summaryrefslogtreecommitdiffstats
path: root/arch/x86/kernel/tsc.c
blob: acb3b606613eb5b4937cdacab17adc74c4ec2c30 (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
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/init.h>
#include <linux/module.h>
#include <linux/timer.h>
#include <linux/acpi_pmtmr.h>
#include <linux/cpufreq.h>
#include <linux/delay.h>
#include <linux/clocksource.h>
#include <linux/percpu.h>
#include <linux/timex.h>
#include <linux/static_key.h>

#include <asm/hpet.h>
#include <asm/timer.h>
#include <asm/vgtod.h>
#include <asm/time.h>
#include <asm/delay.h>
#include <asm/hypervisor.h>
#include <asm/nmi.h>
#include <asm/x86_init.h>

unsigned int __read_mostly cpu_khz;	/* TSC clocks / usec, not used here */
EXPORT_SYMBOL(cpu_khz);

unsigned int __read_mostly tsc_khz;
EXPORT_SYMBOL(tsc_khz);

/*
 * TSC can be unstable due to cpufreq or due to unsynced TSCs
 */
static int __read_mostly tsc_unstable;

/* native_sched_clock() is called before tsc_init(), so
   we must start with the TSC soft disabled to prevent
   erroneous rdtsc usage on !cpu_has_tsc processors */
static int __read_mostly tsc_disabled = -1;

static struct static_key __use_tsc = STATIC_KEY_INIT;

int tsc_clocksource_reliable;

/*
 * Use a ring-buffer like data structure, where a writer advances the head by
 * writing a new data entry and a reader advances the tail when it observes a
 * new entry.
 *
 * Writers are made to wait on readers until there's space to write a new
 * entry.
 *
 * This means that we can always use an {offset, mul} pair to compute a ns
 * value that is 'roughly' in the right direction, even if we're writing a new
 * {offset, mul} pair during the clock read.
 *
 * The down-side is that we can no longer guarantee strict monotonicity anymore
 * (assuming the TSC was that to begin with), because while we compute the
 * intersection point of the two clock slopes and make sure the time is
 * continuous at the point of switching; we can no longer guarantee a reader is
 * strictly before or after the switch point.
 *
 * It does mean a reader no longer needs to disable IRQs in order to avoid
 * CPU-Freq updates messing with his times, and similarly an NMI reader will
 * no longer run the risk of hitting half-written state.
 */

struct cyc2ns {
	struct cyc2ns_data data[2];	/*  0 + 2*24 = 48 */
	struct cyc2ns_data *head;	/* 48 + 8    = 56 */
	struct cyc2ns_data *tail;	/* 56 + 8    = 64 */
}; /* exactly fits one cacheline */

static DEFINE_PER_CPU_ALIGNED(struct cyc2ns, cyc2ns);

struct cyc2ns_data *cyc2ns_read_begin(void)
{
	struct cyc2ns_data *head;

	preempt_disable();

	head = this_cpu_read(cyc2ns.head);
	/*
	 * Ensure we observe the entry when we observe the pointer to it.
	 * matches the wmb from cyc2ns_write_end().
	 */
	smp_read_barrier_depends();
	head->__count++;
	barrier();

	return head;
}

void cyc2ns_read_end(struct cyc2ns_data *head)
{
	barrier();
	/*
	 * If we're the outer most nested read; update the tail pointer
	 * when we're done. This notifies possible pending writers
	 * that we've observed the head pointer and that the other
	 * entry is now free.
	 */
	if (!--head->__count) {
		/*
		 * x86-TSO does not reorder writes with older reads;
		 * therefore once this write becomes visible to another
		 * cpu, we must be finished reading the cyc2ns_data.
		 *
		 * matches with cyc2ns_write_begin().
		 */
		this_cpu_write(cyc2ns.tail, head);
	}
	preempt_enable();
}

/*
 * Begin writing a new @data entry for @cpu.
 *
 * Assumes some sort of write side lock; currently 'provided' by the assumption
 * that cpufreq will call its notifiers sequentially.
 */
static struct cyc2ns_data *cyc2ns_write_begin(int cpu)
{
	struct cyc2ns *c2n = &per_cpu(cyc2ns, cpu);
	struct cyc2ns_data *data = c2n->data;

	if (data == c2n->head)
		data++;

	/* XXX send an IPI to @cpu in order to guarantee a read? */

	/*
	 * When we observe the tail write from cyc2ns_read_end(),
	 * the cpu must be done with that entry and its safe
	 * to start writing to it.
	 */
	while (c2n->tail == data)
		cpu_relax();

	return data;
}

static void cyc2ns_write_end(int cpu, struct cyc2ns_data *data)
{
	struct cyc2ns *c2n = &per_cpu(cyc2ns, cpu);

	/*
	 * Ensure the @data writes are visible before we publish the
	 * entry. Matches the data-depencency in cyc2ns_read_begin().
	 */
	smp_wmb();

