Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
[linux-drm-fsl-dcu.git] / drivers / cpufreq / intel_pstate.c
index eb3fdc755000e0005f43474bb5ae967dd6ab8aa8..f9d561e198ab546249df3facdeced4e6f7477949 100644 (file)
@@ -25,6 +25,7 @@
 #include <linux/types.h>
 #include <linux/fs.h>
 #include <linux/debugfs.h>
+#include <linux/acpi.h>
 #include <trace/events/power.h>
 
 #include <asm/div64.h>
@@ -33,6 +34,8 @@
 
 #define SAMPLE_COUNT           3
 
+#define BYT_RATIOS     0x66a
+
 #define FRAC_BITS 8
 #define int_tofp(X) ((int64_t)(X) << FRAC_BITS)
 #define fp_toint(X) ((X) >> FRAC_BITS)
@@ -78,7 +81,6 @@ struct cpudata {
 
        struct timer_list timer;
 
-       struct pstate_adjust_policy *pstate_policy;
        struct pstate_data pstate;
        struct _pid pid;
 
@@ -100,15 +102,21 @@ struct pstate_adjust_policy {
        int i_gain_pct;
 };
 
-static struct pstate_adjust_policy default_policy = {
-       .sample_rate_ms = 10,
-       .deadband = 0,
-       .setpoint = 97,
-       .p_gain_pct = 20,
-       .d_gain_pct = 0,
-       .i_gain_pct = 0,
+struct pstate_funcs {
+       int (*get_max)(void);
+       int (*get_min)(void);
+       int (*get_turbo)(void);
+       void (*set)(int pstate);
+};
+
+struct cpu_defaults {
+       struct pstate_adjust_policy pid_policy;
+       struct pstate_funcs funcs;
 };
 
+static struct pstate_adjust_policy pid_params;
+static struct pstate_funcs pstate_funcs;
+
 struct perf_limits {
        int no_turbo;
        int max_perf_pct;
@@ -185,14 +193,14 @@ static signed int pid_calc(struct _pid *pid, int32_t busy)
 
 static inline void intel_pstate_busy_pid_reset(struct cpudata *cpu)
 {
-       pid_p_gain_set(&cpu->pid, cpu->pstate_policy->p_gain_pct);
-       pid_d_gain_set(&cpu->pid, cpu->pstate_policy->d_gain_pct);
-       pid_i_gain_set(&cpu->pid, cpu->pstate_policy->i_gain_pct);
+       pid_p_gain_set(&cpu->pid, pid_params.p_gain_pct);
+       pid_d_gain_set(&cpu->pid, pid_params.d_gain_pct);
+       pid_i_gain_set(&cpu->pid, pid_params.i_gain_pct);
 
        pid_reset(&cpu->pid,
-               cpu->pstate_policy->setpoint,
+               pid_params.setpoint,
                100,
-               cpu->pstate_policy->deadband,
+               pid_params.deadband,
                0);
 }
 
@@ -226,12 +234,12 @@ struct pid_param {
 };
 
 static struct pid_param pid_files[] = {
-       {"sample_rate_ms", &default_policy.sample_rate_ms},
-       {"d_gain_pct", &default_policy.d_gain_pct},
-       {"i_gain_pct", &default_policy.i_gain_pct},
-       {"deadband", &default_policy.deadband},
-       {"setpoint", &default_policy.setpoint},
-       {"p_gain_pct", &default_policy.p_gain_pct},
+       {"sample_rate_ms", &pid_params.sample_rate_ms},
+       {"d_gain_pct", &pid_params.d_gain_pct},
+       {"i_gain_pct", &pid_params.i_gain_pct},
+       {"deadband", &pid_params.deadband},
+       {"setpoint", &pid_params.setpoint},
+       {"p_gain_pct", &pid_params.p_gain_pct},
        {NULL, NULL}
 };
 
