aboutsummaryrefslogtreecommitdiffstats
path: root/arch/ia64/kernel/mca.c
blob: 6dc726ad71372954b36c02b93949f240ac794ee4 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
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
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
/*
 * File:	mca.c
 * Purpose:	Generic MCA handling layer
 *
 * Updated for latest kernel
 * Copyright (C) 2003 Hewlett-Packard Co
 *	David Mosberger-Tang <davidm@hpl.hp.com>
 *
 * Copyright (C) 2002 Dell Inc.
 * Copyright (C) Matt Domsch (Matt_Domsch@dell.com)
 *
 * Copyright (C) 2002 Intel
 * Copyright (C) Jenna Hall (jenna.s.hall@intel.com)
 *
 * Copyright (C) 2001 Intel
 * Copyright (C) Fred Lewis (frederick.v.lewis@intel.com)
 *
 * Copyright (C) 2000 Intel
 * Copyright (C) Chuck Fleckenstein (cfleck@co.intel.com)
 *
 * Copyright (C) 1999, 2004 Silicon Graphics, Inc.
 * Copyright (C) Vijay Chander(vijay@engr.sgi.com)
 *
 * 03/04/15 D. Mosberger Added INIT backtrace support.
 * 02/03/25 M. Domsch	GUID cleanups
 *
 * 02/01/04 J. Hall	Aligned MCA stack to 16 bytes, added platform vs. CPU
 *			error flag, set SAL default return values, changed
 *			error record structure to linked list, added init call
 *			to sal_get_state_info_size().
 *
 * 01/01/03 F. Lewis    Added setup of CMCI and CPEI IRQs, logging of corrected
 *                      platform errors, completed code for logging of
 *                      corrected & uncorrected machine check errors, and
 *                      updated for conformance with Nov. 2000 revision of the
 *                      SAL 3.0 spec.
 * 00/03/29 C. Fleckenstein  Fixed PAL/SAL update issues, began MCA bug fixes, logging issues,
 *                           added min save state dump, added INIT handler.
 *
 * 2003-12-08 Keith Owens <kaos@sgi.com>
 *            smp_call_function() must not be called from interrupt context (can
 *            deadlock on tasklist_lock).  Use keventd to call smp_call_function().
 *
 * 2004-02-01 Keith Owens <kaos@sgi.com>
 *            Avoid deadlock when using printk() for MCA and INIT records.
 *            Delete all record printing code, moved to salinfo_decode in user space.
 *            Mark variables and functions static where possible.
 *            Delete dead variables and functions.
 *            Reorder to remove the need for forward declarations and to consolidate
 *            related code.
 *
 * 2005-08-12 Keith Owens <kaos@sgi.com>
 *	      Convert MCA/INIT handlers to use per event stacks and SAL/OS state.
 */
#include <linux/config.h>
#include <linux/types.h>
#include <linux/init.h>
#include <linux/sched.h>
#include <linux/interrupt.h>
#include <linux/irq.h>
#include <linux/kallsyms.h>
#include <linux/smp_lock.h>
#include <linux/bootmem.h>
#include <linux/acpi.h>
#include <linux/timer.h>
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/smp.h>
#include <linux/workqueue.h>

#include <asm/delay.h>
#include <asm/machvec.h>
#include <asm/meminit.h>
#include <asm/page.h>
#include <asm/ptrace.h>
#include <asm/system.h>
#include <asm/sal.h>
#include <asm/mca.h>

#include <asm/irq.h>
#include <asm/hw_irq.h>

#include "entry.h"

#if defined(IA64_MCA_DEBUG_INFO)
# define IA64_MCA_DEBUG(fmt...)	printk(fmt)
#else
# define IA64_MCA_DEBUG(fmt...)
#endif

/* Used by mca_asm.S */
u32				ia64_mca_serialize;
DEFINE_PER_CPU(u64, ia64_mca_data); /* == __per_cpu_mca[smp_processor_id()] */
DEFINE_PER_CPU(u64, ia64_mca_per_cpu_pte); /* PTE to map per-CPU area */
DEFINE_PER_CPU(u64, ia64_mca_pal_pte);	    /* PTE to map PAL code */
DEFINE_PER_CPU(u64, ia64_mca_pal_base);    /* vaddr PAL code granule */

unsigned long __per_cpu_mca[NR_CPUS];

/* In mca_asm.S */
extern void			ia64_os_init_dispatch_monarch (void);
extern void			ia64_os_init_dispatch_slave (void);

static int monarch_cpu = -1;

static ia64_mc_info_t		ia64_mc_info;

#define MAX_CPE_POLL_INTERVAL (15*60*HZ) /* 15 minutes */
#define MIN_CPE_POLL_INTERVAL (2*60*HZ)  /* 2 minutes */
#define CMC_POLL_INTERVAL     (1*60*HZ)  /* 1 minute */
#define CPE_HISTORY_LENGTH    5
#define CMC_HISTORY_LENGTH    5

static struct timer_list cpe_poll_timer;
static struct timer_list cmc_poll_timer;
/*
 * This variable tells whether we are currently in polling mode.
 * Start with this in the wrong state so we won't play w/ timers
 * before the system is ready.
 */
static int cmc_polling_enabled = 1;

/*
 * Clearing this variable prevents CPE polling from getting activated
 * in mca_late_init.  Use it if your system doesn't provide a CPEI,
 * but encounters problems retrieving CPE logs.  This should only be
 * necessary for debugging.
 */
static int cpe_poll_enabled = 1;

extern void salinfo_log_wakeup(int type, u8 *buffer, u64 size, int irqsafe);

static int mca_init;

/*
 * IA64_MCA log support
 */
#define IA64_MAX_LOGS		2	/* Double-buffering for nested MCAs */
#define IA64_MAX_LOG_TYPES      4   /* MCA, INIT, CMC, CPE */

typedef struct ia64_state_log_s
{
	spinlock_t	isl_lock;
	int		isl_index;
	unsigned long	isl_count;
	ia64_err_rec_t  *isl_log[IA64_MAX_LOGS]; /* need space to store header + error log */
} ia64_state_log_t;

static ia64_state_log_t ia64_state_log[IA64_MAX_LOG_TYPES];

#define IA64_LOG_ALLOCATE(it, size) \
	{ia64_state_log[it].isl_log[IA64_LOG_CURR_INDEX(it)] = \
		(ia64_err_rec_t *)alloc_bootmem(size); \
	ia64_state_log[it].isl_log[IA64_LOG_NEXT_INDEX(it)] = \
		(ia64_err_rec_t *)alloc_bootmem(size);}
#define IA64_LOG_LOCK_INIT(it) spin_lock_init(&ia64_state_log[it].isl_lock)
#define IA64_LOG_LOCK(it)      spin_lock_irqsave(&ia64_state_log[it].isl_lock, s)
#define IA64_LOG_UNLOCK(it)    spin_unlock_irqrestore(&ia64_state_log[it].isl_lock,s)
#define IA64_LOG_NEXT_INDEX(it)    ia64_state_log[it].isl_index
#define IA64_LOG_CURR_INDEX(it)    1 - ia64_state_log[it].isl_index
#define IA64_LOG_INDEX_INC(it) \
    {ia64_state_log[it].isl_index = 1 - ia64_state_log[it].isl_index; \
    ia64_state_log[it].isl_count++;}
#define IA64_LOG_INDEX_DEC(it) \
    ia64_state_log[it].isl_index = 1 - ia64_state_log[it].isl_index
#define IA64_LOG_NEXT_BUFFER(it)   (void *)((ia64_state_log[it].isl_log[IA64_LOG_NEXT_INDEX(it)]))
#define IA64_LOG_CURR_BUFFER(it)   (void *)((ia64_state_log[it].isl_log[IA64_LOG_CURR_INDEX(it)]))
#define IA64_LOG_COUNT(it)         ia64_state_log[it].isl_count

/*
 * ia64_log_init
 *	Reset the OS ia64 log buffer
 * Inputs   :   info_type   (SAL_INFO_TYPE_{MCA,INIT,CMC,CPE})
 * Outputs	:	None
 */
static void
ia64_log_init(int sal_info_type)
{
	u64	max_size = 0;

	IA64_LOG_NEXT_INDEX(sal_info_type) = 0;
	IA64_LOG_LOCK_INIT(sal_info_type);

