aboutsummaryrefslogtreecommitdiff
path: root/drivers/infiniband/hw/ipath/ipath_intr.c
blob: 47aa43428fbf90b1284b6af4dee729debdc976bd (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
/*
 * Copyright (c) 2006, 2007 QLogic Corporation. All rights reserved.
 * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */

#include <linux/pci.h>

#include "ipath_kernel.h"
#include "ipath_verbs.h"
#include "ipath_common.h"

/*
 * clear (write) a pio buffer, to clear a parity error.   This routine
 * should only be called when in freeze mode, and the buffer should be
 * canceled afterwards.
 */
static void ipath_clrpiobuf(struct ipath_devdata *dd, u32 pnum)
{
	u32 __iomem *pbuf;
	u32 dwcnt; /* dword count to write */
	if (pnum < dd->ipath_piobcnt2k) {
		pbuf = (u32 __iomem *) (dd->ipath_pio2kbase + pnum *
			dd->ipath_palign);
		dwcnt = dd->ipath_piosize2k >> 2;
	}
	else {
		pbuf = (u32 __iomem *) (dd->ipath_pio4kbase +
			(pnum - dd->ipath_piobcnt2k) * dd->ipath_4kalign);
		dwcnt = dd->ipath_piosize4k >> 2;
	}
	dev_info(&dd->pcidev->dev,
		"Rewrite PIO buffer %u, to recover from parity error\n",
		pnum);
	*pbuf = dwcnt+1; /* no flush required, since already in freeze */
	while(--dwcnt)
		*pbuf++ = 0;
}

/*
 * Called when we might have an error that is specific to a particular
 * PIO buffer, and may need to cancel that buffer, so it can be re-used.
 * If rewrite is true, and bits are set in the sendbufferror registers,
 * we'll write to the buffer, for error recovery on parity errors.
 */
void ipath_disarm_senderrbufs(struct ipath_devdata *dd, int rewrite)
{
	u32 piobcnt;
	unsigned long sbuf[4];
	/*
	 * it's possible that sendbuffererror could have bits set; might
	 * have already done this as a result of hardware error handling
	 */
	piobcnt = dd->ipath_piobcnt2k + dd->ipath_piobcnt4k;
	/* read these before writing errorclear */
	sbuf[0] = ipath_read_kreg64(
		dd, dd->ipath_kregs->kr_sendbuffererror);
	sbuf[1] = ipath_read_kreg64(
		dd, dd->ipath_kregs->kr_sendbuffererror + 1);
	if (piobcnt > 128) {
		sbuf[2] = ipath_read_kreg64(
			dd, dd->ipath_kregs->kr_sendbuffererror + 2);
		sbuf[3] = ipath_read_kreg64(
			dd, dd->ipath_kregs->kr_sendbuffererror + 3);
	}

	if (sbuf[0] || sbuf[1] || (piobcnt > 128 && (sbuf[2] || sbuf[3]))) {
		int i;
		if (ipath_debug & (__IPATH_PKTDBG|__IPATH_DBG) &&
			dd->ipath_lastcancel > jiffies) {
			__IPATH_DBG_WHICH(__IPATH_PKTDBG|__IPATH_DBG,
					  "SendbufErrs %lx %lx", sbuf[0],
					  sbuf[1]);
			if (ipath_debug & __IPATH_PKTDBG && piobcnt > 128)
				printk(" %lx %lx ", sbuf[2], sbuf[3]);
			printk("\n");
		}

		for (i = 0; i < piobcnt; i++)
			if (test_bit(i, sbuf)) {
				if (rewrite)
					ipath_clrpiobuf(dd, i);
				ipath_disarm_piobufs(dd, i, 1);
			}
		/* ignore armlaunch errs for a bit */
		dd->ipath_lastcancel = jiffies+3;
	}
}


/* These are all rcv-related errors which we want to count for stats */
#define E_SUM_PKTERRS \
	(INFINIPATH_E_RHDRLEN | INFINIPATH_E_RBADTID | \
	 INFINIPATH_E_RBADVERSION | INFINIPATH_E_RHDR | \
	 INFINIPATH_E_RLONGPKTLEN | INFINIPATH_E_RSHORTPKTLEN | \
	 INFINIPATH_E_RMAXPKTLEN | INFINIPATH_E_RMINPKTLEN | \
	 INFINIPATH_E_RFORMATERR | INFINIPATH_E_RUNSUPVL | \
	 INFINIPATH_E_RUNEXPCHAR | INFINIPATH_E_REBP)

/* These are all send-related errors which we want to count for stats */
#define E_SUM_ERRS \
	(INFINIPATH_E_SPIOARMLAUNCH | INFINIPATH_E_SUNEXPERRPKTNUM | \
	 INFINIPATH_E_SDROPPEDDATAPKT | INFINIPATH_E_SDROPPEDSMPPKT | \
	 INFINIPATH_E_SMAXPKTLEN | INFINIPATH_E_SUNSUPVL | \
	 INFINIPATH_E_SMINPKTLEN | INFINIPATH_E_SPKTLEN | \
	 INFINIPATH_E_INVALIDADDR)

/*
 * this is similar to E_SUM_ERRS, but can't ignore armlaunch, don't ignore
 * errors not related to freeze and cancelling buffers.  Can't ignore
 * armlaunch because could get more while still cleaning up, and need
 * to cancel those as they happen.
 */
#define E_SPKT_ERRS_IGNORE \
	 (INFINIPATH_E_SDROPPEDDATAPKT | INFINIPATH_E_SDROPPEDSMPPKT | \
	 INFINIPATH_E_SMAXPKTLEN | INFINIPATH_E_SMINPKTLEN | \
	 INFINIPATH_E_SPKTLEN)

