aboutsummaryrefslogtreecommitdiffstats
path: root/arch/x86/kernel/cpu/perfctr-watchdog.c
blob: 2e9bef6e3aa3d8ee0f778164725f024b6b7c388f (plain) (blame)
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
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
/*
 * local apic based NMI watchdog for various CPUs.
 *
 * This file also handles reservation of performance counters for coordination
 * with other users (like oprofile).
 *
 * Note that these events normally don't tick when the CPU idles. This means
 * the frequency varies with CPU load.
 *
 * Original code for K7/P6 written by Keith Owens
 *
 */

#include <linux/percpu.h>
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/bitops.h>
#include <linux/smp.h>
#include <linux/nmi.h>
#include <asm/apic.h>
#include <asm/intel_arch_perfmon.h>

struct nmi_watchdog_ctlblk {
	unsigned int cccr_msr;
	unsigned int perfctr_msr;  /* the MSR to reset in NMI handler */
	unsigned int evntsel_msr;  /* the MSR to select the events to handle */
};

/* Interface defining a CPU specific perfctr watchdog */
struct wd_ops {
	int (*reserve)(void);
	void (*unreserve)(void);
	int (*setup)(unsigned nmi_hz);
	void (*rearm)(struct nmi_watchdog_ctlblk *wd, unsigned nmi_hz);
	void (*stop)(void);
	unsigned perfctr;
	unsigned evntsel;
	u64 checkbit;
};

static const struct wd_ops *wd_ops;

/*
 * this number is calculated from Intel's MSR_P4_CRU_ESCR5 register and it's
 * offset from MSR_P4_BSU_ESCR0.
 *
 * It will be the max for all platforms (for now)
 */
#define NMI_MAX_COUNTER_BITS 66

/*
 * perfctr_nmi_owner tracks the ownership of the perfctr registers:
 * evtsel_nmi_owner tracks the ownership of the event selection
 * - different performance counters/ event selection may be reserved for
 *   different subsystems this reservation system just tries to coordinate
 *   things a little
 */
static DECLARE_BITMAP(perfctr_nmi_owner, NMI_MAX_COUNTER_BITS);
static DECLARE_BITMAP(evntsel_nmi_owner, NMI_MAX_COUNTER_BITS);

static DEFINE_PER_CPU(struct nmi_watchdog_ctlblk, nmi_watchdog_ctlblk);

/* converts an msr to an appropriate reservation bit */
static inline unsigned int nmi_perfctr_msr_to_bit(unsigned int msr)
{
	/* returns the bit offset of the performance counter register */
	switch (boot_cpu_data.x86_vendor) {
	case X86_VENDOR_AMD:
		return (msr - MSR_K7_PERFCTR0);
	case X86_VENDOR_INTEL:
		if (cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON))
			return (msr - MSR_ARCH_PERFMON_PERFCTR0);

		switch (boot_cpu_data.x86) {
		case 6:
			return (msr - MSR_P6_PERFCTR0);
		case 15:
			return (msr - MSR_P4_BPU_PERFCTR0);
		}
	}
	return 0;
}

/*
 * converts an msr to an appropriate reservation bit
 * returns the bit offset of the event selection register
 */
static inline unsigned int nmi_evntsel_msr_to_bit(unsigned int msr)
{
	/* returns the bit offset of the event selection register */
	switch (boot_cpu_data.x86_vendor) {
	case X86_VENDOR_AMD:
		return (msr - MSR_K7_EVNTSEL0);
	case X86_VENDOR_INTEL:
		if (cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON))
			return (msr - MSR_ARCH_PERFMON_EVENTSEL0);

		switch (boot_cpu_data.x86) {
		case 6:
			return (msr - MSR_P6_EVNTSEL0);
		case 15:
			return (msr - MSR_P4_BSU_ESCR0);
		}
	}
	return 0;

