blob: 4d26d514c56f97aa37d8977b280565682fd4a2aa [file] [log] [blame]
Andi Kleen09198e62007-05-02 19:27:20 +02001/* local apic based NMI watchdog for various CPUs.
2 This file also handles reservation of performance counters for coordination
3 with other users (like oprofile).
4
5 Note that these events normally don't tick when the CPU idles. This means
6 the frequency varies with CPU load.
7
8 Original code for K7/P6 written by Keith Owens */
9
10#include <linux/percpu.h>
11#include <linux/module.h>
12#include <linux/kernel.h>
13#include <linux/bitops.h>
14#include <linux/smp.h>
15#include <linux/nmi.h>
16#include <asm/apic.h>
17#include <asm/intel_arch_perfmon.h>
18
19struct nmi_watchdog_ctlblk {
20 unsigned int cccr_msr;
21 unsigned int perfctr_msr; /* the MSR to reset in NMI handler */
22 unsigned int evntsel_msr; /* the MSR to select the events to handle */
23};
24
25/* Interface defining a CPU specific perfctr watchdog */
26struct wd_ops {
27 int (*reserve)(void);
28 void (*unreserve)(void);
29 int (*setup)(unsigned nmi_hz);
30 void (*rearm)(struct nmi_watchdog_ctlblk *wd, unsigned nmi_hz);
Björn Steinbrink54c6ed72007-06-16 10:15:56 -070031 void (*stop)(void);
Andi Kleen09198e62007-05-02 19:27:20 +020032 unsigned perfctr;
33 unsigned evntsel;
34 u64 checkbit;
35};
36
37static struct wd_ops *wd_ops;
38
39/* this number is calculated from Intel's MSR_P4_CRU_ESCR5 register and it's
40 * offset from MSR_P4_BSU_ESCR0. It will be the max for all platforms (for now)
41 */
42#define NMI_MAX_COUNTER_BITS 66
43
44/* perfctr_nmi_owner tracks the ownership of the perfctr registers:
45 * evtsel_nmi_owner tracks the ownership of the event selection
46 * - different performance counters/ event selection may be reserved for
47 * different subsystems this reservation system just tries to coordinate
48 * things a little
49 */
50static DECLARE_BITMAP(perfctr_nmi_owner, NMI_MAX_COUNTER_BITS);
51static DECLARE_BITMAP(evntsel_nmi_owner, NMI_MAX_COUNTER_BITS);
52
53static DEFINE_PER_CPU(struct nmi_watchdog_ctlblk, nmi_watchdog_ctlblk);
54
55/* converts an msr to an appropriate reservation bit */
56static inline unsigned int nmi_perfctr_msr_to_bit(unsigned int msr)
57{
Andi Kleen5dcccd82007-07-04 01:38:13 +020058 /* returns the bit offset of the performance counter register */
59 switch (boot_cpu_data.x86_vendor) {
60 case X86_VENDOR_AMD:
61 return (msr - MSR_K7_PERFCTR0);
62 case X86_VENDOR_INTEL:
63 if (cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON))
64 return (msr - MSR_ARCH_PERFMON_PERFCTR0);
65
66 switch (boot_cpu_data.x86) {
67 case 6:
68 return (msr - MSR_P6_PERFCTR0);
69 case 15:
70 return (msr - MSR_P4_BPU_PERFCTR0);
71 }
72 }
73 return 0;
Andi Kleen09198e62007-05-02 19:27:20 +020074}
75
76/* converts an msr to an appropriate reservation bit */
77/* returns the bit offset of the event selection register */
78static inline unsigned int nmi_evntsel_msr_to_bit(unsigned int msr)
79{
Andi Kleen5dcccd82007-07-04 01:38:13 +020080 /* returns the bit offset of the event selection register */
