Skip to contentSkip to conversation

cpufreq: intel_pstate: Base frequency attribute

PATCH Superseded Srinivas Pandruvada 2 participants 6 messages Oct 12, 2018 – Oct 15, 2018

Subsystems (6)

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Srinivas Pandruvada

author maintainer
#0 lore raw

v3 cover cpufreq: intel_pstate: Base frequency attribute

This series presents base frequency to cpufreq sysfs when intel_pstate is in use in HWP mode.

Changes:
v3:
- Update documentation
v2
- Removed guaranteed attribute addition to acpi_cppc sysfs
- Using the cppc_acpi interface to get base frequency and present

Srinivas Pandruvada (3):
ACPI / CPPC: Add support for guaranteed performance
cpufreq: intel_pstate: Add base_frequency attribute
Documentation: intel_pstate: Add base_frequency information

 Documentation/admin-guide/pm/intel_pstate.rst |  4 ++
 drivers/acpi/cppc_acpi.c                      |  8 +++-
 drivers/cpufreq/intel_pstate.c                | 38 +++++++++++++++++++
 include/acpi/cppc_acpi.h                      |  1 +
 4 files changed, 49 insertions(+), 2 deletions(-)
Signature · 1 line
2.17.1
author maintainer
#1 lore raw

PATCH v3 1/3 ACPI / CPPC: Add support for guaranteed performance

The Continuous Performance Control Package can have guaranteed performance field. Add support to read guaranteed performance.

---
 drivers/acpi/cppc_acpi.c | 8 ++++++--
 include/acpi/cppc_acpi.h | 1 +
 2 files changed, 7 insertions(+), 2 deletions(-)
drivers/acpi/cppc_acpi.c +6 -2
index d9ce4b162e2c..217a782c3e55 100644
@@ -1061,9 +1061,9 @@ int cppc_get_perf_caps(int cpunum, struct cppc_perf_caps *perf_caps)
 {
 	struct cpc_desc *cpc_desc = per_cpu(cpc_desc_ptr, cpunum);
 	struct cpc_register_resource *highest_reg, *lowest_reg,
-		*lowest_non_linear_reg, *nominal_reg,
+		*lowest_non_linear_reg, *nominal_reg, *guaranteed_reg,
 		*low_freq_reg = NULL, *nom_freq_reg = NULL;
-	u64 high, low, nom, min_nonlinear, low_f = 0, nom_f = 0;
+	u64 high, low, guaranteed, nom, min_nonlinear, low_f = 0, nom_f = 0;
 	int pcc_ss_id = per_cpu(cpu_pcc_subspace_idx, cpunum);
 	struct cppc_pcc_data *pcc_ss_data = NULL;
 	int ret = 0, regs_in_pcc = 0;
@@ -1079,6 +1079,7 @@ int cppc_get_perf_caps(int cpunum, struct cppc_perf_caps *perf_caps)
 	nominal_reg = &cpc_desc->cpc_regs[NOMINAL_PERF];
 	low_freq_reg = &cpc_desc->cpc_regs[LOWEST_FREQ];
 	nom_freq_reg = &cpc_desc->cpc_regs[NOMINAL_FREQ];
+	guaranteed_reg = &cpc_desc->cpc_regs[GUARANTEED_PERF];
 
 	/* Are any of the regs PCC ?*/
 	if (CPC_IN_PCC(highest_reg) || CPC_IN_PCC(lowest_reg) ||
@@ -1107,6 +1108,9 @@ int cppc_get_perf_caps(int cpunum, struct cppc_perf_caps *perf_caps)
 	cpc_read(cpunum, nominal_reg, &nom);
 	perf_caps->nominal_perf = nom;
 
+	cpc_read(cpunum, guaranteed_reg, &guaranteed);
+	perf_caps->guaranteed_perf = guaranteed;
+
 	cpc_read(cpunum, lowest_non_linear_reg, &min_nonlinear);
 	perf_caps->lowest_nonlinear_perf = min_nonlinear;
 
include/acpi/cppc_acpi.h +1 -0
index 8e0b8250a139..cf59e6210d27 100644
@@ -104,6 +104,7 @@ enum cppc_regs {
  * today.
  */
 struct cppc_perf_caps {
+	u32 guaranteed_perf;
 	u32 highest_perf;
 	u32 nominal_perf;
 	u32 lowest_perf;
Signature · 1 line
2.17.1
author maintainer
#2 lore raw

PATCH v3 2/3 cpufreq: intel_pstate: Add base_frequency attribute

In reply to Srinivas Pandruvada

Present base_frequency to user space via cpufreq sysfs when HWP is in use.

This HWP base frequency is read from HWP Capabilities MSR, if platform doesn't have ACPI _CPC object. On most of the HWP platforms the _CPC object will point to the HWP Capabilities MSR using address space id as "Functional Fixed Hardware". But the address space id can be simply ACPI_TYPE_INTEGER, where the platform firmware can modify this value based on the system constraints.

