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
|
/*
* Copyright (c) 2006, 2007, 2008, 2009, 2010 QLogic Corporation.
* All rights reserved.
* Copyright (c) 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/err.h>
#include <linux/vmalloc.h>
#include "qib.h"
#define BITS_PER_PAGE (PAGE_SIZE*BITS_PER_BYTE)
#define BITS_PER_PAGE_MASK (BITS_PER_PAGE-1)
static inline unsigned mk_qpn(struct qib_qpn_table *qpt,
struct qpn_map *map, unsigned off)
{
return (map - qpt->map) * BITS_PER_PAGE + off;
}
static inline unsigned find_next_offset(struct qib_qpn_table *qpt,
struct qpn_map *map, unsigned off,
unsigned n)
{
if (qpt->mask) {
off++;
if (((off & qpt->mask) >> 1) >= n)
off = (off | qpt->mask) + 2;
} else
off = find_next_zero_bit(map->page, BITS_PER_PAGE, off);
return off;
}
/*
* Convert the AETH credit code into the number of credits.
*/
static u32 credit_table[31] = {
0, /* 0 */
1, /* 1 */
2, /* 2 */
3, /* 3 */
4, /* 4 */
6, /* 5 */
8, /* 6 */
12, /* 7 */
16, /* 8 */
24, /* 9 */
32, /* A */
48, /* B */
64, /* C */
96, /* D */
128, /* E */
192, /* F */
256, /* 10 */
384, /* 11 */
512, /* 12 */
768, /* 13 */
1024, /* 14 */
1536, /* 15 */
2048, /* 16 */
3072, /* 17 */
4096, /* 18 */
6144, /* 19 */
8192, /* 1A */
12288, /* 1B */
16384, /* 1C */
24576, /* 1D */
32768 /* 1E */
};
static void get_map_page(struct qib_qpn_table *qpt, struct qpn_map *map)
{
unsigned long page = get_zeroed_page(GFP_KERNEL);
/*
* Free the page if someone raced with us installing it.
*/
spin_lock(&qpt->lock);
if (map->page)
free_page(page);
else
map->page = (void *)page;
spin_unlock(&qpt->lock);
}
/*
* Allocate the next available QPN or
* zero/one for QP type IB_QPT_SMI/IB_QPT_GSI.
*/
static int alloc_qpn(struct qib_devdata *dd, struct qib_qpn_table *qpt,
enum ib_qp_type type, u8 port)
{
u32 i, offset, max_scan, qpn;
struct qpn_map *map;
u32 ret;
if (type == IB_QPT_SMI || type == IB_QPT_GSI) {
unsigned n;
ret = type == IB_QPT_GSI;
n = 1 << (ret + 2 * (port - 1));
spin_lock(&qpt->lock);
if (qpt->flags & n)
ret = -EINVAL;
else
qpt->flags |= n;
spin_unlock(&qpt->lock);
goto bail;
}
qpn = qpt->last + 1;
if (qpn >= QPN_MAX)
qpn = 2;
if (qpt->mask && ((qpn & qpt->mask) >> 1) >= dd->n_krcv_queues)
qpn = (qpn | qpt->mask) + 2;
offset = qpn & BITS_PER_PAGE_MASK;
map = &qpt->map[qpn / BITS_PER_PAGE];
max_scan = qpt->nmaps - !offset;
for (i = 0;;) {
if (unlikely(!map->page)) {
get_map_page(qpt, map);
if (unlikely(!map->page))
break;
}
do {
if (!test_and_set_bit(offset, map->page)) {
qpt->last = qpn;
ret = qpn;
goto bail;
}
offset = find_next_offset(qpt, map, offset,
dd->n_krcv_queues);
qpn = mk_qpn(qpt, map, offset);
/*
* This test differs from alloc_pidmap().
* If find_next_offset() does find a zero
* bit, we don't need to check for QPN
* wrapping around past our starting QPN.
* We just need to be sure we don't loop
* forever.
*/
} while (offset < BITS_PER_PAGE && qpn < QPN_MAX);
/*
* In order to keep the number of pages allocated to a
* minimum, we scan the all existing pages before increasing
* the size of the bitmap table.
*/
if (++i > max_scan) {
if (qpt->nmaps == QPNMAP_ENTRIES)
break;
map = &qpt->map[qpt->nmaps++];
offset = 0;
} else if (map < &qpt->map[qpt->nmaps]) {
++map;
offset = 0;
} else {
map = &qpt->map[0];
offset = 2;
}
qpn = mk_qpn(qpt, map, offset);
}
ret = -ENOMEM;
bail:
return ret;
}
static void free_qpn(struct qib_qpn_table *qpt, u32 qpn)
{
struct qpn_map *map;
map = qpt->map + qpn / BITS_PER_PAGE;
if (map->page)
clear_bit(qpn & BITS_PER_PAGE_MASK, map->page);
}
/*
* Put the QP into the hash table.
* The hash table holds a reference to the QP.
