smpboot.c 34.7 KB
Newer Older
1 2 3
/*
 *	x86 SMP booting functions
 *
4
 *	(c) 1995 Alan Cox, Building #3 <alan@lxorguk.ukuu.org.uk>
I
Ingo Molnar 已提交
5
 *	(c) 1998, 1999, 2000, 2009 Ingo Molnar <mingo@redhat.com>
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
 *	Copyright 2001 Andi Kleen, SuSE Labs.
 *
 *	Much of the core SMP work is based on previous work by Thomas Radke, to
 *	whom a great many thanks are extended.
 *
 *	Thanks to Intel for making available several different Pentium,
 *	Pentium Pro and Pentium-II/Xeon MP machines.
 *	Original development of Linux SMP code supported by Caldera.
 *
 *	This code is released under the GNU General Public License version 2 or
 *	later.
 *
 *	Fixes
 *		Felix Koop	:	NR_CPUS used properly
 *		Jose Renau	:	Handle single CPU case.
 *		Alan Cox	:	By repeated request 8) - Total BogoMIPS report.
 *		Greg Wright	:	Fix for kernel stacks panic.
 *		Erich Boleyn	:	MP v1.4 and additional changes.
 *	Matthias Sattler	:	Changes for 2.1 kernel map.
 *	Michel Lespinasse	:	Changes for 2.1 kernel map.
 *	Michael Chastain	:	Change trampoline.S to gnu as.
 *		Alan Cox	:	Dumb bug: 'B' step PPro's are fine
 *		Ingo Molnar	:	Added APIC timers, based on code
 *					from Jose Renau
 *		Ingo Molnar	:	various cleanups and rewrites
 *		Tigran Aivazian	:	fixed "0.00 in /proc/uptime on SMP" bug.
 *	Maciej W. Rozycki	:	Bits for genuine 82489DX APICs
 *	Andi Kleen		:	Changed for SMP boot into long mode.
 *		Martin J. Bligh	: 	Added support for multi-quad systems
 *		Dave Jones	:	Report invalid combinations of Athlon CPUs.
 *		Rusty Russell	:	Hacked into shape for new "hotplug" boot process.
 *      Andi Kleen              :       Converted to new state machine.
 *	Ashok Raj		: 	CPU hotplug support
 *	Glauber Costa		:	i386 and x86_64 integration
 */

42 43
#include <linux/init.h>
#include <linux/smp.h>
44
#include <linux/module.h>
45
#include <linux/sched.h>
46
#include <linux/percpu.h>
G
Glauber Costa 已提交
47
#include <linux/bootmem.h>
48 49
#include <linux/err.h>
#include <linux/nmi.h>
50
#include <linux/tboot.h>
51
#include <linux/stackprotector.h>
52
#include <linux/gfp.h>
53

54
#include <asm/acpi.h>
55
#include <asm/desc.h>
56 57
#include <asm/nmi.h>
#include <asm/irq.h>
58
#include <asm/idle.h>
59
#include <asm/trampoline.h>
60 61
#include <asm/cpu.h>
#include <asm/numa.h>
62 63 64
#include <asm/pgtable.h>
#include <asm/tlbflush.h>
#include <asm/mtrr.h>
65
#include <asm/mwait.h>
I
Ingo Molnar 已提交
66
#include <asm/apic.h>
67
#include <asm/io_apic.h>
68
#include <asm/setup.h>
T
Tejun Heo 已提交
69
#include <asm/uv/uv.h>
70
#include <linux/mc146818rtc.h>
71

72
#include <asm/smpboot_hooks.h>
73
#include <asm/i8259.h>
74

75 76 77
/* State of each CPU */
DEFINE_PER_CPU(int, cpu_state) = { 0 };

78 79 80 81 82 83 84 85 86 87 88 89
/* Store all idle threads, this can be reused instead of creating
* a new thread. Also avoids complicated thread destroy functionality
* for idle threads.
*/
#ifdef CONFIG_HOTPLUG_CPU
/*
 * Needed only for CONFIG_HOTPLUG_CPU because __cpuinitdata is
 * removed after init for !CONFIG_HOTPLUG_CPU.
 */
static DEFINE_PER_CPU(struct task_struct *, idle_thread_array);
#define get_idle_for_cpu(x)      (per_cpu(idle_thread_array, x))
#define set_idle_for_cpu(x, p)   (per_cpu(idle_thread_array, x) = (p))
90 91 92 93 94 95 96

/*
 * We need this for trampoline_base protection from concurrent accesses when
 * off- and onlining cores wildly.
 */
static DEFINE_MUTEX(x86_cpu_hotplug_driver_mutex);

97
void cpu_hotplug_driver_lock(void)
98 99 100 101
{
        mutex_lock(&x86_cpu_hotplug_driver_mutex);
}

102
void cpu_hotplug_driver_unlock(void)
103 104 105 106 107 108
{
        mutex_unlock(&x86_cpu_hotplug_driver_mutex);
}

ssize_t arch_cpu_probe(const char *buf, size_t count) { return -1; }
ssize_t arch_cpu_release(const char *buf, size_t count) { return -1; }
109
#else
110
static struct task_struct *idle_thread_array[NR_CPUS] __cpuinitdata ;
111 112 113
#define get_idle_for_cpu(x)      (idle_thread_array[(x)])
#define set_idle_for_cpu(x, p)   (idle_thread_array[(x)] = (p))
#endif
114

115 116 117 118 119 120 121 122
/* Number of siblings per CPU package */
int smp_num_siblings = 1;
EXPORT_SYMBOL(smp_num_siblings);

/* Last level cache ID of each logical CPU */
DEFINE_PER_CPU(u16, cpu_llc_id) = BAD_APICID;

/* representing HT siblings of each logical CPU */
123
DEFINE_PER_CPU(cpumask_var_t, cpu_sibling_map);
124 125 126
EXPORT_PER_CPU_SYMBOL(cpu_sibling_map);

/* representing HT and core siblings of each logical CPU */
127
DEFINE_PER_CPU(cpumask_var_t, cpu_core_map);
128 129
EXPORT_PER_CPU_SYMBOL(cpu_core_map);

130 131
DEFINE_PER_CPU(cpumask_var_t, cpu_llc_shared_map);

132 133 134
/* Per CPU bogomips and other parameters */
DEFINE_PER_CPU_SHARED_ALIGNED(struct cpuinfo_x86, cpu_info);
EXPORT_PER_CPU_SYMBOL(cpu_info);
135

136
atomic_t init_deasserted;
137 138 139 140 141

/*
 * Report back to the Boot Processor.
 * Running on AP.
 */
I
Ingo Molnar 已提交
142
static void __cpuinit smp_callin(void)
143 144 145 146 147 148 149 150 151 152
{
	int cpuid, phys_id;
	unsigned long timeout;

