Migrating from bedrock-lint¶
spirit_level carries every automatic check of embedded_bedrock's firmware lint, bedrock-lint: 97 checks, each a lint with the same id in kebab-case (EXE-3 is exe-3), the same fixtures and the same findings. They live in the embedded tier, which stands beside the size tiers, and one more lint, firmware-readable, reports the files bedrock-lint printed as loader notes.
Running it¶
| bedrock-lint | spirit_level |
|---|---|
bedrock-lint check fw/board in the firmware crate |
spl check in the workspace: every firmware crate of it is checked |
--elf target/<triple>/release/<name> |
--elf target/<triple>/release/<name>, repeated for several firmware crates; an ELF goes with the crate that has a binary of its file name, or with the only firmware crate |
--only EXE-3,STK |
--only exe-3 --only stk (a lint id or a category) |
--format sarif |
--format json; SARIF is planned (FEATURES ENG-5) |
--fail-on warning (the default) |
a deny finding fails the run, a warn one never does: set the warnings that must fail to deny in spirit_level.toml |
--root (where discovery stops) |
discovery stops at the workspace root |
checks list --all, explain EXE-3 |
spl lints, spl explain exe-3, and the lint pages |
code-size <ELF> --by-crate |
not ported: a report, not a check |
Firmware crates¶
A firmware crate is a workspace member with a binary whose root is #![no_std], or with a memory.x, a bedrock_fw.json or a -none- target in its .cargo/config.toml ([build] target or a [target.<triple>] section). [package.metadata.spirit-level] firmware = true (or false) in the crate's Cargo.toml decides it by hand.
A firmware binary is a firmware crate that builds a binary: the crates an ELF comes from. A workspace without one, such as a host workspace with no_std libraries, has the lints that read the ELF skipped: the run says so and passes.
Configuration¶
bedrock-lint.toml in the firmware crate becomes the workspace's spirit_level.toml:
# bedrock-lint.toml
baseline = "bedrock-lint.baseline.json"
[severity]
"EXE-12" = "warning"
[skip]
"STK-4" = "paint-stack is checked on the bench"
[limits]
region-percent = 95
frame = 4096
# spirit_level.toml
extends = ["small", "embedded"]
[lints.exe-12]
level = "warn"
reason = "the generated server trait wants async handlers"
[lints.stk-4]
level = "allow"
reason = "paint-stack is checked on the bench"
[lints.siz-1]
region-percent = 95
reason = "the bootloader leaves no room to grow"
[lints.stk-6]
frame = 4096
reason = "the USB stack's frame is measured at 3.9 KiB"
| bedrock-lint.toml | spirit_level.toml |
|---|---|
[severity] "ID" = "error" / "warning" / "suggestion" |
[lints.<id>] level = "deny" / "warn" / "warn", with a reason |
[skip] "ID" = "reason" |
[lints.<id>] level = "allow" with the reason; for one crate or some paths, an exception |
[limits] region-percent (default 90) |
parameter region-percent of siz-1 |
[limits] frame (2048) |
parameter frame of stk-6 |
[limits] move-frame (1024) |
parameter move-frame of stk-5 |
[limits] data-init (1024) |
parameter data-init of siz-6 |
[limits] fmt-bytes (1024) |
parameter fmt-bytes of siz-5 |
baseline = "...", --baseline, --write-baseline |
no baseline yet (planned: FEATURES ENG-6); until then a known finding gets a waiver or an exception with its reason, and expires when it is meant to be fixed |
Overrides may expire (expires = "2027-03-31" or a release); see configuration.
Waivers¶
// bedrock-lint: allow(EXE-3) the ROM leaves this vector at P2 becomes // spirit-level: allow(exe-3) the ROM leaves this vector at P2, in Rust sources (build.rs included) and in Cargo.toml files, with the same placement rules (waivers). A waiver without a reason is waiver-invalid, one that allows nothing waiver-unused. For a finding in .cargo/config.toml, memory.x or bedrock_fw.json, use an exception by path in spirit_level.toml.
