smp_64.c 35.6 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
#include <linux/export.h>
L
Linus Torvalds 已提交
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
#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/bootmem.h>
24
#include <linux/vmalloc.h>
25
#include <linux/ftrace.h>
26
#include <linux/cpu.h>
27
#include <linux/slab.h>
L
Linus Torvalds 已提交
28 29 30

#include <asm/head.h>
#include <asm/ptrace.h>
A
Arun Sharma 已提交
31
#include <linux/atomic.h>
L
Linus Torvalds 已提交
32 33 34
#include <asm/tlbflush.h>
#include <asm/mmu_context.h>
#include <asm/cpudata.h>
D
David S. Miller 已提交
35 36
#include <asm/hvtramp.h>
#include <asm/io.h>
37
#include <asm/timer.h>
L
Linus Torvalds 已提交
38 39

#include <asm/irq.h>
A
Al Viro 已提交
40
#include <asm/irq_regs.h>
L
Linus Torvalds 已提交
41 42 43 44 45 46
#include <asm/page.h>
#include <asm/pgtable.h>
#include <asm/oplib.h>
#include <asm/uaccess.h>
#include <asm/starfire.h>
#include <asm/tlb.h>
47
#include <asm/sections.h>
48
#include <asm/prom.h>
49
#include <asm/mdesc.h>
50
#include <asm/ldc.h>
51
#include <asm/hypervisor.h>
52
#include <asm/pcr.h>
L
Linus Torvalds 已提交
53

54 55
#include "cpumap.h"

56 57
int sparc64_multi_core __read_mostly;

58
DEFINE_PER_CPU(cpumask_t, cpu_sibling_map) = CPU_MASK_NONE;
59 60
cpumask_t cpu_core_map[NR_CPUS] __read_mostly =
	{ [0 ... NR_CPUS-1] = CPU_MASK_NONE };
61

62
EXPORT_PER_CPU_SYMBOL(cpu_sibling_map);
63 64
EXPORT_SYMBOL(cpu_core_map);

L
Linus Torvalds 已提交
65 66 67 68 69 70 71
static cpumask_t smp_commenced_mask;

void smp_info(struct seq_file *m)
{
	int i;
	
	seq_printf(m, "State:\n");
72 73
	for_each_online_cpu(i)
		seq_printf(m, "CPU%d:\t\tonline\n", i);
L
Linus Torvalds 已提交
74 75 76 77 78 79
}

void smp_bogo(struct seq_file *m)
{
	int i;
	
80 81 82 83
	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
84 85
}

86
extern void setup_sparc64_timer(void);
L
Linus Torvalds 已提交
87 88 89

static volatile unsigned long callin_flag = 0;

90
void __cpuinit smp_callin(void)
L
Linus Torvalds 已提交
91 92 93
{
	int cpuid = hard_smp_processor_id();

94
	__local_per_cpu_offset = __per_cpu_offset(cpuid);
L
Linus Torvalds 已提交
95

96
	if (tlb_type == hypervisor)
97
		sun4v_ktsb_register();
98

99
	__flush_tlb_all();
L
Linus Torvalds 已提交
100

101
	setup_sparc64_timer();
L
Linus Torvalds 已提交
102

103 104 105
	if (cheetah_pcache_forced_on)
		cheetah_enable_pcache();

L
Linus Torvalds 已提交
106 107 108 109 110 111 112
	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...
	 */
113
	current_thread_info()->new_child = 0;
L
Linus Torvalds 已提交
114 115 116 117 118

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

119 120 121
	/* inform the notifiers about the new cpu */
	notify_cpu_starting(cpuid);

122
	while (!cpumask_test_cpu(cpuid, &smp_commenced_mask))
123
		rmb();
L
Linus Torvalds 已提交
124

125
	set_cpu_online(cpuid, true);
126
	local_irq_enable();
127 128 129

	/* idle thread is expected to have preempt disabled */
	preempt_disable();
L
Linus Torvalds 已提交
130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164
}

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;
165
		membar_safe("#StoreLoad");
L
Linus Torvalds 已提交
166
		while (!(tm = go[SLAVE]))
167
			rmb();
L
Linus Torvalds 已提交
168
		go[SLAVE] = 0;
169
		wmb();
L
Linus Torvalds 已提交
170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188
		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;
189
	unsigned long flags, rt, master_time_stamp;
L
Linus Torvalds 已提交
190 191 192 193 194 195 196 197 198 199 200 201
#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])
202
		rmb();
L
Linus Torvalds 已提交
203 204 205 206 207

	local_irq_save(flags);
	{
		for (i = 0; i < NUM_ROUNDS; i++) {
			delta = get_delta(&rt, &master_time_stamp);
208
			if (delta == 0)
L
Linus Torvalds 已提交
209 210 211 212 213 214 215 216 217
				done = 1;	/* let's lock on to this... */

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

218
				tick_ops->add_tick(adj);
L
Linus Torvalds 已提交
219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235
			}
#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
236 237 238
	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
239 240 241 242 243 244 245 246 247 248 249 250 251 252
}

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])
253
		rmb();
L
Linus Torvalds 已提交
254 255 256

