Merge: arm64: AMU-based CPU frequency

MR: https://gitlab.com/redhat/centos-stream/src/kernel/centos-stream-10/-/merge_requests/462

JIRA: https://issues.redhat.com/browse/RHEL-80968

This is a backport of the patch series "Add support for AArch64 AMUv1-based average freq" (https://lore.kernel.org/all/20250131162439.3843071-1-beata.michalska@arm.com/) and its dependencies.

The first patch in this MR is upstream in v6.14 (since rc3). The rest are now in Linus's master branch and slated for 6.15.

Signed-off-by: Jennifer Berringer <jberring@redhat.com>

Approved-by: David Arcari <darcari@redhat.com>
Approved-by: Mark Langsdorf <mlangsdo@redhat.com>
Approved-by: CKI KWF Bot <cki-ci-bot+kwf-gitlab-com@redhat.com>

Merged-by: Julio Faracco <jfaracco@redhat.com>

Author: jcfaracco

Documentation/admin-guide/pm/cpufreq.rst

@@ -248,6 +248,20 @@ are the following:
 	If that frequency cannot be determined, this attribute should not
 	be present.
 
+``cpuinfo_avg_freq``
+        An average frequency (in KHz) of all CPUs belonging to a given policy,
+        derived from a hardware provided feedback and reported on a time frame
+        spanning at most few milliseconds.
+
+        This is expected to be based on the frequency the hardware actually runs
+        at and, as such, might require specialised hardware support (such as AMU
+        extension on ARM). If one cannot be determined, this attribute should
+        not be present.
+
+        Note, that failed attempt to retrieve current frequency for a given
+        CPU(s) will result in an appropriate error, i.e: EAGAIN for CPU that
+        remains idle (raised on ARM).
+
 ``cpuinfo_max_freq``
 	Maximum possible operating frequency the CPUs belonging to this policy
 	can run at (in kHz).
@@ -293,7 +307,8 @@ are the following:
 	Some architectures (e.g. ``x86``) may attempt to provide information
 	more precisely reflecting the current CPU frequency through this
 	attribute, but that still may not be the exact current CPU frequency as
-	seen by the hardware at the moment.
+	seen by the hardware at the moment. This behavior though, is only
+	available via c:macro:``CPUFREQ_ARCH_CUR_FREQ`` option.
 
 ``scaling_driver``
 	The scaling driver currently in use.

arch/arm64/kernel/topology.c

@@ -17,6 +17,7 @@
 #include <linux/cpufreq.h>
 #include <linux/init.h>
 #include <linux/percpu.h>
+#include <linux/sched/isolation.h>
 
 #include <asm/cpu.h>
 #include <asm/cputype.h>
@@ -88,18 +89,28 @@ int __init parse_acpi_topology(void)
  * initialized.
  */
 static DEFINE_PER_CPU_READ_MOSTLY(unsigned long, arch_max_freq_scale) =  1UL << (2 * SCHED_CAPACITY_SHIFT);
-static DEFINE_PER_CPU(u64, arch_const_cycles_prev);
-static DEFINE_PER_CPU(u64, arch_core_cycles_prev);
 static cpumask_var_t amu_fie_cpus;
 
+struct amu_cntr_sample {
+	u64		arch_const_cycles_prev;
+	u64		arch_core_cycles_prev;
+	unsigned long	last_scale_update;
+};
+
+static DEFINE_PER_CPU_SHARED_ALIGNED(struct amu_cntr_sample, cpu_amu_samples);
+
 void update_freq_counters_refs(void)
 {
-	this_cpu_write(arch_core_cycles_prev, read_corecnt());
-	this_cpu_write(arch_const_cycles_prev, read_constcnt());
+	struct amu_cntr_sample *amu_sample = this_cpu_ptr(&cpu_amu_samples);
+
+	amu_sample->arch_core_cycles_prev = read_corecnt();
+	amu_sample->arch_const_cycles_prev = read_constcnt();
 }
 
 static inline bool freq_counters_valid(int cpu)
 {
+	struct amu_cntr_sample *amu_sample = per_cpu_ptr(&cpu_amu_samples, cpu);
+
 	if ((cpu >= nr_cpu_ids) || !cpumask_test_cpu(cpu, cpu_present_mask))
 		return false;
 
@@ -108,8 +119,8 @@ static inline bool freq_counters_valid(int cpu)
 		return false;
 	}
 
