cpufreq: intel_pstate: Use ACPI perf configuration
authorSrinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Wed, 14 Oct 2015 23:12:01 +0000 (16:12 -0700)
committerRafael J. Wysocki <rafael.j.wysocki@intel.com>
Wed, 14 Oct 2015 23:53:18 +0000 (01:53 +0200)
Use ACPI _PSS to limit the Intel P State turbo, max and min ratios.
This driver uses acpi processor perf lib calls to register performance.
The following logic is used to adjust Intel P state driver limits:
- If there is no turbo entry in _PSS, then disable Intel P state turbo
and limit to non turbo max
- If the non turbo max ratio is more than _PSS max non turbo value, then
set the max non turbo ratio to _PSS non turbo max
- If the min ratio is less than _PSS min then change the min ratio
matching _PSS min
- Scale the _PSS turbo frequency to max turbo frequency based on control
value.
This feature can be disabled by using kernel parameters:
intel_pstate=no_acpi

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Acked-by: Kristen Carlson Accardi <kristen@linux.intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
drivers/cpufreq/Kconfig.x86
drivers/cpufreq/intel_pstate.c

index c59bdcb..adbd1de 100644 (file)
@@ -5,6 +5,7 @@
 config X86_INTEL_PSTATE
        bool "Intel P state control"
        depends on X86
+       select ACPI_PROCESSOR if ACPI
        help
           This driver provides a P state for Intel core processors.
          The driver implements an internal governor and will become
index 1369afd..041cb41 100644 (file)
 #include <asm/cpu_device_id.h>
 #include <asm/cpufeature.h>
 
+#if IS_ENABLED(CONFIG_ACPI)
+#include <acpi/processor.h>
+#endif
+
 #define BYT_RATIOS             0x66a
 #define BYT_VIDS               0x66b
 #define BYT_TURBO_RATIOS       0x66c
@@ -113,6 +117,9 @@ struct cpudata {
        u64     prev_mperf;
        u64     prev_tsc;
        struct sample sample;
+#if IS_ENABLED(CONFIG_ACPI)
+       struct acpi_processor_performance acpi_perf_data;
+#endif
 };
 
 static struct cpudata **all_cpu_data;
@@ -143,6 +150,7 @@ struct cpu_defaults {
 static struct pstate_adjust_policy pid_params;
 static struct pstate_funcs pstate_funcs;
 static int hwp_active;
+static int no_acpi_perf;
 
 struct perf_limits {
        int no_turbo;
@@ -170,6 +178,153 @@ static struct perf_limits limits = {
        .min_sysfs_pct = 0,
 };
 