	// SAL will tell us the maximum size of any error record of this type
	max_size = ia64_sal_get_state_info_size(sal_info_type);
	if (!max_size)
		/* alloc_bootmem() doesn't like zero-sized allocations! */
		return;

	// set up OS data structures to hold error info
	IA64_LOG_ALLOCATE(sal_info_type, max_size);
	memset(IA64_LOG_CURR_BUFFER(sal_info_type), 0, max_size);
	memset(IA64_LOG_NEXT_BUFFER(sal_info_type), 0, max_size);
}

/*
 * ia64_log_get
 *
 *	Get the current MCA log from SAL and copy it into the OS log buffer.
 *
 *  Inputs  :   info_type   (SAL_INFO_TYPE_{MCA,INIT,CMC,CPE})
 *              irq_safe    whether you can use printk at this point
 *  Outputs :   size        (total record length)
 *              *buffer     (ptr to error record)
 *
 */
static u64
ia64_log_get(int sal_info_type, u8 **buffer, int irq_safe)
{
	sal_log_record_header_t     *log_buffer;
	u64                         total_len = 0;
	int                         s;

	IA64_LOG_LOCK(sal_info_type);

	/* Get the process state information */
	log_buffer = IA64_LOG_NEXT_BUFFER(sal_info_type);

	total_len = ia64_sal_get_state_info(sal_info_type, (u64 *)log_buffer);

	if (total_len) {
		IA64_LOG_INDEX_INC(sal_info_type);
		IA64_LOG_UNLOCK(sal_info_type);
		if (irq_safe) {
			IA64_MCA_DEBUG("%s: SAL error record type %d retrieved. "
				       "Record length = %ld\n", __FUNCTION__, sal_info_type, total_len);
		}
		*buffer = (u8 *) log_buffer;
		return total_len;
	} else {
		IA64_LOG_UNLOCK(sal_info_type);
		return 0;
	}
}

/*
 *  ia64_mca_log_sal_error_record
 *
 *  This function retrieves a specified error record type from SAL
 *  and wakes up any processes waiting for error records.
 *
 *  Inputs  :   sal_info_type   (Type of error record MCA/CMC/CPE)
 *              FIXME: remove MCA and irq_safe.
 */
static void
ia64_mca_log_sal_error_record(int sal_info_type)
{
	u8 *buffer;
	sal_log_record_header_t *rh;
	u64 size;
	int irq_safe = sal_info_type != SAL_INFO_TYPE_MCA;
#ifdef IA64_MCA_DEBUG_INFO
	static const char * const rec_name[] = { "MCA", "INIT", "CMC", "CPE" };
#endif

	size = ia64_log_get(sal_info_type, &buffer, irq_safe);
	if (!size)
		return;

	salinfo_log_wakeup(sal_info_type, buffer, size, irq_safe);

	if (irq_safe)
		IA64_MCA_DEBUG("CPU %d: SAL log contains %s error record\n",
			smp_processor_id(),
			sal_info_type < ARRAY_SIZE(rec_name) ? rec_name[sal_info_type] : "UNKNOWN");

	/* Clear logs from corrected errors in case there's no user-level logger */
	rh = (sal_log_record_header_t *)buffer;
	if (rh->severity == sal_log_severity_corrected)
		ia64_sal_clear_state_info(sal_info_type);
}

/*
 * platform dependent error handling
 */
#ifndef PLATFORM_MCA_HANDLERS

#ifdef CONFIG_ACPI

int cpe_vector = -1;

static irqreturn_t
ia64_mca_cpe_int_handler (int cpe_irq, void *arg, struct pt_regs *ptregs)
{
	static unsigned long	cpe_history[CPE_HISTORY_LENGTH];
	static int		index;
	static DEFINE_SPINLOCK(cpe_history_lock);

	IA64_MCA_DEBUG("%s: received interrupt vector = %#x on CPU %d\n",
		       __FUNCTION__, cpe_irq, smp_processor_id());

	/* SAL spec states this should run w/ interrupts enabled */
	local_irq_enable();

	/* Get the CPE error record and log it */
	ia64_mca_log_sal_error_record(SAL_INFO_TYPE_CPE);

	spin_lock(&cpe_history_lock);
	if (!cpe_poll_enabled && cpe_vector >= 0) {

		int i, count = 1; /* we know 1 happened now */
		unsigned long now = jiffies;

		for (i = 0; i < CPE_HISTORY_LENGTH; i++) {
			if (now - cpe_history[i] <= HZ)
				count++;
		}

		IA64_MCA_DEBUG(KERN_INFO "CPE threshold %d/%d\n", count, CPE_HISTORY_LENGTH);
		if (count >= CPE_HISTORY_LENGTH) {

			cpe_poll_enabled = 1;
			spin_unlock(&cpe_history_lock);
			disable_irq_nosync(local_vector_to_irq(IA64_CPE_VECTOR));

			/*
			 * Corrected errors will still be corrected, but
			 * make sure there's a log somewhere that indicates
			 * something is generating more than we can handle.
			 */
			printk(KERN_WARNING "WARNING: Switching to polling CPE handler; error records may be lost\n");

			mod_timer(&cpe_poll_timer, jiffies + MIN_CPE_POLL_INTERVAL);

			/* lock already released, get out now */
			return IRQ_HANDLED;
		} else {
			cpe_history[index++] = now;
			if (index == CPE_HISTORY_LENGTH)
				index = 0;
		}
	}
	spin_unlock(&cpe_history_lock);
	return IRQ_HANDLED;
}

#endif /* CONFIG_ACPI */

#ifdef CONFIG_ACPI
/*
 * ia64_mca_register_cpev
 *
 *  Register the corrected platform error vector with SAL.
 *
 *  Inputs
 *      cpev        Corrected Platform Error Vector number
 *
 *  Outputs
 *      None
 */
static void
ia64_mca_register_cpev (int cpev)
{
	/* Register the CPE interrupt vector with SAL */
	struct ia64_sal_retval isrv;

	isrv = ia64_sal_mc_set_params(SAL_MC_PARAM_CPE_INT, SAL_MC_PARAM_MECHANISM_INT, cpev, 0, 0);
	if (isrv.status) {
		printk(KERN_ERR "Failed to register Corrected Platform "
		       "Error interrupt vector with SAL (status %ld)\n", isrv.status);
		return;
	}

	IA64_MCA_DEBUG("%s: corrected platform error "
		       "vector %#x registered\n", __FUNCTION__, cpev);
}
#endif /* CONFIG_ACPI */

#endif /* PLATFORM_MCA_HANDLERS */

/*
 * ia64_mca_cmc_vector_setup
 *
 *  Setup the corrected machine check vector register in the processor.
 *  (The interrupt is masked on boot. ia64_mca_late_init unmask this.)
 *  This function is invoked on a per-processor basis.
 *
 * Inputs
 *      None
 *
 * Outputs
 *	None
 */
void
ia64_mca_cmc_vector_setup (void)
{
	cmcv_reg_t	cmcv;

	cmcv.cmcv_regval	= 0;
	cmcv.cmcv_mask		= 1;        /* Mask/disable interrupt at first */
	cmcv.cmcv_vector	= IA64_CMC_VECTOR;
	ia64_setreg(_IA64_REG_CR_CMCV, cmcv.cmcv_regval);

	IA64_MCA_DEBUG("%s: CPU %d corrected "
		       "machine check vector %#x registered.\n",
		       __FUNCTION__, smp_processor_id(), IA64_CMC_VECTOR);

	IA64_MCA_DEBUG("%s: CPU %d CMCV = %#016lx\n",
		       __FUNCTION__, smp_processor_id(), ia64_getreg(_IA64_REG_CR_CMCV));
}

/*
 * ia64_mca_cmc_vector_disable
 *
 *  Mask the corrected machine check vector register in the processor.
 *  This function is invoked on a per-processor basis.
 *
 * Inputs
 *      dummy(unused)
 *
 * Outputs
 *	None
 */
static void
ia64_mca_cmc_vector_disable (void *dummy)
{
	cmcv_reg_t	cmcv;

	cmcv.cmcv_regval = ia64_getreg(_IA64_REG_CR_CMCV);

	cmcv.cmcv_mask = 1; /* Mask/disable interrupt */
	ia64_setreg(_IA64_REG_CR_CMCV, cmcv.cmcv_regval);

