[mirror_ubuntu-bionic-kernel.git] / drivers / idle / intel_idle.c

/*
 * intel_idle.c - native hardware idle loop for modern Intel processors
 *
 * Copyright (c) 2010, Intel Corporation.
 * Len Brown <len.brown@intel.com>
 *
 * 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.
 */

/*
 * intel_idle is a cpuidle driver that loads on specific Intel processors
 * in lieu of the legacy ACPI processor_idle driver.  The intent is to
 * make Linux more efficient on these processors, as intel_idle knows
 * more than ACPI, as well as make Linux more immune to ACPI BIOS bugs.
 */

/*
 * Design Assumptions
 *
 * All CPUs have same idle states as boot CPU
 *
 * Chipset BM_STS (bus master status) bit is a NOP
 *	for preventing entry into deep C-stats
 */

/*
 * Known limitations
 *
 * The driver currently initializes for_each_online_cpu() upon modprobe.
 * It it unaware of subsequent processors hot-added to the system.
 * This means that if you boot with maxcpus=n and later online
 * processors above n, those processors will use C1 only.
 *
 * ACPI has a .suspend hack to turn off deep c-statees during suspend
 * to avoid complications with the lapic timer workaround.
 * Have not seen issues with suspend, but may need same workaround here.
 *
 * There is currently no kernel-based automatic probing/loading mechanism
 * if the driver is built as a module.
 */

/* un-comment DEBUG to enable pr_debug() statements */
#define DEBUG

#include <linux/kernel.h>
#include <linux/cpuidle.h>
#include <linux/clockchips.h>
#include <linux/hrtimer.h>	/* ktime_get_real() */
#include <trace/events/power.h>
#include <linux/sched.h>
#include <linux/notifier.h>
#include <linux/cpu.h>
#include <linux/module.h>
#include <asm/mwait.h>
#include <asm/msr.h>

#define INTEL_IDLE_VERSION "0.4"
#define PREFIX "intel_idle: "

static struct cpuidle_driver intel_idle_driver = {
	.name = "intel_idle",
	.owner = THIS_MODULE,
};
/* intel_idle.max_cstate=0 disables driver */
static int max_cstate = MWAIT_MAX_NUM_CSTATES - 1;

static unsigned int mwait_substates;

#define LAPIC_TIMER_ALWAYS_RELIABLE 0xFFFFFFFF
/* Reliable LAPIC Timer States, bit 1 for C1 etc.  */
static unsigned int lapic_timer_reliable_states = (1 << 1);	 /* Default to only C1 */

static struct cpuidle_device __percpu *intel_idle_cpuidle_devices;
static int intel_idle(struct cpuidle_device *dev,
			struct cpuidle_driver *drv, int index);

static struct cpuidle_state *cpuidle_state_table;

/*
 * Hardware C-state auto-demotion may not always be optimal.
 * Indicate which enable bits to clear here.
 */
static unsigned long long auto_demotion_disable_flags;

/*
 * Set this flag for states where the HW flushes the TLB for us
 * and so we don't need cross-calls to keep it consistent.
 * If this flag is set, SW flushes the TLB, so even if the
 * HW doesn't do the flushing, this flag is safe to use.
 */
#define CPUIDLE_FLAG_TLB_FLUSHED	0x10000

/*
 * States are indexed by the cstate number,
 * which is also the index into the MWAIT hint array.
 * Thus C0 is a dummy.
 */
static struct cpuidle_state nehalem_cstates[MWAIT_MAX_NUM_CSTATES] = {
	{ /* MWAIT C0 */ },
	{ /* MWAIT C1 */
		.name = "C1-NHM",
		.desc = "MWAIT 0x00",
		.flags = CPUIDLE_FLAG_TIME_VALID,
		.exit_latency = 3,
		.target_residency = 6,
		.enter = &intel_idle },
	{ /* MWAIT C2 */
		.name = "C3-NHM",
		.desc = "MWAIT 0x10",
		.flags = CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
		.exit_latency = 20,
		.target_residency = 80,
		.enter = &intel_idle },
	{ /* MWAIT C3 */
		.name = "C6-NHM",
		.desc = "MWAIT 0x20",
		.flags = CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
		.exit_latency = 200,
		.target_residency = 800,
		.enter = &intel_idle },
};

static struct cpuidle_state snb_cstates[MWAIT_MAX_NUM_CSTATES] = {
	{ /* MWAIT C0 */ },
	{ /* MWAIT C1 */
		.name = "C1-SNB",
		.desc = "MWAIT 0x00",
		.flags = CPUIDLE_FLAG_TIME_VALID,
		.exit_latency = 1,
		.target_residency = 1,
		.enter = &intel_idle },
	{ /* MWAIT C2 */
		.name = "C3-SNB",
		.desc = "MWAIT 0x10",
		.flags = CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
		.exit_latency = 80,
		.target_residency = 211,
		.enter = &intel_idle },
	{ /* MWAIT C3 */
		.name = "C6-SNB",
		.desc = "MWAIT 0x20",
		.flags = CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
		.exit_latency = 104,
		.target_residency = 345,
		.enter = &intel_idle },
	{ /* MWAIT C4 */
		.name = "C7-SNB",
		.desc = "MWAIT 0x30",
		.flags = CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
		.exit_latency = 109,
		.target_residency = 345,
		.enter = &intel_idle },
};

