Support CPR on MSM8916
Subsystems (9)
v2 cover Support CPR on MSM8916
This patch series adds support for CPR on MSM8916. The first patch exposes a corner voting API to the CPR driver so that we can change the corner for the MX regulator. If possible I would like to make this patch prettier, but I don't have any great ideas right now. The next patch adds support to adjust voltages in the OPP layer, and then hooks that up to cpufreq-dt so that we can adjust the voltage in response to what CPR tells us to do. I've also thrown in a patch to make RCU lockdep warnings go away, but I'm not sure if it's right. There's still work to do.
The final patch adds the CPR driver. This still has some rough edges. With the OPPv2 bindings I'm thinking of moving the frequency tables into DT and adding a custom vendor property to describe which fuse corner to use for each frequency.
Once you have these patches in place along with a CPU clock driver you can eanble enable cpufreq-dt and add the cpufreq-dt device (maybe the CPR driver should add the cpufreq-dt device?) and you'll see interrupts for CPR and OPP voltage adjustments triggering CPUfreq to modify voltages.
Andy Gross (1):
regulator: smd: Add floor and corner operations
Stephen Boyd (4):
PM / OPP: Support adjusting OPP voltages at runtime
OPP: Allow notifiers to call dev_pm_opp_get_{voltage,freq} RCU-free
cpufreq-dt: Handle OPP voltage adjust events
power: avs: Add support for CPR (Core Power Reduction)
.../devicetree/bindings/power/avs/qcom,cpr.txt | 125 ++ drivers/base/power/opp.c | 96 +- drivers/cpufreq/cpufreq-dt.c | 63 +- drivers/power/avs/Kconfig | 15 + drivers/power/avs/Makefile | 1 + drivers/power/avs/qcom-cpr.c | 1999 ++++++++++++++++++++ drivers/regulator/qcom_smd-regulator.c | 52 + include/linux/pm_opp.h | 11 + include/linux/regulator/qcom_smd-regulator.h | 30 + 9 files changed, 2378 insertions(+), 14 deletions(-) create mode 100644 Documentation/devicetree/bindings/power/avs/qcom,cpr.txt create mode 100644 drivers/power/avs/qcom-cpr.c create mode 100644 include/linux/regulator/qcom_smd-regulator.h
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
PATCH v2 1/5 regulator: smd: Add floor and corner operations
From: Andy Gross <[email protected]>
This patch addes the Qualcomm specific functions for setting the floor and corner voltages on the regulators.
--- drivers/regulator/qcom_smd-regulator.c | 52 ++++++++++++++++++++++++++++ include/linux/regulator/qcom_smd-regulator.h | 30 ++++++++++++++++ 2 files changed, 82 insertions(+) create mode 100644 include/linux/regulator/qcom_smd-regulator.h
drivers/regulator/qcom_smd-regulator.c +52 -0
@@ -20,6 +20,9 @@#include <linux/regulator/machine.h> #include <linux/regulator/of_regulator.h> #include <linux/soc/qcom/smd-rpm.h> +#include <linux/regulator/qcom_smd-regulator.h> + +#include "internal.h" struct qcom_rpm_reg {struct device *dev;
@@ -44,6 +47,11 @@ struct rpm_regulator_req {#define RPM_KEY_SWEN 0x6e657773/* "swen" */ #define RPM_KEY_UV 0x00007675/* "uv" */ #define RPM_KEY_MA 0x0000616d/* "ma" */ +#define RPM_KEY_FLOOR 0x00636676/* "vfc" */ +#define RPM_KEY_CORNER 0x6e726f63/* "corn" */ + +#define RPM_MIN_FLOOR_CORNER 0 +#define RPM_MAX_FLOOR_CORNER 6 static int rpm_reg_write_active(struct qcom_rpm_reg *vreg, struct rpm_regulator_req *req,
@@ -56,6 +64,50 @@ static int rpm_reg_write_active(struct qcom_rpm_reg *vreg, req, size); } +int qcom_rpm_set_floor (struct regulator *regulator,int floor) +{ +struct regulator_dev *rdev = regulator->rdev; +struct qcom_rpm_reg *vreg =rdev_get_drvdata (rdev); +struct rpm_regulator_req req; +int ret; + + req.key =RPM_KEY_FLOOR ; + req.nbytes =sizeof (u32); + req.value = floor; + +if (floor <RPM_MIN_FLOOR_CORNER || floor >RPM_MAX_FLOOR_CORNER ) +return -EINVAL ; + + ret =rpm_reg_write_active (vreg, &req,sizeof (req)); +if (ret) +dev_err (rdev_get_dev (rdev),"Failed to set floor %d\n" , floor); + +return ret; +} +EXPORT_SYMBOL (qcom_rpm_set_floor); + +int qcom_rpm_set_corner (struct regulator *regulator,int corner) +{ +struct regulator_dev *rdev = regulator->rdev; +struct qcom_rpm_reg *vreg =rdev_get_drvdata (rdev); +struct rpm_regulator_req req; +int ret; + + req.key =RPM_KEY_CORNER ; + req.nbytes =sizeof (u32); + req.value = corner; + +if (corner <RPM_MIN_FLOOR_CORNER || corner >RPM_MAX_FLOOR_CORNER ) +return -EINVAL ; + + ret =rpm_reg_write_active (vreg, &req,sizeof (req)); +if (ret) +dev_err (rdev_get_dev (rdev),"Failed to set corner %d\n" , corner); + +return ret; +} +EXPORT_SYMBOL (qcom_rpm_set_corner); +static int rpm_reg_enable (struct regulator_dev *rdev) {struct qcom_rpm_reg *vreg =rdev_get_drvdata (rdev);
include/linux/regulator/qcom_smd-regulator.h new file +30 -0
@@ -0,0 +1,30 @@ +/* Copyright (c) 2015, The Linux Foundation. All rights reserved. +* +* This program is free software; you can redistribute it and/or modify +* it under the terms of the GNU General Public License version 2 and +* only version 2 as published by the Free Software Foundation. +* +* This program is distributed in the hope that it will be useful, +* but WITHOUT ANY WARRANTY; without even the implied warranty of +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +* GNU General Public License for more details. +*/ +#ifndef __QCOM_SMD_REGULATOR_H_ +#define __QCOM_SMD_REGULATOR_H_ + +#ifdef CONFIG_REGULATOR_QCOM_SMD_RPM +int qcom_rpm_set_floor (struct regulator *regulator,int floor); +int qcom_rpm_set_corner (struct regulator *regulator,int corner); +#else +static inline int qcom_rpm_set_floor (struct regulator *regulator,int floor) +{ +return -EINVAL ; +} + +static inline int qcom_rpm_set_corner (struct regulator *regulator,int corner) +{ +return -EINVAL ; +} +#endif + +#endif
