summaryrefslogtreecommitdiffstats
path: root/sys/kern/subr_smp.c
blob: b7fbbd1069b55d96e63c4820e19acba40898d122 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
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
162
163
164
165
166
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
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
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
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
/*-
 * Copyright (c) 2001, John Baldwin <jhb@FreeBSD.org>.
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. Neither the name of the author nor the names of any co-contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 */

/*
 * This module holds the global variables and machine independent functions
 * used for the kernel SMP support.
 */

#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/ktr.h>
#include <sys/proc.h>
#include <sys/bus.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/pcpu.h>
#include <sys/smp.h>
#include <sys/sysctl.h>

#include <machine/cpu.h>
#include <machine/smp.h>

#include "opt_sched.h"

#ifdef SMP
volatile cpumask_t stopped_cpus;
volatile cpumask_t started_cpus;
cpumask_t idle_cpus_mask;
cpumask_t hlt_cpus_mask;
cpumask_t logical_cpus_mask;

void (*cpustop_restartfunc)(void);
#endif
/* This is used in modules that need to work in both SMP and UP. */
cpumask_t all_cpus;

int mp_ncpus;
/* export this for libkvm consumers. */
int mp_maxcpus = MAXCPU;

volatile int smp_started;
u_int mp_maxid;

SYSCTL_NODE(_kern, OID_AUTO, smp, CTLFLAG_RD, NULL, "Kernel SMP");

SYSCTL_INT(_kern_smp, OID_AUTO, maxid, CTLFLAG_RD, &mp_maxid, 0,
    "Max CPU ID.");

SYSCTL_INT(_kern_smp, OID_AUTO, maxcpus, CTLFLAG_RD, &mp_maxcpus, 0,
    "Max number of CPUs that the system was compiled for.");

int smp_active = 0;	/* are the APs allowed to run? */
SYSCTL_INT(_kern_smp, OID_AUTO, active, CTLFLAG_RW, &smp_active, 0,
    "Number of Auxillary Processors (APs) that were successfully started");

int smp_disabled = 0;	/* has smp been disabled? */
SYSCTL_INT(_kern_smp, OID_AUTO, disabled, CTLFLAG_RDTUN, &smp_disabled, 0,
    "SMP has been disabled from the loader");
TUNABLE_INT("kern.smp.disabled", &smp_disabled);

int smp_cpus = 1;	/* how many cpu's running */
SYSCTL_INT(_kern_smp, OID_AUTO, cpus, CTLFLAG_RD, &smp_cpus, 0,
    "Number of CPUs online");

int smp_topology = 0;	/* Which topology we're using. */
SYSCTL_INT(_kern_smp, OID_AUTO, topology, CTLFLAG_RD, &smp_topology, 0,
    "Topology override setting; 0 is default provided by hardware.");
TUNABLE_INT("kern.smp.topology", &smp_topology);

#ifdef SMP
/* Enable forwarding of a signal to a process running on a different CPU */
static int forward_signal_enabled = 1;
SYSCTL_INT(_kern_smp, OID_AUTO, forward_signal_enabled, CTLFLAG_RW,
	   &forward_signal_enabled, 0,
	   "Forwarding of a signal to a process on a different CPU");

/* Enable forwarding of roundrobin to all other cpus */
static int forward_roundrobin_enabled = 1;
SYSCTL_INT(_kern_smp, OID_AUTO, forward_roundrobin_enabled, CTLFLAG_RW,
	   &forward_roundrobin_enabled, 0,
	   "Forwarding of roundrobin to all other CPUs");

/* Variables needed for SMP rendezvous. */
static volatile int smp_rv_ncpus;
static void (*volatile smp_rv_setup_func)(void *arg);
static void (*volatile smp_rv_action_func)(void *arg);
static void (*volatile smp_rv_teardown_func)(void *arg);
static void * volatile smp_rv_func_arg;
static volatile int smp_rv_waiters[3];

/* 
 * Shared mutex to restrict busywaits between smp_rendezvous() and
 * smp(_targeted)_tlb_shootdown().  A deadlock occurs if both of these
 * functions trigger at once and cause multiple CPUs to busywait with
 * interrupts disabled. 
 */
struct mtx smp_ipi_mtx;

/*
 * Let the MD SMP code initialize mp_maxid very early if it can.
 */
static void
mp_setmaxid(void *dummy)
{
	cpu_mp_setmaxid();
}
SYSINIT(cpu_mp_setmaxid, SI_SUB_TUNABLES, SI_ORDER_FIRST, mp_setmaxid, NULL);

/*
 * Call the MD SMP initialization code.
 */
static void
mp_start(void *dummy)
{

