[deliverable/linux.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/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;

		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;
			}
			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	*
	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>
979081e7	29	#include <linux/tick.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;
	44
8e0af514 SL	45	static void power_saving_mwait_init(void)
	46	{
	47	unsigned int eax, ebx, ecx, edx;
	48	unsigned int highest_cstate = 0;
	49	unsigned int highest_subcstate = 0;
	50	int i;
	51
	52	if (!boot_cpu_has(X86_FEATURE_MWAIT))
	53	return;
	54	if (boot_cpu_data.cpuid_level < CPUID_MWAIT_LEAF)
	55	return;
	56
	57	cpuid(CPUID_MWAIT_LEAF, &eax, &ebx, &ecx, &edx);
	58
	59	if (!(ecx & CPUID5_ECX_EXTENSIONS_SUPPORTED) \|\|
	60	!(ecx & CPUID5_ECX_INTERRUPT_BREAK))
	61	return;
	62
	63	edx >>= MWAIT_SUBSTATE_SIZE;
	64	for (i = 0; i < 7 && edx; i++, edx >>= MWAIT_SUBSTATE_SIZE) {
	65	if (edx & MWAIT_SUBSTATE_MASK) {
	66	highest_cstate = i;
	67	highest_subcstate = edx & MWAIT_SUBSTATE_MASK;
	68	}
	69	}
	70	power_saving_mwait_eax = (highest_cstate << MWAIT_SUBSTATE_SIZE) \|
	71	(highest_subcstate - 1);
	72
592913ec	73	#if defined(CONFIG_X86)
8e0af514 SL	74	switch (boot_cpu_data.x86_vendor) {
	75	case X86_VENDOR_AMD:
	76	case X86_VENDOR_INTEL:
	77	/*
	78	* AMD Fam10h TSC will tick in all
	79	* C/P/S0/S1 states when this bit is set.
	80	*/
8aa4b14e CG	81	if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC))
8aa4b14e CG	82	tsc_detected_unstable = 1;
8aa4b14e	83	break;
8e0af514	84	default:
3ff70551	85	/* TSC could halt in idle */
0dc698b9	86	tsc_detected_unstable = 1;
8e0af514 SL	87	}
	88	#endif
	89	}
	90
	91	static unsigned long cpu_weight[NR_CPUS];
	92	static int tsk_in_cpu[NR_CPUS] = {[0 ... NR_CPUS-1] = -1};
	93	static DECLARE_BITMAP(pad_busy_cpus_bits, NR_CPUS);
	94	static void round_robin_cpu(unsigned int tsk_index)
	95	{
	96	struct cpumask *pad_busy_cpus = to_cpumask(pad_busy_cpus_bits);
	97	cpumask_var_t tmp;
	98	int cpu;
f67538f8 AM	99	unsigned long min_weight = -1;
f67538f8 AM	100	unsigned long uninitialized_var(preferred_cpu);
8e0af514 SL	101
	102	if (!alloc_cpumask_var(&tmp, GFP_KERNEL))
	103	return;
	104
5f160126	105	mutex_lock(&round_robin_lock);
8e0af514 SL	106	cpumask_clear(tmp);
8e0af514 SL	107	for_each_cpu(cpu, pad_busy_cpus)
06931e62	108	cpumask_or(tmp, tmp, topology_sibling_cpumask(cpu));
8e0af514 SL	109	cpumask_andnot(tmp, cpu_online_mask, tmp);
	110	/* avoid HT sibilings if possible */
	111	if (cpumask_empty(tmp))
	112	cpumask_andnot(tmp, cpu_online_mask, pad_busy_cpus);
	113	if (cpumask_empty(tmp)) {
5f160126	114	mutex_unlock(&round_robin_lock);
8e0af514 SL	115	return;
	116	}
	117	for_each_cpu(cpu, tmp) {
	118	if (cpu_weight[cpu] < min_weight) {
	119	min_weight = cpu_weight[cpu];
	120	preferred_cpu = cpu;
	121	}
	122	}
	123
	124	if (tsk_in_cpu[tsk_index] != -1)