-	if (unlikely(!per_cpu(arch_const_cycles_prev, cpu) ||
-		     !per_cpu(arch_core_cycles_prev, cpu))) {
+	if (unlikely(!amu_sample->arch_const_cycles_prev ||
+		     !amu_sample->arch_core_cycles_prev)) {
 		pr_debug("CPU%d: cycle counters are not enabled.\n", cpu);
 		return false;
 	}
@@ -152,17 +163,22 @@ void freq_inv_set_max_ratio(int cpu, u64 max_rate)
 
 static void amu_scale_freq_tick(void)
 {
+	struct amu_cntr_sample *amu_sample = this_cpu_ptr(&cpu_amu_samples);
 	u64 prev_core_cnt, prev_const_cnt;
 	u64 core_cnt, const_cnt, scale;
 
-	prev_const_cnt = this_cpu_read(arch_const_cycles_prev);
-	prev_core_cnt = this_cpu_read(arch_core_cycles_prev);
+	prev_const_cnt = amu_sample->arch_const_cycles_prev;
+	prev_core_cnt = amu_sample->arch_core_cycles_prev;
 
 	update_freq_counters_refs();
 
-	const_cnt = this_cpu_read(arch_const_cycles_prev);
-	core_cnt = this_cpu_read(arch_core_cycles_prev);
+	const_cnt = amu_sample->arch_const_cycles_prev;
+	core_cnt = amu_sample->arch_core_cycles_prev;
 
+	/*
+	 * This should not happen unless the AMUs have been reset and the
+	 * counter values have not been restored - unlikely
+	 */
 	if (unlikely(core_cnt <= prev_core_cnt ||
 		     const_cnt <= prev_const_cnt))
 		return;
@@ -182,24 +198,123 @@ static void amu_scale_freq_tick(void)
 
 	scale = min_t(unsigned long, scale, SCHED_CAPACITY_SCALE);
 	this_cpu_write(arch_freq_scale, (unsigned long)scale);
+
+	amu_sample->last_scale_update = jiffies;
 }
 
 static struct scale_freq_data amu_sfd = {
 	.source = SCALE_FREQ_SOURCE_ARCH,
 	.set_freq_scale = amu_scale_freq_tick,
 };
 
+static __always_inline bool amu_fie_cpu_supported(unsigned int cpu)
+{
+	return cpumask_available(amu_fie_cpus) &&
+		cpumask_test_cpu(cpu, amu_fie_cpus);
+}
+
+void arch_cpu_idle_enter(void)
+{
+	unsigned int cpu = smp_processor_id();
+
+	if (!amu_fie_cpu_supported(cpu))
+		return;
+
+	/* Kick in AMU update but only if one has not happened already */
+	if (housekeeping_cpu(cpu, HK_TYPE_TICK) &&
+	    time_is_before_jiffies(per_cpu(cpu_amu_samples.last_scale_update, cpu)))
+		amu_scale_freq_tick();
+}
+
+#define AMU_SAMPLE_EXP_MS	20
+
+int arch_freq_get_on_cpu(int cpu)
+{
+	struct amu_cntr_sample *amu_sample;
+	unsigned int start_cpu = cpu;
+	unsigned long last_update;
+	unsigned int freq = 0;
+	u64 scale;
+
+	if (!amu_fie_cpu_supported(cpu) || !arch_scale_freq_ref(cpu))
+		return -EOPNOTSUPP;
+
+	while (1) {
+
+		amu_sample = per_cpu_ptr(&cpu_amu_samples, cpu);
+
+		last_update = amu_sample->last_scale_update;
+
+		/*
+		 * For those CPUs that are in full dynticks mode, or those that have
+		 * not seen tick for a while, try an alternative source for the counters
+		 * (and thus freq scale), if available, for given policy: this boils
+		 * down to identifying an active cpu within the same freq domain, if any.
+		 */
+		if (!housekeeping_cpu(cpu, HK_TYPE_TICK) ||
+		    time_is_before_jiffies(last_update + msecs_to_jiffies(AMU_SAMPLE_EXP_MS))) {
+			struct cpufreq_policy *policy = cpufreq_cpu_get(cpu);
+			int ref_cpu;
+
+			if (!policy)
+				return -EINVAL;
+
+			if (!cpumask_intersects(policy->related_cpus,
+						housekeeping_cpumask(HK_TYPE_TICK))) {
+				cpufreq_cpu_put(policy);
+				return -EOPNOTSUPP;
+			}
+
+			for_each_cpu_wrap(ref_cpu, policy->cpus, cpu + 1) {
+				if (ref_cpu == start_cpu) {
+					/* Prevent verifying same CPU twice */
+					ref_cpu = nr_cpu_ids;
+					break;
+				}
+				if (!idle_cpu(ref_cpu))
+					break;
+			}
+
+			cpufreq_cpu_put(policy);
+
+			if (ref_cpu >= nr_cpu_ids)
+				/* No alternative to pull info from */
+				return -EAGAIN;
+
+			cpu = ref_cpu;
+		} else {
+			break;
+		}
+	}
+	/*
+	 * Reversed computation to the one used to determine
+	 * the arch_freq_scale value
+	 * (see amu_scale_freq_tick for details)
+	 */
+	scale = arch_scale_freq_capacity(cpu);
+	freq = scale * arch_scale_freq_ref(cpu);
+	freq >>= SCHED_CAPACITY_SHIFT;
+	return freq;
+}
+
 static void amu_fie_setup(const struct cpumask *cpus)
 {
 	int cpu;
 
