[linux-2.6-block.git] / drivers / acpi / acpi_pad.c

/*
 * acpi_pad.c ACPI Processor Aggregator Driver
 *
 * Copyright (c) 2009, Intel Corporation.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
 * version 2, as published by the Free Software Foundation.
 *
 * This program is distributed in the hope 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/kernel.h>
#include <linux/cpumask.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/types.h>
#include <linux/kthread.h>
#include <linux/freezer.h>
#include <linux/cpu.h>
#include <linux/tick.h>
#include <linux/slab.h>
#include <linux/acpi.h>
#include <asm/mwait.h>

#define ACPI_PROCESSOR_AGGREGATOR_CLASS	"acpi_pad"
#define ACPI_PROCESSOR_AGGREGATOR_DEVICE_NAME "Processor Aggregator"
#define ACPI_PROCESSOR_AGGREGATOR_NOTIFY 0x80
static DEFINE_MUTEX(isolated_cpus_lock);
static DEFINE_MUTEX(round_robin_lock);

static unsigned long power_saving_mwait_eax;

static unsigned char tsc_detected_unstable;
static unsigned char tsc_marked_unstable;

static void power_saving_mwait_init(void)
{
	unsigned int eax, ebx, ecx, edx;
	unsigned int highest_cstate = 0;
	unsigned int highest_subcstate = 0;
	int i;

	if (!boot_cpu_has(X86_FEATURE_MWAIT))
		return;
	if (boot_cpu_data.cpuid_level < CPUID_MWAIT_LEAF)
		return;

	cpuid(CPUID_MWAIT_LEAF, &eax, &ebx, &ecx, &edx);

	if (!(ecx & CPUID5_ECX_EXTENSIONS_SUPPORTED) ||
	    !(ecx & CPUID5_ECX_INTERRUPT_BREAK))
		return;

	edx >>= MWAIT_SUBSTATE_SIZE;
	for (i = 0; i < 7 && edx; i++, edx >>= MWAIT_SUBSTATE_SIZE) {
		if (edx & MWAIT_SUBSTATE_MASK) {
			highest_cstate = i;
			highest_subcstate = edx & MWAIT_SUBSTATE_MASK;
		}
	}
	power_saving_mwait_eax = (highest_cstate << MWAIT_SUBSTATE_SIZE) |
		(highest_subcstate - 1);

#if defined(CONFIG_X86)
	switch (boot_cpu_data.x86_vendor) {
	case X86_VENDOR_AMD:
	case X86_VENDOR_INTEL:
		/*
		 * AMD Fam10h TSC will tick in all
		 * C/P/S0/S1 states when this bit is set.
		 */
		if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC))
			tsc_detected_unstable = 1;
		break;
	default:
		/* TSC could halt in idle */
		tsc_detected_unstable = 1;
	}
#endif
}

static unsigned long cpu_weight[NR_CPUS];
static int tsk_in_cpu[NR_CPUS] = {[0 ... NR_CPUS-1] = -1};
static DECLARE_BITMAP(pad_busy_cpus_bits, NR_CPUS);
static void round_robin_cpu(unsigned int tsk_index)
{
	struct cpumask *pad_busy_cpus = to_cpumask(pad_busy_cpus_bits);
	cpumask_var_t tmp;
	int cpu;
	unsigned long min_weight = -1;
	unsigned long uninitialized_var(preferred_cpu);

	if (!alloc_cpumask_var(&tmp, GFP_KERNEL))
		return;

	mutex_lock(&round_robin_lock);
	cpumask_clear(tmp);
	for_each_cpu(cpu, pad_busy_cpus)
		cpumask_or(tmp, tmp, topology_sibling_cpumask(cpu));
	cpumask_andnot(tmp, cpu_online_mask, tmp);
	/* avoid HT sibilings if possible */
	if (cpumask_empty(tmp))
		cpumask_andnot(tmp, cpu_online_mask, pad_busy_cpus);
	if (cpumask_empty(tmp)) {
		mutex_unlock(&round_robin_lock);
		return;
	}
	for_each_cpu(cpu, tmp) {
		if (cpu_weight[cpu] < min_weight) {
			min_weight = cpu_weight[cpu];
			preferred_cpu = cpu;
		}
	}

	if (tsk_in_cpu[tsk_index] != -1)
		cpumask_clear_cpu(tsk_in_cpu[tsk_index], pad_busy_cpus);
	tsk_in_cpu[tsk_index] = preferred_cpu;
	cpumask_set_cpu(preferred_cpu, pad_busy_cpus);
	cpu_weight[preferred_cpu]++;
	mutex_unlock(&round_robin_lock);

	set_cpus_allowed_ptr(current, cpumask_of(preferred_cpu));
}

static void exit_round_robin(unsigned int tsk_index)
{
	struct cpumask *pad_busy_cpus = to_cpumask(pad_busy_cpus_bits);
	cpumask_clear_cpu(tsk_in_cpu[tsk_index], pad_busy_cpus);
	tsk_in_cpu[tsk_index] = -1;
}

static unsigned int idle_pct = 5; /* percentage */
static unsigned int round_robin_time = 1; /* second */
static int power_saving_thread(void *data)
{
	struct sched_param param = {.sched_priority = 1};
	int do_sleep;
	unsigned int tsk_index = (unsigned long)data;
	u64 last_jiffies = 0;

	sched_setscheduler(current, SCHED_RR, &param);

	while (!kthread_should_stop()) {
		unsigned long expire_time;

