summaryrefslogtreecommitdiffstats
path: root/net/sunrpc/xprtrdma/rpc_rdma.c
blob: 26d334c83b3861a637916874973aab5cf0a12df8 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
// 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.
 */

/*
 * rpc_rdma.c
 *
 * This file contains the guts of the RPC RDMA protocol, and
 * does marshaling/unmarshaling, etc. It is also where interfacing
 * to the Linux RPC framework lives.
 */

#include <linux/highmem.h>

#include <linux/sunrpc/svc_rdma.h>

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

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

/* Returns size of largest RPC-over-RDMA header in a Call message
 *
 * The largest Call header contains a full-size Read list and a
 * minimal Reply chunk.
 */
static unsigned int rpcrdma_max_call_header_size(unsigned int maxsegs)
{
	unsigned int size;

	/* Fixed header fields and list discriminators */
	size = RPCRDMA_HDRLEN_MIN;

	/* Maximum Read list size */
	size = maxsegs * rpcrdma_readchunk_maxsz * sizeof(__be32);

	/* Minimal Read chunk size */
	size += sizeof(__be32);	/* segment count */
	size += rpcrdma_segment_maxsz * sizeof(__be32);
	size += sizeof(__be32);	/* list discriminator */

	dprintk("RPC:       %s: max call header size = %u\n",
		__func__, size);
	return size;
}

/* Returns size of largest RPC-over-RDMA header in a Reply message
 *
 * There is only one Write list or one Reply chunk per Reply
 * message.  The larger list is the Write list.
 */
static unsigned int rpcrdma_max_reply_header_size(unsigned int maxsegs)
{
	unsigned int size;

	/* Fixed header fields and list discriminators */
	size = RPCRDMA_HDRLEN_MIN;

	/* Maximum Write list size */
	size = sizeof(__be32);		/* segment count */
	size += maxsegs * rpcrdma_segment_maxsz * sizeof(__be32);
	size += sizeof(__be32);	/* list discriminator */

	dprintk("RPC:       %s: max reply header size = %u\n",
		__func__, size);
	return size;
}

/**
 * rpcrdma_set_max_header_sizes - Initialize inline payload sizes
 * @r_xprt: transport instance to initialize
 *
 * The max_inline fields contain the maximum size of an RPC message
 * so the marshaling code doesn't have to repeat this calculation
 * for every RPC.
 */
void rpcrdma_set_max_header_sizes(struct rpcrdma_xprt *r_xprt)
{
	unsigned int maxsegs = r_xprt->rx_ia.ri_max_segs;
	struct rpcrdma_ep *ep = &r_xprt->rx_ep;

	ep->rep_max_inline_send =
		ep->rep_inline_send - rpcrdma_max_call_header_size(maxsegs);
	ep->rep_max_inline_recv =
		ep->rep_inline_recv - rpcrdma_max_reply_header_size(maxsegs);
}

/* The client can send a request inline as long as the RPCRDMA header
 * plus the RPC call fit under the transport's inline limit. If the
 * combined call message size exceeds that limit, the client must use
 * a Read chunk for this operation.
 *
 * A Read chunk is also required if sending the RPC call inline would
 * exceed this device's max_sge limit.
 */
static bool rpcrdma_args_inline(struct rpcrdma_xprt *r_xprt,
				struct rpc_rqst *rqst)
{
	struct xdr_buf *xdr = &rqst->rq_snd_buf;
	unsigned int count, remaining, offset;

	if (xdr->len > r_xprt->rx_ep.rep_max_inline_send)
		return false;

	if (xdr->page_len) {
		remaining = xdr->page_len;
		offset = offset_in_page(xdr->page_base);
		count = RPCRDMA_MIN_SEND_SGES;
		while (remaining) {
			remaining -= min_t(unsigned int,
					   PAGE_SIZE - offset, remaining);
			offset = 0;
			if (++count > r_xprt->rx_ia.ri_max_send_sges)
				return false;
		}
	}

	return true;
}

/* The client can't know how large the actual reply will be. Thus it
 * plans for the largest possible reply for that particular ULP
 * operation. If the maximum combined reply message size exceeds that
 * limit, the client must provide a write list or a reply chunk for
 * this request.
 */
static bool rpcrdma_results_inline(struct rpcrdma_xprt *r_xprt,
				   struct rpc_rqst *rqst)
{
	return rqst->rq_rcv_buf.buflen <= r_xprt->rx_ep.rep_max_inline_recv;
}

/* The client is required to provide a Reply chunk if the maximum
 * size of the non-payload part of the RPC Reply is larger than
 * the inline threshold.
 */
static bool
rpcrdma_nonpayload_inline(const struct rpcrdma_xprt *r_xprt,
			  const struct rpc_rqst *rqst)
{
	const struct xdr_buf *buf = &rqst->rq_rcv_buf;

	return (buf->head[0].iov_len + buf->tail[0].iov_len) <
		r_xprt->rx_ep.rep_max_inline_recv;
}

/* Split @vec on page boundaries into SGEs. FMR registers pages, not
 * a byte range. Other modes coalesce these SGEs into a single MR
 * when they can.
 *
 * Returns pointer to next available SGE, and bumps the total number
 * of SGEs consumed.
 */
static struct rpcrdma_mr_seg *
rpcrdma_convert_kvec(struct kvec *vec, struct rpcrdma_mr_seg *seg,
		     unsigned int *n)
{
	u32 remaining, page_offset;
	char *base;

	base = vec->iov_base;
	page_offset = offset_in_page(base);
	remaining = vec->iov_len;
	while (remaining) {
		seg->mr_page = NULL;
		seg->mr_offset = base;
		seg->mr_len = min_t(u32, PAGE_SIZE - page_offset, remaining);
		remaining -= seg->mr_len;
		base += seg->mr_len;
		++seg;
		++(*n);
		page_offset = 0;
	}
	return seg;
}