	/*
	 * If waken up by an INIT in an 82489DX configuration
	 * we may get here before an INIT-deassert IPI reaches
	 * our local APIC.  We have to wait for the IPI or we'll
	 * lock up on an APIC access.
	 */
153 154
	if (apic->wait_for_init_deassert)
		apic->wait_for_init_deassert(&init_deasserted);
155 156 157 158

	/*
	 * (This works even if the APIC is not enabled.)
	 */
159
	phys_id = read_apic_id();
160
	cpuid = smp_processor_id();
161
	if (cpumask_test_cpu(cpuid, cpu_callin_mask)) {
162 163 164
		panic("%s: phys CPU#%d, CPU#%d already present??\n", __func__,
					phys_id, cpuid);
	}
165
	pr_debug("CPU#%d (phys ID: %d) waiting for CALLOUT\n", cpuid, phys_id);
166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182

	/*
	 * STARTUP IPIs are fragile beasts as they might sometimes
	 * trigger some glue motherboard logic. Complete APIC bus
	 * silence for 1 second, this overestimates the time the
	 * boot CPU is spending to send the up to 2 STARTUP IPIs
	 * by a factor of two. This should be enough.
	 */

	/*
	 * Waiting 2s total for startup (udelay is not yet working)
	 */
	timeout = jiffies + 2*HZ;
	while (time_before(jiffies, timeout)) {
		/*
		 * Has the boot CPU finished it's STARTUP sequence?
		 */
183
		if (cpumask_test_cpu(cpuid, cpu_callout_mask))
184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199
			break;
		cpu_relax();
	}

	if (!time_before(jiffies, timeout)) {
		panic("%s: CPU%d started up but did not get a callout!\n",
		      __func__, cpuid);
	}

	/*
	 * the boot CPU has finished the init stage and is spinning
	 * on callin_map until we finish. We are free to set up this
	 * CPU, first the APIC. (this is probably redundant on most
	 * boards)
	 */

200
	pr_debug("CALLIN, before setup_local_APIC().\n");
201 202
	if (apic->smp_callin_clear_local_apic)
		apic->smp_callin_clear_local_apic();
203 204 205
	setup_local_APIC();
	end_local_APIC_setup();

206 207 208
	/*
	 * Need to setup vector mappings before we enable interrupts.
	 */
209
	setup_vector_irq(smp_processor_id());
210 211 212 213 214 215 216

	/*
	 * Save our processor parameters. Note: this information
	 * is needed for clock calibration.
	 */
	smp_store_cpu_info(cpuid);

217 218
	/*
	 * Get our bogomips.
219 220 221
	 * Update loops_per_jiffy in cpu_data. Previous call to
	 * smp_store_cpu_info() stored a value that is close but not as
	 * accurate as the value just calculated.
222 223
	 */
	calibrate_delay();
224
	cpu_data(cpuid).loops_per_jiffy = loops_per_jiffy;
225
	pr_debug("Stack at about %p\n", &cpuid);
226

227 228 229 230 231 232 233
	/*
	 * This must be done before setting cpu_online_mask
	 * or calling notify_cpu_starting.
	 */
	set_cpu_sibling_map(raw_smp_processor_id());
	wmb();

234 235
	notify_cpu_starting(cpuid);

236 237 238
	/*
	 * Allow the master to continue.
	 */
239
	cpumask_set_cpu(cpuid, cpu_callin_mask);
240 241
}

242 243 244
/*
 * Activate a secondary processor.
 */
245
notrace static void __cpuinit start_secondary(void *unused)
246 247 248 249 250 251
{
	/*
	 * Don't put *anything* before cpu_init(), SMP booting is too
	 * fragile that we want to limit the things done here to the
	 * most necessary things.
	 */
252 253 254
	cpu_init();
	preempt_disable();
	smp_callin();
255 256

#ifdef CONFIG_X86_32
257
	/* switch away from the initial page table */
258 259 260 261
	load_cr3(swapper_pg_dir);
	__flush_tlb_all();
#endif

262 263 264 265 266 267 268 269 270 271 272 273 274 275
	/* otherwise gcc will move up smp_processor_id before the cpu_init */
	barrier();
	/*
	 * Check TSC synchronization with the BP:
	 */
	check_tsc_sync_target();

	/*
	 * We need to hold call_lock, so there is no inconsistency
	 * between the time smp_call_function() determines number of
	 * IPI recipients, and the time when the determination is made
	 * for which cpus receive the IPI. Holding this
	 * lock helps us to not include this cpu in a currently in progress
	 * smp_call_function().
276 277 278 279
	 *
	 * We need to hold vector_lock so there the set of online cpus
	 * does not change while we are assigning vectors to cpus.  Holding
	 * this lock ensures we don't half assign or remove an irq from a cpu.
280
	 */
281
	ipi_call_lock();
282
	lock_vector_lock();
283
	set_cpu_online(smp_processor_id(), true);
284
	unlock_vector_lock();
285
	ipi_call_unlock();
286
	per_cpu(cpu_state, smp_processor_id()) = CPU_ONLINE;
287
	x86_platform.nmi_init();
288

289 290 291
	/* enable local interrupts */
	local_irq_enable();

292 293
	/* to prevent fake stack check failure in clock setup */
	boot_init_stack_canary();
294

295
	x86_cpuinit.setup_percpu_clockev();
296 297 298 299 300

	wmb();
	cpu_idle();
}

301 302 303 304 305 306 307 308 309
/*
 * The bootstrap kernel entry code has set these up. Save them for
 * a given CPU
 */

void __cpuinit smp_store_cpu_info(int id)
{
	struct cpuinfo_x86 *c = &cpu_data(id);

310
	*c = boot_cpu_data;
311 312 313 314 315
	c->cpu_index = id;
	if (id != 0)
		identify_secondary_cpu(c);
}

316 317 318 319 320 321
static void __cpuinit link_thread_siblings(int cpu1, int cpu2)
{
	cpumask_set_cpu(cpu1, cpu_sibling_mask(cpu2));
	cpumask_set_cpu(cpu2, cpu_sibling_mask(cpu1));
	cpumask_set_cpu(cpu1, cpu_core_mask(cpu2));
	cpumask_set_cpu(cpu2, cpu_core_mask(cpu1));
322 323
	cpumask_set_cpu(cpu1, cpu_llc_shared_mask(cpu2));
	cpumask_set_cpu(cpu2, cpu_llc_shared_mask(cpu1));
324 325
}

326

327 328 329 330 331
void __cpuinit set_cpu_sibling_map(int cpu)
{
	int i;
	struct cpuinfo_x86 *c = &cpu_data(cpu);