81 switch (boot_cpu_data.x86_vendor) {
82 case X86_VENDOR_AMD:
83 return (msr - MSR_K7_EVNTSEL0);
84 case X86_VENDOR_INTEL:
85 if (cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON))
86 return (msr - MSR_ARCH_PERFMON_EVENTSEL0);
87
88 switch (boot_cpu_data.x86) {
89 case 6:
90 return (msr - MSR_P6_EVNTSEL0);
91 case 15:
92 return (msr - MSR_P4_BSU_ESCR0);
93 }
94 }
95 return 0;
96
Andi Kleen09198e62007-05-02 19:27:20 +020097}
98
99/* checks for a bit availability (hack for oprofile) */
100int avail_to_resrv_perfctr_nmi_bit(unsigned int counter)
101{
102 BUG_ON(counter > NMI_MAX_COUNTER_BITS);
103
104 return (!test_bit(counter, perfctr_nmi_owner));
105}
106
107/* checks the an msr for availability */
108int avail_to_resrv_perfctr_nmi(unsigned int msr)
109{
110 unsigned int counter;
111
112 counter = nmi_perfctr_msr_to_bit(msr);
113 BUG_ON(counter > NMI_MAX_COUNTER_BITS);
114
115 return (!test_bit(counter, perfctr_nmi_owner));
116}
117
118int reserve_perfctr_nmi(unsigned int msr)
119{
120 unsigned int counter;
121
122 counter = nmi_perfctr_msr_to_bit(msr);
123 BUG_ON(counter > NMI_MAX_COUNTER_BITS);
124
125 if (!test_and_set_bit(counter, perfctr_nmi_owner))
126 return 1;
127 return 0;
128}
129
130void release_perfctr_nmi(unsigned int msr)
131{
132 unsigned int counter;
133
134 counter = nmi_perfctr_msr_to_bit(msr);
135 BUG_ON(counter > NMI_MAX_COUNTER_BITS);
136
137 clear_bit(counter, perfctr_nmi_owner);
138}
139
140int reserve_evntsel_nmi(unsigned int msr)
141{
142 unsigned int counter;
143
144 counter = nmi_evntsel_msr_to_bit(msr);
145 BUG_ON(counter > NMI_MAX_COUNTER_BITS);
146
147 if (!test_and_set_bit(counter, evntsel_nmi_owner))
148 return 1;
149 return 0;
150}
151
152void release_evntsel_nmi(unsigned int msr)
153{
154 unsigned int counter;
155
156 counter = nmi_evntsel_msr_to_bit(msr);
157 BUG_ON(counter > NMI_MAX_COUNTER_BITS);
158
159 clear_bit(counter, evntsel_nmi_owner);
160}
161
162EXPORT_SYMBOL(avail_to_resrv_perfctr_nmi);
163EXPORT_SYMBOL(avail_to_resrv_perfctr_nmi_bit);
164EXPORT_SYMBOL(reserve_perfctr_nmi);
165EXPORT_SYMBOL(release_perfctr_nmi);
166EXPORT_SYMBOL(reserve_evntsel_nmi);
167EXPORT_SYMBOL(release_evntsel_nmi);
168
169void disable_lapic_nmi_watchdog(void)
170{
171 BUG_ON(nmi_watchdog != NMI_LOCAL_APIC);
172
173 if (atomic_read(&nmi_active) <= 0)
174 return;
175
Björn Steinbrink54c6ed72007-06-16 10:15:56 -0700176 on_each_cpu(stop_apic_nmi_watchdog, NULL, 0, 1);
Andi Kleen09198e62007-05-02 19:27:20 +0200177 wd_ops->unreserve();
178
179 BUG_ON(atomic_read(&nmi_active) != 0);
180}
181
182void enable_lapic_nmi_watchdog(void)
183{
184 BUG_ON(nmi_watchdog != NMI_LOCAL_APIC);
185
186 /* are we already enabled */
187 if (atomic_read(&nmi_active) != 0)
188 return;
189
190 /* are we lapic aware */
191 if (!wd_ops)
192 return;
193 if (!wd_ops->reserve()) {
194 printk(KERN_ERR "NMI watchdog: cannot reserve perfctrs\n");
195 return;
196 }
197
198 on_each_cpu(setup_apic_nmi_watchdog, NULL, 0, 1);
199 touch_nmi_watchdog();
200}
201
202/*
203 * Activate the NMI watchdog via the local APIC.
