[PATCH v6 0/2] ACPI / cpufreq: CPPC: Add ospm_nominal_perf support

[PATCH v6 0/2] ACPI / cpufreq: CPPC: Add ospm_nominal_perf support

[Sumit Gupta]

This series adds support for the OSPM Nominal Performance register
(ACPI 6.6, Section 8.4.6.1.2.6) to the CPPC ACPI and cpufreq drivers.

Unlike the read-only Nominal Performance register, OSPM Nominal
Performance is writable and lets OSPM request a nominal level below
the platform-reported nominal. The platform treats performance above
this level as boost and below as throttle for its power and thermal
decisions.

Patch 1: adds the cppc_{get,set}_ospm_nominal_perf() helpers, a
per-policy cpufreq sysfs attribute (ospm_nominal_freq, in kHz), and
hooks the register into the OSPM-set save/restore table.

Patch 2: reflects the OSPM nominal in the cpufreq policy so that boost
and the frequency limits stay consistent with the register.

This series applies on top of the series
"cpufreq: CPPC: Preserve OSPM-set registers across hotplug and unload",
which provides the save/restore table that patch 1 hooks into:
https://lore.kernel.org/lkml/20260716153820.2007095-1-sumitg@nvidia.com/

v5[5] -> v6:
- New patch 2: reflect the OSPM nominal in the policy's boost state
  and frequency limits.
- Patch 1:
  - write only policy->cpu instead of looping over policy->cpus with
    rollback. the register is treated as shared across a policy's CPUs.
  - add the register to the OSPM-set save/restore table.

Sumit Gupta (2):
  ACPI: CPPC: Add ospm_nominal_perf support
  cpufreq: CPPC: Reflect ospm_nominal_perf in boost and limits

 .../ABI/testing/sysfs-devices-system-cpu      |  26 +++
 drivers/acpi/cppc_acpi.c                      |  32 ++++
 drivers/cpufreq/cppc_cpufreq.c                | 166 +++++++++++++++++-
 include/acpi/cppc_acpi.h                      |  10 ++
 4 files changed, 226 insertions(+), 8 deletions(-)

v5: https://lore.kernel.org/lkml/20260615185934.2383514-1-sumitg@nvidia.com/
v4: https://lore.kernel.org/lkml/20260527194626.185286-1-sumitg@nvidia.com/
v3: https://lore.kernel.org/lkml/20260514194822.1841748-1-sumitg@nvidia.com/
v2: https://lore.kernel.org/lkml/20260430142430.755437-1-sumitg@nvidia.com/
v1: https://lore.kernel.org/lkml/20260427051823.280419-1-sumitg@nvidia.com/

-- 
2.34.1

[PATCH v6 1/2] ACPI: CPPC: Add ospm_nominal_perf support

[Sumit Gupta]

Expose the OSPM Nominal Performance register (ACPI 6.6, Section
8.4.6.1.2.6), which conveys the desired nominal performance level
at which the platform may run. Unlike the existing read-only
Nominal Performance register, it is writable and lets OSPM
request a lower nominal level than the platform-reported nominal.
The platform classifies performance above this level as boosted
and below as throttled for its power/thermal decisions.

It is exposed as a per-policy cpufreq sysfs attribute in kHz, to
match the cpufreq sysfs unit convention:

  /sys/devices/system/cpu/cpuX/cpufreq/ospm_nominal_freq

The attribute is documented in
Documentation/ABI/testing/sysfs-devices-system-cpu.

Writes are converted to perf via cppc_khz_to_perf(), validated
against [Lowest Performance, Nominal Performance], and applied to
the policy->cpu. The register is assumed shared across the
policy->cpus.

On read, the current register value is returned, or
"<unsupported>" if the platform does not implement the register.

Also add the register to the OSPM-set register save/restore
table, so its value survives CPU hotplug and reverts to the
firmware value on driver unload, like the other registers in
the table.

Signed-off-by: Sumit Gupta <sumitg@nvidia.com>
---
 .../ABI/testing/sysfs-devices-system-cpu      | 26 ++++++++++
 drivers/acpi/cppc_acpi.c                      | 32 +++++++++++++
 drivers/cpufreq/cppc_cpufreq.c                | 47 +++++++++++++++++++
 include/acpi/cppc_acpi.h                      | 10 ++++
 4 files changed, 115 insertions(+)

diff --git a/Documentation/ABI/testing/sysfs-devices-system-cpu b/Documentation/ABI/testing/sysfs-devices-system-cpu
index 82d10d556cc8..a8d592c08823 100644
--- a/Documentation/ABI/testing/sysfs-devices-system-cpu
+++ b/Documentation/ABI/testing/sysfs-devices-system-cpu
@@ -346,6 +346,32 @@ Description:	Performance Limited
 
 		This file is only present if the cppc-cpufreq driver is in use.
 