static struct cpuidle_state atom_cstates[MWAIT_MAX_NUM_CSTATES] = {
	{ /* MWAIT C0 */ },
	{ /* MWAIT C1 */
		.name = "C1-ATM",
		.desc = "MWAIT 0x00",
		.flags = CPUIDLE_FLAG_TIME_VALID,
		.exit_latency = 1,
		.target_residency = 4,
		.enter = &intel_idle },
	{ /* MWAIT C2 */
		.name = "C2-ATM",
		.desc = "MWAIT 0x10",
		.flags = CPUIDLE_FLAG_TIME_VALID,
		.exit_latency = 20,
		.target_residency = 80,
		.enter = &intel_idle },
	{ /* MWAIT C3 */ },
	{ /* MWAIT C4 */
		.name = "C4-ATM",
		.desc = "MWAIT 0x30",
		.flags = CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
		.exit_latency = 100,
		.target_residency = 400,
		.enter = &intel_idle },
	{ /* MWAIT C5 */ },
	{ /* MWAIT C6 */
		.name = "C6-ATM",
		.desc = "MWAIT 0x52",
		.flags = CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
		.exit_latency = 140,
		.target_residency = 560,
		.enter = &intel_idle },
};

static long get_driver_data(int cstate)
{
	int driver_data;
	switch (cstate) {

	case 1:	/* MWAIT C1 */
		driver_data = 0x00;
		break;
	case 2:	/* MWAIT C2 */
		driver_data = 0x10;
		break;
	case 3:	/* MWAIT C3 */
		driver_data = 0x20;
		break;
	case 4:	/* MWAIT C4 */
		driver_data = 0x30;
		break;
	case 5:	/* MWAIT C5 */
		driver_data = 0x40;
		break;
	case 6:	/* MWAIT C6 */
		driver_data = 0x52;
		break;
	default:
		driver_data = 0x00;
	}
	return driver_data;
}

/**
 * intel_idle
 * @dev: cpuidle_device
 * @drv: cpuidle driver
 * @index: index of cpuidle state
 *
 * Must be called under local_irq_disable().
 */
static int intel_idle(struct cpuidle_device *dev,
		struct cpuidle_driver *drv, int index)
{
	unsigned long ecx = 1; /* break on interrupt flag */
	struct cpuidle_state *state = &drv->states[index];
	struct cpuidle_state_usage *state_usage = &dev->states_usage[index];
	unsigned long eax = (unsigned long)cpuidle_get_statedata(state_usage);
	unsigned int cstate;
	ktime_t kt_before, kt_after;
	s64 usec_delta;
	int cpu = smp_processor_id();

	cstate = (((eax) >> MWAIT_SUBSTATE_SIZE) & MWAIT_CSTATE_MASK) + 1;

	/*
	 * leave_mm() to avoid costly and often unnecessary wakeups
	 * for flushing the user TLB's associated with the active mm.
	 */
	if (state->flags & CPUIDLE_FLAG_TLB_FLUSHED)
		leave_mm(cpu);

	if (!(lapic_timer_reliable_states & (1 << (cstate))))
		clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_ENTER, &cpu);

	kt_before = ktime_get_real();

	stop_critical_timings();
	if (!need_resched()) {

		__monitor((void *)&current_thread_info()->flags, 0, 0);
		smp_mb();
		if (!need_resched())
			__mwait(eax, ecx);
	}

	start_critical_timings();

	kt_after = ktime_get_real();
	usec_delta = ktime_to_us(ktime_sub(kt_after, kt_before));

	local_irq_enable();

	if (!(lapic_timer_reliable_states & (1 << (cstate))))
		clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_EXIT, &cpu);

	/* Update cpuidle counters */
	dev->last_residency = (int)usec_delta;

	return index;
}

static void __setup_broadcast_timer(void *arg)
{
	unsigned long reason = (unsigned long)arg;
	int cpu = smp_processor_id();

	reason = reason ?
		CLOCK_EVT_NOTIFY_BROADCAST_ON : CLOCK_EVT_NOTIFY_BROADCAST_OFF;

	clockevents_notify(reason, &cpu);
}

static int setup_broadcast_cpuhp_notify(struct notifier_block *n,
		unsigned long action, void *hcpu)
{
	int hotcpu = (unsigned long)hcpu;

	switch (action & 0xf) {
	case CPU_ONLINE:
		smp_call_function_single(hotcpu, __setup_broadcast_timer,
			(void *)true, 1);
		break;
	}
	return NOTIFY_OK;
}

static struct notifier_block setup_broadcast_notifier = {
	.notifier_call = setup_broadcast_cpuhp_notify,
};

static void auto_demotion_disable(void *dummy)
{
	unsigned long long msr_bits;

	rdmsrl(MSR_NHM_SNB_PKG_CST_CFG_CTL, msr_bits);
	msr_bits &= ~auto_demotion_disable_flags;
	wrmsrl(MSR_NHM_SNB_PKG_CST_CFG_CTL, msr_bits);
}

/*
 * intel_idle_probe()
 */
static int intel_idle_probe(void)
{
	unsigned int eax, ebx, ecx;

	if (max_cstate == 0) {
		pr_debug(PREFIX "disabled\n");
		return -EPERM;
	}

	if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL)
		return -ENODEV;

	if (!boot_cpu_has(X86_FEATURE_MWAIT))
		return -ENODEV;

	if (boot_cpu_data.cpuid_level < CPUID_MWAIT_LEAF)
		return -ENODEV;