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
PATCH v2 2/5 PM / OPP: Support adjusting OPP voltages at runtime
In reply to Stephen BoydOn some SoCs the Adaptive Voltage Scaling (AVS) technique is employed to optimize the operating voltage of a device. At a given frequency, the hardware monitors dynamic factors and either makes a suggestion for how much to adjust a voltage for the current frequency, or it automatically adjusts the voltage without software intervention. Add an API to the OPP library for the former case, so that AVS type devices can update the voltages for an OPP when the hardware determines the voltage should change. The assumption is that drivers like CPUfreq or devfreq will register for the OPP notifiers and adjust the voltage according to suggestions that AVS makes.
--- drivers/base/power/opp.c | 77 ++++++++++++++++++++++++++++++++++++++++++++++++ include/linux/pm_opp.h | 11 +++++++ 2 files changed, 88 insertions(+)
drivers/base/power/opp.c +77 -0
@@ -1136,6 +1136,83 @@ unlock: }/** +* dev_pm_opp_adjust_voltage() - helper to change the voltage of an OPP +* @dev: device for which we do this operation +* @freq: OPP frequency to adjust voltage of +* @u_volt: new OPP voltage +* +* Change the voltage of an OPP with an RCU operation. +* +* Return: -EINVAL for bad pointers, -ENOMEM if no memory available for the +* copy operation, returns 0 if no modifcation was done OR modification was +* successful. +* +* Locking: The internal device_opp and opp structures are RCU protected. +* Hence this function internally uses RCU updater strategy with mutex locks to +* keep the integrity of the internal data structures. Callers should ensure +* that this function is *NOT* called under RCU protection or in contexts where +* mutex locking or synchronize_rcu() blocking calls cannot be used. +*/ +int dev_pm_opp_adjust_voltage (struct device *dev,unsigned long freq, +unsigned long u_volt) +{ +struct device_opp *dev_opp; +struct dev_pm_opp *new_opp, *tmp_opp, *opp =ERR_PTR (-ENODEV ); +int r =0 ; + +/* keep the node allocated */ + new_opp =kmalloc (sizeof (*new_opp),GFP_KERNEL ); +if (!new_opp) +return -ENOMEM ; + +mutex_lock (&dev_opp_list_lock); + +/* Find the device_opp */ + dev_opp =_find_device_opp (dev); +if (IS_ERR (dev_opp)) { + r =PTR_ERR (dev_opp); +dev_warn (dev,"%s: Device OPP not found (%d)\n" , __func__, r); + goto unlock; + } + +/* Do we have the frequency? */ +list_for_each_entry (tmp_opp, &dev_opp->opp_list, node) { +if (tmp_opp->rate == freq) { + opp = tmp_opp; +break ; + } + } +if (IS_ERR (opp)) { + r =PTR_ERR (opp); + goto unlock; + } + +/* Is update really needed? */ +if (opp->u_volt == u_volt) + goto unlock; +/* copy the old data over */ + *new_opp = *opp; + +/* plug in new node */ + new_opp->u_volt = u_volt; + +list_replace_rcu (&opp->node, &new_opp->node); +mutex_unlock (&dev_opp_list_lock); +call_srcu (&dev_opp->srcu_head.srcu, &opp->rcu_head, _kfree_opp_rcu); + +/* Notify the change of the OPP */ +srcu_notifier_call_chain (&dev_opp->srcu_head,OPP_EVENT_ADJUST_VOLTAGE , + new_opp); + +return 0 ; + +unlock: +mutex_unlock (&dev_opp_list_lock); +kfree (new_opp); +return r; +} + +/** * dev_pm_opp_enable() - Enable a specific OPP * @dev: device for which we do this operation * @freq: OPP frequency to enable
include/linux/pm_opp.h +11 -0
@@ -22,6 +22,7 @@ struct device;enum dev_pm_opp_event {OPP_EVENT_ADD ,OPP_EVENT_REMOVE ,OPP_EVENT_ENABLE ,OPP_EVENT_DISABLE , +OPP_EVENT_ADJUST_VOLTAGE , };#if defined(CONFIG_PM_OPP )
@@ -50,6 +51,9 @@ int dev_pm_opp_add(struct device *dev, unsigned long freq, unsigned long u_volt);void dev_pm_opp_remove (struct device *dev,unsigned long freq); +int dev_pm_opp_adjust_voltage (struct device *dev,unsigned long freq, +unsigned long u_volt); +int dev_pm_opp_enable (struct device *dev,unsigned long freq);int dev_pm_opp_disable (struct device *dev,unsigned long freq);
@@ -114,6 +118,13 @@ static inline void dev_pm_opp_remove(struct device *dev, unsigned long freq) { } +static inline int +dev_pm_opp_adjust_voltage (struct device *dev,unsigned long freq, +unsigned long u_volt) +{ +return 0 ; +} +static inline int dev_pm_opp_enable (struct device *dev,unsigned long freq) {return 0 ;
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
PATCH v2 3/5 OPP: Allow notifiers to call dev_pm_opp_get_{voltage,freq} RCU-free
In reply to Stephen BoydWe pass the dev_pm_opp structure to OPP notifiers but the users of the notifier need to surround calls to dev_pm_opp_get_*() with RCU read locks to avoid lockdep warnings. The notifier is already called with the dev_opp's srcu lock held, so it should be safe to assume the devm_pm_opp structure is already protected inside the notifier. Update the lockdep check for this.