204 */
205
206static unsigned int adjust_for_32bit_ctr(unsigned int hz)
207{
208 u64 counter_val;
209 unsigned int retval = hz;
210
211 /*
212 * On Intel CPUs with P6/ARCH_PERFMON only 32 bits in the counter
213 * are writable, with higher bits sign extending from bit 31.
214 * So, we can only program the counter with 31 bit values and
215 * 32nd bit should be 1, for 33.. to be 1.
216 * Find the appropriate nmi_hz
217 */
218 counter_val = (u64)cpu_khz * 1000;
219 do_div(counter_val, retval);
220 if (counter_val > 0x7fffffffULL) {
221 u64 count = (u64)cpu_khz * 1000;
222 do_div(count, 0x7fffffffUL);
223 retval = count + 1;
224 }
225 return retval;
226}
227
228static void
229write_watchdog_counter(unsigned int perfctr_msr, const char *descr, unsigned nmi_hz)
230{
231 u64 count = (u64)cpu_khz * 1000;
232
233 do_div(count, nmi_hz);
234 if(descr)
235 Dprintk("setting %s to -0x%08Lx\n", descr, count);
236 wrmsrl(perfctr_msr, 0 - count);
237}
238
239static void write_watchdog_counter32(unsigned int perfctr_msr,
240 const char *descr, unsigned nmi_hz)
241{
242 u64 count = (u64)cpu_khz * 1000;
243
244 do_div(count, nmi_hz);
245 if(descr)
246 Dprintk("setting %s to -0x%08Lx\n", descr, count);
247 wrmsr(perfctr_msr, (u32)(-count), 0);
248}
249
250/* AMD K7/K8/Family10h/Family11h support. AMD keeps this interface
251 nicely stable so there is not much variety */
252
253#define K7_EVNTSEL_ENABLE (1 << 22)
254#define K7_EVNTSEL_INT (1 << 20)
255#define K7_EVNTSEL_OS (1 << 17)
256#define K7_EVNTSEL_USR (1 << 16)
257#define K7_EVENT_CYCLES_PROCESSOR_IS_RUNNING 0x76
258#define K7_NMI_EVENT K7_EVENT_CYCLES_PROCESSOR_IS_RUNNING
259
260static int setup_k7_watchdog(unsigned nmi_hz)
261{
262 unsigned int perfctr_msr, evntsel_msr;
263 unsigned int evntsel;
264 struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);
265
266 perfctr_msr = MSR_K7_PERFCTR0;
267 evntsel_msr = MSR_K7_EVNTSEL0;
268
269 wrmsrl(perfctr_msr, 0UL);
270
271 evntsel = K7_EVNTSEL_INT
272 | K7_EVNTSEL_OS
273 | K7_EVNTSEL_USR
274 | K7_NMI_EVENT;
275
276 /* setup the timer */
277 wrmsr(evntsel_msr, evntsel, 0);
278 write_watchdog_counter(perfctr_msr, "K7_PERFCTR0",nmi_hz);
279 apic_write(APIC_LVTPC, APIC_DM_NMI);
280 evntsel |= K7_EVNTSEL_ENABLE;
281 wrmsr(evntsel_msr, evntsel, 0);
282
283 wd->perfctr_msr = perfctr_msr;
284 wd->evntsel_msr = evntsel_msr;
285 wd->cccr_msr = 0; //unused
286 return 1;
287}
288
Björn Steinbrink54c6ed72007-06-16 10:15:56 -0700289static void single_msr_stop_watchdog(void)
Andi Kleen09198e62007-05-02 19:27:20 +0200290{
291 struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);
292
293 wrmsr(wd->evntsel_msr, 0, 0);
294}
295
296static int single_msr_reserve(void)
297{
298 if (!reserve_perfctr_nmi(wd_ops->perfctr))
