summaryrefslogtreecommitdiff
path: root/net/sunrpc/xprtrdma/svc_rdma_rw.c
blob: 48fe3b16b0d9c7903156e19810a5f8909312a7ad (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
// SPDX-License-Identifier: GPL-2.0
/*
 * Copyright (c) 2016-2018 Oracle.  All rights reserved.
 *
 * Use the core R/W API to move RPC-over-RDMA Read and Write chunks.
 */

#include <rdma/rw.h>

#include <linux/sunrpc/rpc_rdma.h>
#include <linux/sunrpc/svc_rdma.h>
#include <linux/sunrpc/debug.h>

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

#define RPCDBG_FACILITY	RPCDBG_SVCXPRT

static void svc_rdma_write_done(struct ib_cq *cq, struct ib_wc *wc);
static void svc_rdma_wc_read_done(struct ib_cq *cq, struct ib_wc *wc);

/* Each R/W context contains state for one chain of RDMA Read or
 * Write Work Requests.
 *
 * Each WR chain handles a single contiguous server-side buffer,
 * because scatterlist entries after the first have to start on
 * page alignment. xdr_buf iovecs cannot guarantee alignment.
 *
 * Each WR chain handles only one R_key. Each RPC-over-RDMA segment
 * from a client may contain a unique R_key, so each WR chain moves
 * up to one segment at a time.
 *
 * The scatterlist makes this data structure over 4KB in size. To
 * make it less likely to fail, and to handle the allocation for
 * smaller I/O requests without disabling bottom-halves, these
 * contexts are created on demand, but cached and reused until the
 * controlling svcxprt_rdma is destroyed.
 */
struct svc_rdma_rw_ctxt {
	struct list_head	rw_list;
	struct rdma_rw_ctx	rw_ctx;
	int			rw_nents;
	struct sg_table		rw_sg_table;
	struct scatterlist	rw_first_sgl[0];
};

static inline struct svc_rdma_rw_ctxt *
svc_rdma_next_ctxt(struct list_head *list)
{
	return list_first_entry_or_null(list, struct svc_rdma_rw_ctxt,
					rw_list);
}

static struct svc_rdma_rw_ctxt *
svc_rdma_get_rw_ctxt(struct svcxprt_rdma *rdma, unsigned int sges)
{
	struct svc_rdma_rw_ctxt *ctxt;

	spin_lock(&rdma->sc_rw_ctxt_lock);

	ctxt = svc_rdma_next_ctxt(&rdma->sc_rw_ctxts);
	if (ctxt) {
		list_del(&ctxt->rw_list);
		spin_unlock(&rdma->sc_rw_ctxt_lock);
	} else {
		spin_unlock(&rdma->sc_rw_ctxt_lock);
		ctxt = kmalloc(struct_size(ctxt, rw_first_sgl, SG_CHUNK_SIZE),
			       GFP_KERNEL);
		if (!ctxt)
			goto out;
		INIT_LIST_HEAD(&ctxt->rw_list);
	}

	ctxt->rw_sg_table.sgl = ctxt->rw_first_sgl;
	if (sg_alloc_table_chained(&ctxt->rw_sg_table, sges,
				   ctxt->rw_sg_table.sgl,
				   SG_CHUNK_SIZE)) {
		kfree(ctxt);
		ctxt = NULL;
	}
out:
	return ctxt;
}

static void svc_rdma_put_rw_ctxt(struct svcxprt_rdma *rdma,
				 struct svc_rdma_rw_ctxt *ctxt)
{
	sg_free_table_chained(&ctxt->rw_sg_table, SG_CHUNK_SIZE);

	spin_lock(&rdma->sc_rw_ctxt_lock);
	list_add(&ctxt->rw_list, &rdma->sc_rw_ctxts);
	spin_unlock(&rdma->sc_rw_ctxt_lock);
}

