smp_64.c 32.6 KB
Newer Older
L
Linus Torvalds 已提交
1 2
/* smp.c: Sparc64 SMP support.
 *
3
 * Copyright (C) 1997, 2007, 2008 David S. Miller (davem@davemloft.net)
L
Linus Torvalds 已提交
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
 */

#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/threads.h>
#include <linux/smp.h>
#include <linux/interrupt.h>
#include <linux/kernel_stat.h>
#include <linux/delay.h>
#include <linux/init.h>
#include <linux/spinlock.h>
#include <linux/fs.h>
#include <linux/seq_file.h>
#include <linux/cache.h>
#include <linux/jiffies.h>
#include <linux/profile.h>
23
#include <linux/lmb.h>
24
#include <linux/cpu.h>
L
Linus Torvalds 已提交
25 26 27 28 29 30 31

#include <asm/head.h>
#include <asm/ptrace.h>
#include <asm/atomic.h>
#include <asm/tlbflush.h>
#include <asm/mmu_context.h>
#include <asm/cpudata.h>
D
David S. Miller 已提交
32 33
#include <asm/hvtramp.h>
#include <asm/io.h>
34
#include <asm/timer.h>
L
Linus Torvalds 已提交
35 36

#include <asm/irq.h>
A
Al Viro 已提交
37
#include <asm/irq_regs.h>
L
Linus Torvalds 已提交
38 39 40 41 42 43
#include <asm/page.h>
#include <asm/pgtable.h>
#include <asm/oplib.h>
#include <asm/uaccess.h>
#include <asm/starfire.h>
#include <asm/tlb.h>
44
#include <asm/sections.h>
45
#include <asm/prom.h>
46
#include <asm/mdesc.h>
47
#include <asm/ldc.h>
48
#include <asm/hypervisor.h>
L
Linus Torvalds 已提交
49

50 51
int sparc64_multi_core __read_mostly;

52
DEFINE_PER_CPU(cpumask_t, cpu_sibling_map) = CPU_MASK_NONE;
53 54
cpumask_t cpu_core_map[NR_CPUS] __read_mostly =
	{ [0 ... NR_CPUS-1] = CPU_MASK_NONE };
55

56
EXPORT_PER_CPU_SYMBOL(cpu_sibling_map);
57 58
EXPORT_SYMBOL(cpu_core_map);

L
Linus Torvalds 已提交
59 60 61 62 63 64 65
static cpumask_t smp_commenced_mask;

void smp_info(struct seq_file *m)
{
	int i;
	
	seq_printf(m, "State:\n");
66 67
	for_each_online_cpu(i)
		seq_printf(m, "CPU%d:\t\tonline\n", i);
L
Linus Torvalds 已提交
68 69 70 71 72 73
}

void smp_bogo(struct seq_file *m)
{
	int i;
	
74 75 76 77
	for_each_online_cpu(i)
		seq_printf(m,
			   "Cpu%dClkTck\t: %016lx\n",
			   i, cpu_data(i).clock_tick);
L
Linus Torvalds 已提交
78 79
}

80
extern void setup_sparc64_timer(void);
L
Linus Torvalds 已提交
81 82 83

static volatile unsigned long callin_flag = 0;

84
void __cpuinit smp_callin(void)
L
Linus Torvalds 已提交
85 86 87
{
	int cpuid = hard_smp_processor_id();

88
	__local_per_cpu_offset = __per_cpu_offset(cpuid);
L
Linus Torvalds 已提交
89

90
	if (tlb_type == hypervisor)
91
		sun4v_ktsb_register();
92

93
	__flush_tlb_all();
L
Linus Torvalds 已提交
94

95
	setup_sparc64_timer();
L
Linus Torvalds 已提交
96

97 98 99
	if (cheetah_pcache_forced_on)
		cheetah_enable_pcache();

L
Linus Torvalds 已提交
100 101 102 103 104 105 106 107 108
	local_irq_enable();

	callin_flag = 1;
	__asm__ __volatile__("membar #Sync\n\t"
			     "flush  %%g6" : : : "memory");

	/* Clear this or we will die instantly when we
	 * schedule back to this idler...
	 */
109
	current_thread_info()->new_child = 0;
L
Linus Torvalds 已提交
110 111 112 113 114

	/* Attach to the address space of init_task. */
	atomic_inc(&init_mm.mm_count);
	current->active_mm = &init_mm;

115 116 117
	/* inform the notifiers about the new cpu */
	notify_cpu_starting(cpuid);

L
Linus Torvalds 已提交
118
	while (!cpu_isset(cpuid, smp_commenced_mask))
119
		rmb();
L
Linus Torvalds 已提交
120

D
David S. Miller 已提交
121
	ipi_call_lock();
L
Linus Torvalds 已提交
122
	cpu_set(cpuid, cpu_online_map);
D
David S. Miller 已提交
123
	ipi_call_unlock();
124 125 126

	/* idle thread is expected to have preempt disabled */
	preempt_disable();
L
Linus Torvalds 已提交
127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161
}

void cpu_panic(void)
{
	printk("CPU[%d]: Returns from cpu_idle!\n", smp_processor_id());
	panic("SMP bolixed\n");
}

/* This tick register synchronization scheme is taken entirely from
 * the ia64 port, see arch/ia64/kernel/smpboot.c for details and credit.
 *
 * The only change I've made is to rework it so that the master
 * initiates the synchonization instead of the slave. -DaveM
 */

#define MASTER	0
#define SLAVE	(SMP_CACHE_BYTES/sizeof(unsigned long))

#define NUM_ROUNDS	64	/* magic value */
#define NUM_ITERS	5	/* likewise */

static DEFINE_SPINLOCK(itc_sync_lock);
static unsigned long go[SLAVE + 1];

#define DEBUG_TICK_SYNC	0

static inline long get_delta (long *rt, long *master)
{
	unsigned long best_t0 = 0, best_t1 = ~0UL, best_tm = 0;
	unsigned long tcenter, t0, t1, tm;
	unsigned long i;

	for (i = 0; i < NUM_ITERS; i++) {
		t0 = tick_ops->get_tick();
		go[MASTER] = 1;
162
		membar_safe("#StoreLoad");
L
Linus Torvalds 已提交
163
		while (!(tm = go[SLAVE]))
164
			rmb();
L
Linus Torvalds 已提交
165
		go[SLAVE] = 0;
166
		wmb();
L
Linus Torvalds 已提交
167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198
		t1 = tick_ops->get_tick();

		if (t1 - t0 < best_t1 - best_t0)
			best_t0 = t0, best_t1 = t1, best_tm = tm;
	}

	*rt = best_t1 - best_t0;
	*master = best_tm - best_t0;

