1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (c) 2016-2018 Oracle.  All rights reserved.
4  *
5  * Use the core R/W API to move RPC-over-RDMA Read and Write chunks.
6  */
7 
8 #include <rdma/rw.h>
9 
10 #include <linux/sunrpc/rpc_rdma.h>
11 #include <linux/sunrpc/svc_rdma.h>
12 #include <linux/sunrpc/debug.h>
13 
14 #include "xprt_rdma.h"
15 #include <trace/events/rpcrdma.h>
16 
17 #define RPCDBG_FACILITY	RPCDBG_SVCXPRT
18 
19 static void svc_rdma_write_done(struct ib_cq *cq, struct ib_wc *wc);
20 static void svc_rdma_wc_read_done(struct ib_cq *cq, struct ib_wc *wc);
21 
22 /* Each R/W context contains state for one chain of RDMA Read or
23  * Write Work Requests.
24  *
25  * Each WR chain handles a single contiguous server-side buffer,
26  * because scatterlist entries after the first have to start on
27  * page alignment. xdr_buf iovecs cannot guarantee alignment.
28  *
29  * Each WR chain handles only one R_key. Each RPC-over-RDMA segment
30  * from a client may contain a unique R_key, so each WR chain moves
31  * up to one segment at a time.
32  *
33  * The scatterlist makes this data structure over 4KB in size. To
34  * make it less likely to fail, and to handle the allocation for
35  * smaller I/O requests without disabling bottom-halves, these
36  * contexts are created on demand, but cached and reused until the
37  * controlling svcxprt_rdma is destroyed.
38  */
39 struct svc_rdma_rw_ctxt {
40 	struct list_head	rw_list;
41 	struct rdma_rw_ctx	rw_ctx;
42 	int			rw_nents;
43 	struct sg_table		rw_sg_table;
44 	struct scatterlist	rw_first_sgl[0];
45 };
46 
47 static inline struct svc_rdma_rw_ctxt *
48 svc_rdma_next_ctxt(struct list_head *list)
49 {
50 	return list_first_entry_or_null(list, struct svc_rdma_rw_ctxt,
51 					rw_list);
52 }
53 
54 static struct svc_rdma_rw_ctxt *
55 svc_rdma_get_rw_ctxt(struct svcxprt_rdma *rdma, unsigned int sges)
56 {
57 	struct svc_rdma_rw_ctxt *ctxt;
58 
59 	spin_lock(&rdma->sc_rw_ctxt_lock);
60 
61 	ctxt = svc_rdma_next_ctxt(&rdma->sc_rw_ctxts);
62 	if (ctxt) {
63 		list_del(&ctxt->rw_list);
64 		spin_unlock(&rdma->sc_rw_ctxt_lock);
65 	} else {
66 		spin_unlock(&rdma->sc_rw_ctxt_lock);
67 		ctxt = kmalloc(struct_size(ctxt, rw_first_sgl, SG_CHUNK_SIZE),
68 			       GFP_KERNEL);
69 		if (!ctxt)
70 			goto out;
71 		INIT_LIST_HEAD(&ctxt->rw_list);
72 	}
73 
74 	ctxt->rw_sg_table.sgl = ctxt->rw_first_sgl;
75 	if (sg_alloc_table_chained(&ctxt->rw_sg_table, sges,
76 				   ctxt->rw_sg_table.sgl,
77 				   SG_CHUNK_SIZE)) {
78 		kfree(ctxt);
79 		ctxt = NULL;
80 	}
81 out:
82 	return ctxt;
83 }
84 
85 static void svc_rdma_put_rw_ctxt(struct svcxprt_rdma *rdma,
86 				 struct svc_rdma_rw_ctxt *ctxt)
87 {
88 	sg_free_table_chained(&ctxt->rw_sg_table, SG_CHUNK_SIZE);
89 
90 	spin_lock(&rdma->sc_rw_ctxt_lock);
91 	list_add(&ctxt->rw_list, &rdma->sc_rw_ctxts);
92 	spin_unlock(&rdma->sc_rw_ctxt_lock);
93 }
94 
95 /**
96  * svc_rdma_destroy_rw_ctxts - Free accumulated R/W contexts
97  * @rdma: transport about to be destroyed
98  *
99  */
100 void svc_rdma_destroy_rw_ctxts(struct svcxprt_rdma *rdma)
101 {
102 	struct svc_rdma_rw_ctxt *ctxt;
103 
104 	while ((ctxt = svc_rdma_next_ctxt(&rdma->sc_rw_ctxts)) != NULL) {
105 		list_del(&ctxt->rw_list);
106 		kfree(ctxt);
107 	}
108 }
109 
110 /* A chunk context tracks all I/O for moving one Read or Write
111  * chunk. This is a a set of rdma_rw's that handle data movement
112  * for all segments of one chunk.