	/* now let the client proceed into his loop */
	go[MASTER] = 0;
257
	membar_safe("#StoreLoad");
L
Linus Torvalds 已提交
258 259 260 261 262

	spin_lock_irqsave(&itc_sync_lock, flags);
	{
		for (i = 0; i < NUM_ROUNDS*NUM_ITERS; i++) {
			while (!go[MASTER])
263
				rmb();
L
Linus Torvalds 已提交
264
			go[MASTER] = 0;
265
			wmb();
L
Linus Torvalds 已提交
266
			go[SLAVE] = tick_ops->get_tick();
267
			membar_safe("#StoreLoad");
L
Linus Torvalds 已提交
268 269 270 271 272
		}
	}
	spin_unlock_irqrestore(&itc_sync_lock, flags);
}

273
#if defined(CONFIG_SUN_LDOMS) && defined(CONFIG_HOTPLUG_CPU)
D
David S. Miller 已提交
274 275 276 277 278 279 280 281
/* 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);
}

282
static void __cpuinit ldom_startcpu_cpuid(unsigned int cpu, unsigned long thread_reg, void **descrp)
283 284 285 286 287 288 289 290
{
	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;
291
	int i;
292

293 294 295 296
	hdesc = kzalloc(sizeof(*hdesc) +
			(sizeof(struct hvtramp_mapping) *
			 num_kernel_image_mappings - 1),
			GFP_KERNEL);
297
	if (!hdesc) {
D
David S. Miller 已提交
298
		printk(KERN_ERR "ldom_startcpu_cpuid: Cannot allocate "
299 300 301
		       "hvtramp_descr.\n");
		return;
	}
302
	*descrp = hdesc;
303 304

	hdesc->cpu = cpu;
305
	hdesc->num_mappings = num_kernel_image_mappings;
306 307 308 309 310 311 312 313 314 315 316

	tb = &trap_block[cpu];

	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;

317 318 319
	for (i = 0; i < hdesc->num_mappings; i++) {
		hdesc->maps[i].vaddr = tte_vaddr;
		hdesc->maps[i].tte   = tte_data;
320 321 322 323 324 325 326 327 328
		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));
329 330 331
	if (hv_err)
		printk(KERN_ERR "ldom_startcpu_cpuid: sun4v_cpu_start() "
		       "gives error %lu\n", hv_err);
332 333 334
}
#endif

L
Linus Torvalds 已提交
335 336 337 338 339 340 341 342
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;

343
static int __cpuinit smp_boot_one_cpu(unsigned int cpu, struct task_struct *idle)
L
Linus Torvalds 已提交
344 345 346 347 348
{
	unsigned long entry =
		(unsigned long)(&sparc64_cpu_startup);
	unsigned long cookie =
		(unsigned long)(&cpu_new_thread);
349
	void *descr = NULL;
350
	int timeout, ret;
L
Linus Torvalds 已提交
351 352

	callin_flag = 0;
353
	cpu_new_thread = task_thread_info(idle);
L
Linus Torvalds 已提交
354

355
	if (tlb_type == hypervisor) {
356
#if defined(CONFIG_SUN_LDOMS) && defined(CONFIG_HOTPLUG_CPU)
357 358
		if (ldom_domaining_enabled)
			ldom_startcpu_cpuid(cpu,
359 360
					    (unsigned long) cpu_new_thread,
					    &descr);
361 362 363
		else
#endif
			prom_startcpu_cpuid(cpu, entry, cookie);
364
	} else {
365
		struct device_node *dp = of_find_node_by_cpuid(cpu);
366

367
		prom_startcpu(dp->phandle, entry, cookie);
368
	}
L
Linus Torvalds 已提交
369

370
	for (timeout = 0; timeout < 50000; timeout++) {
L
Linus Torvalds 已提交
371 372 373 374
		if (callin_flag)
			break;
		udelay(100);
	}
375

L
Linus Torvalds 已提交
376 377 378 379 380 381 382 383
	if (callin_flag) {
		ret = 0;
	} else {
		printk("Processor %d is stuck.\n", cpu);
		ret = -ENODEV;
	}
	cpu_new_thread = NULL;

384
	kfree(descr);
385

L
Linus Torvalds 已提交
386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446
	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) {
447
		printk("CPU[%d]: mondo stuckage result[%016llx]\n",
L
Linus Torvalds 已提交
448 449 450 451 452 453 454
		       smp_processor_id(), result);
	} else {
		udelay(2);
		goto again;
	}
}

455
static void spitfire_xcall_deliver(struct trap_per_cpu *tb, int cnt)
L
Linus Torvalds 已提交
456
{
457 458
	u64 *mondo, data0, data1, data2;
	u16 *cpu_list;
L
Linus Torvalds 已提交
459 460 461 462
	u64 pstate;
	int i;

	__asm__ __volatile__("rdpr %%pstate, %0" : "=r" (pstate));
463 464 465 466 467 468 469
	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
470 471 472 473 474 475
}