		/* round robin to cpus */
		expire_time = last_jiffies + round_robin_time * HZ;
		if (time_before(expire_time, jiffies)) {
			last_jiffies = jiffies;
			round_robin_cpu(tsk_index);
		}

		do_sleep = 0;

		expire_time = jiffies + HZ * (100 - idle_pct) / 100;

		while (!need_resched()) {
			if (tsc_detected_unstable && !tsc_marked_unstable) {
				/* TSC could halt in idle, so notify users */
				mark_tsc_unstable("TSC halts in idle");
				tsc_marked_unstable = 1;
			}
			local_irq_disable();
			tick_broadcast_enable();
			tick_broadcast_enter();
			stop_critical_timings();

			mwait_idle_with_hints(power_saving_mwait_eax, 1);

			start_critical_timings();
			tick_broadcast_exit();
			local_irq_enable();

			if (time_before(expire_time, jiffies)) {
				do_sleep = 1;
				break;
			}
		}

		/*
		 * current sched_rt has threshold for rt task running time.
		 * When a rt task uses 95% CPU time, the rt thread will be
		 * scheduled out for 5% CPU time to not starve other tasks. But
		 * the mechanism only works when all CPUs have RT task running,
		 * as if one CPU hasn't RT task, RT task from other CPUs will
		 * borrow CPU time from this CPU and cause RT task use > 95%
		 * CPU time. To make 'avoid starvation' work, takes a nap here.
		 */
		if (unlikely(do_sleep))
			schedule_timeout_killable(HZ * idle_pct / 100);

		/* If an external event has set the need_resched flag, then
		 * we need to deal with it, or this loop will continue to
		 * spin without calling __mwait().
		 */
		if (unlikely(need_resched()))
			schedule();
	}

	exit_round_robin(tsk_index);
	return 0;
}

static struct task_struct *ps_tsks[NR_CPUS];
static unsigned int ps_tsk_num;
static int create_power_saving_task(void)
{
	int rc;

	ps_tsks[ps_tsk_num] = kthread_run(power_saving_thread,
		(void *)(unsigned long)ps_tsk_num,
		"acpi_pad/%d", ps_tsk_num);

	if (IS_ERR(ps_tsks[ps_tsk_num])) {
		rc = PTR_ERR(ps_tsks[ps_tsk_num]);
		ps_tsks[ps_tsk_num] = NULL;
	} else {
		rc = 0;
		ps_tsk_num++;
	}

	return rc;
}

static void destroy_power_saving_task(void)
{
	if (ps_tsk_num > 0) {
		ps_tsk_num--;
		kthread_stop(ps_tsks[ps_tsk_num]);
		ps_tsks[ps_tsk_num] = NULL;
	}
}

static void set_power_saving_task_num(unsigned int num)
{
	if (num > ps_tsk_num) {
		while (ps_tsk_num < num) {
			if (create_power_saving_task())
				return;
		}
	} else if (num < ps_tsk_num) {
		while (ps_tsk_num > num)
			destroy_power_saving_task();
	}
}

static void acpi_pad_idle_cpus(unsigned int num_cpus)
{
	get_online_cpus();

	num_cpus = min_t(unsigned int, num_cpus, num_online_cpus());
	set_power_saving_task_num(num_cpus);

	put_online_cpus();
}

static uint32_t acpi_pad_idle_cpus_num(void)
{
	return ps_tsk_num;
}

static ssize_t acpi_pad_rrtime_store(struct device *dev,
	struct device_attribute *attr, const char *buf, size_t count)
{
	unsigned long num;
	if (kstrtoul(buf, 0, &num))
		return -EINVAL;
	if (num < 1 || num >= 100)
		return -EINVAL;
	mutex_lock(&isolated_cpus_lock);
	round_robin_time = num;
	mutex_unlock(&isolated_cpus_lock);
	return count;
}

static ssize_t acpi_pad_rrtime_show(struct device *dev,
	struct device_attribute *attr, char *buf)
{
	return scnprintf(buf, PAGE_SIZE, "%d\n", round_robin_time);
}
static DEVICE_ATTR(rrtime, S_IRUGO|S_IWUSR,
	acpi_pad_rrtime_show,
	acpi_pad_rrtime_store);