113  *
114  * These are small, acquired with a single allocator call, and
115  * no more than one is needed per chunk. They are allocated on
116  * demand, and not cached.
117  */
118 struct svc_rdma_chunk_ctxt {
119 	struct ib_cqe		cc_cqe;
120 	struct svcxprt_rdma	*cc_rdma;
121 	struct list_head	cc_rwctxts;
122 	int			cc_sqecount;
123 };
124 
125 static void svc_rdma_cc_init(struct svcxprt_rdma *rdma,
126 			     struct svc_rdma_chunk_ctxt *cc)
127 {
128 	cc->cc_rdma = rdma;
129 	svc_xprt_get(&rdma->sc_xprt);
130 
131 	INIT_LIST_HEAD(&cc->cc_rwctxts);
132 	cc->cc_sqecount = 0;
133 }
134 
135 static void svc_rdma_cc_release(struct svc_rdma_chunk_ctxt *cc,
136 				enum dma_data_direction dir)
137 {
138 	struct svcxprt_rdma *rdma = cc->cc_rdma;
139 	struct svc_rdma_rw_ctxt *ctxt;
140 
141 	while ((ctxt = svc_rdma_next_ctxt(&cc->cc_rwctxts)) != NULL) {
142 		list_del(&ctxt->rw_list);
143 
144 		rdma_rw_ctx_destroy(&ctxt->rw_ctx, rdma->sc_qp,
145 				    rdma->sc_port_num, ctxt->rw_sg_table.sgl,
146 				    ctxt->rw_nents, dir);
147 		svc_rdma_put_rw_ctxt(rdma, ctxt);
148 	}
149 	svc_xprt_put(&rdma->sc_xprt);
150 }
151 
152 /* State for sending a Write or Reply chunk.
153  *  - Tracks progress of writing one chunk over all its segments
154  *  - Stores arguments for the SGL constructor functions
155  */
156 struct svc_rdma_write_info {
157 	/* write state of this chunk */
158 	unsigned int		wi_seg_off;
159 	unsigned int		wi_seg_no;
160 	unsigned int		wi_nsegs;
161 	__be32			*wi_segs;
162 
163 	/* SGL constructor arguments */
164 	struct xdr_buf		*wi_xdr;
165 	unsigned char		*wi_base;
166 	unsigned int		wi_next_off;
167 
168 	struct svc_rdma_chunk_ctxt	wi_cc;
169 };
170 
171 static struct svc_rdma_write_info *
172 svc_rdma_write_info_alloc(struct svcxprt_rdma *rdma, __be32 *chunk)
173 {
174 	struct svc_rdma_write_info *info;
175 
176 	info = kmalloc(sizeof(*info), GFP_KERNEL);
177 	if (!info)
178 		return info;
179 
180 	info->wi_seg_off = 0;
181 	info->wi_seg_no = 0;
182 	info->wi_nsegs = be32_to_cpup(++chunk);
183 	info->wi_segs = ++chunk;
184 	svc_rdma_cc_init(rdma, &info->wi_cc);
185 	info->wi_cc.cc_cqe.done = svc_rdma_write_done;
186 	return info;
187 }
188 
189 static void svc_rdma_write_info_free(struct svc_rdma_write_info *info)
190 {
191 	svc_rdma_cc_release(&info->wi_cc, DMA_TO_DEVICE);
192 	kfree(info);
193 }
194 
195 /**
196  * svc_rdma_write_done - Write chunk completion
197  * @cq: controlling Completion Queue
198  * @wc: Work Completion
199  *
200  * Pages under I/O are freed by a subsequent Send completion.
201  */
202 static void svc_rdma_write_done(struct ib_cq *cq, struct ib_wc *wc)
203 {
204 	struct ib_cqe *cqe = wc->wr_cqe;
205 	struct svc_rdma_chunk_ctxt *cc =
206 			container_of(cqe, struct svc_rdma_chunk_ctxt, cc_cqe);
207 	struct svcxprt_rdma *rdma = cc->cc_rdma;
208 	struct svc_rdma_write_info *info =
209 			container_of(cc, struct svc_rdma_write_info, wi_cc);
210 
211 	trace_svcrdma_wc_write(wc);
212 
213 	atomic_add(cc->cc_sqecount, &rdma->sc_sq_avail);
214 	wake_up(&rdma->sc_send_wait);
215 
216 	if (unlikely(wc->status != IB_WC_SUCCESS))
217 		set_bit(XPT_CLOSE, &rdma->sc_xprt.xpt_flags);
218 
219 	svc_rdma_write_info_free(info);
220 }
221 
222 /* State for pulling a Read chunk.
