Re: [PATCH] perf/x86/intel: Use rdmsrl_safe when initializing RAPL PMU.

[Stephane Eranian]

On Wed, Apr 23, 2014 at 4:45 PM, Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> On Wed, Apr 23, 2014 at 04:31:32PM +0200, Stephane Eranian wrote:
>> On Thu, Mar 13, 2014 at 8:36 PM, Venkatesh Srinivas
>> <venkateshs@xxxxxxxxxx> wrote:
>> > CPUs which should support the RAPL counters according to
>> > Family/Model/Stepping may still issue #GP when attempting to access
>> > the RAPL MSRs. This may happen when Linux is running under KVM and
>> > we are passing-through host F/M/S data, for example. Use rdmsrl_safe
>> > to first access the RAPL_POWER_UNIT MSR; if this fails, do not
>> > attempt to use this PMU.
>> >
>> > Signed-off-by: Venkatesh Srinivas <venkateshs@xxxxxxxxxx>
>> > ---
>> >  arch/x86/kernel/cpu/perf_event_intel_rapl.c | 12 +++++++++---
>> >  1 file changed, 9 insertions(+), 3 deletions(-)
>> >
>> > diff --git a/arch/x86/kernel/cpu/perf_event_intel_rapl.c b/arch/x86/kernel/cpu/perf_event_intel_rapl.c
>> > index 5ad35ad..95700e5 100644
>> > --- a/arch/x86/kernel/cpu/perf_event_intel_rapl.c
>> > +++ b/arch/x86/kernel/cpu/perf_event_intel_rapl.c
>> > @@ -511,6 +511,7 @@ static int rapl_cpu_prepare(int cpu)
>> >         struct rapl_pmu *pmu = per_cpu(rapl_pmu, cpu);
>> >         int phys_id = topology_physical_package_id(cpu);
>> >         u64 ms;
>> > +       u64 msr_rapl_power_unit_bits;
>> >
>> >         if (pmu)
>> >                 return 0;
>> > @@ -518,6 +519,9 @@ static int rapl_cpu_prepare(int cpu)
>> >         if (phys_id < 0)
>> >                 return -1;
>> >
>> > +       if (!rdmsrl_safe(MSR_RAPL_POWER_UNIT, &msr_rapl_power_unit_bits))
>> > +               return -1;
>> > +
>> I have a problem with this patch on native hardware. This
>> rdmsrl_safe() systematically
>> fails when I know the MSR is perfectly valid on the CPU. Consequently, RAPL PMU
>> is disabled when I tried on IvyBridge and Haswell CPUs.
>>
>> I don't know the internals of rdmsrl_safe(). Maybe it is invoked too
>> early in the boot process.
>
> Weird; so the way it works is that it adds an exception table entry for
> the wrmsr instruction, so when the wrmsr generates a fault due to being
> an invalid msr the fault handler looks at the exception table, and finds
> the entry, which instructs it to continue execution at the error path
> and return -EFAULT.
>
On your machine, booted with 3.15-rc2, do you have /sys/devices/power?
If not, and you have at least an SNB, you should have RAPL and that
RAPL_UNIT MSR.

Proof is that if you read that MSR using /dev/cpu/msr it works just fine:
# modprobe msr
# rdmsr 0x606
a1003

So something is broken somewhere.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/