/*
 * these are errors that can occur when the link changes state while
 * a packet is being sent or received.  This doesn't cover things
 * like EBP or VCRC that can be the result of a sending having the
 * link change state, so we receive a "known bad" packet.
 */
#define E_SUM_LINK_PKTERRS \
	(INFINIPATH_E_SDROPPEDDATAPKT | INFINIPATH_E_SDROPPEDSMPPKT | \
	 INFINIPATH_E_SMINPKTLEN | INFINIPATH_E_SPKTLEN | \
	 INFINIPATH_E_RSHORTPKTLEN | INFINIPATH_E_RMINPKTLEN | \
	 INFINIPATH_E_RUNEXPCHAR)

static u64 handle_e_sum_errs(struct ipath_devdata *dd, ipath_err_t errs)
{
	u64 ignore_this_time = 0;

	ipath_disarm_senderrbufs(dd, 0);
	if ((errs & E_SUM_LINK_PKTERRS) &&
	    !(dd->ipath_flags & IPATH_LINKACTIVE)) {
		/*
		 * This can happen when SMA is trying to bring the link
		 * up, but the IB link changes state at the "wrong" time.
		 * The IB logic then complains that the packet isn't
		 * valid.  We don't want to confuse people, so we just
		 * don't print them, except at debug
		 */
		ipath_dbg("Ignoring packet errors %llx, because link not "
			  "ACTIVE\n", (unsigned long long) errs);
		ignore_this_time = errs & E_SUM_LINK_PKTERRS;
	}

	return ignore_this_time;
}

/* generic hw error messages... */
#define INFINIPATH_HWE_TXEMEMPARITYERR_MSG(a) \
	{ \
		.mask = ( INFINIPATH_HWE_TXEMEMPARITYERR_##a <<    \
			  INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT ),   \
		.msg = "TXE " #a " Memory Parity"	     \
	}
#define INFINIPATH_HWE_RXEMEMPARITYERR_MSG(a) \
	{ \
		.mask = ( INFINIPATH_HWE_RXEMEMPARITYERR_##a <<    \
			  INFINIPATH_HWE_RXEMEMPARITYERR_SHIFT ),   \
		.msg = "RXE " #a " Memory Parity"	     \
	}

static const struct ipath_hwerror_msgs ipath_generic_hwerror_msgs[] = {
	INFINIPATH_HWE_MSG(IBCBUSFRSPCPARITYERR, "IPATH2IB Parity"),
	INFINIPATH_HWE_MSG(IBCBUSTOSPCPARITYERR, "IB2IPATH Parity"),

	INFINIPATH_HWE_TXEMEMPARITYERR_MSG(PIOBUF),
	INFINIPATH_HWE_TXEMEMPARITYERR_MSG(PIOPBC),
	INFINIPATH_HWE_TXEMEMPARITYERR_MSG(PIOLAUNCHFIFO),

	INFINIPATH_HWE_RXEMEMPARITYERR_MSG(RCVBUF),
	INFINIPATH_HWE_RXEMEMPARITYERR_MSG(LOOKUPQ),
	INFINIPATH_HWE_RXEMEMPARITYERR_MSG(EAGERTID),
	INFINIPATH_HWE_RXEMEMPARITYERR_MSG(EXPTID),
	INFINIPATH_HWE_RXEMEMPARITYERR_MSG(FLAGBUF),
	INFINIPATH_HWE_RXEMEMPARITYERR_MSG(DATAINFO),
	INFINIPATH_HWE_RXEMEMPARITYERR_MSG(HDRINFO),
};

/**
 * ipath_format_hwmsg - format a single hwerror message
 * @msg message buffer
 * @msgl length of message buffer
 * @hwmsg message to add to message buffer
 */
static void ipath_format_hwmsg(char *msg, size_t msgl, const char *hwmsg)
{
	strlcat(msg, "[", msgl);
	strlcat(msg, hwmsg, msgl);
	strlcat(msg, "]", msgl);
}

/**
 * ipath_format_hwerrors - format hardware error messages for display
 * @hwerrs hardware errors bit vector
 * @hwerrmsgs hardware error descriptions
 * @nhwerrmsgs number of hwerrmsgs
 * @msg message buffer
 * @msgl message buffer length
 */
void ipath_format_hwerrors(u64 hwerrs,
			   const struct ipath_hwerror_msgs *hwerrmsgs,
			   size_t nhwerrmsgs,
			   char *msg, size_t msgl)
{
	int i;
	const int glen =
	    sizeof(ipath_generic_hwerror_msgs) /
	    sizeof(ipath_generic_hwerror_msgs[0]);

	for (i=0; i<glen; i++) {
		if (hwerrs & ipath_generic_hwerror_msgs[i].mask) {
			ipath_format_hwmsg(msg, msgl,
					   ipath_generic_hwerror_msgs[i].msg);
		}
	}

	for (i=0; i<nhwerrmsgs; i++) {
		if (hwerrs & hwerrmsgs[i].mask) {
			ipath_format_hwmsg(msg, msgl, hwerrmsgs[i].msg);
		}
	}
}

/* return the strings for the most common link states */
static char *ib_linkstate(u32 linkstate)
{
	char *ret;

	switch (linkstate) {
	case IPATH_IBSTATE_INIT:
		ret = "Init";
		break;
	case IPATH_IBSTATE_ARM:
		ret = "Arm";
		break;
	case IPATH_IBSTATE_ACTIVE:
		ret = "Active";
		break;
	default:
		ret = "Down";
	}

	return ret;
}

static void handle_e_ibstatuschanged(struct ipath_devdata *dd,
				     ipath_err_t errs, int noprint)
{
	u64 val;
	u32 ltstate, lstate;

	/*
	 * even if diags are enabled, we want to notice LINKINIT, etc.
	 * We just don't want to change the LED state, or
	 * dd->ipath_kregs->kr_ibcctrl
	 */
	val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_ibcstatus);
	lstate = val & IPATH_IBSTATE_MASK;