223  */
224 struct svc_rdma_read_info {
225 	struct svc_rdma_recv_ctxt	*ri_readctxt;
226 	unsigned int			ri_position;
227 	unsigned int			ri_pageno;
228 	unsigned int			ri_pageoff;
229 	unsigned int			ri_chunklen;
230 
231 	struct svc_rdma_chunk_ctxt	ri_cc;
232 };
233 
234 static struct svc_rdma_read_info *
235 svc_rdma_read_info_alloc(struct svcxprt_rdma *rdma)
236 {
237 	struct svc_rdma_read_info *info;
238 
239 	info = kmalloc(sizeof(*info), GFP_KERNEL);
240 	if (!info)
241 		return info;
242 
243 	svc_rdma_cc_init(rdma, &info->ri_cc);
244 	info->ri_cc.cc_cqe.done = svc_rdma_wc_read_done;
245 	return info;
246 }
247 
248 static void svc_rdma_read_info_free(struct svc_rdma_read_info *info)
249 {
250 	svc_rdma_cc_release(&info->ri_cc, DMA_FROM_DEVICE);
251 	kfree(info);
252 }
253 
254 /**
255  * svc_rdma_wc_read_done - Handle completion of an RDMA Read ctx
256  * @cq: controlling Completion Queue
257  * @wc: Work Completion
258  *
259  */
260 static void svc_rdma_wc_read_done(struct ib_cq *cq, struct ib_wc *wc)
261 {
262 	struct ib_cqe *cqe = wc->wr_cqe;
263 	struct svc_rdma_chunk_ctxt *cc =
264 			container_of(cqe, struct svc_rdma_chunk_ctxt, cc_cqe);
265 	struct svcxprt_rdma *rdma = cc->cc_rdma;
266 	struct svc_rdma_read_info *info =
267 			container_of(cc, struct svc_rdma_read_info, ri_cc);
268 
269 	trace_svcrdma_wc_read(wc);
270 
271 	atomic_add(cc->cc_sqecount, &rdma->sc_sq_avail);
272 	wake_up(&rdma->sc_send_wait);
273 
274 	if (unlikely(wc->status != IB_WC_SUCCESS)) {
275 		set_bit(XPT_CLOSE, &rdma->sc_xprt.xpt_flags);
276 		svc_rdma_recv_ctxt_put(rdma, info->ri_readctxt);
277 	} else {
278 		spin_lock(&rdma->sc_rq_dto_lock);
279 		list_add_tail(&info->ri_readctxt->rc_list,
280 			      &rdma->sc_read_complete_q);
281 		/* Note the unlock pairs with the smp_rmb in svc_xprt_ready: */
282 		set_bit(XPT_DATA, &rdma->sc_xprt.xpt_flags);
283 		spin_unlock(&rdma->sc_rq_dto_lock);
284 
285 		svc_xprt_enqueue(&rdma->sc_xprt);
286 	}
287 
288 	svc_rdma_read_info_free(info);
289 }
290 
291 /* This function sleeps when the transport's Send Queue is congested.
292  *
293  * Assumptions:
294  * - If ib_post_send() succeeds, only one completion is expected,
295  *   even if one or more WRs are flushed. This is true when posting
296  *   an rdma_rw_ctx or when posting a single signaled WR.
