intel_idle.c 13.4 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61
/*
 * 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>
62 63
#include <linux/notifier.h>
#include <linux/cpu.h>
64
#include <linux/module.h>
65
#include <asm/mwait.h>
66
#include <asm/msr.h>
67 68 69 70 71 72 73 74 75 76 77

#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;

78
static unsigned int mwait_substates;
79

80
#define LAPIC_TIMER_ALWAYS_RELIABLE 0xFFFFFFFF
81
/* Reliable LAPIC Timer States, bit 1 for C1 etc.  */
82
static unsigned int lapic_timer_reliable_states = (1 << 1);	 /* Default to only C1 */
83

84
static struct cpuidle_device __percpu *intel_idle_cpuidle_devices;
85 86 87 88
static int intel_idle(struct cpuidle_device *dev, struct cpuidle_state *state);

static struct cpuidle_state *cpuidle_state_table;

89 90 91 92 93 94
/*
 * 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;

95 96 97 98 99 100 101 102
/*
 * 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

103 104 105 106 107 108 109 110
/*
 * 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 */
111
		.name = "C1-NHM",
112 113 114 115 116 117 118
		.desc = "MWAIT 0x00",
		.driver_data = (void *) 0x00,
		.flags = CPUIDLE_FLAG_TIME_VALID,
		.exit_latency = 3,
		.target_residency = 6,
		.enter = &intel_idle },
	{ /* MWAIT C2 */
119
		.name = "C3-NHM",
120 121
		.desc = "MWAIT 0x10",
		.driver_data = (void *) 0x10,
122
		.flags = CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
123 124 125 126
		.exit_latency = 20,
		.target_residency = 80,
		.enter = &intel_idle },
	{ /* MWAIT C3 */
127
		.name = "C6-NHM",
128 129
		.desc = "MWAIT 0x20",
		.driver_data = (void *) 0x20,
130
		.flags = CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
131 132 133 134 135
		.exit_latency = 200,
		.target_residency = 800,
		.enter = &intel_idle },
};

136 137 138
static struct cpuidle_state snb_cstates[MWAIT_MAX_NUM_CSTATES] = {
	{ /* MWAIT C0 */ },
	{ /* MWAIT C1 */
139
		.name = "C1-SNB",
140 141 142 143
		.desc = "MWAIT 0x00",
		.driver_data = (void *) 0x00,
		.flags = CPUIDLE_FLAG_TIME_VALID,
		.exit_latency = 1,
144
		.target_residency = 1,
145 146
		.enter = &intel_idle },
	{ /* MWAIT C2 */
147
		.name = "C3-SNB",
148 149
		.desc = "MWAIT 0x10",
		.driver_data = (void *) 0x10,
150
		.flags = CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
151
		.exit_latency = 80,
152
		.target_residency = 211,
153 154
		.enter = &intel_idle },
	{ /* MWAIT C3 */
155
		.name = "C6-SNB",
156 157
		.desc = "MWAIT 0x20",
		.driver_data = (void *) 0x20,
158
		.flags = CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
159
		.exit_latency = 104,
160
		.target_residency = 345,
161 162
		.enter = &intel_idle },
	{ /* MWAIT C4 */
163
		.name = "C7-SNB",
164 165
		.desc = "MWAIT 0x30",
		.driver_data = (void *) 0x30,
166
		.flags = CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
167
		.exit_latency = 109,
168
		.target_residency = 345,
169 170 171
		.enter = &intel_idle },
};

172 173 174
static struct cpuidle_state atom_cstates[MWAIT_MAX_NUM_CSTATES] = {
	{ /* MWAIT C0 */ },
	{ /* MWAIT C1 */
175
		.name = "C1-ATM",
176 177 178 179 180 181 182
		.desc = "MWAIT 0x00",
		.driver_data = (void *) 0x00,
		.flags = CPUIDLE_FLAG_TIME_VALID,
		.exit_latency = 1,
		.target_residency = 4,
		.enter = &intel_idle },
	{ /* MWAIT C2 */
183
		.name = "C2-ATM",
184 185 186 187 188 189 190 191
		.desc = "MWAIT 0x10",
		.driver_data = (void *) 0x10,
		.flags = CPUIDLE_FLAG_TIME_VALID,
		.exit_latency = 20,
		.target_residency = 80,
		.enter = &intel_idle },
	{ /* MWAIT C3 */ },
	{ /* MWAIT C4 */
192
		.name = "C4-ATM",
193 194
		.desc = "MWAIT 0x30",
		.driver_data = (void *) 0x30,
195
		.flags = CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
196 197 198 199 200
		.exit_latency = 100,
		.target_residency = 400,
		.enter = &intel_idle },
	{ /* MWAIT C5 */ },
	{ /* MWAIT C6 */
201
		.name = "C6-ATM",
Len Brown's avatar
Len Brown committed
202 203
		.desc = "MWAIT 0x52",
		.driver_data = (void *) 0x52,
204
		.flags = CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
Len Brown's avatar
Len Brown committed
205 206 207
		.exit_latency = 140,
		.target_residency = 560,
		.enter = &intel_idle },
208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228
};

