cpufreq: intel_pstate: Implement passive mode with HWP enabled

Allow intel_pstate to work in the passive mode with HWP enabled and
make it set the HWP minimum performance limit (HWP floor) to the
P-state value given by the target frequency supplied by the cpufreq
governor, so as to prevent the HWP algorithm and the CPU scheduler
from working against each other, at least when the schedutil governor
is in use, and update the intel_pstate documentation accordingly.

Among other things, this allows utilization clamps to be taken
into account, at least to a certain extent, when intel_pstate is
in use and makes it more likely that sufficient capacity for
deadline tasks will be provided.

After this change, the resulting behavior of an HWP system with
intel_pstate in the passive mode should be close to the behavior
of the analogous non-HWP system with intel_pstate in the passive
mode, except that the HWP algorithm is generally allowed to make the
CPU run at a frequency above the floor P-state set by intel_pstate in
the entire available range of P-states, while without HWP a CPU can
run in a P-state above the requested one if the latter falls into the
range of turbo P-states (referred to as the turbo range) or if the
P-states of all CPUs in one package are coordinated with each other
at the hardware level.

[Note that in principle the HWP floor may not be taken into account
 by the processor if it falls into the turbo range, in which case the
 processor has a license to choose any P-state, either below or above
 the HWP floor, just like a non-HWP processor in the case when the
 target P-state falls into the turbo range.]

With this change applied, intel_pstate in the passive mode assumes
complete control over the HWP request MSR and concurrent changes of
that MSR (eg. via the direct MSR access interface) are overridden by
it.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>

2 files changed, 184 insertions, 67 deletions

diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
index afad06b91c77..02ab56b2a0d8 100644
--- a/drivers/cpufreq/cpufreq.c
+++ b/drivers/cpufreq/cpufreq.c
@@ -73,8 +73,6 @@ static inline bool has_target(void)
 static unsigned int __cpufreq_get(struct cpufreq_policy *policy);
 static int cpufreq_init_governor(struct cpufreq_policy *policy);
 static void cpufreq_exit_governor(struct cpufreq_policy *policy);
-static int cpufreq_start_governor(struct cpufreq_policy *policy);
-static void cpufreq_stop_governor(struct cpufreq_policy *policy);
 static void cpufreq_governor_limits(struct cpufreq_policy *policy);
 static int cpufreq_set_policy(struct cpufreq_policy *policy,
 			      struct cpufreq_governor *new_gov,
@@ -2266,7 +2264,7 @@ static void cpufreq_exit_governor(struct cpufreq_policy *policy)
 	module_put(policy->governor->owner);
 }
 
-static int cpufreq_start_governor(struct cpufreq_policy *policy)
+int cpufreq_start_governor(struct cpufreq_policy *policy)
 {
 	int ret;
 
@@ -2293,7 +2291,7 @@ static int cpufreq_start_governor(struct cpufreq_policy *policy)
 	return 0;
 }
 
-static void cpufreq_stop_governor(struct cpufreq_policy *policy)
+void cpufreq_stop_governor(struct cpufreq_policy *policy)
 {
 	if (cpufreq_suspended || !policy->governor)
 		return;
diff --git a/drivers/cpufreq/intel_pstate.c b/drivers/cpufreq/intel_pstate.c
index fc459c9c00ff..e0220a6fbc69 100644
--- a/drivers/cpufreq/intel_pstate.c
+++ b/drivers/cpufreq/intel_pstate.c
@@ -36,6 +36,7 @@
 #define INTEL_PSTATE_SAMPLING_INTERVAL	(10 * NSEC_PER_MSEC)
 
 #define INTEL_CPUFREQ_TRANSITION_LATENCY	20000
+#define INTEL_CPUFREQ_TRANSITION_DELAY_HWP	5000
 #define INTEL_CPUFREQ_TRANSITION_DELAY		500
 
 #ifdef CONFIG_ACPI
@@ -220,6 +221,7 @@ struct global_params {
  *			preference/bias
  * @epp_saved:		Saved EPP/EPB during system suspend or CPU offline
  *			operation
+ * @epp_cached		Cached HWP energy-performance preference value
  * @hwp_req_cached:	Cached value of the last HWP Request MSR
  * @hwp_cap_cached:	Cached value of the last HWP Capabilities MSR
  * @last_io_update:	Last time when IO wake flag was set
@@ -257,6 +259,7 @@ struct cpudata {
 	s16 epp_policy;
 	s16 epp_default;
 	s16 epp_saved;
+	s16 epp_cached;
 	u64 hwp_req_cached;
 	u64 hwp_cap_cached;
 	u64 last_io_update;
@@ -639,6 +642,26 @@ static int intel_pstate_get_energy_pref_index(struct cpudata *cpu_data, int *raw
 	return index;
 }
 