What runs where¶
bedrock-lint reported a check as not run when the firmware lacked what it needs (no ELF, no memory.x, no chip). In spirit_level only the ELF is a need of the run: with a firmware binary they are enabled on, the lints that read it say "not run" without --elf, and the run exits with 3; without one they are skipped. The other needs are per crate: a lint checks the firmware crates that have a memory.x, a build script, a known chip, Thumb code in the ELF, and finds nothing in the others. The chip comes from bedrock_fw.json, else from the HAL's chip feature (embassy-stm32, embassy-rp, embassy-nrf).
Places are relative to the workspace root (fw/src/main.rs), not to the firmware crate.
Every bedrock check¶
| bedrock | spirit_level | Severity → level | Check |
|---|---|---|---|
EXE-1 |
exe-1 |
error → deny |
A task with a hard answer deadline (a PHY that arms RX DMA only inside receive().await, a software … |
EXE-2 |
exe-2 |
error → deny |
Blocking call (block_for, blocking_* HAL calls, nb::block!, busy-wait loops, … |
EXE-3 |
exe-3 |
error → deny |
Interrupt executor started before its IRQ priority is set: start() unmasks the IRQ at the default … |
EXE-4 |
exe-4 |
warning → warn |
Interrupt executor priority not below the peripheral, DMA and time-driver interrupts that wake its tasks … |
EXE-5 |
exe-5 |
error → deny |
Interrupt executor on a vector that a driver also binds (bind_interrupts!) or that a used peripheral … |
EXE-6 |
exe-6 |
error → deny |
State shared between a task on an interrupt executor and thread-mode code behind ThreadModeRawMutex / … |
EXE-7 |
exe-7 |
warning → warn |
Loop in a shared task that works until a queue is empty (or logs in a loop) without a bound and a yield: … |
EXE-8 |
exe-8 |
warning → warn |
Request server waits on a bus or a slow peripheral inside a handler: defer the transfer to the task that … |
EXE-9 |
exe-9 |
error → deny |
Future polled with a no-op waker (poll_once, noop_waker) that can reach embassy_time: embassy-time … |
EXE-10 |
exe-10 |
error → deny |
Two crates provide the executor and __pender (embassy-rp or embassy-stm32 executor-* features and … |
EXE-11 |
not ported: bedrock has no automatic check for it yet (catalogue: elf hard) |
warning | Task moved to an interrupt executor without a new look at the stack: it runs on top of the deepest … |
EXE-12 |
exe-12 |
suggestion → warn |
async fn that never awaits (handlers especially): each one is a state in the dispatcher future and costs … |
EXE-13 |
exe-13 |
warning → warn |
Ring-buffered reader (RingBufferedUartRx::read) in a loop without batching: a busy line gives 1-byte … |
IRQ-1 |
irq-1 |
error → deny |
Interrupt flags cleared wholesale (all-ones or the read value written back to a W1C register): an event … |
IRQ-2 |
irq-2 |
error → deny |
rc_w1 / W1C flags cleared with modify() (read-modify-write) instead of a write of the one bit: it … |
IRQ-3 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | Handler that doesn't check its source flag before acting, or clears it late (double interrupt) |
IRQ-4 |
irq-4 |
error → deny |
Several drivers on one shared NVIC vector, and a driver that disables "its" interrupt on drop: the others … |
IRQ-5 |
irq-5 |
warning → warn |
Interrupt line routed (ILS) but not enabled (ILE), or enabled without a handler bound |
IRQ-6 |
irq-6 |
warning → warn |
Nested disable / enable of interrupts by hand: an inner enable turns them on under an outer section; use … |
IRQ-7 |
not ported: bedrock has no automatic check for it yet (catalogue: elf mid) |
suggestion | Interrupt does more than capture (formatting, encoding, logging): a few stores and a counter, the rest in … |
IRQ-8 |
irq-8 |
warning → warn |
Unused interrupt vectors not caught: DefaultHandler must record the IRQ number (raw: a static, the crash … |
DMA-1 |
dma-1 |
error → deny |
DMA buffer in memory the DMA master can't reach (H7 DTCM for IDMA, SDMMC, USB OTG; flash for some … |
DMA-2 |
dma-2 |
error → deny |