	IA64_MCA_DEBUG("%s: CPU %d corrected "
		       "machine check vector %#x disabled.\n",
		       __FUNCTION__, smp_processor_id(), cmcv.cmcv_vector);
}

/*
 * ia64_mca_cmc_vector_enable
 *
 *  Unmask the corrected machine check vector register in the processor.
 *  This function is invoked on a per-processor basis.
 *
 * Inputs
 *      dummy(unused)
 *
 * Outputs
 *	None
 */
static void
ia64_mca_cmc_vector_enable (void *dummy)
{
	cmcv_reg_t	cmcv;

	cmcv.cmcv_regval = ia64_getreg(_IA64_REG_CR_CMCV);

	cmcv.cmcv_mask = 0; /* Unmask/enable interrupt */
	ia64_setreg(_IA64_REG_CR_CMCV, cmcv.cmcv_regval);

	IA64_MCA_DEBUG("%s: CPU %d corrected "
		       "machine check vector %#x enabled.\n",
		       __FUNCTION__, smp_processor_id(), cmcv.cmcv_vector);
}

/*
 * ia64_mca_cmc_vector_disable_keventd
 *
 * Called via keventd (smp_call_function() is not safe in interrupt context) to
 * disable the cmc interrupt vector.
 */
static void
ia64_mca_cmc_vector_disable_keventd(void *unused)
{
	on_each_cpu(ia64_mca_cmc_vector_disable, NULL, 1, 0);
}

/*
 * ia64_mca_cmc_vector_enable_keventd
 *
 * Called via keventd (smp_call_function() is not safe in interrupt context) to
 * enable the cmc interrupt vector.
 */
static void
ia64_mca_cmc_vector_enable_keventd(void *unused)
{
	on_each_cpu(ia64_mca_cmc_vector_enable, NULL, 1, 0);
}

/*
 * ia64_mca_wakeup
 *
 *	Send an inter-cpu interrupt to wake-up a particular cpu
 *	and mark that cpu to be out of rendez.
 *
 *  Inputs  :   cpuid
 *  Outputs :   None
 */
static void
ia64_mca_wakeup(int cpu)
{
	platform_send_ipi(cpu, IA64_MCA_WAKEUP_VECTOR, IA64_IPI_DM_INT, 0);
	ia64_mc_info.imi_rendez_checkin[cpu] = IA64_MCA_RENDEZ_CHECKIN_NOTDONE;

}

/*
 * ia64_mca_wakeup_all
 *
 *	Wakeup all the cpus which have rendez'ed previously.
 *
 *  Inputs  :   None
 *  Outputs :   None
 */
static void
ia64_mca_wakeup_all(void)
{
	int cpu;

	/* Clear the Rendez checkin flag for all cpus */
	for(cpu = 0; cpu < NR_CPUS; cpu++) {
		if (!cpu_online(cpu))
			continue;
		if (ia64_mc_info.imi_rendez_checkin[cpu] == IA64_MCA_RENDEZ_CHECKIN_DONE)
			ia64_mca_wakeup(cpu);
	}

}

/*
 * ia64_mca_rendez_interrupt_handler
 *
 *	This is handler used to put slave processors into spinloop
 *	while the monarch processor does the mca handling and later
 *	wake each slave up once the monarch is done.
 *
 *  Inputs  :   None
 *  Outputs :   None
 */
static irqreturn_t
ia64_mca_rendez_int_handler(int rendez_irq, void *arg, struct pt_regs *ptregs)
{
	unsigned long flags;
	int cpu = smp_processor_id();

	/* Mask all interrupts */
	local_irq_save(flags);

	ia64_mc_info.imi_rendez_checkin[cpu] = IA64_MCA_RENDEZ_CHECKIN_DONE;
	/* Register with the SAL monarch that the slave has
	 * reached SAL
	 */
	ia64_sal_mc_rendez();

	/* Wait for the monarch cpu to exit. */
	while (monarch_cpu != -1)
	       cpu_relax();	/* spin until monarch leaves */

	/* Enable all interrupts */
	local_irq_restore(flags);
	return IRQ_HANDLED;
}

/*
 * ia64_mca_wakeup_int_handler
 *
 *	The interrupt handler for processing the inter-cpu interrupt to the
 *	slave cpu which was spinning in the rendez loop.
 *	Since this spinning is done by turning off the interrupts and
 *	polling on the wakeup-interrupt bit in the IRR, there is
 *	nothing useful to be done in the handler.
 *
 *  Inputs  :   wakeup_irq  (Wakeup-interrupt bit)
 *	arg		(Interrupt handler specific argument)
 *	ptregs		(Exception frame at the time of the interrupt)
 *  Outputs :   None
 *
 */
static irqreturn_t
ia64_mca_wakeup_int_handler(int wakeup_irq, void *arg, struct pt_regs *ptregs)
{
	return IRQ_HANDLED;
}

/* Function pointer for extra MCA recovery */
int (*ia64_mca_ucmc_extension)
	(void*,struct ia64_sal_os_state*)
	= NULL;

int
ia64_reg_MCA_extension(int (*fn)(void *, struct ia64_sal_os_state *))
{
	if (ia64_mca_ucmc_extension)
		return 1;

	ia64_mca_ucmc_extension = fn;
	return 0;
}

void
ia64_unreg_MCA_extension(void)
{
	if (ia64_mca_ucmc_extension)
		ia64_mca_ucmc_extension = NULL;
}

EXPORT_SYMBOL(ia64_reg_MCA_extension);
EXPORT_SYMBOL(ia64_unreg_MCA_extension);


static inline void
copy_reg(const u64 *fr, u64 fnat, u64 *tr, u64 *tnat)
{
	u64 fslot, tslot, nat;
	*tr = *fr;
	fslot = ((unsigned long)fr >> 3) & 63;
	tslot = ((unsigned long)tr >> 3) & 63;
	*tnat &= ~(1UL << tslot);
	nat = (fnat >> fslot) & 1;
	*tnat |= (nat << tslot);
}

/* On entry to this routine, we are running on the per cpu stack, see
 * mca_asm.h.  The original stack has not been touched by this event.  Some of
 * the original stack's registers will be in the RBS on this stack.  This stack
 * also contains a partial pt_regs and switch_stack, the rest of the data is in
 * PAL minstate.
 *
 * The first thing to do is modify the original stack to look like a blocked
 * task so we can run backtrace on the original task.  Also mark the per cpu
 * stack as current to ensure that we use the correct task state, it also means
 * that we can do backtrace on the MCA/INIT handler code itself.
 */

static task_t *
ia64_mca_modify_original_stack(struct pt_regs *regs,
		const struct switch_stack *sw,
		struct ia64_sal_os_state *sos,
		const char *type)
{
	char *p, comm[sizeof(current->comm)];
	ia64_va va;
	extern char ia64_leave_kernel[];	/* Need asm address, not function descriptor */
	const pal_min_state_area_t *ms = sos->pal_min_state;
	task_t *previous_current;
	struct pt_regs *old_regs;
	struct switch_stack *old_sw;
	unsigned size = sizeof(struct pt_regs) +
			sizeof(struct switch_stack) + 16;
	u64 *old_bspstore, *old_bsp;
	u64 *new_bspstore, *new_bsp;
	u64 old_unat, old_rnat, new_rnat, nat;
	u64 slots, loadrs = regs->loadrs;
	u64 r12 = ms->pmsa_gr[12-1], r13 = ms->pmsa_gr[13-1];
	u64 ar_bspstore = regs->ar_bspstore;
	u64 ar_bsp = regs->ar_bspstore + (loadrs >> 16);
	const u64 *bank;
	const char *msg;
	int cpu = smp_processor_id();

	previous_current = curr_task(cpu);
	set_curr_task(cpu, current);
	if ((p = strchr(current->comm, ' ')))
		*p = '\0';