+What:		/sys/devices/system/cpu/cpuX/cpufreq/ospm_nominal_freq
+Date:		May 2026
+Contact:	linux-pm@vger.kernel.org
+Description:	OSPM Nominal Performance (kHz)
+
+		OSPM uses this attribute to request a nominal performance
+		level lower than the platform-reported nominal. The
+		platform treats performance above this level as boost
+		and below as throttle for power and thermal decisions.
+
+		Read returns the current value in kHz, or "<unsupported>"
+		if the platform does not implement the register. Write a
+		kHz value in the range [lowest_freq, nominal_freq].
+
+		Note that tasks may be migrated from one CPU to another
+		by the scheduler's load-balancing algorithm, and if
+		different OSPM Nominal Performance values are set for
+		those CPUs (through different cpufreq policies), that may
+		lead to undesirable outcomes. To avoid such issues it is
+		better to set the same value across all policies, or to
+		pin every task potentially sensitive to it to a specific
+		CPU.
+
+		This file is only present if the cppc-cpufreq driver is
+		in use.
+
 What:		/sys/devices/system/cpu/cpu*/cache/index3/cache_disable_{0,1}
 Date:		August 2008
 KernelVersion:	2.6.27
diff --git a/drivers/acpi/cppc_acpi.c b/drivers/acpi/cppc_acpi.c
index a7fec6c93178..681d4fd40c11 100644
--- a/drivers/acpi/cppc_acpi.c
+++ b/drivers/acpi/cppc_acpi.c
@@ -1685,6 +1685,38 @@ int cppc_set_epp(int cpu, u64 epp_val)
 }
 EXPORT_SYMBOL_GPL(cppc_set_epp);
 
+/**
+ * cppc_set_ospm_nominal_perf() - Write OSPM Nominal Performance register.
+ * @cpu: CPU on which to write register.
+ * @ospm_nominal_perf: Value to write to the OSPM Nominal Performance register.
+ *
+ * OSPM Nominal Performance conveys the desired nominal performance level
+ * at which the platform may run. Per ACPI 6.6, s8.4.6.1.2.6, the value
+ * must lie within [Lowest Performance, Nominal Performance] and may be
+ * set independently of Minimum, Maximum and Desired performance. The
+ * caller is responsible for validating the range.
+ *
+ * Return: 0 on success or negative error code.
+ */
+int cppc_set_ospm_nominal_perf(int cpu, u64 ospm_nominal_perf)
+{
+	return cppc_set_reg_val(cpu, OSPM_NOMINAL_PERF, ospm_nominal_perf);
+}
+EXPORT_SYMBOL_GPL(cppc_set_ospm_nominal_perf);
+
+/**
+ * cppc_get_ospm_nominal_perf() - Read OSPM Nominal Performance register.
+ * @cpu: CPU from which to read register.
+ * @ospm_nominal_perf: Pointer to store the OSPM Nominal Performance value.
+ *
+ * Return: 0 on success or negative error code.
+ */
+int cppc_get_ospm_nominal_perf(int cpu, u64 *ospm_nominal_perf)
+{
+	return cppc_get_reg_val(cpu, OSPM_NOMINAL_PERF, ospm_nominal_perf);
+}
+EXPORT_SYMBOL_GPL(cppc_get_ospm_nominal_perf);
+
 /**
  * cppc_get_auto_act_window() - Read autonomous activity window register.
  * @cpu: CPU from which to read register.
diff --git a/drivers/cpufreq/cppc_cpufreq.c b/drivers/cpufreq/cppc_cpufreq.c
index 9c88512d635c..eb6746810fa6 100644
--- a/drivers/cpufreq/cppc_cpufreq.c
+++ b/drivers/cpufreq/cppc_cpufreq.c
@@ -37,6 +37,7 @@ enum cppc_saved_reg_id {
 	CPPC_SAVED_AUTO_SEL,
 	CPPC_SAVED_EPP,
 	CPPC_SAVED_AUTO_ACT_WINDOW,
+	CPPC_SAVED_OSPM_NOMINAL_PERF,
 	CPPC_NR_SAVED_REGS,
 };
 
