[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.
 *
 * You should have received a copy of the GNU General Public License along with
 * this program; if not, write to the Free Software Foundation, Inc.,
 * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
 *
 */

#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/clockchips.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 unsigned char lapic_detected_unstable;
static unsigned char lapic_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;
		if (!boot_cpu_has(X86_FEATURE_ARAT))
			lapic_detected_unstable = 1;
		break;
	default:
		/* TSC & LAPIC could halt in idle */
		tsc_detected_unstable = 1;
		lapic_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_thread_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()) {
		int cpu;
		unsigned long expire_time;

		try_to_freeze();

		/* 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;
			}
			if (lapic_detected_unstable && !lapic_marked_unstable) {
				int i;
				/* LAPIC could halt in idle, so notify users */
				for_each_online_cpu(i)
					clockevents_notify(
						CLOCK_EVT_NOTIFY_BROADCAST_ON,
						&i);
				lapic_marked_unstable = 1;
			}
			local_irq_disable();
			cpu = smp_processor_id();
			if (lapic_marked_unstable)
				clockevents_notify(
					CLOCK_EVT_NOTIFY_BROADCAST_ENTER, &cpu);
			stop_critical_timings();

			mwait_idle_with_hints(power_saving_mwait_eax, 1);

			start_critical_timings();
			if (lapic_marked_unstable)
				clockevents_notify(
					CLOCK_EVT_NOTIFY_BROADCAST_EXIT, &cpu);
			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	*
	15	* You should have received a copy of the GNU General Public License along with
	16	* this program; if not, write to the Free Software Foundation, Inc.,
	17	* 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
	18	*
	19	*/
	20
	21	#include <linux/kernel.h>
	22	#include <linux/cpumask.h>
	23	#include <linux/module.h>
	24	#include <linux/init.h>
	25	#include <linux/types.h>
	26	#include <linux/kthread.h>
	27	#include <linux/freezer.h>
	28	#include <linux/cpu.h>
	29	#include <linux/clockchips.h>
5a0e3ad6	30	#include <linux/slab.h>
8b48463f	31	#include <linux/acpi.h>
bc83cccc	32	#include <asm/mwait.h>
8e0af514	33
a40770a9	34	#define ACPI_PROCESSOR_AGGREGATOR_CLASS "acpi_pad"
8e0af514 SL	35	#define ACPI_PROCESSOR_AGGREGATOR_DEVICE_NAME "Processor Aggregator"
	36	#define ACPI_PROCESSOR_AGGREGATOR_NOTIFY 0x80
	37	static DEFINE_MUTEX(isolated_cpus_lock);
5f160126	38	static DEFINE_MUTEX(round_robin_lock);
8e0af514	39
8e0af514	40	static unsigned long power_saving_mwait_eax;
0dc698b9 VP	41
	42	static unsigned char tsc_detected_unstable;
	43	static unsigned char tsc_marked_unstable;
8aa4b14e CG	44	static unsigned char lapic_detected_unstable;
8aa4b14e CG	45	static unsigned char lapic_marked_unstable;
0dc698b9	46
8e0af514 SL	47	static void power_saving_mwait_init(void)
	48	{
	49	unsigned int eax, ebx, ecx, edx;
	50	unsigned int highest_cstate = 0;
	51	unsigned int highest_subcstate = 0;
	52	int i;
	53
	54	if (!boot_cpu_has(X86_FEATURE_MWAIT))
	55	return;
	56	if (boot_cpu_data.cpuid_level < CPUID_MWAIT_LEAF)
	57	return;
	58
	59	cpuid(CPUID_MWAIT_LEAF, &eax, &ebx, &ecx, &edx);
	60
	61	if (!(ecx & CPUID5_ECX_EXTENSIONS_SUPPORTED) \|\|
	62	!(ecx & CPUID5_ECX_INTERRUPT_BREAK))
	63	return;
	64
	65	edx >>= MWAIT_SUBSTATE_SIZE;
	66	for (i = 0; i < 7 && edx; i++, edx >>= MWAIT_SUBSTATE_SIZE) {
	67	if (edx & MWAIT_SUBSTATE_MASK) {
	68	highest_cstate = i;
	69	highest_subcstate = edx & MWAIT_SUBSTATE_MASK;
	70	}
	71	}
	72	power_saving_mwait_eax = (highest_cstate << MWAIT_SUBSTATE_SIZE) \|
	73	(highest_subcstate - 1);
	74
592913ec	75	#if defined(CONFIG_X86)
8e0af514 SL	76	switch (boot_cpu_data.x86_vendor) {
	77	case X86_VENDOR_AMD:
	78	case X86_VENDOR_INTEL:
	79	/*
	80	* AMD Fam10h TSC will tick in all
	81	* C/P/S0/S1 states when this bit is set.