*/
static void insert_qp(struct qib_ibdev *dev, struct qib_qp *qp)
{
struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
unsigned n = qp->ibqp.qp_num % dev->qp_table_size;
unsigned long flags;
spin_lock_irqsave(&dev->qpt_lock, flags);
if (qp->ibqp.qp_num == 0)
ibp->qp0 = qp;
else if (qp->ibqp.qp_num == 1)
ibp->qp1 = qp;
else {
qp->next = dev->qp_table[n];
dev->qp_table[n] = qp;
}
atomic_inc(&qp->refcount);
spin_unlock_irqrestore(&dev->qpt_lock, flags);
}
/*
* Remove the QP from the table so it can't be found asynchronously by
* the receive interrupt routine.
*/
static void remove_qp(struct qib_ibdev *dev, struct qib_qp *qp)
{
struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
struct qib_qp *q, **qpp;
unsigned long flags;
qpp = &dev->qp_table[qp->ibqp.qp_num % dev->qp_table_size];
spin_lock_irqsave(&dev->qpt_lock, flags);
if (ibp->qp0 == qp) {
ibp->qp0 = NULL;
atomic_dec(&qp->refcount);
} else if (ibp->qp1 == qp) {
ibp->qp1 = NULL;
atomic_dec(&qp->refcount);
} else
for (; (q = *qpp) != NULL; qpp = &q->next)
if (q == qp) {
*qpp = qp->next;
qp->next = NULL;
atomic_dec(&qp->refcount);
break;
}
spin_unlock_irqrestore(&dev->qpt_lock, flags);
}
/**
* qib_free_all_qps - check for QPs still in use
* @qpt: the QP table to empty
*
* There should not be any QPs still in use.
* Free memory for table.
*/
unsigned qib_free_all_qps(struct qib_devdata *dd)
{
struct qib_ibdev *dev = &dd->verbs_dev;
unsigned long flags;
struct qib_qp *qp;
unsigned n, qp_inuse = 0;
for (n = 0; n < dd->num_pports; n++) {
struct qib_ibport *ibp = &dd->pport[n].ibport_data;
if (!qib_mcast_tree_empty(ibp))
qp_inuse++;
if (ibp->qp0)
qp_inuse++;
if (ibp->qp1)
qp_inuse++;
}
spin_lock_irqsave(&dev->qpt_lock, flags);
for (n = 0; n < dev->qp_table_size; n++) {
qp = dev->qp_table[n];
dev->qp_table[n] = NULL;
for (; qp; qp = qp->next)
qp_inuse++;
}
spin_unlock_irqrestore(&dev->qpt_lock, flags);
return qp_inuse;
}
/**
* qib_lookup_qpn - return the QP with the given QPN
* @qpt: the QP table
* @qpn: the QP number to look up
*
* The caller is responsible for decrementing the QP reference count
* when done.
*/
struct qib_qp *qib_lookup_qpn(struct qib_ibport *ibp, u32 qpn)
{
struct qib_ibdev *dev = &ppd_from_ibp(ibp)->dd->verbs_dev;
unsigned long flags;
struct qib_qp *qp;
spin_lock_irqsave(&dev->qpt_lock, flags);
if (qpn == 0)
qp = ibp->qp0;
else if (qpn == 1)
qp = ibp->qp1;
else
for (qp = dev->qp_table[qpn % dev->qp_table_size]; qp;
qp = qp->next)
if (qp->ibqp.qp_num == qpn)
break;
if (qp)
atomic_inc(&qp->refcount);
spin_unlock_irqrestore(&dev->qpt_lock, flags);
return qp;
}
/**
* qib_reset_qp - initialize the QP state to the reset state
* @qp: the QP to reset
* @type: the QP type
*/
static void qib_reset_qp(struct qib_qp *qp, enum ib_qp_type type)
{
qp->remote_qpn = 0;
qp->qkey = 0;
qp->qp_access_flags = 0;
atomic_set(&qp->s_dma_busy, 0);
qp->s_flags &= QIB_S_SIGNAL_REQ_WR;
qp->s_hdrwords = 0;
qp->s_wqe = NULL;
qp->s_draining = 0;
qp->s_next_psn = 0;
qp->s_last_psn = 0;
qp->s_sending_psn = 0;
qp->s_sending_hpsn = 0;
qp->s_psn = 0;
qp->r_psn = 0;
qp->r_msn = 0;
if (type == IB_QPT_RC) {
qp->s_state = IB_OPCODE_RC_SEND_LAST;
qp->r_state = IB_OPCODE_RC_SEND_LAST;
} else {
qp->s_state = IB_OPCODE_UC_SEND_LAST;
qp->r_state = IB_OPCODE_UC_SEND_LAST;
}
qp->s_ack_state = IB_OPCODE_RC_ACKNOWLEDGE;
qp->r_nak_state = 0;
qp->r_aflags = 0;
qp->r_flags = 0;
qp->s_head = 0;
qp->s_tail = 0;
qp->s_cur = 0;
qp->s_acked = 0;
qp->s_last = 0;
qp->s_ssn = 1;
qp->s_lsn = 0;
qp->s_mig_state = IB_MIG_MIGRATED;
memset(qp->s_ack_queue, 0, sizeof(qp->s_ack_queue));
qp->r_head_ack_queue = 0;
qp->s_tail_ack_queue = 0;
qp->s_num_rd_atomic = 0;
if (qp->r_rq.wq) {
qp->r_rq.wq->head = 0;
qp->r_rq.wq->tail = 0;
}
