summaryrefslogtreecommitdiffstats
path: root/drivers/infiniband/hw/hfi1/qp.c
blob: a5aa3517e7d5c537d36786fd0d512cef6d60544c (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
/*
 * Copyright(c) 2015, 2016 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/err.h>
#include <linux/vmalloc.h>
#include <linux/hash.h>
#include <linux/module.h>
#include <linux/seq_file.h>
#include <rdma/rdma_vt.h>
#include <rdma/rdmavt_qp.h>
#include <rdma/ib_verbs.h>

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

unsigned int hfi1_qp_table_size = 256;
module_param_named(qp_table_size, hfi1_qp_table_size, uint, S_IRUGO);
MODULE_PARM_DESC(qp_table_size, "QP table size");

static void flush_tx_list(struct rvt_qp *qp);
static int iowait_sleep(
	struct sdma_engine *sde,
	struct iowait *wait,
	struct sdma_txreq *stx,
	unsigned seq);
static void iowait_wakeup(struct iowait *wait, int reason);
static void iowait_sdma_drained(struct iowait *wait);
static void qp_pio_drain(struct rvt_qp *qp);

static inline unsigned mk_qpn(struct rvt_qpn_table *qpt,
			      struct rvt_qpn_map *map, unsigned off)
{
	return (map - qpt->map) * RVT_BITS_PER_PAGE + off;
}

/*
 * Convert the AETH credit code into the number of credits.
 */
static const u16 credit_table[31] = {
	0,                      /* 0 */
	1,                      /* 1 */
	2,                      /* 2 */
	3,                      /* 3 */
	4,                      /* 4 */
	6,                      /* 5 */
	8,                      /* 6 */
	12,                     /* 7 */
	16,                     /* 8 */
	24,                     /* 9 */
	32,                     /* A */
	48,                     /* B */
	64,                     /* C */
	96,                     /* D */
	128,                    /* E */
	192,                    /* F */
	256,                    /* 10 */
	384,                    /* 11 */
	512,                    /* 12 */
	768,                    /* 13 */
	1024,                   /* 14 */
	1536,                   /* 15 */
	2048,                   /* 16 */
	3072,                   /* 17 */
	4096,                   /* 18 */
	6144,                   /* 19 */
	8192,                   /* 1A */
	12288,                  /* 1B */
	16384,                  /* 1C */
	24576,                  /* 1D */
	32768                   /* 1E */
};

const struct rvt_operation_params hfi1_post_parms[RVT_OPERATION_MAX] = {
[IB_WR_RDMA_WRITE] = {
	.length = sizeof(struct ib_rdma_wr),
	.qpt_support = BIT(IB_QPT_UC) | BIT(IB_QPT_RC),
},

[IB_WR_RDMA_READ] = {
	.length = sizeof(struct ib_rdma_wr),
	.qpt_support = BIT(IB_QPT_RC),
	.flags = RVT_OPERATION_ATOMIC,
},

[IB_WR_ATOMIC_CMP_AND_SWP] = {
	.length = sizeof(struct ib_atomic_wr),
	.qpt_support = BIT(IB_QPT_RC),
	.flags = RVT_OPERATION_ATOMIC | RVT_OPERATION_ATOMIC_SGE,
},

[IB_WR_ATOMIC_FETCH_AND_ADD] = {
	.length = sizeof(struct ib_atomic_wr),
	.qpt_support = BIT(IB_QPT_RC),
	.flags = RVT_OPERATION_ATOMIC | RVT_OPERATION_ATOMIC_SGE,
},

[IB_WR_RDMA_WRITE_WITH_IMM] = {
	.length = sizeof(struct ib_rdma_wr),
	.qpt_support = BIT(IB_QPT_UC) | BIT(IB_QPT_RC),
},

[IB_WR_SEND] = {
	.length = sizeof(struct ib_send_wr),
	.qpt_support = BIT(IB_QPT_UD) | BIT(IB_QPT_SMI) | BIT(IB_QPT_GSI) |
		       BIT(IB_QPT_UC) | BIT(IB_QPT_RC),
},

[IB_WR_SEND_WITH_IMM] = {
	.length = sizeof(struct ib_send_wr),
	.qpt_support = BIT(IB_QPT_UD) | BIT(IB_QPT_SMI) | BIT(IB_QPT_GSI) |
		       BIT(IB_QPT_UC) | BIT(IB_QPT_RC),
},

