[mirror_ubuntu-artful-kernel.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 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;
		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;

	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)
{
	int n = 0;
	n = cpumask_scnprintf(buf, PAGE_SIZE-2, to_cpumask(pad_busy_cpus_bits));
	buf[n++] = '\n';
	buf[n] = '\0';
	return n;
}
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_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>
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);
5f160126	39	static DEFINE_MUTEX(round_robin_lock);
8e0af514	40
8e0af514	41	static unsigned long power_saving_mwait_eax;
0dc698b9 VP	42
	43	static unsigned char tsc_detected_unstable;
	44	static unsigned char tsc_marked_unstable;
8aa4b14e CG	45	static unsigned char lapic_detected_unstable;
8aa4b14e CG	46	static unsigned char lapic_marked_unstable;
0dc698b9	47
8e0af514 SL	48	static void power_saving_mwait_init(void)
	49	{
	50	unsigned int eax, ebx, ecx, edx;
	51	unsigned int highest_cstate = 0;
	52	unsigned int highest_subcstate = 0;
	53	int i;
	54
	55	if (!boot_cpu_has(X86_FEATURE_MWAIT))
	56	return;
	57	if (boot_cpu_data.cpuid_level < CPUID_MWAIT_LEAF)
	58	return;
	59
	60	cpuid(CPUID_MWAIT_LEAF, &eax, &ebx, &ecx, &edx);
	61
	62	if (!(ecx & CPUID5_ECX_EXTENSIONS_SUPPORTED) \|\|
	63	!(ecx & CPUID5_ECX_INTERRUPT_BREAK))
	64	return;
	65
	66	edx >>= MWAIT_SUBSTATE_SIZE;
	67	for (i = 0; i < 7 && edx; i++, edx >>= MWAIT_SUBSTATE_SIZE) {
	68	if (edx & MWAIT_SUBSTATE_MASK) {
	69	highest_cstate = i;
	70	highest_subcstate = edx & MWAIT_SUBSTATE_MASK;
	71	}
	72	}
	73	power_saving_mwait_eax = (highest_cstate << MWAIT_SUBSTATE_SIZE) \|
	74	(highest_subcstate - 1);
	75
592913ec	76	#if defined(CONFIG_X86)
8e0af514 SL	77	switch (boot_cpu_data.x86_vendor) {
	78	case X86_VENDOR_AMD:
	79	case X86_VENDOR_INTEL:
	80	/*
	81	* AMD Fam10h TSC will tick in all
	82	* C/P/S0/S1 states when this bit is set.
	83	*/
8aa4b14e CG	84	if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC))
	85	tsc_detected_unstable = 1;
	86	if (!boot_cpu_has(X86_FEATURE_ARAT))
	87	lapic_detected_unstable = 1;
	88	break;
8e0af514	89	default:
8aa4b14e	90	/* TSC & LAPIC could halt in idle */
0dc698b9	91	tsc_detected_unstable = 1;
8aa4b14e	92	lapic_detected_unstable = 1;
8e0af514 SL	93	}
	94	#endif
	95	}
	96
	97	static unsigned long cpu_weight[NR_CPUS];
	98	static int tsk_in_cpu[NR_CPUS] = {[0 ... NR_CPUS-1] = -1};
	99	static DECLARE_BITMAP(pad_busy_cpus_bits, NR_CPUS);
	100	static void round_robin_cpu(unsigned int tsk_index)
	101	{
	102	struct cpumask *pad_busy_cpus = to_cpumask(pad_busy_cpus_bits);
	103	cpumask_var_t tmp;
	104	int cpu;
f67538f8 AM	105	unsigned long min_weight = -1;
f67538f8 AM	106	unsigned long uninitialized_var(preferred_cpu);
8e0af514 SL	107
	108	if (!alloc_cpumask_var(&tmp, GFP_KERNEL))
