* of the License.
*/
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
#include <linux/kernel.h>
#include <linux/kernel_stat.h>
#include <linux/module.h>
return ret;
}
+/**
+ * struct sample - Store performance sample
+ * @core_pct_busy: Ratio of APERF/MPERF in percent, which is actual
+ * performance during last sample period
+ * @busy_scaled: Scaled busy value which is used to calculate next
+ * P state. This can be different than core_pct_busy
+ * to account for cpu idle period
+ * @aperf: Difference of actual performance frequency clock count
+ * read from APERF MSR between last and current sample
+ * @mperf: Difference of maximum performance frequency clock count
+ * read from MPERF MSR between last and current sample
+ * @tsc: Difference of time stamp counter between last and
+ * current sample
+ * @freq: Effective frequency calculated from APERF/MPERF
+ * @time: Current time from scheduler
+ *
+ * This structure is used in the cpudata structure to store performance sample
+ * data for choosing next P State.
+ */
struct sample {
int32_t core_pct_busy;
int32_t busy_scaled;
u64 time;
};
+/**
+ * struct pstate_data - Store P state data
+ * @current_pstate: Current requested P state
+ * @min_pstate: Min P state possible for this platform
+ * @max_pstate: Max P state possible for this platform
+ * @max_pstate_physical:This is physical Max P state for a processor
+ * This can be higher than the max_pstate which can
+ * be limited by platform thermal design power limits
+ * @scaling: Scaling factor to convert frequency to cpufreq
+ * frequency units
+ * @turbo_pstate: Max Turbo P state possible for this platform
+ *
+ * Stores the per cpu model P state limits and current P state.
+ */
struct pstate_data {
int current_pstate;
int min_pstate;
int turbo_pstate;
};
+/**
+ * struct vid_data - Stores voltage information data
+ * @min: VID data for this platform corresponding to
+ * the lowest P state
+ * @max: VID data corresponding to the highest P State.
+ * @turbo: VID data for turbo P state
+ * @ratio: Ratio of (vid max - vid min) /
+ * (max P state - Min P State)
+ *
+ * Stores the voltage data for DVFS (Dynamic Voltage and Frequency Scaling)
+ * This data is used in Atom platforms, where in addition to target P state,
+ * the voltage data needs to be specified to select next P State.
+ */
struct vid_data {
int min;
int max;
int32_t ratio;
};
+/**
+ * struct _pid - Stores PID data
+ * @setpoint: Target set point for busyness or performance
+ * @integral: Storage for accumulated error values
+ * @p_gain: PID proportional gain
+ * @i_gain: PID integral gain
+ * @d_gain: PID derivative gain
+ * @deadband: PID deadband
+ * @last_err: Last error storage for integral part of PID calculation
+ *
+ * Stores PID coefficients and last error for PID controller.
+ */
struct _pid {
int setpoint;
int32_t integral;
int32_t last_err;
};
+/**
+ * struct cpudata - Per CPU instance data storage
+ * @cpu: CPU number for this instance data
+ * @update_util: CPUFreq utility callback information
+ * @pstate: Stores P state limits for this CPU
+ * @vid: Stores VID limits for this CPU
+ * @pid: Stores PID parameters for this CPU
+ * @last_sample_time: Last Sample time
+ * @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
+ * @prev_cummulative_iowait: IO Wait time difference from last and
+ * current sample
+ * @sample: Storage for storing last Sample data
+ *
+ * This structure stores per CPU instance data for all CPUs.
+ */
struct cpudata {
int cpu;
};
static struct cpudata **all_cpu_data;
+
+/**
+ * struct pid_adjust_policy - Stores static PID configuration data
+ * @sample_rate_ms: PID calculation sample rate in ms
+ * @sample_rate_ns: Sample rate calculation in ns
+ * @deadband: PID deadband
+ * @setpoint: PID Setpoint
+ * @p_gain_pct: PID proportional gain
+ * @i_gain_pct: PID integral gain
+ * @d_gain_pct: PID derivative gain
+ *
+ * Stores per CPU model static PID configuration data.