	/* average best_t0 and best_t1 without overflow: */
	tcenter = (best_t0/2 + best_t1/2);
	if (best_t0 % 2 + best_t1 % 2 == 2)
		tcenter++;
	return tcenter - best_tm;
}

void smp_synchronize_tick_client(void)
{
	long i, delta, adj, adjust_latency = 0, done = 0;
	unsigned long flags, rt, master_time_stamp, bound;
#if DEBUG_TICK_SYNC
	struct {
		long rt;	/* roundtrip time */
		long master;	/* master's timestamp */
		long diff;	/* difference between midpoint and master's timestamp */
		long lat;	/* estimate of itc adjustment latency */
	} t[NUM_ROUNDS];
#endif

	go[MASTER] = 1;

	while (go[MASTER])
199
		rmb();
L
Linus Torvalds 已提交
200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216

	local_irq_save(flags);
	{
		for (i = 0; i < NUM_ROUNDS; i++) {
			delta = get_delta(&rt, &master_time_stamp);
			if (delta == 0) {
				done = 1;	/* let's lock on to this... */
				bound = rt;
			}

			if (!done) {
				if (i > 0) {
					adjust_latency += -delta;
					adj = -delta + adjust_latency/4;
				} else
					adj = -delta;

217
				tick_ops->add_tick(adj);
L
Linus Torvalds 已提交
218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234
			}
#if DEBUG_TICK_SYNC
			t[i].rt = rt;
			t[i].master = master_time_stamp;
			t[i].diff = delta;
			t[i].lat = adjust_latency/4;
#endif
		}
	}
	local_irq_restore(flags);

#if DEBUG_TICK_SYNC
	for (i = 0; i < NUM_ROUNDS; i++)
		printk("rt=%5ld master=%5ld diff=%5ld adjlat=%5ld\n",
		       t[i].rt, t[i].master, t[i].diff, t[i].lat);
#endif

J
Joe Perches 已提交
235 236 237
	printk(KERN_INFO "CPU %d: synchronized TICK with master CPU "
	       "(last diff %ld cycles, maxerr %lu cycles)\n",
	       smp_processor_id(), delta, rt);
L
Linus Torvalds 已提交
238 239 240 241 242 243 244 245 246 247 248 249 250 251
}

static void smp_start_sync_tick_client(int cpu);

static void smp_synchronize_one_tick(int cpu)
{
	unsigned long flags, i;

	go[MASTER] = 0;

	smp_start_sync_tick_client(cpu);

	/* wait for client to be ready */
	while (!go[MASTER])
252
		rmb();
L
Linus Torvalds 已提交
253 254 255

	/* now let the client proceed into his loop */
	go[MASTER] = 0;
256
	membar_safe("#StoreLoad");
L
Linus Torvalds 已提交
257 258 259 260 261

	spin_lock_irqsave(&itc_sync_lock, flags);
	{
		for (i = 0; i < NUM_ROUNDS*NUM_ITERS; i++) {
			while (!go[MASTER])
262
				rmb();
L
Linus Torvalds 已提交
263
			go[MASTER] = 0;
264
			wmb();
L
Linus Torvalds 已提交
265
			go[SLAVE] = tick_ops->get_tick();
266
			membar_safe("#StoreLoad");
L
Linus Torvalds 已提交
267 268 269 270 271
		}
	}
	spin_unlock_irqrestore(&itc_sync_lock, flags);
}

272
#if defined(CONFIG_SUN_LDOMS) && defined(CONFIG_HOTPLUG_CPU)
D
David S. Miller 已提交
273 274 275 276 277 278 279 280
/* XXX Put this in some common place. XXX */
static unsigned long kimage_addr_to_ra(void *p)
{
	unsigned long val = (unsigned long) p;

	return kern_base + (val - KERNBASE);
}

A
Al Viro 已提交
281
static void __cpuinit ldom_startcpu_cpuid(unsigned int cpu, unsigned long thread_reg)
282 283 284 285 286 287 288 289
{
	extern unsigned long sparc64_ttable_tl0;
	extern unsigned long kern_locked_tte_data;
	struct hvtramp_descr *hdesc;
	unsigned long trampoline_ra;
	struct trap_per_cpu *tb;
	u64 tte_vaddr, tte_data;
	unsigned long hv_err;
290
	int i;
291

292 293 294 295
	hdesc = kzalloc(sizeof(*hdesc) +
			(sizeof(struct hvtramp_mapping) *
			 num_kernel_image_mappings - 1),
			GFP_KERNEL);
296
	if (!hdesc) {
D
David S. Miller 已提交
297
		printk(KERN_ERR "ldom_startcpu_cpuid: Cannot allocate "
298 299 300 301 302
		       "hvtramp_descr.\n");
		return;
	}

	hdesc->cpu = cpu;
303
	hdesc->num_mappings = num_kernel_image_mappings;
304 305 306 307 308 309 310 311 312 313 314 315

	tb = &trap_block[cpu];
	tb->hdesc = hdesc;

	hdesc->fault_info_va = (unsigned long) &tb->fault_info;
	hdesc->fault_info_pa = kimage_addr_to_ra(&tb->fault_info);

	hdesc->thread_reg = thread_reg;

	tte_vaddr = (unsigned long) KERNBASE;
	tte_data = kern_locked_tte_data;

316 317 318
	for (i = 0; i < hdesc->num_mappings; i++) {
		hdesc->maps[i].vaddr = tte_vaddr;
		hdesc->maps[i].tte   = tte_data;
319 320 321 322 323 324 325 326 327
		tte_vaddr += 0x400000;
		tte_data  += 0x400000;
	}

	trampoline_ra = kimage_addr_to_ra(hv_cpu_startup);

	hv_err = sun4v_cpu_start(cpu, trampoline_ra,
				 kimage_addr_to_ra(&sparc64_ttable_tl0),
				 __pa(hdesc));
328 329 330
	if (hv_err)
		printk(KERN_ERR "ldom_startcpu_cpuid: sun4v_cpu_start() "
		       "gives error %lu\n", hv_err);
331 332 333
}
#endif

L
Linus Torvalds 已提交
334 335 336 337 338 339 340 341
extern unsigned long sparc64_cpu_startup;

/* The OBP cpu startup callback truncates the 3rd arg cookie to
 * 32-bits (I think) so to be safe we have it read the pointer
 * contained here so we work on >4GB machines. -DaveM
 */
static struct thread_info *cpu_new_thread = NULL;

A
Al Viro 已提交
342
static int __cpuinit smp_boot_one_cpu(unsigned int cpu)
L
Linus Torvalds 已提交
343
{
344
	struct trap_per_cpu *tb = &trap_block[cpu];
L
Linus Torvalds 已提交
345 346 347 348 349
	unsigned long entry =
		(unsigned long)(&sparc64_cpu_startup);
	unsigned long cookie =
		(unsigned long)(&cpu_new_thread);
	struct task_struct *p;
350
	int timeout, ret;
L
Linus Torvalds 已提交
351 352

	p = fork_idle(cpu);
A
Akinobu Mita 已提交
353 354
	if (IS_ERR(p))
		return PTR_ERR(p);
L
Linus Torvalds 已提交
355
	callin_flag = 0;
A
Al Viro 已提交
356
	cpu_new_thread = task_thread_info(p);
L
Linus Torvalds 已提交
357

