summaryrefslogtreecommitdiff
path: root/drivers/infiniband/sw/siw/siw_qp_rx.c
blob: c0a887240325854bb9af0de94d2e86ff6c9e0f4f (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
// SPDX-License-Identifier: GPL-2.0 or BSD-3-Clause

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

#include <linux/errno.h>
#include <linux/types.h>
#include <linux/net.h>
#include <linux/scatterlist.h>
#include <linux/highmem.h>

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

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

/*
 * siw_rx_umem()
 *
 * Receive data of @len into target referenced by @dest_addr.
 *
 * @srx:	Receive Context
 * @umem:	siw representation of target memory
 * @dest_addr:	user virtual address
 * @len:	number of bytes to place
 */
static int siw_rx_umem(struct siw_rx_stream *srx, struct siw_umem *umem,
		       u64 dest_addr, int len)
{
	int copied = 0;

	while (len) {
		struct page *p;
		int pg_off, bytes, rv;
		void *dest;

		p = siw_get_upage(umem, dest_addr);
		if (unlikely(!p)) {
			pr_warn("siw: %s: [QP %u]: bogus addr: %pK, %pK\n",
				__func__, qp_id(rx_qp(srx)),
				(void *)(uintptr_t)dest_addr,
				(void *)(uintptr_t)umem->fp_addr);
			/* siw internal error */
			srx->skb_copied += copied;
			srx->skb_new -= copied;

			return -EFAULT;
		}
		pg_off = dest_addr & ~PAGE_MASK;
		bytes = min(len, (int)PAGE_SIZE - pg_off);

		siw_dbg_qp(rx_qp(srx), "page %pK, bytes=%u\n", p, bytes);

		dest = kmap_atomic(p);
		rv = skb_copy_bits(srx->skb, srx->skb_offset, dest + pg_off,
				   bytes);

		if (unlikely(rv)) {
			kunmap_atomic(dest);
			srx->skb_copied += copied;
			srx->skb_new -= copied;

			pr_warn("siw: [QP %u]: %s, len %d, page %p, rv %d\n",
				qp_id(rx_qp(srx)), __func__, len, p, rv);

			return -EFAULT;
		}
		if (srx->mpa_crc_hd) {
			if (rx_qp(srx)->kernel_verbs) {
				crypto_shash_update(srx->mpa_crc_hd,
					(u8 *)(dest + pg_off), bytes);
				kunmap_atomic(dest);
			} else {
				kunmap_atomic(dest);
				/*
				 * Do CRC on original, not target buffer.
				 * Some user land applications may
				 * concurrently write the target buffer,
				 * which would yield a broken CRC.
				 * Walking the skb twice is very ineffcient.
				 * Folding the CRC into skb_copy_bits()
				 * would be much better, but is currently
				 * not supported.
				 */
				siw_crc_skb(srx, bytes);
			}
		} else {
			kunmap_atomic(dest);
		}
		srx->skb_offset += bytes;
		copied += bytes;
		len -= bytes;
		dest_addr += bytes;
		pg_off = 0;
	}
	srx->skb_copied += copied;
	srx->skb_new -= copied;

	return copied;
}

static int siw_rx_kva(struct siw_rx_stream *srx, void *kva, int len)
{
	int rv;

	siw_dbg_qp(rx_qp(srx), "kva: 0x%pK, len: %u\n", kva, len);

	rv = skb_copy_bits(srx->skb, srx->skb_offset, kva, len);
	if (unlikely(rv)) {
		pr_warn("siw: [QP %u]: %s, len %d, kva 0x%pK, rv %d\n",
			qp_id(rx_qp(srx)), __func__, len, kva, rv);

		return rv;
	}
	if (srx->mpa_crc_hd)
		crypto_shash_update(srx->mpa_crc_hd, (u8 *)kva, len);

	srx->skb_offset += len;
	srx->skb_copied += len;
	srx->skb_new -= len;

	return len;
}

static int siw_rx_pbl(struct siw_rx_stream *srx, int *pbl_idx,
		      struct siw_mem *mem, u64 addr, int len)
{
	struct siw_pbl *pbl = mem->pbl;
	u64 offset = addr - mem->va;
	int copied = 0;

	while (len) {
		int bytes;
		dma_addr_t buf_addr =
			siw_pbl_get_buffer(pbl, offset, &bytes, pbl_idx);
		if (!buf_addr)
			break;

		bytes = min(bytes, len);
		if (siw_rx_kva(srx, (void *)buf_addr, bytes) == bytes) {
			copied += bytes;
			offset += bytes;
			len -= bytes;
		} else {
			break;
		}
	}
	return copied;
}

/*
 * siw_rresp_check_ntoh()
 *
 * Check incoming RRESP fragment header against expected
 * header values and update expected values for potential next
 * fragment.
 *
 * NOTE: This function must be called only if a RRESP DDP segment
 *       starts but not for fragmented consecutive pieces of an
 *       already started DDP segment.
 */
static int siw_rresp_check_ntoh(struct siw_rx_stream *srx,
				struct siw_rx_fpdu *frx)
{
	struct iwarp_rdma_rresp *rresp = &srx->hdr.rresp;
	struct siw_wqe *wqe = &frx->wqe_active;
	enum ddp_ecode ecode;

	u32 sink_stag = be32_to_cpu(rresp->sink_stag);
	u64 sink_to = be64_to_cpu(rresp->sink_to);

	if (frx->first_ddp_seg) {
		srx->ddp_stag = wqe->sqe.sge[0].lkey;
		srx->ddp_to = wqe->sqe.sge[0].laddr;
		frx->pbl_idx = 0;
	}
	/* Below checks extend beyond the semantics of DDP, and
	 * into RDMAP:
	 * We check if the read response matches exactly the
	 * read request which was send to the remote peer to
	 * trigger this read response. RFC5040/5041 do not
	 * always have a proper error code for the detected
	 * error cases. We choose 'base or bounds error' for
	 * cases where the inbound STag is valid, but offset
	 * or length do not match our response receive state.
	 */
	if (unlikely(srx->ddp_stag != sink_stag)) {
		pr_warn("siw: [QP %u]: rresp stag: %08x != %08x\n",
			qp_id(rx_qp(srx)), sink_stag, srx->ddp_stag);
		ecode = DDP_ECODE_T_INVALID_STAG;
		goto error;
	}
	if (unlikely(srx->ddp_to != sink_to)) {
		pr_warn("siw: [QP %u]: rresp off: %016llx != %016llx\n",
			qp_id(rx_qp(srx)), (unsigned long long)sink_to,
			(unsigned long long)srx->ddp_to);
		ecode = DDP_ECODE_T_BASE_BOUNDS;
		goto error;
	}
	if (unlikely(!frx->more_ddp_segs &&
		     (wqe->processed + srx->fpdu_part_rem != wqe->bytes))) {
		pr_warn("siw: [QP %u]: rresp len: %d != %d\n",
			qp_id(rx_qp(srx)),
			wqe->processed + srx->fpdu_part_rem, wqe->bytes);
		ecode = DDP_ECODE_T_BASE_BOUNDS;
		goto error;
	}
	return 0;
error:
	siw_init_terminate(rx_qp(srx), TERM_ERROR_LAYER_DDP,
			   DDP_ETYPE_TAGGED_BUF, ecode, 0);
	return -EINVAL;
}