	ACCESS_ONCE(c2n->head) = data;
}

/*
 * Accelerators for sched_clock()
 * convert from cycles(64bits) => nanoseconds (64bits)
 *  basic equation:
 *              ns = cycles / (freq / ns_per_sec)
 *              ns = cycles * (ns_per_sec / freq)
 *              ns = cycles * (10^9 / (cpu_khz * 10^3))
 *              ns = cycles * (10^6 / cpu_khz)
 *
 *      Then we use scaling math (suggested by george@mvista.com) to get:
 *              ns = cycles * (10^6 * SC / cpu_khz) / SC
 *              ns = cycles * cyc2ns_scale / SC
 *
 *      And since SC is a constant power of two, we can convert the div
 *  into a shift.
 *
 *  We can use khz divisor instead of mhz to keep a better precision, since
 *  cyc2ns_scale is limited to 10^6 * 2^10, which fits in 32 bits.
 *  (mathieu.desnoyers@polymtl.ca)
 *
 *                      -johnstul@us.ibm.com "math is hard, lets go shopping!"
 */

#define CYC2NS_SCALE_FACTOR 10 /* 2^10, carefully chosen */

static void cyc2ns_data_init(struct cyc2ns_data *data)
{
	data->cyc2ns_mul = 0;
	data->cyc2ns_shift = CYC2NS_SCALE_FACTOR;
	data->cyc2ns_offset = 0;
	data->__count = 0;
}

static void cyc2ns_init(int cpu)
{
	struct cyc2ns *c2n = &per_cpu(cyc2ns, cpu);

	cyc2ns_data_init(&c2n->data[0]);
	cyc2ns_data_init(&c2n->data[1]);

	c2n->head = c2n->data;
	c2n->tail = c2n->data;
}

static inline unsigned long long cycles_2_ns(unsigned long long cyc)
{
	struct cyc2ns_data *data, *tail;
	unsigned long long ns;

	/*
	 * See cyc2ns_read_*() for details; replicated in order to avoid
	 * an extra few instructions that came with the abstraction.
	 * Notable, it allows us to only do the __count and tail update
	 * dance when its actually needed.
	 */

	preempt_disable_notrace();
	data = this_cpu_read(cyc2ns.head);
	tail = this_cpu_read(cyc2ns.tail);

	if (likely(data == tail)) {
		ns = data->cyc2ns_offset;
		ns += mul_u64_u32_shr(cyc, data->cyc2ns_mul, CYC2NS_SCALE_FACTOR);
	} else {
		data->__count++;

		barrier();

		ns = data->cyc2ns_offset;
		ns += mul_u64_u32_shr(cyc, data->cyc2ns_mul, CYC2NS_SCALE_FACTOR);

		barrier();

		if (!--data->__count)
			this_cpu_write(cyc2ns.tail, data);
	}
	preempt_enable_notrace();

	return ns;
}

/* XXX surely we already have this someplace in the kernel?! */
#define DIV_ROUND(n, d) (((n) + ((d) / 2)) / (d))

static void set_cyc2ns_scale(unsigned long cpu_khz, int cpu)
{
	unsigned long long tsc_now, ns_now;
	struct cyc2ns_data *data;
	unsigned long flags;

	local_irq_save(flags);
	sched_clock_idle_sleep_event();

	if (!cpu_khz)
		goto done;

	data = cyc2ns_write_begin(cpu);

	rdtscll(tsc_now);
	ns_now = cycles_2_ns(tsc_now);

	/*
	 * Compute a new multiplier as per the above comment and ensure our
	 * time function is continuous; see the comment near struct
	 * cyc2ns_data.
	 */
	data->cyc2ns_mul = DIV_ROUND(NSEC_PER_MSEC << CYC2NS_SCALE_FACTOR, cpu_khz);
	data->cyc2ns_shift = CYC2NS_SCALE_FACTOR;
	data->cyc2ns_offset = ns_now -
		mul_u64_u32_shr(tsc_now, data->cyc2ns_mul, CYC2NS_SCALE_FACTOR);

	cyc2ns_write_end(cpu, data);

done:
	sched_clock_idle_wakeup_event(0);
	local_irq_restore(flags);
}
/*
 * Scheduler clock - returns current time in nanosec units.
 */
u64 native_sched_clock(void)
{
	u64 tsc_now;

	/*
	 * Fall back to jiffies if there's no TSC available:
	 * ( But note that we still use it if the TSC is marked
	 *   unstable. We do this because unlike Time Of Day,
	 *   the scheduler clock tolerates small errors and it's
	 *   very important for it to be as fast as the platform
	 *   can achieve it. )
	 */
	if (!static_key_false(&__use_tsc)) {
		/* No locking but a rare wrong value is not a big deal: */
		return (jiffies_64 - INITIAL_JIFFIES) * (1000000000 / HZ);
	}

	/* read the Time Stamp Counter: */
	rdtscll(tsc_now);

