]> git.karo-electronics.de Git - linux-beck.git/commitdiff
cpufreq: ondemand: allow custom powersave_bias_target handler to be registered
authorJacob Shin <jacob.shin@amd.com>
Tue, 2 Apr 2013 14:56:56 +0000 (09:56 -0500)
committerRafael J. Wysocki <rafael.j.wysocki@intel.com>
Wed, 10 Apr 2013 11:19:26 +0000 (13:19 +0200)
This allows for another [arch specific] driver to hook into existing
powersave bias function of the ondemand governor. i.e. This allows AMD
specific powersave bias function (in a separate AMD specific driver)
to aid ondemand governor's frequency transition decisions.

Signed-off-by: Jacob Shin <jacob.shin@amd.com>
Acked-by: Thomas Renninger <trenn@suse.de>
Acked-by: Borislav Petkov <bp@suse.de>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
drivers/cpufreq/cpufreq_governor.h
drivers/cpufreq/cpufreq_ondemand.c

index 513cc8234e5e712bff8c527f493c756325e5aede..9f668e94dead4bc21439f2a214c4f68c84d9327b 100644 (file)
@@ -263,4 +263,8 @@ int cpufreq_governor_dbs(struct cpufreq_policy *policy,
                struct common_dbs_data *cdata, unsigned int event);
 void gov_queue_work(struct dbs_data *dbs_data, struct cpufreq_policy *policy,
                unsigned int delay, bool all_cpus);
+void od_register_powersave_bias_handler(unsigned int (*f)
+               (struct cpufreq_policy *, unsigned int, unsigned int),
+               unsigned int powersave_bias);
+void od_unregister_powersave_bias_handler(void);
 #endif /* _CPUFREQ_GOVERNER_H */
index 14714787b7240056111674f291159bb6032c9c2a..b0ffef96bf77a2c1bcaf0680c6a500a7430a6858 100644 (file)
@@ -24,6 +24,7 @@
 #include <linux/sysfs.h>
 #include <linux/tick.h>
 #include <linux/types.h>
+#include <linux/cpu.h>
 
 #include "cpufreq_governor.h"
 
@@ -40,6 +41,8 @@
 
 static DEFINE_PER_CPU(struct od_cpu_dbs_info_s, od_cpu_dbs_info);
 
+static struct od_ops od_ops;
+
 #ifndef CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND
 static struct cpufreq_governor cpufreq_gov_ondemand;
 #endif
@@ -80,7 +83,7 @@ static int should_io_be_busy(void)
  * Returns the freq_hi to be used right now and will set freq_hi_jiffies,
  * freq_lo, and freq_lo_jiffies in percpu area for averaging freqs.
  */
-static unsigned int powersave_bias_target(struct cpufreq_policy *policy,
+static unsigned int generic_powersave_bias_target(struct cpufreq_policy *policy,
                unsigned int freq_next, unsigned int relation)
 {
        unsigned int freq_req, freq_reduc, freq_avg;
@@ -145,7 +148,8 @@ static void dbs_freq_increase(struct cpufreq_policy *p, unsigned int freq)
        struct od_dbs_tuners *od_tuners = dbs_data->tuners;
 
        if (od_tuners->powersave_bias)
-               freq = powersave_bias_target(p, freq, CPUFREQ_RELATION_H);
+               freq = od_ops.powersave_bias_target(p, freq,
+                               CPUFREQ_RELATION_H);
        else if (p->cur == p->max)
                return;
 
@@ -205,12 +209,12 @@ static void od_check_cpu(int cpu, unsigned int load_freq)
                if (!od_tuners->powersave_bias) {
                        __cpufreq_driver_target(policy, freq_next,
                                        CPUFREQ_RELATION_L);
-               } else {
-                       int freq = powersave_bias_target(policy, freq_next,
-                                       CPUFREQ_RELATION_L);
-                       __cpufreq_driver_target(policy, freq,
-                                       CPUFREQ_RELATION_L);
+                       return;
                }
+
+               freq_next = od_ops.powersave_bias_target(policy, freq_next,
+                                       CPUFREQ_RELATION_L);
+               __cpufreq_driver_target(policy, freq_next, CPUFREQ_RELATION_L);
        }
 }
 
@@ -557,7 +561,7 @@ define_get_cpu_dbs_routines(od_cpu_dbs_info);
 
 static struct od_ops od_ops = {
        .powersave_bias_init_cpu = ondemand_powersave_bias_init_cpu,
-       .powersave_bias_target = powersave_bias_target,
+       .powersave_bias_target = generic_powersave_bias_target,
        .freq_increase = dbs_freq_increase,
 };
 
@@ -574,6 +578,47 @@ static struct common_dbs_data od_dbs_cdata = {
        .exit = od_exit,
 };
 
+static void od_set_powersave_bias(unsigned int powersave_bias)
+{
+       struct cpufreq_policy *policy;
+       struct dbs_data *dbs_data;
+       struct od_dbs_tuners *od_tuners;
+       unsigned int cpu;
+       cpumask_t done;
+
+       cpumask_clear(&done);
+
+       get_online_cpus();
+       for_each_online_cpu(cpu) {
+               if (cpumask_test_cpu(cpu, &done))
+                       continue;
+
+               policy = per_cpu(od_cpu_dbs_info, cpu).cdbs.cur_policy;
+               dbs_data = policy->governor_data;
+               od_tuners = dbs_data->tuners;
+               od_tuners->powersave_bias = powersave_bias;
+
+               cpumask_or(&done, &done, policy->cpus);
+       }
+       put_online_cpus();
+}
+
+void od_register_powersave_bias_handler(unsigned int (*f)
+               (struct cpufreq_policy *, unsigned int, unsigned int),
+               unsigned int powersave_bias)
+{
+       od_ops.powersave_bias_target = f;
+       od_set_powersave_bias(powersave_bias);
+}
+EXPORT_SYMBOL_GPL(od_register_powersave_bias_handler);
+
+void od_unregister_powersave_bias_handler(void)
+{
+       od_ops.powersave_bias_target = generic_powersave_bias_target;
+       od_set_powersave_bias(0);
+}
+EXPORT_SYMBOL_GPL(od_unregister_powersave_bias_handler);
+
 static int od_cpufreq_governor_dbs(struct cpufreq_policy *policy,
                unsigned int event)
 {