smp.c 19.1 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
/*
 * SMP support for ppc.
 *
 * Written by Cort Dougan (cort@cs.nmt.edu) borrowing a great
 * deal of code from the sparc and intel versions.
 *
 * Copyright (C) 1999 Cort Dougan <cort@cs.nmt.edu>
 *
 * PowerPC-64 Support added by Dave Engebretsen, Peter Bergner, and
 * Mike Corrigan {engebret|bergner|mikec}@us.ibm.com
 *
 *      This program is free software; you can redistribute it and/or
 *      modify it under the terms of the GNU General Public License
 *      as published by the Free Software Foundation; either version
 *      2 of the License, or (at your option) any later version.
 */

#undef DEBUG

#include <linux/kernel.h>
21
#include <linux/export.h>
22
#include <linux/sched/mm.h>
23
#include <linux/sched/topology.h>
L
Linus Torvalds 已提交
24 25 26 27 28 29 30
#include <linux/smp.h>
#include <linux/interrupt.h>
#include <linux/delay.h>
#include <linux/init.h>
#include <linux/spinlock.h>
#include <linux/cache.h>
#include <linux/err.h>
31
#include <linux/device.h>
L
Linus Torvalds 已提交
32 33
#include <linux/cpu.h>
#include <linux/notifier.h>
34
#include <linux/topology.h>
35
#include <linux/profile.h>
L
Linus Torvalds 已提交
36 37

#include <asm/ptrace.h>
A
Arun Sharma 已提交
38
#include <linux/atomic.h>
L
Linus Torvalds 已提交
39
#include <asm/irq.h>
40
#include <asm/hw_irq.h>
41
#include <asm/kvm_ppc.h>
42
#include <asm/dbell.h>
L
Linus Torvalds 已提交
43 44 45 46 47 48
#include <asm/page.h>
#include <asm/pgtable.h>
#include <asm/prom.h>
#include <asm/smp.h>
#include <asm/time.h>
#include <asm/machdep.h>
49
#include <asm/cputhreads.h>
L
Linus Torvalds 已提交
50
#include <asm/cputable.h>
51
#include <asm/mpic.h>
52
#include <asm/vdso_datapage.h>
P
Paul Mackerras 已提交
53 54 55
#ifdef CONFIG_PPC64
#include <asm/paca.h>
#endif
56
#include <asm/vdso.h>
57
#include <asm/debug.h>
58
#include <asm/kexec.h>
59
#include <asm/asm-prototypes.h>
60
#include <asm/cpu_has_feature.h>
P
Paul Mackerras 已提交
61

L
Linus Torvalds 已提交
62
#ifdef DEBUG
63
#include <asm/udbg.h>
L
Linus Torvalds 已提交
64 65 66 67 68
#define DBG(fmt...) udbg_printf(fmt)
#else
#define DBG(fmt...)
#endif

69
#ifdef CONFIG_HOTPLUG_CPU
70 71
/* State of each CPU during hotplug phases */
static DEFINE_PER_CPU(int, cpu_state) = { 0 };
72 73
#endif

74 75
struct thread_info *secondary_ti;

76 77
DEFINE_PER_CPU(cpumask_var_t, cpu_sibling_map);
DEFINE_PER_CPU(cpumask_var_t, cpu_core_map);
L
Linus Torvalds 已提交
78

79
EXPORT_PER_CPU_SYMBOL(cpu_sibling_map);
80
EXPORT_PER_CPU_SYMBOL(cpu_core_map);
L
Linus Torvalds 已提交
81

P
Paul Mackerras 已提交
82
/* SMP operations for this machine */
L
Linus Torvalds 已提交
83 84
struct smp_ops_t *smp_ops;

85 86
/* Can't be static due to PowerMac hackery */
volatile unsigned int cpu_callin_map[NR_CPUS];
L
Linus Torvalds 已提交
87 88 89

int smt_enabled_at_boot = 1;

90 91
static void (*crash_ipi_function_ptr)(struct pt_regs *) = NULL;