	/* Best effort attempt to cope with MCA/INIT delivered while in
	 * physical mode.
	 */
	regs->cr_ipsr = ms->pmsa_ipsr;
	if (ia64_psr(regs)->dt == 0) {
		va.l = r12;
		if (va.f.reg == 0) {
			va.f.reg = 7;
			r12 = va.l;
		}
		va.l = r13;
		if (va.f.reg == 0) {
			va.f.reg = 7;
			r13 = va.l;
		}
	}
	if (ia64_psr(regs)->rt == 0) {
		va.l = ar_bspstore;
		if (va.f.reg == 0) {
			va.f.reg = 7;
			ar_bspstore = va.l;
		}
		va.l = ar_bsp;
		if (va.f.reg == 0) {
			va.f.reg = 7;
			ar_bsp = va.l;
		}
	}

	/* mca_asm.S ia64_old_stack() cannot assume that the dirty registers
	 * have been copied to the old stack, the old stack may fail the
	 * validation tests below.  So ia64_old_stack() must restore the dirty
	 * registers from the new stack.  The old and new bspstore probably
	 * have different alignments, so loadrs calculated on the old bsp
	 * cannot be used to restore from the new bsp.  Calculate a suitable
	 * loadrs for the new stack and save it in the new pt_regs, where
	 * ia64_old_stack() can get it.
	 */
	old_bspstore = (u64 *)ar_bspstore;
	old_bsp = (u64 *)ar_bsp;
	slots = ia64_rse_num_regs(old_bspstore, old_bsp);
	new_bspstore = (u64 *)((u64)current + IA64_RBS_OFFSET);
	new_bsp = ia64_rse_skip_regs(new_bspstore, slots);
	regs->loadrs = (new_bsp - new_bspstore) * 8 << 16;

	/* Verify the previous stack state before we change it */
	if (user_mode(regs)) {
		msg = "occurred in user space";
		goto no_mod;
	}
	if (r13 != sos->prev_IA64_KR_CURRENT) {
		msg = "inconsistent previous current and r13";
		goto no_mod;
	}
	if ((r12 - r13) >= KERNEL_STACK_SIZE) {
		msg = "inconsistent r12 and r13";
		goto no_mod;
	}
	if ((ar_bspstore - r13) >= KERNEL_STACK_SIZE) {
		msg = "inconsistent ar.bspstore and r13";
		goto no_mod;
	}
	va.p = old_bspstore;
	if (va.f.reg < 5) {
		msg = "old_bspstore is in the wrong region";
		goto no_mod;
	}
	if ((ar_bsp - r13) >= KERNEL_STACK_SIZE) {
		msg = "inconsistent ar.bsp and r13";
		goto no_mod;
	}
	size += (ia64_rse_skip_regs(old_bspstore, slots) - old_bspstore) * 8;
	if (ar_bspstore + size > r12) {
		msg = "no room for blocked state";
		goto no_mod;
	}

	/* Change the comm field on the MCA/INT task to include the pid that
	 * was interrupted, it makes for easier debugging.  If that pid was 0
	 * (swapper or nested MCA/INIT) then use the start of the previous comm
	 * field suffixed with its cpu.
	 */
	if (previous_current->pid)
		snprintf(comm, sizeof(comm), "%s %d",
			current->comm, previous_current->pid);
	else {
		int l;
		if ((p = strchr(previous_current->comm, ' ')))
			l = p - previous_current->comm;
		else
			l = strlen(previous_current->comm);
		snprintf(comm, sizeof(comm), "%s %*s %d",
			current->comm, l, previous_current->comm,
			previous_current->thread_info->cpu);
	}
	memcpy(current->comm, comm, sizeof(current->comm));

	/* Make the original task look blocked.  First stack a struct pt_regs,
	 * describing the state at the time of interrupt.  mca_asm.S built a
	 * partial pt_regs, copy it and fill in the blanks using minstate.
	 */
	p = (char *)r12 - sizeof(*regs);
	old_regs = (struct pt_regs *)p;
	memcpy(old_regs, regs, sizeof(*regs));
	/* If ipsr.ic then use pmsa_{iip,ipsr,ifs}, else use
	 * pmsa_{xip,xpsr,xfs}
	 */
	if (ia64_psr(regs)->ic) {
		old_regs->cr_iip = ms->pmsa_iip;
		old_regs->cr_ipsr = ms->pmsa_ipsr;
		old_regs->cr_ifs = ms->pmsa_ifs;
	} else {
		old_regs->cr_iip = ms->pmsa_xip;
		old_regs->cr_ipsr = ms->pmsa_xpsr;
		old_regs->cr_ifs = ms->pmsa_xfs;
	}
	old_regs->pr = ms->pmsa_pr;
	old_regs->b0 = ms->pmsa_br0;
	old_regs->loadrs = loadrs;
	old_regs->ar_rsc = ms->pmsa_rsc;
	old_unat = old_regs->ar_unat;
	copy_reg(&ms->pmsa_gr[1-1], ms->pmsa_nat_bits, &old_regs->r1, &old_unat);
	copy_reg(&ms->pmsa_gr[2-1], ms->pmsa_nat_bits, &old_regs->r2, &old_unat);
	copy_reg(&ms->pmsa_gr[3-1], ms->pmsa_nat_bits, &old_regs->r3, &old_unat);
	copy_reg(&ms->pmsa_gr[8-1], ms->pmsa_nat_bits, &old_regs->r8, &old_unat);
	copy_reg(&ms->pmsa_gr[9-1], ms->pmsa_nat_bits, &old_regs->r9, &old_unat);
	copy_reg(&ms->pmsa_gr[10-1], ms->pmsa_nat_bits, &old_regs->r10, &old_unat);
	copy_reg(&ms->pmsa_gr[11-1], ms->pmsa_nat_bits, &old_regs->r11, &old_unat);
	copy_reg(&ms->pmsa_gr[12-1], ms->pmsa_nat_bits, &old_regs->r12, &old_unat);
	copy_reg(&ms->pmsa_gr[13-1], ms->pmsa_nat_bits, &old_regs->r13, &old_unat);
	copy_reg(&ms->pmsa_gr[14-1], ms->pmsa_nat_bits, &old_regs->r14, &old_unat);
	copy_reg(&ms->pmsa_gr[15-1], ms->pmsa_nat_bits, &old_regs->r15, &old_unat);
	if (ia64_psr(old_regs)->bn)
		bank = ms->pmsa_bank1_gr;
	else
		bank = ms->pmsa_bank0_gr;
	copy_reg(&bank[16-16], ms->pmsa_nat_bits, &old_regs->r16, &old_unat);
	copy_reg(&bank[17-16], ms->pmsa_nat_bits, &old_regs->r17, &old_unat);
	copy_reg(&bank[18-16], ms->pmsa_nat_bits, &old_regs->r18, &old_unat);
	copy_reg(&bank[19-16], ms->pmsa_nat_bits, &old_regs->r19, &old_unat);
	copy_reg(&bank[20-16], ms->pmsa_nat_bits, &old_regs->r20, &old_unat);
	copy_reg(&bank[21-16], ms->pmsa_nat_bits, &old_regs->r21, &old_unat);
	copy_reg(&bank[22-16], ms->pmsa_nat_bits, &old_regs->r22, &old_unat);
	copy_reg(&bank[23-16], ms->pmsa_nat_bits, &old_regs->r23, &old_unat);
	copy_reg(&bank[24-16], ms->pmsa_nat_bits, &old_regs->r24, &old_unat);
	copy_reg(&bank[25-16], ms->pmsa_nat_bits, &old_regs->r25, &old_unat);
	copy_reg(&bank[26-16], ms->pmsa_nat_bits, &old_regs->r26, &old_unat);
	copy_reg(&bank[27-16], ms->pmsa_nat_bits, &old_regs->r27, &old_unat);
	copy_reg(&bank[28-16], ms->pmsa_nat_bits, &old_regs->r28, &old_unat);
	copy_reg(&bank[29-16], ms->pmsa_nat_bits, &old_regs->r29, &old_unat);
	copy_reg(&bank[30-16], ms->pmsa_nat_bits, &old_regs->r30, &old_unat);
	copy_reg(&bank[31-16], ms->pmsa_nat_bits, &old_regs->r31, &old_unat);