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

	if (!(ecx & CPUID5_ECX_EXTENSIONS_SUPPORTED) ||
	    !(ecx & CPUID5_ECX_INTERRUPT_BREAK) ||
	    !mwait_substates)
			return -ENODEV;

	pr_debug(PREFIX "MWAIT substates: 0x%x\n", mwait_substates);


	if (boot_cpu_data.x86 != 6)	/* family 6 */
		return -ENODEV;

	switch (boot_cpu_data.x86_model) {

	case 0x1A:	/* Core i7, Xeon 5500 series */
	case 0x1E:	/* Core i7 and i5 Processor - Lynnfield Jasper Forest */
	case 0x1F:	/* Core i7 and i5 Processor - Nehalem */
	case 0x2E:	/* Nehalem-EX Xeon */
	case 0x2F:	/* Westmere-EX Xeon */
	case 0x25:	/* Westmere */
	case 0x2C:	/* Westmere */
		cpuidle_state_table = nehalem_cstates;
		auto_demotion_disable_flags =
			(NHM_C1_AUTO_DEMOTE | NHM_C3_AUTO_DEMOTE);
		break;

	case 0x1C:	/* 28 - Atom Processor */
		cpuidle_state_table = atom_cstates;
		break;

	case 0x26:	/* 38 - Lincroft Atom Processor */
		cpuidle_state_table = atom_cstates;
		auto_demotion_disable_flags = ATM_LNC_C6_AUTO_DEMOTE;
		break;

	case 0x2A:	/* SNB */
	case 0x2D:	/* SNB Xeon */
		cpuidle_state_table = snb_cstates;
		break;

	default:
		pr_debug(PREFIX "does not run on family %d model %d\n",
			boot_cpu_data.x86, boot_cpu_data.x86_model);
		return -ENODEV;
	}

	if (boot_cpu_has(X86_FEATURE_ARAT))	/* Always Reliable APIC Timer */
		lapic_timer_reliable_states = LAPIC_TIMER_ALWAYS_RELIABLE;
	else {
		on_each_cpu(__setup_broadcast_timer, (void *)true, 1);
		register_cpu_notifier(&setup_broadcast_notifier);
	}

	pr_debug(PREFIX "v" INTEL_IDLE_VERSION
		" model 0x%X\n", boot_cpu_data.x86_model);

	pr_debug(PREFIX "lapic_timer_reliable_states 0x%x\n",
		lapic_timer_reliable_states);
	return 0;
}

/*
 * intel_idle_cpuidle_devices_uninit()
 * unregister, free cpuidle_devices
 */
static void intel_idle_cpuidle_devices_uninit(void)
{
	int i;
	struct cpuidle_device *dev;

	for_each_online_cpu(i) {
		dev = per_cpu_ptr(intel_idle_cpuidle_devices, i);
		cpuidle_unregister_device(dev);
	}

	free_percpu(intel_idle_cpuidle_devices);
	return;
}
/*
 * intel_idle_cpuidle_driver_init()
 * allocate, initialize cpuidle_states
 */
static int intel_idle_cpuidle_driver_init(void)
{
	int cstate;
	struct cpuidle_driver *drv = &intel_idle_driver;

	drv->state_count = 1;

	for (cstate = 1; cstate < MWAIT_MAX_NUM_CSTATES; ++cstate) {
		int num_substates;

		if (cstate > max_cstate) {
			printk(PREFIX "max_cstate %d reached\n",
				max_cstate);
			break;
		}

		/* does the state exist in CPUID.MWAIT? */
		num_substates = (mwait_substates >> ((cstate) * 4))
					& MWAIT_SUBSTATE_MASK;
		if (num_substates == 0)
			continue;
		/* is the state not enabled? */
		if (cpuidle_state_table[cstate].enter == NULL) {
			/* does the driver not know about the state? */
			if (*cpuidle_state_table[cstate].name == '\0')
				pr_debug(PREFIX "unaware of model 0x%x"
					" MWAIT %d please"
					" contact lenb@kernel.org",
				boot_cpu_data.x86_model, cstate);
			continue;
		}

		if ((cstate > 2) &&
			!boot_cpu_has(X86_FEATURE_NONSTOP_TSC))
			mark_tsc_unstable("TSC halts in idle"
					" states deeper than C2");

		drv->states[drv->state_count] =	/* structure copy */
			cpuidle_state_table[cstate];

		drv->state_count += 1;
	}

	if (auto_demotion_disable_flags)
		on_each_cpu(auto_demotion_disable, NULL, 1);

	return 0;
}


/*
 * intel_idle_cpu_init()
 * allocate, initialize, register cpuidle_devices
 * @cpu: cpu/core to initialize
 */
int intel_idle_cpu_init(int cpu)
{
	int cstate;
	struct cpuidle_device *dev;

	dev = per_cpu_ptr(intel_idle_cpuidle_devices, cpu);

	dev->state_count = 1;

	for (cstate = 1; cstate < MWAIT_MAX_NUM_CSTATES; ++cstate) {
		int num_substates;

		if (cstate > max_cstate) {
			printk(PREFIX "max_cstate %d reached\n",
			       max_cstate);
			break;
		}

		/* does the state exist in CPUID.MWAIT? */
		num_substates = (mwait_substates >> ((cstate) * 4))
			& MWAIT_SUBSTATE_MASK;
		if (num_substates == 0)
			continue;
		/* is the state not enabled? */
		if (cpuidle_state_table[cstate].enter == NULL)
			continue;

		dev->states_usage[dev->state_count].driver_data =
			(void *)get_driver_data(cstate);

			dev->state_count += 1;
		}
	dev->cpu = cpu;

	if (cpuidle_register_device(dev)) {
		pr_debug(PREFIX "cpuidle_register_device %d failed!\n", cpu);
		intel_idle_cpuidle_devices_uninit();
		return -EIO;
	}

	if (auto_demotion_disable_flags)
		smp_call_function_single(cpu, auto_demotion_disable, NULL, 1);

	return 0;
}


static int __init intel_idle_init(void)
{
	int retval, i;