332
	cpumask_set_cpu(cpu, cpu_sibling_setup_mask);
333 334

	if (smp_num_siblings > 1) {
335 336 337
		for_each_cpu(i, cpu_sibling_setup_mask) {
			struct cpuinfo_x86 *o = &cpu_data(i);

338 339
			if (cpu_has(c, X86_FEATURE_TOPOEXT)) {
				if (c->phys_proc_id == o->phys_proc_id &&
340
				    per_cpu(cpu_llc_id, cpu) == per_cpu(cpu_llc_id, i) &&
341 342 343 344 345
				    c->compute_unit_id == o->compute_unit_id)
					link_thread_siblings(cpu, i);
			} else if (c->phys_proc_id == o->phys_proc_id &&
				   c->cpu_core_id == o->cpu_core_id) {
				link_thread_siblings(cpu, i);
346 347 348
			}
		}
	} else {
349
		cpumask_set_cpu(cpu, cpu_sibling_mask(cpu));
350 351
	}

352
	cpumask_set_cpu(cpu, cpu_llc_shared_mask(cpu));
353

354
	if (__this_cpu_read(cpu_info.x86_max_cores) == 1) {
355
		cpumask_copy(cpu_core_mask(cpu), cpu_sibling_mask(cpu));
356 357 358 359
		c->booted_cores = 1;
		return;
	}

360
	for_each_cpu(i, cpu_sibling_setup_mask) {
361 362
		if (per_cpu(cpu_llc_id, cpu) != BAD_APICID &&
		    per_cpu(cpu_llc_id, cpu) == per_cpu(cpu_llc_id, i)) {
363 364
			cpumask_set_cpu(i, cpu_llc_shared_mask(cpu));
			cpumask_set_cpu(cpu, cpu_llc_shared_mask(i));
365 366
		}
		if (c->phys_proc_id == cpu_data(i).phys_proc_id) {
367 368
			cpumask_set_cpu(i, cpu_core_mask(cpu));
			cpumask_set_cpu(cpu, cpu_core_mask(i));
369 370 371
			/*
			 *  Does this new cpu bringup a new core?
			 */
372
			if (cpumask_weight(cpu_sibling_mask(cpu)) == 1) {
373 374 375 376
				/*
				 * for each core in package, increment
				 * the booted_cores for this new cpu
				 */
377
				if (cpumask_first(cpu_sibling_mask(i)) == i)
378 379 380 381 382 383 384 385 386 387 388 389 390
					c->booted_cores++;
				/*
				 * increment the core count for all
				 * the other cpus in this package
				 */
				if (i != cpu)
					cpu_data(i).booted_cores++;
			} else if (i != cpu && !c->booted_cores)
				c->booted_cores = cpu_data(i).booted_cores;
		}
	}
}

391
/* maps the cpu to the sched domain representing multi-core */
R
Rusty Russell 已提交
392
const struct cpumask *cpu_coregroup_mask(int cpu)
393 394 395 396 397 398
{
	struct cpuinfo_x86 *c = &cpu_data(cpu);
	/*
	 * For perf, we return last level cache shared map.
	 * And for power savings, we return cpu_core_map
	 */
399 400
	if ((sched_mc_power_savings || sched_smt_power_savings) &&
	    !(cpu_has(c, X86_FEATURE_AMD_DCM)))
401
		return cpu_core_mask(cpu);
402
	else
403
		return cpu_llc_shared_mask(cpu);
R
Rusty Russell 已提交
404 405
}

I
Ingo Molnar 已提交
406
static void impress_friends(void)
407 408 409 410 411 412
{
	int cpu;
	unsigned long bogosum = 0;
	/*
	 * Allow the user to impress friends.
	 */
413
	pr_debug("Before bogomips.\n");
414
	for_each_possible_cpu(cpu)
415
		if (cpumask_test_cpu(cpu, cpu_callout_mask))
416 417 418
			bogosum += cpu_data(cpu).loops_per_jiffy;
	printk(KERN_INFO
		"Total of %d processors activated (%lu.%02lu BogoMIPS).\n",
419
		num_online_cpus(),
420 421 422
		bogosum/(500000/HZ),
		(bogosum/(5000/HZ))%100);

423
	pr_debug("Before bogocount - setting activated=1.\n");
424 425
}

426
void __inquire_remote_apic(int apicid)
427 428
{
	unsigned i, regs[] = { APIC_ID >> 4, APIC_LVR >> 4, APIC_SPIV >> 4 };
429
	const char * const names[] = { "ID", "VERSION", "SPIV" };
430 431 432
	int timeout;
	u32 status;

Y
Yinghai Lu 已提交
433
	printk(KERN_INFO "Inquiring remote APIC 0x%x...\n", apicid);
434 435

	for (i = 0; i < ARRAY_SIZE(regs); i++) {
Y
Yinghai Lu 已提交
436
		printk(KERN_INFO "... APIC 0x%x %s: ", apicid, names[i]);
437 438 439 440 441 442 443 444 445

		/*
		 * Wait for idle.
		 */
		status = safe_apic_wait_icr_idle();
		if (status)
			printk(KERN_CONT
			       "a previous APIC delivery may have failed\n");

446
		apic_icr_write(APIC_DM_REMRD | regs[i], apicid);
447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469

		timeout = 0;
		do {
			udelay(100);
			status = apic_read(APIC_ICR) & APIC_ICR_RR_MASK;
		} while (status == APIC_ICR_RR_INPROG && timeout++ < 1000);

		switch (status) {
		case APIC_ICR_RR_VALID:
			status = apic_read(APIC_RRR);
			printk(KERN_CONT "%08x\n", status);
			break;
		default:
			printk(KERN_CONT "failed\n");
		}
	}
}

/*
 * Poke the other CPU in the eye via NMI to wake it up. Remember that the normal
 * INIT, INIT, STARTUP sequence will reset the chip hard for us, and this
 * won't ... remember to clear down the APIC, etc later.
 */
470
int __cpuinit
471
wakeup_secondary_cpu_via_nmi(int logical_apicid, unsigned long start_eip)
472 473 474 475 476 477 478
{
	unsigned long send_status, accept_status = 0;
	int maxlvt;

	/* Target chip */
	/* Boot on the stack */
	/* Kick the second */
479
	apic_icr_write(APIC_DM_NMI | apic->dest_logical, logical_apicid);
480

481
	pr_debug("Waiting for send to finish...\n");
482 483 484 485 486 487
	send_status = safe_apic_wait_icr_idle();

