summaryrefslogtreecommitdiff
path: root/drivers/infiniband/sw/siw/siw_verbs.c
blob: b18a677832e105ded578bd74d29694c45e55d031 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
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
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
// SPDX-License-Identifier: GPL-2.0 or BSD-3-Clause

/* Authors: Bernard Metzler <bmt@zurich.ibm.com> */
/* Copyright (c) 2008-2019, IBM Corporation */

#include <linux/errno.h>
#include <linux/types.h>
#include <linux/uaccess.h>
#include <linux/vmalloc.h>
#include <linux/xarray.h>

#include <rdma/iw_cm.h>
#include <rdma/ib_verbs.h>
#include <rdma/ib_user_verbs.h>
#include <rdma/uverbs_ioctl.h>

#include "siw.h"
#include "siw_verbs.h"
#include "siw_mem.h"

static int ib_qp_state_to_siw_qp_state[IB_QPS_ERR + 1] = {
	[IB_QPS_RESET] = SIW_QP_STATE_IDLE,
	[IB_QPS_INIT] = SIW_QP_STATE_IDLE,
	[IB_QPS_RTR] = SIW_QP_STATE_RTR,
	[IB_QPS_RTS] = SIW_QP_STATE_RTS,
	[IB_QPS_SQD] = SIW_QP_STATE_CLOSING,
	[IB_QPS_SQE] = SIW_QP_STATE_TERMINATE,
	[IB_QPS_ERR] = SIW_QP_STATE_ERROR
};

static char ib_qp_state_to_string[IB_QPS_ERR + 1][sizeof("RESET")] = {
	[IB_QPS_RESET] = "RESET", [IB_QPS_INIT] = "INIT", [IB_QPS_RTR] = "RTR",
	[IB_QPS_RTS] = "RTS",     [IB_QPS_SQD] = "SQD",   [IB_QPS_SQE] = "SQE",
	[IB_QPS_ERR] = "ERR"
};

static u32 siw_create_uobj(struct siw_ucontext *uctx, void *vaddr, u32 size)
{
	struct siw_uobj *uobj;
	struct xa_limit limit = XA_LIMIT(0, SIW_UOBJ_MAX_KEY);
	u32 key;

	uobj = kzalloc(sizeof(*uobj), GFP_KERNEL);
	if (!uobj)
		return SIW_INVAL_UOBJ_KEY;

	if (xa_alloc_cyclic(&uctx->xa, &key, uobj, limit, &uctx->uobj_nextkey,
			    GFP_KERNEL) < 0) {
		kfree(uobj);
		return SIW_INVAL_UOBJ_KEY;
	}
	uobj->size = PAGE_ALIGN(size);
	uobj->addr = vaddr;

	return key;
}

static struct siw_uobj *siw_get_uobj(struct siw_ucontext *uctx,
				     unsigned long off, u32 size)
{
	struct siw_uobj *uobj = xa_load(&uctx->xa, off);

	if (uobj && uobj->size == size)
		return uobj;

	return NULL;
}

int siw_mmap(struct ib_ucontext *ctx, struct vm_area_struct *vma)
{
	struct siw_ucontext *uctx = to_siw_ctx(ctx);
	struct siw_uobj *uobj;
	unsigned long off = vma->vm_pgoff;
	int size = vma->vm_end - vma->vm_start;
	int rv = -EINVAL;

	/*
	 * Must be page aligned
	 */
	if (vma->vm_start & (PAGE_SIZE - 1)) {
		pr_warn("siw: mmap not page aligned\n");
		goto out;
	}
	uobj = siw_get_uobj(uctx, off, size);
	if (!uobj) {
		siw_dbg(&uctx->sdev->base_dev, "mmap lookup failed: %lu, %u\n",
			off, size);
		goto out;
	}
	rv = remap_vmalloc_range(vma, uobj->addr, 0);
	if (rv)
		pr_warn("remap_vmalloc_range failed: %lu, %u\n", off, size);
out:
	return rv;
}

int siw_alloc_ucontext(struct ib_ucontext *base_ctx, struct ib_udata *udata)
{
	struct siw_device *sdev = to_siw_dev(base_ctx->device);
	struct siw_ucontext *ctx = to_siw_ctx(base_ctx);
	struct siw_uresp_alloc_ctx uresp = {};
	int rv;

	if (atomic_inc_return(&sdev->num_ctx) > SIW_MAX_CONTEXT) {
		rv = -ENOMEM;
		goto err_out;
	}
	xa_init_flags(&ctx->xa, XA_FLAGS_ALLOC);
	ctx->uobj_nextkey = 0;
	ctx->sdev = sdev;

	uresp.dev_id = sdev->vendor_part_id;

	if (udata->outlen < sizeof(uresp)) {
		rv = -EINVAL;
		goto err_out;
	}
	rv = ib_copy_to_udata(udata, &uresp, sizeof(uresp));
	if (rv)
		goto err_out;

	siw_dbg(base_ctx->device, "success. now %d context(s)\n",
		atomic_read(&sdev->num_ctx));

	return 0;

err_out:
	atomic_dec(&sdev->num_ctx);
	siw_dbg(base_ctx->device, "failure %d. now %d context(s)\n", rv,
		atomic_read(&sdev->num_ctx));

	return rv;
}

void siw_dealloc_ucontext(struct ib_ucontext *base_ctx)
{
	struct siw_ucontext *uctx = to_siw_ctx(base_ctx);
	void *entry;
	unsigned long index;

	/*
	 * Make sure all user mmap objects are gone. Since QP, CQ
	 * and SRQ destroy routines destroy related objects, nothing
	 * should be found here.
	 */
	xa_for_each(&uctx->xa, index, entry) {
		kfree(xa_erase(&uctx->xa, index));
		pr_warn("siw: dropping orphaned uobj at %lu\n", index);
	}
	xa_destroy(&uctx->xa);
	atomic_dec(&uctx->sdev->num_ctx);
}

int siw_query_device(struct ib_device *base_dev, struct ib_device_attr *attr,
		     struct ib_udata *udata)
{
	struct siw_device *sdev = to_siw_dev(base_dev);

	if (udata->inlen || udata->outlen)
		return -EINVAL;

	memset(attr, 0, sizeof(*attr));

	/* Revisit atomic caps if RFC 7306 gets supported */
	attr->atomic_cap = 0;
	attr->device_cap_flags =
		IB_DEVICE_MEM_MGT_EXTENSIONS | IB_DEVICE_ALLOW_USER_UNREG;
	attr->max_cq = sdev->attrs.max_cq;
	attr->max_cqe = sdev->attrs.max_cqe;
	attr->max_fast_reg_page_list_len = SIW_MAX_SGE_PBL;
	attr->max_fmr = sdev->attrs.max_fmr;
	attr->max_mr = sdev->attrs.max_mr;
	attr->max_mw = sdev->attrs.max_mw;
	attr->max_mr_size = ~0ull;
	attr->max_pd = sdev->attrs.max_pd;
	attr->max_qp = sdev->attrs.max_qp;
	attr->max_qp_init_rd_atom = sdev->attrs.max_ird;
	attr->max_qp_rd_atom = sdev->attrs.max_ord;
	attr->max_qp_wr = sdev->attrs.max_qp_wr;
	attr->max_recv_sge = sdev->attrs.max_sge;
	attr->max_res_rd_atom = sdev->attrs.max_qp * sdev->attrs.max_ird;
	attr->max_send_sge = sdev->attrs.max_sge;
	attr->max_sge_rd = sdev->attrs.max_sge_rd;
	attr->max_srq = sdev->attrs.max_srq;
	attr->max_srq_sge = sdev->attrs.max_srq_sge;
	attr->max_srq_wr = sdev->attrs.max_srq_wr;
	attr->page_size_cap = PAGE_SIZE;
	attr->vendor_id = SIW_VENDOR_ID;
	attr->vendor_part_id = sdev->vendor_part_id;