92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113
/*
 * Returns 1 if the specified cpu should be brought up during boot.
 * Used to inhibit booting threads if they've been disabled or
 * limited on the command line
 */
int smp_generic_cpu_bootable(unsigned int nr)
{
	/* Special case - we inhibit secondary thread startup
	 * during boot if the user requests it.
	 */
	if (system_state == SYSTEM_BOOTING && cpu_has_feature(CPU_FTR_SMT)) {
		if (!smt_enabled_at_boot && cpu_thread_in_core(nr) != 0)
			return 0;
		if (smt_enabled_at_boot
		    && cpu_thread_in_core(nr) >= smt_enabled_at_boot)
			return 0;
	}

	return 1;
}


P
Paul Mackerras 已提交
114
#ifdef CONFIG_PPC64
115
int smp_generic_kick_cpu(int nr)
L
Linus Torvalds 已提交
116 117 118 119 120 121 122 123
{
	BUG_ON(nr < 0 || nr >= NR_CPUS);

	/*
	 * The processor is currently spinning, waiting for the
	 * cpu_start field to become non-zero After we set cpu_start,
	 * the processor will continue on to secondary_start
	 */
124 125 126 127 128 129 130 131 132 133 134
	if (!paca[nr].cpu_start) {
		paca[nr].cpu_start = 1;
		smp_mb();
		return 0;
	}

#ifdef CONFIG_HOTPLUG_CPU
	/*
	 * Ok it's not there, so it might be soft-unplugged, let's
	 * try to bring it back
	 */
135
	generic_set_cpu_up(nr);
136 137 138
	smp_wmb();
	smp_send_reschedule(nr);
#endif /* CONFIG_HOTPLUG_CPU */
139 140

	return 0;
L
Linus Torvalds 已提交
141
}
142
#endif /* CONFIG_PPC64 */
L
Linus Torvalds 已提交
143

144 145 146 147 148 149 150 151
static irqreturn_t call_function_action(int irq, void *data)
{
	generic_smp_call_function_interrupt();
	return IRQ_HANDLED;
}

static irqreturn_t reschedule_action(int irq, void *data)
{
152
	scheduler_ipi();
153 154 155
	return IRQ_HANDLED;
}

156
static irqreturn_t tick_broadcast_ipi_action(int irq, void *data)
157
{
158
	tick_broadcast_ipi_handler();
159 160 161
	return IRQ_HANDLED;
}

162
static irqreturn_t debug_ipi_action(int irq, void *data)
163
{
164 165 166 167 168 169 170 171 172
	if (crash_ipi_function_ptr) {
		crash_ipi_function_ptr(get_irq_regs());
		return IRQ_HANDLED;
	}

#ifdef CONFIG_DEBUGGER
	debugger_ipi(get_irq_regs());
#endif /* CONFIG_DEBUGGER */

173 174 175 176 177 178
	return IRQ_HANDLED;
}

static irq_handler_t smp_ipi_action[] = {
	[PPC_MSG_CALL_FUNCTION] =  call_function_action,
	[PPC_MSG_RESCHEDULE] = reschedule_action,
179
	[PPC_MSG_TICK_BROADCAST] = tick_broadcast_ipi_action,
180 181 182 183 184 185
	[PPC_MSG_DEBUGGER_BREAK] = debug_ipi_action,
};

const char *smp_ipi_name[] = {
	[PPC_MSG_CALL_FUNCTION] =  "ipi call function",
	[PPC_MSG_RESCHEDULE] = "ipi reschedule",
186
	[PPC_MSG_TICK_BROADCAST] = "ipi tick-broadcast",
187 188 189 190 191 192 193 194 195 196 197
	[PPC_MSG_DEBUGGER_BREAK] = "ipi debugger",
};

/* optional function to request ipi, for controllers with >= 4 ipis */
int smp_request_message_ipi(int virq, int msg)
{
	int err;

	if (msg < 0 || msg > PPC_MSG_DEBUGGER_BREAK) {
		return -EINVAL;
	}
198
#if !defined(CONFIG_DEBUGGER) && !defined(CONFIG_KEXEC_CORE)
199 200 201 202
	if (msg == PPC_MSG_DEBUGGER_BREAK) {
		return 1;
	}
#endif
203
	err = request_irq(virq, smp_ipi_action[msg],
204
			  IRQF_PERCPU | IRQF_NO_THREAD | IRQF_NO_SUSPEND,
205
			  smp_ipi_name[msg], NULL);
206 207 208 209 210 211
	WARN(err < 0, "unable to request_irq %d for %s (rc %d)\n",
		virq, smp_ipi_name[msg], err);

	return err;
}

212
#ifdef CONFIG_PPC_SMP_MUXED_IPI
213
struct cpu_messages {
214
	long messages;			/* current messages */
215 216 217
};
static DEFINE_PER_CPU_SHARED_ALIGNED(struct cpu_messages, ipi_message);