	/* return the value in ns */
	return cycles_2_ns(tsc_now);
}

/* We need to define a real function for sched_clock, to override the
   weak default version */
#ifdef CONFIG_PARAVIRT
unsigned long long sched_clock(void)
{
	return paravirt_sched_clock();
}
#else
unsigned long long
sched_clock(void) __attribute__((alias("native_sched_clock")));
#endif

unsigned long long native_read_tsc(void)
{
	return __native_read_tsc();
}
EXPORT_SYMBOL(native_read_tsc);

int check_tsc_unstable(void)
{
	return tsc_unstable;
}
EXPORT_SYMBOL_GPL(check_tsc_unstable);

int check_tsc_disabled(void)
{
	return tsc_disabled;
}
EXPORT_SYMBOL_GPL(check_tsc_disabled);

#ifdef CONFIG_X86_TSC
int __init notsc_setup(char *str)
{
	pr_warn("Kernel compiled with CONFIG_X86_TSC, cannot disable TSC completely\n");
	tsc_disabled = 1;
	return 1;
}
#else
/*
 * disable flag for tsc. Takes effect by clearing the TSC cpu flag
 * in cpu/common.c
 */
int __init notsc_setup(char *str)
{
	setup_clear_cpu_cap(X86_FEATURE_TSC);
	return 1;
}
#endif

__setup("notsc", notsc_setup);

static int no_sched_irq_time;

static int __init tsc_setup(char *str)
{
	if (!strcmp(str, "reliable"))
		tsc_clocksource_reliable = 1;
	if (!strncmp(str, "noirqtime", 9))
		no_sched_irq_time = 1;
	return 1;
}

__setup("tsc=", tsc_setup);

#define MAX_RETRIES     5
#define SMI_TRESHOLD    50000

/*
 * Read TSC and the reference counters. Take care of SMI disturbance
 */
static u64 tsc_read_refs(u64 *p, int hpet)
{
	u64 t1, t2;
	int i;

	for (i = 0; i < MAX_RETRIES; i++) {
		t1 = get_cycles();
		if (hpet)
			*p = hpet_readl(HPET_COUNTER) & 0xFFFFFFFF;
		else
			*p = acpi_pm_read_early();
		t2 = get_cycles();
		if ((t2 - t1) < SMI_TRESHOLD)
			return t2;
	}
	return ULLONG_MAX;
}

/*
 * Calculate the TSC frequency from HPET reference
 */
static unsigned long calc_hpet_ref(u64 deltatsc, u64 hpet1, u64 hpet2)
{
	u64 tmp;

	if (hpet2 < hpet1)
		hpet2 += 0x100000000ULL;
	hpet2 -= hpet1;
	tmp = ((u64)hpet2 * hpet_readl(HPET_PERIOD));
	do_div(tmp, 1000000);
	do_div(deltatsc, tmp);

	return (unsigned long) deltatsc;
}

/*
 * Calculate the TSC frequency from PMTimer reference
 */
static unsigned long calc_pmtimer_ref(u64 deltatsc, u64 pm1, u64 pm2)
{
	u64 tmp;

	if (!pm1 && !pm2)
		return ULONG_MAX;

	if (pm2 < pm1)
		pm2 += (u64)ACPI_PM_OVRRUN;
	pm2 -= pm1;
	tmp = pm2 * 1000000000LL;
	do_div(tmp, PMTMR_TICKS_PER_SEC);
	do_div(deltatsc, tmp);

	return (unsigned long) deltatsc;
}

#define CAL_MS		10
#define CAL_LATCH	(PIT_TICK_RATE / (1000 / CAL_MS))
#define CAL_PIT_LOOPS	1000

#define CAL2_MS		50
#define CAL2_LATCH	(PIT_TICK_RATE / (1000 / CAL2_MS))
#define CAL2_PIT_LOOPS	5000


/*
 * Try to calibrate the TSC against the Programmable
 * Interrupt Timer and return the frequency of the TSC
 * in kHz.
 *
 * Return ULONG_MAX on failure to calibrate.
 */
static unsigned long pit_calibrate_tsc(u32 latch, unsigned long ms, int loopmin)
{
	u64 tsc, t1, t2, delta;
	unsigned long tscmin, tscmax;
	int pitcnt;

	/* Set the Gate high, disable speaker */
	outb((inb(0x61) & ~0x02) | 0x01, 0x61);

	/*
	 * Setup CTC channel 2* for mode 0, (interrupt on terminal
	 * count mode), binary count. Set the latch register to 50ms
	 * (LSB then MSB) to begin countdown.
	 */
	outb(0xb0, 0x43);
	outb(latch & 0xff, 0x42);
	outb(latch >> 8, 0x42);

	tsc = t1 = t2 = get_cycles();

	pitcnt = 0;
	tscmax = 0;
	tscmin = ULONG_MAX;
	while ((inb(0x61) & 0x20) == 0) {
		t2 = get_cycles();
		delta = t2 - tsc;
		tsc = t2;
		if ((unsigned long) delta < tscmin)
			tscmin = (unsigned int) delta;
		if ((unsigned long) delta > tscmax)
			tscmax = (unsigned int) delta;
		pitcnt++;
	}

	/*
	 * Sanity checks:
	 *
	 * If we were not able to read the PIT more than loopmin
	 * times, then we have been hit by a massive SMI
	 *
	 * If the maximum is 10 times larger than the minimum,
	 * then we got hit by an SMI as well.
	 */
	if (pitcnt < loopmin || tscmax > 10 * tscmin)
		return ULONG_MAX;

