summaryrefslogtreecommitdiff
path: root/net/sunrpc/xprtrdma/verbs.c
blob: 3f86d039875c9778b1fa7a29f019ae03e3f07a61 (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
// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause
/*
 * Copyright (c) 2014-2017 Oracle.  All rights reserved.
 * Copyright (c) 2003-2007 Network Appliance, 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 BSD-type
 * 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.
 *
 *      Neither the name of the Network Appliance, Inc. nor the names of
 *      its contributors may be used to endorse or promote products
 *      derived from this software without specific prior written
 *      permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */

/*
 * verbs.c
 *
 * Encapsulates the major functions managing:
 *  o adapters
 *  o endpoints
 *  o connections
 *  o buffer memory
 */

#include <linux/interrupt.h>
#include <linux/slab.h>
#include <linux/sunrpc/addr.h>
#include <linux/sunrpc/svc_rdma.h>
#include <linux/log2.h>

#include <asm-generic/barrier.h>
#include <asm/bitops.h>

#include <rdma/ib_cm.h>

#include "xprt_rdma.h"
#include <trace/events/rpcrdma.h>

/*
 * Globals/Macros
 */

#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
# define RPCDBG_FACILITY	RPCDBG_TRANS
#endif

/*
 * internal functions
 */
static int rpcrdma_sendctxs_create(struct rpcrdma_xprt *r_xprt);
static void rpcrdma_sendctxs_destroy(struct rpcrdma_xprt *r_xprt);
static void rpcrdma_sendctx_put_locked(struct rpcrdma_xprt *r_xprt,
				       struct rpcrdma_sendctx *sc);
static int rpcrdma_reqs_setup(struct rpcrdma_xprt *r_xprt);
static void rpcrdma_reqs_reset(struct rpcrdma_xprt *r_xprt);
static void rpcrdma_rep_destroy(struct rpcrdma_rep *rep);
static void rpcrdma_reps_unmap(struct rpcrdma_xprt *r_xprt);
static void rpcrdma_mrs_create(struct rpcrdma_xprt *r_xprt);
static void rpcrdma_mrs_destroy(struct rpcrdma_xprt *r_xprt);
static void rpcrdma_ep_get(struct rpcrdma_ep *ep);
static int rpcrdma_ep_put(struct rpcrdma_ep *ep);
static struct rpcrdma_regbuf *
rpcrdma_regbuf_alloc(size_t size, enum dma_data_direction direction,
		     gfp_t flags);
static void rpcrdma_regbuf_dma_unmap(struct rpcrdma_regbuf *rb);
static void rpcrdma_regbuf_free(struct rpcrdma_regbuf *rb);

/* Wait for outstanding transport work to finish. ib_drain_qp
 * handles the drains in the wrong order for us, so open code
 * them here.
 */
static void rpcrdma_xprt_drain(struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_ep *ep = r_xprt->rx_ep;
	struct rdma_cm_id *id = ep->re_id;

	/* Flush Receives, then wait for deferred Reply work
	 * to complete.
	 */
	ib_drain_rq(id->qp);

	/* Deferred Reply processing might have scheduled
	 * local invalidations.
	 */
	ib_drain_sq(id->qp);

	rpcrdma_ep_put(ep);
}

/**
 * rpcrdma_qp_event_handler - Handle one QP event (error notification)
 * @event: details of the event
 * @context: ep that owns QP where event occurred
 *
 * Called from the RDMA provider (device driver) possibly in an interrupt
 * context. The QP is always destroyed before the ID, so the ID will be
 * reliably available when this handler is invoked.
 */
static void rpcrdma_qp_event_handler(struct ib_event *event, void *context)
{
	struct rpcrdma_ep *ep = context;

	trace_xprtrdma_qp_event(ep, event);
}

/* Ensure xprt_force_disconnect() is invoked exactly once when a
 * connection is closed or lost. (The important thing is it needs
 * to be invoked "at least" once).
 */
static void rpcrdma_force_disconnect(struct rpcrdma_ep *ep)
{
	if (atomic_add_unless(&ep->re_force_disconnect, 1, 1))
		xprt_force_disconnect(ep->re_xprt);
}

/**
 * rpcrdma_flush_disconnect - Disconnect on flushed completion
 * @r_xprt: transport to disconnect
 * @wc: work completion entry
 *
 * Must be called in process context.
 */
void rpcrdma_flush_disconnect(struct rpcrdma_xprt *r_xprt, struct ib_wc *wc)
{
	if (wc->status != IB_WC_SUCCESS)
		rpcrdma_force_disconnect(r_xprt->rx_ep);
}

/**
 * rpcrdma_wc_send - Invoked by RDMA provider for each polled Send WC
 * @cq:	completion queue
 * @wc:	WCE for a completed Send WR
 *
 */
static void rpcrdma_wc_send(struct ib_cq *cq, struct ib_wc *wc)
{
	struct ib_cqe *cqe = wc->wr_cqe;
	struct rpcrdma_sendctx *sc =
		container_of(cqe, struct rpcrdma_sendctx, sc_cqe);
	struct rpcrdma_xprt *r_xprt = cq->cq_context;

	/* WARNING: Only wr_cqe and status are reliable at this point */
	trace_xprtrdma_wc_send(sc, wc);
	rpcrdma_sendctx_put_locked(r_xprt, sc);
	rpcrdma_flush_disconnect(r_xprt, wc);
}

/**
 * rpcrdma_wc_receive - Invoked by RDMA provider for each polled Receive WC
 * @cq:	completion queue
 * @wc:	WCE for a completed Receive WR
 *
 */
static void rpcrdma_wc_receive(struct ib_cq *cq, struct ib_wc *wc)
{
	struct ib_cqe *cqe = wc->wr_cqe;
	struct rpcrdma_rep *rep = container_of(cqe, struct rpcrdma_rep,
					       rr_cqe);
	struct rpcrdma_xprt *r_xprt = cq->cq_context;

	/* WARNING: Only wr_cqe and status are reliable at this point */
	trace_xprtrdma_wc_receive(wc);
	--r_xprt->rx_ep->re_receive_count;
	if (wc->status != IB_WC_SUCCESS)
		goto out_flushed;

