aboutsummaryrefslogtreecommitdiff
path: root/drivers/infiniband/hw/hfi1/ruc.c
blob: 5f56f3c1b4c4e42d9de35f014fb554252fa3f758 (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
/*
 * Copyright(c) 2015 - 2018 Intel Corporation.
 *
 * This file is provided under a dual BSD/GPLv2 license.  When using or
 * redistributing this file, you may do so under either license.
 *
 * GPL LICENSE SUMMARY
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of version 2 of the GNU General Public License as
 * published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * BSD LICENSE
 *
 * 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 Intel Corporation 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.
 *
 */

#include <linux/spinlock.h>

#include "hfi.h"
#include "mad.h"
#include "qp.h"
#include "verbs_txreq.h"
#include "trace.h"

static int gid_ok(union ib_gid *gid, __be64 gid_prefix, __be64 id)
{
	return (gid->global.interface_id == id &&
		(gid->global.subnet_prefix == gid_prefix ||
		 gid->global.subnet_prefix == IB_DEFAULT_GID_PREFIX));
}

/*
 *
 * This should be called with the QP r_lock held.
 *
 * The s_lock will be acquired around the hfi1_migrate_qp() call.
 */
int hfi1_ruc_check_hdr(struct hfi1_ibport *ibp, struct hfi1_packet *packet)
{
	__be64 guid;
	unsigned long flags;
	struct rvt_qp *qp = packet->qp;
	u8 sc5 = ibp->sl_to_sc[rdma_ah_get_sl(&qp->remote_ah_attr)];
	u32 dlid = packet->dlid;
	u32 slid = packet->slid;
	u32 sl = packet->sl;
	bool migrated = packet->migrated;
	u16 pkey = packet->pkey;

	if (qp->s_mig_state == IB_MIG_ARMED && migrated) {
		if (!packet->grh) {
			if ((rdma_ah_get_ah_flags(&qp->alt_ah_attr) &
			     IB_AH_GRH) &&
			    (packet->etype != RHF_RCV_TYPE_BYPASS))
				return 1;
		} else {
			const struct ib_global_route *grh;

			if (!(rdma_ah_get_ah_flags(&qp->alt_ah_attr) &
			      IB_AH_GRH))
				return 1;
			grh = rdma_ah_read_grh(&qp->alt_ah_attr);
			guid = get_sguid(ibp, grh->sgid_index);
			if (!gid_ok(&packet->grh->dgid, ibp->rvp.gid_prefix,
				    guid))
				return 1;
			if (!gid_ok(
				&packet->grh->sgid,
				grh->dgid.global.subnet_prefix,
				grh->dgid.global.interface_id))
				return 1;
		}
		if (unlikely(rcv_pkey_check(ppd_from_ibp(ibp), pkey,
					    sc5, slid))) {
			hfi1_bad_pkey(ibp, pkey, sl, 0, qp->ibqp.qp_num,
				      slid, dlid);
			return 1;
		}
		/* Validate the SLID. See Ch. 9.6.1.5 and 17.2.8 */
		if (slid != rdma_ah_get_dlid(&qp->alt_ah_attr) ||
		    ppd_from_ibp(ibp)->port !=
			rdma_ah_get_port_num(&qp->alt_ah_attr))
			return 1;
		spin_lock_irqsave(&qp->s_lock, flags);
		hfi1_migrate_qp(qp);
		spin_unlock_irqrestore(&qp->s_lock, flags);
	} else {
		if (!packet->grh) {
			if ((rdma_ah_get_ah_flags(&qp->remote_ah_attr) &
			     IB_AH_GRH) &&
			    (packet->etype != RHF_RCV_TYPE_BYPASS))
				return 1;
		} else {
			const struct ib_global_route *grh;

			if (!(rdma_ah_get_ah_flags(&qp->remote_ah_attr) &
						   IB_AH_GRH))
				return 1;
			grh = rdma_ah_read_grh(&qp->remote_ah_attr);
			guid = get_sguid(ibp, grh->sgid_index);
			if (!gid_ok(&packet->grh->dgid, ibp->rvp.gid_prefix,
				    guid))
				return 1;
			if (!gid_ok(
			     &packet->grh->sgid,
			     grh->dgid.global.subnet_prefix,
			     grh->dgid.global.interface_id))
				return 1;
		}
		if (unlikely(rcv_pkey_check(ppd_from_ibp(ibp), pkey,
					    sc5, slid))) {
			hfi1_bad_pkey(ibp, pkey, sl, 0, qp->ibqp.qp_num,
				      slid, dlid);
			return 1;
		}
		/* Validate the SLID. See Ch. 9.6.1.5 */
		if ((slid != rdma_ah_get_dlid(&qp->remote_ah_attr)) ||
		    ppd_from_ibp(ibp)->port != qp->port_num)
			return 1;
		if (qp->s_mig_state == IB_MIG_REARM && !migrated)
			qp->s_mig_state = IB_MIG_ARMED;
	}