/*
 * siw_write_check_ntoh()
 *
 * Check incoming WRITE fragment header against expected
 * header values and update expected values for potential next
 * fragment
 *
 * NOTE: This function must be called only if a WRITE DDP segment
 *       starts but not for fragmented consecutive pieces of an
 *       already started DDP segment.
 */
static int siw_write_check_ntoh(struct siw_rx_stream *srx,
				struct siw_rx_fpdu *frx)
{
	struct iwarp_rdma_write *write = &srx->hdr.rwrite;
	enum ddp_ecode ecode;

	u32 sink_stag = be32_to_cpu(write->sink_stag);
	u64 sink_to = be64_to_cpu(write->sink_to);

	if (frx->first_ddp_seg) {
		srx->ddp_stag = sink_stag;
		srx->ddp_to = sink_to;
		frx->pbl_idx = 0;
	} else {
		if (unlikely(srx->ddp_stag != sink_stag)) {
			pr_warn("siw: [QP %u]: write stag: %08x != %08x\n",
				qp_id(rx_qp(srx)), sink_stag,
				srx->ddp_stag);
			ecode = DDP_ECODE_T_INVALID_STAG;
			goto error;
		}
		if (unlikely(srx->ddp_to != sink_to)) {
			pr_warn("siw: [QP %u]: write off: %016llx != %016llx\n",
				qp_id(rx_qp(srx)),
				(unsigned long long)sink_to,
				(unsigned long long)srx->ddp_to);
			ecode = DDP_ECODE_T_BASE_BOUNDS;
			goto error;
		}
	}
	return 0;
error:
	siw_init_terminate(rx_qp(srx), TERM_ERROR_LAYER_DDP,
			   DDP_ETYPE_TAGGED_BUF, ecode, 0);
	return -EINVAL;
}

/*
 * siw_send_check_ntoh()
 *
 * Check incoming SEND fragment header against expected
 * header values and update expected MSN if no next
 * fragment expected
 *
 * NOTE: This function must be called only if a SEND DDP segment
 *       starts but not for fragmented consecutive pieces of an
 *       already started DDP segment.
 */
static int siw_send_check_ntoh(struct siw_rx_stream *srx,
			       struct siw_rx_fpdu *frx)
{
	struct iwarp_send_inv *send = &srx->hdr.send_inv;
	struct siw_wqe *wqe = &frx->wqe_active;
	enum ddp_ecode ecode;

	u32 ddp_msn = be32_to_cpu(send->ddp_msn);
	u32 ddp_mo = be32_to_cpu(send->ddp_mo);
	u32 ddp_qn = be32_to_cpu(send->ddp_qn);

	if (unlikely(ddp_qn != RDMAP_UNTAGGED_QN_SEND)) {
		pr_warn("siw: [QP %u]: invalid ddp qn %d for send\n",
			qp_id(rx_qp(srx)), ddp_qn);
		ecode = DDP_ECODE_UT_INVALID_QN;
		goto error;
	}
	if (unlikely(ddp_msn != srx->ddp_msn[RDMAP_UNTAGGED_QN_SEND])) {
		pr_warn("siw: [QP %u]: send msn: %u != %u\n",
			qp_id(rx_qp(srx)), ddp_msn,
			srx->ddp_msn[RDMAP_UNTAGGED_QN_SEND]);
		ecode = DDP_ECODE_UT_INVALID_MSN_RANGE;
		goto error;
	}
	if (unlikely(ddp_mo != wqe->processed)) {
		pr_warn("siw: [QP %u], send mo: %u != %u\n",
			qp_id(rx_qp(srx)), ddp_mo, wqe->processed);
		ecode = DDP_ECODE_UT_INVALID_MO;
		goto error;
	}
	if (frx->first_ddp_seg) {
		/* initialize user memory write position */
		frx->sge_idx = 0;
		frx->sge_off = 0;
		frx->pbl_idx = 0;

		/* only valid for SEND_INV and SEND_SE_INV operations */
		srx->inval_stag = be32_to_cpu(send->inval_stag);
	}
	if (unlikely(wqe->bytes < wqe->processed + srx->fpdu_part_rem)) {
		siw_dbg_qp(rx_qp(srx), "receive space short: %d - %d < %d\n",
			   wqe->bytes, wqe->processed, srx->fpdu_part_rem);
		wqe->wc_status = SIW_WC_LOC_LEN_ERR;
		ecode = DDP_ECODE_UT_INVALID_MSN_NOBUF;
		goto error;
	}
	return 0;
error:
	siw_init_terminate(rx_qp(srx), TERM_ERROR_LAYER_DDP,
			   DDP_ETYPE_UNTAGGED_BUF, ecode, 0);
	return -EINVAL;
}

static struct siw_wqe *siw_rqe_get(struct siw_qp *qp)
{
	struct siw_rqe *rqe;
	struct siw_srq *srq;
	struct siw_wqe *wqe = NULL;
	bool srq_event = false;
	unsigned long flags;

	srq = qp->srq;
	if (srq) {
		spin_lock_irqsave(&srq->lock, flags);
		if (unlikely(!srq->num_rqe))
			goto out;

		rqe = &srq->recvq[srq->rq_get % srq->num_rqe];
	} else {
		if (unlikely(!qp->recvq))
			goto out;

		rqe = &qp->recvq[qp->rq_get % qp->attrs.rq_size];
	}
	if (likely(rqe->flags == SIW_WQE_VALID)) {
		int num_sge = rqe->num_sge;

		if (likely(num_sge <= SIW_MAX_SGE)) {
			int i = 0;

			wqe = rx_wqe(&qp->rx_untagged);
			rx_type(wqe) = SIW_OP_RECEIVE;
			wqe->wr_status = SIW_WR_INPROGRESS;
			wqe->bytes = 0;
			wqe->processed = 0;

			wqe->rqe.id = rqe->id;
			wqe->rqe.num_sge = num_sge;