	memcpy(&attr->sys_image_guid, sdev->netdev->dev_addr, 6);

	return 0;
}

int siw_query_port(struct ib_device *base_dev, u8 port,
		   struct ib_port_attr *attr)
{
	struct siw_device *sdev = to_siw_dev(base_dev);

	memset(attr, 0, sizeof(*attr));

	attr->active_mtu = attr->max_mtu;
	attr->active_speed = 2;
	attr->active_width = 2;
	attr->gid_tbl_len = 1;
	attr->max_msg_sz = -1;
	attr->max_mtu = ib_mtu_int_to_enum(sdev->netdev->mtu);
	attr->phys_state = sdev->state == IB_PORT_ACTIVE ?
		IB_PORT_PHYS_STATE_LINK_UP : IB_PORT_PHYS_STATE_DISABLED;
	attr->pkey_tbl_len = 1;
	attr->port_cap_flags = IB_PORT_CM_SUP | IB_PORT_DEVICE_MGMT_SUP;
	attr->state = sdev->state;
	/*
	 * All zero
	 *
	 * attr->lid = 0;
	 * attr->bad_pkey_cntr = 0;
	 * attr->qkey_viol_cntr = 0;
	 * attr->sm_lid = 0;
	 * attr->lmc = 0;
	 * attr->max_vl_num = 0;
	 * attr->sm_sl = 0;
	 * attr->subnet_timeout = 0;
	 * attr->init_type_repy = 0;
	 */
	return 0;
}

int siw_get_port_immutable(struct ib_device *base_dev, u8 port,
			   struct ib_port_immutable *port_immutable)
{
	struct ib_port_attr attr;
	int rv = siw_query_port(base_dev, port, &attr);

	if (rv)
		return rv;

	port_immutable->pkey_tbl_len = attr.pkey_tbl_len;
	port_immutable->gid_tbl_len = attr.gid_tbl_len;
	port_immutable->core_cap_flags = RDMA_CORE_PORT_IWARP;

	return 0;
}

int siw_query_pkey(struct ib_device *base_dev, u8 port, u16 idx, u16 *pkey)
{
	/* Report the default pkey */
	*pkey = 0xffff;
	return 0;
}

int siw_query_gid(struct ib_device *base_dev, u8 port, int idx,
		  union ib_gid *gid)
{
	struct siw_device *sdev = to_siw_dev(base_dev);

	/* subnet_prefix == interface_id == 0; */
	memset(gid, 0, sizeof(*gid));
	memcpy(&gid->raw[0], sdev->netdev->dev_addr, 6);

	return 0;
}

int siw_alloc_pd(struct ib_pd *pd, struct ib_udata *udata)
{
	struct siw_device *sdev = to_siw_dev(pd->device);

	if (atomic_inc_return(&sdev->num_pd) > SIW_MAX_PD) {
		atomic_dec(&sdev->num_pd);
		return -ENOMEM;
	}
	siw_dbg_pd(pd, "now %d PD's(s)\n", atomic_read(&sdev->num_pd));

	return 0;
}

void siw_dealloc_pd(struct ib_pd *pd, struct ib_udata *udata)
{
	struct siw_device *sdev = to_siw_dev(pd->device);

	siw_dbg_pd(pd, "free PD\n");
	atomic_dec(&sdev->num_pd);
}

void siw_qp_get_ref(struct ib_qp *base_qp)
{
	siw_qp_get(to_siw_qp(base_qp));
}

void siw_qp_put_ref(struct ib_qp *base_qp)
{
	siw_qp_put(to_siw_qp(base_qp));
}

/*
 * siw_create_qp()
 *
 * Create QP of requested size on given device.
 *
 * @pd:		Protection Domain
 * @attrs:	Initial QP attributes.
 * @udata:	used to provide QP ID, SQ and RQ size back to user.
 */

struct ib_qp *siw_create_qp(struct ib_pd *pd,
			    struct ib_qp_init_attr *attrs,
			    struct ib_udata *udata)
{
	struct siw_qp *qp = NULL;
	struct siw_base_qp *siw_base_qp = NULL;
	struct ib_device *base_dev = pd->device;
	struct siw_device *sdev = to_siw_dev(base_dev);
	struct siw_ucontext *uctx =
		rdma_udata_to_drv_context(udata, struct siw_ucontext,
					  base_ucontext);
	struct siw_cq *scq = NULL, *rcq = NULL;
	unsigned long flags;
	int num_sqe, num_rqe, rv = 0;

	siw_dbg(base_dev, "create new QP\n");

	if (atomic_inc_return(&sdev->num_qp) > SIW_MAX_QP) {
		siw_dbg(base_dev, "too many QP's\n");
		rv = -ENOMEM;
		goto err_out;
	}
	if (attrs->qp_type != IB_QPT_RC) {
		siw_dbg(base_dev, "only RC QP's supported\n");
		rv = -EINVAL;
		goto err_out;
	}
	if ((attrs->cap.max_send_wr > SIW_MAX_QP_WR) ||
	    (attrs->cap.max_recv_wr > SIW_MAX_QP_WR) ||
	    (attrs->cap.max_send_sge > SIW_MAX_SGE) ||
	    (attrs->cap.max_recv_sge > SIW_MAX_SGE)) {
		siw_dbg(base_dev, "QP size error\n");
		rv = -EINVAL;
		goto err_out;
	}
	if (attrs->cap.max_inline_data > SIW_MAX_INLINE) {
		siw_dbg(base_dev, "max inline send: %d > %d\n",
			attrs->cap.max_inline_data, (int)SIW_MAX_INLINE);
		rv = -EINVAL;
		goto err_out;
	}
	/*
	 * NOTE: we allow for zero element SQ and RQ WQE's SGL's
	 * but not for a QP unable to hold any WQE (SQ + RQ)
	 */
	if (attrs->cap.max_send_wr + attrs->cap.max_recv_wr == 0) {
		siw_dbg(base_dev, "QP must have send or receive queue\n");
		rv = -EINVAL;
		goto err_out;
	}
	scq = to_siw_cq(attrs->send_cq);
	rcq = to_siw_cq(attrs->recv_cq);

	if (!scq || (!rcq && !attrs->srq)) {
		siw_dbg(base_dev, "send CQ or receive CQ invalid\n");
		rv = -EINVAL;
		goto err_out;
	}
	siw_base_qp = kzalloc(sizeof(*siw_base_qp), GFP_KERNEL);
	if (!siw_base_qp) {
		rv = -ENOMEM;
		goto err_out;
	}
	qp = kzalloc(sizeof(*qp), GFP_KERNEL);
	if (!qp) {
		rv = -ENOMEM;
		goto err_out;
	}
	siw_base_qp->qp = qp;
	qp->ib_qp = &siw_base_qp->base_qp;

	init_rwsem(&qp->state_lock);
	spin_lock_init(&qp->sq_lock);
	spin_lock_init(&qp->rq_lock);
	spin_lock_init(&qp->orq_lock);

	qp->kernel_verbs = !udata;
	qp->xa_sq_index = SIW_INVAL_UOBJ_KEY;
	qp->xa_rq_index = SIW_INVAL_UOBJ_KEY;

	rv = siw_qp_add(sdev, qp);
	if (rv)
		goto err_out;