@@ -336,33 +344,92 @@ static void intel_pstate_sysfs_expose_params(void)
 }
 
 /************************** sysfs end ************************/
+static int byt_get_min_pstate(void)
+{
+       u64 value;
+       rdmsrl(BYT_RATIOS, value);
+       return value & 0xFF;
+}
+
+static int byt_get_max_pstate(void)
+{
+       u64 value;
+       rdmsrl(BYT_RATIOS, value);
+       return (value >> 16) & 0xFF;
+}
 
-static int intel_pstate_min_pstate(void)
+static int core_get_min_pstate(void)
 {
        u64 value;
        rdmsrl(MSR_PLATFORM_INFO, value);
        return (value >> 40) & 0xFF;
 }
 
-static int intel_pstate_max_pstate(void)
+static int core_get_max_pstate(void)
 {
        u64 value;
        rdmsrl(MSR_PLATFORM_INFO, value);
        return (value >> 8) & 0xFF;
 }
 
-static int intel_pstate_turbo_pstate(void)
+static int core_get_turbo_pstate(void)
 {
        u64 value;
        int nont, ret;
        rdmsrl(MSR_NHM_TURBO_RATIO_LIMIT, value);
-       nont = intel_pstate_max_pstate();
+       nont = core_get_max_pstate();
        ret = ((value) & 255);
        if (ret <= nont)
                ret = nont;
        return ret;
 }
 
+static void core_set_pstate(int pstate)
+{
+       u64 val;
+
+       val = pstate << 8;
+       if (limits.no_turbo)
+               val |= (u64)1 << 32;
+
+       wrmsrl(MSR_IA32_PERF_CTL, val);
+}
+
+static struct cpu_defaults core_params = {
+       .pid_policy = {
+               .sample_rate_ms = 10,
+               .deadband = 0,
+               .setpoint = 97,
+               .p_gain_pct = 20,
+               .d_gain_pct = 0,
+               .i_gain_pct = 0,
+       },
+       .funcs = {
+               .get_max = core_get_max_pstate,
+               .get_min = core_get_min_pstate,
+               .get_turbo = core_get_turbo_pstate,
+               .set = core_set_pstate,
+       },
+};
+
+static struct cpu_defaults byt_params = {
+       .pid_policy = {
+               .sample_rate_ms = 10,
+               .deadband = 0,
+               .setpoint = 97,
+               .p_gain_pct = 14,
+               .d_gain_pct = 0,
+               .i_gain_pct = 4,
+       },
+       .funcs = {
+               .get_max = byt_get_max_pstate,
+               .get_min = byt_get_min_pstate,
+               .get_turbo = byt_get_max_pstate,
+               .set = core_set_pstate,
+       },
+};
+
+
 static void intel_pstate_get_min_max(struct cpudata *cpu, int *min, int *max)
 {
        int max_perf = cpu->pstate.turbo_pstate;
@@ -383,7 +450,6 @@ static void intel_pstate_get_min_max(struct cpudata *cpu, int *min, int *max)
 static void intel_pstate_set_pstate(struct cpudata *cpu, int pstate)
 {
        int max_perf, min_perf;
-       u64 val;
 
        intel_pstate_get_min_max(cpu, &min_perf, &max_perf);
 
@@ -395,11 +461,8 @@ static void intel_pstate_set_pstate(struct cpudata *cpu, int pstate)
        trace_cpu_frequency(pstate * 100000, cpu->cpu);
 
        cpu->pstate.current_pstate = pstate;
-       val = pstate << 8;
-       if (limits.no_turbo)
-               val |= (u64)1 << 32;
 
-       wrmsrl(MSR_IA32_PERF_CTL, val);
+       pstate_funcs.set(pstate);
 }
 
 static inline void intel_pstate_pstate_increase(struct cpudata *cpu, int steps)
@@ -421,9 +484,9 @@ static void intel_pstate_get_cpu_pstates(struct cpudata *cpu)
 {
        sprintf(cpu->name, "Intel 2nd generation core");
 