			while (i < num_sge) {
				wqe->rqe.sge[i].laddr = rqe->sge[i].laddr;
				wqe->rqe.sge[i].lkey = rqe->sge[i].lkey;
				wqe->rqe.sge[i].length = rqe->sge[i].length;
				wqe->bytes += wqe->rqe.sge[i].length;
				wqe->mem[i] = NULL;
				i++;
			}
			/* can be re-used by appl */
			smp_store_mb(rqe->flags, 0);
		} else {
			siw_dbg_qp(qp, "too many sge's: %d\n", rqe->num_sge);
			if (srq)
				spin_unlock_irqrestore(&srq->lock, flags);
			return NULL;
		}
		if (!srq) {
			qp->rq_get++;
		} else {
			if (srq->armed) {
				/* Test SRQ limit */
				u32 off = (srq->rq_get + srq->limit) %
					  srq->num_rqe;
				struct siw_rqe *rqe2 = &srq->recvq[off];

				if (!(rqe2->flags & SIW_WQE_VALID)) {
					srq->armed = 0;
					srq_event = true;
				}
			}
			srq->rq_get++;
		}
	}
out:
	if (srq) {
		spin_unlock_irqrestore(&srq->lock, flags);
		if (srq_event)
			siw_srq_event(srq, IB_EVENT_SRQ_LIMIT_REACHED);
	}
	return wqe;
}

/*
 * siw_proc_send:
 *
 * Process one incoming SEND and place data into memory referenced by
 * receive wqe.
 *
 * Function supports partially received sends (suspending/resuming
 * current receive wqe processing)
 *
 * return value:
 *	0:       reached the end of a DDP segment
 *	-EAGAIN: to be called again to finish the DDP segment
 */
int siw_proc_send(struct siw_qp *qp)
{
	struct siw_rx_stream *srx = &qp->rx_stream;
	struct siw_rx_fpdu *frx = &qp->rx_untagged;
	struct siw_wqe *wqe;
	u32 data_bytes; /* all data bytes available */
	u32 rcvd_bytes; /* sum of data bytes rcvd */
	int rv = 0;

	if (frx->first_ddp_seg) {
		wqe = siw_rqe_get(qp);
		if (unlikely(!wqe)) {
			siw_init_terminate(qp, TERM_ERROR_LAYER_DDP,
					   DDP_ETYPE_UNTAGGED_BUF,
					   DDP_ECODE_UT_INVALID_MSN_NOBUF, 0);
			return -ENOENT;
		}
	} else {
		wqe = rx_wqe(frx);
	}
	if (srx->state == SIW_GET_DATA_START) {
		rv = siw_send_check_ntoh(srx, frx);
		if (unlikely(rv)) {
			siw_qp_event(qp, IB_EVENT_QP_FATAL);
			return rv;
		}
		if (!srx->fpdu_part_rem) /* zero length SEND */
			return 0;
	}
	data_bytes = min(srx->fpdu_part_rem, srx->skb_new);
	rcvd_bytes = 0;

	/* A zero length SEND will skip below loop */
	while (data_bytes) {
		struct ib_pd *pd;
		struct siw_mem **mem, *mem_p;
		struct siw_sge *sge;
		u32 sge_bytes; /* data bytes avail for SGE */

		sge = &wqe->rqe.sge[frx->sge_idx];

		if (!sge->length) {
			/* just skip empty sge's */
			frx->sge_idx++;
			frx->sge_off = 0;
			frx->pbl_idx = 0;
			continue;
		}
		sge_bytes = min(data_bytes, sge->length - frx->sge_off);
		mem = &wqe->mem[frx->sge_idx];

		/*
		 * check with QP's PD if no SRQ present, SRQ's PD otherwise
		 */
		pd = qp->srq == NULL ? qp->pd : qp->srq->base_srq.pd;

		rv = siw_check_sge(pd, sge, mem, IB_ACCESS_LOCAL_WRITE,
				   frx->sge_off, sge_bytes);
		if (unlikely(rv)) {
			siw_init_terminate(qp, TERM_ERROR_LAYER_DDP,
					   DDP_ETYPE_CATASTROPHIC,
					   DDP_ECODE_CATASTROPHIC, 0);

			siw_qp_event(qp, IB_EVENT_QP_ACCESS_ERR);
			break;
		}
		mem_p = *mem;
		if (mem_p->mem_obj == NULL)
			rv = siw_rx_kva(srx,
				(void *)(uintptr_t)(sge->laddr + frx->sge_off),
				sge_bytes);
		else if (!mem_p->is_pbl)
			rv = siw_rx_umem(srx, mem_p->umem,
					 sge->laddr + frx->sge_off, sge_bytes);
		else
			rv = siw_rx_pbl(srx, &frx->pbl_idx, mem_p,
					sge->laddr + frx->sge_off, sge_bytes);

		if (unlikely(rv != sge_bytes)) {
			wqe->processed += rcvd_bytes;

			siw_init_terminate(qp, TERM_ERROR_LAYER_DDP,
					   DDP_ETYPE_CATASTROPHIC,
					   DDP_ECODE_CATASTROPHIC, 0);
			return -EINVAL;
		}
		frx->sge_off += rv;

		if (frx->sge_off == sge->length) {
			frx->sge_idx++;
			frx->sge_off = 0;
			frx->pbl_idx = 0;
		}
		data_bytes -= rv;
		rcvd_bytes += rv;

		srx->fpdu_part_rem -= rv;
		srx->fpdu_part_rcvd += rv;
	}
	wqe->processed += rcvd_bytes;

	if (!srx->fpdu_part_rem)
		return 0;

	return (rv < 0) ? rv : -EAGAIN;
}

/*
 * siw_proc_write:
 *
 * Place incoming WRITE after referencing and checking target buffer

 * Function supports partially received WRITEs (suspending/resuming
 * current receive processing)
 *
 * return value:
 *	0:       reached the end of a DDP segment
 *	-EAGAIN: to be called again to finish the DDP segment
 */
int siw_proc_write(struct siw_qp *qp)
{
	struct siw_rx_stream *srx = &qp->rx_stream;
	struct siw_rx_fpdu *frx = &qp->rx_tagged;
	struct siw_mem *mem;
	int bytes, rv;

	if (srx->state == SIW_GET_DATA_START) {
		if (!srx->fpdu_part_rem) /* zero length WRITE */
			return 0;

		rv = siw_write_check_ntoh(srx, frx);
		if (unlikely(rv)) {
			siw_qp_event(qp, IB_EVENT_QP_FATAL);
			return rv;
		}
	}
	bytes = min(srx->fpdu_part_rem, srx->skb_new);

	if (frx->first_ddp_seg) {
		struct siw_wqe *wqe = rx_wqe(frx);

		rx_mem(frx) = siw_mem_id2obj(qp->sdev, srx->ddp_stag >> 8);
		if (unlikely(!rx_mem(frx))) {
			siw_dbg_qp(qp,
				   "sink stag not found/invalid, stag 0x%08x\n",
				   srx->ddp_stag);

			siw_init_terminate(qp, TERM_ERROR_LAYER_DDP,
					   DDP_ETYPE_TAGGED_BUF,
					   DDP_ECODE_T_INVALID_STAG, 0);
			return -EINVAL;
		}
		wqe->rqe.num_sge = 1;
		rx_type(wqe) = SIW_OP_WRITE;
		wqe->wr_status = SIW_WR_INPROGRESS;
	}
	mem = rx_mem(frx);