}

/* checks for a bit availability (hack for oprofile) */
int avail_to_resrv_perfctr_nmi_bit(unsigned int counter)
{
	BUG_ON(counter > NMI_MAX_COUNTER_BITS);

	return (!test_bit(counter, perfctr_nmi_owner));
}

/* checks the an msr for availability */
int avail_to_resrv_perfctr_nmi(unsigned int msr)
{
	unsigned int counter;

	counter = nmi_perfctr_msr_to_bit(msr);
	BUG_ON(counter > NMI_MAX_COUNTER_BITS);

	return (!test_bit(counter, perfctr_nmi_owner));
}
EXPORT_SYMBOL(avail_to_resrv_perfctr_nmi_bit);

int reserve_perfctr_nmi(unsigned int msr)
{
	unsigned int counter;

	counter = nmi_perfctr_msr_to_bit(msr);
	/* register not managed by the allocator? */
	if (counter > NMI_MAX_COUNTER_BITS)
		return 1;

	if (!test_and_set_bit(counter, perfctr_nmi_owner))
		return 1;
	return 0;
}
EXPORT_SYMBOL(reserve_perfctr_nmi);

void release_perfctr_nmi(unsigned int msr)
{
	unsigned int counter;

	counter = nmi_perfctr_msr_to_bit(msr);
	/* register not managed by the allocator? */
	if (counter > NMI_MAX_COUNTER_BITS)
		return;

	clear_bit(counter, perfctr_nmi_owner);
}
EXPORT_SYMBOL(release_perfctr_nmi);

int reserve_evntsel_nmi(unsigned int msr)
{
	unsigned int counter;

	counter = nmi_evntsel_msr_to_bit(msr);
	/* register not managed by the allocator? */
	if (counter > NMI_MAX_COUNTER_BITS)
		return 1;

	if (!test_and_set_bit(counter, evntsel_nmi_owner))
		return 1;
	return 0;
}
EXPORT_SYMBOL(reserve_evntsel_nmi);

void release_evntsel_nmi(unsigned int msr)
{
	unsigned int counter;

	counter = nmi_evntsel_msr_to_bit(msr);
	/* register not managed by the allocator? */
	if (counter > NMI_MAX_COUNTER_BITS)
		return;

	clear_bit(counter, evntsel_nmi_owner);
}
EXPORT_SYMBOL(release_evntsel_nmi);

void disable_lapic_nmi_watchdog(void)
{
	BUG_ON(nmi_watchdog != NMI_LOCAL_APIC);

	if (atomic_read(&nmi_active) <= 0)
		return;

	on_each_cpu(stop_apic_nmi_watchdog, NULL, 0, 1);

	if (wd_ops)
		wd_ops->unreserve();

	BUG_ON(atomic_read(&nmi_active) != 0);
}

void enable_lapic_nmi_watchdog(void)
{
	BUG_ON(nmi_watchdog != NMI_LOCAL_APIC);

	/* are we already enabled */
	if (atomic_read(&nmi_active) != 0)
		return;

	/* are we lapic aware */
	if (!wd_ops)
		return;
	if (!wd_ops->reserve()) {
		printk(KERN_ERR "NMI watchdog: cannot reserve perfctrs\n");
		return;
	}

	on_each_cpu(setup_apic_nmi_watchdog, NULL, 0, 1);
	touch_nmi_watchdog();
}

/*
 * Activate the NMI watchdog via the local APIC.
 */

static unsigned int adjust_for_32bit_ctr(unsigned int hz)
{
	u64 counter_val;
	unsigned int retval = hz;

	/*
	 * On Intel CPUs with P6/ARCH_PERFMON only 32 bits in the counter
	 * are writable, with higher bits sign extending from bit 31.
	 * So, we can only program the counter with 31 bit values and
	 * 32nd bit should be 1, for 33.. to be 1.
	 * Find the appropriate nmi_hz
	 */
	counter_val = (u64)cpu_khz * 1000;
	do_div(counter_val, retval);
 	if (counter_val > 0x7fffffffULL) {
		u64 count = (u64)cpu_khz * 1000;
		do_div(count, 0x7fffffffUL);
		retval = count + 1;
	}
	return retval;
}

static void write_watchdog_counter(unsigned int perfctr_msr,
				const char *descr, unsigned nmi_hz)
{
	u64 count = (u64)cpu_khz * 1000;

	do_div(count, nmi_hz);
	if(descr)
		Dprintk("setting %s to -0x%08Lx\n", descr, count);
	wrmsrl(perfctr_msr, 0 - count);
}

static void write_watchdog_counter32(unsigned int perfctr_msr,
				const char *descr, unsigned nmi_hz)
{
	u64 count = (u64)cpu_khz * 1000;

	do_div(count, nmi_hz);
	if(descr)
		Dprintk("setting %s to -0x%08Lx\n", descr, count);
	wrmsr(perfctr_msr, (u32)(-count), 0);
}