static ssize_t acpi_pad_idlepct_store(struct device *dev,
	struct device_attribute *attr, const char *buf, size_t count)
{
	unsigned long num;
	if (kstrtoul(buf, 0, &num))
		return -EINVAL;
	if (num < 1 || num >= 100)
		return -EINVAL;
	mutex_lock(&isolated_cpus_lock);
	idle_pct = num;
	mutex_unlock(&isolated_cpus_lock);
	return count;
}

static ssize_t acpi_pad_idlepct_show(struct device *dev,
	struct device_attribute *attr, char *buf)
{
	return scnprintf(buf, PAGE_SIZE, "%d\n", idle_pct);
}
static DEVICE_ATTR(idlepct, S_IRUGO|S_IWUSR,
	acpi_pad_idlepct_show,
	acpi_pad_idlepct_store);

static ssize_t acpi_pad_idlecpus_store(struct device *dev,
	struct device_attribute *attr, const char *buf, size_t count)
{
	unsigned long num;
	if (kstrtoul(buf, 0, &num))
		return -EINVAL;
	mutex_lock(&isolated_cpus_lock);
	acpi_pad_idle_cpus(num);
	mutex_unlock(&isolated_cpus_lock);
	return count;
}

static ssize_t acpi_pad_idlecpus_show(struct device *dev,
	struct device_attribute *attr, char *buf)
{
	return cpumap_print_to_pagebuf(false, buf,
				       to_cpumask(pad_busy_cpus_bits));
}

static DEVICE_ATTR(idlecpus, S_IRUGO|S_IWUSR,
	acpi_pad_idlecpus_show,
	acpi_pad_idlecpus_store);

static int acpi_pad_add_sysfs(struct acpi_device *device)
{
	int result;

	result = device_create_file(&device->dev, &dev_attr_idlecpus);
	if (result)
		return -ENODEV;
	result = device_create_file(&device->dev, &dev_attr_idlepct);
	if (result) {
		device_remove_file(&device->dev, &dev_attr_idlecpus);
		return -ENODEV;
	}
	result = device_create_file(&device->dev, &dev_attr_rrtime);
	if (result) {
		device_remove_file(&device->dev, &dev_attr_idlecpus);
		device_remove_file(&device->dev, &dev_attr_idlepct);
		return -ENODEV;
	}
	return 0;
}

static void acpi_pad_remove_sysfs(struct acpi_device *device)
{
	device_remove_file(&device->dev, &dev_attr_idlecpus);
	device_remove_file(&device->dev, &dev_attr_idlepct);
	device_remove_file(&device->dev, &dev_attr_rrtime);
}

/*
 * Query firmware how many CPUs should be idle
 * return -1 on failure
 */
static int acpi_pad_pur(acpi_handle handle)
{
	struct acpi_buffer buffer = {ACPI_ALLOCATE_BUFFER, NULL};
	union acpi_object *package;
	int num = -1;

	if (ACPI_FAILURE(acpi_evaluate_object(handle, "_PUR", NULL, &buffer)))
		return num;

	if (!buffer.length || !buffer.pointer)
		return num;

	package = buffer.pointer;

	if (package->type == ACPI_TYPE_PACKAGE &&
		package->package.count == 2 &&
		package->package.elements[0].integer.value == 1) /* rev 1 */

		num = package->package.elements[1].integer.value;

	kfree(buffer.pointer);
	return num;
}

static void acpi_pad_handle_notify(acpi_handle handle)
{
	int num_cpus;
	uint32_t idle_cpus;
	struct acpi_buffer param = {
		.length = 4,
		.pointer = (void *)&idle_cpus,
	};

	mutex_lock(&isolated_cpus_lock);
	num_cpus = acpi_pad_pur(handle);
	if (num_cpus < 0) {
		mutex_unlock(&isolated_cpus_lock);
		return;
	}
	acpi_pad_idle_cpus(num_cpus);
	idle_cpus = acpi_pad_idle_cpus_num();
	acpi_evaluate_ost(handle, ACPI_PROCESSOR_AGGREGATOR_NOTIFY, 0, &param);
	mutex_unlock(&isolated_cpus_lock);
}

static void acpi_pad_notify(acpi_handle handle, u32 event,
	void *data)
{
	struct acpi_device *device = data;

	switch (event) {
	case ACPI_PROCESSOR_AGGREGATOR_NOTIFY:
		acpi_pad_handle_notify(handle);
		acpi_bus_generate_netlink_event(device->pnp.device_class,
			dev_name(&device->dev), event, 0);
		break;
	default:
		pr_warn("Unsupported event [0x%x]\n", event);
		break;
	}
}

static int acpi_pad_add(struct acpi_device *device)
{
	acpi_status status;