	/* Do not load intel_idle at all for now if idle= is passed */
	if (boot_option_idle_override != IDLE_NO_OVERRIDE)
		return -ENODEV;

	retval = intel_idle_probe();
	if (retval)
		return retval;

	intel_idle_cpuidle_driver_init();
	retval = cpuidle_register_driver(&intel_idle_driver);
	if (retval) {
		printk(KERN_DEBUG PREFIX "intel_idle yielding to %s",
			cpuidle_get_driver()->name);
		return retval;
	}

	intel_idle_cpuidle_devices = alloc_percpu(struct cpuidle_device);
	if (intel_idle_cpuidle_devices == NULL)
		return -ENOMEM;

	for_each_online_cpu(i) {
		retval = intel_idle_cpu_init(i);
		if (retval) {
			cpuidle_unregister_driver(&intel_idle_driver);
			return retval;
		}
	}

	return 0;
}

static void __exit intel_idle_exit(void)
{
	intel_idle_cpuidle_devices_uninit();
	cpuidle_unregister_driver(&intel_idle_driver);

	if (lapic_timer_reliable_states != LAPIC_TIMER_ALWAYS_RELIABLE) {
		on_each_cpu(__setup_broadcast_timer, (void *)false, 1);
		unregister_cpu_notifier(&setup_broadcast_notifier);
	}

	return;
}

module_init(intel_idle_init);
module_exit(intel_idle_exit);

module_param(max_cstate, int, 0444);