+ */
struct pstate_adjust_policy {
int sample_rate_ms;
s64 sample_rate_ns;
int i_gain_pct;
};
+/**
+ * struct pstate_funcs - Per CPU model specific callbacks
+ * @get_max: Callback to get maximum non turbo effective P state
+ * @get_max_physical: Callback to get maximum non turbo physical P state
+ * @get_min: Callback to get minimum P state
+ * @get_turbo: Callback to get turbo P state
+ * @get_scaling: Callback to get frequency scaling factor
+ * @get_val: Callback to convert P state to actual MSR write value
+ * @get_vid: Callback to get VID data for Atom platforms
+ * @get_target_pstate: Callback to a function to calculate next P state to use
+ *
+ * Core and Atom CPU models have different way to get P State limits. This
+ * structure is used to store those callbacks.
+ */
struct pstate_funcs {
int (*get_max)(void);
int (*get_max_physical)(void);
int (*get_min)(void);
int (*get_turbo)(void);
int (*get_scaling)(void);
- void (*set)(struct cpudata*, int pstate);
+ u64 (*get_val)(struct cpudata*, int pstate);
void (*get_vid)(struct cpudata *);
int32_t (*get_target_pstate)(struct cpudata *);
};
+/**
+ * struct cpu_defaults- Per CPU model default config data
+ * @pid_policy: PID config data
+ * @funcs: Callback function data
+ */
struct cpu_defaults {
struct pstate_adjust_policy pid_policy;
struct pstate_funcs funcs;
static struct pstate_funcs pstate_funcs;
static int hwp_active;
+
+/**
+ * struct perf_limits - Store user and policy limits
+ * @no_turbo: User requested turbo state from intel_pstate sysfs
+ * @turbo_disabled: Platform turbo status either from msr
+ * MSR_IA32_MISC_ENABLE or when maximum available pstate
+ * matches the maximum turbo pstate
+ * @max_perf_pct: Effective maximum performance limit in percentage, this
+ * is minimum of either limits enforced by cpufreq policy
+ * or limits from user set limits via intel_pstate sysfs
+ * @min_perf_pct: Effective minimum performance limit in percentage, this
+ * is maximum of either limits enforced by cpufreq policy
+ * or limits from user set limits via intel_pstate sysfs
+ * @max_perf: This is a scaled value between 0 to 255 for max_perf_pct
+ * This value is used to limit max pstate
+ * @min_perf: This is a scaled value between 0 to 255 for min_perf_pct
+ * This value is used to limit min pstate
+ * @max_policy_pct: The maximum performance in percentage enforced by
+ * cpufreq setpolicy interface
+ * @max_sysfs_pct: The maximum performance in percentage enforced by
+ * intel pstate sysfs interface
+ * @min_policy_pct: The minimum performance in percentage enforced by
+ * cpufreq setpolicy interface
+ * @min_sysfs_pct: The minimum performance in percentage enforced by
+ * intel pstate sysfs interface
+ *
+ * Storage for user and policy defined limits.
+ */
struct perf_limits {
int no_turbo;
int turbo_disabled;
static inline void pid_p_gain_set(struct _pid *pid, int percent)
{
- pid->p_gain = div_fp(int_tofp(percent), int_tofp(100));
+ pid->p_gain = div_fp(percent, 100);
}
static inline void pid_i_gain_set(struct _pid *pid, int percent)
{
- pid->i_gain = div_fp(int_tofp(percent), int_tofp(100));
+ pid->i_gain = div_fp(percent, 100);
}
static inline void pid_d_gain_set(struct _pid *pid, int percent)
{
- pid->d_gain = div_fp(int_tofp(percent), int_tofp(100));
+ pid->d_gain = div_fp(percent, 100);
}
static signed int pid_calc(struct _pid *pid, int32_t busy)
total = cpu->pstate.turbo_pstate - cpu->pstate.min_pstate + 1;
no_turbo = cpu->pstate.max_pstate - cpu->pstate.min_pstate + 1;
- turbo_fp = div_fp(int_tofp(no_turbo), int_tofp(total));