358
	if (tlb_type == hypervisor) {
359
#if defined(CONFIG_SUN_LDOMS) && defined(CONFIG_HOTPLUG_CPU)
360 361 362 363 364 365
		if (ldom_domaining_enabled)
			ldom_startcpu_cpuid(cpu,
					    (unsigned long) cpu_new_thread);
		else
#endif
			prom_startcpu_cpuid(cpu, entry, cookie);
366
	} else {
367
		struct device_node *dp = of_find_node_by_cpuid(cpu);
368

369
		prom_startcpu(dp->node, entry, cookie);
370
	}
L
Linus Torvalds 已提交
371

372
	for (timeout = 0; timeout < 50000; timeout++) {
L
Linus Torvalds 已提交
373 374 375 376
		if (callin_flag)
			break;
		udelay(100);
	}
377

L
Linus Torvalds 已提交
378 379 380 381 382 383 384 385
	if (callin_flag) {
		ret = 0;
	} else {
		printk("Processor %d is stuck.\n", cpu);
		ret = -ENODEV;
	}
	cpu_new_thread = NULL;

386 387 388 389 390
	if (tb->hdesc) {
		kfree(tb->hdesc);
		tb->hdesc = NULL;
	}

L
Linus Torvalds 已提交
391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
	return ret;
}

static void spitfire_xcall_helper(u64 data0, u64 data1, u64 data2, u64 pstate, unsigned long cpu)
{
	u64 result, target;
	int stuck, tmp;

	if (this_is_starfire) {
		/* map to real upaid */
		cpu = (((cpu & 0x3c) << 1) |
			((cpu & 0x40) >> 4) |
			(cpu & 0x3));
	}

	target = (cpu << 14) | 0x70;
again:
	/* Ok, this is the real Spitfire Errata #54.
	 * One must read back from a UDB internal register
	 * after writes to the UDB interrupt dispatch, but
	 * before the membar Sync for that write.
	 * So we use the high UDB control register (ASI 0x7f,
	 * ADDR 0x20) for the dummy read. -DaveM
	 */
	tmp = 0x40;
	__asm__ __volatile__(
	"wrpr	%1, %2, %%pstate\n\t"
	"stxa	%4, [%0] %3\n\t"
	"stxa	%5, [%0+%8] %3\n\t"
	"add	%0, %8, %0\n\t"
	"stxa	%6, [%0+%8] %3\n\t"
	"membar	#Sync\n\t"
	"stxa	%%g0, [%7] %3\n\t"
	"membar	#Sync\n\t"
	"mov	0x20, %%g1\n\t"
	"ldxa	[%%g1] 0x7f, %%g0\n\t"
	"membar	#Sync"
	: "=r" (tmp)
	: "r" (pstate), "i" (PSTATE_IE), "i" (ASI_INTR_W),
	  "r" (data0), "r" (data1), "r" (data2), "r" (target),
	  "r" (0x10), "0" (tmp)
        : "g1");

	/* NOTE: PSTATE_IE is still clear. */
	stuck = 100000;
	do {
		__asm__ __volatile__("ldxa [%%g0] %1, %0"
			: "=r" (result)
			: "i" (ASI_INTR_DISPATCH_STAT));
		if (result == 0) {
			__asm__ __volatile__("wrpr %0, 0x0, %%pstate"
					     : : "r" (pstate));
			return;
		}
		stuck -= 1;
		if (stuck == 0)
			break;
	} while (result & 0x1);
	__asm__ __volatile__("wrpr %0, 0x0, %%pstate"
			     : : "r" (pstate));
	if (stuck == 0) {
		printk("CPU[%d]: mondo stuckage result[%016lx]\n",
		       smp_processor_id(), result);
	} else {
		udelay(2);
		goto again;
	}
}

460
static void spitfire_xcall_deliver(struct trap_per_cpu *tb, int cnt)
L
Linus Torvalds 已提交
461
{
462 463
	u64 *mondo, data0, data1, data2;
	u16 *cpu_list;
L
Linus Torvalds 已提交
464 465 466 467
	u64 pstate;
	int i;

	__asm__ __volatile__("rdpr %%pstate, %0" : "=r" (pstate));
468 469 470 471 472 473 474
	cpu_list = __va(tb->cpu_list_pa);
	mondo = __va(tb->cpu_mondo_block_pa);
	data0 = mondo[0];
	data1 = mondo[1];
	data2 = mondo[2];
	for (i = 0; i < cnt; i++)
		spitfire_xcall_helper(data0, data1, data2, pstate, cpu_list[i]);
L
Linus Torvalds 已提交
475 476 477 478 479 480
}

/* Cheetah now allows to send the whole 64-bytes of data in the interrupt
 * packet, but we have no use for that.  However we do take advantage of
 * the new pipelining feature (ie. dispatch to multiple cpus simultaneously).
 */
481
static void cheetah_xcall_deliver(struct trap_per_cpu *tb, int cnt)
L
Linus Torvalds 已提交
482
{
483
	int nack_busy_id, is_jbus, need_more;
484 485
	u64 *mondo, pstate, ver, busy_mask;
	u16 *cpu_list;
L
Linus Torvalds 已提交
486

487 488
	cpu_list = __va(tb->cpu_list_pa);
	mondo = __va(tb->cpu_mondo_block_pa);
489

L
Linus Torvalds 已提交
490 491 492 493 494
	/* Unfortunately, someone at Sun had the brilliant idea to make the
	 * busy/nack fields hard-coded by ITID number for this Ultra-III
	 * derivative processor.
	 */
	__asm__ ("rdpr %%ver, %0" : "=r" (ver));
495 496
	is_jbus = ((ver >> 32) == __JALAPENO_ID ||
		   (ver >> 32) == __SERRANO_ID);
L
Linus Torvalds 已提交
497 498 499 500

	__asm__ __volatile__("rdpr %%pstate, %0" : "=r" (pstate));

retry:
501
	need_more = 0;
L
Linus Torvalds 已提交
502 503 504 505 506 507 508 509 510
	__asm__ __volatile__("wrpr %0, %1, %%pstate\n\t"
			     : : "r" (pstate), "i" (PSTATE_IE));

	/* Setup the dispatch data registers. */
	__asm__ __volatile__("stxa	%0, [%3] %6\n\t"
			     "stxa	%1, [%4] %6\n\t"
			     "stxa	%2, [%5] %6\n\t"
			     "membar	#Sync\n\t"
			     : /* no outputs */
511
			     : "r" (mondo[0]), "r" (mondo[1]), "r" (mondo[2]),
L
Linus Torvalds 已提交
512 513 514 515
			       "r" (0x40), "r" (0x50), "r" (0x60),
			       "i" (ASI_INTR_W));

	nack_busy_id = 0;
516
	busy_mask = 0;
L
Linus Torvalds 已提交
517 518 519
	{
		int i;