--- drivers/base/power/opp.c | 19 +++++++++++-------- 1 file changed, 11 insertions(+), 8 deletions(-)
drivers/base/power/opp.c +11 -8
@@ -146,9 +146,10 @@ static LIST_HEAD(dev_opp_list);/* Lock to allow exclusive modification to the device and opp lists */ static DEFINE_MUTEX (dev_opp_list_lock); -#define opp_rcu_lockdep_assert () \ +#define opp_rcu_lockdep_assert (s) \ do { \ RCU_LOCKDEP_WARN(!rcu_read_lock_held() && \ + !(s && srcu_read_lock_held(s)) && \ !lockdep_is_held(&dev_opp_list_lock), \ "Missing rcu_read_lock() or " \ "dev_opp_list_lock protection"); \
@@ -236,9 +237,10 @@ unsigned long dev_pm_opp_get_voltage(struct dev_pm_opp *opp)struct dev_pm_opp *tmp_opp;unsigned long v =0 ; -opp_rcu_lockdep_assert (); +opp_rcu_lockdep_assert (&opp->dev_opp->srcu_head.srcu); - tmp_opp =rcu_dereference (opp); + tmp_opp =srcu_dereference_check (opp, &opp->dev_opp->srcu_head.srcu, +rcu_read_lock_held ());if (IS_ERR_OR_NULL (tmp_opp) || !tmp_opp->available)pr_err ("%s: Invalid parameters\n" , __func__);else
@@ -268,9 +270,10 @@ unsigned long dev_pm_opp_get_freq(struct dev_pm_opp *opp)struct dev_pm_opp *tmp_opp;unsigned long f =0 ; -opp_rcu_lockdep_assert (); +opp_rcu_lockdep_assert (&opp->dev_opp->srcu_head.srcu); - tmp_opp =rcu_dereference (opp); + tmp_opp =srcu_dereference_check (opp, &opp->dev_opp->srcu_head.srcu, +rcu_read_lock_held ());if (IS_ERR_OR_NULL (tmp_opp) || !tmp_opp->available)pr_err ("%s: Invalid parameters\n" , __func__);else
@@ -437,7 +440,7 @@ struct dev_pm_opp *dev_pm_opp_find_freq_exact(struct device *dev,struct device_opp *dev_opp;struct dev_pm_opp *temp_opp, *opp =ERR_PTR (-ERANGE ); -opp_rcu_lockdep_assert (); +opp_rcu_lockdep_assert (NULL ); dev_opp =_find_device_opp (dev);if (IS_ERR (dev_opp)) {
@@ -485,7 +488,7 @@ struct dev_pm_opp *dev_pm_opp_find_freq_ceil(struct device *dev,struct device_opp *dev_opp;struct dev_pm_opp *temp_opp, *opp =ERR_PTR (-ERANGE ); -opp_rcu_lockdep_assert (); +opp_rcu_lockdep_assert (NULL );if (!dev || !freq) {dev_err (dev,"%s: Invalid argument freq=%p\n" , __func__, freq);
@@ -535,7 +538,7 @@ struct dev_pm_opp *dev_pm_opp_find_freq_floor(struct device *dev,struct device_opp *dev_opp;struct dev_pm_opp *temp_opp, *opp =ERR_PTR (-ERANGE ); -opp_rcu_lockdep_assert (); +opp_rcu_lockdep_assert (NULL );if (!dev || !freq) {dev_err (dev,"%s: Invalid argument freq=%p\n" , __func__, freq);
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
PATCH v2 4/5 cpufreq-dt: Handle OPP voltage adjust events
In reply to Stephen BoydOn some SoCs the Adaptive Voltage Scaling (AVS) technique is employed to optimize the operating voltage of a device. At a given frequency, the hardware monitors dynamic factors and either makes a suggestion for how much to adjust a voltage for the current frequency, or it automatically adjusts the voltage without software intervention.
In the former case, an AVS driver will call dev_pm_opp_modify_voltage() and update the voltage for the particular OPP the CPUs are using. Add an OPP notifier to cpufreq-dt so that we can adjust the voltage of the CPU when AVS updates the OPP.