	strcpy(acpi_device_name(device), ACPI_PROCESSOR_AGGREGATOR_DEVICE_NAME);
	strcpy(acpi_device_class(device), ACPI_PROCESSOR_AGGREGATOR_CLASS);

	if (acpi_pad_add_sysfs(device))
		return -ENODEV;

	status = acpi_install_notify_handler(device->handle,
		ACPI_DEVICE_NOTIFY, acpi_pad_notify, device);
	if (ACPI_FAILURE(status)) {
		acpi_pad_remove_sysfs(device);
		return -ENODEV;
	}

	return 0;
}

static int acpi_pad_remove(struct acpi_device *device)
{
	mutex_lock(&isolated_cpus_lock);
	acpi_pad_idle_cpus(0);
	mutex_unlock(&isolated_cpus_lock);

	acpi_remove_notify_handler(device->handle,
		ACPI_DEVICE_NOTIFY, acpi_pad_notify);
	acpi_pad_remove_sysfs(device);
	return 0;
}

static const struct acpi_device_id pad_device_ids[] = {
	{"ACPI000C", 0},
	{"", 0},
};
MODULE_DEVICE_TABLE(acpi, pad_device_ids);

static struct acpi_driver acpi_pad_driver = {
	.name = "processor_aggregator",
	.class = ACPI_PROCESSOR_AGGREGATOR_CLASS,
	.ids = pad_device_ids,
	.ops = {
		.add = acpi_pad_add,
		.remove = acpi_pad_remove,
	},
};

static int __init acpi_pad_init(void)
{
	power_saving_mwait_init();
	if (power_saving_mwait_eax == 0)
		return -EINVAL;

	return acpi_bus_register_driver(&acpi_pad_driver);
}

static void __exit acpi_pad_exit(void)
{
	acpi_bus_unregister_driver(&acpi_pad_driver);
}

