summaryrefslogtreecommitdiff
path: root/arch/powerpc/kernel/perf_counter.c
blob: df007fe0cc0bb6e6c989dc273a82349df8e5af99 (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
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
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
/*
 * Performance counter support - powerpc architecture code
 *
 * Copyright 2008-2009 Paul Mackerras, IBM Corporation.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; either version
 * 2 of the License, or (at your option) any later version.
 */
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/perf_counter.h>
#include <linux/percpu.h>
#include <linux/hardirq.h>
#include <asm/reg.h>
#include <asm/pmc.h>
#include <asm/machdep.h>
#include <asm/firmware.h>

struct cpu_hw_counters {
	int n_counters;
	int n_percpu;
	int disabled;
	int n_added;
	struct perf_counter *counter[MAX_HWCOUNTERS];
	unsigned int events[MAX_HWCOUNTERS];
	u64 mmcr[3];
	u8 pmcs_enabled;
};
DEFINE_PER_CPU(struct cpu_hw_counters, cpu_hw_counters);

struct power_pmu *ppmu;

/*
 * Normally, to ignore kernel events we set the FCS (freeze counters
 * in supervisor mode) bit in MMCR0, but if the kernel runs with the
 * hypervisor bit set in the MSR, or if we are running on a processor
 * where the hypervisor bit is forced to 1 (as on Apple G5 processors),
 * then we need to use the FCHV bit to ignore kernel events.
 */
static unsigned int freeze_counters_kernel = MMCR0_FCS;

void perf_counter_print_debug(void)
{
}

/*
 * Read one performance monitor counter (PMC).
 */
static unsigned long read_pmc(int idx)
{
	unsigned long val;

	switch (idx) {
	case 1:
		val = mfspr(SPRN_PMC1);
		break;
	case 2:
		val = mfspr(SPRN_PMC2);
		break;
	case 3:
		val = mfspr(SPRN_PMC3);
		break;
	case 4:
		val = mfspr(SPRN_PMC4);
		break;
	case 5:
		val = mfspr(SPRN_PMC5);
		break;
	case 6:
		val = mfspr(SPRN_PMC6);
		break;
	case 7:
		val = mfspr(SPRN_PMC7);
		break;
	case 8:
		val = mfspr(SPRN_PMC8);
		break;
	default:
		printk(KERN_ERR "oops trying to read PMC%d\n", idx);
		val = 0;
	}
	return val;
}

/*
 * Write one PMC.
 */
static void write_pmc(int idx, unsigned long val)
{
	switch (idx) {
	case 1:
		mtspr(SPRN_PMC1, val);
		break;
	case 2:
		mtspr(SPRN_PMC2, val);
		break;
	case 3:
		mtspr(SPRN_PMC3, val);
		break;
	case 4:
		mtspr(SPRN_PMC4, val);
		break;
	case 5:
		mtspr(SPRN_PMC5, val);
		break;
	case 6:
		mtspr(SPRN_PMC6, val);
		break;
	case 7:
		mtspr(SPRN_PMC7, val);
		break;
	case 8:
		mtspr(SPRN_PMC8, val);
		break;
	default:
		printk(KERN_ERR "oops trying to write PMC%d\n", idx);
	}
}

/*
 * Check if a set of events can all go on the PMU at once.
 * If they can't, this will look at alternative codes for the events
 * and see if any combination of alternative codes is feasible.
 * The feasible set is returned in event[].
 */
static int power_check_constraints(unsigned int event[], int n_ev)
{
	u64 mask, value, nv;
	unsigned int alternatives[MAX_HWCOUNTERS][MAX_EVENT_ALTERNATIVES];
	u64 amasks[MAX_HWCOUNTERS][MAX_EVENT_ALTERNATIVES];
	u64 avalues[MAX_HWCOUNTERS][MAX_EVENT_ALTERNATIVES];
	u64 smasks[MAX_HWCOUNTERS], svalues[MAX_HWCOUNTERS];
	int n_alt[MAX_HWCOUNTERS], choice[MAX_HWCOUNTERS];
	int i, j;
	u64 addf = ppmu->add_fields;
	u64 tadd = ppmu->test_adder;

	if (n_ev > ppmu->n_counter)
		return -1;