297  */
298 static int svc_rdma_post_chunk_ctxt(struct svc_rdma_chunk_ctxt *cc)
299 {
300 	struct svcxprt_rdma *rdma = cc->cc_rdma;
301 	struct svc_xprt *xprt = &rdma->sc_xprt;
302 	struct ib_send_wr *first_wr;
303 	const struct ib_send_wr *bad_wr;
304 	struct list_head *tmp;
305 	struct ib_cqe *cqe;
306 	int ret;
307 
308 	if (cc->cc_sqecount > rdma->sc_sq_depth)
309 		return -EINVAL;
310 
311 	first_wr = NULL;
312 	cqe = &cc->cc_cqe;
313 	list_for_each(tmp, &cc->cc_rwctxts) {
314 		struct svc_rdma_rw_ctxt *ctxt;
315 
316 		ctxt = list_entry(tmp, struct svc_rdma_rw_ctxt, rw_list);
317 		first_wr = rdma_rw_ctx_wrs(&ctxt->rw_ctx, rdma->sc_qp,
318 					   rdma->sc_port_num, cqe, first_wr);
319 		cqe = NULL;
320 	}
321 
322 	do {
323 		if (atomic_sub_return(cc->cc_sqecount,
324 				      &rdma->sc_sq_avail) > 0) {
325 			ret = ib_post_send(rdma->sc_qp, first_wr, &bad_wr);
326 			trace_svcrdma_post_rw(&cc->cc_cqe,
327 					      cc->cc_sqecount, ret);
328 			if (ret)
329 				break;
330 			return 0;
331 		}
332 
333 		trace_svcrdma_sq_full(rdma);
334 		atomic_add(cc->cc_sqecount, &rdma->sc_sq_avail);
335 		wait_event(rdma->sc_send_wait,
336 			   atomic_read(&rdma->sc_sq_avail) > cc->cc_sqecount);
337 		trace_svcrdma_sq_retry(rdma);
338 	} while (1);
339 
340 	set_bit(XPT_CLOSE, &xprt->xpt_flags);
341 
342 	/* If even one was posted, there will be a completion. */
343 	if (bad_wr != first_wr)
344 		return 0;
345 
346 	atomic_add(cc->cc_sqecount, &rdma->sc_sq_avail);
347 	wake_up(&rdma->sc_send_wait);
348 	return -ENOTCONN;
349 }
350 
351 /* Build and DMA-map an SGL that covers one kvec in an xdr_buf
352  */
353 static void svc_rdma_vec_to_sg(struct svc_rdma_write_info *info,
354 			       unsigned int len,
355 			       struct svc_rdma_rw_ctxt *ctxt)
356 {
357 	struct scatterlist *sg = ctxt->rw_sg_table.sgl;
358 
359 	sg_set_buf(&sg[0], info->wi_base, len);
360 	info->wi_base += len;
361 
362 	ctxt->rw_nents = 1;
363 }
364 
365 /* Build and DMA-map an SGL that covers part of an xdr_buf's pagelist.
366  */
367 static void svc_rdma_pagelist_to_sg(struct svc_rdma_write_info *info,
368 				    unsigned int remaining,
369 				    struct svc_rdma_rw_ctxt *ctxt)
370 {
371 	unsigned int sge_no, sge_bytes, page_off, page_no;
372 	struct xdr_buf *xdr = info->wi_xdr;
373 	struct scatterlist *sg;
374 	struct page **page;
375 
376 	page_off = info->wi_next_off + xdr->page_base;
377 	page_no = page_off >> PAGE_SHIFT;
378 	page_off = offset_in_page(page_off);
379 	page = xdr->pages + page_no;
380 	info->wi_next_off += remaining;
381 	sg = ctxt->rw_sg_table.sgl;
382 	sge_no = 0;
383 	do {
384 		sge_bytes = min_t(unsigned int, remaining,
385 				  PAGE_SIZE - page_off);
386 		sg_set_page(sg, *page, sge_bytes, page_off);
387 
388 		remaining -= sge_bytes;
389 		sg = sg_next(sg);
390 		page_off = 0;
391 		sge_no++;
392 		page++;
393 	} while (remaining);
394 
395 	ctxt->rw_nents = sge_no;
396 }
397 
398 /* Construct RDMA Write WRs to send a portion of an xdr_buf containing
399  * an RPC Reply.