/* Convert @xdrbuf into SGEs no larger than a page each. As they
 * are registered, these SGEs are then coalesced into RDMA segments
 * when the selected memreg mode supports it.
 *
 * Returns positive number of SGEs consumed, or a negative errno.
 */

static int
rpcrdma_convert_iovs(struct rpcrdma_xprt *r_xprt, struct xdr_buf *xdrbuf,
		     unsigned int pos, enum rpcrdma_chunktype type,
		     struct rpcrdma_mr_seg *seg)
{
	unsigned long page_base;
	unsigned int len, n;
	struct page **ppages;

	n = 0;
	if (pos == 0)
		seg = rpcrdma_convert_kvec(&xdrbuf->head[0], seg, &n);

	len = xdrbuf->page_len;
	ppages = xdrbuf->pages + (xdrbuf->page_base >> PAGE_SHIFT);
	page_base = offset_in_page(xdrbuf->page_base);
	while (len) {
		/* ACL likes to be lazy in allocating pages - ACLs
		 * are small by default but can get huge.
		 */
		if (unlikely(xdrbuf->flags & XDRBUF_SPARSE_PAGES)) {
			if (!*ppages)
				*ppages = alloc_page(GFP_NOWAIT | __GFP_NOWARN);
			if (!*ppages)
				return -ENOBUFS;
		}
		seg->mr_page = *ppages;
		seg->mr_offset = (char *)page_base;
		seg->mr_len = min_t(u32, PAGE_SIZE - page_base, len);
		len -= seg->mr_len;
		++ppages;
		++seg;
		++n;
		page_base = 0;
	}

	/* When encoding a Read chunk, the tail iovec contains an
	 * XDR pad and may be omitted.
	 */
	if (type == rpcrdma_readch && r_xprt->rx_ia.ri_implicit_roundup)
		goto out;

	/* When encoding a Write chunk, some servers need to see an
	 * extra segment for non-XDR-aligned Write chunks. The upper
	 * layer provides space in the tail iovec that may be used
	 * for this purpose.
	 */
	if (type == rpcrdma_writech && r_xprt->rx_ia.ri_implicit_roundup)
		goto out;

	if (xdrbuf->tail[0].iov_len)
		seg = rpcrdma_convert_kvec(&xdrbuf->tail[0], seg, &n);

out:
	if (unlikely(n > RPCRDMA_MAX_SEGS))
		return -EIO;
	return n;
}

static inline int
encode_item_present(struct xdr_stream *xdr)
{
	__be32 *p;

	p = xdr_reserve_space(xdr, sizeof(*p));
	if (unlikely(!p))
		return -EMSGSIZE;

	*p = xdr_one;
	return 0;
}

static inline int
encode_item_not_present(struct xdr_stream *xdr)
{
	__be32 *p;

	p = xdr_reserve_space(xdr, sizeof(*p));
	if (unlikely(!p))
		return -EMSGSIZE;

	*p = xdr_zero;
	return 0;
}

static void
xdr_encode_rdma_segment(__be32 *iptr, struct rpcrdma_mr *mr)
{
	*iptr++ = cpu_to_be32(mr->mr_handle);
	*iptr++ = cpu_to_be32(mr->mr_length);
	xdr_encode_hyper(iptr, mr->mr_offset);
}

static int
encode_rdma_segment(struct xdr_stream *xdr, struct rpcrdma_mr *mr)
{
	__be32 *p;

	p = xdr_reserve_space(xdr, 4 * sizeof(*p));
	if (unlikely(!p))
		return -EMSGSIZE;

	xdr_encode_rdma_segment(p, mr);
	return 0;
}

static int
encode_read_segment(struct xdr_stream *xdr, struct rpcrdma_mr *mr,
		    u32 position)
{
	__be32 *p;

	p = xdr_reserve_space(xdr, 6 * sizeof(*p));
	if (unlikely(!p))
		return -EMSGSIZE;

	*p++ = xdr_one;			/* Item present */
	*p++ = cpu_to_be32(position);
	xdr_encode_rdma_segment(p, mr);
	return 0;
}

static struct rpcrdma_mr_seg *rpcrdma_mr_prepare(struct rpcrdma_xprt *r_xprt,
						 struct rpcrdma_req *req,
						 struct rpcrdma_mr_seg *seg,
						 int nsegs, bool writing,
						 struct rpcrdma_mr **mr)
{
	*mr = rpcrdma_mr_pop(&req->rl_free_mrs);
	if (!*mr) {
		*mr = rpcrdma_mr_get(r_xprt);
		if (!*mr)
			goto out_getmr_err;
		trace_xprtrdma_mr_get(req);
		(*mr)->mr_req = req;
	}

	rpcrdma_mr_push(*mr, &req->rl_registered);
	return frwr_map(r_xprt, seg, nsegs, writing, req->rl_slot.rq_xid, *mr);

out_getmr_err:
	trace_xprtrdma_nomrs(req);
	xprt_wait_for_buffer_space(&r_xprt->rx_xprt);
	rpcrdma_mrs_refresh(r_xprt);
	return ERR_PTR(-EAGAIN);
}

/* Register and XDR encode the Read list. Supports encoding a list of read
 * segments that belong to a single read chunk.
 *
 * Encoding key for single-list chunks (HLOO = Handle32 Length32 Offset64):
 *
 *  Read chunklist (a linked list):
 *   N elements, position P (same P for all chunks of same arg!):
 *    1 - PHLOO - 1 - PHLOO - ... - 1 - PHLOO - 0
 *
 * Returns zero on success, or a negative errno if a failure occurred.
 * @xdr is advanced to the next position in the stream.
 *
 * Only a single @pos value is currently supported.
 */
static int rpcrdma_encode_read_list(struct rpcrdma_xprt *r_xprt,
				    struct rpcrdma_req *req,
				    struct rpc_rqst *rqst,
				    enum rpcrdma_chunktype rtype)
{
	struct xdr_stream *xdr = &req->rl_stream;
	struct rpcrdma_mr_seg *seg;
	struct rpcrdma_mr *mr;
	unsigned int pos;
	int nsegs;

	if (rtype == rpcrdma_noch_pullup || rtype == rpcrdma_noch_mapped)
		goto done;

	pos = rqst->rq_snd_buf.head[0].iov_len;
	if (rtype == rpcrdma_areadch)
		pos = 0;
	seg = req->rl_segments;
	nsegs = rpcrdma_convert_iovs(r_xprt, &rqst->rq_snd_buf, pos,
				     rtype, seg);
	if (nsegs < 0)
		return nsegs;

	do {
		seg = rpcrdma_mr_prepare(r_xprt, req, seg, nsegs, false, &mr);
		if (IS_ERR(seg))
			return PTR_ERR(seg);

		if (encode_read_segment(xdr, mr, pos) < 0)
			return -EMSGSIZE;

		trace_xprtrdma_chunk_read(rqst->rq_task, pos, mr, nsegs);
		r_xprt->rx_stats.read_chunk_count++;
		nsegs -= mr->mr_nents;
	} while (nsegs);