299 return 0;
300
301 if (!reserve_evntsel_nmi(wd_ops->evntsel)) {
302 release_perfctr_nmi(wd_ops->perfctr);
303 return 0;
304 }
305 return 1;
306}
307
308static void single_msr_unreserve(void)
309{
Björn Steinbrinkda88ba12007-06-16 10:16:04 -0700310 release_evntsel_nmi(wd_ops->evntsel);
311 release_perfctr_nmi(wd_ops->perfctr);
Andi Kleen09198e62007-05-02 19:27:20 +0200312}
313
314static void single_msr_rearm(struct nmi_watchdog_ctlblk *wd, unsigned nmi_hz)
315{
316 /* start the cycle over again */
317 write_watchdog_counter(wd->perfctr_msr, NULL, nmi_hz);
318}
319
320static struct wd_ops k7_wd_ops = {
321 .reserve = single_msr_reserve,
322 .unreserve = single_msr_unreserve,
323 .setup = setup_k7_watchdog,
324 .rearm = single_msr_rearm,
325 .stop = single_msr_stop_watchdog,
326 .perfctr = MSR_K7_PERFCTR0,
327 .evntsel = MSR_K7_EVNTSEL0,
328 .checkbit = 1ULL<<63,
329};
330
331/* Intel Model 6 (PPro+,P2,P3,P-M,Core1) */
332
333#define P6_EVNTSEL0_ENABLE (1 << 22)
334#define P6_EVNTSEL_INT (1 << 20)
335#define P6_EVNTSEL_OS (1 << 17)
336#define P6_EVNTSEL_USR (1 << 16)
337#define P6_EVENT_CPU_CLOCKS_NOT_HALTED 0x79
338#define P6_NMI_EVENT P6_EVENT_CPU_CLOCKS_NOT_HALTED
339
340static int setup_p6_watchdog(unsigned nmi_hz)
341{
342 unsigned int perfctr_msr, evntsel_msr;
343 unsigned int evntsel;
344 struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);
345
346 perfctr_msr = MSR_P6_PERFCTR0;
347 evntsel_msr = MSR_P6_EVNTSEL0;
348
349 wrmsrl(perfctr_msr, 0UL);
350
351 evntsel = P6_EVNTSEL_INT
352 | P6_EVNTSEL_OS
353 | P6_EVNTSEL_USR
354 | P6_NMI_EVENT;
355
356 /* setup the timer */
357 wrmsr(evntsel_msr, evntsel, 0);
358 nmi_hz = adjust_for_32bit_ctr(nmi_hz);
359 write_watchdog_counter32(perfctr_msr, "P6_PERFCTR0",nmi_hz);
360 apic_write(APIC_LVTPC, APIC_DM_NMI);
361 evntsel |= P6_EVNTSEL0_ENABLE;
362 wrmsr(evntsel_msr, evntsel, 0);
363
364 wd->perfctr_msr = perfctr_msr;
365 wd->evntsel_msr = evntsel_msr;
366 wd->cccr_msr = 0; //unused
367 return 1;
368}
369
370static void p6_rearm(struct nmi_watchdog_ctlblk *wd, unsigned nmi_hz)
371{
372 /* P6 based Pentium M need to re-unmask
373 * the apic vector but it doesn't hurt
374 * other P6 variant.
375 * ArchPerfom/Core Duo also needs this */
376 apic_write(APIC_LVTPC, APIC_DM_NMI);
377 /* P6/ARCH_PERFMON has 32 bit counter write */
378 write_watchdog_counter32(wd->perfctr_msr, NULL,nmi_hz);
379}
380
381static struct wd_ops p6_wd_ops = {
382 .reserve = single_msr_reserve,
383 .unreserve = single_msr_unreserve,
384 .setup = setup_p6_watchdog,
385 .rearm = p6_rearm,
386 .stop = single_msr_stop_watchdog,
387 .perfctr = MSR_P6_PERFCTR0,
388 .evntsel = MSR_P6_EVNTSEL0,
389 .checkbit = 1ULL<<39,
390};
391