	/* status == SUCCESS means all fields in wc are trustworthy */
	rpcrdma_set_xdrlen(&rep->rr_hdrbuf, wc->byte_len);
	rep->rr_wc_flags = wc->wc_flags;
	rep->rr_inv_rkey = wc->ex.invalidate_rkey;

	ib_dma_sync_single_for_cpu(rdmab_device(rep->rr_rdmabuf),
				   rdmab_addr(rep->rr_rdmabuf),
				   wc->byte_len, DMA_FROM_DEVICE);

	rpcrdma_reply_handler(rep);
	return;

out_flushed:
	rpcrdma_flush_disconnect(r_xprt, wc);
	rpcrdma_rep_destroy(rep);
}

static void rpcrdma_update_cm_private(struct rpcrdma_ep *ep,
				      struct rdma_conn_param *param)
{
	const struct rpcrdma_connect_private *pmsg = param->private_data;
	unsigned int rsize, wsize;

	/* Default settings for RPC-over-RDMA Version One */
	ep->re_implicit_roundup = xprt_rdma_pad_optimize;
	rsize = RPCRDMA_V1_DEF_INLINE_SIZE;
	wsize = RPCRDMA_V1_DEF_INLINE_SIZE;

	if (pmsg &&
	    pmsg->cp_magic == rpcrdma_cmp_magic &&
	    pmsg->cp_version == RPCRDMA_CMP_VERSION) {
		ep->re_implicit_roundup = true;
		rsize = rpcrdma_decode_buffer_size(pmsg->cp_send_size);
		wsize = rpcrdma_decode_buffer_size(pmsg->cp_recv_size);
	}

	if (rsize < ep->re_inline_recv)
		ep->re_inline_recv = rsize;
	if (wsize < ep->re_inline_send)
		ep->re_inline_send = wsize;

	rpcrdma_set_max_header_sizes(ep);
}

/**
 * rpcrdma_cm_event_handler - Handle RDMA CM events
 * @id: rdma_cm_id on which an event has occurred
 * @event: details of the event
 *
 * Called with @id's mutex held. Returns 1 if caller should
 * destroy @id, otherwise 0.
 */
static int
rpcrdma_cm_event_handler(struct rdma_cm_id *id, struct rdma_cm_event *event)
{
	struct sockaddr *sap = (struct sockaddr *)&id->route.addr.dst_addr;
	struct rpcrdma_ep *ep = id->context;

	might_sleep();

	switch (event->event) {
	case RDMA_CM_EVENT_ADDR_RESOLVED:
	case RDMA_CM_EVENT_ROUTE_RESOLVED:
		ep->re_async_rc = 0;
		complete(&ep->re_done);
		return 0;
	case RDMA_CM_EVENT_ADDR_ERROR:
		ep->re_async_rc = -EPROTO;
		complete(&ep->re_done);
		return 0;
	case RDMA_CM_EVENT_ROUTE_ERROR:
		ep->re_async_rc = -ENETUNREACH;
		complete(&ep->re_done);
		return 0;
	case RDMA_CM_EVENT_DEVICE_REMOVAL:
		pr_info("rpcrdma: removing device %s for %pISpc\n",
			ep->re_id->device->name, sap);
		fallthrough;
	case RDMA_CM_EVENT_ADDR_CHANGE:
		ep->re_connect_status = -ENODEV;
		goto disconnected;
	case RDMA_CM_EVENT_ESTABLISHED:
		rpcrdma_ep_get(ep);
		ep->re_connect_status = 1;
		rpcrdma_update_cm_private(ep, &event->param.conn);
		trace_xprtrdma_inline_thresh(ep);
		wake_up_all(&ep->re_connect_wait);
		break;
	case RDMA_CM_EVENT_CONNECT_ERROR:
		ep->re_connect_status = -ENOTCONN;
		goto wake_connect_worker;
	case RDMA_CM_EVENT_UNREACHABLE:
		ep->re_connect_status = -ENETUNREACH;
		goto wake_connect_worker;
	case RDMA_CM_EVENT_REJECTED:
		dprintk("rpcrdma: connection to %pISpc rejected: %s\n",
			sap, rdma_reject_msg(id, event->status));
		ep->re_connect_status = -ECONNREFUSED;
		if (event->status == IB_CM_REJ_STALE_CONN)
			ep->re_connect_status = -ENOTCONN;
wake_connect_worker:
		wake_up_all(&ep->re_connect_wait);
		return 0;
	case RDMA_CM_EVENT_DISCONNECTED:
		ep->re_connect_status = -ECONNABORTED;
disconnected:
		rpcrdma_force_disconnect(ep);
		return rpcrdma_ep_put(ep);
	default:
		break;
	}

	dprintk("RPC:       %s: %pISpc on %s/frwr: %s\n", __func__, sap,
		ep->re_id->device->name, rdma_event_msg(event->event));
	return 0;
}

static struct rdma_cm_id *rpcrdma_create_id(struct rpcrdma_xprt *r_xprt,
					    struct rpcrdma_ep *ep)
{
	unsigned long wtimeout = msecs_to_jiffies(RDMA_RESOLVE_TIMEOUT) + 1;
	struct rpc_xprt *xprt = &r_xprt->rx_xprt;
	struct rdma_cm_id *id;
	int rc;

	init_completion(&ep->re_done);

	id = rdma_create_id(xprt->xprt_net, rpcrdma_cm_event_handler, ep,
			    RDMA_PS_TCP, IB_QPT_RC);
	if (IS_ERR(id))
		return id;

	ep->re_async_rc = -ETIMEDOUT;
	rc = rdma_resolve_addr(id, NULL, (struct sockaddr *)&xprt->addr,
			       RDMA_RESOLVE_TIMEOUT);
	if (rc)
		goto out;
	rc = wait_for_completion_interruptible_timeout(&ep->re_done, wtimeout);
	if (rc < 0)
		goto out;

	rc = ep->re_async_rc;
	if (rc)
		goto out;