/* 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).
 */
476
static void cheetah_xcall_deliver(struct trap_per_cpu *tb, int cnt)
L
Linus Torvalds 已提交
477
{
478
	int nack_busy_id, is_jbus, need_more;
479 480
	u64 *mondo, pstate, ver, busy_mask;
	u16 *cpu_list;
L
Linus Torvalds 已提交
481

482 483
	cpu_list = __va(tb->cpu_list_pa);
	mondo = __va(tb->cpu_mondo_block_pa);
484

L
Linus Torvalds 已提交
485 486 487 488 489
	/* 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));
490 491
	is_jbus = ((ver >> 32) == __JALAPENO_ID ||
		   (ver >> 32) == __SERRANO_ID);
L
Linus Torvalds 已提交
492 493 494 495

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

retry:
496
	need_more = 0;
L
Linus Torvalds 已提交
497 498 499 500 501 502 503 504 505
	__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 */
506
			     : "r" (mondo[0]), "r" (mondo[1]), "r" (mondo[2]),
L
Linus Torvalds 已提交
507 508 509 510
			       "r" (0x40), "r" (0x50), "r" (0x60),
			       "i" (ASI_INTR_W));

	nack_busy_id = 0;
511
	busy_mask = 0;
L
Linus Torvalds 已提交
512 513 514
	{
		int i;

515 516 517 518 519 520
		for (i = 0; i < cnt; i++) {
			u64 target, nr;

			nr = cpu_list[i];
			if (nr == 0xffff)
				continue;
L
Linus Torvalds 已提交
521

522
			target = (nr << 14) | 0x70;
523
			if (is_jbus) {
524
				busy_mask |= (0x1UL << (nr * 2));
525
			} else {
L
Linus Torvalds 已提交
526
				target |= (nack_busy_id << 24);
527 528 529
				busy_mask |= (0x1UL <<
					      (nack_busy_id * 2));
			}
L
Linus Torvalds 已提交
530 531 532 533 534 535
			__asm__ __volatile__(
				"stxa	%%g0, [%0] %1\n\t"
				"membar	#Sync\n\t"
				: /* no outputs */
				: "r" (target), "i" (ASI_INTR_W));
			nack_busy_id++;
536 537 538 539
			if (nack_busy_id == 32) {
				need_more = 1;
				break;
			}
L
Linus Torvalds 已提交
540 541 542 543 544
		}
	}

	/* Now, poll for completion. */
	{
545
		u64 dispatch_stat, nack_mask;
L
Linus Torvalds 已提交
546 547 548
		long stuck;

		stuck = 100000 * nack_busy_id;
549
		nack_mask = busy_mask << 1;
L
Linus Torvalds 已提交
550 551 552 553
		do {
			__asm__ __volatile__("ldxa	[%%g0] %1, %0"
					     : "=r" (dispatch_stat)
					     : "i" (ASI_INTR_DISPATCH_STAT));
554
			if (!(dispatch_stat & (busy_mask | nack_mask))) {
L
Linus Torvalds 已提交
555 556
				__asm__ __volatile__("wrpr %0, 0x0, %%pstate"
						     : : "r" (pstate));
557
				if (unlikely(need_more)) {
558 559 560 561 562 563 564
					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)
565 566 567 568
							break;
					}
					goto retry;
				}
L
Linus Torvalds 已提交
569 570 571 572
				return;
			}
			if (!--stuck)
				break;
573
		} while (dispatch_stat & busy_mask);
L
Linus Torvalds 已提交
574 575 576 577

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

578
		if (dispatch_stat & busy_mask) {
L
Linus Torvalds 已提交
579 580 581
			/* Busy bits will not clear, continue instead
			 * of freezing up on this cpu.
			 */
582
			printk("CPU[%d]: mondo stuckage result[%016llx]\n",
L
Linus Torvalds 已提交
583 584 585 586 587 588 589 590 591 592 593 594
			       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.
			 */
595 596 597 598 599 600
			for (i = 0; i < cnt; i++) {
				u64 check_mask, nr;

				nr = cpu_list[i];
				if (nr == 0xffff)
					continue;
L
Linus Torvalds 已提交
601

602
				if (is_jbus)
603
					check_mask = (0x2UL << (2*nr));
L
Linus Torvalds 已提交
604 605 606 607
				else
					check_mask = (0x2UL <<
						      this_busy_nack);
				if ((dispatch_stat & check_mask) == 0)
608
					cpu_list[i] = 0xffff;
L
Linus Torvalds 已提交
609
				this_busy_nack += 2;
610 611
				if (this_busy_nack == 64)
					break;
L
Linus Torvalds 已提交
612 613 614 615 616 617 618
			}

			goto retry;
		}
	}
}

619
/* Multi-cpu list version.  */
620
static void hypervisor_xcall_deliver(struct trap_per_cpu *tb, int cnt)
621
{
622
	int retries, this_cpu, prev_sent, i, saw_cpu_error;
623
	unsigned long status;
624
	u16 *cpu_list;
625

626
	this_cpu = smp_processor_id();
627

628 629
	cpu_list = __va(tb->cpu_list_pa);

630
	saw_cpu_error = 0;
631
	retries = 0;
632
	prev_sent = 0;
633
	do {
634
		int forward_progress, n_sent;
635

