AVR Freaks

And just to point out that:

delay_ms(100);

is really probably nothing much more than:

for (uint64_t counter = 0; counter < some_big_number_caculated_to_be_100ms; counter++) {
    asm("nop"):
}

So this "holds execution". (though in a pre-emptive OS something will likely task switch out of this long loop a number of times - that in itself will make the 100ms calculation wrong!)

Having said that what OS is this and where does the delay_ms() come from? For all we know maybe it is already implemented as:

void delay_ms(int n) {
    while (n--) {
        os_sleep(1);
    }
}

or something like that? If so then it *should* work as the name suggests I guess? (assuming in my example that os_sleep() really can deliver a 1ms granularity?)

EDIT: From the xSemaphoreTake() call it looks very like this is FreeRTOS. If so then I can't see that it has delay_ms() so I guess this is something in your own code? What it does have is:

https://www.freertos.org/a00127.html

EDIT2: having said that I can't see that it natively has os_sleep() either so are delay_ms() and os_sleep() some kind of wrappers around lower FreeRTOS provide functions? Is this your own implementation or are you using something dreadful like ASF ??

EDIT3: clearly more coffee needed here - just spotted that this post is in the ASF forum - so I guess these things are ASF nonsense then?

EDIT4: I found an ASF document that mentions both os_sleep() and delay_ms():  https://usermanual.wiki/Document/ASF420API20Reference20Manual.198802149.pdf  it says it is an API reference and yet it only has one mention of each API - each in example code with no explanation of what the functions do. However other digging suggests "delay_ms()"  is nothing to do with FreeRTOS but is an independent cycle waster perhaps.