[IB_WR_REG_MR] = {
	.length = sizeof(struct ib_reg_wr),
	.qpt_support = BIT(IB_QPT_UC) | BIT(IB_QPT_RC),
	.flags = RVT_OPERATION_LOCAL,
},

[IB_WR_LOCAL_INV] = {
	.length = sizeof(struct ib_send_wr),
	.qpt_support = BIT(IB_QPT_UC) | BIT(IB_QPT_RC),
	.flags = RVT_OPERATION_LOCAL,
},

[IB_WR_SEND_WITH_INV] = {
	.length = sizeof(struct ib_send_wr),
	.qpt_support = BIT(IB_QPT_RC),
},

};

static void flush_tx_list(struct rvt_qp *qp)
{
	struct hfi1_qp_priv *priv = qp->priv;

	while (!list_empty(&priv->s_iowait.tx_head)) {
		struct sdma_txreq *tx;

		tx = list_first_entry(
			&priv->s_iowait.tx_head,
			struct sdma_txreq,
			list);
		list_del_init(&tx->list);
		hfi1_put_txreq(
			container_of(tx, struct verbs_txreq, txreq));
	}
}

static void flush_iowait(struct rvt_qp *qp)
{
	struct hfi1_qp_priv *priv = qp->priv;
	struct hfi1_ibdev *dev = to_idev(qp->ibqp.device);
	unsigned long flags;

	write_seqlock_irqsave(&dev->iowait_lock, flags);
	if (!list_empty(&priv->s_iowait.list)) {
		list_del_init(&priv->s_iowait.list);
		if (atomic_dec_and_test(&qp->refcount))
			wake_up(&qp->wait);
	}
	write_sequnlock_irqrestore(&dev->iowait_lock, flags);
}

static inline int opa_mtu_enum_to_int(int mtu)
{
	switch (mtu) {
	case OPA_MTU_8192:  return 8192;
	case OPA_MTU_10240: return 10240;
	default:            return -1;
	}
}

/**
 * This function is what we would push to the core layer if we wanted to be a
 * "first class citizen".  Instead we hide this here and rely on Verbs ULPs
 * to blindly pass the MTU enum value from the PathRecord to us.
 */
static inline int verbs_mtu_enum_to_int(struct ib_device *dev, enum ib_mtu mtu)
{
	int val;

	/* Constraining 10KB packets to 8KB packets */
	if (mtu == (enum ib_mtu)OPA_MTU_10240)
		mtu = OPA_MTU_8192;
	val = opa_mtu_enum_to_int((int)mtu);
	if (val > 0)
		return val;
	return ib_mtu_enum_to_int(mtu);
}

int hfi1_check_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
			 int attr_mask, struct ib_udata *udata)
{
	struct ib_qp *ibqp = &qp->ibqp;
	struct hfi1_ibdev *dev = to_idev(ibqp->device);
	struct hfi1_devdata *dd = dd_from_dev(dev);
	u8 sc;

	if (attr_mask & IB_QP_AV) {
		sc = ah_to_sc(ibqp->device, &attr->ah_attr);
		if (sc == 0xf)
			return -EINVAL;

		if (!qp_to_sdma_engine(qp, sc) &&
		    dd->flags & HFI1_HAS_SEND_DMA)
			return -EINVAL;

		if (!qp_to_send_context(qp, sc))
			return -EINVAL;
	}

	if (attr_mask & IB_QP_ALT_PATH) {
		sc = ah_to_sc(ibqp->device, &attr->alt_ah_attr);
		if (sc == 0xf)
			return -EINVAL;

		if (!qp_to_sdma_engine(qp, sc) &&
		    dd->flags & HFI1_HAS_SEND_DMA)
			return -EINVAL;

		if (!qp_to_send_context(qp, sc))
			return -EINVAL;
	}

	return 0;
}

void hfi1_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr,
		    int attr_mask, struct ib_udata *udata)
{
	struct ib_qp *ibqp = &qp->ibqp;
	struct hfi1_qp_priv *priv = qp->priv;

	if (attr_mask & IB_QP_AV) {
		priv->s_sc = ah_to_sc(ibqp->device, &qp->remote_ah_attr);
		priv->s_sde = qp_to_sdma_engine(qp, priv->s_sc);
		priv->s_sendcontext = qp_to_send_context(qp, priv->s_sc);
	}

	if (attr_mask & IB_QP_PATH_MIG_STATE &&
	    attr->path_mig_state == IB_MIG_MIGRATED &&
	    qp->s_mig_state == IB_MIG_ARMED) {
		qp->s_flags |= RVT_S_AHG_CLEAR;
		priv->s_sc = ah_to_sc(ibqp->device, &qp->remote_ah_attr);
		priv->s_sde = qp_to_sdma_engine(qp, priv->s_sc);
		priv->s_sendcontext = qp_to_send_context(qp, priv->s_sc);
	}
}