	/*
	 * this is confusing enough when it happens that I want to always put it
	 * on the console and in the logs.  If it was a requested state change,
	 * we'll have already cleared the flags, so we won't print this warning
	 */
	if ((lstate != IPATH_IBSTATE_ARM && lstate != IPATH_IBSTATE_ACTIVE)
		&& (dd->ipath_flags & (IPATH_LINKARMED | IPATH_LINKACTIVE))) {
		dev_info(&dd->pcidev->dev, "Link state changed from %s to %s\n",
				 (dd->ipath_flags & IPATH_LINKARMED) ? "ARM" : "ACTIVE",
				 ib_linkstate(lstate));
		/*
		 * Flush all queued sends when link went to DOWN or INIT,
		 * to be sure that they don't block SMA and other MAD packets
		 */
		ipath_cancel_sends(dd);
	}
	else if (lstate == IPATH_IBSTATE_INIT || lstate == IPATH_IBSTATE_ARM ||
	    lstate == IPATH_IBSTATE_ACTIVE) {
		/*
		 * only print at SMA if there is a change, debug if not
		 * (sometimes we want to know that, usually not).
		 */
		if (lstate == ((unsigned) dd->ipath_lastibcstat
			       & IPATH_IBSTATE_MASK)) {
			ipath_dbg("Status change intr but no change (%s)\n",
				  ib_linkstate(lstate));
		}
		else
			ipath_cdbg(VERBOSE, "Unit %u link state %s, last "
				   "was %s\n", dd->ipath_unit,
				   ib_linkstate(lstate),
				   ib_linkstate((unsigned)
						dd->ipath_lastibcstat
						& IPATH_IBSTATE_MASK));
	}
	else {
		lstate = dd->ipath_lastibcstat & IPATH_IBSTATE_MASK;
		if (lstate == IPATH_IBSTATE_INIT ||
		    lstate == IPATH_IBSTATE_ARM ||
		    lstate == IPATH_IBSTATE_ACTIVE)
			ipath_cdbg(VERBOSE, "Unit %u link state down"
				   " (state 0x%x), from %s\n",
				   dd->ipath_unit,
				   (u32)val & IPATH_IBSTATE_MASK,
				   ib_linkstate(lstate));
		else
			ipath_cdbg(VERBOSE, "Unit %u link state changed "
				   "to 0x%x from down (%x)\n",
				   dd->ipath_unit, (u32) val, lstate);
	}
	ltstate = (val >> INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) &
		INFINIPATH_IBCS_LINKTRAININGSTATE_MASK;
	lstate = (val >> INFINIPATH_IBCS_LINKSTATE_SHIFT) &
		INFINIPATH_IBCS_LINKSTATE_MASK;

	if (ltstate == INFINIPATH_IBCS_LT_STATE_POLLACTIVE ||
	    ltstate == INFINIPATH_IBCS_LT_STATE_POLLQUIET) {
		u32 last_ltstate;

		/*
		 * Ignore cycling back and forth from Polling.Active
		 * to Polling.Quiet while waiting for the other end of
		 * the link to come up. We will cycle back and forth
		 * between them if no cable is plugged in,
		 * the other device is powered off or disabled, etc.
		 */
		last_ltstate = (dd->ipath_lastibcstat >>
				INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT)
			& INFINIPATH_IBCS_LINKTRAININGSTATE_MASK;
		if (last_ltstate == INFINIPATH_IBCS_LT_STATE_POLLACTIVE
		    || last_ltstate ==
		    INFINIPATH_IBCS_LT_STATE_POLLQUIET) {
			if (dd->ipath_ibpollcnt > 40) {
				dd->ipath_flags |= IPATH_NOCABLE;
				*dd->ipath_statusp |=
					IPATH_STATUS_IB_NOCABLE;
			} else
				dd->ipath_ibpollcnt++;
			goto skip_ibchange;
		}
	}
	dd->ipath_ibpollcnt = 0;	/* some state other than 2 or 3 */
	ipath_stats.sps_iblink++;
	if (ltstate != INFINIPATH_IBCS_LT_STATE_LINKUP) {
		dd->ipath_flags |= IPATH_LINKDOWN;
		dd->ipath_flags &= ~(IPATH_LINKUNK | IPATH_LINKINIT
				     | IPATH_LINKACTIVE |
				     IPATH_LINKARMED);
		*dd->ipath_statusp &= ~IPATH_STATUS_IB_READY;
		dd->ipath_lli_counter = 0;
		if (!noprint) {
			if (((dd->ipath_lastibcstat >>
			      INFINIPATH_IBCS_LINKSTATE_SHIFT) &
			     INFINIPATH_IBCS_LINKSTATE_MASK)
			    == INFINIPATH_IBCS_L_STATE_ACTIVE)
				/* if from up to down be more vocal */
				ipath_cdbg(VERBOSE,
					   "Unit %u link now down (%s)\n",
					   dd->ipath_unit,
					   ipath_ibcstatus_str[ltstate]);
			else
				ipath_cdbg(VERBOSE, "Unit %u link is "
					   "down (%s)\n", dd->ipath_unit,
					   ipath_ibcstatus_str[ltstate]);
		}

		dd->ipath_f_setextled(dd, lstate, ltstate);
	} else if ((val & IPATH_IBSTATE_MASK) == IPATH_IBSTATE_ACTIVE) {
		dd->ipath_flags |= IPATH_LINKACTIVE;
		dd->ipath_flags &=
			~(IPATH_LINKUNK | IPATH_LINKINIT | IPATH_LINKDOWN |
			  IPATH_LINKARMED | IPATH_NOCABLE);
		*dd->ipath_statusp &= ~IPATH_STATUS_IB_NOCABLE;
		*dd->ipath_statusp |=
			IPATH_STATUS_IB_READY | IPATH_STATUS_IB_CONF;
		dd->ipath_f_setextled(dd, lstate, ltstate);
	} else if ((val & IPATH_IBSTATE_MASK) == IPATH_IBSTATE_INIT) {
		/*
		 * set INIT and DOWN.  Down is checked by most of the other
		 * code, but INIT is useful to know in a few places.
		 */
		dd->ipath_flags |= IPATH_LINKINIT | IPATH_LINKDOWN;
		dd->ipath_flags &=
			~(IPATH_LINKUNK | IPATH_LINKACTIVE | IPATH_LINKARMED
			  | IPATH_NOCABLE);
		*dd->ipath_statusp &= ~(IPATH_STATUS_IB_NOCABLE
					| IPATH_STATUS_IB_READY);
		dd->ipath_f_setextled(dd, lstate, ltstate);
	} else if ((val & IPATH_IBSTATE_MASK) == IPATH_IBSTATE_ARM) {
		dd->ipath_flags |= IPATH_LINKARMED;
		dd->ipath_flags &=
			~(IPATH_LINKUNK | IPATH_LINKDOWN | IPATH_LINKINIT |
			  IPATH_LINKACTIVE | IPATH_NOCABLE);
		*dd->ipath_statusp &= ~(IPATH_STATUS_IB_NOCABLE
					| IPATH_STATUS_IB_READY);
		dd->ipath_f_setextled(dd, lstate, ltstate);
	} else {
		if (!noprint)
			ipath_dbg("IBstatuschange unit %u: %s (%x)\n",
				  dd->ipath_unit,
				  ipath_ibcstatus_str[ltstate], ltstate);
	}
skip_ibchange:
	dd->ipath_lastibcstat = val;
}