	82	*/
8aa4b14e CG	83	if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC))
	84	tsc_detected_unstable = 1;
	85	if (!boot_cpu_has(X86_FEATURE_ARAT))
	86	lapic_detected_unstable = 1;
	87	break;
8e0af514	88	default:
8aa4b14e	89	/* TSC & LAPIC could halt in idle */
0dc698b9	90	tsc_detected_unstable = 1;
8aa4b14e	91	lapic_detected_unstable = 1;
8e0af514 SL	92	}
	93	#endif
	94	}
	95
	96	static unsigned long cpu_weight[NR_CPUS];
	97	static int tsk_in_cpu[NR_CPUS] = {[0 ... NR_CPUS-1] = -1};
	98	static DECLARE_BITMAP(pad_busy_cpus_bits, NR_CPUS);
	99	static void round_robin_cpu(unsigned int tsk_index)
	100	{
	101	struct cpumask *pad_busy_cpus = to_cpumask(pad_busy_cpus_bits);
	102	cpumask_var_t tmp;
	103	int cpu;
f67538f8 AM	104	unsigned long min_weight = -1;
f67538f8 AM	105	unsigned long uninitialized_var(preferred_cpu);
8e0af514 SL	106
	107	if (!alloc_cpumask_var(&tmp, GFP_KERNEL))
	108	return;
	109
5f160126	110	mutex_lock(&round_robin_lock);
8e0af514 SL	111	cpumask_clear(tmp);
	112	for_each_cpu(cpu, pad_busy_cpus)
	113	cpumask_or(tmp, tmp, topology_thread_cpumask(cpu));
	114	cpumask_andnot(tmp, cpu_online_mask, tmp);
	115	/* avoid HT sibilings if possible */
	116	if (cpumask_empty(tmp))
	117	cpumask_andnot(tmp, cpu_online_mask, pad_busy_cpus);
	118	if (cpumask_empty(tmp)) {
5f160126	119	mutex_unlock(&round_robin_lock);
8e0af514 SL	120	return;
	121	}
	122	for_each_cpu(cpu, tmp) {
	123	if (cpu_weight[cpu] < min_weight) {
	124	min_weight = cpu_weight[cpu];
	125	preferred_cpu = cpu;
	126	}
	127	}
	128
	129	if (tsk_in_cpu[tsk_index] != -1)
	130	cpumask_clear_cpu(tsk_in_cpu[tsk_index], pad_busy_cpus);
	131	tsk_in_cpu[tsk_index] = preferred_cpu;
	132	cpumask_set_cpu(preferred_cpu, pad_busy_cpus);
	133	cpu_weight[preferred_cpu]++;
5f160126	134	mutex_unlock(&round_robin_lock);
8e0af514 SL	135
	136	set_cpus_allowed_ptr(current, cpumask_of(preferred_cpu));
	137	}
	138
	139	static void exit_round_robin(unsigned int tsk_index)
	140	{
	141	struct cpumask *pad_busy_cpus = to_cpumask(pad_busy_cpus_bits);
	142	cpumask_clear_cpu(tsk_in_cpu[tsk_index], pad_busy_cpus);
	143	tsk_in_cpu[tsk_index] = -1;
	144	}
	145
	146	static unsigned int idle_pct = 5; /* percentage */
fa7584e1	147	static unsigned int round_robin_time = 1; /* second */
8e0af514 SL	148	static int power_saving_thread(void *data)
	149	{
	150	struct sched_param param = {.sched_priority = 1};
	151	int do_sleep;
	152	unsigned int tsk_index = (unsigned long)data;
	153	u64 last_jiffies = 0;
	154
