[net-next-2.6.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 <acpi/acpi_bus.h>
#include <acpi/acpi_drivers.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 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(&isolated_cpus_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(&isolated_cpus_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(&isolated_cpus_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 = 10; /* 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;
		u64 expire_time;

		try_to_freeze();

		/* round robin to cpus */
		if (last_jiffies + round_robin_time * HZ < 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();

			__monitor((void *)&current_thread_info()->flags, 0, 0);
			smp_mb();
			if (!need_resched())
				__mwait(power_saving_mwait_eax, 1);

			start_critical_timings();
			if (lapic_marked_unstable)
				clockevents_notify(
					CLOCK_EVT_NOTIFY_BROADCAST_EXIT, &cpu);
			local_irq_enable();

			if (jiffies > expire_time) {
				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 (do_sleep)
			schedule_timeout_killable(HZ * idle_pct / 100);
	}

	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 = -ENOMEM;

	ps_tsks[ps_tsk_num] = kthread_run(power_saving_thread,
		(void *)(unsigned long)ps_tsk_num,
		"power_saving/%d", ps_tsk_num);
	rc = IS_ERR(ps_tsks[ps_tsk_num]) ? PTR_ERR(ps_tsks[ps_tsk_num]) : 0;
	if (!rc)
		ps_tsk_num++;
	else
		ps_tsks[ps_tsk_num] = NULL;

	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 (strict_strtoul(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", 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 (strict_strtoul(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", 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 (strict_strtoul(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 cpumask_scnprintf(buf, PAGE_SIZE,
		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;
}

/* Notify firmware how many CPUs are idle */
static void acpi_pad_ost(acpi_handle handle, int stat,
	uint32_t idle_cpus)
{
	union acpi_object params[3] = {
		{.type = ACPI_TYPE_INTEGER,},
		{.type = ACPI_TYPE_INTEGER,},
		{.type = ACPI_TYPE_BUFFER,},
	};
	struct acpi_object_list arg_list = {3, params};

	params[0].integer.value = ACPI_PROCESSOR_AGGREGATOR_NOTIFY;
	params[1].integer.value =  stat;
	params[2].buffer.length = 4;
	params[2].buffer.pointer = (void *)&idle_cpus;
	acpi_evaluate_object(handle, "_OST", &arg_list, NULL);
}