	125	cpumask_clear_cpu(tsk_in_cpu[tsk_index], pad_busy_cpus);
	126	tsk_in_cpu[tsk_index] = preferred_cpu;
	127	cpumask_set_cpu(preferred_cpu, pad_busy_cpus);
	128	cpu_weight[preferred_cpu]++;
5f160126	129	mutex_unlock(&round_robin_lock);
8e0af514 SL	130
	131	set_cpus_allowed_ptr(current, cpumask_of(preferred_cpu));
	132	}
	133
	134	static void exit_round_robin(unsigned int tsk_index)
	135	{
	136	struct cpumask *pad_busy_cpus = to_cpumask(pad_busy_cpus_bits);
	137	cpumask_clear_cpu(tsk_in_cpu[tsk_index], pad_busy_cpus);
	138	tsk_in_cpu[tsk_index] = -1;
	139	}
	140
	141	static unsigned int idle_pct = 5; /* percentage */
fa7584e1	142	static unsigned int round_robin_time = 1; /* second */
8e0af514 SL	143	static int power_saving_thread(void *data)
	144	{
	145	struct sched_param param = {.sched_priority = 1};
	146	int do_sleep;
	147	unsigned int tsk_index = (unsigned long)data;
	148	u64 last_jiffies = 0;
	149
	150	sched_setscheduler(current, SCHED_RR, &param);
	151
	152	while (!kthread_should_stop()) {
4ff248f3	153	unsigned long expire_time;
8e0af514 SL	154
	155	try_to_freeze();
	156
	157	/* round robin to cpus */
4ff248f3 MS	158	expire_time = last_jiffies + round_robin_time * HZ;
4ff248f3 MS	159	if (time_before(expire_time, jiffies)) {
8e0af514 SL	160	last_jiffies = jiffies;
	161	round_robin_cpu(tsk_index);
	162	}
	163
	164	do_sleep = 0;
	165
8e0af514 SL	166	expire_time = jiffies + HZ * (100 - idle_pct) / 100;
	167
	168	while (!need_resched()) {
0dc698b9 VP	169	if (tsc_detected_unstable && !tsc_marked_unstable) {
	170	/* TSC could halt in idle, so notify users */
	171	mark_tsc_unstable("TSC halts in idle");
	172	tsc_marked_unstable = 1;
	173	}
8e0af514	174	local_irq_disable();
979081e7	175	tick_broadcast_enable();
c7952135	176	tick_broadcast_enter();
8e0af514 SL	177	stop_critical_timings();
8e0af514 SL	178
16824255	179	mwait_idle_with_hints(power_saving_mwait_eax, 1);
8e0af514 SL	180
8e0af514 SL	181	start_critical_timings();
c7952135	182	tick_broadcast_exit();
8e0af514 SL	183	local_irq_enable();
8e0af514 SL	184
4ff248f3	185	if (time_before(expire_time, jiffies)) {
8e0af514 SL	186	do_sleep = 1;
	187	break;
	188	}
	189	}
	190
8e0af514 SL	191	/*
	192	* current sched_rt has threshold for rt task running time.
	193	* When a rt task uses 95% CPU time, the rt thread will be
	194	* scheduled out for 5% CPU time to not starve other tasks. But
	195	* the mechanism only works when all CPUs have RT task running,
	196	* as if one CPU hasn't RT task, RT task from other CPUs will
	197	* borrow CPU time from this CPU and cause RT task use > 95%
3b8cb427	198	* CPU time. To make 'avoid starvation' work, takes a nap here.
8e0af514	199	*/
5b59c69e	200	if (unlikely(do_sleep))
8e0af514	201	schedule_timeout_killable(HZ * idle_pct / 100);
5b59c69e TC	202
	203	/* If an external event has set the need_resched flag, then
	204	* we need to deal with it, or this loop will continue to
	205	* spin without calling __mwait().