qp->r_sge.num_sge = 0;
}
static void clear_mr_refs(struct qib_qp *qp, int clr_sends)
{
unsigned n;
if (test_and_clear_bit(QIB_R_REWIND_SGE, &qp->r_aflags))
while (qp->s_rdma_read_sge.num_sge) {
atomic_dec(&qp->s_rdma_read_sge.sge.mr->refcount);
if (--qp->s_rdma_read_sge.num_sge)
qp->s_rdma_read_sge.sge =
*qp->s_rdma_read_sge.sg_list++;
}
while (qp->r_sge.num_sge) {
atomic_dec(&qp->r_sge.sge.mr->refcount);
if (--qp->r_sge.num_sge)
qp->r_sge.sge = *qp->r_sge.sg_list++;
}
if (clr_sends) {
while (qp->s_last != qp->s_head) {
struct qib_swqe *wqe = get_swqe_ptr(qp, qp->s_last);
unsigned i;
for (i = 0; i < wqe->wr.num_sge; i++) {
struct qib_sge *sge = &wqe->sg_list[i];
atomic_dec(&sge->mr->refcount);
}
if (qp->ibqp.qp_type == IB_QPT_UD ||
qp->ibqp.qp_type == IB_QPT_SMI ||
qp->ibqp.qp_type == IB_QPT_GSI)
atomic_dec(&to_iah(wqe->wr.wr.ud.ah)->refcount);
if (++qp->s_last >= qp->s_size)
qp->s_last = 0;
}
if (qp->s_rdma_mr) {
atomic_dec(&qp->s_rdma_mr->refcount);
qp->s_rdma_mr = NULL;
}
}
if (qp->ibqp.qp_type != IB_QPT_RC)
return;
for (n = 0; n < ARRAY_SIZE(qp->s_ack_queue); n++) {
struct qib_ack_entry *e = &qp->s_ack_queue[n];
if (e->opcode == IB_OPCODE_RC_RDMA_READ_REQUEST &&
e->rdma_sge.mr) {
atomic_dec(&e->rdma_sge.mr->refcount);
e->rdma_sge.mr = NULL;
}
}
}
/**
* qib_error_qp - put a QP into the error state
* @qp: the QP to put into the error state
* @err: the receive completion error to signal if a RWQE is active
*
* Flushes both send and receive work queues.
* Returns true if last WQE event should be generated.
* The QP r_lock and s_lock should be held and interrupts disabled.
* If we are already in error state, just return.
*/
int qib_error_qp(struct qib_qp *qp, enum ib_wc_status err)
{
struct qib_ibdev *dev = to_idev(qp->ibqp.device);
struct ib_wc wc;
int ret = 0;
if (qp->state == IB_QPS_ERR || qp->state == IB_QPS_RESET)
goto bail;
qp->state = IB_QPS_ERR;
if (qp->s_flags & (QIB_S_TIMER | QIB_S_WAIT_RNR)) {
qp->s_flags &= ~(QIB_S_TIMER | QIB_S_WAIT_RNR);
del_timer(&qp->s_timer);
}
if (qp->s_flags & QIB_S_ANY_WAIT_SEND)
qp->s_flags &= ~QIB_S_ANY_WAIT_SEND;
spin_lock(&dev->pending_lock);
if (!list_empty(&qp->iowait) && !(qp->s_flags & QIB_S_BUSY)) {
qp->s_flags &= ~QIB_S_ANY_WAIT_IO;
list_del_init(&qp->iowait);
}
spin_unlock(&dev->pending_lock);
if (!(qp->s_flags & QIB_S_BUSY)) {
qp->s_hdrwords = 0;
if (qp->s_rdma_mr) {
atomic_dec(&qp->s_rdma_mr->refcount);
qp->s_rdma_mr = NULL;
}
if (qp->s_tx) {
qib_put_txreq(qp->s_tx);
qp->s_tx = NULL;
}
}
/* Schedule the sending tasklet to drain the send work queue. */
if (qp->s_last != qp->s_head)
qib_schedule_send(qp);
clear_mr_refs(qp, 0);
memset(&wc, 0, sizeof(wc));
wc.qp = &qp->ibqp;
wc.opcode = IB_WC_RECV;
if (test_and_clear_bit(QIB_R_WRID_VALID, &qp->r_aflags)) {
wc.wr_id = qp->r_wr_id;
wc.status = err;
qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, 1);
}
wc.status = IB_WC_WR_FLUSH_ERR;
if (qp->r_rq.wq) {
struct qib_rwq *wq;
u32 head;
u32 tail;
spin_lock(&qp->r_rq.lock);
/* sanity check pointers before trusting them */
wq = qp->r_rq.wq;
head = wq->head;
if (head >= qp->r_rq.size)
head = 0;
tail = wq->tail;
if (tail >= qp->r_rq.size)
tail = 0;
while (tail != head) {
wc.wr_id = get_rwqe_ptr(&qp->r_rq, tail)->wr_id;
if (++tail >= qp->r_rq.size)
tail = 0;
qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, 1);
}
wq->tail = tail;
spin_unlock(&qp->r_rq.lock);
} else if (qp->ibqp.event_handler)
ret = 1;
bail:
return ret;
}
/**
* qib_modify_qp - modify the attributes of a queue pair
* @ibqp: the queue pair who's attributes we're modifying
* @attr: the new attributes
* @attr_mask: the mask of attributes to modify
* @udata: user data for libibverbs.so
*
* Returns 0 on success, otherwise returns an errno.