MODULE_AUTHOR("Len Brown <len.brown@intel.com>");
MODULE_DESCRIPTION("Cpuidle driver for Intel Hardware v" INTEL_IDLE_VERSION);
MODULE_LICENSE("GPL");
Commit	Line	Data
26717172 LB	1	/*
	2	* intel_idle.c - native hardware idle loop for modern Intel processors
	3	*
	4	* Copyright (c) 2010, Intel Corporation.
	5	* Len Brown <len.brown@intel.com>
	6	*
	7	* This program is free software; you can redistribute it and/or modify it
	8	* under the terms and conditions of the GNU General Public License,
	9	* version 2, as published by the Free Software Foundation.
	10	*
	11	* This program is distributed in the hope it will be useful, but WITHOUT
	12	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	13	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
	14	* more details.
	15	*
	16	* You should have received a copy of the GNU General Public License along with
	17	* this program; if not, write to the Free Software Foundation, Inc.,
	18	* 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
	19	*/
	20
	21	/*
	22	* intel_idle is a cpuidle driver that loads on specific Intel processors
	23	* in lieu of the legacy ACPI processor_idle driver. The intent is to
	24	* make Linux more efficient on these processors, as intel_idle knows
	25	* more than ACPI, as well as make Linux more immune to ACPI BIOS bugs.
	26	*/
	27
	28	/*
	29	* Design Assumptions
	30	*
	31	* All CPUs have same idle states as boot CPU
	32	*
	33	* Chipset BM_STS (bus master status) bit is a NOP
	34	* for preventing entry into deep C-stats
	35	*/
	36
	37	/*
	38	* Known limitations
	39	*
	40	* The driver currently initializes for_each_online_cpu() upon modprobe.
	41	* It it unaware of subsequent processors hot-added to the system.
	42	* This means that if you boot with maxcpus=n and later online
	43	* processors above n, those processors will use C1 only.
	44	*
	45	* ACPI has a .suspend hack to turn off deep c-statees during suspend
	46	* to avoid complications with the lapic timer workaround.
	47	* Have not seen issues with suspend, but may need same workaround here.
	48	*
	49	* There is currently no kernel-based automatic probing/loading mechanism
	50	* if the driver is built as a module.
	51	*/
	52
	53	/* un-comment DEBUG to enable pr_debug() statements */
	54	#define DEBUG
	55
	56	#include <linux/kernel.h>
	57	#include <linux/cpuidle.h>
	58	#include <linux/clockchips.h>
	59	#include <linux/hrtimer.h> /* ktime_get_real() */
	60	#include <trace/events/power.h>
	61	#include <linux/sched.h>
2a2d31c8 SL	62	#include <linux/notifier.h>
2a2d31c8 SL	63	#include <linux/cpu.h>
7c52d551	64	#include <linux/module.h>
bc83cccc	65	#include <asm/mwait.h>
14796fca	66	#include <asm/msr.h>
26717172 LB	67
	68	#define INTEL_IDLE_VERSION "0.4"
	69	#define PREFIX "intel_idle: "
	70
26717172 LB	71	static struct cpuidle_driver intel_idle_driver = {
	72	.name = "intel_idle",
	73	.owner = THIS_MODULE,
	74	};
	75	/* intel_idle.max_cstate=0 disables driver */
	76	static int max_cstate = MWAIT_MAX_NUM_CSTATES - 1;
26717172	77
c4236282	78	static unsigned int mwait_substates;
26717172	79
2a2d31c8	80	#define LAPIC_TIMER_ALWAYS_RELIABLE 0xFFFFFFFF
26717172	81	/* Reliable LAPIC Timer States, bit 1 for C1 etc. */
d13780d4	82	static unsigned int lapic_timer_reliable_states = (1 << 1); /* Default to only C1 */
26717172	83
3265eba0	84	static struct cpuidle_device __percpu *intel_idle_cpuidle_devices;
46bcfad7 DD	85	static int intel_idle(struct cpuidle_device *dev,
46bcfad7 DD	86	struct cpuidle_driver *drv, int index);
26717172 LB	87
	88	static struct cpuidle_state *cpuidle_state_table;
	89
14796fca LB	90	/*
	91	* Hardware C-state auto-demotion may not always be optimal.
	92	* Indicate which enable bits to clear here.
	93	*/
	94	static unsigned long long auto_demotion_disable_flags;
	95
956d033f LB	96	/*
	97	* Set this flag for states where the HW flushes the TLB for us
	98	* and so we don't need cross-calls to keep it consistent.
	99	* If this flag is set, SW flushes the TLB, so even if the
	100	* HW doesn't do the flushing, this flag is safe to use.
	101	*/
	102	#define CPUIDLE_FLAG_TLB_FLUSHED 0x10000
	103
26717172 LB	104	/*
	105	* States are indexed by the cstate number,
	106	* which is also the index into the MWAIT hint array.
	107	* Thus C0 is a dummy.
	108	*/
	109	static struct cpuidle_state nehalem_cstates[MWAIT_MAX_NUM_CSTATES] = {
	110	{ /* MWAIT C0 */ },
	111	{ /* MWAIT C1 */
15e123e5	112	.name = "C1-NHM",
26717172	113	.desc = "MWAIT 0x00",
26717172 LB	114	.flags = CPUIDLE_FLAG_TIME_VALID,
26717172 LB	115	.exit_latency = 3,
26717172 LB	116	.target_residency = 6,
	117	.enter = &intel_idle },
	118	{ /* MWAIT C2 */
15e123e5	119	.name = "C3-NHM",
26717172	120	.desc = "MWAIT 0x10",
6110a1f4	121	.flags = CPUIDLE_FLAG_TIME_VALID \| CPUIDLE_FLAG_TLB_FLUSHED,
26717172	122	.exit_latency = 20,
