Free Electrons

Embedded Linux Experts

   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
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
/*
 * Copyright (c) 2006, 2007 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 <rdma/ib_mad.h>
#include <rdma/ib_user_verbs.h>
#include <linux/io.h>
#include <linux/utsname.h>

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

static unsigned int ib_ipath_qp_table_size = 251;
module_param_named(qp_table_size, ib_ipath_qp_table_size, uint, S_IRUGO);
MODULE_PARM_DESC(qp_table_size, "QP table size");

unsigned int ib_ipath_lkey_table_size = 12;
module_param_named(lkey_table_size, ib_ipath_lkey_table_size, uint,
		   S_IRUGO);
MODULE_PARM_DESC(lkey_table_size,
		 "LKEY table size in bits (2^n, 1 <= n <= 23)");

static unsigned int ib_ipath_max_pds = 0xFFFF;
module_param_named(max_pds, ib_ipath_max_pds, uint, S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_pds,
		 "Maximum number of protection domains to support");

static unsigned int ib_ipath_max_ahs = 0xFFFF;
module_param_named(max_ahs, ib_ipath_max_ahs, uint, S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_ahs, "Maximum number of address handles to support");

unsigned int ib_ipath_max_cqes = 0x2FFFF;
module_param_named(max_cqes, ib_ipath_max_cqes, uint, S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_cqes,
		 "Maximum number of completion queue entries to support");

unsigned int ib_ipath_max_cqs = 0x1FFFF;
module_param_named(max_cqs, ib_ipath_max_cqs, uint, S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_cqs, "Maximum number of completion queues to support");

unsigned int ib_ipath_max_qp_wrs = 0x3FFF;
module_param_named(max_qp_wrs, ib_ipath_max_qp_wrs, uint,
		   S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_qp_wrs, "Maximum number of QP WRs to support");

unsigned int ib_ipath_max_qps = 16384;
module_param_named(max_qps, ib_ipath_max_qps, uint, S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_qps, "Maximum number of QPs to support");

unsigned int ib_ipath_max_sges = 0x60;
module_param_named(max_sges, ib_ipath_max_sges, uint, S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_sges, "Maximum number of SGEs to support");

unsigned int ib_ipath_max_mcast_grps = 16384;
module_param_named(max_mcast_grps, ib_ipath_max_mcast_grps, uint,
		   S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_mcast_grps,
		 "Maximum number of multicast groups to support");

unsigned int ib_ipath_max_mcast_qp_attached = 16;
module_param_named(max_mcast_qp_attached, ib_ipath_max_mcast_qp_attached,
		   uint, S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_mcast_qp_attached,
		 "Maximum number of attached QPs to support");

unsigned int ib_ipath_max_srqs = 1024;
module_param_named(max_srqs, ib_ipath_max_srqs, uint, S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_srqs, "Maximum number of SRQs to support");

unsigned int ib_ipath_max_srq_sges = 128;
module_param_named(max_srq_sges, ib_ipath_max_srq_sges,
		   uint, S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_srq_sges, "Maximum number of SRQ SGEs to support");

unsigned int ib_ipath_max_srq_wrs = 0x1FFFF;
module_param_named(max_srq_wrs, ib_ipath_max_srq_wrs,
		   uint, S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_srq_wrs, "Maximum number of SRQ WRs support");

static unsigned int ib_ipath_disable_sma;
module_param_named(disable_sma, ib_ipath_disable_sma, uint, S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(ib_ipath_disable_sma, "Disable the SMA");

const int ib_ipath_state_ops[IB_QPS_ERR + 1] = {
	[IB_QPS_RESET] = 0,
	[IB_QPS_INIT] = IPATH_POST_RECV_OK,
	[IB_QPS_RTR] = IPATH_POST_RECV_OK | IPATH_PROCESS_RECV_OK,
	[IB_QPS_RTS] = IPATH_POST_RECV_OK | IPATH_PROCESS_RECV_OK |
	    IPATH_POST_SEND_OK | IPATH_PROCESS_SEND_OK,
	[IB_QPS_SQD] = IPATH_POST_RECV_OK | IPATH_PROCESS_RECV_OK |
	    IPATH_POST_SEND_OK,
	[IB_QPS_SQE] = IPATH_POST_RECV_OK | IPATH_PROCESS_RECV_OK,
	[IB_QPS_ERR] = 0,
};

struct ipath_ucontext {
	struct ib_ucontext ibucontext;
};

static inline struct ipath_ucontext *to_iucontext(struct ib_ucontext
						  *ibucontext)
{
	return container_of(ibucontext, struct ipath_ucontext, ibucontext);
}

/*
 * Translate ib_wr_opcode into ib_wc_opcode.
 */
const enum ib_wc_opcode ib_ipath_wc_opcode[] = {
	[IB_WR_RDMA_WRITE] = IB_WC_RDMA_WRITE,
	[IB_WR_RDMA_WRITE_WITH_IMM] = IB_WC_RDMA_WRITE,
	[IB_WR_SEND] = IB_WC_SEND,
	[IB_WR_SEND_WITH_IMM] = IB_WC_SEND,
	[IB_WR_RDMA_READ] = IB_WC_RDMA_READ,
	[IB_WR_ATOMIC_CMP_AND_SWP] = IB_WC_COMP_SWAP,
	[IB_WR_ATOMIC_FETCH_AND_ADD] = IB_WC_FETCH_ADD
};

/*
 * System image GUID.
 */
static __be64 sys_image_guid;

/**
 * ipath_copy_sge - copy data to SGE memory
 * @ss: the SGE state
 * @data: the data to copy
 * @length: the length of the data
 */
void ipath_copy_sge(struct ipath_sge_state *ss, void *data, u32 length)
{
	struct ipath_sge *sge = &ss->sge;

	while (length) {
		u32 len = sge->length;

		if (len > length)
			len = length;
		if (len > sge->sge_length)
			len = sge->sge_length;
		BUG_ON(len == 0);
		memcpy(sge->vaddr, data, len);
		sge->vaddr += len;
		sge->length -= len;
		sge->sge_length -= len;
		if (sge->sge_length == 0) {
			if (--ss->num_sge)
				*sge = *ss->sg_list++;
		} else if (sge->length == 0 && sge->mr != NULL) {
			if (++sge->n >= IPATH_SEGSZ) {
				if (++sge->m >= sge->mr->mapsz)
					break;
				sge->n = 0;
			}
			sge->vaddr =
				sge->mr->map[sge->m]->segs[sge->n].vaddr;
			sge->length =
				sge->mr->map[sge->m]->segs[sge->n].length;
		}
		data += len;
		length -= len;
	}
}

/**
 * ipath_skip_sge - skip over SGE memory - XXX almost dup of prev func
 * @ss: the SGE state
 * @length: the number of bytes to skip
 */
void ipath_skip_sge(struct ipath_sge_state *ss, u32 length)
{
	struct ipath_sge *sge = &ss->sge;

	while (length) {
		u32 len = sge->length;

		if (len > length)
			len = length;
		if (len > sge->sge_length)
			len = sge->sge_length;
		BUG_ON(len == 0);
		sge->vaddr += len;
		sge->length -= len;
		sge->sge_length -= len;
		if (sge->sge_length == 0) {
			if (--ss->num_sge)
				*sge = *ss->sg_list++;
		} else if (sge->length == 0 && sge->mr != NULL) {
			if (++sge->n >= IPATH_SEGSZ) {
				if (++sge->m >= sge->mr->mapsz)
					break;
				sge->n = 0;
			}
			sge->vaddr =
				sge->mr->map[sge->m]->segs[sge->n].vaddr;
			sge->length =
				sge->mr->map[sge->m]->segs[sge->n].length;
		}
		length -= len;
	}
}

static void ipath_flush_wqe(struct ipath_qp *qp, struct ib_send_wr *wr)
{
	struct ib_wc wc;

	memset(&wc, 0, sizeof(wc));
	wc.wr_id = wr->wr_id;
	wc.status = IB_WC_WR_FLUSH_ERR;
	wc.opcode = ib_ipath_wc_opcode[wr->opcode];
	wc.qp = &qp->ibqp;
	ipath_cq_enter(to_icq(qp->ibqp.send_cq), &wc, 1);
}

/**
 * ipath_post_one_send - post one RC, UC, or UD send work request
 * @qp: the QP to post on
 * @wr: the work request to send
 */
static int ipath_post_one_send(struct ipath_qp *qp, struct ib_send_wr *wr)
{
	struct ipath_swqe *wqe;
	u32 next;
	int i;
	int j;
	int acc;
	int ret;
	unsigned long flags;

	spin_lock_irqsave(&qp->s_lock, flags);