636 637 638 639 640 641
		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))
642
			break;
643

644 645 646 647
		/* First, see if we made any forward progress.
		 *
		 * The hypervisor indicates successful sends by setting
		 * cpu list entries to the value 0xffff.
648
		 */
649
		n_sent = 0;
650
		for (i = 0; i < cnt; i++) {
651 652
			if (likely(cpu_list[i] == 0xffff))
				n_sent++;
653 654
		}

655 656 657 658 659 660
		forward_progress = 0;
		if (n_sent > prev_sent)
			forward_progress = 1;

		prev_sent = n_sent;

661 662 663 664 665 666 667 668 669 670 671 672 673 674
		/* 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);
675 676
				if (err == HV_CPU_STATE_ERROR) {
					saw_cpu_error = (cpu + 1);
677
					cpu_list[i] = 0xffff;
678 679 680 681 682
				}
			}
		} else if (unlikely(status != HV_EWOULDBLOCK))
			goto fatal_mondo_error;

683 684 685 686 687 688 689
		/* 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.
		 */
690 691 692 693 694 695 696 697 698
		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);
		}
699 700
	} while (1);

701
	if (unlikely(saw_cpu_error))
702 703 704 705 706 707
		goto fatal_mondo_cpu_error;

	return;

fatal_mondo_cpu_error:
	printk(KERN_CRIT "CPU[%d]: SUN4V mondo cpu error, some target cpus "
708 709
	       "(including %d) were in error state\n",
	       this_cpu, saw_cpu_error - 1);
710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729
	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");
730
}
731

732
static void (*xcall_deliver_impl)(struct trap_per_cpu *, int);
733 734 735

static void xcall_deliver(u64 data0, u64 data1, u64 data2, const cpumask_t *mask)
{
736 737
	struct trap_per_cpu *tb;
	int this_cpu, i, cnt;
738
	unsigned long flags;
739 740
	u16 *cpu_list;
	u64 *mondo;
741 742 743 744 745 746 747 748 749 750 751 752

	/* 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);
753 754 755 756 757 758 759 760 761 762 763 764 765 766

	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;
767
	for_each_cpu(i, mask) {
768 769 770 771 772 773 774 775
		if (i == this_cpu || !cpu_online(i))
			continue;
		cpu_list[cnt++] = i;
	}

	if (cnt)
		xcall_deliver_impl(tb, cnt);

776
	local_irq_restore(flags);
777
}
778

779 780
/* Send cross call to all processors mentioned in MASK_P
 * except self.  Really, there are only two cases currently,
781
 * "cpu_online_mask" and "mm_cpumask(mm)".
L
Linus Torvalds 已提交
782
 */
783
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
784 785 786
{
	u64 data0 = (((u64)ctx)<<32 | (((u64)func) & 0xffffffff));

787 788
	xcall_deliver(data0, data1, data2, mask);
}
L
Linus Torvalds 已提交
789

790 791 792
/* Send cross call to all processors except self. */
static void smp_cross_call(unsigned long *func, u32 ctx, u64 data1, u64 data2)
{
793
	smp_cross_call_masked(func, ctx, data1, data2, cpu_online_mask);
L
Linus Torvalds 已提交
794 795 796 797 798 799
}

extern unsigned long xcall_sync_tick;

static void smp_start_sync_tick_client(int cpu)
{
800
	xcall_deliver((u64) &xcall_sync_tick, 0, 0,
801
		      cpumask_of(cpu));
L
Linus Torvalds 已提交
802 803 804 805
}

extern unsigned long xcall_call_function;

806
void arch_send_call_function_ipi_mask(const struct cpumask *mask)
L
Linus Torvalds 已提交
807
{
808
	xcall_deliver((u64) &xcall_call_function, 0, 0, mask);
809
}
L
Linus Torvalds 已提交
810

811
extern unsigned long xcall_call_function_single;
L
Linus Torvalds 已提交
812

813 814
void arch_send_call_function_single_ipi(int cpu)
{
815
	xcall_deliver((u64) &xcall_call_function_single, 0, 0,
816
		      cpumask_of(cpu));
L
Linus Torvalds 已提交
817 818
}

819
void __irq_entry smp_call_function_client(int irq, struct pt_regs *regs)
L
Linus Torvalds 已提交
820
{
821 822 823
	clear_softint(1 << irq);
	generic_smp_call_function_interrupt();
}
L
Linus Torvalds 已提交
824

825
void __irq_entry smp_call_function_single_client(int irq, struct pt_regs *regs)
826
{
L
Linus Torvalds 已提交
827
	clear_softint(1 << irq);
828
	generic_smp_call_function_single_interrupt();
L
Linus Torvalds 已提交
829 830
}

831 832
static void tsb_sync(void *info)
{
833
	struct trap_per_cpu *tp = &trap_block[raw_smp_processor_id()];
834 835
	struct mm_struct *mm = info;

