On Fri, 2013-08-02 at 16:36 +0200, Peter Zijlstra wrote: > On Fri, Aug 02, 2013 at 10:14:39AM -0400, Alan Stern wrote: > > URL Clas-3286 2d.h. 1us : local_clock <-perf_event_update_userpage > > URL Clas-3286 2d.h. 2us : watchdog_overflow_callback <-__perf_event_overflow > > URL Clas-3286 2d.h. 3us : arch_trigger_all_cpu_backtrace_handler <-nmi_handle.isra.0 > > URL Clas-3286 2d.h. 3us : perf_ibs_nmi_handler <-nmi_handle.isra.0 > > URL Clas-3286 2d.h. 3us : perf_ibs_handle_irq <-perf_ibs_nmi_handler > > URL Clas-3286 2d.h. 4us : perf_ibs_handle_irq <-perf_ibs_nmi_handler > > URL Clas-3286 2d.h. 4us!: rcu_nmi_exit <-do_nmi > > What's cute is that the trace starts when the NMI handler does > local_irq_save(), _not_ when the NMI starts, which is where the hardware > actually disabled interrupts. Yeah, the NMI can be messing with the tracer. It's built on top of lockdep, which does not handle NMIs. Perhaps we can add NMI handling to the latency tracer, but that may need a bit of work to do that. -- Steve > > > URL Clas-3286 2d... 879us : smp_apic_timer_interrupt <-apic_timer_interrupt > > And then we have hit massive jump.. out of nowhere. weirdness. -- 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