static void acpi_pad_handle_notify(acpi_handle handle)
{
	int num_cpus;
	uint32_t 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_pad_ost(handle, 0, idle_cpus);
	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_proc_event(device, event, 0);
		acpi_bus_generate_netlink_event(device->pnp.device_class,
			dev_name(&device->dev), event, 0);
		break;
	default:
		printk(KERN_WARNING"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,
	int type)
{
	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>
8e0af514 SL	31	#include <acpi/acpi_bus.h>
8e0af514 SL	32	#include <acpi/acpi_drivers.h>
bc83cccc	33	#include <asm/mwait.h>
8e0af514	34
a40770a9	35	#define ACPI_PROCESSOR_AGGREGATOR_CLASS "acpi_pad"
8e0af514 SL	36	#define ACPI_PROCESSOR_AGGREGATOR_DEVICE_NAME "Processor Aggregator"
	37	#define ACPI_PROCESSOR_AGGREGATOR_NOTIFY 0x80
	38	static DEFINE_MUTEX(isolated_cpus_lock);
	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
	110	mutex_lock(&isolated_cpus_lock);
	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)) {
	119	mutex_unlock(&isolated_cpus_lock);
	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]++;
	134	mutex_unlock(&isolated_cpus_lock);
	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 */
	147	static unsigned int round_robin_time = 10; /* second */
	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;
	159	u64 expire_time;
	160
	161	try_to_freeze();
	162
	163	/* round robin to cpus */
	164	if (last_jiffies + round_robin_time * HZ < jiffies) {
	165	last_jiffies = jiffies;
	166	round_robin_cpu(tsk_index);
	167	}
	168
	169	do_sleep = 0;
170
8e0af514 SL	171	expire_time = jiffies + HZ * (100 - idle_pct) / 100;
	172
	173	while (!need_resched()) {
0dc698b9 VP	174	if (tsc_detected_unstable && !tsc_marked_unstable) {
	175	/* TSC could halt in idle, so notify users */
	176	mark_tsc_unstable("TSC halts in idle");
	177	tsc_marked_unstable = 1;
	178	}
8aa4b14e CG	179	if (lapic_detected_unstable && !lapic_marked_unstable) {
	180	int i;
	181	/* LAPIC could halt in idle, so notify users */
	182	for_each_online_cpu(i)
	183	clockevents_notify(
	184	CLOCK_EVT_NOTIFY_BROADCAST_ON,
	185	&i);
	186	lapic_marked_unstable = 1;
	187	}
8e0af514 SL	188	local_irq_disable();
8e0af514 SL	189	cpu = smp_processor_id();
8aa4b14e CG	190	if (lapic_marked_unstable)
	191	clockevents_notify(
	192	CLOCK_EVT_NOTIFY_BROADCAST_ENTER, &cpu);
8e0af514 SL	193	stop_critical_timings();
	194
	195	__monitor((void *)&current_thread_info()->flags, 0, 0);
	196	smp_mb();
	197	if (!need_resched())
	198	__mwait(power_saving_mwait_eax, 1);
	199
	200	start_critical_timings();
8aa4b14e CG	201	if (lapic_marked_unstable)
	202	clockevents_notify(
	203	CLOCK_EVT_NOTIFY_BROADCAST_EXIT, &cpu);
8e0af514 SL	204	local_irq_enable();
	205
	206	if (jiffies > expire_time) {
	207	do_sleep = 1;
	208	break;
	209	}
	210	}
	211
8e0af514 SL	212	/*
	213	* current sched_rt has threshold for rt task running time.
	214	* When a rt task uses 95% CPU time, the rt thread will be
	215	* scheduled out for 5% CPU time to not starve other tasks. But
	216	* the mechanism only works when all CPUs have RT task running,
	217	* as if one CPU hasn't RT task, RT task from other CPUs will
	218	* borrow CPU time from this CPU and cause RT task use > 95%
3b8cb427	219	* CPU time. To make 'avoid starvation' work, takes a nap here.
8e0af514 SL	220	*/
	221	if (do_sleep)
	222	schedule_timeout_killable(HZ * idle_pct / 100);
	223	}
	224
	225	exit_round_robin(tsk_index);
	226	return 0;
	227	}
	228
	229	static struct task_struct *ps_tsks[NR_CPUS];
	230	static unsigned int ps_tsk_num;
	231	static int create_power_saving_task(void)
	232	{
3b8cb427 CG	233	int rc = -ENOMEM;
3b8cb427 CG	234
8e0af514 SL	235	ps_tsks[ps_tsk_num] = kthread_run(power_saving_thread,
	236	(void *)(unsigned long)ps_tsk_num,
	237	"power_saving/%d", ps_tsk_num);
3b8cb427 CG	238	rc = IS_ERR(ps_tsks[ps_tsk_num]) ? PTR_ERR(ps_tsks[ps_tsk_num]) : 0;
3b8cb427 CG	239	if (!rc)
8e0af514	240	ps_tsk_num++;
3b8cb427 CG	241	else
	242	ps_tsks[ps_tsk_num] = NULL;
	243
	244	return rc;
8e0af514 SL	245	}
	246