/**
 * svc_rdma_destroy_rw_ctxts - Free accumulated R/W contexts
 * @rdma: transport about to be destroyed
 *
 */
void svc_rdma_destroy_rw_ctxts(struct svcxprt_rdma *rdma)
{
	struct svc_rdma_rw_ctxt *ctxt;

	while ((ctxt = svc_rdma_next_ctxt(&rdma->sc_rw_ctxts)) != NULL) {
		list_del(&ctxt->rw_list);
		kfree(ctxt);
	}
}

/* A chunk context tracks all I/O for moving one Read or Write
 * chunk. This is a a set of rdma_rw's that handle data movement
 * for all segments of one chunk.
 *
 * These are small, acquired with a single allocator call, and
 * no more than one is needed per chunk. They are allocated on
 * demand, and not cached.
 */
struct svc_rdma_chunk_ctxt {
	struct ib_cqe		cc_cqe;
	struct svcxprt_rdma	*cc_rdma;
	struct list_head	cc_rwctxts;
	int			cc_sqecount;
};

static void svc_rdma_cc_init(struct svcxprt_rdma *rdma,
			     struct svc_rdma_chunk_ctxt *cc)
{
	cc->cc_rdma = rdma;
	svc_xprt_get(&rdma->sc_xprt);

	INIT_LIST_HEAD(&cc->cc_rwctxts);
	cc->cc_sqecount = 0;
}

static void svc_rdma_cc_release(struct svc_rdma_chunk_ctxt *cc,
				enum dma_data_direction dir)
{
	struct svcxprt_rdma *rdma = cc->cc_rdma;
	struct svc_rdma_rw_ctxt *ctxt;

	while ((ctxt = svc_rdma_next_ctxt(&cc->cc_rwctxts)) != NULL) {
		list_del(&ctxt->rw_list);

		rdma_rw_ctx_destroy(&ctxt->rw_ctx, rdma->sc_qp,
				    rdma->sc_port_num, ctxt->rw_sg_table.sgl,
				    ctxt->rw_nents, dir);
		svc_rdma_put_rw_ctxt(rdma, ctxt);
	}
	svc_xprt_put(&rdma->sc_xprt);
}

/* State for sending a Write or Reply chunk.
 *  - Tracks progress of writing one chunk over all its segments
 *  - Stores arguments for the SGL constructor functions
 */
struct svc_rdma_write_info {
	/* write state of this chunk */
	unsigned int		wi_seg_off;
	unsigned int		wi_seg_no;
	unsigned int		wi_nsegs;
	__be32			*wi_segs;

	/* SGL constructor arguments */
	struct xdr_buf		*wi_xdr;
	unsigned char		*wi_base;
	unsigned int		wi_next_off;

	struct svc_rdma_chunk_ctxt	wi_cc;
};

static struct svc_rdma_write_info *
svc_rdma_write_info_alloc(struct svcxprt_rdma *rdma, __be32 *chunk)
{
	struct svc_rdma_write_info *info;

	info = kmalloc(sizeof(*info), GFP_KERNEL);
	if (!info)
		return info;

	info->wi_seg_off = 0;
	info->wi_seg_no = 0;
	info->wi_nsegs = be32_to_cpup(++chunk);
	info->wi_segs = ++chunk;
	svc_rdma_cc_init(rdma, &info->wi_cc);
	info->wi_cc.cc_cqe.done = svc_rdma_write_done;
	return info;
}

static void svc_rdma_write_info_free(struct svc_rdma_write_info *info)
{
	svc_rdma_cc_release(&info->wi_cc, DMA_TO_DEVICE);
	kfree(info);
}

