On Wed, Aug 16, 2023 at 11:08:21PM -0500, Eric W. Biederman wrote: > Kees Cook <keescook@xxxxxxxxxxxx> writes: > > > This is a continuation of the proposal[1] for mocking init_task for > > KUnit testing. Changing the behavior of kill_something_info() is moving > > forward[2] and I'd _really_ like to have some unit tests in place to > > actually test the behavioral changes. > > > > I tried to incorporate feedback from Daniel and David, and I think the > > result is fairly workable -- the only tricky part is building valid > > task_struct instances. :) > > > > Notably, I haven't actually gotten as far as testing the actual proposed > > behavioral change since I wanted to make sure this approach wasn't going > > to totally crash and burn. > > > > Thoughts? > > Overall this looks like a nice start. More comments below. > > I do wonder though. Would it perhaps be easier to create a pid > namespace with just the processes you want in it? Do you have a short example of how I could do this correctly? It's not obvious to me how to actually set all that up (and tear it down). > I am wondering because you failed to mock find_vpid and so you > are actually testing sending signals to kernel threads. Hah. Eek. -- Kees Cook