	/*
	 * Give the other CPU some time to accept the IPI.
	 */
	udelay(200);
488
	if (APIC_INTEGRATED(apic_version[boot_cpu_physical_apicid])) {
489 490 491 492 493
		maxlvt = lapic_get_maxlvt();
		if (maxlvt > 3)			/* Due to the Pentium erratum 3AP.  */
			apic_write(APIC_ESR, 0);
		accept_status = (apic_read(APIC_ESR) & 0xEF);
	}
494
	pr_debug("NMI sent.\n");
495 496 497 498 499 500 501 502 503

	if (send_status)
		printk(KERN_ERR "APIC never delivered???\n");
	if (accept_status)
		printk(KERN_ERR "APIC delivery error (%lx).\n", accept_status);

	return (send_status | accept_status);
}

504
static int __cpuinit
505
wakeup_secondary_cpu_via_init(int phys_apicid, unsigned long start_eip)
506 507 508 509
{
	unsigned long send_status, accept_status = 0;
	int maxlvt, num_starts, j;

510 511
	maxlvt = lapic_get_maxlvt();

512 513 514 515
	/*
	 * Be paranoid about clearing APIC errors.
	 */
	if (APIC_INTEGRATED(apic_version[phys_apicid])) {
516 517
		if (maxlvt > 3)		/* Due to the Pentium erratum 3AP.  */
			apic_write(APIC_ESR, 0);
518 519 520
		apic_read(APIC_ESR);
	}

521
	pr_debug("Asserting INIT.\n");
522 523 524 525 526 527 528

	/*
	 * Turn INIT on target chip
	 */
	/*
	 * Send IPI
	 */
529 530
	apic_icr_write(APIC_INT_LEVELTRIG | APIC_INT_ASSERT | APIC_DM_INIT,
		       phys_apicid);
531

532
	pr_debug("Waiting for send to finish...\n");
533 534 535 536
	send_status = safe_apic_wait_icr_idle();

	mdelay(10);

537
	pr_debug("Deasserting INIT.\n");
538 539 540

	/* Target chip */
	/* Send IPI */
541
	apic_icr_write(APIC_INT_LEVELTRIG | APIC_DM_INIT, phys_apicid);
542

543
	pr_debug("Waiting for send to finish...\n");
544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564
	send_status = safe_apic_wait_icr_idle();

	mb();
	atomic_set(&init_deasserted, 1);

	/*
	 * Should we send STARTUP IPIs ?
	 *
	 * Determine this based on the APIC version.
	 * If we don't have an integrated APIC, don't send the STARTUP IPIs.
	 */
	if (APIC_INTEGRATED(apic_version[phys_apicid]))
		num_starts = 2;
	else
		num_starts = 0;

	/*
	 * Paravirt / VMI wants a startup IPI hook here to set up the
	 * target processor state.
	 */
	startup_ipi_hook(phys_apicid, (unsigned long) start_secondary,
565
			 stack_start);
566 567 568 569

	/*
	 * Run STARTUP IPI loop.
	 */
570
	pr_debug("#startup loops: %d.\n", num_starts);
571 572

	for (j = 1; j <= num_starts; j++) {
573
		pr_debug("Sending STARTUP #%d.\n", j);
574 575
		if (maxlvt > 3)		/* Due to the Pentium erratum 3AP.  */
			apic_write(APIC_ESR, 0);
576
		apic_read(APIC_ESR);
577
		pr_debug("After apic_write.\n");
578 579 580 581 582 583 584 585

		/*
		 * STARTUP IPI
		 */

		/* Target chip */
		/* Boot on the stack */
		/* Kick the second */
586 587
		apic_icr_write(APIC_DM_STARTUP | (start_eip >> 12),
			       phys_apicid);
588 589 590 591 592 593

		/*
		 * Give the other CPU some time to accept the IPI.
		 */
		udelay(300);

594
		pr_debug("Startup point 1.\n");
595

596
		pr_debug("Waiting for send to finish...\n");
597 598 599 600 601 602
		send_status = safe_apic_wait_icr_idle();

		/*
		 * Give the other CPU some time to accept the IPI.
		 */
		udelay(200);
603
		if (maxlvt > 3)		/* Due to the Pentium erratum 3AP.  */
604 605 606 607 608
			apic_write(APIC_ESR, 0);
		accept_status = (apic_read(APIC_ESR) & 0xEF);
		if (send_status || accept_status)
			break;
	}
609
	pr_debug("After Startup.\n");
610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634

	if (send_status)
		printk(KERN_ERR "APIC never delivered???\n");
	if (accept_status)
		printk(KERN_ERR "APIC delivery error (%lx).\n", accept_status);

	return (send_status | accept_status);
}

struct create_idle {
	struct work_struct work;
	struct task_struct *idle;
	struct completion done;
	int cpu;
};

static void __cpuinit do_fork_idle(struct work_struct *work)
{
	struct create_idle *c_idle =
		container_of(work, struct create_idle, work);

	c_idle->idle = fork_idle(c_idle->cpu);
	complete(&c_idle->done);
}

635 636 637 638
/* reduce the number of lines printed when booting a large cpu count system */
static void __cpuinit announce_cpu(int cpu, int apicid)
{
	static int current_node = -1;
639
	int node = early_cpu_to_node(cpu);
640 641 642 643 644 645 646 647 648 649 650 651 652 653 654

	if (system_state == SYSTEM_BOOTING) {
		if (node != current_node) {
			if (current_node > (-1))
				pr_cont(" Ok.\n");
			current_node = node;
			pr_info("Booting Node %3d, Processors ", node);
		}
		pr_cont(" #%d%s", cpu, cpu == (nr_cpu_ids - 1) ? " Ok.\n" : "");
		return;
	} else
		pr_info("Booting Node %d Processor %d APIC 0x%x\n",
			node, cpu, apicid);
}

655 656 657
/*
 * NOTE - on most systems this is a PHYSICAL apic ID, but on multiquad
 * (ie clustered apic addressing mode), this is a LOGICAL apic ID.
658 659
 * Returns zero if CPU booted OK, else error code from
 * ->wakeup_secondary_cpu.
660
 */
661
static int __cpuinit do_boot_cpu(int apicid, int cpu)
662 663 664
{
	unsigned long boot_error = 0;
	unsigned long start_ip;
665
	int timeout;
666
	struct create_idle c_idle = {
667 668
		.cpu	= cpu,
		.done	= COMPLETION_INITIALIZER_ONSTACK(c_idle.done),
669
	};
670

A
Andrew Morton 已提交
671
	INIT_WORK_ONSTACK(&c_idle.work, do_fork_idle);
672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687

	alternatives_smp_switch(1);

	c_idle.idle = get_idle_for_cpu(cpu);

	/*
	 * We can't use kernel_thread since we must avoid to
	 * reschedule the child.
	 */
	if (c_idle.idle) {
		c_idle.idle->thread.sp = (unsigned long) (((struct pt_regs *)
			(THREAD_SIZE +  task_stack_page(c_idle.idle))) - 1);
		init_idle(c_idle.idle, cpu);
		goto do_rest;
	}