D-cache on with DMA buffers in cacheable RAM and no clean / invalidate or MPU non-cacheable region |
DMA-3 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | Receive DMA armed only inside an await (embassy UCPD receive(), one-shot UART reads): data that arrives … |
DMA-4 |
dma-4 |
warning → warn |
DMA ring overrun not counted: every ring consumer counts overruns with cnt! |
DMA-5 |
not ported: bedrock has no automatic check for it yet (catalogue: elf mid) |
suggestion | A per-element % in the ring copy on a core without a divider (embassy ReadableDmaRingBuffer per-byte … |
DMA-6 |
dma-6 |
warning → warn |
Erratum asks for DMA where the code polls (G0 UCPD TX underflow then TXHRST corrupts the next CRC: ES0548 … |
DMA-7 |
not ported: a review check (a person or an agent reads the code or the datasheet) | suggestion | "Things just hang": suspect DMA into a wrong memory region first |
STK-1 |
stk-1 |
error → deny |
flip-link not the linker: linker = "flip-link" under [target.<triple>] in .cargo/config.toml |
STK-2 |
stk-2 |
error → deny |
cargo:rustc-linker= (or another unknown cargo: key) printed by build.rs: not a build-script directive, … |
STK-3 |
stk-3 |
error → deny |
Stack not below .data / .bss in the ELF (_stack_start != start of .data), or a zero-size stack |
STK-4 |
stk-4 |
error → deny |
cortex-m-rt paint-stack on before its BusFault is understood |
STK-5 |
stk-5 |
warning → warn |
Big state built by value and moved into a static (StaticCell::init(Big::new())): the frame holds it … |
STK-6 |
stk-6 |
warning → warn |
Stack frame over a budget (per function) or a worst-case stack depth over the stack region |
STK-7 |
not ported: checked on hardware | error | Stack overflow never provoked on hardware after a RAM layout change: it must end in a fault, not in … |
MEM-1 |
mem-1 |
error → deny |
FLASH length in memory.x covers a region kept for something else (settings, bootloader state, DFU): the … |
MEM-2 |
mem-2 |
warning → warn |
Extra RAM section with a tail . = ALIGN(n) or a symbol in it: lld keeps an empty section that flip-link … |
MEM-3 |
mem-3 |
error → deny |
RAM init or zeroing range overlaps .uninit (crash record, handoff data) |
MEM-4 |
not ported: bedrock has no automatic check for it yet (catalogue: src mid) |
error | Extra SRAM bank used before its RCC clock enable |
MEM-5 |
not ported: bedrock has no automatic check for it yet (catalogue: cfg mid + elf mid) |
error | Backup-register map overlaps: bkp_cnt! buffer, boot word, boot reason and board state on the same … |
MEM-6 |
mem-6 |
warning → warn |
.cnt_bkp_buffer not at the backup-register address |
MEM-7 |
mem-7 |
warning → warn |
Flash partition symbols for embassy-boot written as absolute addresses where offsets are expected (`- … |
MEM-8 |
not ported: bedrock has no automatic check for it yet (catalogue: cfg mid, chip table) |
warning | Settings store smaller than two erase pages, or on the code's bank of a dual-bank part (an erase stalls … |
MEM-9 |
not ported: bedrock has no automatic check for it yet (catalogue: elf mid, chip table) |
error | Data that must survive the ROM bootloader kept in RAM the ROM uses (H7: AXI SRAM from 0x24000000, DTCM … |
MEM-10 |
mem-10 |
warning → warn |
RWX LOAD segment or a writable symbol in .rodata (static mut in a read-only section) |
MEM-11 |
not ported: bedrock has no automatic check for it yet (catalogue: cfg easy) |
warning | Linker warnings ignored (section alignment, .text not 8-aligned after .start_block on RP235x) |
MEM-12 |
mem-12 |
error → deny |
defmt in the dependencies without -Tdefmt.x (no .defmt section) |
SIZ-1 |
siz-1 |
warning → warn |
Flash or RAM use over a threshold (default 90 % of the region) in a shipped profile |
SIZ-2 |
not ported: bedrock has no automatic check for it yet (catalogue: cfg easy) |
warning | Dev profile not linked in CI: debug assertions and overflow checks add 18-30 KiB and overflow unnoticed |
SIZ-3 |
siz-3 |
suggestion → warn |