392/* Intel P4 performance counters. By far the most complicated of all. */
393
394#define MSR_P4_MISC_ENABLE_PERF_AVAIL (1<<7)
395#define P4_ESCR_EVENT_SELECT(N) ((N)<<25)
396#define P4_ESCR_OS (1<<3)
397#define P4_ESCR_USR (1<<2)
398#define P4_CCCR_OVF_PMI0 (1<<26)
399#define P4_CCCR_OVF_PMI1 (1<<27)
400#define P4_CCCR_THRESHOLD(N) ((N)<<20)
401#define P4_CCCR_COMPLEMENT (1<<19)
402#define P4_CCCR_COMPARE (1<<18)
403#define P4_CCCR_REQUIRED (3<<16)
404#define P4_CCCR_ESCR_SELECT(N) ((N)<<13)
405#define P4_CCCR_ENABLE (1<<12)
406#define P4_CCCR_OVF (1<<31)
407
408/* Set up IQ_COUNTER0 to behave like a clock, by having IQ_CCCR0 filter
409 CRU_ESCR0 (with any non-null event selector) through a complemented
410 max threshold. [IA32-Vol3, Section 14.9.9] */
411
412static int setup_p4_watchdog(unsigned nmi_hz)
413{
414 unsigned int perfctr_msr, evntsel_msr, cccr_msr;
415 unsigned int evntsel, cccr_val;
416 unsigned int misc_enable, dummy;
417 unsigned int ht_num;
418 struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);
419
420 rdmsr(MSR_IA32_MISC_ENABLE, misc_enable, dummy);
421 if (!(misc_enable & MSR_P4_MISC_ENABLE_PERF_AVAIL))
422 return 0;
423
424#ifdef CONFIG_SMP
425 /* detect which hyperthread we are on */
426 if (smp_num_siblings == 2) {
427 unsigned int ebx, apicid;
428
429 ebx = cpuid_ebx(1);
430 apicid = (ebx >> 24) & 0xff;
431 ht_num = apicid & 1;
432 } else
433#endif
434 ht_num = 0;
435
436 /* performance counters are shared resources
437 * assign each hyperthread its own set
438 * (re-use the ESCR0 register, seems safe
439 * and keeps the cccr_val the same)
440 */
441 if (!ht_num) {
442 /* logical cpu 0 */
443 perfctr_msr = MSR_P4_IQ_PERFCTR0;
444 evntsel_msr = MSR_P4_CRU_ESCR0;
445 cccr_msr = MSR_P4_IQ_CCCR0;
446 cccr_val = P4_CCCR_OVF_PMI0 | P4_CCCR_ESCR_SELECT(4);
447 } else {
448 /* logical cpu 1 */
449 perfctr_msr = MSR_P4_IQ_PERFCTR1;
450 evntsel_msr = MSR_P4_CRU_ESCR0;
451 cccr_msr = MSR_P4_IQ_CCCR1;
452 cccr_val = P4_CCCR_OVF_PMI1 | P4_CCCR_ESCR_SELECT(4);
453 }
454
455 evntsel = P4_ESCR_EVENT_SELECT(0x3F)
456 | P4_ESCR_OS
457 | P4_ESCR_USR;
458
459 cccr_val |= P4_CCCR_THRESHOLD(15)
460 | P4_CCCR_COMPLEMENT
461 | P4_CCCR_COMPARE
462 | P4_CCCR_REQUIRED;
463
464 wrmsr(evntsel_msr, evntsel, 0);
465 wrmsr(cccr_msr, cccr_val, 0);
466 write_watchdog_counter(perfctr_msr, "P4_IQ_COUNTER0", nmi_hz);
467 apic_write(APIC_LVTPC, APIC_DM_NMI);
468 cccr_val |= P4_CCCR_ENABLE;
469 wrmsr(cccr_msr, cccr_val, 0);
470 wd->perfctr_msr = perfctr_msr;
471 wd->evntsel_msr = evntsel_msr;
472 wd->cccr_msr = cccr_msr;
473 return 1;
474}
475
Björn Steinbrink54c6ed72007-06-16 10:15:56 -0700476static void stop_p4_watchdog(void)
Andi Kleen09198e62007-05-02 19:27:20 +0200477{