	/* Next stack a struct switch_stack.  mca_asm.S built a partial
	 * switch_stack, copy it and fill in the blanks using pt_regs and
	 * minstate.
	 *
	 * In the synthesized switch_stack, b0 points to ia64_leave_kernel,
	 * ar.pfs is set to 0.
	 *
	 * unwind.c::unw_unwind() does special processing for interrupt frames.
	 * It checks if the PRED_NON_SYSCALL predicate is set, if the predicate
	 * is clear then unw_unwind() does _not_ adjust bsp over pt_regs.  Not
	 * that this is documented, of course.  Set PRED_NON_SYSCALL in the
	 * switch_stack on the original stack so it will unwind correctly when
	 * unwind.c reads pt_regs.
	 *
	 * thread.ksp is updated to point to the synthesized switch_stack.
	 */
	p -= sizeof(struct switch_stack);
	old_sw = (struct switch_stack *)p;
	memcpy(old_sw, sw, sizeof(*sw));
	old_sw->caller_unat = old_unat;
	old_sw->ar_fpsr = old_regs->ar_fpsr;
	copy_reg(&ms->pmsa_gr[4-1], ms->pmsa_nat_bits, &old_sw->r4, &old_unat);
	copy_reg(&ms->pmsa_gr[5-1], ms->pmsa_nat_bits, &old_sw->r5, &old_unat);
	copy_reg(&ms->pmsa_gr[6-1], ms->pmsa_nat_bits, &old_sw->r6, &old_unat);
	copy_reg(&ms->pmsa_gr[7-1], ms->pmsa_nat_bits, &old_sw->r7, &old_unat);
	old_sw->b0 = (u64)ia64_leave_kernel;
	old_sw->b1 = ms->pmsa_br1;
	old_sw->ar_pfs = 0;
	old_sw->ar_unat = old_unat;
	old_sw->pr = old_regs->pr | (1UL << PRED_NON_SYSCALL);
	previous_current->thread.ksp = (u64)p - 16;

	/* Finally copy the original stack's registers back to its RBS.
	 * Registers from ar.bspstore through ar.bsp at the time of the event
	 * are in the current RBS, copy them back to the original stack.  The
	 * copy must be done register by register because the original bspstore
	 * and the current one have different alignments, so the saved RNAT
	 * data occurs at different places.
	 *
	 * mca_asm does cover, so the old_bsp already includes all registers at
	 * the time of MCA/INIT.  It also does flushrs, so all registers before
	 * this function have been written to backing store on the MCA/INIT
	 * stack.
	 */
	new_rnat = ia64_get_rnat(ia64_rse_rnat_addr(new_bspstore));
	old_rnat = regs->ar_rnat;
	while (slots--) {
		if (ia64_rse_is_rnat_slot(new_bspstore)) {
			new_rnat = ia64_get_rnat(new_bspstore++);
		}
		if (ia64_rse_is_rnat_slot(old_bspstore)) {
			*old_bspstore++ = old_rnat;
			old_rnat = 0;
		}
		nat = (new_rnat >> ia64_rse_slot_num(new_bspstore)) & 1UL;
		old_rnat &= ~(1UL << ia64_rse_slot_num(old_bspstore));
		old_rnat |= (nat << ia64_rse_slot_num(old_bspstore));
		*old_bspstore++ = *new_bspstore++;
	}
	old_sw->ar_bspstore = (unsigned long)old_bspstore;
	old_sw->ar_rnat = old_rnat;

	sos->prev_task = previous_current;
	return previous_current;

no_mod:
	printk(KERN_INFO "cpu %d, %s %s, original stack not modified\n",
			smp_processor_id(), type, msg);
	return previous_current;
}

/* The monarch/slave interaction is based on monarch_cpu and requires that all
 * slaves have entered rendezvous before the monarch leaves.  If any cpu has
 * not entered rendezvous yet then wait a bit.  The assumption is that any
 * slave that has not rendezvoused after a reasonable time is never going to do
 * so.  In this context, slave includes cpus that respond to the MCA rendezvous
 * interrupt, as well as cpus that receive the INIT slave event.
 */

static void
ia64_wait_for_slaves(int monarch)
{
	int c, wait = 0;
	for_each_online_cpu(c) {
		if (c == monarch)
			continue;
		if (ia64_mc_info.imi_rendez_checkin[c] == IA64_MCA_RENDEZ_CHECKIN_NOTDONE) {
			udelay(1000);		/* short wait first */
			wait = 1;
			break;
		}
	}
	if (!wait)
		return;
	for_each_online_cpu(c) {
		if (c == monarch)
			continue;
		if (ia64_mc_info.imi_rendez_checkin[c] == IA64_MCA_RENDEZ_CHECKIN_NOTDONE) {
			udelay(5*1000000);	/* wait 5 seconds for slaves (arbitrary) */
			break;
		}
	}
}

/*
 * ia64_mca_handler
 *
 *	This is uncorrectable machine check handler called from OS_MCA
 *	dispatch code which is in turn called from SAL_CHECK().
 *	This is the place where the core of OS MCA handling is done.
 *	Right now the logs are extracted and displayed in a well-defined
 *	format. This handler code is supposed to be run only on the
 *	monarch processor. Once the monarch is done with MCA handling
 *	further MCA logging is enabled by clearing logs.
 *	Monarch also has the duty of sending wakeup-IPIs to pull the
 *	slave processors out of rendezvous spinloop.
 */
void
ia64_mca_handler(struct pt_regs *regs, struct switch_stack *sw,
		 struct ia64_sal_os_state *sos)
{
	pal_processor_state_info_t *psp = (pal_processor_state_info_t *)
		&sos->proc_state_param;
	int recover, cpu = smp_processor_id();
	task_t *previous_current;

	oops_in_progress = 1;	/* FIXME: make printk NMI/MCA/INIT safe */
	previous_current = ia64_mca_modify_original_stack(regs, sw, sos, "MCA");
	monarch_cpu = cpu;
	ia64_wait_for_slaves(cpu);

	/* Wakeup all the processors which are spinning in the rendezvous loop.
	 * They will leave SAL, then spin in the OS with interrupts disabled
	 * until this monarch cpu leaves the MCA handler.  That gets control
	 * back to the OS so we can backtrace the other cpus, backtrace when
	 * spinning in SAL does not work.
	 */
	ia64_mca_wakeup_all();

	/* Get the MCA error record and log it */
	ia64_mca_log_sal_error_record(SAL_INFO_TYPE_MCA);

	/* TLB error is only exist in this SAL error record */
	recover = (psp->tc && !(psp->cc || psp->bc || psp->rc || psp->uc))
	/* other error recovery */
	   || (ia64_mca_ucmc_extension
		&& ia64_mca_ucmc_extension(
			IA64_LOG_CURR_BUFFER(SAL_INFO_TYPE_MCA),
			sos));

	if (recover) {
		sal_log_record_header_t *rh = IA64_LOG_CURR_BUFFER(SAL_INFO_TYPE_MCA);
		rh->severity = sal_log_severity_corrected;
		ia64_sal_clear_state_info(SAL_INFO_TYPE_MCA);
		sos->os_status = IA64_MCA_CORRECTED;
	}

	set_curr_task(cpu, previous_current);
	monarch_cpu = -1;
}

static DECLARE_WORK(cmc_disable_work, ia64_mca_cmc_vector_disable_keventd, NULL);
static DECLARE_WORK(cmc_enable_work, ia64_mca_cmc_vector_enable_keventd, NULL);

/*
 * ia64_mca_cmc_int_handler
 *
 *  This is corrected machine check interrupt handler.
 *	Right now the logs are extracted and displayed in a well-defined
 *	format.
 *
 * Inputs
 *      interrupt number
 *      client data arg ptr
 *      saved registers ptr
 *
 * Outputs
 *	None
 */
static irqreturn_t
ia64_mca_cmc_int_handler(int cmc_irq, void *arg, struct pt_regs *ptregs)
{
	static unsigned long	cmc_history[CMC_HISTORY_LENGTH];
	static int		index;
	static DEFINE_SPINLOCK(cmc_history_lock);

	IA64_MCA_DEBUG("%s: received interrupt vector = %#x on CPU %d\n",
		       __FUNCTION__, cmc_irq, smp_processor_id());

	/* SAL spec states this should run w/ interrupts enabled */
	local_irq_enable();

	/* Get the CMC error record and log it */
	ia64_mca_log_sal_error_record(SAL_INFO_TYPE_CMC);

	spin_lock(&cmc_history_lock);
	if (!cmc_polling_enabled) {
		int i, count = 1; /* we know 1 happened now */
		unsigned long now = jiffies;

		for (i = 0; i < CMC_HISTORY_LENGTH; i++) {
			if (now - cmc_history[i] <= HZ)
				count++;
		}