	247	static void destroy_power_saving_task(void)
	248	{
	249	if (ps_tsk_num > 0) {
	250	ps_tsk_num--;
	251	kthread_stop(ps_tsks[ps_tsk_num]);
3b8cb427	252	ps_tsks[ps_tsk_num] = NULL;
8e0af514 SL	253	}
	254	}
	255
	256	static void set_power_saving_task_num(unsigned int num)
	257	{
	258	if (num > ps_tsk_num) {
	259	while (ps_tsk_num < num) {
	260	if (create_power_saving_task())
	261	return;
	262	}
	263	} else if (num < ps_tsk_num) {
	264	while (ps_tsk_num > num)
	265	destroy_power_saving_task();
	266	}
	267	}
	268
3b8cb427	269	static void acpi_pad_idle_cpus(unsigned int num_cpus)
8e0af514 SL	270	{
	271	get_online_cpus();
	272
	273	num_cpus = min_t(unsigned int, num_cpus, num_online_cpus());
	274	set_power_saving_task_num(num_cpus);
	275
	276	put_online_cpus();
8e0af514 SL	277	}
	278
	279	static uint32_t acpi_pad_idle_cpus_num(void)
	280	{
	281	return ps_tsk_num;
	282	}
	283
	284	static ssize_t acpi_pad_rrtime_store(struct device *dev,
	285	struct device_attribute attr, const char buf, size_t count)
	286	{
	287	unsigned long num;
	288	if (strict_strtoul(buf, 0, &num))
	289	return -EINVAL;
	290	if (num < 1 \|\| num >= 100)
	291	return -EINVAL;
	292	mutex_lock(&isolated_cpus_lock);
	293	round_robin_time = num;
	294	mutex_unlock(&isolated_cpus_lock);
	295	return count;
	296	}
	297
	298	static ssize_t acpi_pad_rrtime_show(struct device *dev,
	299	struct device_attribute attr, char buf)
	300	{
	301	return scnprintf(buf, PAGE_SIZE, "%d", round_robin_time);
	302	}
	303	static DEVICE_ATTR(rrtime, S_IRUGO\|S_IWUSR,
	304	acpi_pad_rrtime_show,
	305	acpi_pad_rrtime_store);
	306
	307	static ssize_t acpi_pad_idlepct_store(struct device *dev,
	308	struct device_attribute attr, const char buf, size_t count)
	309	{
	310	unsigned long num;
	311	if (strict_strtoul(buf, 0, &num))
	312	return -EINVAL;
	313	if (num < 1 \|\| num >= 100)
	314	return -EINVAL;
	315	mutex_lock(&isolated_cpus_lock);
	316	idle_pct = num;
	317	mutex_unlock(&isolated_cpus_lock);
	318	return count;
	319	}
	320
	321	static ssize_t acpi_pad_idlepct_show(struct device *dev,
	322	struct device_attribute attr, char buf)
	323	{
	324	return scnprintf(buf, PAGE_SIZE, "%d", idle_pct);
	325	}
	326	static DEVICE_ATTR(idlepct, S_IRUGO\|S_IWUSR,
	327	acpi_pad_idlepct_show,
	328	acpi_pad_idlepct_store);
	329
	330	static ssize_t acpi_pad_idlecpus_store(struct device *dev,
	331	struct device_attribute attr, const char buf, size_t count)
	332	{
	333	unsigned long num;
	334	if (strict_strtoul(buf, 0, &num))
	335	return -EINVAL;
	336	mutex_lock(&isolated_cpus_lock);
	337	acpi_pad_idle_cpus(num);
	338	mutex_unlock(&isolated_cpus_lock);
	339	return count;
	340	}
341
342	static ssize_t acpi_pad_idlecpus_show(struct device *dev,
343	struct device_attribute attr, char buf)
344	{
345	return cpumask_scnprintf(buf, PAGE_SIZE,
346	to_cpumask(pad_busy_cpus_bits));
347	}
348	static DEVICE_ATTR(idlecpus, S_IRUGO\|S_IWUSR,
349	acpi_pad_idlecpus_show,
350	acpi_pad_idlecpus_store);
351
352	static int acpi_pad_add_sysfs(struct acpi_device *device)
353	{
354	int result;
355
356	result = device_create_file(&device->dev, &dev_attr_idlecpus);
357	if (result)
358	return -ENODEV;
359	result = device_create_file(&device->dev, &dev_attr_idlepct);
360	if (result) {
361	device_remove_file(&device->dev, &dev_attr_idlecpus);
362	return -ENODEV;
363	}
364	result = device_create_file(&device->dev, &dev_attr_rrtime);
365	if (result) {
366	device_remove_file(&device->dev, &dev_attr_idlecpus);
367	device_remove_file(&device->dev, &dev_attr_idlepct);
368	return -ENODEV;
369	}
370	return 0;
371	}
372
373	static void acpi_pad_remove_sysfs(struct acpi_device *device)
374	{
375	device_remove_file(&device->dev, &dev_attr_idlecpus);
376	device_remove_file(&device->dev, &dev_attr_idlepct);
377	device_remove_file(&device->dev, &dev_attr_rrtime);
378	}
379
c9ad8e06 LB	380	/*
	381	* Query firmware how many CPUs should be idle
	382	* return -1 on failure
	383	*/
	384	static int acpi_pad_pur(acpi_handle handle)
8e0af514 SL	385	{
8e0af514 SL	386	struct acpi_buffer buffer = {ACPI_ALLOCATE_BUFFER, NULL};
8e0af514	387	union acpi_object *package;