	/* First see if the events will go on as-is */
	for (i = 0; i < n_ev; ++i) {
		alternatives[i][0] = event[i];
		if (ppmu->get_constraint(event[i], &amasks[i][0],
					 &avalues[i][0]))
			return -1;
		choice[i] = 0;
	}
	value = mask = 0;
	for (i = 0; i < n_ev; ++i) {
		nv = (value | avalues[i][0]) + (value & avalues[i][0] & addf);
		if ((((nv + tadd) ^ value) & mask) != 0 ||
		    (((nv + tadd) ^ avalues[i][0]) & amasks[i][0]) != 0)
			break;
		value = nv;
		mask |= amasks[i][0];
	}
	if (i == n_ev)
		return 0;	/* all OK */

	/* doesn't work, gather alternatives... */
	if (!ppmu->get_alternatives)
		return -1;
	for (i = 0; i < n_ev; ++i) {
		n_alt[i] = ppmu->get_alternatives(event[i], alternatives[i]);
		for (j = 1; j < n_alt[i]; ++j)
			ppmu->get_constraint(alternatives[i][j],
					     &amasks[i][j], &avalues[i][j]);
	}

	/* enumerate all possibilities and see if any will work */
	i = 0;
	j = -1;
	value = mask = nv = 0;
	while (i < n_ev) {
		if (j >= 0) {
			/* we're backtracking, restore context */
			value = svalues[i];
			mask = smasks[i];
			j = choice[i];
		}
		/*
		 * See if any alternative k for event i,
		 * where k > j, will satisfy the constraints.
		 */
		while (++j < n_alt[i]) {
			nv = (value | avalues[i][j]) +
				(value & avalues[i][j] & addf);
			if ((((nv + tadd) ^ value) & mask) == 0 &&
			    (((nv + tadd) ^ avalues[i][j])
			     & amasks[i][j]) == 0)
				break;
		}
		if (j >= n_alt[i]) {
			/*
			 * No feasible alternative, backtrack
			 * to event i-1 and continue enumerating its
			 * alternatives from where we got up to.
			 */
			if (--i < 0)
				return -1;
		} else {
			/*
			 * Found a feasible alternative for event i,
			 * remember where we got up to with this event,
			 * go on to the next event, and start with
			 * the first alternative for it.
			 */
			choice[i] = j;
			svalues[i] = value;
			smasks[i] = mask;
			value = nv;
			mask |= amasks[i][j];
			++i;
			j = -1;
		}
	}

	/* OK, we have a feasible combination, tell the caller the solution */
	for (i = 0; i < n_ev; ++i)
		event[i] = alternatives[i][choice[i]];
	return 0;
}

/*
 * Check if newly-added counters have consistent settings for
 * exclude_{user,kernel,hv} with each other and any previously
 * added counters.
 */
static int check_excludes(struct perf_counter **ctrs, int n_prev, int n_new)
{
	int eu, ek, eh;
	int i, n;
	struct perf_counter *counter;

	n = n_prev + n_new;
	if (n <= 1)
		return 0;

	eu = ctrs[0]->hw_event.exclude_user;
	ek = ctrs[0]->hw_event.exclude_kernel;
	eh = ctrs[0]->hw_event.exclude_hv;
	if (n_prev == 0)
		n_prev = 1;
	for (i = n_prev; i < n; ++i) {
		counter = ctrs[i];
		if (counter->hw_event.exclude_user != eu ||
		    counter->hw_event.exclude_kernel != ek ||
		    counter->hw_event.exclude_hv != eh)
			return -EAGAIN;
	}
	return 0;
}

static void power_perf_read(struct perf_counter *counter)
{
	long val, delta, prev;

	if (!counter->hw.idx)
		return;
	/*
	 * Performance monitor interrupts come even when interrupts
	 * are soft-disabled, as long as interrupts are hard-enabled.
	 * Therefore we treat them like NMIs.
	 */
	do {
		prev = atomic64_read(&counter->hw.prev_count);
		barrier();
		val = read_pmc(counter->hw.idx);
	} while (atomic64_cmpxchg(&counter->hw.prev_count, prev, val) != prev);

	/* The counters are only 32 bits wide */
	delta = (val - prev) & 0xfffffffful;
	atomic64_add(delta, &counter->count);
	atomic64_sub(delta, &counter->hw.period_left);
}