	return 0;
}

/**
 * ruc_loopback - handle UC and RC loopback requests
 * @sqp: the sending QP
 *
 * This is called from hfi1_do_send() to
 * forward a WQE addressed to the same HFI.
 * Note that although we are single threaded due to the send engine, we still
 * have to protect against post_send().  We don't have to worry about
 * receive interrupts since this is a connected protocol and all packets
 * will pass through here.
 */
static void ruc_loopback(struct rvt_qp *sqp)
{
	struct hfi1_ibport *ibp = to_iport(sqp->ibqp.device, sqp->port_num);
	struct rvt_qp *qp;
	struct rvt_swqe *wqe;
	struct rvt_sge *sge;
	unsigned long flags;
	struct ib_wc wc;
	u64 sdata;
	atomic64_t *maddr;
	enum ib_wc_status send_status;
	bool release;
	int ret;
	bool copy_last = false;
	int local_ops = 0;

	rcu_read_lock();

	/*
	 * Note that we check the responder QP state after
	 * checking the requester's state.
	 */
	qp = rvt_lookup_qpn(ib_to_rvt(sqp->ibqp.device), &ibp->rvp,
			    sqp->remote_qpn);

	spin_lock_irqsave(&sqp->s_lock, flags);

	/* Return if we are already busy processing a work request. */
	if ((sqp->s_flags & (RVT_S_BUSY | HFI1_S_ANY_WAIT)) ||
	    !(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_OR_FLUSH_SEND))
		goto unlock;

	sqp->s_flags |= RVT_S_BUSY;

again:
	if (sqp->s_last == READ_ONCE(sqp->s_head))
		goto clr_busy;
	wqe = rvt_get_swqe_ptr(sqp, sqp->s_last);

	/* Return if it is not OK to start a new work request. */
	if (!(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_NEXT_SEND_OK)) {
		if (!(ib_rvt_state_ops[sqp->state] & RVT_FLUSH_SEND))
			goto clr_busy;
		/* We are in the error state, flush the work request. */
		send_status = IB_WC_WR_FLUSH_ERR;
		goto flush_send;
	}

	/*
	 * We can rely on the entry not changing without the s_lock
	 * being held until we update s_last.
	 * We increment s_cur to indicate s_last is in progress.
	 */
	if (sqp->s_last == sqp->s_cur) {
		if (++sqp->s_cur >= sqp->s_size)
			sqp->s_cur = 0;
	}
	spin_unlock_irqrestore(&sqp->s_lock, flags);

	if (!qp || !(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) ||
	    qp->ibqp.qp_type != sqp->ibqp.qp_type) {
		ibp->rvp.n_pkt_drops++;
		/*
		 * For RC, the requester would timeout and retry so
		 * shortcut the timeouts and just signal too many retries.
		 */
		if (sqp->ibqp.qp_type == IB_QPT_RC)
			send_status = IB_WC_RETRY_EXC_ERR;
		else
			send_status = IB_WC_SUCCESS;
		goto serr;
	}

	memset(&wc, 0, sizeof(wc));
	send_status = IB_WC_SUCCESS;

	release = true;
	sqp->s_sge.sge = wqe->sg_list[0];
	sqp->s_sge.sg_list = wqe->sg_list + 1;
	sqp->s_sge.num_sge = wqe->wr.num_sge;
	sqp->s_len = wqe->length;
	switch (wqe->wr.opcode) {
	case IB_WR_REG_MR:
		goto send_comp;

	case IB_WR_LOCAL_INV:
		if (!(wqe->wr.send_flags & RVT_SEND_COMPLETION_ONLY)) {
			if (rvt_invalidate_rkey(sqp,
						wqe->wr.ex.invalidate_rkey))
				send_status = IB_WC_LOC_PROT_ERR;
			local_ops = 1;
		}
		goto send_comp;

	case IB_WR_SEND_WITH_INV:
		if (!rvt_invalidate_rkey(qp, wqe->wr.ex.invalidate_rkey)) {
			wc.wc_flags = IB_WC_WITH_INVALIDATE;
			wc.ex.invalidate_rkey = wqe->wr.ex.invalidate_rkey;
		}
		goto send;