688 689
	schedule_work(&c_idle.work);
	wait_for_completion(&c_idle.done);
690 691 692

	if (IS_ERR(c_idle.idle)) {
		printk("failed fork for CPU %d\n", cpu);
693
		destroy_work_on_stack(&c_idle.work);
694 695 696 697 698 699
		return PTR_ERR(c_idle.idle);
	}

	set_idle_for_cpu(cpu, c_idle.idle);
do_rest:
	per_cpu(current_task, cpu) = c_idle.idle;
700
#ifdef CONFIG_X86_32
701 702 703 704
	/* Stack for startup_32 can be just as for start_secondary onwards */
	irq_ctx_init(cpu);
#else
	clear_tsk_thread_flag(c_idle.idle, TIF_FORK);
705
	initial_gs = per_cpu_offset(cpu);
706 707 708
	per_cpu(kernel_stack, cpu) =
		(unsigned long)task_stack_page(c_idle.idle) -
		KERNEL_STACK_OFFSET + THREAD_SIZE;
709
#endif
710
	early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(cpu);
711
	initial_code = (unsigned long)start_secondary;
712
	stack_start  = c_idle.idle->thread.sp;
713 714

	/* start_ip had better be page-aligned! */
715
	start_ip = trampoline_address();
716

717 718
	/* So we see what's up */
	announce_cpu(cpu, apicid);
719 720 721 722 723 724

	/*
	 * This grunge runs the startup process for
	 * the targeted processor.
	 */

725 726
	printk(KERN_DEBUG "smpboot cpu %d: start_ip = %lx\n", cpu, start_ip);

727 728
	atomic_set(&init_deasserted, 0);

J
Jack Steiner 已提交
729
	if (get_uv_system_type() != UV_NON_UNIQUE_APIC) {
730

731
		pr_debug("Setting warm reset code and vector.\n");
732

J
Jack Steiner 已提交
733 734 735
		smpboot_setup_warm_reset_vector(start_ip);
		/*
		 * Be paranoid about clearing APIC errors.
736 737 738 739 740
		*/
		if (APIC_INTEGRATED(apic_version[boot_cpu_physical_apicid])) {
			apic_write(APIC_ESR, 0);
			apic_read(APIC_ESR);
		}
J
Jack Steiner 已提交
741
	}
742 743

	/*
744 745
	 * Kick the secondary CPU. Use the method in the APIC driver
	 * if it's defined - or use an INIT boot APIC message otherwise:
746
	 */
747 748 749 750
	if (apic->wakeup_secondary_cpu)
		boot_error = apic->wakeup_secondary_cpu(apicid, start_ip);
	else
		boot_error = wakeup_secondary_cpu_via_init(apicid, start_ip);
751 752 753 754 755

	if (!boot_error) {
		/*
		 * allow APs to start initializing.
		 */
756
		pr_debug("Before Callout %d.\n", cpu);
757
		cpumask_set_cpu(cpu, cpu_callout_mask);
758
		pr_debug("After Callout %d.\n", cpu);
759 760 761 762 763

		/*
		 * Wait 5s total for a response
		 */
		for (timeout = 0; timeout < 50000; timeout++) {
764
			if (cpumask_test_cpu(cpu, cpu_callin_mask))
765 766
				break;	/* It has booted */
			udelay(100);
767 768 769 770 771 772 773
			/*
			 * Allow other tasks to run while we wait for the
			 * AP to come online. This also gives a chance
			 * for the MTRR work(triggered by the AP coming online)
			 * to be completed in the stop machine context.
			 */
			schedule();
774 775
		}

776 777 778
		if (cpumask_test_cpu(cpu, cpu_callin_mask))
			pr_debug("CPU%d: has booted.\n", cpu);
		else {
779
			boot_error = 1;
780 781
			if (*(volatile u32 *)TRAMPOLINE_SYM(trampoline_status)
			    == 0xA5A5A5A5)
782
				/* trampoline started but...? */
783
				pr_err("CPU%d: Stuck ??\n", cpu);
784 785
			else
				/* trampoline code not run */
786
				pr_err("CPU%d: Not responding.\n", cpu);
787 788
			if (apic->inquire_remote_apic)
				apic->inquire_remote_apic(apicid);
789 790
		}
	}
791

792 793
	if (boot_error) {
		/* Try to put things back the way they were before ... */
794
		numa_remove_cpu(cpu); /* was set by numa_add_cpu */
795 796 797 798 799 800 801 802

		/* was set by do_boot_cpu() */
		cpumask_clear_cpu(cpu, cpu_callout_mask);

		/* was set by cpu_init() */
		cpumask_clear_cpu(cpu, cpu_initialized_mask);

		set_cpu_present(cpu, false);
803 804 805 806
		per_cpu(x86_cpu_to_apicid, cpu) = BAD_APICID;
	}

	/* mark "stuck" area as not stuck */
807
	*(volatile u32 *)TRAMPOLINE_SYM(trampoline_status) = 0;
808

809 810 811 812 813 814
	if (get_uv_system_type() != UV_NON_UNIQUE_APIC) {
		/*
		 * Cleanup possible dangling ends...
		 */
		smpboot_restore_warm_reset_vector();
	}
815

816
	destroy_work_on_stack(&c_idle.work);
817 818 819 820 821
	return boot_error;
}

int __cpuinit native_cpu_up(unsigned int cpu)
{
822
	int apicid = apic->cpu_present_to_apicid(cpu);
823 824 825 826 827
	unsigned long flags;
	int err;

	WARN_ON(irqs_disabled());

828
	pr_debug("++++++++++++++++++++=_---CPU UP  %u\n", cpu);
829 830

	if (apicid == BAD_APICID || apicid == boot_cpu_physical_apicid ||
831 832
	    !physid_isset(apicid, phys_cpu_present_map) ||
	    (!x2apic_mode && apicid >= 255)) {
833 834 835 836 837 838 839
		printk(KERN_ERR "%s: bad cpu %d\n", __func__, cpu);
		return -EINVAL;
	}

	/*
	 * Already booted CPU?
	 */
840
	if (cpumask_test_cpu(cpu, cpu_callin_mask)) {
841
		pr_debug("do_boot_cpu %d Already started\n", cpu);
842 843 844 845 846 847 848 849 850 851 852 853
		return -ENOSYS;
	}

	/*
	 * Save current MTRR state in case it was changed since early boot
	 * (e.g. by the ACPI SMI) to initialize new CPUs with MTRRs in sync:
	 */
	mtrr_save_state();

	per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;

	err = do_boot_cpu(apicid, cpu);
854
	if (err) {
855
		pr_debug("do_boot_cpu failed %d\n", err);
856
		return -EIO;
857 858 859 860 861 862 863 864 865 866
	}

