1 /*
2  * Copyright (c) 2005-2006 Network Appliance, Inc. All rights reserved.
3  *
4  * This software is available to you under a choice of one of two
5  * licenses.  You may choose to be licensed under the terms of the GNU
6  * General Public License (GPL) Version 2, available from the file
7  * COPYING in the main directory of this source tree, or the BSD-type
8  * license below:
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  *
14  *      Redistributions of source code must retain the above copyright
15  *      notice, this list of conditions and the following disclaimer.
16  *
17  *      Redistributions in binary form must reproduce the above
18  *      copyright notice, this list of conditions and the following
19  *      disclaimer in the documentation and/or other materials provided
20  *      with the distribution.
21  *
22  *      Neither the name of the Network Appliance, Inc. nor the names of
23  *      its contributors may be used to endorse or promote products
24  *      derived from this software without specific prior written
25  *      permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
28  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
29  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
30  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
31  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
32  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
33  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
34  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
35  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
36  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
37  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
38  *
39  * Author: Tom Tucker <tom@opengridcomputing.com>
40  */
41 
42 #include <linux/sunrpc/debug.h>
43 #include <linux/sunrpc/rpc_rdma.h>
44 #include <linux/spinlock.h>
45 #include <asm/unaligned.h>
46 #include <rdma/ib_verbs.h>
47 #include <rdma/rdma_cm.h>
48 #include <linux/sunrpc/svc_rdma.h>
49 
50 #define RPCDBG_FACILITY	RPCDBG_SVCXPRT
51 
52 /*
53  * Replace the pages in the rq_argpages array with the pages from the SGE in
54  * the RDMA_RECV completion. The SGL should contain full pages up until the
55  * last one.
56  */
57 static void rdma_build_arg_xdr(struct svc_rqst *rqstp,
58 			       struct svc_rdma_op_ctxt *ctxt,
59 			       u32 byte_count)
60 {
61 	struct page *page;
62 	u32 bc;
63 	int sge_no;
64 
65 	/* Swap the page in the SGE with the page in argpages */
66 	page = ctxt->pages[0];
67 	put_page(rqstp->rq_pages[0]);
68 	rqstp->rq_pages[0] = page;
69 
70 	/* Set up the XDR head */
71 	rqstp->rq_arg.head[0].iov_base = page_address(page);
72 	rqstp->rq_arg.head[0].iov_len = min(byte_count, ctxt->sge[0].length);
73 	rqstp->rq_arg.len = byte_count;
74 	rqstp->rq_arg.buflen = byte_count;
75 
76 	/* Compute bytes past head in the SGL */
77 	bc = byte_count - rqstp->rq_arg.head[0].iov_len;
78 
79 	/* If data remains, store it in the pagelist */
80 	rqstp->rq_arg.page_len = bc;
81 	rqstp->rq_arg.page_base = 0;
82 	rqstp->rq_arg.pages = &rqstp->rq_pages[1];
83 	sge_no = 1;
84 	while (bc && sge_no < ctxt->count) {
85 		page = ctxt->pages[sge_no];
86 		put_page(rqstp->rq_pages[sge_no]);
87 		rqstp->rq_pages[sge_no] = page;
88 		bc -= min(bc, ctxt->sge[sge_no].length);
89 		rqstp->rq_arg.buflen += ctxt->sge[sge_no].length;
90 		sge_no++;
91 	}
92 	rqstp->rq_respages = &rqstp->rq_pages[sge_no];
93 
94 	/* We should never run out of SGE because the limit is defined to
95 	 * support the max allowed RPC data length
96 	 */
97 	BUG_ON(bc && (sge_no == ctxt->count));
98 	BUG_ON((rqstp->rq_arg.head[0].iov_len + rqstp->rq_arg.page_len)
99 	       != byte_count);
100 	BUG_ON(rqstp->rq_arg.len != byte_count);
101 
102 	/* If not all pages were used from the SGL, free the remaining ones */
103 	bc = sge_no;
104 	while (sge_no < ctxt->count) {
105 		page = ctxt->pages[sge_no++];
106 		put_page(page);
107 	}
108 	ctxt->count = bc;
109 
110 	/* Set up tail */
111 	rqstp->rq_arg.tail[0].iov_base = NULL;
112 	rqstp->rq_arg.tail[0].iov_len = 0;
113 }
114 
115 struct chunk_sge {
116 	int start;		/* sge no for this chunk */
117 	int count;		/* sge count for this chunk */
118 };
119 
120 /* Encode a read-chunk-list as an array of IB SGE
121  *
122  * Assumptions:
123  * - chunk[0]->position points to pages[0] at an offset of 0
124  * - pages[] is not physically or virtually contigous and consists of
125  *   PAGE_SIZE elements.