-       cpu->pstate.min_pstate = intel_pstate_min_pstate();
-       cpu->pstate.max_pstate = intel_pstate_max_pstate();
-       cpu->pstate.turbo_pstate = intel_pstate_turbo_pstate();
+       cpu->pstate.min_pstate = pstate_funcs.get_min();
+       cpu->pstate.max_pstate = pstate_funcs.get_max();
+       cpu->pstate.turbo_pstate = pstate_funcs.get_turbo();
 
        /*
         * goto max pstate so we don't slow up boot if we are built-in if we are
@@ -465,7 +528,7 @@ static inline void intel_pstate_set_sample_time(struct cpudata *cpu)
 {
        int sample_time, delay;
 
-       sample_time = cpu->pstate_policy->sample_rate_ms;
+       sample_time = pid_params.sample_rate_ms;
        delay = msecs_to_jiffies(sample_time);
        mod_timer_pinned(&cpu->timer, jiffies + delay);
 }
@@ -521,14 +584,15 @@ static void intel_pstate_timer_func(unsigned long __data)
        { X86_VENDOR_INTEL, 6, model, X86_FEATURE_ANY, (unsigned long)&policy }
 
 static const struct x86_cpu_id intel_pstate_cpu_ids[] = {
-       ICPU(0x2a, default_policy),
-       ICPU(0x2d, default_policy),
-       ICPU(0x3a, default_policy),
-       ICPU(0x3c, default_policy),
-       ICPU(0x3e, default_policy),
-       ICPU(0x3f, default_policy),
-       ICPU(0x45, default_policy),
-       ICPU(0x46, default_policy),
+       ICPU(0x2a, core_params),
+       ICPU(0x2d, core_params),
+       ICPU(0x37, byt_params),
+       ICPU(0x3a, core_params),
+       ICPU(0x3c, core_params),
+       ICPU(0x3e, core_params),
+       ICPU(0x3f, core_params),
+       ICPU(0x45, core_params),
+       ICPU(0x46, core_params),
        {}
 };
 MODULE_DEVICE_TABLE(x86cpu, intel_pstate_cpu_ids);
@@ -550,10 +614,14 @@ static int intel_pstate_init_cpu(unsigned int cpunum)
        cpu = all_cpu_data[cpunum];
 
        intel_pstate_get_cpu_pstates(cpu);
+       if (!cpu->pstate.current_pstate) {
+               all_cpu_data[cpunum] = NULL;
+               kfree(cpu);
+               return -ENODATA;
+       }
 
        cpu->cpu = cpunum;
-       cpu->pstate_policy =
-               (struct pstate_adjust_policy *)id->driver_data;
+
        init_timer_deferrable(&cpu->timer);
        cpu->timer.function = intel_pstate_timer_func;
        cpu->timer.data =
@@ -613,9 +681,7 @@ static int intel_pstate_set_policy(struct cpufreq_policy *policy)
 
 static int intel_pstate_verify_policy(struct cpufreq_policy *policy)
 {
-       cpufreq_verify_within_limits(policy,
-                               policy->cpuinfo.min_freq,
-                               policy->cpuinfo.max_freq);
+       cpufreq_verify_within_cpu_limits(policy);
 
        if ((policy->policy != CPUFREQ_POLICY_POWERSAVE) &&
                (policy->policy != CPUFREQ_POLICY_PERFORMANCE))
@@ -683,9 +749,9 @@ static int intel_pstate_msrs_not_valid(void)
        rdmsrl(MSR_IA32_APERF, aperf);
        rdmsrl(MSR_IA32_MPERF, mperf);
 