218
void smp_muxed_ipi_set_message(int cpu, int msg)
219 220
{
	struct cpu_messages *info = &per_cpu(ipi_message, cpu);
221
	char *message = (char *)&info->messages;
222

223 224 225 226
	/*
	 * Order previous accesses before accesses in the IPI handler.
	 */
	smp_mb();
227
	message[msg] = 1;
228 229 230 231 232
}

void smp_muxed_ipi_message_pass(int cpu, int msg)
{
	smp_muxed_ipi_set_message(cpu, msg);
233

234 235 236 237
	/*
	 * cause_ipi functions are required to include a full barrier
	 * before doing whatever causes the IPI.
	 */
238
	smp_ops->cause_ipi(cpu);
239 240
}

241
#ifdef __BIG_ENDIAN__
242
#define IPI_MESSAGE(A) (1uL << ((BITS_PER_LONG - 8) - 8 * (A)))
243
#else
244
#define IPI_MESSAGE(A) (1uL << (8 * (A)))
245 246
#endif

247 248
irqreturn_t smp_ipi_demux(void)
{
249
	struct cpu_messages *info;
250
	unsigned long all;
251 252

	mb();	/* order any irq clear */
253

254
	info = this_cpu_ptr(&ipi_message);
255
	do {
256
		all = xchg(&info->messages, 0);
257 258 259 260 261 262 263 264 265 266 267
#if defined(CONFIG_KVM_XICS) && defined(CONFIG_KVM_BOOK3S_HV_POSSIBLE)
		/*
		 * Must check for PPC_MSG_RM_HOST_ACTION messages
		 * before PPC_MSG_CALL_FUNCTION messages because when
		 * a VM is destroyed, we call kick_all_cpus_sync()
		 * to ensure that any pending PPC_MSG_RM_HOST_ACTION
		 * messages have completed before we free any VCPUs.
		 */
		if (all & IPI_MESSAGE(PPC_MSG_RM_HOST_ACTION))
			kvmppc_xics_ipi_action();
#endif
268
		if (all & IPI_MESSAGE(PPC_MSG_CALL_FUNCTION))
269
			generic_smp_call_function_interrupt();
270
		if (all & IPI_MESSAGE(PPC_MSG_RESCHEDULE))
271
			scheduler_ipi();
272 273
		if (all & IPI_MESSAGE(PPC_MSG_TICK_BROADCAST))
			tick_broadcast_ipi_handler();
274
		if (all & IPI_MESSAGE(PPC_MSG_DEBUGGER_BREAK))
275
			debug_ipi_action(0, NULL);
276 277
	} while (info->messages);

278 279
	return IRQ_HANDLED;
}
280
#endif /* CONFIG_PPC_SMP_MUXED_IPI */
281

282 283 284 285 286 287 288 289 290 291
static inline void do_message_pass(int cpu, int msg)
{
	if (smp_ops->message_pass)
		smp_ops->message_pass(cpu, msg);
#ifdef CONFIG_PPC_SMP_MUXED_IPI
	else
		smp_muxed_ipi_message_pass(cpu, msg);
#endif
}

L
Linus Torvalds 已提交
292 293
void smp_send_reschedule(int cpu)
{
294
	if (likely(smp_ops))
295
		do_message_pass(cpu, PPC_MSG_RESCHEDULE);
L
Linus Torvalds 已提交
296
}
297
EXPORT_SYMBOL_GPL(smp_send_reschedule);
L
Linus Torvalds 已提交
298

299 300
void arch_send_call_function_single_ipi(int cpu)
{
301
	do_message_pass(cpu, PPC_MSG_CALL_FUNCTION);
302 303
}

304
void arch_send_call_function_ipi_mask(const struct cpumask *mask)
305 306 307
{
	unsigned int cpu;