	ep->re_async_rc = -ETIMEDOUT;
	rc = rdma_resolve_route(id, RDMA_RESOLVE_TIMEOUT);
	if (rc)
		goto out;
	rc = wait_for_completion_interruptible_timeout(&ep->re_done, wtimeout);
	if (rc < 0)
		goto out;
	rc = ep->re_async_rc;
	if (rc)
		goto out;

	return id;

out:
	rdma_destroy_id(id);
	return ERR_PTR(rc);
}

static void rpcrdma_ep_destroy(struct kref *kref)
{
	struct rpcrdma_ep *ep = container_of(kref, struct rpcrdma_ep, re_kref);

	if (ep->re_id->qp) {
		rdma_destroy_qp(ep->re_id);
		ep->re_id->qp = NULL;
	}

	if (ep->re_attr.recv_cq)
		ib_free_cq(ep->re_attr.recv_cq);
	ep->re_attr.recv_cq = NULL;
	if (ep->re_attr.send_cq)
		ib_free_cq(ep->re_attr.send_cq);
	ep->re_attr.send_cq = NULL;

	if (ep->re_pd)
		ib_dealloc_pd(ep->re_pd);
	ep->re_pd = NULL;

	kfree(ep);
	module_put(THIS_MODULE);
}

static noinline void rpcrdma_ep_get(struct rpcrdma_ep *ep)
{
	kref_get(&ep->re_kref);
}

/* Returns:
 *     %0 if @ep still has a positive kref count, or
 *     %1 if @ep was destroyed successfully.
 */
static noinline int rpcrdma_ep_put(struct rpcrdma_ep *ep)
{
	return kref_put(&ep->re_kref, rpcrdma_ep_destroy);
}

static int rpcrdma_ep_create(struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_connect_private *pmsg;
	struct ib_device *device;
	struct rdma_cm_id *id;
	struct rpcrdma_ep *ep;
	int rc;

	ep = kzalloc(sizeof(*ep), GFP_NOFS);
	if (!ep)
		return -ENOTCONN;
	ep->re_xprt = &r_xprt->rx_xprt;
	kref_init(&ep->re_kref);

	id = rpcrdma_create_id(r_xprt, ep);
	if (IS_ERR(id)) {
		kfree(ep);
		return PTR_ERR(id);
	}
	__module_get(THIS_MODULE);
	device = id->device;
	ep->re_id = id;

	ep->re_max_requests = r_xprt->rx_xprt.max_reqs;
	ep->re_inline_send = xprt_rdma_max_inline_write;
	ep->re_inline_recv = xprt_rdma_max_inline_read;
	rc = frwr_query_device(ep, device);
	if (rc)
		goto out_destroy;

	r_xprt->rx_buf.rb_max_requests = cpu_to_be32(ep->re_max_requests);

	ep->re_attr.event_handler = rpcrdma_qp_event_handler;
	ep->re_attr.qp_context = ep;
	ep->re_attr.srq = NULL;
	ep->re_attr.cap.max_inline_data = 0;
	ep->re_attr.sq_sig_type = IB_SIGNAL_REQ_WR;
	ep->re_attr.qp_type = IB_QPT_RC;
	ep->re_attr.port_num = ~0;

	dprintk("RPC:       %s: requested max: dtos: send %d recv %d; "
		"iovs: send %d recv %d\n",
		__func__,
		ep->re_attr.cap.max_send_wr,
		ep->re_attr.cap.max_recv_wr,
		ep->re_attr.cap.max_send_sge,
		ep->re_attr.cap.max_recv_sge);

	ep->re_send_batch = ep->re_max_requests >> 3;
	ep->re_send_count = ep->re_send_batch;
	init_waitqueue_head(&ep->re_connect_wait);

	ep->re_attr.send_cq = ib_alloc_cq_any(device, r_xprt,
					      ep->re_attr.cap.max_send_wr,
					      IB_POLL_WORKQUEUE);
	if (IS_ERR(ep->re_attr.send_cq)) {
		rc = PTR_ERR(ep->re_attr.send_cq);
		goto out_destroy;
	}

	ep->re_attr.recv_cq = ib_alloc_cq_any(device, r_xprt,
					      ep->re_attr.cap.max_recv_wr,
					      IB_POLL_WORKQUEUE);
	if (IS_ERR(ep->re_attr.recv_cq)) {
		rc = PTR_ERR(ep->re_attr.recv_cq);
		goto out_destroy;
	}
	ep->re_receive_count = 0;

	/* Initialize cma parameters */
	memset(&ep->re_remote_cma, 0, sizeof(ep->re_remote_cma));

	/* Prepare RDMA-CM private message */
	pmsg = &ep->re_cm_private;
	pmsg->cp_magic = rpcrdma_cmp_magic;
	pmsg->cp_version = RPCRDMA_CMP_VERSION;
	pmsg->cp_flags |= RPCRDMA_CMP_F_SND_W_INV_OK;
	pmsg->cp_send_size = rpcrdma_encode_buffer_size(ep->re_inline_send);
	pmsg->cp_recv_size = rpcrdma_encode_buffer_size(ep->re_inline_recv);
	ep->re_remote_cma.private_data = pmsg;
	ep->re_remote_cma.private_data_len = sizeof(*pmsg);

	/* Client offers RDMA Read but does not initiate */
	ep->re_remote_cma.initiator_depth = 0;
	ep->re_remote_cma.responder_resources =
		min_t(int, U8_MAX, device->attrs.max_qp_rd_atom);

	/* Limit transport retries so client can detect server
	 * GID changes quickly. RPC layer handles re-establishing
	 * transport connection and retransmission.
	 */
	ep->re_remote_cma.retry_count = 6;