--- drivers/cpufreq/cpufreq-dt.c | 63 +++++++++++++++++++++++++++++++++++++++----- 1 file changed, 57 insertions(+), 6 deletions(-)
drivers/cpufreq/cpufreq-dt.c +57 -6
@@ -34,6 +34,9 @@ struct private_data {struct regulator *cpu_reg;struct thermal_cooling_device *cdev;unsigned int voltage_tolerance;/* in percentage */ +struct notifier_block opp_nb; +struct mutex lock; +unsigned long opp_freq; };static struct freq_attr *cpufreq_dt_attr[] = {
@@ -42,6 +45,33 @@ static struct freq_attr *cpufreq_dt_attr[] = { NULL, }; +static int opp_notifier (struct notifier_block *nb,unsigned long event, +void *data) +{ +struct dev_pm_opp *opp = data; + struct private_data *priv = container_of(nb, struct private_data, + opp_nb); +struct device *cpu_dev = priv->cpu_dev; +struct regulator *cpu_reg = priv->cpu_reg; +unsigned long volt, tol, freq; +int ret =0 ; + +if (event ==OPP_EVENT_ADJUST_VOLTAGE ) { + volt =dev_pm_opp_get_voltage (opp); + freq =dev_pm_opp_get_freq (opp); + tol = volt * priv->voltage_tolerance /100 ; + +mutex_lock (&priv->lock); +if (freq == priv->opp_freq) + ret =regulator_set_voltage_tol (cpu_reg, volt, tol); +mutex_unlock (&priv->lock); +if (ret) +dev_err (cpu_dev,"failed to scale voltage: %d\n" , ret); + } + +return notifier_from_errno (ret); +} +static int set_target (struct cpufreq_policy *policy,unsigned int index) {struct dev_pm_opp *opp;
@@ -53,6 +83,7 @@ static int set_target(struct cpufreq_policy *policy, unsigned int index)unsigned long volt =0 , volt_old =0 , tol =0 ;unsigned int old_freq, new_freq;long freq_Hz, freq_exact; +unsigned long opp_freq =0 ;int ret; freq_Hz =clk_round_rate (cpu_clk, freq_table[index].frequency *1000 );
@@ -63,8 +94,8 @@ static int set_target(struct cpufreq_policy *policy, unsigned int index) new_freq = freq_Hz /1000 ; old_freq =clk_get_rate (cpu_clk) /1000 ; +mutex_lock (&priv->lock);if (!IS_ERR (cpu_reg)) { -unsigned long opp_freq;rcu_read_lock (); opp =dev_pm_opp_find_freq_ceil (cpu_dev, &freq_Hz);
@@ -72,7 +103,8 @@ static int set_target(struct cpufreq_policy *policy, unsigned int index)rcu_read_unlock ();dev_err (cpu_dev,"failed to find OPP for %ld\n" , freq_Hz); -return PTR_ERR (opp); + ret =PTR_ERR (opp); + goto out; } volt =dev_pm_opp_get_voltage (opp); opp_freq =dev_pm_opp_get_freq (opp);
@@ -93,7 +125,7 @@ static int set_target(struct cpufreq_policy *policy, unsigned int index)if (ret) {dev_err (cpu_dev,"failed to scale voltage up: %d\n" , ret); -return ret; + goto out; } }
@@ -102,7 +134,7 @@ static int set_target(struct cpufreq_policy *policy, unsigned int index)dev_err (cpu_dev,"failed to set clock rate: %d\n" , ret);if (!IS_ERR (cpu_reg) && volt_old >0 )regulator_set_voltage_tol (cpu_reg, volt_old, tol); -return ret; + goto out; }/* scaling down? scale voltage after frequency */
@@ -112,9 +144,12 @@ static int set_target(struct cpufreq_policy *policy, unsigned int index)dev_err (cpu_dev,"failed to scale voltage down: %d\n" , ret);clk_set_rate (cpu_clk, old_freq *1000 ); + goto out; } } - + priv->opp_freq = opp_freq; +out: +mutex_unlock (&priv->lock);return ret; }
@@ -201,6 +236,7 @@ static int cpufreq_init(struct cpufreq_policy *policy)unsigned int transition_latency;bool need_update = false;int ret; +struct srcu_notifier_head *opp_srcu_head; ret =allocate_resources (policy->cpu, &cpu_dev, &cpu_reg, &cpu_clk);if (ret) {
@@ -277,6 +313,19 @@ static int cpufreq_init(struct cpufreq_policy *policy) goto out_free_opp; } +mutex_init (&priv->lock); + + opp_srcu_head =dev_pm_opp_get_notifier (cpu_dev); +if (IS_ERR (opp_srcu_head)) { + ret =PTR_ERR (opp_srcu_head); + goto out_free_priv; + } + + priv->opp_nb.notifier_call = opp_notifier; + ret =srcu_notifier_chain_register (opp_srcu_head, &priv->opp_nb); +if (ret) + goto out_free_priv; +of_property_read_u32 (np,"voltage-tolerance" , &priv->voltage_tolerance);if (!transition_latency)
@@ -326,7 +375,7 @@ static int cpufreq_init(struct cpufreq_policy *policy) ret =dev_pm_opp_init_cpufreq_table (cpu_dev, &freq_table);if (ret) {pr_err ("failed to init cpufreq table: %d\n" , ret); - goto out_free_priv; + goto out_unregister_nb; } priv->cpu_dev = cpu_dev;
@@ -365,6 +414,8 @@ static int cpufreq_init(struct cpufreq_policy *policy) out_free_cpufreq_table:dev_pm_opp_free_cpufreq_table (cpu_dev, &freq_table); +out_unregister_nb: +srcu_notifier_chain_unregister (opp_srcu_head, &priv->opp_nb); out_free_priv:kfree (priv); out_free_opp:
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
PATCH v2 5/5 power: avs: Add support for CPR (Core Power Reduction)
In reply to Stephen BoydCPR (Core Power Reduction) is a technology that reduces core power on a CPU or other device. It reads voltage settings in efuse from product test process as initial settings and populates OPPs for the device being "monitored". Each OPP corresponds to a "corner" that has a range of valid voltages for a particular frequency. While the device is running at a particular frequency, CPR monitors dynamic factors such as temperature, etc. and adjusts the voltage for that frequency accordingly to save power and meet silicon characteristic requirements.
