[PATCH] cpufreq: schedutil: Fix sugov_start versus sugov_update_shared race

[PATCH] cpufreq: schedutil: Fix sugov_start versus sugov_update_shared race

[Vikram Mulukutla]

With a shared policy in place, when one of the CPUs in the policy is
hotplugged out and then brought back online, sugov_stop and
sugov_start are called in order.

sugov_stop removes utilization hooks for each CPU in the policy and
does nothing else in the for_each_cpu loop. sugov_start on the other
hand iterates through the CPUs in the policy and re-initializes the
per-cpu structure _and_ adds the utilization hook. This implies that
the scheduler is allowed to invoke a CPU's utilization update hook
when the rest of the per-cpu structures have yet to be re-inited.

Apart from some strange values in tracepoints this doesn't cause a
problem, but if we do end up accessing a pointer from the per-cpu
sugov_cpu structure somewhere in the sugov_update_shared path,
we will likely see crashes since the memset for another CPU in the
policy is free to race with sugov_update_shared from the CPU that is
ready to go. So let's fix this now to first init all per-cpu
structures, and then add the per-cpu utilization update hooks all at
once.

Signed-off-by: Vikram Mulukutla <markivx@codeaurora.org>
---
 kernel/sched/cpufreq_schedutil.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/kernel/sched/cpufreq_schedutil.c b/kernel/sched/cpufreq_schedutil.c
index 076a2e3..29a3970 100644
--- a/kernel/sched/cpufreq_schedutil.c
+++ b/kernel/sched/cpufreq_schedutil.c
@@ -610,6 +610,11 @@ static int sugov_start(struct cpufreq_policy *policy)
 		sg_cpu->sg_policy = sg_policy;
 		sg_cpu->flags = SCHED_CPUFREQ_RT;
 		sg_cpu->iowait_boost_max = policy->cpuinfo.max_freq;
+	}
+
+	for_each_cpu(cpu, policy->cpus) {
+		struct sugov_cpu *sg_cpu = &per_cpu(sugov_cpu, cpu);
+
 		cpufreq_add_update_util_hook(cpu, &sg_cpu->update_util,
 					     policy_is_shared(policy) ?
 							sugov_update_shared :
-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project

Re: [PATCH] cpufreq: schedutil: Fix sugov_start versus sugov_update_shared race

[Rafael J. Wysocki]

On Thursday, July 06, 2017 10:53:20 AM Vikram Mulukutla wrote:
> With a shared policy in place, when one of the CPUs in the policy is
> hotplugged out and then brought back online, sugov_stop and
> sugov_start are called in order.
> 
> sugov_stop removes utilization hooks for each CPU in the policy and
> does nothing else in the for_each_cpu loop. sugov_start on the other
> hand iterates through the CPUs in the policy and re-initializes the
> per-cpu structure _and_ adds the utilization hook. This implies that
> the scheduler is allowed to invoke a CPU's utilization update hook
> when the rest of the per-cpu structures have yet to be re-inited.
> 
> Apart from some strange values in tracepoints this doesn't cause a
> problem, but if we do end up accessing a pointer from the per-cpu
> sugov_cpu structure somewhere in the sugov_update_shared path,
> we will likely see crashes since the memset for another CPU in the
> policy is free to race with sugov_update_shared from the CPU that is
> ready to go. So let's fix this now to first init all per-cpu
> structures, and then add the per-cpu utilization update hooks all at
> once.
> 
> Signed-off-by: Vikram Mulukutla <markivx@codeaurora.org>

OK

Any objections anyone?

> ---
>  kernel/sched/cpufreq_schedutil.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/kernel/sched/cpufreq_schedutil.c b/kernel/sched/cpufreq_schedutil.c
> index 076a2e3..29a3970 100644
> --- a/kernel/sched/cpufreq_schedutil.c
> +++ b/kernel/sched/cpufreq_schedutil.c
> @@ -610,6 +610,11 @@ static int sugov_start(struct cpufreq_policy *policy)
>  		sg_cpu->sg_policy = sg_policy;
>  		sg_cpu->flags = SCHED_CPUFREQ_RT;
>  		sg_cpu->iowait_boost_max = policy->cpuinfo.max_freq;
> +	}
> +
> +	for_each_cpu(cpu, policy->cpus) {
> +		struct sugov_cpu *sg_cpu = &per_cpu(sugov_cpu, cpu);
> +
>  		cpufreq_add_update_util_hook(cpu, &sg_cpu->update_util,
>  					     policy_is_shared(policy) ?
>  							sugov_update_shared :
> 

Re: [PATCH] cpufreq: schedutil: Fix sugov_start versus sugov_update_shared race

[Viresh Kumar]