	/* RPC-over-RDMA handles its own flow control. In addition,
	 * make all RNR NAKs visible so we know that RPC-over-RDMA
	 * flow control is working correctly (no NAKs should be seen).
	 */
	ep->re_remote_cma.flow_control = 0;
	ep->re_remote_cma.rnr_retry_count = 0;

	ep->re_pd = ib_alloc_pd(device, 0);
	if (IS_ERR(ep->re_pd)) {
		rc = PTR_ERR(ep->re_pd);
		goto out_destroy;
	}

	rc = rdma_create_qp(id, ep->re_pd, &ep->re_attr);
	if (rc)
		goto out_destroy;

	r_xprt->rx_ep = ep;
	return 0;

out_destroy:
	rpcrdma_ep_put(ep);
	rdma_destroy_id(id);
	return rc;
}

/**
 * rpcrdma_xprt_connect - Connect an unconnected transport
 * @r_xprt: controlling transport instance
 *
 * Returns 0 on success or a negative errno.
 */
int rpcrdma_xprt_connect(struct rpcrdma_xprt *r_xprt)
{
	struct rpc_xprt *xprt = &r_xprt->rx_xprt;
	struct rpcrdma_ep *ep;
	int rc;

	rc = rpcrdma_ep_create(r_xprt);
	if (rc)
		return rc;
	ep = r_xprt->rx_ep;

	xprt_clear_connected(xprt);
	rpcrdma_reset_cwnd(r_xprt);

	/* Bump the ep's reference count while there are
	 * outstanding Receives.
	 */
	rpcrdma_ep_get(ep);
	rpcrdma_post_recvs(r_xprt, true);

	rc = rdma_connect(ep->re_id, &ep->re_remote_cma);
	if (rc)
		goto out;

	if (xprt->reestablish_timeout < RPCRDMA_INIT_REEST_TO)
		xprt->reestablish_timeout = RPCRDMA_INIT_REEST_TO;
	wait_event_interruptible(ep->re_connect_wait,
				 ep->re_connect_status != 0);
	if (ep->re_connect_status <= 0) {
		rc = ep->re_connect_status;
		goto out;
	}

	rc = rpcrdma_sendctxs_create(r_xprt);
	if (rc) {
		rc = -ENOTCONN;
		goto out;
	}

	rc = rpcrdma_reqs_setup(r_xprt);
	if (rc) {
		rc = -ENOTCONN;
		goto out;
	}
	rpcrdma_mrs_create(r_xprt);

out:
	trace_xprtrdma_connect(r_xprt, rc);
	return rc;
}

/**
 * rpcrdma_xprt_disconnect - Disconnect underlying transport
 * @r_xprt: controlling transport instance
 *
 * Caller serializes. Either the transport send lock is held,
 * or we're being called to destroy the transport.
 *
 * On return, @r_xprt is completely divested of all hardware
 * resources and prepared for the next ->connect operation.
 */
void rpcrdma_xprt_disconnect(struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_ep *ep = r_xprt->rx_ep;
	struct rdma_cm_id *id;
	int rc;

	if (!ep)
		return;

	id = ep->re_id;
	rc = rdma_disconnect(id);
	trace_xprtrdma_disconnect(r_xprt, rc);

	rpcrdma_xprt_drain(r_xprt);
	rpcrdma_reps_unmap(r_xprt);
	rpcrdma_reqs_reset(r_xprt);
	rpcrdma_mrs_destroy(r_xprt);
	rpcrdma_sendctxs_destroy(r_xprt);

	if (rpcrdma_ep_put(ep))
		rdma_destroy_id(id);

	r_xprt->rx_ep = NULL;
}

/* Fixed-size circular FIFO queue. This implementation is wait-free and
 * lock-free.
 *
 * Consumer is the code path that posts Sends. This path dequeues a
 * sendctx for use by a Send operation. Multiple consumer threads
 * are serialized by the RPC transport lock, which allows only one
 * ->send_request call at a time.
 *
 * Producer is the code path that handles Send completions. This path
 * enqueues a sendctx that has been completed. Multiple producer
 * threads are serialized by the ib_poll_cq() function.
 */

/* rpcrdma_sendctxs_destroy() assumes caller has already quiesced
 * queue activity, and rpcrdma_xprt_drain has flushed all remaining
 * Send requests.
 */
static void rpcrdma_sendctxs_destroy(struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	unsigned long i;

	if (!buf->rb_sc_ctxs)
		return;
	for (i = 0; i <= buf->rb_sc_last; i++)
		kfree(buf->rb_sc_ctxs[i]);
	kfree(buf->rb_sc_ctxs);
	buf->rb_sc_ctxs = NULL;
}

static struct rpcrdma_sendctx *rpcrdma_sendctx_create(struct rpcrdma_ep *ep)
{
	struct rpcrdma_sendctx *sc;

	sc = kzalloc(struct_size(sc, sc_sges, ep->re_attr.cap.max_send_sge),
		     GFP_KERNEL);
	if (!sc)
		return NULL;

	sc->sc_cqe.done = rpcrdma_wc_send;
	return sc;
}

static int rpcrdma_sendctxs_create(struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	struct rpcrdma_sendctx *sc;
	unsigned long i;

	/* Maximum number of concurrent outstanding Send WRs. Capping
	 * the circular queue size stops Send Queue overflow by causing
	 * the ->send_request call to fail temporarily before too many
	 * Sends are posted.
	 */
	i = r_xprt->rx_ep->re_max_requests + RPCRDMA_MAX_BC_REQUESTS;
	buf->rb_sc_ctxs = kcalloc(i, sizeof(sc), GFP_KERNEL);
	if (!buf->rb_sc_ctxs)
		return -ENOMEM;

	buf->rb_sc_last = i - 1;
	for (i = 0; i <= buf->rb_sc_last; i++) {
		sc = rpcrdma_sendctx_create(r_xprt->rx_ep);
		if (!sc)
			return -ENOMEM;

		buf->rb_sc_ctxs[i] = sc;
	}

	buf->rb_sc_head = 0;
	buf->rb_sc_tail = 0;
	return 0;
}

/* The sendctx queue is not guaranteed to have a size that is a
 * power of two, thus the helpers in circ_buf.h cannot be used.
 * The other option is to use modulus (%), which can be expensive.
 */
static unsigned long rpcrdma_sendctx_next(struct rpcrdma_buffer *buf,
					  unsigned long item)
{
	return likely(item < buf->rb_sc_last) ? item + 1 : 0;
}