-       if (!intel_pstate_min_pstate() ||
-               !intel_pstate_max_pstate() ||
-               !intel_pstate_turbo_pstate())
+       if (!pstate_funcs.get_max() ||
+               !pstate_funcs.get_min() ||
+               !pstate_funcs.get_turbo())
                return -ENODEV;
 
        rdmsrl(MSR_IA32_APERF, tmp);
@@ -698,10 +764,96 @@ static int intel_pstate_msrs_not_valid(void)
 
        return 0;
 }
+
+static void copy_pid_params(struct pstate_adjust_policy *policy)
+{
+       pid_params.sample_rate_ms = policy->sample_rate_ms;
+       pid_params.p_gain_pct = policy->p_gain_pct;
+       pid_params.i_gain_pct = policy->i_gain_pct;
+       pid_params.d_gain_pct = policy->d_gain_pct;
+       pid_params.deadband = policy->deadband;
+       pid_params.setpoint = policy->setpoint;
+}
+
+static void copy_cpu_funcs(struct pstate_funcs *funcs)
+{
+       pstate_funcs.get_max   = funcs->get_max;
+       pstate_funcs.get_min   = funcs->get_min;
+       pstate_funcs.get_turbo = funcs->get_turbo;
+       pstate_funcs.set       = funcs->set;
+}
+
+#if IS_ENABLED(CONFIG_ACPI)
+#include <acpi/processor.h>
+
+static bool intel_pstate_no_acpi_pss(void)
+{
+       int i;
+
+       for_each_possible_cpu(i) {
+               acpi_status status;
+               union acpi_object *pss;
+               struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL };
+               struct acpi_processor *pr = per_cpu(processors, i);
+
+               if (!pr)
+                       continue;
+
+               status = acpi_evaluate_object(pr->handle, "_PSS", NULL, &buffer);
+               if (ACPI_FAILURE(status))
+                       continue;
+
+               pss = buffer.pointer;
+               if (pss && pss->type == ACPI_TYPE_PACKAGE) {
+                       kfree(pss);
+                       return false;
+               }
+
+               kfree(pss);
+       }
+
+       return true;
+}
+
+struct hw_vendor_info {
+       u16  valid;
+       char oem_id[ACPI_OEM_ID_SIZE];
+       char oem_table_id[ACPI_OEM_TABLE_ID_SIZE];
+};
+
+/* Hardware vendor-specific info that has its own power management modes */
+static struct hw_vendor_info vendor_info[] = {
+       {1, "HP    ", "ProLiant"},
+       {0, "", ""},
+};
+
+static bool intel_pstate_platform_pwr_mgmt_exists(void)
+{
+       struct acpi_table_header hdr;
+       struct hw_vendor_info *v_info;
+
+       if (acpi_disabled
+           || ACPI_FAILURE(acpi_get_table_header(ACPI_SIG_FADT, 0, &hdr)))
+               return false;
+
+       for (v_info = vendor_info; v_info->valid; v_info++) {
+               if (!strncmp(hdr.oem_id, v_info->oem_id, ACPI_OEM_ID_SIZE)
+                   && !strncmp(hdr.oem_table_id, v_info->oem_table_id, ACPI_OEM_TABLE_ID_SIZE)
+                   && intel_pstate_no_acpi_pss())
+                       return true;
+       }
+
+       return false;
+}
+#else /* CONFIG_ACPI not enabled */
+static inline bool intel_pstate_platform_pwr_mgmt_exists(void) { return false; }
+#endif /* CONFIG_ACPI */
+
 static int __init intel_pstate_init(void)
 {
        int cpu, rc = 0;
        const struct x86_cpu_id *id;
+       struct cpu_defaults *cpu_info;
 
        if (no_load)
                return -ENODEV;
@@ -710,6 +862,18 @@ static int __init intel_pstate_init(void)
        if (!id)
                return -ENODEV;
 
+       /*
+        * The Intel pstate driver will be ignored if the platform
+        * firmware has its own power management modes.
+        */
+       if (intel_pstate_platform_pwr_mgmt_exists())
+               return -ENODEV;
+
+       cpu_info = (struct cpu_defaults *)id->driver_data;
+
+       copy_pid_params(&cpu_info->pid_policy);
+       copy_cpu_funcs(&cpu_info->funcs);
+
        if (intel_pstate_msrs_not_valid())
                return -ENODEV;