126  *
127  * Output:
128  * - sge array pointing into pages[] array.
129  * - chunk_sge array specifying sge index and count for each
130  *   chunk in the read list
131  *
132  */
133 static int rdma_rcl_to_sge(struct svcxprt_rdma *xprt,
134 			   struct svc_rqst *rqstp,
135 			   struct svc_rdma_op_ctxt *head,
136 			   struct rpcrdma_msg *rmsgp,
137 			   struct ib_sge *sge,
138 			   struct chunk_sge *ch_sge_ary,
139 			   int ch_count,
140 			   int byte_count)
141 {
142 	int sge_no;
143 	int sge_bytes;
144 	int page_off;
145 	int page_no;
146 	int ch_bytes;
147 	int ch_no;
148 	struct rpcrdma_read_chunk *ch;
149 
150 	sge_no = 0;
151 	page_no = 0;
152 	page_off = 0;
153 	ch = (struct rpcrdma_read_chunk *)&rmsgp->rm_body.rm_chunks[0];
154 	ch_no = 0;
155 	ch_bytes = ch->rc_target.rs_length;
156 	head->arg.head[0] = rqstp->rq_arg.head[0];
157 	head->arg.tail[0] = rqstp->rq_arg.tail[0];
158 	head->arg.pages = &head->pages[head->count];
159 	head->sge[0].length = head->count; /* save count of hdr pages */
160 	head->arg.page_base = 0;
161 	head->arg.page_len = ch_bytes;
162 	head->arg.len = rqstp->rq_arg.len + ch_bytes;
163 	head->arg.buflen = rqstp->rq_arg.buflen + ch_bytes;
164 	head->count++;
165 	ch_sge_ary[0].start = 0;
166 	while (byte_count) {
167 		sge_bytes = min_t(int, PAGE_SIZE-page_off, ch_bytes);
168 		sge[sge_no].addr =
169 			ib_dma_map_page(xprt->sc_cm_id->device,
170 					rqstp->rq_arg.pages[page_no],
171 					page_off, sge_bytes,
172 					DMA_FROM_DEVICE);
173 		sge[sge_no].length = sge_bytes;
174 		sge[sge_no].lkey = xprt->sc_phys_mr->lkey;
175 		/*
176 		 * Don't bump head->count here because the same page
177 		 * may be used by multiple SGE.
178 		 */
179 		head->arg.pages[page_no] = rqstp->rq_arg.pages[page_no];
180 		rqstp->rq_respages = &rqstp->rq_arg.pages[page_no+1];
181 
182 		byte_count -= sge_bytes;
183 		ch_bytes -= sge_bytes;
184 		sge_no++;
185 		/*
186 		 * If all bytes for this chunk have been mapped to an
187 		 * SGE, move to the next SGE
188 		 */
189 		if (ch_bytes == 0) {
190 			ch_sge_ary[ch_no].count =
191 				sge_no - ch_sge_ary[ch_no].start;
192 			ch_no++;
193 			ch++;
194 			ch_sge_ary[ch_no].start = sge_no;
195 			ch_bytes = ch->rc_target.rs_length;
196 			/* If bytes remaining account for next chunk */
197 			if (byte_count) {
198 				head->arg.page_len += ch_bytes;
199 				head->arg.len += ch_bytes;
200 				head->arg.buflen += ch_bytes;
201 			}
202 		}
203 		/*
204 		 * If this SGE consumed all of the page, move to the
205 		 * next page
206 		 */
207 		if ((sge_bytes + page_off) == PAGE_SIZE) {
208 			page_no++;
209 			page_off = 0;
210 			/*
211 			 * If there are still bytes left to map, bump
212 			 * the page count
213 			 */
214 			if (byte_count)
215 				head->count++;
216 		} else
217 			page_off += sge_bytes;
218 	}
219 	BUG_ON(byte_count != 0);
220 	return sge_no;
221 }
222 
223 static void rdma_set_ctxt_sge(struct svc_rdma_op_ctxt *ctxt,
224 			      struct ib_sge *sge,
