On 02/02, Oleg Nesterov wrote: > > And another question. Lets forget about this ->mm check. I simply can not > understand this > > ret = (p->flags & PF_EXITPIDONE) ? -ESRCH : -EAGAIN > > logic in attach_to_pi_owner(). First of all, why do we need to retry if > PF_EXITING is set but PF_EXITPIDONE is not? Why we can not simply ignore > PF_EXITING and rely on exit_pi_state_list() if PF_EXITPIDONE is not set? > > I must have missed something but this looks buggy, I do not see any > preemption point ^^^^^^^^^^
I meant synchronization point, sorry for confusion. > in this "retry" loop. Suppose that max_cpus=1 and rt_task() > preempts the non-rt PF_EXITING owner. Looks like futex_lock_pi() can spin > forever in this case? (OK, ignoring RT throttling). Finally I forced myself to try to make the 1st patch ;) To remind, we have more problems with robust+pi futexes, this needs another patch(es). Otherwise we could (probably) even kill PF_EXITPIDONE. Peter. I have no idea how to test it (except it obviously fixes the test- case I sent before). IOW: please review. And I still fail to understand why this PF_EXITING logic was added in the first place. So I also have the problem with the changelog, it merely says "don't do this because this is not needed". On top of "check PF_KTHREAD rather than !p->mm to filter out kthreads" but doesn't depend on it. Oleg. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/