	/* Calculate the PIT value */
	delta = t2 - t1;
	do_div(delta, ms);
	return delta;
}

/*
 * This reads the current MSB of the PIT counter, and
 * checks if we are running on sufficiently fast and
 * non-virtualized hardware.
 *
 * Our expectations are:
 *
 *  - the PIT is running at roughly 1.19MHz
 *
 *  - each IO is going to take about 1us on real hardware,
 *    but we allow it to be much faster (by a factor of 10) or
 *    _slightly_ slower (ie we allow up to a 2us read+counter
 *    update - anything else implies a unacceptably slow CPU
 *    or PIT for the fast calibration to work.
 *
 *  - with 256 PIT ticks to read the value, we have 214us to
 *    see the same MSB (and overhead like doing a single TSC
 *    read per MSB value etc).
 *
 *  - We're doing 2 reads per loop (LSB, MSB), and we expect
 *    them each to take about a microsecond on real hardware.
 *    So we expect a count value of around 100. But we'll be
 *    generous, and accept anything over 50.
 *
 *  - if the PIT is stuck, and we see *many* more reads, we
 *    return early (and the next caller of pit_expect_msb()
 *    then consider it a failure when they don't see the
 *    next expected value).
 *
 * These expectations mean that we know that we have seen the
 * transition from one expected value to another with a fairly
 * high accuracy, and we didn't miss any events. We can thus
 * use the TSC value at the transitions to calculate a pretty
 * good value for the TSC frequencty.
 */
static inline int pit_verify_msb(unsigned char val)
{
	/* Ignore LSB */
	inb(0x42);
	return inb(0x42) == val;
}

static inline int pit_expect_msb(unsigned char val, u64 *tscp, unsigned long *deltap)
{
	int count;
	u64 tsc = 0, prev_tsc = 0;

	for (count = 0; count < 50000; count++) {
		if (!pit_verify_msb(val))
			break;
		prev_tsc = tsc;
		tsc = get_cycles();
	}
	*deltap = get_cycles() - prev_tsc;
	*tscp = tsc;

	/*
	 * We require _some_ success, but the quality control
	 * will be based on the error terms on the TSC values.
	 */
	return count > 5;
}

/*
 * How many MSB values do we want to see? We aim for
 * a maximum error rate of 500ppm (in practice the
 * real error is much smaller), but refuse to spend
 * more than 50ms on it.
 */
#define MAX_QUICK_PIT_MS 50
#define MAX_QUICK_PIT_ITERATIONS (MAX_QUICK_PIT_MS * PIT_TICK_RATE / 1000 / 256)

static unsigned long quick_pit_calibrate(void)
{
	int i;
	u64 tsc, delta;
	unsigned long d1, d2;

	/* Set the Gate high, disable speaker */
	outb((inb(0x61) & ~0x02) | 0x01, 0x61);

	/*
	 * Counter 2, mode 0 (one-shot), binary count
	 *
	 * NOTE! Mode 2 decrements by two (and then the
	 * output is flipped each time, giving the same
	 * final output frequency as a decrement-by-one),
	 * so mode 0 is much better when looking at the
	 * individual counts.
	 */
	outb(0xb0, 0x43);

	/* Start at 0xffff */
	outb(0xff, 0x42);
	outb(0xff, 0x42);

	/*
	 * The PIT starts counting at the next edge, so we
	 * need to delay for a microsecond. The easiest way
	 * to do that is to just read back the 16-bit counter
	 * once from the PIT.
	 */
	pit_verify_msb(0);

	if (pit_expect_msb(0xff, &tsc, &d1)) {
		for (i = 1; i <= MAX_QUICK_PIT_ITERATIONS; i++) {
			if (!pit_expect_msb(0xff-i, &delta, &d2))
				break;

			/*
			 * Iterate until the error is less than 500 ppm
			 */
			delta -= tsc;
			if (d1+d2 >= delta >> 11)
				continue;

			/*
			 * Check the PIT one more time to verify that
			 * all TSC reads were stable wrt the PIT.
			 *
			 * This also guarantees serialization of the
			 * last cycle read ('d2') in pit_expect_msb.
			 */
			if (!pit_verify_msb(0xfe - i))
				break;
			goto success;
		}
	}
	pr_err("Fast TSC calibration failed\n");
	return 0;

success:
	/*
	 * Ok, if we get here, then we've seen the
	 * MSB of the PIT decrement 'i' times, and the
	 * error has shrunk to less than 500 ppm.
	 *
	 * As a result, we can depend on there not being
	 * any odd delays anywhere, and the TSC reads are
	 * reliable (within the error).
	 *
	 * kHz = ticks / time-in-seconds / 1000;
	 * kHz = (t2 - t1) / (I * 256 / PIT_TICK_RATE) / 1000
	 * kHz = ((t2 - t1) * PIT_TICK_RATE) / (I * 256 * 1000)
	 */
	delta *= PIT_TICK_RATE;
	do_div(delta, i*256*1000);
	pr_info("Fast TSC calibration using PIT\n");
	return delta;
}