	case IB_WR_SEND_WITH_IMM:
		wc.wc_flags = IB_WC_WITH_IMM;
		wc.ex.imm_data = wqe->wr.ex.imm_data;
		/* FALLTHROUGH */
	case IB_WR_SEND:
send:
		ret = rvt_get_rwqe(qp, false);
		if (ret < 0)
			goto op_err;
		if (!ret)
			goto rnr_nak;
		break;

	case IB_WR_RDMA_WRITE_WITH_IMM:
		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
			goto inv_err;
		wc.wc_flags = IB_WC_WITH_IMM;
		wc.ex.imm_data = wqe->wr.ex.imm_data;
		ret = rvt_get_rwqe(qp, true);
		if (ret < 0)
			goto op_err;
		if (!ret)
			goto rnr_nak;
		/* skip copy_last set and qp_access_flags recheck */
		goto do_write;
	case IB_WR_RDMA_WRITE:
		copy_last = rvt_is_user_qp(qp);
		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
			goto inv_err;
do_write:
		if (wqe->length == 0)
			break;
		if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, wqe->length,
					  wqe->rdma_wr.remote_addr,
					  wqe->rdma_wr.rkey,
					  IB_ACCESS_REMOTE_WRITE)))
			goto acc_err;
		qp->r_sge.sg_list = NULL;
		qp->r_sge.num_sge = 1;
		qp->r_sge.total_len = wqe->length;
		break;

	case IB_WR_RDMA_READ:
		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_READ)))
			goto inv_err;
		if (unlikely(!rvt_rkey_ok(qp, &sqp->s_sge.sge, wqe->length,
					  wqe->rdma_wr.remote_addr,
					  wqe->rdma_wr.rkey,
					  IB_ACCESS_REMOTE_READ)))
			goto acc_err;
		release = false;
		sqp->s_sge.sg_list = NULL;
		sqp->s_sge.num_sge = 1;
		qp->r_sge.sge = wqe->sg_list[0];
		qp->r_sge.sg_list = wqe->sg_list + 1;
		qp->r_sge.num_sge = wqe->wr.num_sge;
		qp->r_sge.total_len = wqe->length;
		break;

	case IB_WR_ATOMIC_CMP_AND_SWP:
	case IB_WR_ATOMIC_FETCH_AND_ADD:
		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_ATOMIC)))
			goto inv_err;
		if (unlikely(!rvt_rkey_ok(qp, &qp->r_sge.sge, sizeof(u64),
					  wqe->atomic_wr.remote_addr,
					  wqe->atomic_wr.rkey,
					  IB_ACCESS_REMOTE_ATOMIC)))
			goto acc_err;
		/* Perform atomic OP and save result. */
		maddr = (atomic64_t *)qp->r_sge.sge.vaddr;
		sdata = wqe->atomic_wr.compare_add;
		*(u64 *)sqp->s_sge.sge.vaddr =
			(wqe->wr.opcode == IB_WR_ATOMIC_FETCH_AND_ADD) ?
			(u64)atomic64_add_return(sdata, maddr) - sdata :
			(u64)cmpxchg((u64 *)qp->r_sge.sge.vaddr,
				      sdata, wqe->atomic_wr.swap);
		rvt_put_mr(qp->r_sge.sge.mr);
		qp->r_sge.num_sge = 0;
		goto send_comp;

	default:
		send_status = IB_WC_LOC_QP_OP_ERR;
		goto serr;
	}

	sge = &sqp->s_sge.sge;
	while (sqp->s_len) {
		u32 len = sqp->s_len;

		if (len > sge->length)
			len = sge->length;
		if (len > sge->sge_length)
			len = sge->sge_length;
		WARN_ON_ONCE(len == 0);
		hfi1_copy_sge(&qp->r_sge, sge->vaddr, len, release, copy_last);
		sge->vaddr += len;
		sge->length -= len;
		sge->sge_length -= len;
		if (sge->sge_length == 0) {
			if (!release)
				rvt_put_mr(sge->mr);
			if (--sqp->s_sge.num_sge)
				*sge = *sqp->s_sge.sg_list++;
		} else if (sge->length == 0 && sge->mr->lkey) {
			if (++sge->n >= RVT_SEGSZ) {
				if (++sge->m >= sge->mr->mapsz)
					break;
				sge->n = 0;
			}
			sge->vaddr =
				sge->mr->map[sge->m]->segs[sge->n].vaddr;
			sge->length =
				sge->mr->map[sge->m]->segs[sge->n].length;
		}
		sqp->s_len -= len;
	}
	if (release)
		rvt_put_ss(&qp->r_sge);

	if (!test_and_clear_bit(RVT_R_WRID_VALID, &qp->r_aflags))
		goto send_comp;

