Merge branches 'intel_pstate' and 'pm-domains'
authorRafael J. Wysocki <rafael.j.wysocki@intel.com>
Thu, 20 Jul 2017 16:57:15 +0000 (18:57 +0200)
committerRafael J. Wysocki <rafael.j.wysocki@intel.com>
Thu, 20 Jul 2017 16:57:15 +0000 (18:57 +0200)
* intel_pstate:
  cpufreq: intel_pstate: Correct the busy calculation for KNL

* pm-domains:
  PM / Domains: defer dev_pm_domain_set() until genpd->attach_dev succeeds if present

drivers/base/power/domain.c
drivers/cpufreq/intel_pstate.c

index 3b8210ebb50ebfd1844622a883193fecdacaa282..60303aa28587b9079317a8fc3500ef1aa70ec247 100644 (file)
@@ -1222,8 +1222,6 @@ static struct generic_pm_domain_data *genpd_alloc_dev_data(struct device *dev,
 
        spin_unlock_irq(&dev->power.lock);
 
-       dev_pm_domain_set(dev, &genpd->domain);
-
        return gpd_data;
 
  err_free:
@@ -1237,8 +1235,6 @@ static struct generic_pm_domain_data *genpd_alloc_dev_data(struct device *dev,
 static void genpd_free_dev_data(struct device *dev,
                                struct generic_pm_domain_data *gpd_data)
 {
-       dev_pm_domain_set(dev, NULL);
-
        spin_lock_irq(&dev->power.lock);
 
        dev->power.subsys_data->domain_data = NULL;
@@ -1275,6 +1271,8 @@ static int genpd_add_device(struct generic_pm_domain *genpd, struct device *dev,
        if (ret)
                goto out;
 
+       dev_pm_domain_set(dev, &genpd->domain);
+
        genpd->device_count++;
        genpd->max_off_time_changed = true;
 
@@ -1336,6 +1334,8 @@ static int genpd_remove_device(struct generic_pm_domain *genpd,
        if (genpd->detach_dev)
                genpd->detach_dev(genpd, dev);
 
+       dev_pm_domain_set(dev, NULL);
+
        list_del_init(&pdd->list_node);
 
        genpd_unlock(genpd);
index b7fb8b7c980dc5d2eb2001e4790910cb5b399514..6cd50352563894bea24866d2d36d6b815d0abc20 100644 (file)
@@ -225,6 +225,9 @@ struct global_params {
  * @vid:               Stores VID limits for this CPU
  * @pid:               Stores PID parameters for this CPU
  * @last_sample_time:  Last Sample time
+ * @aperf_mperf_shift: Number of clock cycles after aperf, merf is incremented
+ *                     This shift is a multiplier to mperf delta to
+ *                     calculate CPU busy.
  * @prev_aperf:                Last APERF value read from APERF MSR
  * @prev_mperf:                Last MPERF value read from MPERF MSR
  * @prev_tsc:          Last timestamp counter (TSC) value
@@ -259,6 +262,7 @@ struct cpudata {
 
        u64     last_update;
        u64     last_sample_time;
+       u64     aperf_mperf_shift;
        u64     prev_aperf;
        u64     prev_mperf;
        u64     prev_tsc;
@@ -321,6 +325,7 @@ struct pstate_funcs {
        int (*get_min)(void);
        int (*get_turbo)(void);
        int (*get_scaling)(void);
+       int (*get_aperf_mperf_shift)(void);
        u64 (*get_val)(struct cpudata*, int pstate);
        void (*get_vid)(struct cpudata *);
        void (*update_util)(struct update_util_data *data, u64 time,
@@ -1486,6 +1491,11 @@ static u64 core_get_val(struct cpudata *cpudata, int pstate)
        return val;
 }
 
+static int knl_get_aperf_mperf_shift(void)
+{
+       return 10;
+}
+
 static int knl_get_turbo_pstate(void)
 {
        u64 value;
@@ -1543,6 +1553,9 @@ static void intel_pstate_get_cpu_pstates(struct cpudata *cpu)
        cpu->pstate.max_freq = cpu->pstate.max_pstate * cpu->pstate.scaling;
        cpu->pstate.turbo_freq = cpu->pstate.turbo_pstate * cpu->pstate.scaling;
 
+       if (pstate_funcs.get_aperf_mperf_shift)
+               cpu->aperf_mperf_shift = pstate_funcs.get_aperf_mperf_shift();
+
        if (pstate_funcs.get_vid)
                pstate_funcs.get_vid(cpu);
 
@@ -1616,7 +1629,8 @@ static inline int32_t get_target_pstate_use_cpu_load(struct cpudata *cpu)
        int32_t busy_frac, boost;
        int target, avg_pstate;
 
-       busy_frac = div_fp(sample->mperf, sample->tsc);
+       busy_frac = div_fp(sample->mperf << cpu->aperf_mperf_shift,
+                          sample->tsc);
 
        boost = cpu->iowait_boost;
        cpu->iowait_boost >>= 1;
@@ -1675,7 +1689,8 @@ static inline int32_t get_target_pstate_use_performance(struct cpudata *cpu)
                sample_ratio = div_fp(pid_params.sample_rate_ns, duration_ns);
                perf_scaled = mul_fp(perf_scaled, sample_ratio);
        } else {
-               sample_ratio = div_fp(100 * cpu->sample.mperf, cpu->sample.tsc);
+               sample_ratio = div_fp(100 * (cpu->sample.mperf << cpu->aperf_mperf_shift),
+                                     cpu->sample.tsc);
                if (sample_ratio < int_tofp(1))
                        perf_scaled = 0;
        }
@@ -1807,6 +1822,7 @@ static const struct pstate_funcs knl_funcs = {
        .get_max_physical = core_get_max_pstate_physical,
        .get_min = core_get_min_pstate,
        .get_turbo = knl_get_turbo_pstate,
+       .get_aperf_mperf_shift = knl_get_aperf_mperf_shift,
        .get_scaling = core_get_scaling,
        .get_val = core_get_val,
        .update_util = intel_pstate_update_util_pid,
@@ -2403,6 +2419,7 @@ static void __init copy_cpu_funcs(struct pstate_funcs *funcs)
        pstate_funcs.get_val   = funcs->get_val;
        pstate_funcs.get_vid   = funcs->get_vid;
        pstate_funcs.update_util = funcs->update_util;
+       pstate_funcs.get_aperf_mperf_shift = funcs->get_aperf_mperf_shift;
 
        intel_pstate_use_acpi_profile();
 }