On 06-07-17, 10:53, Vikram Mulukutla wrote:
> With a shared policy in place, when one of the CPUs in the policy is
> hotplugged out and then brought back online, sugov_stop and
> sugov_start are called in order.
> 
> sugov_stop removes utilization hooks for each CPU in the policy and
> does nothing else in the for_each_cpu loop. sugov_start on the other
> hand iterates through the CPUs in the policy and re-initializes the
> per-cpu structure _and_ adds the utilization hook. This implies that
> the scheduler is allowed to invoke a CPU's utilization update hook
> when the rest of the per-cpu structures have yet to be re-inited.
> 
> Apart from some strange values in tracepoints this doesn't cause a
> problem, but if we do end up accessing a pointer from the per-cpu
> sugov_cpu structure somewhere in the sugov_update_shared path,
> we will likely see crashes since the memset for another CPU in the
> policy is free to race with sugov_update_shared from the CPU that is
> ready to go. So let's fix this now to first init all per-cpu
> structures, and then add the per-cpu utilization update hooks all at
> once.
> 
> Signed-off-by: Vikram Mulukutla <markivx@codeaurora.org>
> ---
>  kernel/sched/cpufreq_schedutil.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/kernel/sched/cpufreq_schedutil.c b/kernel/sched/cpufreq_schedutil.c
> index 076a2e3..29a3970 100644
> --- a/kernel/sched/cpufreq_schedutil.c
> +++ b/kernel/sched/cpufreq_schedutil.c
> @@ -610,6 +610,11 @@ static int sugov_start(struct cpufreq_policy *policy)
>  		sg_cpu->sg_policy = sg_policy;
>  		sg_cpu->flags = SCHED_CPUFREQ_RT;
>  		sg_cpu->iowait_boost_max = policy->cpuinfo.max_freq;
> +	}
> +
> +	for_each_cpu(cpu, policy->cpus) {
> +		struct sugov_cpu *sg_cpu = &per_cpu(sugov_cpu, cpu);
> +
>  		cpufreq_add_update_util_hook(cpu, &sg_cpu->update_util,
>  					     policy_is_shared(policy) ?
>  							sugov_update_shared :

Acked-by: Viresh Kumar <viresh.kumar@linaro.org>

-- 
viresh

Re: [PATCH] cpufreq: schedutil: Fix sugov_start versus sugov_update_shared race

[Rafael J. Wysocki]

On Friday, July 07, 2017 08:55:09 AM Viresh Kumar wrote:
> On 06-07-17, 10:53, Vikram Mulukutla wrote:
> > With a shared policy in place, when one of the CPUs in the policy is
> > hotplugged out and then brought back online, sugov_stop and
> > sugov_start are called in order.
> > 
> > sugov_stop removes utilization hooks for each CPU in the policy and
> > does nothing else in the for_each_cpu loop. sugov_start on the other
> > hand iterates through the CPUs in the policy and re-initializes the
> > per-cpu structure _and_ adds the utilization hook. This implies that
> > the scheduler is allowed to invoke a CPU's utilization update hook
> > when the rest of the per-cpu structures have yet to be re-inited.
> > 
> > Apart from some strange values in tracepoints this doesn't cause a
> > problem, but if we do end up accessing a pointer from the per-cpu
> > sugov_cpu structure somewhere in the sugov_update_shared path,
> > we will likely see crashes since the memset for another CPU in the
> > policy is free to race with sugov_update_shared from the CPU that is
> > ready to go. So let's fix this now to first init all per-cpu
> > structures, and then add the per-cpu utilization update hooks all at
> > once.
> > 
> > Signed-off-by: Vikram Mulukutla <markivx@codeaurora.org>
> > ---
> >  kernel/sched/cpufreq_schedutil.c | 5 +++++
> >  1 file changed, 5 insertions(+)
> > 
> > diff --git a/kernel/sched/cpufreq_schedutil.c b/kernel/sched/cpufreq_schedutil.c
> > index 076a2e3..29a3970 100644
> > --- a/kernel/sched/cpufreq_schedutil.c
> > +++ b/kernel/sched/cpufreq_schedutil.c
> > @@ -610,6 +610,11 @@ static int sugov_start(struct cpufreq_policy *policy)
> >  		sg_cpu->sg_policy = sg_policy;
> >  		sg_cpu->flags = SCHED_CPUFREQ_RT;
> >  		sg_cpu->iowait_boost_max = policy->cpuinfo.max_freq;
> > +	}
> > +
> > +	for_each_cpu(cpu, policy->cpus) {
> > +		struct sugov_cpu *sg_cpu = &per_cpu(sugov_cpu, cpu);
> > +
> >  		cpufreq_add_update_util_hook(cpu, &sg_cpu->update_util,
> >  					     policy_is_shared(policy) ?
> >  							sugov_update_shared :
> 
> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>

Applied, thanks!