Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit a40741e2 authored by Mahesh Sivasubramanian's avatar Mahesh Sivasubramanian Committed by Lina Iyer
Browse files

drivers: soc: qcom: Snapshot of idle/sleep driver as of msm-4.4 



This is a snapshot of the Sleep driver and realted functionality as of
'commit b67bb93f565e ("power: reset: Do not disable SDI during reset")'
on msm-4.4 branch.

Moved Energy Awareness driver and Idle stats driver from
drivers/power/qcom to drivers/soc/qcom.

Remove lpm-workarounds file that no longer is applicable on newer
branches.

Updated the patch to fix some checkpatch issues
	- Use WARN
	- White space issues
	- Replace S_IRUGO etc
	- Remove return on void functions
	- Add function arguments
	- unsigned -> unsigned int
	- Add comments around wmb, mb.
	- Fix comment styles
	- kzalloc to kcalloc
	- Updated MSM_PM Kconfig help
	- Remove references to spm.h and file

Change-Id: I98fec26849898c5c66abbb1b094439780c23964d
Signed-off-by: default avatarMahesh Sivasubramanian <msivasub@codeaurora.org>
Signed-off-by: default avatarLina Iyer <ilina@codeaurora.org>
parent d4dfec3a

Documentation/devicetree/bindings/arm/msm/lpm-levels.txt

0 → 100644
+325 −0
Original line number Diff line number Diff line 
* Low Power Management Levels



The application processor in MSM can do a variety of C-States for low power

management. The LPM module performs the System low power modes based on

the latency/residency information of the individual CPUs and clusters.



LPM-levels defines a hierarchy of low power modes that a cluster and

clusters/cpus within that cluster can enter. The bottom hierarchy level

represents the low power modes that a CPU can enter. The CPU low power nodes

are associated with a cluster that defines the low power modes that a cluster

can enter. For system involving a hierarchy of clusters, the cluster low power

modes can be contained within another cluster.



[Top Level Node]

Required properties:



- compatible: "qcom,lpm-levels"



[Node bindings for qcom,pm-cluster]

 Required properties:

	- reg - The numeric cluster id

	- label: Identifies the cluster name. The name will be

	used when reporting the stats for each low power mode.

	- qcom,spm-device-names: List of  SPM device names which control the

	low power modes for this driver. The lpm driver uses the device name

	to obtain a handle to the SPM driver that controls the cluster's low

	power mode. This is only required if "qcom,use-psci" is not defined.

	- qcom,default-level: The default low power level that a cluster is

	programmed. The SPM of the corresponding device is configured at this

	low power mode by default.



	qcom,pm-cluster contains qcom,pm-cluster-level nodes which identify

	the various low power modes that the cluster can enter. The

	qcom,pm-cluster node should also include another cluster node or a cpu

	node that defines their respective low power modes.



[Node bindings for qcom,pm-cluster-level]

 Required properties:

	- reg: The numeric cluster level id

	- label: Name to identify the low power mode in stats

	module.

	- qcom,spm-<device-name>-mode: For each SPM device defined in

	qcom,spm-devices-names, a corresponding entry identifying the low

	power mode is expected. For example, the qcom,pm-cluster node contains

	a SPM device by name "l2" then the cluster level should contain a

	qcom,spm-l2-mode.  When a cluster level is chosen ,the SPM device is

	programmed with its

	corresponding low power mode. The accepted values for this property

	are:

		- "active"

		- "wfi"

		- "retention"

		- "gdhs"

		- "pc"

		- "fpc"

	- qcom,min-child-idx: The minimum level that a child CPU should be in

	before this level can be chosen. This property is required for all

        non-default level.

	- qcom,latency-us: The latency in handling the interrupt if this level

	was chosen, in uSec

	- qcom,ss-power: The steady state power expelled when the processor is

	in this level in mWatts

	- qcom,energy-overhead: The energy used up in entering and exiting

	this level in mWatts.uSec

	- qcom,time-overhead: The time spent in entering and exiting this

	level in uS

 Optional properties:

	- qcom,notify-rpm: When set, the driver flushes the RPM sleep set and

	configures the virtual MPM driver in prepration for a RPM assisted

	sleep.

	- qcom,last-level - When set, the cluster level is applied only when

	there is 1 online core.

	- qcom,disable-dynamic-int-routing: When set disables the dynamic

	routing of rpm-smd and mpm interrupts to next wake up core.

	- qcom,use-psci: This boolean property allows the LPM modules to

	terminate in PSCI to configure SPM for low power modes.

	- qcom,psci-mode-shift: The property is used to determine with bit

	location of the cluster mode in the composite state ID used to define

	cluster low power modes in PSCI v1.0. Required only if qcom,use-psci

	is defined at the lpm-levels root node.

