Blob Blame History Raw
From: Srikar Srimath Tirumala <srikars@nvidia.com>
Date: Thu, 23 Nov 2023 17:44:33 +0530
Subject: ACPI: processor: reduce CPUFREQ thermal reduction pctg for Tegra241
Git-commit: 310293a2b94197f3d75e65ab22672287a7938a00
Patch-mainline: v6.7 or v6.7-rc9 (next release)
References: bsc#1214377

Current implementation of processor_thermal performs software throttling
in fixed steps of "20%" which can be too coarse for some platforms.
We observed some performance gain after reducing the throttle percentage.
Change the CPUFREQ thermal reduction percentage and maximum thermal steps
to be configurable. Also, update the default values of both for Nvidia
Tegra241 (Grace) SoC. The thermal reduction percentage is reduced to "5%"
and accordingly the maximum number of thermal steps are increased as they
are derived from the reduction percentage.

Signed-off-by: Srikar Srimath Tirumala <srikars@nvidia.com>
Co-developed-by: Sumit Gupta <sumitg@nvidia.com>
Signed-off-by: Sumit Gupta <sumitg@nvidia.com>
Acked-by: Sudeep Holla <sudeep.holla@arm.com>
Acked-by: Hanjun Guo <guohanjun@huawei.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Ivan T. Ivanov <iivanov@suse.de>
---
 drivers/acpi/arm64/Makefile          |    1 
 drivers/acpi/arm64/thermal_cpufreq.c |   20 ++++++++++++++
 drivers/acpi/internal.h              |    9 ++++++
 drivers/acpi/processor_thermal.c     |   49 ++++++++++++++++++++++++++++-------
 4 files changed, 70 insertions(+), 9 deletions(-)

--- a/drivers/acpi/arm64/Makefile
+++ b/drivers/acpi/arm64/Makefile
@@ -3,3 +3,4 @@ obj-$(CONFIG_ACPI_AGDI) 	+= agdi.o
 obj-$(CONFIG_ACPI_IORT) 	+= iort.o
 obj-$(CONFIG_ACPI_GTDT) 	+= gtdt.o
 obj-y				+= dma.o
+obj-y				+= thermal_cpufreq.o
--- /dev/null
+++ b/drivers/acpi/arm64/thermal_cpufreq.c
@@ -0,0 +1,20 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include <linux/acpi.h>
+
+#include "../internal.h"
+
+#define SMCCC_SOC_ID_T241      0x036b0241
+
+int acpi_arch_thermal_cpufreq_pctg(void)
+{
+	s32 soc_id = arm_smccc_get_soc_id_version();
+
+	/*
+	 * Check JEP106 code for NVIDIA Tegra241 chip (036b:0241) and
+	 * reduce the CPUFREQ Thermal reduction percentage to 5%.
+	 */
+	if (soc_id == SMCCC_SOC_ID_T241)
+		return 5;
+
+	return 0;
+}
--- a/drivers/acpi/internal.h
+++ b/drivers/acpi/internal.h
@@ -88,6 +88,15 @@ bool acpi_scan_is_offline(struct acpi_de
 acpi_status acpi_sysfs_table_handler(u32 event, void *table, void *context);
 void acpi_scan_table_notify(void);
 
+#ifdef CONFIG_ARM64
+int acpi_arch_thermal_cpufreq_pctg(void);
+#else
+static inline int acpi_arch_thermal_cpufreq_pctg(void)
+{
+	return 0;
+}
+#endif
+
 /* --------------------------------------------------------------------------
                      Device Node Initialization / Removal
    -------------------------------------------------------------------------- */
--- a/drivers/acpi/processor_thermal.c
+++ b/drivers/acpi/processor_thermal.c
@@ -17,6 +17,8 @@
 #include <acpi/processor.h>
 #include <linux/uaccess.h>
 
+#include "internal.h"
+
 #ifdef CONFIG_CPU_FREQ
 
 /* If a passive cooling situation is detected, primarily CPUfreq is used, as it
@@ -26,12 +28,21 @@
  */
 