/*
 * Disable all counters to prevent PMU interrupts and to allow
 * counters to be added or removed.
 */
u64 hw_perf_save_disable(void)
{
	struct cpu_hw_counters *cpuhw;
	unsigned long ret;
	unsigned long flags;

	local_irq_save(flags);
	cpuhw = &__get_cpu_var(cpu_hw_counters);

	ret = cpuhw->disabled;
	if (!ret) {
		cpuhw->disabled = 1;
		cpuhw->n_added = 0;

		/*
		 * Check if we ever enabled the PMU on this cpu.
		 */
		if (!cpuhw->pmcs_enabled) {
			if (ppc_md.enable_pmcs)
				ppc_md.enable_pmcs();
			cpuhw->pmcs_enabled = 1;
		}

		/*
		 * Set the 'freeze counters' bit.
		 * The barrier is to make sure the mtspr has been
		 * executed and the PMU has frozen the counters
		 * before we return.
		 */
		mtspr(SPRN_MMCR0, mfspr(SPRN_MMCR0) | MMCR0_FC);
		mb();
	}
	local_irq_restore(flags);
	return ret;
}

/*
 * Re-enable all counters if disable == 0.
 * If we were previously disabled and counters were added, then
 * put the new config on the PMU.
 */
void hw_perf_restore(u64 disable)
{
	struct perf_counter *counter;
	struct cpu_hw_counters *cpuhw;
	unsigned long flags;
	long i;
	unsigned long val;
	s64 left;
	unsigned int hwc_index[MAX_HWCOUNTERS];

	if (disable)
		return;
	local_irq_save(flags);
	cpuhw = &__get_cpu_var(cpu_hw_counters);
	cpuhw->disabled = 0;

	/*
	 * If we didn't change anything, or only removed counters,
	 * no need to recalculate MMCR* settings and reset the PMCs.
	 * Just reenable the PMU with the current MMCR* settings
	 * (possibly updated for removal of counters).
	 */
	if (!cpuhw->n_added) {
		mtspr(SPRN_MMCRA, cpuhw->mmcr[2]);
		mtspr(SPRN_MMCR1, cpuhw->mmcr[1]);
		mtspr(SPRN_MMCR0, cpuhw->mmcr[0]);
		if (cpuhw->n_counters == 0)
			get_lppaca()->pmcregs_in_use = 0;
		goto out;
	}

	/*
	 * Compute MMCR* values for the new set of counters
	 */
	if (ppmu->compute_mmcr(cpuhw->events, cpuhw->n_counters, hwc_index,
			       cpuhw->mmcr)) {
		/* shouldn't ever get here */
		printk(KERN_ERR "oops compute_mmcr failed\n");
		goto out;
	}

	/*
	 * Add in MMCR0 freeze bits corresponding to the
	 * hw_event.exclude_* bits for the first counter.
	 * We have already checked that all counters have the
	 * same values for these bits as the first counter.
	 */
	counter = cpuhw->counter[0];
	if (counter->hw_event.exclude_user)
		cpuhw->mmcr[0] |= MMCR0_FCP;
	if (counter->hw_event.exclude_kernel)
		cpuhw->mmcr[0] |= freeze_counters_kernel;
	if (counter->hw_event.exclude_hv)
		cpuhw->mmcr[0] |= MMCR0_FCHV;

	/*
	 * Write the new configuration to MMCR* with the freeze
	 * bit set and set the hardware counters to their initial values.
	 * Then unfreeze the counters.
	 */
	get_lppaca()->pmcregs_in_use = 1;
	mtspr(SPRN_MMCRA, cpuhw->mmcr[2]);
	mtspr(SPRN_MMCR1, cpuhw->mmcr[1]);
	mtspr(SPRN_MMCR0, (cpuhw->mmcr[0] & ~(MMCR0_PMC1CE | MMCR0_PMCjCE))
				| MMCR0_FC);

	/*
	 * Read off any pre-existing counters that need to move
	 * to another PMC.
	 */
	for (i = 0; i < cpuhw->n_counters; ++i) {
		counter = cpuhw->counter[i];
		if (counter->hw.idx && counter->hw.idx != hwc_index[i] + 1) {
			power_perf_read(counter);
			write_pmc(counter->hw.idx, 0);
			counter->hw.idx = 0;
		}
	}