*/
int qib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
int attr_mask, struct ib_udata *udata)
{
struct qib_ibdev *dev = to_idev(ibqp->device);
struct qib_qp *qp = to_iqp(ibqp);
enum ib_qp_state cur_state, new_state;
struct ib_event ev;
int lastwqe = 0;
int mig = 0;
int ret;
u32 pmtu = 0; /* for gcc warning only */
spin_lock_irq(&qp->r_lock);
spin_lock(&qp->s_lock);
cur_state = attr_mask & IB_QP_CUR_STATE ?
attr->cur_qp_state : qp->state;
new_state = attr_mask & IB_QP_STATE ? attr->qp_state : cur_state;
if (!ib_modify_qp_is_ok(cur_state, new_state, ibqp->qp_type,
attr_mask))
goto inval;
if (attr_mask & IB_QP_AV) {
if (attr->ah_attr.dlid >= QIB_MULTICAST_LID_BASE)
goto inval;
if (qib_check_ah(qp->ibqp.device, &attr->ah_attr))
goto inval;
}
if (attr_mask & IB_QP_ALT_PATH) {
if (attr->alt_ah_attr.dlid >= QIB_MULTICAST_LID_BASE)
goto inval;
if (qib_check_ah(qp->ibqp.device, &attr->alt_ah_attr))
goto inval;
if (attr->alt_pkey_index >= qib_get_npkeys(dd_from_dev(dev)))
goto inval;
}
if (attr_mask & IB_QP_PKEY_INDEX)
if (attr->pkey_index >= qib_get_npkeys(dd_from_dev(dev)))
goto inval;
if (attr_mask & IB_QP_MIN_RNR_TIMER)
if (attr->min_rnr_timer > 31)
goto inval;
if (attr_mask & IB_QP_PORT)
if (qp->ibqp.qp_type == IB_QPT_SMI ||
qp->ibqp.qp_type == IB_QPT_GSI ||
attr->port_num == 0 ||
attr->port_num > ibqp->device->phys_port_cnt)
goto inval;
if (attr_mask & IB_QP_DEST_QPN)
if (attr->dest_qp_num > QIB_QPN_MASK)
goto inval;
if (attr_mask & IB_QP_RETRY_CNT)
if (attr->retry_cnt > 7)
goto inval;
if (attr_mask & IB_QP_RNR_RETRY)
if (attr->rnr_retry > 7)
goto inval;
/*
* Don't allow invalid path_mtu values. OK to set greater
* than the active mtu (or even the max_cap, if we have tuned
* that to a small mtu. We'll set qp->path_mtu
* to the lesser of requested attribute mtu and active,
* for packetizing messages.
* Note that the QP port has to be set in INIT and MTU in RTR.
*/
if (attr_mask & IB_QP_PATH_MTU) {
struct qib_devdata *dd = dd_from_dev(dev);
int mtu, pidx = qp->port_num - 1;
mtu = ib_mtu_enum_to_int(attr->path_mtu);
if (mtu == -1)
goto inval;
if (mtu > dd->pport[pidx].ibmtu) {
switch (dd->pport[pidx].ibmtu) {
case 4096:
pmtu = IB_MTU_4096;
break;
case 2048:
pmtu = IB_MTU_2048;
break;
case 1024:
pmtu = IB_MTU_1024;
break;
case 512:
pmtu = IB_MTU_512;
break;
case 256:
pmtu = IB_MTU_256;
break;
default:
pmtu = IB_MTU_2048;
}
} else
pmtu = attr->path_mtu;
}
if (attr_mask & IB_QP_PATH_MIG_STATE) {
if (attr->path_mig_state == IB_MIG_REARM) {
if (qp->s_mig_state == IB_MIG_ARMED)
goto inval;
if (new_state != IB_QPS_RTS)
goto inval;
} else if (attr->path_mig_state == IB_MIG_MIGRATED) {
if (qp->s_mig_state == IB_MIG_REARM)
goto inval;
if (new_state != IB_QPS_RTS && new_state != IB_QPS_SQD)
goto inval;
if (qp->s_mig_state == IB_MIG_ARMED)
mig = 1;
} else
goto inval;
}
if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
if (attr->max_dest_rd_atomic > QIB_MAX_RDMA_ATOMIC)
goto inval;
switch (new_state) {
case IB_QPS_RESET:
if (qp->state != IB_QPS_RESET) {
qp->state = IB_QPS_RESET;
spin_lock(&dev->pending_lock);
if (!list_empty(&qp->iowait))
list_del_init(&qp->iowait);
spin_unlock(&dev->pending_lock);
qp->s_flags &= ~(QIB_S_TIMER | QIB_S_ANY_WAIT);