400  */
401 static int
402 svc_rdma_build_writes(struct svc_rdma_write_info *info,
403 		      void (*constructor)(struct svc_rdma_write_info *info,
404 					  unsigned int len,
405 					  struct svc_rdma_rw_ctxt *ctxt),
406 		      unsigned int remaining)
407 {
408 	struct svc_rdma_chunk_ctxt *cc = &info->wi_cc;
409 	struct svcxprt_rdma *rdma = cc->cc_rdma;
410 	struct svc_rdma_rw_ctxt *ctxt;
411 	__be32 *seg;
412 	int ret;
413 
414 	seg = info->wi_segs + info->wi_seg_no * rpcrdma_segment_maxsz;
415 	do {
416 		unsigned int write_len;
417 		u32 seg_length, seg_handle;
418 		u64 seg_offset;
419 
420 		if (info->wi_seg_no >= info->wi_nsegs)
421 			goto out_overflow;
422 
423 		seg_handle = be32_to_cpup(seg);
424 		seg_length = be32_to_cpup(seg + 1);
425 		xdr_decode_hyper(seg + 2, &seg_offset);
426 		seg_offset += info->wi_seg_off;
427 
428 		write_len = min(remaining, seg_length - info->wi_seg_off);
429 		ctxt = svc_rdma_get_rw_ctxt(rdma,
430 					    (write_len >> PAGE_SHIFT) + 2);
431 		if (!ctxt)
432 			goto out_noctx;
433 
434 		constructor(info, write_len, ctxt);
435 		ret = rdma_rw_ctx_init(&ctxt->rw_ctx, rdma->sc_qp,
436 				       rdma->sc_port_num, ctxt->rw_sg_table.sgl,
437 				       ctxt->rw_nents, 0, seg_offset,
438 				       seg_handle, DMA_TO_DEVICE);
439 		if (ret < 0)
440 			goto out_initerr;
441 
442 		trace_svcrdma_encode_wseg(seg_handle, write_len, seg_offset);
443 		list_add(&ctxt->rw_list, &cc->cc_rwctxts);
444 		cc->cc_sqecount += ret;
445 		if (write_len == seg_length - info->wi_seg_off) {
446 			seg += 4;
447 			info->wi_seg_no++;
448 			info->wi_seg_off = 0;
449 		} else {
450 			info->wi_seg_off += write_len;
451 		}
452 		remaining -= write_len;
453 	} while (remaining);
454 
455 	return 0;
456 
457 out_overflow:
458 	dprintk("svcrdma: inadequate space in Write chunk (%u)\n",
459 		info->wi_nsegs);
460 	return -E2BIG;
461 
462 out_noctx:
463 	dprintk("svcrdma: no R/W ctxs available\n");
464 	return -ENOMEM;
465 
466 out_initerr:
467 	svc_rdma_put_rw_ctxt(rdma, ctxt);
468 	trace_svcrdma_dma_map_rwctx(rdma, ret);
469 	return -EIO;
470 }
471 
472 /* Send one of an xdr_buf's kvecs by itself. To send a Reply
473  * chunk, the whole RPC Reply is written back to the client.
474  * This function writes either the head or tail of the xdr_buf
475  * containing the Reply.
476  */
477 static int svc_rdma_send_xdr_kvec(struct svc_rdma_write_info *info,
478 				  struct kvec *vec)
479 {
480 	info->wi_base = vec->iov_base;
481 	return svc_rdma_build_writes(info, svc_rdma_vec_to_sg,
482 				     vec->iov_len);
483 }
484 
485 /* Send an xdr_buf's page list by itself. A Write chunk is
486  * just the page list. a Reply chunk is the head, page list,
487  * and tail. This function is shared between the two types
488  * of chunk.
489  */
490 static int svc_rdma_send_xdr_pagelist(struct svc_rdma_write_info *info,
491 				      struct xdr_buf *xdr)
492 {
493 	info->wi_xdr = xdr;
494 	info->wi_next_off = 0;
495 	return svc_rdma_build_writes(info, svc_rdma_pagelist_to_sg,
496 				     xdr->page_len);
497 }
498 
499 /**
500  * svc_rdma_send_write_chunk - Write all segments in a Write chunk
501  * @rdma: controlling RDMA transport
502  * @wr_ch: Write chunk provided by client
503  * @xdr: xdr_buf containing the data payload
504  *
505  * Returns a non-negative number of bytes the chunk consumed, or
506  *	%-E2BIG if the payload was larger than the Write chunk,
507  *	%-EINVAL if client provided too many segments,
508  *	%-ENOMEM if rdma_rw context pool was exhausted,
509  *	%-ENOTCONN if posting failed (connection is lost),
510  *	%-EIO if rdma_rw initialization failed (DMA mapping, etc).
