nmi.c 16.3 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
/*
 *  linux/arch/x86_64/nmi.c
 *
 *  NMI watchdog support on APIC systems
 *
 *  Started by Ingo Molnar <mingo@redhat.com>
 *
 *  Fixes:
 *  Mikael Pettersson	: AMD K7 support for local APIC NMI watchdog.
 *  Mikael Pettersson	: Power Management for local APIC NMI watchdog.
 *  Pavel Machek and
 *  Mikael Pettersson	: PM converted to driver model. Disable/enable API.
 */

#include <linux/config.h>
#include <linux/mm.h>
#include <linux/delay.h>
#include <linux/interrupt.h>
#include <linux/module.h>
#include <linux/sysdev.h>
#include <linux/nmi.h>
#include <linux/sysctl.h>
23
#include <linux/kprobes.h>
Linus Torvalds's avatar
Linus Torvalds committed
24
25
26
27
28

#include <asm/smp.h>
#include <asm/nmi.h>
#include <asm/proto.h>
#include <asm/kdebug.h>
29
#include <asm/mce.h>
30
#include <asm/intel_arch_perfmon.h>
Linus Torvalds's avatar
Linus Torvalds committed
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

/*
 * lapic_nmi_owner tracks the ownership of the lapic NMI hardware:
 * - it may be reserved by some other driver, or not
 * - when not reserved by some other driver, it may be used for
 *   the NMI watchdog, or not
 *
 * This is maintained separately from nmi_active because the NMI
 * watchdog may also be driven from the I/O APIC timer.
 */
static DEFINE_SPINLOCK(lapic_nmi_owner_lock);
static unsigned int lapic_nmi_owner;
#define LAPIC_NMI_WATCHDOG	(1<<0)
#define LAPIC_NMI_RESERVED	(1<<1)

/* nmi_active:
 * +1: the lapic NMI watchdog is active, but can be disabled
 *  0: the lapic NMI watchdog has not been set up, and cannot
 *     be enabled
 * -1: the lapic NMI watchdog is disabled, but can be enabled
 */
int nmi_active;		/* oprofile uses this */
int panic_on_timeout;

unsigned int nmi_watchdog = NMI_DEFAULT;
static unsigned int nmi_hz = HZ;
57
58
static unsigned int nmi_perfctr_msr;	/* the MSR to reset in NMI handler */
static unsigned int nmi_p4_cccr_val;
Linus Torvalds's avatar
Linus Torvalds committed
59
60
61
62
63
64
65
66
67
68
69

/* Note that these events don't tick when the CPU idles. This means
   the frequency varies with CPU load. */

#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

70
71
72
#define ARCH_PERFMON_NMI_EVENT_SEL	ARCH_PERFMON_UNHALTED_CORE_CYCLES_SEL
#define ARCH_PERFMON_NMI_EVENT_UMASK	ARCH_PERFMON_UNHALTED_CORE_CYCLES_UMASK

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
#define MSR_P4_MISC_ENABLE	0x1A0
#define MSR_P4_MISC_ENABLE_PERF_AVAIL	(1<<7)
#define MSR_P4_MISC_ENABLE_PEBS_UNAVAIL	(1<<12)
#define MSR_P4_PERFCTR0		0x300
#define MSR_P4_CCCR0		0x360
#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)
/* 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] */
#define MSR_P4_IQ_COUNTER0	0x30C
#define P4_NMI_CRU_ESCR0	(P4_ESCR_EVENT_SELECT(0x3F)|P4_ESCR_OS|P4_ESCR_USR)
#define P4_NMI_IQ_CCCR0	\
	(P4_CCCR_OVF_PMI0|P4_CCCR_THRESHOLD(15)|P4_CCCR_COMPLEMENT|	\
	 P4_CCCR_COMPARE|P4_CCCR_REQUIRED|P4_CCCR_ESCR_SELECT(4)|P4_CCCR_ENABLE)