/**
 * hfi1_check_send_wqe - validate wqe
 * @qp - The qp
 * @wqe - The built wqe
 *
 * validate wqe.  This is called
 * prior to inserting the wqe into
 * the ring but after the wqe has been
 * setup.
 *
 * Returns 0 on success, -EINVAL on failure
 *
 */
int hfi1_check_send_wqe(struct rvt_qp *qp,
			struct rvt_swqe *wqe)
{
	struct hfi1_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
	struct rvt_ah *ah;

	switch (qp->ibqp.qp_type) {
	case IB_QPT_RC:
	case IB_QPT_UC:
		if (wqe->length > 0x80000000U)
			return -EINVAL;
		break;
	case IB_QPT_SMI:
		ah = ibah_to_rvtah(wqe->ud_wr.ah);
		if (wqe->length > (1 << ah->log_pmtu))
			return -EINVAL;
		break;
	case IB_QPT_GSI:
	case IB_QPT_UD:
		ah = ibah_to_rvtah(wqe->ud_wr.ah);
		if (wqe->length > (1 << ah->log_pmtu))
			return -EINVAL;
		if (ibp->sl_to_sc[ah->attr.sl] == 0xf)
			return -EINVAL;
	default:
		break;
	}
	return wqe->length <= piothreshold;
}

/**
 * hfi1_compute_aeth - compute the AETH (syndrome + MSN)
 * @qp: the queue pair to compute the AETH for
 *
 * Returns the AETH.
 */
__be32 hfi1_compute_aeth(struct rvt_qp *qp)
{
	u32 aeth = qp->r_msn & HFI1_MSN_MASK;

	if (qp->ibqp.srq) {
		/*
		 * Shared receive queues don't generate credits.
		 * Set the credit field to the invalid value.
		 */
		aeth |= HFI1_AETH_CREDIT_INVAL << HFI1_AETH_CREDIT_SHIFT;
	} else {
		u32 min, max, x;
		u32 credits;
		struct rvt_rwq *wq = qp->r_rq.wq;
		u32 head;
		u32 tail;

		/* sanity check pointers before trusting them */
		head = wq->head;
		if (head >= qp->r_rq.size)
			head = 0;
		tail = wq->tail;
		if (tail >= qp->r_rq.size)
			tail = 0;
		/*
		 * Compute the number of credits available (RWQEs).
		 * There is a small chance that the pair of reads are
		 * not atomic, which is OK, since the fuzziness is
		 * resolved as further ACKs go out.
		 */
		credits = head - tail;
		if ((int)credits < 0)
			credits += qp->r_rq.size;
		/*
		 * Binary search the credit table to find the code to
		 * use.
		 */
		min = 0;
		max = 31;
		for (;;) {
			x = (min + max) / 2;
			if (credit_table[x] == credits)
				break;
			if (credit_table[x] > credits) {
				max = x;
			} else {
				if (min == x)
					break;
				min = x;
			}
		}
		aeth |= x << HFI1_AETH_CREDIT_SHIFT;
	}
	return cpu_to_be32(aeth);
}

/**
 * _hfi1_schedule_send - schedule progress
 * @qp: the QP
 *
 * This schedules qp progress w/o regard to the s_flags.
 *
 * It is only used in the post send, which doesn't hold
 * the s_lock.
 */
void _hfi1_schedule_send(struct rvt_qp *qp)
{
	struct hfi1_qp_priv *priv = qp->priv;
	struct hfi1_ibport *ibp =
		to_iport(qp->ibqp.device, qp->port_num);
	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
	struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);

	iowait_schedule(&priv->s_iowait, ppd->hfi1_wq,
			priv->s_sde ?
			priv->s_sde->cpu :
			cpumask_first(cpumask_of_node(dd->node)));
}

