pub struct SystemResources<'a> {
pub sw_int: SoftwareInterrupts<'a>,
pub timer0_0: AnyTimer<'a>,
pub timer0_1: AnyTimer<'a>,
pub timer1_0: AnyTimer<'a>,
pub timer1_1: AnyTimer<'a>,
pub cpu_ctrl: CPU_CTRL<'a>,
pub lpwr: LPWR<'a>,
pub flash: FLASH<'a>,
pub aes: AES<'a>,
pub rtc_timer: RTC_TIMER<'a>,
}Expand description
Chip-level (board-independent) system resources: executor timers, software
interrupts, the second core, and the RTC domain (e.g. the production RWDT,
crate::io::watchdog).
Fields§
§sw_int: SoftwareInterrupts<'a>§timer0_0: AnyTimer<'a>§timer0_1: AnyTimer<'a>§timer1_0: AnyTimer<'a>§timer1_1: AnyTimer<'a>§cpu_ctrl: CPU_CTRL<'a>§lpwr: LPWR<'a>RTC peripheral — RTC domain registers (reset reason, RTC-fast memory).
flash: FLASH<'a>Internal SPI flash — what esp_hal::flash::Flash::new takes, so a
consumer can persist settings in its own partition.
Handed out rather than kept because the BSP has no opinion about what belongs in flash. Note the cost that comes with it: on a dual-core part an erase requires the other core to stop fetching from flash, so a write parks it for tens of milliseconds. That is the caller’s problem to schedule, and it is why this is a peripheral rather than a convenience method.
aes: AES<'a>AES accelerator — one 16-byte block per encrypt/decrypt call.
No board wiring, handed out like Self::flash: a consumer that
decrypts on a hot path wants it instead of a software cipher, which on
xtensa is kilobytes of code plus a key schedule per message.
rtc_timer: RTC_TIMER<'a>RTC timer singleton — what Rtc::new takes as of esp-hal 1.2, which
split it out of LPWR. Handed out separately because the two are now
distinct peripherals: a consumer arming the production RWDT
(crate::io::watchdog) needs THIS, while lpwr still covers the
RTC-domain registers. Before 1.2 Rtc::new(lpwr) covered both, so a
board that only exposed lpwr silently stopped being able to build an
Rtc at all.