	/*
	 * Check if application re-registered memory with different
	 * key field of STag.
	 */
	if (unlikely(mem->stag != srx->ddp_stag)) {
		siw_init_terminate(qp, TERM_ERROR_LAYER_DDP,
				   DDP_ETYPE_TAGGED_BUF,
				   DDP_ECODE_T_INVALID_STAG, 0);
		return -EINVAL;
	}
	rv = siw_check_mem(qp->pd, mem, srx->ddp_to + srx->fpdu_part_rcvd,
			   IB_ACCESS_REMOTE_WRITE, bytes);
	if (unlikely(rv)) {
		siw_init_terminate(qp, TERM_ERROR_LAYER_DDP,
				   DDP_ETYPE_TAGGED_BUF, siw_tagged_error(-rv),
				   0);

		siw_qp_event(qp, IB_EVENT_QP_ACCESS_ERR);

		return -EINVAL;
	}

	if (mem->mem_obj == NULL)
		rv = siw_rx_kva(srx,
			(void *)(uintptr_t)(srx->ddp_to + srx->fpdu_part_rcvd),
			bytes);
	else if (!mem->is_pbl)
		rv = siw_rx_umem(srx, mem->umem,
				 srx->ddp_to + srx->fpdu_part_rcvd, bytes);
	else
		rv = siw_rx_pbl(srx, &frx->pbl_idx, mem,
				srx->ddp_to + srx->fpdu_part_rcvd, bytes);

	if (unlikely(rv != bytes)) {
		siw_init_terminate(qp, TERM_ERROR_LAYER_DDP,
				   DDP_ETYPE_CATASTROPHIC,
				   DDP_ECODE_CATASTROPHIC, 0);
		return -EINVAL;
	}
	srx->fpdu_part_rem -= rv;
	srx->fpdu_part_rcvd += rv;

	if (!srx->fpdu_part_rem) {
		srx->ddp_to += srx->fpdu_part_rcvd;
		return 0;
	}
	return -EAGAIN;
}

/*
 * Inbound RREQ's cannot carry user data.
 */
int siw_proc_rreq(struct siw_qp *qp)
{
	struct siw_rx_stream *srx = &qp->rx_stream;

	if (!srx->fpdu_part_rem)
		return 0;

	pr_warn("siw: [QP %u]: rreq with mpa len %d\n", qp_id(qp),
		be16_to_cpu(srx->hdr.ctrl.mpa_len));

	return -EPROTO;
}

/*
 * siw_init_rresp:
 *
 * Process inbound RDMA READ REQ. Produce a pseudo READ RESPONSE WQE.
 * Put it at the tail of the IRQ, if there is another WQE currently in
 * transmit processing. If not, make it the current WQE to be processed
 * and schedule transmit processing.
 *
 * Can be called from softirq context and from process
 * context (RREAD socket loopback case!)
 *
 * return value:
 *	0:      success,
 *		failure code otherwise
 */

static int siw_init_rresp(struct siw_qp *qp, struct siw_rx_stream *srx)
{
	struct siw_wqe *tx_work = tx_wqe(qp);
	struct siw_sqe *resp;

	uint64_t raddr = be64_to_cpu(srx->hdr.rreq.sink_to),
		 laddr = be64_to_cpu(srx->hdr.rreq.source_to);
	uint32_t length = be32_to_cpu(srx->hdr.rreq.read_size),
		 lkey = be32_to_cpu(srx->hdr.rreq.source_stag),
		 rkey = be32_to_cpu(srx->hdr.rreq.sink_stag),
		 msn = be32_to_cpu(srx->hdr.rreq.ddp_msn);

	int run_sq = 1, rv = 0;
	unsigned long flags;

	if (unlikely(msn != srx->ddp_msn[RDMAP_UNTAGGED_QN_RDMA_READ])) {
		siw_init_terminate(qp, TERM_ERROR_LAYER_DDP,
				   DDP_ETYPE_UNTAGGED_BUF,
				   DDP_ECODE_UT_INVALID_MSN_RANGE, 0);
		return -EPROTO;
	}
	spin_lock_irqsave(&qp->sq_lock, flags);

	if (tx_work->wr_status == SIW_WR_IDLE) {
		/*
		 * immediately schedule READ response w/o
		 * consuming IRQ entry: IRQ must be empty.
		 */
		tx_work->processed = 0;
		tx_work->mem[0] = NULL;
		tx_work->wr_status = SIW_WR_QUEUED;
		resp = &tx_work->sqe;
	} else {
		resp = irq_alloc_free(qp);
		run_sq = 0;
	}
	if (likely(resp)) {
		resp->opcode = SIW_OP_READ_RESPONSE;

		resp->sge[0].length = length;
		resp->sge[0].laddr = laddr;
		resp->sge[0].lkey = lkey;

		/* Keep aside message sequence number for potential
		 * error reporting during Read Response generation.
		 */
		resp->sge[1].length = msn;

		resp->raddr = raddr;
		resp->rkey = rkey;
		resp->num_sge = length ? 1 : 0;

		/* RRESP now valid as current TX wqe or placed into IRQ */
		smp_store_mb(resp->flags, SIW_WQE_VALID);
	} else {
		pr_warn("siw: [QP %u]: irq %d exceeded %d\n", qp_id(qp),
			qp->irq_put % qp->attrs.irq_size, qp->attrs.irq_size);

		siw_init_terminate(qp, TERM_ERROR_LAYER_RDMAP,
				   RDMAP_ETYPE_REMOTE_OPERATION,
				   RDMAP_ECODE_CATASTROPHIC_STREAM, 0);
		rv = -EPROTO;
	}

	spin_unlock_irqrestore(&qp->sq_lock, flags);

	if (run_sq)
		rv = siw_sq_start(qp);

	return rv;
}

/*
 * Only called at start of Read.Resonse processing.
 * Transfer pending Read from tip of ORQ into currrent rx wqe,
 * but keep ORQ entry valid until Read.Response processing done.
 * No Queue locking needed.
 */
static int siw_orqe_start_rx(struct siw_qp *qp)
{
	struct siw_sqe *orqe;
	struct siw_wqe *wqe = NULL;