	- qcom,psci-mode-mask: The property is used to determine with bit

	mask of the cluster mode in the composite state ID used to define

	cluster low power modes in PSCI v1.0. Required only if qcom,use-psci

	is defined at the lpm-levels root node.

	- qcom,psci-mode: ID to be passed into the PSCI firmware. Required

	only if qcom,use-psci is defined at the lpm-levels root node.

	- qcom,is-reset: This boolean property will tell whether

	cluster level need power management notifications to be sent out

	or not for the drivers to prepare for cluster collapse.

	- qcom,hyp-psci: This property is used to determine if the cpu

        enters the low power mode within hypervisor.

	- qcom,reset-level: This property is used to determine in this

	low power mode only control logic power collapse happens or memory

	logic power collapse aswell happens or retention state.

	The accepted values for this property are:

		"LPM_RESET_LVL_NONE" - No power collapse

		"LPM_RESET_LVL_RET"  - Retention state

		"LPM_RESET_LVL_GDHS" - Only control logic power collapse (GDHS)

		"LPM_RESET_LVL_PC" - Control logic and memory logic

					power collapse (PC)



[Node bindings for qcom,pm-cpu]

qcom,pm-cpu contains the low power modes that a cpu could enter and the CPUs

that share the parameters.It contains the following properties.

	- qcom,cpu: List of CPU phandles to identify the CPUs associated with

	this cluster.

	- qcom,pm-cpu-levels: The different low power modes that a CPU could

	enter. The following section explains the required properties of this

	node.



[Node bindings for qcom,pm-cpu-levels]

 Required properties:

	- reg: The numeric cpu level id

	- qcom,spm-cpu-mode: The sleep mode of the processor, values for the

	property are:

		"wfi" - Wait for Interrupt

		"retention" - Retention

		"standalone_pc" - Standalone power collapse

		"pc" - Power Collapse

	- qcom,latency-us: The latency in handling the interrupt if this level

	was chosen, in uSec

	- qcom,ss-power: The steady state power expelled when the processor is

	in this level in mWatts

	- qcom,energy-overhead: The energy used up in entering and exiting

	this level in mWatts.uSec

	- qcom,time-overhead: The time spent in entering and exiting this

	level in uS

	- qcom,use-broadcast-timer: Indicates that the timer gets reset during

	power collapse and the cpu relies on Broadcast timer for scheduled

	wakeups. Required only for states where the CPUs internal timer state

	is lost.



 Optional properties:

	- qcom,psci-mode-shift: Same as cluster level fields.

	- qcom,psci-mode-mask: Same as cluster level fields.

	- qcom,psci-cpu-mode: ID to be passed into PSCI firmware.

	- qcom,jtag-save-restore: A boolean specifying jtag registers save and restore

	required are not.

	- qcom,is-reset: This boolean property maps to "power state" bit in PSCI

	state_id configuration. This property will tell whether CPU get reset for

	a particular LPM or not. This property will also be used to notify the

	drivers in case of cpu reset.



[Example dts]



qcom,lpm-levels {

	#address-cells = <1>;

	#size-cells = <0>;

	compatible = "qcom,lpm-levels";



	qcom,pm-cluster@0 {

		#address-cells = <1>;

		#size-cells = <0>;

		reg = <0>;

		label = "system";

		qcom,spm-device-names = "cci";

		qcom,default-level = <0>;



		qcom,pm-cluster-level@0{

			reg = <0>;

			label = "system-cci-retention";

			qcom,spm-cci-mode = "retention";

			qcom,latency-us = <100>;

			qcom,ss-power = <1000>;

			qcom,energy-overhead = <300000>;

			qcom,time-overhead = <100>;

		};



		qcom,pm-cluster-level@2{

			reg = <1>;

			label = "system-cci-pc";

			qcom,spm-cci-mode = "pc";

			qcom,latency-us = <30000>;

			qcom,ss-power = <83>;

			qcom,energy-overhead = <2274420>;

			qcom,time-overhead = <6605>;

			qcom,min-child-idx = <1>;

			qcom,notify-rpm;

		};



		qcom,pm-cluster@0{

			#address-cells = <1>;

			#size-cells = <0>;

			reg = <0>;

			label = "a53";

			qcom,spm-device-names = "l2";

			qcom,default-level=<0>;



			qcom,pm-cluster-level@0{

				reg = <0>;

				label = "a53-l2-retention";

				qcom,spm-l2-mode = "retention";

				qcom,latency-us = <100>;

				qcom,ss-power = <1000>;

				qcom,energy-overhead = <300000>;

				qcom,time-overhead = <100>;