	/* Check that state is OK to post send. */
	if (unlikely(!(ib_ipath_state_ops[qp->state] & IPATH_POST_SEND_OK))) {
		if (qp->state != IB_QPS_SQE && qp->state != IB_QPS_ERR)
			goto bail_inval;
		/* C10-96 says generate a flushed completion entry. */
		ipath_flush_wqe(qp, wr);
		ret = 0;
		goto bail;
	}

	/* IB spec says that num_sge == 0 is OK. */
	if (wr->num_sge > qp->s_max_sge)
		goto bail_inval;

	/*
	 * Don't allow RDMA reads or atomic operations on UC or
	 * undefined operations.
	 * Make sure buffer is large enough to hold the result for atomics.
	 */
	if (qp->ibqp.qp_type == IB_QPT_UC) {
		if ((unsigned) wr->opcode >= IB_WR_RDMA_READ)
			goto bail_inval;
	} else if (qp->ibqp.qp_type == IB_QPT_UD) {
		/* Check UD opcode */
		if (wr->opcode != IB_WR_SEND &&
		    wr->opcode != IB_WR_SEND_WITH_IMM)
			goto bail_inval;
		/* Check UD destination address PD */
		if (qp->ibqp.pd != wr->wr.ud.ah->pd)
			goto bail_inval;
	} else if ((unsigned) wr->opcode > IB_WR_ATOMIC_FETCH_AND_ADD)
		goto bail_inval;
	else if (wr->opcode >= IB_WR_ATOMIC_CMP_AND_SWP &&
		   (wr->num_sge == 0 ||
		    wr->sg_list[0].length < sizeof(u64) ||
		    wr->sg_list[0].addr & (sizeof(u64) - 1)))
		goto bail_inval;
	else if (wr->opcode >= IB_WR_RDMA_READ && !qp->s_max_rd_atomic)
		goto bail_inval;

	next = qp->s_head + 1;
	if (next >= qp->s_size)
		next = 0;
	if (next == qp->s_last) {
		ret = -ENOMEM;
		goto bail;
	}

	wqe = get_swqe_ptr(qp, qp->s_head);
	wqe->wr = *wr;
	wqe->ssn = qp->s_ssn++;
	wqe->length = 0;
	if (wr->num_sge) {
		acc = wr->opcode >= IB_WR_RDMA_READ ?
			IB_ACCESS_LOCAL_WRITE : 0;
		for (i = 0, j = 0; i < wr->num_sge; i++) {
			u32 length = wr->sg_list[i].length;
			int ok;

			if (length == 0)
				continue;
			ok = ipath_lkey_ok(qp, &wqe->sg_list[j],
					   &wr->sg_list[i], acc);
			if (!ok)
				goto bail_inval;
			wqe->length += length;
			j++;
		}
		wqe->wr.num_sge = j;
	}
	if (qp->ibqp.qp_type == IB_QPT_UC ||
	    qp->ibqp.qp_type == IB_QPT_RC) {
		if (wqe->length > 0x80000000U)
			goto bail_inval;
	} else if (wqe->length > to_idev(qp->ibqp.device)->dd->ipath_ibmtu)
		goto bail_inval;
	qp->s_head = next;

	ret = 0;
	goto bail;

bail_inval:
	ret = -EINVAL;
bail:
	spin_unlock_irqrestore(&qp->s_lock, flags);
	return ret;
}

/**
 * ipath_post_send - post a send on a QP
 * @ibqp: the QP to post the send on
 * @wr: the list of work requests to post
 * @bad_wr: the first bad WR is put here
 *
 * This may be called from interrupt context.
 */
static int ipath_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
			   struct ib_send_wr **bad_wr)
{
	struct ipath_qp *qp = to_iqp(ibqp);
	int err = 0;

	for (; wr; wr = wr->next) {
		err = ipath_post_one_send(qp, wr);
		if (err) {
			*bad_wr = wr;
			goto bail;
		}
	}

	/* Try to do the send work in the caller's context. */
	ipath_do_send((unsigned long) qp);

bail:
	return err;
}

/**
 * ipath_post_receive - post a receive on a QP
 * @ibqp: the QP to post the receive on
 * @wr: the WR to post
 * @bad_wr: the first bad WR is put here
 *
 * This may be called from interrupt context.
 */
static int ipath_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *wr,
			      struct ib_recv_wr **bad_wr)
{
	struct ipath_qp *qp = to_iqp(ibqp);
	struct ipath_rwq *wq = qp->r_rq.wq;
	unsigned long flags;
	int ret;

	/* Check that state is OK to post receive. */
	if (!(ib_ipath_state_ops[qp->state] & IPATH_POST_RECV_OK) || !wq) {
		*bad_wr = wr;
		ret = -EINVAL;
		goto bail;
	}

	for (; wr; wr = wr->next) {
		struct ipath_rwqe *wqe;
		u32 next;
		int i;

		if ((unsigned) wr->num_sge > qp->r_rq.max_sge) {
			*bad_wr = wr;
			ret = -EINVAL;
			goto bail;
		}

		spin_lock_irqsave(&qp->r_rq.lock, flags);
		next = wq->head + 1;
		if (next >= qp->r_rq.size)
			next = 0;
		if (next == wq->tail) {
			spin_unlock_irqrestore(&qp->r_rq.lock, flags);
			*bad_wr = wr;
			ret = -ENOMEM;
			goto bail;
		}

		wqe = get_rwqe_ptr(&qp->r_rq, wq->head);
		wqe->wr_id = wr->wr_id;
		wqe->num_sge = wr->num_sge;
		for (i = 0; i < wr->num_sge; i++)
			wqe->sg_list[i] = wr->sg_list[i];
		/* Make sure queue entry is written before the head index. */
		smp_wmb();
		wq->head = next;
		spin_unlock_irqrestore(&qp->r_rq.lock, flags);
	}
	ret = 0;

bail:
	return ret;
}

/**
 * ipath_qp_rcv - processing an incoming packet on a QP
 * @dev: the device the packet came on
 * @hdr: the packet header
 * @has_grh: true if the packet has a GRH
 * @data: the packet data
 * @tlen: the packet length
 * @qp: the QP the packet came on
 *
 * This is called from ipath_ib_rcv() to process an incoming packet
 * for the given QP.
 * Called at interrupt level.
 */
static void ipath_qp_rcv(struct ipath_ibdev *dev,
			 struct ipath_ib_header *hdr, int has_grh,
			 void *data, u32 tlen, struct ipath_qp *qp)
{
	/* Check for valid receive state. */
	if (!(ib_ipath_state_ops[qp->state] & IPATH_PROCESS_RECV_OK)) {
		dev->n_pkt_drops++;
		return;
	}

	switch (qp->ibqp.qp_type) {
	case IB_QPT_SMI:
	case IB_QPT_GSI:
		if (ib_ipath_disable_sma)
			break;
		/* FALLTHROUGH */
	case IB_QPT_UD:
		ipath_ud_rcv(dev, hdr, has_grh, data, tlen, qp);
		break;

	case IB_QPT_RC:
		ipath_rc_rcv(dev, hdr, has_grh, data, tlen, qp);
		break;

	case IB_QPT_UC:
		ipath_uc_rcv(dev, hdr, has_grh, data, tlen, qp);
		break;

	default:
		break;
	}
}

/**
 * ipath_ib_rcv - process an incoming packet
 * @arg: the device pointer
 * @rhdr: the header of the packet
 * @data: the packet data
 * @tlen: the packet length
 *
 * This is called from ipath_kreceive() to process an incoming packet at
 * interrupt level. Tlen is the length of the header + data + CRC in bytes.
 */
void ipath_ib_rcv(struct ipath_ibdev *dev, void *rhdr, void *data,
		  u32 tlen)
{
	struct ipath_ib_header *hdr = rhdr;
	struct ipath_other_headers *ohdr;
	struct ipath_qp *qp;
	u32 qp_num;
	int lnh;
	u8 opcode;
	u16 lid;

	if (unlikely(dev == NULL))
		goto bail;

	if (unlikely(tlen < 24)) {	/* LRH+BTH+CRC */
		dev->rcv_errors++;
		goto bail;
	}

	/* Check for a valid destination LID (see ch. 7.11.1). */
	lid = be16_to_cpu(hdr->lrh[1]);
	if (lid < IPATH_MULTICAST_LID_BASE) {
		lid &= ~((1 << dev->dd->ipath_lmc) - 1);
		if (unlikely(lid != dev->dd->ipath_lid)) {
			dev->rcv_errors++;
			goto bail;
		}
	}