308
	for_each_cpu(cpu, mask)
309
		do_message_pass(cpu, PPC_MSG_CALL_FUNCTION);
310 311
}

312 313 314 315 316 317 318 319 320 321
#ifdef CONFIG_GENERIC_CLOCKEVENTS_BROADCAST
void tick_broadcast(const struct cpumask *mask)
{
	unsigned int cpu;

	for_each_cpu(cpu, mask)
		do_message_pass(cpu, PPC_MSG_TICK_BROADCAST);
}
#endif

322
#if defined(CONFIG_DEBUGGER) || defined(CONFIG_KEXEC_CORE)
323
void smp_send_debugger_break(void)
L
Linus Torvalds 已提交
324
{
325 326 327 328 329 330 331 332
	int cpu;
	int me = raw_smp_processor_id();

	if (unlikely(!smp_ops))
		return;

	for_each_online_cpu(cpu)
		if (cpu != me)
333
			do_message_pass(cpu, PPC_MSG_DEBUGGER_BREAK);
L
Linus Torvalds 已提交
334 335 336
}
#endif

337
#ifdef CONFIG_KEXEC_CORE
338 339 340
void crash_send_ipi(void (*crash_ipi_callback)(struct pt_regs *))
{
	crash_ipi_function_ptr = crash_ipi_callback;
341
	if (crash_ipi_callback) {
342
		mb();
343
		smp_send_debugger_break();
344 345 346 347
	}
}
#endif

L
Linus Torvalds 已提交
348 349
static void stop_this_cpu(void *dummy)
{
350 351 352
	/* Remove this CPU */
	set_cpu_online(smp_processor_id(), false);

L
Linus Torvalds 已提交
353 354 355 356 357
	local_irq_disable();
	while (1)
		;
}

358 359
void smp_send_stop(void)
{
360
	smp_call_function(stop_this_cpu, NULL, 0);
L
Linus Torvalds 已提交
361 362 363 364
}

struct thread_info *current_set[NR_CPUS];

365
static void smp_store_cpu_info(int id)
L
Linus Torvalds 已提交
366
{
367
	per_cpu(cpu_pvr, id) = mfspr(SPRN_PVR);
368 369 370 371
#ifdef CONFIG_PPC_FSL_BOOK3E
	per_cpu(next_tlbcam_idx, id)
		= (mfspr(SPRN_TLB1CFG) & TLBnCFG_N_ENTRY) - 1;
#endif
L
Linus Torvalds 已提交
372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389
}

void __init smp_prepare_cpus(unsigned int max_cpus)
{
	unsigned int cpu;

	DBG("smp_prepare_cpus\n");

	/* 
	 * setup_cpu may need to be called on the boot cpu. We havent
	 * spun any cpus up but lets be paranoid.
	 */
	BUG_ON(boot_cpuid != smp_processor_id());

	/* Fixup boot cpu */
	smp_store_cpu_info(boot_cpuid);
	cpu_callin_map[boot_cpuid] = 1;

390 391 392 393 394
	for_each_possible_cpu(cpu) {
		zalloc_cpumask_var_node(&per_cpu(cpu_sibling_map, cpu),
					GFP_KERNEL, cpu_to_node(cpu));
		zalloc_cpumask_var_node(&per_cpu(cpu_core_map, cpu),
					GFP_KERNEL, cpu_to_node(cpu));
395 396 397
		/*
		 * numa_node_id() works after this.
		 */
398 399 400 401 402
		if (cpu_present(cpu)) {
			set_cpu_numa_node(cpu, numa_cpu_lookup_table[cpu]);
			set_cpu_numa_mem(cpu,
				local_memory_node(numa_cpu_lookup_table[cpu]));
		}
403 404 405 406 407
	}

	cpumask_set_cpu(boot_cpuid, cpu_sibling_mask(boot_cpuid));
	cpumask_set_cpu(boot_cpuid, cpu_core_mask(boot_cpuid));

408 409
	if (smp_ops && smp_ops->probe)
		smp_ops->probe();
L
Linus Torvalds 已提交
410 411
}

