M7350/kernel/drivers/power/qcom/msm-pm.c
2024-09-09 08:57:42 +00:00

890 lines
20 KiB
C

/* Copyright (c) 2010-2015, 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.
*
*/
#include <linux/debugfs.h>
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/clk.h>
#include <linux/clkdev.h>
#include <linux/io.h>
#include <linux/ktime.h>
#include <linux/smp.h>
#include <linux/tick.h>
#include <linux/delay.h>
#include <linux/platform_device.h>
#include <linux/of_platform.h>
#include <linux/of_address.h>
#include <linux/msm-bus.h>
#include <linux/uaccess.h>
#include <linux/dma-mapping.h>
#include <soc/qcom/spm.h>
#include <soc/qcom/pm.h>
#include <soc/qcom/scm.h>
#include <soc/qcom/scm-boot.h>
#include <asm/suspend.h>
#include <asm/cacheflush.h>
#include <asm/cputype.h>
#include <asm/system_misc.h>
#ifdef CONFIG_VFP
#include <asm/vfp.h>
#endif
#include <soc/qcom/jtag.h>
#include "idle.h"
#include "pm-boot.h"
#define SCM_CMD_TERMINATE_PC (0x2)
#define SCM_CMD_CORE_HOTPLUGGED (0x10)
#define SCM_FLUSH_FLAG_MASK (0x3)
#define SCLK_HZ (32768)
#define MAX_BUF_SIZE 1024
static int msm_pm_debug_mask = 1;
module_param_named(
debug_mask, msm_pm_debug_mask, int, S_IRUGO | S_IWUSR | S_IWGRP
);
enum {
MSM_PM_DEBUG_SUSPEND = BIT(0),
MSM_PM_DEBUG_POWER_COLLAPSE = BIT(1),
MSM_PM_DEBUG_SUSPEND_LIMITS = BIT(2),
MSM_PM_DEBUG_CLOCK = BIT(3),
MSM_PM_DEBUG_RESET_VECTOR = BIT(4),
MSM_PM_DEBUG_IDLE = BIT(5),
MSM_PM_DEBUG_IDLE_LIMITS = BIT(6),
MSM_PM_DEBUG_HOTPLUG = BIT(7),
};
enum msm_pc_count_offsets {
MSM_PC_ENTRY_COUNTER,
MSM_PC_EXIT_COUNTER,
MSM_PC_FALLTHRU_COUNTER,
MSM_PC_UNUSED,
MSM_PC_NUM_COUNTERS,
};
static bool msm_pm_ldo_retention_enabled = true;
static bool msm_pm_tz_flushes_cache;
static bool msm_pm_ret_no_pll_switch;
static bool msm_no_ramp_down_pc;
static struct msm_pm_sleep_status_data *msm_pm_slp_sts;
DEFINE_PER_CPU(struct clk *, cpu_clks);
static struct clk *l2_clk;
static long *msm_pc_debug_counters;
static cpumask_t retention_cpus;
static DEFINE_SPINLOCK(retention_lock);
static bool msm_pm_is_L1_writeback(void)
{
u32 cache_id = 0;
#if defined(CONFIG_CPU_V7)
u32 sel = 0;
asm volatile ("mcr p15, 2, %[ccselr], c0, c0, 0\n\t"
"isb\n\t"
"mrc p15, 1, %[ccsidr], c0, c0, 0\n\t"
:[ccsidr]"=r" (cache_id)
:[ccselr]"r" (sel)
);
return cache_id & BIT(30);
#elif defined(CONFIG_ARM64)
u32 sel = 0;
asm volatile("msr csselr_el1, %[ccselr]\n\t"
"isb\n\t"
"mrs %[ccsidr],ccsidr_el1\n\t"
:[ccsidr]"=r" (cache_id)
:[ccselr]"r" (sel)
);
return cache_id & BIT(30);
#else
#error No valid CPU arch selected
#endif
}
static bool msm_pm_swfi(bool from_idle)
{
msm_arch_idle();
return true;
}
static bool msm_pm_retention(bool from_idle)
{
int ret = 0;
unsigned int cpu = smp_processor_id();