This driver is based on work by others on codeaurora.org[1].
--- .../devicetree/bindings/power/avs/qcom,cpr.txt | 125 ++ drivers/power/avs/Kconfig | 15 + drivers/power/avs/Makefile | 1 + drivers/power/avs/qcom-cpr.c | 1999 ++++++++++++++++++++ 4 files changed, 2140 insertions(+) create mode 100644 Documentation/devicetree/bindings/power/avs/qcom,cpr.txt create mode 100644 drivers/power/avs/qcom-cpr.c
Documentation/devicetree/bindings/power/avs/qcom,cpr.txt new file +125 -0
@@ -0,0 +1,125 @@ +QCOM CPR (Core Power Reduction) + +CPR (Core Power Reduction) is a technology to reduce core power on a CPU +or other device. Each OPP of a device corresponds to a "corner" that has +a range of valid voltages for a particular frequency. While the device is +running at a particular frequency, CPR monitors dynamic factors such as +temperature, etc. and suggests adjustments to the voltage to save power +and meet silicon characteristic requirements. + +- compatible: + Usage: required + Value type: <string> + Definition: must be "qcom,cpr" + +- reg: + Usage: required + Value type: <prop-encoded-array> + Definition: base address and size of the rbcpr register region + +- interrupts: + Usage: required + Value type: <prop-encoded-array> + Definition: list of three interrupts in order of irq0, irq1, irq2 + +- acc-syscon: + Usage: optional + Value type: <phandle> + Definition: phandle to syscon for writing ACC settings + +- nvmem: + Usage: required + Value type: <phandle> + Definition: phandle to nvmem provider containing efuse settings + +- nvmem-names: + Usage: required + Value type: <string> + Definition: must be "qfprom" + +vdd-mx-supply = <&pm8916_l3>; + +- qcom,cpr-ref-clk: + Usage: required + Value type: <u32> + Definition: rate of reference clock in kHz + +- qcom,cpr-timer-delay-us: + Usage: required + Value type: <u32> + Definition: delay in uS for the timer interval + +- qcom,cpr-timer-cons-up: + Usage: required + Value type: <u32> + Definition: Consecutive number of timer intervals, or units of + qcom,cpr-timer-delay-us, that occur before issuing an up + interrupt + +- qcom,cpr-timer-cons-down: + Usage: required + Value type: <u32> + Definition: Consecutive number of timer intervals, or units of + qcom,cpr-timer-delay-us, that occur before issuing a down + interrupt + +- qcom,cpr-up-threshold: + Usage: optional + Value type: <u32> + Definition: The threshold for CPR to issue interrupt when error_steps + is greater than it when stepping up + +- qcom,cpr-down-threshold: + Usage: optional + Value type: <u32> + Definition: The threshold for CPR to issue interrdownt when error_steps + is greater than it when stepping down + +- qcom,cpr-down-threshold: + Usage: optional + Value type: <u32> + Definition: Idle clock cycles ring oscillator can be in + +- qcom,cpr-gcnt-us: + Usage: required + Value type: <u32> + Definition: The time for gate count in uS + +- qcom,vdd-apc-step-up-limit: + Usage: required + Value type: <u32> + Definition: Limit of vdd-apc-supply steps for scaling up + +- qcom,vdd-apc-step-down-limit: + Usage: required + Value type: <u32> + Definition: Limit of vdd-apc-supply steps for scaling down + +- qcom,cpr-cpus: + Usage: required + Value type: <prop-encoded-array> + Definition: List of CPUs that are being monitored + +Example: + + avs at b018000 { + compatible = "qcom,cpr"; + reg = <0xb018000 0x1000>; + interrupts = <0 15 1>, <0 16 1>, <0 17 1>; + vdd-mx-supply = <&pm8916_l3>; + acc-syscon = <&tcsr>; + nvmem = <&qfprom>; + nvmem-names = "qfprom"; + + qcom,cpr-ref-clk = <19200>; + qcom,cpr-timer-delay-us = <5000>; + qcom,cpr-timer-cons-up = <0>; + qcom,cpr-timer-cons-down = <2>; + qcom,cpr-up-threshold = <0>; + qcom,cpr-down-threshold = <2>; + qcom,cpr-idle-clocks = <15>; + qcom,cpr-gcnt-us = <1>; + qcom,vdd-apc-step-up-limit = <1>; + qcom,vdd-apc-step-down-limit = <1>; + qcom,cpr-cpus = <&CPU0 &CPU1 &CPU2 &CPU3>; + };
drivers/power/avs/Kconfig +15 -0
@@ -11,6 +11,21 @@ menuconfig POWER_AVS Say Y here to enable Adaptive Voltage Scaling class support. +config QCOM_CPR + tristate "QCOM Core Power Reduction (CPR) support" + depends on POWER_AVS + select PM_OPP + help + Say Y here to enable support for the CPR hardware found on Qualcomm + SoCs like MSM8916. + + This driver populates CPU OPPs tables and makes adjustments to the + tables based on feedback from the CPR hardware. If you want to do + CPUfrequency scaling say Y here. + + To compile this driver as a module, choose M here: the module will + be called qcom-cpr + config ROCKCHIP_IODOMAIN tristate "Rockchip IO domain support" depends on POWER_AVS && ARCH_ROCKCHIP && OF
drivers/power/avs/Makefile +1 -0
@@ -1,2 +1,3 @@ obj-$(CONFIG_POWER_AVS_OMAP) += smartreflex.o obj-$(CONFIG_ROCKCHIP_IODOMAIN) += rockchip-io-domain.o +obj-$(CONFIG_QCOM_CPR) += qcom-cpr.o
drivers/power/avs/qcom-cpr.c new file +1999 -0
1,999 lines not on this page · raw message
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
Mark Brown
maintainerQuoted 2 lines · Stephen Boyd
This patch series adds support for CPR on MSM8916. The first patch exposes a corner voting API to the CPR driver so that we can
You might want to expand on the acronym CPR since I'm reasonably sure
that you don't in fact mean the normal expansion cardiopulmonary
resuscitation.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 473 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20150919/e88aa77e/attachment.sig>
Mark Brown
maintainerRe: [PATCH v2 1/5] regulator: smd: Add floor and corner operations
In reply to Stephen BoydQuoted 1 line · Stephen Boyd
+EXPORT_SYMBOL(qcom_rpm_set_floor);
The entire regulator API is EXPORT_SYMBOL_GPL(), adding plainEXPORT_SYMBOL() exports is not good.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 473 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20150919/ed392835/attachment.sig>
Re: [PATCH v2 1/5] regulator: smd: Add floor and corner operations
Quoted 6 lines · Mark Brown
Quoted 1 line · Stephen Boyd
+EXPORT_SYMBOL(qcom_rpm_set_floor);
The entire regulator API is
EXPORT_SYMBOL_GPL(), adding plainEXPORT_SYMBOL()exports is not good.