		IA64_MCA_DEBUG(KERN_INFO "CMC threshold %d/%d\n", count, CMC_HISTORY_LENGTH);
		if (count >= CMC_HISTORY_LENGTH) {

			cmc_polling_enabled = 1;
			spin_unlock(&cmc_history_lock);
			schedule_work(&cmc_disable_work);

			/*
			 * Corrected errors will still be corrected, but
			 * make sure there's a log somewhere that indicates
			 * something is generating more than we can handle.
			 */
			printk(KERN_WARNING "WARNING: Switching to polling CMC handler; error records may be lost\n");

			mod_timer(&cmc_poll_timer, jiffies + CMC_POLL_INTERVAL);

			/* lock already released, get out now */
			return IRQ_HANDLED;
		} else {
			cmc_history[index++] = now;
			if (index == CMC_HISTORY_LENGTH)
				index = 0;
		}
	}
	spin_unlock(&cmc_history_lock);
	return IRQ_HANDLED;
}

/*
 *  ia64_mca_cmc_int_caller
 *
 * 	Triggered by sw interrupt from CMC polling routine.  Calls
 * 	real interrupt handler and either triggers a sw interrupt
 * 	on the next cpu or does cleanup at the end.
 *
 * Inputs
 *	interrupt number
 *	client data arg ptr
 *	saved registers ptr
 * Outputs
 * 	handled
 */
static irqreturn_t
ia64_mca_cmc_int_caller(int cmc_irq, void *arg, struct pt_regs *ptregs)
{
	static int start_count = -1;
	unsigned int cpuid;

	cpuid = smp_processor_id();

	/* If first cpu, update count */
	if (start_count == -1)
		start_count = IA64_LOG_COUNT(SAL_INFO_TYPE_CMC);

	ia64_mca_cmc_int_handler(cmc_irq, arg, ptregs);

	for (++cpuid ; cpuid < NR_CPUS && !cpu_online(cpuid) ; cpuid++);

	if (cpuid < NR_CPUS) {
		platform_send_ipi(cpuid, IA64_CMCP_VECTOR, IA64_IPI_DM_INT, 0);
	} else {
		/* If no log record, switch out of polling mode */
		if (start_count == IA64_LOG_COUNT(SAL_INFO_TYPE_CMC)) {

			printk(KERN_WARNING "Returning to interrupt driven CMC handler\n");
			schedule_work(&cmc_enable_work);
			cmc_polling_enabled = 0;

		} else {

			mod_timer(&cmc_poll_timer, jiffies + CMC_POLL_INTERVAL);
		}

		start_count = -1;
	}

	return IRQ_HANDLED;
}

/*
 *  ia64_mca_cmc_poll
 *
 *	Poll for Corrected Machine Checks (CMCs)
 *
 * Inputs   :   dummy(unused)
 * Outputs  :   None
 *
 */
static void
ia64_mca_cmc_poll (unsigned long dummy)
{
	/* Trigger a CMC interrupt cascade  */
	platform_send_ipi(first_cpu(cpu_online_map), IA64_CMCP_VECTOR, IA64_IPI_DM_INT, 0);
}

/*
 *  ia64_mca_cpe_int_caller
 *
 * 	Triggered by sw interrupt from CPE polling routine.  Calls
 * 	real interrupt handler and either triggers a sw interrupt
 * 	on the next cpu or does cleanup at the end.
 *
 * Inputs
 *	interrupt number
 *	client data arg ptr
 *	saved registers ptr
 * Outputs
 * 	handled
 */
#ifdef CONFIG_ACPI

static irqreturn_t
ia64_mca_cpe_int_caller(int cpe_irq, void *arg, struct pt_regs *ptregs)
{
	static int start_count = -1;
	static int poll_time = MIN_CPE_POLL_INTERVAL;
	unsigned int cpuid;

	cpuid = smp_processor_id();

	/* If first cpu, update count */
	if (start_count == -1)
		start_count = IA64_LOG_COUNT(SAL_INFO_TYPE_CPE);

	ia64_mca_cpe_int_handler(cpe_irq, arg, ptregs);

	for (++cpuid ; cpuid < NR_CPUS && !cpu_online(cpuid) ; cpuid++);

	if (cpuid < NR_CPUS) {
		platform_send_ipi(cpuid, IA64_CPEP_VECTOR, IA64_IPI_DM_INT, 0);
	} else {
		/*
		 * If a log was recorded, increase our polling frequency,
		 * otherwise, backoff or return to interrupt mode.
		 */
		if (start_count != IA64_LOG_COUNT(SAL_INFO_TYPE_CPE)) {
			poll_time = max(MIN_CPE_POLL_INTERVAL, poll_time / 2);
		} else if (cpe_vector < 0) {
			poll_time = min(MAX_CPE_POLL_INTERVAL, poll_time * 2);
		} else {
			poll_time = MIN_CPE_POLL_INTERVAL;

			printk(KERN_WARNING "Returning to interrupt driven CPE handler\n");
			enable_irq(local_vector_to_irq(IA64_CPE_VECTOR));
			cpe_poll_enabled = 0;
		}

		if (cpe_poll_enabled)
			mod_timer(&cpe_poll_timer, jiffies + poll_time);
		start_count = -1;
	}

	return IRQ_HANDLED;
}

/*
 *  ia64_mca_cpe_poll
 *
 *	Poll for Corrected Platform Errors (CPEs), trigger interrupt
 *	on first cpu, from there it will trickle through all the cpus.
 *
 * Inputs   :   dummy(unused)
 * Outputs  :   None
 *
 */
static void
ia64_mca_cpe_poll (unsigned long dummy)
{
	/* Trigger a CPE interrupt cascade  */
	platform_send_ipi(first_cpu(cpu_online_map), IA64_CPEP_VECTOR, IA64_IPI_DM_INT, 0);
}

#endif /* CONFIG_ACPI */

/*
 * C portion of the OS INIT handler
 *
 * Called from ia64_os_init_dispatch
 *
 * Inputs: pointer to pt_regs where processor info was saved.  SAL/OS state for
 * this event.  This code is used for both monarch and slave INIT events, see
 * sos->monarch.
 *
 * All INIT events switch to the INIT stack and change the previous process to
 * blocked status.  If one of the INIT events is the monarch then we are
 * probably processing the nmi button/command.  Use the monarch cpu to dump all
 * the processes.  The slave INIT events all spin until the monarch cpu
 * returns.  We can also get INIT slave events for MCA, in which case the MCA
 * process is the monarch.
 */

void
ia64_init_handler(struct pt_regs *regs, struct switch_stack *sw,
		  struct ia64_sal_os_state *sos)
{
	static atomic_t slaves;
	static atomic_t monarchs;
	task_t *previous_current;
	int cpu = smp_processor_id(), c;
	struct task_struct *g, *t;

	oops_in_progress = 1;	/* FIXME: make printk NMI/MCA/INIT safe */
	console_loglevel = 15;	/* make sure printks make it to console */

	printk(KERN_INFO "Entered OS INIT handler. PSP=%lx cpu=%d monarch=%ld\n",
		sos->proc_state_param, cpu, sos->monarch);
	salinfo_log_wakeup(SAL_INFO_TYPE_INIT, NULL, 0, 0);

	previous_current = ia64_mca_modify_original_stack(regs, sw, sos, "INIT");
	sos->os_status = IA64_INIT_RESUME;

	/* FIXME: Workaround for broken proms that drive all INIT events as
	 * slaves.  The last slave that enters is promoted to be a monarch.
	 * Remove this code in September 2006, that gives platforms a year to
	 * fix their proms and get their customers updated.
	 */
	if (!sos->monarch && atomic_add_return(1, &slaves) == num_online_cpus()) {
		printk(KERN_WARNING "%s: Promoting cpu %d to monarch.\n",
		       __FUNCTION__, cpu);
		atomic_dec(&slaves);
		sos->monarch = 1;
	}