			};



			qcom,pm-cluster-level@1{

				reg = <1>;

				label = "a53-l2-pc";

				qcom,spm-l2-mode = "pc";

				qcom,latency-us = <30000>;

				qcom,ss-power = <83>;

				qcom,energy-overhead = <2274420>;

				qcom,time-overhead = <6605>;

				qcom,min-child-idx = <3>;

			};



			qcom,pm-cpu {

				#address-cells = <1>;

				#size-cells = <0>;

				qcom,cpu = <&CPU0 &CPU1 &CPU2 &CPU3>;

				qcom,pm-cpu-level@0 {

					reg = <0>;

					qcom,spm-cpu-mode = "wfi";

					qcom,latency-us = <1>;

					qcom,ss-power = <715>;

					qcom,energy-overhead = <17700>;

					qcom,time-overhead = <2>;

				};



				qcom,pm-cpu-level@1 {

					reg = <1>;

					qcom,spm-cpu-mode = "retention";

					qcom,latency-us = <35>;

					qcom,ss-power = <542>;

					qcom,energy-overhead = <34920>;

					qcom,time-overhead = <40>;

				};



				qcom,pm-cpu-level@2 {

					reg = <2>;

					qcom,spm-cpu-mode = "standalone_pc";

					qcom,latency-us = <300>;

					qcom,ss-power = <476>;

					qcom,energy-overhead = <225300>;

					qcom,time-overhead = <350>;

				};



				qcom,pm-cpu-level@3 {

					reg = <3>;

					qcom,spm-cpu-mode = "pc";

					qcom,latency-us = <500>;

					qcom,ss-power = <163>;

					qcom,energy-overhead = <577736>;

					qcom,time-overhead = <1000>;

				};

			};

		};



		qcom,pm-cluster@1{

			#address-cells = <1>;

			#size-cells = <0>;

			reg = <1>;

			label = "a57";

			qcom,spm-device-names = "l2";

			qcom,default-level=<0>;



			qcom,pm-cluster-level@0{

				reg = <0>;

				label = "a57-l2-retention";

				qcom,spm-l2-mode = "retention";

				qcom,latency-us = <100>;

				qcom,ss-power = <1000>;

				qcom,energy-overhead = <300000>;

				qcom,time-overhead = <100>;

			};



			qcom,pm-cluster-level@2{

				reg = <1>;

				label = "a57-l2-pc";

				qcom,spm-l2-mode = "pc";

				qcom,latency-us = <30000>;

				qcom,ss-power = <83>;

				qcom,energy-overhead = <2274420>;

				qcom,time-overhead = <6605>;

				qcom,min-child-idx = <3>;

			};



			qcom,pm-cpu {

				#address-cells = <1>;

				#size-cells = <0>;

				qcom,cpu = <&CPU4 &CPU5 &CPU6 &CPU7>;

				qcom,pm-cpu-level@0 {

					reg = <0>;

					qcom,spm-cpu-mode = "wfi";

					qcom,latency-us = <1>;

					qcom,ss-power = <715>;

					qcom,energy-overhead = <17700>;

					qcom,time-overhead = <2>;

				};



				qcom,pm-cpu-level@1 {

					reg = <1>;

					qcom,spm-cpu-mode = "retention";

					qcom,latency-us = <35>;

					qcom,ss-power = <542>;

					qcom,energy-overhead = <34920>;

					qcom,time-overhead = <40>;

				};



				qcom,pm-cpu-level@2 {

					reg = <2>;

					qcom,spm-cpu-mode = "standalone_pc";

					qcom,latency-us = <300>;

					qcom,ss-power = <476>;

					qcom,energy-overhead = <225300>;

					qcom,time-overhead = <350>;

				};



				qcom,pm-cpu-level@3 {

					reg = <3>;

					qcom,spm-cpu-mode = "pc";

					qcom,latency-us = <500>;

					qcom,ss-power = <163>;

					qcom,energy-overhead = <577736>;

					qcom,time-overhead = <1000>;

				};

			};

		};

	};





};

drivers/cpuidle/Makefile

+1 −0
Original line number Diff line number Diff line
@@ -29,3 +29,4 @@ obj-$(CONFIG_MIPS_CPS_CPUIDLE) += cpuidle-cps.o 
# POWERPC drivers

obj-$(CONFIG_PSERIES_CPUIDLE)		+= cpuidle-pseries.o

obj-$(CONFIG_POWERNV_CPUIDLE)		+= cpuidle-powernv.o

obj-$(CONFIG_MSM_PM) += lpm-levels.o lpm-levels-of.o

drivers/cpuidle/lpm-levels-of.c

0 → 100644
+929 −0

File added.