	/*
	 * Check TSC synchronization with the AP (keep irqs disabled
	 * while doing so):
	 */
	local_irq_save(flags);
	check_tsc_sync_source(cpu);
	local_irq_restore(flags);

867
	while (!cpu_online(cpu)) {
868 869 870 871 872 873 874
		cpu_relax();
		touch_nmi_watchdog();
	}

	return 0;
}

875 876 877 878 879 880 881 882
/**
 * arch_disable_smp_support() - disables SMP support for x86 at runtime
 */
void arch_disable_smp_support(void)
{
	disable_ioapic_support();
}

883 884 885 886 887 888 889
/*
 * Fall back to non SMP mode after errors.
 *
 * RED-PEN audit/test this more. I bet there is more state messed up here.
 */
static __init void disable_smp(void)
{
890 891
	init_cpu_present(cpumask_of(0));
	init_cpu_possible(cpumask_of(0));
892
	smpboot_clear_io_apic_irqs();
893

894
	if (smp_found_config)
895
		physid_set_mask_of_physid(boot_cpu_physical_apicid, &phys_cpu_present_map);
896
	else
897
		physid_set_mask_of_physid(0, &phys_cpu_present_map);
898 899
	cpumask_set_cpu(0, cpu_sibling_mask(0));
	cpumask_set_cpu(0, cpu_core_mask(0));
900 901 902 903 904 905 906
}

/*
 * Various sanity checks.
 */
static int __init smp_sanity_check(unsigned max_cpus)
{
J
Jack Steiner 已提交
907
	preempt_disable();
908

909
#if !defined(CONFIG_X86_BIGSMP) && defined(CONFIG_X86_32)
910 911 912 913 914 915
	if (def_to_bigsmp && nr_cpu_ids > 8) {
		unsigned int cpu;
		unsigned nr;

		printk(KERN_WARNING
		       "More than 8 CPUs detected - skipping them.\n"
916
		       "Use CONFIG_X86_BIGSMP.\n");
917 918 919 920

		nr = 0;
		for_each_present_cpu(cpu) {
			if (nr >= 8)
921
				set_cpu_present(cpu, false);
922 923 924 925 926 927
			nr++;
		}

		nr = 0;
		for_each_possible_cpu(cpu) {
			if (nr >= 8)
928
				set_cpu_possible(cpu, false);
929 930 931 932 933 934 935
			nr++;
		}

		nr_cpu_ids = 8;
	}
#endif

936
	if (!physid_isset(hard_smp_processor_id(), phys_cpu_present_map)) {
M
Michael Tokarev 已提交
937 938 939 940
		printk(KERN_WARNING
			"weird, boot CPU (#%d) not listed by the BIOS.\n",
			hard_smp_processor_id());

941 942 943 944 945 946 947 948
		physid_set(hard_smp_processor_id(), phys_cpu_present_map);
	}

	/*
	 * If we couldn't find an SMP configuration at boot time,
	 * get out of here now!
	 */
	if (!smp_found_config && !acpi_lapic) {
J
Jack Steiner 已提交
949
		preempt_enable();
950 951 952 953 954 955 956 957 958 959 960 961
		printk(KERN_NOTICE "SMP motherboard not detected.\n");
		disable_smp();
		if (APIC_init_uniprocessor())
			printk(KERN_NOTICE "Local APIC not detected."
					   " Using dummy APIC emulation.\n");
		return -1;
	}

	/*
	 * Should not be necessary because the MP table should list the boot
	 * CPU too, but we do it for the sake of robustness anyway.
	 */
962
	if (!apic->check_phys_apicid_present(boot_cpu_physical_apicid)) {
963 964 965 966 967
		printk(KERN_NOTICE
			"weird, boot CPU (#%d) not listed by the BIOS.\n",
			boot_cpu_physical_apicid);
		physid_set(hard_smp_processor_id(), phys_cpu_present_map);
	}
J
Jack Steiner 已提交
968
	preempt_enable();
969 970 971 972 973 974

	/*
	 * If we couldn't find a local APIC, then get out of here now!
	 */
	if (APIC_INTEGRATED(apic_version[boot_cpu_physical_apicid]) &&
	    !cpu_has_apic) {
975 976 977 978
		if (!disable_apic) {
			pr_err("BIOS bug, local APIC #%d not detected!...\n",
				boot_cpu_physical_apicid);
			pr_err("... forcing use of dummy APIC emulation."
979
				"(tell your hw vendor)\n");
980
		}
981
		smpboot_clear_io_apic();
982
		disable_ioapic_support();
983 984 985 986 987 988 989 990 991
		return -1;
	}

	verify_local_APIC();

	/*
	 * If SMP should be disabled, then really disable it!
	 */
	if (!max_cpus) {
992
		printk(KERN_INFO "SMP mode deactivated.\n");
993
		smpboot_clear_io_apic();
994

995 996
		connect_bsp_APIC();
		setup_local_APIC();
997
		bsp_end_local_APIC_setup();
998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008
		return -1;
	}

	return 0;
}

static void __init smp_cpu_index_default(void)
{
	int i;
	struct cpuinfo_x86 *c;

1009
	for_each_possible_cpu(i) {
1010 1011
		c = &cpu_data(i);
		/* mark all to hotplug */
1012
		c->cpu_index = nr_cpu_ids;
1013 1014 1015 1016 1017 1018 1019 1020 1021
	}
}

/*
 * Prepare for SMP bootup.  The MP table or ACPI has been read
 * earlier.  Just do some sanity checking here and enable APIC mode.
 */
void __init native_smp_prepare_cpus(unsigned int max_cpus)
{
1022 1023
	unsigned int i;

1024
	preempt_disable();
1025
	smp_cpu_index_default();
1026

1027 1028 1029 1030
	/*
	 * Setup boot CPU information
	 */
	smp_store_cpu_info(0); /* Final full version of the data */
1031 1032
	cpumask_copy(cpu_callin_mask, cpumask_of(0));
	mb();
1033

1034
	current_thread_info()->cpu = 0;  /* needed? */
1035
	for_each_possible_cpu(i) {
1036 1037
		zalloc_cpumask_var(&per_cpu(cpu_sibling_map, i), GFP_KERNEL);
		zalloc_cpumask_var(&per_cpu(cpu_core_map, i), GFP_KERNEL);
1038
		zalloc_cpumask_var(&per_cpu(cpu_llc_shared_map, i), GFP_KERNEL);
1039
	}
1040 1041
	set_cpu_sibling_map(0);