	/*
	 * Initialize the PMCs for all the new and moved counters.
	 */
	for (i = 0; i < cpuhw->n_counters; ++i) {
		counter = cpuhw->counter[i];
		if (counter->hw.idx)
			continue;
		val = 0;
		if (counter->hw_event.irq_period) {
			left = atomic64_read(&counter->hw.period_left);
			if (left < 0x80000000L)
				val = 0x80000000L - left;
		}
		atomic64_set(&counter->hw.prev_count, val);
		counter->hw.idx = hwc_index[i] + 1;
		write_pmc(counter->hw.idx, val);
		perf_counter_update_userpage(counter);
	}
	mb();
	cpuhw->mmcr[0] |= MMCR0_PMXE | MMCR0_FCECE;
	mtspr(SPRN_MMCR0, cpuhw->mmcr[0]);

 out:
	local_irq_restore(flags);
}

static int collect_events(struct perf_counter *group, int max_count,
			  struct perf_counter *ctrs[], unsigned int *events)
{
	int n = 0;
	struct perf_counter *counter;

	if (!is_software_counter(group)) {
		if (n >= max_count)
			return -1;
		ctrs[n] = group;
		events[n++] = group->hw.config;
	}
	list_for_each_entry(counter, &group->sibling_list, list_entry) {
		if (!is_software_counter(counter) &&
		    counter->state != PERF_COUNTER_STATE_OFF) {
			if (n >= max_count)
				return -1;
			ctrs[n] = counter;
			events[n++] = counter->hw.config;
		}
	}
	return n;
}

static void counter_sched_in(struct perf_counter *counter, int cpu)
{
	counter->state = PERF_COUNTER_STATE_ACTIVE;
	counter->oncpu = cpu;
	counter->tstamp_running += counter->ctx->time_now -
		counter->tstamp_stopped;
	if (is_software_counter(counter))
		counter->hw_ops->enable(counter);
}

/*
 * Called to enable a whole group of counters.
 * Returns 1 if the group was enabled, or -EAGAIN if it could not be.
 * Assumes the caller has disabled interrupts and has
 * frozen the PMU with hw_perf_save_disable.
 */
int hw_perf_group_sched_in(struct perf_counter *group_leader,
	       struct perf_cpu_context *cpuctx,
	       struct perf_counter_context *ctx, int cpu)
{
	struct cpu_hw_counters *cpuhw;
	long i, n, n0;
	struct perf_counter *sub;

	cpuhw = &__get_cpu_var(cpu_hw_counters);
	n0 = cpuhw->n_counters;
	n = collect_events(group_leader, ppmu->n_counter - n0,
			   &cpuhw->counter[n0], &cpuhw->events[n0]);
	if (n < 0)
		return -EAGAIN;
	if (check_excludes(cpuhw->counter, n0, n))
		return -EAGAIN;
	if (power_check_constraints(cpuhw->events, n + n0))
		return -EAGAIN;
	cpuhw->n_counters = n0 + n;
	cpuhw->n_added += n;

	/*
	 * OK, this group can go on; update counter states etc.,
	 * and enable any software counters
	 */
	for (i = n0; i < n0 + n; ++i)
		cpuhw->counter[i]->hw.config = cpuhw->events[i];
	cpuctx->active_oncpu += n;
	n = 1;
	counter_sched_in(group_leader, cpu);
	list_for_each_entry(sub, &group_leader->sibling_list, list_entry) {
		if (sub->state != PERF_COUNTER_STATE_OFF) {
			counter_sched_in(sub, cpu);
			++n;
		}
	}
	ctx->nr_active += n;

	return 1;
}

/*
 * Add a counter to the PMU.
 * If all counters are not already frozen, then we disable and
 * re-enable the PMU in order to get hw_perf_restore to do the
 * actual work of reconfiguring the PMU.
 */
static int power_perf_enable(struct perf_counter *counter)
{
	struct cpu_hw_counters *cpuhw;
	unsigned long flags;
	u64 pmudis;
	int n0;
	int ret = -EAGAIN;

	local_irq_save(flags);
	pmudis = hw_perf_save_disable();

	/*
	 * Add the counter to the list (if there is room)
	 * and check whether the total set is still feasible.
	 */
	cpuhw = &__get_cpu_var(cpu_hw_counters);
	n0 = cpuhw->n_counters;
	if (n0 >= ppmu->n_counter)
		goto out;
	cpuhw->counter[n0] = counter;
	cpuhw->events[n0] = counter->hw.config;
	if (check_excludes(cpuhw->counter, n0, 1))
		goto out;
	if (power_check_constraints(cpuhw->events, n0 + 1))
		goto out;

	counter->hw.config = cpuhw->events[n0];
	++cpuhw->n_counters;
	++cpuhw->n_added;

	ret = 0;
 out:
	hw_perf_restore(pmudis);
	local_irq_restore(flags);
	return ret;
}