+#if IS_ENABLED(CONFIG_ACPI)
+/*
+ * The max target pstate ratio is a 8 bit value in both PLATFORM_INFO MSR and
+ * in TURBO_RATIO_LIMIT MSR, which pstate driver stores in max_pstate and
+ * max_turbo_pstate fields. The PERF_CTL MSR contains 16 bit value for P state
+ * ratio, out of it only high 8 bits are used. For example 0x1700 is setting
+ * target ratio 0x17. The _PSS control value stores in a format which can be
+ * directly written to PERF_CTL MSR. But in intel_pstate driver this shift
+ * occurs during write to PERF_CTL (E.g. for cores core_set_pstate()).
+ * This function converts the _PSS control value to intel pstate driver format
+ * for comparison and assignment.
+ */
+static int convert_to_native_pstate_format(struct cpudata *cpu, int index)
+{
+       return cpu->acpi_perf_data.states[index].control >> 8;
+}
+
+static int intel_pstate_init_perf_limits(struct cpufreq_policy *policy)
+{
+       struct cpudata *cpu;
+       int ret;
+       bool turbo_absent = false;
+       int max_pstate_index;
+       int min_pss_ctl, max_pss_ctl, turbo_pss_ctl;
+       int i;
+
+       cpu = all_cpu_data[policy->cpu];
+
+       pr_debug("intel_pstate: default limits 0x%x 0x%x 0x%x\n",
+                cpu->pstate.min_pstate, cpu->pstate.max_pstate,
+                cpu->pstate.turbo_pstate);
+
+       if (!cpu->acpi_perf_data.shared_cpu_map &&
+           zalloc_cpumask_var_node(&cpu->acpi_perf_data.shared_cpu_map,
+                                   GFP_KERNEL, cpu_to_node(policy->cpu))) {
+               return -ENOMEM;
+       }
+
+       ret = acpi_processor_register_performance(&cpu->acpi_perf_data,
+                                                 policy->cpu);
+       if (ret)
+               return ret;
+
+       /*
+        * Check if the control value in _PSS is for PERF_CTL MSR, which should
+        * guarantee that the states returned by it map to the states in our
+        * list directly.
+        */
+       if (cpu->acpi_perf_data.control_register.space_id !=
+                                               ACPI_ADR_SPACE_FIXED_HARDWARE)
+               return -EIO;
+
+       pr_debug("intel_pstate: CPU%u - ACPI _PSS perf data\n", policy->cpu);
+       for (i = 0; i < cpu->acpi_perf_data.state_count; i++)
+               pr_debug("     %cP%d: %u MHz, %u mW, 0x%x\n",
+                        (i == cpu->acpi_perf_data.state ? '*' : ' '), i,
+                        (u32) cpu->acpi_perf_data.states[i].core_frequency,
+                        (u32) cpu->acpi_perf_data.states[i].power,
+                        (u32) cpu->acpi_perf_data.states[i].control);
+
+       /*
+        * If there is only one entry _PSS, simply ignore _PSS and continue as
+        * usual without taking _PSS into account
+        */
+       if (cpu->acpi_perf_data.state_count < 2)
+               return 0;
+
+       turbo_pss_ctl = convert_to_native_pstate_format(cpu, 0);
+       min_pss_ctl = convert_to_native_pstate_format(cpu,
+                                       cpu->acpi_perf_data.state_count - 1);
+       /* Check if there is a turbo freq in _PSS */
+       if (turbo_pss_ctl <= cpu->pstate.max_pstate &&
+           turbo_pss_ctl > cpu->pstate.min_pstate) {
+               pr_debug("intel_pstate: no turbo range exists in _PSS\n");
+               limits.no_turbo = limits.turbo_disabled = 1;
+               cpu->pstate.turbo_pstate = cpu->pstate.max_pstate;
+               turbo_absent = true;
+       }
+
+       /* Check if the max non turbo p state < Intel P state max */
+       max_pstate_index = turbo_absent ? 0 : 1;
+       max_pss_ctl = convert_to_native_pstate_format(cpu, max_pstate_index);
+       if (max_pss_ctl < cpu->pstate.max_pstate &&
+           max_pss_ctl > cpu->pstate.min_pstate)
+               cpu->pstate.max_pstate = max_pss_ctl;
+
+       /* check If min perf > Intel P State min */
+       if (min_pss_ctl > cpu->pstate.min_pstate &&
+           min_pss_ctl < cpu->pstate.max_pstate) {
+               cpu->pstate.min_pstate = min_pss_ctl;
+               policy->cpuinfo.min_freq = min_pss_ctl * cpu->pstate.scaling;
+       }
+
+       if (turbo_absent)
+               policy->cpuinfo.max_freq = cpu->pstate.max_pstate *
+                                               cpu->pstate.scaling;
+       else {
+               policy->cpuinfo.max_freq = cpu->pstate.turbo_pstate *
+                                               cpu->pstate.scaling;
+               /*
+                * The _PSS table doesn't contain whole turbo frequency range.
+                * This just contains +1 MHZ above the max non turbo frequency,
+                * with control value corresponding to max turbo ratio. But
+                * when cpufreq set policy is called, it will call with this
+                * max frequency, which will cause a reduced performance as
+                * this driver uses real max turbo frequency as the max
+                * frequeny. So correct this frequency in _PSS table to
+                * correct max turbo frequency based on the turbo ratio.
+                * Also need to convert to MHz as _PSS freq is in MHz.
+                */
+               cpu->acpi_perf_data.states[0].core_frequency =
+                                               turbo_pss_ctl * 100;
+       }
+
+       pr_debug("intel_pstate: Updated limits using _PSS 0x%x 0x%x 0x%x\n",
+                cpu->pstate.min_pstate, cpu->pstate.max_pstate,
+                cpu->pstate.turbo_pstate);
+       pr_debug("intel_pstate: policy max_freq=%d Khz min_freq = %d KHz\n",
+                policy->cpuinfo.max_freq, policy->cpuinfo.min_freq);
+
+       return 0;
+}
+
+static int intel_pstate_exit_perf_limits(struct cpufreq_policy *policy)
+{
+       struct cpudata *cpu;
+
+       if (!no_acpi_perf)
+               return 0;
+
+       cpu = all_cpu_data[policy->cpu];
+       acpi_processor_unregister_performance(policy->cpu);
+       return 0;
+}
+
+#else
+static int intel_pstate_init_perf_limits(struct cpufreq_policy *policy)
+{
+       return 0;
+}
+
+static int intel_pstate_exit_perf_limits(struct cpufreq_policy *policy)
+{
+       return 0;
+}
+#endif
+
 static inline void pid_reset(struct _pid *pid, int setpoint, int busy,
                             int deadband, int integral) {
        pid->setpoint = setpoint;
@@ -1115,18 +1270,30 @@ static int intel_pstate_cpu_init(struct cpufreq_policy *policy)
        policy->cpuinfo.min_freq = cpu->pstate.min_pstate * cpu->pstate.scaling;
        policy->cpuinfo.max_freq =
                cpu->pstate.turbo_pstate * cpu->pstate.scaling;
+       if (!no_acpi_perf)
+               intel_pstate_init_perf_limits(policy);
+       /*
+        * If there is no acpi perf data or error, we ignore and use Intel P
+        * state calculated limits, So this is not fatal error.
+        */
        policy->cpuinfo.transition_latency = CPUFREQ_ETERNAL;
        cpumask_set_cpu(policy->cpu, policy->cpus);
 
        return 0;
 }
 
+static int intel_pstate_cpu_exit(struct cpufreq_policy *policy)
+{
+       return intel_pstate_exit_perf_limits(policy);
+}
+
 static struct cpufreq_driver intel_pstate_driver = {
        .flags          = CPUFREQ_CONST_LOOPS,
        .verify         = intel_pstate_verify_policy,
        .setpolicy      = intel_pstate_set_policy,
        .get            = intel_pstate_get,
        .init           = intel_pstate_cpu_init,
+       .exit           = intel_pstate_cpu_exit,
        .stop_cpu       = intel_pstate_stop_cpu,
        .name           = "intel_pstate",
 };
@@ -1168,7 +1335,6 @@ static void copy_cpu_funcs(struct pstate_funcs *funcs)
 }
 
 #if IS_ENABLED(CONFIG_ACPI)
-#include <acpi/processor.h>
 
 static bool intel_pstate_no_acpi_pss(void)
 {
@@ -1360,6 +1526,9 @@ static int __init intel_pstate_setup(char *str)
                force_load = 1;
        if (!strcmp(str, "hwp_only"))
                hwp_only = 1;
+       if (!strcmp(str, "no_acpi"))
+               no_acpi_perf = 1;
+
        return 0;
 }
 early_param("intel_pstate", intel_pstate_setup);