Merge tag 'pm-5.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm

Pull power management fixes from Rafael Wysocki:
 "Make the intel_pstate driver behave as expected when it operates in
  the passive mode with HWP enabled and the 'powersave' governor on top
  of it"

* tag 'pm-5.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  cpufreq: intel_pstate: Take CPUFREQ_GOV_STRICT_TARGET into account
  cpufreq: Add strict_target to struct cpufreq_policy
  cpufreq: Introduce CPUFREQ_GOV_STRICT_TARGET
  cpufreq: Introduce governor flags

Author: torvalds

drivers/cpufreq/cpufreq.c

@@ -2254,7 +2254,7 @@ static int cpufreq_init_governor(struct cpufreq_policy *policy)
 		return -EINVAL;
 
 	/* Platform doesn't want dynamic frequency switching ? */
-	if (policy->governor->dynamic_switching &&
+	if (policy->governor->flags & CPUFREQ_GOV_DYNAMIC_SWITCHING &&
 	    cpufreq_driver->flags & CPUFREQ_NO_AUTO_DYNAMIC_SWITCHING) {
 		struct cpufreq_governor *gov = cpufreq_fallback_governor();
 
@@ -2280,6 +2280,8 @@ static int cpufreq_init_governor(struct cpufreq_policy *policy)
 		}
 	}
 
+	policy->strict_target = !!(policy->governor->flags & CPUFREQ_GOV_STRICT_TARGET);
+
 	return 0;
 }
 

drivers/cpufreq/cpufreq_governor.h

@@ -156,7 +156,7 @@ void cpufreq_dbs_governor_limits(struct cpufreq_policy *policy);
 #define CPUFREQ_DBS_GOVERNOR_INITIALIZER(_name_)			\
 	{								\
 		.name = _name_,						\
-		.dynamic_switching = true,				\
+		.flags = CPUFREQ_GOV_DYNAMIC_SWITCHING,			\
 		.owner = THIS_MODULE,					\
 		.init = cpufreq_dbs_governor_init,			\
 		.exit = cpufreq_dbs_governor_exit,			\

drivers/cpufreq/cpufreq_performance.c

@@ -20,6 +20,7 @@ static void cpufreq_gov_performance_limits(struct cpufreq_policy *policy)
 static struct cpufreq_governor cpufreq_gov_performance = {
 	.name		= "performance",
 	.owner		= THIS_MODULE,
+	.flags		= CPUFREQ_GOV_STRICT_TARGET,
 	.limits		= cpufreq_gov_performance_limits,
 };
 

drivers/cpufreq/cpufreq_powersave.c

@@ -21,6 +21,7 @@ static struct cpufreq_governor cpufreq_gov_powersave = {
 	.name		= "powersave",
 	.limits		= cpufreq_gov_powersave_limits,
 	.owner		= THIS_MODULE,
+	.flags		= CPUFREQ_GOV_STRICT_TARGET,
 };
 
 MODULE_AUTHOR("Dominik Brodowski <linux@brodo.de>");

drivers/cpufreq/intel_pstate.c

@@ -2527,7 +2527,7 @@ static void intel_cpufreq_trace(struct cpudata *cpu, unsigned int trace_type, in
 }
 
 static void intel_cpufreq_adjust_hwp(struct cpudata *cpu, u32 target_pstate,
-				     bool fast_switch)
+				     bool strict, bool fast_switch)
 {
 	u64 prev = READ_ONCE(cpu->hwp_req_cached), value = prev;
 
@@ -2539,7 +2539,7 @@ static void intel_cpufreq_adjust_hwp(struct cpudata *cpu, u32 target_pstate,
 	 * field in it, so opportunistically update the max too if needed.
 	 */
 	value &= ~HWP_MAX_PERF(~0L);
-	value |= HWP_MAX_PERF(cpu->max_perf_ratio);
+	value |= HWP_MAX_PERF(strict ? target_pstate : cpu->max_perf_ratio);
 
 	if (value == prev)
 		return;
@@ -2562,14 +2562,16 @@ static void intel_cpufreq_adjust_perf_ctl(struct cpudata *cpu,
 			      pstate_funcs.get_val(cpu, target_pstate));
 }
 