/*
 * Remove a counter from the PMU.
 */
static void power_perf_disable(struct perf_counter *counter)
{
	struct cpu_hw_counters *cpuhw;
	long i;
	u64 pmudis;
	unsigned long flags;

	local_irq_save(flags);
	pmudis = hw_perf_save_disable();

	power_perf_read(counter);

	cpuhw = &__get_cpu_var(cpu_hw_counters);
	for (i = 0; i < cpuhw->n_counters; ++i) {
		if (counter == cpuhw->counter[i]) {
			while (++i < cpuhw->n_counters)
				cpuhw->counter[i-1] = cpuhw->counter[i];
			--cpuhw->n_counters;
			ppmu->disable_pmc(counter->hw.idx - 1, cpuhw->mmcr);
			write_pmc(counter->hw.idx, 0);
			counter->hw.idx = 0;
			perf_counter_update_userpage(counter);
			break;
		}
	}
	if (cpuhw->n_counters == 0) {
		/* disable exceptions if no counters are running */
		cpuhw->mmcr[0] &= ~(MMCR0_PMXE | MMCR0_FCECE);
	}

	hw_perf_restore(pmudis);
	local_irq_restore(flags);
}

struct hw_perf_counter_ops power_perf_ops = {
	.enable = power_perf_enable,
	.disable = power_perf_disable,
	.read = power_perf_read
};

const struct hw_perf_counter_ops *
hw_perf_counter_init(struct perf_counter *counter)
{
	unsigned long ev;
	struct perf_counter *ctrs[MAX_HWCOUNTERS];
	unsigned int events[MAX_HWCOUNTERS];
	int n;

	if (!ppmu)
		return NULL;
	if ((s64)counter->hw_event.irq_period < 0)
		return NULL;
	if (!perf_event_raw(&counter->hw_event)) {
		ev = perf_event_id(&counter->hw_event);
		if (ev >= ppmu->n_generic || ppmu->generic_events[ev] == 0)
			return NULL;
		ev = ppmu->generic_events[ev];
	} else {
		ev = perf_event_config(&counter->hw_event);
	}
	counter->hw.config_base = ev;
	counter->hw.idx = 0;

	/*
	 * If we are not running on a hypervisor, force the
	 * exclude_hv bit to 0 so that we don't care what
	 * the user set it to.
	 */
	if (!firmware_has_feature(FW_FEATURE_LPAR))
		counter->hw_event.exclude_hv = 0;
	
	/*
	 * If this is in a group, check if it can go on with all the
	 * other hardware counters in the group.  We assume the counter
	 * hasn't been linked into its leader's sibling list at this point.
	 */
	n = 0;
	if (counter->group_leader != counter) {
		n = collect_events(counter->group_leader, ppmu->n_counter - 1,
				   ctrs, events);
		if (n < 0)
			return NULL;
	}
	events[n] = ev;
	ctrs[n] = counter;
	if (check_excludes(ctrs, n, 1))
		return NULL;
	if (power_check_constraints(events, n + 1))
		return NULL;

	counter->hw.config = events[n];
	atomic64_set(&counter->hw.period_left, counter->hw_event.irq_period);
	return &power_perf_ops;
}

/*
 * Handle wakeups.
 */
void perf_counter_do_pending(void)
{
	int i;
	struct cpu_hw_counters *cpuhw = &__get_cpu_var(cpu_hw_counters);
	struct perf_counter *counter;

	for (i = 0; i < cpuhw->n_counters; ++i) {
		counter = cpuhw->counter[i];
		if (counter && counter->wakeup_pending) {
			counter->wakeup_pending = 0;
			wake_up(&counter->waitq);
		}
	}
}

/*
 * A counter has overflowed; update its count and record
 * things if requested.  Note that interrupts are hard-disabled
 * here so there is no possibility of being interrupted.
 */
static void record_and_restart(struct perf_counter *counter, long val,
			       struct pt_regs *regs)
{
	s64 prev, delta, left;
	int record = 0;

	/* we don't have to worry about interrupts here */
	prev = atomic64_read(&counter->hw.prev_count);
	delta = (val - prev) & 0xfffffffful;
	atomic64_add(delta, &counter->count);