struct clk *cpu_clk = per_cpu(cpu_clks, cpu);
spin_lock(&retention_lock);
if (!msm_pm_ldo_retention_enabled)
goto bailout;
cpumask_set_cpu(cpu, &retention_cpus);
spin_unlock(&retention_lock);
if (!msm_pm_ret_no_pll_switch)
clk_disable(cpu_clk);
ret = msm_spm_set_low_power_mode(MSM_SPM_MODE_RETENTION, false);
WARN_ON(ret);
msm_arch_idle();
ret = msm_spm_set_low_power_mode(MSM_SPM_MODE_CLOCK_GATING, false);
WARN_ON(ret);
if (!msm_pm_ret_no_pll_switch)
if (clk_enable(cpu_clk))
pr_err("%s(): Error restore cpu clk\n", __func__);
spin_lock(&retention_lock);
cpumask_clear_cpu(cpu, &retention_cpus);
bailout:
spin_unlock(&retention_lock);
return true;
}
static inline void msm_pc_inc_debug_count(uint32_t cpu,
enum msm_pc_count_offsets offset)
{
int cntr_offset;
uint32_t cluster_id = MPIDR_AFFINITY_LEVEL(cpu_logical_map(cpu), 1);
uint32_t cpu_id = MPIDR_AFFINITY_LEVEL(cpu_logical_map(cpu), 0);
if (cluster_id >= MAX_NUM_CLUSTER || cpu_id >= MAX_CPUS_PER_CLUSTER)
BUG();
cntr_offset = (cluster_id * MAX_CPUS_PER_CLUSTER * MSM_PC_NUM_COUNTERS)
+ (cpu_id * MSM_PC_NUM_COUNTERS) + offset;
if (!msm_pc_debug_counters)
return;
msm_pc_debug_counters[cntr_offset]++;
}
static bool msm_pm_pc_hotplug(void)
{
uint32_t cpu = smp_processor_id();
enum msm_pm_l2_scm_flag flag;
struct scm_desc desc;
flag = lpm_cpu_pre_pc_cb(cpu);
if (!msm_pm_tz_flushes_cache) {
if (flag == MSM_SCM_L2_OFF)
flush_cache_all();
else if (msm_pm_is_L1_writeback())
flush_cache_louis();
}
msm_pc_inc_debug_count(cpu, MSM_PC_ENTRY_COUNTER);
if (is_scm_armv8()) {
desc.args[0] = SCM_CMD_CORE_HOTPLUGGED |
(flag & SCM_FLUSH_FLAG_MASK);
desc.arginfo = SCM_ARGS(1);
scm_call2_atomic(SCM_SIP_FNID(SCM_SVC_BOOT,
SCM_CMD_TERMINATE_PC), &desc);
} else {
scm_call_atomic1(SCM_SVC_BOOT, SCM_CMD_TERMINATE_PC,
SCM_CMD_CORE_HOTPLUGGED | (flag & SCM_FLUSH_FLAG_MASK));
}
/* Should not return here */
msm_pc_inc_debug_count(cpu, MSM_PC_FALLTHRU_COUNTER);
return 0;
}
static bool msm_pm_fastpc(bool from_idle)
{
int ret = 0;
unsigned int cpu = smp_processor_id();
ret = msm_spm_set_low_power_mode(MSM_SPM_MODE_FASTPC, false);
WARN_ON(ret);
if (from_idle || cpu_online(cpu))
msm_arch_idle();
else
msm_pm_pc_hotplug();
ret = msm_spm_set_low_power_mode(MSM_SPM_MODE_CLOCK_GATING, false);
WARN_ON(ret);
return true;
}
int msm_pm_collapse(unsigned long unused)
{
uint32_t cpu = smp_processor_id();
enum msm_pm_l2_scm_flag flag;
struct scm_desc desc;
flag = lpm_cpu_pre_pc_cb(cpu);
if (!msm_pm_tz_flushes_cache) {
if (flag == MSM_SCM_L2_OFF)
flush_cache_all();
else if (msm_pm_is_L1_writeback())
flush_cache_louis();
}
msm_pc_inc_debug_count(cpu, MSM_PC_ENTRY_COUNTER);
if (is_scm_armv8()) {
desc.args[0] = flag;
desc.arginfo = SCM_ARGS(1);
scm_call2_atomic(SCM_SIP_FNID(SCM_SVC_BOOT,