836
	/* It is not valid to test "current->active_mm == mm" here.
837 838 839 840 841 842
	 *
	 * 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))
843 844 845 846 847
		tsb_context_switch(mm);
}

void smp_tsb_sync(struct mm_struct *mm)
{
848
	smp_call_function_many(mm_cpumask(mm), tsb_sync, mm, 1);
849 850
}

L
Linus Torvalds 已提交
851
extern unsigned long xcall_flush_tlb_mm;
852
extern unsigned long xcall_flush_tlb_page;
L
Linus Torvalds 已提交
853
extern unsigned long xcall_flush_tlb_kernel_range;
854
extern unsigned long xcall_fetch_glob_regs;
855 856
extern unsigned long xcall_fetch_glob_pmu;
extern unsigned long xcall_fetch_glob_pmu_n4;
L
Linus Torvalds 已提交
857
extern unsigned long xcall_receive_signal;
858
extern unsigned long xcall_new_mmu_context_version;
D
David S. Miller 已提交
859 860 861
#ifdef CONFIG_KGDB
extern unsigned long xcall_kgdb_capture;
#endif
L
Linus Torvalds 已提交
862 863 864 865 866 867 868 869 870 871 872

#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
873
static inline void __local_flush_dcache_page(struct page *page)
L
Linus Torvalds 已提交
874 875 876 877 878 879 880 881 882 883 884 885 886 887
{
#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)
{
888 889 890 891
	int this_cpu;

	if (tlb_type == hypervisor)
		return;
L
Linus Torvalds 已提交
892 893 894 895

#ifdef CONFIG_DEBUG_DCFLUSH
	atomic_inc(&dcpage_flushes);
#endif
896 897 898

	this_cpu = get_cpu();

L
Linus Torvalds 已提交
899 900 901 902
	if (cpu == this_cpu) {
		__local_flush_dcache_page(page);
	} else if (cpu_online(cpu)) {
		void *pg_addr = page_address(page);
903
		u64 data0 = 0;
L
Linus Torvalds 已提交
904 905

		if (tlb_type == spitfire) {
906
			data0 = ((u64)&xcall_flush_dcache_page_spitfire);
L
Linus Torvalds 已提交
907 908
			if (page_mapping(page) != NULL)
				data0 |= ((u64)1 << 32);
909
		} else if (tlb_type == cheetah || tlb_type == cheetah_plus) {
L
Linus Torvalds 已提交
910
#ifdef DCACHE_ALIASING_POSSIBLE
911
			data0 =	((u64)&xcall_flush_dcache_page_cheetah);
L
Linus Torvalds 已提交
912 913
#endif
		}
914 915
		if (data0) {
			xcall_deliver(data0, __pa(pg_addr),
916
				      (u64) pg_addr, cpumask_of(cpu));
L
Linus Torvalds 已提交
917
#ifdef CONFIG_DEBUG_DCFLUSH
918
			atomic_inc(&dcpage_flushes_xcall);
L
Linus Torvalds 已提交
919
#endif
920
		}
L
Linus Torvalds 已提交
921 922 923 924 925 926 927
	}

	put_cpu();
}

void flush_dcache_page_all(struct mm_struct *mm, struct page *page)
{
928 929
	void *pg_addr;
	u64 data0;
930 931 932 933

	if (tlb_type == hypervisor)
		return;

934
	preempt_disable();
L
Linus Torvalds 已提交
935 936 937 938

#ifdef CONFIG_DEBUG_DCFLUSH
	atomic_inc(&dcpage_flushes);
#endif
939 940
	data0 = 0;
	pg_addr = page_address(page);
L
Linus Torvalds 已提交
941 942 943 944
	if (tlb_type == spitfire) {
		data0 = ((u64)&xcall_flush_dcache_page_spitfire);
		if (page_mapping(page) != NULL)
			data0 |= ((u64)1 << 32);
945
	} else if (tlb_type == cheetah || tlb_type == cheetah_plus) {
L
Linus Torvalds 已提交
946 947 948 949
#ifdef DCACHE_ALIASING_POSSIBLE
		data0 = ((u64)&xcall_flush_dcache_page_cheetah);
#endif
	}
950 951
	if (data0) {
		xcall_deliver(data0, __pa(pg_addr),
952
			      (u64) pg_addr, cpu_online_mask);
L
Linus Torvalds 已提交
953
#ifdef CONFIG_DEBUG_DCFLUSH
954
		atomic_inc(&dcpage_flushes_xcall);
L
Linus Torvalds 已提交
955
#endif
956
	}
L
Linus Torvalds 已提交
957 958
	__local_flush_dcache_page(page);

959
	preempt_enable();
L
Linus Torvalds 已提交
960 961
}

962
void __irq_entry smp_new_mmu_context_version_client(int irq, struct pt_regs *regs)
L
Linus Torvalds 已提交
963
{
964
	struct mm_struct *mm;
965
	unsigned long flags;
966

L
Linus Torvalds 已提交
967
	clear_softint(1 << irq);
968 969 970 971 972

	/* 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;
973 974
	if (unlikely(!mm || (mm == &init_mm)))
		return;
975