static void qp_pio_drain(struct rvt_qp *qp)
{
	struct hfi1_ibdev *dev;
	struct hfi1_qp_priv *priv = qp->priv;

	if (!priv->s_sendcontext)
		return;
	dev = to_idev(qp->ibqp.device);
	while (iowait_pio_pending(&priv->s_iowait)) {
		write_seqlock_irq(&dev->iowait_lock);
		hfi1_sc_wantpiobuf_intr(priv->s_sendcontext, 1);
		write_sequnlock_irq(&dev->iowait_lock);
		iowait_pio_drain(&priv->s_iowait);
		write_seqlock_irq(&dev->iowait_lock);
		hfi1_sc_wantpiobuf_intr(priv->s_sendcontext, 0);
		write_sequnlock_irq(&dev->iowait_lock);
	}
}

/**
 * hfi1_schedule_send - schedule progress
 * @qp: the QP
 *
 * This schedules qp progress and caller should hold
 * the s_lock.
 */
void hfi1_schedule_send(struct rvt_qp *qp)
{
	if (hfi1_send_ok(qp))
		_hfi1_schedule_send(qp);
}

/**
 * hfi1_get_credit - flush the send work queue of a QP
 * @qp: the qp who's send work queue to flush
 * @aeth: the Acknowledge Extended Transport Header
 *
 * The QP s_lock should be held.
 */
void hfi1_get_credit(struct rvt_qp *qp, u32 aeth)
{
	u32 credit = (aeth >> HFI1_AETH_CREDIT_SHIFT) & HFI1_AETH_CREDIT_MASK;

	/*
	 * If the credit is invalid, we can send
	 * as many packets as we like.  Otherwise, we have to
	 * honor the credit field.
	 */
	if (credit == HFI1_AETH_CREDIT_INVAL) {
		if (!(qp->s_flags & RVT_S_UNLIMITED_CREDIT)) {
			qp->s_flags |= RVT_S_UNLIMITED_CREDIT;
			if (qp->s_flags & RVT_S_WAIT_SSN_CREDIT) {
				qp->s_flags &= ~RVT_S_WAIT_SSN_CREDIT;
				hfi1_schedule_send(qp);
			}
		}
	} else if (!(qp->s_flags & RVT_S_UNLIMITED_CREDIT)) {
		/* Compute new LSN (i.e., MSN + credit) */
		credit = (aeth + credit_table[credit]) & HFI1_MSN_MASK;
		if (cmp_msn(credit, qp->s_lsn) > 0) {
			qp->s_lsn = credit;
			if (qp->s_flags & RVT_S_WAIT_SSN_CREDIT) {
				qp->s_flags &= ~RVT_S_WAIT_SSN_CREDIT;
				hfi1_schedule_send(qp);
			}
		}
	}
}

void hfi1_qp_wakeup(struct rvt_qp *qp, u32 flag)
{
	unsigned long flags;

	spin_lock_irqsave(&qp->s_lock, flags);
	if (qp->s_flags & flag) {
		qp->s_flags &= ~flag;
		trace_hfi1_qpwakeup(qp, flag);
		hfi1_schedule_send(qp);
	}
	spin_unlock_irqrestore(&qp->s_lock, flags);
	/* Notify hfi1_destroy_qp() if it is waiting. */
	if (atomic_dec_and_test(&qp->refcount))
		wake_up(&qp->wait);
}

static int iowait_sleep(
	struct sdma_engine *sde,
	struct iowait *wait,
	struct sdma_txreq *stx,
	unsigned seq)
{
	struct verbs_txreq *tx = container_of(stx, struct verbs_txreq, txreq);
	struct rvt_qp *qp;
	struct hfi1_qp_priv *priv;
	unsigned long flags;
	int ret = 0;
	struct hfi1_ibdev *dev;

	qp = tx->qp;
	priv = qp->priv;

	spin_lock_irqsave(&qp->s_lock, flags);
	if (ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) {
		/*
		 * If we couldn't queue the DMA request, save the info
		 * and try again later rather than destroying the
		 * buffer and undoing the side effects of the copy.
		 */
		/* Make a common routine? */
		dev = &sde->dd->verbs_dev;
		list_add_tail(&stx->list, &wait->tx_head);
		write_seqlock(&dev->iowait_lock);
		if (sdma_progress(sde, seq, stx))
			goto eagain;
		if (list_empty(&priv->s_iowait.list)) {
			struct hfi1_ibport *ibp =
				to_iport(qp->ibqp.device, qp->port_num);