SCM_CMD_TERMINATE_PC), &desc);
} else {
scm_call_atomic1(SCM_SVC_BOOT, SCM_CMD_TERMINATE_PC, flag);
}
msm_pc_inc_debug_count(cpu, MSM_PC_FALLTHRU_COUNTER);
return 0;
}
EXPORT_SYMBOL(msm_pm_collapse);
static bool __ref msm_pm_spm_power_collapse(
unsigned int cpu, int mode, bool from_idle, bool notify_rpm)
{
void *entry;
bool collapsed = 0;
int ret;
bool save_cpu_regs = (cpu_online(cpu) || from_idle);
if (MSM_PM_DEBUG_POWER_COLLAPSE & msm_pm_debug_mask)
pr_info("CPU%u: %s: notify_rpm %d\n",
cpu, __func__, (int) notify_rpm);
ret = msm_spm_set_low_power_mode(mode, notify_rpm);
WARN_ON(ret);
entry = save_cpu_regs ? cpu_resume : msm_secondary_startup;
msm_pm_boot_config_before_pc(cpu, virt_to_phys(entry));
if (MSM_PM_DEBUG_RESET_VECTOR & msm_pm_debug_mask)
pr_info("CPU%u: %s: program vector to %p\n",
cpu, __func__, entry);
msm_jtag_save_state();
collapsed = save_cpu_regs ?
!__cpu_suspend(0, msm_pm_collapse) : msm_pm_pc_hotplug();
msm_jtag_restore_state();
if (collapsed)
local_fiq_enable();
msm_pm_boot_config_after_pc(cpu);
if (MSM_PM_DEBUG_POWER_COLLAPSE & msm_pm_debug_mask)
pr_info("CPU%u: %s: msm_pm_collapse returned, collapsed %d\n",
cpu, __func__, collapsed);
ret = msm_spm_set_low_power_mode(MSM_SPM_MODE_CLOCK_GATING, false);
WARN_ON(ret);
return collapsed;
}
static bool msm_pm_power_collapse_standalone(
bool from_idle)
{
unsigned int cpu = smp_processor_id();
bool collapsed;
collapsed = msm_pm_spm_power_collapse(cpu,
MSM_SPM_MODE_STANDALONE_POWER_COLLAPSE,
from_idle, false);
return collapsed;
}
static int ramp_down_last_cpu(int cpu)
{
struct clk *cpu_clk = per_cpu(cpu_clks, cpu);
int ret = 0;
clk_disable(cpu_clk);
clk_disable(l2_clk);
return ret;
}
static int ramp_up_first_cpu(int cpu, int saved_rate)
{
struct clk *cpu_clk = per_cpu(cpu_clks, cpu);
int rc = 0;
if (MSM_PM_DEBUG_CLOCK & msm_pm_debug_mask)
pr_info("CPU%u: %s: restore clock rate\n",
cpu, __func__);
clk_enable(l2_clk);
if (cpu_clk) {
int ret = clk_enable(cpu_clk);
if (ret) {
pr_err("%s(): Error restoring cpu clk\n",
__func__);
return ret;
}
}
return rc;
}
static bool msm_pm_power_collapse(bool from_idle)
{
unsigned int cpu = smp_processor_id();
unsigned long saved_acpuclk_rate = 0;
bool collapsed;
if (MSM_PM_DEBUG_POWER_COLLAPSE & msm_pm_debug_mask)
pr_info("CPU%u: %s: idle %d\n",
cpu, __func__, (int)from_idle);
if (MSM_PM_DEBUG_POWER_COLLAPSE & msm_pm_debug_mask)
pr_info("CPU%u: %s: pre power down\n", cpu, __func__);
if (cpu_online(cpu) && !msm_no_ramp_down_pc)
saved_acpuclk_rate = ramp_down_last_cpu(cpu);
collapsed = msm_pm_spm_power_collapse(cpu, MSM_SPM_MODE_POWER_COLLAPSE,
from_idle, true);
if (cpu_online(cpu) && !msm_no_ramp_down_pc)
ramp_up_first_cpu(cpu, saved_acpuclk_rate);