+ turbo_fp = div_fp(no_turbo, total);
turbo_pct = 100 - fp_toint(mul_fp(turbo_fp, int_tofp(100)));
return sprintf(buf, "%u\n", turbo_pct);
}
update_turbo_state();
if (limits->turbo_disabled) {
- pr_warn("intel_pstate: Turbo disabled by BIOS or unavailable on processor\n");
+ pr_warn("Turbo disabled by BIOS or unavailable on processor\n");
return -EPERM;
}
limits->max_perf_pct);
limits->max_perf_pct = max(limits->min_perf_pct,
limits->max_perf_pct);
- limits->max_perf = div_fp(int_tofp(limits->max_perf_pct),
- int_tofp(100));
+ limits->max_perf = div_fp(limits->max_perf_pct, 100);
if (hwp_active)
intel_pstate_hwp_set_online_cpus();
limits->min_perf_pct);
limits->min_perf_pct = min(limits->max_perf_pct,
limits->min_perf_pct);
- limits->min_perf = div_fp(int_tofp(limits->min_perf_pct),
- int_tofp(100));
+ limits->min_perf = div_fp(limits->min_perf_pct, 100);
if (hwp_active)
intel_pstate_hwp_set_online_cpus();
return value & 0x7F;
}
-static void atom_set_pstate(struct cpudata *cpudata, int pstate)
+static u64 atom_get_val(struct cpudata *cpudata, int pstate)
{
u64 val;
int32_t vid_fp;
if (pstate > cpudata->pstate.max_pstate)
vid = cpudata->vid.turbo;
- val |= vid;
-
- wrmsrl_on_cpu(cpudata->cpu, MSR_IA32_PERF_CTL, val);
+ return val | vid;
}
static int silvermont_get_scaling(void)
return 100000;
}
-static void core_set_pstate(struct cpudata *cpudata, int pstate)
+static u64 core_get_val(struct cpudata *cpudata, int pstate)
{
u64 val;
if (limits->no_turbo && !limits->turbo_disabled)
val |= (u64)1 << 32;
- wrmsrl(MSR_IA32_PERF_CTL, val);
+ return val;
}
static int knl_get_turbo_pstate(void)
.get_min = core_get_min_pstate,
.get_turbo = core_get_turbo_pstate,
.get_scaling = core_get_scaling,
- .set = core_set_pstate,
+ .get_val = core_get_val,
.get_target_pstate = get_target_pstate_use_performance,
},
};
.get_max_physical = atom_get_max_pstate,
.get_min = atom_get_min_pstate,
.get_turbo = atom_get_turbo_pstate,
- .set = atom_set_pstate,
+ .get_val = atom_get_val,
.get_scaling = silvermont_get_scaling,
.get_vid = atom_get_vid,
.get_target_pstate = get_target_pstate_use_cpu_load,
.get_max_physical = atom_get_max_pstate,
.get_min = atom_get_min_pstate,
.get_turbo = atom_get_turbo_pstate,
- .set = atom_set_pstate,
+ .get_val = atom_get_val,
.get_scaling = airmont_get_scaling,
.get_vid = atom_get_vid,
.get_target_pstate = get_target_pstate_use_cpu_load,
.get_min = core_get_min_pstate,
.get_turbo = knl_get_turbo_pstate,
.get_scaling = core_get_scaling,
- .set = core_set_pstate,
+ .get_val = core_get_val,
.get_target_pstate = get_target_pstate_use_performance,
},
};
*min = clamp_t(int, min_perf, cpu->pstate.min_pstate, max_perf);
}
-static void intel_pstate_set_pstate(struct cpudata *cpu, int pstate, bool force)
+static inline void intel_pstate_record_pstate(struct cpudata *cpu, int pstate)
{
- int max_perf, min_perf;
-
- if (force) {
- update_turbo_state();
-
- intel_pstate_get_min_max(cpu, &min_perf, &max_perf);
-
- pstate = clamp_t(int, pstate, min_perf, max_perf);
-
- if (pstate == cpu->pstate.current_pstate)
- return;
- }
trace_cpu_frequency(pstate * cpu->pstate.scaling, cpu->cpu);
-
cpu->pstate.current_pstate = pstate;
+}
+
+static void intel_pstate_set_min_pstate(struct cpudata *cpu)
+{
+ int pstate = cpu->pstate.min_pstate;
- pstate_funcs.set(cpu, pstate);
+ intel_pstate_record_pstate(cpu, pstate);
+ /*
+ * Generally, there is no guarantee that this code will always run on
+ * the CPU being updated, so force the register update to run on the
+ * right CPU.
