commit via the workflow's preview gate. Do not edit ODMDATA — /​jetson-customize-uphy already emitted pcie@N_status=…, pcie@N_max-link-speed, pcie@N_pcie-mode, pcie@N_clk-scheme, and pcie-cN-endpoint-enable in its single atomic ODMDATA commit. This skill only translates the per-controller plan into kernel-DT overlay fragments.
  7. (Step folded into Step 6 — overlay-only emission.)
  8. Cross-check ODMDATA vs overlay consistency. On a contradictory row, stop and ask the user how to recover the two commits. Never run git reset --hard autonomously.
  9. Write run-state JSON sidecar at <workspace>/​target-platform/<profile-stem>.jetson-customize-pcie.json
    • summary, then drive the downstream next-step chain via sequential AskUserQuestion prompts per references/​procedure.md Step 9. Never substitute a printed "Next step: …" line for the prompts.

Limitations

  • Mode hard-pinned to RC. Endpoint mode is only emitted when the operator passes mode_override="ep" in Step 5c.
  • enable is derived, not asked. UPHY-allocated controllers are mandatorily okay; non-allocated are mandatorily disabled.
  • No upstream BSP edits. Output lands in Linux_for_Tegra/ + bsp_sources/ only.
  • Pre-flight overlay merge is a sanity check, not the production build. /​jetson-build-source is authoritative.
  • Flash-conf overlay registration is out of scope. Owned by /​jetson-build-source Step 5.0a.

Troubleshooting

  • <uphy-state> missing → run /​jetson-customize-uphy first.
  • Slot doesn't enumerate after flash → check dmesg | grep pcie; re-verify ODMDATA pcie@<N>_status=okay and the overlay fragment agree (Step 8 table in references/​procedure.md).
  • Link trains at wrong width → confirm UPHY config in <uphy-state> allocates the expected lane count; the kernel fragment's num-lanes must match.
  • compatible mismatch → fix the composite root, not the fragment. UEFI plugin-manager silently skips on mismatch.
  • Contradictory ODMDATA-vs-overlay row → ask the user; do not auto-git reset --hard. See gotchas.
  • Common pitfalls — see references/​gotchas.md (RC pinning, node- address sourcing, stock-disabled controllers, intra-file handoff with jetson-customize-uphy).

References

More skills from nvidia

cudaq-guide logo
nvidia/skills

cudaq-guide

CUDA-Q onboarding guide for installation, test programs, GPU simulation, QPU hardware, and quantum applications.

3.5K 2.5K
View
data-designer logo
nvidia/skills

data-designer

Use when the user wants to create a dataset, generate synthetic data, or build a data generation pipeline.

3.5K 2.4K
View
deepstream-dev logo
nvidia/skills

deepstream-dev

NVIDIA DeepStream SDK development with Python pyservicemaker API. Use when building video analytics pipelines, GStreamer-based video processing, TensorRT inference integration, object detection/tracking, or Kafka/message broker integration.

3.5K 2.4K
View
rag-eval logo
nvidia/skills

rag-eval

Filesystem RAG benchmarks: corpus/, train.json, evaluate_rag.py (RAGAS quality). Not for prod monitoring, latency/throughput benchmarking (use rag-perf), or evals outside this repo layout.

3.5K 2.3K
View
nemo-retriever logo
nvidia/skills

nemo-retriever

Use when the user wants to search, query, extract, transcribe, describe, quote, filter, or aggregate across documents — PDFs, scanned forms / images (`.jpg` `.png` `.tiff`), Office (`.docx` `.pptx`), text (`.html` `.txt`), audio (`.mp3` `.wav` `.m4a`), or video (`.mp4` `.mov`). Prefer this over native Read / Grep for multi-file or non-PDF corpora. Not for: editing files, web browsing, single-file plain-text lookups, fine-tuning.

3.5K 2.3K
View

Coordinate the end-to-end CAD/source-asset to SimReady workflow. Use for broad requests such as CAD to SimReady, source asset to simulation-ready USD, or prop packaging that require conversion, material/physics assignment, SimReady conformance, validation, and optional package creation; deploy or verify Content Agents services first when property assignment is enabled; route single-stage work through nested references.

3.5K 2.3K
View

Use this skill to bring any vision model from HuggingFace or NVIDIA NGC into an NVIDIA DeepStream pipeline with end-to-end automation: ONNX download, SafeTensors export, TRT engine build, custom nvinfer bbox parser, multi-stream benchmark, and PDF report. Object detection models only.

3.5K 2.3K
View

Top-level workflow skill for USD performance diagnosis and optimization. Handles slow loading, high memory, low FPS, and broad scene-optimization requests; delegates auth/runtime setup to Phase 0 owners.

3.5K 2.3K
View

Vehicle routing (VRP, TSP, PDP) with cuOpt — Python API only. Use when the user is building or solving routing in Python.

3.5K 2.2K
View

Use only to generate or update a governance skill card for a specified existing agent skill directory. Do not use for explaining, listing, comparing, or discussing skill capabilities.

3.5K 2.2K
View

Use as the top-level router for Omniverse Realtime Viewer USD app requests and focused viewer reference documents.

3.5K 2.2K
View
rag-perf logo
nvidia/skills

rag-perf

Performance benchmarking for a deployed NVIDIA RAG Blueprint server: profiling pass + aiperf load test driven by a single YAML config. Not for accuracy / RAGAS scoring (use rag-eval) or for deploying / repairing services (use rag-blueprint).

3.5K 2.1K
View

Popular AI tools

Kaiber logo
Video

Kaiber

Generate, edit, and beat-sync AI video with leading models in one workspace.

Paid
View
Vimcal logo
Productivity

Vimcal

The world's fastest calendar for remote work

Free
View

Transform Your Design with AI Designer by ImgCreator.ai

Freemium
View
Akool AI logo
Content & writing

Akool AI

Revolutionizing Video Production with AI-Powered Creativity

Paid
View

Extend an image past the frame and let AI fill the new aspect ratio.

Freemium
View
StarByFace logo
Security

StarByFace

Discover your celebrity doppelgänger with StarByFace!

Free
View
C

ChainClarity explains 700+ crypto whitepapers in plain English, with layered summaries, comparisons, research tools, alerts, and a $4.99 Pro plan.

Freemium
View
Opus Clip logo
Coding & apps

Opus Clip

Opus.ai: Revolutionize Your Web Experience

Free
View