/**
 * native_calibrate_tsc - calibrate the tsc on boot
 */
unsigned long native_calibrate_tsc(void)
{
	u64 tsc1, tsc2, delta, ref1, ref2;
	unsigned long tsc_pit_min = ULONG_MAX, tsc_ref_min = ULONG_MAX;
	unsigned long flags, latch, ms, fast_calibrate;
	int hpet = is_hpet_enabled(), i, loopmin;

	/* Calibrate TSC using MSR for Intel Atom SoCs */
	local_irq_save(flags);
	i = try_msr_calibrate_tsc(&fast_calibrate);
	local_irq_restore(flags);
	if (i >= 0) {
		if (i == 0)
			pr_warn("Fast TSC calibration using MSR failed\n");
		return fast_calibrate;
	}

	local_irq_save(flags);
	fast_calibrate = quick_pit_calibrate();
	local_irq_restore(flags);
	if (fast_calibrate)
		return fast_calibrate;

	/*
	 * Run 5 calibration loops to get the lowest frequency value
	 * (the best estimate). We use two different calibration modes
	 * here:
	 *
	 * 1) PIT loop. We set the PIT Channel 2 to oneshot mode and
	 * load a timeout of 50ms. We read the time right after we
	 * started the timer and wait until the PIT count down reaches
	 * zero. In each wait loop iteration we read the TSC and check
	 * the delta to the previous read. We keep track of the min
	 * and max values of that delta. The delta is mostly defined
	 * by the IO time of the PIT access, so we can detect when a
	 * SMI/SMM disturbance happened between the two reads. If the
	 * maximum time is significantly larger than the minimum time,
	 * then we discard the result and have another try.
	 *
	 * 2) Reference counter. If available we use the HPET or the
	 * PMTIMER as a reference to check the sanity of that value.
	 * We use separate TSC readouts and check inside of the
	 * reference read for a SMI/SMM disturbance. We dicard
	 * disturbed values here as well. We do that around the PIT
	 * calibration delay loop as we have to wait for a certain
	 * amount of time anyway.
	 */

	/* Preset PIT loop values */
	latch = CAL_LATCH;
	ms = CAL_MS;
	loopmin = CAL_PIT_LOOPS;

	for (i = 0; i < 3; i++) {
		unsigned long tsc_pit_khz;

		/*
		 * Read the start value and the reference count of
		 * hpet/pmtimer when available. Then do the PIT
		 * calibration, which will take at least 50ms, and
		 * read the end value.
		 */
		local_irq_save(flags);
		tsc1 = tsc_read_refs(&ref1, hpet);
		tsc_pit_khz = pit_calibrate_tsc(latch, ms, loopmin);
		tsc2 = tsc_read_refs(&ref2, hpet);
		local_irq_restore(flags);

		/* Pick the lowest PIT TSC calibration so far */
		tsc_pit_min = min(tsc_pit_min, tsc_pit_khz);

		/* hpet or pmtimer available ? */
		if (ref1 == ref2)
			continue;

		/* Check, whether the sampling was disturbed by an SMI */
		if (tsc1 == ULLONG_MAX || tsc2 == ULLONG_MAX)
			continue;

		tsc2 = (tsc2 - tsc1) * 1000000LL;
		if (hpet)
			tsc2 = calc_hpet_ref(tsc2, ref1, ref2);
		else
			tsc2 = calc_pmtimer_ref(tsc2, ref1, ref2);

		tsc_ref_min = min(tsc_ref_min, (unsigned long) tsc2);

		/* Check the reference deviation */
		delta = ((u64) tsc_pit_min) * 100;
		do_div(delta, tsc_ref_min);

		/*
		 * If both calibration results are inside a 10% window
		 * then we can be sure, that the calibration
		 * succeeded. We break out of the loop right away. We
		 * use the reference value, as it is more precise.
		 */
		if (delta >= 90 && delta <= 110) {
			pr_info("PIT calibration matches %s. %d loops\n",
				hpet ? "HPET" : "PMTIMER", i + 1);
			return tsc_ref_min;
		}

		/*
		 * Check whether PIT failed more than once. This
		 * happens in virtualized environments. We need to
		 * give the virtual PC a slightly longer timeframe for
		 * the HPET/PMTIMER to make the result precise.
		 */
		if (i == 1 && tsc_pit_min == ULONG_MAX) {
			latch = CAL2_LATCH;
			ms = CAL2_MS;
			loopmin = CAL2_PIT_LOOPS;
		}
	}

	/*
	 * Now check the results.
	 */
	if (tsc_pit_min == ULONG_MAX) {
		/* PIT gave no useful value */
		pr_warn("Unable to calibrate against PIT\n");

		/* We don't have an alternative source, disable TSC */
		if (!hpet && !ref1 && !ref2) {
			pr_notice("No reference (HPET/PMTIMER) available\n");
			return 0;
		}

		/* The alternative source failed as well, disable TSC */
		if (tsc_ref_min == ULONG_MAX) {
			pr_warn("HPET/PMTIMER calibration failed\n");
			return 0;
		}