+ */
+ wrmsrl_on_cpu(cpu->cpu, MSR_IA32_PERF_CTL,
+ pstate_funcs.get_val(cpu, pstate));
}
static void intel_pstate_get_cpu_pstates(struct cpudata *cpu)
if (pstate_funcs.get_vid)
pstate_funcs.get_vid(cpu);
- intel_pstate_set_pstate(cpu, cpu->pstate.min_pstate, false);
+
+ intel_pstate_set_min_pstate(cpu);
}
static inline void intel_pstate_calc_busy(struct cpudata *cpu)
struct sample *sample = &cpu->sample;
int64_t core_pct;
- core_pct = int_tofp(sample->aperf) * int_tofp(100);
- core_pct = div64_u64(core_pct, int_tofp(sample->mperf));
+ core_pct = sample->aperf * int_tofp(100);
+ core_pct = div64_u64(core_pct, sample->mperf);
sample->core_pct_busy = (int32_t)core_pct;
}
cpu->prev_aperf = aperf;
cpu->prev_mperf = mperf;
cpu->prev_tsc = tsc;
- return true;
+ /*
+ * First time this function is invoked in a given cycle, all of the
+ * previous sample data fields are equal to zero or stale and they must
+ * be populated with meaningful numbers for things to work, so assume
+ * that sample.time will always be reset before setting the utilization
+ * update hook and make the caller skip the sample then.
+ */
+ return !!cpu->last_sample_time;
}
static inline int32_t get_avg_frequency(struct cpudata *cpu)
* specified pstate.
*/
core_busy = cpu->sample.core_pct_busy;
- max_pstate = int_tofp(cpu->pstate.max_pstate_physical);
- current_pstate = int_tofp(cpu->pstate.current_pstate);
+ max_pstate = cpu->pstate.max_pstate_physical;
+ current_pstate = cpu->pstate.current_pstate;
core_busy = mul_fp(core_busy, div_fp(max_pstate, current_pstate));
/*
* enough period of time to adjust our busyness.
*/
duration_ns = cpu->sample.time - cpu->last_sample_time;
- if ((s64)duration_ns > pid_params.sample_rate_ns * 3
- && cpu->last_sample_time > 0) {
- sample_ratio = div_fp(int_tofp(pid_params.sample_rate_ns),
- int_tofp(duration_ns));
+ if ((s64)duration_ns > pid_params.sample_rate_ns * 3) {
+ sample_ratio = div_fp(pid_params.sample_rate_ns, duration_ns);
core_busy = mul_fp(core_busy, sample_ratio);
}
return cpu->pstate.current_pstate - pid_calc(&cpu->pid, core_busy);
}
+static inline void intel_pstate_update_pstate(struct cpudata *cpu, int pstate)
+{
+ int max_perf, min_perf;
+
+ update_turbo_state();
+
+ intel_pstate_get_min_max(cpu, &min_perf, &max_perf);
+ pstate = clamp_t(int, pstate, min_perf, max_perf);
+ if (pstate == cpu->pstate.current_pstate)
+ return;
+
+ intel_pstate_record_pstate(cpu, pstate);
+ wrmsrl(MSR_IA32_PERF_CTL, pstate_funcs.get_val(cpu, pstate));
+}
+
static inline void intel_pstate_adjust_busy_pstate(struct cpudata *cpu)
{
int from, target_pstate;
target_pstate = pstate_funcs.get_target_pstate(cpu);
- intel_pstate_set_pstate(cpu, target_pstate, true);
+ intel_pstate_update_pstate(cpu, target_pstate);
sample = &cpu->sample;
trace_pstate_sample(fp_toint(sample->core_pct_busy),
intel_pstate_get_cpu_pstates(cpu);
intel_pstate_busy_pid_reset(cpu);
- intel_pstate_sample(cpu, 0);
- cpu->update_util.func = intel_pstate_update_util;
- cpufreq_set_update_util_data(cpunum, &cpu->update_util);
-
- pr_debug("intel_pstate: controlling: cpu %d\n", cpunum);
+ pr_debug("controlling: cpu %d\n", cpunum);
return 0;
}
return get_avg_frequency(cpu);
}
+static void intel_pstate_set_update_util_hook(unsigned int cpu_num)
+{
+ struct cpudata *cpu = all_cpu_data[cpu_num];
+
+ /* Prevent intel_pstate_update_util() from using stale data. */
+ cpu->sample.time = 0;
+ cpufreq_add_update_util_hook(cpu_num, &cpu->update_util,
+ intel_pstate_update_util);
+}
+