	109	return;
	110
5f160126	111	mutex_lock(&round_robin_lock);
8e0af514 SL	112	cpumask_clear(tmp);
	113	for_each_cpu(cpu, pad_busy_cpus)
	114	cpumask_or(tmp, tmp, topology_thread_cpumask(cpu));
	115	cpumask_andnot(tmp, cpu_online_mask, tmp);
	116	/* avoid HT sibilings if possible */
	117	if (cpumask_empty(tmp))
	118	cpumask_andnot(tmp, cpu_online_mask, pad_busy_cpus);
	119	if (cpumask_empty(tmp)) {
5f160126	120	mutex_unlock(&round_robin_lock);
8e0af514 SL	121	return;
	122	}
	123	for_each_cpu(cpu, tmp) {
	124	if (cpu_weight[cpu] < min_weight) {
	125	min_weight = cpu_weight[cpu];
	126	preferred_cpu = cpu;
	127	}
	128	}
	129
	130	if (tsk_in_cpu[tsk_index] != -1)
	131	cpumask_clear_cpu(tsk_in_cpu[tsk_index], pad_busy_cpus);
	132	tsk_in_cpu[tsk_index] = preferred_cpu;
	133	cpumask_set_cpu(preferred_cpu, pad_busy_cpus);
	134	cpu_weight[preferred_cpu]++;
5f160126	135	mutex_unlock(&round_robin_lock);
8e0af514 SL	136
	137	set_cpus_allowed_ptr(current, cpumask_of(preferred_cpu));
	138	}
	139
	140	static void exit_round_robin(unsigned int tsk_index)
	141	{
	142	struct cpumask *pad_busy_cpus = to_cpumask(pad_busy_cpus_bits);
	143	cpumask_clear_cpu(tsk_in_cpu[tsk_index], pad_busy_cpus);
	144	tsk_in_cpu[tsk_index] = -1;
	145	}
	146
	147	static unsigned int idle_pct = 5; /* percentage */
fa7584e1	148	static unsigned int round_robin_time = 1; /* second */
8e0af514 SL	149	static int power_saving_thread(void *data)
	150	{
	151	struct sched_param param = {.sched_priority = 1};
	152	int do_sleep;
	153	unsigned int tsk_index = (unsigned long)data;
	154	u64 last_jiffies = 0;
	155
	156	sched_setscheduler(current, SCHED_RR, &param);
	157
	158	while (!kthread_should_stop()) {
	159	int cpu;
	160	u64 expire_time;
	161
	162	try_to_freeze();
	163
	164	/* round robin to cpus */
	165	if (last_jiffies + round_robin_time * HZ < jiffies) {
	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();
	195
	196	__monitor((void *)&current_thread_info()->flags, 0, 0);
	197	smp_mb();
	198	if (!need_resched())
	199	__mwait(power_saving_mwait_eax, 1);
	200
	201	start_critical_timings();
8aa4b14e CG	202	if (lapic_marked_unstable)
	203	clockevents_notify(
	204	CLOCK_EVT_NOTIFY_BROADCAST_EXIT, &cpu);
8e0af514 SL	205	local_irq_enable();
	206
	207	if (jiffies > expire_time) {
	208	do_sleep = 1;
	209	break;
	210	}
	211	}
	212
8e0af514 SL	213	/*
	214	* current sched_rt has threshold for rt task running time.
	215	* When a rt task uses 95% CPU time, the rt thread will be
	216	* scheduled out for 5% CPU time to not starve other tasks. But
	217	* the mechanism only works when all CPUs have RT task running,
	218	* as if one CPU hasn't RT task, RT task from other CPUs will
	219	* borrow CPU time from this CPU and cause RT task use > 95%
3b8cb427	220	* CPU time. To make 'avoid starvation' work, takes a nap here.