520 521 522 523 524 525
		for (i = 0; i < cnt; i++) {
			u64 target, nr;

			nr = cpu_list[i];
			if (nr == 0xffff)
				continue;
L
Linus Torvalds 已提交
526

527
			target = (nr << 14) | 0x70;
528
			if (is_jbus) {
529
				busy_mask |= (0x1UL << (nr * 2));
530
			} else {
L
Linus Torvalds 已提交
531
				target |= (nack_busy_id << 24);
532 533 534
				busy_mask |= (0x1UL <<
					      (nack_busy_id * 2));
			}
L
Linus Torvalds 已提交
535 536 537 538 539 540
			__asm__ __volatile__(
				"stxa	%%g0, [%0] %1\n\t"
				"membar	#Sync\n\t"
				: /* no outputs */
				: "r" (target), "i" (ASI_INTR_W));
			nack_busy_id++;
541 542 543 544
			if (nack_busy_id == 32) {
				need_more = 1;
				break;
			}
L
Linus Torvalds 已提交
545 546 547 548 549
		}
	}

	/* Now, poll for completion. */
	{
550
		u64 dispatch_stat, nack_mask;
L
Linus Torvalds 已提交
551 552 553
		long stuck;

		stuck = 100000 * nack_busy_id;
554
		nack_mask = busy_mask << 1;
L
Linus Torvalds 已提交
555 556 557 558
		do {
			__asm__ __volatile__("ldxa	[%%g0] %1, %0"
					     : "=r" (dispatch_stat)
					     : "i" (ASI_INTR_DISPATCH_STAT));
559
			if (!(dispatch_stat & (busy_mask | nack_mask))) {
L
Linus Torvalds 已提交
560 561
				__asm__ __volatile__("wrpr %0, 0x0, %%pstate"
						     : : "r" (pstate));
562
				if (unlikely(need_more)) {
563 564 565 566 567 568 569
					int i, this_cnt = 0;
					for (i = 0; i < cnt; i++) {
						if (cpu_list[i] == 0xffff)
							continue;
						cpu_list[i] = 0xffff;
						this_cnt++;
						if (this_cnt == 32)
570 571 572 573
							break;
					}
					goto retry;
				}
L
Linus Torvalds 已提交
574 575 576 577
				return;
			}
			if (!--stuck)
				break;
578
		} while (dispatch_stat & busy_mask);
L
Linus Torvalds 已提交
579 580 581 582

		__asm__ __volatile__("wrpr %0, 0x0, %%pstate"
				     : : "r" (pstate));

583
		if (dispatch_stat & busy_mask) {
L
Linus Torvalds 已提交
584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599
			/* Busy bits will not clear, continue instead
			 * of freezing up on this cpu.
			 */
			printk("CPU[%d]: mondo stuckage result[%016lx]\n",
			       smp_processor_id(), dispatch_stat);
		} else {
			int i, this_busy_nack = 0;

			/* Delay some random time with interrupts enabled
			 * to prevent deadlock.
			 */
			udelay(2 * nack_busy_id);

			/* Clear out the mask bits for cpus which did not
			 * NACK us.
			 */
600 601 602 603 604 605
			for (i = 0; i < cnt; i++) {
				u64 check_mask, nr;

				nr = cpu_list[i];
				if (nr == 0xffff)
					continue;
L
Linus Torvalds 已提交
606

607
				if (is_jbus)
608
					check_mask = (0x2UL << (2*nr));
L
Linus Torvalds 已提交
609 610 611 612
				else
					check_mask = (0x2UL <<
						      this_busy_nack);
				if ((dispatch_stat & check_mask) == 0)
613
					cpu_list[i] = 0xffff;
L
Linus Torvalds 已提交
614
				this_busy_nack += 2;
615 616
				if (this_busy_nack == 64)
					break;
L
Linus Torvalds 已提交
617 618 619 620 621 622 623
			}

			goto retry;
		}
	}
}

624
/* Multi-cpu list version.  */
625
static void hypervisor_xcall_deliver(struct trap_per_cpu *tb, int cnt)
626
{
627
	int retries, this_cpu, prev_sent, i, saw_cpu_error;
628
	unsigned long status;
629
	u16 *cpu_list;
630

631
	this_cpu = smp_processor_id();
632

633 634
	cpu_list = __va(tb->cpu_list_pa);

635
	saw_cpu_error = 0;
636
	retries = 0;
637
	prev_sent = 0;
638
	do {
639
		int forward_progress, n_sent;
640

641 642 643 644 645 646
		status = sun4v_cpu_mondo_send(cnt,
					      tb->cpu_list_pa,
					      tb->cpu_mondo_block_pa);

		/* HV_EOK means all cpus received the xcall, we're done.  */
		if (likely(status == HV_EOK))
647
			break;
648

649 650 651 652
		/* First, see if we made any forward progress.
		 *
		 * The hypervisor indicates successful sends by setting
		 * cpu list entries to the value 0xffff.
653
		 */
654
		n_sent = 0;
655
		for (i = 0; i < cnt; i++) {
656 657
			if (likely(cpu_list[i] == 0xffff))
				n_sent++;
658 659
		}

660 661 662 663 664 665
		forward_progress = 0;
		if (n_sent > prev_sent)
			forward_progress = 1;

		prev_sent = n_sent;

666 667 668 669 670 671 672 673 674 675 676 677 678 679
		/* If we get a HV_ECPUERROR, then one or more of the cpus
		 * in the list are in error state.  Use the cpu_state()
		 * hypervisor call to find out which cpus are in error state.
		 */
		if (unlikely(status == HV_ECPUERROR)) {
			for (i = 0; i < cnt; i++) {
				long err;
				u16 cpu;

				cpu = cpu_list[i];
				if (cpu == 0xffff)
					continue;

				err = sun4v_cpu_state(cpu);
680 681
				if (err == HV_CPU_STATE_ERROR) {
					saw_cpu_error = (cpu + 1);
682
					cpu_list[i] = 0xffff;
683 684 685 686 687
				}
			}
		} else if (unlikely(status != HV_EWOULDBLOCK))
			goto fatal_mondo_error;

688 689 690 691 692 693 694
		/* Don't bother rewriting the CPU list, just leave the
		 * 0xffff and non-0xffff entries in there and the
		 * hypervisor will do the right thing.
		 *
		 * Only advance timeout state if we didn't make any
		 * forward progress.
		 */
695 696 697 698 699 700 701 702 703
		if (unlikely(!forward_progress)) {
			if (unlikely(++retries > 10000))
				goto fatal_mondo_timeout;

			/* Delay a little bit to let other cpus catch up
			 * on their cpu mondo queue work.
			 */
			udelay(2 * cnt);
		}
704 705
	} while (1);