	/* Probe for MP hardware. */
	if (smp_disabled != 0 || cpu_mp_probe() == 0) {
		mp_ncpus = 1;
		all_cpus = PCPU_GET(cpumask);
		return;
	}

	mtx_init(&smp_ipi_mtx, "smp rendezvous", NULL, MTX_SPIN);
	cpu_mp_start();
	printf("FreeBSD/SMP: Multiprocessor System Detected: %d CPUs\n",
	    mp_ncpus);
	cpu_mp_announce();
}
SYSINIT(cpu_mp, SI_SUB_CPU, SI_ORDER_THIRD, mp_start, NULL);

void
forward_signal(struct thread *td)
{
	int id;

	/*
	 * signotify() has already set TDF_ASTPENDING and TDF_NEEDSIGCHECK on
	 * this thread, so all we need to do is poke it if it is currently
	 * executing so that it executes ast().
	 */
	THREAD_LOCK_ASSERT(td, MA_OWNED);
	KASSERT(TD_IS_RUNNING(td),
	    ("forward_signal: thread is not TDS_RUNNING"));

	CTR1(KTR_SMP, "forward_signal(%p)", td->td_proc);

	if (!smp_started || cold || panicstr)
		return;
	if (!forward_signal_enabled)
		return;

	/* No need to IPI ourself. */
	if (td == curthread)
		return;

	id = td->td_oncpu;
	if (id == NOCPU)
		return;
	ipi_selected(1 << id, IPI_AST);
}

void
forward_roundrobin(void)
{
	struct pcpu *pc;
	struct thread *td;
	cpumask_t id, map, me;

	CTR0(KTR_SMP, "forward_roundrobin()");

	if (!smp_started || cold || panicstr)
		return;
	if (!forward_roundrobin_enabled)
		return;
	map = 0;
	me = PCPU_GET(cpumask);
	SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
		td = pc->pc_curthread;
		id = pc->pc_cpumask;
		if (id != me && (id & stopped_cpus) == 0 &&
		    !TD_IS_IDLETHREAD(td)) {
			td->td_flags |= TDF_NEEDRESCHED;
			map |= id;
		}
	}
	ipi_selected(map, IPI_AST);
}

/*
 * When called the executing CPU will send an IPI to all other CPUs
 *  requesting that they halt execution.
 *
 * Usually (but not necessarily) called with 'other_cpus' as its arg.
 *
 *  - Signals all CPUs in map to stop.
 *  - Waits for each to stop.
 *
 * Returns:
 *  -1: error
 *   0: NA
 *   1: ok
 *
 * XXX FIXME: this is not MP-safe, needs a lock to prevent multiple CPUs
 *            from executing at same time.
 */
int
stop_cpus(cpumask_t map)
{
	int i;

	if (!smp_started)
		return 0;

	CTR1(KTR_SMP, "stop_cpus(%x)", map);

	/* send the stop IPI to all CPUs in map */
	ipi_selected(map, IPI_STOP);

	i = 0;
	while ((stopped_cpus & map) != map) {
		/* spin */
		cpu_spinwait();
		i++;
#ifdef DIAGNOSTIC
		if (i == 100000) {
			printf("timeout stopping cpus\n");
			break;
		}
#endif
	}

	return 1;
}

/*
 * Called by a CPU to restart stopped CPUs. 
 *
 * Usually (but not necessarily) called with 'stopped_cpus' as its arg.
 *
 *  - Signals all CPUs in map to restart.
 *  - Waits for each to restart.
 *
 * Returns:
 *  -1: error
 *   0: NA
 *   1: ok
 */
int
restart_cpus(cpumask_t map)
{

	if (!smp_started)
		return 0;

	CTR1(KTR_SMP, "restart_cpus(%x)", map);

	/* signal other cpus to restart */
	atomic_store_rel_int(&started_cpus, map);

	/* wait for each to clear its bit */
	while ((stopped_cpus & map) != 0)
		cpu_spinwait();

	return 1;
}

/*
 * All-CPU rendezvous.  CPUs are signalled, all execute the setup function 
 * (if specified), rendezvous, execute the action function (if specified),
 * rendezvous again, execute the teardown function (if specified), and then
 * resume.
 *
 * Note that the supplied external functions _must_ be reentrant and aware
 * that they are running in parallel and in an unknown lock context.
 */
void
smp_rendezvous_action(void)
{
	void* local_func_arg = smp_rv_func_arg;
	void (*local_setup_func)(void*)   = smp_rv_setup_func;
	void (*local_action_func)(void*)   = smp_rv_action_func;
	void (*local_teardown_func)(void*) = smp_rv_teardown_func;

	/* Ensure we have up-to-date values. */
	atomic_add_acq_int(&smp_rv_waiters[0], 1);
	while (smp_rv_waiters[0] < smp_rv_ncpus)
		cpu_spinwait();