if (MSM_PM_DEBUG_POWER_COLLAPSE & msm_pm_debug_mask)
pr_info("CPU%u: %s: post power up\n", cpu, __func__);
if (MSM_PM_DEBUG_POWER_COLLAPSE & msm_pm_debug_mask)
pr_info("CPU%u: %s: return\n", cpu, __func__);
return collapsed;
}
/******************************************************************************
* External Idle/Suspend Functions
*****************************************************************************/
void arch_idle(void)
{
return;
}
static bool (*execute[MSM_PM_SLEEP_MODE_NR])(bool idle) = {
[MSM_PM_SLEEP_MODE_WAIT_FOR_INTERRUPT] = msm_pm_swfi,
[MSM_PM_SLEEP_MODE_POWER_COLLAPSE_STANDALONE] =
msm_pm_power_collapse_standalone,
[MSM_PM_SLEEP_MODE_RETENTION] = msm_pm_retention,
[MSM_PM_SLEEP_MODE_FASTPC] = msm_pm_fastpc,
[MSM_PM_SLEEP_MODE_POWER_COLLAPSE] = msm_pm_power_collapse,
};
/**
* msm_cpu_pm_enter_sleep(): Enter a low power mode on current cpu
*
* @mode - sleep mode to enter
* @from_idle - bool to indicate that the mode is exercised during idle/suspend
*
* returns none
*
* The code should be with interrupts disabled and on the core on which the
* low power is to be executed.
*
*/
bool msm_cpu_pm_enter_sleep(enum msm_pm_sleep_mode mode, bool from_idle)
{
bool exit_stat = false;
unsigned int cpu = smp_processor_id();
if ((!from_idle && cpu_online(cpu))
|| (MSM_PM_DEBUG_IDLE & msm_pm_debug_mask))
pr_info("CPU%u:%s mode:%d during %s\n", cpu, __func__,
mode, from_idle ? "idle" : "suspend");
if (execute[mode])
exit_stat = execute[mode](from_idle);
return exit_stat;
}
/**
* msm_pm_wait_cpu_shutdown() - Wait for a core to be power collapsed during
* hotplug
*
* @ cpu - cpu to wait on.
*
* Blocking function call that waits on the core to be power collapsed. This
* function is called from platform_cpu_die to ensure that a core is power
* collapsed before sending the CPU_DEAD notification so the drivers could
* remove the resource votes for this CPU(regulator and clock)
*/
int msm_pm_wait_cpu_shutdown(unsigned int cpu)
{
int timeout = 0;
if (!msm_pm_slp_sts)
return 0;
if (!msm_pm_slp_sts[cpu].base_addr)
return 0;
while (1) {
/*
* Check for the SPM of the core being hotplugged to set
* its sleep state.The SPM sleep state indicates that the
* core has been power collapsed.
*/
int acc_sts = __raw_readl(msm_pm_slp_sts[cpu].base_addr);
if (acc_sts & msm_pm_slp_sts[cpu].mask)
return 0;
udelay(100);
/*
* Dump spm registers for debugging
*/
if (++timeout == 20) {
msm_spm_dump_regs(cpu);
__WARN_printf("CPU%u didn't collapse in 2ms, sleep status: 0x%x\n",
cpu, acc_sts);
}
}
return -EBUSY;
}
static void msm_pm_ack_retention_disable(void *data)
{
/*
* This is a NULL function to ensure that the core has woken up
* and is safe to disable retention.