706
	if (unlikely(saw_cpu_error))
707 708 709 710 711 712
		goto fatal_mondo_cpu_error;

	return;

fatal_mondo_cpu_error:
	printk(KERN_CRIT "CPU[%d]: SUN4V mondo cpu error, some target cpus "
713 714
	       "(including %d) were in error state\n",
	       this_cpu, saw_cpu_error - 1);
715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734
	return;

fatal_mondo_timeout:
	printk(KERN_CRIT "CPU[%d]: SUN4V mondo timeout, no forward "
	       " progress after %d retries.\n",
	       this_cpu, retries);
	goto dump_cpu_list_and_out;

fatal_mondo_error:
	printk(KERN_CRIT "CPU[%d]: Unexpected SUN4V mondo error %lu\n",
	       this_cpu, status);
	printk(KERN_CRIT "CPU[%d]: Args were cnt(%d) cpulist_pa(%lx) "
	       "mondo_block_pa(%lx)\n",
	       this_cpu, cnt, tb->cpu_list_pa, tb->cpu_mondo_block_pa);

dump_cpu_list_and_out:
	printk(KERN_CRIT "CPU[%d]: CPU list [ ", this_cpu);
	for (i = 0; i < cnt; i++)
		printk("%u ", cpu_list[i]);
	printk("]\n");
735
}
736

737
static void (*xcall_deliver_impl)(struct trap_per_cpu *, int);
738 739 740

static void xcall_deliver(u64 data0, u64 data1, u64 data2, const cpumask_t *mask)
{
741 742
	struct trap_per_cpu *tb;
	int this_cpu, i, cnt;
743
	unsigned long flags;
744 745
	u16 *cpu_list;
	u64 *mondo;
746 747 748 749 750 751 752 753 754 755 756 757

	/* We have to do this whole thing with interrupts fully disabled.
	 * Otherwise if we send an xcall from interrupt context it will
	 * corrupt both our mondo block and cpu list state.
	 *
	 * One consequence of this is that we cannot use timeout mechanisms
	 * that depend upon interrupts being delivered locally.  So, for
	 * example, we cannot sample jiffies and expect it to advance.
	 *
	 * Fortunately, udelay() uses %stick/%tick so we can use that.
	 */
	local_irq_save(flags);
758 759 760 761 762 763 764 765 766 767 768 769 770 771

	this_cpu = smp_processor_id();
	tb = &trap_block[this_cpu];

	mondo = __va(tb->cpu_mondo_block_pa);
	mondo[0] = data0;
	mondo[1] = data1;
	mondo[2] = data2;
	wmb();

	cpu_list = __va(tb->cpu_list_pa);

	/* Setup the initial cpu list.  */
	cnt = 0;
772
	for_each_cpu(i, mask) {
773 774 775 776 777 778 779 780
		if (i == this_cpu || !cpu_online(i))
			continue;
		cpu_list[cnt++] = i;
	}

	if (cnt)
		xcall_deliver_impl(tb, cnt);

781
	local_irq_restore(flags);
782
}
783

784 785 786
/* Send cross call to all processors mentioned in MASK_P
 * except self.  Really, there are only two cases currently,
 * "&cpu_online_map" and "&mm->cpu_vm_mask".
L
Linus Torvalds 已提交
787
 */
788
static void smp_cross_call_masked(unsigned long *func, u32 ctx, u64 data1, u64 data2, const cpumask_t *mask)
L
Linus Torvalds 已提交
789 790 791
{
	u64 data0 = (((u64)ctx)<<32 | (((u64)func) & 0xffffffff));

792 793
	xcall_deliver(data0, data1, data2, mask);
}
L
Linus Torvalds 已提交
794

795 796 797 798
/* Send cross call to all processors except self. */
static void smp_cross_call(unsigned long *func, u32 ctx, u64 data1, u64 data2)
{
	smp_cross_call_masked(func, ctx, data1, data2, &cpu_online_map);
L
Linus Torvalds 已提交
799 800 801 802 803 804
}

extern unsigned long xcall_sync_tick;

static void smp_start_sync_tick_client(int cpu)
{
805 806
	xcall_deliver((u64) &xcall_sync_tick, 0, 0,
		      &cpumask_of_cpu(cpu));
L
Linus Torvalds 已提交
807 808 809 810
}

extern unsigned long xcall_call_function;

811
void arch_send_call_function_ipi(cpumask_t mask)
L
Linus Torvalds 已提交
812
{
813
	xcall_deliver((u64) &xcall_call_function, 0, 0, &mask);
814
}
L
Linus Torvalds 已提交
815

816
extern unsigned long xcall_call_function_single;
L
Linus Torvalds 已提交
817

818 819
void arch_send_call_function_single_ipi(int cpu)
{
820 821
	xcall_deliver((u64) &xcall_call_function_single, 0, 0,
		      &cpumask_of_cpu(cpu));
L
Linus Torvalds 已提交
822 823 824 825
}

void smp_call_function_client(int irq, struct pt_regs *regs)
{
826 827 828
	clear_softint(1 << irq);
	generic_smp_call_function_interrupt();
}
L
Linus Torvalds 已提交
829

830 831
void smp_call_function_single_client(int irq, struct pt_regs *regs)
{
L
Linus Torvalds 已提交
832
	clear_softint(1 << irq);
833
	generic_smp_call_function_single_interrupt();
L
Linus Torvalds 已提交
834 835
}

836 837
static void tsb_sync(void *info)
{
838
	struct trap_per_cpu *tp = &trap_block[raw_smp_processor_id()];
839 840
	struct mm_struct *mm = info;

841 842 843 844 845 846 847
	/* It is not valid to test "currrent->active_mm == mm" here.
	 *
	 * The value of "current" is not changed atomically with
	 * switch_mm().  But that's OK, we just need to check the
	 * current cpu's trap block PGD physical address.
	 */
	if (tp->pgd_paddr == __pa(mm->pgd))
848 849 850 851 852
		tsb_context_switch(mm);
}

void smp_tsb_sync(struct mm_struct *mm)
{
853
	smp_call_function_mask(mm->cpu_vm_mask, tsb_sync, mm, 1);
854 855
}

L
Linus Torvalds 已提交
856 857 858
extern unsigned long xcall_flush_tlb_mm;
extern unsigned long xcall_flush_tlb_pending;
extern unsigned long xcall_flush_tlb_kernel_range;
859
extern unsigned long xcall_fetch_glob_regs;
L
Linus Torvalds 已提交
860
extern unsigned long xcall_receive_signal;
861
extern unsigned long xcall_new_mmu_context_version;
D
David S. Miller 已提交
862 863 864
#ifdef CONFIG_KGDB
extern unsigned long xcall_kgdb_capture;
#endif
L
Linus Torvalds 已提交
865 866 867 868 869 870 871 872 873 874 875

#ifdef DCACHE_ALIASING_POSSIBLE
extern unsigned long xcall_flush_dcache_page_cheetah;
#endif
extern unsigned long xcall_flush_dcache_page_spitfire;