478 struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);
479 wrmsr(wd->cccr_msr, 0, 0);
480 wrmsr(wd->evntsel_msr, 0, 0);
481}
482
483static int p4_reserve(void)
484{
485 if (!reserve_perfctr_nmi(MSR_P4_IQ_PERFCTR0))
486 return 0;
487#ifdef CONFIG_SMP
488 if (smp_num_siblings > 1 && !reserve_perfctr_nmi(MSR_P4_IQ_PERFCTR1))
489 goto fail1;
490#endif
491 if (!reserve_evntsel_nmi(MSR_P4_CRU_ESCR0))
492 goto fail2;
493 /* RED-PEN why is ESCR1 not reserved here? */
494 return 1;
495 fail2:
496#ifdef CONFIG_SMP
497 if (smp_num_siblings > 1)
498 release_perfctr_nmi(MSR_P4_IQ_PERFCTR1);
499 fail1:
500#endif
501 release_perfctr_nmi(MSR_P4_IQ_PERFCTR0);
502 return 0;
503}
504
505static void p4_unreserve(void)
506{
507#ifdef CONFIG_SMP
508 if (smp_num_siblings > 1)
Björn Steinbrinkda88ba12007-06-16 10:16:04 -0700509 release_perfctr_nmi(MSR_P4_IQ_PERFCTR1);
Andi Kleen09198e62007-05-02 19:27:20 +0200510#endif
Björn Steinbrinkda88ba12007-06-16 10:16:04 -0700511 release_evntsel_nmi(MSR_P4_CRU_ESCR0);
512 release_perfctr_nmi(MSR_P4_IQ_PERFCTR0);
Andi Kleen09198e62007-05-02 19:27:20 +0200513}
514
515static void p4_rearm(struct nmi_watchdog_ctlblk *wd, unsigned nmi_hz)
516{
517 unsigned dummy;
518 /*
519 * P4 quirks:
520 * - An overflown perfctr will assert its interrupt
521 * until the OVF flag in its CCCR is cleared.
522 * - LVTPC is masked on interrupt and must be
523 * unmasked by the LVTPC handler.
524 */
525 rdmsrl(wd->cccr_msr, dummy);
526 dummy &= ~P4_CCCR_OVF;
527 wrmsrl(wd->cccr_msr, dummy);
528 apic_write(APIC_LVTPC, APIC_DM_NMI);
529 /* start the cycle over again */
530 write_watchdog_counter(wd->perfctr_msr, NULL, nmi_hz);
531}
532
533static struct wd_ops p4_wd_ops = {
534 .reserve = p4_reserve,
535 .unreserve = p4_unreserve,
536 .setup = setup_p4_watchdog,
537 .rearm = p4_rearm,
538 .stop = stop_p4_watchdog,
539 /* RED-PEN this is wrong for the other sibling */
540 .perfctr = MSR_P4_BPU_PERFCTR0,
541 .evntsel = MSR_P4_BSU_ESCR0,
542 .checkbit = 1ULL<<39,
543};
544
545/* Watchdog using the Intel architected PerfMon. Used for Core2 and hopefully
546 all future Intel CPUs. */
547
548#define ARCH_PERFMON_NMI_EVENT_SEL ARCH_PERFMON_UNHALTED_CORE_CYCLES_SEL
549#define ARCH_PERFMON_NMI_EVENT_UMASK ARCH_PERFMON_UNHALTED_CORE_CYCLES_UMASK
550
551static int setup_intel_arch_watchdog(unsigned nmi_hz)
552{
553 unsigned int ebx;
554 union cpuid10_eax eax;
555 unsigned int unused;
556 unsigned int perfctr_msr, evntsel_msr;
557 unsigned int evntsel;
558 struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);
559
560 /*
561 * Check whether the Architectural PerfMon supports
562 * Unhalted Core Cycles Event or not.
563 * NOTE: Corresponding bit = 0 in ebx indicates event present.