static void handle_supp_msgs(struct ipath_devdata *dd,
			     unsigned supp_msgs, char msg[512])
{
	/*
	 * Print the message unless it's ibc status change only, which
	 * happens so often we never want to count it.
	 */
	if (dd->ipath_lasterror & ~INFINIPATH_E_IBSTATUSCHANGED) {
		int iserr;
		iserr = ipath_decode_err(msg, sizeof msg,
				dd->ipath_lasterror &
				~INFINIPATH_E_IBSTATUSCHANGED);
		if (dd->ipath_lasterror &
			~(INFINIPATH_E_RRCVEGRFULL |
			INFINIPATH_E_RRCVHDRFULL | INFINIPATH_E_PKTERRS))
			ipath_dev_err(dd, "Suppressed %u messages for "
				      "fast-repeating errors (%s) (%llx)\n",
				      supp_msgs, msg,
				      (unsigned long long)
				      dd->ipath_lasterror);
		else {
			/*
			 * rcvegrfull and rcvhdrqfull are "normal", for some
			 * types of processes (mostly benchmarks) that send
			 * huge numbers of messages, while not processing
			 * them. So only complain about these at debug
			 * level.
			 */
			if (iserr)
				ipath_dbg("Suppressed %u messages for %s\n",
					  supp_msgs, msg);
			else
				ipath_cdbg(ERRPKT,
					"Suppressed %u messages for %s\n",
					  supp_msgs, msg);
		}
	}
}

static unsigned handle_frequent_errors(struct ipath_devdata *dd,
				       ipath_err_t errs, char msg[512],
				       int *noprint)
{
	unsigned long nc;
	static unsigned long nextmsg_time;
	static unsigned nmsgs, supp_msgs;

	/*
	 * Throttle back "fast" messages to no more than 10 per 5 seconds.
	 * This isn't perfect, but it's a reasonable heuristic. If we get
	 * more than 10, give a 6x longer delay.
	 */
	nc = jiffies;
	if (nmsgs > 10) {
		if (time_before(nc, nextmsg_time)) {
			*noprint = 1;
			if (!supp_msgs++)
				nextmsg_time = nc + HZ * 3;
		}
		else if (supp_msgs) {
			handle_supp_msgs(dd, supp_msgs, msg);
			supp_msgs = 0;
			nmsgs = 0;
		}
	}
	else if (!nmsgs++ || time_after(nc, nextmsg_time))
		nextmsg_time = nc + HZ / 2;

	return supp_msgs;
}

static int handle_errors(struct ipath_devdata *dd, ipath_err_t errs)
{
	char msg[512];
	u64 ignore_this_time = 0;
	int i, iserr = 0;
	int chkerrpkts = 0, noprint = 0;
	unsigned supp_msgs;
	int log_idx;

	supp_msgs = handle_frequent_errors(dd, errs, msg, &noprint);

	/*
	 * don't report errors that are masked (includes those always
	 * ignored)
	 */
	errs &= ~dd->ipath_maskederrs;

	/* do these first, they are most important */
	if (errs & INFINIPATH_E_HARDWARE) {
		/* reuse same msg buf */
		dd->ipath_f_handle_hwerrors(dd, msg, sizeof msg);
	} else {
		u64 mask;
		for (log_idx = 0; log_idx < IPATH_EEP_LOG_CNT; ++log_idx) {
			mask = dd->ipath_eep_st_masks[log_idx].errs_to_log;
			if (errs & mask)
				ipath_inc_eeprom_err(dd, log_idx, 1);
		}
	}

	if (!noprint && (errs & ~dd->ipath_e_bitsextant))
		ipath_dev_err(dd, "error interrupt with unknown errors "
			      "%llx set\n", (unsigned long long)
			      (errs & ~dd->ipath_e_bitsextant));

	if (errs & E_SUM_ERRS)
		ignore_this_time = handle_e_sum_errs(dd, errs);
	else if ((errs & E_SUM_LINK_PKTERRS) &&
	    !(dd->ipath_flags & IPATH_LINKACTIVE)) {
		/*
		 * This can happen when SMA is trying to bring the link
		 * up, but the IB link changes state at the "wrong" time.
		 * The IB logic then complains that the packet isn't
		 * valid.  We don't want to confuse people, so we just
		 * don't print them, except at debug
		 */
		ipath_dbg("Ignoring packet errors %llx, because link not "
			  "ACTIVE\n", (unsigned long long) errs);
		ignore_this_time = errs & E_SUM_LINK_PKTERRS;
	}

	if (supp_msgs == 250000) {
		int s_iserr;
		/*
		 * It's not entirely reasonable assuming that the errors set
		 * in the last clear period are all responsible for the
		 * problem, but the alternative is to assume it's the only
		 * ones on this particular interrupt, which also isn't great
		 */
		dd->ipath_maskederrs |= dd->ipath_lasterror | errs;
		ipath_write_kreg(dd, dd->ipath_kregs->kr_errormask,
				 ~dd->ipath_maskederrs);
		s_iserr = ipath_decode_err(msg, sizeof msg,
				 (dd->ipath_maskederrs & ~dd->
				  ipath_ignorederrs));

		if ((dd->ipath_maskederrs & ~dd->ipath_ignorederrs) &
			~(INFINIPATH_E_RRCVEGRFULL |
			INFINIPATH_E_RRCVHDRFULL | INFINIPATH_E_PKTERRS))
			ipath_dev_err(dd, "Temporarily disabling "
			    "error(s) %llx reporting; too frequent (%s)\n",
				(unsigned long long) (dd->ipath_maskederrs &
				~dd->ipath_ignorederrs), msg);
		else {
			/*
			 * rcvegrfull and rcvhdrqfull are "normal",
			 * for some types of processes (mostly benchmarks)
			 * that send huge numbers of messages, while not
			 * processing them.  So only complain about
			 * these at debug level.
			 */
			if (s_iserr)
				ipath_dbg("Temporarily disabling reporting "
				    "too frequent queue full errors (%s)\n",
				    msg);
			else
				ipath_cdbg(ERRPKT,
				    "Temporarily disabling reporting too"
				    " frequent packet errors (%s)\n",
				    msg);
		}