That's fine. I can make it GPL.
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project
Re: [PATCH v2 2/5] PM / OPP: Support adjusting OPP voltages at runtime
In reply to Stephen BoydQuoted 10 lines · Stephen Boyd
/**
+ *dev_pm_opp_adjust_voltage()- helper to change the voltage of an OPP
+ * @dev: device for which we do this operation
+ * @freq: OPP frequency to adjust voltage of
+ * @u_volt: new OPP voltage
+ *
+ * Change the voltage of an OPP with an RCU operation.
+ *
+ * Return:-EINVALfor bad pointers,-ENOMEMif no memory available for the
+ * copy operation, returns 0 if no modifcation was done OR modification was
modification
Quoted 1 line
+ * successful.
Otherwise,
Signature · 1 line
viresh
Re: [PATCH v2 3/5] OPP: Allow notifiers to call dev_pm_opp_get_{voltage,freq} RCU-free
In reply to Stephen BoydQuoted 74 lines · Stephen Boyd
We pass the
dev_pm_oppstructure to OPP notifiers but the users of the notifier need to surround calls to dev_pm_opp_get_*() with RCU read locks to avoid lockdep warnings. The notifier is already called with the dev_opp's srcu lock held, so it should be safe to assume thedevm_pm_oppstructure is already protected inside the notifier. Update the lockdep check for this.Cc Krzysztof Kozlowski <[email protected]>Signed-off-by Stephen Boyd <[email protected]>--- drivers/base/power/opp.c | 19 +++++++++++-------- 1 file changed, 11 insertions(+), 8 deletions(-)drivers/base/power/opp.c +10
-7@@ -146,9 +146,10 @@ static LIST_HEAD(dev_opp_list);/* Lock to allow exclusive modification to the device and opp lists */ static DEFINE_MUTEX (dev_opp_list_lock); -#define opp_rcu_lockdep_assert () \ +#define opp_rcu_lockdep_assert (s) \ do { \ RCU_LOCKDEP_WARN(!rcu_read_lock_held() && \ + !(s && srcu_read_lock_held(s)) && \ !lockdep_is_held(&dev_opp_list_lock), \ "Missing rcu_read_lock() or " \ "dev_opp_list_lock protection"); \@@ -236,9 +237,10 @@ unsigned long dev_pm_opp_get_voltage(struct dev_pm_opp *opp)struct dev_pm_opp *tmp_opp;unsigned long v =0 ; -opp_rcu_lockdep_assert (); +opp_rcu_lockdep_assert (&opp->dev_opp->srcu_head.srcu); - tmp_opp =rcu_dereference (opp); + tmp_opp =srcu_dereference_check (opp, &opp->dev_opp->srcu_head.srcu, +rcu_read_lock_held ());if (IS_ERR_OR_NULL (tmp_opp) || !tmp_opp->available)pr_err ("%s: Invalid parameters\n" , __func__);else @@ -268,9 +270,10 @@ unsigned long dev_pm_opp_get_freq(struct dev_pm_opp *opp)struct dev_pm_opp *tmp_opp;unsigned long f =0 ; -opp_rcu_lockdep_assert (); +opp_rcu_lockdep_assert (&opp->dev_opp->srcu_head.srcu); - tmp_opp =rcu_dereference (opp); + tmp_opp =srcu_dereference_check (opp, &opp->dev_opp->srcu_head.srcu, +rcu_read_lock_held ());if (IS_ERR_OR_NULL (tmp_opp) || !tmp_opp->available)pr_err ("%s: Invalid parameters\n" , __func__);else @@ -437,7 +440,7 @@ struct dev_pm_opp *dev_pm_opp_find_freq_exact(struct device *dev,struct device_opp *dev_opp;struct dev_pm_opp *temp_opp, *opp =ERR_PTR (-ERANGE ); -opp_rcu_lockdep_assert (); +opp_rcu_lockdep_assert (NULL ); dev_opp =_find_device_opp (dev);if (IS_ERR (dev_opp)) {@@ -485,7 +488,7 @@ struct dev_pm_opp *dev_pm_opp_find_freq_ceil(struct device *dev,struct device_opp *dev_opp;struct dev_pm_opp *temp_opp, *opp =ERR_PTR (-ERANGE ); -opp_rcu_lockdep_assert (); +opp_rcu_lockdep_assert (NULL );if (!dev || !freq) {dev_err (dev,"%s: Invalid argument freq=%p\n" , __func__, freq);
drivers/base/power/opp.c struct dev_pm_opp *dev_pm_opp_find_freq_floor(struct device *dev, · line 544
Quoted 8 lines
drivers/base/power/opp.c +1
-1@@ -535,7 +538,7 @@ struct dev_pm_opp *dev_pm_opp_find_freq_floor(struct device *dev,struct device_opp *dev_opp;struct dev_pm_opp *temp_opp, *opp =ERR_PTR (-ERANGE ); -opp_rcu_lockdep_assert (); +opp_rcu_lockdep_assert (NULL );if (!dev || !freq) {dev_err (dev,"%s: Invalid argument freq=%p\n" , __func__, freq);
Signature · 1 line
viresh
Re: [PATCH v2 4/5] cpufreq-dt: Handle OPP voltage adjust events
In reply to Stephen BoydQuoted 164 lines · Stephen Boyd
On some SoCs the Adaptive Voltage Scaling (AVS) technique is employed to optimize the operating voltage of a device. At a given frequency, the hardware monitors dynamic factors and either makes a suggestion for how much to adjust a voltage for the current frequency, or it automatically adjusts the voltage without software intervention.