@@ -55,6 +56,9 @@ static const struct cppc_saved_reg cppc_saved_regs[CPPC_NR_SAVED_REGS] = {
 	[CPPC_SAVED_AUTO_ACT_WINDOW] = {
 		cppc_get_auto_act_window, cppc_set_auto_act_window,
 	},
+	[CPPC_SAVED_OSPM_NOMINAL_PERF] = {
+		cppc_get_ospm_nominal_perf, cppc_set_ospm_nominal_perf,
+	},
 };
 
 /*
@@ -1166,11 +1170,53 @@ static int cppc_get_perf_limited_filtered(int cpu, u64 *perf_limited)
 CPPC_CPUFREQ_ATTR_RW_U64(perf_limited, cppc_get_perf_limited_filtered,
 			 cppc_set_perf_limited)
 
+static ssize_t show_ospm_nominal_freq(struct cpufreq_policy *policy, char *buf)
+{
+	struct cppc_cpudata *cpu_data = policy->driver_data;
+	u64 perf;
+	int ret;
+
+	ret = cppc_get_ospm_nominal_perf(policy->cpu, &perf);
+	if (ret == -EOPNOTSUPP)
+		return sysfs_emit(buf, "<unsupported>\n");
+	if (ret)
+		return ret;
+
+	return sysfs_emit(buf, "%u\n",
+			  cppc_perf_to_khz(&cpu_data->perf_caps, perf));
+}
+
+static ssize_t store_ospm_nominal_freq(struct cpufreq_policy *policy,
+				       const char *buf, size_t count)
+{
+	struct cppc_cpudata *cpu_data = policy->driver_data;
+	unsigned int freq_khz;
+	u32 perf;
+	int ret;
+
+	ret = kstrtouint(buf, 0, &freq_khz);
+	if (ret)
+		return ret;
+
+	perf = cppc_khz_to_perf(&cpu_data->perf_caps, freq_khz);
+	if (perf < cpu_data->perf_caps.lowest_perf ||
+	    perf > cpu_data->perf_caps.nominal_perf)
+		return -EINVAL;
+
+	/* The register is assumed shared across the policy's CPUs. */
+	ret = cppc_set_ospm_nominal_perf(policy->cpu, perf);
+	if (ret)
+		return ret;
+
+	return count;
+}
+
 cpufreq_freq_attr_ro(freqdomain_cpus);
 cpufreq_freq_attr_rw(auto_select);
 cpufreq_freq_attr_rw(auto_act_window);
 cpufreq_freq_attr_rw(energy_performance_preference_val);
 cpufreq_freq_attr_rw(perf_limited);
+cpufreq_freq_attr_rw(ospm_nominal_freq);
 
 static struct freq_attr *cppc_cpufreq_attr[] = {
 	&freqdomain_cpus,
@@ -1178,6 +1224,7 @@ static struct freq_attr *cppc_cpufreq_attr[] = {
 	&auto_act_window,
 	&energy_performance_preference_val,
 	&perf_limited,
+	&ospm_nominal_freq,
 	NULL,
 };
 
diff --git a/include/acpi/cppc_acpi.h b/include/acpi/cppc_acpi.h
index cd07e1e92bf4..4c9e59643cc0 100644
--- a/include/acpi/cppc_acpi.h
+++ b/include/acpi/cppc_acpi.h
@@ -180,6 +180,8 @@ extern int cpc_write_ffh(int cpunum, struct cpc_reg *reg, u64 val);
 extern int cppc_get_epp_perf(int cpunum, u64 *epp_perf);
 extern int cppc_set_epp_perf(int cpu, struct cppc_perf_ctrls *perf_ctrls, bool enable);
 extern int cppc_set_epp(int cpu, u64 epp_val);
+extern int cppc_set_ospm_nominal_perf(int cpu, u64 ospm_nominal_perf);
+extern int cppc_get_ospm_nominal_perf(int cpu, u64 *ospm_nominal_perf);
 extern int cppc_get_auto_act_window(int cpu, u64 *auto_act_window);
 extern int cppc_set_auto_act_window(int cpu, u64 auto_act_window);
 extern int cppc_get_auto_sel(int cpu, bool *enable);
@@ -268,6 +270,14 @@ static inline int cppc_set_epp(int cpu, u64 epp_val)
 {
 	return -EOPNOTSUPP;
 }
+static inline int cppc_set_ospm_nominal_perf(int cpu, u64 ospm_nominal_perf)
+{
+	return -EOPNOTSUPP;
+}
+static inline int cppc_get_ospm_nominal_perf(int cpu, u64 *ospm_nominal_perf)
+{
+	return -EOPNOTSUPP;
+}
 static inline int cppc_get_auto_act_window(int cpu, u64 *auto_act_window)
 {
 	return -EOPNOTSUPP;
-- 
2.34.1

[PATCH v6 2/2] cpufreq: CPPC: Reflect ospm_nominal_perf in boost and limits

[Sumit Gupta]

OSPM Nominal Performance lets OSPM request a nominal level below the
platform-reported nominal. Since boost is the range above nominal,
lowering the nominal enlarges the boost range and drops the non-boost
ceiling.

cppc_cpufreq currently uses the platform nominal as the non-boost
ceiling. Changing the OSPM nominal then leaves the policy limits
unchanged, so the boost and non-boost ranges no longer match the
register.