		/*
		 * Re-enable the masked errors after around 3 minutes.  in
		 * ipath_get_faststats().  If we have a series of fast
		 * repeating but different errors, the interval will keep
		 * stretching out, but that's OK, as that's pretty
		 * catastrophic.
		 */
		dd->ipath_unmasktime = jiffies + HZ * 180;
	}

	ipath_write_kreg(dd, dd->ipath_kregs->kr_errorclear, errs);
	if (ignore_this_time)
		errs &= ~ignore_this_time;
	if (errs & ~dd->ipath_lasterror) {
		errs &= ~dd->ipath_lasterror;
		/* never suppress duplicate hwerrors or ibstatuschange */
		dd->ipath_lasterror |= errs &
			~(INFINIPATH_E_HARDWARE |
			  INFINIPATH_E_IBSTATUSCHANGED);
	}

	/* likely due to cancel, so suppress */
	if ((errs & (INFINIPATH_E_SPKTLEN | INFINIPATH_E_SPIOARMLAUNCH)) &&
		dd->ipath_lastcancel > jiffies) {
		ipath_dbg("Suppressed armlaunch/spktlen after error send cancel\n");
		errs &= ~(INFINIPATH_E_SPIOARMLAUNCH | INFINIPATH_E_SPKTLEN);
	}

	if (!errs)
		return 0;

	if (!noprint)
		/*
		 * the ones we mask off are handled specially below or above
		 */
		ipath_decode_err(msg, sizeof msg,
				 errs & ~(INFINIPATH_E_IBSTATUSCHANGED |
					  INFINIPATH_E_RRCVEGRFULL |
					  INFINIPATH_E_RRCVHDRFULL |
					  INFINIPATH_E_HARDWARE));
	else
		/* so we don't need if (!noprint) at strlcat's below */
		*msg = 0;

	if (errs & E_SUM_PKTERRS) {
		ipath_stats.sps_pkterrs++;
		chkerrpkts = 1;
	}
	if (errs & E_SUM_ERRS)
		ipath_stats.sps_errs++;

	if (errs & (INFINIPATH_E_RICRC | INFINIPATH_E_RVCRC)) {
		ipath_stats.sps_crcerrs++;
		chkerrpkts = 1;
	}
	iserr = errs & ~(E_SUM_PKTERRS | INFINIPATH_E_PKTERRS);


	/*
	 * We don't want to print these two as they happen, or we can make
	 * the situation even worse, because it takes so long to print
	 * messages to serial consoles.  Kernel ports get printed from
	 * fast_stats, no more than every 5 seconds, user ports get printed
	 * on close
	 */
	if (errs & INFINIPATH_E_RRCVHDRFULL) {
		u32 hd, tl;
		ipath_stats.sps_hdrqfull++;
		for (i = 0; i < dd->ipath_cfgports; i++) {
			struct ipath_portdata *pd = dd->ipath_pd[i];
			if (i == 0) {
				hd = dd->ipath_port0head;
				tl = (u32) le64_to_cpu(
					*dd->ipath_hdrqtailptr);
			} else if (pd && pd->port_cnt &&
				   pd->port_rcvhdrtail_kvaddr) {
				/*
				 * don't report same point multiple times,
				 * except kernel
				 */
				tl = *(u64 *) pd->port_rcvhdrtail_kvaddr;
				if (tl == dd->ipath_lastrcvhdrqtails[i])
					continue;
				hd = ipath_read_ureg32(dd, ur_rcvhdrhead,
						       i);
			} else
				continue;
			if (hd == (tl + 1) ||
			    (!hd && tl == dd->ipath_hdrqlast)) {
				if (i == 0)
					chkerrpkts = 1;
				dd->ipath_lastrcvhdrqtails[i] = tl;
				pd->port_hdrqfull++;
				if (test_bit(IPATH_PORT_WAITING_OVERFLOW,
					     &pd->port_flag)) {
					clear_bit(
					  IPATH_PORT_WAITING_OVERFLOW,
					  &pd->port_flag);
					set_bit(
					  IPATH_PORT_WAITING_OVERFLOW,
					  &pd->int_flag);
					wake_up_interruptible(
					  &pd->port_wait);
				}
			}
		}
	}
	if (errs & INFINIPATH_E_RRCVEGRFULL) {
		/*
		 * since this is of less importance and not likely to
		 * happen without also getting hdrfull, only count
		 * occurrences; don't check each port (or even the kernel
		 * vs user)
		 */
		ipath_stats.sps_etidfull++;
		if (dd->ipath_port0head !=
		    (u32) le64_to_cpu(*dd->ipath_hdrqtailptr))
			chkerrpkts = 1;
	}

	/*
	 * do this before IBSTATUSCHANGED, in case both bits set in a single
	 * interrupt; we want the STATUSCHANGE to "win", so we do our
	 * internal copy of state machine correctly
	 */
	if (errs & INFINIPATH_E_RIBLOSTLINK) {
		/*
		 * force through block below
		 */
		errs |= INFINIPATH_E_IBSTATUSCHANGED;
		ipath_stats.sps_iblink++;
		dd->ipath_flags |= IPATH_LINKDOWN;
		dd->ipath_flags &= ~(IPATH_LINKUNK | IPATH_LINKINIT
				     | IPATH_LINKARMED | IPATH_LINKACTIVE);
		*dd->ipath_statusp &= ~IPATH_STATUS_IB_READY;
		if (!noprint) {
			u64 st = ipath_read_kreg64(
				dd, dd->ipath_kregs->kr_ibcstatus);

			ipath_dbg("Lost link, link now down (%s)\n",
				  ipath_ibcstatus_str[st & 0xf]);
		}
	}
	if (errs & INFINIPATH_E_IBSTATUSCHANGED)
		handle_e_ibstatuschanged(dd, errs, noprint);