done:
	return encode_item_not_present(xdr);
}

/* Register and XDR encode the Write list. Supports encoding a list
 * containing one array of plain segments that belong to a single
 * write chunk.
 *
 * Encoding key for single-list chunks (HLOO = Handle32 Length32 Offset64):
 *
 *  Write chunklist (a list of (one) counted array):
 *   N elements:
 *    1 - N - HLOO - HLOO - ... - HLOO - 0
 *
 * Returns zero on success, or a negative errno if a failure occurred.
 * @xdr is advanced to the next position in the stream.
 *
 * Only a single Write chunk is currently supported.
 */
static int rpcrdma_encode_write_list(struct rpcrdma_xprt *r_xprt,
				     struct rpcrdma_req *req,
				     struct rpc_rqst *rqst,
				     enum rpcrdma_chunktype wtype)
{
	struct xdr_stream *xdr = &req->rl_stream;
	struct rpcrdma_mr_seg *seg;
	struct rpcrdma_mr *mr;
	int nsegs, nchunks;
	__be32 *segcount;

	if (wtype != rpcrdma_writech)
		goto done;

	seg = req->rl_segments;
	nsegs = rpcrdma_convert_iovs(r_xprt, &rqst->rq_rcv_buf,
				     rqst->rq_rcv_buf.head[0].iov_len,
				     wtype, seg);
	if (nsegs < 0)
		return nsegs;

	if (encode_item_present(xdr) < 0)
		return -EMSGSIZE;
	segcount = xdr_reserve_space(xdr, sizeof(*segcount));
	if (unlikely(!segcount))
		return -EMSGSIZE;
	/* Actual value encoded below */

	nchunks = 0;
	do {
		seg = rpcrdma_mr_prepare(r_xprt, req, seg, nsegs, true, &mr);
		if (IS_ERR(seg))
			return PTR_ERR(seg);

		if (encode_rdma_segment(xdr, mr) < 0)
			return -EMSGSIZE;

		trace_xprtrdma_chunk_write(rqst->rq_task, mr, nsegs);
		r_xprt->rx_stats.write_chunk_count++;
		r_xprt->rx_stats.total_rdma_request += mr->mr_length;
		nchunks++;
		nsegs -= mr->mr_nents;
	} while (nsegs);

	/* Update count of segments in this Write chunk */
	*segcount = cpu_to_be32(nchunks);

done:
	return encode_item_not_present(xdr);
}

/* Register and XDR encode the Reply chunk. Supports encoding an array
 * of plain segments that belong to a single write (reply) chunk.
 *
 * Encoding key for single-list chunks (HLOO = Handle32 Length32 Offset64):
 *
 *  Reply chunk (a counted array):
 *   N elements:
 *    1 - N - HLOO - HLOO - ... - HLOO
 *
 * Returns zero on success, or a negative errno if a failure occurred.
 * @xdr is advanced to the next position in the stream.
 */
static int rpcrdma_encode_reply_chunk(struct rpcrdma_xprt *r_xprt,
				      struct rpcrdma_req *req,
				      struct rpc_rqst *rqst,
				      enum rpcrdma_chunktype wtype)
{
	struct xdr_stream *xdr = &req->rl_stream;
	struct rpcrdma_mr_seg *seg;
	struct rpcrdma_mr *mr;
	int nsegs, nchunks;
	__be32 *segcount;

	if (wtype != rpcrdma_replych)
		return encode_item_not_present(xdr);

	seg = req->rl_segments;
	nsegs = rpcrdma_convert_iovs(r_xprt, &rqst->rq_rcv_buf, 0, wtype, seg);
	if (nsegs < 0)
		return nsegs;

	if (encode_item_present(xdr) < 0)
		return -EMSGSIZE;
	segcount = xdr_reserve_space(xdr, sizeof(*segcount));
	if (unlikely(!segcount))
		return -EMSGSIZE;
	/* Actual value encoded below */

	nchunks = 0;
	do {
		seg = rpcrdma_mr_prepare(r_xprt, req, seg, nsegs, true, &mr);
		if (IS_ERR(seg))
			return PTR_ERR(seg);

		if (encode_rdma_segment(xdr, mr) < 0)
			return -EMSGSIZE;

		trace_xprtrdma_chunk_reply(rqst->rq_task, mr, nsegs);
		r_xprt->rx_stats.reply_chunk_count++;
		r_xprt->rx_stats.total_rdma_request += mr->mr_length;
		nchunks++;
		nsegs -= mr->mr_nents;
	} while (nsegs);

	/* Update count of segments in the Reply chunk */
	*segcount = cpu_to_be32(nchunks);

	return 0;
}

static void rpcrdma_sendctx_done(struct kref *kref)
{
	struct rpcrdma_req *req =
		container_of(kref, struct rpcrdma_req, rl_kref);
	struct rpcrdma_rep *rep = req->rl_reply;

	rpcrdma_complete_rqst(rep);
	rep->rr_rxprt->rx_stats.reply_waits_for_send++;
}

/**
 * rpcrdma_sendctx_unmap - DMA-unmap Send buffer
 * @sc: sendctx containing SGEs to unmap
 *
 */
void rpcrdma_sendctx_unmap(struct rpcrdma_sendctx *sc)
{
	struct rpcrdma_regbuf *rb = sc->sc_req->rl_sendbuf;
	struct ib_sge *sge;

	if (!sc->sc_unmap_count)
		return;

	/* The first two SGEs contain the transport header and
	 * the inline buffer. These are always left mapped so
	 * they can be cheaply re-used.
	 */
	for (sge = &sc->sc_sges[2]; sc->sc_unmap_count;
	     ++sge, --sc->sc_unmap_count)
		ib_dma_unmap_page(rdmab_device(rb), sge->addr, sge->length,
				  DMA_TO_DEVICE);

	kref_put(&sc->sc_req->rl_kref, rpcrdma_sendctx_done);
}

/* Prepare an SGE for the RPC-over-RDMA transport header.
 */
static bool rpcrdma_prepare_hdr_sge(struct rpcrdma_xprt *r_xprt,
				    struct rpcrdma_req *req, u32 len)
{
	struct rpcrdma_sendctx *sc = req->rl_sendctx;
	struct rpcrdma_regbuf *rb = req->rl_rdmabuf;
	struct ib_sge *sge = &sc->sc_sges[req->rl_wr.num_sge++];

	if (!rpcrdma_regbuf_dma_map(r_xprt, rb))
		return false;
	sge->addr = rdmab_addr(rb);
	sge->length = len;
	sge->lkey = rdmab_lkey(rb);

	ib_dma_sync_single_for_device(rdmab_device(rb), sge->addr, sge->length,
				      DMA_TO_DEVICE);
	return true;
}