8e0af514 SL	221	*/
	222	if (do_sleep)
	223	schedule_timeout_killable(HZ * idle_pct / 100);
	224	}
	225
	226	exit_round_robin(tsk_index);
	227	return 0;
	228	}
	229
	230	static struct task_struct *ps_tsks[NR_CPUS];
	231	static unsigned int ps_tsk_num;
	232	static int create_power_saving_task(void)
	233	{
5d7e4386	234	int rc;
3b8cb427	235
8e0af514 SL	236	ps_tsks[ps_tsk_num] = kthread_run(power_saving_thread,
8e0af514 SL	237	(void *)(unsigned long)ps_tsk_num,
150ed86f	238	"acpi_pad/%d", ps_tsk_num);
5d7e4386 RR	239
	240	if (IS_ERR(ps_tsks[ps_tsk_num])) {
	241	rc = PTR_ERR(ps_tsks[ps_tsk_num]);
3b8cb427	242	ps_tsks[ps_tsk_num] = NULL;
5d7e4386 RR	243	} else {
	244	rc = 0;
	245	ps_tsk_num++;
	246	}
3b8cb427 CG	247
3b8cb427 CG	248	return rc;
8e0af514 SL	249	}
	250
	251	static void destroy_power_saving_task(void)
	252	{
	253	if (ps_tsk_num > 0) {
	254	ps_tsk_num--;
	255	kthread_stop(ps_tsks[ps_tsk_num]);
3b8cb427	256	ps_tsks[ps_tsk_num] = NULL;
8e0af514 SL	257	}
	258	}
	259
	260	static void set_power_saving_task_num(unsigned int num)
	261	{
	262	if (num > ps_tsk_num) {
	263	while (ps_tsk_num < num) {
	264	if (create_power_saving_task())
	265	return;
	266	}
	267	} else if (num < ps_tsk_num) {
	268	while (ps_tsk_num > num)
	269	destroy_power_saving_task();
	270	}
	271	}
	272
3b8cb427	273	static void acpi_pad_idle_cpus(unsigned int num_cpus)
8e0af514 SL	274	{
	275	get_online_cpus();
	276
	277	num_cpus = min_t(unsigned int, num_cpus, num_online_cpus());
	278	set_power_saving_task_num(num_cpus);
	279
	280	put_online_cpus();
8e0af514 SL	281	}
	282
	283	static uint32_t acpi_pad_idle_cpus_num(void)
	284	{
	285	return ps_tsk_num;
	286	}
	287
	288	static ssize_t acpi_pad_rrtime_store(struct device *dev,
	289	struct device_attribute attr, const char buf, size_t count)
	290	{
	291	unsigned long num;
73d4511a	292	if (kstrtoul(buf, 0, &num))
8e0af514 SL	293	return -EINVAL;
	294	if (num < 1 \|\| num >= 100)
	295	return -EINVAL;
	296	mutex_lock(&isolated_cpus_lock);
	297	round_robin_time = num;
	298	mutex_unlock(&isolated_cpus_lock);
	299	return count;
	300	}
	301
	302	static ssize_t acpi_pad_rrtime_show(struct device *dev,
	303	struct device_attribute attr, char buf)
	304	{
32297abd	305	return scnprintf(buf, PAGE_SIZE, "%d\n", round_robin_time);
8e0af514 SL	306	}
	307	static DEVICE_ATTR(rrtime, S_IRUGO\|S_IWUSR,
	308	acpi_pad_rrtime_show,
	309	acpi_pad_rrtime_store);
	310
	311	static ssize_t acpi_pad_idlepct_store(struct device *dev,
	312	struct device_attribute attr, const char buf, size_t count)
	313	{
	314	unsigned long num;
73d4511a	315	if (kstrtoul(buf, 0, &num))
8e0af514 SL	316	return -EINVAL;
	317	if (num < 1 \|\| num >= 100)
	318	return -EINVAL;
	319	mutex_lock(&isolated_cpus_lock);
	320	idle_pct = num;
	321	mutex_unlock(&isolated_cpus_lock);
	322	return count;
	323	}
	324
	325	static ssize_t acpi_pad_idlepct_show(struct device *dev,
	326	struct device_attribute attr, char buf)
	327	{
32297abd	328	return scnprintf(buf, PAGE_SIZE, "%d\n", idle_pct);
8e0af514 SL	329	}
	330	static DEVICE_ATTR(idlepct, S_IRUGO\|S_IWUSR,