	/* make sure ORQ indices are current */
	smp_mb();

	orqe = orq_get_current(qp);
	if (READ_ONCE(orqe->flags) & SIW_WQE_VALID) {
		/* RRESP is a TAGGED RDMAP operation */
		wqe = rx_wqe(&qp->rx_tagged);
		wqe->sqe.id = orqe->id;
		wqe->sqe.opcode = orqe->opcode;
		wqe->sqe.sge[0].laddr = orqe->sge[0].laddr;
		wqe->sqe.sge[0].lkey = orqe->sge[0].lkey;
		wqe->sqe.sge[0].length = orqe->sge[0].length;
		wqe->sqe.flags = orqe->flags;
		wqe->sqe.num_sge = 1;
		wqe->bytes = orqe->sge[0].length;
		wqe->processed = 0;
		wqe->mem[0] = NULL;
		/* make sure WQE is completely written before valid */
		smp_wmb();
		wqe->wr_status = SIW_WR_INPROGRESS;

		return 0;
	}
	return -EPROTO;
}

/*
 * siw_proc_rresp:
 *
 * Place incoming RRESP data into memory referenced by RREQ WQE
 * which is at the tip of the ORQ
 *
 * Function supports partially received RRESP's (suspending/resuming
 * current receive processing)
 */
int siw_proc_rresp(struct siw_qp *qp)
{
	struct siw_rx_stream *srx = &qp->rx_stream;
	struct siw_rx_fpdu *frx = &qp->rx_tagged;
	struct siw_wqe *wqe = rx_wqe(frx);
	struct siw_mem **mem, *mem_p;
	struct siw_sge *sge;
	int bytes, rv;

	if (frx->first_ddp_seg) {
		if (unlikely(wqe->wr_status != SIW_WR_IDLE)) {
			pr_warn("siw: [QP %u]: proc RRESP: status %d, op %d\n",
				qp_id(qp), wqe->wr_status, wqe->sqe.opcode);
			rv = -EPROTO;
			goto error_term;
		}
		/*
		 * fetch pending RREQ from orq
		 */
		rv = siw_orqe_start_rx(qp);
		if (rv) {
			pr_warn("siw: [QP %u]: ORQ empty at idx %d\n",
				qp_id(qp), qp->orq_get % qp->attrs.orq_size);
			goto error_term;
		}
		rv = siw_rresp_check_ntoh(srx, frx);
		if (unlikely(rv)) {
			siw_qp_event(qp, IB_EVENT_QP_FATAL);
			return rv;
		}
	} else {
		if (unlikely(wqe->wr_status != SIW_WR_INPROGRESS)) {
			pr_warn("siw: [QP %u]: resume RRESP: status %d\n",
				qp_id(qp), wqe->wr_status);
			rv = -EPROTO;
			goto error_term;
		}
	}
	if (!srx->fpdu_part_rem) /* zero length RRESPONSE */
		return 0;

	sge = wqe->sqe.sge; /* there is only one */
	mem = &wqe->mem[0];

	if (!(*mem)) {
		/*
		 * check target memory which resolves memory on first fragment
		 */
		rv = siw_check_sge(qp->pd, sge, mem, IB_ACCESS_LOCAL_WRITE, 0,
				   wqe->bytes);
		if (unlikely(rv)) {
			siw_dbg_qp(qp, "target mem check: %d\n", rv);
			wqe->wc_status = SIW_WC_LOC_PROT_ERR;

			siw_init_terminate(qp, TERM_ERROR_LAYER_DDP,
					   DDP_ETYPE_TAGGED_BUF,
					   siw_tagged_error(-rv), 0);

			siw_qp_event(qp, IB_EVENT_QP_ACCESS_ERR);

			return -EINVAL;
		}
	}
	mem_p = *mem;

	bytes = min(srx->fpdu_part_rem, srx->skb_new);

	if (mem_p->mem_obj == NULL)
		rv = siw_rx_kva(srx,
			(void *)(uintptr_t)(sge->laddr + wqe->processed),
			bytes);
	else if (!mem_p->is_pbl)
		rv = siw_rx_umem(srx, mem_p->umem, sge->laddr + wqe->processed,
				 bytes);
	else
		rv = siw_rx_pbl(srx, &frx->pbl_idx, mem_p,
				sge->laddr + wqe->processed, bytes);
	if (rv != bytes) {
		wqe->wc_status = SIW_WC_GENERAL_ERR;
		rv = -EINVAL;
		goto error_term;
	}
	srx->fpdu_part_rem -= rv;
	srx->fpdu_part_rcvd += rv;
	wqe->processed += rv;

	if (!srx->fpdu_part_rem) {
		srx->ddp_to += srx->fpdu_part_rcvd;
		return 0;
	}
	return -EAGAIN;

error_term:
	siw_init_terminate(qp, TERM_ERROR_LAYER_DDP, DDP_ETYPE_CATASTROPHIC,
			   DDP_ECODE_CATASTROPHIC, 0);
	return rv;
}

int siw_proc_terminate(struct siw_qp *qp)
{
	struct siw_rx_stream *srx = &qp->rx_stream;
	struct sk_buff *skb = srx->skb;
	struct iwarp_terminate *term = &srx->hdr.terminate;
	union iwarp_hdr term_info;
	u8 *infop = (u8 *)&term_info;
	enum rdma_opcode op;
	u16 to_copy = sizeof(struct iwarp_ctrl);

	pr_warn("siw: got TERMINATE. layer %d, type %d, code %d\n",
		__rdmap_term_layer(term), __rdmap_term_etype(term),
		__rdmap_term_ecode(term));

	if (be32_to_cpu(term->ddp_qn) != RDMAP_UNTAGGED_QN_TERMINATE ||
	    be32_to_cpu(term->ddp_msn) !=
		    qp->rx_stream.ddp_msn[RDMAP_UNTAGGED_QN_TERMINATE] ||
	    be32_to_cpu(term->ddp_mo) != 0) {
		pr_warn("siw: rx bogus TERM [QN x%08x, MSN x%08x, MO x%08x]\n",
			be32_to_cpu(term->ddp_qn), be32_to_cpu(term->ddp_msn),
			be32_to_cpu(term->ddp_mo));
		return -ECONNRESET;
	}
	/*
	 * Receive remaining pieces of TERM if indicated
	 */
	if (!term->flag_m)
		return -ECONNRESET;

