Skip to content

Day 41 – Notes: IIO Control Plane v3

🧭 Overview

Day 41 focuses on completing a clean and extensible control plane across:

  • STM32 firmware (register model v3)
  • Linux IIO driver (myadc_iio.c)

Key idea:

Separate what the sensor does from how it notifies events.

🧱 Control Plane Architecture (v3)

Components

ACQ_MODE  → acquisition behavior
IRQ_MASK  → interrupt policy
CMD       → explicit trigger

ACQ_MODE

Defines how the sensor acquires data:

Mode Description
IDLE No sampling
CONTINUOUS Periodic sampling
FIFO Buffered sampling
ONE_SHOT Single sample on command

IRQ_MASK

Selects which event generates an interrupt:

Bit Event
DATA_READY New sample available
FIFO_WATERMARK FIFO reached threshold
FIFO_FULL FIFO full
ONE_SHOT_DONE One-shot completed

CMD

Explicit command interface:

ONE_SHOT_START → trigger one-shot acquisition

🔄 v2 vs v3 Comparison

v2 (coupled design)

CONFIG:
  ENABLE
  FIFO_ENABLE
  DRDY_ENABLE
  IRQ_ENABLE

Problems:

  • Mode + IRQ tightly coupled
  • Hard to extend
  • Driver must understand bit combinations

v3 (decoupled design)

ACQ_MODE + IRQ_MASK + CMD

Advantages:

  • Clear separation of concerns
  • Easier to extend
  • Cleaner Linux driver mapping

🔗 Linux Driver Mapping

Linux Interface Register
mode ACQ_MODE
one_shot CMD
internal policy IRQ_MASK

⚙️ Driver Design Principles

1. Mode is passive

mode = one-shot → no sampling happens

2. Command triggers action

CMD → start acquisition

3. IRQ policy is independent

mode ≠ irq behavior

4. Data path unchanged

  • FIFO path untouched
  • Buffer path untouched

Only control plane updated.

🧠 Key Concepts

One-shot is command-driven

Correct:

mode = one-shot
CMD = start

Incorrect:

mode = one-shot → auto sample ❌

Decoupling

Firmware:

  • Fully decoupled

Linux:

  • Currently uses mode-based default IRQ policy
  • Can be extended later

Deterministic behavior

Every action must be explicit:

  • Set mode
  • Set IRQ mask
  • Trigger acquisition

⚠️ Design Pitfalls

Mixing background task in register tests

Issue:

  • FIFO test corrupted by timer-driven updates

Fix:

  • Separate:

  • register model tests

  • task integration tests

Implicit side effects

Avoid:

  • Mode causing sampling
  • IRQ tied to mode implicitly

📌 Summary

Day 41 establishes:

  • Clean control plane
  • Explicit command-based design
  • Proper separation between mode and IRQ
  • Stable integration between firmware and Linux driver

🚀 Next Topics

  • IRQ policy exposure to user space
  • Blocking one-shot helper
  • IIO trigger integration
  • Driver refinement