	155	sched_setscheduler(current, SCHED_RR, &param);
	156
	157	while (!kthread_should_stop()) {
	158	int cpu;
4ff248f3	159	unsigned long expire_time;
8e0af514 SL	160
	161	try_to_freeze();
	162
	163	/* round robin to cpus */
4ff248f3 MS	164	expire_time = last_jiffies + round_robin_time * HZ;
4ff248f3 MS	165	if (time_before(expire_time, jiffies)) {
8e0af514 SL	166	last_jiffies = jiffies;
	167	round_robin_cpu(tsk_index);
	168	}
	169
	170	do_sleep = 0;
	171
8e0af514 SL	172	expire_time = jiffies + HZ * (100 - idle_pct) / 100;
	173
	174	while (!need_resched()) {
0dc698b9 VP	175	if (tsc_detected_unstable && !tsc_marked_unstable) {
	176	/* TSC could halt in idle, so notify users */
	177	mark_tsc_unstable("TSC halts in idle");
	178	tsc_marked_unstable = 1;
	179	}
8aa4b14e CG	180	if (lapic_detected_unstable && !lapic_marked_unstable) {
	181	int i;
	182	/* LAPIC could halt in idle, so notify users */
	183	for_each_online_cpu(i)
	184	clockevents_notify(
	185	CLOCK_EVT_NOTIFY_BROADCAST_ON,
	186	&i);
	187	lapic_marked_unstable = 1;
	188	}
8e0af514 SL	189	local_irq_disable();
8e0af514 SL	190	cpu = smp_processor_id();
8aa4b14e CG	191	if (lapic_marked_unstable)
	192	clockevents_notify(
	193	CLOCK_EVT_NOTIFY_BROADCAST_ENTER, &cpu);
8e0af514 SL	194	stop_critical_timings();
8e0af514 SL	195
16824255	196	mwait_idle_with_hints(power_saving_mwait_eax, 1);
8e0af514 SL	197
8e0af514 SL	198	start_critical_timings();
8aa4b14e CG	199	if (lapic_marked_unstable)
	200	clockevents_notify(
	201	CLOCK_EVT_NOTIFY_BROADCAST_EXIT, &cpu);
8e0af514 SL	202	local_irq_enable();
8e0af514 SL	203
4ff248f3	204	if (time_before(expire_time, jiffies)) {
8e0af514 SL	205	do_sleep = 1;
	206	break;
	207	}
	208	}
	209
8e0af514 SL	210	/*
	211	* current sched_rt has threshold for rt task running time.
	212	* When a rt task uses 95% CPU time, the rt thread will be
	213	* scheduled out for 5% CPU time to not starve other tasks. But
	214	* the mechanism only works when all CPUs have RT task running,
	215	* as if one CPU hasn't RT task, RT task from other CPUs will
	216	* borrow CPU time from this CPU and cause RT task use > 95%
3b8cb427	217	* CPU time. To make 'avoid starvation' work, takes a nap here.
8e0af514	218	*/
5b59c69e	219	if (unlikely(do_sleep))
8e0af514	220	schedule_timeout_killable(HZ * idle_pct / 100);
5b59c69e TC	221
	222	/* If an external event has set the need_resched flag, then
	223	* we need to deal with it, or this loop will continue to
	224	* spin without calling __mwait().
