Skip to content
jetson-customize-pinmux logo

Jetson Customize Pinmux

jetson-customize-pinmux

Per-pin SFIO / direction / initial-state configurator for a Jetson Orin or Thor custom carrier from the pinmux XLSM. Do NOT use for kernel-DT overlay or ODMDATA edits.

nvidia/skills1.1Kinstalls3.5Kstars

SKILL.md

Full skill instructions

Customize pinmux (per-pin SFIO / direction / state)

Overview

The Tegra pinmux spreadsheet (.xlsm) is the ground truth for every CVM ball: SoC pin name, supported SFIOs, customer-selected function, direction, and initial state. This skill parses that XLSM, runs a per-pin Q1–Q6 interactive loop, and emits the three BCT DTSIs (pinmux, gpio, padvoltage) in one shot into the overlay tracker at <source.root_path>/​Linux_for_Tegra/​bootloader/.

Unlike sibling skills jetson-customize-uphy / jetson-customize-pcie / jetson-customize-camera, pinmux has no kernel-DT overlay surface and no ODMDATA edit. The XLSM is the source of truth; the three emitted DTSIs land at flash time via the carrier conf's PINMUX_CONFIG= / GPIOINT_CONFIG= / PMC_CONFIG= references (which /​jetson-derive-carrier set up).

Pad classification (silicon-fixed): only BD* and BI* pads have configurable pull / drive / open-drain attributes. LP5XA_*, UPHYDS_*, DP_SINGLE_*, BDMIPI16X_*, BDUSB2_*, OSCI27_* are fixed-function and skip Q4–Q6 (configurable: no).

The bundled scripts/​modify_pinmux.py is the workhorse: it parses the XLSM via openpyxl>=3.1, builds the per-carrier pinmap JSON, captures pin edits into a session shim, and (on generate) writes the three DTSIs.

When to invoke

  • The user says "configure pin", "set SFIO", "edit pinmux DTSI", "set pin direction", "set initial state", or asks to repurpose a CVM ball (e.g. flip a pin between GPIO and a peripheral function).
  • A sibling skill (jetson-customize-camera, jetson-customize-pcie, jetson-customize-usb, jetson-customize-mgbe) reports an HSIO pin mismatch via pin_verifier.py and the user wants to fix it.
  • The user pre-derived a custom carrier with /​jetson-derive-carrier and now wants to author the pinmux from a freshly-edited .xlsm.

Prerequisites:

  • Active profile selected (target-platform/​active_target.yml → <profile>.yaml with reference_devkit: AND custom_carrier:).
  • <source.root_path>/​Linux_for_Tegra/ exists as a git repo (/​jetson-init-source).
  • /​jetson-derive-carrier has run — the three pinmux-side BCT DTSIs (PINMUX_CONFIG, GPIOINT_CONFIG, PMC_CONFIG references in the carrier conf) exist in the overlay tracker.
  • A pinmux .xlsm is registered in the active profile at documents.custom_carrier_pinmux_xls (preferred when custom-carrier- specific) or documents.ref_devkit_pinmux_xls (fallback). The bundled modify_pinmux.py requires openpyxl>=3.1.

Procedure

See references/​procedure.md for the full step-by-step procedure (Steps 1–8). Summary:

  1. Resolve active target + XLSM. Validate active profile, custom_carrier:, overlay-tracker prerequisites; resolve the pinmux .xlsm path from documents.custom_carrier_pinmux_xls → documents.ref_devkit_pinmux_xls → single XLSM under documents.root_path → user prompt.
  2. Probe. Run modify_pinmux.py probe to parse the XLSM into the per-skill scratch <KB>/​pinmap/<custom-carrier>.json plus session.json shim.
  3. Lookup. Resolve a free-form user query (CVM ball, Verilog name, signal, DT pin) via modify_pinmux.py lookup; surface supported SFIO list, defaults, and configurable: yes/​no.
  4. Set-pin (HARD GATE — Q1–Q6 via AskUserQuestion). Q1–Q3 (sfio / direction / initial_state) always asked; Q4–Q6 (pull / drive_type / open_drain) only when configurable: yes. tristate and e_input are derived from direction, never asked.
  5. Generate. modify_pinmux.py generate --out-dir <source.root_path>/​Linux_for_Tegra/​bootloader/ (root, not bootloader/​generic/​BCT/ — derive-carrier .dts forks live there, do not colocate). Emits: tegra<soc>-mb1-bct-{pinmux,gpio,padvoltage}-<carrier-key>.dtsi. <carrier-key> comes from the carrier conf's PINMUX_CONFIG= reference, NOT the kebab-cased carrier name.
  6. Commit (single batched commit per workflow rule). All three DTSIs are one logical edit → one customization commit. Run the commit-preview gate before each commit.
  7. Run-state sidecar + session shim. Write the user-facing <profile-stem>.jetson-customize-pinmux.json sidecar and the transient session.json shim under <workspace>/​target-platform/.
  8. Summary. Emit the standard one-line + table summary.