/*
 * AMD K7/K8/Family10h/Family11h support.
 * AMD keeps this interface nicely stable so there is not much variety
 */
#define K7_EVNTSEL_ENABLE	(1 << 22)
#define K7_EVNTSEL_INT		(1 << 20)
#define K7_EVNTSEL_OS		(1 << 17)
#define K7_EVNTSEL_USR		(1 << 16)
#define K7_EVENT_CYCLES_PROCESSOR_IS_RUNNING	0x76
#define K7_NMI_EVENT		K7_EVENT_CYCLES_PROCESSOR_IS_RUNNING

static int setup_k7_watchdog(unsigned nmi_hz)
{
	unsigned int perfctr_msr, evntsel_msr;
	unsigned int evntsel;
	struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);

	perfctr_msr = wd_ops->perfctr;
	evntsel_msr = wd_ops->evntsel;

	wrmsrl(perfctr_msr, 0UL);

	evntsel = K7_EVNTSEL_INT
		| K7_EVNTSEL_OS
		| K7_EVNTSEL_USR
		| K7_NMI_EVENT;

	/* setup the timer */
	wrmsr(evntsel_msr, evntsel, 0);
	write_watchdog_counter(perfctr_msr, "K7_PERFCTR0",nmi_hz);
	apic_write(APIC_LVTPC, APIC_DM_NMI);
	evntsel |= K7_EVNTSEL_ENABLE;
	wrmsr(evntsel_msr, evntsel, 0);

	wd->perfctr_msr = perfctr_msr;
	wd->evntsel_msr = evntsel_msr;
	wd->cccr_msr = 0;  /* unused */
	return 1;
}

static void single_msr_stop_watchdog(void)
{
	struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);

	wrmsr(wd->evntsel_msr, 0, 0);
}

static int single_msr_reserve(void)
{
	if (!reserve_perfctr_nmi(wd_ops->perfctr))
		return 0;

	if (!reserve_evntsel_nmi(wd_ops->evntsel)) {
		release_perfctr_nmi(wd_ops->perfctr);
		return 0;
	}
	return 1;
}

static void single_msr_unreserve(void)
{
	release_evntsel_nmi(wd_ops->evntsel);
	release_perfctr_nmi(wd_ops->perfctr);
}

static void single_msr_rearm(struct nmi_watchdog_ctlblk *wd, unsigned nmi_hz)
{
	/* start the cycle over again */
	write_watchdog_counter(wd->perfctr_msr, NULL, nmi_hz);
}

static const struct wd_ops k7_wd_ops = {
	.reserve	= single_msr_reserve,
	.unreserve	= single_msr_unreserve,
	.setup		= setup_k7_watchdog,
	.rearm		= single_msr_rearm,
	.stop		= single_msr_stop_watchdog,
	.perfctr	= MSR_K7_PERFCTR0,
	.evntsel	= MSR_K7_EVNTSEL0,
	.checkbit	= 1ULL << 47,
};

/*
 * Intel Model 6 (PPro+,P2,P3,P-M,Core1)
 */
#define P6_EVNTSEL0_ENABLE	(1 << 22)
#define P6_EVNTSEL_INT		(1 << 20)
#define P6_EVNTSEL_OS		(1 << 17)
#define P6_EVNTSEL_USR		(1 << 16)
#define P6_EVENT_CPU_CLOCKS_NOT_HALTED	0x79
#define P6_NMI_EVENT		P6_EVENT_CPU_CLOCKS_NOT_HALTED

static int setup_p6_watchdog(unsigned nmi_hz)
{
	unsigned int perfctr_msr, evntsel_msr;
	unsigned int evntsel;
	struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);

	perfctr_msr = wd_ops->perfctr;
	evntsel_msr = wd_ops->evntsel;

	/* KVM doesn't implement this MSR */
	if (wrmsr_safe(perfctr_msr, 0, 0) < 0)
		return 0;

	evntsel = P6_EVNTSEL_INT
		| P6_EVNTSEL_OS
		| P6_EVNTSEL_USR
		| P6_NMI_EVENT;

	/* setup the timer */
	wrmsr(evntsel_msr, evntsel, 0);
	nmi_hz = adjust_for_32bit_ctr(nmi_hz);
	write_watchdog_counter32(perfctr_msr, "P6_PERFCTR0",nmi_hz);
	apic_write(APIC_LVTPC, APIC_DM_NMI);
	evntsel |= P6_EVNTSEL0_ENABLE;
	wrmsr(evntsel_msr, evntsel, 0);

	wd->perfctr_msr = perfctr_msr;
	wd->evntsel_msr = evntsel_msr;
	wd->cccr_msr = 0;  /* unused */
	return 1;
}

static void p6_rearm(struct nmi_watchdog_ctlblk *wd, unsigned nmi_hz)
{
	/*
	 * P6 based Pentium M need to re-unmask
	 * the apic vector but it doesn't hurt
	 * other P6 variant.
	 * ArchPerfom/Core Duo also needs this
	 */
	apic_write(APIC_LVTPC, APIC_DM_NMI);

