On Tue, Mar 19, 2019 at 6:50 AM Viresh Kumar <viresh.kumar@xxxxxxxxxx> wrote:
>
> On 18-03-19, 12:49, Rafael J. Wysocki wrote:
> > To summarize, I think that it would be sufficient to do this just for
> > policy->cpu and, as Peter said, warn once if there are more CPUs in
> > the policy or policy->cpu is not the CPU running this code. And mark
> > the TSC as unstable in both of these cases.
>
> How about this ?
We guarantee that this will always run on policy->cpu IIRC, so it LGTM
overall, but the new message is missing "one".
> diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
> index 3fae23834069..4d3681cfb6e0 100644
> --- a/arch/x86/kernel/tsc.c
> +++ b/arch/x86/kernel/tsc.c
> @@ -958,10 +958,13 @@ static int time_cpufreq_notifier(struct notifier_block *nb, unsigned long val,
> struct cpufreq_freqs *freq = data;
> unsigned long *lpj;
>
> + if (WARN_ON_ONCE(cpumask_weight(freq->policy->related_cpus) != 1))
> + mark_tsc_unstable("cpufreq policy has more than CPU");
Also I would check policy->cpus here. After all, we don't care about
CPUs that are never online.
And the message could be something like "cpufreq changes: related CPUs
affected".
> +
> lpj = &boot_cpu_data.loops_per_jiffy;
> #ifdef CONFIG_SMP
> if (!(freq->flags & CPUFREQ_CONST_LOOPS))
> - lpj = &cpu_data(freq->cpu).loops_per_jiffy;
> + lpj = &cpu_data(freq->policy->cpu).loops_per_jiffy;
> #endif
>
> if (!ref_freq) {
> @@ -977,7 +980,7 @@ static int time_cpufreq_notifier(struct notifier_block *nb, unsigned long val,
> if (!(freq->flags & CPUFREQ_CONST_LOOPS))
> mark_tsc_unstable("cpufreq changes");
>
> - set_cyc2ns_scale(tsc_khz, freq->cpu, rdtsc());
> + set_cyc2ns_scale(tsc_khz, freq->policy->cpu, rdtsc());
> }
>
> return 0;