	/* All queue indices are derived from modulo operations
	 * on a free running 'get' (consumer) and 'put' (producer)
	 * unsigned counter. Having queue sizes at power of two
	 * avoids handling counter wrap around.
	 */
	num_sqe = roundup_pow_of_two(attrs->cap.max_send_wr);
	num_rqe = roundup_pow_of_two(attrs->cap.max_recv_wr);

	if (qp->kernel_verbs)
		qp->sendq = vzalloc(num_sqe * sizeof(struct siw_sqe));
	else
		qp->sendq = vmalloc_user(num_sqe * sizeof(struct siw_sqe));

	if (qp->sendq == NULL) {
		siw_dbg(base_dev, "SQ size %d alloc failed\n", num_sqe);
		rv = -ENOMEM;
		goto err_out_xa;
	}
	if (attrs->sq_sig_type != IB_SIGNAL_REQ_WR) {
		if (attrs->sq_sig_type == IB_SIGNAL_ALL_WR)
			qp->attrs.flags |= SIW_SIGNAL_ALL_WR;
		else {
			rv = -EINVAL;
			goto err_out_xa;
		}
	}
	qp->pd = pd;
	qp->scq = scq;
	qp->rcq = rcq;

	if (attrs->srq) {
		/*
		 * SRQ support.
		 * Verbs 6.3.7: ignore RQ size, if SRQ present
		 * Verbs 6.3.5: do not check PD of SRQ against PD of QP
		 */
		qp->srq = to_siw_srq(attrs->srq);
		qp->attrs.rq_size = 0;
		siw_dbg(base_dev, "QP [%u]: SRQ attached\n", qp->qp_num);
	} else if (num_rqe) {
		if (qp->kernel_verbs)
			qp->recvq = vzalloc(num_rqe * sizeof(struct siw_rqe));
		else
			qp->recvq =
				vmalloc_user(num_rqe * sizeof(struct siw_rqe));

		if (qp->recvq == NULL) {
			siw_dbg(base_dev, "RQ size %d alloc failed\n", num_rqe);
			rv = -ENOMEM;
			goto err_out_xa;
		}
		qp->attrs.rq_size = num_rqe;
	}
	qp->attrs.sq_size = num_sqe;
	qp->attrs.sq_max_sges = attrs->cap.max_send_sge;
	qp->attrs.rq_max_sges = attrs->cap.max_recv_sge;

	/* Make those two tunables fixed for now. */
	qp->tx_ctx.gso_seg_limit = 1;
	qp->tx_ctx.zcopy_tx = zcopy_tx;

	qp->attrs.state = SIW_QP_STATE_IDLE;

	if (udata) {
		struct siw_uresp_create_qp uresp = {};

		uresp.num_sqe = num_sqe;
		uresp.num_rqe = num_rqe;
		uresp.qp_id = qp_id(qp);

		if (qp->sendq) {
			qp->xa_sq_index =
				siw_create_uobj(uctx, qp->sendq,
					num_sqe * sizeof(struct siw_sqe));
		}
		if (qp->recvq) {
			qp->xa_rq_index =
				 siw_create_uobj(uctx, qp->recvq,
					num_rqe * sizeof(struct siw_rqe));
		}
		if (qp->xa_sq_index == SIW_INVAL_UOBJ_KEY ||
		    qp->xa_rq_index == SIW_INVAL_UOBJ_KEY) {
			rv = -ENOMEM;
			goto err_out_xa;
		}
		uresp.sq_key = qp->xa_sq_index << PAGE_SHIFT;
		uresp.rq_key = qp->xa_rq_index << PAGE_SHIFT;

		if (udata->outlen < sizeof(uresp)) {
			rv = -EINVAL;
			goto err_out_xa;
		}
		rv = ib_copy_to_udata(udata, &uresp, sizeof(uresp));
		if (rv)
			goto err_out_xa;
	}
	qp->tx_cpu = siw_get_tx_cpu(sdev);
	if (qp->tx_cpu < 0) {
		rv = -EINVAL;
		goto err_out_xa;
	}
	INIT_LIST_HEAD(&qp->devq);
	spin_lock_irqsave(&sdev->lock, flags);
	list_add_tail(&qp->devq, &sdev->qp_list);
	spin_unlock_irqrestore(&sdev->lock, flags);

	return qp->ib_qp;

err_out_xa:
	xa_erase(&sdev->qp_xa, qp_id(qp));
err_out:
	kfree(siw_base_qp);

	if (qp) {
		if (qp->xa_sq_index != SIW_INVAL_UOBJ_KEY)
			kfree(xa_erase(&uctx->xa, qp->xa_sq_index));
		if (qp->xa_rq_index != SIW_INVAL_UOBJ_KEY)
			kfree(xa_erase(&uctx->xa, qp->xa_rq_index));

		vfree(qp->sendq);
		vfree(qp->recvq);
		kfree(qp);
	}
	atomic_dec(&sdev->num_qp);

	return ERR_PTR(rv);
}

/*
 * Minimum siw_query_qp() verb interface.
 *
 * @qp_attr_mask is not used but all available information is provided
 */
int siw_query_qp(struct ib_qp *base_qp, struct ib_qp_attr *qp_attr,
		 int qp_attr_mask, struct ib_qp_init_attr *qp_init_attr)
{
	struct siw_qp *qp;
	struct siw_device *sdev;

	if (base_qp && qp_attr && qp_init_attr) {
		qp = to_siw_qp(base_qp);
		sdev = to_siw_dev(base_qp->device);
	} else {
		return -EINVAL;
	}
	qp_attr->cap.max_inline_data = SIW_MAX_INLINE;
	qp_attr->cap.max_send_wr = qp->attrs.sq_size;
	qp_attr->cap.max_send_sge = qp->attrs.sq_max_sges;
	qp_attr->cap.max_recv_wr = qp->attrs.rq_size;
	qp_attr->cap.max_recv_sge = qp->attrs.rq_max_sges;
	qp_attr->path_mtu = ib_mtu_int_to_enum(sdev->netdev->mtu);
	qp_attr->max_rd_atomic = qp->attrs.irq_size;
	qp_attr->max_dest_rd_atomic = qp->attrs.orq_size;

	qp_attr->qp_access_flags = IB_ACCESS_LOCAL_WRITE |
				   IB_ACCESS_REMOTE_WRITE |
				   IB_ACCESS_REMOTE_READ;

	qp_init_attr->qp_type = base_qp->qp_type;
	qp_init_attr->send_cq = base_qp->send_cq;
	qp_init_attr->recv_cq = base_qp->recv_cq;
	qp_init_attr->srq = base_qp->srq;

	qp_init_attr->cap = qp_attr->cap;

	return 0;
}

int siw_verbs_modify_qp(struct ib_qp *base_qp, struct ib_qp_attr *attr,
			int attr_mask, struct ib_udata *udata)
{
	struct siw_qp_attrs new_attrs;
	enum siw_qp_attr_mask siw_attr_mask = 0;
	struct siw_qp *qp = to_siw_qp(base_qp);
	int rv = 0;

	if (!attr_mask)
		return 0;

	memset(&new_attrs, 0, sizeof(new_attrs));

	if (attr_mask & IB_QP_ACCESS_FLAGS) {
		siw_attr_mask = SIW_QP_ATTR_ACCESS_FLAGS;

		if (attr->qp_access_flags & IB_ACCESS_REMOTE_READ)
			new_attrs.flags |= SIW_RDMA_READ_ENABLED;
		if (attr->qp_access_flags & IB_ACCESS_REMOTE_WRITE)
			new_attrs.flags |= SIW_RDMA_WRITE_ENABLED;
		if (attr->qp_access_flags & IB_ACCESS_MW_BIND)
			new_attrs.flags |= SIW_RDMA_BIND_ENABLED;
	}
	if (attr_mask & IB_QP_STATE) {
		siw_dbg_qp(qp, "desired IB QP state: %s\n",
			   ib_qp_state_to_string[attr->qp_state]);

		new_attrs.state = ib_qp_state_to_siw_qp_state[attr->qp_state];

		if (new_attrs.state > SIW_QP_STATE_RTS)
			qp->tx_ctx.tx_suspend = 1;

		siw_attr_mask |= SIW_QP_ATTR_STATE;
	}
	if (!siw_attr_mask)
		goto out;

	down_write(&qp->state_lock);

	rv = siw_qp_modify(qp, &new_attrs, siw_attr_mask);

	up_write(&qp->state_lock);
out:
	return rv;
}

int siw_destroy_qp(struct ib_qp *base_qp, struct ib_udata *udata)
{
	struct siw_qp *qp = to_siw_qp(base_qp);
	struct siw_ucontext *uctx =
		rdma_udata_to_drv_context(udata, struct siw_ucontext,
					  base_ucontext);
	struct siw_qp_attrs qp_attrs;

	siw_dbg_qp(qp, "state %d\n", qp->attrs.state);