	206	*/
	207	if (unlikely(need_resched()))
	208	schedule();
8e0af514 SL	209	}
	210
	211	exit_round_robin(tsk_index);
	212	return 0;
	213	}
	214
	215	static struct task_struct *ps_tsks[NR_CPUS];
	216	static unsigned int ps_tsk_num;
	217	static int create_power_saving_task(void)
	218	{
5d7e4386	219	int rc;
3b8cb427	220
8e0af514 SL	221	ps_tsks[ps_tsk_num] = kthread_run(power_saving_thread,
8e0af514 SL	222	(void *)(unsigned long)ps_tsk_num,
150ed86f	223	"acpi_pad/%d", ps_tsk_num);
5d7e4386 RR	224
	225	if (IS_ERR(ps_tsks[ps_tsk_num])) {
	226	rc = PTR_ERR(ps_tsks[ps_tsk_num]);
3b8cb427	227	ps_tsks[ps_tsk_num] = NULL;
5d7e4386 RR	228	} else {
	229	rc = 0;
	230	ps_tsk_num++;
	231	}
3b8cb427 CG	232
3b8cb427 CG	233	return rc;
8e0af514 SL	234	}
	235
	236	static void destroy_power_saving_task(void)
	237	{
	238	if (ps_tsk_num > 0) {
	239	ps_tsk_num--;
	240	kthread_stop(ps_tsks[ps_tsk_num]);
3b8cb427	241	ps_tsks[ps_tsk_num] = NULL;
8e0af514 SL	242	}
	243	}
	244
	245	static void set_power_saving_task_num(unsigned int num)
	246	{
	247	if (num > ps_tsk_num) {
	248	while (ps_tsk_num < num) {
	249	if (create_power_saving_task())
	250	return;
	251	}
	252	} else if (num < ps_tsk_num) {
	253	while (ps_tsk_num > num)
	254	destroy_power_saving_task();
	255	}
	256	}
	257
3b8cb427	258	static void acpi_pad_idle_cpus(unsigned int num_cpus)
8e0af514 SL	259	{
	260	get_online_cpus();
	261
	262	num_cpus = min_t(unsigned int, num_cpus, num_online_cpus());
	263	set_power_saving_task_num(num_cpus);
	264
	265	put_online_cpus();
8e0af514 SL	266	}
	267
	268	static uint32_t acpi_pad_idle_cpus_num(void)
	269	{
	270	return ps_tsk_num;
	271	}
	272
	273	static ssize_t acpi_pad_rrtime_store(struct device *dev,
	274	struct device_attribute attr, const char buf, size_t count)
	275	{
	276	unsigned long num;
73d4511a	277	if (kstrtoul(buf, 0, &num))
8e0af514 SL	278	return -EINVAL;
	279	if (num < 1 \|\| num >= 100)
	280	return -EINVAL;
	281	mutex_lock(&isolated_cpus_lock);
	282	round_robin_time = num;
	283	mutex_unlock(&isolated_cpus_lock);
	284	return count;
	285	}
	286
	287	static ssize_t acpi_pad_rrtime_show(struct device *dev,
	288	struct device_attribute attr, char buf)
	289	{
32297abd	290	return scnprintf(buf, PAGE_SIZE, "%d\n", round_robin_time);
8e0af514 SL	291	}
	292	static DEVICE_ATTR(rrtime, S_IRUGO\|S_IWUSR,
	293	acpi_pad_rrtime_show,
	294	acpi_pad_rrtime_store);
	295
	296	static ssize_t acpi_pad_idlepct_store(struct device *dev,
	297	struct device_attribute attr, const char buf, size_t count)
	298	{
	299	unsigned long num;
73d4511a	300	if (kstrtoul(buf, 0, &num))
8e0af514 SL	301	return -EINVAL;
	302	if (num < 1 \|\| num >= 100)
	303	return -EINVAL;
	304	mutex_lock(&isolated_cpus_lock);
	305	idle_pct = num;
	306	mutex_unlock(&isolated_cpus_lock);
	307	return count;
	308	}
	309
	310	static ssize_t acpi_pad_idlepct_show(struct device *dev,
	311	struct device_attribute attr, char buf)
	312	{