	/* setup function */
	if (local_setup_func != smp_no_rendevous_barrier) {
		if (smp_rv_setup_func != NULL)
			smp_rv_setup_func(smp_rv_func_arg);

		/* spin on entry rendezvous */
		atomic_add_int(&smp_rv_waiters[1], 1);
		while (smp_rv_waiters[1] < smp_rv_ncpus)
                	cpu_spinwait();
	}

	/* action function */
	if (local_action_func != NULL)
		local_action_func(local_func_arg);

	/* spin on exit rendezvous */
	atomic_add_int(&smp_rv_waiters[2], 1);
	if (local_teardown_func == smp_no_rendevous_barrier)
                return;
	while (smp_rv_waiters[2] < smp_rv_ncpus)
		cpu_spinwait();

	/* teardown function */
	if (local_teardown_func != NULL)
		local_teardown_func(local_func_arg);
}

void
smp_rendezvous_cpus(cpumask_t map,
	void (* setup_func)(void *), 
	void (* action_func)(void *),
	void (* teardown_func)(void *),
	void *arg)
{
	int i, ncpus = 0;

	if (!smp_started) {
		if (setup_func != NULL)
			setup_func(arg);
		if (action_func != NULL)
			action_func(arg);
		if (teardown_func != NULL)
			teardown_func(arg);
		return;
	}

	for (i = 0; i < mp_maxid; i++)
		if (((1 << i) & map) != 0 && !CPU_ABSENT(i))
			ncpus++;

	/* obtain rendezvous lock */
	mtx_lock_spin(&smp_ipi_mtx);

	/* set static function pointers */
	smp_rv_ncpus = ncpus;
	smp_rv_setup_func = setup_func;
	smp_rv_action_func = action_func;
	smp_rv_teardown_func = teardown_func;
	smp_rv_func_arg = arg;
	smp_rv_waiters[1] = 0;
	smp_rv_waiters[2] = 0;
	atomic_store_rel_int(&smp_rv_waiters[0], 0);

	/* signal other processors, which will enter the IPI with interrupts off */
	ipi_selected(map & ~(1 << curcpu), IPI_RENDEZVOUS);

	/* Check if the current CPU is in the map */
	if ((map & (1 << curcpu)) != 0)
		smp_rendezvous_action();

	if (teardown_func == smp_no_rendevous_barrier)
		while (atomic_load_acq_int(&smp_rv_waiters[2]) < ncpus)
			cpu_spinwait();

	/* release lock */
	mtx_unlock_spin(&smp_ipi_mtx);
}

void
smp_rendezvous(void (* setup_func)(void *), 
	       void (* action_func)(void *),
	       void (* teardown_func)(void *),
	       void *arg)
{
	smp_rendezvous_cpus(all_cpus, setup_func, action_func, teardown_func, arg);
}

static struct cpu_group group[MAXCPU];

struct cpu_group *
smp_topo(void)
{
	struct cpu_group *top;

	/*
	 * Check for a fake topology request for debugging purposes.
	 */
	switch (smp_topology) {
	case 1:
		/* Dual core with no sharing.  */
		top = smp_topo_1level(CG_SHARE_NONE, 2, 0);
		break;
	case 2:
		/* No topology, all cpus are equal. */
		top = smp_topo_none();
		break;
	case 3:
		/* Dual core with shared L2.  */
		top = smp_topo_1level(CG_SHARE_L2, 2, 0);
		break;
	case 4:
		/* quad core, shared l3 among each package, private l2.  */
		top = smp_topo_1level(CG_SHARE_L3, 4, 0);
		break;
	case 5:
		/* quad core,  2 dualcore parts on each package share l2.  */
		top = smp_topo_2level(CG_SHARE_NONE, 2, CG_SHARE_L2, 2, 0);
		break;
	case 6:
		/* Single-core 2xHTT */
		top = smp_topo_1level(CG_SHARE_L1, 2, CG_FLAG_HTT);
		break;
	case 7:
		/* quad core with a shared l3, 8 threads sharing L2.  */
		top = smp_topo_2level(CG_SHARE_L3, 4, CG_SHARE_L2, 8,
		    CG_FLAG_THREAD);
		break;
	default:
		/* Default, ask the system what it wants. */
		top = cpu_topo();
		break;
	}
	/*
	 * Verify the returned topology.
	 */
	if (top->cg_count != mp_ncpus)
		panic("Built bad topology at %p.  CPU count %d != %d",
		    top, top->cg_count, mp_ncpus);
	if (top->cg_mask != all_cpus)
		panic("Built bad topology at %p.  CPU mask 0x%X != 0x%X",
		    top, top->cg_mask, all_cpus);
	return (top);
}

struct cpu_group *
smp_topo_none(void)
{
	struct cpu_group *top;

