Hi David,
On 2024-11-22, David Gow davidgow@google.com wrote:
It's a little unusual for a KUnit test -- particularly since it is time-based and uses lots of threads. This isn't a problem, but it's definitely a good thing that it's marked as slow. Additionally, KUnit doesn't track any extra threads spawned, so it requires a bit more care.
There are a couple of issues (e.g., it crashes on non-SMP systems, a potential race, etc) and some minor suggestions below. In short, it'd be a good idea to move some of the initialisation and checks into the main test function, rather than the helper threads.
Equally, it looks like there are a bunch of variables shared between kthreads — do these need to be checked with READ_ONCE()/WRITE_ONCE(), or made volatile, or something?
Agreed.
In fact, I'm not sure why there's a separate start_test() and test_readerwriter() function -- or indeed, a separate kthread? Am I missing something, or could everything start_test() does be done from the main test function/kthread?
You are not missing anything. It is definitely awkward, mostly because it was taken from parts of my own personal testing software. I will implement all your suggestions. Thanks for the detailed review!
John