976
	spin_lock_irqsave(&mm->context.lock, flags);
977

978 979
	if (unlikely(!CTX_VALID(mm->context)))
		get_new_mmu_context(mm);
980

981
	spin_unlock_irqrestore(&mm->context.lock, flags);
982

983 984 985
	load_secondary_context(mm);
	__flush_tlb_mm(CTX_HWBITS(mm->context),
		       SECONDARY_CONTEXT);
986 987 988 989
}

void smp_new_mmu_context_version(void)
{
990
	smp_cross_call(&xcall_new_mmu_context_version, 0, 0, 0);
L
Linus Torvalds 已提交
991 992
}

D
David S. Miller 已提交
993 994 995 996 997 998 999
#ifdef CONFIG_KGDB
void kgdb_roundup_cpus(unsigned long flags)
{
	smp_cross_call(&xcall_kgdb_capture, 0, 0, 0);
}
#endif

1000 1001 1002 1003 1004
void smp_fetch_global_regs(void)
{
	smp_cross_call(&xcall_fetch_glob_regs, 0, 0, 0);
}

1005 1006 1007 1008 1009 1010 1011 1012 1013
void smp_fetch_global_pmu(void)
{
	if (tlb_type == hypervisor &&
	    sun4v_chip_type >= SUN4V_CHIP_NIAGARA4)
		smp_cross_call(&xcall_fetch_glob_pmu_n4, 0, 0, 0);
	else
		smp_cross_call(&xcall_fetch_glob_pmu, 0, 0, 0);
}

L
Linus Torvalds 已提交
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
/* 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
1039
 *    processor's in current->mm->cpu_vm_mask and performing the
L
Linus Torvalds 已提交
1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051
 *    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).
 */
1052 1053 1054 1055 1056

/* 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
1057 1058
void smp_flush_tlb_mm(struct mm_struct *mm)
{
1059 1060
	u32 ctx = CTX_HWBITS(mm->context);
	int cpu = get_cpu();
L
Linus Torvalds 已提交
1061

1062
	if (atomic_read(&mm->mm_users) == 1) {
1063
		cpumask_copy(mm_cpumask(mm), cpumask_of(cpu));
1064 1065
		goto local_flush_and_out;
	}
L
Linus Torvalds 已提交
1066

1067 1068
	smp_cross_call_masked(&xcall_flush_tlb_mm,
			      ctx, 0, 0,
1069
			      mm_cpumask(mm));
L
Linus Torvalds 已提交
1070

1071 1072
local_flush_and_out:
	__flush_tlb_mm(ctx, SECONDARY_CONTEXT);
L
Linus Torvalds 已提交
1073

1074
	put_cpu();
L
Linus Torvalds 已提交
1075 1076
}

1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089
struct tlb_pending_info {
	unsigned long ctx;
	unsigned long nr;
	unsigned long *vaddrs;
};

static void tlb_pending_func(void *info)
{
	struct tlb_pending_info *t = info;

	__flush_tlb_pending(t->ctx, t->nr, t->vaddrs);
}

L
Linus Torvalds 已提交
1090 1091 1092
void smp_flush_tlb_pending(struct mm_struct *mm, unsigned long nr, unsigned long *vaddrs)
{
	u32 ctx = CTX_HWBITS(mm->context);
1093
	struct tlb_pending_info info;
L
Linus Torvalds 已提交
1094 1095
	int cpu = get_cpu();

1096 1097 1098 1099
	info.ctx = ctx;
	info.nr = nr;
	info.vaddrs = vaddrs;

1100
	if (mm == current->mm && atomic_read(&mm->mm_users) == 1)
1101
		cpumask_copy(mm_cpumask(mm), cpumask_of(cpu));
H
Hugh Dickins 已提交
1102
	else
1103 1104
		smp_call_function_many(mm_cpumask(mm), tlb_pending_func,
				       &info, 1);
L
Linus Torvalds 已提交
1105 1106 1107 1108 1109 1110

	__flush_tlb_pending(ctx, nr, vaddrs);

	put_cpu();
}

1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126
void smp_flush_tlb_page(struct mm_struct *mm, unsigned long vaddr)
{
	unsigned long context = CTX_HWBITS(mm->context);
	int cpu = get_cpu();

	if (mm == current->mm && atomic_read(&mm->mm_users) == 1)
		cpumask_copy(mm_cpumask(mm), cpumask_of(cpu));
	else
		smp_cross_call_masked(&xcall_flush_tlb_page,
				      context, vaddr, 0,
				      mm_cpumask(mm));
	__flush_tlb_page(context, vaddr);

	put_cpu();
}

L
Linus Torvalds 已提交
1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161
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)
1162
			rmb();
L
Linus Torvalds 已提交
1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177
#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;
1178
		membar_safe("#StoreLoad");
L
Linus Torvalds 已提交
1179 1180 1181 1182
		atomic_dec(&smp_capture_registry);
	}
}

1183 1184
/* 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
1185 1186
 */
extern void prom_world(int);
1187

1188
void __irq_entry smp_penguin_jailcell(int irq, struct pt_regs *regs)
L
Linus Torvalds 已提交
1189 1190 1191 1192 1193 1194 1195 1196
{
	clear_softint(1 << irq);

	preempt_disable();