*/
}
/**
* msm_pm_enable_retention() - Disable/Enable retention on all cores
* @enable: Enable/Disable retention
*
*/
void msm_pm_enable_retention(bool enable)
{
if (enable == msm_pm_ldo_retention_enabled)
return;
msm_pm_ldo_retention_enabled = enable;
/*
* If retention is being disabled, wakeup all online core to ensure
* that it isn't executing retention. Offlined cores need not be woken
* up as they enter the deepest sleep mode, namely RPM assited power
* collapse
*/
if (!enable) {
preempt_disable();
smp_call_function_many(&retention_cpus,
msm_pm_ack_retention_disable,
NULL, true);
preempt_enable();
}
}
EXPORT_SYMBOL(msm_pm_enable_retention);
/**
* msm_pm_retention_enabled() - Check if retention is enabled
*
* returns true if retention is enabled
*/
bool msm_pm_retention_enabled(void)
{
return msm_pm_ldo_retention_enabled;
}
EXPORT_SYMBOL(msm_pm_retention_enabled);
static int msm_pm_snoc_client_probe(struct platform_device *pdev)
{
int rc = 0;
static struct msm_bus_scale_pdata *msm_pm_bus_pdata;
static uint32_t msm_pm_bus_client;
msm_pm_bus_pdata = msm_bus_cl_get_pdata(pdev);
if (msm_pm_bus_pdata) {
msm_pm_bus_client =
msm_bus_scale_register_client(msm_pm_bus_pdata);
if (!msm_pm_bus_client) {
pr_err("%s: Failed to register SNOC client", __func__);
rc = -ENXIO;
goto snoc_cl_probe_done;
}
rc = msm_bus_scale_client_update_request(msm_pm_bus_client, 1);
if (rc)
pr_err("%s: Error setting bus rate", __func__);
}
snoc_cl_probe_done:
return rc;
}
static int msm_cpu_status_probe(struct platform_device *pdev)
{
u32 cpu;
int rc;
if (!pdev | !pdev->dev.of_node)
return -EFAULT;
msm_pm_slp_sts = devm_kzalloc(&pdev->dev,
sizeof(*msm_pm_slp_sts) * num_possible_cpus(),
GFP_KERNEL);
if (!msm_pm_slp_sts)
return -ENOMEM;
for_each_possible_cpu(cpu) {
struct device_node *cpun, *node;
char *key;
cpun = of_get_cpu_node(cpu, NULL);
if (!cpun) {
__WARN();
continue;
}
node = of_parse_phandle(cpun, "qcom,sleep-status", 0);
if (!node)
return -ENODEV;
msm_pm_slp_sts[cpu].base_addr = of_iomap(node, 0);
if (!msm_pm_slp_sts[cpu].base_addr) {
pr_err("%s: Can't find base addr\n", __func__);
return -ENODEV;
}
key = "qcom,sleep-status-mask";
rc = of_property_read_u32(node, key, &msm_pm_slp_sts[cpu].mask);
if (rc) {
pr_err("%s: Can't find %s property\n", __func__, key);
iounmap(msm_pm_slp_sts[cpu].base_addr);
return rc;
}
}
return 0;
};
static struct of_device_id msm_slp_sts_match_tbl[] = {
{.compatible = "qcom,cpu-sleep-status"},
{},
};
static struct platform_driver msm_cpu_status_driver = {
.probe = msm_cpu_status_probe,
.driver = {
.name = "cpu_slp_status",
.owner = THIS_MODULE,
.of_match_table = msm_slp_sts_match_tbl,
},
};
static struct of_device_id msm_snoc_clnt_match_tbl[] = {
{.compatible = "qcom,pm-snoc-client"},
{},
};
static struct platform_driver msm_cpu_pm_snoc_client_driver = {
.probe = msm_pm_snoc_client_probe,