 	/* We are already set since the last insmod of cpufreq driver */
-	if (unlikely(cpumask_subset(cpus, amu_fie_cpus)))
+	if (cpumask_available(amu_fie_cpus) &&
+	    unlikely(cpumask_subset(cpus, amu_fie_cpus)))
 		return;
 
-	for_each_cpu(cpu, cpus) {
+	for_each_cpu(cpu, cpus)
 		if (!freq_counters_valid(cpu))
 			return;
+
+	if (!cpumask_available(amu_fie_cpus) &&
+	    !zalloc_cpumask_var(&amu_fie_cpus, GFP_KERNEL)) {
+		WARN_ONCE(1, "Failed to allocate FIE cpumask for CPUs[%*pbl]\n",
+			  cpumask_pr_args(cpus));
+		return;
 	}
 
 	cpumask_or(amu_fie_cpus, amu_fie_cpus, cpus);
@@ -237,17 +352,8 @@ static struct notifier_block init_amu_fie_notifier = {
 
 static int __init init_amu_fie(void)
 {
-	int ret;
-
-	if (!zalloc_cpumask_var(&amu_fie_cpus, GFP_KERNEL))
-		return -ENOMEM;
-
-	ret = cpufreq_register_notifier(&init_amu_fie_notifier,
+	return cpufreq_register_notifier(&init_amu_fie_notifier,
 					CPUFREQ_POLICY_NOTIFIER);
-	if (ret)
-		free_cpumask_var(amu_fie_cpus);
-
-	return ret;
 }
 core_initcall(init_amu_fie);
 

arch/x86/kernel/cpu/aperfmperf.c

@@ -498,7 +498,7 @@ void arch_scale_freq_tick(void)
  */
 #define MAX_SAMPLE_AGE	((unsigned long)HZ / 50)
 
-unsigned int arch_freq_get_on_cpu(int cpu)
+int arch_freq_get_on_cpu(int cpu)
 {
 	struct aperfmperf *s = per_cpu_ptr(&cpu_samples, cpu);
 	unsigned int seq, freq;

arch/x86/kernel/cpu/proc.c

@@ -86,9 +86,12 @@ static int show_cpuinfo(struct seq_file *m, void *v)
 		seq_printf(m, "microcode\t: 0x%x\n", c->microcode);
 
 	if (cpu_has(c, X86_FEATURE_TSC)) {
-		unsigned int freq = arch_freq_get_on_cpu(cpu);
+		int freq = arch_freq_get_on_cpu(cpu);
 
-		seq_printf(m, "cpu MHz\t\t: %u.%03u\n", freq / 1000, (freq % 1000));
+		if (freq < 0)
+			seq_puts(m, "cpu MHz\t\t: Unknown\n");
+		else
+			seq_printf(m, "cpu MHz\t\t: %u.%03u\n", freq / 1000, (freq % 1000));
 	}
 
 	/* Cache size */

drivers/base/arch_topology.c

@@ -28,7 +28,7 @@
 static DEFINE_PER_CPU(struct scale_freq_data __rcu *, sft_data);
 static struct cpumask scale_freq_counters_mask;
 static bool scale_freq_invariant;
-DEFINE_PER_CPU(unsigned long, capacity_freq_ref) = 1;
+DEFINE_PER_CPU(unsigned long, capacity_freq_ref) = 0;
 EXPORT_PER_CPU_SYMBOL_GPL(capacity_freq_ref);
 
 static bool supports_scale_freq_counters(const struct cpumask *cpus)
@@ -293,13 +293,15 @@ void topology_normalize_cpu_scale(void)
 
 	capacity_scale = 1;
 	for_each_possible_cpu(cpu) {
-		capacity = raw_capacity[cpu] * per_cpu(capacity_freq_ref, cpu);
+		capacity = raw_capacity[cpu] *
+			   (per_cpu(capacity_freq_ref, cpu) ?: 1);
 		capacity_scale = max(capacity, capacity_scale);
 	}
 