564 */
565 cpuid(10, &(eax.full), &ebx, &unused, &unused);
566 if ((eax.split.mask_length < (ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX+1)) ||
567 (ebx & ARCH_PERFMON_UNHALTED_CORE_CYCLES_PRESENT))
568 return 0;
569
570 perfctr_msr = MSR_ARCH_PERFMON_PERFCTR1;
571 evntsel_msr = MSR_ARCH_PERFMON_EVENTSEL1;
572
573 wrmsrl(perfctr_msr, 0UL);
574
575 evntsel = ARCH_PERFMON_EVENTSEL_INT
576 | ARCH_PERFMON_EVENTSEL_OS
577 | ARCH_PERFMON_EVENTSEL_USR
578 | ARCH_PERFMON_NMI_EVENT_SEL
579 | ARCH_PERFMON_NMI_EVENT_UMASK;
580
581 /* setup the timer */
582 wrmsr(evntsel_msr, evntsel, 0);
583 nmi_hz = adjust_for_32bit_ctr(nmi_hz);
584 write_watchdog_counter32(perfctr_msr, "INTEL_ARCH_PERFCTR0", nmi_hz);
585 apic_write(APIC_LVTPC, APIC_DM_NMI);
586 evntsel |= ARCH_PERFMON_EVENTSEL0_ENABLE;
587 wrmsr(evntsel_msr, evntsel, 0);
588
589 wd->perfctr_msr = perfctr_msr;
590 wd->evntsel_msr = evntsel_msr;
591 wd->cccr_msr = 0; //unused
592 wd_ops->checkbit = 1ULL << (eax.split.bit_width - 1);
593 return 1;
594}
595
596static struct wd_ops intel_arch_wd_ops = {
597 .reserve = single_msr_reserve,
598 .unreserve = single_msr_unreserve,
599 .setup = setup_intel_arch_watchdog,
600 .rearm = p6_rearm,
601 .stop = single_msr_stop_watchdog,
602 .perfctr = MSR_ARCH_PERFMON_PERFCTR0,
603 .evntsel = MSR_ARCH_PERFMON_EVENTSEL0,
604};
605
606static void probe_nmi_watchdog(void)
607{
608 switch (boot_cpu_data.x86_vendor) {
609 case X86_VENDOR_AMD:
610 if (boot_cpu_data.x86 != 6 && boot_cpu_data.x86 != 15 &&
611 boot_cpu_data.x86 != 16)
612 return;
613 wd_ops = &k7_wd_ops;
614 break;
615 case X86_VENDOR_INTEL:
616 if (cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON)) {
617 wd_ops = &intel_arch_wd_ops;
618 break;
619 }
620 switch (boot_cpu_data.x86) {
621 case 6:
622 if (boot_cpu_data.x86_model > 0xd)
623 return;
624
625 wd_ops = &p6_wd_ops;
626 break;
627 case 15:
628 if (boot_cpu_data.x86_model > 0x4)
629 return;
630
631 wd_ops = &p4_wd_ops;
632 break;
633 default:
634 return;
635 }
636 break;
637 }
638}
639
640/* Interface to nmi.c */
641
642int lapic_watchdog_init(unsigned nmi_hz)
643{
644 if (!wd_ops) {
645 probe_nmi_watchdog();
646 if (!wd_ops)
647 return -1;
Björn Steinbrinkfaa4cfa2007-06-16 10:15:55 -0700648
649 if (!wd_ops->reserve()) {
650 printk(KERN_ERR
651 "NMI watchdog: cannot reserve perfctrs\n");
652 return -1;
653 }
Andi Kleen09198e62007-05-02 19:27:20 +0200654 }
655
656 if (!(wd_ops->setup(nmi_hz))) {
657 printk(KERN_ERR "Cannot setup NMI watchdog on CPU %d\n",
658 raw_smp_processor_id());
659 return -1;
660 }
661
662 return 0;
663}
664
665void lapic_watchdog_stop(void)
666{
667 if (wd_ops)
Björn Steinbrink54c6ed72007-06-16 10:15:56 -0700668 wd_ops->stop();
Andi Kleen09198e62007-05-02 19:27:20 +0200669}
670
671unsigned lapic_adjust_nmi_hz(unsigned hz)
672{
673 struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);
674 if (wd->perfctr_msr == MSR_P6_PERFCTR0 ||
675 wd->perfctr_msr == MSR_ARCH_PERFMON_PERFCTR1)
676 hz = adjust_for_32bit_ctr(hz);
677 return hz;
678}
679
680int lapic_wd_event(unsigned nmi_hz)
681{
682 struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);
683 u64 ctr;
684 rdmsrl(wd->perfctr_msr, ctr);
685 if (ctr & wd_ops->checkbit) { /* perfctr still running? */
686 return 0;
687 }
688 wd_ops->rearm(wd, nmi_hz);
689 return 1;
690}
691
692int lapic_watchdog_ok(void)
693{
694 return wd_ops != NULL;
695}