	/* Do not take the effort to reassemble a network fragmented
	 * TERM message
	 */
	if (srx->skb_new < sizeof(struct iwarp_ctrl_tagged))
		return -ECONNRESET;

	memset(infop, 0, sizeof(term_info));

	skb_copy_bits(skb, srx->skb_offset, infop, to_copy);

	op = __rdmap_get_opcode(&term_info.ctrl);
	if (op >= RDMAP_TERMINATE)
		goto out;

	infop += to_copy;
	srx->skb_offset += to_copy;
	srx->skb_new -= to_copy;
	srx->skb_copied += to_copy;
	srx->fpdu_part_rcvd += to_copy;
	srx->fpdu_part_rem -= to_copy;

	to_copy = iwarp_pktinfo[op].hdr_len - to_copy;

	/* Again, no network fragmented TERM's */
	if (to_copy + MPA_CRC_SIZE > srx->skb_new)
		return -ECONNRESET;

	skb_copy_bits(skb, srx->skb_offset, infop, to_copy);

	if (term->flag_r) {
		siw_dbg_qp(qp, "TERM reports RDMAP hdr type %u, len %u (%s)\n",
			   op, be16_to_cpu(term_info.ctrl.mpa_len),
			   term->flag_m ? "valid" : "invalid");
	} else if (term->flag_d) {
		siw_dbg_qp(qp, "TERM reports DDP hdr type %u, len %u (%s)\n",
			   op, be16_to_cpu(term_info.ctrl.mpa_len),
			   term->flag_m ? "valid" : "invalid");
	}
out:
	srx->skb_new -= to_copy;
	srx->skb_offset += to_copy;
	srx->skb_copied += to_copy;
	srx->fpdu_part_rcvd += to_copy;
	srx->fpdu_part_rem -= to_copy;

	return -ECONNRESET;
}

static int siw_get_trailer(struct siw_qp *qp, struct siw_rx_stream *srx)
{
	struct sk_buff *skb = srx->skb;
	u8 *tbuf = (u8 *)&srx->trailer.crc - srx->pad;
	__wsum crc_in, crc_own = 0;

	siw_dbg_qp(qp, "expected %d, available %d, pad %u\n",
		   srx->fpdu_part_rem, srx->skb_new, srx->pad);

	if (srx->skb_new < srx->fpdu_part_rem)
		return -EAGAIN;

	skb_copy_bits(skb, srx->skb_offset, tbuf, srx->fpdu_part_rem);

	if (srx->mpa_crc_hd && srx->pad)
		crypto_shash_update(srx->mpa_crc_hd, tbuf, srx->pad);

	srx->skb_new -= srx->fpdu_part_rem;
	srx->skb_offset += srx->fpdu_part_rem;
	srx->skb_copied += srx->fpdu_part_rem;

	if (!srx->mpa_crc_hd)
		return 0;

	/*
	 * CRC32 is computed, transmitted and received directly in NBO,
	 * so there's never a reason to convert byte order.
	 */
	crypto_shash_final(srx->mpa_crc_hd, (u8 *)&crc_own);
	crc_in = (__force __wsum)srx->trailer.crc;

	if (unlikely(crc_in != crc_own)) {
		pr_warn("siw: crc error. in: %08x, own %08x, op %u\n",
			crc_in, crc_own, qp->rx_stream.rdmap_op);

		siw_init_terminate(qp, TERM_ERROR_LAYER_LLP,
				   LLP_ETYPE_MPA,
				   LLP_ECODE_RECEIVED_CRC, 0);
		return -EINVAL;
	}
	return 0;
}

#define MIN_DDP_HDR sizeof(struct iwarp_ctrl_tagged)

static int siw_get_hdr(struct siw_rx_stream *srx)
{
	struct sk_buff *skb = srx->skb;
	struct siw_qp *qp = rx_qp(srx);
	struct iwarp_ctrl *c_hdr = &srx->hdr.ctrl;
	struct siw_rx_fpdu *frx;
	u8 opcode;
	int bytes;

	if (srx->fpdu_part_rcvd < MIN_DDP_HDR) {
		/*
		 * copy a mimimum sized (tagged) DDP frame control part
		 */
		bytes = min_t(int, srx->skb_new,
			      MIN_DDP_HDR - srx->fpdu_part_rcvd);

		skb_copy_bits(skb, srx->skb_offset,
			      (char *)c_hdr + srx->fpdu_part_rcvd, bytes);

		srx->fpdu_part_rcvd += bytes;

		srx->skb_new -= bytes;
		srx->skb_offset += bytes;
		srx->skb_copied += bytes;

		if (srx->fpdu_part_rcvd < MIN_DDP_HDR)
			return -EAGAIN;

		if (unlikely(__ddp_get_version(c_hdr) != DDP_VERSION)) {
			enum ddp_etype etype;
			enum ddp_ecode ecode;

			pr_warn("siw: received ddp version unsupported %d\n",
				__ddp_get_version(c_hdr));

			if (c_hdr->ddp_rdmap_ctrl & DDP_FLAG_TAGGED) {
				etype = DDP_ETYPE_TAGGED_BUF;
				ecode = DDP_ECODE_T_VERSION;
			} else {
				etype = DDP_ETYPE_UNTAGGED_BUF;
				ecode = DDP_ECODE_UT_VERSION;
			}
			siw_init_terminate(rx_qp(srx), TERM_ERROR_LAYER_DDP,
					   etype, ecode, 0);
			return -EINVAL;
		}
		if (unlikely(__rdmap_get_version(c_hdr) != RDMAP_VERSION)) {
			pr_warn("siw: received rdmap version unsupported %d\n",
				__rdmap_get_version(c_hdr));

			siw_init_terminate(rx_qp(srx), TERM_ERROR_LAYER_RDMAP,
					   RDMAP_ETYPE_REMOTE_OPERATION,
					   RDMAP_ECODE_VERSION, 0);
			return -EINVAL;
		}
		opcode = __rdmap_get_opcode(c_hdr);

		if (opcode > RDMAP_TERMINATE) {
			pr_warn("siw: received unknown packet type %u\n",
				opcode);

			siw_init_terminate(rx_qp(srx), TERM_ERROR_LAYER_RDMAP,
					   RDMAP_ETYPE_REMOTE_OPERATION,
					   RDMAP_ECODE_OPCODE, 0);
			return -EINVAL;
		}
		siw_dbg_qp(rx_qp(srx), "new header, opcode %u\n", opcode);
	} else {
		opcode = __rdmap_get_opcode(c_hdr);
	}
	set_rx_fpdu_context(qp, opcode);
	frx = qp->rx_fpdu;