1042

1043 1044 1045
	if (smp_sanity_check(max_cpus) < 0) {
		printk(KERN_INFO "SMP disabled\n");
		disable_smp();
1046
		goto out;
1047 1048
	}

1049 1050
	default_setup_apic_routing();

J
Jack Steiner 已提交
1051
	preempt_disable();
1052
	if (read_apic_id() != boot_cpu_physical_apicid) {
1053
		panic("Boot APIC ID in local APIC unexpected (%d vs %d)",
1054
		     read_apic_id(), boot_cpu_physical_apicid);
1055 1056
		/* Or can we switch back to PIC here? */
	}
J
Jack Steiner 已提交
1057
	preempt_enable();
1058 1059

	connect_bsp_APIC();
1060

1061 1062 1063 1064 1065 1066 1067 1068 1069 1070
	/*
	 * Switch from PIC to APIC mode.
	 */
	setup_local_APIC();

	/*
	 * Enable IO APIC before setting up error vector
	 */
	if (!skip_ioapic_setup && nr_ioapics)
		enable_IO_APIC();
1071

1072
	bsp_end_local_APIC_setup();
1073

1074 1075
	if (apic->setup_portio_remap)
		apic->setup_portio_remap();
1076 1077 1078 1079 1080 1081 1082 1083

	smpboot_setup_io_apic();
	/*
	 * Set up local APIC timer on boot CPU.
	 */

	printk(KERN_INFO "CPU%d: ", 0);
	print_cpu_info(&cpu_data(0));
1084
	x86_init.timers.setup_percpu_clockev();
1085 1086 1087

	if (is_uv_system())
		uv_system_init();
1088 1089

	set_mtrr_aps_delayed_init();
1090 1091
out:
	preempt_enable();
1092
}
1093

1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107
void arch_disable_nonboot_cpus_begin(void)
{
	/*
	 * Avoid the smp alternatives switch during the disable_nonboot_cpus().
	 * In the suspend path, we will be back in the SMP mode shortly anyways.
	 */
	skip_smp_alternatives = true;
}

void arch_disable_nonboot_cpus_end(void)
{
	skip_smp_alternatives = false;
}

1108 1109 1110 1111 1112 1113 1114 1115 1116 1117
void arch_enable_nonboot_cpus_begin(void)
{
	set_mtrr_aps_delayed_init();
}

void arch_enable_nonboot_cpus_end(void)
{
	mtrr_aps_init();
}

1118 1119 1120 1121 1122 1123
/*
 * Early setup to make printk work.
 */
void __init native_smp_prepare_boot_cpu(void)
{
	int me = smp_processor_id();
1124
	switch_to_new_gdt(me);
1125 1126
	/* already set me in cpu_online_mask in boot_cpu_init() */
	cpumask_set_cpu(me, cpu_callout_mask);
1127 1128 1129
	per_cpu(cpu_state, me) = CPU_ONLINE;
}

1130 1131
void __init native_smp_cpus_done(unsigned int max_cpus)
{
1132
	pr_debug("Boot done.\n");
1133

D
Don Zickus 已提交
1134
	nmi_selftest();
1135 1136 1137 1138
	impress_friends();
#ifdef CONFIG_X86_IO_APIC
	setup_ioapic_dest();
#endif
1139
	mtrr_aps_init();
1140 1141
}

1142 1143 1144 1145 1146 1147 1148 1149 1150
static int __initdata setup_possible_cpus = -1;
static int __init _setup_possible_cpus(char *str)
{
	get_option(&str, &setup_possible_cpus);
	return 0;
}
early_param("possible_cpus", _setup_possible_cpus);


1151
/*
1152
 * cpu_possible_mask should be static, it cannot change as cpu's
1153 1154 1155
 * are onlined, or offlined. The reason is per-cpu data-structures
 * are allocated by some modules at init time, and dont expect to
 * do this dynamically on cpu arrival/departure.
1156
 * cpu_present_mask on the other hand can change dynamically.
1157 1158 1159 1160 1161 1162
 * In case when cpu_hotplug is not compiled, then we resort to current
 * behaviour, which is cpu_possible == cpu_present.
 * - Ashok Raj
 *
 * Three ways to find out the number of additional hotplug CPUs:
 * - If the BIOS specified disabled CPUs in ACPI/mptables use that.
1163
 * - The user can overwrite it with possible_cpus=NUM
1164 1165 1166 1167 1168 1169
 * - Otherwise don't reserve additional CPUs.
 * We do this because additional CPUs waste a lot of memory.
 * -AK
 */
__init void prefill_possible_map(void)
{
T
Thomas Gleixner 已提交
1170
	int i, possible;
1171

1172 1173 1174 1175
	/* no processor from mptable or madt */
	if (!num_processors)
		num_processors = 1;

1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186
	i = setup_max_cpus ?: 1;
	if (setup_possible_cpus == -1) {
		possible = num_processors;
#ifdef CONFIG_HOTPLUG_CPU
		if (setup_max_cpus)
			possible += disabled_cpus;
#else
		if (possible > i)
			possible = i;
#endif
	} else
1187 1188
		possible = setup_possible_cpus;

1189 1190
	total_cpus = max_t(int, possible, num_processors + disabled_cpus);

1191 1192
	/* nr_cpu_ids could be reduced via nr_cpus= */
	if (possible > nr_cpu_ids) {
1193 1194
		printk(KERN_WARNING
			"%d Processors exceeds NR_CPUS limit of %d\n",
1195 1196
			possible, nr_cpu_ids);
		possible = nr_cpu_ids;
1197
	}
1198

1199 1200 1201 1202 1203 1204 1205 1206 1207 1208
#ifdef CONFIG_HOTPLUG_CPU
	if (!setup_max_cpus)
#endif
	if (possible > i) {
		printk(KERN_WARNING
			"%d Processors exceeds max_cpus limit of %u\n",
			possible, setup_max_cpus);
		possible = i;
	}

1209 1210 1211 1212
	printk(KERN_INFO "SMP: Allowing %d CPUs, %d hotplug CPUs\n",
		possible, max_t(int, possible - num_processors, 0));

	for (i = 0; i < possible; i++)
1213
		set_cpu_possible(i, true);
1214 1215
	for (; i < NR_CPUS; i++)
		set_cpu_possible(i, false);
1216 1217

	nr_cpu_ids = possible;
1218
}
1219

1220 1221 1222 1223 1224 1225 1226
#ifdef CONFIG_HOTPLUG_CPU

static void remove_siblinginfo(int cpu)
{
	int sibling;
	struct cpuinfo_x86 *c = &cpu_data(cpu);

