2026-07-20 by Igor De Souza
Pi4J is built for the Raspberry Pi, but the underlying libgpiod-based approach works on any Linux board that exposes its GPIOs through the kernel’s gpiochip interface. In a new article on dev.to, Igor De Souza puts that to the test on an NVIDIA Jetson Nano (an A02 board running the 2019-era JetPack 4.2 / L4T R32.1), with the same simple goal as the recent Banana Pi experiments: blink an LED.
The Jetson Nano’s Tegra SoC doesn’t number its GPIOs the way a Raspberry Pi does. Pins are grouped into named ports (AA, BB, CC, DD, …) rather than a flat BCM range, and NVIDIA’s documentation on how those map to Linux gpiochip line numbers turned out to be scattered across multiple documents, source files, and forum posts, with no single authoritative reference. A newer NVIDIA utility, jetson-gpio-pinmux-lookup, would have made this easier, but it requires a JetPack version newer than what this older board was running.
Working from NVIDIA’s Jetson.GPIO repository, the article derives the offset formula for Tegra: (port_index × 8) + line_number. For header J41, physical pin 12 maps to PJ.07 in Tegra’s own nomenclature, and with port J at index 9, that resolves to line 9 × 8 + 7 = 79 on gpiochip0. As with the Banana Pi board, that offset was verified independently before writing any Java code, using sudo gpioset --mode=wait gpiochip0 79=1.
With the correct offset confirmed, the article uses a draft LineHelper utility to translate the Tegra port/pin notation directly into a Pi4J GPIO address:
LineHelper.getAddress(GpioLineFamily.TEGRA, 'J', 7)
That resolved address is then used the same way any BCM value is elsewhere in Pi4J, and the LED blinked.
Each SoC family — Allwinner, Tegra, Rockchip, and others — turns out to follow its own GPIO addressing rules, but once the pattern for a given family is understood, it’s manageable. The article’s broader point is a proposal for Pi4J’s core: a pluggable LineMapper interface that would let SoC-specific address translation (like the Tegra formula above) live behind a clean extension point, so Pi4J can stay platform-agnostic instead of hardcoding knowledge of every board family it might run on.
Read the full article on dev.to: Running Pi4J on the NVIDIA Jetson Nano.