#ifdef CONFIG_DEBUG_DCFLUSH
extern atomic_t dcpage_flushes;
extern atomic_t dcpage_flushes_xcall;
#endif

D
David S. Miller 已提交
876
static inline void __local_flush_dcache_page(struct page *page)
L
Linus Torvalds 已提交
877 878 879 880 881 882 883 884 885 886 887 888 889 890
{
#ifdef DCACHE_ALIASING_POSSIBLE
	__flush_dcache_page(page_address(page),
			    ((tlb_type == spitfire) &&
			     page_mapping(page) != NULL));
#else
	if (page_mapping(page) != NULL &&
	    tlb_type == spitfire)
		__flush_icache_page(__pa(page_address(page)));
#endif
}

void smp_flush_dcache_page_impl(struct page *page, int cpu)
{
891 892 893 894
	int this_cpu;

	if (tlb_type == hypervisor)
		return;
L
Linus Torvalds 已提交
895 896 897 898

#ifdef CONFIG_DEBUG_DCFLUSH
	atomic_inc(&dcpage_flushes);
#endif
899 900 901

	this_cpu = get_cpu();

L
Linus Torvalds 已提交
902 903 904 905
	if (cpu == this_cpu) {
		__local_flush_dcache_page(page);
	} else if (cpu_online(cpu)) {
		void *pg_addr = page_address(page);
906
		u64 data0 = 0;
L
Linus Torvalds 已提交
907 908

		if (tlb_type == spitfire) {
909
			data0 = ((u64)&xcall_flush_dcache_page_spitfire);
L
Linus Torvalds 已提交
910 911
			if (page_mapping(page) != NULL)
				data0 |= ((u64)1 << 32);
912
		} else if (tlb_type == cheetah || tlb_type == cheetah_plus) {
L
Linus Torvalds 已提交
913
#ifdef DCACHE_ALIASING_POSSIBLE
914
			data0 =	((u64)&xcall_flush_dcache_page_cheetah);
L
Linus Torvalds 已提交
915 916
#endif
		}
917 918
		if (data0) {
			xcall_deliver(data0, __pa(pg_addr),
919
				      (u64) pg_addr, &cpumask_of_cpu(cpu));
L
Linus Torvalds 已提交
920
#ifdef CONFIG_DEBUG_DCFLUSH
921
			atomic_inc(&dcpage_flushes_xcall);
L
Linus Torvalds 已提交
922
#endif
923
		}
L
Linus Torvalds 已提交
924 925 926 927 928 929 930
	}

	put_cpu();
}

void flush_dcache_page_all(struct mm_struct *mm, struct page *page)
{
931
	void *pg_addr;
932
	int this_cpu;
933
	u64 data0;
934 935 936 937 938

	if (tlb_type == hypervisor)
		return;

	this_cpu = get_cpu();
L
Linus Torvalds 已提交
939 940 941 942

#ifdef CONFIG_DEBUG_DCFLUSH
	atomic_inc(&dcpage_flushes);
#endif
943 944
	data0 = 0;
	pg_addr = page_address(page);
L
Linus Torvalds 已提交
945 946 947 948
	if (tlb_type == spitfire) {
		data0 = ((u64)&xcall_flush_dcache_page_spitfire);
		if (page_mapping(page) != NULL)
			data0 |= ((u64)1 << 32);
949
	} else if (tlb_type == cheetah || tlb_type == cheetah_plus) {
L
Linus Torvalds 已提交
950 951 952 953
#ifdef DCACHE_ALIASING_POSSIBLE
		data0 = ((u64)&xcall_flush_dcache_page_cheetah);
#endif
	}
954 955
	if (data0) {
		xcall_deliver(data0, __pa(pg_addr),
956
			      (u64) pg_addr, &cpu_online_map);
L
Linus Torvalds 已提交
957
#ifdef CONFIG_DEBUG_DCFLUSH
958
		atomic_inc(&dcpage_flushes_xcall);
L
Linus Torvalds 已提交
959
#endif
960
	}
L
Linus Torvalds 已提交
961 962 963 964 965
	__local_flush_dcache_page(page);

	put_cpu();
}

966
void smp_new_mmu_context_version_client(int irq, struct pt_regs *regs)
L
Linus Torvalds 已提交
967
{
968
	struct mm_struct *mm;
969
	unsigned long flags;
970

L
Linus Torvalds 已提交
971
	clear_softint(1 << irq);
972 973 974 975 976

	/* See if we need to allocate a new TLB context because
	 * the version of the one we are using is now out of date.
	 */
	mm = current->active_mm;
977 978
	if (unlikely(!mm || (mm == &init_mm)))
		return;
979

980
	spin_lock_irqsave(&mm->context.lock, flags);
981

982 983
	if (unlikely(!CTX_VALID(mm->context)))
		get_new_mmu_context(mm);
984

985
	spin_unlock_irqrestore(&mm->context.lock, flags);
986

987 988 989
	load_secondary_context(mm);
	__flush_tlb_mm(CTX_HWBITS(mm->context),
		       SECONDARY_CONTEXT);
990 991 992 993
}

void smp_new_mmu_context_version(void)
{
994
	smp_cross_call(&xcall_new_mmu_context_version, 0, 0, 0);
L
Linus Torvalds 已提交
995 996
}

D
David S. Miller 已提交
997 998 999 1000 1001 1002 1003
#ifdef CONFIG_KGDB
void kgdb_roundup_cpus(unsigned long flags)
{
	smp_cross_call(&xcall_kgdb_capture, 0, 0, 0);
}
#endif

1004 1005 1006 1007 1008
void smp_fetch_global_regs(void)
{
	smp_cross_call(&xcall_fetch_glob_regs, 0, 0, 0);
}

L
Linus Torvalds 已提交
1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046
/* We know that the window frames of the user have been flushed
 * to the stack before we get here because all callers of us
 * are flush_tlb_*() routines, and these run after flush_cache_*()
 * which performs the flushw.
 *
 * The SMP TLB coherency scheme we use works as follows:
 *
 * 1) mm->cpu_vm_mask is a bit mask of which cpus an address
 *    space has (potentially) executed on, this is the heuristic
 *    we use to avoid doing cross calls.
 *
 *    Also, for flushing from kswapd and also for clones, we
 *    use cpu_vm_mask as the list of cpus to make run the TLB.
 *
 * 2) TLB context numbers are shared globally across all processors
 *    in the system, this allows us to play several games to avoid
 *    cross calls.
 *
 *    One invariant is that when a cpu switches to a process, and
 *    that processes tsk->active_mm->cpu_vm_mask does not have the
 *    current cpu's bit set, that tlb context is flushed locally.
 *
 *    If the address space is non-shared (ie. mm->count == 1) we avoid
 *    cross calls when we want to flush the currently running process's
 *    tlb state.  This is done by clearing all cpu bits except the current
 *    processor's in current->active_mm->cpu_vm_mask and performing the
 *    flush locally only.  This will force any subsequent cpus which run
 *    this task to flush the context from the local tlb if the process
 *    migrates to another cpu (again).
 *
 * 3) For shared address spaces (threads) and swapping we bite the
 *    bullet for most cases and perform the cross call (but only to
 *    the cpus listed in cpu_vm_mask).
 *
 *    The performance gain from "optimizing" away the cross call for threads is
 *    questionable (in theory the big win for threads is the massive sharing of
 *    address space state across processors).
 */
1047 1048 1049 1050 1051