			ibp->rvp.n_dmawait++;
			qp->s_flags |= RVT_S_WAIT_DMA_DESC;
			list_add_tail(&priv->s_iowait.list, &sde->dmawait);
			trace_hfi1_qpsleep(qp, RVT_S_WAIT_DMA_DESC);
			atomic_inc(&qp->refcount);
		}
		write_sequnlock(&dev->iowait_lock);
		qp->s_flags &= ~RVT_S_BUSY;
		spin_unlock_irqrestore(&qp->s_lock, flags);
		ret = -EBUSY;
	} else {
		spin_unlock_irqrestore(&qp->s_lock, flags);
		hfi1_put_txreq(tx);
	}
	return ret;
eagain:
	write_sequnlock(&dev->iowait_lock);
	spin_unlock_irqrestore(&qp->s_lock, flags);
	list_del_init(&stx->list);
	return -EAGAIN;
}

static void iowait_wakeup(struct iowait *wait, int reason)
{
	struct rvt_qp *qp = iowait_to_qp(wait);

	WARN_ON(reason != SDMA_AVAIL_REASON);
	hfi1_qp_wakeup(qp, RVT_S_WAIT_DMA_DESC);
}

static void iowait_sdma_drained(struct iowait *wait)
{
	struct rvt_qp *qp = iowait_to_qp(wait);
	unsigned long flags;

	/*
	 * This happens when the send engine notes
	 * a QP in the error state and cannot
	 * do the flush work until that QP's
	 * sdma work has finished.
	 */
	spin_lock_irqsave(&qp->s_lock, flags);
	if (qp->s_flags & RVT_S_WAIT_DMA) {
		qp->s_flags &= ~RVT_S_WAIT_DMA;
		hfi1_schedule_send(qp);
	}
	spin_unlock_irqrestore(&qp->s_lock, flags);
}

/**
 *
 * qp_to_sdma_engine - map a qp to a send engine
 * @qp: the QP
 * @sc5: the 5 bit sc
 *
 * Return:
 * A send engine for the qp or NULL for SMI type qp.
 */
struct sdma_engine *qp_to_sdma_engine(struct rvt_qp *qp, u8 sc5)
{
	struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);
	struct sdma_engine *sde;

	if (!(dd->flags & HFI1_HAS_SEND_DMA))
		return NULL;
	switch (qp->ibqp.qp_type) {
	case IB_QPT_SMI:
		return NULL;
	default:
		break;
	}
	sde = sdma_select_engine_sc(dd, qp->ibqp.qp_num >> dd->qos_shift, sc5);
	return sde;
}

/*
 * qp_to_send_context - map a qp to a send context
 * @qp: the QP
 * @sc5: the 5 bit sc
 *
 * Return:
 * A send context for the qp
 */
struct send_context *qp_to_send_context(struct rvt_qp *qp, u8 sc5)
{
	struct hfi1_devdata *dd = dd_from_ibdev(qp->ibqp.device);

	switch (qp->ibqp.qp_type) {
	case IB_QPT_SMI:
		/* SMA packets to VL15 */
		return dd->vld[15].sc;
	default:
		break;
	}

	return pio_select_send_context_sc(dd, qp->ibqp.qp_num >> dd->qos_shift,
					  sc5);
}

struct qp_iter {
	struct hfi1_ibdev *dev;
	struct rvt_qp *qp;
	int specials;
	int n;
};

struct qp_iter *qp_iter_init(struct hfi1_ibdev *dev)
{
	struct qp_iter *iter;

	iter = kzalloc(sizeof(*iter), GFP_KERNEL);
	if (!iter)
		return NULL;

	iter->dev = dev;
	iter->specials = dev->rdi.ibdev.phys_port_cnt * 2;
	if (qp_iter_next(iter)) {
		kfree(iter);
		return NULL;
	}

	return iter;
}

int qp_iter_next(struct qp_iter *iter)
{
	struct hfi1_ibdev *dev = iter->dev;
	int n = iter->n;
	int ret = 1;
	struct rvt_qp *pqp = iter->qp;
	struct rvt_qp *qp;