/* The head iovec is straightforward, as it is usually already
 * DMA-mapped. Sync the content that has changed.
 */
static bool rpcrdma_prepare_head_iov(struct rpcrdma_xprt *r_xprt,
				     struct rpcrdma_req *req, unsigned int len)
{
	struct rpcrdma_sendctx *sc = req->rl_sendctx;
	struct ib_sge *sge = &sc->sc_sges[req->rl_wr.num_sge++];
	struct rpcrdma_regbuf *rb = req->rl_sendbuf;

	if (!rpcrdma_regbuf_dma_map(r_xprt, rb))
		return false;

	sge->addr = rdmab_addr(rb);
	sge->length = len;
	sge->lkey = rdmab_lkey(rb);

	ib_dma_sync_single_for_device(rdmab_device(rb), sge->addr, sge->length,
				      DMA_TO_DEVICE);
	return true;
}

/* If there is a page list present, DMA map and prepare an
 * SGE for each page to be sent.
 */
static bool rpcrdma_prepare_pagelist(struct rpcrdma_req *req,
				     struct xdr_buf *xdr)
{
	struct rpcrdma_sendctx *sc = req->rl_sendctx;
	struct rpcrdma_regbuf *rb = req->rl_sendbuf;
	unsigned int page_base, len, remaining;
	struct page **ppages;
	struct ib_sge *sge;

	ppages = xdr->pages + (xdr->page_base >> PAGE_SHIFT);
	page_base = offset_in_page(xdr->page_base);
	remaining = xdr->page_len;
	while (remaining) {
		sge = &sc->sc_sges[req->rl_wr.num_sge++];
		len = min_t(unsigned int, PAGE_SIZE - page_base, remaining);
		sge->addr = ib_dma_map_page(rdmab_device(rb), *ppages,
					    page_base, len, DMA_TO_DEVICE);
		if (ib_dma_mapping_error(rdmab_device(rb), sge->addr))
			goto out_mapping_err;

		sge->length = len;
		sge->lkey = rdmab_lkey(rb);

		sc->sc_unmap_count++;
		ppages++;
		remaining -= len;
		page_base = 0;
	}

	return true;

out_mapping_err:
	trace_xprtrdma_dma_maperr(sge->addr);
	return false;
}

/* The tail iovec may include an XDR pad for the page list,
 * as well as additional content, and may not reside in the
 * same page as the head iovec.
 */
static bool rpcrdma_prepare_tail_iov(struct rpcrdma_req *req,
				     struct xdr_buf *xdr,
				     unsigned int page_base, unsigned int len)
{
	struct rpcrdma_sendctx *sc = req->rl_sendctx;
	struct ib_sge *sge = &sc->sc_sges[req->rl_wr.num_sge++];
	struct rpcrdma_regbuf *rb = req->rl_sendbuf;
	struct page *page = virt_to_page(xdr->tail[0].iov_base);

	sge->addr = ib_dma_map_page(rdmab_device(rb), page, page_base, len,
				    DMA_TO_DEVICE);
	if (ib_dma_mapping_error(rdmab_device(rb), sge->addr))
		goto out_mapping_err;

	sge->length = len;
	sge->lkey = rdmab_lkey(rb);
	++sc->sc_unmap_count;
	return true;

out_mapping_err:
	trace_xprtrdma_dma_maperr(sge->addr);
	return false;
}

/* Copy the tail to the end of the head buffer.
 */
static void rpcrdma_pullup_tail_iov(struct rpcrdma_xprt *r_xprt,
				    struct rpcrdma_req *req,
				    struct xdr_buf *xdr)
{
	unsigned char *dst;

	dst = (unsigned char *)xdr->head[0].iov_base;
	dst += xdr->head[0].iov_len + xdr->page_len;
	memmove(dst, xdr->tail[0].iov_base, xdr->tail[0].iov_len);
	r_xprt->rx_stats.pullup_copy_count += xdr->tail[0].iov_len;
}

/* Copy pagelist content into the head buffer.
 */
static void rpcrdma_pullup_pagelist(struct rpcrdma_xprt *r_xprt,
				    struct rpcrdma_req *req,
				    struct xdr_buf *xdr)
{
	unsigned int len, page_base, remaining;
	struct page **ppages;
	unsigned char *src, *dst;

	dst = (unsigned char *)xdr->head[0].iov_base;
	dst += xdr->head[0].iov_len;
	ppages = xdr->pages + (xdr->page_base >> PAGE_SHIFT);
	page_base = offset_in_page(xdr->page_base);
	remaining = xdr->page_len;
	while (remaining) {
		src = page_address(*ppages);
		src += page_base;
		len = min_t(unsigned int, PAGE_SIZE - page_base, remaining);
		memcpy(dst, src, len);
		r_xprt->rx_stats.pullup_copy_count += len;

		ppages++;
		dst += len;
		remaining -= len;
		page_base = 0;
	}
}

/* Copy the contents of @xdr into @rl_sendbuf and DMA sync it.
 * When the head, pagelist, and tail are small, a pull-up copy
 * is considerably less costly than DMA mapping the components
 * of @xdr.
 *
 * Assumptions:
 *  - the caller has already verified that the total length
 *    of the RPC Call body will fit into @rl_sendbuf.
 */
static bool rpcrdma_prepare_noch_pullup(struct rpcrdma_xprt *r_xprt,
					struct rpcrdma_req *req,
					struct xdr_buf *xdr)
{
	if (unlikely(xdr->tail[0].iov_len))
		rpcrdma_pullup_tail_iov(r_xprt, req, xdr);

	if (unlikely(xdr->page_len))
		rpcrdma_pullup_pagelist(r_xprt, req, xdr);