	/* FIXME: Workaround for broken proms that drive all INIT events as
	 * monarchs.  Second and subsequent monarchs are demoted to slaves.
	 * Remove this code in September 2006, that gives platforms a year to
	 * fix their proms and get their customers updated.
	 */
	if (sos->monarch && atomic_add_return(1, &monarchs) > 1) {
		printk(KERN_WARNING "%s: Demoting cpu %d to slave.\n",
			       __FUNCTION__, cpu);
		atomic_dec(&monarchs);
		sos->monarch = 0;
	}

	if (!sos->monarch) {
		ia64_mc_info.imi_rendez_checkin[cpu] = IA64_MCA_RENDEZ_CHECKIN_INIT;
		while (monarch_cpu == -1)
		       cpu_relax();	/* spin until monarch enters */
		while (monarch_cpu != -1)
		       cpu_relax();	/* spin until monarch leaves */
		printk("Slave on cpu %d returning to normal service.\n", cpu);
		set_curr_task(cpu, previous_current);
		ia64_mc_info.imi_rendez_checkin[cpu] = IA64_MCA_RENDEZ_CHECKIN_NOTDONE;
		atomic_dec(&slaves);
		return;
	}

	monarch_cpu = cpu;

	/*
	 * Wait for a bit.  On some machines (e.g., HP's zx2000 and zx6000, INIT can be
	 * generated via the BMC's command-line interface, but since the console is on the
	 * same serial line, the user will need some time to switch out of the BMC before
	 * the dump begins.
	 */
	printk("Delaying for 5 seconds...\n");
	udelay(5*1000000);
	ia64_wait_for_slaves(cpu);
	printk(KERN_ERR "Processes interrupted by INIT -");
	for_each_online_cpu(c) {
		struct ia64_sal_os_state *s;
		t = __va(__per_cpu_mca[c] + IA64_MCA_CPU_INIT_STACK_OFFSET);
		s = (struct ia64_sal_os_state *)((char *)t + MCA_SOS_OFFSET);
		g = s->prev_task;
		if (g) {
			if (g->pid)
				printk(" %d", g->pid);
			else
				printk(" %d (cpu %d task 0x%p)", g->pid, task_cpu(g), g);
		}
	}
	printk("\n\n");
	if (read_trylock(&tasklist_lock)) {
		do_each_thread (g, t) {
			printk("\nBacktrace of pid %d (%s)\n", t->pid, t->comm);
			show_stack(t, NULL);
		} while_each_thread (g, t);
		read_unlock(&tasklist_lock);
	}
	printk("\nINIT dump complete.  Monarch on cpu %d returning to normal service.\n", cpu);
	atomic_dec(&monarchs);
	set_curr_task(cpu, previous_current);
	monarch_cpu = -1;
	return;
}

static int __init
ia64_mca_disable_cpe_polling(char *str)
{
	cpe_poll_enabled = 0;
	return 1;
}

__setup("disable_cpe_poll", ia64_mca_disable_cpe_polling);

static struct irqaction cmci_irqaction = {
	.handler =	ia64_mca_cmc_int_handler,
	.flags =	SA_INTERRUPT,
	.name =		"cmc_hndlr"
};

static struct irqaction cmcp_irqaction = {
	.handler =	ia64_mca_cmc_int_caller,
	.flags =	SA_INTERRUPT,
	.name =		"cmc_poll"
};

static struct irqaction mca_rdzv_irqaction = {
	.handler =	ia64_mca_rendez_int_handler,
	.flags =	SA_INTERRUPT,
	.name =		"mca_rdzv"
};

static struct irqaction mca_wkup_irqaction = {
	.handler =	ia64_mca_wakeup_int_handler,
	.flags =	SA_INTERRUPT,
	.name =		"mca_wkup"
};

#ifdef CONFIG_ACPI
static struct irqaction mca_cpe_irqaction = {
	.handler =	ia64_mca_cpe_int_handler,
	.flags =	SA_INTERRUPT,
	.name =		"cpe_hndlr"
};

static struct irqaction mca_cpep_irqaction = {
	.handler =	ia64_mca_cpe_int_caller,
	.flags =	SA_INTERRUPT,
	.name =		"cpe_poll"
};
#endif /* CONFIG_ACPI */

/* Minimal format of the MCA/INIT stacks.  The pseudo processes that run on
 * these stacks can never sleep, they cannot return from the kernel to user
 * space, they do not appear in a normal ps listing.  So there is no need to
 * format most of the fields.
 */

static void
format_mca_init_stack(void *mca_data, unsigned long offset,
		const char *type, int cpu)
{
	struct task_struct *p = (struct task_struct *)((char *)mca_data + offset);
	struct thread_info *ti;
	memset(p, 0, KERNEL_STACK_SIZE);
	ti = (struct thread_info *)((char *)p + IA64_TASK_SIZE);
	ti->flags = _TIF_MCA_INIT;
	ti->preempt_count = 1;
	ti->task = p;
	ti->cpu = cpu;
	p->thread_info = ti;
	p->state = TASK_UNINTERRUPTIBLE;
	__set_bit(cpu, &p->cpus_allowed);
	INIT_LIST_HEAD(&p->tasks);
	p->parent = p->real_parent = p->group_leader = p;
	INIT_LIST_HEAD(&p->children);
	INIT_LIST_HEAD(&p->sibling);
	strncpy(p->comm, type, sizeof(p->comm)-1);
}

/* Do per-CPU MCA-related initialization.  */

void __devinit
ia64_mca_cpu_init(void *cpu_data)
{
	void *pal_vaddr;

	if (smp_processor_id() == 0) {
		void *mca_data;
		int cpu;

		mca_data = alloc_bootmem(sizeof(struct ia64_mca_cpu)
					 * NR_CPUS + KERNEL_STACK_SIZE);
		mca_data = (void *)(((unsigned long)mca_data +
					KERNEL_STACK_SIZE - 1) &
				(-KERNEL_STACK_SIZE));
		for (cpu = 0; cpu < NR_CPUS; cpu++) {
			format_mca_init_stack(mca_data,
					offsetof(struct ia64_mca_cpu, mca_stack),
					"MCA", cpu);
			format_mca_init_stack(mca_data,
					offsetof(struct ia64_mca_cpu, init_stack),
					"INIT", cpu);
			__per_cpu_mca[cpu] = __pa(mca_data);
			mca_data += sizeof(struct ia64_mca_cpu);
		}
	}

	/*
	 * The MCA info structure was allocated earlier and its
	 * physical address saved in __per_cpu_mca[cpu].  Copy that
	 * address * to ia64_mca_data so we can access it as a per-CPU
	 * variable.
	 */
	__get_cpu_var(ia64_mca_data) = __per_cpu_mca[smp_processor_id()];

	/*
	 * Stash away a copy of the PTE needed to map the per-CPU page.
	 * We may need it during MCA recovery.
	 */
	__get_cpu_var(ia64_mca_per_cpu_pte) =
		pte_val(mk_pte_phys(__pa(cpu_data), PAGE_KERNEL));

	/*
	 * Also, stash away a copy of the PAL address and the PTE
	 * needed to map it.
	 */
	pal_vaddr = efi_get_pal_addr();
	if (!pal_vaddr)
		return;
	__get_cpu_var(ia64_mca_pal_base) =
		GRANULEROUNDDOWN((unsigned long) pal_vaddr);
	__get_cpu_var(ia64_mca_pal_pte) = pte_val(mk_pte_phys(__pa(pal_vaddr),
							      PAGE_KERNEL));
}

/*
 * ia64_mca_init
 *
 *  Do all the system level mca specific initialization.
 *
 *	1. Register spinloop and wakeup request interrupt vectors
 *
 *	2. Register OS_MCA handler entry point
 *
 *	3. Register OS_INIT handler entry point
 *
 *  4. Initialize MCA/CMC/INIT related log buffers maintained by the OS.
 *
 *  Note that this initialization is done very early before some kernel
 *  services are available.
 *
 *  Inputs  :   None
 *
 *  Outputs :   None
 */
void __init
ia64_mca_init(void)
{
	ia64_fptr_t *init_hldlr_ptr_monarch = (ia64_fptr_t *)ia64_os_init_dispatch_monarch;
	ia64_fptr_t *init_hldlr_ptr_slave = (ia64_fptr_t *)ia64_os_init_dispatch_slave;
	ia64_fptr_t *mca_hldlr_ptr = (ia64_fptr_t *)ia64_os_mca_dispatch;
	int i;
	s64 rc;
	struct ia64_sal_retval isrv;
	u64 timeout = IA64_MCA_RENDEZ_TIMEOUT;	/* platform specific */