.driver = {
.name = "pm_snoc_client",
.owner = THIS_MODULE,
.of_match_table = msm_snoc_clnt_match_tbl,
},
};
struct msm_pc_debug_counters_buffer {
long *reg;
u32 len;
char buf[MAX_BUF_SIZE];
};
char *counter_name[MSM_PC_NUM_COUNTERS] = {
"PC Entry Counter",
"Warmboot Entry Counter",
"PC Bailout Counter"
};
static int msm_pc_debug_counters_copy(
struct msm_pc_debug_counters_buffer *data)
{
int j;
u32 stat;
unsigned int cpu;
unsigned int len;
uint32_t cluster_id;
uint32_t cpu_id;
uint32_t offset;
for_each_possible_cpu(cpu) {
len = scnprintf(data->buf + data->len,
sizeof(data->buf)-data->len,
"CPU%d\n", cpu);
cluster_id = MPIDR_AFFINITY_LEVEL(cpu_logical_map(cpu), 1);
cpu_id = MPIDR_AFFINITY_LEVEL(cpu_logical_map(cpu), 0);
offset = (cluster_id * MAX_CPUS_PER_CLUSTER
* MSM_PC_NUM_COUNTERS)
+ (cpu_id * MSM_PC_NUM_COUNTERS);
data->len += len;
for (j = 0; j < MSM_PC_NUM_COUNTERS - 1; j++) {
stat = data->reg[offset + j];
len = scnprintf(data->buf + data->len,
sizeof(data->buf) - data->len,
"\t%s: %d", counter_name[j], stat);
data->len += len;
}
len = scnprintf(data->buf + data->len,
sizeof(data->buf) - data->len,
"\n");
data->len += len;
}
return data->len;
}
static ssize_t msm_pc_debug_counters_file_read(struct file *file,
char __user *bufu, size_t count, loff_t *ppos)
{
struct msm_pc_debug_counters_buffer *data;
data = file->private_data;
if (!data)
return -EINVAL;
if (!bufu)
return -EINVAL;
if (!access_ok(VERIFY_WRITE, bufu, count))
return -EFAULT;
if (*ppos >= data->len && data->len == 0)
data->len = msm_pc_debug_counters_copy(data);
return simple_read_from_buffer(bufu, count, ppos,
data->buf, data->len);
}
static int msm_pc_debug_counters_file_open(struct inode *inode,
struct file *file)
{
struct msm_pc_debug_counters_buffer *buf;
if (!inode->i_private)
return -EINVAL;
file->private_data = kzalloc(
sizeof(struct msm_pc_debug_counters_buffer), GFP_KERNEL);
if (!file->private_data) {
pr_err("%s: ERROR kmalloc failed to allocate %zu bytes\n",
__func__, sizeof(struct msm_pc_debug_counters_buffer));
return -ENOMEM;
}
buf = file->private_data;
buf->reg = (long *)inode->i_private;
return 0;
}
static int msm_pc_debug_counters_file_close(struct inode *inode,
struct file *file)
{
kfree(file->private_data);
return 0;
}
static const struct file_operations msm_pc_debug_counters_fops = {
.open = msm_pc_debug_counters_file_open,
.read = msm_pc_debug_counters_file_read,
.release = msm_pc_debug_counters_file_close,
.llseek = no_llseek,
};
static int msm_pm_clk_init(struct platform_device *pdev)
{
bool synced_clocks;
u32 cpu;
char clk_name[] = "cpu??_clk";
char *key;
key = "qcom,saw-turns-off-pll";
if (of_property_read_bool(pdev->dev.of_node, key))
return 0;
key = "qcom,synced-clocks";
synced_clocks = of_property_read_bool(pdev->dev.of_node, key);
for_each_possible_cpu(cpu) {
struct clk *clk;