	/*
	 * Mark QP as in process of destruction to prevent from
	 * any async callbacks to RDMA core
	 */
	qp->attrs.flags |= SIW_QP_IN_DESTROY;
	qp->rx_stream.rx_suspend = 1;

	if (uctx && qp->xa_sq_index != SIW_INVAL_UOBJ_KEY)
		kfree(xa_erase(&uctx->xa, qp->xa_sq_index));
	if (uctx && qp->xa_rq_index != SIW_INVAL_UOBJ_KEY)
		kfree(xa_erase(&uctx->xa, qp->xa_rq_index));

	down_write(&qp->state_lock);

	qp_attrs.state = SIW_QP_STATE_ERROR;
	siw_qp_modify(qp, &qp_attrs, SIW_QP_ATTR_STATE);

	if (qp->cep) {
		siw_cep_put(qp->cep);
		qp->cep = NULL;
	}
	up_write(&qp->state_lock);

	kfree(qp->tx_ctx.mpa_crc_hd);
	kfree(qp->rx_stream.mpa_crc_hd);

	qp->scq = qp->rcq = NULL;

	siw_qp_put(qp);

	return 0;
}

/*
 * siw_copy_inline_sgl()
 *
 * Prepare sgl of inlined data for sending. For userland callers
 * function checks if given buffer addresses and len's are within
 * process context bounds.
 * Data from all provided sge's are copied together into the wqe,
 * referenced by a single sge.
 */
static int siw_copy_inline_sgl(const struct ib_send_wr *core_wr,
			       struct siw_sqe *sqe)
{
	struct ib_sge *core_sge = core_wr->sg_list;
	void *kbuf = &sqe->sge[1];
	int num_sge = core_wr->num_sge, bytes = 0;

	sqe->sge[0].laddr = (uintptr_t)kbuf;
	sqe->sge[0].lkey = 0;

	while (num_sge--) {
		if (!core_sge->length) {
			core_sge++;
			continue;
		}
		bytes += core_sge->length;
		if (bytes > SIW_MAX_INLINE) {
			bytes = -EINVAL;
			break;
		}
		memcpy(kbuf, (void *)(uintptr_t)core_sge->addr,
		       core_sge->length);

		kbuf += core_sge->length;
		core_sge++;
	}
	sqe->sge[0].length = bytes > 0 ? bytes : 0;
	sqe->num_sge = bytes > 0 ? 1 : 0;

	return bytes;
}

/*
 * siw_post_send()
 *
 * Post a list of S-WR's to a SQ.
 *
 * @base_qp:	Base QP contained in siw QP
 * @wr:		Null terminated list of user WR's
 * @bad_wr:	Points to failing WR in case of synchronous failure.
 */
int siw_post_send(struct ib_qp *base_qp, const struct ib_send_wr *wr,
		  const struct ib_send_wr **bad_wr)
{
	struct siw_qp *qp = to_siw_qp(base_qp);
	struct siw_wqe *wqe = tx_wqe(qp);

	unsigned long flags;
	int rv = 0;

	/*
	 * Try to acquire QP state lock. Must be non-blocking
	 * to accommodate kernel clients needs.
	 */
	if (!down_read_trylock(&qp->state_lock)) {
		*bad_wr = wr;
		siw_dbg_qp(qp, "QP locked, state %d\n", qp->attrs.state);
		return -ENOTCONN;
	}
	if (unlikely(qp->attrs.state != SIW_QP_STATE_RTS)) {
		up_read(&qp->state_lock);
		*bad_wr = wr;
		siw_dbg_qp(qp, "QP out of state %d\n", qp->attrs.state);
		return -ENOTCONN;
	}
	if (wr && !qp->kernel_verbs) {
		siw_dbg_qp(qp, "wr must be empty for user mapped sq\n");
		up_read(&qp->state_lock);
		*bad_wr = wr;
		return -EINVAL;
	}
	spin_lock_irqsave(&qp->sq_lock, flags);

	while (wr) {
		u32 idx = qp->sq_put % qp->attrs.sq_size;
		struct siw_sqe *sqe = &qp->sendq[idx];

		if (sqe->flags) {
			siw_dbg_qp(qp, "sq full\n");
			rv = -ENOMEM;
			break;
		}
		if (wr->num_sge > qp->attrs.sq_max_sges) {
			siw_dbg_qp(qp, "too many sge's: %d\n", wr->num_sge);
			rv = -EINVAL;
			break;
		}
		sqe->id = wr->wr_id;

		if ((wr->send_flags & IB_SEND_SIGNALED) ||
		    (qp->attrs.flags & SIW_SIGNAL_ALL_WR))
			sqe->flags |= SIW_WQE_SIGNALLED;

		if (wr->send_flags & IB_SEND_FENCE)
			sqe->flags |= SIW_WQE_READ_FENCE;

		switch (wr->opcode) {
		case IB_WR_SEND:
		case IB_WR_SEND_WITH_INV:
			if (wr->send_flags & IB_SEND_SOLICITED)
				sqe->flags |= SIW_WQE_SOLICITED;

			if (!(wr->send_flags & IB_SEND_INLINE)) {
				siw_copy_sgl(wr->sg_list, sqe->sge,
					     wr->num_sge);
				sqe->num_sge = wr->num_sge;
			} else {
				rv = siw_copy_inline_sgl(wr, sqe);
				if (rv <= 0) {
					rv = -EINVAL;
					break;
				}
				sqe->flags |= SIW_WQE_INLINE;
				sqe->num_sge = 1;
			}
			if (wr->opcode == IB_WR_SEND)
				sqe->opcode = SIW_OP_SEND;
			else {
				sqe->opcode = SIW_OP_SEND_REMOTE_INV;
				sqe->rkey = wr->ex.invalidate_rkey;
			}
			break;

		case IB_WR_RDMA_READ_WITH_INV:
		case IB_WR_RDMA_READ:
			/*
			 * iWarp restricts RREAD sink to SGL containing
			 * 1 SGE only. we could relax to SGL with multiple
			 * elements referring the SAME ltag or even sending
			 * a private per-rreq tag referring to a checked
			 * local sgl with MULTIPLE ltag's.
			 */
			if (unlikely(wr->num_sge != 1)) {
				rv = -EINVAL;
				break;
			}
			siw_copy_sgl(wr->sg_list, &sqe->sge[0], 1);
			/*
			 * NOTE: zero length RREAD is allowed!
			 */
			sqe->raddr = rdma_wr(wr)->remote_addr;
			sqe->rkey = rdma_wr(wr)->rkey;
			sqe->num_sge = 1;

			if (wr->opcode == IB_WR_RDMA_READ)
				sqe->opcode = SIW_OP_READ;
			else
				sqe->opcode = SIW_OP_READ_LOCAL_INV;
			break;

		case IB_WR_RDMA_WRITE:
			if (!(wr->send_flags & IB_SEND_INLINE)) {
				siw_copy_sgl(wr->sg_list, &sqe->sge[0],
					     wr->num_sge);
				sqe->num_sge = wr->num_sge;
			} else {
				rv = siw_copy_inline_sgl(wr, sqe);
				if (unlikely(rv < 0)) {
					rv = -EINVAL;
					break;
				}
				sqe->flags |= SIW_WQE_INLINE;
				sqe->num_sge = 1;
			}
			sqe->raddr = rdma_wr(wr)->remote_addr;
			sqe->rkey = rdma_wr(wr)->rkey;
			sqe->opcode = SIW_OP_WRITE;
			break;

		case IB_WR_REG_MR:
			sqe->base_mr = (uintptr_t)reg_wr(wr)->mr;
			sqe->rkey = reg_wr(wr)->key;
			sqe->access = reg_wr(wr)->access & IWARP_ACCESS_MASK;
			sqe->opcode = SIW_OP_REG_MR;
			break;

		case IB_WR_LOCAL_INV:
			sqe->rkey = wr->ex.invalidate_rkey;
			sqe->opcode = SIW_OP_INVAL_STAG;
			break;

		default:
			siw_dbg_qp(qp, "ib wr type %d unsupported\n",
				   wr->opcode);
			rv = -EINVAL;
			break;
		}
		siw_dbg_qp(qp, "opcode %d, flags 0x%x, wr_id 0x%pK\n",
			   sqe->opcode, sqe->flags,
			   (void *)(uintptr_t)sqe->id);

		if (unlikely(rv < 0))
			break;