	/* The whole RPC message resides in the head iovec now */
	return rpcrdma_prepare_head_iov(r_xprt, req, xdr->len);
}

static bool rpcrdma_prepare_noch_mapped(struct rpcrdma_xprt *r_xprt,
					struct rpcrdma_req *req,
					struct xdr_buf *xdr)
{
	struct kvec *tail = &xdr->tail[0];

	if (!rpcrdma_prepare_head_iov(r_xprt, req, xdr->head[0].iov_len))
		return false;
	if (xdr->page_len)
		if (!rpcrdma_prepare_pagelist(req, xdr))
			return false;
	if (tail->iov_len)
		if (!rpcrdma_prepare_tail_iov(req, xdr,
					      offset_in_page(tail->iov_base),
					      tail->iov_len))
			return false;

	if (req->rl_sendctx->sc_unmap_count)
		kref_get(&req->rl_kref);
	return true;
}

static bool rpcrdma_prepare_readch(struct rpcrdma_xprt *r_xprt,
				   struct rpcrdma_req *req,
				   struct xdr_buf *xdr)
{
	if (!rpcrdma_prepare_head_iov(r_xprt, req, xdr->head[0].iov_len))
		return false;

	/* If there is a Read chunk, the page list is being handled
	 * via explicit RDMA, and thus is skipped here.
	 */

	/* Do not include the tail if it is only an XDR pad */
	if (xdr->tail[0].iov_len > 3) {
		unsigned int page_base, len;

		/* If the content in the page list is an odd length,
		 * xdr_write_pages() adds a pad at the beginning of
		 * the tail iovec. Force the tail's non-pad content to
		 * land at the next XDR position in the Send message.
		 */
		page_base = offset_in_page(xdr->tail[0].iov_base);
		len = xdr->tail[0].iov_len;
		page_base += len & 3;
		len -= len & 3;
		if (!rpcrdma_prepare_tail_iov(req, xdr, page_base, len))
			return false;
		kref_get(&req->rl_kref);
	}

	return true;
}

/**
 * rpcrdma_prepare_send_sges - Construct SGEs for a Send WR
 * @r_xprt: controlling transport
 * @req: context of RPC Call being marshalled
 * @hdrlen: size of transport header, in bytes
 * @xdr: xdr_buf containing RPC Call
 * @rtype: chunk type being encoded
 *
 * Returns 0 on success; otherwise a negative errno is returned.
 */
inline int rpcrdma_prepare_send_sges(struct rpcrdma_xprt *r_xprt,
				     struct rpcrdma_req *req, u32 hdrlen,
				     struct xdr_buf *xdr,
				     enum rpcrdma_chunktype rtype)
{
	int ret;

	ret = -EAGAIN;
	req->rl_sendctx = rpcrdma_sendctx_get_locked(r_xprt);
	if (!req->rl_sendctx)
		goto out_nosc;
	req->rl_sendctx->sc_unmap_count = 0;
	req->rl_sendctx->sc_req = req;
	kref_init(&req->rl_kref);
	req->rl_wr.wr_cqe = &req->rl_sendctx->sc_cqe;
	req->rl_wr.sg_list = req->rl_sendctx->sc_sges;
	req->rl_wr.num_sge = 0;
	req->rl_wr.opcode = IB_WR_SEND;

	ret = -EIO;
	if (!rpcrdma_prepare_hdr_sge(r_xprt, req, hdrlen))
		goto out_unmap;

	switch (rtype) {
	case rpcrdma_noch_pullup:
		if (!rpcrdma_prepare_noch_pullup(r_xprt, req, xdr))
			goto out_unmap;
		break;
	case rpcrdma_noch_mapped:
		if (!rpcrdma_prepare_noch_mapped(r_xprt, req, xdr))
			goto out_unmap;
		break;
	case rpcrdma_readch:
		if (!rpcrdma_prepare_readch(r_xprt, req, xdr))
			goto out_unmap;
		break;
	case rpcrdma_areadch:
		break;
	default:
		goto out_unmap;
	}

	return 0;

out_unmap:
	rpcrdma_sendctx_unmap(req->rl_sendctx);
out_nosc:
	trace_xprtrdma_prepsend_failed(&req->rl_slot, ret);
	return ret;
}

/**
 * rpcrdma_marshal_req - Marshal and send one RPC request
 * @r_xprt: controlling transport
 * @rqst: RPC request to be marshaled
 *
 * For the RPC in "rqst", this function:
 *  - Chooses the transfer mode (eg., RDMA_MSG or RDMA_NOMSG)
 *  - Registers Read, Write, and Reply chunks
 *  - Constructs the transport header
 *  - Posts a Send WR to send the transport header and request
 *
 * Returns:
 *	%0 if the RPC was sent successfully,
 *	%-ENOTCONN if the connection was lost,
 *	%-EAGAIN if the caller should call again with the same arguments,
 *	%-ENOBUFS if the caller should call again after a delay,
 *	%-EMSGSIZE if the transport header is too small,
 *	%-EIO if a permanent problem occurred while marshaling.
 */
int
rpcrdma_marshal_req(struct rpcrdma_xprt *r_xprt, struct rpc_rqst *rqst)
{
	struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
	struct xdr_stream *xdr = &req->rl_stream;
	enum rpcrdma_chunktype rtype, wtype;
	struct xdr_buf *buf = &rqst->rq_snd_buf;
	bool ddp_allowed;
	__be32 *p;
	int ret;

	rpcrdma_set_xdrlen(&req->rl_hdrbuf, 0);
	xdr_init_encode(xdr, &req->rl_hdrbuf, rdmab_data(req->rl_rdmabuf),
			rqst);

	/* Fixed header fields */
	ret = -EMSGSIZE;
	p = xdr_reserve_space(xdr, 4 * sizeof(*p));
	if (!p)
		goto out_err;
	*p++ = rqst->rq_xid;
	*p++ = rpcrdma_version;
	*p++ = cpu_to_be32(r_xprt->rx_buf.rb_max_requests);

	/* When the ULP employs a GSS flavor that guarantees integrity
	 * or privacy, direct data placement of individual data items
	 * is not allowed.
	 */
	ddp_allowed = !(rqst->rq_cred->cr_auth->au_flags &
						RPCAUTH_AUTH_DATATOUCH);

	/*
	 * Chunks needed for results?
	 *
	 * o If the expected result is under the inline threshold, all ops
	 *   return as inline.
	 * o Large read ops return data as write chunk(s), header as
	 *   inline.
	 * o Large non-read ops return as a single reply chunk.
	 */
	if (rpcrdma_results_inline(r_xprt, rqst))
		wtype = rpcrdma_noch;
	else if ((ddp_allowed && rqst->rq_rcv_buf.flags & XDRBUF_READ) &&
		 rpcrdma_nonpayload_inline(r_xprt, rqst))
		wtype = rpcrdma_writech;
	else
		wtype = rpcrdma_replych;