511  */
512 int svc_rdma_send_write_chunk(struct svcxprt_rdma *rdma, __be32 *wr_ch,
513 			      struct xdr_buf *xdr)
514 {
515 	struct svc_rdma_write_info *info;
516 	int ret;
517 
518 	if (!xdr->page_len)
519 		return 0;
520 
521 	info = svc_rdma_write_info_alloc(rdma, wr_ch);
522 	if (!info)
523 		return -ENOMEM;
524 
525 	ret = svc_rdma_send_xdr_pagelist(info, xdr);
526 	if (ret < 0)
527 		goto out_err;
528 
529 	ret = svc_rdma_post_chunk_ctxt(&info->wi_cc);
530 	if (ret < 0)
531 		goto out_err;
532 
533 	trace_svcrdma_encode_write(xdr->page_len);
534 	return xdr->page_len;
535 
536 out_err:
537 	svc_rdma_write_info_free(info);
538 	return ret;
539 }
540 
541 /**
542  * svc_rdma_send_reply_chunk - Write all segments in the Reply chunk
543  * @rdma: controlling RDMA transport
544  * @rp_ch: Reply chunk provided by client
545  * @writelist: true if client provided a Write list
546  * @xdr: xdr_buf containing an RPC Reply
547  *
548  * Returns a non-negative number of bytes the chunk consumed, or
549  *	%-E2BIG if the payload was larger than the Reply chunk,
550  *	%-EINVAL if client provided too many segments,
551  *	%-ENOMEM if rdma_rw context pool was exhausted,
552  *	%-ENOTCONN if posting failed (connection is lost),
553  *	%-EIO if rdma_rw initialization failed (DMA mapping, etc).
554  */
555 int svc_rdma_send_reply_chunk(struct svcxprt_rdma *rdma, __be32 *rp_ch,
556 			      bool writelist, struct xdr_buf *xdr)
557 {
558 	struct svc_rdma_write_info *info;
559 	int consumed, ret;
560 
561 	info = svc_rdma_write_info_alloc(rdma, rp_ch);
562 	if (!info)
563 		return -ENOMEM;
564 
565 	ret = svc_rdma_send_xdr_kvec(info, &xdr->head[0]);
566 	if (ret < 0)
567 		goto out_err;
568 	consumed = xdr->head[0].iov_len;
569 
570 	/* Send the page list in the Reply chunk only if the
571 	 * client did not provide Write chunks.
572 	 */
573 	if (!writelist && xdr->page_len) {
574 		ret = svc_rdma_send_xdr_pagelist(info, xdr);
575 		if (ret < 0)
576 			goto out_err;
577 		consumed += xdr->page_len;
578 	}
579 
580 	if (xdr->tail[0].iov_len) {
581 		ret = svc_rdma_send_xdr_kvec(info, &xdr->tail[0]);
582 		if (ret < 0)
583 			goto out_err;
584 		consumed += xdr->tail[0].iov_len;
585 	}
586 
587 	ret = svc_rdma_post_chunk_ctxt(&info->wi_cc);
588 	if (ret < 0)
589 		goto out_err;
590 
591 	trace_svcrdma_encode_reply(consumed);
592 	return consumed;
593 
594 out_err:
595 	svc_rdma_write_info_free(info);
596 	return ret;
597 }
598 
599 static int svc_rdma_build_read_segment(struct svc_rdma_read_info *info,
600 				       struct svc_rqst *rqstp,
601 				       u32 rkey, u32 len, u64 offset)
602 {
603 	struct svc_rdma_recv_ctxt *head = info->ri_readctxt;
604 	struct svc_rdma_chunk_ctxt *cc = &info->ri_cc;
605 	struct svc_rdma_rw_ctxt *ctxt;
606 	unsigned int sge_no, seg_len;
607 	struct scatterlist *sg;
608 	int ret;
609 
610 	sge_no = PAGE_ALIGN(info->ri_pageoff + len) >> PAGE_SHIFT;
611 	ctxt = svc_rdma_get_rw_ctxt(cc->cc_rdma, sge_no);
612 	if (!ctxt)
613 		goto out_noctx;
614 	ctxt->rw_nents = sge_no;
615 
616 	sg = ctxt->rw_sg_table.sgl;
617 	for (sge_no = 0; sge_no < ctxt->rw_nents; sge_no++) {
618 		seg_len = min_t(unsigned int, len,
619 				PAGE_SIZE - info->ri_pageoff);
620 
621 		head->rc_arg.pages[info->ri_pageno] =
622 			rqstp->rq_pages[info->ri_pageno];
623 		if (!info->ri_pageoff)
624 			head->rc_page_count++;
625 
626 		sg_set_page(sg, rqstp->rq_pages[info->ri_pageno],
627 			    seg_len, info->ri_pageoff);
628 		sg = sg_next(sg);
629 
630 		info->ri_pageoff += seg_len;
631 		if (info->ri_pageoff == PAGE_SIZE) {
632 			info->ri_pageno++;
633 			info->ri_pageoff = 0;
634 		}
635 		len -= seg_len;
636 
637 		/* Safety check */
638 		if (len &&
639 		    &rqstp->rq_pages[info->ri_pageno + 1] > rqstp->rq_page_end)
640 			goto out_overrun;
641 	}
642 
643 	ret = rdma_rw_ctx_init(&ctxt->rw_ctx, cc->cc_rdma->sc_qp,
644 			       cc->cc_rdma->sc_port_num,
645 			       ctxt->rw_sg_table.sgl, ctxt->rw_nents,
646 			       0, offset, rkey, DMA_FROM_DEVICE);
647 	if (ret < 0)
648 		goto out_initerr;
649 
650 	list_add(&ctxt->rw_list, &cc->cc_rwctxts);
651 	cc->cc_sqecount += ret;
652 	return 0;
653 
654 out_noctx:
655 	dprintk("svcrdma: no R/W ctxs available\n");
656 	return -ENOMEM;
657 
658 out_overrun:
659 	dprintk("svcrdma: request overruns rq_pages\n");
660 	return -EINVAL;
661 
662 out_initerr:
663 	trace_svcrdma_dma_map_rwctx(cc->cc_rdma, ret);
664 	svc_rdma_put_rw_ctxt(cc->cc_rdma, ctxt);
665 	return -EIO;
666 }
667 
668 /* Walk the segments in the Read chunk starting at @p and construct
669  * RDMA Read operations to pull the chunk to the server.