		/* make SQE only valid after completely written */
		smp_wmb();
		sqe->flags |= SIW_WQE_VALID;

		qp->sq_put++;
		wr = wr->next;
	}

	/*
	 * Send directly if SQ processing is not in progress.
	 * Eventual immediate errors (rv < 0) do not affect the involved
	 * RI resources (Verbs, 8.3.1) and thus do not prevent from SQ
	 * processing, if new work is already pending. But rv must be passed
	 * to caller.
	 */
	if (wqe->wr_status != SIW_WR_IDLE) {
		spin_unlock_irqrestore(&qp->sq_lock, flags);
		goto skip_direct_sending;
	}
	rv = siw_activate_tx(qp);
	spin_unlock_irqrestore(&qp->sq_lock, flags);

	if (rv <= 0)
		goto skip_direct_sending;

	if (qp->kernel_verbs) {
		rv = siw_sq_start(qp);
	} else {
		qp->tx_ctx.in_syscall = 1;

		if (siw_qp_sq_process(qp) != 0 && !(qp->tx_ctx.tx_suspend))
			siw_qp_cm_drop(qp, 0);

		qp->tx_ctx.in_syscall = 0;
	}
skip_direct_sending:

	up_read(&qp->state_lock);

	if (rv >= 0)
		return 0;
	/*
	 * Immediate error
	 */
	siw_dbg_qp(qp, "error %d\n", rv);

	*bad_wr = wr;
	return rv;
}

/*
 * siw_post_receive()
 *
 * Post a list of R-WR's to a RQ.
 *
 * @base_qp:	Base QP contained in siw QP
 * @wr:		Null terminated list of user WR's
 * @bad_wr:	Points to failing WR in case of synchronous failure.
 */
int siw_post_receive(struct ib_qp *base_qp, const struct ib_recv_wr *wr,
		     const struct ib_recv_wr **bad_wr)
{
	struct siw_qp *qp = to_siw_qp(base_qp);
	unsigned long flags;
	int rv = 0;

	if (qp->srq) {
		*bad_wr = wr;
		return -EOPNOTSUPP; /* what else from errno.h? */
	}
	/*
	 * Try to acquire QP state lock. Must be non-blocking
	 * to accommodate kernel clients needs.
	 */
	if (!down_read_trylock(&qp->state_lock)) {
		*bad_wr = wr;
		return -ENOTCONN;
	}
	if (!qp->kernel_verbs) {
		siw_dbg_qp(qp, "no kernel post_recv for user mapped sq\n");
		up_read(&qp->state_lock);
		*bad_wr = wr;
		return -EINVAL;
	}
	if (qp->attrs.state > SIW_QP_STATE_RTS) {
		up_read(&qp->state_lock);
		*bad_wr = wr;
		return -EINVAL;
	}
	/*
	 * Serialize potentially multiple producers.
	 * Not needed for single threaded consumer side.
	 */
	spin_lock_irqsave(&qp->rq_lock, flags);

	while (wr) {
		u32 idx = qp->rq_put % qp->attrs.rq_size;
		struct siw_rqe *rqe = &qp->recvq[idx];

		if (rqe->flags) {
			siw_dbg_qp(qp, "RQ full\n");
			rv = -ENOMEM;
			break;
		}
		if (wr->num_sge > qp->attrs.rq_max_sges) {
			siw_dbg_qp(qp, "too many sge's: %d\n", wr->num_sge);
			rv = -EINVAL;
			break;
		}
		rqe->id = wr->wr_id;
		rqe->num_sge = wr->num_sge;
		siw_copy_sgl(wr->sg_list, rqe->sge, wr->num_sge);

		/* make sure RQE is completely written before valid */
		smp_wmb();

		rqe->flags = SIW_WQE_VALID;

		qp->rq_put++;
		wr = wr->next;
	}
	spin_unlock_irqrestore(&qp->rq_lock, flags);

	up_read(&qp->state_lock);

	if (rv < 0) {
		siw_dbg_qp(qp, "error %d\n", rv);
		*bad_wr = wr;
	}
	return rv > 0 ? 0 : rv;
}

void siw_destroy_cq(struct ib_cq *base_cq, struct ib_udata *udata)
{
	struct siw_cq *cq = to_siw_cq(base_cq);
	struct siw_device *sdev = to_siw_dev(base_cq->device);
	struct siw_ucontext *ctx =
		rdma_udata_to_drv_context(udata, struct siw_ucontext,
					  base_ucontext);

	siw_dbg_cq(cq, "free CQ resources\n");

	siw_cq_flush(cq);

	if (ctx && cq->xa_cq_index != SIW_INVAL_UOBJ_KEY)
		kfree(xa_erase(&ctx->xa, cq->xa_cq_index));

	atomic_dec(&sdev->num_cq);

	vfree(cq->queue);
}

/*
 * siw_create_cq()
 *
 * Populate CQ of requested size
 *
 * @base_cq: CQ as allocated by RDMA midlayer
 * @attr: Initial CQ attributes
 * @udata: relates to user context
 */

int siw_create_cq(struct ib_cq *base_cq, const struct ib_cq_init_attr *attr,
		  struct ib_udata *udata)
{
	struct siw_device *sdev = to_siw_dev(base_cq->device);
	struct siw_cq *cq = to_siw_cq(base_cq);
	int rv, size = attr->cqe;

	if (atomic_inc_return(&sdev->num_cq) > SIW_MAX_CQ) {
		siw_dbg(base_cq->device, "too many CQ's\n");
		rv = -ENOMEM;
		goto err_out;
	}
	if (size < 1 || size > sdev->attrs.max_cqe) {
		siw_dbg(base_cq->device, "CQ size error: %d\n", size);
		rv = -EINVAL;
		goto err_out;
	}
	size = roundup_pow_of_two(size);
	cq->base_cq.cqe = size;
	cq->num_cqe = size;
	cq->xa_cq_index = SIW_INVAL_UOBJ_KEY;

	if (!udata) {
		cq->kernel_verbs = 1;
		cq->queue = vzalloc(size * sizeof(struct siw_cqe) +
				    sizeof(struct siw_cq_ctrl));
	} else {
		cq->queue = vmalloc_user(size * sizeof(struct siw_cqe) +
					 sizeof(struct siw_cq_ctrl));
	}
	if (cq->queue == NULL) {
		rv = -ENOMEM;
		goto err_out;
	}
	get_random_bytes(&cq->id, 4);
	siw_dbg(base_cq->device, "new CQ [%u]\n", cq->id);

	spin_lock_init(&cq->lock);

	cq->notify = (struct siw_cq_ctrl *)&cq->queue[size];