/* This currently is only used by the hugetlb arch pre-fault
 * hook on UltraSPARC-III+ and later when changing the pagesize
 * bits of the context register for an address space.
 */
L
Linus Torvalds 已提交
1052 1053
void smp_flush_tlb_mm(struct mm_struct *mm)
{
1054 1055
	u32 ctx = CTX_HWBITS(mm->context);
	int cpu = get_cpu();
L
Linus Torvalds 已提交
1056

1057 1058 1059 1060
	if (atomic_read(&mm->mm_users) == 1) {
		mm->cpu_vm_mask = cpumask_of_cpu(cpu);
		goto local_flush_and_out;
	}
L
Linus Torvalds 已提交
1061

1062 1063
	smp_cross_call_masked(&xcall_flush_tlb_mm,
			      ctx, 0, 0,
1064
			      &mm->cpu_vm_mask);
L
Linus Torvalds 已提交
1065

1066 1067
local_flush_and_out:
	__flush_tlb_mm(ctx, SECONDARY_CONTEXT);
L
Linus Torvalds 已提交
1068

1069
	put_cpu();
L
Linus Torvalds 已提交
1070 1071 1072 1073 1074 1075 1076
}

void smp_flush_tlb_pending(struct mm_struct *mm, unsigned long nr, unsigned long *vaddrs)
{
	u32 ctx = CTX_HWBITS(mm->context);
	int cpu = get_cpu();

H
Hugh Dickins 已提交
1077
	if (mm == current->active_mm && atomic_read(&mm->mm_users) == 1)
L
Linus Torvalds 已提交
1078
		mm->cpu_vm_mask = cpumask_of_cpu(cpu);
H
Hugh Dickins 已提交
1079 1080 1081
	else
		smp_cross_call_masked(&xcall_flush_tlb_pending,
				      ctx, nr, (unsigned long) vaddrs,
1082
				      &mm->cpu_vm_mask);
L
Linus Torvalds 已提交
1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123

	__flush_tlb_pending(ctx, nr, vaddrs);

	put_cpu();
}

void smp_flush_tlb_kernel_range(unsigned long start, unsigned long end)
{
	start &= PAGE_MASK;
	end    = PAGE_ALIGN(end);
	if (start != end) {
		smp_cross_call(&xcall_flush_tlb_kernel_range,
			       0, start, end);

		__flush_tlb_kernel_range(start, end);
	}
}

/* CPU capture. */
/* #define CAPTURE_DEBUG */
extern unsigned long xcall_capture;

static atomic_t smp_capture_depth = ATOMIC_INIT(0);
static atomic_t smp_capture_registry = ATOMIC_INIT(0);
static unsigned long penguins_are_doing_time;

void smp_capture(void)
{
	int result = atomic_add_ret(1, &smp_capture_depth);

	if (result == 1) {
		int ncpus = num_online_cpus();

#ifdef CAPTURE_DEBUG
		printk("CPU[%d]: Sending penguins to jail...",
		       smp_processor_id());
#endif
		penguins_are_doing_time = 1;
		atomic_inc(&smp_capture_registry);
		smp_cross_call(&xcall_capture, 0, 0, 0);
		while (atomic_read(&smp_capture_registry) != ncpus)
1124
			rmb();
L
Linus Torvalds 已提交
1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139
#ifdef CAPTURE_DEBUG
		printk("done\n");
#endif
	}
}

void smp_release(void)
{
	if (atomic_dec_and_test(&smp_capture_depth)) {
#ifdef CAPTURE_DEBUG
		printk("CPU[%d]: Giving pardon to "
		       "imprisoned penguins\n",
		       smp_processor_id());
#endif
		penguins_are_doing_time = 0;
1140
		membar_safe("#StoreLoad");
L
Linus Torvalds 已提交
1141 1142 1143 1144
		atomic_dec(&smp_capture_registry);
	}
}

1145 1146
/* Imprisoned penguins run with %pil == PIL_NORMAL_MAX, but PSTATE_IE
 * set, so they can service tlb flush xcalls...
L
Linus Torvalds 已提交
1147 1148
 */
extern void prom_world(int);
1149

L
Linus Torvalds 已提交
1150 1151 1152 1153 1154 1155 1156 1157 1158
void smp_penguin_jailcell(int irq, struct pt_regs *regs)
{
	clear_softint(1 << irq);

	preempt_disable();

	__asm__ __volatile__("flushw");
	prom_world(1);
	atomic_inc(&smp_capture_registry);
1159
	membar_safe("#StoreLoad");
L
Linus Torvalds 已提交
1160
	while (penguins_are_doing_time)
1161
		rmb();
L
Linus Torvalds 已提交
1162 1163 1164 1165 1166 1167 1168 1169 1170
	atomic_dec(&smp_capture_registry);
	prom_world(0);

	preempt_enable();
}

/* /proc/profile writes can call this, don't __init it please. */
int setup_profiling_timer(unsigned int multiplier)
{
1171
	return -EINVAL;
L
Linus Torvalds 已提交
1172 1173 1174 1175 1176 1177
}

void __init smp_prepare_cpus(unsigned int max_cpus)
{
}

1178
void __devinit smp_prepare_boot_cpu(void)
1179 1180 1181
{
}

1182 1183 1184
void __init smp_setup_processor_id(void)
{
	if (tlb_type == spitfire)
1185
		xcall_deliver_impl = spitfire_xcall_deliver;
1186
	else if (tlb_type == cheetah || tlb_type == cheetah_plus)
1187
		xcall_deliver_impl = cheetah_xcall_deliver;
1188
	else
1189
		xcall_deliver_impl = hypervisor_xcall_deliver;
1190 1191
}

1192
void __devinit smp_fill_in_sib_core_maps(void)
L
Linus Torvalds 已提交
1193
{
1194 1195
	unsigned int i;

1196
	for_each_present_cpu(i) {
1197 1198
		unsigned int j;

1199
		cpus_clear(cpu_core_map[i]);
1200
		if (cpu_data(i).core_id == 0) {
1201
			cpu_set(i, cpu_core_map[i]);
1202 1203 1204
			continue;
		}

1205
		for_each_present_cpu(j) {
1206 1207
			if (cpu_data(i).core_id ==
			    cpu_data(j).core_id)
1208 1209 1210 1211
				cpu_set(j, cpu_core_map[i]);
		}
	}

1212
	for_each_present_cpu(i) {
1213 1214
		unsigned int j;