26717172 LB	123	.target_residency = 80,
	124	.enter = &intel_idle },
	125	{ /* MWAIT C3 */
15e123e5	126	.name = "C6-NHM",
26717172	127	.desc = "MWAIT 0x20",
6110a1f4	128	.flags = CPUIDLE_FLAG_TIME_VALID \| CPUIDLE_FLAG_TLB_FLUSHED,
26717172	129	.exit_latency = 200,
26717172 LB	130	.target_residency = 800,
	131	.enter = &intel_idle },
	132	};
	133
d13780d4 LB	134	static struct cpuidle_state snb_cstates[MWAIT_MAX_NUM_CSTATES] = {
	135	{ /* MWAIT C0 */ },
	136	{ /* MWAIT C1 */
15e123e5	137	.name = "C1-SNB",
d13780d4	138	.desc = "MWAIT 0x00",
d13780d4 LB	139	.flags = CPUIDLE_FLAG_TIME_VALID,
d13780d4 LB	140	.exit_latency = 1,
ddbd550d	141	.target_residency = 1,
d13780d4 LB	142	.enter = &intel_idle },
d13780d4 LB	143	{ /* MWAIT C2 */
15e123e5	144	.name = "C3-SNB",
d13780d4	145	.desc = "MWAIT 0x10",
00527cc6	146	.flags = CPUIDLE_FLAG_TIME_VALID \| CPUIDLE_FLAG_TLB_FLUSHED,
d13780d4	147	.exit_latency = 80,
ddbd550d	148	.target_residency = 211,
d13780d4 LB	149	.enter = &intel_idle },
d13780d4 LB	150	{ /* MWAIT C3 */
15e123e5	151	.name = "C6-SNB",
d13780d4	152	.desc = "MWAIT 0x20",
00527cc6	153	.flags = CPUIDLE_FLAG_TIME_VALID \| CPUIDLE_FLAG_TLB_FLUSHED,
d13780d4	154	.exit_latency = 104,
ddbd550d	155	.target_residency = 345,
d13780d4 LB	156	.enter = &intel_idle },
d13780d4 LB	157	{ /* MWAIT C4 */
15e123e5	158	.name = "C7-SNB",
d13780d4	159	.desc = "MWAIT 0x30",
00527cc6	160	.flags = CPUIDLE_FLAG_TIME_VALID \| CPUIDLE_FLAG_TLB_FLUSHED,
d13780d4	161	.exit_latency = 109,
ddbd550d	162	.target_residency = 345,
d13780d4 LB	163	.enter = &intel_idle },
	164	};
	165
26717172 LB	166	static struct cpuidle_state atom_cstates[MWAIT_MAX_NUM_CSTATES] = {
	167	{ /* MWAIT C0 */ },
	168	{ /* MWAIT C1 */
15e123e5	169	.name = "C1-ATM",
26717172	170	.desc = "MWAIT 0x00",
26717172 LB	171	.flags = CPUIDLE_FLAG_TIME_VALID,
26717172 LB	172	.exit_latency = 1,
26717172 LB	173	.target_residency = 4,
	174	.enter = &intel_idle },
	175	{ /* MWAIT C2 */
15e123e5	176	.name = "C2-ATM",
26717172	177	.desc = "MWAIT 0x10",
26717172 LB	178	.flags = CPUIDLE_FLAG_TIME_VALID,
26717172 LB	179	.exit_latency = 20,
26717172 LB	180	.target_residency = 80,
	181	.enter = &intel_idle },
	182	{ /* MWAIT C3 */ },
	183	{ /* MWAIT C4 */
15e123e5	184	.name = "C4-ATM",
26717172	185	.desc = "MWAIT 0x30",
6110a1f4	186	.flags = CPUIDLE_FLAG_TIME_VALID \| CPUIDLE_FLAG_TLB_FLUSHED,
26717172	187	.exit_latency = 100,
26717172 LB	188	.target_residency = 400,
	189	.enter = &intel_idle },
	190	{ /* MWAIT C5 */ },
	191	{ /* MWAIT C6 */
15e123e5	192	.name = "C6-ATM",
7fcca7d9	193	.desc = "MWAIT 0x52",
6110a1f4	194	.flags = CPUIDLE_FLAG_TIME_VALID \| CPUIDLE_FLAG_TLB_FLUSHED,
7fcca7d9	195	.exit_latency = 140,
7fcca7d9 LB	196	.target_residency = 560,
7fcca7d9 LB	197	.enter = &intel_idle },
26717172 LB	198	};
26717172 LB	199
95e3ec11	200	static long get_driver_data(int cstate)
4202735e DD	201	{
	202	int driver_data;
	203	switch (cstate) {
	204
	205	case 1: /* MWAIT C1 */
	206	driver_data = 0x00;
	207	break;
	208	case 2: /* MWAIT C2 */
	209	driver_data = 0x10;
	210	break;
	211	case 3: /* MWAIT C3 */
	212	driver_data = 0x20;
	213	break;
	214	case 4: /* MWAIT C4 */
	215	driver_data = 0x30;
	216	break;
	217	case 5: /* MWAIT C5 */
	218	driver_data = 0x40;
	219	break;
	220	case 6: /* MWAIT C6 */
	221	driver_data = 0x52;
	222	break;
	223	default:
	224	driver_data = 0x00;
	225	}
	226	return driver_data;
	227	}
	228
26717172 LB	229	/**
	230	* intel_idle
	231	* @dev: cpuidle_device
46bcfad7	232	* @drv: cpuidle driver
e978aa7d	233	* @index: index of cpuidle state
26717172	234	*
63ff07be	235	* Must be called under local_irq_disable().
26717172	236	*/
46bcfad7 DD	237	static int intel_idle(struct cpuidle_device *dev,
46bcfad7 DD	238	struct cpuidle_driver *drv, int index)
26717172 LB	239	{
26717172 LB	240	unsigned long ecx = 1; /* break on interrupt flag */
46bcfad7	241	struct cpuidle_state *state = &drv->states[index];
4202735e DD	242	struct cpuidle_state_usage *state_usage = &dev->states_usage[index];
4202735e DD	243	unsigned long eax = (unsigned long)cpuidle_get_statedata(state_usage);
26717172 LB	244	unsigned int cstate;
	245	ktime_t kt_before, kt_after;
	246	s64 usec_delta;
	247	int cpu = smp_processor_id();
	248
	249	cstate = (((eax) >> MWAIT_SUBSTATE_SIZE) & MWAIT_CSTATE_MASK) + 1;
	250
6110a1f4	251	/*
c8381cc3 LB	252	* leave_mm() to avoid costly and often unnecessary wakeups
c8381cc3 LB	253	* for flushing the user TLB's associated with the active mm.
6110a1f4	254	*/
c8381cc3	255	if (state->flags & CPUIDLE_FLAG_TLB_FLUSHED)
6110a1f4 SS	256	leave_mm(cpu);
6110a1f4 SS	257
26717172 LB	258	if (!(lapic_timer_reliable_states & (1 << (cstate))))
	259	clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_ENTER, &cpu);
	260
	261	kt_before = ktime_get_real();
	262
	263	stop_critical_timings();
