Re: [PATCH v6 1/4] cpufreq: Remove per-CPU QoS constraint

[Pierre Gondois]

Hello Viresh,

On 3/20/26 10:18, Viresh Kumar wrote:
> On 19-03-26, 10:30, Pierre Gondois wrote:
> > On 3/18/26 12:13, Viresh Kumar wrote:
> > > On 17-03-26, 11:17, Pierre Gondois wrote:
> > > > policy->max_freq_req represents the maximum allowed frequency as
> > > > requested by the policyX/scaling_max_freq sysfs file. This request
> > > > applies to all CPUs of the policy. It is not possible to request
> > > > a per-CPU maximum frequency.
> > > >
> > > > Thus, the interaction between the policy boost and scaling_max_freq
> > > > settings should be handled by adding a boost specific QoS constraint.
> > > > This will be handled in the following patches.
> > > I don't think the above is required anymore. This patch is removing stale code
> > > now which isn't useful anymore. It has nothing to do with a boost specific QOS
> > > constraint.
> > Yes ok
> > > And it would be better to know for sure why this isn't required anymore and
> > > which patch exactly fixed this issue.
> > On a kernel based on 1608f0230510~, and replicating the
> > process described in the commit message of
> >
> > commit 1608f0230510 ("cpufreq: Fix re-boost issue after hotplugging
> > a CPU")
> >
> > I could not see any issue regarding the values of:
> >
> > - policy1/cpuinfo_max_freq
> > - policy1/scaling_max_freq
> The commit message (of 1608f0230510) is confusing. The issue was discussed
> properly in the following thread.
>
> https://lore.kernel.org/all/20250120082723.am7rxujmdvzz4eky@vireshk-i7/
>
> The problem is that policy->max and policy->cpuinfo_max_freq are incorrect after
> the sequence mentioned in the commit, while max_freq_req is correct.

I experimented a bit more and it seems the following happens:

1. boost all CPUs: echo 1 > /sys/devices/system/cpu/cpufreq/boost
2. offline one CPU: echo 0 > /sys/devices/system/cpu/cpuX/online
3. deboost all CPUs: echo 0 > /sys/devices/system/cpu/cpufreq/boost

cpufreq_boost_trigger_state()
\-for_each_active_policy()
  \-cpufreq_driver->set_boost()
doesn't act on the policy where there are no more online CPUs,
so the max/cpuinfo.max/max_freq_req is left to the actual
boost freq.

4. online CPUX: echo 1 > /sys/devices/system/cpu/cpuX/online

cpufreq_online()
\-cpufreq_driver->init()
  \-cppc_cpufreq_cpu_init()
There:
- policy->max
- policy->cpuinfo.max_freq
are set to the maximal non-boost freq., which is the correct value.

However, max_freq_req is left to the boosted frequency, so this
is effectively an incorrect state.
Also in cpufreq_set_policy(), policy->max is set to:
  min(max_freq_req, cpuinfo.max_freq)
(cf. verify() cb), so the incorrect state of max_freq_req is not
visible.

5. boost all CPUs again: echo 1 > /sys/devices/system/cpu/cpufreq/boost

As the max_freq_req value and the new boost value are equal,
cpufreq_notifier_max() won't be called, which means that if the
CPU needed to raise its freq., it won't be notified until another
event trigger a re-evaluation of the freq. selection.

To observe that, I had to:
- use the performance governor to be sure to select the max. available
  freq.
- observe scaling_cur_freq to see the last requested freq. for the CPU

So IMO the issue was actually fixed by:
dd016f379ebc ("cpufreq: Introduce a more generic way to set default
per-policy boost flag")
which sets the correct max_freq_req value when putting back an
inactive policy.


> I think another commit has fixed that (incorrectly and unintentionally):
> commit 6db0f533d320 ("cpufreq: preserve freq_table_sorted across suspend/hibernate")
>
> @@ -1421,9 +1421,12 @@ static int cpufreq_policy_online(struct cpufreq_policy *policy,
>                   * If there is a problem with its frequency table, take it
>                   * offline and drop it.
>                   */
> -               ret = cpufreq_table_validate_and_sort(policy);
> -               if (ret)
> -                       goto out_offline_policy;
> +               if (policy->freq_table_sorted != CPUFREQ_TABLE_SORTED_ASCENDING &&
> +                   policy->freq_table_sorted != CPUFREQ_TABLE_SORTED_DESCENDING) {
> +                       ret = cpufreq_table_validate_and_sort(policy);
> +                       if (ret)
> +                               goto out_offline_policy;
> +               }
>
> This skipped calling cpufreq_table_validate_and_sort() completely on online and
> so max/cpuinfo_max_freq, max_freq_req are all in sync.
>
> That change should be fixed with:
>
> diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
> index 277884d91913..1f794524a1d9 100644
> --- a/drivers/cpufreq/cpufreq.c
> +++ b/drivers/cpufreq/cpufreq.c
> @@ -1427,12 +1427,9 @@ static int cpufreq_policy_online(struct cpufreq_policy *policy,
>                   * If there is a problem with its frequency table, take it
>                   * offline and drop it.
>                   */
> -               if (policy->freq_table_sorted != CPUFREQ_TABLE_SORTED_ASCENDING &&
> -                   policy->freq_table_sorted != CPUFREQ_TABLE_SORTED_DESCENDING) {
> -                       ret = cpufreq_table_validate_and_sort(policy);
> -                       if (ret)
> -                               goto out_offline_policy;
> -               }
> +               ret = cpufreq_table_validate_and_sort(policy);
> +               if (ret)
> +                       goto out_offline_policy;
>
>                  /* related_cpus should at least include policy->cpus. */
>                  cpumask_copy(policy->related_cpus, policy->cpus);
> diff --git a/drivers/cpufreq/freq_table.c b/drivers/cpufreq/freq_table.c
> index 7f251daf03ce..5b364d8da4f9 100644
> --- a/drivers/cpufreq/freq_table.c
> +++ b/drivers/cpufreq/freq_table.c
> @@ -360,6 +360,10 @@ int cpufreq_table_validate_and_sort(struct cpufreq_policy *policy)
>          if (policy_has_boost_freq(policy))
>                  policy->boost_supported = true;
>
> +       if (policy->freq_table_sorted == CPUFREQ_TABLE_SORTED_ASCENDING ||
> +           policy->freq_table_sorted == CPUFREQ_TABLE_SORTED_DESCENDING)
> +               return 0;
> +
>          return set_freq_table_sorted(policy);
>   }