19677: boards/nucleo-l432k: provide three periph_timer instances r=maribu a=maribu
### Contribution description
- `cpu/stm32/periph_timer`: Generalize to also work with timers that do not have 4 channels
- `boards/common/stm32`: Add timer config for three timers based on TIM2, TIM15, and TIM16 (the three general-purpose timers of the STM32L4)
- `boards/nucleo-l432kc`: Make use of the new timer config
19683: cpu/sam0_eth: clean up init() r=maribu a=benpicco
Co-authored-by: Marian Buschsieweke <marian.buschsieweke@ovgu.de>
Co-authored-by: Benjamin Valentin <benjamin.valentin@ml-pa.com>
- point out behavior on denormalized time stamps
- use errno codes to indicate errors (and adapt the few instances of
actual error handling to use them)
The macros CONCAT(), MIN(), and MAX() are defined over and over again in
RIOT's code base. This de-duplicates the code by moving the macros to a
common place.
rtc_set_alarm() / rtt_set_alarm() are heavily used by ztimer during ISR. This will reduce time spent during ISR drastically. We trust that the peripheral is able to propagate the alarm asynchronously.
Due to the RIOT_EPOCH of 2020 this overflow will happen in year 2084. It would be scary if IoT devices are still around then.
We can save RAM and ROM. Furthermore, this overflow handling should block BACKUP power mode in order to keep track of the reference year.
`rtt_get_counter()` already waits for syncbusy before reading the time,
but we also have to do this in RTC mode (`rtc_get_time()`) to avoid
reading old values.
Thus, always wait for syncbusy to clear when accessing the COUNT register.
On SAM L21 only 8 EXTWAKE pins can wake the CPU from Backup sleep.
Handle this analogous to the RTC tamper pins on SAM D5x/E5x where
configuring them as an interrupt will also cause them to wake the
device from Deep Sleep.
Setting up a DMA transfer can take longer than sending out a buffer
byte by byte if the buffer is small.
DMA only shows advantages for large buffers, using it for every transfer
will cause a net slowdown.
Since we did not come up with a good way to determine the treshold based
on the SPI frequency, just use a fixed buffer for now so that DMA can be
used without slowing things down overall.