	225	*/
	226	if (unlikely(need_resched()))
	227	schedule();
8e0af514 SL	228	}
	229
	230	exit_round_robin(tsk_index);
	231	return 0;
	232	}
	233
	234	static struct task_struct *ps_tsks[NR_CPUS];
	235	static unsigned int ps_tsk_num;
	236	static int create_power_saving_task(void)
	237	{
5d7e4386	238	int rc;
3b8cb427	239
8e0af514 SL	240	ps_tsks[ps_tsk_num] = kthread_run(power_saving_thread,
8e0af514 SL	241	(void *)(unsigned long)ps_tsk_num,
150ed86f	242	"acpi_pad/%d", ps_tsk_num);
5d7e4386 RR	243
	244	if (IS_ERR(ps_tsks[ps_tsk_num])) {
	245	rc = PTR_ERR(ps_tsks[ps_tsk_num]);
3b8cb427	246	ps_tsks[ps_tsk_num] = NULL;
5d7e4386 RR	247	} else {
	248	rc = 0;
	249	ps_tsk_num++;
	250	}
3b8cb427 CG	251
3b8cb427 CG	252	return rc;
8e0af514 SL	253	}
	254
	255	static void destroy_power_saving_task(void)
	256	{
	257	if (ps_tsk_num > 0) {
	258	ps_tsk_num--;
	259	kthread_stop(ps_tsks[ps_tsk_num]);
3b8cb427	260	ps_tsks[ps_tsk_num] = NULL;
8e0af514 SL	261	}
	262	}
	263
	264	static void set_power_saving_task_num(unsigned int num)
	265	{
	266	if (num > ps_tsk_num) {
	267	while (ps_tsk_num < num) {
	268	if (create_power_saving_task())
	269	return;
	270	}
	271	} else if (num < ps_tsk_num) {
	272	while (ps_tsk_num > num)
	273	destroy_power_saving_task();
	274	}
	275	}
	276
3b8cb427	277	static void acpi_pad_idle_cpus(unsigned int num_cpus)
8e0af514 SL	278	{
	279	get_online_cpus();
	280
	281	num_cpus = min_t(unsigned int, num_cpus, num_online_cpus());
	282	set_power_saving_task_num(num_cpus);
	283
	284	put_online_cpus();
8e0af514 SL	285	}
	286
	287	static uint32_t acpi_pad_idle_cpus_num(void)
	288	{
	289	return ps_tsk_num;
	290	}
	291
	292	static ssize_t acpi_pad_rrtime_store(struct device *dev,
	293	struct device_attribute attr, const char buf, size_t count)
	294	{
	295	unsigned long num;
73d4511a	296	if (kstrtoul(buf, 0, &num))
8e0af514 SL	297	return -EINVAL;
	298	if (num < 1 \|\| num >= 100)
	299	return -EINVAL;
	300	mutex_lock(&isolated_cpus_lock);
	301	round_robin_time = num;
	302	mutex_unlock(&isolated_cpus_lock);
	303	return count;
	304	}
	305
	306	static ssize_t acpi_pad_rrtime_show(struct device *dev,
	307	struct device_attribute attr, char buf)
	308	{
32297abd	309	return scnprintf(buf, PAGE_SIZE, "%d\n", round_robin_time);
8e0af514 SL	310	}
	311	static DEVICE_ATTR(rrtime, S_IRUGO\|S_IWUSR,
	312	acpi_pad_rrtime_show,
	313	acpi_pad_rrtime_store);
	314
	315	static ssize_t acpi_pad_idlepct_store(struct device *dev,
	316	struct device_attribute attr, const char buf, size_t count)
	317	{
	318	unsigned long num;
73d4511a	319	if (kstrtoul(buf, 0, &num))
8e0af514 SL	320	return -EINVAL;
	321	if (num < 1 \|\| num >= 100)
	322	return -EINVAL;
	323	mutex_lock(&isolated_cpus_lock);
	324	idle_pct = num;
	325	mutex_unlock(&isolated_cpus_lock);
	326	return count;
	327	}
	328
	329	static ssize_t acpi_pad_idlepct_show(struct device *dev,
	330	struct device_attribute attr, char buf)
	331	{