/**
 * intel_idle
 * @dev: cpuidle_device
 * @state: cpuidle state
 *
 */
static int intel_idle(struct cpuidle_device *dev, struct cpuidle_state *state)
{
	unsigned long ecx = 1; /* break on interrupt flag */
	unsigned long eax = (unsigned long)cpuidle_get_statedata(state);
	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;

	local_irq_disable();

229
	/*
230 231
	 * leave_mm() to avoid costly and often unnecessary wakeups
	 * for flushing the user TLB's associated with the active mm.
232
	 */
233
	if (state->flags & CPUIDLE_FLAG_TLB_FLUSHED)
234 235
		leave_mm(cpu);

236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262
	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);

	return usec_delta;
}

263 264 265 266 267 268 269 270 271 272 273
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);
}

274
static int setup_broadcast_cpuhp_notify(struct notifier_block *n,
275 276 277 278 279 280 281 282 283 284 285 286 287
		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;
}

288
static struct notifier_block setup_broadcast_notifier = {
289 290 291
	.notifier_call = setup_broadcast_cpuhp_notify,
};

292 293 294 295 296 297 298 299 300
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);
}

301 302 303 304 305
/*
 * intel_idle_probe()
 */
static int intel_idle_probe(void)
{
306
	unsigned int eax, ebx, ecx;
307 308 309 310 311 312 313 314 315 316 317 318 319 320 321

	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;

322
	cpuid(CPUID_MWAIT_LEAF, &eax, &ebx, &ecx, &mwait_substates);
323 324 325 326 327

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

328
	pr_debug(PREFIX "MWAIT substates: 0x%x\n", mwait_substates);
329 330 331 332 333 334 335 336 337 338 339


	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 */
340
	case 0x2F:	/* Westmere-EX Xeon */
341 342 343
	case 0x25:	/* Westmere */
	case 0x2C:	/* Westmere */
		cpuidle_state_table = nehalem_cstates;
344 345
		auto_demotion_disable_flags =
			(NHM_C1_AUTO_DEMOTE | NHM_C3_AUTO_DEMOTE);
346 347 348
		break;

	case 0x1C:	/* 28 - Atom Processor */
349 350 351
		cpuidle_state_table = atom_cstates;
		break;

352
	case 0x26:	/* 38 - Lincroft Atom Processor */
353
		cpuidle_state_table = atom_cstates;
354
		auto_demotion_disable_flags = ATM_LNC_C6_AUTO_DEMOTE;
355
		break;
356 357 358 359 360

	case 0x2A:	/* SNB */
	case 0x2D:	/* SNB Xeon */
		cpuidle_state_table = snb_cstates;
		break;
361 362 363 364 365 366 367

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

368
	if (boot_cpu_has(X86_FEATURE_ARAT))	/* Always Reliable APIC Timer */
369 370 371 372 373
		lapic_timer_reliable_states = LAPIC_TIMER_ALWAYS_RELIABLE;
	else {
		smp_call_function(__setup_broadcast_timer, (void *)true, 1);
		register_cpu_notifier(&setup_broadcast_notifier);
	}
374

375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427
	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_devices_init()
 * allocate, initialize, register cpuidle_devices
 */
static int intel_idle_cpuidle_devices_init(void)
{
	int i, cstate;
	struct cpuidle_device *dev;

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

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

		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? */
428
			num_substates = (mwait_substates >> ((cstate) * 4))
429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461
						& 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");

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

			dev->state_count += 1;
		}

		dev->cpu = i;
		if (cpuidle_register_device(dev)) {
			pr_debug(PREFIX "cpuidle_register_device %d failed!\n",
				 i);
			intel_idle_cpuidle_devices_uninit();
			return -EIO;
		}
	}
462 463
	if (auto_demotion_disable_flags)
		smp_call_function(auto_demotion_disable, NULL, 1);
464 465 466 467 468 469 470 471 472

	return 0;
}


static int __init intel_idle_init(void)
{
	int retval;

473 474 475 476
	/* Do not load intel_idle at all for now if idle= is passed */
	if (boot_option_idle_override != IDLE_NO_OVERRIDE)
		return -ENODEV;

477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501
	retval = intel_idle_probe();
	if (retval)
		return retval;

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

	retval = intel_idle_cpuidle_devices_init();
	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);

502 503 504 505 506
	if (lapic_timer_reliable_states != LAPIC_TIMER_ALWAYS_RELIABLE) {
		smp_call_function(__setup_broadcast_timer, (void *)false, 1);
		unregister_cpu_notifier(&setup_broadcast_notifier);
	}

507 508 509 510 511 512 513 514 515 516 517
	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");