	if (wqe->wr.opcode == IB_WR_RDMA_WRITE_WITH_IMM)
		wc.opcode = IB_WC_RECV_RDMA_WITH_IMM;
	else
		wc.opcode = IB_WC_RECV;
	wc.wr_id = qp->r_wr_id;
	wc.status = IB_WC_SUCCESS;
	wc.byte_len = wqe->length;
	wc.qp = &qp->ibqp;
	wc.src_qp = qp->remote_qpn;
	wc.slid = rdma_ah_get_dlid(&qp->remote_ah_attr) & U16_MAX;
	wc.sl = rdma_ah_get_sl(&qp->remote_ah_attr);
	wc.port_num = 1;
	/* Signal completion event if the solicited bit is set. */
	rvt_cq_enter(ibcq_to_rvtcq(qp->ibqp.recv_cq), &wc,
		     wqe->wr.send_flags & IB_SEND_SOLICITED);

send_comp:
	spin_lock_irqsave(&sqp->s_lock, flags);
	ibp->rvp.n_loop_pkts++;
flush_send:
	sqp->s_rnr_retry = sqp->s_rnr_retry_cnt;
	hfi1_send_complete(sqp, wqe, send_status);
	if (local_ops) {
		atomic_dec(&sqp->local_ops_pending);
		local_ops = 0;
	}
	goto again;

rnr_nak:
	/* Handle RNR NAK */
	if (qp->ibqp.qp_type == IB_QPT_UC)
		goto send_comp;
	ibp->rvp.n_rnr_naks++;
	/*
	 * Note: we don't need the s_lock held since the BUSY flag
	 * makes this single threaded.
	 */
	if (sqp->s_rnr_retry == 0) {
		send_status = IB_WC_RNR_RETRY_EXC_ERR;
		goto serr;
	}
	if (sqp->s_rnr_retry_cnt < 7)
		sqp->s_rnr_retry--;
	spin_lock_irqsave(&sqp->s_lock, flags);
	if (!(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_RECV_OK))
		goto clr_busy;
	rvt_add_rnr_timer(sqp, qp->r_min_rnr_timer <<
				IB_AETH_CREDIT_SHIFT);
	goto clr_busy;

op_err:
	send_status = IB_WC_REM_OP_ERR;
	wc.status = IB_WC_LOC_QP_OP_ERR;
	goto err;

inv_err:
	send_status = IB_WC_REM_INV_REQ_ERR;
	wc.status = IB_WC_LOC_QP_OP_ERR;
	goto err;

acc_err:
	send_status = IB_WC_REM_ACCESS_ERR;
	wc.status = IB_WC_LOC_PROT_ERR;
err:
	/* responder goes to error state */
	rvt_rc_error(qp, wc.status);

serr:
	spin_lock_irqsave(&sqp->s_lock, flags);
	hfi1_send_complete(sqp, wqe, send_status);
	if (sqp->ibqp.qp_type == IB_QPT_RC) {
		int lastwqe = rvt_error_qp(sqp, IB_WC_WR_FLUSH_ERR);

		sqp->s_flags &= ~RVT_S_BUSY;
		spin_unlock_irqrestore(&sqp->s_lock, flags);
		if (lastwqe) {
			struct ib_event ev;

			ev.device = sqp->ibqp.device;
			ev.element.qp = &sqp->ibqp;
			ev.event = IB_EVENT_QP_LAST_WQE_REACHED;
			sqp->ibqp.event_handler(&ev, sqp->ibqp.qp_context);
		}
		goto done;
	}
clr_busy:
	sqp->s_flags &= ~RVT_S_BUSY;
unlock:
	spin_unlock_irqrestore(&sqp->s_lock, flags);
done:
	rcu_read_unlock();
}

/**
 * hfi1_make_grh - construct a GRH header
 * @ibp: a pointer to the IB port
 * @hdr: a pointer to the GRH header being constructed
 * @grh: the global route address to send to
 * @hwords: size of header after grh being sent in dwords
 * @nwords: the number of 32 bit words of data being sent
 *
 * Return the size of the header in 32 bit words.
 */
u32 hfi1_make_grh(struct hfi1_ibport *ibp, struct ib_grh *hdr,
		  const struct ib_global_route *grh, u32 hwords, u32 nwords)
{
	hdr->version_tclass_flow =
		cpu_to_be32((IB_GRH_VERSION << IB_GRH_VERSION_SHIFT) |
			    (grh->traffic_class << IB_GRH_TCLASS_SHIFT) |
			    (grh->flow_label << IB_GRH_FLOW_SHIFT));
	hdr->paylen = cpu_to_be16((hwords + nwords) << 2);
	/* next_hdr is defined by C8-7 in ch. 8.4.1 */
	hdr->next_hdr = IB_GRH_NEXT_HDR;
	hdr->hop_limit = grh->hop_limit;
	/* The SGID is 32-bit aligned. */
	hdr->sgid.global.subnet_prefix = ibp->rvp.gid_prefix;
	hdr->sgid.global.interface_id =
		grh->sgid_index < HFI1_GUIDS_PER_PORT ?
		get_sguid(ibp, grh->sgid_index) :
		get_sguid(ibp, HFI1_PORT_GUID_INDEX);
	hdr->dgid = grh->dgid;