	/* Check for GRH */
	lnh = be16_to_cpu(hdr->lrh[0]) & 3;
	if (lnh == IPATH_LRH_BTH)
		ohdr = &hdr->u.oth;
	else if (lnh == IPATH_LRH_GRH)
		ohdr = &hdr->u.l.oth;
	else {
		dev->rcv_errors++;
		goto bail;
	}

	opcode = be32_to_cpu(ohdr->bth[0]) >> 24;
	dev->opstats[opcode].n_bytes += tlen;
	dev->opstats[opcode].n_packets++;

	/* Get the destination QP number. */
	qp_num = be32_to_cpu(ohdr->bth[1]) & IPATH_QPN_MASK;
	if (qp_num == IPATH_MULTICAST_QPN) {
		struct ipath_mcast *mcast;
		struct ipath_mcast_qp *p;

		if (lnh != IPATH_LRH_GRH) {
			dev->n_pkt_drops++;
			goto bail;
		}
		mcast = ipath_mcast_find(&hdr->u.l.grh.dgid);
		if (mcast == NULL) {
			dev->n_pkt_drops++;
			goto bail;
		}
		dev->n_multicast_rcv++;
		list_for_each_entry_rcu(p, &mcast->qp_list, list)
			ipath_qp_rcv(dev, hdr, 1, data, tlen, p->qp);
		/*
		 * Notify ipath_multicast_detach() if it is waiting for us
		 * to finish.
		 */
		if (atomic_dec_return(&mcast->refcount) <= 1)
			wake_up(&mcast->wait);
	} else {
		qp = ipath_lookup_qpn(&dev->qp_table, qp_num);
		if (qp) {
			dev->n_unicast_rcv++;
			ipath_qp_rcv(dev, hdr, lnh == IPATH_LRH_GRH, data,
				     tlen, qp);
			/*
			 * Notify ipath_destroy_qp() if it is waiting
			 * for us to finish.
			 */
			if (atomic_dec_and_test(&qp->refcount))
				wake_up(&qp->wait);
		} else
			dev->n_pkt_drops++;
	}

bail:;
}

/**
 * ipath_ib_timer - verbs timer
 * @arg: the device pointer
 *
 * This is called from ipath_do_rcv_timer() at interrupt level to check for
 * QPs which need retransmits and to collect performance numbers.
 */
static void ipath_ib_timer(struct ipath_ibdev *dev)
{
	struct ipath_qp *resend = NULL;
	struct list_head *last;
	struct ipath_qp *qp;
	unsigned long flags;

	if (dev == NULL)
		return;

	spin_lock_irqsave(&dev->pending_lock, flags);
	/* Start filling the next pending queue. */
	if (++dev->pending_index >= ARRAY_SIZE(dev->pending))
		dev->pending_index = 0;
	/* Save any requests still in the new queue, they have timed out. */
	last = &dev->pending[dev->pending_index];
	while (!list_empty(last)) {
		qp = list_entry(last->next, struct ipath_qp, timerwait);
		list_del_init(&qp->timerwait);
		qp->timer_next = resend;
		resend = qp;
		atomic_inc(&qp->refcount);
	}
	last = &dev->rnrwait;
	if (!list_empty(last)) {
		qp = list_entry(last->next, struct ipath_qp, timerwait);
		if (--qp->s_rnr_timeout == 0) {
			do {
				list_del_init(&qp->timerwait);
				tasklet_hi_schedule(&qp->s_task);
				if (list_empty(last))
					break;
				qp = list_entry(last->next, struct ipath_qp,
						timerwait);
			} while (qp->s_rnr_timeout == 0);
		}
	}
	/*
	 * We should only be in the started state if pma_sample_start != 0
	 */
	if (dev->pma_sample_status == IB_PMA_SAMPLE_STATUS_STARTED &&
	    --dev->pma_sample_start == 0) {
		dev->pma_sample_status = IB_PMA_SAMPLE_STATUS_RUNNING;
		ipath_snapshot_counters(dev->dd, &dev->ipath_sword,
					&dev->ipath_rword,
					&dev->ipath_spkts,
					&dev->ipath_rpkts,
					&dev->ipath_xmit_wait);
	}
	if (dev->pma_sample_status == IB_PMA_SAMPLE_STATUS_RUNNING) {
		if (dev->pma_sample_interval == 0) {
			u64 ta, tb, tc, td, te;

			dev->pma_sample_status = IB_PMA_SAMPLE_STATUS_DONE;
			ipath_snapshot_counters(dev->dd, &ta, &tb,
						&tc, &td, &te);

			dev->ipath_sword = ta - dev->ipath_sword;
			dev->ipath_rword = tb - dev->ipath_rword;
			dev->ipath_spkts = tc - dev->ipath_spkts;
			dev->ipath_rpkts = td - dev->ipath_rpkts;
			dev->ipath_xmit_wait = te - dev->ipath_xmit_wait;
		}
		else
			dev->pma_sample_interval--;
	}
	spin_unlock_irqrestore(&dev->pending_lock, flags);

	/* XXX What if timer fires again while this is running? */
	for (qp = resend; qp != NULL; qp = qp->timer_next) {
		struct ib_wc wc;

		spin_lock_irqsave(&qp->s_lock, flags);
		if (qp->s_last != qp->s_tail && qp->state == IB_QPS_RTS) {
			dev->n_timeouts++;
			ipath_restart_rc(qp, qp->s_last_psn + 1, &wc);
		}
		spin_unlock_irqrestore(&qp->s_lock, flags);

		/* Notify ipath_destroy_qp() if it is waiting. */
		if (atomic_dec_and_test(&qp->refcount))
			wake_up(&qp->wait);
	}
}

static void update_sge(struct ipath_sge_state *ss, u32 length)
{
	struct ipath_sge *sge = &ss->sge;

	sge->vaddr += length;
	sge->length -= length;
	sge->sge_length -= length;
	if (sge->sge_length == 0) {
		if (--ss->num_sge)
			*sge = *ss->sg_list++;
	} else if (sge->length == 0 && sge->mr != NULL) {
		if (++sge->n >= IPATH_SEGSZ) {
			if (++sge->m >= sge->mr->mapsz)
				return;
			sge->n = 0;
		}
		sge->vaddr = sge->mr->map[sge->m]->segs[sge->n].vaddr;
		sge->length = sge->mr->map[sge->m]->segs[sge->n].length;
	}
}

#ifdef __LITTLE_ENDIAN
static inline u32 get_upper_bits(u32 data, u32 shift)
{
	return data >> shift;
}

static inline u32 set_upper_bits(u32 data, u32 shift)
{
	return data << shift;
}

static inline u32 clear_upper_bytes(u32 data, u32 n, u32 off)
{
	data <<= ((sizeof(u32) - n) * BITS_PER_BYTE);
	data >>= ((sizeof(u32) - n - off) * BITS_PER_BYTE);
	return data;
}
#else
static inline u32 get_upper_bits(u32 data, u32 shift)
{
	return data << shift;
}

static inline u32 set_upper_bits(u32 data, u32 shift)
{
	return data >> shift;
}

static inline u32 clear_upper_bytes(u32 data, u32 n, u32 off)
{
	data >>= ((sizeof(u32) - n) * BITS_PER_BYTE);
	data <<= ((sizeof(u32) - n - off) * BITS_PER_BYTE);
	return data;
}
#endif

static void copy_io(u32 __iomem *piobuf, struct ipath_sge_state *ss,
		    u32 length, unsigned flush_wc)
{
	u32 extra = 0;
	u32 data = 0;
	u32 last;

	while (1) {
		u32 len = ss->sge.length;
		u32 off;

		if (len > length)
			len = length;
		if (len > ss->sge.sge_length)
			len = ss->sge.sge_length;
		BUG_ON(len == 0);
		/* If the source address is not aligned, try to align it. */
		off = (unsigned long)ss->sge.vaddr & (sizeof(u32) - 1);
		if (off) {
			u32 *addr = (u32 *)((unsigned long)ss->sge.vaddr &
					    ~(sizeof(u32) - 1));
			u32 v = get_upper_bits(*addr, off * BITS_PER_BYTE);
			u32 y;

			y = sizeof(u32) - off;
			if (len > y)
				len = y;
			if (len + extra >= sizeof(u32)) {
				data |= set_upper_bits(v, extra *
						       BITS_PER_BYTE);
				len = sizeof(u32) - extra;
				if (len == length) {
					last = data;
					break;
				}
				__raw_writel(data, piobuf);
				piobuf++;
				extra = 0;
				data = 0;
			} else {
				/* Clear unused upper bytes */
				data |= clear_upper_bytes(v, len, extra);
				if (len == length) {
					last = data;
					break;
				}
				extra += len;
			}
		} else if (extra) {
			/* Source address is aligned. */
			u32 *addr = (u32 *) ss->sge.vaddr;
			int shift = extra * BITS_PER_BYTE;
			int ushift = 32 - shift;
			u32 l = len;

			while (l >= sizeof(u32)) {
				u32 v = *addr;

				data |= set_upper_bits(v, shift);
				__raw_writel(data, piobuf);
				data = get_upper_bits(v, ushift);
				piobuf++;
				addr++;
				l -= sizeof(u32);
			}
			/*
			 * We still have 'extra' number of bytes leftover.
			 */
			if (l) {
				u32 v = *addr;

				if (l + extra >= sizeof(u32)) {
					data |= set_upper_bits(v, shift);
					len -= l + extra - sizeof(u32);
					if (len == length) {
						last = data;
						break;
					}
					__raw_writel(data, piobuf);
					piobuf++;
					extra = 0;
					data = 0;
				} else {
					/* Clear unused upper bytes */
					data |= clear_upper_bytes(v, l,
								  extra);
					if (len == length) {
						last = data;
						break;
					}
					extra += l;
				}
			} else if (len == length) {
				last = data;
				break;
			}
		} else if (len == length) {
			u32 w;