/**
 * rpcrdma_sendctx_get_locked - Acquire a send context
 * @r_xprt: controlling transport instance
 *
 * Returns pointer to a free send completion context; or NULL if
 * the queue is empty.
 *
 * Usage: Called to acquire an SGE array before preparing a Send WR.
 *
 * The caller serializes calls to this function (per transport), and
 * provides an effective memory barrier that flushes the new value
 * of rb_sc_head.
 */
struct rpcrdma_sendctx *rpcrdma_sendctx_get_locked(struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	struct rpcrdma_sendctx *sc;
	unsigned long next_head;

	next_head = rpcrdma_sendctx_next(buf, buf->rb_sc_head);

	if (next_head == READ_ONCE(buf->rb_sc_tail))
		goto out_emptyq;

	/* ORDER: item must be accessed _before_ head is updated */
	sc = buf->rb_sc_ctxs[next_head];

	/* Releasing the lock in the caller acts as a memory
	 * barrier that flushes rb_sc_head.
	 */
	buf->rb_sc_head = next_head;

	return sc;

out_emptyq:
	/* The queue is "empty" if there have not been enough Send
	 * completions recently. This is a sign the Send Queue is
	 * backing up. Cause the caller to pause and try again.
	 */
	xprt_wait_for_buffer_space(&r_xprt->rx_xprt);
	r_xprt->rx_stats.empty_sendctx_q++;
	return NULL;
}

/**
 * rpcrdma_sendctx_put_locked - Release a send context
 * @r_xprt: controlling transport instance
 * @sc: send context to release
 *
 * Usage: Called from Send completion to return a sendctxt
 * to the queue.
 *
 * The caller serializes calls to this function (per transport).
 */
static void rpcrdma_sendctx_put_locked(struct rpcrdma_xprt *r_xprt,
				       struct rpcrdma_sendctx *sc)
{
	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	unsigned long next_tail;

	/* Unmap SGEs of previously completed but unsignaled
	 * Sends by walking up the queue until @sc is found.
	 */
	next_tail = buf->rb_sc_tail;
	do {
		next_tail = rpcrdma_sendctx_next(buf, next_tail);

		/* ORDER: item must be accessed _before_ tail is updated */
		rpcrdma_sendctx_unmap(buf->rb_sc_ctxs[next_tail]);

	} while (buf->rb_sc_ctxs[next_tail] != sc);

	/* Paired with READ_ONCE */
	smp_store_release(&buf->rb_sc_tail, next_tail);

	xprt_write_space(&r_xprt->rx_xprt);
}

static void
rpcrdma_mrs_create(struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	struct rpcrdma_ep *ep = r_xprt->rx_ep;
	unsigned int count;

	for (count = 0; count < ep->re_max_rdma_segs; count++) {
		struct rpcrdma_mr *mr;
		int rc;

		mr = kzalloc(sizeof(*mr), GFP_NOFS);
		if (!mr)
			break;

		rc = frwr_mr_init(r_xprt, mr);
		if (rc) {
			kfree(mr);
			break;
		}

		spin_lock(&buf->rb_lock);
		rpcrdma_mr_push(mr, &buf->rb_mrs);
		list_add(&mr->mr_all, &buf->rb_all_mrs);
		spin_unlock(&buf->rb_lock);
	}

	r_xprt->rx_stats.mrs_allocated += count;
	trace_xprtrdma_createmrs(r_xprt, count);
}

static void
rpcrdma_mr_refresh_worker(struct work_struct *work)
{
	struct rpcrdma_buffer *buf = container_of(work, struct rpcrdma_buffer,
						  rb_refresh_worker);
	struct rpcrdma_xprt *r_xprt = container_of(buf, struct rpcrdma_xprt,
						   rx_buf);

	rpcrdma_mrs_create(r_xprt);
	xprt_write_space(&r_xprt->rx_xprt);
}

/**
 * rpcrdma_mrs_refresh - Wake the MR refresh worker
 * @r_xprt: controlling transport instance
 *
 */
void rpcrdma_mrs_refresh(struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	struct rpcrdma_ep *ep = r_xprt->rx_ep;

	/* If there is no underlying connection, it's no use
	 * to wake the refresh worker.
	 */
	if (ep->re_connect_status == 1) {
		/* The work is scheduled on a WQ_MEM_RECLAIM
		 * workqueue in order to prevent MR allocation
		 * from recursing into NFS during direct reclaim.
		 */
		queue_work(xprtiod_workqueue, &buf->rb_refresh_worker);
	}
}

/**
 * rpcrdma_req_create - Allocate an rpcrdma_req object
 * @r_xprt: controlling r_xprt
 * @size: initial size, in bytes, of send and receive buffers
 * @flags: GFP flags passed to memory allocators
 *
 * Returns an allocated and fully initialized rpcrdma_req or NULL.
 */
struct rpcrdma_req *rpcrdma_req_create(struct rpcrdma_xprt *r_xprt, size_t size,
				       gfp_t flags)
{
	struct rpcrdma_buffer *buffer = &r_xprt->rx_buf;
	struct rpcrdma_req *req;

	req = kzalloc(sizeof(*req), flags);
	if (req == NULL)
		goto out1;

	req->rl_sendbuf = rpcrdma_regbuf_alloc(size, DMA_TO_DEVICE, flags);
	if (!req->rl_sendbuf)
		goto out2;

	req->rl_recvbuf = rpcrdma_regbuf_alloc(size, DMA_NONE, flags);
	if (!req->rl_recvbuf)
		goto out3;

	INIT_LIST_HEAD(&req->rl_free_mrs);
	INIT_LIST_HEAD(&req->rl_registered);
	spin_lock(&buffer->rb_lock);
	list_add(&req->rl_all, &buffer->rb_allreqs);
	spin_unlock(&buffer->rb_lock);
	return req;

out3:
	kfree(req->rl_sendbuf);
out2:
	kfree(req);
out1:
	return NULL;
}

/**
 * rpcrdma_req_setup - Per-connection instance setup of an rpcrdma_req object
 * @r_xprt: controlling transport instance
 * @req: rpcrdma_req object to set up
 *
 * Returns zero on success, and a negative errno on failure.
 */
int rpcrdma_req_setup(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
{
	struct rpcrdma_regbuf *rb;
	size_t maxhdrsize;