+static void intel_pstate_clear_update_util_hook(unsigned int cpu)
+{
+ cpufreq_remove_update_util_hook(cpu);
+ synchronize_sched();
+}
+
+static void intel_pstate_set_performance_limits(struct perf_limits *limits)
+{
+ limits->no_turbo = 0;
+ limits->turbo_disabled = 0;
+ limits->max_perf_pct = 100;
+ limits->max_perf = int_tofp(1);
+ limits->min_perf_pct = 100;
+ limits->min_perf = int_tofp(1);
+ limits->max_policy_pct = 100;
+ limits->max_sysfs_pct = 100;
+ limits->min_policy_pct = 0;
+ limits->min_sysfs_pct = 0;
+}
+
static int intel_pstate_set_policy(struct cpufreq_policy *policy)
{
if (!policy->cpuinfo.max_freq)
return -ENODEV;
- if (policy->policy == CPUFREQ_POLICY_PERFORMANCE &&
- policy->max >= policy->cpuinfo.max_freq) {
- pr_debug("intel_pstate: set performance\n");
+ intel_pstate_clear_update_util_hook(policy->cpu);
+
+ if (policy->policy == CPUFREQ_POLICY_PERFORMANCE) {
limits = &performance_limits;
- if (hwp_active)
- intel_pstate_hwp_set(policy->cpus);
- return 0;
+ if (policy->max >= policy->cpuinfo.max_freq) {
+ pr_debug("set performance\n");
+ intel_pstate_set_performance_limits(limits);
+ goto out;
+ }
+ } else {
+ pr_debug("set powersave\n");
+ limits = &powersave_limits;
}
- pr_debug("intel_pstate: set powersave\n");
- limits = &powersave_limits;
limits->min_policy_pct = (policy->min * 100) / policy->cpuinfo.max_freq;
limits->min_policy_pct = clamp_t(int, limits->min_policy_pct, 0 , 100);
limits->max_policy_pct = DIV_ROUND_UP(policy->max * 100,
/* Make sure min_perf_pct <= max_perf_pct */
limits->min_perf_pct = min(limits->max_perf_pct, limits->min_perf_pct);
- limits->min_perf = div_fp(int_tofp(limits->min_perf_pct),
- int_tofp(100));
- limits->max_perf = div_fp(int_tofp(limits->max_perf_pct),
- int_tofp(100));
+ limits->min_perf = div_fp(limits->min_perf_pct, 100);
+ limits->max_perf = div_fp(limits->max_perf_pct, 100);
+
+ out:
+ intel_pstate_set_update_util_hook(policy->cpu);
if (hwp_active)
intel_pstate_hwp_set(policy->cpus);
int cpu_num = policy->cpu;
struct cpudata *cpu = all_cpu_data[cpu_num];
- pr_debug("intel_pstate: CPU %d exiting\n", cpu_num);
+ pr_debug("CPU %d exiting\n", cpu_num);
- cpufreq_set_update_util_data(cpu_num, NULL);
- synchronize_sched();
+ intel_pstate_clear_update_util_hook(cpu_num);
if (hwp_active)
return;
- intel_pstate_set_pstate(cpu, cpu->pstate.min_pstate, false);
+ intel_pstate_set_min_pstate(cpu);
}
static int intel_pstate_cpu_init(struct cpufreq_policy *policy)
pstate_funcs.get_min = funcs->get_min;
pstate_funcs.get_turbo = funcs->get_turbo;
pstate_funcs.get_scaling = funcs->get_scaling;
- pstate_funcs.set = funcs->set;
+ pstate_funcs.get_val = funcs->get_val;
pstate_funcs.get_vid = funcs->get_vid;
pstate_funcs.get_target_pstate = funcs->get_target_pstate;
if (intel_pstate_platform_pwr_mgmt_exists())
return -ENODEV;
- pr_info("Intel P-state driver initializing.\n");
+ pr_info("Intel P-state driver initializing\n");
all_cpu_data = vzalloc(sizeof(void *) * num_possible_cpus());
if (!all_cpu_data)
intel_pstate_sysfs_expose_params();
if (hwp_active)
- pr_info("intel_pstate: HWP enabled\n");
+ pr_info("HWP enabled\n");
return rc;
out:
get_online_cpus();
for_each_online_cpu(cpu) {
if (all_cpu_data[cpu]) {
- cpufreq_set_update_util_data(cpu, NULL);
- synchronize_sched();
+ intel_pstate_clear_update_util_hook(cpu);
kfree(all_cpu_data[cpu]);
}
}
if (!strcmp(str, "disable"))
no_load = 1;
if (!strcmp(str, "no_hwp")) {
- pr_info("intel_pstate: HWP disabled\n");
+ pr_info("HWP disabled\n");
no_hwp = 1;
}
if (!strcmp(str, "force"))