snprintf(clk_name, sizeof(clk_name), "cpu%d_clk", cpu);
clk = clk_get(&pdev->dev, clk_name);
if (IS_ERR(clk)) {
if (cpu && synced_clocks)
return 0;
else
clk = NULL;
}
per_cpu(cpu_clks, cpu) = clk;
}
if (synced_clocks)
return 0;
l2_clk = clk_get(&pdev->dev, "l2_clk");
if (IS_ERR(l2_clk))
pr_warn("%s: Could not get l2_clk (-%ld)\n", __func__,
PTR_ERR(l2_clk));
return 0;
}
static int msm_cpu_pm_probe(struct platform_device *pdev)
{
struct dentry *dent = NULL;
struct resource *res = NULL;
int ret = 0;
void __iomem *msm_pc_debug_counters_imem;
char *key;
int alloc_size = (MAX_NUM_CLUSTER * MAX_CPUS_PER_CLUSTER
* MSM_PC_NUM_COUNTERS
* sizeof(*msm_pc_debug_counters));
msm_pc_debug_counters = dma_alloc_coherent(&pdev->dev, alloc_size,
&msm_pc_debug_counters_phys, GFP_KERNEL);
if (msm_pc_debug_counters) {
memset(msm_pc_debug_counters, 0, alloc_size);
dent = debugfs_create_file("pc_debug_counter", S_IRUGO, NULL,
msm_pc_debug_counters,
&msm_pc_debug_counters_fops);
if (!dent)
pr_err("%s: ERROR debugfs_create_file failed\n",
__func__);
res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
if (!res)
goto skip_save_imem;
msm_pc_debug_counters_imem = devm_ioremap(&pdev->dev,
res->start, resource_size(res));
if (msm_pc_debug_counters_imem) {
writel_relaxed(msm_pc_debug_counters_phys,
msm_pc_debug_counters_imem);
mb();
devm_iounmap(&pdev->dev,
msm_pc_debug_counters_imem);
}
} else {
msm_pc_debug_counters = 0;
msm_pc_debug_counters_phys = 0;
}
skip_save_imem:
if (pdev->dev.of_node) {
key = "qcom,tz-flushes-cache";
msm_pm_tz_flushes_cache =
of_property_read_bool(pdev->dev.of_node, key);
key = "qcom,no-pll-switch-for-retention";
msm_pm_ret_no_pll_switch =
of_property_read_bool(pdev->dev.of_node, key);
ret = msm_pm_clk_init(pdev);
if (ret) {
pr_info("msm_pm_clk_init returned error\n");
return ret;
}
}
if (pdev->dev.of_node)
of_platform_populate(pdev->dev.of_node, NULL, NULL, &pdev->dev);
return ret;
}
static struct of_device_id msm_cpu_pm_table[] = {
{.compatible = "qcom,pm"},
{},
};
static struct platform_driver msm_cpu_pm_driver = {
.probe = msm_cpu_pm_probe,
.driver = {
.name = "msm-pm",
.owner = THIS_MODULE,
.of_match_table = msm_cpu_pm_table,
},
};
static int __init msm_pm_drv_init(void)
{
int rc;
cpumask_clear(&retention_cpus);
rc = platform_driver_register(&msm_cpu_pm_snoc_client_driver);
if (rc) {
pr_err("%s(): failed to register driver %s\n", __func__,
msm_cpu_pm_snoc_client_driver.driver.name);
return rc;
}
return platform_driver_register(&msm_cpu_pm_driver);
}
late_initcall(msm_pm_drv_init);
int __init msm_pm_sleep_status_init(void)
{
static bool registered;
if (registered)
return 0;
registered = true;
return platform_driver_register(&msm_cpu_status_driver);
}
arch_initcall(msm_pm_sleep_status_init);
#ifdef CONFIG_ARM
static int idle_initialize(void)
{
arm_pm_idle = arch_idle;
return 0;
}
early_initcall(idle_initialize);
#endif