 	pr_debug("cpu_capacity: capacity_scale=%llu\n", capacity_scale);
 	for_each_possible_cpu(cpu) {
-		capacity = raw_capacity[cpu] * per_cpu(capacity_freq_ref, cpu);
+		capacity = raw_capacity[cpu] *
+			   (per_cpu(capacity_freq_ref, cpu) ?: 1);
 		capacity = div64_u64(capacity << SCHED_CAPACITY_SHIFT,
 			capacity_scale);
 		topology_set_cpu_scale(cpu, capacity);

drivers/cpufreq/Kconfig.x86

@@ -340,3 +340,15 @@ config X86_SPEEDSTEP_RELAXED_CAP_CHECK
 	  option lets the probing code bypass some of those checks if the
 	  parameter "relaxed_check=1" is passed to the module.
 
+config CPUFREQ_ARCH_CUR_FREQ
+	default y
+	bool "Current frequency derived from HW provided feedback"
+	help
+	  This determines whether the scaling_cur_freq sysfs attribute returns
+	  the last requested frequency or a more precise value based on hardware
+	  provided feedback (as architected counters).
+	  Given that a more precise frequency can now be provided via the
+	  cpuinfo_avg_freq attribute, by enabling this option,
+	  scaling_cur_freq maintains the provision of a counter based frequency,
+	  for compatibility reasons.
+

drivers/cpufreq/cpufreq.c

@@ -728,18 +728,26 @@ show_one(cpuinfo_transition_latency, cpuinfo.transition_latency);
 show_one(scaling_min_freq, min);
 show_one(scaling_max_freq, max);
 
-__weak unsigned int arch_freq_get_on_cpu(int cpu)
+__weak int arch_freq_get_on_cpu(int cpu)
 {
-	return 0;
+	return -EOPNOTSUPP;
+}
+
+static inline bool cpufreq_avg_freq_supported(struct cpufreq_policy *policy)
+{
+	return arch_freq_get_on_cpu(policy->cpu) != -EOPNOTSUPP;
 }
 
 static ssize_t show_scaling_cur_freq(struct cpufreq_policy *policy, char *buf)
 {
 	ssize_t ret;
-	unsigned int freq;
+	int freq;
+
+	freq = IS_ENABLED(CONFIG_CPUFREQ_ARCH_CUR_FREQ)
+		? arch_freq_get_on_cpu(policy->cpu)
+		: 0;
 
-	freq = arch_freq_get_on_cpu(policy->cpu);
-	if (freq)
+	if (freq > 0)
 		ret = sysfs_emit(buf, "%u\n", freq);
 	else if (cpufreq_driver->setpolicy && cpufreq_driver->get)
 		ret = sysfs_emit(buf, "%u\n", cpufreq_driver->get(policy->cpu));
@@ -783,6 +791,19 @@ static ssize_t show_cpuinfo_cur_freq(struct cpufreq_policy *policy,
 	return sysfs_emit(buf, "<unknown>\n");
 }
 
+/*
+ * show_cpuinfo_avg_freq - average CPU frequency as detected by hardware
+ */
+static ssize_t show_cpuinfo_avg_freq(struct cpufreq_policy *policy,
+				     char *buf)
+{
+	int avg_freq = arch_freq_get_on_cpu(policy->cpu);
+
+	if (avg_freq > 0)
+		return sysfs_emit(buf, "%u\n", avg_freq);
+	return avg_freq != 0 ? avg_freq : -EINVAL;
+}
+
 /*
  * show_scaling_governor - show the current policy for the specified CPU
  */
@@ -945,6 +966,7 @@ static ssize_t show_bios_limit(struct cpufreq_policy *policy, char *buf)
 }
 
 cpufreq_freq_attr_ro_perm(cpuinfo_cur_freq, 0400);
+cpufreq_freq_attr_ro(cpuinfo_avg_freq);
 cpufreq_freq_attr_ro(cpuinfo_min_freq);
 cpufreq_freq_attr_ro(cpuinfo_max_freq);
 cpufreq_freq_attr_ro(cpuinfo_transition_latency);
@@ -1072,6 +1094,12 @@ static int cpufreq_add_dev_interface(struct cpufreq_policy *policy)
 			return ret;
 	}
 
+	if (cpufreq_avg_freq_supported(policy)) {
+		ret = sysfs_create_file(&policy->kobj, &cpuinfo_avg_freq.attr);
+		if (ret)
+			return ret;
+	}
+
 	ret = sysfs_create_file(&policy->kobj, &scaling_cur_freq.attr);
 	if (ret)
 		return ret;