On 05-12-15, 11:01, Paul E. McKenney wrote: > And it was getting lucky. In a set of 24 two-hour runs (triple parallel) > on an earlier commit (not 3497d206c4d9, no clue what I was thinking) got > me two failed runs, for a total of 49 reports of one of RCU's grace-period > kthreads being starved, no reports of rcutorture's kthreads being starved, > and no hangs on shutdown. So much lower failure rate, but still failures. > > At this point, I am a bit disgusted with bisection, so my next test cycle > (36 two-hour runs on a system capable of doing three concurrently) is on > the most recent -rcu, but with CPU hotplug disabled. If that shows failures, > then I hammer 3497d206c4d9 hard. > > Anyway, if you have any ideas as to what might be happening, please don't > keep them a secret!
I can be the least helpful here (based on knowledge), but I am not able to find a reason for this diff in 3497d206c4d9: - if (!hrtimer_callback_running(hr)) - __hrtimer_start_range_ns(hr, cpuctx->hrtimer_interval, - 0, HRTIMER_MODE_REL_PINNED, 0); + hrtimer_start(hr, cpuctx->hrtimer_interval, HRTIMER_MODE_REL_PINNED); The commit talks *only* about s/__hrtimer_start_range_ns/hrtimer_start but not at all on why !hrtimer_callback_running(hr) was removed. Perhaps there was a reason :) -- viresh -- 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/