412
void smp_prepare_boot_cpu(void)
L
Linus Torvalds 已提交
413 414
{
	BUG_ON(smp_processor_id() != boot_cpuid);
P
Paul Mackerras 已提交
415
#ifdef CONFIG_PPC64
L
Linus Torvalds 已提交
416
	paca[boot_cpuid].__current = current;
P
Paul Mackerras 已提交
417
#endif
418
	set_numa_node(numa_cpu_lookup_table[boot_cpuid]);
A
Al Viro 已提交
419
	current_set[boot_cpuid] = task_thread_info(current);
L
Linus Torvalds 已提交
420 421 422 423 424 425 426 427 428 429 430
}

#ifdef CONFIG_HOTPLUG_CPU

int generic_cpu_disable(void)
{
	unsigned int cpu = smp_processor_id();

	if (cpu == boot_cpuid)
		return -EBUSY;

431
	set_cpu_online(cpu, false);
432
#ifdef CONFIG_PPC64
433
	vdso_data->processorCount--;
434
#endif
435 436 437 438 439 440 441 442
	/* Update affinity of all IRQs previously aimed at this CPU */
	irq_migrate_all_off_this_cpu();

	/* Give the CPU time to drain in-flight ones */
	local_irq_enable();
	mdelay(1);
	local_irq_disable();

L
Linus Torvalds 已提交
443 444 445 446 447 448 449 450
	return 0;
}

void generic_cpu_die(unsigned int cpu)
{
	int i;

	for (i = 0; i < 100; i++) {
451
		smp_rmb();
452
		if (is_cpu_dead(cpu))
L
Linus Torvalds 已提交
453 454 455 456 457 458
			return;
		msleep(100);
	}
	printk(KERN_ERR "CPU%d didn't die...\n", cpu);
}

459 460 461 462
void generic_set_cpu_dead(unsigned int cpu)
{
	per_cpu(cpu_state, cpu) = CPU_DEAD;
}
463

464 465 466 467 468 469 470 471 472 473
/*
 * The cpu_state should be set to CPU_UP_PREPARE in kick_cpu(), otherwise
 * the cpu_state is always CPU_DEAD after calling generic_set_cpu_dead(),
 * which makes the delay in generic_cpu_die() not happen.
 */
void generic_set_cpu_up(unsigned int cpu)
{
	per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;
}

474 475 476 477
int generic_check_cpu_restart(unsigned int cpu)
{
	return per_cpu(cpu_state, cpu) == CPU_UP_PREPARE;
}
478

479 480 481 482 483
int is_cpu_dead(unsigned int cpu)
{
	return per_cpu(cpu_state, cpu) == CPU_DEAD;
}

484
static bool secondaries_inhibited(void)
485
{
486
	return kvm_hv_mode_active();
487 488 489 490 491 492
}

#else /* HOTPLUG_CPU */

#define secondaries_inhibited()		0

L
Linus Torvalds 已提交
493 494
#endif

495
static void cpu_idle_thread_init(unsigned int cpu, struct task_struct *idle)
496
{
497
	struct thread_info *ti = task_thread_info(idle);
498 499

#ifdef CONFIG_PPC64
500
	paca[cpu].__current = idle;
501 502 503
	paca[cpu].kstack = (unsigned long)ti + THREAD_SIZE - STACK_FRAME_OVERHEAD;
#endif
	ti->cpu = cpu;
504
	secondary_ti = current_set[cpu] = ti;
505 506
}

507
int __cpu_up(unsigned int cpu, struct task_struct *tidle)
L
Linus Torvalds 已提交
508
{
509
	int rc, c;
L
Linus Torvalds 已提交
510

511 512 513 514
	/*
	 * Don't allow secondary threads to come online if inhibited
	 */
	if (threads_per_core > 1 && secondaries_inhibited() &&
515
	    cpu_thread_in_subcore(cpu))
516 517
		return -EBUSY;

518 519
	if (smp_ops == NULL ||
	    (smp_ops->cpu_bootable && !smp_ops->cpu_bootable(cpu)))
L
Linus Torvalds 已提交
520 521
		return -EINVAL;

522
	cpu_idle_thread_init(cpu, tidle);
523

524 525 526 527 528 529 530 531 532 533
	/*
	 * The platform might need to allocate resources prior to bringing
	 * up the CPU
	 */
	if (smp_ops->prepare_cpu) {
		rc = smp_ops->prepare_cpu(cpu);
		if (rc)
			return rc;
	}