			/*
			 * Need to round up for the last dword in the
			 * packet.
			 */
			w = (len + 3) >> 2;
			__iowrite32_copy(piobuf, ss->sge.vaddr, w - 1);
			piobuf += w - 1;
			last = ((u32 *) ss->sge.vaddr)[w - 1];
			break;
		} else {
			u32 w = len >> 2;

			__iowrite32_copy(piobuf, ss->sge.vaddr, w);
			piobuf += w;

			extra = len & (sizeof(u32) - 1);
			if (extra) {
				u32 v = ((u32 *) ss->sge.vaddr)[w];

				/* Clear unused upper bytes */
				data = clear_upper_bytes(v, extra, 0);
			}
		}
		update_sge(ss, len);
		length -= len;
	}
	/* Update address before sending packet. */
	update_sge(ss, length);
	if (flush_wc) {
		/* must flush early everything before trigger word */
		ipath_flush_wc();
		__raw_writel(last, piobuf);
		/* be sure trigger word is written */
		ipath_flush_wc();
	} else
		__raw_writel(last, piobuf);
}

static int ipath_verbs_send_pio(struct ipath_qp *qp, u32 *hdr, u32 hdrwords,
				struct ipath_sge_state *ss, u32 len,
				u32 plen, u32 dwords)
{
	struct ipath_devdata *dd = to_idev(qp->ibqp.device)->dd;
	u32 __iomem *piobuf;
	unsigned flush_wc;
	int ret;

	piobuf = ipath_getpiobuf(dd, NULL);
	if (unlikely(piobuf == NULL)) {
		ret = -EBUSY;
		goto bail;
	}

	/*
	 * Write len to control qword, no flags.
	 * We have to flush after the PBC for correctness on some cpus
	 * or WC buffer can be written out of order.
	 */
	writeq(plen, piobuf);
	piobuf += 2;

	flush_wc = dd->ipath_flags & IPATH_PIO_FLUSH_WC;
	if (len == 0) {
		/*
		 * If there is just the header portion, must flush before
		 * writing last word of header for correctness, and after
		 * the last header word (trigger word).
		 */
		if (flush_wc) {
			ipath_flush_wc();
			__iowrite32_copy(piobuf, hdr, hdrwords - 1);
			ipath_flush_wc();
			__raw_writel(hdr[hdrwords - 1], piobuf + hdrwords - 1);
			ipath_flush_wc();
		} else
			__iowrite32_copy(piobuf, hdr, hdrwords);
		goto done;
	}

	if (flush_wc)
		ipath_flush_wc();
	__iowrite32_copy(piobuf, hdr, hdrwords);
	piobuf += hdrwords;

	/* The common case is aligned and contained in one segment. */
	if (likely(ss->num_sge == 1 && len <= ss->sge.length &&
		   !((unsigned long)ss->sge.vaddr & (sizeof(u32) - 1)))) {
		u32 *addr = (u32 *) ss->sge.vaddr;

		/* Update address before sending packet. */
		update_sge(ss, len);
		if (flush_wc) {
			__iowrite32_copy(piobuf, addr, dwords - 1);
			/* must flush early everything before trigger word */
			ipath_flush_wc();
			__raw_writel(addr[dwords - 1], piobuf + dwords - 1);
			/* be sure trigger word is written */
			ipath_flush_wc();
		} else
			__iowrite32_copy(piobuf, addr, dwords);
		goto done;
	}
	copy_io(piobuf, ss, len, flush_wc);
done:
	if (qp->s_wqe)
		ipath_send_complete(qp, qp->s_wqe, IB_WC_SUCCESS);
	ret = 0;
bail:
	return ret;
}

/**
 * ipath_verbs_send - send a packet
 * @qp: the QP to send on
 * @hdr: the packet header
 * @hdrwords: the number of 32-bit words in the header
 * @ss: the SGE to send
 * @len: the length of the packet in bytes
 */
int ipath_verbs_send(struct ipath_qp *qp, struct ipath_ib_header *hdr,
		     u32 hdrwords, struct ipath_sge_state *ss, u32 len)
{
	struct ipath_devdata *dd = to_idev(qp->ibqp.device)->dd;
	u32 plen;
	int ret;
	u32 dwords = (len + 3) >> 2;

	/*
	 * Calculate the send buffer trigger address.
	 * The +1 counts for the pbc control dword following the pbc length.
	 */
	plen = hdrwords + dwords + 1;

	/* Drop non-VL15 packets if we are not in the active state */
	if (!(dd->ipath_flags & IPATH_LINKACTIVE) &&
	    qp->ibqp.qp_type != IB_QPT_SMI) {
		if (qp->s_wqe)
			ipath_send_complete(qp, qp->s_wqe, IB_WC_SUCCESS);
		ret = 0;
	} else
		ret = ipath_verbs_send_pio(qp, (u32 *) hdr, hdrwords,
					   ss, len, plen, dwords);

	return ret;
}

int ipath_snapshot_counters(struct ipath_devdata *dd, u64 *swords,
			    u64 *rwords, u64 *spkts, u64 *rpkts,
			    u64 *xmit_wait)
{
	int ret;

	if (!(dd->ipath_flags & IPATH_INITTED)) {
		/* no hardware, freeze, etc. */
		ret = -EINVAL;
		goto bail;
	}
	*swords = ipath_snap_cntr(dd, dd->ipath_cregs->cr_wordsendcnt);
	*rwords = ipath_snap_cntr(dd, dd->ipath_cregs->cr_wordrcvcnt);
	*spkts = ipath_snap_cntr(dd, dd->ipath_cregs->cr_pktsendcnt);
	*rpkts = ipath_snap_cntr(dd, dd->ipath_cregs->cr_pktrcvcnt);
	*xmit_wait = ipath_snap_cntr(dd, dd->ipath_cregs->cr_sendstallcnt);

	ret = 0;

bail:
	return ret;
}

/**
 * ipath_get_counters - get various chip counters
 * @dd: the infinipath device
 * @cntrs: counters are placed here
 *
 * Return the counters needed by recv_pma_get_portcounters().
 */
int ipath_get_counters(struct ipath_devdata *dd,
		       struct ipath_verbs_counters *cntrs)
{
	struct ipath_cregs const *crp = dd->ipath_cregs;
	int ret;

	if (!(dd->ipath_flags & IPATH_INITTED)) {
		/* no hardware, freeze, etc. */
		ret = -EINVAL;
		goto bail;
	}
	cntrs->symbol_error_counter =
		ipath_snap_cntr(dd, crp->cr_ibsymbolerrcnt);
	cntrs->link_error_recovery_counter =
		ipath_snap_cntr(dd, crp->cr_iblinkerrrecovcnt);
	/*
	 * The link downed counter counts when the other side downs the
	 * connection.  We add in the number of times we downed the link
	 * due to local link integrity errors to compensate.
	 */
	cntrs->link_downed_counter =
		ipath_snap_cntr(dd, crp->cr_iblinkdowncnt);
	cntrs->port_rcv_errors =
		ipath_snap_cntr(dd, crp->cr_rxdroppktcnt) +
		ipath_snap_cntr(dd, crp->cr_rcvovflcnt) +
		ipath_snap_cntr(dd, crp->cr_portovflcnt) +
		ipath_snap_cntr(dd, crp->cr_err_rlencnt) +
		ipath_snap_cntr(dd, crp->cr_invalidrlencnt) +
		ipath_snap_cntr(dd, crp->cr_errlinkcnt) +
		ipath_snap_cntr(dd, crp->cr_erricrccnt) +
		ipath_snap_cntr(dd, crp->cr_errvcrccnt) +
		ipath_snap_cntr(dd, crp->cr_errlpcrccnt) +
		ipath_snap_cntr(dd, crp->cr_badformatcnt) +
		dd->ipath_rxfc_unsupvl_errs;
	cntrs->port_rcv_remphys_errors =
		ipath_snap_cntr(dd, crp->cr_rcvebpcnt);
	cntrs->port_xmit_discards = ipath_snap_cntr(dd, crp->cr_unsupvlcnt);
	cntrs->port_xmit_data = ipath_snap_cntr(dd, crp->cr_wordsendcnt);
	cntrs->port_rcv_data = ipath_snap_cntr(dd, crp->cr_wordrcvcnt);
	cntrs->port_xmit_packets = ipath_snap_cntr(dd, crp->cr_pktsendcnt);
	cntrs->port_rcv_packets = ipath_snap_cntr(dd, crp->cr_pktrcvcnt);
	cntrs->local_link_integrity_errors =
		(dd->ipath_flags & IPATH_GPIO_ERRINTRS) ?
		dd->ipath_lli_errs : dd->ipath_lli_errors;
	cntrs->excessive_buffer_overrun_errors = dd->ipath_overrun_thresh_errs;