	/* Compute maximum header buffer size in bytes */
	maxhdrsize = rpcrdma_fixed_maxsz + 3 +
		     r_xprt->rx_ep->re_max_rdma_segs * rpcrdma_readchunk_maxsz;
	maxhdrsize *= sizeof(__be32);
	rb = rpcrdma_regbuf_alloc(__roundup_pow_of_two(maxhdrsize),
				  DMA_TO_DEVICE, GFP_KERNEL);
	if (!rb)
		goto out;

	if (!__rpcrdma_regbuf_dma_map(r_xprt, rb))
		goto out_free;

	req->rl_rdmabuf = rb;
	xdr_buf_init(&req->rl_hdrbuf, rdmab_data(rb), rdmab_length(rb));
	return 0;

out_free:
	rpcrdma_regbuf_free(rb);
out:
	return -ENOMEM;
}

/* ASSUMPTION: the rb_allreqs list is stable for the duration,
 * and thus can be walked without holding rb_lock. Eg. the
 * caller is holding the transport send lock to exclude
 * device removal or disconnection.
 */
static int rpcrdma_reqs_setup(struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	struct rpcrdma_req *req;
	int rc;

	list_for_each_entry(req, &buf->rb_allreqs, rl_all) {
		rc = rpcrdma_req_setup(r_xprt, req);
		if (rc)
			return rc;
	}
	return 0;
}

static void rpcrdma_req_reset(struct rpcrdma_req *req)
{
	/* Credits are valid for only one connection */
	req->rl_slot.rq_cong = 0;

	rpcrdma_regbuf_free(req->rl_rdmabuf);
	req->rl_rdmabuf = NULL;

	rpcrdma_regbuf_dma_unmap(req->rl_sendbuf);
	rpcrdma_regbuf_dma_unmap(req->rl_recvbuf);
}

/* ASSUMPTION: the rb_allreqs list is stable for the duration,
 * and thus can be walked without holding rb_lock. Eg. the
 * caller is holding the transport send lock to exclude
 * device removal or disconnection.
 */
static void rpcrdma_reqs_reset(struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	struct rpcrdma_req *req;

	list_for_each_entry(req, &buf->rb_allreqs, rl_all)
		rpcrdma_req_reset(req);
}

/* No locking needed here. This function is called only by the
 * Receive completion handler.
 */
static noinline
struct rpcrdma_rep *rpcrdma_rep_create(struct rpcrdma_xprt *r_xprt,
				       bool temp)
{
	struct rpcrdma_rep *rep;

	rep = kzalloc(sizeof(*rep), GFP_KERNEL);
	if (rep == NULL)
		goto out;

	rep->rr_rdmabuf = rpcrdma_regbuf_alloc(r_xprt->rx_ep->re_inline_recv,
					       DMA_FROM_DEVICE, GFP_KERNEL);
	if (!rep->rr_rdmabuf)
		goto out_free;

	if (!rpcrdma_regbuf_dma_map(r_xprt, rep->rr_rdmabuf))
		goto out_free_regbuf;

	xdr_buf_init(&rep->rr_hdrbuf, rdmab_data(rep->rr_rdmabuf),
		     rdmab_length(rep->rr_rdmabuf));
	rep->rr_cqe.done = rpcrdma_wc_receive;
	rep->rr_rxprt = r_xprt;
	rep->rr_recv_wr.next = NULL;
	rep->rr_recv_wr.wr_cqe = &rep->rr_cqe;
	rep->rr_recv_wr.sg_list = &rep->rr_rdmabuf->rg_iov;
	rep->rr_recv_wr.num_sge = 1;
	rep->rr_temp = temp;
	list_add(&rep->rr_all, &r_xprt->rx_buf.rb_all_reps);
	return rep;

out_free_regbuf:
	rpcrdma_regbuf_free(rep->rr_rdmabuf);
out_free:
	kfree(rep);
out:
	return NULL;
}

/* No locking needed here. This function is invoked only by the
 * Receive completion handler, or during transport shutdown.
 */
static void rpcrdma_rep_destroy(struct rpcrdma_rep *rep)
{
	list_del(&rep->rr_all);
	rpcrdma_regbuf_free(rep->rr_rdmabuf);
	kfree(rep);
}

static struct rpcrdma_rep *rpcrdma_rep_get_locked(struct rpcrdma_buffer *buf)
{
	struct llist_node *node;

	/* Calls to llist_del_first are required to be serialized */
	node = llist_del_first(&buf->rb_free_reps);
	if (!node)
		return NULL;
	return llist_entry(node, struct rpcrdma_rep, rr_node);
}

static void rpcrdma_rep_put(struct rpcrdma_buffer *buf,
			    struct rpcrdma_rep *rep)
{
	llist_add(&rep->rr_node, &buf->rb_free_reps);
}

static void rpcrdma_reps_unmap(struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	struct rpcrdma_rep *rep;

	list_for_each_entry(rep, &buf->rb_all_reps, rr_all) {
		rpcrdma_regbuf_dma_unmap(rep->rr_rdmabuf);
		rep->rr_temp = true;
	}
}

static void rpcrdma_reps_destroy(struct rpcrdma_buffer *buf)
{
	struct rpcrdma_rep *rep;

	while ((rep = rpcrdma_rep_get_locked(buf)) != NULL)
		rpcrdma_rep_destroy(rep);
}

/**
 * rpcrdma_buffer_create - Create initial set of req/rep objects
 * @r_xprt: transport instance to (re)initialize
 *
 * Returns zero on success, otherwise a negative errno.
 */
int rpcrdma_buffer_create(struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	int i, rc;

	buf->rb_bc_srv_max_requests = 0;
	spin_lock_init(&buf->rb_lock);
	INIT_LIST_HEAD(&buf->rb_mrs);
	INIT_LIST_HEAD(&buf->rb_all_mrs);
	INIT_WORK(&buf->rb_refresh_worker, rpcrdma_mr_refresh_worker);