Per-package opt-level together with lto = "fat": the override does nothing |
SIZ-4 |
siz-4 |
suggestion → warn |
Release profile without codegen-units = 1, lto, panic = "abort", and an opt-level chosen by … |
SIZ-5 |
siz-5 |
suggestion → warn |
core::fmt pulled in: Result::unwrap / expect (through &dyn Debug), Debug2Format, a panic handler … |
SIZ-6 |
siz-6 |
suggestion → warn |
Large .data initializer (a static with a non-zero const value): zero-init and set it up in place |
SIZ-7 |
not ported: bedrock has no automatic check for it yet (catalogue: cfg mid) |
suggestion | A shared crate turns on wide types or features for every firmware (ww_si u64: 7.7 KB) |
SIZ-8 |
not ported: bedrock has no automatic check for it yet (catalogue: elf mid) |
suggestion | A cargo feature meant to cut code saves nothing (bodies gated, dispatch and ser/des not): measure what … |
SIZ-9 |
not ported: bedrock has no automatic check for it yet (catalogue: elf hard) |
suggestion | Generic code compiled once per const parameter (ObserveTable<N>), unique constants that stop identical … |
SIZ-10 |
not ported: a review check (a person or an agent reads the code or the datasheet) | suggestion | Size comments in Cargo.toml or code without a date and profile: they go stale and mislead |
FLT-1 |
flt-1 |
error → deny |
HardFault handler that spins (loop {}) in a shipped build: record (crash capture) and reset |
FLT-2 |
flt-2 |
error → deny |
defmt, cnt!, unwrap or anything that can panic or nest-log inside HardFault / DefaultHandler: … |
FLT-3 |
flt-3 |
warning → warn |
bedrock_crash::exc_return!() not the first statement of HardFault (LR changes) |
FLT-4 |
flt-4 |
warning → warn |
No UsageFault, MemManage and BusFault handlers (or not enabled in SHCSR) where the core has them |
FLT-5 |
flt-5 |
warning → warn |
Reset cause not read and reported at boot (power-on, brown-out, watchdog, software, pin, lockup) |
FLT-6 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | Crash record or RAM handoff read after a power-on or brown-out (RAM content undefined, ECC RAM faults) or … |
FLT-7 |
flt-7 |
error → deny |
No watchdog, or a watchdog fed from an interrupt or from a loop not bounded by a timeout; window not … |
FLT-8 |
flt-8 |
warning → warn |
No boot-loop rescue: count boots without a healthy mark in a backup register, fall back to the ROM … |
FLT-9 |
flt-9 |
warning → warn |
Healthy mark never set on the transport in use (usb_configured() not called with USB transport): every … |
FLT-10 |
not ported: checked on hardware | error | Forced HardFault and panic not tried on hardware: the next boot reports them (last reset: Fault / Panic) … |
FLT-11 |
not ported: a review check (a person or an agent reads the code or the datasheet) | suggestion | exc_return relied on for crash capture on Cortex-M0+ (reads 0 there) |
FLT-12 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | Runaway conditions not handled (AN1015): control registers of critical peripherals never re-written or … |
FLT-13 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | Critical data used without a validity check (CRC, redundant copy, range) |
FLT-14 |
not ported: checked on hardware | error | Reset-cause flags read as fresh when a core-only reset leaves them: RP2040 / RP2350 CHIP_RESET keeps … |
BOOT-1 |
boot-1 |
error → deny |
Firmware entered by a jump (ROM DfuSe leave, bootloader) not detected: VTOR, PLL, SysTick, enabled IRQs of … |
BOOT-2 |
boot-2 |
error → deny |
H7 supply config written blind: an old ROM bootloader (before V9.3) locks PWR_CR3 until a power cycle; … |
BOOT-3 |
not ported: a review check (a person or an agent reads the code or the datasheet) | error | Control signal on a pin the ROM bootloader drives (AN2606: USART, USB, CAN, I2C pins of the ROM's interfaces) |
BOOT-4 |
not ported: a review check (a person or an agent reads the code or the datasheet) | error | No probe-less recovery path: ROM bootloader (BOOT pin, boot reboot --updater) or a DFU in the firmware … |
BOOT-5 |