	if (errs & INFINIPATH_E_RESET) {
		if (!noprint)
			ipath_dev_err(dd, "Got reset, requires re-init "
				      "(unload and reload driver)\n");
		dd->ipath_flags &= ~IPATH_INITTED;	/* needs re-init */
		/* mark as having had error */
		*dd->ipath_statusp |= IPATH_STATUS_HWERROR;
		*dd->ipath_statusp &= ~IPATH_STATUS_IB_CONF;
	}

	if (!noprint && *msg) {
		if (iserr)
			ipath_dev_err(dd, "%s error\n", msg);
		else
			dev_info(&dd->pcidev->dev, "%s packet problems\n",
				msg);
	}
	if (dd->ipath_state_wanted & dd->ipath_flags) {
		ipath_cdbg(VERBOSE, "driver wanted state %x, iflags now %x, "
			   "waking\n", dd->ipath_state_wanted,
			   dd->ipath_flags);
		wake_up_interruptible(&ipath_state_wait);
	}

	return chkerrpkts;
}


/*
 * try to cleanup as much as possible for anything that might have gone
 * wrong while in freeze mode, such as pio buffers being written by user
 * processes (causing armlaunch), send errors due to going into freeze mode,
 * etc., and try to avoid causing extra interrupts while doing so.
 * Forcibly update the in-memory pioavail register copies after cleanup
 * because the chip won't do it for anything changing while in freeze mode
 * (we don't want to wait for the next pio buffer state change).
 * Make sure that we don't lose any important interrupts by using the chip
 * feature that says that writing 0 to a bit in *clear that is set in
 * *status will cause an interrupt to be generated again (if allowed by
 * the *mask value).
 */
void ipath_clear_freeze(struct ipath_devdata *dd)
{
	int i, im;
	__le64 val;

	/* disable error interrupts, to avoid confusion */
	ipath_write_kreg(dd, dd->ipath_kregs->kr_errormask, 0ULL);

	/*
	 * clear all sends, because they have may been
	 * completed by usercode while in freeze mode, and
	 * therefore would not be sent, and eventually
	 * might cause the process to run out of bufs
	 */
	ipath_cancel_sends(dd);
	ipath_write_kreg(dd, dd->ipath_kregs->kr_control,
			 dd->ipath_control);

	/* ensure pio avail updates continue */
	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
		 dd->ipath_sendctrl & ~IPATH_S_PIOBUFAVAILUPD);
	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
		 dd->ipath_sendctrl);

	/*
	 * We just enabled pioavailupdate, so dma copy is almost certainly
	 * not yet right, so read the registers directly.  Similar to init
	 */
	for (i = 0; i < dd->ipath_pioavregs; i++) {
		/* deal with 6110 chip bug */
		im = i > 3 ? ((i&1) ? i-1 : i+1) : i;
		val = ipath_read_kreg64(dd, 0x1000+(im*sizeof(u64)));
		dd->ipath_pioavailregs_dma[i] = dd->ipath_pioavailshadow[i]
			= le64_to_cpu(val);
	}

	/*
	 * force new interrupt if any hwerr, error or interrupt bits are
	 * still set, and clear "safe" send packet errors related to freeze
	 * and cancelling sends.  Re-enable error interrupts before possible
	 * force of re-interrupt on pending interrupts.
	 */
	ipath_write_kreg(dd, dd->ipath_kregs->kr_hwerrclear, 0ULL);
	ipath_write_kreg(dd, dd->ipath_kregs->kr_errorclear,
		E_SPKT_ERRS_IGNORE);
	ipath_write_kreg(dd, dd->ipath_kregs->kr_errormask,
		~dd->ipath_maskederrs);
	ipath_write_kreg(dd, dd->ipath_kregs->kr_intclear, 0ULL);
}


/* this is separate to allow for better optimization of ipath_intr() */

static void ipath_bad_intr(struct ipath_devdata *dd, u32 * unexpectp)
{
	/*
	 * sometimes happen during driver init and unload, don't want
	 * to process any interrupts at that point
	 */

	/* this is just a bandaid, not a fix, if something goes badly
	 * wrong */
	if (++*unexpectp > 100) {
		if (++*unexpectp > 105) {
			/*
			 * ok, we must be taking somebody else's interrupts,
			 * due to a messed up mptable and/or PIRQ table, so
			 * unregister the interrupt.  We've seen this during
			 * linuxbios development work, and it may happen in
			 * the future again.
			 */
			if (dd->pcidev && dd->ipath_irq) {
				ipath_dev_err(dd, "Now %u unexpected "
					      "interrupts, unregistering "
					      "interrupt handler\n",
					      *unexpectp);
				ipath_dbg("free_irq of irq %d\n",
					  dd->ipath_irq);
				dd->ipath_f_free_irq(dd);
			}
		}
		if (ipath_read_kreg32(dd, dd->ipath_kregs->kr_intmask)) {
			ipath_dev_err(dd, "%u unexpected interrupts, "
				      "disabling interrupts completely\n",
				      *unexpectp);
			/*
			 * disable all interrupts, something is very wrong
			 */
			ipath_write_kreg(dd, dd->ipath_kregs->kr_intmask,
					 0ULL);
		}
	} else if (*unexpectp > 1)
		ipath_dbg("Interrupt when not ready, should not happen, "
			  "ignoring\n");
}

static void ipath_bad_regread(struct ipath_devdata *dd)
{
	static int allbits;

