Linux Power Management development
 help / color / mirror / Atom feed
From: Mario Limonciello <superm1@kernel.org>
To: Zhongqiu Han <zhongqiu.han@oss.qualcomm.com>,
	"Rafael J. Wysocki" <rafael@kernel.org>,
	Linux PM <linux-pm@vger.kernel.org>
Cc: LKML <linux-kernel@vger.kernel.org>,
	Viresh Kumar <viresh.kumar@linaro.org>,
	Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>,
	Christian Loehle <christian.loehle@arm.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Vincent Guittot <vincent.guittot@linaro.org>
Subject: Re: [PATCH v2] cpufreq: intel_pstate: Adjust the .adjust_perf() driver callback
Date: Sun, 21 Jun 2026 12:07:17 -0700	[thread overview]
Message-ID: <645895b6-78d5-4ef9-890b-a032059d304e@kernel.org> (raw)
In-Reply-To: <56a6b35a-407f-4529-b655-27d6cde5e595@oss.qualcomm.com>

On 6/19/26 08:43, Zhongqiu Han wrote:
> On 6/19/2026 10:52 PM, Rafael J. Wysocki wrote:
>> From: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
>>
>> In some cases, the processor may not actually stick to the "desired"
>> performance level programmed through the driver's .adjust_perf()
>> callback and may go above it, which may not be desirable (for instance,
>> there may be a UCLAMP_MAX limit set for the task currently running on
>> the given CPU which should be respected).
>>
>> Address that by adjusting the .adjust_perf() callback to take an
>> additional argument, max_perf, representing the maximum allowed
>> performance level of the CPU and update the intel_pstate driver to
>> take that argument into account as appropriate.
>>
>> Accordingly, adjust cpufreq_driver_adjust_perf() and the other existing
>> user of .adjust_perf(), which is the amd-pstate driver (but the behavior
>> of that driver is not changed).
>>
>> While at it, also update the cpufreq_driver_adjust_perf()
>> documentation to reflect this change and some previous code
>> changes that have not been taken into account in it.
>>
>> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
>> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
> 
> Looks good to me. Thanks
> 
> Reviewed-by: Zhongqiu Han <zhongqiu.han@oss.qualcomm.com>

Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
> 
>> ---
>>
>> This is an update of
>>
>> https://lore.kernel.org/linux-pm/14060154.uLZWGnKmhe@rafael.j.wysocki/
>>
>> sent mainly because the v1 did not update the Rust bindings by
>> omission.  It also fixes a few typos present in the v1.
>>
>> Thanks!
>>
>> ---
>>   drivers/cpufreq/amd-pstate.c     |    1 +
>>   drivers/cpufreq/cpufreq.c        |   14 +++++++++-----
>>   drivers/cpufreq/intel_pstate.c   |    9 ++++++++-
>>   include/linux/cpufreq.h          |    2 ++
>>   kernel/sched/cpufreq_schedutil.c |    4 +++-
>>   rust/kernel/cpufreq.rs           |    6 ++++--
>>   6 files changed, 27 insertions(+), 9 deletions(-)
>>
>> --- a/drivers/cpufreq/amd-pstate.c
>> +++ b/drivers/cpufreq/amd-pstate.c
>> @@ -781,6 +781,7 @@ static unsigned int amd_pstate_fast_swit
>>   static void amd_pstate_adjust_perf(struct cpufreq_policy *policy,
>>                      unsigned long _min_perf,
>>                      unsigned long target_perf,
>> +                   unsigned long _max_perf,
>>                      unsigned long capacity)
>>   {
>>       u8 max_perf, min_perf, des_perf, cap_perf;
>> --- a/drivers/cpufreq/cpufreq.c
>> +++ b/drivers/cpufreq/cpufreq.c
>> @@ -2252,14 +2252,17 @@ EXPORT_SYMBOL_GPL(cpufreq_driver_fast_sw
>>    * @policy: cpufreq policy object of the target CPU.
>>    * @min_perf: Minimum (required) performance level (units of 
>> @capacity).
>>    * @target_perf: Target (desired) performance level (units of 
>> @capacity).
>> + * @max_perf: Maximum (allowed) performance level (units of @capacity).
>>    * @capacity: Capacity of the target CPU.
>>    *
>> - * Carry out a fast performance level switch of @cpu without sleeping.
>> + * Carry out a fast performance level adjustment for the CPU 
>> represented by
>> + * @policy without sleeping.
>>    *
>>    * The driver's ->adjust_perf() callback invoked by this function 
>> must be
>> - * suitable for being called from within RCU-sched read-side critical 
>> sections
>> - * and it is expected to select a suitable performance level equal to 
>> or above
>> - * @min_perf and preferably equal to or below @target_perf.
>> + * suitable for calling from within RCU-sched read-side critical 
>> sections and
>> + * it is expected to program the processor to select suitable 
>> performance
>> + * levels between @min_perf and @max_perf inclusive and preferably 
>> close to
>> + * @target_perf going forward for the CPU represented by @policy.
>>    *
>>    * This function must not be called if policy->fast_switch_enabled 
>> is unset.
>>    *
>> @@ -2271,9 +2274,10 @@ EXPORT_SYMBOL_GPL(cpufreq_driver_fast_sw
>>   void cpufreq_driver_adjust_perf(struct cpufreq_policy *policy,
>>                    unsigned long min_perf,
>>                    unsigned long target_perf,
>> +                 unsigned long max_perf,
>>                    unsigned long capacity)
>>   {
>> -    cpufreq_driver->adjust_perf(policy, min_perf, target_perf, 
>> capacity);
>> +    cpufreq_driver->adjust_perf(policy, min_perf, target_perf, 
>> max_perf, capacity);
>>   }
>>   /**
>> --- a/drivers/cpufreq/intel_pstate.c
>> +++ b/drivers/cpufreq/intel_pstate.c
>> @@ -3241,6 +3241,7 @@ static unsigned int intel_cpufreq_fast_s
>>   static void intel_cpufreq_adjust_perf(struct cpufreq_policy *policy,
>>                         unsigned long min_perf,
>>                         unsigned long target_perf,
>> +                      unsigned long max_perf,
>>                         unsigned long capacity)
>>   {
>>       struct cpudata *cpu = all_cpu_data[policy->cpu];
>> @@ -3271,7 +3272,13 @@ static void intel_cpufreq_adjust_perf(st
>>       if (min_pstate > cpu->max_perf_ratio)
>>           min_pstate = cpu->max_perf_ratio;
>> -    max_pstate = min(cap_pstate, cpu->max_perf_ratio);
>> +    max_pstate = cap_pstate;
>> +    if (max_perf < capacity)
>> +        max_pstate = DIV_ROUND_UP(cap_pstate * max_perf, capacity);
>> +
>> +    if (max_pstate > cpu->max_perf_ratio)
>> +        max_pstate = cpu->max_perf_ratio;
>> +
>>       if (max_pstate < min_pstate)
>>           max_pstate = min_pstate;
>> --- a/include/linux/cpufreq.h
>> +++ b/include/linux/cpufreq.h
>> @@ -379,6 +379,7 @@ struct cpufreq_driver {
>>       void        (*adjust_perf)(struct cpufreq_policy *policy,
>>                          unsigned long min_perf,
>>                          unsigned long target_perf,
>> +                       unsigned long max_perf,
>>                          unsigned long capacity);
>>       /*
>> @@ -624,6 +625,7 @@ unsigned int cpufreq_driver_fast_switch(
>>   void cpufreq_driver_adjust_perf(struct cpufreq_policy *policy,
>>                   unsigned long min_perf,
>>                   unsigned long target_perf,
>> +                unsigned long max_perf,
>>                   unsigned long capacity);
>>   bool cpufreq_driver_has_adjust_perf(void);
>>   int cpufreq_driver_target(struct cpufreq_policy *policy,
>> --- a/kernel/sched/cpufreq_schedutil.c
>> +++ b/kernel/sched/cpufreq_schedutil.c
>> @@ -50,6 +50,7 @@ struct sugov_cpu {
>>       unsigned long        util;
>>       unsigned long        bw_min;
>> +    unsigned long        bw_max;
>>       /* The field below is for single-CPU policies only: */
>>   #ifdef CONFIG_NO_HZ_COMMON
>> @@ -232,6 +233,7 @@ static void sugov_get_util(struct sugov_
>>       util = effective_cpu_util(sg_cpu->cpu, util, &min, &max);
>>       util = max(util, boost);
>>       sg_cpu->bw_min = min;
>> +    sg_cpu->bw_max = max;
>>       sg_cpu->util = sugov_effective_cpu_perf(sg_cpu->cpu, util, min, 
>> max);
>>   }
>> @@ -484,7 +486,7 @@ static void sugov_update_single_perf(str
>>           sg_cpu->util = prev_util;
>>       cpufreq_driver_adjust_perf(sg_policy->policy, sg_cpu->bw_min,
>> -                   sg_cpu->util, max_cap);
>> +                   sg_cpu->util, sg_cpu->bw_max, max_cap);
>>       sg_policy->last_freq_update_time = time;
>>   }
>> --- a/rust/kernel/cpufreq.rs
>> +++ b/rust/kernel/cpufreq.rs
>> @@ -792,7 +792,8 @@ pub trait Driver {
>>       }
>>       /// Driver's `adjust_perf` callback.
>> -    fn adjust_perf(_policy: &mut Policy, _min_perf: usize, 
>> _target_perf: usize, _capacity: usize) {
>> +    fn adjust_perf(_policy: &mut Policy, _min_perf: usize, 
>> _target_perf: usize,
>> +                   _max_perf: usize, _capacity: usize) {
>>           build_error!(VTABLE_DEFAULT_ERROR)
>>       }
>> @@ -1262,12 +1263,13 @@ impl<T: Driver> Registration<T> {
>>           ptr: *mut bindings::cpufreq_policy,
>>           min_perf: c_ulong,
>>           target_perf: c_ulong,
>> +        max_perf: c_ulong,
>>           capacity: c_ulong,
>>       ) {
>>           // SAFETY: The `ptr` is guaranteed to be valid by the 
>> contract with the C code for the
>>           // lifetime of `policy`.
>>           let policy = unsafe { Policy::from_raw_mut(ptr) };
>> -        T::adjust_perf(policy, min_perf, target_perf, capacity);
>> +        T::adjust_perf(policy, min_perf, target_perf, max_perf, 
>> capacity);
>>       }
>>       /// Driver's `get_intermediate` callback.
>>
>>
>>
> 
> 



      reply	other threads:[~2026-06-21 19:07 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-06-19 14:52 [PATCH v2] cpufreq: intel_pstate: Adjust the .adjust_perf() driver callback Rafael J. Wysocki
2026-06-19 15:43 ` Zhongqiu Han
2026-06-21 19:07   ` Mario Limonciello [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=645895b6-78d5-4ef9-890b-a032059d304e@kernel.org \
    --to=superm1@kernel.org \
    --cc=christian.loehle@arm.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-pm@vger.kernel.org \
    --cc=peterz@infradead.org \
    --cc=rafael@kernel.org \
    --cc=srinivas.pandruvada@linux.intel.com \
    --cc=vincent.guittot@linaro.org \
    --cc=viresh.kumar@linaro.org \
    --cc=zhongqiu.han@oss.qualcomm.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox