By now, you know how to use the *delay() and *sleep() APIs to construct delays and sleeps (non-blocking and blocking, respectively). Hang on, though – we haven't really tried it out in a kernel module. Not only that, are the delays and sleeps as accurate as we have been led to believe? Let's, as usual, be empirical (this is important!) and not make any assumptions. Let's actually try it out for ourselves!
The demo kernel module we'll be looking at in this subsection performs two kinds of delays, in order:
- First, it employs the *delay() routines (which you learned about in the Understanding how to use the *delay() atomic APIs section) to implement atomic non-blocking delays of 10 ns, 10 us, and 10 ms.
- Next, it employs the *sleep() routines (which you learned about in the Understanding how to use the *sleep() blocking APIs section) to implement...