32297abd	332	return scnprintf(buf, PAGE_SIZE, "%d\n", idle_pct);
8e0af514 SL	333	}
	334	static DEVICE_ATTR(idlepct, S_IRUGO\|S_IWUSR,
	335	acpi_pad_idlepct_show,
	336	acpi_pad_idlepct_store);
	337
	338	static ssize_t acpi_pad_idlecpus_store(struct device *dev,
	339	struct device_attribute attr, const char buf, size_t count)
	340	{
	341	unsigned long num;
73d4511a	342	if (kstrtoul(buf, 0, &num))
8e0af514 SL	343	return -EINVAL;
	344	mutex_lock(&isolated_cpus_lock);
	345	acpi_pad_idle_cpus(num);
	346	mutex_unlock(&isolated_cpus_lock);
	347	return count;
	348	}
	349
	350	static ssize_t acpi_pad_idlecpus_show(struct device *dev,
	351	struct device_attribute attr, char buf)
	352	{
5aaba363 SH	353	return cpumap_print_to_pagebuf(false, buf,
5aaba363 SH	354	to_cpumask(pad_busy_cpus_bits));
8e0af514	355	}
5aaba363	356
8e0af514 SL	357	static DEVICE_ATTR(idlecpus, S_IRUGO\|S_IWUSR,
	358	acpi_pad_idlecpus_show,
	359	acpi_pad_idlecpus_store);
	360
	361	static int acpi_pad_add_sysfs(struct acpi_device *device)
	362	{
	363	int result;
	364
	365	result = device_create_file(&device->dev, &dev_attr_idlecpus);
	366	if (result)
	367	return -ENODEV;
	368	result = device_create_file(&device->dev, &dev_attr_idlepct);
	369	if (result) {
	370	device_remove_file(&device->dev, &dev_attr_idlecpus);
	371	return -ENODEV;
	372	}
	373	result = device_create_file(&device->dev, &dev_attr_rrtime);
	374	if (result) {
	375	device_remove_file(&device->dev, &dev_attr_idlecpus);
	376	device_remove_file(&device->dev, &dev_attr_idlepct);
	377	return -ENODEV;
	378	}
	379	return 0;
	380	}
	381
	382	static void acpi_pad_remove_sysfs(struct acpi_device *device)
	383	{
	384	device_remove_file(&device->dev, &dev_attr_idlecpus);
	385	device_remove_file(&device->dev, &dev_attr_idlepct);
	386	device_remove_file(&device->dev, &dev_attr_rrtime);
	387	}
	388
c9ad8e06 LB	389	/*
	390	* Query firmware how many CPUs should be idle
	391	* return -1 on failure
	392	*/
	393	static int acpi_pad_pur(acpi_handle handle)
8e0af514 SL	394	{
8e0af514 SL	395	struct acpi_buffer buffer = {ACPI_ALLOCATE_BUFFER, NULL};
8e0af514	396	union acpi_object *package;
c9ad8e06	397	int num = -1;
8e0af514	398
3b8cb427	399	if (ACPI_FAILURE(acpi_evaluate_object(handle, "_PUR", NULL, &buffer)))
c9ad8e06	400	return num;
3b8cb427 CG	401
3b8cb427 CG	402	if (!buffer.length \|\| !buffer.pointer)
c9ad8e06	403	return num;
3b8cb427	404
8e0af514	405	package = buffer.pointer;
c9ad8e06 LB	406
	407	if (package->type == ACPI_TYPE_PACKAGE &&
	408	package->package.count == 2 &&
	409	package->package.elements[0].integer.value == 1) /* rev 1 */
	410
	411	num = package->package.elements[1].integer.value;
	412
8e0af514	413	kfree(buffer.pointer);
c9ad8e06	414	return num;
8e0af514 SL	415	}
8e0af514 SL	416
8e0af514 SL	417	static void acpi_pad_handle_notify(acpi_handle handle)
8e0af514 SL	418	{
3b8cb427	419	int num_cpus;
8e0af514	420	uint32_t idle_cpus;