 #define CPUFREQ_THERMAL_MIN_STEP 0
-#define CPUFREQ_THERMAL_MAX_STEP 3
 
-static DEFINE_PER_CPU(unsigned int, cpufreq_thermal_reduction_pctg);
+static int cpufreq_thermal_max_step __read_mostly = 3;
+
+/*
+ * Minimum throttle percentage for processor_thermal cooling device.
+ * The processor_thermal driver uses it to calculate the percentage amount by
+ * which cpu frequency must be reduced for each cooling state. This is also used
+ * to calculate the maximum number of throttling steps or cooling states.
+ */
+static int cpufreq_thermal_reduction_pctg __read_mostly = 20;
 
-#define reduction_pctg(cpu) \
-	per_cpu(cpufreq_thermal_reduction_pctg, phys_package_first_cpu(cpu))
+static DEFINE_PER_CPU(unsigned int, cpufreq_thermal_reduction_step);
+
+#define reduction_step(cpu) \
+	per_cpu(cpufreq_thermal_reduction_step, phys_package_first_cpu(cpu))
 
 /*
  * Emulate "per package data" using per cpu data (which should really be
@@ -71,7 +82,7 @@ static int cpufreq_get_max_state(unsigne
 	if (!cpu_has_cpufreq(cpu))
 		return 0;
 
-	return CPUFREQ_THERMAL_MAX_STEP;
+	return cpufreq_thermal_max_step;
 }
 
 static int cpufreq_get_cur_state(unsigned int cpu)
@@ -79,7 +90,7 @@ static int cpufreq_get_cur_state(unsigne
 	if (!cpu_has_cpufreq(cpu))
 		return 0;
 
-	return reduction_pctg(cpu);
+	return reduction_step(cpu);
 }
 
 static int cpufreq_set_cur_state(unsigned int cpu, int state)
@@ -92,7 +103,7 @@ static int cpufreq_set_cur_state(unsigne
 	if (!cpu_has_cpufreq(cpu))
 		return 0;
 
-	reduction_pctg(cpu) = state;
+	reduction_step(cpu) = state;
 
 	/*
 	 * Update all the CPUs in the same package because they all
@@ -113,7 +124,8 @@ static int cpufreq_set_cur_state(unsigne
 		if (!policy)
 			return -EINVAL;
 
-		max_freq = (policy->cpuinfo.max_freq * (100 - reduction_pctg(i) * 20)) / 100;
+		max_freq = (policy->cpuinfo.max_freq *
+			    (100 - reduction_step(i) * cpufreq_thermal_reduction_pctg)) / 100;
 
 		cpufreq_cpu_put(policy);
 
@@ -126,10 +138,29 @@ static int cpufreq_set_cur_state(unsigne
 	return 0;
 }
 
+static void acpi_thermal_cpufreq_config(void)
+{
+	int cpufreq_pctg = acpi_arch_thermal_cpufreq_pctg();
+
+	if (!cpufreq_pctg)
+		return;
+
+	cpufreq_thermal_reduction_pctg = cpufreq_pctg;
+
+	/*
+	 * Derive the MAX_STEP from minimum throttle percentage so that the reduction
+	 * percentage doesn't end up becoming negative. Also, cap the MAX_STEP so that
+	 * the CPU performance doesn't become 0.
+	 */
+	cpufreq_thermal_max_step = (100 / cpufreq_pctg) - 2;
+}
+
 void acpi_thermal_cpufreq_init(struct cpufreq_policy *policy)
 {
 	unsigned int cpu;
 
+	acpi_thermal_cpufreq_config();
+
 	for_each_cpu(cpu, policy->related_cpus) {
 		struct acpi_processor *pr = per_cpu(processors, cpu);
 		int ret;
@@ -182,7 +213,7 @@ static int acpi_processor_max_state(stru
 
 	/*
 	 * There exists four states according to
-	 * cpufreq_thermal_reduction_pctg. 0, 1, 2, 3
+	 * cpufreq_thermal_reduction_step. 0, 1, 2, 3
 	 */
 	max_state += cpufreq_get_max_state(pr->id);
 	if (pr->flags.throttling)