	/*
	 * Chunks needed for arguments?
	 *
	 * o If the total request is under the inline threshold, all ops
	 *   are sent as inline.
	 * o Large write ops transmit data as read chunk(s), header as
	 *   inline.
	 * o Large non-write ops are sent with the entire message as a
	 *   single read chunk (protocol 0-position special case).
	 *
	 * This assumes that the upper layer does not present a request
	 * that both has a data payload, and whose non-data arguments
	 * by themselves are larger than the inline threshold.
	 */
	if (rpcrdma_args_inline(r_xprt, rqst)) {
		*p++ = rdma_msg;
		rtype = buf->len < rdmab_length(req->rl_sendbuf) ?
			rpcrdma_noch_pullup : rpcrdma_noch_mapped;
	} else if (ddp_allowed && buf->flags & XDRBUF_WRITE) {
		*p++ = rdma_msg;
		rtype = rpcrdma_readch;
	} else {
		r_xprt->rx_stats.nomsg_call_count++;
		*p++ = rdma_nomsg;
		rtype = rpcrdma_areadch;
	}

	/* This implementation supports the following combinations
	 * of chunk lists in one RPC-over-RDMA Call message:
	 *
	 *   - Read list
	 *   - Write list
	 *   - Reply chunk
	 *   - Read list + Reply chunk
	 *
	 * It might not yet support the following combinations:
	 *
	 *   - Read list + Write list
	 *
	 * It does not support the following combinations:
	 *
	 *   - Write list + Reply chunk
	 *   - Read list + Write list + Reply chunk
	 *
	 * This implementation supports only a single chunk in each
	 * Read or Write list. Thus for example the client cannot
	 * send a Call message with a Position Zero Read chunk and a
	 * regular Read chunk at the same time.
	 */
	ret = rpcrdma_encode_read_list(r_xprt, req, rqst, rtype);
	if (ret)
		goto out_err;
	ret = rpcrdma_encode_write_list(r_xprt, req, rqst, wtype);
	if (ret)
		goto out_err;
	ret = rpcrdma_encode_reply_chunk(r_xprt, req, rqst, wtype);
	if (ret)
		goto out_err;

	ret = rpcrdma_prepare_send_sges(r_xprt, req, req->rl_hdrbuf.len,
					buf, rtype);
	if (ret)
		goto out_err;

	trace_xprtrdma_marshal(req, rtype, wtype);
	return 0;

out_err:
	trace_xprtrdma_marshal_failed(rqst, ret);
	r_xprt->rx_stats.failed_marshal_count++;
	frwr_reset(req);
	return ret;
}

static void __rpcrdma_update_cwnd_locked(struct rpc_xprt *xprt,
					 struct rpcrdma_buffer *buf,
					 u32 grant)
{
	buf->rb_credits = grant;
	xprt->cwnd = grant << RPC_CWNDSHIFT;
}

static void rpcrdma_update_cwnd(struct rpcrdma_xprt *r_xprt, u32 grant)
{
	struct rpc_xprt *xprt = &r_xprt->rx_xprt;

	spin_lock(&xprt->transport_lock);
	__rpcrdma_update_cwnd_locked(xprt, &r_xprt->rx_buf, grant);
	spin_unlock(&xprt->transport_lock);
}

/**
 * rpcrdma_reset_cwnd - Reset the xprt's congestion window
 * @r_xprt: controlling transport instance
 *
 * Prepare @r_xprt for the next connection by reinitializing
 * its credit grant to one (see RFC 8166, Section 3.3.3).
 */
void rpcrdma_reset_cwnd(struct rpcrdma_xprt *r_xprt)
{
	struct rpc_xprt *xprt = &r_xprt->rx_xprt;

	spin_lock(&xprt->transport_lock);
	xprt->cong = 0;
	__rpcrdma_update_cwnd_locked(xprt, &r_xprt->rx_buf, 1);
	spin_unlock(&xprt->transport_lock);
}

/**
 * rpcrdma_inline_fixup - Scatter inline received data into rqst's iovecs
 * @rqst: controlling RPC request
 * @srcp: points to RPC message payload in receive buffer
 * @copy_len: remaining length of receive buffer content
 * @pad: Write chunk pad bytes needed (zero for pure inline)
 *
 * The upper layer has set the maximum number of bytes it can
 * receive in each component of rq_rcv_buf. These values are set in
 * the head.iov_len, page_len, tail.iov_len, and buflen fields.
 *
 * Unlike the TCP equivalent (xdr_partial_copy_from_skb), in
 * many cases this function simply updates iov_base pointers in
 * rq_rcv_buf to point directly to the received reply data, to
 * avoid copying reply data.
 *
 * Returns the count of bytes which had to be memcopied.
 */
static unsigned long
rpcrdma_inline_fixup(struct rpc_rqst *rqst, char *srcp, int copy_len, int pad)
{
	unsigned long fixup_copy_count;
	int i, npages, curlen;
	char *destp;
	struct page **ppages;
	int page_base;

	/* The head iovec is redirected to the RPC reply message
	 * in the receive buffer, to avoid a memcopy.
	 */
	rqst->rq_rcv_buf.head[0].iov_base = srcp;
	rqst->rq_private_buf.head[0].iov_base = srcp;

	/* The contents of the receive buffer that follow
	 * head.iov_len bytes are copied into the page list.
	 */
	curlen = rqst->rq_rcv_buf.head[0].iov_len;
	if (curlen > copy_len)
		curlen = copy_len;
	srcp += curlen;
	copy_len -= curlen;

	ppages = rqst->rq_rcv_buf.pages +
		(rqst->rq_rcv_buf.page_base >> PAGE_SHIFT);
	page_base = offset_in_page(rqst->rq_rcv_buf.page_base);
	fixup_copy_count = 0;
	if (copy_len && rqst->rq_rcv_buf.page_len) {
		int pagelist_len;

		pagelist_len = rqst->rq_rcv_buf.page_len;
		if (pagelist_len > copy_len)
			pagelist_len = copy_len;
		npages = PAGE_ALIGN(page_base + pagelist_len) >> PAGE_SHIFT;
		for (i = 0; i < npages; i++) {
			curlen = PAGE_SIZE - page_base;
			if (curlen > pagelist_len)
				curlen = pagelist_len;

			destp = kmap_atomic(ppages[i]);
			memcpy(destp + page_base, srcp, curlen);
			flush_dcache_page(ppages[i]);
			kunmap_atomic(destp);
			srcp += curlen;
			copy_len -= curlen;
			fixup_copy_count += curlen;
			pagelist_len -= curlen;
			if (!pagelist_len)
				break;
			page_base = 0;
		}