	/*
	 * Figure out len of current hdr: variable length of
	 * iwarp hdr may force us to copy hdr information in
	 * two steps. Only tagged DDP messages are already
	 * completely received.
	 */
	if (iwarp_pktinfo[opcode].hdr_len > sizeof(struct iwarp_ctrl_tagged)) {
		bytes = iwarp_pktinfo[opcode].hdr_len - MIN_DDP_HDR;

		if (srx->skb_new < bytes)
			return -EAGAIN;

		skb_copy_bits(skb, srx->skb_offset,
			      (char *)c_hdr + srx->fpdu_part_rcvd, bytes);

		srx->fpdu_part_rcvd += bytes;

		srx->skb_new -= bytes;
		srx->skb_offset += bytes;
		srx->skb_copied += bytes;
	}

	/*
	 * DDP/RDMAP header receive completed. Check if the current
	 * DDP segment starts a new RDMAP message or continues a previously
	 * started RDMAP message.
	 *
	 * Alternating reception of DDP segments (or FPDUs) from incomplete
	 * tagged and untagged RDMAP messages is supported, as long as
	 * the current tagged or untagged message gets eventually completed
	 * w/o intersection from another message of the same type
	 * (tagged/untagged). E.g., a WRITE can get intersected by a SEND,
	 * but not by a READ RESPONSE etc.
	 */
	if (srx->mpa_crc_hd) {
		/*
		 * Restart CRC computation
		 */
		crypto_shash_init(srx->mpa_crc_hd);
		crypto_shash_update(srx->mpa_crc_hd, (u8 *)c_hdr,
				    srx->fpdu_part_rcvd);
	}
	if (frx->more_ddp_segs) {
		frx->first_ddp_seg = 0;
		if (frx->prev_rdmap_op != opcode) {
			pr_warn("siw: packet intersection: %u : %u\n",
				frx->prev_rdmap_op, opcode);
			/*
			 * The last inbound RDMA operation of same type
			 * (tagged or untagged) is left unfinished.
			 * To complete it in error, make it the current
			 * operation again, even with the header already
			 * overwritten. For error handling, only the opcode
			 * and current rx context are relevant.
			 */
			set_rx_fpdu_context(qp, frx->prev_rdmap_op);
			__rdmap_set_opcode(c_hdr, frx->prev_rdmap_op);
			return -EPROTO;
		}
	} else {
		frx->prev_rdmap_op = opcode;
		frx->first_ddp_seg = 1;
	}
	frx->more_ddp_segs = c_hdr->ddp_rdmap_ctrl & DDP_FLAG_LAST ? 0 : 1;

	return 0;
}

static int siw_check_tx_fence(struct siw_qp *qp)
{
	struct siw_wqe *tx_waiting = tx_wqe(qp);
	struct siw_sqe *rreq;
	int resume_tx = 0, rv = 0;
	unsigned long flags;

	spin_lock_irqsave(&qp->orq_lock, flags);

	rreq = orq_get_current(qp);

	/* free current orq entry */
	WRITE_ONCE(rreq->flags, 0);

	if (qp->tx_ctx.orq_fence) {
		if (unlikely(tx_waiting->wr_status != SIW_WR_QUEUED)) {
			pr_warn("siw: [QP %u]: fence resume: bad status %d\n",
				qp_id(qp), tx_waiting->wr_status);
			rv = -EPROTO;
			goto out;
		}
		/* resume SQ processing */
		if (tx_waiting->sqe.opcode == SIW_OP_READ ||
		    tx_waiting->sqe.opcode == SIW_OP_READ_LOCAL_INV) {
			rreq = orq_get_tail(qp);
			if (unlikely(!rreq)) {
				pr_warn("siw: [QP %u]: no ORQE\n", qp_id(qp));
				rv = -EPROTO;
				goto out;
			}
			siw_read_to_orq(rreq, &tx_waiting->sqe);

			qp->orq_put++;
			qp->tx_ctx.orq_fence = 0;
			resume_tx = 1;

		} else if (siw_orq_empty(qp)) {
			qp->tx_ctx.orq_fence = 0;
			resume_tx = 1;
		} else {
			pr_warn("siw: [QP %u]: fence resume: orq idx: %d:%d\n",
				qp_id(qp), qp->orq_get, qp->orq_put);
			rv = -EPROTO;
		}
	}
	qp->orq_get++;
out:
	spin_unlock_irqrestore(&qp->orq_lock, flags);

	if (resume_tx)
		rv = siw_sq_start(qp);

	return rv;
}

/*
 * siw_rdmap_complete()
 *
 * Complete processing of an RDMA message after receiving all
 * DDP segmens or ABort processing after encountering error case.
 *
 *   o SENDs + RRESPs will need for completion,
 *   o RREQs need for  READ RESPONSE initialization
 *   o WRITEs need memory dereferencing
 *
 * TODO: Failed WRITEs need local error to be surfaced.
 */
static int siw_rdmap_complete(struct siw_qp *qp, int error)
{
	struct siw_rx_stream *srx = &qp->rx_stream;
	struct siw_wqe *wqe = rx_wqe(qp->rx_fpdu);
	enum siw_wc_status wc_status = wqe->wc_status;
	u8 opcode = __rdmap_get_opcode(&srx->hdr.ctrl);
	int rv = 0;

	switch (opcode) {
	case RDMAP_SEND_SE:
	case RDMAP_SEND_SE_INVAL:
		wqe->rqe.flags |= SIW_WQE_SOLICITED;
		/* Fall through */

	case RDMAP_SEND:
	case RDMAP_SEND_INVAL:
		if (wqe->wr_status == SIW_WR_IDLE)
			break;

		srx->ddp_msn[RDMAP_UNTAGGED_QN_SEND]++;

		if (error != 0 && wc_status == SIW_WC_SUCCESS)
			wc_status = SIW_WC_GENERAL_ERR;
		/*
		 * Handle STag invalidation request
		 */
		if (wc_status == SIW_WC_SUCCESS &&
		    (opcode == RDMAP_SEND_INVAL ||
		     opcode == RDMAP_SEND_SE_INVAL)) {
			rv = siw_invalidate_stag(qp->pd, srx->inval_stag);
			if (rv) {
				siw_init_terminate(
					qp, TERM_ERROR_LAYER_RDMAP,
					rv == -EACCES ?
						RDMAP_ETYPE_REMOTE_PROTECTION :
						RDMAP_ETYPE_REMOTE_OPERATION,
					RDMAP_ECODE_CANNOT_INVALIDATE, 0);

				wc_status = SIW_WC_REM_INV_REQ_ERR;
			}
			rv = siw_rqe_complete(qp, &wqe->rqe, wqe->processed,
					      rv ? 0 : srx->inval_stag,
					      wc_status);
		} else {
			rv = siw_rqe_complete(qp, &wqe->rqe, wqe->processed,
					      0, wc_status);
		}
		siw_wqe_put_mem(wqe, SIW_OP_RECEIVE);
		break;

