F0046 fls-3¶
Rule: RP flash is erased and written from core 0 with nothing else reading flash.
| Code | F0046 (stable, never reused) |
| Id | fls-3 |
| Category | fls |
| 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: RP flash erase or write from core 1, or with core 1 started outside spawn_core1, or while DMA reads flash (XIP stalls) |
| ui | off | |
| egui | off | |
| units-strict | off | |
| units-paranoid | off |
Why¶
RP2040/RP2350 flash erase or write issued from core 1, or with core 1 running outside spawn_core1's controlled start-up, or while DMA is reading flash through XIP.
RP's flash program/erase routines execute from RAM precisely because flash can't be read (including XIP-mapped code fetches) while it's being erased or written; core 1 running arbitrary code during a flash op can stall on an XIP read exactly when the flash isn't available to answer it, and DMA reading flash during the op hits the same stall from a different direction.
Example¶
Embedded_bedrock TPL-25 is this exact hazard worked through for the RP flash path the template generates.
How to fix¶
Park core 1 (or route its own flash needs through the same serialised owner, fls-2) and make sure no DMA channel is mid-transfer from flash before issuing an erase or write.
How it is checked¶
src mid, chip RP: flash erase/write call sites, checked against whether core 1 is confirmed parked (via spawn_core1's synchronisation) and whether any DMA channel is configured to read flash concurrently.
Source¶
Ported from embedded_bedrock's firmware check catalogue, FLS-3 (book/checks.md, detail in book/checks/fls.md). Found on: embedded_bedrock TPL-25.
Configure¶
Allow one finding in the code with // spirit-level: allow(fls-3) <reason> (waivers); the code works there too, as in [lints.F0046].