	INIT_LIST_HEAD(&buf->rb_send_bufs);
	INIT_LIST_HEAD(&buf->rb_allreqs);
	INIT_LIST_HEAD(&buf->rb_all_reps);

	rc = -ENOMEM;
	for (i = 0; i < r_xprt->rx_xprt.max_reqs; i++) {
		struct rpcrdma_req *req;

		req = rpcrdma_req_create(r_xprt, RPCRDMA_V1_DEF_INLINE_SIZE * 2,
					 GFP_KERNEL);
		if (!req)
			goto out;
		list_add(&req->rl_list, &buf->rb_send_bufs);
	}

	init_llist_head(&buf->rb_free_reps);

	return 0;
out:
	rpcrdma_buffer_destroy(buf);
	return rc;
}

/**
 * rpcrdma_req_destroy - Destroy an rpcrdma_req object
 * @req: unused object to be destroyed
 *
 * Relies on caller holding the transport send lock to protect
 * removing req->rl_all from buf->rb_all_reqs safely.
 */
void rpcrdma_req_destroy(struct rpcrdma_req *req)
{
	struct rpcrdma_mr *mr;

	list_del(&req->rl_all);

	while ((mr = rpcrdma_mr_pop(&req->rl_free_mrs))) {
		struct rpcrdma_buffer *buf = &mr->mr_xprt->rx_buf;

		spin_lock(&buf->rb_lock);
		list_del(&mr->mr_all);
		spin_unlock(&buf->rb_lock);

		frwr_release_mr(mr);
	}

	rpcrdma_regbuf_free(req->rl_recvbuf);
	rpcrdma_regbuf_free(req->rl_sendbuf);
	rpcrdma_regbuf_free(req->rl_rdmabuf);
	kfree(req);
}

/**
 * rpcrdma_mrs_destroy - Release all of a transport's MRs
 * @r_xprt: controlling transport instance
 *
 * Relies on caller holding the transport send lock to protect
 * removing mr->mr_list from req->rl_free_mrs safely.
 */
static void rpcrdma_mrs_destroy(struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	struct rpcrdma_mr *mr;

	cancel_work_sync(&buf->rb_refresh_worker);

	spin_lock(&buf->rb_lock);
	while ((mr = list_first_entry_or_null(&buf->rb_all_mrs,
					      struct rpcrdma_mr,
					      mr_all)) != NULL) {
		list_del(&mr->mr_list);
		list_del(&mr->mr_all);
		spin_unlock(&buf->rb_lock);

		frwr_release_mr(mr);

		spin_lock(&buf->rb_lock);
	}
	spin_unlock(&buf->rb_lock);
}

/**
 * rpcrdma_buffer_destroy - Release all hw resources
 * @buf: root control block for resources
 *
 * ORDERING: relies on a prior rpcrdma_xprt_drain :
 * - No more Send or Receive completions can occur
 * - All MRs, reps, and reqs are returned to their free lists
 */
void
rpcrdma_buffer_destroy(struct rpcrdma_buffer *buf)
{
	rpcrdma_reps_destroy(buf);

	while (!list_empty(&buf->rb_send_bufs)) {
		struct rpcrdma_req *req;

		req = list_first_entry(&buf->rb_send_bufs,
				       struct rpcrdma_req, rl_list);
		list_del(&req->rl_list);
		rpcrdma_req_destroy(req);
	}
}

/**
 * rpcrdma_mr_get - Allocate an rpcrdma_mr object
 * @r_xprt: controlling transport
 *
 * Returns an initialized rpcrdma_mr or NULL if no free
 * rpcrdma_mr objects are available.
 */
struct rpcrdma_mr *
rpcrdma_mr_get(struct rpcrdma_xprt *r_xprt)
{
	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	struct rpcrdma_mr *mr;

	spin_lock(&buf->rb_lock);
	mr = rpcrdma_mr_pop(&buf->rb_mrs);
	spin_unlock(&buf->rb_lock);
	return mr;
}

/**
 * rpcrdma_mr_put - DMA unmap an MR and release it
 * @mr: MR to release
 *
 */
void rpcrdma_mr_put(struct rpcrdma_mr *mr)
{
	struct rpcrdma_xprt *r_xprt = mr->mr_xprt;

	if (mr->mr_dir != DMA_NONE) {
		trace_xprtrdma_mr_unmap(mr);
		ib_dma_unmap_sg(r_xprt->rx_ep->re_id->device,
				mr->mr_sg, mr->mr_nents, mr->mr_dir);
		mr->mr_dir = DMA_NONE;
	}

	rpcrdma_mr_push(mr, &mr->mr_req->rl_free_mrs);
}

/**
 * rpcrdma_buffer_get - Get a request buffer
 * @buffers: Buffer pool from which to obtain a buffer
 *
 * Returns a fresh rpcrdma_req, or NULL if none are available.
 */
struct rpcrdma_req *
rpcrdma_buffer_get(struct rpcrdma_buffer *buffers)
{
	struct rpcrdma_req *req;

	spin_lock(&buffers->rb_lock);
	req = list_first_entry_or_null(&buffers->rb_send_bufs,
				       struct rpcrdma_req, rl_list);
	if (req)
		list_del_init(&req->rl_list);
	spin_unlock(&buffers->rb_lock);
	return req;
}

/**
 * rpcrdma_buffer_put - Put request/reply buffers back into pool
 * @buffers: buffer pool
 * @req: object to return
 *
 */
void rpcrdma_buffer_put(struct rpcrdma_buffer *buffers, struct rpcrdma_req *req)
{
	if (req->rl_reply)
		rpcrdma_rep_put(buffers, req->rl_reply);
	req->rl_reply = NULL;

	spin_lock(&buffers->rb_lock);
	list_add(&req->rl_list, &buffers->rb_send_bufs);
	spin_unlock(&buffers->rb_lock);
}

/**
 * rpcrdma_recv_buffer_put - Release rpcrdma_rep back to free list
 * @rep: rep to release
 *
 * Used after error conditions.
 */
void rpcrdma_recv_buffer_put(struct rpcrdma_rep *rep)
{
	rpcrdma_rep_put(&rep->rr_rxprt->rx_buf, rep);
}