+static int intel_pstate_set_epp(struct cpudata *cpu, u32 epp)
+{
+	/*
+	 * Use the cached HWP Request MSR value, because in the active mode the
+	 * register itself may be updated by intel_pstate_hwp_boost_up() or
+	 * intel_pstate_hwp_boost_down() at any time.
+	 */
+	u64 value = READ_ONCE(cpu->hwp_req_cached);
+
+	value &= ~GENMASK_ULL(31, 24);
+	value |= (u64)epp << 24;
+	/*
+	 * The only other updater of hwp_req_cached in the active mode,
+	 * intel_pstate_hwp_set(), is called under the same lock as this
+	 * function, so it cannot run in parallel with the update below.
+	 */
+	WRITE_ONCE(cpu->hwp_req_cached, value);
+	return wrmsrl_on_cpu(cpu->cpu, MSR_HWP_REQUEST, value);
+}
+
 static int intel_pstate_set_energy_pref_index(struct cpudata *cpu_data,
 					      int pref_index, bool use_raw,
 					      u32 raw_epp)
@@ -650,28 +673,12 @@ static int intel_pstate_set_energy_pref_index(struct cpudata *cpu_data,
 		epp = cpu_data->epp_default;
 
 	if (boot_cpu_has(X86_FEATURE_HWP_EPP)) {
-		/*
-		 * Use the cached HWP Request MSR value, because the register
-		 * itself may be updated by intel_pstate_hwp_boost_up() or
-		 * intel_pstate_hwp_boost_down() at any time.
-		 */
-		u64 value = READ_ONCE(cpu_data->hwp_req_cached);
-
-		value &= ~GENMASK_ULL(31, 24);
-
 		if (use_raw)
 			epp = raw_epp;
 		else if (epp == -EINVAL)
 			epp = epp_values[pref_index - 1];
 
-		value |= (u64)epp << 24;
-		/*
-		 * The only other updater of hwp_req_cached in the active mode,
-		 * intel_pstate_hwp_set(), is called under the same lock as this
-		 * function, so it cannot run in parallel with the update below.
-		 */
-		WRITE_ONCE(cpu_data->hwp_req_cached, value);
-		ret = wrmsrl_on_cpu(cpu_data->cpu, MSR_HWP_REQUEST, value);
+		ret = intel_pstate_set_epp(cpu_data, epp);
 	} else {
 		if (epp == -EINVAL)
 			epp = (pref_index - 1) << 2;
@@ -697,10 +704,12 @@ static ssize_t show_energy_performance_available_preferences(
 
 cpufreq_freq_attr_ro(energy_performance_available_preferences);
 
+static struct cpufreq_driver intel_pstate;
+
 static ssize_t store_energy_performance_preference(
 		struct cpufreq_policy *policy, const char *buf, size_t count)
 {
-	struct cpudata *cpu_data = all_cpu_data[policy->cpu];
+	struct cpudata *cpu = all_cpu_data[policy->cpu];
 	char str_preference[21];
 	bool raw = false;
 	ssize_t ret;
@@ -725,15 +734,44 @@ static ssize_t store_energy_performance_preference(
 		raw = true;
 	}
 
+	/*
+	 * This function runs with the policy R/W semaphore held, which
+	 * guarantees that the driver pointer will not change while it is
+	 * running.
+	 */
+	if (!intel_pstate_driver)
+		return -EAGAIN;
+
 	mutex_lock(&intel_pstate_limits_lock);
 
-	ret = intel_pstate_set_energy_pref_index(cpu_data, ret, raw, epp);
-	if (!ret)
-		ret = count;
+	if (intel_pstate_driver == &intel_pstate) {
+		ret = intel_pstate_set_energy_pref_index(cpu, ret, raw, epp);
+	} else {
+		/*
+		 * In the passive mode the governor needs to be stopped on the
+		 * target CPU before the EPP update and restarted after it,
+		 * which is super-heavy-weight, so make sure it is worth doing
+		 * upfront.
+		 */
+		if (!raw)
+			epp = ret ? epp_values[ret - 1] : cpu->epp_default;
+
+		if (cpu->epp_cached != epp) {
+			int err;
+
+			cpufreq_stop_governor(policy);
+			ret = intel_pstate_set_epp(cpu, epp);
+			err = cpufreq_start_governor(policy);
+			if (!ret) {
+				cpu->epp_cached = epp;
+				ret = err;
+			}
+		}
+	}
 
 	mutex_unlock(&intel_pstate_limits_lock);
 