spin_unlock(&qp->s_lock);
spin_unlock_irq(&qp->r_lock);
/* Stop the sending work queue and retry timer */
cancel_work_sync(&qp->s_work);
del_timer_sync(&qp->s_timer);
wait_event(qp->wait_dma, !atomic_read(&qp->s_dma_busy));
if (qp->s_tx) {
qib_put_txreq(qp->s_tx);
qp->s_tx = NULL;
}
remove_qp(dev, qp);
wait_event(qp->wait, !atomic_read(&qp->refcount));
spin_lock_irq(&qp->r_lock);
spin_lock(&qp->s_lock);
clear_mr_refs(qp, 1);
qib_reset_qp(qp, ibqp->qp_type);
}
break;
case IB_QPS_RTR:
/* Allow event to retrigger if QP set to RTR more than once */
qp->r_flags &= ~QIB_R_COMM_EST;
qp->state = new_state;
break;
case IB_QPS_SQD:
qp->s_draining = qp->s_last != qp->s_cur;
qp->state = new_state;
break;
case IB_QPS_SQE:
if (qp->ibqp.qp_type == IB_QPT_RC)
goto inval;
qp->state = new_state;
break;
case IB_QPS_ERR:
lastwqe = qib_error_qp(qp, IB_WC_WR_FLUSH_ERR);
break;
default:
qp->state = new_state;
break;
}
if (attr_mask & IB_QP_PKEY_INDEX)
qp->s_pkey_index = attr->pkey_index;
if (attr_mask & IB_QP_PORT)
qp->port_num = attr->port_num;
if (attr_mask & IB_QP_DEST_QPN)
qp->remote_qpn = attr->dest_qp_num;
if (attr_mask & IB_QP_SQ_PSN) {
qp->s_next_psn = attr->sq_psn & QIB_PSN_MASK;
qp->s_psn = qp->s_next_psn;
qp->s_sending_psn = qp->s_next_psn;
qp->s_last_psn = qp->s_next_psn - 1;
qp->s_sending_hpsn = qp->s_last_psn;
}
if (attr_mask & IB_QP_RQ_PSN)
qp->r_psn = attr->rq_psn & QIB_PSN_MASK;
if (attr_mask & IB_QP_ACCESS_FLAGS)
qp->qp_access_flags = attr->qp_access_flags;
if (attr_mask & IB_QP_AV) {
qp->remote_ah_attr = attr->ah_attr;
qp->s_srate = attr->ah_attr.static_rate;
}
if (attr_mask & IB_QP_ALT_PATH) {
qp->alt_ah_attr = attr->alt_ah_attr;
qp->s_alt_pkey_index = attr->alt_pkey_index;
}
if (attr_mask & IB_QP_PATH_MIG_STATE) {
qp->s_mig_state = attr->path_mig_state;
if (mig) {
qp->remote_ah_attr = qp->alt_ah_attr;
qp->port_num = qp->alt_ah_attr.port_num;
qp->s_pkey_index = qp->s_alt_pkey_index;
}
}
if (attr_mask & IB_QP_PATH_MTU)
qp->path_mtu = pmtu;
if (attr_mask & IB_QP_RETRY_CNT) {
qp->s_retry_cnt = attr->retry_cnt;
qp->s_retry = attr->retry_cnt;
}
if (attr_mask & IB_QP_RNR_RETRY) {
qp->s_rnr_retry_cnt = attr->rnr_retry;
qp->s_rnr_retry = attr->rnr_retry;
}
if (attr_mask & IB_QP_MIN_RNR_TIMER)
qp->r_min_rnr_timer = attr->min_rnr_timer;
if (attr_mask & IB_QP_TIMEOUT)
qp->timeout = attr->timeout;
if (attr_mask & IB_QP_QKEY)
qp->qkey = attr->qkey;
if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
qp->r_max_rd_atomic = attr->max_dest_rd_atomic;
if (attr_mask & IB_QP_MAX_QP_RD_ATOMIC)
qp->s_max_rd_atomic = attr->max_rd_atomic;
spin_unlock(&qp->s_lock);
spin_unlock_irq(&qp->r_lock);
if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT)
insert_qp(dev, qp);
if (lastwqe) {
ev.device = qp->ibqp.device;
ev.element.qp = &qp->ibqp;
ev.event = IB_EVENT_QP_LAST_WQE_REACHED;
qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
}
if (mig) {
ev.device = qp->ibqp.device;
ev.element.qp = &qp->ibqp;
ev.event = IB_EVENT_PATH_MIG;
qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
}
ret = 0;
goto bail;
inval:
spin_unlock(&qp->s_lock);
spin_unlock_irq(&qp->r_lock);
ret = -EINVAL;
bail:
return ret;
}
int qib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