-static int intel_cpufreq_update_pstate(struct cpudata *cpu, int target_pstate,
-				       bool fast_switch)
+static int intel_cpufreq_update_pstate(struct cpufreq_policy *policy,
+				       int target_pstate, bool fast_switch)
 {
+	struct cpudata *cpu = all_cpu_data[policy->cpu];
 	int old_pstate = cpu->pstate.current_pstate;
 
 	target_pstate = intel_pstate_prepare_request(cpu, target_pstate);
 	if (hwp_active) {
-		intel_cpufreq_adjust_hwp(cpu, target_pstate, fast_switch);
+		intel_cpufreq_adjust_hwp(cpu, target_pstate,
+					 policy->strict_target, fast_switch);
 		cpu->pstate.current_pstate = target_pstate;
 	} else if (target_pstate != old_pstate) {
 		intel_cpufreq_adjust_perf_ctl(cpu, target_pstate, fast_switch);
@@ -2609,7 +2611,7 @@ static int intel_cpufreq_target(struct cpufreq_policy *policy,
 		break;
 	}
 
-	target_pstate = intel_cpufreq_update_pstate(cpu, target_pstate, false);
+	target_pstate = intel_cpufreq_update_pstate(policy, target_pstate, false);
 
 	freqs.new = target_pstate * cpu->pstate.scaling;
 
@@ -2628,7 +2630,7 @@ static unsigned int intel_cpufreq_fast_switch(struct cpufreq_policy *policy,
 
 	target_pstate = DIV_ROUND_UP(target_freq, cpu->pstate.scaling);
 
-	target_pstate = intel_cpufreq_update_pstate(cpu, target_pstate, true);
+	target_pstate = intel_cpufreq_update_pstate(policy, target_pstate, true);
 
 	return target_pstate * cpu->pstate.scaling;
 }

include/linux/cpufreq.h

@@ -109,6 +109,12 @@ struct cpufreq_policy {
 	bool			fast_switch_possible;
 	bool			fast_switch_enabled;
 
+	/*
+	 * Set if the CPUFREQ_GOV_STRICT_TARGET flag is set for the current
+	 * governor.
+	 */
+	bool			strict_target;
+
 	/*
 	 * Preferred average time interval between consecutive invocations of
 	 * the driver to set the frequency for this policy.  To be set by the
@@ -570,12 +576,20 @@ struct cpufreq_governor {
 					 char *buf);
 	int	(*store_setspeed)	(struct cpufreq_policy *policy,
 					 unsigned int freq);
-	/* For governors which change frequency dynamically by themselves */
-	bool			dynamic_switching;
 	struct list_head	governor_list;
 	struct module		*owner;
+	u8			flags;
 };
 
+/* Governor flags */
+
+/* For governors which change frequency dynamically by themselves */
+#define CPUFREQ_GOV_DYNAMIC_SWITCHING	BIT(0)
+
+/* For governors wanting the target frequency to be set exactly */
+#define CPUFREQ_GOV_STRICT_TARGET	BIT(1)
+
+
 /* Pass a target to the cpufreq driver */
 unsigned int cpufreq_driver_fast_switch(struct cpufreq_policy *policy,
 					unsigned int target_freq);

kernel/sched/cpufreq_schedutil.c

@@ -881,7 +881,7 @@ static void sugov_limits(struct cpufreq_policy *policy)
 struct cpufreq_governor schedutil_gov = {
 	.name			= "schedutil",
 	.owner			= THIS_MODULE,
-	.dynamic_switching	= true,
+	.flags			= CPUFREQ_GOV_DYNAMIC_SWITCHING,
 	.init			= sugov_init,
 	.exit			= sugov_exit,
 	.start			= sugov_start,