-	return ret;
+	return ret ?: count;
 }
 
 static ssize_t show_energy_performance_preference(
@@ -1145,8 +1183,6 @@ static ssize_t store_no_turbo(struct kobject *a, struct kobj_attribute *b,
 	return count;
 }
 
-static struct cpufreq_driver intel_pstate;
-
 static void update_qos_request(enum freq_qos_req_type type)
 {
 	int max_state, turbo_max, freq, i, perf_pct;
@@ -1330,9 +1366,10 @@ static const struct attribute_group intel_pstate_attr_group = {
 
 static const struct x86_cpu_id intel_pstate_cpu_ee_disable_ids[];
 
+static struct kobject *intel_pstate_kobject;
+
 static void __init intel_pstate_sysfs_expose_params(void)
 {
-	struct kobject *intel_pstate_kobject;
 	int rc;
 
 	intel_pstate_kobject = kobject_create_and_add("intel_pstate",
@@ -1357,17 +1394,31 @@ static void __init intel_pstate_sysfs_expose_params(void)
 	rc = sysfs_create_file(intel_pstate_kobject, &min_perf_pct.attr);
 	WARN_ON(rc);
 
-	if (hwp_active) {
-		rc = sysfs_create_file(intel_pstate_kobject,
-				       &hwp_dynamic_boost.attr);
-		WARN_ON(rc);
-	}
-
 	if (x86_match_cpu(intel_pstate_cpu_ee_disable_ids)) {
 		rc = sysfs_create_file(intel_pstate_kobject, &energy_efficiency.attr);
 		WARN_ON(rc);
 	}
 }
+
+static void intel_pstate_sysfs_expose_hwp_dynamic_boost(void)
+{
+	int rc;
+
+	if (!hwp_active)
+		return;
+
+	rc = sysfs_create_file(intel_pstate_kobject, &hwp_dynamic_boost.attr);
+	WARN_ON_ONCE(rc);
+}
+
+static void intel_pstate_sysfs_hide_hwp_dynamic_boost(void)
+{
+	if (!hwp_active)
+		return;
+
+	sysfs_remove_file(intel_pstate_kobject, &hwp_dynamic_boost.attr);
+}
+
 /************************** sysfs end ************************/
 
 static void intel_pstate_hwp_enable(struct cpudata *cpudata)
@@ -2247,7 +2298,10 @@ static int intel_pstate_verify_policy(struct cpufreq_policy_data *policy)
 
 static void intel_cpufreq_stop_cpu(struct cpufreq_policy *policy)
 {
-	intel_pstate_set_min_pstate(all_cpu_data[policy->cpu]);
+	if (hwp_active)
+		intel_pstate_hwp_force_min_perf(policy->cpu);
+	else
+		intel_pstate_set_min_pstate(all_cpu_data[policy->cpu]);
 }
 
 static void intel_pstate_stop_cpu(struct cpufreq_policy *policy)
@@ -2255,12 +2309,10 @@ static void intel_pstate_stop_cpu(struct cpufreq_policy *policy)
 	pr_debug("CPU %d exiting\n", policy->cpu);
 
 	intel_pstate_clear_update_util_hook(policy->cpu);
-	if (hwp_active) {
+	if (hwp_active)
 		intel_pstate_hwp_save_state(policy);
-		intel_pstate_hwp_force_min_perf(policy->cpu);
-	} else {
-		intel_cpufreq_stop_cpu(policy);
-	}
+
+	intel_cpufreq_stop_cpu(policy);
 }
 
 static int intel_pstate_cpu_exit(struct cpufreq_policy *policy)
@@ -2390,13 +2442,71 @@ static void intel_cpufreq_trace(struct cpudata *cpu, unsigned int trace_type, in
 		fp_toint(cpu->iowait_boost * 100));
 }
 