		/* Implicit padding for the last segment in a Write
		 * chunk is inserted inline at the front of the tail
		 * iovec. The upper layer ignores the content of
		 * the pad. Simply ensure inline content in the tail
		 * that follows the Write chunk is properly aligned.
		 */
		if (pad)
			srcp -= pad;
	}

	/* The tail iovec is redirected to the remaining data
	 * in the receive buffer, to avoid a memcopy.
	 */
	if (copy_len || pad) {
		rqst->rq_rcv_buf.tail[0].iov_base = srcp;
		rqst->rq_private_buf.tail[0].iov_base = srcp;
	}

	if (fixup_copy_count)
		trace_xprtrdma_fixup(rqst, fixup_copy_count);
	return fixup_copy_count;
}

/* By convention, backchannel calls arrive via rdma_msg type
 * messages, and never populate the chunk lists. This makes
 * the RPC/RDMA header small and fixed in size, so it is
 * straightforward to check the RPC header's direction field.
 */
static bool
rpcrdma_is_bcall(struct rpcrdma_xprt *r_xprt, struct rpcrdma_rep *rep)
#if defined(CONFIG_SUNRPC_BACKCHANNEL)
{
	struct xdr_stream *xdr = &rep->rr_stream;
	__be32 *p;

	if (rep->rr_proc != rdma_msg)
		return false;

	/* Peek at stream contents without advancing. */
	p = xdr_inline_decode(xdr, 0);

	/* Chunk lists */
	if (*p++ != xdr_zero)
		return false;
	if (*p++ != xdr_zero)
		return false;
	if (*p++ != xdr_zero)
		return false;

	/* RPC header */
	if (*p++ != rep->rr_xid)
		return false;
	if (*p != cpu_to_be32(RPC_CALL))
		return false;

	/* Now that we are sure this is a backchannel call,
	 * advance to the RPC header.
	 */
	p = xdr_inline_decode(xdr, 3 * sizeof(*p));
	if (unlikely(!p))
		goto out_short;

	rpcrdma_bc_receive_call(r_xprt, rep);
	return true;

out_short:
	pr_warn("RPC/RDMA short backward direction call\n");
	return true;
}
#else	/* CONFIG_SUNRPC_BACKCHANNEL */
{
	return false;
}
#endif	/* CONFIG_SUNRPC_BACKCHANNEL */

static int decode_rdma_segment(struct xdr_stream *xdr, u32 *length)
{
	u32 handle;
	u64 offset;
	__be32 *p;

	p = xdr_inline_decode(xdr, 4 * sizeof(*p));
	if (unlikely(!p))
		return -EIO;

	handle = be32_to_cpup(p++);
	*length = be32_to_cpup(p++);
	xdr_decode_hyper(p, &offset);

	trace_xprtrdma_decode_seg(handle, *length, offset);
	return 0;
}

static int decode_write_chunk(struct xdr_stream *xdr, u32 *length)
{
	u32 segcount, seglength;
	__be32 *p;

	p = xdr_inline_decode(xdr, sizeof(*p));
	if (unlikely(!p))
		return -EIO;

	*length = 0;
	segcount = be32_to_cpup(p);
	while (segcount--) {
		if (decode_rdma_segment(xdr, &seglength))
			return -EIO;
		*length += seglength;
	}

	return 0;
}

/* In RPC-over-RDMA Version One replies, a Read list is never
 * expected. This decoder is a stub that returns an error if
 * a Read list is present.
 */
static int decode_read_list(struct xdr_stream *xdr)
{
	__be32 *p;

	p = xdr_inline_decode(xdr, sizeof(*p));
	if (unlikely(!p))
		return -EIO;
	if (unlikely(*p != xdr_zero))
		return -EIO;
	return 0;
}

/* Supports only one Write chunk in the Write list
 */
static int decode_write_list(struct xdr_stream *xdr, u32 *length)
{
	u32 chunklen;
	bool first;
	__be32 *p;

	*length = 0;
	first = true;
	do {
		p = xdr_inline_decode(xdr, sizeof(*p));
		if (unlikely(!p))
			return -EIO;
		if (*p == xdr_zero)
			break;
		if (!first)
			return -EIO;

		if (decode_write_chunk(xdr, &chunklen))
			return -EIO;
		*length += chunklen;
		first = false;
	} while (true);
	return 0;
}

static int decode_reply_chunk(struct xdr_stream *xdr, u32 *length)
{
	__be32 *p;

	p = xdr_inline_decode(xdr, sizeof(*p));
	if (unlikely(!p))
		return -EIO;

	*length = 0;
	if (*p != xdr_zero)
		if (decode_write_chunk(xdr, length))
			return -EIO;
	return 0;
}

static int
rpcrdma_decode_msg(struct rpcrdma_xprt *r_xprt, struct rpcrdma_rep *rep,
		   struct rpc_rqst *rqst)
{
	struct xdr_stream *xdr = &rep->rr_stream;
	u32 writelist, replychunk, rpclen;
	char *base;

	/* Decode the chunk lists */
	if (decode_read_list(xdr))
		return -EIO;
	if (decode_write_list(xdr, &writelist))
		return -EIO;
	if (decode_reply_chunk(xdr, &replychunk))
		return -EIO;

	/* RDMA_MSG sanity checks */
	if (unlikely(replychunk))
		return -EIO;

	/* Build the RPC reply's Payload stream in rqst->rq_rcv_buf */
	base = (char *)xdr_inline_decode(xdr, 0);
	rpclen = xdr_stream_remaining(xdr);
	r_xprt->rx_stats.fixup_copy_count +=
		rpcrdma_inline_fixup(rqst, base, rpclen, writelist & 3);

	r_xprt->rx_stats.total_rdma_reply += writelist;
	return rpclen + xdr_align_size(writelist);
}

static noinline int
rpcrdma_decode_nomsg(struct rpcrdma_xprt *r_xprt, struct rpcrdma_rep *rep)
{
	struct xdr_stream *xdr = &rep->rr_stream;
	u32 writelist, replychunk;

	/* Decode the chunk lists */
	if (decode_read_list(xdr))
		return -EIO;
	if (decode_write_list(xdr, &writelist))
		return -EIO;
	if (decode_reply_chunk(xdr, &replychunk))
		return -EIO;