98
static __cpuinit inline int nmi_known_cpu(void)
99
100
101
102
103
{
	switch (boot_cpu_data.x86_vendor) {
	case X86_VENDOR_AMD:
		return boot_cpu_data.x86 == 15;
	case X86_VENDOR_INTEL:
104
105
106
107
		if (cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON))
			return 1;
		else
			return (boot_cpu_data.x86 == 15);
108
109
110
	}
	return 0;
}
Linus Torvalds's avatar
Linus Torvalds committed
111
112

/* Run after command line and cpu_init init, but before all other checks */
113
void __cpuinit nmi_watchdog_default(void)
Linus Torvalds's avatar
Linus Torvalds committed
114
115
116
{
	if (nmi_watchdog != NMI_DEFAULT)
		return;
117
118
119
	if (nmi_known_cpu())
		nmi_watchdog = NMI_LOCAL_APIC;
	else
Linus Torvalds's avatar
Linus Torvalds committed
120
121
122
		nmi_watchdog = NMI_IO_APIC;
}

123
124
125
126
127
128
#ifdef CONFIG_SMP
/* The performance counters used by NMI_LOCAL_APIC don't trigger when
 * the CPU is idle. To make sure the NMI watchdog really ticks on all
 * CPUs during the test make them busy.
 */
static __init void nmi_cpu_busy(void *data)
Linus Torvalds's avatar
Linus Torvalds committed
129
{
130
131
132
133
134
135
136
137
138
139
	volatile int *endflag = data;
	local_irq_enable();
	/* Intentionally don't use cpu_relax here. This is
	   to make sure that the performance counter really ticks,
	   even if there is a simulator or similar that catches the
	   pause instruction. On a real HT machine this is fine because
	   all other CPUs are busy with "useless" delay loops and don't
	   care if they get somewhat less cycles. */
	while (*endflag == 0)
		barrier();
Linus Torvalds's avatar
Linus Torvalds committed
140
}
141
#endif
Linus Torvalds's avatar
Linus Torvalds committed
142

143
int __init check_nmi_watchdog (void)
Linus Torvalds's avatar
Linus Torvalds committed
144
{
145
	volatile int endflag = 0;
146
	int *counts;
Linus Torvalds's avatar
Linus Torvalds committed
147
148
	int cpu;

149
150
151
	counts = kmalloc(NR_CPUS * sizeof(int), GFP_KERNEL);
	if (!counts)
		return -1;
Linus Torvalds's avatar
Linus Torvalds committed
152

153
	printk(KERN_INFO "testing NMI watchdog ... ");
154

155
#ifdef CONFIG_SMP
156
157
	if (nmi_watchdog == NMI_LOCAL_APIC)
		smp_call_function(nmi_cpu_busy, (void *)&endflag, 0, 0);
158
#endif
Linus Torvalds's avatar
Linus Torvalds committed
159
160

	for (cpu = 0; cpu < NR_CPUS; cpu++)
161
		counts[cpu] = cpu_pda(cpu)->__nmi_count;
Linus Torvalds's avatar
Linus Torvalds committed
162
163
164
	local_irq_enable();
	mdelay((10*1000)/nmi_hz); // wait 10 ticks

165
	for_each_online_cpu(cpu) {
166
		if (cpu_pda(cpu)->__nmi_count - counts[cpu] <= 5) {
167
168
			endflag = 1;
			printk("CPU#%d: NMI appears to be stuck (%d->%d)!\n",
Linus Torvalds's avatar
Linus Torvalds committed
169
			       cpu,
170
			       counts[cpu],
171
			       cpu_pda(cpu)->__nmi_count);
Linus Torvalds's avatar
Linus Torvalds committed
172
173
			nmi_active = 0;
			lapic_nmi_owner &= ~LAPIC_NMI_WATCHDOG;
174
			nmi_perfctr_msr = 0;
175
			kfree(counts);
Linus Torvalds's avatar
Linus Torvalds committed
176
177
178
			return -1;
		}
	}
179
	endflag = 1;
Linus Torvalds's avatar
Linus Torvalds committed
180
181
182
183
184
185
186
	printk("OK.\n");

