After cleaning up the default config of SPI, this commit cleans up the configuration for `sipeed-longan-nano` a bit:
- PA4 is no longer used as default CS for `SPI_DEV(1)`
- At least one DAC channel PA4 is always available if DAC is used.
- PA4 is always available as ADC channel if DAC is not used.
- The order of ADC and DAC channels has been changed accordingly.
19256: pkg/tinyusb: add GD32VF103 support r=gschorcht a=gschorcht
### Contribution description
This PR provides the tinyUSB support for GD32VF103 and enables the `tinyusb_device` feature as well as `stdio_tinyusb_cdc_acm` for GD32VF103 boards.
### Testing procedure
```
BOARD=sipeeed-longan-nano make -C tests/shell flash term
```
should work
### Issues/PRs references
Co-authored-by: Gunar Schorcht <gunar@schorcht.net>
The ADC configuration was too complex. It was hard to follow when certain ADC lines are available. Furthermore, the order of ADC lines did depend on the use of other peripherals. Now, either the TFT display is not connected and all ADC lines are available or the TFT display is connected and the second SPI device is used so that only the first 4 ADC lines are available.
19199: sys/suit: Ensure previous thread is stopped before reusing its stack r=benpicco a=chrysn
### Contribution description
Closes: https://github.com/RIOT-OS/RIOT/issues/19195
If the thread has released the mutex but the thread has not terminated (which happens in the situation that would previously have overwritten a still active thread's state), then a warning is shown and the trigger is ignored.
### Testing procedure
This should work before and after:
* `make -C examples/suit_update BOARD=native all term`
* `aiocoap-client coap://'[fe80::3c63:beff:fe85:ca96%tapbr0]'/suit/trigger -m POST --payload 'coap://[2001:db8::]/foo'`
* In parallel, on the RIOT shell, run `suit fetch coap://[2001:db8::]/foo`
* After the first download fails, the second one starts right away ("suit_worker: update failed, hdr invalid" / "suit_worker: started").
Run again with the worker thread on low priority:
```patch
diff --git a/sys/suit/transport/worker.c b/sys/suit/transport/worker.c
index a54022fb28..e26701a64c 100644
--- a/sys/suit/transport/worker.c
+++ b/sys/suit/transport/worker.c
`@@` -70 +70 `@@`
-#define SUIT_COAP_WORKER_PRIO THREAD_PRIORITY_MAIN - 1
+#define SUIT_COAP_WORKER_PRIO THREAD_PRIORITY_MAIN + 1
```
Before, this runs the download once silently (no clue why it doesn't run it twice, but then again, I claim there's concurrent memory access from two thread, so who knows what happens). After, it runs a single download and shows an error message for the second one once the first is complete ("Ignoring SUIT trigger: worker is still busy.").
### Issues/PRs references
This may be made incrementally easier by https://github.com/RIOT-OS/RIOT/pull/19197 -- that PR as it is now would spare us the zombification (because returning would do that), and having a `wait` function would allow us to turn the new error case into a success.
19205: boards/common: add common timer config for GD32VF103 boards r=benpicco a=benpicco
19207: examples/gnrc_border_router: static: use router from advertisements by default r=benpicco a=benpicco
Co-authored-by: chrysn <chrysn@fsfe.org>
Co-authored-by: Benjamin Valentin <benjamin.valentin@bht-berlin.de>
Co-authored-by: Benjamin Valentin <benpicco@beuth-hochschule.de>