	ret = 0;

bail:
	return ret;
}

/**
 * ipath_ib_piobufavail - callback when a PIO buffer is available
 * @arg: the device pointer
 *
 * This is called from ipath_intr() at interrupt level when a PIO buffer is
 * available after ipath_verbs_send() returned an error that no buffers were
 * available.  Return 1 if we consumed all the PIO buffers and we still have
 * QPs waiting for buffers (for now, just do a tasklet_hi_schedule and
 * return zero).
 */
int ipath_ib_piobufavail(struct ipath_ibdev *dev)
{
	struct ipath_qp *qp;
	unsigned long flags;

	if (dev == NULL)
		goto bail;

	spin_lock_irqsave(&dev->pending_lock, flags);
	while (!list_empty(&dev->piowait)) {
		qp = list_entry(dev->piowait.next, struct ipath_qp,
				piowait);
		list_del_init(&qp->piowait);
		clear_bit(IPATH_S_BUSY, &qp->s_busy);
		tasklet_hi_schedule(&qp->s_task);
	}
	spin_unlock_irqrestore(&dev->pending_lock, flags);

bail:
	return 0;
}

static int ipath_query_device(struct ib_device *ibdev,
			      struct ib_device_attr *props)
{
	struct ipath_ibdev *dev = to_idev(ibdev);

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

	props->device_cap_flags = IB_DEVICE_BAD_PKEY_CNTR |
		IB_DEVICE_BAD_QKEY_CNTR | IB_DEVICE_SHUTDOWN_PORT |
		IB_DEVICE_SYS_IMAGE_GUID;
	props->page_size_cap = PAGE_SIZE;
	props->vendor_id = dev->dd->ipath_vendorid;
	props->vendor_part_id = dev->dd->ipath_deviceid;
	props->hw_ver = dev->dd->ipath_pcirev;

	props->sys_image_guid = dev->sys_image_guid;

	props->max_mr_size = ~0ull;
	props->max_qp = ib_ipath_max_qps;
	props->max_qp_wr = ib_ipath_max_qp_wrs;
	props->max_sge = ib_ipath_max_sges;
	props->max_cq = ib_ipath_max_cqs;
	props->max_ah = ib_ipath_max_ahs;
	props->max_cqe = ib_ipath_max_cqes;
	props->max_mr = dev->lk_table.max;
	props->max_fmr = dev->lk_table.max;
	props->max_map_per_fmr = 32767;
	props->max_pd = ib_ipath_max_pds;
	props->max_qp_rd_atom = IPATH_MAX_RDMA_ATOMIC;
	props->max_qp_init_rd_atom = 255;
	/* props->max_res_rd_atom */
	props->max_srq = ib_ipath_max_srqs;
	props->max_srq_wr = ib_ipath_max_srq_wrs;
	props->max_srq_sge = ib_ipath_max_srq_sges;
	/* props->local_ca_ack_delay */
	props->atomic_cap = IB_ATOMIC_GLOB;
	props->max_pkeys = ipath_get_npkeys(dev->dd);
	props->max_mcast_grp = ib_ipath_max_mcast_grps;
	props->max_mcast_qp_attach = ib_ipath_max_mcast_qp_attached;
	props->max_total_mcast_qp_attach = props->max_mcast_qp_attach *
		props->max_mcast_grp;

	return 0;
}

const u8 ipath_cvt_physportstate[32] = {
	[INFINIPATH_IBCS_LT_STATE_DISABLED] = IB_PHYSPORTSTATE_DISABLED,
	[INFINIPATH_IBCS_LT_STATE_LINKUP] = IB_PHYSPORTSTATE_LINKUP,
	[INFINIPATH_IBCS_LT_STATE_POLLACTIVE] = IB_PHYSPORTSTATE_POLL,
	[INFINIPATH_IBCS_LT_STATE_POLLQUIET] = IB_PHYSPORTSTATE_POLL,
	[INFINIPATH_IBCS_LT_STATE_SLEEPDELAY] = IB_PHYSPORTSTATE_SLEEP,
	[INFINIPATH_IBCS_LT_STATE_SLEEPQUIET] = IB_PHYSPORTSTATE_SLEEP,
	[INFINIPATH_IBCS_LT_STATE_CFGDEBOUNCE] =
		IB_PHYSPORTSTATE_CFG_TRAIN,
	[INFINIPATH_IBCS_LT_STATE_CFGRCVFCFG] =
		IB_PHYSPORTSTATE_CFG_TRAIN,
	[INFINIPATH_IBCS_LT_STATE_CFGWAITRMT] =
		IB_PHYSPORTSTATE_CFG_TRAIN,
	[INFINIPATH_IBCS_LT_STATE_CFGIDLE] = IB_PHYSPORTSTATE_CFG_TRAIN,
	[INFINIPATH_IBCS_LT_STATE_RECOVERRETRAIN] =
		IB_PHYSPORTSTATE_LINK_ERR_RECOVER,
	[INFINIPATH_IBCS_LT_STATE_RECOVERWAITRMT] =
		IB_PHYSPORTSTATE_LINK_ERR_RECOVER,
	[INFINIPATH_IBCS_LT_STATE_RECOVERIDLE] =
		IB_PHYSPORTSTATE_LINK_ERR_RECOVER,
	[0x10] = IB_PHYSPORTSTATE_CFG_TRAIN,
	[0x11] = IB_PHYSPORTSTATE_CFG_TRAIN,
	[0x12] = IB_PHYSPORTSTATE_CFG_TRAIN,
	[0x13] = IB_PHYSPORTSTATE_CFG_TRAIN,
	[0x14] = IB_PHYSPORTSTATE_CFG_TRAIN,
	[0x15] = IB_PHYSPORTSTATE_CFG_TRAIN,
	[0x16] = IB_PHYSPORTSTATE_CFG_TRAIN,
	[0x17] = IB_PHYSPORTSTATE_CFG_TRAIN
};

u32 ipath_get_cr_errpkey(struct ipath_devdata *dd)
{
	return ipath_read_creg32(dd, dd->ipath_cregs->cr_errpkey);
}

static int ipath_query_port(struct ib_device *ibdev,
			    u8 port, struct ib_port_attr *props)
{
	struct ipath_ibdev *dev = to_idev(ibdev);
	struct ipath_devdata *dd = dev->dd;
	enum ib_mtu mtu;
	u16 lid = dd->ipath_lid;
	u64 ibcstat;

	memset(props, 0, sizeof(*props));
	props->lid = lid ? lid : __constant_be16_to_cpu(IB_LID_PERMISSIVE);
	props->lmc = dd->ipath_lmc;
	props->sm_lid = dev->sm_lid;
	props->sm_sl = dev->sm_sl;
	ibcstat = dd->ipath_lastibcstat;
	props->state = ((ibcstat >> 4) & 0x3) + 1;
	/* See phys_state_show() */
	props->phys_state = /* MEA: assumes shift == 0 */
		ipath_cvt_physportstate[dd->ipath_lastibcstat &
		dd->ibcs_lts_mask];
	props->port_cap_flags = dev->port_cap_flags;
	props->gid_tbl_len = 1;
	props->max_msg_sz = 0x80000000;
	props->pkey_tbl_len = ipath_get_npkeys(dd);
	props->bad_pkey_cntr = ipath_get_cr_errpkey(dd) -
		dev->z_pkey_violations;
	props->qkey_viol_cntr = dev->qkey_violations;
	props->active_width = IB_WIDTH_4X;
	/* See rate_show() */
	props->active_speed = 1;	/* Regular 10Mbs speed. */
	props->max_vl_num = 1;		/* VLCap = VL0 */
	props->init_type_reply = 0;