module_init(acpi_pad_init);
module_exit(acpi_pad_exit);
MODULE_AUTHOR("Shaohua Li<shaohua.li@intel.com>");
MODULE_DESCRIPTION("ACPI Processor Aggregator Driver");
MODULE_LICENSE("GPL");
Commit	Line	Data
8e0af514 SL	1	/*
	2	* acpi_pad.c ACPI Processor Aggregator Driver
	3	*
	4	* Copyright (c) 2009, Intel Corporation.
	5	*
	6	* This program is free software; you can redistribute it and/or modify it
	7	* under the terms and conditions of the GNU General Public License,
	8	* version 2, as published by the Free Software Foundation.
	9	*
	10	* This program is distributed in the hope it will be useful, but WITHOUT
	11	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	12	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
	13	* more details.
	14	*
8e0af514 SL	15	*/
	16
	17	#include <linux/kernel.h>
	18	#include <linux/cpumask.h>
	19	#include <linux/module.h>
	20	#include <linux/init.h>
	21	#include <linux/types.h>
	22	#include <linux/kthread.h>
	23	#include <linux/freezer.h>
	24	#include <linux/cpu.h>
979081e7	25	#include <linux/tick.h>
5a0e3ad6	26	#include <linux/slab.h>
8b48463f	27	#include <linux/acpi.h>
bc83cccc	28	#include <asm/mwait.h>
8e0af514	29
a40770a9	30	#define ACPI_PROCESSOR_AGGREGATOR_CLASS "acpi_pad"
8e0af514 SL	31	#define ACPI_PROCESSOR_AGGREGATOR_DEVICE_NAME "Processor Aggregator"
	32	#define ACPI_PROCESSOR_AGGREGATOR_NOTIFY 0x80
	33	static DEFINE_MUTEX(isolated_cpus_lock);
5f160126	34	static DEFINE_MUTEX(round_robin_lock);
8e0af514	35
8e0af514	36	static unsigned long power_saving_mwait_eax;
0dc698b9 VP	37
	38	static unsigned char tsc_detected_unstable;
	39	static unsigned char tsc_marked_unstable;
	40
8e0af514 SL	41	static void power_saving_mwait_init(void)
	42	{
	43	unsigned int eax, ebx, ecx, edx;
	44	unsigned int highest_cstate = 0;
	45	unsigned int highest_subcstate = 0;
	46	int i;
	47
	48	if (!boot_cpu_has(X86_FEATURE_MWAIT))
	49	return;
	50	if (boot_cpu_data.cpuid_level < CPUID_MWAIT_LEAF)
	51	return;
	52
	53	cpuid(CPUID_MWAIT_LEAF, &eax, &ebx, &ecx, &edx);
	54
	55	if (!(ecx & CPUID5_ECX_EXTENSIONS_SUPPORTED) \|\|
	56	!(ecx & CPUID5_ECX_INTERRUPT_BREAK))
	57	return;
	58
	59	edx >>= MWAIT_SUBSTATE_SIZE;
	60	for (i = 0; i < 7 && edx; i++, edx >>= MWAIT_SUBSTATE_SIZE) {
	61	if (edx & MWAIT_SUBSTATE_MASK) {
	62	highest_cstate = i;
	63	highest_subcstate = edx & MWAIT_SUBSTATE_MASK;
	64	}
	65	}
	66	power_saving_mwait_eax = (highest_cstate << MWAIT_SUBSTATE_SIZE) \|
	67	(highest_subcstate - 1);
	68
592913ec	69	#if defined(CONFIG_X86)
8e0af514 SL	70	switch (boot_cpu_data.x86_vendor) {
	71	case X86_VENDOR_AMD:
	72	case X86_VENDOR_INTEL:
	73	/*
	74	* AMD Fam10h TSC will tick in all
	75	* C/P/S0/S1 states when this bit is set.
	76	*/
8aa4b14e CG	77	if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC))
8aa4b14e CG	78	tsc_detected_unstable = 1;
8aa4b14e	79	break;
8e0af514	80	default:
3ff70551	81	/* TSC could halt in idle */
0dc698b9	82	tsc_detected_unstable = 1;
8e0af514 SL	83	}
	84	#endif
	85	}
	86
	87	static unsigned long cpu_weight[NR_CPUS];
	88	static int tsk_in_cpu[NR_CPUS] = {[0 ... NR_CPUS-1] = -1};
	89	static DECLARE_BITMAP(pad_busy_cpus_bits, NR_CPUS);
	90	static void round_robin_cpu(unsigned int tsk_index)
	91	{
	92	struct cpumask *pad_busy_cpus = to_cpumask(pad_busy_cpus_bits);
	93	cpumask_var_t tmp;
	94	int cpu;
f67538f8 AM	95	unsigned long min_weight = -1;
f67538f8 AM	96	unsigned long uninitialized_var(preferred_cpu);
8e0af514 SL	97
	98	if (!alloc_cpumask_var(&tmp, GFP_KERNEL))
	99	return;
	100
5f160126	101	mutex_lock(&round_robin_lock);
8e0af514 SL	102	cpumask_clear(tmp);
8e0af514 SL	103	for_each_cpu(cpu, pad_busy_cpus)
06931e62	104	cpumask_or(tmp, tmp, topology_sibling_cpumask(cpu));
8e0af514 SL	105	cpumask_andnot(tmp, cpu_online_mask, tmp);
	106	/* avoid HT sibilings if possible */
	107	if (cpumask_empty(tmp))
	108	cpumask_andnot(tmp, cpu_online_mask, pad_busy_cpus);
	109	if (cpumask_empty(tmp)) {
5f160126	110	mutex_unlock(&round_robin_lock);