+static void intel_cpufreq_adjust_hwp(struct cpudata *cpu, u32 target_pstate,
+				     bool fast_switch)
+{
+	u64 prev = READ_ONCE(cpu->hwp_req_cached), value = prev;
+
+	value &= ~HWP_MIN_PERF(~0L);
+	value |= HWP_MIN_PERF(target_pstate);
+
+	/*
+	 * The entire MSR needs to be updated in order to update the HWP min
+	 * field in it, so opportunistically update the max too if needed.
+	 */
+	value &= ~HWP_MAX_PERF(~0L);
+	value |= HWP_MAX_PERF(cpu->max_perf_ratio);
+
+	if (value == prev)
+		return;
+
+	WRITE_ONCE(cpu->hwp_req_cached, value);
+	if (fast_switch)
+		wrmsrl(MSR_HWP_REQUEST, value);
+	else
+		wrmsrl_on_cpu(cpu->cpu, MSR_HWP_REQUEST, value);
+}
+
+static void intel_cpufreq_adjust_perf_ctl(struct cpudata *cpu,
+					  u32 target_pstate, bool fast_switch)
+{
+	if (fast_switch)
+		wrmsrl(MSR_IA32_PERF_CTL,
+		       pstate_funcs.get_val(cpu, target_pstate));
+	else
+		wrmsrl_on_cpu(cpu->cpu, MSR_IA32_PERF_CTL,
+			      pstate_funcs.get_val(cpu, target_pstate));
+}
+
+static int intel_cpufreq_update_pstate(struct cpudata *cpu, int target_pstate,
+				       bool fast_switch)
+{
+	int old_pstate = cpu->pstate.current_pstate;
+
+	target_pstate = intel_pstate_prepare_request(cpu, target_pstate);
+	if (target_pstate != old_pstate) {
+		cpu->pstate.current_pstate = target_pstate;
+		if (hwp_active)
+			intel_cpufreq_adjust_hwp(cpu, target_pstate,
+						 fast_switch);
+		else
+			intel_cpufreq_adjust_perf_ctl(cpu, target_pstate,
+						      fast_switch);
+	}
+
+	intel_cpufreq_trace(cpu, fast_switch ? INTEL_PSTATE_TRACE_FAST_SWITCH :
+			    INTEL_PSTATE_TRACE_TARGET, old_pstate);
+
+	return target_pstate;
+}
+
 static int intel_cpufreq_target(struct cpufreq_policy *policy,
 				unsigned int target_freq,
 				unsigned int relation)
 {
 	struct cpudata *cpu = all_cpu_data[policy->cpu];
 	struct cpufreq_freqs freqs;
-	int target_pstate, old_pstate;
+	int target_pstate;
 
 	update_turbo_state();
 
@@ -2404,6 +2514,7 @@ static int intel_cpufreq_target(struct cpufreq_policy *policy,
 	freqs.new = target_freq;
 
 	cpufreq_freq_transition_begin(policy, &freqs);
+
 	switch (relation) {
 	case CPUFREQ_RELATION_L:
 		target_pstate = DIV_ROUND_UP(freqs.new, cpu->pstate.scaling);
@@ -2415,15 +2526,11 @@ static int intel_cpufreq_target(struct cpufreq_policy *policy,
 		target_pstate = DIV_ROUND_CLOSEST(freqs.new, cpu->pstate.scaling);
 		break;
 	}
-	target_pstate = intel_pstate_prepare_request(cpu, target_pstate);
-	old_pstate = cpu->pstate.current_pstate;
-	if (target_pstate != cpu->pstate.current_pstate) {
-		cpu->pstate.current_pstate = target_pstate;
-		wrmsrl_on_cpu(policy->cpu, MSR_IA32_PERF_CTL,
-			      pstate_funcs.get_val(cpu, target_pstate));
-	}
+
+	target_pstate = intel_cpufreq_update_pstate(cpu, target_pstate, false);
+
 	freqs.new = target_pstate * cpu->pstate.scaling;
-	intel_cpufreq_trace(cpu, INTEL_PSTATE_TRACE_TARGET, old_pstate);
+
 	cpufreq_freq_transition_end(policy, &freqs, false);
 
 	return 0;
@@ -2433,15 +2540,14 @@ static unsigned int intel_cpufreq_fast_switch(struct cpufreq_policy *policy,
 					      unsigned int target_freq)
 {
 	struct cpudata *cpu = all_cpu_data[policy->cpu];
-	int target_pstate, old_pstate;
+	int target_pstate;
 
 	update_turbo_state();
 
 	target_pstate = DIV_ROUND_UP(target_freq, cpu->pstate.scaling);
-	target_pstate = intel_pstate_prepare_request(cpu, target_pstate);
-	old_pstate = cpu->pstate.current_pstate;
-	intel_pstate_update_pstate(cpu, target_pstate);
-	intel_cpufreq_trace(cpu, INTEL_PSTATE_TRACE_FAST_SWITCH, old_pstate);
+
+	target_pstate = intel_cpufreq_update_pstate(cpu, target_pstate, true);
+
 	return target_pstate * cpu->pstate.scaling;
 }
 