26717172 LB	264	if (!need_resched()) {
	265
	266	__monitor((void *)&current_thread_info()->flags, 0, 0);
	267	smp_mb();
	268	if (!need_resched())
	269	__mwait(eax, ecx);
	270	}
	271
	272	start_critical_timings();
	273
	274	kt_after = ktime_get_real();
	275	usec_delta = ktime_to_us(ktime_sub(kt_after, kt_before));
	276
	277	local_irq_enable();
	278
	279	if (!(lapic_timer_reliable_states & (1 << (cstate))))
	280	clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_EXIT, &cpu);
	281
e978aa7d DD	282	/* Update cpuidle counters */
	283	dev->last_residency = (int)usec_delta;
	284
	285	return index;
26717172 LB	286	}
26717172 LB	287
2a2d31c8 SL	288	static void __setup_broadcast_timer(void *arg)
	289	{
	290	unsigned long reason = (unsigned long)arg;
	291	int cpu = smp_processor_id();
	292
	293	reason = reason ?
	294	CLOCK_EVT_NOTIFY_BROADCAST_ON : CLOCK_EVT_NOTIFY_BROADCAST_OFF;
	295
	296	clockevents_notify(reason, &cpu);
	297	}
	298
ec30f343	299	static int setup_broadcast_cpuhp_notify(struct notifier_block *n,
2a2d31c8 SL	300	unsigned long action, void *hcpu)
	301	{
	302	int hotcpu = (unsigned long)hcpu;
	303
	304	switch (action & 0xf) {
	305	case CPU_ONLINE:
	306	smp_call_function_single(hotcpu, __setup_broadcast_timer,
	307	(void *)true, 1);
	308	break;
2a2d31c8 SL	309	}
	310	return NOTIFY_OK;
	311	}
	312
ec30f343	313	static struct notifier_block setup_broadcast_notifier = {
2a2d31c8 SL	314	.notifier_call = setup_broadcast_cpuhp_notify,
	315	};
	316
14796fca LB	317	static void auto_demotion_disable(void *dummy)
	318	{
	319	unsigned long long msr_bits;
	320
	321	rdmsrl(MSR_NHM_SNB_PKG_CST_CFG_CTL, msr_bits);
	322	msr_bits &= ~auto_demotion_disable_flags;
	323	wrmsrl(MSR_NHM_SNB_PKG_CST_CFG_CTL, msr_bits);
	324	}
	325
26717172 LB	326	/*
	327	* intel_idle_probe()
	328	*/
	329	static int intel_idle_probe(void)
	330	{
c4236282	331	unsigned int eax, ebx, ecx;
26717172 LB	332
	333	if (max_cstate == 0) {
	334	pr_debug(PREFIX "disabled\n");
	335	return -EPERM;
	336	}
	337
	338	if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL)
	339	return -ENODEV;
	340
	341	if (!boot_cpu_has(X86_FEATURE_MWAIT))
	342	return -ENODEV;
	343
	344	if (boot_cpu_data.cpuid_level < CPUID_MWAIT_LEAF)
	345	return -ENODEV;
	346
c4236282	347	cpuid(CPUID_MWAIT_LEAF, &eax, &ebx, &ecx, &mwait_substates);
26717172 LB	348
26717172 LB	349	if (!(ecx & CPUID5_ECX_EXTENSIONS_SUPPORTED) \|\|
5c2a9f06 TR	350	!(ecx & CPUID5_ECX_INTERRUPT_BREAK) \|\|
5c2a9f06 TR	351	!mwait_substates)
26717172	352	return -ENODEV;
26717172	353
c4236282	354	pr_debug(PREFIX "MWAIT substates: 0x%x\n", mwait_substates);
26717172	355
26717172 LB	356
	357	if (boot_cpu_data.x86 != 6) /* family 6 */
	358	return -ENODEV;
	359
	360	switch (boot_cpu_data.x86_model) {
	361
	362	case 0x1A: /* Core i7, Xeon 5500 series */
	363	case 0x1E: /* Core i7 and i5 Processor - Lynnfield Jasper Forest */
	364	case 0x1F: /* Core i7 and i5 Processor - Nehalem */
	365	case 0x2E: /* Nehalem-EX Xeon */
ec67a2ba	366	case 0x2F: /* Westmere-EX Xeon */
26717172 LB	367	case 0x25: /* Westmere */
	368	case 0x2C: /* Westmere */
	369	cpuidle_state_table = nehalem_cstates;
14796fca LB	370	auto_demotion_disable_flags =
14796fca LB	371	(NHM_C1_AUTO_DEMOTE \| NHM_C3_AUTO_DEMOTE);
26717172 LB	372	break;
	373
	374	case 0x1C: /* 28 - Atom Processor */
bfb53ccf LB	375	cpuidle_state_table = atom_cstates;
	376	break;
	377
4725fd3c	378	case 0x26: /* 38 - Lincroft Atom Processor */
26717172	379	cpuidle_state_table = atom_cstates;
bfb53ccf	380	auto_demotion_disable_flags = ATM_LNC_C6_AUTO_DEMOTE;
26717172	381	break;
d13780d4 LB	382
	383	case 0x2A: /* SNB */
	384	case 0x2D: /* SNB Xeon */
	385	cpuidle_state_table = snb_cstates;
d13780d4	386	break;
26717172 LB	387
	388	default:
	389	pr_debug(PREFIX "does not run on family %d model %d\n",
	390	boot_cpu_data.x86, boot_cpu_data.x86_model);
	391	return -ENODEV;
	392	}
	393
56b9aea3	394	if (boot_cpu_has(X86_FEATURE_ARAT)) /* Always Reliable APIC Timer */
2a2d31c8 SL	395	lapic_timer_reliable_states = LAPIC_TIMER_ALWAYS_RELIABLE;
2a2d31c8 SL	396	else {
39a74fde	397	on_each_cpu(__setup_broadcast_timer, (void *)true, 1);
2a2d31c8 SL	398	register_cpu_notifier(&setup_broadcast_notifier);
2a2d31c8 SL	399	}
56b9aea3	400
26717172 LB	401	pr_debug(PREFIX "v" INTEL_IDLE_VERSION
	402	" model 0x%X\n", boot_cpu_data.x86_model);
	403
	404	pr_debug(PREFIX "lapic_timer_reliable_states 0x%x\n",
	405	lapic_timer_reliable_states);
	406	return 0;
	407	}
	408
	409	/*
	410	* intel_idle_cpuidle_devices_uninit()
	411	* unregister, free cpuidle_devices
	412	*/
	413	static void intel_idle_cpuidle_devices_uninit(void)
	414	{
	415	int i;
	416	struct cpuidle_device *dev;
	417
	418	for_each_online_cpu(i) {
	419	dev = per_cpu_ptr(intel_idle_cpuidle_devices, i);
	420	cpuidle_unregister_device(dev);
	421	}
	422
	423	free_percpu(intel_idle_cpuidle_devices);