Preview size limit exceeded, changes collapsed.

drivers/cpuidle/lpm-levels.c

0 → 100644
+1736 −0

File added.

Preview size limit exceeded, changes collapsed.

drivers/cpuidle/lpm-levels.h

0 → 100644
+153 −0
Original line number Diff line number Diff line 
/* Copyright (c) 2014-2018, 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.

 */



#include <soc/qcom/pm.h>



#define NR_LPM_LEVELS 8

#define MAXSAMPLES 5

#define CLUST_SMPL_INVLD_TIME 40000



struct lpm_lookup_table {

	uint32_t modes;

	const char *mode_name;

};



struct power_params {

	uint32_t latency_us;		/* Enter + Exit latency */

	uint32_t ss_power;		/* Steady state power */

	uint32_t energy_overhead;	/* Enter + exit over head */

	uint32_t time_overhead_us;	/* Enter + exit overhead */

	uint32_t residencies[NR_LPM_LEVELS];

	uint32_t min_residency;

	uint32_t max_residency;

};



struct lpm_cpu_level {

	const char *name;

	enum msm_pm_sleep_mode mode;

	bool use_bc_timer;

	struct power_params pwr;

	unsigned int psci_id;

	bool is_reset;

	bool jtag_save_restore;

	bool hyp_psci;

	int reset_level;

};



struct lpm_cpu {

	struct list_head list;

	struct cpumask related_cpus;

	struct lpm_cpu_level levels[NR_LPM_LEVELS];

	int nlevels;

	unsigned int psci_mode_shift;

	unsigned int psci_mode_mask;

	struct cpuidle_driver *drv;

	struct lpm_cluster *parent;

};



struct lpm_level_avail {

	bool idle_enabled;

	bool suspend_enabled;

	uint32_t latency_us;

	struct kobject *kobj;

	struct kobj_attribute idle_enabled_attr;

	struct kobj_attribute suspend_enabled_attr;

	struct kobj_attribute latency_attr;

	void *data;

	int idx;

	bool cpu_node;

};



struct lpm_cluster_level {

	const char *level_name;

	int *mode;			/* SPM mode to enter */

	int min_child_level;

	struct cpumask num_cpu_votes;

	struct power_params pwr;

	bool notify_rpm;

	bool sync_level;

	struct lpm_level_avail available;

	unsigned int psci_id;

	bool is_reset;

	int reset_level;

};



struct cluster_history {

	uint32_t resi[MAXSAMPLES];

	int mode[MAXSAMPLES];

	int64_t stime[MAXSAMPLES];

	uint32_t hptr;

	uint32_t hinvalid;

	uint32_t htmr_wkup;

	uint64_t entry_time;

	int entry_idx;

	int nsamp;

	int flag;

};



struct lpm_cluster {

	struct list_head list;

	struct list_head child;

	const char *cluster_name;

	const char **name;

	unsigned long aff_level; /* Affinity level of the node */

	int ndevices;

	struct lpm_cluster_level levels[NR_LPM_LEVELS];

	int nlevels;

	int min_child_level;

	int default_level;

	int last_level;

	struct list_head cpu;

	spinlock_t sync_lock;

	struct cpumask child_cpus;

	struct cpumask num_children_in_sync;

	struct lpm_cluster *parent;

	struct lpm_stats *stats;

	unsigned int psci_mode_shift;

	unsigned int psci_mode_mask;

	struct cluster_history history;

	struct hrtimer histtimer;

};



void lpm_suspend_wake_time(uint64_t wakeup_time);



struct lpm_cluster *lpm_of_parse_cluster(struct platform_device *pdev);

void free_cluster_node(struct lpm_cluster *cluster);

void cluster_dt_walkthrough(struct lpm_cluster *cluster);



int create_cluster_lvl_nodes(struct lpm_cluster *p, struct kobject *kobj);

bool lpm_cpu_mode_allow(unsigned int cpu,

		unsigned int mode, bool from_idle);

bool lpm_cluster_mode_allow(struct lpm_cluster *cluster,

		unsigned int mode, bool from_idle);

uint32_t *get_per_cpu_max_residency(int cpu);

uint32_t *get_per_cpu_min_residency(int cpu);

extern struct lpm_cluster *lpm_root_node;



#if CONFIG_SMP

extern DEFINE_PER_CPU(bool, pending_ipi);

static inline bool is_IPI_pending(const struct cpumask *mask)

{

	unsigned int cpu;



	for_each_cpu(cpu, mask) {

		if per_cpu(pending_ipi, cpu)

			return true;

	}

	return false;

}

#else

static inline bool is_IPI_pending(const struct cpumask *mask)

{

	return false;

}

#endif