L
Linus Torvalds 已提交
534 535 536 537 538 539 540 541 542
	/* Make sure callin-map entry is 0 (can be leftover a CPU
	 * hotplug
	 */
	cpu_callin_map[cpu] = 0;

	/* The information for processor bringup must
	 * be written out to main store before we release
	 * the processor.
	 */
543
	smp_mb();
L
Linus Torvalds 已提交
544 545 546

	/* wake up cpus */
	DBG("smp: kicking cpu %d\n", cpu);
547 548 549 550 551
	rc = smp_ops->kick_cpu(cpu);
	if (rc) {
		pr_err("smp: failed starting cpu %d (rc %d)\n", cpu, rc);
		return rc;
	}
L
Linus Torvalds 已提交
552 553 554 555 556 557 558

	/*
	 * wait to see if the cpu made a callin (is actually up).
	 * use this value that I found through experimentation.
	 * -- Cort
	 */
	if (system_state < SYSTEM_RUNNING)
559
		for (c = 50000; c && !cpu_callin_map[cpu]; c--)
L
Linus Torvalds 已提交
560 561 562 563 564 565 566
			udelay(100);
#ifdef CONFIG_HOTPLUG_CPU
	else
		/*
		 * CPUs can take much longer to come up in the
		 * hotplug case.  Wait five seconds.
		 */
567 568
		for (c = 5000; c && !cpu_callin_map[cpu]; c--)
			msleep(1);
L
Linus Torvalds 已提交
569 570 571
#endif

	if (!cpu_callin_map[cpu]) {
572
		printk(KERN_ERR "Processor %u is stuck.\n", cpu);
L
Linus Torvalds 已提交
573 574 575
		return -ENOENT;
	}

576
	DBG("Processor %u found.\n", cpu);
L
Linus Torvalds 已提交
577 578 579 580

	if (smp_ops->give_timebase)
		smp_ops->give_timebase();

581
	/* Wait until cpu puts itself in the online & active maps */
582
	while (!cpu_online(cpu))
L
Linus Torvalds 已提交
583 584 585 586 587
		cpu_relax();

	return 0;
}

588 589 590 591 592 593
/* Return the value of the reg property corresponding to the given
 * logical cpu.
 */
int cpu_to_core_id(int cpu)
{
	struct device_node *np;
594
	const __be32 *reg;
595 596 597 598 599 600 601 602 603 604
	int id = -1;

	np = of_get_cpu_node(cpu, NULL);
	if (!np)
		goto out;

	reg = of_get_property(np, "reg", NULL);
	if (!reg)
		goto out;

605
	id = be32_to_cpup(reg);
606 607 608 609
out:
	of_node_put(np);
	return id;
}
610
EXPORT_SYMBOL_GPL(cpu_to_core_id);
611

612 613 614 615 616 617 618 619 620 621 622 623 624
/* Helper routines for cpu to core mapping */
int cpu_core_index_of_thread(int cpu)
{
	return cpu >> threads_shift;
}
EXPORT_SYMBOL_GPL(cpu_core_index_of_thread);

int cpu_first_thread_of_core(int core)
{
	return core << threads_shift;
}
EXPORT_SYMBOL_GPL(cpu_first_thread_of_core);

625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651
static void traverse_siblings_chip_id(int cpu, bool add, int chipid)
{
	const struct cpumask *mask;
	struct device_node *np;
	int i, plen;
	const __be32 *prop;

	mask = add ? cpu_online_mask : cpu_present_mask;
	for_each_cpu(i, mask) {
		np = of_get_cpu_node(i, NULL);
		if (!np)
			continue;
		prop = of_get_property(np, "ibm,chip-id", &plen);
		if (prop && plen == sizeof(int) &&
		    of_read_number(prop, 1) == chipid) {
			if (add) {
				cpumask_set_cpu(cpu, cpu_core_mask(i));
				cpumask_set_cpu(i, cpu_core_mask(cpu));
			} else {
				cpumask_clear_cpu(cpu, cpu_core_mask(i));
				cpumask_clear_cpu(i, cpu_core_mask(cpu));
			}
		}
		of_node_put(np);
	}
}