/**
 * svc_rdma_write_done - Write chunk completion
 * @cq: controlling Completion Queue
 * @wc: Work Completion
 *
 * Pages under I/O are freed by a subsequent Send completion.
 */
static void svc_rdma_write_done(struct ib_cq *cq, struct ib_wc *wc)
{
	struct ib_cqe *cqe = wc->wr_cqe;
	struct svc_rdma_chunk_ctxt *cc =
			container_of(cqe, struct svc_rdma_chunk_ctxt, cc_cqe);
	struct svcxprt_rdma *rdma = cc->cc_rdma;
	struct svc_rdma_write_info *info =
			container_of(cc, struct svc_rdma_write_info, wi_cc);

	trace_svcrdma_wc_write(wc);

	atomic_add(cc->cc_sqecount, &rdma->sc_sq_avail);
	wake_up(&rdma->sc_send_wait);

	if (unlikely(wc->status != IB_WC_SUCCESS))
		set_bit(XPT_CLOSE, &rdma->sc_xprt.xpt_flags);

	svc_rdma_write_info_free(info);
}

/* State for pulling a Read chunk.
 */
struct svc_rdma_read_info {
	struct svc_rdma_recv_ctxt	*ri_readctxt;
	unsigned int			ri_position;
	unsigned int			ri_pageno;
	unsigned int			ri_pageoff;
	unsigned int			ri_chunklen;

	struct svc_rdma_chunk_ctxt	ri_cc;
};

static struct svc_rdma_read_info *
svc_rdma_read_info_alloc(struct svcxprt_rdma *rdma)
{
	struct svc_rdma_read_info *info;

	info = kmalloc(sizeof(*info), GFP_KERNEL);
	if (!info)
		return info;

	svc_rdma_cc_init(rdma, &info->ri_cc);
	info->ri_cc.cc_cqe.done = svc_rdma_wc_read_done;
	return info;
}

static void svc_rdma_read_info_free(struct svc_rdma_read_info *info)
{
	svc_rdma_cc_release(&info->ri_cc, DMA_FROM_DEVICE);
	kfree(info);
}

/**
 * svc_rdma_wc_read_done - Handle completion of an RDMA Read ctx
 * @cq: controlling Completion Queue
 * @wc: Work Completion
 *
 */
static void svc_rdma_wc_read_done(struct ib_cq *cq, struct ib_wc *wc)
{
	struct ib_cqe *cqe = wc->wr_cqe;
	struct svc_rdma_chunk_ctxt *cc =
			container_of(cqe, struct svc_rdma_chunk_ctxt, cc_cqe);
	struct svcxprt_rdma *rdma = cc->cc_rdma;
	struct svc_rdma_read_info *info =
			container_of(cc, struct svc_rdma_read_info, ri_cc);

	trace_svcrdma_wc_read(wc);

	atomic_add(cc->cc_sqecount, &rdma->sc_sq_avail);
	wake_up(&rdma->sc_send_wait);

	if (unlikely(wc->status != IB_WC_SUCCESS)) {
		set_bit(XPT_CLOSE, &rdma->sc_xprt.xpt_flags);
		svc_rdma_recv_ctxt_put(rdma, info->ri_readctxt);
	} else {
		spin_lock(&rdma->sc_rq_dto_lock);
		list_add_tail(&info->ri_readctxt->rc_list,
			      &rdma->sc_read_complete_q);
		/* Note the unlock pairs with the smp_rmb in svc_xprt_ready: */
		set_bit(XPT_DATA, &rdma->sc_xprt.xpt_flags);
		spin_unlock(&rdma->sc_rq_dto_lock);

		svc_xprt_enqueue(&rdma->sc_xprt);
	}

	svc_rdma_read_info_free(info);
}

/* This function sleeps when the transport's Send Queue is congested.
 *
 * Assumptions:
 * - If ib_post_send() succeeds, only one completion is expected,
 *   even if one or more WRs are flushed. This is true when posting
 *   an rdma_rw_ctx or when posting a single signaled WR.
 */
static int svc_rdma_post_chunk_ctxt(struct svc_rdma_chunk_ctxt *cc)
{
	struct svcxprt_rdma *rdma = cc->cc_rdma;
	struct svc_xprt *xprt = &rdma->sc_xprt;
	struct ib_send_wr *first_wr;
	const struct ib_send_wr *bad_wr;
	struct list_head *tmp;
	struct ib_cqe *cqe;
	int ret;

	if (cc->cc_sqecount > rdma->sc_sq_depth)
		return -EINVAL;