	/* GRH header size in 32-bit words. */
	return sizeof(struct ib_grh) / sizeof(u32);
}

#define BTH2_OFFSET (offsetof(struct hfi1_sdma_header, \
			      hdr.ibh.u.oth.bth[2]) / 4)

/**
 * build_ahg - create ahg in s_ahg
 * @qp: a pointer to QP
 * @npsn: the next PSN for the request/response
 *
 * This routine handles the AHG by allocating an ahg entry and causing the
 * copy of the first middle.
 *
 * Subsequent middles use the copied entry, editing the
 * PSN with 1 or 2 edits.
 */
static inline void build_ahg(struct rvt_qp *qp, u32 npsn)
{
	struct hfi1_qp_priv *priv = qp->priv;

	if (unlikely(qp->s_flags & HFI1_S_AHG_CLEAR))
		clear_ahg(qp);
	if (!(qp->s_flags & HFI1_S_AHG_VALID)) {
		/* first middle that needs copy  */
		if (qp->s_ahgidx < 0)
			qp->s_ahgidx = sdma_ahg_alloc(priv->s_sde);
		if (qp->s_ahgidx >= 0) {
			qp->s_ahgpsn = npsn;
			priv->s_ahg->tx_flags |= SDMA_TXREQ_F_AHG_COPY;
			/* save to protect a change in another thread */
			priv->s_ahg->ahgidx = qp->s_ahgidx;
			qp->s_flags |= HFI1_S_AHG_VALID;
		}
	} else {
		/* subsequent middle after valid */
		if (qp->s_ahgidx >= 0) {
			priv->s_ahg->tx_flags |= SDMA_TXREQ_F_USE_AHG;
			priv->s_ahg->ahgidx = qp->s_ahgidx;
			priv->s_ahg->ahgcount++;
			priv->s_ahg->ahgdesc[0] =
				sdma_build_ahg_descriptor(
					(__force u16)cpu_to_be16((u16)npsn),
					BTH2_OFFSET,
					16,
					16);
			if ((npsn & 0xffff0000) !=
					(qp->s_ahgpsn & 0xffff0000)) {
				priv->s_ahg->ahgcount++;
				priv->s_ahg->ahgdesc[1] =
					sdma_build_ahg_descriptor(
						(__force u16)cpu_to_be16(
							(u16)(npsn >> 16)),
						BTH2_OFFSET,
						0,
						16);
			}
		}
	}
}

static inline void hfi1_make_ruc_bth(struct rvt_qp *qp,
				     struct ib_other_headers *ohdr,
				     u32 bth0, u32 bth1, u32 bth2)
{
	bth1 |= qp->remote_qpn;
	ohdr->bth[0] = cpu_to_be32(bth0);
	ohdr->bth[1] = cpu_to_be32(bth1);
	ohdr->bth[2] = cpu_to_be32(bth2);
}

/**
 * hfi1_make_ruc_header_16B - build a 16B header
 * @qp: the queue pair
 * @ohdr: a pointer to the destination header memory
 * @bth0: bth0 passed in from the RC/UC builder
 * @bth2: bth2 passed in from the RC/UC builder
 * @middle: non zero implies indicates ahg "could" be used
 * @ps: the current packet state
 *
 * This routine may disarm ahg under these situations:
 * - packet needs a GRH
 * - BECN needed
 * - migration state not IB_MIG_MIGRATED
 */
static inline void hfi1_make_ruc_header_16B(struct rvt_qp *qp,
					    struct ib_other_headers *ohdr,
					    u32 bth0, u32 bth2, int middle,
					    struct hfi1_pkt_state *ps)
{
	struct hfi1_qp_priv *priv = qp->priv;
	struct hfi1_ibport *ibp = ps->ibp;
	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
	u32 bth1 = 0;
	u32 slid;
	u16 pkey = hfi1_get_pkey(ibp, qp->s_pkey_index);
	u8 l4 = OPA_16B_L4_IB_LOCAL;
	u8 extra_bytes = hfi1_get_16b_padding(
				(ps->s_txreq->hdr_dwords << 2),
				ps->s_txreq->s_cur_size);
	u32 nwords = SIZE_OF_CRC + ((ps->s_txreq->s_cur_size +
				 extra_bytes + SIZE_OF_LT) >> 2);
	bool becn = false;