	/*
	 * Note: the chip supports a maximum MTU of 4096, but the driver
	 * hasn't implemented this feature yet, so set the maximum value
	 * to 2048.
	 */
	props->max_mtu = IB_MTU_2048;
	switch (dd->ipath_ibmtu) {
	case 4096:
		mtu = IB_MTU_4096;
		break;
	case 2048:
		mtu = IB_MTU_2048;
		break;
	case 1024:
		mtu = IB_MTU_1024;
		break;
	case 512:
		mtu = IB_MTU_512;
		break;
	case 256:
		mtu = IB_MTU_256;
		break;
	default:
		mtu = IB_MTU_2048;
	}
	props->active_mtu = mtu;
	props->subnet_timeout = dev->subnet_timeout;

	return 0;
}

static int ipath_modify_device(struct ib_device *device,
			       int device_modify_mask,
			       struct ib_device_modify *device_modify)
{
	int ret;

	if (device_modify_mask & ~(IB_DEVICE_MODIFY_SYS_IMAGE_GUID |
				   IB_DEVICE_MODIFY_NODE_DESC)) {
		ret = -EOPNOTSUPP;
		goto bail;
	}

	if (device_modify_mask & IB_DEVICE_MODIFY_NODE_DESC)
		memcpy(device->node_desc, device_modify->node_desc, 64);

	if (device_modify_mask & IB_DEVICE_MODIFY_SYS_IMAGE_GUID)
		to_idev(device)->sys_image_guid =
			cpu_to_be64(device_modify->sys_image_guid);

	ret = 0;

bail:
	return ret;
}

static int ipath_modify_port(struct ib_device *ibdev,
			     u8 port, int port_modify_mask,
			     struct ib_port_modify *props)
{
	struct ipath_ibdev *dev = to_idev(ibdev);

	dev->port_cap_flags |= props->set_port_cap_mask;
	dev->port_cap_flags &= ~props->clr_port_cap_mask;
	if (port_modify_mask & IB_PORT_SHUTDOWN)
		ipath_set_linkstate(dev->dd, IPATH_IB_LINKDOWN);
	if (port_modify_mask & IB_PORT_RESET_QKEY_CNTR)
		dev->qkey_violations = 0;
	return 0;
}

static int ipath_query_gid(struct ib_device *ibdev, u8 port,
			   int index, union ib_gid *gid)
{
	struct ipath_ibdev *dev = to_idev(ibdev);
	int ret;

	if (index >= 1) {
		ret = -EINVAL;
		goto bail;
	}
	gid->global.subnet_prefix = dev->gid_prefix;
	gid->global.interface_id = dev->dd->ipath_guid;

	ret = 0;

bail:
	return ret;
}

static struct ib_pd *ipath_alloc_pd(struct ib_device *ibdev,
				    struct ib_ucontext *context,
				    struct ib_udata *udata)
{
	struct ipath_ibdev *dev = to_idev(ibdev);
	struct ipath_pd *pd;
	struct ib_pd *ret;

	/*
	 * This is actually totally arbitrary.	Some correctness tests
	 * assume there's a maximum number of PDs that can be allocated.
	 * We don't actually have this limit, but we fail the test if
	 * we allow allocations of more than we report for this value.
	 */

	pd = kmalloc(sizeof *pd, GFP_KERNEL);
	if (!pd) {
		ret = ERR_PTR(-ENOMEM);
		goto bail;
	}

	spin_lock(&dev->n_pds_lock);
	if (dev->n_pds_allocated == ib_ipath_max_pds) {
		spin_unlock(&dev->n_pds_lock);
		kfree(pd);
		ret = ERR_PTR(-ENOMEM);
		goto bail;
	}

	dev->n_pds_allocated++;
	spin_unlock(&dev->n_pds_lock);

	/* ib_alloc_pd() will initialize pd->ibpd. */
	pd->user = udata != NULL;

	ret = &pd->ibpd;

bail:
	return ret;
}

static int ipath_dealloc_pd(struct ib_pd *ibpd)
{
	struct ipath_pd *pd = to_ipd(ibpd);
	struct ipath_ibdev *dev = to_idev(ibpd->device);

	spin_lock(&dev->n_pds_lock);
	dev->n_pds_allocated--;
	spin_unlock(&dev->n_pds_lock);

	kfree(pd);

	return 0;
}

/**
 * ipath_create_ah - create an address handle
 * @pd: the protection domain
 * @ah_attr: the attributes of the AH
 *
 * This may be called from interrupt context.
 */
static struct ib_ah *ipath_create_ah(struct ib_pd *pd,
				     struct ib_ah_attr *ah_attr)
{
	struct ipath_ah *ah;
	struct ib_ah *ret;
	struct ipath_ibdev *dev = to_idev(pd->device);
	unsigned long flags;

	/* A multicast address requires a GRH (see ch. 8.4.1). */
	if (ah_attr->dlid >= IPATH_MULTICAST_LID_BASE &&
	    ah_attr->dlid != IPATH_PERMISSIVE_LID &&
	    !(ah_attr->ah_flags & IB_AH_GRH)) {
		ret = ERR_PTR(-EINVAL);
		goto bail;
	}

	if (ah_attr->dlid == 0) {
		ret = ERR_PTR(-EINVAL);
		goto bail;
	}

	if (ah_attr->port_num < 1 ||
	    ah_attr->port_num > pd->device->phys_port_cnt) {
		ret = ERR_PTR(-EINVAL);
		goto bail;
	}

	ah = kmalloc(sizeof *ah, GFP_ATOMIC);
	if (!ah) {
		ret = ERR_PTR(-ENOMEM);
		goto bail;
	}

	spin_lock_irqsave(&dev->n_ahs_lock, flags);
	if (dev->n_ahs_allocated == ib_ipath_max_ahs) {
		spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
		kfree(ah);
		ret = ERR_PTR(-ENOMEM);
		goto bail;
	}

	dev->n_ahs_allocated++;
	spin_unlock_irqrestore(&dev->n_ahs_lock, flags);

	/* ib_create_ah() will initialize ah->ibah. */
	ah->attr = *ah_attr;

	ret = &ah->ibah;

bail:
	return ret;
}

/**
 * ipath_destroy_ah - destroy an address handle
 * @ibah: the AH to destroy
 *
 * This may be called from interrupt context.
 */
static int ipath_destroy_ah(struct ib_ah *ibah)
{
	struct ipath_ibdev *dev = to_idev(ibah->device);
	struct ipath_ah *ah = to_iah(ibah);
	unsigned long flags;

	spin_lock_irqsave(&dev->n_ahs_lock, flags);
	dev->n_ahs_allocated--;
	spin_unlock_irqrestore(&dev->n_ahs_lock, flags);

	kfree(ah);

	return 0;
}

static int ipath_query_ah(struct ib_ah *ibah, struct ib_ah_attr *ah_attr)
{
	struct ipath_ah *ah = to_iah(ibah);

	*ah_attr = ah->attr;

	return 0;
}

/**
 * ipath_get_npkeys - return the size of the PKEY table for port 0
 * @dd: the infinipath device
 */
unsigned ipath_get_npkeys(struct ipath_devdata *dd)
{
	return ARRAY_SIZE(dd->ipath_pd[0]->port_pkeys);
}

/**
 * ipath_get_pkey - return the indexed PKEY from the port 0 PKEY table
 * @dd: the infinipath device
 * @index: the PKEY index
 */
unsigned ipath_get_pkey(struct ipath_devdata *dd, unsigned index)
{
	unsigned ret;

	if (index >= ARRAY_SIZE(dd->ipath_pd[0]->port_pkeys))
		ret = 0;
	else
		ret = dd->ipath_pd[0]->port_pkeys[index];

	return ret;
}

static int ipath_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
			    u16 *pkey)
{
	struct ipath_ibdev *dev = to_idev(ibdev);
	int ret;

	if (index >= ipath_get_npkeys(dev->dd)) {
		ret = -EINVAL;
		goto bail;
	}

	*pkey = ipath_get_pkey(dev->dd, index);
	ret = 0;

bail:
	return ret;
}

/**
 * ipath_alloc_ucontext - allocate a ucontest
 * @ibdev: the infiniband device
 * @udata: not used by the InfiniPath driver
 */

static struct ib_ucontext *ipath_alloc_ucontext(struct ib_device *ibdev,
						struct ib_udata *udata)
{
	struct ipath_ucontext *context;
	struct ib_ucontext *ret;

	context = kmalloc(sizeof *context, GFP_KERNEL);
	if (!context) {
		ret = ERR_PTR(-ENOMEM);
		goto bail;
	}

	ret = &context->ibucontext;

bail:
	return ret;
}

static int ipath_dealloc_ucontext(struct ib_ucontext *context)
{
	kfree(to_iucontext(context));
	return 0;
}

static int ipath_verbs_register_sysfs(struct ib_device *dev);

static void __verbs_timer(unsigned long arg)
{
	struct ipath_devdata *dd = (struct ipath_devdata *) arg;