652
/* Must be called when no change can occur to cpu_present_mask,
653 654 655 656 657
 * i.e. during cpu online or offline.
 */
static struct device_node *cpu_to_l2cache(int cpu)
{
	struct device_node *np;
658
	struct device_node *cache;
659 660 661 662 663 664 665 666

	if (!cpu_present(cpu))
		return NULL;

	np = of_get_cpu_node(cpu, NULL);
	if (np == NULL)
		return NULL;

667 668
	cache = of_find_next_cache_node(np);

669 670
	of_node_put(np);

671
	return cache;
672
}
L
Linus Torvalds 已提交
673

674 675
static void traverse_core_siblings(int cpu, bool add)
{
676
	struct device_node *l2_cache, *np;
677
	const struct cpumask *mask;
678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693
	int i, chip, plen;
	const __be32 *prop;

	/* First see if we have ibm,chip-id properties in cpu nodes */
	np = of_get_cpu_node(cpu, NULL);
	if (np) {
		chip = -1;
		prop = of_get_property(np, "ibm,chip-id", &plen);
		if (prop && plen == sizeof(int))
			chip = of_read_number(prop, 1);
		of_node_put(np);
		if (chip >= 0) {
			traverse_siblings_chip_id(cpu, add, chip);
			return;
		}
	}
694 695 696 697

	l2_cache = cpu_to_l2cache(cpu);
	mask = add ? cpu_online_mask : cpu_present_mask;
	for_each_cpu(i, mask) {
698
		np = cpu_to_l2cache(i);
699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714
		if (!np)
			continue;
		if (np == l2_cache) {
			if (add) {
				cpumask_set_cpu(cpu, cpu_core_mask(i));
				cpumask_set_cpu(i, cpu_core_mask(cpu));
			} else {
				cpumask_clear_cpu(cpu, cpu_core_mask(i));
				cpumask_clear_cpu(i, cpu_core_mask(cpu));
			}
		}
		of_node_put(np);
	}
	of_node_put(l2_cache);
}

L
Linus Torvalds 已提交
715
/* Activate a secondary processor. */
716
void start_secondary(void *unused)
L
Linus Torvalds 已提交
717 718
{
	unsigned int cpu = smp_processor_id();
719
	int i, base;
L
Linus Torvalds 已提交
720

V
Vegard Nossum 已提交
721
	mmgrab(&init_mm);
L
Linus Torvalds 已提交
722 723 724
	current->active_mm = &init_mm;

	smp_store_cpu_info(cpu);
P
Paul Mackerras 已提交
725
	set_dec(tb_ticks_per_jiffy);
A
Andrew Morton 已提交
726
	preempt_disable();
727
	cpu_callin_map[cpu] = 1;
L
Linus Torvalds 已提交
728

729 730
	if (smp_ops->setup_cpu)
		smp_ops->setup_cpu(cpu);
L
Linus Torvalds 已提交
731 732 733
	if (smp_ops->take_timebase)
		smp_ops->take_timebase();

734 735
	secondary_cpu_time_init();

736 737 738
#ifdef CONFIG_PPC64
	if (system_state == SYSTEM_RUNNING)
		vdso_data->processorCount++;
739 740

	vdso_getcpu_init();
741
#endif
742
	/* Update sibling maps */
743
	base = cpu_first_thread_sibling(cpu);
744
	for (i = 0; i < threads_per_core; i++) {
745
		if (cpu_is_offline(base + i) && (cpu != base + i))
746
			continue;
747 748
		cpumask_set_cpu(cpu, cpu_sibling_mask(base + i));
		cpumask_set_cpu(base + i, cpu_sibling_mask(cpu));
749 750 751 752 753

		/* cpu_core_map should be a superset of
		 * cpu_sibling_map even if we don't have cache
		 * information, so update the former here, too.
		 */
754 755
		cpumask_set_cpu(cpu, cpu_core_mask(base + i));
		cpumask_set_cpu(base + i, cpu_core_mask(cpu));
756
	}
757
	traverse_core_siblings(cpu, true);
L
Linus Torvalds 已提交
758

759 760 761
	set_numa_node(numa_cpu_lookup_table[cpu]);
	set_numa_mem(local_memory_node(numa_cpu_lookup_table[cpu]));

