hal_rng_get_random_in_range example implementation can lead to division by zero #127
Description
A fairly basic problem, but a bit tricky nonetheless. So, in Random Number Generator Hardware Abstraction Layer implementation there is:
In porting tests random is tested with following values:
0xffffffff-0+1 = 0x100000000 but because Cortex-M registers are 32 bit wide it is truncated, which causes division by zero. Disassembly on STM32L4 looks like this:
08010260: subs r4, r4, r6
08010262: adds r4, #1
08010264: ldr r3, [sp, #4]
08010266: udiv r2, r3, r4
The problem is that somehow hardfault does not occur when running the code without SWD or JTAG debugger attached, but it appears when debugging.
I suggest to fix it like this:
uint32_t min_val, max_val, range;
if (val_1 <= val_2) {
min_val = val_1;
max_val = val_2;
} else {
min_val = val_2;
max_val = val_1;
}
if (min_val == max_val) {
return min_val;
}
if (max_val == UINT32_MAX && min_val == 0) {
return rand_nb;
}
range = max_val - min_val + 1;
return min_val + (rand_nb % range);