	if (unlikely(rdma_ah_get_ah_flags(&qp->remote_ah_attr) & IB_AH_GRH) &&
	    hfi1_check_mcast(rdma_ah_get_dlid(&qp->remote_ah_attr))) {
		struct ib_grh *grh;
		struct ib_global_route *grd =
			rdma_ah_retrieve_grh(&qp->remote_ah_attr);
		/*
		 * Ensure OPA GIDs are transformed to IB gids
		 * before creating the GRH.
		 */
		if (grd->sgid_index == OPA_GID_INDEX)
			grd->sgid_index = 0;
		grh = &ps->s_txreq->phdr.hdr.opah.u.l.grh;
		l4 = OPA_16B_L4_IB_GLOBAL;
		ps->s_txreq->hdr_dwords +=
			hfi1_make_grh(ibp, grh, grd,
				      ps->s_txreq->hdr_dwords - LRH_16B_DWORDS,
				      nwords);
		middle = 0;
	}

	if (qp->s_mig_state == IB_MIG_MIGRATED)
		bth1 |= OPA_BTH_MIG_REQ;
	else
		middle = 0;

	if (qp->s_flags & RVT_S_ECN) {
		qp->s_flags &= ~RVT_S_ECN;
		/* we recently received a FECN, so return a BECN */
		becn = true;
		middle = 0;
	}
	if (middle)
		build_ahg(qp, bth2);
	else
		qp->s_flags &= ~HFI1_S_AHG_VALID;

	bth0 |= pkey;
	bth0 |= extra_bytes << 20;
	hfi1_make_ruc_bth(qp, ohdr, bth0, bth1, bth2);

	if (!ppd->lid)
		slid = be32_to_cpu(OPA_LID_PERMISSIVE);
	else
		slid = ppd->lid |
			(rdma_ah_get_path_bits(&qp->remote_ah_attr) &
			((1 << ppd->lmc) - 1));

	hfi1_make_16b_hdr(&ps->s_txreq->phdr.hdr.opah,
			  slid,
			  opa_get_lid(rdma_ah_get_dlid(&qp->remote_ah_attr),
				      16B),
			  (ps->s_txreq->hdr_dwords + nwords) >> 1,
			  pkey, becn, 0, l4, priv->s_sc);
}

/**
 * hfi1_make_ruc_header_9B - build a 9B header
 * @qp: the queue pair
 * @ohdr: a pointer to the destination header memory
 * @bth0: bth0 passed in from the RC/UC builder
 * @bth2: bth2 passed in from the RC/UC builder
 * @middle: non zero implies indicates ahg "could" be used
 * @ps: the current packet state
 *
 * This routine may disarm ahg under these situations:
 * - packet needs a GRH
 * - BECN needed
 * - migration state not IB_MIG_MIGRATED
 */
static inline void hfi1_make_ruc_header_9B(struct rvt_qp *qp,
					   struct ib_other_headers *ohdr,
					   u32 bth0, u32 bth2, int middle,
					   struct hfi1_pkt_state *ps)
{
	struct hfi1_qp_priv *priv = qp->priv;
	struct hfi1_ibport *ibp = ps->ibp;
	u32 bth1 = 0;
	u16 pkey = hfi1_get_pkey(ibp, qp->s_pkey_index);
	u16 lrh0 = HFI1_LRH_BTH;
	u8 extra_bytes = -ps->s_txreq->s_cur_size & 3;
	u32 nwords = SIZE_OF_CRC + ((ps->s_txreq->s_cur_size +
					 extra_bytes) >> 2);

	if (unlikely(rdma_ah_get_ah_flags(&qp->remote_ah_attr) & IB_AH_GRH)) {
		struct ib_grh *grh = &ps->s_txreq->phdr.hdr.ibh.u.l.grh;

		lrh0 = HFI1_LRH_GRH;
		ps->s_txreq->hdr_dwords +=
			hfi1_make_grh(ibp, grh,
				      rdma_ah_read_grh(&qp->remote_ah_attr),
				      ps->s_txreq->hdr_dwords - LRH_9B_DWORDS,
				      nwords);
		middle = 0;
	}
	lrh0 |= (priv->s_sc & 0xf) << 12 |
		(rdma_ah_get_sl(&qp->remote_ah_attr) & 0xf) << 4;