	case RDMAP_RDMA_READ_RESP:
		if (wqe->wr_status == SIW_WR_IDLE)
			break;

		if (error != 0) {
			if ((srx->state == SIW_GET_HDR &&
			     qp->rx_fpdu->first_ddp_seg) || error == -ENODATA)
				/* possible RREQ in ORQ left untouched */
				break;

			if (wc_status == SIW_WC_SUCCESS)
				wc_status = SIW_WC_GENERAL_ERR;
		} else if (qp->kernel_verbs &&
			   rx_type(wqe) == SIW_OP_READ_LOCAL_INV) {
			/*
			 * Handle any STag invalidation request
			 */
			rv = siw_invalidate_stag(qp->pd, wqe->sqe.sge[0].lkey);
			if (rv) {
				siw_init_terminate(qp, TERM_ERROR_LAYER_RDMAP,
						   RDMAP_ETYPE_CATASTROPHIC,
						   RDMAP_ECODE_UNSPECIFIED, 0);

				if (wc_status == SIW_WC_SUCCESS) {
					wc_status = SIW_WC_GENERAL_ERR;
					error = rv;
				}
			}
		}
		/*
		 * All errors turn the wqe into signalled.
		 */
		if ((wqe->sqe.flags & SIW_WQE_SIGNALLED) || error != 0)
			rv = siw_sqe_complete(qp, &wqe->sqe, wqe->processed,
					      wc_status);
		siw_wqe_put_mem(wqe, SIW_OP_READ);

		if (!error)
			rv = siw_check_tx_fence(qp);
		else
			/* Disable current ORQ eleement */
			WRITE_ONCE(orq_get_current(qp)->flags, 0);
		break;

	case RDMAP_RDMA_READ_REQ:
		if (!error) {
			rv = siw_init_rresp(qp, srx);
			srx->ddp_msn[RDMAP_UNTAGGED_QN_RDMA_READ]++;
		}
		break;

	case RDMAP_RDMA_WRITE:
		if (wqe->wr_status == SIW_WR_IDLE)
			break;

		/*
		 * Free References from memory object if
		 * attached to receive context (inbound WRITE).
		 * While a zero-length WRITE is allowed,
		 * no memory reference got created.
		 */
		if (rx_mem(&qp->rx_tagged)) {
			siw_mem_put(rx_mem(&qp->rx_tagged));
			rx_mem(&qp->rx_tagged) = NULL;
		}
		break;

	default:
		break;
	}
	wqe->wr_status = SIW_WR_IDLE;

	return rv;
}

/*
 * siw_tcp_rx_data()
 *
 * Main routine to consume inbound TCP payload
 *
 * @rd_desc:	read descriptor
 * @skb:	socket buffer
 * @off:	offset in skb
 * @len:	skb->len - offset : payload in skb
 */
int siw_tcp_rx_data(read_descriptor_t *rd_desc, struct sk_buff *skb,
		    unsigned int off, size_t len)
{
	struct siw_qp *qp = rd_desc->arg.data;
	struct siw_rx_stream *srx = &qp->rx_stream;
	int rv;

	srx->skb = skb;
	srx->skb_new = skb->len - off;
	srx->skb_offset = off;
	srx->skb_copied = 0;

	siw_dbg_qp(qp, "new data, len %d\n", srx->skb_new);

	while (srx->skb_new) {
		int run_completion = 1;

		if (unlikely(srx->rx_suspend)) {
			/* Do not process any more data */
			srx->skb_copied += srx->skb_new;
			break;
		}
		switch (srx->state) {
		case SIW_GET_HDR:
			rv = siw_get_hdr(srx);
			if (!rv) {
				srx->fpdu_part_rem =
					be16_to_cpu(srx->hdr.ctrl.mpa_len) -
					srx->fpdu_part_rcvd + MPA_HDR_SIZE;

				if (srx->fpdu_part_rem)
					srx->pad = -srx->fpdu_part_rem & 0x3;
				else
					srx->pad = 0;

				srx->state = SIW_GET_DATA_START;
				srx->fpdu_part_rcvd = 0;
			}
			break;

		case SIW_GET_DATA_MORE:
			/*
			 * Another data fragment of the same DDP segment.
			 * Setting first_ddp_seg = 0 avoids repeating
			 * initializations that shall occur only once per
			 * DDP segment.
			 */
			qp->rx_fpdu->first_ddp_seg = 0;
			/* Fall through */

		case SIW_GET_DATA_START:
			/*
			 * Headers will be checked by the opcode-specific
			 * data receive function below.
			 */
			rv = iwarp_pktinfo[qp->rx_stream.rdmap_op].rx_data(qp);
			if (!rv) {
				int mpa_len =
					be16_to_cpu(srx->hdr.ctrl.mpa_len)
					+ MPA_HDR_SIZE;

				srx->fpdu_part_rem = (-mpa_len & 0x3)
						      + MPA_CRC_SIZE;
				srx->fpdu_part_rcvd = 0;
				srx->state = SIW_GET_TRAILER;
			} else {
				if (unlikely(rv == -ECONNRESET))
					run_completion = 0;
				else
					srx->state = SIW_GET_DATA_MORE;
			}
			break;

		case SIW_GET_TRAILER:
			/*
			 * read CRC + any padding
			 */
			rv = siw_get_trailer(qp, srx);
			if (likely(!rv)) {
				/*
				 * FPDU completed.
				 * complete RDMAP message if last fragment
				 */
				srx->state = SIW_GET_HDR;
				srx->fpdu_part_rcvd = 0;

				if (!(srx->hdr.ctrl.ddp_rdmap_ctrl &
				      DDP_FLAG_LAST))
					/* more frags */
					break;

				rv = siw_rdmap_complete(qp, 0);
				run_completion = 0;
			}
			break;

		default:
			pr_warn("QP[%u]: RX out of state\n", qp_id(qp));
			rv = -EPROTO;
			run_completion = 0;
		}
		if (unlikely(rv != 0 && rv != -EAGAIN)) {
			if ((srx->state > SIW_GET_HDR ||
			     qp->rx_fpdu->more_ddp_segs) && run_completion)
				siw_rdmap_complete(qp, rv);

			siw_dbg_qp(qp, "rx error %d, rx state %d\n", rv,
				   srx->state);

			siw_qp_cm_drop(qp, 1);

			break;
		}
		if (rv) {
			siw_dbg_qp(qp, "fpdu fragment, state %d, missing %d\n",
				   srx->state, srx->fpdu_part_rem);
			break;
		}
	}
	return srx->skb_copied;
}