	first_wr = NULL;
	cqe = &cc->cc_cqe;
	list_for_each(tmp, &cc->cc_rwctxts) {
		struct svc_rdma_rw_ctxt *ctxt;

		ctxt = list_entry(tmp, struct svc_rdma_rw_ctxt, rw_list);
		first_wr = rdma_rw_ctx_wrs(&ctxt->rw_ctx, rdma->sc_qp,
					   rdma->sc_port_num, cqe, first_wr);
		cqe = NULL;
	}

	do {
		if (atomic_sub_return(cc->cc_sqecount,
				      &rdma->sc_sq_avail) > 0) {
			ret = ib_post_send(rdma->sc_qp, first_wr, &bad_wr);
			trace_svcrdma_post_rw(&cc->cc_cqe,
					      cc->cc_sqecount, ret);
			if (ret)
				break;
			return 0;
		}

		trace_svcrdma_sq_full(rdma);
		atomic_add(cc->cc_sqecount, &rdma->sc_sq_avail);
		wait_event(rdma->sc_send_wait,
			   atomic_read(&rdma->sc_sq_avail) > cc->cc_sqecount);
		trace_svcrdma_sq_retry(rdma);
	} while (1);

	set_bit(XPT_CLOSE, &xprt->xpt_flags);

	/* If even one was posted, there will be a completion. */
	if (bad_wr != first_wr)
		return 0;

	atomic_add(cc->cc_sqecount, &rdma->sc_sq_avail);
	wake_up(&rdma->sc_send_wait);
	return -ENOTCONN;
}

/* Build and DMA-map an SGL that covers one kvec in an xdr_buf
 */
static void svc_rdma_vec_to_sg(struct svc_rdma_write_info *info,
			       unsigned int len,
			       struct svc_rdma_rw_ctxt *ctxt)
{
	struct scatterlist *sg = ctxt->rw_sg_table.sgl;

	sg_set_buf(&sg[0], info->wi_base, len);
	info->wi_base += len;

	ctxt->rw_nents = 1;
}

/* Build and DMA-map an SGL that covers part of an xdr_buf's pagelist.
 */
static void svc_rdma_pagelist_to_sg(struct svc_rdma_write_info *info,
				    unsigned int remaining,
				    struct svc_rdma_rw_ctxt *ctxt)
{
	unsigned int sge_no, sge_bytes, page_off, page_no;
	struct xdr_buf *xdr = info->wi_xdr;
	struct scatterlist *sg;
	struct page **page;

	page_off = info->wi_next_off + xdr->page_base;
	page_no = page_off >> PAGE_SHIFT;
	page_off = offset_in_page(page_off);
	page = xdr->pages + page_no;
	info->wi_next_off += remaining;
	sg = ctxt->rw_sg_table.sgl;
	sge_no = 0;
	do {
		sge_bytes = min_t(unsigned int, remaining,
				  PAGE_SIZE - page_off);
		sg_set_page(sg, *page, sge_bytes, page_off);

		remaining -= sge_bytes;
		sg = sg_next(sg);
		page_off = 0;
		sge_no++;
		page++;
	} while (remaining);

	ctxt->rw_nents = sge_no;
}

/* Construct RDMA Write WRs to send a portion of an xdr_buf containing
 * an RPC Reply.
 */
static int
svc_rdma_build_writes(struct svc_rdma_write_info *info,
		      void (*constructor)(struct svc_rdma_write_info *info,
					  unsigned int len,
					  struct svc_rdma_rw_ctxt *ctxt),
		      unsigned int remaining)
{
	struct svc_rdma_chunk_ctxt *cc = &info->wi_cc;
	struct svcxprt_rdma *rdma = cc->cc_rdma;
	struct svc_rdma_rw_ctxt *ctxt;
	__be32 *seg;
	int ret;

	seg = info->wi_segs + info->wi_seg_no * rpcrdma_segment_maxsz;
	do {
		unsigned int write_len;
		u32 seg_length, seg_handle;
		u64 seg_offset;