	if (udata) {
		struct siw_uresp_create_cq uresp = {};
		struct siw_ucontext *ctx =
			rdma_udata_to_drv_context(udata, struct siw_ucontext,
						  base_ucontext);

		cq->xa_cq_index =
			siw_create_uobj(ctx, cq->queue,
					size * sizeof(struct siw_cqe) +
						sizeof(struct siw_cq_ctrl));
		if (cq->xa_cq_index == SIW_INVAL_UOBJ_KEY) {
			rv = -ENOMEM;
			goto err_out;
		}
		uresp.cq_key = cq->xa_cq_index << PAGE_SHIFT;
		uresp.cq_id = cq->id;
		uresp.num_cqe = size;

		if (udata->outlen < sizeof(uresp)) {
			rv = -EINVAL;
			goto err_out;
		}
		rv = ib_copy_to_udata(udata, &uresp, sizeof(uresp));
		if (rv)
			goto err_out;
	}
	return 0;

err_out:
	siw_dbg(base_cq->device, "CQ creation failed: %d", rv);

	if (cq && cq->queue) {
		struct siw_ucontext *ctx =
			rdma_udata_to_drv_context(udata, struct siw_ucontext,
						  base_ucontext);
		if (cq->xa_cq_index != SIW_INVAL_UOBJ_KEY)
			kfree(xa_erase(&ctx->xa, cq->xa_cq_index));
		vfree(cq->queue);
	}
	atomic_dec(&sdev->num_cq);

	return rv;
}

/*
 * siw_poll_cq()
 *
 * Reap CQ entries if available and copy work completion status into
 * array of WC's provided by caller. Returns number of reaped CQE's.
 *
 * @base_cq:	Base CQ contained in siw CQ.
 * @num_cqe:	Maximum number of CQE's to reap.
 * @wc:		Array of work completions to be filled by siw.
 */
int siw_poll_cq(struct ib_cq *base_cq, int num_cqe, struct ib_wc *wc)
{
	struct siw_cq *cq = to_siw_cq(base_cq);
	int i;

	for (i = 0; i < num_cqe; i++) {
		if (!siw_reap_cqe(cq, wc))
			break;
		wc++;
	}
	return i;
}

/*
 * siw_req_notify_cq()
 *
 * Request notification for new CQE's added to that CQ.
 * Defined flags:
 * o SIW_CQ_NOTIFY_SOLICITED lets siw trigger a notification
 *   event if a WQE with notification flag set enters the CQ
 * o SIW_CQ_NOTIFY_NEXT_COMP lets siw trigger a notification
 *   event if a WQE enters the CQ.
 * o IB_CQ_REPORT_MISSED_EVENTS: return value will provide the
 *   number of not reaped CQE's regardless of its notification
 *   type and current or new CQ notification settings.
 *
 * @base_cq:	Base CQ contained in siw CQ.
 * @flags:	Requested notification flags.
 */
int siw_req_notify_cq(struct ib_cq *base_cq, enum ib_cq_notify_flags flags)
{
	struct siw_cq *cq = to_siw_cq(base_cq);

	siw_dbg_cq(cq, "flags: 0x%02x\n", flags);

	if ((flags & IB_CQ_SOLICITED_MASK) == IB_CQ_SOLICITED)
		/*
		 * Enable CQ event for next solicited completion.
		 * and make it visible to all associated producers.
		 */
		smp_store_mb(cq->notify->flags, SIW_NOTIFY_SOLICITED);
	else
		/*
		 * Enable CQ event for any signalled completion.
		 * and make it visible to all associated producers.
		 */
		smp_store_mb(cq->notify->flags, SIW_NOTIFY_ALL);

	if (flags & IB_CQ_REPORT_MISSED_EVENTS)
		return cq->cq_put - cq->cq_get;

	return 0;
}

/*
 * siw_dereg_mr()
 *
 * Release Memory Region.
 *
 * @base_mr: Base MR contained in siw MR.
 * @udata: points to user context, unused.
 */
int siw_dereg_mr(struct ib_mr *base_mr, struct ib_udata *udata)
{
	struct siw_mr *mr = to_siw_mr(base_mr);
	struct siw_device *sdev = to_siw_dev(base_mr->device);

	siw_dbg_mem(mr->mem, "deregister MR\n");

	atomic_dec(&sdev->num_mr);

	siw_mr_drop_mem(mr);
	kfree_rcu(mr, rcu);

	return 0;
}

/*
 * siw_reg_user_mr()
 *
 * Register Memory Region.
 *
 * @pd:		Protection Domain
 * @start:	starting address of MR (virtual address)
 * @len:	len of MR
 * @rnic_va:	not used by siw
 * @rights:	MR access rights
 * @udata:	user buffer to communicate STag and Key.
 */
struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,
			      u64 rnic_va, int rights, struct ib_udata *udata)
{
	struct siw_mr *mr = NULL;
	struct siw_umem *umem = NULL;
	struct siw_ureq_reg_mr ureq;
	struct siw_device *sdev = to_siw_dev(pd->device);

	unsigned long mem_limit = rlimit(RLIMIT_MEMLOCK);
	int rv;

	siw_dbg_pd(pd, "start: 0x%pK, va: 0x%pK, len: %llu\n",
		   (void *)(uintptr_t)start, (void *)(uintptr_t)rnic_va,
		   (unsigned long long)len);

	if (atomic_inc_return(&sdev->num_mr) > SIW_MAX_MR) {
		siw_dbg_pd(pd, "too many mr's\n");
		rv = -ENOMEM;
		goto err_out;
	}
	if (!len) {
		rv = -EINVAL;
		goto err_out;
	}
	if (mem_limit != RLIM_INFINITY) {
		unsigned long num_pages =
			(PAGE_ALIGN(len + (start & ~PAGE_MASK))) >> PAGE_SHIFT;
		mem_limit >>= PAGE_SHIFT;

		if (num_pages > mem_limit - current->mm->locked_vm) {
			siw_dbg_pd(pd, "pages req %lu, max %lu, lock %lu\n",
				   num_pages, mem_limit,
				   current->mm->locked_vm);
			rv = -ENOMEM;
			goto err_out;
		}
	}
	umem = siw_umem_get(start, len, ib_access_writable(rights));
	if (IS_ERR(umem)) {
		rv = PTR_ERR(umem);
		siw_dbg_pd(pd, "getting user memory failed: %d\n", rv);
		umem = NULL;
		goto err_out;
	}
	mr = kzalloc(sizeof(*mr), GFP_KERNEL);
	if (!mr) {
		rv = -ENOMEM;
		goto err_out;
	}
	rv = siw_mr_add_mem(mr, pd, umem, start, len, rights);
	if (rv)
		goto err_out;

	if (udata) {
		struct siw_uresp_reg_mr uresp = {};
		struct siw_mem *mem = mr->mem;

		if (udata->inlen < sizeof(ureq)) {
			rv = -EINVAL;
			goto err_out;
		}
		rv = ib_copy_from_udata(&ureq, udata, sizeof(ureq));
		if (rv)
			goto err_out;

		mr->base_mr.lkey |= ureq.stag_key;
		mr->base_mr.rkey |= ureq.stag_key;
		mem->stag |= ureq.stag_key;
		uresp.stag = mem->stag;

		if (udata->outlen < sizeof(uresp)) {
			rv = -EINVAL;
			goto err_out;
		}
		rv = ib_copy_to_udata(udata, &uresp, sizeof(uresp));
		if (rv)
			goto err_out;
	}
	mr->mem->stag_valid = 1;

	return &mr->base_mr;

err_out:
	atomic_dec(&sdev->num_mr);
	if (mr) {
		if (mr->mem)
			siw_mr_drop_mem(mr);
		kfree_rcu(mr, rcu);
	} else {
		if (umem)
			siw_umem_release(umem, false);
	}
	return ERR_PTR(rv);
}