	/*
	 * The approach is to consider the special qps
	 * as an additional table entries before the
	 * real hash table.  Since the qp code sets
	 * the qp->next hash link to NULL, this works just fine.
	 *
	 * iter->specials is 2 * # ports
	 *
	 * n = 0..iter->specials is the special qp indices
	 *
	 * n = iter->specials..dev->rdi.qp_dev->qp_table_size+iter->specials are
	 * the potential hash bucket entries
	 *
	 */
	for (; n <  dev->rdi.qp_dev->qp_table_size + iter->specials; n++) {
		if (pqp) {
			qp = rcu_dereference(pqp->next);
		} else {
			if (n < iter->specials) {
				struct hfi1_pportdata *ppd;
				struct hfi1_ibport *ibp;
				int pidx;

				pidx = n % dev->rdi.ibdev.phys_port_cnt;
				ppd = &dd_from_dev(dev)->pport[pidx];
				ibp = &ppd->ibport_data;

				if (!(n & 1))
					qp = rcu_dereference(ibp->rvp.qp[0]);
				else
					qp = rcu_dereference(ibp->rvp.qp[1]);
			} else {
				qp = rcu_dereference(
					dev->rdi.qp_dev->qp_table[
						(n - iter->specials)]);
			}
		}
		pqp = qp;
		if (qp) {
			iter->qp = qp;
			iter->n = n;
			return 0;
		}
	}
	return ret;
}

static const char * const qp_type_str[] = {
	"SMI", "GSI", "RC", "UC", "UD",
};

static int qp_idle(struct rvt_qp *qp)
{
	return
		qp->s_last == qp->s_acked &&
		qp->s_acked == qp->s_cur &&
		qp->s_cur == qp->s_tail &&
		qp->s_tail == qp->s_head;
}

void qp_iter_print(struct seq_file *s, struct qp_iter *iter)
{
	struct rvt_swqe *wqe;
	struct rvt_qp *qp = iter->qp;
	struct hfi1_qp_priv *priv = qp->priv;
	struct sdma_engine *sde;
	struct send_context *send_context;

	sde = qp_to_sdma_engine(qp, priv->s_sc);
	wqe = rvt_get_swqe_ptr(qp, qp->s_last);
	send_context = qp_to_send_context(qp, priv->s_sc);
	seq_printf(s,
		   "N %d %s QP %x R %u %s %u %u %u f=%x %u %u %u %u %u %u PSN %x %x %x %x %x (%u %u %u %u %u %u %u) RQP %x LID %x SL %u MTU %u %u %u %u SDE %p,%u SC %p,%u SCQ %u %u PID %d\n",
		   iter->n,
		   qp_idle(qp) ? "I" : "B",
		   qp->ibqp.qp_num,
		   atomic_read(&qp->refcount),
		   qp_type_str[qp->ibqp.qp_type],
		   qp->state,
		   wqe ? wqe->wr.opcode : 0,
		   qp->s_hdrwords,
		   qp->s_flags,
		   iowait_sdma_pending(&priv->s_iowait),
		   iowait_pio_pending(&priv->s_iowait),
		   !list_empty(&priv->s_iowait.list),
		   qp->timeout,
		   wqe ? wqe->ssn : 0,
		   qp->s_lsn,
		   qp->s_last_psn,
		   qp->s_psn, qp->s_next_psn,
		   qp->s_sending_psn, qp->s_sending_hpsn,
		   qp->s_last, qp->s_acked, qp->s_cur,
		   qp->s_tail, qp->s_head, qp->s_size,
		   qp->s_avail,
		   qp->remote_qpn,
		   qp->remote_ah_attr.dlid,
		   qp->remote_ah_attr.sl,
		   qp->pmtu,
		   qp->s_retry,
		   qp->s_retry_cnt,
		   qp->s_rnr_retry_cnt,
		   sde,
		   sde ? sde->this_idx : 0,
		   send_context,
		   send_context ? send_context->sw_index : 0,
		   ibcq_to_rvtcq(qp->ibqp.send_cq)->queue->head,
		   ibcq_to_rvtcq(qp->ibqp.send_cq)->queue->tail,
		   qp->pid);
}

void qp_comm_est(struct rvt_qp *qp)
{
	qp->r_flags |= RVT_R_COMM_EST;
	if (qp->ibqp.event_handler) {
		struct ib_event ev;

		ev.device = qp->ibqp.device;
		ev.element.qp = &qp->ibqp;
		ev.event = IB_EVENT_COMM_EST;
		qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
	}
}