---
 drivers/cpufreq/intel_pstate.c | 38 ++++++++++++++++++++++++++++++++++
 1 file changed, 38 insertions(+)
drivers/cpufreq/intel_pstate.c +38 -0
index b6a1aadaff9f..2a99e2fd9412 100644
@@ -373,10 +373,28 @@ static void intel_pstate_set_itmt_prio(int cpu)
 		}
 	}
 }
+
+static int intel_pstate_get_cppc_guranteed(int cpu)
+{
+	struct cppc_perf_caps cppc_perf;
+	int ret;
+
+	ret = cppc_get_perf_caps(cpu, &cppc_perf);
+	if (ret)
+		return ret;
+
+	return cppc_perf.guaranteed_perf;
+}
+
 #else
 static void intel_pstate_set_itmt_prio(int cpu)
 {
 }
+
+static int intel_pstate_get_cppc_guranteed(int cpu)
+{
+	return -ENOTSUPP;
+}
 #endif
 
 static void intel_pstate_init_acpi_perf_limits(struct cpufreq_policy *policy)
@@ -699,9 +717,29 @@ static ssize_t show_energy_performance_preference(
 
 cpufreq_freq_attr_rw(energy_performance_preference);
 
+static ssize_t show_base_frequency(struct cpufreq_policy *policy, char *buf)
+{
+	struct cpudata *cpu;
+	u64 cap;
+	int ratio;
+
+	ratio = intel_pstate_get_cppc_guranteed(policy->cpu);
+	if (ratio <= 0) {
+		rdmsrl_on_cpu(policy->cpu, MSR_HWP_CAPABILITIES, &cap);
+		ratio = HWP_GUARANTEED_PERF(cap);
+	}
+
+	cpu = all_cpu_data[policy->cpu];
+
+	return sprintf(buf, "%d\n", ratio * cpu->pstate.scaling);
+}
+
+cpufreq_freq_attr_ro(base_frequency);
+
 static struct freq_attr *hwp_cpufreq_attrs[] = {
 	&energy_performance_preference,
 	&energy_performance_available_preferences,
+	&base_frequency,
 	NULL,
 };
 
Signature · 1 line
2.17.1
author maintainer
#3 lore raw

PATCH v3 3/3 Documentation: intel_pstate: Add base_frequency information

In reply to Srinivas Pandruvada

Updated documentation to explain base_frequency attribute.

---
 Documentation/admin-guide/pm/intel_pstate.rst | 4 ++++
 1 file changed, 4 insertions(+)
Documentation/admin-guide/pm/intel_pstate.rst +4 -0
index 8f1d3de449b5..14a5505e073e 100644
@@ -465,6 +465,10 @@ Next, the following policy attributes have special meaning if
 	policy for the time interval between the last two invocations of the
 	driver's utilization update callback by the CPU scheduler for that CPU.
 
+``base_frequency``
+	When present, shows the base frequency of the CPU. Any frequency above
+	this will be in the turbo frequency range.
+
 The meaning of these attributes in the `passive mode <Passive Mode_>`_ is the
 same as for other scaling drivers.
 
Signature · 1 line
2.17.1
#4 lore raw

Re: [PATCH v3 3/3] Documentation: intel_pstate: Add base_frequency information

Quoted 13 lines · Srinivas Pandruvada

Updated documentation to explain base_frequency attribute.

Signed-off-by Srinivas Pandruvada <srinivas.pandruvada(a)linux.intel.com>
---
 Documentation/admin-guide/pm/intel_pstate.rst | 4 ++++
 1 file changed, 4 insertions(+)
Documentation/admin-guide/pm/intel_pstate.rst +0 -0
index 8f1d3de449b5..14a5505e073e 100644

Documentation/admin-guide/pm/intel_pstate.rst Next, the following policy attributes have special meaning if · line 471

Quoted 7 lines
Documentation/admin-guide/pm/intel_pstate.rst +4 -0
@@ -465,6 +465,10 @@ Next, the following policy attributes have special meaning if
        policy for the time interval between the last two invocations of the
        driver's utilization update callback by the CPU scheduler for that CPU.
 
+``base_frequency``
+       When present, shows the base frequency of the CPU. Any frequency above
+       this will be in the turbo frequency range.
+

This isn't entirely correct, because base_frequency is not present in the passive mode (and it is not present for the other drivers for that matter).

Documentation/admin-guide/pm/intel_pstate.rst Next, the following policy attributes have special meaning if · line 473

Quoted 2 lines
Documentation/admin-guide/pm/intel_pstate.rst +0 -0
 The meaning of these attributes in the `passive mode <Passive Mode_>`_ is the
 same as for other scaling drivers.

Instead, I would say:

"One more policy attribute is present if the HWP feature is enabled in the processor <Active Mode With HWP_>_:"

here and I would add the description of the attribute below.

Thanks,
Rafael

author maintainer
#5 lore raw

Re: [PATCH v3 3/3] Documentation: intel_pstate: Add base_frequency information

Quoted 23 lines · Rafael J. Wysocki
Quoted 21 lines · Srinivas Pandruvada

Updated documentation to explain base_frequency attribute.

Signed-off-by: Srinivas Pandruvada <
[email protected]>

---
 Documentation/admin-guide/pm/intel_pstate.rst | 4 ++++
 1 file changed, 4 insertions(+)
Documentation/admin-guide/pm/intel_pstate.rst +0 -0
index 8f1d3de449b5..14a5505e073e 100644
@@ -465,6 +465,10 @@ Next, the following policy attributes have

special meaning if
policy for the time interval between the last two
invocations of the
driver's utilization update callback by the CPU scheduler
for that CPU.

Quoted 10 lines

Documentation/admin-guide/pm/intel_pstate.rst Next, the following policy attributes have special meaning if · line 471

Quoted 5 lines

+``base_frequency``
+ When present, shows the base frequency of the CPU. Any
frequency above

Documentation/admin-guide/pm/intel_pstate.rst +2 -0
+       this will be in the turbo frequency range.
+

This isn't entirely correct, because base_frequency is not present in the passive mode (and it is not present for the other drivers for that
matter).

That's why it said, "when present". This keeps door open to add to other modes too in future if needed.

Thanks,
Srinivas

Quoted 14 lines
Quoted 4 lines

The meaning of these attributes in the `passive mode <Passive
Mode_>`_ is the
same as for other scaling drivers.

Instead, I would say:

"One more policy attribute is present if the HWP feature is enabled in the processor <Active Mode With HWP_>_:"

here and I would add the description of the attribute below.

Thanks,
Rafael