	/* Handle verbs layer timeouts. */
	ipath_ib_timer(dd->verbs_dev);

	mod_timer(&dd->verbs_timer, jiffies + 1);
}

static int enable_timer(struct ipath_devdata *dd)
{
	/*
	 * Early chips had a design flaw where the chip and kernel idea
	 * of the tail register don't always agree, and therefore we won't
	 * get an interrupt on the next packet received.
	 * If the board supports per packet receive interrupts, use it.
	 * Otherwise, the timer function periodically checks for packets
	 * to cover this case.
	 * Either way, the timer is needed for verbs layer related
	 * processing.
	 */
	if (dd->ipath_flags & IPATH_GPIO_INTR) {
		ipath_write_kreg(dd, dd->ipath_kregs->kr_debugportselect,
				 0x2074076542310ULL);
		/* Enable GPIO bit 2 interrupt */
		dd->ipath_gpio_mask |= (u64) (1 << IPATH_GPIO_PORT0_BIT);
		ipath_write_kreg(dd, dd->ipath_kregs->kr_gpio_mask,
				 dd->ipath_gpio_mask);
	}

	init_timer(&dd->verbs_timer);
	dd->verbs_timer.function = __verbs_timer;
	dd->verbs_timer.data = (unsigned long)dd;
	dd->verbs_timer.expires = jiffies + 1;
	add_timer(&dd->verbs_timer);

	return 0;
}

static int disable_timer(struct ipath_devdata *dd)
{
	/* Disable GPIO bit 2 interrupt */
	if (dd->ipath_flags & IPATH_GPIO_INTR) {
                /* Disable GPIO bit 2 interrupt */
		dd->ipath_gpio_mask &= ~((u64) (1 << IPATH_GPIO_PORT0_BIT));
		ipath_write_kreg(dd, dd->ipath_kregs->kr_gpio_mask,
				 dd->ipath_gpio_mask);
		/*
		 * We might want to undo changes to debugportselect,
		 * but how?
		 */
	}

	del_timer_sync(&dd->verbs_timer);

	return 0;
}

/**
 * ipath_register_ib_device - register our device with the infiniband core
 * @dd: the device data structure
 * Return the allocated ipath_ibdev pointer or NULL on error.
 */
int ipath_register_ib_device(struct ipath_devdata *dd)
{
	struct ipath_verbs_counters cntrs;
	struct ipath_ibdev *idev;
	struct ib_device *dev;
	int ret;

	idev = (struct ipath_ibdev *)ib_alloc_device(sizeof *idev);
	if (idev == NULL) {
		ret = -ENOMEM;
		goto bail;
	}

	dev = &idev->ibdev;

	/* Only need to initialize non-zero fields. */
	spin_lock_init(&idev->n_pds_lock);
	spin_lock_init(&idev->n_ahs_lock);
	spin_lock_init(&idev->n_cqs_lock);
	spin_lock_init(&idev->n_qps_lock);
	spin_lock_init(&idev->n_srqs_lock);
	spin_lock_init(&idev->n_mcast_grps_lock);

	spin_lock_init(&idev->qp_table.lock);
	spin_lock_init(&idev->lk_table.lock);
	idev->sm_lid = __constant_be16_to_cpu(IB_LID_PERMISSIVE);
	/* Set the prefix to the default value (see ch. 4.1.1) */
	idev->gid_prefix = __constant_cpu_to_be64(0xfe80000000000000ULL);

	ret = ipath_init_qp_table(idev, ib_ipath_qp_table_size);
	if (ret)
		goto err_qp;

	/*
	 * The top ib_ipath_lkey_table_size bits are used to index the
	 * table.  The lower 8 bits can be owned by the user (copied from
	 * the LKEY).  The remaining bits act as a generation number or tag.
	 */
	idev->lk_table.max = 1 << ib_ipath_lkey_table_size;
	idev->lk_table.table = kzalloc(idev->lk_table.max *
				       sizeof(*idev->lk_table.table),
				       GFP_KERNEL);
	if (idev->lk_table.table == NULL) {
		ret = -ENOMEM;
		goto err_lk;
	}
	INIT_LIST_HEAD(&idev->pending_mmaps);
	spin_lock_init(&idev->pending_lock);
	idev->mmap_offset = PAGE_SIZE;
	spin_lock_init(&idev->mmap_offset_lock);
	INIT_LIST_HEAD(&idev->pending[0]);
	INIT_LIST_HEAD(&idev->pending[1]);
	INIT_LIST_HEAD(&idev->pending[2]);
	INIT_LIST_HEAD(&idev->piowait);
	INIT_LIST_HEAD(&idev->rnrwait);
	idev->pending_index = 0;
	idev->port_cap_flags =
		IB_PORT_SYS_IMAGE_GUID_SUP | IB_PORT_CLIENT_REG_SUP;
	idev->pma_counter_select[0] = IB_PMA_PORT_XMIT_DATA;
	idev->pma_counter_select[1] = IB_PMA_PORT_RCV_DATA;
	idev->pma_counter_select[2] = IB_PMA_PORT_XMIT_PKTS;
	idev->pma_counter_select[3] = IB_PMA_PORT_RCV_PKTS;
	idev->pma_counter_select[4] = IB_PMA_PORT_XMIT_WAIT;
	idev->link_width_enabled = 3;	/* 1x or 4x */

	/* Snapshot current HW counters to "clear" them. */
	ipath_get_counters(dd, &cntrs);
	idev->z_symbol_error_counter = cntrs.symbol_error_counter;
	idev->z_link_error_recovery_counter =
		cntrs.link_error_recovery_counter;
	idev->z_link_downed_counter = cntrs.link_downed_counter;
	idev->z_port_rcv_errors = cntrs.port_rcv_errors;
	idev->z_port_rcv_remphys_errors =
		cntrs.port_rcv_remphys_errors;
	idev->z_port_xmit_discards = cntrs.port_xmit_discards;
	idev->z_port_xmit_data = cntrs.port_xmit_data;
	idev->z_port_rcv_data = cntrs.port_rcv_data;
	idev->z_port_xmit_packets = cntrs.port_xmit_packets;
	idev->z_port_rcv_packets = cntrs.port_rcv_packets;
	idev->z_local_link_integrity_errors =
		cntrs.local_link_integrity_errors;
	idev->z_excessive_buffer_overrun_errors =
		cntrs.excessive_buffer_overrun_errors;
	idev->z_vl15_dropped = cntrs.vl15_dropped;

	/*
	 * The system image GUID is supposed to be the same for all
	 * IB HCAs in a single system but since there can be other
	 * device types in the system, we can't be sure this is unique.
	 */
	if (!sys_image_guid)
		sys_image_guid = dd->ipath_guid;
	idev->sys_image_guid = sys_image_guid;
	idev->ib_unit = dd->ipath_unit;
	idev->dd = dd;