void *qp_priv_alloc(struct rvt_dev_info *rdi, struct rvt_qp *qp,
		    gfp_t gfp)
{
	struct hfi1_qp_priv *priv;

	priv = kzalloc_node(sizeof(*priv), gfp, rdi->dparms.node);
	if (!priv)
		return ERR_PTR(-ENOMEM);

	priv->owner = qp;

	priv->s_ahg = kzalloc_node(sizeof(*priv->s_ahg), gfp,
				   rdi->dparms.node);
	if (!priv->s_ahg) {
		kfree(priv);
		return ERR_PTR(-ENOMEM);
	}
	setup_timer(&priv->s_rnr_timer, hfi1_rc_rnr_retry, (unsigned long)qp);
	qp->s_timer.function = hfi1_rc_timeout;
	return priv;
}

void qp_priv_free(struct rvt_dev_info *rdi, struct rvt_qp *qp)
{
	struct hfi1_qp_priv *priv = qp->priv;

	kfree(priv->s_ahg);
	kfree(priv);
}

unsigned free_all_qps(struct rvt_dev_info *rdi)
{
	struct hfi1_ibdev *verbs_dev = container_of(rdi,
						    struct hfi1_ibdev,
						    rdi);
	struct hfi1_devdata *dd = container_of(verbs_dev,
					       struct hfi1_devdata,
					       verbs_dev);
	int n;
	unsigned qp_inuse = 0;

	for (n = 0; n < dd->num_pports; n++) {
		struct hfi1_ibport *ibp = &dd->pport[n].ibport_data;

		rcu_read_lock();
		if (rcu_dereference(ibp->rvp.qp[0]))
			qp_inuse++;
		if (rcu_dereference(ibp->rvp.qp[1]))
			qp_inuse++;
		rcu_read_unlock();
	}

	return qp_inuse;
}

void flush_qp_waiters(struct rvt_qp *qp)
{
	flush_iowait(qp);
	hfi1_stop_rc_timers(qp);
}

void stop_send_queue(struct rvt_qp *qp)
{
	struct hfi1_qp_priv *priv = qp->priv;

	cancel_work_sync(&priv->s_iowait.iowork);
	hfi1_del_timers_sync(qp);
}

void quiesce_qp(struct rvt_qp *qp)
{
	struct hfi1_qp_priv *priv = qp->priv;

	iowait_sdma_drain(&priv->s_iowait);
	qp_pio_drain(qp);
	flush_tx_list(qp);
}

void notify_qp_reset(struct rvt_qp *qp)
{
	struct hfi1_qp_priv *priv = qp->priv;

	iowait_init(
		&priv->s_iowait,
		1,
		_hfi1_do_send,
		iowait_sleep,
		iowait_wakeup,
		iowait_sdma_drained);
	priv->r_adefered = 0;
	clear_ahg(qp);
}

/*
 * Switch to alternate path.
 * The QP s_lock should be held and interrupts disabled.
 */
void hfi1_migrate_qp(struct rvt_qp *qp)
{
	struct hfi1_qp_priv *priv = qp->priv;
	struct ib_event ev;

	qp->s_mig_state = IB_MIG_MIGRATED;
	qp->remote_ah_attr = qp->alt_ah_attr;
	qp->port_num = qp->alt_ah_attr.port_num;
	qp->s_pkey_index = qp->s_alt_pkey_index;
	qp->s_flags |= RVT_S_AHG_CLEAR;
	priv->s_sc = ah_to_sc(qp->ibqp.device, &qp->remote_ah_attr);
	priv->s_sde = qp_to_sdma_engine(qp, priv->s_sc);

	ev.device = qp->ibqp.device;
	ev.element.qp = &qp->ibqp;
	ev.event = IB_EVENT_PATH_MIG;
	qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
}

int mtu_to_path_mtu(u32 mtu)
{
	return mtu_to_enum(mtu, OPA_MTU_8192);
}

u32 mtu_from_qp(struct rvt_dev_info *rdi, struct rvt_qp *qp, u32 pmtu)
{
	u32 mtu;
	struct hfi1_ibdev *verbs_dev = container_of(rdi,
						    struct hfi1_ibdev,
						    rdi);
	struct hfi1_devdata *dd = container_of(verbs_dev,
					       struct hfi1_devdata,
					       verbs_dev);
	struct hfi1_ibport *ibp;
	u8 sc, vl;