		if (info->wi_seg_no >= info->wi_nsegs)
			goto out_overflow;

		seg_handle = be32_to_cpup(seg);
		seg_length = be32_to_cpup(seg + 1);
		xdr_decode_hyper(seg + 2, &seg_offset);
		seg_offset += info->wi_seg_off;

		write_len = min(remaining, seg_length - info->wi_seg_off);
		ctxt = svc_rdma_get_rw_ctxt(rdma,
					    (write_len >> PAGE_SHIFT) + 2);
		if (!ctxt)
			goto out_noctx;

		constructor(info, write_len, ctxt);
		ret = rdma_rw_ctx_init(&ctxt->rw_ctx, rdma->sc_qp,
				       rdma->sc_port_num, ctxt->rw_sg_table.sgl,
				       ctxt->rw_nents, 0, seg_offset,
				       seg_handle, DMA_TO_DEVICE);
		if (ret < 0)
			goto out_initerr;

		trace_svcrdma_encode_wseg(seg_handle, write_len, seg_offset);
		list_add(&ctxt->rw_list, &cc->cc_rwctxts);
		cc->cc_sqecount += ret;
		if (write_len == seg_length - info->wi_seg_off) {
			seg += 4;
			info->wi_seg_no++;
			info->wi_seg_off = 0;
		} else {
			info->wi_seg_off += write_len;
		}
		remaining -= write_len;
	} while (remaining);

	return 0;

out_overflow:
	dprintk("svcrdma: inadequate space in Write chunk (%u)\n",
		info->wi_nsegs);
	return -E2BIG;

out_noctx:
	dprintk("svcrdma: no R/W ctxs available\n");
	return -ENOMEM;

out_initerr:
	svc_rdma_put_rw_ctxt(rdma, ctxt);
	trace_svcrdma_dma_map_rwctx(rdma, ret);
	return -EIO;
}

/* Send one of an xdr_buf's kvecs by itself. To send a Reply
 * chunk, the whole RPC Reply is written back to the client.
 * This function writes either the head or tail of the xdr_buf
 * containing the Reply.
 */
static int svc_rdma_send_xdr_kvec(struct svc_rdma_write_info *info,
				  struct kvec *vec)
{
	info->wi_base = vec->iov_base;
	return svc_rdma_build_writes(info, svc_rdma_vec_to_sg,
				     vec->iov_len);
}

/* Send an xdr_buf's page list by itself. A Write chunk is
 * just the page list. a Reply chunk is the head, page list,
 * and tail. This function is shared between the two types
 * of chunk.
 */
static int svc_rdma_send_xdr_pagelist(struct svc_rdma_write_info *info,
				      struct xdr_buf *xdr)
{
	info->wi_xdr = xdr;
	info->wi_next_off = 0;
	return svc_rdma_build_writes(info, svc_rdma_pagelist_to_sg,
				     xdr->page_len);
}

/**
 * svc_rdma_send_write_chunk - Write all segments in a Write chunk
 * @rdma: controlling RDMA transport
 * @wr_ch: Write chunk provided by client
 * @xdr: xdr_buf containing the data payload
 *
 * Returns a non-negative number of bytes the chunk consumed, or
 *	%-E2BIG if the payload was larger than the Write chunk,
 *	%-EINVAL if client provided too many segments,
 *	%-ENOMEM if rdma_rw context pool was exhausted,
 *	%-ENOTCONN if posting failed (connection is lost),
 *	%-EIO if rdma_rw initialization failed (DMA mapping, etc).
 */
int svc_rdma_send_write_chunk(struct svcxprt_rdma *rdma, __be32 *wr_ch,
			      struct xdr_buf *xdr)
{
	struct svc_rdma_write_info *info;
	int ret;

	if (!xdr->page_len)
		return 0;

	info = svc_rdma_write_info_alloc(rdma, wr_ch);
	if (!info)
		return -ENOMEM;

	ret = svc_rdma_send_xdr_pagelist(info, xdr);
	if (ret < 0)
		goto out_err;