	IA64_MCA_DEBUG("%s: begin\n", __FUNCTION__);

	/* Clear the Rendez checkin flag for all cpus */
	for(i = 0 ; i < NR_CPUS; i++)
		ia64_mc_info.imi_rendez_checkin[i] = IA64_MCA_RENDEZ_CHECKIN_NOTDONE;

	/*
	 * Register the rendezvous spinloop and wakeup mechanism with SAL
	 */

	/* Register the rendezvous interrupt vector with SAL */
	while (1) {
		isrv = ia64_sal_mc_set_params(SAL_MC_PARAM_RENDEZ_INT,
					      SAL_MC_PARAM_MECHANISM_INT,
					      IA64_MCA_RENDEZ_VECTOR,
					      timeout,
					      SAL_MC_PARAM_RZ_ALWAYS);
		rc = isrv.status;
		if (rc == 0)
			break;
		if (rc == -2) {
			printk(KERN_INFO "Increasing MCA rendezvous timeout from "
				"%ld to %ld milliseconds\n", timeout, isrv.v0);
			timeout = isrv.v0;
			continue;
		}
		printk(KERN_ERR "Failed to register rendezvous interrupt "
		       "with SAL (status %ld)\n", rc);
		return;
	}

	/* Register the wakeup interrupt vector with SAL */
	isrv = ia64_sal_mc_set_params(SAL_MC_PARAM_RENDEZ_WAKEUP,
				      SAL_MC_PARAM_MECHANISM_INT,
				      IA64_MCA_WAKEUP_VECTOR,
				      0, 0);
	rc = isrv.status;
	if (rc) {
		printk(KERN_ERR "Failed to register wakeup interrupt with SAL "
		       "(status %ld)\n", rc);
		return;
	}

	IA64_MCA_DEBUG("%s: registered MCA rendezvous spinloop and wakeup mech.\n", __FUNCTION__);

	ia64_mc_info.imi_mca_handler        = ia64_tpa(mca_hldlr_ptr->fp);
	/*
	 * XXX - disable SAL checksum by setting size to 0; should be
	 *	ia64_tpa(ia64_os_mca_dispatch_end) - ia64_tpa(ia64_os_mca_dispatch);
	 */
	ia64_mc_info.imi_mca_handler_size	= 0;

	/* Register the os mca handler with SAL */
	if ((rc = ia64_sal_set_vectors(SAL_VECTOR_OS_MCA,
				       ia64_mc_info.imi_mca_handler,
				       ia64_tpa(mca_hldlr_ptr->gp),
				       ia64_mc_info.imi_mca_handler_size,
				       0, 0, 0)))
	{
		printk(KERN_ERR "Failed to register OS MCA handler with SAL "
		       "(status %ld)\n", rc);
		return;
	}

	IA64_MCA_DEBUG("%s: registered OS MCA handler with SAL at 0x%lx, gp = 0x%lx\n", __FUNCTION__,
		       ia64_mc_info.imi_mca_handler, ia64_tpa(mca_hldlr_ptr->gp));

	/*
	 * XXX - disable SAL checksum by setting size to 0, should be
	 * size of the actual init handler in mca_asm.S.
	 */
	ia64_mc_info.imi_monarch_init_handler		= ia64_tpa(init_hldlr_ptr_monarch->fp);
	ia64_mc_info.imi_monarch_init_handler_size	= 0;
	ia64_mc_info.imi_slave_init_handler		= ia64_tpa(init_hldlr_ptr_slave->fp);
	ia64_mc_info.imi_slave_init_handler_size	= 0;

	IA64_MCA_DEBUG("%s: OS INIT handler at %lx\n", __FUNCTION__,
		       ia64_mc_info.imi_monarch_init_handler);

	/* Register the os init handler with SAL */
	if ((rc = ia64_sal_set_vectors(SAL_VECTOR_OS_INIT,
				       ia64_mc_info.imi_monarch_init_handler,
				       ia64_tpa(ia64_getreg(_IA64_REG_GP)),
				       ia64_mc_info.imi_monarch_init_handler_size,
				       ia64_mc_info.imi_slave_init_handler,
				       ia64_tpa(ia64_getreg(_IA64_REG_GP)),
				       ia64_mc_info.imi_slave_init_handler_size)))
	{
		printk(KERN_ERR "Failed to register m/s INIT handlers with SAL "
		       "(status %ld)\n", rc);
		return;
	}

	IA64_MCA_DEBUG("%s: registered OS INIT handler with SAL\n", __FUNCTION__);

	/*
	 *  Configure the CMCI/P vector and handler. Interrupts for CMC are
	 *  per-processor, so AP CMC interrupts are setup in smp_callin() (smpboot.c).
	 */
	register_percpu_irq(IA64_CMC_VECTOR, &cmci_irqaction);
	register_percpu_irq(IA64_CMCP_VECTOR, &cmcp_irqaction);
	ia64_mca_cmc_vector_setup();       /* Setup vector on BSP */

	/* Setup the MCA rendezvous interrupt vector */
	register_percpu_irq(IA64_MCA_RENDEZ_VECTOR, &mca_rdzv_irqaction);

	/* Setup the MCA wakeup interrupt vector */
	register_percpu_irq(IA64_MCA_WAKEUP_VECTOR, &mca_wkup_irqaction);

#ifdef CONFIG_ACPI
	/* Setup the CPEI/P handler */
	register_percpu_irq(IA64_CPEP_VECTOR, &mca_cpep_irqaction);
#endif

	/* Initialize the areas set aside by the OS to buffer the
	 * platform/processor error states for MCA/INIT/CMC
	 * handling.
	 */
	ia64_log_init(SAL_INFO_TYPE_MCA);
	ia64_log_init(SAL_INFO_TYPE_INIT);
	ia64_log_init(SAL_INFO_TYPE_CMC);
	ia64_log_init(SAL_INFO_TYPE_CPE);

	mca_init = 1;
	printk(KERN_INFO "MCA related initialization done\n");
}

/*
 * ia64_mca_late_init
 *
 *	Opportunity to setup things that require initialization later
 *	than ia64_mca_init.  Setup a timer to poll for CPEs if the
 *	platform doesn't support an interrupt driven mechanism.
 *
 *  Inputs  :   None
 *  Outputs :   Status
 */
static int __init
ia64_mca_late_init(void)
{
	if (!mca_init)
		return 0;

	/* Setup the CMCI/P vector and handler */
	init_timer(&cmc_poll_timer);
	cmc_poll_timer.function = ia64_mca_cmc_poll;

	/* Unmask/enable the vector */
	cmc_polling_enabled = 0;
	schedule_work(&cmc_enable_work);

	IA64_MCA_DEBUG("%s: CMCI/P setup and enabled.\n", __FUNCTION__);

#ifdef CONFIG_ACPI
	/* Setup the CPEI/P vector and handler */
	cpe_vector = acpi_request_vector(ACPI_INTERRUPT_CPEI);
	init_timer(&cpe_poll_timer);
	cpe_poll_timer.function = ia64_mca_cpe_poll;

	{
		irq_desc_t *desc;
		unsigned int irq;

		if (cpe_vector >= 0) {
			/* If platform supports CPEI, enable the irq. */
			cpe_poll_enabled = 0;
			for (irq = 0; irq < NR_IRQS; ++irq)
				if (irq_to_vector(irq) == cpe_vector) {
					desc = irq_descp(irq);
					desc->status |= IRQ_PER_CPU;
					setup_irq(irq, &mca_cpe_irqaction);
				}
			ia64_mca_register_cpev(cpe_vector);
			IA64_MCA_DEBUG("%s: CPEI/P setup and enabled.\n", __FUNCTION__);
		} else {
			/* If platform doesn't support CPEI, get the timer going. */
			if (cpe_poll_enabled) {
				ia64_mca_cpe_poll(0UL);
				IA64_MCA_DEBUG("%s: CPEP setup and enabled.\n", __FUNCTION__);
			}
		}
	}
#endif

	return 0;
}

device_initcall(ia64_mca_late_init);