		/* Use the alternative source */
		pr_info("using %s reference calibration\n",
			hpet ? "HPET" : "PMTIMER");

		return tsc_ref_min;
	}

	/* We don't have an alternative source, use the PIT calibration value */
	if (!hpet && !ref1 && !ref2) {
		pr_info("Using PIT calibration value\n");
		return tsc_pit_min;
	}

	/* The alternative source failed, use the PIT calibration value */
	if (tsc_ref_min == ULONG_MAX) {
		pr_warn("HPET/PMTIMER calibration failed. Using PIT calibration.\n");
		return tsc_pit_min;
	}

	/*
	 * The calibration values differ too much. In doubt, we use
	 * the PIT value as we know that there are PMTIMERs around
	 * running at double speed. At least we let the user know:
	 */
	pr_warn("PIT calibration deviates from %s: %lu %lu\n",
		hpet ? "HPET" : "PMTIMER", tsc_pit_min, tsc_ref_min);
	pr_info("Using PIT calibration value\n");
	return tsc_pit_min;
}

int recalibrate_cpu_khz(void)
{
#ifndef CONFIG_SMP
	unsigned long cpu_khz_old = cpu_khz;

	if (cpu_has_tsc) {
		tsc_khz = x86_platform.calibrate_tsc();
		cpu_khz = tsc_khz;
		cpu_data(0).loops_per_jiffy =
			cpufreq_scale(cpu_data(0).loops_per_jiffy,
					cpu_khz_old, cpu_khz);
		return 0;
	} else
		return -ENODEV;
#else
	return -ENODEV;
#endif
}

EXPORT_SYMBOL(recalibrate_cpu_khz);


static unsigned long long cyc2ns_suspend;

void tsc_save_sched_clock_state(void)
{
	if (!sched_clock_stable())
		return;

	cyc2ns_suspend = sched_clock();
}

/*
 * Even on processors with invariant TSC, TSC gets reset in some the
 * ACPI system sleep states. And in some systems BIOS seem to reinit TSC to
 * arbitrary value (still sync'd across cpu's) during resume from such sleep
 * states. To cope up with this, recompute the cyc2ns_offset for each cpu so
 * that sched_clock() continues from the point where it was left off during
 * suspend.
 */
void tsc_restore_sched_clock_state(void)
{
	unsigned long long offset;
	unsigned long flags;
	int cpu;

	if (!sched_clock_stable())
		return;

	local_irq_save(flags);

	/*
	 * We're comming out of suspend, there's no concurrency yet; don't
	 * bother being nice about the RCU stuff, just write to both
	 * data fields.
	 */

	this_cpu_write(cyc2ns.data[0].cyc2ns_offset, 0);
	this_cpu_write(cyc2ns.data[1].cyc2ns_offset, 0);

	offset = cyc2ns_suspend - sched_clock();

	for_each_possible_cpu(cpu) {
		per_cpu(cyc2ns.data[0].cyc2ns_offset, cpu) = offset;
		per_cpu(cyc2ns.data[1].cyc2ns_offset, cpu) = offset;
	}

	local_irq_restore(flags);
}

#ifdef CONFIG_CPU_FREQ

/* Frequency scaling support. Adjust the TSC based timer when the cpu frequency
 * changes.
 *
 * RED-PEN: On SMP we assume all CPUs run with the same frequency.  It's
 * not that important because current Opteron setups do not support
 * scaling on SMP anyroads.
 *
 * Should fix up last_tsc too. Currently gettimeofday in the
 * first tick after the change will be slightly wrong.
 */

static unsigned int  ref_freq;
static unsigned long loops_per_jiffy_ref;
static unsigned long tsc_khz_ref;

static int time_cpufreq_notifier(struct notifier_block *nb, unsigned long val,
				void *data)
{
	struct cpufreq_freqs *freq = data;
	unsigned long *lpj;

	if (cpu_has(&cpu_data(freq->cpu), X86_FEATURE_CONSTANT_TSC))
		return 0;

	lpj = &boot_cpu_data.loops_per_jiffy;
#ifdef CONFIG_SMP
	if (!(freq->flags & CPUFREQ_CONST_LOOPS))
		lpj = &cpu_data(freq->cpu).loops_per_jiffy;
#endif

	if (!ref_freq) {
		ref_freq = freq->old;
		loops_per_jiffy_ref = *lpj;
		tsc_khz_ref = tsc_khz;
	}
	if ((val == CPUFREQ_PRECHANGE  && freq->old < freq->new) ||
			(val == CPUFREQ_POSTCHANGE && freq->old > freq->new) ||
			(val == CPUFREQ_RESUMECHANGE)) {
		*lpj = cpufreq_scale(loops_per_jiffy_ref, ref_freq, freq->new);

		tsc_khz = cpufreq_scale(tsc_khz_ref, ref_freq, freq->new);
		if (!(freq->flags & CPUFREQ_CONST_LOOPS))
			mark_tsc_unstable("cpufreq changes");
	}

	set_cyc2ns_scale(tsc_khz, freq->cpu);

	return 0;
}

static struct notifier_block time_cpufreq_notifier_block = {
	.notifier_call  = time_cpufreq_notifier
};

static int __init cpufreq_tsc(void)
{
	if (!cpu_has_tsc)
		return 0;
	if (boot_cpu_has(X86_FEATURE_CONSTANT_TSC))
		return 0;
	cpufreq_register_notifier(&time_cpufreq_notifier_block,
				CPUFREQ_TRANSITION_NOTIFIER);
	return 0;
}

core_initcall(cpufreq_tsc);

