Interrupt masking
Temporarily stopping the CPU from taking interrupts, usually to protect a few lines of code that cannot survive being interrupted.
A masked interrupt isn't lost; it waits as pending until the mask lifts. But if the same interrupt fires twice in the meantime, only one survives, so long masking windows drop events. On Cortex-M, PRIMASK (cpsid i, __disable_irq()) blocks everything except NMI and HardFault, while BASEPRI only shuts out interrupts below a given priority.
That difference is why FreeRTOS uses BASEPRI: kernel critical sections block only the interrupts allowed to call into the RTOS, and a higher-priority motor control loop carries on as if nothing happened. Every microsecond spent masked is added straight onto worst-case interrupt latency.