	/* separate routine, for better optimization of ipath_intr() */

	/*
	 * We print the message and disable interrupts, in hope of
	 * having a better chance of debugging the problem.
	 */
	ipath_dev_err(dd,
		      "Read of interrupt status failed (all bits set)\n");
	if (allbits++) {
		/* disable all interrupts, something is very wrong */
		ipath_write_kreg(dd, dd->ipath_kregs->kr_intmask, 0ULL);
		if (allbits == 2) {
			ipath_dev_err(dd, "Still bad interrupt status, "
				      "unregistering interrupt\n");
			dd->ipath_f_free_irq(dd);
		} else if (allbits > 2) {
			if ((allbits % 10000) == 0)
				printk(".");
		} else
			ipath_dev_err(dd, "Disabling interrupts, "
				      "multiple errors\n");
	}
}

static void handle_port_pioavail(struct ipath_devdata *dd)
{
	u32 i;
	/*
	 * start from port 1, since for now port 0  is never using
	 * wait_event for PIO
	 */
	for (i = 1; dd->ipath_portpiowait && i < dd->ipath_cfgports; i++) {
		struct ipath_portdata *pd = dd->ipath_pd[i];

		if (pd && pd->port_cnt &&
		    dd->ipath_portpiowait & (1U << i)) {
			clear_bit(i, &dd->ipath_portpiowait);
			if (test_bit(IPATH_PORT_WAITING_PIO,
				     &pd->port_flag)) {
				clear_bit(IPATH_PORT_WAITING_PIO,
					  &pd->port_flag);
				wake_up_interruptible(&pd->port_wait);
			}
		}
	}
}

static void handle_layer_pioavail(struct ipath_devdata *dd)
{
	int ret;

	ret = ipath_ib_piobufavail(dd->verbs_dev);
	if (ret > 0)
		goto set;

	return;
set:
	set_bit(IPATH_S_PIOINTBUFAVAIL, &dd->ipath_sendctrl);
	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
			 dd->ipath_sendctrl);
}

/*
 * Handle receive interrupts for user ports; this means a user
 * process was waiting for a packet to arrive, and didn't want
 * to poll
 */
static void handle_urcv(struct ipath_devdata *dd, u32 istat)
{
	u64 portr;
	int i;
	int rcvdint = 0;

	portr = ((istat >> INFINIPATH_I_RCVAVAIL_SHIFT) &
		 dd->ipath_i_rcvavail_mask)
		| ((istat >> INFINIPATH_I_RCVURG_SHIFT) &
		   dd->ipath_i_rcvurg_mask);
	for (i = 1; i < dd->ipath_cfgports; i++) {
		struct ipath_portdata *pd = dd->ipath_pd[i];
		if (portr & (1 << i) && pd && pd->port_cnt) {
			if (test_bit(IPATH_PORT_WAITING_RCV,
				     &pd->port_flag)) {
				clear_bit(IPATH_PORT_WAITING_RCV,
					  &pd->port_flag);
				set_bit(IPATH_PORT_WAITING_RCV,
					&pd->int_flag);
				clear_bit(i + INFINIPATH_R_INTRAVAIL_SHIFT,
					  &dd->ipath_rcvctrl);
				wake_up_interruptible(&pd->port_wait);
				rcvdint = 1;
			} else if (test_bit(IPATH_PORT_WAITING_URG,
					    &pd->port_flag)) {
				clear_bit(IPATH_PORT_WAITING_URG,
					  &pd->port_flag);
				set_bit(IPATH_PORT_WAITING_URG,
					&pd->int_flag);
				wake_up_interruptible(&pd->port_wait);
			}
		}
	}
	if (rcvdint) {
		/* only want to take one interrupt, so turn off the rcv
		 * interrupt for all the ports that we did the wakeup on
		 * (but never for kernel port)
		 */
		ipath_write_kreg(dd, dd->ipath_kregs->kr_rcvctrl,
				 dd->ipath_rcvctrl);
	}
}

irqreturn_t ipath_intr(int irq, void *data)
{
	struct ipath_devdata *dd = data;
	u32 istat, chk0rcv = 0;
	ipath_err_t estat = 0;
	irqreturn_t ret;
	u32 oldhead, curtail;
	static unsigned unexpected = 0;
	static const u32 port0rbits = (1U<<INFINIPATH_I_RCVAVAIL_SHIFT) |
		 (1U<<INFINIPATH_I_RCVURG_SHIFT);

	ipath_stats.sps_ints++;

	if (dd->ipath_int_counter != (u32) -1)
		dd->ipath_int_counter++;

	if (!(dd->ipath_flags & IPATH_PRESENT)) {
		/*
		 * This return value is not great, but we do not want the
		 * interrupt core code to remove our interrupt handler
		 * because we don't appear to be handling an interrupt
		 * during a chip reset.
		 */
		return IRQ_HANDLED;
	}

	/*
	 * this needs to be flags&initted, not statusp, so we keep
	 * taking interrupts even after link goes down, etc.
	 * Also, we *must* clear the interrupt at some point, or we won't
	 * take it again, which can be real bad for errors, etc...
	 */

	if (!(dd->ipath_flags & IPATH_INITTED)) {
		ipath_bad_intr(dd, &unexpected);
		ret = IRQ_NONE;
		goto bail;
	}

	/*
	 * We try to avoid reading the interrupt status register, since
	 * that's a PIO read, and stalls the processor for up to about
	 * ~0.25 usec. The idea is that if we processed a port0 packet,
	 * we blindly clear the  port 0 receive interrupt bits, and nothing
	 * else, then return.  If other interrupts are pending, the chip
	 * will re-interrupt us as soon as we write the intclear register.
	 * We then won't process any more kernel packets (if not the 2nd
	 * time, then the 3rd or 4th) and we'll then handle the other
	 * interrupts.   We clear the interrupts first so that we don't
	 * lose intr for later packets that arrive while we are processing.
	 */
	oldhead = dd->ipath_port0head;
	curtail = (u32)le64_to_cpu(*dd->ipath_hdrqtailptr);
	if (oldhead != curtail) {
		if (dd->ipath_flags & IPATH_GPIO_INTR) {
			ipath_write_kreg(dd, dd->ipath_kregs->kr_gpio_clear,
					 (u64) (1 << IPATH_GPIO_PORT0_BIT));
			istat = port0rbits | INFINIPATH_I_GPIO;
		}
		else
			istat = port0rbits;
		ipath_write_kreg(dd, dd->ipath_kregs->kr_intclear, istat);
		ipath_kreceive(dd);
		if (oldhead != dd->ipath_port0head) {
			ipath_stats.sps_fastrcvint++;
			goto done;
		}
	}

	istat = ipath_read_kreg32(dd, dd->ipath_kregs->kr_intstatus);