	ret = svc_rdma_post_chunk_ctxt(&info->wi_cc);
	if (ret < 0)
		goto out_err;

	trace_svcrdma_encode_write(xdr->page_len);
	return xdr->page_len;

out_err:
	svc_rdma_write_info_free(info);
	return ret;
}

/**
 * svc_rdma_send_reply_chunk - Write all segments in the Reply chunk
 * @rdma: controlling RDMA transport
 * @rp_ch: Reply chunk provided by client
 * @writelist: true if client provided a Write list
 * @xdr: xdr_buf containing an RPC Reply
 *
 * Returns a non-negative number of bytes the chunk consumed, or
 *	%-E2BIG if the payload was larger than the Reply chunk,
 *	%-EINVAL if client provided too many segments,
 *	%-ENOMEM if rdma_rw context pool was exhausted,
 *	%-ENOTCONN if posting failed (connection is lost),
 *	%-EIO if rdma_rw initialization failed (DMA mapping, etc).
 */
int svc_rdma_send_reply_chunk(struct svcxprt_rdma *rdma, __be32 *rp_ch,
			      bool writelist, struct xdr_buf *xdr)
{
	struct svc_rdma_write_info *info;
	int consumed, ret;

	info = svc_rdma_write_info_alloc(rdma, rp_ch);
	if (!info)
		return -ENOMEM;

	ret = svc_rdma_send_xdr_kvec(info, &xdr->head[0]);
	if (ret < 0)
		goto out_err;
	consumed = xdr->head[0].iov_len;

	/* Send the page list in the Reply chunk only if the
	 * client did not provide Write chunks.
	 */
	if (!writelist && xdr->page_len) {
		ret = svc_rdma_send_xdr_pagelist(info, xdr);
		if (ret < 0)
			goto out_err;
		consumed += xdr->page_len;
	}

	if (xdr->tail[0].iov_len) {
		ret = svc_rdma_send_xdr_kvec(info, &xdr->tail[0]);
		if (ret < 0)
			goto out_err;
		consumed += xdr->tail[0].iov_len;
	}

	ret = svc_rdma_post_chunk_ctxt(&info->wi_cc);
	if (ret < 0)
		goto out_err;

	trace_svcrdma_encode_reply(consumed);
	return consumed;

out_err:
	svc_rdma_write_info_free(info);
	return ret;
}

static int svc_rdma_build_read_segment(struct svc_rdma_read_info *info,
				       struct svc_rqst *rqstp,
				       u32 rkey, u32 len, u64 offset)
{
	struct svc_rdma_recv_ctxt *head = info->ri_readctxt;
	struct svc_rdma_chunk_ctxt *cc = &info->ri_cc;
	struct svc_rdma_rw_ctxt *ctxt;
	unsigned int sge_no, seg_len;
	struct scatterlist *sg;
	int ret;

	sge_no = PAGE_ALIGN(info->ri_pageoff + len) >> PAGE_SHIFT;
	ctxt = svc_rdma_get_rw_ctxt(cc->cc_rdma, sge_no);
	if (!ctxt)
		goto out_noctx;
	ctxt->rw_nents = sge_no;

	sg = ctxt->rw_sg_table.sgl;
	for (sge_no = 0; sge_no < ctxt->rw_nents; sge_no++) {
		seg_len = min_t(unsigned int, len,
				PAGE_SIZE - info->ri_pageoff);

		head->rc_arg.pages[info->ri_pageno] =
			rqstp->rq_pages[info->ri_pageno];
		if (!info->ri_pageoff)
			head->rc_page_count++;

		sg_set_page(sg, rqstp->rq_pages[info->ri_pageno],
			    seg_len, info->ri_pageoff);
		sg = sg_next(sg);

		info->ri_pageoff += seg_len;
		if (info->ri_pageoff == PAGE_SIZE) {
			info->ri_pageno++;
			info->ri_pageoff = 0;
		}
		len -= seg_len;