Reflect the OSPM nominal in the cpufreq policy so that boost and the
frequency limits stay consistent with the register:

  - Add cppc_cpufreq_get_effective_nominal(), which yields the OSPM
    Nominal Performance when set and the platform nominal otherwise.
  - Use it to derive cpuinfo.max_freq for the non-boost ceiling in init()
    and in set_boost() while boost is disabled.

The cpufreq core includes a per-policy boost QoS request among the
constraints that cap scaling_max_freq. It updates that request to match
cpuinfo.max_freq, but only when boost is toggled. Writing
ospm_nominal_freq changes cpuinfo.max_freq without toggling boost, so the
request (boost_freq_req) is left stale and keeps scaling_max_freq at the
old nominal. cppc_cpufreq_reflect_nominal() therefore updates the request
directly when the nominal changes while boost is disabled.

At boot, when highest_perf == nominal_perf there is no boost range, but
lowering the OSPM nominal later from sysfs can create one if
highest_perf > lowest_perf. Since the core registers its boost
FREQ_QOS_MAX request at policy setup only when boost_supported is already
set, set boost_supported in init() when the OSPM register is supported
and such a range is possible. Boost can then be enabled once the nominal
is lowered.

Signed-off-by: Sumit Gupta <sumitg@nvidia.com>
---
 drivers/cpufreq/cppc_cpufreq.c | 119 ++++++++++++++++++++++++++++++---
 1 file changed, 111 insertions(+), 8 deletions(-)

diff --git a/drivers/cpufreq/cppc_cpufreq.c b/drivers/cpufreq/cppc_cpufreq.c
index eb6746810fa6..048bb567ec45 100644
--- a/drivers/cpufreq/cppc_cpufreq.c
+++ b/drivers/cpufreq/cppc_cpufreq.c
@@ -761,11 +761,63 @@ static void cppc_cpufreq_put_cpu_data(struct cpufreq_policy *policy)
 	policy->driver_data = NULL;
 }
 
+/**
+ * cppc_cpufreq_get_effective_nominal() - Get the effective nominal performance
+ * @policy: cpufreq policy associated with the CPU
+ * @nominal: Updated with the effective nominal performance
+ * @supported: Updated with whether OSPM Nominal Performance is supported.
+ *	       Pass NULL if not needed
+ *
+ * Use the OSPM Nominal Performance value when the register is supported and
+ * contains a nonzero value. Otherwise, use the platform-reported Nominal
+ * Performance. The resulting value is the non-boost performance ceiling.
+ *
+ * Return: 0 on success, or a negative error code if the register cannot be
+ * read or contains a value outside the supported performance range.
+ */
+static int cppc_cpufreq_get_effective_nominal(struct cpufreq_policy *policy,
+					      u32 *nominal, bool *supported)
+{
+	struct cppc_cpudata *cpu_data = policy->driver_data;
+	struct cppc_perf_caps *caps = &cpu_data->perf_caps;
+	u32 effective_nominal = caps->nominal_perf;
+	bool ospm_supported = false;
+	u64 ospm_nominal;
+	int ret;
+
+	ret = cppc_get_ospm_nominal_perf(policy->cpu, &ospm_nominal);
+	if (ret == -EOPNOTSUPP)
+		goto out;
+	if (ret)
+		return ret;
+
+	ospm_supported = true;
+
+	/* A zero value means OSPM has not selected a nominal level. */
+	if (!ospm_nominal)
+		goto out;
+
+	if (ospm_nominal < caps->lowest_perf ||
+	    ospm_nominal > caps->nominal_perf)
+		return -EINVAL;
+
+	effective_nominal = (u32)ospm_nominal;
+
+out:
+	*nominal = effective_nominal;
+	if (supported)
+		*supported = ospm_supported;
+
+	return 0;
+}
+
 static int cppc_cpufreq_cpu_init(struct cpufreq_policy *policy)
 {
 	unsigned int cpu = policy->cpu;
 	struct cppc_cpudata *cpu_data;
 	struct cppc_perf_caps *caps;
+	bool ospm_supported;
+	u32 nominal, perf;
 	int ret;
 