	top = &group[0];
	top->cg_parent = NULL;
	top->cg_child = NULL;
	top->cg_mask = (1 << mp_ncpus) - 1;
	top->cg_count = mp_ncpus;
	top->cg_children = 0;
	top->cg_level = CG_SHARE_NONE;
	top->cg_flags = 0;
	
	return (top);
}

static int
smp_topo_addleaf(struct cpu_group *parent, struct cpu_group *child, int share,
    int count, int flags, int start)
{
	cpumask_t mask;
	int i;

	for (mask = 0, i = 0; i < count; i++, start++)
		mask |= (1 << start);
	child->cg_parent = parent;
	child->cg_child = NULL;
	child->cg_children = 0;
	child->cg_level = share;
	child->cg_count = count;
	child->cg_flags = flags;
	child->cg_mask = mask;
	parent->cg_children++;
	for (; parent != NULL; parent = parent->cg_parent) {
		if ((parent->cg_mask & child->cg_mask) != 0)
			panic("Duplicate children in %p.  mask 0x%X child 0x%X",
			    parent, parent->cg_mask, child->cg_mask);
		parent->cg_mask |= child->cg_mask;
		parent->cg_count += child->cg_count;
	}

	return (start);
}

struct cpu_group *
smp_topo_1level(int share, int count, int flags)
{
	struct cpu_group *child;
	struct cpu_group *top;
	int packages;
	int cpu;
	int i;

	cpu = 0;
	top = &group[0];
	packages = mp_ncpus / count;
	top->cg_child = child = &group[1];
	top->cg_level = CG_SHARE_NONE;
	for (i = 0; i < packages; i++, child++)
		cpu = smp_topo_addleaf(top, child, share, count, flags, cpu);
	return (top);
}

struct cpu_group *
smp_topo_2level(int l2share, int l2count, int l1share, int l1count,
    int l1flags)
{
	struct cpu_group *top;
	struct cpu_group *l1g;
	struct cpu_group *l2g;
	int cpu;
	int i;
	int j;

	cpu = 0;
	top = &group[0];
	l2g = &group[1];
	top->cg_child = l2g;
	top->cg_level = CG_SHARE_NONE;
	top->cg_children = mp_ncpus / (l2count * l1count);
	l1g = l2g + top->cg_children;
	for (i = 0; i < top->cg_children; i++, l2g++) {
		l2g->cg_parent = top;
		l2g->cg_child = l1g;
		l2g->cg_level = l2share;
		for (j = 0; j < l2count; j++, l1g++)
			cpu = smp_topo_addleaf(l2g, l1g, l1share, l1count,
			    l1flags, cpu);
	}
	return (top);
}


struct cpu_group *
smp_topo_find(struct cpu_group *top, int cpu)
{
	struct cpu_group *cg;
	cpumask_t mask;
	int children;
	int i;

	mask = (1 << cpu);
	cg = top;
	for (;;) {
		if ((cg->cg_mask & mask) == 0)
			return (NULL);
		if (cg->cg_children == 0)
			return (cg);
		children = cg->cg_children;
		for (i = 0, cg = cg->cg_child; i < children; cg++, i++)
			if ((cg->cg_mask & mask) != 0)
				break;
	}
	return (NULL);
}
#else /* !SMP */

void
smp_rendezvous_cpus(cpumask_t map,
	void (*setup_func)(void *), 
	void (*action_func)(void *),
	void (*teardown_func)(void *),
	void *arg)
{
	if (setup_func != NULL)
		setup_func(arg);
	if (action_func != NULL)
		action_func(arg);
	if (teardown_func != NULL)
		teardown_func(arg);
}

void
smp_rendezvous(void (*setup_func)(void *), 
	       void (*action_func)(void *),
	       void (*teardown_func)(void *),
	       void *arg)
{

	if (setup_func != NULL)
		setup_func(arg);
	if (action_func != NULL)
		action_func(arg);
	if (teardown_func != NULL)
		teardown_func(arg);
}

/*
 * Provide dummy SMP support for UP kernels.  Modules that need to use SMP
 * APIs will still work using this dummy support.
 */
static void
mp_setvariables_for_up(void *dummy)
{
	mp_ncpus = 1;
	mp_maxid = PCPU_GET(cpuid);
	all_cpus = PCPU_GET(cpumask);
	KASSERT(PCPU_GET(cpuid) == 0, ("UP must have a CPU ID of zero"));
}
SYSINIT(cpu_mp_setvariables, SI_SUB_TUNABLES, SI_ORDER_FIRST,
    mp_setvariables_for_up, NULL);
#endif /* SMP */

void
smp_no_rendevous_barrier(void *dummy)
{
#ifdef SMP
	KASSERT((!smp_started),("smp_no_rendevous called and smp is started"));
#endif
}
OpenPOWER on IntegriCloud