		/* Safety check */
		if (len &&
		    &rqstp->rq_pages[info->ri_pageno + 1] > rqstp->rq_page_end)
			goto out_overrun;
	}

	ret = rdma_rw_ctx_init(&ctxt->rw_ctx, cc->cc_rdma->sc_qp,
			       cc->cc_rdma->sc_port_num,
			       ctxt->rw_sg_table.sgl, ctxt->rw_nents,
			       0, offset, rkey, DMA_FROM_DEVICE);
	if (ret < 0)
		goto out_initerr;

	list_add(&ctxt->rw_list, &cc->cc_rwctxts);
	cc->cc_sqecount += ret;
	return 0;

out_noctx:
	dprintk("svcrdma: no R/W ctxs available\n");
	return -ENOMEM;

out_overrun:
	dprintk("svcrdma: request overruns rq_pages\n");
	return -EINVAL;

out_initerr:
	trace_svcrdma_dma_map_rwctx(cc->cc_rdma, ret);
	svc_rdma_put_rw_ctxt(cc->cc_rdma, ctxt);
	return -EIO;
}

/* Walk the segments in the Read chunk starting at @p and construct
 * RDMA Read operations to pull the chunk to the server.
 */
static int svc_rdma_build_read_chunk(struct svc_rqst *rqstp,
				     struct svc_rdma_read_info *info,
				     __be32 *p)
{
	unsigned int i;
	int ret;

	ret = -EINVAL;
	info->ri_chunklen = 0;
	while (*p++ != xdr_zero && be32_to_cpup(p++) == info->ri_position) {
		u32 rs_handle, rs_length;
		u64 rs_offset;

		rs_handle = be32_to_cpup(p++);
		rs_length = be32_to_cpup(p++);
		p = xdr_decode_hyper(p, &rs_offset);

		ret = svc_rdma_build_read_segment(info, rqstp,
						  rs_handle, rs_length,
						  rs_offset);
		if (ret < 0)
			break;

		trace_svcrdma_encode_rseg(rs_handle, rs_length, rs_offset);
		info->ri_chunklen += rs_length;
	}

	/* Pages under I/O have been copied to head->rc_pages.
	 * Prevent their premature release by svc_xprt_release() .
	 */
	for (i = 0; i < info->ri_readctxt->rc_page_count; i++)
		rqstp->rq_pages[i] = NULL;

	return ret;
}

/* Construct RDMA Reads to pull over a normal Read chunk. The chunk
 * data lands in the page list of head->rc_arg.pages.
 *
 * Currently NFSD does not look at the head->rc_arg.tail[0] iovec.
 * Therefore, XDR round-up of the Read chunk and trailing
 * inline content must both be added at the end of the pagelist.
 */
static int svc_rdma_build_normal_read_chunk(struct svc_rqst *rqstp,
					    struct svc_rdma_read_info *info,
					    __be32 *p)
{
	struct svc_rdma_recv_ctxt *head = info->ri_readctxt;
	int ret;

	ret = svc_rdma_build_read_chunk(rqstp, info, p);
	if (ret < 0)
		goto out;

	trace_svcrdma_encode_read(info->ri_chunklen, info->ri_position);

	head->rc_hdr_count = 0;

	/* Split the Receive buffer between the head and tail
	 * buffers at Read chunk's position. XDR roundup of the
	 * chunk is not included in either the pagelist or in
	 * the tail.
	 */
	head->rc_arg.tail[0].iov_base =
		head->rc_arg.head[0].iov_base + info->ri_position;
	head->rc_arg.tail[0].iov_len =
		head->rc_arg.head[0].iov_len - info->ri_position;
	head->rc_arg.head[0].iov_len = info->ri_position;