not ported: a test the firmware should have | error | Bootloader state written in an order a power cut can tear (embassy-boot STATE erase before zeroing the … |
BOOT-6 |
not ported: a test the firmware should have | warning | A confirm or mark write that rewrites state when nothing changed |
BOOT-7 |
not ported: checked on hardware | error | Update not tried from the previous released version, over the field path (no probe) |
BOOT-8 |
not ported: bedrock has no automatic check for it yet (catalogue: cfg mid (API snapshots)) |
warning | Wire format of the update path changed without a version bump: an older firmware must still understand … |
BOOT-9 |
not ported: bedrock has no automatic check for it yet (catalogue: cfg mid) |
warning | Dual-bank bootloader passed bank1_region only while the DFU region sits in bank 2 |
BOOT-10 |
boot-10 |
warning → warn |
Backup-domain helper that turns the RTC clock or PWR_CR1.DBP off afterwards instead of restoring what it … |
PWR-1 |
pwr-1 |
warning → warn |
No brown-out reset level or programmable voltage detector (PVD) for the supply the board really has |
PWR-2 |
not ported: bedrock has no automatic check for it yet (catalogue: cfg mid + review) |
error | H7 SMPS / LDO supply config not the one of the schematic: the MCU may not start or may be damaged |
PWR-3 |
pwr-3 |
warning → warn |
Backup domain not reset at boot when the RTC and backup registers are unused (corrupted after a brown-out … |
PWR-4 |
pwr-4 |
warning → warn |
RTC clock source changed in the clock config of a firmware that keeps state in the backup domain: embassy … |
PWR-5 |
not ported: checked on hardware | warning | How long the MCU runs after a supply cut not measured where it matters (writes, erases): B129 G0 keeps … |
FLS-1 |
fls-1 |
error → deny |
Flash that a power cut can tear (settings, bootloader state) read raw on an ECC part: a double-ECC word is … |
FLS-2 |
fls-2 |
error → deny |
More than one owner of the flash peripheral (updater and settings each take FLASH) |
FLS-3 |
fls-3 |
error → deny |
RP flash erase or write from core 1, or with core 1 started outside spawn_core1, or while DMA reads … |
FLS-4 |
fls-4 |
warning → warn |
No periodic CRC of the image in flash, no ECC on flash and RAM where the part has it |
FLS-5 |
not ported: a review check (a person or an agent reads the code or the datasheet) | suggestion | Slow erase: check the programming parallelism (higher peak current) |
FLS-6 |
not ported: a test the firmware should have | warning | Power-fail behaviour of the flash store not tested with cuts during erase and write on the family in use … |
FLS-7 |
not ported: bedrock has no automatic check for it yet (catalogue: elf easy) |
warning | Unused program memory not filled with a trap (or protected) |
PIN-1 |
pin-1 |
error → deny |
Pin driver dropped (let _ = Output::new(..), a temporary, end of main's scope): embassy disconnects … |
PIN-2 |
pin-2 |
error → deny |
Board not put in its safe state first thing after the HAL init (output enables, shift-register chains, … |
PIN-3 |
not ported: a test the firmware should have | error | A level translator or bus switch enabled before the MCU side is ready (pins released, at the right level, … |
PIN-4 |
pin-4 |
error → deny |
Peripheral pins moved to GPIO and back under an enabled peripheral (RP2350: the peripheral reads the OR of … |
PIN-5 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | Input relies on an external pull the BOM may not fit (DNP), with the pad pull off |
PIN-6 |
pin-6 |
error → deny |
Two owners of a pin (a diagnostic or test API and a driver task): one owner per pin, the other refused |
PIN-7 |
not ported: a test the firmware should have | error | Direction pin of a level translator turned while its IO pin drives: stop driving first |
PIN-8 |
pin-8 |
warning → warn |
Unused pins left floating inputs: analog (STM32) or a defined pull |
PIN-9 |
pin-9 |
error → deny |
Pin map chosen by a cargo feature for a board variant: a build for one variant flashed onto another drives … |