int attr_mask, struct ib_qp_init_attr *init_attr)
{
struct qib_qp *qp = to_iqp(ibqp);
attr->qp_state = qp->state;
attr->cur_qp_state = attr->qp_state;
attr->path_mtu = qp->path_mtu;
attr->path_mig_state = qp->s_mig_state;
attr->qkey = qp->qkey;
attr->rq_psn = qp->r_psn & QIB_PSN_MASK;
attr->sq_psn = qp->s_next_psn & QIB_PSN_MASK;
attr->dest_qp_num = qp->remote_qpn;
attr->qp_access_flags = qp->qp_access_flags;
attr->cap.max_send_wr = qp->s_size - 1;
attr->cap.max_recv_wr = qp->ibqp.srq ? 0 : qp->r_rq.size - 1;
attr->cap.max_send_sge = qp->s_max_sge;
attr->cap.max_recv_sge = qp->r_rq.max_sge;
attr->cap.max_inline_data = 0;
attr->ah_attr = qp->remote_ah_attr;
attr->alt_ah_attr = qp->alt_ah_attr;
attr->pkey_index = qp->s_pkey_index;
attr->alt_pkey_index = qp->s_alt_pkey_index;
attr->en_sqd_async_notify = 0;
attr->sq_draining = qp->s_draining;
attr->max_rd_atomic = qp->s_max_rd_atomic;
attr->max_dest_rd_atomic = qp->r_max_rd_atomic;
attr->min_rnr_timer = qp->r_min_rnr_timer;
attr->port_num = qp->port_num;
attr->timeout = qp->timeout;
attr->retry_cnt = qp->s_retry_cnt;
attr->rnr_retry = qp->s_rnr_retry_cnt;
attr->alt_port_num = qp->alt_ah_attr.port_num;
attr->alt_timeout = qp->alt_timeout;
init_attr->event_handler = qp->ibqp.event_handler;
init_attr->qp_context = qp->ibqp.qp_context;
init_attr->send_cq = qp->ibqp.send_cq;
init_attr->recv_cq = qp->ibqp.recv_cq;
init_attr->srq = qp->ibqp.srq;
init_attr->cap = attr->cap;
if (qp->s_flags & QIB_S_SIGNAL_REQ_WR)
init_attr->sq_sig_type = IB_SIGNAL_REQ_WR;
else
init_attr->sq_sig_type = IB_SIGNAL_ALL_WR;
init_attr->qp_type = qp->ibqp.qp_type;
init_attr->port_num = qp->port_num;
return 0;
}
/**
* qib_compute_aeth - compute the AETH (syndrome + MSN)
* @qp: the queue pair to compute the AETH for
*
* Returns the AETH.
*/
__be32 qib_compute_aeth(struct qib_qp *qp)
{
u32 aeth = qp->r_msn & QIB_MSN_MASK;
if (qp->ibqp.srq) {
/*
* Shared receive queues don't generate credits.
* Set the credit field to the invalid value.
*/
aeth |= QIB_AETH_CREDIT_INVAL << QIB_AETH_CREDIT_SHIFT;
} else {
u32 min, max, x;
u32 credits;
struct qib_rwq *wq = qp->r_rq.wq;
u32 head;
u32 tail;
/* sanity check pointers before trusting them */
head = wq->head;
if (head >= qp->r_rq.size)
head = 0;
tail = wq->tail;
if (tail >= qp->r_rq.size)
tail = 0;
/*
* Compute the number of credits available (RWQEs).
* XXX Not holding the r_rq.lock here so there is a small
* chance that the pair of reads are not atomic.
*/
credits = head - tail;
if ((int)credits < 0)
credits += qp->r_rq.size;
/*
* Binary search the credit table to find the code to
* use.
*/
min = 0;
max = 31;
for (;;) {
x = (min + max) / 2;
if (credit_table[x] == credits)
break;
if (credit_table[x] > credits)
max = x;
else if (min == x)
break;
else
min = x;
}
aeth |= x << QIB_AETH_CREDIT_SHIFT;
}
return cpu_to_be32(aeth);
}
/**
* qib_create_qp - create a queue pair for a device
* @ibpd: the protection domain who's device we create the queue pair for
* @init_attr: the attributes of the queue pair
* @udata: user data for libibverbs.so
*
* Returns the queue pair on success, otherwise returns an errno.
*
* Called by the ib_create_qp() core verbs function.