Gotchas

  • No kernel-DT overlay; no OVERLAY_DTB_FILE edit; no render_conf.py hand-off. This skill ends at the three BCT DTSIs — the carrier conf already references them via PINMUX_CONFIG= / GPIOINT_CONFIG= / PMC_CONFIG= (set up by /​jetson-derive-carrier).
  • Re-point the .dts wrapper's #include after generate. /​jetson-derive-carrier forks the .dts wrappers at bootloader/​generic/​BCT/, but their #include lines may still pull the upstream devkit .dtsi (e.g. …-p3834-xxxx-p4071-0000.dtsi). After generate writes the new <CARRIER_KEY>.dtsi to bootloader/ root, edit each wrapper's #include to the new filename — by bare basename (#include "tegra<soc>-mb1-bct-pinmux-<CARRIER_KEY>.dtsi"), not ../​../… filesystem-relative. The BCT build's cpp -I bootloader/ resolves bare basenames; that's the convention every other BCT include in the tree follows. Roll the wrapper edits into the same customization commit as the three DTSIs. See references/​procedure.md Step 5 ("Sanity-check the carrier .dts wrapper").
  • Q4–Q6 gated on configurable: yes. Asking pull / drive_type / open_drain on a fixed-function pad (LP5XA_*, UPHYDS_*, BDMIPI16X_*, etc.) is silently dropped by the script and confuses the user. lookup prints configurable: yes/​no — always check it before prompting Q4–Q6.
  • tristate and e_input are derived, never asked. unused → tristate=ENABLE; input / bidirectional → enable-input=ENABLE. Exposing them as separate prompts produces inconsistent DTSIs.
  • sfio=gpio requires a parseable gpio=GPIOn_PD.NN entry in the pinmap row's sfio list. Pins without one are GPIO-incapable silicon; set-pin rejects the call. Surface the rejection — don't silently fall back to a non-GPIO SFIO.
  • Marker idempotency. Every per-pin edit carries // custom-bsp: pinmux on the closing brace; gpio default-state entries carry the same marker as a trailing comment. Re-running generate must detect and update — never duplicate.
  • modify_pinmux.py is unchanged from the original framework — it reads its own session.json shim under --kb-dir. The shim is regenerated each run from the active profile + the user-facing sidecar. Do not hand-edit the shim; it's transient.
  • Multiple pinmux DTSI variants per Thor module SKU. Some carrier pins live in a different DTSI variant than the one the carrier conf references. modify_pinmux.py commit (legacy patch-in-place flow) tolerates missing per-pin blocks via pinmux.warnings[] rather than failing. Surface the warning; point at the alternate DTSI variant.
  • Don't touch the upstream BSP at <bsp_image.root_path>. All edits land in <source.root_path>/​Linux_for_Tegra/​bootloader/ under the pristine + customization commit pattern.

Available Scripts

ScriptPurposeArguments
scripts/​modify_pinmux.pyXLSM parser + per-pin DTSI generator. Invoked via run_script() from Steps 3-6 with the subcommand of the current phase.probe | lookup | set-pin | apply | generate | commit [...] (see --help)
scripts/​generate_dtsi.pyRenders pinmux/​GPIO/​padvoltage DTSI fragments from the bundled session state. Called by modify_pinmux.py generate.--session <path> --out-dir <dir>

Invoke from the skill body as a subprocess via run_script():

# run_script: probe the carrier pinmux XLSM and write a session state
scripts/​modify_pinmux.py probe --xlsm carrier.xlsm --session .pinmux-session.json

# run_script: render DTSI fragments from the final session state
scripts/​modify_pinmux.py generate --session .pinmux-session.json --out-dir bsp_sources/​pinmux/

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