670  */
671 static int svc_rdma_build_read_chunk(struct svc_rqst *rqstp,
672 				     struct svc_rdma_read_info *info,
673 				     __be32 *p)
674 {
675 	unsigned int i;
676 	int ret;
677 
678 	ret = -EINVAL;
679 	info->ri_chunklen = 0;
680 	while (*p++ != xdr_zero && be32_to_cpup(p++) == info->ri_position) {
681 		u32 rs_handle, rs_length;
682 		u64 rs_offset;
683 
684 		rs_handle = be32_to_cpup(p++);
685 		rs_length = be32_to_cpup(p++);
686 		p = xdr_decode_hyper(p, &rs_offset);
687 
688 		ret = svc_rdma_build_read_segment(info, rqstp,
689 						  rs_handle, rs_length,
690 						  rs_offset);
691 		if (ret < 0)
692 			break;
693 
694 		trace_svcrdma_encode_rseg(rs_handle, rs_length, rs_offset);
695 		info->ri_chunklen += rs_length;
696 	}
697 
698 	/* Pages under I/O have been copied to head->rc_pages.
699 	 * Prevent their premature release by svc_xprt_release() .
700 	 */
701 	for (i = 0; i < info->ri_readctxt->rc_page_count; i++)
702 		rqstp->rq_pages[i] = NULL;
703 
704 	return ret;
705 }
706 
707 /* Construct RDMA Reads to pull over a normal Read chunk. The chunk
708  * data lands in the page list of head->rc_arg.pages.
709  *
710  * Currently NFSD does not look at the head->rc_arg.tail[0] iovec.
711  * Therefore, XDR round-up of the Read chunk and trailing
712  * inline content must both be added at the end of the pagelist.
713  */
714 static int svc_rdma_build_normal_read_chunk(struct svc_rqst *rqstp,
715 					    struct svc_rdma_read_info *info,
716 					    __be32 *p)
717 {
718 	struct svc_rdma_recv_ctxt *head = info->ri_readctxt;
719 	int ret;
720 
721 	ret = svc_rdma_build_read_chunk(rqstp, info, p);
722 	if (ret < 0)
723 		goto out;
724 
725 	trace_svcrdma_encode_read(info->ri_chunklen, info->ri_position);
726 
727 	head->rc_hdr_count = 0;
728 
729 	/* Split the Receive buffer between the head and tail
730 	 * buffers at Read chunk's position. XDR roundup of the
731 	 * chunk is not included in either the pagelist or in
732 	 * the tail.
733 	 */
734 	head->rc_arg.tail[0].iov_base =
735 		head->rc_arg.head[0].iov_base + info->ri_position;
736 	head->rc_arg.tail[0].iov_len =
737 		head->rc_arg.head[0].iov_len - info->ri_position;
738 	head->rc_arg.head[0].iov_len = info->ri_position;
739 
740 	/* Read chunk may need XDR roundup (see RFC 8166, s. 3.4.5.2).
741 	 *
742 	 * If the client already rounded up the chunk length, the
743 	 * length does not change. Otherwise, the length of the page
744 	 * list is increased to include XDR round-up.
745 	 *
746 	 * Currently these chunks always start at page offset 0,
747 	 * thus the rounded-up length never crosses a page boundary.