PIN-10 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | Lines driven onto a device that may be unpowered (back-feed), and not released to Hi-Z on every session … |
PIN-11 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | Hardware feature used without checking the board record says the rework or variant is there |
PIN-12 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | Pin direction set only by an optional protocol command instead of on every path that uses it |
PIN-13 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | A pin that must hold across an MCU reset driven by a GPIO (goes to reset state): use a backup-domain … |
PIN-14 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | Input sampled once: filter by sampling several times where noise can fake an edge |
PIN-15 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | IO speed, drive and load against the frequency used (edge rates, EMC) |
PIN-16 |
pin-16 |
warning → warn |
Pin read that a dead-battery pull-down still holds (G0 UCPD DBCC pins until the SYSCFG strobe; never clear … |
CLK-1 |
not ported: checked on hardware | warning | Core clock not measured once (DWT cycles against the time driver): prescaler math goes wrong silently (H7 … |
CLK-2 |
clk-2 |
error → deny |
embassy time driver time-driver-any, or the named timer also used for PWM or capture |
CLK-3 |
clk-3 |
warning → warn |
No clock security: CSS on HSE (and LSE) with a handler |
CLK-4 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | LSE start-up (seconds on H7) blocks boot or a feature |
CLK-5 |
clk-5 |
error → deny |
USB enabled on a part without HSI48 / CRS and no exact 48 MHz source configured |
CLK-6 |
not ported: a test the firmware should have | warning | A kernel clock that can't divide the rates needed exactly (FDCAN 135 MHz for 2/4/8 Mbit/s, USART OVER8 … |
CLK-7 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | Clock run at the part's maximum just because it can (lifetime at small nodes, ECC RAM limits: H7 TCM ECC … |
CLK-8 |
not ported: a test the firmware should have | warning | Fractional PLL math not checked: embassy refuses a core above the limit at boot, a FRACN at 0 is integer … |
CLK-10 |
not ported: bedrock has no automatic check for it yet (catalogue: src mid) |
warning | FRACN changed on the fly with PLLxFRACEN cleared and set back to back: the new value is sometimes not … |
ERR-1 |
not ported: a review check (a person or an agent reads the code or the datasheet) | error | Errata sheet for the silicon revision on the board not read for the peripherals, clocks, flash, DMA, low … |
ERR-2 |
err-2 |
error → deny |
Register poke for an erratum without the erratum quoted (number, title, revisions, workaround) next to it |
ERR-3 |
err-3 |
error → deny |
G0B0/G0B1/G0C1 flash prefetch on while code crosses into bank 2 (ES0548 2.2.10, ES0547 2.2.8: HardFault on … |
ERR-4 |
err-4 |
error → deny |
FDCAN edge filtering (CCCR.EFBI) on, or dedicated Tx buffers beside the Tx FIFO (ES0548 2.15.½.15.2, … |
ERR-5 |
err-5 |
warning → warn |
FDCAN DAR (one-shot) without retransmit after a lost arbitration on the first two ID bits (ES0491 2.22.3) |
ERR-6 |
err-6 |
warning → warn |
RP2350 A2 pad with pull-down and input buffer left on (RP2350-E9: the pin stays high): enable the input … |
ERR-7 |
err-7 |
warning → warn |
RP2350 ROM reboot with delay 0 or FRCE_OFF bits beyond PROC1 (RP2350-E30, E19) |
ERR-8 |
err-8 |
warning → warn |
G0 ADC calibration without the wait of 2 ADC kernel clocks after disable (freezes) |
ERR-9 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | I2C timing not checked against the kernel / APB clock ratio errata (ES0548 2.11.1, 2.11.4) |
ERR-10 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | Cache on without reading the cache errata of the part |
ERR-11 |
err-11 |
suggestion → warn |
Silicon revision read from a register that can't tell all revisions apart (RP2350 CHIP_ID.REVISION: A3 … |