	/* P6/ARCH_PERFMON has 32 bit counter write */
	write_watchdog_counter32(wd->perfctr_msr, NULL,nmi_hz);
}

static const struct wd_ops p6_wd_ops = {
	.reserve	= single_msr_reserve,
	.unreserve	= single_msr_unreserve,
	.setup		= setup_p6_watchdog,
	.rearm		= p6_rearm,
	.stop		= single_msr_stop_watchdog,
	.perfctr	= MSR_P6_PERFCTR0,
	.evntsel	= MSR_P6_EVNTSEL0,
	.checkbit	= 1ULL << 39,
};

/*
 * Intel P4 performance counters.
 * By far the most complicated of all.
 */
#define MSR_P4_MISC_ENABLE_PERF_AVAIL	(1 << 7)
#define P4_ESCR_EVENT_SELECT(N)	((N) << 25)
#define P4_ESCR_OS		(1 << 3)
#define P4_ESCR_USR		(1 << 2)
#define P4_CCCR_OVF_PMI0	(1 << 26)
#define P4_CCCR_OVF_PMI1	(1 << 27)
#define P4_CCCR_THRESHOLD(N)	((N) << 20)
#define P4_CCCR_COMPLEMENT	(1 << 19)
#define P4_CCCR_COMPARE		(1 << 18)
#define P4_CCCR_REQUIRED	(3 << 16)
#define P4_CCCR_ESCR_SELECT(N)	((N) << 13)
#define P4_CCCR_ENABLE		(1 << 12)
#define P4_CCCR_OVF 		(1 << 31)

/*
 * Set up IQ_COUNTER0 to behave like a clock, by having IQ_CCCR0 filter
 * CRU_ESCR0 (with any non-null event selector) through a complemented
 * max threshold. [IA32-Vol3, Section 14.9.9]
 */
static int setup_p4_watchdog(unsigned nmi_hz)
{
	unsigned int perfctr_msr, evntsel_msr, cccr_msr;
	unsigned int evntsel, cccr_val;
	unsigned int misc_enable, dummy;
	unsigned int ht_num;
	struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);

	rdmsr(MSR_IA32_MISC_ENABLE, misc_enable, dummy);
	if (!(misc_enable & MSR_P4_MISC_ENABLE_PERF_AVAIL))
		return 0;

#ifdef CONFIG_SMP
	/* detect which hyperthread we are on */
	if (smp_num_siblings == 2) {
		unsigned int ebx, apicid;

        	ebx = cpuid_ebx(1);
	        apicid = (ebx >> 24) & 0xff;
        	ht_num = apicid & 1;
	} else
#endif
		ht_num = 0;

	/*
	 * performance counters are shared resources
	 * assign each hyperthread its own set
	 * (re-use the ESCR0 register, seems safe
	 * and keeps the cccr_val the same)
	 */
	if (!ht_num) {
		/* logical cpu 0 */
		perfctr_msr = MSR_P4_IQ_PERFCTR0;
		evntsel_msr = MSR_P4_CRU_ESCR0;
		cccr_msr = MSR_P4_IQ_CCCR0;
		cccr_val = P4_CCCR_OVF_PMI0 | P4_CCCR_ESCR_SELECT(4);
	} else {
		/* logical cpu 1 */
		perfctr_msr = MSR_P4_IQ_PERFCTR1;
		evntsel_msr = MSR_P4_CRU_ESCR0;
		cccr_msr = MSR_P4_IQ_CCCR1;
		cccr_val = P4_CCCR_OVF_PMI1 | P4_CCCR_ESCR_SELECT(4);
	}

	evntsel = P4_ESCR_EVENT_SELECT(0x3F)
	 	| P4_ESCR_OS
		| P4_ESCR_USR;

	cccr_val |= P4_CCCR_THRESHOLD(15)
		 | P4_CCCR_COMPLEMENT
		 | P4_CCCR_COMPARE
		 | P4_CCCR_REQUIRED;