1215
		cpus_clear(per_cpu(cpu_sibling_map, i));
1216
		if (cpu_data(i).proc_id == -1) {
1217
			cpu_set(i, per_cpu(cpu_sibling_map, i));
1218 1219 1220
			continue;
		}

1221
		for_each_present_cpu(j) {
1222 1223
			if (cpu_data(i).proc_id ==
			    cpu_data(j).proc_id)
1224
				cpu_set(j, per_cpu(cpu_sibling_map, i));
1225 1226
		}
	}
L
Linus Torvalds 已提交
1227 1228
}

1229
int __cpuinit __cpu_up(unsigned int cpu)
L
Linus Torvalds 已提交
1230 1231 1232 1233 1234 1235 1236 1237 1238 1239
{
	int ret = smp_boot_one_cpu(cpu);

	if (!ret) {
		cpu_set(cpu, smp_commenced_mask);
		while (!cpu_isset(cpu, cpu_online_map))
			mb();
		if (!cpu_isset(cpu, cpu_online_map)) {
			ret = -ENODEV;
		} else {
1240 1241 1242 1243 1244
			/* On SUN4V, writes to %tick and %stick are
			 * not allowed.
			 */
			if (tlb_type != hypervisor)
				smp_synchronize_one_tick(cpu);
L
Linus Torvalds 已提交
1245 1246 1247 1248 1249
		}
	}
	return ret;
}

1250
#ifdef CONFIG_HOTPLUG_CPU
1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285
void cpu_play_dead(void)
{
	int cpu = smp_processor_id();
	unsigned long pstate;

	idle_task_exit();

	if (tlb_type == hypervisor) {
		struct trap_per_cpu *tb = &trap_block[cpu];

		sun4v_cpu_qconf(HV_CPU_QUEUE_CPU_MONDO,
				tb->cpu_mondo_pa, 0);
		sun4v_cpu_qconf(HV_CPU_QUEUE_DEVICE_MONDO,
				tb->dev_mondo_pa, 0);
		sun4v_cpu_qconf(HV_CPU_QUEUE_RES_ERROR,
				tb->resum_mondo_pa, 0);
		sun4v_cpu_qconf(HV_CPU_QUEUE_NONRES_ERROR,
				tb->nonresum_mondo_pa, 0);
	}

	cpu_clear(cpu, smp_commenced_mask);
	membar_safe("#Sync");

	local_irq_disable();

	__asm__ __volatile__(
		"rdpr	%%pstate, %0\n\t"
		"wrpr	%0, %1, %%pstate"
		: "=r" (pstate)
		: "i" (PSTATE_IE));

	while (1)
		barrier();
}

1286 1287
int __cpu_disable(void)
{
1288 1289 1290 1291 1292 1293 1294 1295
	int cpu = smp_processor_id();
	cpuinfo_sparc *c;
	int i;

	for_each_cpu_mask(i, cpu_core_map[cpu])
		cpu_clear(cpu, cpu_core_map[i]);
	cpus_clear(cpu_core_map[cpu]);

1296 1297 1298
	for_each_cpu_mask(i, per_cpu(cpu_sibling_map, cpu))
		cpu_clear(cpu, per_cpu(cpu_sibling_map, i));
	cpus_clear(per_cpu(cpu_sibling_map, cpu));
1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313

	c = &cpu_data(cpu);

	c->core_id = 0;
	c->proc_id = -1;

	smp_wmb();

	/* Make sure no interrupts point to this cpu.  */
	fixup_irqs();

	local_irq_enable();
	mdelay(1);
	local_irq_disable();

1314 1315 1316 1317
	ipi_call_lock();
	cpu_clear(cpu, cpu_online_map);
	ipi_call_unlock();

1318
	return 0;
1319 1320 1321 1322
}

void __cpu_die(unsigned int cpu)
{
1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350
	int i;

	for (i = 0; i < 100; i++) {
		smp_rmb();
		if (!cpu_isset(cpu, smp_commenced_mask))
			break;
		msleep(100);
	}
	if (cpu_isset(cpu, smp_commenced_mask)) {
		printk(KERN_ERR "CPU %u didn't die...\n", cpu);
	} else {
#if defined(CONFIG_SUN_LDOMS)
		unsigned long hv_err;
		int limit = 100;

		do {
			hv_err = sun4v_cpu_stop(cpu);
			if (hv_err == HV_EOK) {
				cpu_clear(cpu, cpu_present_map);
				break;
			}
		} while (--limit > 0);
		if (limit <= 0) {
			printk(KERN_ERR "sun4v_cpu_stop() fails err=%lu\n",
			       hv_err);
		}
#endif
	}
1351 1352 1353
}
#endif

L
Linus Torvalds 已提交
1354 1355 1356 1357 1358 1359
void __init smp_cpus_done(unsigned int max_cpus)
{
}

void smp_send_reschedule(int cpu)
{
1360 1361 1362 1363 1364 1365 1366
	xcall_deliver((u64) &xcall_receive_signal, 0, 0,
		      &cpumask_of_cpu(cpu));
}

void smp_receive_signal_client(int irq, struct pt_regs *regs)
{
	clear_softint(1 << irq);
L
Linus Torvalds 已提交
1367 1368 1369 1370 1371 1372 1373 1374 1375
}

/* This is a nop because we capture all other cpus
 * anyways when making the PROM active.
 */
void smp_send_stop(void)
{
}

1376 1377
unsigned long __per_cpu_base __read_mostly;
unsigned long __per_cpu_shift __read_mostly;
L
Linus Torvalds 已提交
1378 1379 1380 1381

EXPORT_SYMBOL(__per_cpu_base);
EXPORT_SYMBOL(__per_cpu_shift);

1382
void __init real_setup_per_cpu_areas(void)
L
Linus Torvalds 已提交
1383
{
1384
	unsigned long paddr, goal, size, i;
L
Linus Torvalds 已提交
1385 1386 1387
	char *ptr;

	/* Copy section for each CPU (we discard the original) */
1388 1389
	goal = PERCPU_ENOUGH_ROOM;

1390 1391
	__per_cpu_shift = PAGE_SHIFT;
	for (size = PAGE_SIZE; size < goal; size <<= 1UL)
L
Linus Torvalds 已提交
1392 1393
		__per_cpu_shift++;

1394 1395 1396 1397 1398
	paddr = lmb_alloc(size * NR_CPUS, PAGE_SIZE);
	if (!paddr) {
		prom_printf("Cannot allocate per-cpu memory.\n");
		prom_halt();
	}
L
Linus Torvalds 已提交
1399

1400
	ptr = __va(paddr);
L
Linus Torvalds 已提交
1401 1402 1403 1404
	__per_cpu_base = ptr - __per_cpu_start;

	for (i = 0; i < NR_CPUS; i++, ptr += size)
		memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
1405 1406 1407

	/* Setup %g5 for the boot cpu.  */
	__local_per_cpu_offset = __per_cpu_offset(smp_processor_id());
L
Linus Torvalds 已提交
1408
}