	if (unlikely(!istat)) {
		ipath_stats.sps_nullintr++;
		ret = IRQ_NONE; /* not our interrupt, or already handled */
		goto bail;
	}
	if (unlikely(istat == -1)) {
		ipath_bad_regread(dd);
		/* don't know if it was our interrupt or not */
		ret = IRQ_NONE;
		goto bail;
	}

	if (unexpected)
		unexpected = 0;

	if (unlikely(istat & ~dd->ipath_i_bitsextant))
		ipath_dev_err(dd,
			      "interrupt with unknown interrupts %x set\n",
			      istat & (u32) ~ dd->ipath_i_bitsextant);
	else
		ipath_cdbg(VERBOSE, "intr stat=0x%x\n", istat);

	if (unlikely(istat & INFINIPATH_I_ERROR)) {
		ipath_stats.sps_errints++;
		estat = ipath_read_kreg64(dd,
					  dd->ipath_kregs->kr_errorstatus);
		if (!estat)
			dev_info(&dd->pcidev->dev, "error interrupt (%x), "
				 "but no error bits set!\n", istat);
		else if (estat == -1LL)
			/*
			 * should we try clearing all, or hope next read
			 * works?
			 */
			ipath_dev_err(dd, "Read of error status failed "
				      "(all bits set); ignoring\n");
		else
			if (handle_errors(dd, estat))
				/* force calling ipath_kreceive() */
				chk0rcv = 1;
	}

	if (istat & INFINIPATH_I_GPIO) {
		/*
		 * GPIO interrupts fall in two broad classes:
		 * GPIO_2 indicates (on some HT4xx boards) that a packet
		 *        has arrived for Port 0. Checking for this
		 *        is controlled by flag IPATH_GPIO_INTR.
		 * GPIO_3..5 on IBA6120 Rev2 chips indicate errors
		 *        that we need to count. Checking for this
		 *        is controlled by flag IPATH_GPIO_ERRINTRS.
		 */
		u32 gpiostatus;
		u32 to_clear = 0;

		gpiostatus = ipath_read_kreg32(
			dd, dd->ipath_kregs->kr_gpio_status);
		/* First the error-counter case.
		 */
		if ((gpiostatus & IPATH_GPIO_ERRINTR_MASK) &&
		    (dd->ipath_flags & IPATH_GPIO_ERRINTRS)) {
			/* want to clear the bits we see asserted. */
			to_clear |= (gpiostatus & IPATH_GPIO_ERRINTR_MASK);

			/*
			 * Count appropriately, clear bits out of our copy,
			 * as they have been "handled".
			 */
			if (gpiostatus & (1 << IPATH_GPIO_RXUVL_BIT)) {
				ipath_dbg("FlowCtl on UnsupVL\n");
				dd->ipath_rxfc_unsupvl_errs++;
			}
			if (gpiostatus & (1 << IPATH_GPIO_OVRUN_BIT)) {
				ipath_dbg("Overrun Threshold exceeded\n");
				dd->ipath_overrun_thresh_errs++;
			}
			if (gpiostatus & (1 << IPATH_GPIO_LLI_BIT)) {
				ipath_dbg("Local Link Integrity error\n");
				dd->ipath_lli_errs++;
			}
			gpiostatus &= ~IPATH_GPIO_ERRINTR_MASK;
		}
		/* Now the Port0 Receive case */
		if ((gpiostatus & (1 << IPATH_GPIO_PORT0_BIT)) &&
		    (dd->ipath_flags & IPATH_GPIO_INTR)) {
			/*
			 * GPIO status bit 2 is set, and we expected it.
			 * clear it and indicate in p0bits.
			 * This probably only happens if a Port0 pkt
			 * arrives at _just_ the wrong time, and we
			 * handle that by seting chk0rcv;
			 */
			to_clear |= (1 << IPATH_GPIO_PORT0_BIT);
			gpiostatus &= ~(1 << IPATH_GPIO_PORT0_BIT);
			chk0rcv = 1;
		}
		if (gpiostatus) {
			/*
			 * Some unexpected bits remain. If they could have
			 * caused the interrupt, complain and clear.
			 * MEA: this is almost certainly non-ideal.
			 * we should look into auto-disable of unexpected
			 * GPIO interrupts, possibly on a "three strikes"
			 * basis.
			 */
			const u32 mask = (u32) dd->ipath_gpio_mask;

			if (mask & gpiostatus) {
				ipath_dbg("Unexpected GPIO IRQ bits %x\n",
				  gpiostatus & mask);
				to_clear |= (gpiostatus & mask);
			}
		}
		if (to_clear) {
			ipath_write_kreg(dd, dd->ipath_kregs->kr_gpio_clear,
					(u64) to_clear);
		}
	}
	chk0rcv |= istat & port0rbits;

	/*
	 * Clear the interrupt bits we found set, unless they are receive
	 * related, in which case we already cleared them above, and don't
	 * want to clear them again, because we might lose an interrupt.
	 * Clear it early, so we "know" know the chip will have seen this by
	 * the time we process the queue, and will re-interrupt if necessary.
	 * The processor itself won't take the interrupt again until we return.
	 */
	ipath_write_kreg(dd, dd->ipath_kregs->kr_intclear, istat);

	/*
	 * handle port0 receive  before checking for pio buffers available,
	 * since receives can overflow; piobuf waiters can afford a few
	 * extra cycles, since they were waiting anyway, and user's waiting
	 * for receive are at the bottom.
	 */
	if (chk0rcv) {
		ipath_kreceive(dd);
		istat &= ~port0rbits;
	}

	if (istat & ((dd->ipath_i_rcvavail_mask <<
		      INFINIPATH_I_RCVAVAIL_SHIFT)
		     | (dd->ipath_i_rcvurg_mask <<
			INFINIPATH_I_RCVURG_SHIFT)))
		handle_urcv(dd, istat);

	if (istat & INFINIPATH_I_SPIOBUFAVAIL) {
		clear_bit(IPATH_S_PIOINTBUFAVAIL, &dd->ipath_sendctrl);
		ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
				 dd->ipath_sendctrl);

		if (dd->ipath_portpiowait)
			handle_port_pioavail(dd);

		handle_layer_pioavail(dd);
	}

done:
	ret = IRQ_HANDLED;

bail:
	return ret;
}