On Tue, Jul 30, 2013 at 02:05:29PM -0400, Steven Rostedt wrote: > On Tue, 2013-07-30 at 11:42 -0400, Alan Stern wrote: > > On Mon, 29 Jul 2013, James Stone wrote: > > Just an FYI, it is usually better to email my rostedt@xxxxxxxxxxx > account. I don't always read my redhat email. That's reserved for > bugzillas assigned to me ;-) > > > > > > OK, having said that, I just got it to happen - listening to audacity > > > and just logging into Facebook (of all things!! Meh!) > > > > > > This is the contents of the trace file (as per instructions on bug #1191603) > > > > > > # tracer: irqsoff > > > # > > > # irqsoff latency trace v1.1.5 on 3.10.0-ver5 > > > # -------------------------------------------------------------------- > > > # latency: 2173 us, #4/4, CPU#0 | (M:desktop VP:0, KP:0, SP:0 HP:0 #P:4) > > > # ----------------- > > > # | task: apt-check-3628 (uid:1000 nice:19 policy:0 rt_prio:0) > > > # ----------------- > > > # => started at: perf_event_update_userpage > > > # => ended at: retint_careful > > > # > > > # > > > # _------=> CPU# > > > # / _-----=> irqs-off > > > # | / _----=> need-resched > > > # || / _---=> hardirq/softirq > > > # ||| / _--=> preempt-depth > > > # |||| / delay > > > # cmd pid ||||| time | caller > > > # \ / ||||| \ | / > > > apt-chec-3628 0d.h. 0us!: local_clock <-perf_event_update_userpage > > perf_event_update_userpage should not be taking 2ms to complete. > > > This suggests that the IRQ line was disabled or interrupts were > > disabled on all four CPUs for 7 ms. However, the irqsoff trace shows > > that the maximum latency was 2 ms. Something strange is going on, but > > I don't know what or how to find it. > > > > Steve, do you have any ideas about this? Or suggestions for other > > people to ask? > > Is there SMIs on this box? Those will not be detected by the latency > tracers directly. But we do have other tools to detect them. Right, NMI shouldn't be on all CPUs concurrently, and while its possible to get very long NMIs with perf (callchains tend to take a while) 7ms sounds like very long on a regular machine (there's issues with SGI class NUMA machines). -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