	ibp = &dd->pport[qp->port_num - 1].ibport_data;
	sc = ibp->sl_to_sc[qp->remote_ah_attr.sl];
	vl = sc_to_vlt(dd, sc);

	mtu = verbs_mtu_enum_to_int(qp->ibqp.device, pmtu);
	if (vl < PER_VL_SEND_CONTEXTS)
		mtu = min_t(u32, mtu, dd->vld[vl].mtu);
	return mtu;
}

int get_pmtu_from_attr(struct rvt_dev_info *rdi, struct rvt_qp *qp,
		       struct ib_qp_attr *attr)
{
	int mtu, pidx = qp->port_num - 1;
	struct hfi1_ibdev *verbs_dev = container_of(rdi,
						    struct hfi1_ibdev,
						    rdi);
	struct hfi1_devdata *dd = container_of(verbs_dev,
					       struct hfi1_devdata,
					       verbs_dev);
	mtu = verbs_mtu_enum_to_int(qp->ibqp.device, attr->path_mtu);
	if (mtu == -1)
		return -1; /* values less than 0 are error */

	if (mtu > dd->pport[pidx].ibmtu)
		return mtu_to_enum(dd->pport[pidx].ibmtu, IB_MTU_2048);
	else
		return attr->path_mtu;
}

void notify_error_qp(struct rvt_qp *qp)
{
	struct hfi1_ibdev *dev = to_idev(qp->ibqp.device);
	struct hfi1_qp_priv *priv = qp->priv;

	write_seqlock(&dev->iowait_lock);
	if (!list_empty(&priv->s_iowait.list) && !(qp->s_flags & RVT_S_BUSY)) {
		qp->s_flags &= ~RVT_S_ANY_WAIT_IO;
		list_del_init(&priv->s_iowait.list);
		if (atomic_dec_and_test(&qp->refcount))
			wake_up(&qp->wait);
	}
	write_sequnlock(&dev->iowait_lock);

	if (!(qp->s_flags & RVT_S_BUSY)) {
		qp->s_hdrwords = 0;
		if (qp->s_rdma_mr) {
			rvt_put_mr(qp->s_rdma_mr);
			qp->s_rdma_mr = NULL;
		}
		flush_tx_list(qp);
	}
}

/**
 * hfi1_error_port_qps - put a port's RC/UC qps into error state
 * @ibp: the ibport.
 * @sl: the service level.
 *
 * This function places all RC/UC qps with a given service level into error
 * state. It is generally called to force upper lay apps to abandon stale qps
 * after an sl->sc mapping change.
 */
void hfi1_error_port_qps(struct hfi1_ibport *ibp, u8 sl)
{
	struct rvt_qp *qp = NULL;
	struct hfi1_pportdata *ppd = ppd_from_ibp(ibp);
	struct hfi1_ibdev *dev = &ppd->dd->verbs_dev;
	int n;
	int lastwqe;
	struct ib_event ev;

	rcu_read_lock();

	/* Deal only with RC/UC qps that use the given SL. */
	for (n = 0; n < dev->rdi.qp_dev->qp_table_size; n++) {
		for (qp = rcu_dereference(dev->rdi.qp_dev->qp_table[n]); qp;
			qp = rcu_dereference(qp->next)) {
			if (qp->port_num == ppd->port &&
			    (qp->ibqp.qp_type == IB_QPT_UC ||
			     qp->ibqp.qp_type == IB_QPT_RC) &&
			    qp->remote_ah_attr.sl == sl &&
			    (ib_rvt_state_ops[qp->state] &
			     RVT_POST_SEND_OK)) {
				spin_lock_irq(&qp->r_lock);
				spin_lock(&qp->s_hlock);
				spin_lock(&qp->s_lock);
				lastwqe = rvt_error_qp(qp,
						       IB_WC_WR_FLUSH_ERR);
				spin_unlock(&qp->s_lock);
				spin_unlock(&qp->s_hlock);
				spin_unlock_irq(&qp->r_lock);
				if (lastwqe) {
					ev.device = qp->ibqp.device;
					ev.element.qp = &qp->ibqp;
					ev.event =
						IB_EVENT_QP_LAST_WQE_REACHED;
					qp->ibqp.event_handler(&ev,
						qp->ibqp.qp_context);
				}
			}
		}
	}

	rcu_read_unlock();
}