#endif /* CONFIG_CPU_FREQ */

/* clocksource code */

static struct clocksource clocksource_tsc;

/*
 * We compare the TSC to the cycle_last value in the clocksource
 * structure to avoid a nasty time-warp. This can be observed in a
 * very small window right after one CPU updated cycle_last under
 * xtime/vsyscall_gtod lock and the other CPU reads a TSC value which
 * is smaller than the cycle_last reference value due to a TSC which
 * is slighty behind. This delta is nowhere else observable, but in
 * that case it results in a forward time jump in the range of hours
 * due to the unsigned delta calculation of the time keeping core
 * code, which is necessary to support wrapping clocksources like pm
 * timer.
 */
static cycle_t read_tsc(struct clocksource *cs)
{
	cycle_t ret = (cycle_t)get_cycles();

	return ret >= clocksource_tsc.cycle_last ?
		ret : clocksource_tsc.cycle_last;
}

static void resume_tsc(struct clocksource *cs)
{
	if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC_S3))
		clocksource_tsc.cycle_last = 0;
}

static struct clocksource clocksource_tsc = {
	.name                   = "tsc",
	.rating                 = 300,
	.read                   = read_tsc,
	.resume			= resume_tsc,
	.mask                   = CLOCKSOURCE_MASK(64),
	.flags                  = CLOCK_SOURCE_IS_CONTINUOUS |
				  CLOCK_SOURCE_MUST_VERIFY,
#ifdef CONFIG_X86_64
	.archdata               = { .vclock_mode = VCLOCK_TSC },
#endif
};

void mark_tsc_unstable(char *reason)
{
	if (!tsc_unstable) {
		tsc_unstable = 1;
		clear_sched_clock_stable();
		disable_sched_clock_irqtime();
		pr_info("Marking TSC unstable due to %s\n", reason);
		/* Change only the rating, when not registered */
		if (clocksource_tsc.mult)
			clocksource_mark_unstable(&clocksource_tsc);
		else {
			clocksource_tsc.flags |= CLOCK_SOURCE_UNSTABLE;
			clocksource_tsc.rating = 0;
		}
	}
}

EXPORT_SYMBOL_GPL(mark_tsc_unstable);

static void __init check_system_tsc_reliable(void)
{
#ifdef CONFIG_MGEODE_LX
	/* RTSC counts during suspend */
#define RTSC_SUSP 0x100
	unsigned long res_low, res_high;

	rdmsr_safe(MSR_GEODE_BUSCONT_CONF0, &res_low, &res_high);
	/* Geode_LX - the OLPC CPU has a very reliable TSC */
	if (res_low & RTSC_SUSP)
		tsc_clocksource_reliable = 1;
#endif
	if (boot_cpu_has(X86_FEATURE_TSC_RELIABLE))
		tsc_clocksource_reliable = 1;
}

/*
 * Make an educated guess if the TSC is trustworthy and synchronized
 * over all CPUs.
 */
int unsynchronized_tsc(void)
{
	if (!cpu_has_tsc || tsc_unstable)
		return 1;

#ifdef CONFIG_SMP
	if (apic_is_clustered_box())
		return 1;
#endif

	if (boot_cpu_has(X86_FEATURE_CONSTANT_TSC))
		return 0;

	if (tsc_clocksource_reliable)
		return 0;
	/*
	 * Intel systems are normally all synchronized.
	 * Exceptions must mark TSC as unstable:
	 */
	if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL) {
		/* assume multi socket systems are not synchronized: */
		if (num_possible_cpus() > 1)
			return 1;
	}

	return 0;
}


static void tsc_refine_calibration_work(struct work_struct *work);
static DECLARE_DELAYED_WORK(tsc_irqwork, tsc_refine_calibration_work);
/**
 * tsc_refine_calibration_work - Further refine tsc freq calibration
 * @work - ignored.
 *
 * This functions uses delayed work over a period of a
 * second to further refine the TSC freq value. Since this is
 * timer based, instead of loop based, we don't block the boot
 * process while this longer calibration is done.
 *
 * If there are any calibration anomalies (too many SMIs, etc),
 * or the refined calibration is off by 1% of the fast early
 * calibration, we throw out the new calibration and use the
 * early calibration.
 */
static void tsc_refine_calibration_work(struct work_struct *work)
{
	static u64 tsc_start = -1, ref_start;
	static int hpet;
	u64 tsc_stop, ref_stop, delta;
	unsigned long freq;

	/* Don't bother refining TSC on unstable systems */
	if (check_tsc_unstable())
		goto out;

