boards/nrf52dk: Minimal Arduino pinout support

[chrysn]

Contribution description

This defines the Arduino GPIO pins, sets the shield I2C bus and the form factor.

Testing procedure

Run tests/periph/selftest_shield with an nrf52dk and a Peripheral Selftest Shield in "all off, D-3-4 on" configuration.

Issues/PRs references

This simplifies testing #20282

PR status

I'm keeping this as Draft PR because I'm still looking at whether there are cheap extras to enable (maybe SPI bus, maybe the ADC pins). If I find none, this will be changed to "ready for review" without further comment.

[chrysn]

On the good side: I found something that can be added easily.

Now, test with jumpers 5, 4 and 2 set (D7-CS, D3-D4, Side SPI).

On the bad side, there seems to be somethign wrong with the nRF52 SPI driver, as the test's main.c needs this patch to run through successfully:

--- a/tests/periph/selftest_shield/main.c
+++ b/tests/periph/selftest_shield/main.c
@@ -857,4 +857,5 @@ static bool periph_spi_rxtx_test(spi_t bus, spi_mode_t mode, spi_clk_t clk,
         }
         uint8_t received = spi_transfer_byte(bus, cs, true, i);
+        received = spi_transfer_byte(bus, cs, true, i);
         uint16_t stop = 0;
         if (IS_USED(MODULE_PERIPH_TIMER)) {

Without this, every other (literally: one out of two in alternation) SPI byte transfer will not read back the written value, but the value written in the last iteration. I suspected premature resumption (like, maybe there's some register buffering and writes are too fast), but large delays in the outer loop don't fix it either.

If all other properties of this PR are to the reviewers' satisfaction, I'd prefer to have this merged, and then start digging for where the SPI goes wrong -- after all, this PR doesn't cause the trouble, it just makes them testable.

[chrysn]

I'm observing that doing 2-byte writes also fixes the many failures:

@@ -857,3 +857,6 @@ static bool periph_spi_rxtx_test(spi_t bus, spi_mode_t mode, spi_clk_t clk,
         }
-        uint8_t received = spi_transfer_byte(bus, cs, true, i);
+        const uint8_t outbuf[2] = {i, 0xff};
+        uint8_t inbuf[2] = {0x55, 0x55};
+        spi_transfer_bytes(bus, cs, true, &outbuf, &inbuf, 2);
+        uint8_t received = inbuf[0];
         uint16_t stop = 0;

CC'ing @bergzand who contributed the SPI driver in #14057, including a workaround for single-byte transfers which this test uses.

[riot-ci]

Murdock results

✔️ PASSED

7452a61 fixup! boards/nrf52dk: Minimal Arduino pinout support

Success	Failures	Total	Runtime
1128	0	1128	03m:46s

Artifacts

• Documentation preview

[maribu]

Thx :)

[chrysn]

Brief follow-up on the SPI trouble: I have no clue how that went wrong, but I can't reproduce the SPI trouble any more.

All shiny, tests pass.