	/*
	 * See if the total period for this counter has expired,
	 * and update for the next period.
	 */
	val = 0;
	left = atomic64_read(&counter->hw.period_left) - delta;
	if (counter->hw_event.irq_period) {
		if (left <= 0) {
			left += counter->hw_event.irq_period;
			if (left <= 0)
				left = counter->hw_event.irq_period;
			record = 1;
		}
		if (left < 0x80000000L)
			val = 0x80000000L - left;
	}
	write_pmc(counter->hw.idx, val);
	atomic64_set(&counter->hw.prev_count, val);
	atomic64_set(&counter->hw.period_left, left);
	perf_counter_update_userpage(counter);

	/*
	 * Finally record data if requested.
	 */
	if (record)
		perf_counter_output(counter, 1, regs);
}

/*
 * Performance monitor interrupt stuff
 */
static void perf_counter_interrupt(struct pt_regs *regs)
{
	int i;
	struct cpu_hw_counters *cpuhw = &__get_cpu_var(cpu_hw_counters);
	struct perf_counter *counter;
	long val;
	int need_wakeup = 0, found = 0;

	for (i = 0; i < cpuhw->n_counters; ++i) {
		counter = cpuhw->counter[i];
		val = read_pmc(counter->hw.idx);
		if ((int)val < 0) {
			/* counter has overflowed */
			found = 1;
			record_and_restart(counter, val, regs);
		}
	}

	/*
	 * In case we didn't find and reset the counter that caused
	 * the interrupt, scan all counters and reset any that are
	 * negative, to avoid getting continual interrupts.
	 * Any that we processed in the previous loop will not be negative.
	 */
	if (!found) {
		for (i = 0; i < ppmu->n_counter; ++i) {
			val = read_pmc(i + 1);
			if ((int)val < 0)
				write_pmc(i + 1, 0);
		}
	}

	/*
	 * Reset MMCR0 to its normal value.  This will set PMXE and
	 * clear FC (freeze counters) and PMAO (perf mon alert occurred)
	 * and thus allow interrupts to occur again.
	 * XXX might want to use MSR.PM to keep the counters frozen until
	 * we get back out of this interrupt.
	 */
	mtspr(SPRN_MMCR0, cpuhw->mmcr[0]);

	/*
	 * If we need a wakeup, check whether interrupts were soft-enabled
	 * when we took the interrupt.  If they were, we can wake stuff up
	 * immediately; otherwise we'll have do the wakeup when interrupts
	 * get soft-enabled.
	 */
	if (get_perf_counter_pending() && regs->softe) {
		irq_enter();
		clear_perf_counter_pending();
		perf_counter_do_pending();
		irq_exit();
	}
}

void hw_perf_counter_setup(int cpu)
{
	struct cpu_hw_counters *cpuhw = &per_cpu(cpu_hw_counters, cpu);

	memset(cpuhw, 0, sizeof(*cpuhw));
	cpuhw->mmcr[0] = MMCR0_FC;
}

extern struct power_pmu power4_pmu;
extern struct power_pmu ppc970_pmu;
extern struct power_pmu power5_pmu;
extern struct power_pmu power5p_pmu;
extern struct power_pmu power6_pmu;

static int init_perf_counters(void)
{
	unsigned long pvr;

	if (reserve_pmc_hardware(perf_counter_interrupt)) {
		printk(KERN_ERR "Couldn't init performance monitor subsystem\n");
		return -EBUSY;
	}

	/* XXX should get this from cputable */
	pvr = mfspr(SPRN_PVR);
	switch (PVR_VER(pvr)) {
	case PV_POWER4:
	case PV_POWER4p:
		ppmu = &power4_pmu;
		break;
	case PV_970:
	case PV_970FX:
	case PV_970MP:
		ppmu = &ppc970_pmu;
		break;
	case PV_POWER5:
		ppmu = &power5_pmu;
		break;
	case PV_POWER5p:
		ppmu = &power5p_pmu;
		break;
	case 0x3e:
		ppmu = &power6_pmu;
		break;
	}

	/*
	 * Use FCHV to ignore kernel events if MSR.HV is set.
	 */
	if (mfmsr() & MSR_HV)
		freeze_counters_kernel = MMCR0_FCHV;

	return 0;
}

arch_initcall(init_perf_counters);