	/*
	 * Since the work is started early in boot, we may be
	 * delayed the first time we expire. So set the workqueue
	 * again once we know timers are working.
	 */
	if (tsc_start == -1) {
		/*
		 * Only set hpet once, to avoid mixing hardware
		 * if the hpet becomes enabled later.
		 */
		hpet = is_hpet_enabled();
		schedule_delayed_work(&tsc_irqwork, HZ);
		tsc_start = tsc_read_refs(&ref_start, hpet);
		return;
	}

	tsc_stop = tsc_read_refs(&ref_stop, hpet);

	/* hpet or pmtimer available ? */
	if (ref_start == ref_stop)
		goto out;

	/* Check, whether the sampling was disturbed by an SMI */
	if (tsc_start == ULLONG_MAX || tsc_stop == ULLONG_MAX)
		goto out;

	delta = tsc_stop - tsc_start;
	delta *= 1000000LL;
	if (hpet)
		freq = calc_hpet_ref(delta, ref_start, ref_stop);
	else
		freq = calc_pmtimer_ref(delta, ref_start, ref_stop);

	/* Make sure we're within 1% */
	if (abs(tsc_khz - freq) > tsc_khz/100)
		goto out;

	tsc_khz = freq;
	pr_info("Refined TSC clocksource calibration: %lu.%03lu MHz\n",
		(unsigned long)tsc_khz / 1000,
		(unsigned long)tsc_khz % 1000);

out:
	clocksource_register_khz(&clocksource_tsc, tsc_khz);
}


static int __init init_tsc_clocksource(void)
{
	if (!cpu_has_tsc || tsc_disabled > 0 || !tsc_khz)
		return 0;

	if (tsc_clocksource_reliable)
		clocksource_tsc.flags &= ~CLOCK_SOURCE_MUST_VERIFY;
	/* lower the rating if we already know its unstable: */
	if (check_tsc_unstable()) {
		clocksource_tsc.rating = 0;
		clocksource_tsc.flags &= ~CLOCK_SOURCE_IS_CONTINUOUS;
	}

	if (boot_cpu_has(X86_FEATURE_NONSTOP_TSC_S3))
		clocksource_tsc.flags |= CLOCK_SOURCE_SUSPEND_NONSTOP;

	/*
	 * Trust the results of the earlier calibration on systems
	 * exporting a reliable TSC.
	 */
	if (boot_cpu_has(X86_FEATURE_TSC_RELIABLE)) {
		clocksource_register_khz(&clocksource_tsc, tsc_khz);
		return 0;
	}

	schedule_delayed_work(&tsc_irqwork, 0);
	return 0;
}
/*
 * We use device_initcall here, to ensure we run after the hpet
 * is fully initialized, which may occur at fs_initcall time.
 */
device_initcall(init_tsc_clocksource);

void __init tsc_init(void)
{
	u64 lpj;
	int cpu;

	x86_init.timers.tsc_pre_init();

	if (!cpu_has_tsc)
		return;

	tsc_khz = x86_platform.calibrate_tsc();
	cpu_khz = tsc_khz;

	if (!tsc_khz) {
		mark_tsc_unstable("could not calculate TSC khz");
		return;
	}

	pr_info("Detected %lu.%03lu MHz processor\n",
		(unsigned long)cpu_khz / 1000,
		(unsigned long)cpu_khz % 1000);

	/*
	 * Secondary CPUs do not run through tsc_init(), so set up
	 * all the scale factors for all CPUs, assuming the same
	 * speed as the bootup CPU. (cpufreq notifiers will fix this
	 * up if their speed diverges)
	 */
	for_each_possible_cpu(cpu) {
		cyc2ns_init(cpu);
		set_cyc2ns_scale(cpu_khz, cpu);
	}

	if (tsc_disabled > 0)
		return;

	/* now allow native_sched_clock() to use rdtsc */

	tsc_disabled = 0;
	static_key_slow_inc(&__use_tsc);

	if (!no_sched_irq_time)
		enable_sched_clock_irqtime();

	lpj = ((u64)tsc_khz * 1000);
	do_div(lpj, HZ);
	lpj_fine = lpj;

	use_tsc_delay();

	if (unsynchronized_tsc())
		mark_tsc_unstable("TSCs unsynchronized");

	check_system_tsc_reliable();
}

#ifdef CONFIG_SMP
/*
 * If we have a constant TSC and are using the TSC for the delay loop,
 * we can skip clock calibration if another cpu in the same socket has already
 * been calibrated. This assumes that CONSTANT_TSC applies to all
 * cpus in the socket - this should be a safe assumption.
 */
unsigned long calibrate_delay_is_known(void)
{
	int i, cpu = smp_processor_id();

	if (!tsc_disabled && !cpu_has(&cpu_data(cpu), X86_FEATURE_CONSTANT_TSC))
		return 0;

	for_each_online_cpu(i)
		if (cpu_data(i).phys_proc_id == cpu_data(cpu).phys_proc_id)
			return cpu_data(i).loops_per_jiffy;
	return 0;
}
#endif