8e0af514 SL	111	return;
	112	}
	113	for_each_cpu(cpu, tmp) {
	114	if (cpu_weight[cpu] < min_weight) {
	115	min_weight = cpu_weight[cpu];
	116	preferred_cpu = cpu;
	117	}
	118	}
	119
	120	if (tsk_in_cpu[tsk_index] != -1)
	121	cpumask_clear_cpu(tsk_in_cpu[tsk_index], pad_busy_cpus);
	122	tsk_in_cpu[tsk_index] = preferred_cpu;
	123	cpumask_set_cpu(preferred_cpu, pad_busy_cpus);
	124	cpu_weight[preferred_cpu]++;
5f160126	125	mutex_unlock(&round_robin_lock);
8e0af514 SL	126
	127	set_cpus_allowed_ptr(current, cpumask_of(preferred_cpu));
	128	}
	129
	130	static void exit_round_robin(unsigned int tsk_index)
	131	{
	132	struct cpumask *pad_busy_cpus = to_cpumask(pad_busy_cpus_bits);
	133	cpumask_clear_cpu(tsk_in_cpu[tsk_index], pad_busy_cpus);
	134	tsk_in_cpu[tsk_index] = -1;
	135	}
	136
	137	static unsigned int idle_pct = 5; /* percentage */
fa7584e1	138	static unsigned int round_robin_time = 1; /* second */
8e0af514 SL	139	static int power_saving_thread(void *data)
	140	{
	141	struct sched_param param = {.sched_priority = 1};
	142	int do_sleep;
	143	unsigned int tsk_index = (unsigned long)data;
	144	u64 last_jiffies = 0;
	145
	146	sched_setscheduler(current, SCHED_RR, &param);
	147
	148	while (!kthread_should_stop()) {
4ff248f3	149	unsigned long expire_time;
8e0af514	150
8e0af514	151	/* round robin to cpus */
4ff248f3 MS	152	expire_time = last_jiffies + round_robin_time * HZ;
4ff248f3 MS	153	if (time_before(expire_time, jiffies)) {
8e0af514 SL	154	last_jiffies = jiffies;
	155	round_robin_cpu(tsk_index);
	156	}
	157
	158	do_sleep = 0;
	159
8e0af514 SL	160	expire_time = jiffies + HZ * (100 - idle_pct) / 100;
	161
	162	while (!need_resched()) {
0dc698b9 VP	163	if (tsc_detected_unstable && !tsc_marked_unstable) {
	164	/* TSC could halt in idle, so notify users */
	165	mark_tsc_unstable("TSC halts in idle");
	166	tsc_marked_unstable = 1;
	167	}
8e0af514	168	local_irq_disable();
979081e7	169	tick_broadcast_enable();
c7952135	170	tick_broadcast_enter();
8e0af514 SL	171	stop_critical_timings();
8e0af514 SL	172
16824255	173	mwait_idle_with_hints(power_saving_mwait_eax, 1);
8e0af514 SL	174
8e0af514 SL	175	start_critical_timings();
c7952135	176	tick_broadcast_exit();
8e0af514 SL	177	local_irq_enable();
8e0af514 SL	178
4ff248f3	179	if (time_before(expire_time, jiffies)) {
8e0af514 SL	180	do_sleep = 1;
	181	break;
	182	}
	183	}
	184
8e0af514 SL	185	/*
	186	* current sched_rt has threshold for rt task running time.
	187	* When a rt task uses 95% CPU time, the rt thread will be
	188	* scheduled out for 5% CPU time to not starve other tasks. But
	189	* the mechanism only works when all CPUs have RT task running,
	190	* as if one CPU hasn't RT task, RT task from other CPUs will
	191	* borrow CPU time from this CPU and cause RT task use > 95%
3b8cb427	192	* CPU time. To make 'avoid starvation' work, takes a nap here.
8e0af514	193	*/
5b59c69e	194	if (unlikely(do_sleep))
8e0af514	195	schedule_timeout_killable(HZ * idle_pct / 100);
5b59c69e TC	196
	197	/* If an external event has set the need_resched flag, then
	198	* we need to deal with it, or this loop will continue to
	199	* spin without calling __mwait().
	200	*/
	201	if (unlikely(need_resched()))
	202	schedule();
8e0af514 SL	203	}
	204
	205	exit_round_robin(tsk_index);
	206	return 0;
	207	}
	208
	209	static struct task_struct *ps_tsks[NR_CPUS];
	210	static unsigned int ps_tsk_num;
	211	static int create_power_saving_task(void)
	212	{
5d7e4386	213	int rc;
3b8cb427	214
8e0af514 SL	215	ps_tsks[ps_tsk_num] = kthread_run(power_saving_thread,
8e0af514 SL	216	(void *)(unsigned long)ps_tsk_num,
150ed86f	217	"acpi_pad/%d", ps_tsk_num);
5d7e4386 RR	218
	219	if (IS_ERR(ps_tsks[ps_tsk_num])) {
	220	rc = PTR_ERR(ps_tsks[ps_tsk_num]);
3b8cb427	221	ps_tsks[ps_tsk_num] = NULL;
5d7e4386 RR	222	} else {
	223	rc = 0;
	224	ps_tsk_num++;
	225	}
3b8cb427 CG	226
3b8cb427 CG	227	return rc;
8e0af514 SL	228	}
	229
	230	static void destroy_power_saving_task(void)
	231	{