748 	 */
749 	info->ri_chunklen = XDR_QUADLEN(info->ri_chunklen) << 2;
750 
751 	head->rc_arg.page_len = info->ri_chunklen;
752 	head->rc_arg.len += info->ri_chunklen;
753 	head->rc_arg.buflen += info->ri_chunklen;
754 
755 out:
756 	return ret;
757 }
758 
759 /* Construct RDMA Reads to pull over a Position Zero Read chunk.
760  * The start of the data lands in the first page just after
761  * the Transport header, and the rest lands in the page list of
762  * head->rc_arg.pages.
763  *
764  * Assumptions:
765  *	- A PZRC has an XDR-aligned length (no implicit round-up).
766  *	- There can be no trailing inline content (IOW, we assume
767  *	  a PZRC is never sent in an RDMA_MSG message, though it's
768  *	  allowed by spec).
769  */
770 static int svc_rdma_build_pz_read_chunk(struct svc_rqst *rqstp,
771 					struct svc_rdma_read_info *info,
772 					__be32 *p)
773 {
774 	struct svc_rdma_recv_ctxt *head = info->ri_readctxt;
775 	int ret;
776 
777 	ret = svc_rdma_build_read_chunk(rqstp, info, p);
778 	if (ret < 0)
779 		goto out;
780 
781 	trace_svcrdma_encode_pzr(info->ri_chunklen);
782 
783 	head->rc_arg.len += info->ri_chunklen;
784 	head->rc_arg.buflen += info->ri_chunklen;
785 
786 	head->rc_hdr_count = 1;
787 	head->rc_arg.head[0].iov_base = page_address(head->rc_pages[0]);
788 	head->rc_arg.head[0].iov_len = min_t(size_t, PAGE_SIZE,
789 					     info->ri_chunklen);
790 
791 	head->rc_arg.page_len = info->ri_chunklen -
792 				head->rc_arg.head[0].iov_len;
793 
794 out:
795 	return ret;
796 }
797 
798 /**
799  * svc_rdma_recv_read_chunk - Pull a Read chunk from the client
800  * @rdma: controlling RDMA transport
801  * @rqstp: set of pages to use as Read sink buffers
802  * @head: pages under I/O collect here
803  * @p: pointer to start of Read chunk
804  *
805  * Returns:
806  *	%0 if all needed RDMA Reads were posted successfully,
807  *	%-EINVAL if client provided too many segments,
808  *	%-ENOMEM if rdma_rw context pool was exhausted,
809  *	%-ENOTCONN if posting failed (connection is lost),
810  *	%-EIO if rdma_rw initialization failed (DMA mapping, etc).
811  *
812  * Assumptions:
813  * - All Read segments in @p have the same Position value.
814  */
815 int svc_rdma_recv_read_chunk(struct svcxprt_rdma *rdma, struct svc_rqst *rqstp,
816 			     struct svc_rdma_recv_ctxt *head, __be32 *p)
817 {
818 	struct svc_rdma_read_info *info;
819 	int ret;
820 
821 	/* The request (with page list) is constructed in
822 	 * head->rc_arg. Pages involved with RDMA Read I/O are
823 	 * transferred there.
824 	 */
825 	head->rc_arg.head[0] = rqstp->rq_arg.head[0];
826 	head->rc_arg.tail[0] = rqstp->rq_arg.tail[0];
827 	head->rc_arg.pages = head->rc_pages;
828 	head->rc_arg.page_base = 0;
829 	head->rc_arg.page_len = 0;
830 	head->rc_arg.len = rqstp->rq_arg.len;
831 	head->rc_arg.buflen = rqstp->rq_arg.buflen;
832 
833 	info = svc_rdma_read_info_alloc(rdma);
834 	if (!info)
835 		return -ENOMEM;
836 	info->ri_readctxt = head;
837 	info->ri_pageno = 0;
838 	info->ri_pageoff = 0;
839 
840 	info->ri_position = be32_to_cpup(p + 1);
841 	if (info->ri_position)
842 		ret = svc_rdma_build_normal_read_chunk(rqstp, info, p);
843 	else
844 		ret = svc_rdma_build_pz_read_chunk(rqstp, info, p);
845 	if (ret < 0)
846 		goto out_err;
847 
848 	ret = svc_rdma_post_chunk_ctxt(&info->ri_cc);
849 	if (ret < 0)
850 		goto out_err;
851 	return 0;
852 
853 out_err:
854 	svc_rdma_read_info_free(info);
855 	return ret;
856 }
857