	331	acpi_pad_idlepct_show,
	332	acpi_pad_idlepct_store);
	333
	334	static ssize_t acpi_pad_idlecpus_store(struct device *dev,
	335	struct device_attribute attr, const char buf, size_t count)
	336	{
	337	unsigned long num;
73d4511a	338	if (kstrtoul(buf, 0, &num))
8e0af514 SL	339	return -EINVAL;
	340	mutex_lock(&isolated_cpus_lock);
	341	acpi_pad_idle_cpus(num);
	342	mutex_unlock(&isolated_cpus_lock);
	343	return count;
	344	}
	345
	346	static ssize_t acpi_pad_idlecpus_show(struct device *dev,
	347	struct device_attribute attr, char buf)
	348	{
32297abd NC	349	int n = 0;
	350	n = cpumask_scnprintf(buf, PAGE_SIZE-2, to_cpumask(pad_busy_cpus_bits));
	351	buf[n++] = '\n';
	352	buf[n] = '\0';
	353	return n;
8e0af514 SL	354	}
	355	static DEVICE_ATTR(idlecpus, S_IRUGO\|S_IWUSR,
	356	acpi_pad_idlecpus_show,
	357	acpi_pad_idlecpus_store);
	358
	359	static int acpi_pad_add_sysfs(struct acpi_device *device)
	360	{
	361	int result;
	362
	363	result = device_create_file(&device->dev, &dev_attr_idlecpus);
	364	if (result)
	365	return -ENODEV;
	366	result = device_create_file(&device->dev, &dev_attr_idlepct);
	367	if (result) {
	368	device_remove_file(&device->dev, &dev_attr_idlecpus);
	369	return -ENODEV;
	370	}
	371	result = device_create_file(&device->dev, &dev_attr_rrtime);
	372	if (result) {
	373	device_remove_file(&device->dev, &dev_attr_idlecpus);
	374	device_remove_file(&device->dev, &dev_attr_idlepct);
	375	return -ENODEV;
	376	}
	377	return 0;
	378	}
	379
	380	static void acpi_pad_remove_sysfs(struct acpi_device *device)
	381	{
	382	device_remove_file(&device->dev, &dev_attr_idlecpus);
	383	device_remove_file(&device->dev, &dev_attr_idlepct);
	384	device_remove_file(&device->dev, &dev_attr_rrtime);
	385	}
	386
c9ad8e06 LB	387	/*
	388	* Query firmware how many CPUs should be idle
	389	* return -1 on failure
	390	*/
	391	static int acpi_pad_pur(acpi_handle handle)
8e0af514 SL	392	{
8e0af514 SL	393	struct acpi_buffer buffer = {ACPI_ALLOCATE_BUFFER, NULL};
8e0af514	394	union acpi_object *package;
c9ad8e06	395	int num = -1;
8e0af514	396
3b8cb427	397	if (ACPI_FAILURE(acpi_evaluate_object(handle, "_PUR", NULL, &buffer)))
c9ad8e06	398	return num;
3b8cb427 CG	399
3b8cb427 CG	400	if (!buffer.length \|\| !buffer.pointer)
c9ad8e06	401	return num;
3b8cb427	402
8e0af514	403	package = buffer.pointer;
c9ad8e06 LB	404
	405	if (package->type == ACPI_TYPE_PACKAGE &&
	406	package->package.count == 2 &&
	407	package->package.elements[0].integer.value == 1) /* rev 1 */
	408
	409	num = package->package.elements[1].integer.value;
	410
8e0af514	411	kfree(buffer.pointer);
c9ad8e06	412	return num;
8e0af514 SL	413	}
	414
	415	/* Notify firmware how many CPUs are idle */
	416	static void acpi_pad_ost(acpi_handle handle, int stat,
	417	uint32_t idle_cpus)
	418	{
	419	union acpi_object params[3] = {
	420	{.type = ACPI_TYPE_INTEGER,},
	421	{.type = ACPI_TYPE_INTEGER,},
	422	{.type = ACPI_TYPE_BUFFER,},
	423	};
	424	struct acpi_object_list arg_list = {3, params};
	425
	426	params[0].integer.value = ACPI_PROCESSOR_AGGREGATOR_NOTIFY;