	232	if (ps_tsk_num > 0) {
	233	ps_tsk_num--;
	234	kthread_stop(ps_tsks[ps_tsk_num]);
3b8cb427	235	ps_tsks[ps_tsk_num] = NULL;
8e0af514 SL	236	}
	237	}
	238
	239	static void set_power_saving_task_num(unsigned int num)
	240	{
	241	if (num > ps_tsk_num) {
	242	while (ps_tsk_num < num) {
	243	if (create_power_saving_task())
	244	return;
	245	}
	246	} else if (num < ps_tsk_num) {
	247	while (ps_tsk_num > num)
	248	destroy_power_saving_task();
	249	}
	250	}
	251
3b8cb427	252	static void acpi_pad_idle_cpus(unsigned int num_cpus)
8e0af514 SL	253	{
	254	get_online_cpus();
	255
	256	num_cpus = min_t(unsigned int, num_cpus, num_online_cpus());
	257	set_power_saving_task_num(num_cpus);
	258
	259	put_online_cpus();
8e0af514 SL	260	}
	261
	262	static uint32_t acpi_pad_idle_cpus_num(void)
	263	{
	264	return ps_tsk_num;
	265	}
	266
	267	static ssize_t acpi_pad_rrtime_store(struct device *dev,
	268	struct device_attribute attr, const char buf, size_t count)
	269	{
	270	unsigned long num;
73d4511a	271	if (kstrtoul(buf, 0, &num))
8e0af514 SL	272	return -EINVAL;
	273	if (num < 1 \|\| num >= 100)
	274	return -EINVAL;
	275	mutex_lock(&isolated_cpus_lock);
	276	round_robin_time = num;
	277	mutex_unlock(&isolated_cpus_lock);
	278	return count;
	279	}
	280
	281	static ssize_t acpi_pad_rrtime_show(struct device *dev,
	282	struct device_attribute attr, char buf)
	283	{
32297abd	284	return scnprintf(buf, PAGE_SIZE, "%d\n", round_robin_time);
8e0af514 SL	285	}
	286	static DEVICE_ATTR(rrtime, S_IRUGO\|S_IWUSR,
	287	acpi_pad_rrtime_show,
	288	acpi_pad_rrtime_store);
	289
	290	static ssize_t acpi_pad_idlepct_store(struct device *dev,
	291	struct device_attribute attr, const char buf, size_t count)
	292	{
	293	unsigned long num;
73d4511a	294	if (kstrtoul(buf, 0, &num))
8e0af514 SL	295	return -EINVAL;
	296	if (num < 1 \|\| num >= 100)
	297	return -EINVAL;
	298	mutex_lock(&isolated_cpus_lock);
	299	idle_pct = num;
	300	mutex_unlock(&isolated_cpus_lock);
	301	return count;
	302	}
	303
	304	static ssize_t acpi_pad_idlepct_show(struct device *dev,
	305	struct device_attribute attr, char buf)
	306	{
32297abd	307	return scnprintf(buf, PAGE_SIZE, "%d\n", idle_pct);
8e0af514 SL	308	}
	309	static DEVICE_ATTR(idlepct, S_IRUGO\|S_IWUSR,
	310	acpi_pad_idlepct_show,
	311	acpi_pad_idlepct_store);
	312
	313	static ssize_t acpi_pad_idlecpus_store(struct device *dev,
	314	struct device_attribute attr, const char buf, size_t count)
	315	{
	316	unsigned long num;
73d4511a	317	if (kstrtoul(buf, 0, &num))
8e0af514 SL	318	return -EINVAL;
	319	mutex_lock(&isolated_cpus_lock);
	320	acpi_pad_idle_cpus(num);
	321	mutex_unlock(&isolated_cpus_lock);
	322	return count;
	323	}
	324
	325	static ssize_t acpi_pad_idlecpus_show(struct device *dev,
	326	struct device_attribute attr, char buf)
	327	{
5aaba363 SH	328	return cpumap_print_to_pagebuf(false, buf,
5aaba363 SH	329	to_cpumask(pad_busy_cpus_bits));
8e0af514	330	}
5aaba363	331
8e0af514 SL	332	static DEVICE_ATTR(idlecpus, S_IRUGO\|S_IWUSR,
	333	acpi_pad_idlecpus_show,
	334	acpi_pad_idlecpus_store);
	335
	336	static int acpi_pad_add_sysfs(struct acpi_device *device)
	337	{
	338	int result;
	339
	340	result = device_create_file(&device->dev, &dev_attr_idlecpus);
	341	if (result)
	342	return -ENODEV;
	343	result = device_create_file(&device->dev, &dev_attr_idlepct);
	344	if (result) {
	345	device_remove_file(&device->dev, &dev_attr_idlecpus);
	346	return -ENODEV;
	347	}
	348	result = device_create_file(&device->dev, &dev_attr_rrtime);
	349	if (result) {
	350	device_remove_file(&device->dev, &dev_attr_idlecpus);
	351	device_remove_file(&device->dev, &dev_attr_idlepct);
	352	return -ENODEV;
	353	}
	354	return 0;
	355	}
	356
	357	static void acpi_pad_remove_sysfs(struct acpi_device *device)
	358	{
	359	device_remove_file(&device->dev, &dev_attr_idlecpus);
	360	device_remove_file(&device->dev, &dev_attr_idlepct);
	361	device_remove_file(&device->dev, &dev_attr_rrtime);
	362	}
	363
c9ad8e06 LB	364	/*