	if (qp->s_mig_state == IB_MIG_MIGRATED)
		bth0 |= IB_BTH_MIG_REQ;
	else
		middle = 0;

	if (qp->s_flags & RVT_S_ECN) {
		qp->s_flags &= ~RVT_S_ECN;
		/* we recently received a FECN, so return a BECN */
		bth1 |= (IB_BECN_MASK << IB_BECN_SHIFT);
		middle = 0;
	}
	if (middle)
		build_ahg(qp, bth2);
	else
		qp->s_flags &= ~HFI1_S_AHG_VALID;

	bth0 |= pkey;
	bth0 |= extra_bytes << 20;
	hfi1_make_ruc_bth(qp, ohdr, bth0, bth1, bth2);
	hfi1_make_ib_hdr(&ps->s_txreq->phdr.hdr.ibh,
			 lrh0,
			 ps->s_txreq->hdr_dwords + nwords,
			 opa_get_lid(rdma_ah_get_dlid(&qp->remote_ah_attr), 9B),
			 ppd_from_ibp(ibp)->lid |
				rdma_ah_get_path_bits(&qp->remote_ah_attr));
}

typedef void (*hfi1_make_ruc_hdr)(struct rvt_qp *qp,
				  struct ib_other_headers *ohdr,
				  u32 bth0, u32 bth2, int middle,
				  struct hfi1_pkt_state *ps);

/* We support only two types - 9B and 16B for now */
static const hfi1_make_ruc_hdr hfi1_ruc_header_tbl[2] = {
	[HFI1_PKT_TYPE_9B] = &hfi1_make_ruc_header_9B,
	[HFI1_PKT_TYPE_16B] = &hfi1_make_ruc_header_16B
};

void hfi1_make_ruc_header(struct rvt_qp *qp, struct ib_other_headers *ohdr,
			  u32 bth0, u32 bth2, int middle,
			  struct hfi1_pkt_state *ps)
{
	struct hfi1_qp_priv *priv = qp->priv;

	/*
	 * reset s_ahg/AHG fields
	 *
	 * This insures that the ahgentry/ahgcount
	 * are at a non-AHG default to protect
	 * build_verbs_tx_desc() from using
	 * an include ahgidx.
	 *
	 * build_ahg() will modify as appropriate
	 * to use the AHG feature.
	 */
	priv->s_ahg->tx_flags = 0;
	priv->s_ahg->ahgcount = 0;
	priv->s_ahg->ahgidx = 0;

	/* Make the appropriate header */
	hfi1_ruc_header_tbl[priv->hdr_type](qp, ohdr, bth0, bth2, middle, ps);
}

/* when sending, force a reschedule every one of these periods */
#define SEND_RESCHED_TIMEOUT (5 * HZ)  /* 5s in jiffies */

/**
 * schedule_send_yield - test for a yield required for QP send engine
 * @timeout: Final time for timeout slice for jiffies
 * @qp: a pointer to QP
 * @ps: a pointer to a structure with commonly lookup values for
 *      the the send engine progress
 *
 * This routine checks if the time slice for the QP has expired
 * for RC QPs, if so an additional work entry is queued. At this
 * point, other QPs have an opportunity to be scheduled. It
 * returns true if a yield is required, otherwise, false
 * is returned.
 */
static bool schedule_send_yield(struct rvt_qp *qp,
				struct hfi1_pkt_state *ps)
{
	ps->pkts_sent = true;

	if (unlikely(time_after(jiffies, ps->timeout))) {
		if (!ps->in_thread ||
		    workqueue_congested(ps->cpu, ps->ppd->hfi1_wq)) {
			spin_lock_irqsave(&qp->s_lock, ps->flags);
			qp->s_flags &= ~RVT_S_BUSY;
			hfi1_schedule_send(qp);
			spin_unlock_irqrestore(&qp->s_lock, ps->flags);
			this_cpu_inc(*ps->ppd->dd->send_schedule);
			trace_hfi1_rc_expired_time_slice(qp, true);
			return true;
		}

		cond_resched();
		this_cpu_inc(*ps->ppd->dd->send_schedule);
		ps->timeout = jiffies + ps->timeout_int;
	}

	trace_hfi1_rc_expired_time_slice(qp, false);
	return false;
}

void hfi1_do_send_from_rvt(struct rvt_qp *qp)
{
	hfi1_do_send(qp, false);
}

void _hfi1_do_send(struct work_struct *work)
{
	struct iowait *wait = container_of(work, struct iowait, iowork);
	struct rvt_qp *qp = iowait_to_qp(wait);