	427	params[1].integer.value = stat;
	428	params[2].buffer.length = 4;
	429	params[2].buffer.pointer = (void *)&idle_cpus;
	430	acpi_evaluate_object(handle, "_OST", &arg_list, NULL);
	431	}
	432
	433	static void acpi_pad_handle_notify(acpi_handle handle)
	434	{
3b8cb427	435	int num_cpus;
8e0af514 SL	436	uint32_t idle_cpus;
	437
	438	mutex_lock(&isolated_cpus_lock);
c9ad8e06 LB	439	num_cpus = acpi_pad_pur(handle);
c9ad8e06 LB	440	if (num_cpus < 0) {
8e0af514 SL	441	mutex_unlock(&isolated_cpus_lock);
	442	return;
	443	}
3b8cb427	444	acpi_pad_idle_cpus(num_cpus);
8e0af514	445	idle_cpus = acpi_pad_idle_cpus_num();
3b8cb427	446	acpi_pad_ost(handle, 0, idle_cpus);
8e0af514 SL	447	mutex_unlock(&isolated_cpus_lock);
	448	}
	449
	450	static void acpi_pad_notify(acpi_handle handle, u32 event,
	451	void *data)
	452	{
	453	struct acpi_device *device = data;
	454
	455	switch (event) {
	456	case ACPI_PROCESSOR_AGGREGATOR_NOTIFY:
	457	acpi_pad_handle_notify(handle);
8e0af514 SL	458	acpi_bus_generate_netlink_event(device->pnp.device_class,
	459	dev_name(&device->dev), event, 0);
	460	break;
	461	default:
73d4511a	462	pr_warn("Unsupported event [0x%x]\n", event);
8e0af514 SL	463	break;
	464	}
	465	}
	466
	467	static int acpi_pad_add(struct acpi_device *device)
	468	{
	469	acpi_status status;
	470
	471	strcpy(acpi_device_name(device), ACPI_PROCESSOR_AGGREGATOR_DEVICE_NAME);
	472	strcpy(acpi_device_class(device), ACPI_PROCESSOR_AGGREGATOR_CLASS);
	473
	474	if (acpi_pad_add_sysfs(device))
	475	return -ENODEV;
	476
	477	status = acpi_install_notify_handler(device->handle,
	478	ACPI_DEVICE_NOTIFY, acpi_pad_notify, device);
	479	if (ACPI_FAILURE(status)) {
	480	acpi_pad_remove_sysfs(device);
	481	return -ENODEV;
	482	}
	483
	484	return 0;
	485	}
	486
51fac838	487	static int acpi_pad_remove(struct acpi_device *device)
8e0af514 SL	488	{
	489	mutex_lock(&isolated_cpus_lock);
	490	acpi_pad_idle_cpus(0);
	491	mutex_unlock(&isolated_cpus_lock);
	492
	493	acpi_remove_notify_handler(device->handle,
	494	ACPI_DEVICE_NOTIFY, acpi_pad_notify);
	495	acpi_pad_remove_sysfs(device);
	496	return 0;
	497	}
	498
	499	static const struct acpi_device_id pad_device_ids[] = {
	500	{"ACPI000C", 0},
	501	{"", 0},
	502	};
	503	MODULE_DEVICE_TABLE(acpi, pad_device_ids);
	504
	505	static struct acpi_driver acpi_pad_driver = {
	506	.name = "processor_aggregator",
	507	.class = ACPI_PROCESSOR_AGGREGATOR_CLASS,
	508	.ids = pad_device_ids,
	509	.ops = {
	510	.add = acpi_pad_add,
	511	.remove = acpi_pad_remove,
	512	},
	513	};
	514
	515	static int __init acpi_pad_init(void)
	516	{
	517	power_saving_mwait_init();
	518	if (power_saving_mwait_eax == 0)
	519	return -EINVAL;
	520
	521	return acpi_bus_register_driver(&acpi_pad_driver);
	522	}
	523
	524	static void __exit acpi_pad_exit(void)
	525	{
	526	acpi_bus_unregister_driver(&acpi_pad_driver);
	527	}
	528
	529	module_init(acpi_pad_init);
	530	module_exit(acpi_pad_exit);
	531	MODULE_AUTHOR("Shaohua Li<shaohua.li@intel.com>");
	532	MODULE_DESCRIPTION("ACPI Processor Aggregator Driver");
	533	MODULE_LICENSE("GPL");