	strlcpy(dev->name, "ipath%d", IB_DEVICE_NAME_MAX);
	dev->owner = THIS_MODULE;
	dev->node_guid = dd->ipath_guid;
	dev->uverbs_abi_ver = IPATH_UVERBS_ABI_VERSION;
	dev->uverbs_cmd_mask =
		(1ull << IB_USER_VERBS_CMD_GET_CONTEXT)		|
		(1ull << IB_USER_VERBS_CMD_QUERY_DEVICE)	|
		(1ull << IB_USER_VERBS_CMD_QUERY_PORT)		|
		(1ull << IB_USER_VERBS_CMD_ALLOC_PD)		|
		(1ull << IB_USER_VERBS_CMD_DEALLOC_PD)		|
		(1ull << IB_USER_VERBS_CMD_CREATE_AH)		|
		(1ull << IB_USER_VERBS_CMD_DESTROY_AH)		|
		(1ull << IB_USER_VERBS_CMD_QUERY_AH)		|
		(1ull << IB_USER_VERBS_CMD_REG_MR)		|
		(1ull << IB_USER_VERBS_CMD_DEREG_MR)		|
		(1ull << IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL) |
		(1ull << IB_USER_VERBS_CMD_CREATE_CQ)		|
		(1ull << IB_USER_VERBS_CMD_RESIZE_CQ)		|
		(1ull << IB_USER_VERBS_CMD_DESTROY_CQ)		|
		(1ull << IB_USER_VERBS_CMD_POLL_CQ)		|
		(1ull << IB_USER_VERBS_CMD_REQ_NOTIFY_CQ)	|
		(1ull << IB_USER_VERBS_CMD_CREATE_QP)		|
		(1ull << IB_USER_VERBS_CMD_QUERY_QP)		|
		(1ull << IB_USER_VERBS_CMD_MODIFY_QP)		|
		(1ull << IB_USER_VERBS_CMD_DESTROY_QP)		|
		(1ull << IB_USER_VERBS_CMD_POST_SEND)		|
		(1ull << IB_USER_VERBS_CMD_POST_RECV)		|
		(1ull << IB_USER_VERBS_CMD_ATTACH_MCAST)	|
		(1ull << IB_USER_VERBS_CMD_DETACH_MCAST)	|
		(1ull << IB_USER_VERBS_CMD_CREATE_SRQ)		|
		(1ull << IB_USER_VERBS_CMD_MODIFY_SRQ)		|
		(1ull << IB_USER_VERBS_CMD_QUERY_SRQ)		|
		(1ull << IB_USER_VERBS_CMD_DESTROY_SRQ)		|
		(1ull << IB_USER_VERBS_CMD_POST_SRQ_RECV);
	dev->node_type = RDMA_NODE_IB_CA;
	dev->phys_port_cnt = 1;
	dev->num_comp_vectors = 1;
	dev->dma_device = &dd->pcidev->dev;
	dev->query_device = ipath_query_device;
	dev->modify_device = ipath_modify_device;
	dev->query_port = ipath_query_port;
	dev->modify_port = ipath_modify_port;
	dev->query_pkey = ipath_query_pkey;
	dev->query_gid = ipath_query_gid;
	dev->alloc_ucontext = ipath_alloc_ucontext;
	dev->dealloc_ucontext = ipath_dealloc_ucontext;
	dev->alloc_pd = ipath_alloc_pd;
	dev->dealloc_pd = ipath_dealloc_pd;
	dev->create_ah = ipath_create_ah;
	dev->destroy_ah = ipath_destroy_ah;
	dev->query_ah = ipath_query_ah;
	dev->create_srq = ipath_create_srq;
	dev->modify_srq = ipath_modify_srq;
	dev->query_srq = ipath_query_srq;
	dev->destroy_srq = ipath_destroy_srq;
	dev->create_qp = ipath_create_qp;
	dev->modify_qp = ipath_modify_qp;
	dev->query_qp = ipath_query_qp;
	dev->destroy_qp = ipath_destroy_qp;
	dev->post_send = ipath_post_send;
	dev->post_recv = ipath_post_receive;
	dev->post_srq_recv = ipath_post_srq_receive;
	dev->create_cq = ipath_create_cq;
	dev->destroy_cq = ipath_destroy_cq;
	dev->resize_cq = ipath_resize_cq;
	dev->poll_cq = ipath_poll_cq;
	dev->req_notify_cq = ipath_req_notify_cq;
	dev->get_dma_mr = ipath_get_dma_mr;
	dev->reg_phys_mr = ipath_reg_phys_mr;
	dev->reg_user_mr = ipath_reg_user_mr;
	dev->dereg_mr = ipath_dereg_mr;
	dev->alloc_fmr = ipath_alloc_fmr;
	dev->map_phys_fmr = ipath_map_phys_fmr;
	dev->unmap_fmr = ipath_unmap_fmr;
	dev->dealloc_fmr = ipath_dealloc_fmr;
	dev->attach_mcast = ipath_multicast_attach;
	dev->detach_mcast = ipath_multicast_detach;
	dev->process_mad = ipath_process_mad;
	dev->mmap = ipath_mmap;
	dev->dma_ops = &ipath_dma_mapping_ops;

	snprintf(dev->node_desc, sizeof(dev->node_desc),
		 IPATH_IDSTR " %s", init_utsname()->nodename);

	ret = ib_register_device(dev);
	if (ret)
		goto err_reg;

	if (ipath_verbs_register_sysfs(dev))
		goto err_class;

	enable_timer(dd);

	goto bail;

err_class:
	ib_unregister_device(dev);
err_reg:
	kfree(idev->lk_table.table);
err_lk:
	kfree(idev->qp_table.table);
err_qp:
	ib_dealloc_device(dev);
	ipath_dev_err(dd, "cannot register verbs: %d!\n", -ret);
	idev = NULL;

bail:
	dd->verbs_dev = idev;
	return ret;
}

void ipath_unregister_ib_device(struct ipath_ibdev *dev)
{
	struct ib_device *ibdev = &dev->ibdev;

	disable_timer(dev->dd);

	ib_unregister_device(ibdev);

	if (!list_empty(&dev->pending[0]) ||
	    !list_empty(&dev->pending[1]) ||
	    !list_empty(&dev->pending[2]))
		ipath_dev_err(dev->dd, "pending list not empty!\n");
	if (!list_empty(&dev->piowait))
		ipath_dev_err(dev->dd, "piowait list not empty!\n");
	if (!list_empty(&dev->rnrwait))
		ipath_dev_err(dev->dd, "rnrwait list not empty!\n");
	if (!ipath_mcast_tree_empty())
		ipath_dev_err(dev->dd, "multicast table memory leak!\n");
	/*
	 * Note that ipath_unregister_ib_device() can be called before all
	 * the QPs are destroyed!
	 */
	ipath_free_all_qps(&dev->qp_table);
	kfree(dev->qp_table.table);
	kfree(dev->lk_table.table);
	ib_dealloc_device(ibdev);
}

static ssize_t show_rev(struct class_device *cdev, char *buf)
{
	struct ipath_ibdev *dev =
		container_of(cdev, struct ipath_ibdev, ibdev.class_dev);

	return sprintf(buf, "%x\n", dev->dd->ipath_pcirev);
}

static ssize_t show_hca(struct class_device *cdev, char *buf)
{
	struct ipath_ibdev *dev =
		container_of(cdev, struct ipath_ibdev, ibdev.class_dev);
	int ret;

	ret = dev->dd->ipath_f_get_boardname(dev->dd, buf, 128);
	if (ret < 0)
		goto bail;
	strcat(buf, "\n");
	ret = strlen(buf);

bail:
	return ret;
}

static ssize_t show_stats(struct class_device *cdev, char *buf)
{
	struct ipath_ibdev *dev =
		container_of(cdev, struct ipath_ibdev, ibdev.class_dev);
	int i;
	int len;

	len = sprintf(buf,
		      "RC resends  %d\n"
		      "RC no QACK  %d\n"
		      "RC ACKs     %d\n"
		      "RC SEQ NAKs %d\n"
		      "RC RDMA seq %d\n"
		      "RC RNR NAKs %d\n"
		      "RC OTH NAKs %d\n"
		      "RC timeouts %d\n"
		      "RC RDMA dup %d\n"
		      "RC stalls   %d\n"
		      "piobuf wait %d\n"
		      "no piobuf   %d\n"
		      "PKT drops   %d\n"
		      "WQE errs    %d\n",
		      dev->n_rc_resends, dev->n_rc_qacks, dev->n_rc_acks,
		      dev->n_seq_naks, dev->n_rdma_seq, dev->n_rnr_naks,
		      dev->n_other_naks, dev->n_timeouts,
		      dev->n_rdma_dup_busy, dev->n_rc_stalls, dev->n_piowait,
		      dev->n_no_piobuf, dev->n_pkt_drops, dev->n_wqe_errs);
	for (i = 0; i < ARRAY_SIZE(dev->opstats); i++) {
		const struct ipath_opcode_stats *si = &dev->opstats[i];

		if (!si->n_packets && !si->n_bytes)
			continue;
		len += sprintf(buf + len, "%02x %llu/%llu\n", i,
			       (unsigned long long) si->n_packets,
			       (unsigned long long) si->n_bytes);
	}
	return len;
}

static CLASS_DEVICE_ATTR(hw_rev, S_IRUGO, show_rev, NULL);
static CLASS_DEVICE_ATTR(hca_type, S_IRUGO, show_hca, NULL);
static CLASS_DEVICE_ATTR(board_id, S_IRUGO, show_hca, NULL);
static CLASS_DEVICE_ATTR(stats, S_IRUGO, show_stats, NULL);

static struct class_device_attribute *ipath_class_attributes[] = {
	&class_device_attr_hw_rev,
	&class_device_attr_hca_type,
	&class_device_attr_board_id,
	&class_device_attr_stats
};

static int ipath_verbs_register_sysfs(struct ib_device *dev)
{
	int i;
	int ret;

	for (i = 0; i < ARRAY_SIZE(ipath_class_attributes); ++i)
		if (class_device_create_file(&dev->class_dev,
					     ipath_class_attributes[i])) {
			ret = 1;
			goto bail;
		}

	ret = 0;

bail:
	return ret;
}