struct ib_mr *siw_alloc_mr(struct ib_pd *pd, enum ib_mr_type mr_type,
			   u32 max_sge, struct ib_udata *udata)
{
	struct siw_device *sdev = to_siw_dev(pd->device);
	struct siw_mr *mr = NULL;
	struct siw_pbl *pbl = NULL;
	int rv;

	if (atomic_inc_return(&sdev->num_mr) > SIW_MAX_MR) {
		siw_dbg_pd(pd, "too many mr's\n");
		rv = -ENOMEM;
		goto err_out;
	}
	if (mr_type != IB_MR_TYPE_MEM_REG) {
		siw_dbg_pd(pd, "mr type %d unsupported\n", mr_type);
		rv = -EOPNOTSUPP;
		goto err_out;
	}
	if (max_sge > SIW_MAX_SGE_PBL) {
		siw_dbg_pd(pd, "too many sge's: %d\n", max_sge);
		rv = -ENOMEM;
		goto err_out;
	}
	pbl = siw_pbl_alloc(max_sge);
	if (IS_ERR(pbl)) {
		rv = PTR_ERR(pbl);
		siw_dbg_pd(pd, "pbl allocation failed: %d\n", rv);
		pbl = NULL;
		goto err_out;
	}
	mr = kzalloc(sizeof(*mr), GFP_KERNEL);
	if (!mr) {
		rv = -ENOMEM;
		goto err_out;
	}
	rv = siw_mr_add_mem(mr, pd, pbl, 0, max_sge * PAGE_SIZE, 0);
	if (rv)
		goto err_out;

	mr->mem->is_pbl = 1;

	siw_dbg_pd(pd, "[MEM %u]: success\n", mr->mem->stag);

	return &mr->base_mr;

err_out:
	atomic_dec(&sdev->num_mr);

	if (!mr) {
		kfree(pbl);
	} else {
		if (mr->mem)
			siw_mr_drop_mem(mr);
		kfree_rcu(mr, rcu);
	}
	siw_dbg_pd(pd, "failed: %d\n", rv);

	return ERR_PTR(rv);
}

/* Just used to count number of pages being mapped */
static int siw_set_pbl_page(struct ib_mr *base_mr, u64 buf_addr)
{
	return 0;
}

int siw_map_mr_sg(struct ib_mr *base_mr, struct scatterlist *sl, int num_sle,
		  unsigned int *sg_off)
{
	struct scatterlist *slp;
	struct siw_mr *mr = to_siw_mr(base_mr);
	struct siw_mem *mem = mr->mem;
	struct siw_pbl *pbl = mem->pbl;
	struct siw_pble *pble;
	unsigned long pbl_size;
	int i, rv;

	if (!pbl) {
		siw_dbg_mem(mem, "no PBL allocated\n");
		return -EINVAL;
	}
	pble = pbl->pbe;

	if (pbl->max_buf < num_sle) {
		siw_dbg_mem(mem, "too many SGE's: %d > %d\n",
			    mem->pbl->max_buf, num_sle);
		return -ENOMEM;
	}
	for_each_sg(sl, slp, num_sle, i) {
		if (sg_dma_len(slp) == 0) {
			siw_dbg_mem(mem, "empty SGE\n");
			return -EINVAL;
		}
		if (i == 0) {
			pble->addr = sg_dma_address(slp);
			pble->size = sg_dma_len(slp);
			pble->pbl_off = 0;
			pbl_size = pble->size;
			pbl->num_buf = 1;
		} else {
			/* Merge PBL entries if adjacent */
			if (pble->addr + pble->size == sg_dma_address(slp)) {
				pble->size += sg_dma_len(slp);
			} else {
				pble++;
				pbl->num_buf++;
				pble->addr = sg_dma_address(slp);
				pble->size = sg_dma_len(slp);
				pble->pbl_off = pbl_size;
			}
			pbl_size += sg_dma_len(slp);
		}
		siw_dbg_mem(mem,
			"sge[%d], size %u, addr 0x%p, total %lu\n",
			i, pble->size, (void *)(uintptr_t)pble->addr,
			pbl_size);
	}
	rv = ib_sg_to_pages(base_mr, sl, num_sle, sg_off, siw_set_pbl_page);
	if (rv > 0) {
		mem->len = base_mr->length;
		mem->va = base_mr->iova;
		siw_dbg_mem(mem,
			"%llu bytes, start 0x%pK, %u SLE to %u entries\n",
			mem->len, (void *)(uintptr_t)mem->va, num_sle,
			pbl->num_buf);
	}
	return rv;
}

/*
 * siw_get_dma_mr()
 *
 * Create a (empty) DMA memory region, where no umem is attached.
 */
struct ib_mr *siw_get_dma_mr(struct ib_pd *pd, int rights)
{
	struct siw_device *sdev = to_siw_dev(pd->device);
	struct siw_mr *mr = NULL;
	int rv;

	if (atomic_inc_return(&sdev->num_mr) > SIW_MAX_MR) {
		siw_dbg_pd(pd, "too many mr's\n");
		rv = -ENOMEM;
		goto err_out;
	}
	mr = kzalloc(sizeof(*mr), GFP_KERNEL);
	if (!mr) {
		rv = -ENOMEM;
		goto err_out;
	}
	rv = siw_mr_add_mem(mr, pd, NULL, 0, ULONG_MAX, rights);
	if (rv)
		goto err_out;

	mr->mem->stag_valid = 1;

	siw_dbg_pd(pd, "[MEM %u]: success\n", mr->mem->stag);

	return &mr->base_mr;

err_out:
	if (rv)
		kfree(mr);

	atomic_dec(&sdev->num_mr);

	return ERR_PTR(rv);
}

/*
 * siw_create_srq()
 *
 * Create Shared Receive Queue of attributes @init_attrs
 * within protection domain given by @pd.
 *
 * @base_srq:	Base SRQ contained in siw SRQ.
 * @init_attrs:	SRQ init attributes.
 * @udata:	points to user context
 */
int siw_create_srq(struct ib_srq *base_srq,
		   struct ib_srq_init_attr *init_attrs, struct ib_udata *udata)
{
	struct siw_srq *srq = to_siw_srq(base_srq);
	struct ib_srq_attr *attrs = &init_attrs->attr;
	struct siw_device *sdev = to_siw_dev(base_srq->device);
	struct siw_ucontext *ctx =
		rdma_udata_to_drv_context(udata, struct siw_ucontext,
					  base_ucontext);
	int rv;

	if (atomic_inc_return(&sdev->num_srq) > SIW_MAX_SRQ) {
		siw_dbg_pd(base_srq->pd, "too many SRQ's\n");
		rv = -ENOMEM;
		goto err_out;
	}
	if (attrs->max_wr == 0 || attrs->max_wr > SIW_MAX_SRQ_WR ||
	    attrs->max_sge > SIW_MAX_SGE || attrs->srq_limit > attrs->max_wr) {
		rv = -EINVAL;
		goto err_out;
	}
	srq->max_sge = attrs->max_sge;
	srq->num_rqe = roundup_pow_of_two(attrs->max_wr);
	srq->xa_srq_index = SIW_INVAL_UOBJ_KEY;
	srq->limit = attrs->srq_limit;
	if (srq->limit)
		srq->armed = 1;

	srq->kernel_verbs = !udata;

	if (udata)
		srq->recvq =
			vmalloc_user(srq->num_rqe * sizeof(struct siw_rqe));
	else
		srq->recvq = vzalloc(srq->num_rqe * sizeof(struct siw_rqe));

	if (srq->recvq == NULL) {
		rv = -ENOMEM;
		goto err_out;
	}
	if (udata) {
		struct siw_uresp_create_srq uresp = {};

		srq->xa_srq_index = siw_create_uobj(
			ctx, srq->recvq, srq->num_rqe * sizeof(struct siw_rqe));