*/
struct ib_qp *qib_create_qp(struct ib_pd *ibpd,
struct ib_qp_init_attr *init_attr,
struct ib_udata *udata)
{
struct qib_qp *qp;
int err;
struct qib_swqe *swq = NULL;
struct qib_ibdev *dev;
struct qib_devdata *dd;
size_t sz;
size_t sg_list_sz;
struct ib_qp *ret;
if (init_attr->cap.max_send_sge > ib_qib_max_sges ||
init_attr->cap.max_send_wr > ib_qib_max_qp_wrs) {
ret = ERR_PTR(-EINVAL);
goto bail;
}
/* Check receive queue parameters if no SRQ is specified. */
if (!init_attr->srq) {
if (init_attr->cap.max_recv_sge > ib_qib_max_sges ||
init_attr->cap.max_recv_wr > ib_qib_max_qp_wrs) {
ret = ERR_PTR(-EINVAL);
goto bail;
}
if (init_attr->cap.max_send_sge +
init_attr->cap.max_send_wr +
init_attr->cap.max_recv_sge +
init_attr->cap.max_recv_wr == 0) {
ret = ERR_PTR(-EINVAL);
goto bail;
}
}
switch (init_attr->qp_type) {
case IB_QPT_SMI:
case IB_QPT_GSI:
if (init_attr->port_num == 0 ||
init_attr->port_num > ibpd->device->phys_port_cnt) {
ret = ERR_PTR(-EINVAL);
goto bail;
}
case IB_QPT_UC:
case IB_QPT_RC:
case IB_QPT_UD:
sz = sizeof(struct qib_sge) *
init_attr->cap.max_send_sge +
sizeof(struct qib_swqe);
swq = vmalloc((init_attr->cap.max_send_wr + 1) * sz);
if (swq == NULL) {
ret = ERR_PTR(-ENOMEM);
goto bail;
}
sz = sizeof(*qp);
sg_list_sz = 0;
if (init_attr->srq) {
struct qib_srq *srq = to_isrq(init_attr->srq);
if (srq->rq.max_sge > 1)
sg_list_sz = sizeof(*qp->r_sg_list) *
(srq->rq.max_sge - 1);
} else if (init_attr->cap.max_recv_sge > 1)
sg_list_sz = sizeof(*qp->r_sg_list) *
(init_attr->cap.max_recv_sge - 1);
qp = kzalloc(sz + sg_list_sz, GFP_KERNEL);
if (!qp) {
ret = ERR_PTR(-ENOMEM);
goto bail_swq;
}
if (init_attr->srq)
sz = 0;
else {
qp->r_rq.size = init_attr->cap.max_recv_wr + 1;
qp->r_rq.max_sge = init_attr->cap.max_recv_sge;
sz = (sizeof(struct ib_sge) * qp->r_rq.max_sge) +
sizeof(struct qib_rwqe);
qp->r_rq.wq = vmalloc_user(sizeof(struct qib_rwq) +
qp->r_rq.size * sz);
if (!qp->r_rq.wq) {
ret = ERR_PTR(-ENOMEM);
goto bail_qp;
}
}
/*
* ib_create_qp() will initialize qp->ibqp
* except for qp->ibqp.qp_num.
*/
spin_lock_init(&qp->r_lock);
spin_lock_init(&qp->s_lock);
spin_lock_init(&qp->r_rq.lock);
atomic_set(&qp->refcount, 0);
init_waitqueue_head(&qp->wait);
init_waitqueue_head(&qp->wait_dma);
init_timer(&qp->s_timer);
qp->s_timer.data = (unsigned long)qp;
INIT_WORK(&qp->s_work, qib_do_send);
INIT_LIST_HEAD(&qp->iowait);
INIT_LIST_HEAD(&qp->rspwait);
qp->state = IB_QPS_RESET;
qp->s_wq = swq;
qp->s_size = init_attr->cap.max_send_wr + 1;
qp->s_max_sge = init_attr->cap.max_send_sge;
if (init_attr->sq_sig_type == IB_SIGNAL_REQ_WR)
qp->s_flags = QIB_S_SIGNAL_REQ_WR;
dev = to_idev(ibpd->device);
dd = dd_from_dev(dev);
err = alloc_qpn(dd, &dev->qpn_table, init_attr->qp_type,
init_attr->port_num);
if (err < 0) {
ret = ERR_PTR(err);
vfree(qp->r_rq.wq);
goto bail_qp;
}
qp->ibqp.qp_num = err;
qp->port_num = init_attr->port_num;
qib_reset_qp(qp, init_attr->qp_type);
break;
default:
/* Don't support raw QPs */
ret = ERR_PTR(-ENOSYS);
goto bail;
}
init_attr->cap.max_inline_data = 0;
/*
* Return the address of the RWQ as the offset to mmap.
* See qib_mmap() for details.