	/* now that we know it works we can reduce NMI frequency to
	   something more reasonable; makes a difference in some configs */
	if (nmi_watchdog == NMI_LOCAL_APIC)
		nmi_hz = 1;

187
	kfree(counts);
Linus Torvalds's avatar
Linus Torvalds committed
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
	return 0;
}

int __init setup_nmi_watchdog(char *str)
{
	int nmi;

	if (!strncmp(str,"panic",5)) {
		panic_on_timeout = 1;
		str = strchr(str, ',');
		if (!str)
			return 1;
		++str;
	}

	get_option(&str, &nmi);

	if (nmi >= NMI_INVALID)
		return 0;
207
	nmi_watchdog = nmi;
Linus Torvalds's avatar
Linus Torvalds committed
208
209
210
211
212
	return 1;
}

__setup("nmi_watchdog=", setup_nmi_watchdog);

213
214
static void disable_intel_arch_watchdog(void);

Linus Torvalds's avatar
Linus Torvalds committed
215
216
217
218
219
220
221
222
223
static void disable_lapic_nmi_watchdog(void)
{
	if (nmi_active <= 0)
		return;
	switch (boot_cpu_data.x86_vendor) {
	case X86_VENDOR_AMD:
		wrmsr(MSR_K7_EVNTSEL0, 0, 0);
		break;
	case X86_VENDOR_INTEL:
224
225
226
		if (boot_cpu_data.x86 == 15) {
			wrmsr(MSR_P4_IQ_CCCR0, 0, 0);
			wrmsr(MSR_P4_CRU_ESCR0, 0, 0);
227
228
		} else if (cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON)) {
			disable_intel_arch_watchdog();
229
		}
Linus Torvalds's avatar
Linus Torvalds committed
230
231
232
233
234
235
236
237
238
239
240
		break;
	}
	nmi_active = -1;
	/* tell do_nmi() and others that we're not active any more */
	nmi_watchdog = 0;
}

static void enable_lapic_nmi_watchdog(void)
{
	if (nmi_active < 0) {
		nmi_watchdog = NMI_LOCAL_APIC;
241
		touch_nmi_watchdog();
Linus Torvalds's avatar
Linus Torvalds committed
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
		setup_apic_nmi_watchdog();
	}
}

int reserve_lapic_nmi(void)
{
	unsigned int old_owner;

	spin_lock(&lapic_nmi_owner_lock);
	old_owner = lapic_nmi_owner;
	lapic_nmi_owner |= LAPIC_NMI_RESERVED;
	spin_unlock(&lapic_nmi_owner_lock);
	if (old_owner & LAPIC_NMI_RESERVED)
		return -EBUSY;
	if (old_owner & LAPIC_NMI_WATCHDOG)
		disable_lapic_nmi_watchdog();
	return 0;
}

void release_lapic_nmi(void)
{
	unsigned int new_owner;

	spin_lock(&lapic_nmi_owner_lock);
	new_owner = lapic_nmi_owner & ~LAPIC_NMI_RESERVED;
	lapic_nmi_owner = new_owner;
	spin_unlock(&lapic_nmi_owner_lock);
	if (new_owner & LAPIC_NMI_WATCHDOG)
		enable_lapic_nmi_watchdog();
}

void disable_timer_nmi_watchdog(void)
{
	if ((nmi_watchdog != NMI_IO_APIC) || (nmi_active <= 0))
		return;

	disable_irq(0);
	unset_nmi_callback();
	nmi_active = -1;
	nmi_watchdog = NMI_NONE;
}

void enable_timer_nmi_watchdog(void)
{
	if (nmi_active < 0) {
		nmi_watchdog = NMI_IO_APIC;
		touch_nmi_watchdog();
		nmi_active = 1;
		enable_irq(0);
	}
}