EXT-1 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | An I2C NAK taken as "device lost": chips NAK while busy; use a positive signature (ID reads all-zero … |
EXT-2 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | External chip on a rail that can sag without an MCU reset, and no watchdog that notices it lost its config … |
EXT-3 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | A configurable chip identified by its ID registers before it is configured (no factory default): use a strap |
EXT-4 |
ext-4 |
warning → warn |
Read-modify-write of an external chip's registers over the bus, clear-on-read registers read more than … |
EXT-5 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | A sense line assumed wired (PHY VBUS pin, card detect polarity): read the board and the connector datasheet |
EXT-6 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | A settle or start-up time skipped on one path (VBUS after a hard reset, reset to first command, power-on … |
EXT-7 |
ext-7 |
warning → warn |
Register field encoded without a range check (x - 1 underflows, masked to 0) |
EXT-8 |
not ported: a test the firmware should have | warning | Config value set but never written to the hardware by the apply path, or not stored and re-applied stale … |
EXT-9 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | Shared peripheral memory (H7 FDCAN message RAM) zeroed whole by one instance's init |
EXT-10 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | Register map or masks of another IP variant (G0 FDCAN with the H7 map, reserved bits in masks) |
EXT-11 |
not ported: a review check (a person or an agent reads the code or the datasheet) | suggestion | A HAL driver taken without checking it against what the product needs (timing accuracy, NAK index, … |
EXT-12 |
ext-12 |
warning → warn |
RCC enable not read back before the first access to the peripheral |
USB-1 |
usb-1 |
error → deny |
High-speed bulk endpoint max packet not 512, bInterval out of range for the speed |
USB-2 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | A transfer that is an exact multiple of the max packet without a ZLP, or framing that waits for a short … |
USB-3 |
usb-3 |
suggestion → warn |
Single-buffered bulk endpoints on a hot path behind a hub TT (G0 usb_v4: 500 vs 947 kB/s) |
USB-4 |
not ported: checked on hardware | warning | Sending stuck when the host driver closes or the host dies without a disconnect |
USB-5 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | OUT packet read from USB SRAM within 800 ns of its CTR on a part with the "buffer description table update … |
HOST-1 |
host-1 |
error → deny |
Panic reachable from host input: unwrap, expect, slice indexing, unreachable!, arithmetic overflow … |
HOST-2 |
not ported: a review check (a person or an agent reads the code or the datasheet) | error | Host can drive a pin into a level or mode that damages the board: every electrical rule enforced by the … |
HOST-3 |
host-3 |
warning → warn |
Unknown enum value from a newer host mapped to a default (_ => ..) instead of refused |
HOST-4 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | A host that stops reading can stall a release or cleanup path: every wait on the host has a timeout |
HOST-5 |
host-5 |
warning → warn |
Truncated or oversized request not counted and dropped |
HOST-6 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | A host-settable range wider than the firmware can serve (bounded by DMA geometry, buffers): limits derived … |
HOST-7 |
not ported: a test the firmware should have | warning | One unimplemented field voids a batched update; a deferred answer path (credit, grant) only sent from … |
HOST-8 |
not ported: a test the firmware should have | warning | Loss reported out of order with the data it belongs to |
LOG-1 |
log-1 |
error → deny |
RTT that can block: probe-rs sets the channel to blocking on attach; force NoBlockSkip, whole frames or … |
LOG-2 |
log-2 |
warning → warn |
