Messages in this thread | | | Date | Fri, 15 Jul 2016 17:16:54 +0100 | From | Mark Rutland <> | Subject | Re: [patch V2 35/67] arm/perf: Convert to hotplug state machine |
| |
On Fri, Jul 15, 2016 at 05:26:30PM +0200, Sebastian Andrzej Siewior wrote: > * Mark Rutland | 2016-07-15 14:08:18 [+0100]: > > >Hi, > Hi Mark, > > >We may have multiple PMUs (e.g. two in big.LITTLE systems), and > >__oprofile_cpu_pmu only contains one of these. So this conversion is not > >correct. > > > >We were relying on the notifier list implicitly containing a list of > >those PMUs. It seems like we need an explicit list here. > > > >We keep __oprofile_cpu_pmu around for legacy 32-bit users of OProfile > >(on non-hetereogeneous systems), and that's all that the variable should > >be used for. > > By the time I'e been looking there was only one node in the .dts. Now > you say it is not only possible but likely to have more than one node.
Yes. For example, arch/arm/boot/dts/vexpress-v2p-ca15_a7.dts is a big.LITTLE system with a Cortex-A7 PMU and a Cortex-A15 PMU.
> So what we need is a list which gets extended in cpu_pmu_init() instead > of using the global pointer we have now. Correct?
Yes, that sounds right to me, with matching removal in cpu_pmu_destroy.
Thanks, Mark.
| |