#ifdef CONFIG_PM

static int nmi_pm_active; /* nmi_active before suspend */

298
static int lapic_nmi_suspend(struct sys_device *dev, pm_message_t state)
Linus Torvalds's avatar
Linus Torvalds committed
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
{
	nmi_pm_active = nmi_active;
	disable_lapic_nmi_watchdog();
	return 0;
}

static int lapic_nmi_resume(struct sys_device *dev)
{
	if (nmi_pm_active > 0)
	enable_lapic_nmi_watchdog();
	return 0;
}

static struct sysdev_class nmi_sysclass = {
	set_kset_name("lapic_nmi"),
	.resume		= lapic_nmi_resume,
	.suspend	= lapic_nmi_suspend,
};

static struct sys_device device_lapic_nmi = {
	.id		= 0,
	.cls	= &nmi_sysclass,
};

static int __init init_lapic_nmi_sysfs(void)
{
	int error;

	if (nmi_active == 0 || nmi_watchdog != NMI_LOCAL_APIC)
		return 0;

	error = sysdev_class_register(&nmi_sysclass);
	if (!error)
		error = sysdev_register(&device_lapic_nmi);
	return error;
}
/* must come after the local APIC's device_initcall() */
late_initcall(init_lapic_nmi_sysfs);

#endif	/* CONFIG_PM */

/*
 * Activate the NMI watchdog via the local APIC.
 * Original code written by Keith Owens.
 */

345
346
347
348
349
350
351
352
static void clear_msr_range(unsigned int base, unsigned int n)
{
	unsigned int i;

	for(i = 0; i < n; ++i)
		wrmsr(base+i, 0, 0);
}

Linus Torvalds's avatar
Linus Torvalds committed
353
354
355
356
357
358
359
360
361
static void setup_k7_watchdog(void)
{
	int i;
	unsigned int evntsel;

	nmi_perfctr_msr = MSR_K7_PERFCTR0;

	for(i = 0; i < 4; ++i) {
		/* Simulator may not support it */
362
363
		if (checking_wrmsrl(MSR_K7_EVNTSEL0+i, 0UL)) {
			nmi_perfctr_msr = 0;
Linus Torvalds's avatar
Linus Torvalds committed
364
			return;
365
		}
Linus Torvalds's avatar
Linus Torvalds committed
366
367
368
369
370
371
372
373
374
		wrmsrl(MSR_K7_PERFCTR0+i, 0UL);
	}

	evntsel = K7_EVNTSEL_INT
		| K7_EVNTSEL_OS
		| K7_EVNTSEL_USR
		| K7_NMI_EVENT;

	wrmsr(MSR_K7_EVNTSEL0, evntsel, 0);
375
	wrmsrl(MSR_K7_PERFCTR0, -((u64)cpu_khz * 1000 / nmi_hz));
Linus Torvalds's avatar
Linus Torvalds committed
376
377
378
379
380
	apic_write(APIC_LVTPC, APIC_DM_NMI);
	evntsel |= K7_EVNTSEL_ENABLE;
	wrmsr(MSR_K7_EVNTSEL0, evntsel, 0);
}

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
static void disable_intel_arch_watchdog(void)
{
	unsigned ebx;

	/*
	 * Check whether the Architectural PerfMon supports
	 * Unhalted Core Cycles Event or not.
	 * NOTE: Corresponding bit = 0 in ebp indicates event present.
	 */
	ebx = cpuid_ebx(10);
	if (!(ebx & ARCH_PERFMON_UNHALTED_CORE_CYCLES_PRESENT))
		wrmsr(MSR_ARCH_PERFMON_EVENTSEL0, 0, 0);
}

static int setup_intel_arch_watchdog(void)
{
	unsigned int evntsel;
	unsigned ebx;