DEFMT_RTT_BUFFER_SIZE and DEFMT_LOG not set explicitly in .cargo/config.toml (init bursts lost: 773 … |
LOG-3 |
not ported: a review check (a person or an agent reads the code or the datasheet) | suggestion | Hot-path logging at info or above |
CNT-1 |
cnt-1 |
warning → warn |
Error branch (an Err handled, a timeout, a retry, a dropped frame, an overrun) without a cnt! counter: … |
CNT-2 |
cnt-2 |
warning → warn |
&mut to a counter buffer that an interrupt can update, unaligned u64 access |
CNT-3 |
cnt-3 |
warning → warn |
Proc-macro crate and its runtime crate not pinned to the exact same version (=x.y.z) |
CFG-1 |
cfg-1 |
error → deny |
probe-rs runner or a recipe without --probe VID:PID: with two probes on the PC another board gets flashed |
CFG-2 |
cfg-2 |
warning → warn |
probe-rs download then reset on F0/G0/L0: probe-rs writes RCC_APBENR1 back on a running core (USB … |
CFG-3 |
not ported: bedrock has no automatic check for it yet (catalogue: cfg easy) |
warning | Feature sets not all built and linted (default, --all-features, --no-default-features): gated code rots |
CFG-4 |
cfg-4 |
error → deny |
A crate in the tree twice (git and path copies of WireWeaver, embassy): patch it to one source |
CFG-5 |
cfg-5 |
error → deny |
Chip feature (package, flash size) of the HAL not the part on the board |
CFG-6 |
cfg-6 |
warning → warn |
thumbv6m without portable-atomic + critical-section where a crate needs CAS (static_cell) |
CFG-7 |
cfg-7 |
warning → warn |
Dirty build flashed without a mark: build info carries the git SHA and a dirty flag, read back from the device |
CFG-8 |
not ported: bedrock has no automatic check for it yet (catalogue: cfg mid) |
warning | Git consumers not built: a dependency from git (not path) breaks where path builds pass |
CFG-9 |
not ported: bedrock has no automatic check for it yet (catalogue: cfg easy) |
warning | Committed path dependency or [patch] to a sibling checkout (../wire_weaver): the lock follows each … |
PERF-1 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | Speed or size claims without numbers at the opt-level that ships, in their own units with the clock … |
PERF-2 |
perf-2 |
suggestion → warn |
64-bit or i128 maths, float formatting, or % / / by a runtime value per element on a hot path of a … |
PERF-3 |
perf-3 |
warning → warn |
RAM function (#[link_section = ".data.*"]) without #[inline(never)]: LTO inlines it back into flash |
PERF-4 |
not ported: a review check (a person or an agent reads the code or the datasheet) | suggestion | Hot path through a bitwise CRC, a bit-by-bit serializer, or GPIO wrappers per edge (embassy Flex ~50 … |
PERF-5 |
not ported: a review check (a person or an agent reads the code or the datasheet) | suggestion | Wake cost of an event loop with nothing to do (select4 with timers and waker registrations per turn on M0+) |
TEST-1 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | Host test that races a free-running tick or sleeps "long enough" |
TEST-2 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | A simulator that answers inline hides bugs of the deferred path: test both |
TEST-3 |
not ported: a test the firmware should have | warning | Sequencing rules of the board (enable order, safe off) without host tests of the transition model |
TEST-4 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | A board left running its last test after probe-rs exits: reset before each run |
TEST-5 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | One probe, one owner: another session's probe-rs or gdb on the same board makes faults that aren't the … |
TEST-6 |
not ported: a review check (a person or an agent reads the code or the datasheet) | suggestion | A fault that depends on timing or layout bisected before the errata were checked |
TEST-7 |
not ported: a review check (a person or an agent reads the code or the datasheet) | warning | A peripheral or chip blamed before our own side was measured (M1 blamed the sink's FUSB302B for our RXDR … |