Let's look at another table that quickly summarizes the available (to us module authors) blocking *sleep*() kernel APIs; these are only meant to be used in the process context when it's safe to sleep; that is, where the invocation of schedule() is not a problem. In other words, the delay is implemented by the process context actually going to sleep for the duration of the delay and is then woke up when it's done:
| API | Internally "backed by" | Comment |
| usleep_range(umin, umax); | hrtimers (high-resolution timers) | Sleep for between umin and umax microseconds. Use where the wakeup time is flexible. This is the recommended API to use. |
| msleep(ms); | jiffies/legacy_timers | Sleep for ms milliseconds. Typically meant for a sleep with a duration of 10 ms or more. |
| msleep_interruptible(ms); | jiffies/legacy_timers | An interruptible variant of msleep(ms);. |
| ssleep(s); | jiffies/legacy_timers | Sleep for ... |