	/*
	 * Check whether the Architectural PerfMon supports
	 * Unhalted Core Cycles Event or not.
	 * NOTE: Corresponding bit = 0 in ebp indicates event present.
	 */
	ebx = cpuid_ebx(10);
	if ((ebx & ARCH_PERFMON_UNHALTED_CORE_CYCLES_PRESENT))
		return 0;

	nmi_perfctr_msr = MSR_ARCH_PERFMON_PERFCTR0;

	clear_msr_range(MSR_ARCH_PERFMON_EVENTSEL0, 2);
	clear_msr_range(MSR_ARCH_PERFMON_PERFCTR0, 2);

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

	wrmsr(MSR_ARCH_PERFMON_EVENTSEL0, evntsel, 0);
	wrmsrl(MSR_ARCH_PERFMON_PERFCTR0, -((u64)cpu_khz * 1000 / nmi_hz));
	apic_write(APIC_LVTPC, APIC_DM_NMI);
	evntsel |= ARCH_PERFMON_EVENTSEL0_ENABLE;
	wrmsr(MSR_ARCH_PERFMON_EVENTSEL0, evntsel, 0);
	return 1;
}

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

static int setup_p4_watchdog(void)
{
	unsigned int misc_enable, dummy;

	rdmsr(MSR_P4_MISC_ENABLE, misc_enable, dummy);
	if (!(misc_enable & MSR_P4_MISC_ENABLE_PERF_AVAIL))
		return 0;

	nmi_perfctr_msr = MSR_P4_IQ_COUNTER0;
	nmi_p4_cccr_val = P4_NMI_IQ_CCCR0;
#ifdef CONFIG_SMP
	if (smp_num_siblings == 2)
		nmi_p4_cccr_val |= P4_CCCR_OVF_PMI1;
#endif

	if (!(misc_enable & MSR_P4_MISC_ENABLE_PEBS_UNAVAIL))
		clear_msr_range(0x3F1, 2);
	/* MSR 0x3F0 seems to have a default value of 0xFC00, but current
	   docs doesn't fully define it, so leave it alone for now. */
	if (boot_cpu_data.x86_model >= 0x3) {
		/* MSR_P4_IQ_ESCR0/1 (0x3ba/0x3bb) removed */
		clear_msr_range(0x3A0, 26);
		clear_msr_range(0x3BC, 3);
	} else {
		clear_msr_range(0x3A0, 31);
	}
	clear_msr_range(0x3C0, 6);
	clear_msr_range(0x3C8, 6);
	clear_msr_range(0x3E0, 2);
	clear_msr_range(MSR_P4_CCCR0, 18);
	clear_msr_range(MSR_P4_PERFCTR0, 18);

	wrmsr(MSR_P4_CRU_ESCR0, P4_NMI_CRU_ESCR0, 0);
	wrmsr(MSR_P4_IQ_CCCR0, P4_NMI_IQ_CCCR0 & ~P4_CCCR_ENABLE, 0);
463
464
	Dprintk("setting P4_IQ_COUNTER0 to 0x%08lx\n", -(cpu_khz * 1000UL / nmi_hz));
	wrmsrl(MSR_P4_IQ_COUNTER0, -((u64)cpu_khz * 1000 / nmi_hz));
465
466
467
468
469
	apic_write(APIC_LVTPC, APIC_DM_NMI);
	wrmsr(MSR_P4_IQ_CCCR0, nmi_p4_cccr_val, 0);
	return 1;
}

Linus Torvalds's avatar
Linus Torvalds committed
470
471
472
473
void setup_apic_nmi_watchdog(void)
{
	switch (boot_cpu_data.x86_vendor) {
	case X86_VENDOR_AMD:
474
		if (boot_cpu_data.x86 != 15)
Linus Torvalds's avatar
Linus Torvalds committed
475
476
477
478
479
			return;
		if (strstr(boot_cpu_data.x86_model_id, "Screwdriver"))
			return;
		setup_k7_watchdog();
		break;
480
	case X86_VENDOR_INTEL:
481
482
483
484
485
486
487
		if (cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON)) {
			if (!setup_intel_arch_watchdog())
				return;
		} else if (boot_cpu_data.x86 == 15) {
			if (!setup_p4_watchdog())
				return;
		} else {
488
			return;
489
490
		}

