代码

void bsp_us_delay(uint32_t us) {
    uint32_t start, now, delta, reload, us_tick;
    start = SysTick->VAL;
    reload = SysTick->LOAD;
    us_tick = SystemCoreClock / 1000000UL;
    do {
        now = SysTick->VAL;
        delta = start > now ? start - now : reload + start - now;
    } while (delta < us_tick * us);
}

测试（主要逻辑）

 stm32_pin_mode(GPIOB,GPIO_PIN_6,pin_mode_output);

    while (1){
        pin_high(GPIOB,GPIO_PIN_6);
        bsp_us_delay(20);
        pin_low(GPIOB,GPIO_PIN_6);
        bsp_us_delay(20);

    }

逻辑分析仪查看结果

大概1us左右的误差