Jetson ESP-Hosted Host Code — lectures¶
Source boundary — reviewed October 18, 2026: this is the historical NG fork at ce36dec9294988e598f04e84c9d28092ca9e9786. Board observations are attributed to that fork, not rerun here. Complete the source-reading lab; it explains the current migration path and defects in these abbreviated source excerpts. Snippets containing ... are reading excerpts, not standalone compilable programs.
Work through these in order. Each file is one session built around the real esp_hosted_ng/host/ Linux host code.
Hub: Jetson ESP-Hosted Host Code guide
| Lecture | Topic | Main kernel interfaces | Main concepts |
|---|---|---|---|
| Lecture-01 | Linux mental model for this host stack | cfg80211, HCI, netdev, SPI transport split |
subsystem boundaries |
| Lecture-02 | Build, load, and board policy on Jetson | kernel modules, module_param, spidev ownership |
board policy vs driver code |
| Lecture-03 | SPI transport, GPIOs, and IRQ-driven bring-up | gpio_to_irq, request_irq, SPI device reuse |
interrupt-driven transport |
| Lecture-04 | How Wi-Fi becomes wlan0 |
wiphy_new, wiphy_register, register_netdevice |
Linux wireless subsystem integration |
| Lecture-05 | How BLE becomes hci0 and how to validate the full path |
hci_alloc_dev, hci_register_dev, hci_recv_frame |
HCI and BLE host/controller split |
Use one evidence chain across the five lectures¶
L1 produces a subsystem map; L2 produces a build/board manifest; L3 produces an IRQ-to-transfer trace; L4 verifies Wi-Fi registration and traffic separately; L5 verifies Bluetooth controller registration, an actual exchange, and recovery. Each result names the host kernel, fork commit, peer firmware, carrier/pinmux, and connection used.
At the final review, restart the peer while a bounded test is running and explain whether pending buffers are drained, cancelled, or retried. Do not mix Wi-Fi data-plane success with Bluetooth success just because both share a transport.
Carry forward: A board-to-subsystem evidence chain with separate Wi-Fi/Bluetooth results. Continue with the next course and reuse that evidence.