In the former case, an AVS driver will call
dev_pm_opp_modify_voltage()and update the voltage for the particular OPP the CPUs are using. Add an OPP notifier to cpufreq-dt so that we can adjust the voltage of the CPU when AVS updates the OPP.Signed-off-by Stephen Boyd <[email protected]>--- drivers/cpufreq/cpufreq-dt.c | 63 +++++++++++++++++++++++++++++++++++++++----- 1 file changed, 57 insertions(+), 6 deletions(-)drivers/cpufreq/cpufreq-dt.c +55
-6@@ -34,6 +34,9 @@ struct private_data {struct regulator *cpu_reg;struct thermal_cooling_device *cdev;unsigned int voltage_tolerance;/* in percentage */ +struct notifier_block opp_nb; +struct mutex lock; +unsigned long opp_freq; };static struct freq_attr *cpufreq_dt_attr[] = {@@ -42,6 +45,33 @@ static struct freq_attr *cpufreq_dt_attr[] = { NULL, }; +static int opp_notifier (struct notifier_block *nb,unsigned long event, +void *data) +{ +struct dev_pm_opp *opp = data; + struct private_data *priv = container_of(nb, struct private_data, + opp_nb); +struct device *cpu_dev = priv->cpu_dev; +struct regulator *cpu_reg = priv->cpu_reg; +unsigned long volt, tol, freq; +int ret =0 ; + +if (event ==OPP_EVENT_ADJUST_VOLTAGE ) { + volt =dev_pm_opp_get_voltage (opp); + freq =dev_pm_opp_get_freq (opp); + tol = volt * priv->voltage_tolerance /100 ; + +mutex_lock (&priv->lock); +if (freq == priv->opp_freq) + ret =regulator_set_voltage_tol (cpu_reg, volt, tol); +mutex_unlock (&priv->lock); +if (ret) +dev_err (cpu_dev,"failed to scale voltage: %d\n" , ret); + } + +return notifier_from_errno (ret); +} +static int set_target (struct cpufreq_policy *policy,unsigned int index) {struct dev_pm_opp *opp;@@ -53,6 +83,7 @@ static int set_target(struct cpufreq_policy *policy, unsigned int index)unsigned long volt =0 , volt_old =0 , tol =0 ;unsigned int old_freq, new_freq;long freq_Hz, freq_exact; +unsigned long opp_freq =0 ;int ret; freq_Hz =clk_round_rate (cpu_clk, freq_table[index].frequency *1000 );@@ -63,8 +94,8 @@ static int set_target(struct cpufreq_policy *policy, unsigned int index) new_freq = freq_Hz /1000 ; old_freq =clk_get_rate (cpu_clk) /1000 ; +mutex_lock (&priv->lock);if (!IS_ERR (cpu_reg)) { -unsigned long opp_freq;rcu_read_lock (); opp =dev_pm_opp_find_freq_ceil (cpu_dev, &freq_Hz);@@ -72,7 +103,8 @@ static int set_target(struct cpufreq_policy *policy, unsigned int index)rcu_read_unlock ();dev_err (cpu_dev,"failed to find OPP for %ld\n" , freq_Hz); -return PTR_ERR (opp); + ret =PTR_ERR (opp); + goto out; } volt =dev_pm_opp_get_voltage (opp); opp_freq =dev_pm_opp_get_freq (opp);@@ -93,7 +125,7 @@ static int set_target(struct cpufreq_policy *policy, unsigned int index)if (ret) {dev_err (cpu_dev,"failed to scale voltage up: %d\n" , ret); -return ret; + goto out; } }@@ -102,7 +134,7 @@ static int set_target(struct cpufreq_policy *policy, unsigned int index)dev_err (cpu_dev,"failed to set clock rate: %d\n" , ret);if (!IS_ERR (cpu_reg) && volt_old >0 )regulator_set_voltage_tol (cpu_reg, volt_old, tol); -return ret; + goto out; }/* scaling down? scale voltage after frequency */ @@ -112,9 +144,12 @@ static int set_target(struct cpufreq_policy *policy, unsigned int index)dev_err (cpu_dev,"failed to scale voltage down: %d\n" , ret);clk_set_rate (cpu_clk, old_freq *1000 ); + goto out; } } - + priv->opp_freq = opp_freq; +out: +mutex_unlock (&priv->lock);return ret; }@@ -201,6 +236,7 @@ static int cpufreq_init(struct cpufreq_policy *policy)unsigned int transition_latency;bool need_update = false;int ret; +struct srcu_notifier_head *opp_srcu_head; ret =allocate_resources (policy->cpu, &cpu_dev, &cpu_reg, &cpu_clk);if (ret) {@@ -277,6 +313,19 @@ static int cpufreq_init(struct cpufreq_policy *policy) goto out_free_opp; } +mutex_init (&priv->lock); + + opp_srcu_head =dev_pm_opp_get_notifier (cpu_dev); +if (IS_ERR (opp_srcu_head)) { + ret =PTR_ERR (opp_srcu_head); + goto out_free_priv; + } + + priv->opp_nb.notifier_call = opp_notifier; + ret =srcu_notifier_chain_register (opp_srcu_head, &priv->opp_nb); +if (ret) + goto out_free_priv; +of_property_read_u32 (np,"voltage-tolerance" , &priv->voltage_tolerance);if (!transition_latency)@@ -326,7 +375,7 @@ static int cpufreq_init(struct cpufreq_policy *policy) ret =dev_pm_opp_init_cpufreq_table (cpu_dev, &freq_table);if (ret) {pr_err ("failed to init cpufreq table: %d\n" , ret); - goto out_free_priv; + goto out_unregister_nb; } priv->cpu_dev = cpu_dev;