	wrmsr(evntsel_msr, evntsel, 0);
	wrmsr(cccr_msr, cccr_val, 0);
	write_watchdog_counter(perfctr_msr, "P4_IQ_COUNTER0", nmi_hz);
	apic_write(APIC_LVTPC, APIC_DM_NMI);
	cccr_val |= P4_CCCR_ENABLE;
	wrmsr(cccr_msr, cccr_val, 0);
	wd->perfctr_msr = perfctr_msr;
	wd->evntsel_msr = evntsel_msr;
	wd->cccr_msr = cccr_msr;
	return 1;
}

static void stop_p4_watchdog(void)
{
	struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);
	wrmsr(wd->cccr_msr, 0, 0);
	wrmsr(wd->evntsel_msr, 0, 0);
}

static int p4_reserve(void)
{
	if (!reserve_perfctr_nmi(MSR_P4_IQ_PERFCTR0))
		return 0;
#ifdef CONFIG_SMP
	if (smp_num_siblings > 1 && !reserve_perfctr_nmi(MSR_P4_IQ_PERFCTR1))
		goto fail1;
#endif
	if (!reserve_evntsel_nmi(MSR_P4_CRU_ESCR0))
		goto fail2;
	/* RED-PEN why is ESCR1 not reserved here? */
	return 1;
 fail2:
#ifdef CONFIG_SMP
	if (smp_num_siblings > 1)
		release_perfctr_nmi(MSR_P4_IQ_PERFCTR1);
 fail1:
#endif
	release_perfctr_nmi(MSR_P4_IQ_PERFCTR0);
	return 0;
}

static void p4_unreserve(void)
{
#ifdef CONFIG_SMP
	if (smp_num_siblings > 1)
		release_perfctr_nmi(MSR_P4_IQ_PERFCTR1);
#endif
	release_evntsel_nmi(MSR_P4_CRU_ESCR0);
	release_perfctr_nmi(MSR_P4_IQ_PERFCTR0);
}

static void p4_rearm(struct nmi_watchdog_ctlblk *wd, unsigned nmi_hz)
{
	unsigned dummy;
	/*
 	 * P4 quirks:
	 * - An overflown perfctr will assert its interrupt
	 *   until the OVF flag in its CCCR is cleared.
	 * - LVTPC is masked on interrupt and must be
	 *   unmasked by the LVTPC handler.
	 */
	rdmsrl(wd->cccr_msr, dummy);
	dummy &= ~P4_CCCR_OVF;
	wrmsrl(wd->cccr_msr, dummy);
	apic_write(APIC_LVTPC, APIC_DM_NMI);
	/* start the cycle over again */
	write_watchdog_counter(wd->perfctr_msr, NULL, nmi_hz);
}

static const struct wd_ops p4_wd_ops = {
	.reserve	= p4_reserve,
	.unreserve	= p4_unreserve,
	.setup		= setup_p4_watchdog,
	.rearm		= p4_rearm,
	.stop		= stop_p4_watchdog,
	/* RED-PEN this is wrong for the other sibling */
	.perfctr	= MSR_P4_BPU_PERFCTR0,
	.evntsel	= MSR_P4_BSU_ESCR0,
	.checkbit	= 1ULL << 39,
};

/*
 * Watchdog using the Intel architected PerfMon.
 * Used for Core2 and hopefully all future Intel CPUs.
 */
#define ARCH_PERFMON_NMI_EVENT_SEL	ARCH_PERFMON_UNHALTED_CORE_CYCLES_SEL
#define ARCH_PERFMON_NMI_EVENT_UMASK	ARCH_PERFMON_UNHALTED_CORE_CYCLES_UMASK

static struct wd_ops intel_arch_wd_ops;

static int setup_intel_arch_watchdog(unsigned nmi_hz)
{
	unsigned int ebx;
	union cpuid10_eax eax;
	unsigned int unused;
	unsigned int perfctr_msr, evntsel_msr;
	unsigned int evntsel;
	struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);

	/*
	 * Check whether the Architectural PerfMon supports
	 * Unhalted Core Cycles Event or not.
	 * NOTE: Corresponding bit = 0 in ebx indicates event present.
	 */
	cpuid(10, &(eax.full), &ebx, &unused, &unused);
	if ((eax.split.mask_length < (ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX+1)) ||
	    (ebx & ARCH_PERFMON_UNHALTED_CORE_CYCLES_PRESENT))
		return 0;

	perfctr_msr = wd_ops->perfctr;
	evntsel_msr = wd_ops->evntsel;

	wrmsrl(perfctr_msr, 0UL);

	evntsel = ARCH_PERFMON_EVENTSEL_INT
		| ARCH_PERFMON_EVENTSEL_OS
		| ARCH_PERFMON_EVENTSEL_USR
		| ARCH_PERFMON_NMI_EVENT_SEL
		| ARCH_PERFMON_NMI_EVENT_UMASK;