32297abd	313	return scnprintf(buf, PAGE_SIZE, "%d\n", idle_pct);
8e0af514 SL	314	}
	315	static DEVICE_ATTR(idlepct, S_IRUGO\|S_IWUSR,
	316	acpi_pad_idlepct_show,
	317	acpi_pad_idlepct_store);
	318
	319	static ssize_t acpi_pad_idlecpus_store(struct device *dev,
	320	struct device_attribute attr, const char buf, size_t count)
	321	{
	322	unsigned long num;
73d4511a	323	if (kstrtoul(buf, 0, &num))
8e0af514 SL	324	return -EINVAL;
	325	mutex_lock(&isolated_cpus_lock);
	326	acpi_pad_idle_cpus(num);
	327	mutex_unlock(&isolated_cpus_lock);
	328	return count;
	329	}
	330
	331	static ssize_t acpi_pad_idlecpus_show(struct device *dev,
	332	struct device_attribute attr, char buf)
	333	{
5aaba363 SH	334	return cpumap_print_to_pagebuf(false, buf,
5aaba363 SH	335	to_cpumask(pad_busy_cpus_bits));
8e0af514	336	}
5aaba363	337
8e0af514 SL	338	static DEVICE_ATTR(idlecpus, S_IRUGO\|S_IWUSR,
	339	acpi_pad_idlecpus_show,
	340	acpi_pad_idlecpus_store);
	341
	342	static int acpi_pad_add_sysfs(struct acpi_device *device)
	343	{
	344	int result;
	345
	346	result = device_create_file(&device->dev, &dev_attr_idlecpus);
	347	if (result)
	348	return -ENODEV;
	349	result = device_create_file(&device->dev, &dev_attr_idlepct);
	350	if (result) {
	351	device_remove_file(&device->dev, &dev_attr_idlecpus);
	352	return -ENODEV;
	353	}
	354	result = device_create_file(&device->dev, &dev_attr_rrtime);
	355	if (result) {
	356	device_remove_file(&device->dev, &dev_attr_idlecpus);
	357	device_remove_file(&device->dev, &dev_attr_idlepct);
	358	return -ENODEV;
	359	}
	360	return 0;
	361	}
	362
	363	static void acpi_pad_remove_sysfs(struct acpi_device *device)
	364	{
	365	device_remove_file(&device->dev, &dev_attr_idlecpus);
	366	device_remove_file(&device->dev, &dev_attr_idlepct);
	367	device_remove_file(&device->dev, &dev_attr_rrtime);
	368	}
	369
c9ad8e06 LB	370	/*
	371	* Query firmware how many CPUs should be idle
	372	* return -1 on failure
	373	*/
	374	static int acpi_pad_pur(acpi_handle handle)
8e0af514 SL	375	{
8e0af514 SL	376	struct acpi_buffer buffer = {ACPI_ALLOCATE_BUFFER, NULL};
8e0af514	377	union acpi_object *package;
c9ad8e06	378	int num = -1;
8e0af514	379
3b8cb427	380	if (ACPI_FAILURE(acpi_evaluate_object(handle, "_PUR", NULL, &buffer)))
c9ad8e06	381	return num;
3b8cb427 CG	382
3b8cb427 CG	383	if (!buffer.length \|\| !buffer.pointer)
c9ad8e06	384	return num;
3b8cb427	385
8e0af514	386	package = buffer.pointer;
c9ad8e06 LB	387
	388	if (package->type == ACPI_TYPE_PACKAGE &&
	389	package->package.count == 2 &&
	390	package->package.elements[0].integer.value == 1) /* rev 1 */
	391
	392	num = package->package.elements[1].integer.value;
	393
8e0af514	394	kfree(buffer.pointer);
c9ad8e06	395	return num;
8e0af514 SL	396	}
8e0af514 SL	397
8e0af514 SL	398	static void acpi_pad_handle_notify(acpi_handle handle)
8e0af514 SL	399	{
3b8cb427	400	int num_cpus;
8e0af514	401	uint32_t idle_cpus;