c9ad8e06	388	int num = -1;
8e0af514	389
3b8cb427	390	if (ACPI_FAILURE(acpi_evaluate_object(handle, "_PUR", NULL, &buffer)))
c9ad8e06	391	return num;
3b8cb427 CG	392
3b8cb427 CG	393	if (!buffer.length \|\| !buffer.pointer)
c9ad8e06	394	return num;
3b8cb427	395
8e0af514	396	package = buffer.pointer;
c9ad8e06 LB	397
	398	if (package->type == ACPI_TYPE_PACKAGE &&
	399	package->package.count == 2 &&
	400	package->package.elements[0].integer.value == 1) /* rev 1 */
	401
	402	num = package->package.elements[1].integer.value;
	403
8e0af514	404	kfree(buffer.pointer);
c9ad8e06	405	return num;
8e0af514 SL	406	}
	407
	408	/* Notify firmware how many CPUs are idle */
	409	static void acpi_pad_ost(acpi_handle handle, int stat,
	410	uint32_t idle_cpus)
	411	{
	412	union acpi_object params[3] = {
	413	{.type = ACPI_TYPE_INTEGER,},
	414	{.type = ACPI_TYPE_INTEGER,},
	415	{.type = ACPI_TYPE_BUFFER,},
	416	};
	417	struct acpi_object_list arg_list = {3, params};
	418
	419	params[0].integer.value = ACPI_PROCESSOR_AGGREGATOR_NOTIFY;
	420	params[1].integer.value = stat;
	421	params[2].buffer.length = 4;
	422	params[2].buffer.pointer = (void *)&idle_cpus;
	423	acpi_evaluate_object(handle, "_OST", &arg_list, NULL);
	424	}
	425
	426	static void acpi_pad_handle_notify(acpi_handle handle)
	427	{
3b8cb427	428	int num_cpus;
8e0af514 SL	429	uint32_t idle_cpus;
	430
	431	mutex_lock(&isolated_cpus_lock);
c9ad8e06 LB	432	num_cpus = acpi_pad_pur(handle);
c9ad8e06 LB	433	if (num_cpus < 0) {
8e0af514 SL	434	mutex_unlock(&isolated_cpus_lock);
	435	return;
	436	}
3b8cb427	437	acpi_pad_idle_cpus(num_cpus);
8e0af514	438	idle_cpus = acpi_pad_idle_cpus_num();
3b8cb427	439	acpi_pad_ost(handle, 0, idle_cpus);
8e0af514 SL	440	mutex_unlock(&isolated_cpus_lock);
	441	}
	442
	443	static void acpi_pad_notify(acpi_handle handle, u32 event,
	444	void *data)
	445	{
	446	struct acpi_device *device = data;
	447
	448	switch (event) {
	449	case ACPI_PROCESSOR_AGGREGATOR_NOTIFY:
	450	acpi_pad_handle_notify(handle);
	451	acpi_bus_generate_proc_event(device, event, 0);
	452	acpi_bus_generate_netlink_event(device->pnp.device_class,
	453	dev_name(&device->dev), event, 0);
	454	break;
	455	default:
	456	printk(KERN_WARNING"Unsupported event [0x%x]\n", event);
	457	break;
	458	}
	459	}
	460
	461	static int acpi_pad_add(struct acpi_device *device)
	462	{
	463	acpi_status status;
	464
	465	strcpy(acpi_device_name(device), ACPI_PROCESSOR_AGGREGATOR_DEVICE_NAME);
	466	strcpy(acpi_device_class(device), ACPI_PROCESSOR_AGGREGATOR_CLASS);
	467
	468	if (acpi_pad_add_sysfs(device))
	469	return -ENODEV;
	470
	471	status = acpi_install_notify_handler(device->handle,
	472	ACPI_DEVICE_NOTIFY, acpi_pad_notify, device);
	473	if (ACPI_FAILURE(status)) {
	474	acpi_pad_remove_sysfs(device);
	475	return -ENODEV;
	476	}
	477
	478	return 0;
	479	}
	480
	481	static int acpi_pad_remove(struct acpi_device *device,
	482	int type)
	483	{
	484	mutex_lock(&isolated_cpus_lock);
	485	acpi_pad_idle_cpus(0);
	486	mutex_unlock(&isolated_cpus_lock);
	487
	488	acpi_remove_notify_handler(device->handle,
	489	ACPI_DEVICE_NOTIFY, acpi_pad_notify);
	490	acpi_pad_remove_sysfs(device);
	491	return 0;
	492	}
	493
	494	static const struct acpi_device_id pad_device_ids[] = {
	495	{"ACPI000C", 0},
	496	{"", 0},
	497	};
	498	MODULE_DEVICE_TABLE(acpi, pad_device_ids);
	499
	500	static struct acpi_driver acpi_pad_driver = {
	501	.name = "processor_aggregator",
	502	.class = ACPI_PROCESSOR_AGGREGATOR_CLASS,
	503	.ids = pad_device_ids,
504	.ops = {
505	.add = acpi_pad_add,
506	.remove = acpi_pad_remove,
507	},
508	};
509
510	static int __init acpi_pad_init(void)
511	{
512	power_saving_mwait_init();
513	if (power_saving_mwait_eax == 0)
514	return -EINVAL;
515
516	return acpi_bus_register_driver(&acpi_pad_driver);
517	}
518
519	static void __exit acpi_pad_exit(void)
520	{
521	acpi_bus_unregister_driver(&acpi_pad_driver);
522	}
523
524	module_init(acpi_pad_init);
525	module_exit(acpi_pad_exit);
526	MODULE_AUTHOR("Shaohua Li<shaohua.li@intel.com>");
527	MODULE_DESCRIPTION("ACPI Processor Aggregator Driver");
528	MODULE_LICENSE("GPL");