		if (srq->xa_srq_index == SIW_INVAL_UOBJ_KEY) {
			rv = -ENOMEM;
			goto err_out;
		}
		uresp.srq_key = srq->xa_srq_index;
		uresp.num_rqe = srq->num_rqe;

		if (udata->outlen < sizeof(uresp)) {
			rv = -EINVAL;
			goto err_out;
		}
		rv = ib_copy_to_udata(udata, &uresp, sizeof(uresp));
		if (rv)
			goto err_out;
	}
	spin_lock_init(&srq->lock);

	siw_dbg_pd(base_srq->pd, "[SRQ]: success\n");

	return 0;

err_out:
	if (srq->recvq) {
		if (ctx && srq->xa_srq_index != SIW_INVAL_UOBJ_KEY)
			kfree(xa_erase(&ctx->xa, srq->xa_srq_index));
		vfree(srq->recvq);
	}
	atomic_dec(&sdev->num_srq);

	return rv;
}

/*
 * siw_modify_srq()
 *
 * Modify SRQ. The caller may resize SRQ and/or set/reset notification
 * limit and (re)arm IB_EVENT_SRQ_LIMIT_REACHED notification.
 *
 * NOTE: it is unclear if RDMA core allows for changing the MAX_SGE
 * parameter. siw_modify_srq() does not check the attrs->max_sge param.
 */
int siw_modify_srq(struct ib_srq *base_srq, struct ib_srq_attr *attrs,
		   enum ib_srq_attr_mask attr_mask, struct ib_udata *udata)
{
	struct siw_srq *srq = to_siw_srq(base_srq);
	unsigned long flags;
	int rv = 0;

	spin_lock_irqsave(&srq->lock, flags);

	if (attr_mask & IB_SRQ_MAX_WR) {
		/* resize request not yet supported */
		rv = -EOPNOTSUPP;
		goto out;
	}
	if (attr_mask & IB_SRQ_LIMIT) {
		if (attrs->srq_limit) {
			if (unlikely(attrs->srq_limit > srq->num_rqe)) {
				rv = -EINVAL;
				goto out;
			}
			srq->armed = 1;
		} else {
			srq->armed = 0;
		}
		srq->limit = attrs->srq_limit;
	}
out:
	spin_unlock_irqrestore(&srq->lock, flags);

	return rv;
}

/*
 * siw_query_srq()
 *
 * Query SRQ attributes.
 */
int siw_query_srq(struct ib_srq *base_srq, struct ib_srq_attr *attrs)
{
	struct siw_srq *srq = to_siw_srq(base_srq);
	unsigned long flags;

	spin_lock_irqsave(&srq->lock, flags);

	attrs->max_wr = srq->num_rqe;
	attrs->max_sge = srq->max_sge;
	attrs->srq_limit = srq->limit;

	spin_unlock_irqrestore(&srq->lock, flags);

	return 0;
}

/*
 * siw_destroy_srq()
 *
 * Destroy SRQ.
 * It is assumed that the SRQ is not referenced by any
 * QP anymore - the code trusts the RDMA core environment to keep track
 * of QP references.
 */
void siw_destroy_srq(struct ib_srq *base_srq, struct ib_udata *udata)
{
	struct siw_srq *srq = to_siw_srq(base_srq);
	struct siw_device *sdev = to_siw_dev(base_srq->device);
	struct siw_ucontext *ctx =
		rdma_udata_to_drv_context(udata, struct siw_ucontext,
					  base_ucontext);

	if (ctx && srq->xa_srq_index != SIW_INVAL_UOBJ_KEY)
		kfree(xa_erase(&ctx->xa, srq->xa_srq_index));

	vfree(srq->recvq);
	atomic_dec(&sdev->num_srq);
}

/*
 * siw_post_srq_recv()
 *
 * Post a list of receive queue elements to SRQ.
 * NOTE: The function does not check or lock a certain SRQ state
 *       during the post operation. The code simply trusts the
 *       RDMA core environment.
 *
 * @base_srq:	Base SRQ contained in siw SRQ
 * @wr:		List of R-WR's
 * @bad_wr:	Updated to failing WR if posting fails.
 */
int siw_post_srq_recv(struct ib_srq *base_srq, const struct ib_recv_wr *wr,
		      const struct ib_recv_wr **bad_wr)
{
	struct siw_srq *srq = to_siw_srq(base_srq);
	unsigned long flags;
	int rv = 0;

	if (unlikely(!srq->kernel_verbs)) {
		siw_dbg_pd(base_srq->pd,
			   "[SRQ]: no kernel post_recv for mapped srq\n");
		rv = -EINVAL;
		goto out;
	}
	/*
	 * Serialize potentially multiple producers.
	 * Also needed to serialize potentially multiple
	 * consumers.
	 */
	spin_lock_irqsave(&srq->lock, flags);

	while (wr) {
		u32 idx = srq->rq_put % srq->num_rqe;
		struct siw_rqe *rqe = &srq->recvq[idx];

		if (rqe->flags) {
			siw_dbg_pd(base_srq->pd, "SRQ full\n");
			rv = -ENOMEM;
			break;
		}
		if (unlikely(wr->num_sge > srq->max_sge)) {
			siw_dbg_pd(base_srq->pd,
				   "[SRQ]: too many sge's: %d\n", wr->num_sge);
			rv = -EINVAL;
			break;
		}
		rqe->id = wr->wr_id;
		rqe->num_sge = wr->num_sge;
		siw_copy_sgl(wr->sg_list, rqe->sge, wr->num_sge);

		/* Make sure S-RQE is completely written before valid */
		smp_wmb();

		rqe->flags = SIW_WQE_VALID;

		srq->rq_put++;
		wr = wr->next;
	}
	spin_unlock_irqrestore(&srq->lock, flags);
out:
	if (unlikely(rv < 0)) {
		siw_dbg_pd(base_srq->pd, "[SRQ]: error %d\n", rv);
		*bad_wr = wr;
	}
	return rv;
}

void siw_qp_event(struct siw_qp *qp, enum ib_event_type etype)
{
	struct ib_event event;
	struct ib_qp *base_qp = qp->ib_qp;

	/*
	 * Do not report asynchronous errors on QP which gets
	 * destroyed via verbs interface (siw_destroy_qp())
	 */
	if (qp->attrs.flags & SIW_QP_IN_DESTROY)
		return;

	event.event = etype;
	event.device = base_qp->device;
	event.element.qp = base_qp;

	if (base_qp->event_handler) {
		siw_dbg_qp(qp, "reporting event %d\n", etype);
		base_qp->event_handler(&event, base_qp->qp_context);
	}
}

void siw_cq_event(struct siw_cq *cq, enum ib_event_type etype)
{
	struct ib_event event;
	struct ib_cq *base_cq = &cq->base_cq;

	event.event = etype;
	event.device = base_cq->device;
	event.element.cq = base_cq;

	if (base_cq->event_handler) {
		siw_dbg_cq(cq, "reporting CQ event %d\n", etype);
		base_cq->event_handler(&event, base_cq->cq_context);
	}
}

void siw_srq_event(struct siw_srq *srq, enum ib_event_type etype)
{
	struct ib_event event;
	struct ib_srq *base_srq = &srq->base_srq;

	event.event = etype;
	event.device = base_srq->device;
	event.element.srq = base_srq;

	if (base_srq->event_handler) {
		siw_dbg_pd(srq->base_srq.pd,
			   "reporting SRQ event %d\n", etype);
		base_srq->event_handler(&event, base_srq->srq_context);
	}
}

void siw_port_event(struct siw_device *sdev, u8 port, enum ib_event_type etype)
{
	struct ib_event event;

	event.event = etype;
	event.device = &sdev->base_dev;
	event.element.port_num = port;

	siw_dbg(&sdev->base_dev, "reporting port event %d\n", etype);

	ib_dispatch_event(&event);
}