491
492
		break;

Linus Torvalds's avatar
Linus Torvalds committed
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
	default:
		return;
	}
	lapic_nmi_owner = LAPIC_NMI_WATCHDOG;
	nmi_active = 1;
}

/*
 * the best way to detect whether a CPU has a 'hard lockup' problem
 * is to check it's local APIC timer IRQ counts. If they are not
 * changing then that CPU has some problem.
 *
 * as these watchdog NMI IRQs are generated on every CPU, we only
 * have to check the current processor.
 */

509
510
511
static DEFINE_PER_CPU(unsigned, last_irq_sum);
static DEFINE_PER_CPU(local_t, alert_counter);
static DEFINE_PER_CPU(int, nmi_touch);
Linus Torvalds's avatar
Linus Torvalds committed
512
513
514

void touch_nmi_watchdog (void)
{
515
516
	if (nmi_watchdog > 0) {
		unsigned cpu;
Linus Torvalds's avatar
Linus Torvalds committed
517

518
519
520
521
522
523
524
525
		/*
 		 * Tell other CPUs to reset their alert counters. We cannot
		 * do it ourselves because the alert count increase is not
		 * atomic.
		 */
		for_each_present_cpu (cpu)
			per_cpu(nmi_touch, cpu) = 1;
	}
Ingo Molnar's avatar
Ingo Molnar committed
526
527

 	touch_softlockup_watchdog();
Linus Torvalds's avatar
Linus Torvalds committed
528
529
}

530
void __kprobes nmi_watchdog_tick(struct pt_regs * regs, unsigned reason)
Linus Torvalds's avatar
Linus Torvalds committed
531
{
532
533
	int sum;
	int touched = 0;
Linus Torvalds's avatar
Linus Torvalds committed
534
535

	sum = read_pda(apic_timer_irqs);
536
537
538
539
	if (__get_cpu_var(nmi_touch)) {
		__get_cpu_var(nmi_touch) = 0;
		touched = 1;
	}
540
541
542
543
544
545
#ifdef CONFIG_X86_MCE
	/* Could check oops_in_progress here too, but it's safer
	   not too */
	if (atomic_read(&mce_entry) > 0)
		touched = 1;
#endif
546
	if (!touched && __get_cpu_var(last_irq_sum) == sum) {
Linus Torvalds's avatar
Linus Torvalds committed
547
548
549
550
		/*
		 * Ayiee, looks like this CPU is stuck ...
		 * wait a few IRQs (5 seconds) before doing the oops ...
		 */
551
552
		local_inc(&__get_cpu_var(alert_counter));
		if (local_read(&__get_cpu_var(alert_counter)) == 5*nmi_hz) {
Linus Torvalds's avatar
Linus Torvalds committed
553
554
			if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT)
							== NOTIFY_STOP) {
555
				local_set(&__get_cpu_var(alert_counter), 0);
Linus Torvalds's avatar
Linus Torvalds committed
556
				return;
557
558
			}
			die_nmi("NMI Watchdog detected LOCKUP on CPU %d\n", regs);
Linus Torvalds's avatar
Linus Torvalds committed
559
560
		}
	} else {
561
562
		__get_cpu_var(last_irq_sum) = sum;
		local_set(&__get_cpu_var(alert_counter), 0);
Linus Torvalds's avatar
Linus Torvalds committed
563
	}
564
565
566
567
568
569
570
571
572
573
574
	if (nmi_perfctr_msr) {
 		if (nmi_perfctr_msr == MSR_P4_IQ_COUNTER0) {
 			/*
 			 * 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.
 			 */
 			wrmsr(MSR_P4_IQ_CCCR0, nmi_p4_cccr_val, 0);
 			apic_write(APIC_LVTPC, APIC_DM_NMI);
575
576
577
578
579
580
581
582
 		} else if (nmi_perfctr_msr == MSR_ARCH_PERFMON_PERFCTR0) {
			/*
			 * For Intel based architectural perfmon
			 * - LVTPC is masked on interrupt and must be
			 *   unmasked by the LVTPC handler.
			 */
			apic_write(APIC_LVTPC, APIC_DM_NMI);
		}
583
		wrmsrl(nmi_perfctr_msr, -((u64)cpu_khz * 1000 / nmi_hz));
584
	}
Linus Torvalds's avatar
Linus Torvalds committed
585
586
}