	/* Read chunk may need XDR roundup (see RFC 8166, s. 3.4.5.2).
	 *
	 * If the client already rounded up the chunk length, the
	 * length does not change. Otherwise, the length of the page
	 * list is increased to include XDR round-up.
	 *
	 * Currently these chunks always start at page offset 0,
	 * thus the rounded-up length never crosses a page boundary.
	 */
	info->ri_chunklen = XDR_QUADLEN(info->ri_chunklen) << 2;

	head->rc_arg.page_len = info->ri_chunklen;
	head->rc_arg.len += info->ri_chunklen;
	head->rc_arg.buflen += info->ri_chunklen;

out:
	return ret;
}

/* Construct RDMA Reads to pull over a Position Zero Read chunk.
 * The start of the data lands in the first page just after
 * the Transport header, and the rest lands in the page list of
 * head->rc_arg.pages.
 *
 * Assumptions:
 *	- A PZRC has an XDR-aligned length (no implicit round-up).
 *	- There can be no trailing inline content (IOW, we assume
 *	  a PZRC is never sent in an RDMA_MSG message, though it's
 *	  allowed by spec).
 */
static int svc_rdma_build_pz_read_chunk(struct svc_rqst *rqstp,
					struct svc_rdma_read_info *info,
					__be32 *p)
{
	struct svc_rdma_recv_ctxt *head = info->ri_readctxt;
	int ret;

	ret = svc_rdma_build_read_chunk(rqstp, info, p);
	if (ret < 0)
		goto out;

	trace_svcrdma_encode_pzr(info->ri_chunklen);

	head->rc_arg.len += info->ri_chunklen;
	head->rc_arg.buflen += info->ri_chunklen;

	head->rc_hdr_count = 1;
	head->rc_arg.head[0].iov_base = page_address(head->rc_pages[0]);
	head->rc_arg.head[0].iov_len = min_t(size_t, PAGE_SIZE,
					     info->ri_chunklen);

	head->rc_arg.page_len = info->ri_chunklen -
				head->rc_arg.head[0].iov_len;

out:
	return ret;
}

/**
 * svc_rdma_recv_read_chunk - Pull a Read chunk from the client
 * @rdma: controlling RDMA transport
 * @rqstp: set of pages to use as Read sink buffers
 * @head: pages under I/O collect here
 * @p: pointer to start of Read chunk
 *
 * Returns:
 *	%0 if all needed RDMA Reads were posted successfully,
 *	%-EINVAL if client provided too many segments,
 *	%-ENOMEM if rdma_rw context pool was exhausted,
 *	%-ENOTCONN if posting failed (connection is lost),
 *	%-EIO if rdma_rw initialization failed (DMA mapping, etc).
 *
 * Assumptions:
 * - All Read segments in @p have the same Position value.
 */
int svc_rdma_recv_read_chunk(struct svcxprt_rdma *rdma, struct svc_rqst *rqstp,
			     struct svc_rdma_recv_ctxt *head, __be32 *p)
{
	struct svc_rdma_read_info *info;
	int ret;

	/* The request (with page list) is constructed in
	 * head->rc_arg. Pages involved with RDMA Read I/O are
	 * transferred there.
	 */
	head->rc_arg.head[0] = rqstp->rq_arg.head[0];
	head->rc_arg.tail[0] = rqstp->rq_arg.tail[0];
	head->rc_arg.pages = head->rc_pages;
	head->rc_arg.page_base = 0;
	head->rc_arg.page_len = 0;
	head->rc_arg.len = rqstp->rq_arg.len;
	head->rc_arg.buflen = rqstp->rq_arg.buflen;

	info = svc_rdma_read_info_alloc(rdma);
	if (!info)
		return -ENOMEM;
	info->ri_readctxt = head;
	info->ri_pageno = 0;
	info->ri_pageoff = 0;

	info->ri_position = be32_to_cpup(p + 1);
	if (info->ri_position)
		ret = svc_rdma_build_normal_read_chunk(rqstp, info, p);
	else
		ret = svc_rdma_build_pz_read_chunk(rqstp, info, p);
	if (ret < 0)
		goto out_err;

	ret = svc_rdma_post_chunk_ctxt(&info->ri_cc);
	if (ret < 0)
		goto out_err;
	return 0;

out_err:
	svc_rdma_read_info_free(info);
	return ret;
}