F0059 irq-1¶
Rule: Interrupt handlers clear exactly the flags they handled.
| Code | F0059 (stable, never reused) |
| Id | irq-1 |
| Category | irq |
| Source | spirit_level |
| Fix | by hand |
Level per tier¶
| Tier | Level | Reason |
|---|---|---|
| small | off | |
| small-workspace | off | |
| medium | off | |
| big | off | |
| ai | off | |
| embedded | deny |
bedrock error: Interrupt flags cleared wholesale (all-ones or the read value written back to a W1C register): an event between the read and the clear is lost; clear exactly the flags handled |
| ui | off | |
| egui | off | |
| units-strict | off | |
| units-paranoid | off |
Why¶
A status-register clear that writes all-ones (or writes back the value it just read) instead of exactly the flags the handler is about to act on.
A W1C (write-1-to-clear) register clears every bit written as 1, not just the ones the code cared about; an event flag that sets between the read and the wholesale clear is lost without the handler ever seeing it. On a ring-style peripheral that means a dropped element nothing else notices.
Example¶
Mcan finding I1/Y6: clearing the whole interrupt status register lost a pending RFxL (Rx FIFO lost message) flag that arrived in the gap between the read and the clear.
How to fix¶
Clear only the bits the handler is about to act on, computed from what it actually read and handled, not a blanket mask.
How it is checked¶
src mid: needs the register's field layout to tell "clear everything" from "clear exactly the bits handled"; a literal all-ones write or a write of the just-read value is the pattern to flag.
Source¶
Ported from embedded_bedrock's firmware check catalogue, IRQ-1 (book/checks.md, detail in book/checks/irq.md). Found on: mcan I1/Y6 (RFxL lost); notion.
Configure¶
Allow one finding in the code with // spirit-level: allow(irq-1) <reason> (waivers); the code works there too, as in [lints.F0059].