@@ -2461,7 +2567,6 @@ static int intel_cpufreq_cpu_init(struct cpufreq_policy *policy)
 		return ret;
 
 	policy->cpuinfo.transition_latency = INTEL_CPUFREQ_TRANSITION_LATENCY;
-	policy->transition_delay_us = INTEL_CPUFREQ_TRANSITION_DELAY;
 	/* This reflects the intel_pstate_get_cpu_pstates() setting. */
 	policy->cur = policy->cpuinfo.min_freq;
 
@@ -2473,10 +2578,18 @@ static int intel_cpufreq_cpu_init(struct cpufreq_policy *policy)
 
 	cpu = all_cpu_data[policy->cpu];
 
-	if (hwp_active)
+	if (hwp_active) {
+		u64 value;
+
 		intel_pstate_get_hwp_max(policy->cpu, &turbo_max, &max_state);
-	else
+		policy->transition_delay_us = INTEL_CPUFREQ_TRANSITION_DELAY_HWP;
+		rdmsrl_on_cpu(cpu->cpu, MSR_HWP_REQUEST, &value);
+		WRITE_ONCE(cpu->hwp_req_cached, value);
+		cpu->epp_cached = (value & GENMASK_ULL(31, 24)) >> 24;
+	} else {
 		turbo_max = cpu->pstate.turbo_pstate;
+		policy->transition_delay_us = INTEL_CPUFREQ_TRANSITION_DELAY;
+	}
 
 	min_freq = DIV_ROUND_UP(turbo_max * global.min_perf_pct, 100);
 	min_freq *= cpu->pstate.scaling;
@@ -2553,6 +2666,10 @@ static void intel_pstate_driver_cleanup(void)
 		}
 	}
 	put_online_cpus();
+
+	if (intel_pstate_driver == &intel_pstate)
+		intel_pstate_sysfs_hide_hwp_dynamic_boost();
+
 	intel_pstate_driver = NULL;
 }
 
@@ -2560,6 +2677,9 @@ static int intel_pstate_register_driver(struct cpufreq_driver *driver)
 {
 	int ret;
 
+	if (driver == &intel_pstate)
+		intel_pstate_sysfs_expose_hwp_dynamic_boost();
+
 	memset(&global, 0, sizeof(global));
 	global.max_perf_pct = 100;
 
@@ -2577,9 +2697,6 @@ static int intel_pstate_register_driver(struct cpufreq_driver *driver)
 
 static int intel_pstate_unregister_driver(void)
 {
-	if (hwp_active)
-		return -EBUSY;
-
 	cpufreq_unregister_driver(intel_pstate_driver);
 	intel_pstate_driver_cleanup();
 
@@ -2835,7 +2952,10 @@ static int __init intel_pstate_init(void)
 			hwp_active++;
 			hwp_mode_bdw = id->driver_data;
 			intel_pstate.attr = hwp_cpufreq_attrs;
-			default_driver = &intel_pstate;
+			intel_cpufreq.attr = hwp_cpufreq_attrs;
+			if (!default_driver)
+				default_driver = &intel_pstate;
+
 			goto hwp_cpu_matched;
 		}
 	} else {
@@ -2906,14 +3026,13 @@ static int __init intel_pstate_setup(char *str)
 	if (!str)
 		return -EINVAL;
 
-	if (!strcmp(str, "disable")) {
+	if (!strcmp(str, "disable"))
 		no_load = 1;
-	} else if (!strcmp(str, "active")) {
+	else if (!strcmp(str, "active"))
 		default_driver = &intel_pstate;
-	} else if (!strcmp(str, "passive")) {
+	else if (!strcmp(str, "passive"))
 		default_driver = &intel_cpufreq;
-		no_hwp = 1;
-	}
+
 	if (!strcmp(str, "no_hwp")) {
 		pr_info("HWP disabled\n");
 		no_hwp = 1;