	365	* Query firmware how many CPUs should be idle
	366	* return -1 on failure
	367	*/
	368	static int acpi_pad_pur(acpi_handle handle)
8e0af514 SL	369	{
8e0af514 SL	370	struct acpi_buffer buffer = {ACPI_ALLOCATE_BUFFER, NULL};
8e0af514	371	union acpi_object *package;
c9ad8e06	372	int num = -1;
8e0af514	373
3b8cb427	374	if (ACPI_FAILURE(acpi_evaluate_object(handle, "_PUR", NULL, &buffer)))
c9ad8e06	375	return num;
3b8cb427 CG	376
3b8cb427 CG	377	if (!buffer.length \|\| !buffer.pointer)
c9ad8e06	378	return num;
3b8cb427	379
8e0af514	380	package = buffer.pointer;
c9ad8e06 LB	381
	382	if (package->type == ACPI_TYPE_PACKAGE &&
	383	package->package.count == 2 &&
	384	package->package.elements[0].integer.value == 1) /* rev 1 */
	385
	386	num = package->package.elements[1].integer.value;
	387
8e0af514	388	kfree(buffer.pointer);
c9ad8e06	389	return num;
8e0af514 SL	390	}
8e0af514 SL	391
8e0af514 SL	392	static void acpi_pad_handle_notify(acpi_handle handle)
8e0af514 SL	393	{
3b8cb427	394	int num_cpus;
8e0af514	395	uint32_t idle_cpus;
8b296d94 JL	396	struct acpi_buffer param = {
	397	.length = 4,
	398	.pointer = (void *)&idle_cpus,
	399	};
8e0af514 SL	400
8e0af514 SL	401	mutex_lock(&isolated_cpus_lock);
c9ad8e06 LB	402	num_cpus = acpi_pad_pur(handle);
c9ad8e06 LB	403	if (num_cpus < 0) {
8e0af514 SL	404	mutex_unlock(&isolated_cpus_lock);
	405	return;
	406	}
3b8cb427	407	acpi_pad_idle_cpus(num_cpus);
8e0af514	408	idle_cpus = acpi_pad_idle_cpus_num();
8b296d94	409	acpi_evaluate_ost(handle, ACPI_PROCESSOR_AGGREGATOR_NOTIFY, 0, &param);
8e0af514 SL	410	mutex_unlock(&isolated_cpus_lock);
	411	}
	412
	413	static void acpi_pad_notify(acpi_handle handle, u32 event,
	414	void *data)
	415	{
	416	struct acpi_device *device = data;
	417
	418	switch (event) {
	419	case ACPI_PROCESSOR_AGGREGATOR_NOTIFY:
	420	acpi_pad_handle_notify(handle);
8e0af514 SL	421	acpi_bus_generate_netlink_event(device->pnp.device_class,
	422	dev_name(&device->dev), event, 0);
	423	break;
	424	default:
73d4511a	425	pr_warn("Unsupported event [0x%x]\n", event);
8e0af514 SL	426	break;
	427	}
	428	}
	429
	430	static int acpi_pad_add(struct acpi_device *device)
	431	{
	432	acpi_status status;
	433
	434	strcpy(acpi_device_name(device), ACPI_PROCESSOR_AGGREGATOR_DEVICE_NAME);
	435	strcpy(acpi_device_class(device), ACPI_PROCESSOR_AGGREGATOR_CLASS);
	436
	437	if (acpi_pad_add_sysfs(device))
	438	return -ENODEV;
	439
	440	status = acpi_install_notify_handler(device->handle,
	441	ACPI_DEVICE_NOTIFY, acpi_pad_notify, device);
	442	if (ACPI_FAILURE(status)) {
	443	acpi_pad_remove_sysfs(device);
	444	return -ENODEV;
	445	}
	446
	447	return 0;
	448	}
	449
51fac838	450	static int acpi_pad_remove(struct acpi_device *device)
8e0af514 SL	451	{
	452	mutex_lock(&isolated_cpus_lock);
	453	acpi_pad_idle_cpus(0);
	454	mutex_unlock(&isolated_cpus_lock);
	455
	456	acpi_remove_notify_handler(device->handle,
	457	ACPI_DEVICE_NOTIFY, acpi_pad_notify);
	458	acpi_pad_remove_sysfs(device);
	459	return 0;
	460	}
	461
	462	static const struct acpi_device_id pad_device_ids[] = {
	463	{"ACPI000C", 0},
	464	{"", 0},
	465	};
	466	MODULE_DEVICE_TABLE(acpi, pad_device_ids);
	467
	468	static struct acpi_driver acpi_pad_driver = {
	469	.name = "processor_aggregator",
	470	.class = ACPI_PROCESSOR_AGGREGATOR_CLASS,
	471	.ids = pad_device_ids,
	472	.ops = {
	473	.add = acpi_pad_add,
	474	.remove = acpi_pad_remove,
	475	},
	476	};
	477
	478	static int __init acpi_pad_init(void)
	479	{
	480	power_saving_mwait_init();
	481	if (power_saving_mwait_eax == 0)
	482	return -EINVAL;
	483
	484	return acpi_bus_register_driver(&acpi_pad_driver);
	485	}
	486
	487	static void __exit acpi_pad_exit(void)
	488	{
	489	acpi_bus_unregister_driver(&acpi_pad_driver);
	490	}
	491
	492	module_init(acpi_pad_init);
	493	module_exit(acpi_pad_exit);
	494	MODULE_AUTHOR("Shaohua Li<shaohua.li@intel.com>");
	495	MODULE_DESCRIPTION("ACPI Processor Aggregator Driver");
	496	MODULE_LICENSE("GPL");