762 763 764 765
	smp_wmb();
	notify_cpu_starting(cpu);
	set_cpu_online(cpu, true);

L
Linus Torvalds 已提交
766 767
	local_irq_enable();

768
	cpu_startup_entry(CPUHP_AP_ONLINE_IDLE);
769 770

	BUG();
L
Linus Torvalds 已提交
771 772 773 774 775 776 777
}

int setup_profiling_timer(unsigned int multiplier)
{
	return 0;
}

778 779
#ifdef CONFIG_SCHED_SMT
/* cpumask of CPUs with asymetric SMT dependancy */
G
Guenter Roeck 已提交
780
static int powerpc_smt_flags(void)
781
{
782
	int flags = SD_SHARE_CPUCAPACITY | SD_SHARE_PKG_RESOURCES;
783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799

	if (cpu_has_feature(CPU_FTR_ASYM_SMT)) {
		printk_once(KERN_INFO "Enabling Asymmetric SMT scheduling\n");
		flags |= SD_ASYM_PACKING;
	}
	return flags;
}
#endif

static struct sched_domain_topology_level powerpc_topology[] = {
#ifdef CONFIG_SCHED_SMT
	{ cpu_smt_mask, powerpc_smt_flags, SD_INIT_NAME(SMT) },
#endif
	{ cpu_cpu_mask, SD_INIT_NAME(DIE) },
	{ NULL, },
};

L
Linus Torvalds 已提交
800 801
void __init smp_cpus_done(unsigned int max_cpus)
{
802
	cpumask_var_t old_mask;
L
Linus Torvalds 已提交
803 804 805 806 807

	/* We want the setup_cpu() here to be called from CPU 0, but our
	 * init thread may have been "borrowed" by another CPU in the meantime
	 * se we pin us down to CPU 0 for a short while
	 */
808
	alloc_cpumask_var(&old_mask, GFP_NOWAIT);
809
	cpumask_copy(old_mask, &current->cpus_allowed);
J
Julia Lawall 已提交
810
	set_cpus_allowed_ptr(current, cpumask_of(boot_cpuid));
L
Linus Torvalds 已提交
811
	
812
	if (smp_ops && smp_ops->setup_cpu)
813
		smp_ops->setup_cpu(boot_cpuid);
L
Linus Torvalds 已提交
814

815 816 817
	set_cpus_allowed_ptr(current, old_mask);

	free_cpumask_var(old_mask);
818

819 820 821
	if (smp_ops && smp_ops->bringup_done)
		smp_ops->bringup_done();

822
	dump_numa_cpu_topology();
823

824
	set_sched_topology(powerpc_topology);
L
Linus Torvalds 已提交
825

826 827
}

L
Linus Torvalds 已提交
828 829 830
#ifdef CONFIG_HOTPLUG_CPU
int __cpu_disable(void)
{
831 832 833
	int cpu = smp_processor_id();
	int base, i;
	int err;
L
Linus Torvalds 已提交
834

835 836 837 838 839 840 841 842
	if (!smp_ops->cpu_disable)
		return -ENOSYS;

	err = smp_ops->cpu_disable();
	if (err)
		return err;

	/* Update sibling maps */
843
	base = cpu_first_thread_sibling(cpu);
844
	for (i = 0; i < threads_per_core && base + i < nr_cpu_ids; i++) {
845 846 847 848
		cpumask_clear_cpu(cpu, cpu_sibling_mask(base + i));
		cpumask_clear_cpu(base + i, cpu_sibling_mask(cpu));
		cpumask_clear_cpu(cpu, cpu_core_mask(base + i));
		cpumask_clear_cpu(base + i, cpu_core_mask(cpu));
849
	}
850
	traverse_core_siblings(cpu, false);
851 852

	return 0;
L
Linus Torvalds 已提交
853 854 855 856 857 858 859
}

void __cpu_die(unsigned int cpu)
{
	if (smp_ops->cpu_die)
		smp_ops->cpu_die(cpu);
}
860

861 862 863 864
void cpu_die(void)
{
	if (ppc_md.cpu_die)
		ppc_md.cpu_die();
865 866 867

	/* If we return, we re-enter start_secondary */
	start_secondary_resume();
868
}
869

L
Linus Torvalds 已提交
870
#endif