	hfi1_do_send(qp, true);
}

/**
 * hfi1_do_send - perform a send on a QP
 * @work: contains a pointer to the QP
 * @in_thread: true if in a workqueue thread
 *
 * Process entries in the send work queue until credit or queue is
 * exhausted.  Only allow one CPU to send a packet per QP.
 * Otherwise, two threads could send packets out of order.
 */
void hfi1_do_send(struct rvt_qp *qp, bool in_thread)
{
	struct hfi1_pkt_state ps;
	struct hfi1_qp_priv *priv = qp->priv;
	int (*make_req)(struct rvt_qp *qp, struct hfi1_pkt_state *ps);

	ps.dev = to_idev(qp->ibqp.device);
	ps.ibp = to_iport(qp->ibqp.device, qp->port_num);
	ps.ppd = ppd_from_ibp(ps.ibp);
	ps.in_thread = in_thread;

	trace_hfi1_rc_do_send(qp, in_thread);

	switch (qp->ibqp.qp_type) {
	case IB_QPT_RC:
		if (!loopback && ((rdma_ah_get_dlid(&qp->remote_ah_attr) &
				   ~((1 << ps.ppd->lmc) - 1)) ==
				  ps.ppd->lid)) {
			ruc_loopback(qp);
			return;
		}
		make_req = hfi1_make_rc_req;
		ps.timeout_int = qp->timeout_jiffies;
		break;
	case IB_QPT_UC:
		if (!loopback && ((rdma_ah_get_dlid(&qp->remote_ah_attr) &
				   ~((1 << ps.ppd->lmc) - 1)) ==
				  ps.ppd->lid)) {
			ruc_loopback(qp);
			return;
		}
		make_req = hfi1_make_uc_req;
		ps.timeout_int = SEND_RESCHED_TIMEOUT;
		break;
	default:
		make_req = hfi1_make_ud_req;
		ps.timeout_int = SEND_RESCHED_TIMEOUT;
	}

	spin_lock_irqsave(&qp->s_lock, ps.flags);

	/* Return if we are already busy processing a work request. */
	if (!hfi1_send_ok(qp)) {
		spin_unlock_irqrestore(&qp->s_lock, ps.flags);
		return;
	}

	qp->s_flags |= RVT_S_BUSY;

	ps.timeout_int = ps.timeout_int / 8;
	ps.timeout = jiffies + ps.timeout_int;
	ps.cpu = priv->s_sde ? priv->s_sde->cpu :
			cpumask_first(cpumask_of_node(ps.ppd->dd->node));
	ps.pkts_sent = false;

	/* insure a pre-built packet is handled  */
	ps.s_txreq = get_waiting_verbs_txreq(qp);
	do {
		/* Check for a constructed packet to be sent. */
		if (ps.s_txreq) {
			spin_unlock_irqrestore(&qp->s_lock, ps.flags);
			/*
			 * If the packet cannot be sent now, return and
			 * the send engine will be woken up later.
			 */
			if (hfi1_verbs_send(qp, &ps))
				return;
			/* allow other tasks to run */
			if (schedule_send_yield(qp, &ps))
				return;

			spin_lock_irqsave(&qp->s_lock, ps.flags);
		}
	} while (make_req(qp, &ps));
	iowait_starve_clear(ps.pkts_sent, &priv->s_iowait);
	spin_unlock_irqrestore(&qp->s_lock, ps.flags);
}

/*
 * This should be called with s_lock held.
 */
void hfi1_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe,
			enum ib_wc_status status)
{
	u32 old_last, last;

	if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_OR_FLUSH_SEND))
		return;

	last = qp->s_last;
	old_last = last;
	trace_hfi1_qp_send_completion(qp, wqe, last);
	if (++last >= qp->s_size)
		last = 0;
	trace_hfi1_qp_send_completion(qp, wqe, last);
	qp->s_last = last;
	/* See post_send() */
	barrier();
	rvt_put_swqe(wqe);
	if (qp->ibqp.qp_type == IB_QPT_UD ||
	    qp->ibqp.qp_type == IB_QPT_SMI ||
	    qp->ibqp.qp_type == IB_QPT_GSI)
		atomic_dec(&ibah_to_rvtah(wqe->ud_wr.ah)->refcount);

	rvt_qp_swqe_complete(qp,
			     wqe,
			     ib_hfi1_wc_opcode[wqe->wr.opcode],
			     status);

	if (qp->s_acked == old_last)
		qp->s_acked = last;
	if (qp->s_cur == old_last)
		qp->s_cur = last;
	if (qp->s_tail == old_last)
		qp->s_tail = last;
	if (qp->state == IB_QPS_SQD && last == qp->s_cur)
		qp->s_draining = 0;
}