8b296d94 JL	402	struct acpi_buffer param = {
	403	.length = 4,
	404	.pointer = (void *)&idle_cpus,
	405	};
8e0af514 SL	406
8e0af514 SL	407	mutex_lock(&isolated_cpus_lock);
c9ad8e06 LB	408	num_cpus = acpi_pad_pur(handle);
c9ad8e06 LB	409	if (num_cpus < 0) {
8e0af514 SL	410	mutex_unlock(&isolated_cpus_lock);
	411	return;
	412	}
3b8cb427	413	acpi_pad_idle_cpus(num_cpus);
8e0af514	414	idle_cpus = acpi_pad_idle_cpus_num();
8b296d94	415	acpi_evaluate_ost(handle, ACPI_PROCESSOR_AGGREGATOR_NOTIFY, 0, &param);
8e0af514 SL	416	mutex_unlock(&isolated_cpus_lock);
	417	}
	418
	419	static void acpi_pad_notify(acpi_handle handle, u32 event,
	420	void *data)
	421	{
	422	struct acpi_device *device = data;
	423
	424	switch (event) {
	425	case ACPI_PROCESSOR_AGGREGATOR_NOTIFY:
	426	acpi_pad_handle_notify(handle);
8e0af514 SL	427	acpi_bus_generate_netlink_event(device->pnp.device_class,
	428	dev_name(&device->dev), event, 0);
	429	break;
	430	default:
73d4511a	431	pr_warn("Unsupported event [0x%x]\n", event);
8e0af514 SL	432	break;
	433	}
	434	}
	435
	436	static int acpi_pad_add(struct acpi_device *device)
	437	{
	438	acpi_status status;
	439
	440	strcpy(acpi_device_name(device), ACPI_PROCESSOR_AGGREGATOR_DEVICE_NAME);
	441	strcpy(acpi_device_class(device), ACPI_PROCESSOR_AGGREGATOR_CLASS);
	442
	443	if (acpi_pad_add_sysfs(device))
	444	return -ENODEV;
	445
	446	status = acpi_install_notify_handler(device->handle,
	447	ACPI_DEVICE_NOTIFY, acpi_pad_notify, device);
	448	if (ACPI_FAILURE(status)) {
	449	acpi_pad_remove_sysfs(device);
	450	return -ENODEV;
	451	}
	452
	453	return 0;
	454	}
	455
51fac838	456	static int acpi_pad_remove(struct acpi_device *device)
8e0af514 SL	457	{
	458	mutex_lock(&isolated_cpus_lock);
	459	acpi_pad_idle_cpus(0);
	460	mutex_unlock(&isolated_cpus_lock);
	461
	462	acpi_remove_notify_handler(device->handle,
	463	ACPI_DEVICE_NOTIFY, acpi_pad_notify);
	464	acpi_pad_remove_sysfs(device);
	465	return 0;
	466	}
	467
	468	static const struct acpi_device_id pad_device_ids[] = {
	469	{"ACPI000C", 0},
	470	{"", 0},
	471	};
	472	MODULE_DEVICE_TABLE(acpi, pad_device_ids);
	473
	474	static struct acpi_driver acpi_pad_driver = {
	475	.name = "processor_aggregator",
	476	.class = ACPI_PROCESSOR_AGGREGATOR_CLASS,
	477	.ids = pad_device_ids,
	478	.ops = {
	479	.add = acpi_pad_add,
	480	.remove = acpi_pad_remove,
	481	},
	482	};
	483
	484	static int __init acpi_pad_init(void)
	485	{
	486	power_saving_mwait_init();
	487	if (power_saving_mwait_eax == 0)
	488	return -EINVAL;
	489
	490	return acpi_bus_register_driver(&acpi_pad_driver);
	491	}
	492
	493	static void __exit acpi_pad_exit(void)
	494	{
	495	acpi_bus_unregister_driver(&acpi_pad_driver);
	496	}
	497
	498	module_init(acpi_pad_init);
	499	module_exit(acpi_pad_exit);
	500	MODULE_AUTHOR("Shaohua Li<shaohua.li@intel.com>");
	501	MODULE_DESCRIPTION("ACPI Processor Aggregator Driver");
	502	MODULE_LICENSE("GPL");