	__asm__ __volatile__("flushw");
	prom_world(1);
	atomic_inc(&smp_capture_registry);
1197
	membar_safe("#StoreLoad");
L
Linus Torvalds 已提交
1198
	while (penguins_are_doing_time)
1199
		rmb();
L
Linus Torvalds 已提交
1200 1201 1202 1203 1204 1205 1206 1207 1208
	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)
{
1209
	return -EINVAL;
L
Linus Torvalds 已提交
1210 1211 1212 1213 1214 1215
}

void __init smp_prepare_cpus(unsigned int max_cpus)
{
}

1216
void smp_prepare_boot_cpu(void)
1217 1218 1219
{
}

1220 1221 1222
void __init smp_setup_processor_id(void)
{
	if (tlb_type == spitfire)
1223
		xcall_deliver_impl = spitfire_xcall_deliver;
1224
	else if (tlb_type == cheetah || tlb_type == cheetah_plus)
1225
		xcall_deliver_impl = cheetah_xcall_deliver;
1226
	else
1227
		xcall_deliver_impl = hypervisor_xcall_deliver;
1228 1229
}

1230
void smp_fill_in_sib_core_maps(void)
L
Linus Torvalds 已提交
1231
{
1232 1233
	unsigned int i;

1234
	for_each_present_cpu(i) {
1235 1236
		unsigned int j;

1237
		cpumask_clear(&cpu_core_map[i]);
1238
		if (cpu_data(i).core_id == 0) {
1239
			cpumask_set_cpu(i, &cpu_core_map[i]);
1240 1241 1242
			continue;
		}

1243
		for_each_present_cpu(j) {
1244 1245
			if (cpu_data(i).core_id ==
			    cpu_data(j).core_id)
1246
				cpumask_set_cpu(j, &cpu_core_map[i]);
1247 1248 1249
		}
	}

1250
	for_each_present_cpu(i) {
1251 1252
		unsigned int j;

1253
		cpumask_clear(&per_cpu(cpu_sibling_map, i));
1254
		if (cpu_data(i).proc_id == -1) {
1255
			cpumask_set_cpu(i, &per_cpu(cpu_sibling_map, i));
1256 1257 1258
			continue;
		}

1259
		for_each_present_cpu(j) {
1260 1261
			if (cpu_data(i).proc_id ==
			    cpu_data(j).proc_id)
1262
				cpumask_set_cpu(j, &per_cpu(cpu_sibling_map, i));
1263 1264
		}
	}
L
Linus Torvalds 已提交
1265 1266
}

1267
int __cpuinit __cpu_up(unsigned int cpu, struct task_struct *tidle)
L
Linus Torvalds 已提交
1268
{
1269
	int ret = smp_boot_one_cpu(cpu, tidle);
L
Linus Torvalds 已提交
1270 1271

	if (!ret) {
1272 1273
		cpumask_set_cpu(cpu, &smp_commenced_mask);
		while (!cpu_online(cpu))
L
Linus Torvalds 已提交
1274
			mb();
1275
		if (!cpu_online(cpu)) {
L
Linus Torvalds 已提交
1276 1277
			ret = -ENODEV;
		} else {
1278 1279 1280 1281 1282
			/* 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
1283 1284 1285 1286 1287
		}
	}
	return ret;
}

1288
#ifdef CONFIG_HOTPLUG_CPU
1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308
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);
	}

1309
	cpumask_clear_cpu(cpu, &smp_commenced_mask);
1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323
	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();
}

1324 1325
int __cpu_disable(void)
{
1326 1327 1328 1329
	int cpu = smp_processor_id();
	cpuinfo_sparc *c;
	int i;

1330 1331 1332
	for_each_cpu(i, &cpu_core_map[cpu])
		cpumask_clear_cpu(cpu, &cpu_core_map[i]);
	cpumask_clear(&cpu_core_map[cpu]);
1333

1334 1335 1336
	for_each_cpu(i, &per_cpu(cpu_sibling_map, cpu))
		cpumask_clear_cpu(cpu, &per_cpu(cpu_sibling_map, i));
	cpumask_clear(&per_cpu(cpu_sibling_map, cpu));
1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351

	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();

1352
	set_cpu_online(cpu, false);
1353

1354 1355
	cpu_map_rebuild();

1356
	return 0;
1357 1358 1359 1360
}

void __cpu_die(unsigned int cpu)
{
1361 1362 1363 1364
	int i;

	for (i = 0; i < 100; i++) {
		smp_rmb();
1365
		if (!cpumask_test_cpu(cpu, &smp_commenced_mask))
1366 1367 1368
			break;
		msleep(100);
	}
1369
	if (cpumask_test_cpu(cpu, &smp_commenced_mask)) {
1370 1371 1372 1373 1374 1375 1376 1377 1378
		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) {
1379
				set_cpu_present(cpu, false);
1380 1381 1382 1383 1384 1385 1386 1387 1388
				break;
			}
		} while (--limit > 0);
		if (limit <= 0) {
			printk(KERN_ERR "sun4v_cpu_stop() fails err=%lu\n",
			       hv_err);
		}
#endif
	}
1389 1390 1391
}
#endif