	/* setup the timer */
	wrmsr(evntsel_msr, evntsel, 0);
	nmi_hz = adjust_for_32bit_ctr(nmi_hz);
	write_watchdog_counter32(perfctr_msr, "INTEL_ARCH_PERFCTR0", nmi_hz);
	apic_write(APIC_LVTPC, APIC_DM_NMI);
	evntsel |= ARCH_PERFMON_EVENTSEL0_ENABLE;
	wrmsr(evntsel_msr, evntsel, 0);

	wd->perfctr_msr = perfctr_msr;
	wd->evntsel_msr = evntsel_msr;
	wd->cccr_msr = 0;  /* unused */
	intel_arch_wd_ops.checkbit = 1ULL << (eax.split.bit_width - 1);
	return 1;
}

static struct wd_ops intel_arch_wd_ops __read_mostly = {
	.reserve	= single_msr_reserve,
	.unreserve	= single_msr_unreserve,
	.setup		= setup_intel_arch_watchdog,
	.rearm		= p6_rearm,
	.stop		= single_msr_stop_watchdog,
	.perfctr	= MSR_ARCH_PERFMON_PERFCTR1,
	.evntsel	= MSR_ARCH_PERFMON_EVENTSEL1,
};

static void probe_nmi_watchdog(void)
{
	switch (boot_cpu_data.x86_vendor) {
	case X86_VENDOR_AMD:
		if (boot_cpu_data.x86 != 6 && boot_cpu_data.x86 != 15 &&
		    boot_cpu_data.x86 != 16)
			return;
		wd_ops = &k7_wd_ops;
		break;
	case X86_VENDOR_INTEL:
		/*
		 * Work around Core Duo (Yonah) errata AE49 where perfctr1
		 * doesn't have a working enable bit.
		 */
		if (boot_cpu_data.x86 == 6 && boot_cpu_data.x86_model == 14) {
			intel_arch_wd_ops.perfctr = MSR_ARCH_PERFMON_PERFCTR0;
			intel_arch_wd_ops.evntsel = MSR_ARCH_PERFMON_EVENTSEL0;
		}
		if (cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON)) {
			wd_ops = &intel_arch_wd_ops;
			break;
		}
		switch (boot_cpu_data.x86) {
		case 6:
			if (boot_cpu_data.x86_model > 13)
				return;

			wd_ops = &p6_wd_ops;
			break;
		case 15:
			wd_ops = &p4_wd_ops;
			break;
		default:
			return;
		}
		break;
	}
}

/* Interface to nmi.c */

int lapic_watchdog_init(unsigned nmi_hz)
{
	if (!wd_ops) {
		probe_nmi_watchdog();
		if (!wd_ops) {
			printk(KERN_INFO "NMI watchdog: CPU not supported\n");
			return -1;
		}

		if (!wd_ops->reserve()) {
			printk(KERN_ERR
				"NMI watchdog: cannot reserve perfctrs\n");
			return -1;
		}
	}

	if (!(wd_ops->setup(nmi_hz))) {
		printk(KERN_ERR "Cannot setup NMI watchdog on CPU %d\n",
		       raw_smp_processor_id());
		return -1;
	}

	return 0;
}

void lapic_watchdog_stop(void)
{
	if (wd_ops)
		wd_ops->stop();
}

unsigned lapic_adjust_nmi_hz(unsigned hz)
{
	struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);
	if (wd->perfctr_msr == MSR_P6_PERFCTR0 ||
	    wd->perfctr_msr == MSR_ARCH_PERFMON_PERFCTR1)
		hz = adjust_for_32bit_ctr(hz);
	return hz;
}

int lapic_wd_event(unsigned nmi_hz)
{
	struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);
	u64 ctr;

	rdmsrl(wd->perfctr_msr, ctr);
	if (ctr & wd_ops->checkbit) /* perfctr still running? */
		return 0;

	wd_ops->rearm(wd, nmi_hz);
	return 1;
}

int lapic_watchdog_ok(void)
{
	return wd_ops != NULL;
}