drivers/cpufreq/cpufreq-dt.c static int cpufreq_init(struct cpufreq_policy *policy) · line 421
Quoted 8 lines
drivers/cpufreq/cpufreq-dt.c +2
-0@@ -365,6 +414,8 @@ static int cpufreq_init(struct cpufreq_policy *policy) out_free_cpufreq_table:dev_pm_opp_free_cpufreq_table (cpu_dev, &freq_table); +out_unregister_nb: +srcu_notifier_chain_unregister (opp_srcu_head, &priv->opp_nb); out_free_priv:kfree (priv); out_free_opp:
Signature · 1 line
viresh
Re: [PATCH v2 2/5] PM / OPP: Support adjusting OPP voltages at runtime
In reply to Viresh KumarQuoted 12 lines · Viresh Kumar
Quoted 10 lines · Stephen Boyd
/**
+ *dev_pm_opp_adjust_voltage()- helper to change the voltage of an OPP
+ * @dev: device for which we do this operation
+ * @freq: OPP frequency to adjust voltage of
+ * @u_volt: new OPP voltage
+ *
+ * Change the voltage of an OPP with an RCU operation.
+ *
+ * Return:-EINVALfor bad pointers,-ENOMEMif no memory available for the
+ * copy operation, returns 0 if no modifcation was done OR modification was
Quoted 8 lines
modification
Quoted 1 line
+ * successful.
Otherwise,
Acked-by Viresh Kumar <[email protected]>
This typo is all over the file! I'll send a patch for that too.
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project
Quoted 23 lines · Stephen Boyd
This patch series adds support for CPR on MSM8916. The first patch exposes a corner voting API to the CPR driver so that we can change the corner for the MX regulator. If possible I would like to make this patch prettier, but I don't have any great ideas right now. The next patch adds support to adjust voltages in the OPP layer, and then hooks that up to cpufreq-dt so that we can adjust the voltage in response to what CPR tells us to do. I've also thrown in a patch to make RCU lockdep warnings go away, but I'm not sure if it's right. There's still work to do.
The final patch adds the CPR driver. This still has some rough edges. With the OPPv2 bindings I'm thinking of moving the frequency tables into DT and adding a custom vendor property to describe which fuse corner to use for each frequency.
Once you have these patches in place along with a CPU clock driver you can eanble enable cpufreq-dt and add the cpufreq-dt device (maybe the CPR driver should add the cpufreq-dt device?) and you'll see interrupts for CPR and OPP voltage adjustments triggering CPUfreq to modify voltages.
Andy Gross (1):
regulator: smd: Add floor and corner operations
Quoted 7 lines
Stephen Boyd (4):
PM / OPP: Support adjusting OPP voltages at runtime
OPP: Allow notifiers to call dev_pm_opp_get_{voltage,freq} RCU-free
cpufreq-dt: Handle OPP voltage adjust events
power: avs: Add support for CPR (Core Power Reduction)From the responses so far I gather there will be a v3?
Thanks,
Rafael
Quoted 24 lines · Rafael J. Wysocki
Quoted 23 lines · Stephen Boyd
This patch series adds support for CPR on MSM8916. The first patch exposes a corner voting API to the CPR driver so that we can change the corner for the MX regulator. If possible I would like to make this patch prettier, but I don't have any great ideas right now. The next patch adds support to adjust voltages in the OPP layer, and then hooks that up to cpufreq-dt so that we can adjust the voltage in response to what CPR tells us to do. I've also thrown in a patch to make RCU lockdep warnings go away, but I'm not sure if it's right. There's still work to do.
The final patch adds the CPR driver. This still has some rough edges. With the OPPv2 bindings I'm thinking of moving the frequency tables into DT and adding a custom vendor property to describe which fuse corner to use for each frequency.
Once you have these patches in place along with a CPU clock driver you can eanble enable cpufreq-dt and add the cpufreq-dt device (maybe the CPR driver should add the cpufreq-dt device?) and you'll see interrupts for CPR and OPP voltage adjustments triggering CPUfreq to modify voltages.
Andy Gross (1):
regulator: smd: Add floor and corner operations
Quoted 8 lines
Quoted 5 lines
Stephen Boyd (4):
PM / OPP: Support adjusting OPP voltages at runtime
OPP: Allow notifiers to call dev_pm_opp_get_{voltage,freq} RCU-free
cpufreq-dt: Handle OPP voltage adjust events
power: avs: Add support for CPR (Core Power Reduction)From the responses so far I gather there will be a v3?
Yes. Definitely.
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project