L
Linus Torvalds 已提交
1392 1393
void __init smp_cpus_done(unsigned int max_cpus)
{
1394
	pcr_arch_init();
L
Linus Torvalds 已提交
1395 1396 1397 1398
}

void smp_send_reschedule(int cpu)
{
1399
	xcall_deliver((u64) &xcall_receive_signal, 0, 0,
1400
		      cpumask_of(cpu));
1401 1402
}

1403
void __irq_entry smp_receive_signal_client(int irq, struct pt_regs *regs)
1404 1405
{
	clear_softint(1 << irq);
1406
	scheduler_ipi();
L
Linus Torvalds 已提交
1407 1408 1409 1410 1411 1412 1413 1414 1415
}

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

1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428
/**
 * pcpu_alloc_bootmem - NUMA friendly alloc_bootmem wrapper for percpu
 * @cpu: cpu to allocate for
 * @size: size allocation in bytes
 * @align: alignment
 *
 * Allocate @size bytes aligned at @align for cpu @cpu.  This wrapper
 * does the right thing for NUMA regardless of the current
 * configuration.
 *
 * RETURNS:
 * Pointer to the allocated area on success, NULL on failure.
 */
1429 1430
static void * __init pcpu_alloc_bootmem(unsigned int cpu, size_t size,
					size_t align)
1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454
{
	const unsigned long goal = __pa(MAX_DMA_ADDRESS);
#ifdef CONFIG_NEED_MULTIPLE_NODES
	int node = cpu_to_node(cpu);
	void *ptr;

	if (!node_online(node) || !NODE_DATA(node)) {
		ptr = __alloc_bootmem(size, align, goal);
		pr_info("cpu %d has no node %d or node-local memory\n",
			cpu, node);
		pr_debug("per cpu data for cpu%d %lu bytes at %016lx\n",
			 cpu, size, __pa(ptr));
	} else {
		ptr = __alloc_bootmem_node(NODE_DATA(node),
					   size, align, goal);
		pr_debug("per cpu data for cpu%d %lu bytes on node%d at "
			 "%016lx\n", cpu, size, node, __pa(ptr));
	}
	return ptr;
#else
	return __alloc_bootmem(size, align, goal);
#endif
}

1455
static void __init pcpu_free_bootmem(void *ptr, size_t size)
1456
{
1457 1458
	free_bootmem(__pa(ptr), size);
}
1459

1460
static int __init pcpu_cpu_distance(unsigned int from, unsigned int to)
1461 1462 1463 1464 1465
{
	if (cpu_to_node(from) == cpu_to_node(to))
		return LOCAL_DISTANCE;
	else
		return REMOTE_DISTANCE;
1466 1467
}

1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490
static void __init pcpu_populate_pte(unsigned long addr)
{
	pgd_t *pgd = pgd_offset_k(addr);
	pud_t *pud;
	pmd_t *pmd;

	pud = pud_offset(pgd, addr);
	if (pud_none(*pud)) {
		pmd_t *new;

		new = __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, PAGE_SIZE);
		pud_populate(&init_mm, pud, new);
	}

	pmd = pmd_offset(pud, addr);
	if (!pmd_present(*pmd)) {
		pte_t *new;

		new = __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, PAGE_SIZE);
		pmd_populate_kernel(&init_mm, pmd, new);
	}
}

1491
void __init setup_per_cpu_areas(void)
L
Linus Torvalds 已提交
1492
{
1493 1494
	unsigned long delta;
	unsigned int cpu;
1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514
	int rc = -EINVAL;

	if (pcpu_chosen_fc != PCPU_FC_PAGE) {
		rc = pcpu_embed_first_chunk(PERCPU_MODULE_RESERVE,
					    PERCPU_DYNAMIC_RESERVE, 4 << 20,
					    pcpu_cpu_distance,
					    pcpu_alloc_bootmem,
					    pcpu_free_bootmem);
		if (rc)
			pr_warning("PERCPU: %s allocator failed (%d), "
				   "falling back to page size\n",
				   pcpu_fc_names[pcpu_chosen_fc], rc);
	}
	if (rc < 0)
		rc = pcpu_page_first_chunk(PERCPU_MODULE_RESERVE,
					   pcpu_alloc_bootmem,
					   pcpu_free_bootmem,
					   pcpu_populate_pte);
	if (rc < 0)
		panic("cannot initialize percpu area (err=%d)", rc);
1515

1516
	delta = (unsigned long)pcpu_base_addr - (unsigned long)__per_cpu_start;
T
Tejun Heo 已提交
1517 1518
	for_each_possible_cpu(cpu)
		__per_cpu_offset(cpu) = delta + pcpu_unit_offsets[cpu];
1519 1520 1521

	/* Setup %g5 for the boot cpu.  */
	__local_per_cpu_offset = __per_cpu_offset(smp_processor_id());
1522 1523 1524

	of_fill_in_cpu_data();
	if (tlb_type == hypervisor)
1525
		mdesc_fill_in_cpu_data(cpu_all_mask);
L
Linus Torvalds 已提交
1526
}