[hal] SPI: fixes potential deadlock on Gen3 platforms.

[XuGuohui]

Problem

Gen3 Device may get blocked when priority of the thread doing SPI transmission higher than the thread doing USB Serial logging.

Solution

Check and handle SPI NRF_SPIM_EVENT_END while waiting SPI transaction to be completed.

Example App

#include "Particle.h"

SerialLogHandler logHandler(115200, LOG_LEVEL_ALL);

SYSTEM_THREAD(ENABLED);
SYSTEM_MODE(SEMI_AUTOMATIC);

Thread *spi_thread;

__SPISettings spi_settings(5*MHZ, MSBFIRST, SPI_MODE0);

uint32_t spi_count = 0;

void spi_thread_f(void) {
    while (true) {
        uint8_t rx_buf[64];
        SPI.beginTransaction(spi_settings);
        SPI.transfer(NULL, rx_buf, sizeof(rx_buf), NULL);
        SPI.endTransaction();
        spi_count++;
        delay(1);
    }
}

void setup() {
    SPI.setDataMode(SPI_MODE0);
    SPI.setClockSpeed(5, MHZ);
    SPI.begin();

    Serial.begin(115200);
    while(!Serial.isConnected());

    spi_thread = new Thread("spiTest", spi_thread_f, OS_THREAD_PRIORITY_CRITICAL);

    pinMode(D7, OUTPUT);
}

void loop() {
    Log.info("heartbeat: %lu", spi_count);
    delay(1);

    static system_tick_t start = millis();
    static bool toggle = false;
    if (millis() - start >= 1000) {
        start = millis();
        if (toggle) {
            digitalWrite(D7, HIGH);
        } else {
            digitalWrite(D7, LOW);
        }
        toggle = !toggle;
    }
}

Completeness

• User is totes amazing for contributing!
• Contributor has signed CLA (Info here)
• Problem and Solution clearly stated
• Run unit/integration/application tests on device
• Added documentation
• Added to CHANGELOG.md after merging (add links to docs and issues)

[avtolstoy]

@XuGuohui Let's fix this properly as we've discussed previously by overriding CRITICAL_REGION_ENTER()/EXIT() to use PRIMASK instead of altering NVIC registers.

[YutingYou]

@XuGuohui Let's fix this properly as we've discussed previously by overriding CRITICAL_REGION_ENTER()/EXIT() to use PRIMASK instead of altering NVIC registers.

It might potentially affect BLE protocol which is based on the timer0 and other interrupts.

[avtolstoy]

It might potentially affect BLE protocol which is based on the timer0 and other interrupts.

Which will not get blocked. We are already using this approach throughout DeviceOS. See how HAL_disable_irq() is implemented for Gen 3. We may need to use priority 4 vs 2 here specifically though.

[XuGuohui]

We may need to use priority 4 vs 2 here specifically though.

I still got a SOS when use priority 2 building against the latest develop branch. I tested on Argon.

[XuGuohui]

We have another workaround that makes more sense to fix the issue. This PR will be closed and superseded by another PR when it is ready.