8b296d94 JL	421	struct acpi_buffer param = {
	422	.length = 4,
	423	.pointer = (void *)&idle_cpus,
	424	};
8e0af514 SL	425
8e0af514 SL	426	mutex_lock(&isolated_cpus_lock);
c9ad8e06 LB	427	num_cpus = acpi_pad_pur(handle);
c9ad8e06 LB	428	if (num_cpus < 0) {
8e0af514 SL	429	mutex_unlock(&isolated_cpus_lock);
	430	return;
	431	}
3b8cb427	432	acpi_pad_idle_cpus(num_cpus);
8e0af514	433	idle_cpus = acpi_pad_idle_cpus_num();
8b296d94	434	acpi_evaluate_ost(handle, ACPI_PROCESSOR_AGGREGATOR_NOTIFY, 0, &param);
8e0af514 SL	435	mutex_unlock(&isolated_cpus_lock);
	436	}
	437
	438	static void acpi_pad_notify(acpi_handle handle, u32 event,
	439	void *data)
	440	{
	441	struct acpi_device *device = data;
	442
	443	switch (event) {
	444	case ACPI_PROCESSOR_AGGREGATOR_NOTIFY:
	445	acpi_pad_handle_notify(handle);
8e0af514 SL	446	acpi_bus_generate_netlink_event(device->pnp.device_class,
	447	dev_name(&device->dev), event, 0);
	448	break;
	449	default:
73d4511a	450	pr_warn("Unsupported event [0x%x]\n", event);
8e0af514 SL	451	break;
	452	}
	453	}
	454
	455	static int acpi_pad_add(struct acpi_device *device)
	456	{
	457	acpi_status status;
	458
	459	strcpy(acpi_device_name(device), ACPI_PROCESSOR_AGGREGATOR_DEVICE_NAME);
	460	strcpy(acpi_device_class(device), ACPI_PROCESSOR_AGGREGATOR_CLASS);
	461
	462	if (acpi_pad_add_sysfs(device))
	463	return -ENODEV;
	464
	465	status = acpi_install_notify_handler(device->handle,
	466	ACPI_DEVICE_NOTIFY, acpi_pad_notify, device);
	467	if (ACPI_FAILURE(status)) {
	468	acpi_pad_remove_sysfs(device);
	469	return -ENODEV;
	470	}
	471
	472	return 0;
	473	}
	474
51fac838	475	static int acpi_pad_remove(struct acpi_device *device)
8e0af514 SL	476	{
	477	mutex_lock(&isolated_cpus_lock);
	478	acpi_pad_idle_cpus(0);
	479	mutex_unlock(&isolated_cpus_lock);
	480
	481	acpi_remove_notify_handler(device->handle,
	482	ACPI_DEVICE_NOTIFY, acpi_pad_notify);
	483	acpi_pad_remove_sysfs(device);
	484	return 0;
	485	}
	486
	487	static const struct acpi_device_id pad_device_ids[] = {
	488	{"ACPI000C", 0},
	489	{"", 0},
	490	};
	491	MODULE_DEVICE_TABLE(acpi, pad_device_ids);
	492
	493	static struct acpi_driver acpi_pad_driver = {
	494	.name = "processor_aggregator",
	495	.class = ACPI_PROCESSOR_AGGREGATOR_CLASS,
	496	.ids = pad_device_ids,
	497	.ops = {
	498	.add = acpi_pad_add,
	499	.remove = acpi_pad_remove,
	500	},
	501	};
	502
	503	static int __init acpi_pad_init(void)
	504	{
	505	power_saving_mwait_init();
	506	if (power_saving_mwait_eax == 0)
	507	return -EINVAL;
	508
	509	return acpi_bus_register_driver(&acpi_pad_driver);
	510	}
	511
	512	static void __exit acpi_pad_exit(void)
	513	{
	514	acpi_bus_unregister_driver(&acpi_pad_driver);
	515	}
	516
	517	module_init(acpi_pad_init);
	518	module_exit(acpi_pad_exit);
	519	MODULE_AUTHOR("Shaohua Li<shaohua.li@intel.com>");
	520	MODULE_DESCRIPTION("ACPI Processor Aggregator Driver");
	521	MODULE_LICENSE("GPL");