2021-08-27 08:22:10

by Schspa Shi

[permalink] [raw]
Subject: [PATCH] cpufreq_ondemand: fix bad auto calculated frequency.

We can litmit cpufreq range by change min & max from cpufreq_policy.
So cpu frequency target should be in range [policy->min, policy->max].

Signed-off-by: schspa <[email protected]>
---
drivers/cpufreq/cpufreq_ondemand.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/cpufreq/cpufreq_ondemand.c b/drivers/cpufreq/cpufreq_ondemand.c
index ac361a8b1d3b..8afb2c84c38c 100644
--- a/drivers/cpufreq/cpufreq_ondemand.c
+++ b/drivers/cpufreq/cpufreq_ondemand.c
@@ -151,8 +151,8 @@ static void od_update(struct cpufreq_policy *policy)
/* Calculate the next frequency proportional to load */
unsigned int freq_next, min_f, max_f;

- min_f = policy->cpuinfo.min_freq;
- max_f = policy->cpuinfo.max_freq;
+ min_f = policy->min;
+ max_f = policy->max;
freq_next = min_f + load * (max_f - min_f) / 100;

/* No longer fully busy, reset rate_mult */
--
2.29.0


2021-08-31 05:01:26

by Viresh Kumar

[permalink] [raw]
Subject: Re: [PATCH] cpufreq_ondemand: fix bad auto calculated frequency.

On 27-08-21, 16:17, schspa wrote:
> We can litmit cpufreq range by change min & max from cpufreq_policy.
> So cpu frequency target should be in range [policy->min, policy->max].
>
> Signed-off-by: schspa <[email protected]>
> ---
> drivers/cpufreq/cpufreq_ondemand.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/cpufreq/cpufreq_ondemand.c b/drivers/cpufreq/cpufreq_ondemand.c
> index ac361a8b1d3b..8afb2c84c38c 100644
> --- a/drivers/cpufreq/cpufreq_ondemand.c
> +++ b/drivers/cpufreq/cpufreq_ondemand.c
> @@ -151,8 +151,8 @@ static void od_update(struct cpufreq_policy *policy)
> /* Calculate the next frequency proportional to load */
> unsigned int freq_next, min_f, max_f;
>
> - min_f = policy->cpuinfo.min_freq;
> - max_f = policy->cpuinfo.max_freq;
> + min_f = policy->min;
> + max_f = policy->max;
> freq_next = min_f + load * (max_f - min_f) / 100;

No. The current calculations are right. The new frequency is proportional to
current load and it needs to take into account the entire freq range of the
CPUs. Note that we will eventually try to get the resultant frequency within
policy->min/max range in __cpufreq_driver_target().

--
viresh