Re: [PATCH] cpufreq: acpi_cpufreq: base frequency attribute support

[Viresh Kumar <viresh.kumar@linaro.org>]

On 01-03-16, 10:10, Srinivas Pandruvada wrote:
> > What about:
> > 
> >                                 acpi_cpufreq_attr[ARRAY_SIZE(acpi_cpu
> > freq_attr) - 2] = &base_frequency;
> > 
> >                                 and a comment to describe that ?
> > 
> Will it work in this case?
> "
> Even if CONFIG_X86_ACPI_CPUFREQ_CPB is defined, 
> acpi_cpufreq_attr[2] will be set to NULL in the init callback when
> some condition fails. So we will have a NULL in the acpi_cpufreq_attr[]
> before &base_frequency if we add at fixed place. 
> "

Oh dude, you already have this hack in place in a different way. Though its
doable this way:

diff --git a/drivers/cpufreq/acpi-cpufreq.c b/drivers/cpufreq/acpi-cpufreq.c
index 51eef87bbc37..26c026a34e1e 100644
--- a/drivers/cpufreq/acpi-cpufreq.c
+++ b/drivers/cpufreq/acpi-cpufreq.c
@@ -937,7 +937,7 @@ static void acpi_cpufreq_boost_exit(void)
 
 static int __init acpi_cpufreq_init(void)
 {
-       int ret;
+       int ret, cpb_removed = 0;
 
        if (acpi_disabled)
                return -ENODEV;
@@ -967,6 +967,7 @@ static int __init acpi_cpufreq_init(void)
                for (attr = acpi_cpufreq_attr; *attr; attr++)
                        if (*attr == &cpb) {
                                *attr = NULL;
+                               cpb_removed = 1;
                                break;
                        }
        }

And then the above statement:

acpi_cpufreq_attr[ARRAY_SIZE(acpi_cpufreq_attr) - 2 - cpb_removed] = &base_frequency;

But I would rather add another patch first to set the CPB thing in the array to
NULL and remove the ifdef from within the array. And then set it to a valid
attribute only if we are making an addition. And then you can easily keep a
count of the next NULL entry you can update.

-- 
viresh