*/
if (udata && udata->outlen >= sizeof(__u64)) {
if (!qp->r_rq.wq) {
__u64 offset = 0;
err = ib_copy_to_udata(udata, &offset,
sizeof(offset));
if (err) {
ret = ERR_PTR(err);
goto bail_ip;
}
} else {
u32 s = sizeof(struct qib_rwq) + qp->r_rq.size * sz;
qp->ip = qib_create_mmap_info(dev, s,
ibpd->uobject->context,
qp->r_rq.wq);
if (!qp->ip) {
ret = ERR_PTR(-ENOMEM);
goto bail_ip;
}
err = ib_copy_to_udata(udata, &(qp->ip->offset),
sizeof(qp->ip->offset));
if (err) {
ret = ERR_PTR(err);
goto bail_ip;
}
}
}
spin_lock(&dev->n_qps_lock);
if (dev->n_qps_allocated == ib_qib_max_qps) {
spin_unlock(&dev->n_qps_lock);
ret = ERR_PTR(-ENOMEM);
goto bail_ip;
}
dev->n_qps_allocated++;
spin_unlock(&dev->n_qps_lock);
if (qp->ip) {
spin_lock_irq(&dev->pending_lock);
list_add(&qp->ip->pending_mmaps, &dev->pending_mmaps);
spin_unlock_irq(&dev->pending_lock);
}
ret = &qp->ibqp;
goto bail;
bail_ip:
if (qp->ip)
kref_put(&qp->ip->ref, qib_release_mmap_info);
else
vfree(qp->r_rq.wq);
free_qpn(&dev->qpn_table, qp->ibqp.qp_num);
bail_qp:
kfree(qp);
bail_swq:
vfree(swq);
bail:
return ret;
}
/**
* qib_destroy_qp - destroy a queue pair
* @ibqp: the queue pair to destroy
*
* Returns 0 on success.
*
* Note that this can be called while the QP is actively sending or
* receiving!
*/
int qib_destroy_qp(struct ib_qp *ibqp)
{
struct qib_qp *qp = to_iqp(ibqp);
struct qib_ibdev *dev = to_idev(ibqp->device);
/* Make sure HW and driver activity is stopped. */
spin_lock_irq(&qp->s_lock);
if (qp->state != IB_QPS_RESET) {
qp->state = IB_QPS_RESET;
spin_lock(&dev->pending_lock);
if (!list_empty(&qp->iowait))
list_del_init(&qp->iowait);
spin_unlock(&dev->pending_lock);
qp->s_flags &= ~(QIB_S_TIMER | QIB_S_ANY_WAIT);
spin_unlock_irq(&qp->s_lock);
cancel_work_sync(&qp->s_work);
del_timer_sync(&qp->s_timer);
wait_event(qp->wait_dma, !atomic_read(&qp->s_dma_busy));
if (qp->s_tx) {
qib_put_txreq(qp->s_tx);
qp->s_tx = NULL;
}
remove_qp(dev, qp);
wait_event(qp->wait, !atomic_read(&qp->refcount));
clear_mr_refs(qp, 1);
} else
spin_unlock_irq(&qp->s_lock);
/* all user's cleaned up, mark it available */
free_qpn(&dev->qpn_table, qp->ibqp.qp_num);
spin_lock(&dev->n_qps_lock);
dev->n_qps_allocated--;
spin_unlock(&dev->n_qps_lock);
if (qp->ip)
kref_put(&qp->ip->ref, qib_release_mmap_info);
else
vfree(qp->r_rq.wq);
vfree(qp->s_wq);
kfree(qp);
return 0;
}
/**
* qib_init_qpn_table - initialize the QP number table for a device
* @qpt: the QPN table
*/
void qib_init_qpn_table(struct qib_devdata *dd, struct qib_qpn_table *qpt)
{
spin_lock_init(&qpt->lock);
qpt->last = 1; /* start with QPN 2 */
qpt->nmaps = 1;
qpt->mask = dd->qpn_mask;
}
/**
* qib_free_qpn_table - free the QP number table for a device
* @qpt: the QPN table
*/
void qib_free_qpn_table(struct qib_qpn_table *qpt)
{
int i;
for (i = 0; i < ARRAY_SIZE(qpt->map); i++)
if (qpt->map[i].page)
free_page((unsigned long) qpt->map[i].page);
}
/**
* qib_get_credit - flush the send work queue of a QP
* @qp: the qp who's send work queue to flush
* @aeth: the Acknowledge Extended Transport Header
*
* The QP s_lock should be held.
*/
void qib_get_credit(struct qib_qp *qp, u32 aeth)
{
u32 credit = (aeth >> QIB_AETH_CREDIT_SHIFT) & QIB_AETH_CREDIT_MASK;
/*
* If the credit is invalid, we can send
* as many packets as we like. Otherwise, we have to
* honor the credit field.
*/
if (credit == QIB_AETH_CREDIT_INVAL) {
if (!(qp->s_flags & QIB_S_UNLIMITED_CREDIT)) {
qp->s_flags |= QIB_S_UNLIMITED_CREDIT;
if (qp->s_flags & QIB_S_WAIT_SSN_CREDIT) {
qp->s_flags &= ~QIB_S_WAIT_SSN_CREDIT;
qib_schedule_send(qp);
}
}
} else if (!(qp->s_flags & QIB_S_UNLIMITED_CREDIT)) {
/* Compute new LSN (i.e., MSN + credit) */
credit = (aeth + credit_table[credit]) & QIB_MSN_MASK;
if (qib_cmp24(credit, qp->s_lsn) > 0) {
qp->s_lsn = credit;
if (qp->s_flags & QIB_S_WAIT_SSN_CREDIT) {
qp->s_flags &= ~QIB_S_WAIT_SSN_CREDIT;
qib_schedule_send(qp);
}
}
}
}
|