 	cpu_data = cppc_cpufreq_get_cpu_data(cpu);
@@ -788,8 +840,17 @@ static int cppc_cpufreq_cpu_init(struct cpufreq_policy *policy)
 	 * nonlinear perf
 	 */
 	policy->cpuinfo.min_freq = cppc_perf_to_khz(caps, caps->lowest_perf);
-	policy->cpuinfo.max_freq = cppc_perf_to_khz(caps, policy->boost_enabled ?
-						    caps->highest_perf : caps->nominal_perf);
+
+	ret = cppc_cpufreq_get_effective_nominal(policy, &nominal,
+						 &ospm_supported);
+	if (ret) {
+		pr_debug("CPU%u: failed to get effective nominal: %d\n",
+			 cpu, ret);
+		goto out;
+	}
+
+	perf = policy->boost_enabled ? caps->highest_perf : nominal;
+	policy->cpuinfo.max_freq = cppc_perf_to_khz(caps, perf);
 
 	policy->transition_delay_us = cppc_cpufreq_get_transition_delay_us(cpu);
 	policy->shared_type = cpu_data->shared_type;
@@ -819,9 +880,13 @@ static int cppc_cpufreq_cpu_init(struct cpufreq_policy *policy)
 
 	/*
 	 * If 'highest_perf' is greater than 'nominal_perf', we assume CPU Boost
-	 * is supported.
+	 * is supported. A writable OSPM Nominal Performance register can also
+	 * open a boost range at runtime by lowering the nominal, so assume
+	 * boost is supported in that case too, letting the core register its
+	 * QoS request up front.
 	 */
-	if (caps->highest_perf > caps->nominal_perf)
+	if (caps->highest_perf > caps->nominal_perf ||
+	    (ospm_supported && caps->highest_perf > caps->lowest_perf))
 		policy->boost_supported = true;
 
 	/* Set policy->cur to max now. The governors will adjust later. */
@@ -992,11 +1057,16 @@ static int cppc_cpufreq_set_boost(struct cpufreq_policy *policy, int state)
 {
 	struct cppc_cpudata *cpu_data = policy->driver_data;
 	struct cppc_perf_caps *caps = &cpu_data->perf_caps;
+	u32 perf = caps->highest_perf;
+	int ret;
 
-	if (state)
-		policy->cpuinfo.max_freq = cppc_perf_to_khz(caps, caps->highest_perf);
-	else
-		policy->cpuinfo.max_freq = cppc_perf_to_khz(caps, caps->nominal_perf);
+	if (!state) {
+		ret = cppc_cpufreq_get_effective_nominal(policy, &perf, NULL);
+		if (ret)
+			return ret;
+	}
+
+	policy->cpuinfo.max_freq = cppc_perf_to_khz(caps, perf);
 
 	return 0;
 }
@@ -1025,6 +1095,35 @@ static ssize_t show_auto_select(struct cpufreq_policy *policy, char *buf)
 	return sysfs_emit(buf, "%d\n", val);
 }
 
+static int cppc_cpufreq_reflect_nominal(struct cpufreq_policy *policy,
+					u32 nominal)
+{
+	struct cppc_cpudata *cpu_data = policy->driver_data;
+	int ret;
+
+	/*
+	 * While boost is disabled, the nominal is the ceiling. Set
+	 * cpuinfo.max_freq to it and update the core's boost_freq_req to
+	 * match. The core only syncs boost_freq_req when boost is enabled or
+	 * disabled, so a plain nominal change must update it here.
+	 */
+	if (!policy->boost_enabled) {
+		policy->cpuinfo.max_freq =
+			cppc_perf_to_khz(&cpu_data->perf_caps, nominal);
+
+		if (freq_qos_request_active(&policy->boost_freq_req)) {
+			ret = freq_qos_update_request(&policy->boost_freq_req,
+						      policy->cpuinfo.max_freq);
+			if (ret < 0)
+				return ret;
+		}
+	}
+
+	refresh_frequency_limits(policy);
+
+	return 0;
+}
+
 static ssize_t store_auto_select(struct cpufreq_policy *policy,
 				 const char *buf, size_t count)
 {
@@ -1208,6 +1307,10 @@ static ssize_t store_ospm_nominal_freq(struct cpufreq_policy *policy,
 	if (ret)
 		return ret;
 
+	ret = cppc_cpufreq_reflect_nominal(policy, perf);
+	if (ret)
+		return ret;
+
 	return count;
 }
 
-- 
2.34.1