587
static __kprobes int dummy_nmi_callback(struct pt_regs * regs, int cpu)
Linus Torvalds's avatar
Linus Torvalds committed
588
589
590
591
592
593
{
	return 0;
}
 
static nmi_callback_t nmi_callback = dummy_nmi_callback;
 
594
asmlinkage __kprobes void do_nmi(struct pt_regs * regs, long error_code)
Linus Torvalds's avatar
Linus Torvalds committed
595
596
597
598
599
{
	int cpu = safe_smp_processor_id();

	nmi_enter();
	add_pda(__nmi_count,1);
600
	if (!rcu_dereference(nmi_callback)(regs, cpu))
Linus Torvalds's avatar
Linus Torvalds committed
601
602
603
604
605
606
		default_do_nmi(regs);
	nmi_exit();
}

void set_nmi_callback(nmi_callback_t callback)
{
607
	vmalloc_sync_all();
608
	rcu_assign_pointer(nmi_callback, callback);
Linus Torvalds's avatar
Linus Torvalds committed
609
}
610
EXPORT_SYMBOL_GPL(set_nmi_callback);
Linus Torvalds's avatar
Linus Torvalds committed
611
612
613
614
615

void unset_nmi_callback(void)
{
	nmi_callback = dummy_nmi_callback;
}
616
EXPORT_SYMBOL_GPL(unset_nmi_callback);
Linus Torvalds's avatar
Linus Torvalds committed
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

#ifdef CONFIG_SYSCTL

static int unknown_nmi_panic_callback(struct pt_regs *regs, int cpu)
{
	unsigned char reason = get_nmi_reason();
	char buf[64];

	if (!(reason & 0xc0)) {
		sprintf(buf, "NMI received for unknown reason %02x\n", reason);
		die_nmi(buf,regs);
	}
	return 0;
}

/*
 * proc handler for /proc/sys/kernel/unknown_nmi_panic
 */
int proc_unknown_nmi_panic(struct ctl_table *table, int write, struct file *file,
			void __user *buffer, size_t *length, loff_t *ppos)
{
	int old_state;

	old_state = unknown_nmi_panic;
	proc_dointvec(table, write, file, buffer, length, ppos);
	if (!!old_state == !!unknown_nmi_panic)
		return 0;

	if (unknown_nmi_panic) {
		if (reserve_lapic_nmi() < 0) {
			unknown_nmi_panic = 0;
			return -EBUSY;
		} else {
			set_nmi_callback(unknown_nmi_panic_callback);
		}
	} else {
		release_lapic_nmi();
		unset_nmi_callback();
	}
	return 0;
}

#endif

EXPORT_SYMBOL(nmi_active);
EXPORT_SYMBOL(nmi_watchdog);
EXPORT_SYMBOL(reserve_lapic_nmi);
EXPORT_SYMBOL(release_lapic_nmi);
EXPORT_SYMBOL(disable_timer_nmi_watchdog);
EXPORT_SYMBOL(enable_timer_nmi_watchdog);
EXPORT_SYMBOL(touch_nmi_watchdog);