/* Returns a pointer to a rpcrdma_regbuf object, or NULL.
 *
 * xprtrdma uses a regbuf for posting an outgoing RDMA SEND, or for
 * receiving the payload of RDMA RECV operations. During Long Calls
 * or Replies they may be registered externally via frwr_map.
 */
static struct rpcrdma_regbuf *
rpcrdma_regbuf_alloc(size_t size, enum dma_data_direction direction,
		     gfp_t flags)
{
	struct rpcrdma_regbuf *rb;

	rb = kmalloc(sizeof(*rb), flags);
	if (!rb)
		return NULL;
	rb->rg_data = kmalloc(size, flags);
	if (!rb->rg_data) {
		kfree(rb);
		return NULL;
	}

	rb->rg_device = NULL;
	rb->rg_direction = direction;
	rb->rg_iov.length = size;
	return rb;
}

/**
 * rpcrdma_regbuf_realloc - re-allocate a SEND/RECV buffer
 * @rb: regbuf to reallocate
 * @size: size of buffer to be allocated, in bytes
 * @flags: GFP flags
 *
 * Returns true if reallocation was successful. If false is
 * returned, @rb is left untouched.
 */
bool rpcrdma_regbuf_realloc(struct rpcrdma_regbuf *rb, size_t size, gfp_t flags)
{
	void *buf;

	buf = kmalloc(size, flags);
	if (!buf)
		return false;

	rpcrdma_regbuf_dma_unmap(rb);
	kfree(rb->rg_data);

	rb->rg_data = buf;
	rb->rg_iov.length = size;
	return true;
}

/**
 * __rpcrdma_regbuf_dma_map - DMA-map a regbuf
 * @r_xprt: controlling transport instance
 * @rb: regbuf to be mapped
 *
 * Returns true if the buffer is now DMA mapped to @r_xprt's device
 */
bool __rpcrdma_regbuf_dma_map(struct rpcrdma_xprt *r_xprt,
			      struct rpcrdma_regbuf *rb)
{
	struct ib_device *device = r_xprt->rx_ep->re_id->device;

	if (rb->rg_direction == DMA_NONE)
		return false;

	rb->rg_iov.addr = ib_dma_map_single(device, rdmab_data(rb),
					    rdmab_length(rb), rb->rg_direction);
	if (ib_dma_mapping_error(device, rdmab_addr(rb))) {
		trace_xprtrdma_dma_maperr(rdmab_addr(rb));
		return false;
	}

	rb->rg_device = device;
	rb->rg_iov.lkey = r_xprt->rx_ep->re_pd->local_dma_lkey;
	return true;
}

static void rpcrdma_regbuf_dma_unmap(struct rpcrdma_regbuf *rb)
{
	if (!rb)
		return;

	if (!rpcrdma_regbuf_is_mapped(rb))
		return;

	ib_dma_unmap_single(rb->rg_device, rdmab_addr(rb), rdmab_length(rb),
			    rb->rg_direction);
	rb->rg_device = NULL;
}

static void rpcrdma_regbuf_free(struct rpcrdma_regbuf *rb)
{
	rpcrdma_regbuf_dma_unmap(rb);
	if (rb)
		kfree(rb->rg_data);
	kfree(rb);
}

/**
 * rpcrdma_post_sends - Post WRs to a transport's Send Queue
 * @r_xprt: controlling transport instance
 * @req: rpcrdma_req containing the Send WR to post
 *
 * Returns 0 if the post was successful, otherwise -ENOTCONN
 * is returned.
 */
int rpcrdma_post_sends(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
{
	struct ib_send_wr *send_wr = &req->rl_wr;
	struct rpcrdma_ep *ep = r_xprt->rx_ep;
	int rc;

	if (!ep->re_send_count || kref_read(&req->rl_kref) > 1) {
		send_wr->send_flags |= IB_SEND_SIGNALED;
		ep->re_send_count = ep->re_send_batch;
	} else {
		send_wr->send_flags &= ~IB_SEND_SIGNALED;
		--ep->re_send_count;
	}

	trace_xprtrdma_post_send(req);
	rc = frwr_send(r_xprt, req);
	if (rc)
		return -ENOTCONN;
	return 0;
}

/**
 * rpcrdma_post_recvs - Refill the Receive Queue
 * @r_xprt: controlling transport instance
 * @temp: mark Receive buffers to be deleted after use
 *
 */
void rpcrdma_post_recvs(struct rpcrdma_xprt *r_xprt, bool temp)
{
	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	struct rpcrdma_ep *ep = r_xprt->rx_ep;
	struct ib_recv_wr *wr, *bad_wr;
	struct rpcrdma_rep *rep;
	int needed, count, rc;

	rc = 0;
	count = 0;

	needed = buf->rb_credits + (buf->rb_bc_srv_max_requests << 1);
	if (likely(ep->re_receive_count > needed))
		goto out;
	needed -= ep->re_receive_count;
	if (!temp)
		needed += RPCRDMA_MAX_RECV_BATCH;

	/* fast path: all needed reps can be found on the free list */
	wr = NULL;
	while (needed) {
		rep = rpcrdma_rep_get_locked(buf);
		if (rep && rep->rr_temp) {
			rpcrdma_rep_destroy(rep);
			continue;
		}
		if (!rep)
			rep = rpcrdma_rep_create(r_xprt, temp);
		if (!rep)
			break;

		trace_xprtrdma_post_recv(rep);
		rep->rr_recv_wr.next = wr;
		wr = &rep->rr_recv_wr;
		--needed;
		++count;
	}
	if (!wr)
		goto out;

	rc = ib_post_recv(ep->re_id->qp, wr,
			  (const struct ib_recv_wr **)&bad_wr);
out:
	trace_xprtrdma_post_recvs(r_xprt, count, rc);
	if (rc) {
		for (wr = bad_wr; wr;) {
			struct rpcrdma_rep *rep;

			rep = container_of(wr, struct rpcrdma_rep, rr_recv_wr);
			wr = wr->next;
			rpcrdma_recv_buffer_put(rep);
			--count;
		}
	}
	ep->re_receive_count += count;
	return;
}