	424	return;
	425	}
46bcfad7 DD	426	/*
	427	* intel_idle_cpuidle_driver_init()
	428	* allocate, initialize cpuidle_states
	429	*/
	430	static int intel_idle_cpuidle_driver_init(void)
	431	{
	432	int cstate;
	433	struct cpuidle_driver *drv = &intel_idle_driver;
	434
	435	drv->state_count = 1;
	436
	437	for (cstate = 1; cstate < MWAIT_MAX_NUM_CSTATES; ++cstate) {
	438	int num_substates;
	439
	440	if (cstate > max_cstate) {
	441	printk(PREFIX "max_cstate %d reached\n",
	442	max_cstate);
	443	break;
	444	}
	445
	446	/* does the state exist in CPUID.MWAIT? */
	447	num_substates = (mwait_substates >> ((cstate) * 4))
	448	& MWAIT_SUBSTATE_MASK;
	449	if (num_substates == 0)
	450	continue;
	451	/* is the state not enabled? */
	452	if (cpuidle_state_table[cstate].enter == NULL) {
	453	/* does the driver not know about the state? */
	454	if (*cpuidle_state_table[cstate].name == '\0')
	455	pr_debug(PREFIX "unaware of model 0x%x"
	456	" MWAIT %d please"
	457	" contact lenb@kernel.org",
	458	boot_cpu_data.x86_model, cstate);
	459	continue;
	460	}
	461
	462	if ((cstate > 2) &&
	463	!boot_cpu_has(X86_FEATURE_NONSTOP_TSC))
	464	mark_tsc_unstable("TSC halts in idle"
	465	" states deeper than C2");
	466
	467	drv->states[drv->state_count] = /* structure copy */
	468	cpuidle_state_table[cstate];
	469
	470	drv->state_count += 1;
	471	}
	472
	473	if (auto_demotion_disable_flags)
39a74fde	474	on_each_cpu(auto_demotion_disable, NULL, 1);
46bcfad7 DD	475
	476	return 0;
	477	}
	478
	479
26717172	480	/*
65b7f839	481	* intel_idle_cpu_init()
26717172	482	* allocate, initialize, register cpuidle_devices
65b7f839	483	* @cpu: cpu/core to initialize
26717172	484	*/
65b7f839	485	int intel_idle_cpu_init(int cpu)
26717172	486	{
65b7f839	487	int cstate;
26717172 LB	488	struct cpuidle_device *dev;
26717172 LB	489
65b7f839	490	dev = per_cpu_ptr(intel_idle_cpuidle_devices, cpu);
26717172	491
65b7f839	492	dev->state_count = 1;
26717172	493
65b7f839 TR	494	for (cstate = 1; cstate < MWAIT_MAX_NUM_CSTATES; ++cstate) {
65b7f839 TR	495	int num_substates;
26717172	496
65b7f839 TR	497	if (cstate > max_cstate) {
	498	printk(PREFIX "max_cstate %d reached\n",
	499	max_cstate);
	500	break;
	501	}
26717172	502
65b7f839 TR	503	/* does the state exist in CPUID.MWAIT? */
	504	num_substates = (mwait_substates >> ((cstate) * 4))
	505	& MWAIT_SUBSTATE_MASK;
	506	if (num_substates == 0)
	507	continue;
	508	/* is the state not enabled? */
	509	if (cpuidle_state_table[cstate].enter == NULL)
	510	continue;
26717172	511
65b7f839 TR	512	dev->states_usage[dev->state_count].driver_data =
65b7f839 TR	513	(void *)get_driver_data(cstate);
26717172 LB	514
	515	dev->state_count += 1;
	516	}
65b7f839	517	dev->cpu = cpu;
26717172	518
65b7f839 TR	519	if (cpuidle_register_device(dev)) {
	520	pr_debug(PREFIX "cpuidle_register_device %d failed!\n", cpu);
	521	intel_idle_cpuidle_devices_uninit();
	522	return -EIO;
26717172 LB	523	}
26717172 LB	524
65b7f839 TR	525	if (auto_demotion_disable_flags)
	526	smp_call_function_single(cpu, auto_demotion_disable, NULL, 1);
	527
26717172 LB	528	return 0;
	529	}
	530
	531
	532	static int __init intel_idle_init(void)
	533	{
65b7f839	534	int retval, i;
26717172	535
d1896049 TR	536	/* Do not load intel_idle at all for now if idle= is passed */
	537	if (boot_option_idle_override != IDLE_NO_OVERRIDE)
	538	return -ENODEV;
	539
26717172 LB	540	retval = intel_idle_probe();
	541	if (retval)
	542	return retval;
	543
46bcfad7	544	intel_idle_cpuidle_driver_init();
26717172 LB	545	retval = cpuidle_register_driver(&intel_idle_driver);
	546	if (retval) {
	547	printk(KERN_DEBUG PREFIX "intel_idle yielding to %s",
	548	cpuidle_get_driver()->name);
	549	return retval;
	550	}
	551
65b7f839 TR	552	intel_idle_cpuidle_devices = alloc_percpu(struct cpuidle_device);
	553	if (intel_idle_cpuidle_devices == NULL)
	554	return -ENOMEM;
	555
	556	for_each_online_cpu(i) {
	557	retval = intel_idle_cpu_init(i);
	558	if (retval) {
	559	cpuidle_unregister_driver(&intel_idle_driver);
	560	return retval;
	561	}
26717172 LB	562	}
	563
	564	return 0;
	565	}
	566
	567	static void __exit intel_idle_exit(void)
	568	{
	569	intel_idle_cpuidle_devices_uninit();
	570	cpuidle_unregister_driver(&intel_idle_driver);
	571
2a2d31c8	572	if (lapic_timer_reliable_states != LAPIC_TIMER_ALWAYS_RELIABLE) {
39a74fde	573	on_each_cpu(__setup_broadcast_timer, (void *)false, 1);
2a2d31c8 SL	574	unregister_cpu_notifier(&setup_broadcast_notifier);
	575	}
	576
26717172 LB	577	return;
	578	}
	579
	580	module_init(intel_idle_init);
	581	module_exit(intel_idle_exit);
	582
26717172	583	module_param(max_cstate, int, 0444);
26717172 LB	584
	585	MODULE_AUTHOR("Len Brown <len.brown@intel.com>");
	586	MODULE_DESCRIPTION("Cpuidle driver for Intel Hardware v" INTEL_IDLE_VERSION);
	587	MODULE_LICENSE("GPL");