	/* RDMA_NOMSG sanity checks */
	if (unlikely(writelist))
		return -EIO;
	if (unlikely(!replychunk))
		return -EIO;

	/* Reply chunk buffer already is the reply vector */
	r_xprt->rx_stats.total_rdma_reply += replychunk;
	return replychunk;
}

static noinline int
rpcrdma_decode_error(struct rpcrdma_xprt *r_xprt, struct rpcrdma_rep *rep,
		     struct rpc_rqst *rqst)
{
	struct xdr_stream *xdr = &rep->rr_stream;
	__be32 *p;

	p = xdr_inline_decode(xdr, sizeof(*p));
	if (unlikely(!p))
		return -EIO;

	switch (*p) {
	case err_vers:
		p = xdr_inline_decode(xdr, 2 * sizeof(*p));
		if (!p)
			break;
		dprintk("RPC:       %s: server reports "
			"version error (%u-%u), xid %08x\n", __func__,
			be32_to_cpup(p), be32_to_cpu(*(p + 1)),
			be32_to_cpu(rep->rr_xid));
		break;
	case err_chunk:
		dprintk("RPC:       %s: server reports "
			"header decoding error, xid %08x\n", __func__,
			be32_to_cpu(rep->rr_xid));
		break;
	default:
		dprintk("RPC:       %s: server reports "
			"unrecognized error %d, xid %08x\n", __func__,
			be32_to_cpup(p), be32_to_cpu(rep->rr_xid));
	}

	r_xprt->rx_stats.bad_reply_count++;
	return -EREMOTEIO;
}

/* Perform XID lookup, reconstruction of the RPC reply, and
 * RPC completion while holding the transport lock to ensure
 * the rep, rqst, and rq_task pointers remain stable.
 */
void rpcrdma_complete_rqst(struct rpcrdma_rep *rep)
{
	struct rpcrdma_xprt *r_xprt = rep->rr_rxprt;
	struct rpc_xprt *xprt = &r_xprt->rx_xprt;
	struct rpc_rqst *rqst = rep->rr_rqst;
	int status;

	switch (rep->rr_proc) {
	case rdma_msg:
		status = rpcrdma_decode_msg(r_xprt, rep, rqst);
		break;
	case rdma_nomsg:
		status = rpcrdma_decode_nomsg(r_xprt, rep);
		break;
	case rdma_error:
		status = rpcrdma_decode_error(r_xprt, rep, rqst);
		break;
	default:
		status = -EIO;
	}
	if (status < 0)
		goto out_badheader;

out:
	spin_lock(&xprt->queue_lock);
	xprt_complete_rqst(rqst->rq_task, status);
	xprt_unpin_rqst(rqst);
	spin_unlock(&xprt->queue_lock);
	return;

/* If the incoming reply terminated a pending RPC, the next
 * RPC call will post a replacement receive buffer as it is
 * being marshaled.
 */
out_badheader:
	trace_xprtrdma_reply_hdr(rep);
	r_xprt->rx_stats.bad_reply_count++;
	goto out;
}

static void rpcrdma_reply_done(struct kref *kref)
{
	struct rpcrdma_req *req =
		container_of(kref, struct rpcrdma_req, rl_kref);

	rpcrdma_complete_rqst(req->rl_reply);
}

/**
 * rpcrdma_reply_handler - Process received RPC/RDMA messages
 * @rep: Incoming rpcrdma_rep object to process
 *
 * Errors must result in the RPC task either being awakened, or
 * allowed to timeout, to discover the errors at that time.
 */
void rpcrdma_reply_handler(struct rpcrdma_rep *rep)
{
	struct rpcrdma_xprt *r_xprt = rep->rr_rxprt;
	struct rpc_xprt *xprt = &r_xprt->rx_xprt;
	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
	struct rpcrdma_req *req;
	struct rpc_rqst *rqst;
	u32 credits;
	__be32 *p;

	/* Any data means we had a useful conversation, so
	 * then we don't need to delay the next reconnect.
	 */
	if (xprt->reestablish_timeout)
		xprt->reestablish_timeout = 0;

	/* Fixed transport header fields */
	xdr_init_decode(&rep->rr_stream, &rep->rr_hdrbuf,
			rep->rr_hdrbuf.head[0].iov_base, NULL);
	p = xdr_inline_decode(&rep->rr_stream, 4 * sizeof(*p));
	if (unlikely(!p))
		goto out_shortreply;
	rep->rr_xid = *p++;
	rep->rr_vers = *p++;
	credits = be32_to_cpu(*p++);
	rep->rr_proc = *p++;

	if (rep->rr_vers != rpcrdma_version)
		goto out_badversion;

	if (rpcrdma_is_bcall(r_xprt, rep))
		return;

	/* Match incoming rpcrdma_rep to an rpcrdma_req to
	 * get context for handling any incoming chunks.
	 */
	spin_lock(&xprt->queue_lock);
	rqst = xprt_lookup_rqst(xprt, rep->rr_xid);
	if (!rqst)
		goto out_norqst;
	xprt_pin_rqst(rqst);
	spin_unlock(&xprt->queue_lock);

	if (credits == 0)
		credits = 1;	/* don't deadlock */
	else if (credits > buf->rb_max_requests)
		credits = buf->rb_max_requests;
	if (buf->rb_credits != credits)
		rpcrdma_update_cwnd(r_xprt, credits);
	rpcrdma_post_recvs(r_xprt, false);

	req = rpcr_to_rdmar(rqst);
	if (req->rl_reply) {
		trace_xprtrdma_leaked_rep(rqst, req->rl_reply);
		rpcrdma_recv_buffer_put(req->rl_reply);
	}
	req->rl_reply = rep;
	rep->rr_rqst = rqst;

	trace_xprtrdma_reply(rqst->rq_task, rep, req, credits);

	if (rep->rr_wc_flags & IB_WC_WITH_INVALIDATE)
		frwr_reminv(rep, &req->rl_registered);
	if (!list_empty(&req->rl_registered))
		frwr_unmap_async(r_xprt, req);
		/* LocalInv completion will complete the RPC */
	else
		kref_put(&req->rl_kref, rpcrdma_reply_done);
	return;

out_badversion:
	trace_xprtrdma_reply_vers(rep);
	goto out;

out_norqst:
	spin_unlock(&xprt->queue_lock);
	trace_xprtrdma_reply_rqst(rep);
	goto out;

out_shortreply:
	trace_xprtrdma_reply_short(rep);

out:
	rpcrdma_recv_buffer_put(rep);
}