1227 1228
	for_each_cpu(sibling, cpu_core_mask(cpu)) {
		cpumask_clear_cpu(cpu, cpu_core_mask(sibling));
1229 1230 1231
		/*/
		 * last thread sibling in this cpu core going down
		 */
1232
		if (cpumask_weight(cpu_sibling_mask(cpu)) == 1)
1233 1234 1235
			cpu_data(sibling).booted_cores--;
	}

1236 1237 1238 1239
	for_each_cpu(sibling, cpu_sibling_mask(cpu))
		cpumask_clear_cpu(cpu, cpu_sibling_mask(sibling));
	cpumask_clear(cpu_sibling_mask(cpu));
	cpumask_clear(cpu_core_mask(cpu));
1240 1241
	c->phys_proc_id = 0;
	c->cpu_core_id = 0;
1242
	cpumask_clear_cpu(cpu, cpu_sibling_setup_mask);
1243 1244
}

1245 1246
static void __ref remove_cpu_from_maps(int cpu)
{
1247 1248 1249
	set_cpu_online(cpu, false);
	cpumask_clear_cpu(cpu, cpu_callout_mask);
	cpumask_clear_cpu(cpu, cpu_callin_mask);
1250
	/* was set by cpu_init() */
1251
	cpumask_clear_cpu(cpu, cpu_initialized_mask);
1252
	numa_remove_cpu(cpu);
1253 1254
}

1255
void cpu_disable_common(void)
1256 1257 1258 1259 1260 1261
{
	int cpu = smp_processor_id();

	remove_siblinginfo(cpu);

	/* It's now safe to remove this processor from the online map */
1262
	lock_vector_lock();
1263
	remove_cpu_from_maps(cpu);
1264
	unlock_vector_lock();
1265
	fixup_irqs();
1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285
}

int native_cpu_disable(void)
{
	int cpu = smp_processor_id();

	/*
	 * Perhaps use cpufreq to drop frequency, but that could go
	 * into generic code.
	 *
	 * We won't take down the boot processor on i386 due to some
	 * interrupts only being able to be serviced by the BSP.
	 * Especially so if we're not using an IOAPIC	-zwane
	 */
	if (cpu == 0)
		return -EBUSY;

	clear_local_APIC();

	cpu_disable_common();
1286 1287 1288
	return 0;
}

1289
void native_cpu_die(unsigned int cpu)
1290 1291 1292 1293 1294 1295 1296
{
	/* We don't do anything here: idle task is faking death itself. */
	unsigned int i;

	for (i = 0; i < 10; i++) {
		/* They ack this in play_dead by setting CPU_DEAD */
		if (per_cpu(cpu_state, cpu) == CPU_DEAD) {
1297 1298 1299
			if (system_state == SYSTEM_RUNNING)
				pr_info("CPU %u is now offline\n", cpu);

1300 1301 1302 1303 1304 1305
			if (1 == num_online_cpus())
				alternatives_smp_switch(0);
			return;
		}
		msleep(100);
	}
1306
	pr_err("CPU %u didn't die...\n", cpu);
1307
}
1308 1309 1310 1311 1312

void play_dead_common(void)
{
	idle_task_exit();
	reset_lazy_tlbstate();
1313
	amd_e400_remove_cpu(raw_smp_processor_id());
1314 1315 1316

	mb();
	/* Ack it */
T
Tejun Heo 已提交
1317
	__this_cpu_write(cpu_state, CPU_DEAD);
1318 1319 1320 1321 1322 1323 1324

	/*
	 * With physical CPU hotplug, we should halt the cpu
	 */
	local_irq_disable();
}

1325 1326 1327 1328 1329 1330 1331 1332 1333 1334
/*
 * We need to flush the caches before going to sleep, lest we have
 * dirty data in our caches when we come back up.
 */
static inline void mwait_play_dead(void)
{
	unsigned int eax, ebx, ecx, edx;
	unsigned int highest_cstate = 0;
	unsigned int highest_subcstate = 0;
	int i;
1335
	void *mwait_ptr;
1336
	struct cpuinfo_x86 *c = __this_cpu_ptr(&cpu_info);
1337

1338
	if (!(this_cpu_has(X86_FEATURE_MWAIT) && mwait_usable(c)))
1339
		return;
1340
	if (!this_cpu_has(X86_FEATURE_CLFLSH))
1341
		return;
1342
	if (__this_cpu_read(cpu_info.cpuid_level) < CPUID_MWAIT_LEAF)
1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366
		return;

	eax = CPUID_MWAIT_LEAF;
	ecx = 0;
	native_cpuid(&eax, &ebx, &ecx, &edx);

	/*
	 * eax will be 0 if EDX enumeration is not valid.
	 * Initialized below to cstate, sub_cstate value when EDX is valid.
	 */
	if (!(ecx & CPUID5_ECX_EXTENSIONS_SUPPORTED)) {
		eax = 0;
	} else {
		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;
			}
		}
		eax = (highest_cstate << MWAIT_SUBSTATE_SIZE) |
			(highest_subcstate - 1);
	}

1367 1368 1369 1370 1371 1372 1373
	/*
	 * This should be a memory location in a cache line which is
	 * unlikely to be touched by other processors.  The actual
	 * content is immaterial as it is not actually modified in any way.
	 */
	mwait_ptr = &current_thread_info()->flags;

1374 1375
	wbinvd();

1376
	while (1) {
1377 1378 1379 1380 1381 1382 1383 1384 1385
		/*
		 * The CLFLUSH is a workaround for erratum AAI65 for
		 * the Xeon 7400 series.  It's not clear it is actually
		 * needed, but it should be harmless in either case.
		 * The WBINVD is insufficient due to the spurious-wakeup
		 * case where we return around the loop.
		 */
		clflush(mwait_ptr);
		__monitor(mwait_ptr, 0, 0);
1386 1387 1388 1389 1390 1391 1392
		mb();
		__mwait(eax, 0);
	}
}

static inline void hlt_play_dead(void)
{
1393
	if (__this_cpu_read(cpu_info.x86) >= 4)
1394 1395
		wbinvd();

1396 1397 1398 1399 1400
	while (1) {
		native_halt();
	}
}

1401 1402 1403
void native_play_dead(void)
{
	play_dead_common();
1404
	tboot_shutdown(TB_SHUTDOWN_WFS);
1405 1406 1407

	mwait_play_dead();	/* Only returns on failure */
	hlt_play_dead();
1408 1409
}

1410
#else /* ... !CONFIG_HOTPLUG_CPU */
1411
int native_cpu_disable(void)
1412 1413 1414 1415
{
	return -ENOSYS;
}

1416
void native_cpu_die(unsigned int cpu)
1417 1418 1419 1420
{
	/* We said "no" in __cpu_disable */
	BUG();
}
1421 1422 1423 1424 1425 1426

void native_play_dead(void)
{
	BUG();
}

1427
#endif