225 			      u64 *sgl_offset,
226 			      int count)
227 {
228 	int i;
229 
230 	ctxt->count = count;
231 	for (i = 0; i < count; i++) {
232 		ctxt->sge[i].addr = sge[i].addr;
233 		ctxt->sge[i].length = sge[i].length;
234 		*sgl_offset = *sgl_offset + sge[i].length;
235 	}
236 }
237 
238 static int rdma_read_max_sge(struct svcxprt_rdma *xprt, int sge_count)
239 {
240 	if ((RDMA_TRANSPORT_IWARP ==
241 	     rdma_node_get_transport(xprt->sc_cm_id->
242 				     device->node_type))
243 	    && sge_count > 1)
244 		return 1;
245 	else
246 		return min_t(int, sge_count, xprt->sc_max_sge);
247 }
248 
249 /*
250  * Use RDMA_READ to read data from the advertised client buffer into the
251  * XDR stream starting at rq_arg.head[0].iov_base.
252  * Each chunk in the array
253  * contains the following fields:
254  * discrim      - '1', This isn't used for data placement
255  * position     - The xdr stream offset (the same for every chunk)
256  * handle       - RMR for client memory region
257  * length       - data transfer length
258  * offset       - 64 bit tagged offset in remote memory region
259  *
260  * On our side, we need to read into a pagelist. The first page immediately
261  * follows the RPC header.
262  *
263  * This function returns 1 to indicate success. The data is not yet in
264  * the pagelist and therefore the RPC request must be deferred. The
265  * I/O completion will enqueue the transport again and
266  * svc_rdma_recvfrom will complete the request.
267  *
268  * NOTE: The ctxt must not be touched after the last WR has been posted
269  * because the I/O completion processing may occur on another
270  * processor and free / modify the context. Ne touche pas!
271  */
272 static int rdma_read_xdr(struct svcxprt_rdma *xprt,
273 			 struct rpcrdma_msg *rmsgp,
274 			 struct svc_rqst *rqstp,
275 			 struct svc_rdma_op_ctxt *hdr_ctxt)
276 {
277 	struct ib_send_wr read_wr;
278 	int err = 0;
279 	int ch_no;
280 	struct ib_sge *sge;
281 	int ch_count;
282 	int byte_count;
283 	int sge_count;
284 	u64 sgl_offset;
285 	struct rpcrdma_read_chunk *ch;
286 	struct svc_rdma_op_ctxt *ctxt = NULL;
287 	struct svc_rdma_op_ctxt *head;
288 	struct svc_rdma_op_ctxt *tmp_sge_ctxt;
289 	struct svc_rdma_op_ctxt *tmp_ch_ctxt;
290 	struct chunk_sge *ch_sge_ary;
291 
292 	/* If no read list is present, return 0 */
293 	ch = svc_rdma_get_read_chunk(rmsgp);
294 	if (!ch)
295 		return 0;
296 
297 	/* Allocate temporary contexts to keep SGE */
298 	BUG_ON(sizeof(struct ib_sge) < sizeof(struct chunk_sge));
299 	tmp_sge_ctxt = svc_rdma_get_context(xprt);
300 	sge = tmp_sge_ctxt->sge;
301 	tmp_ch_ctxt = svc_rdma_get_context(xprt);
302 	ch_sge_ary = (struct chunk_sge *)tmp_ch_ctxt->sge;
303 
304 	svc_rdma_rcl_chunk_counts(ch, &ch_count, &byte_count);
305 	sge_count = rdma_rcl_to_sge(xprt, rqstp, hdr_ctxt, rmsgp,
306 				    sge, ch_sge_ary,
307 				    ch_count, byte_count);
308 	head = svc_rdma_get_context(xprt);
309 	sgl_offset = 0;
310 	ch_no = 0;
311 
312 	for (ch = (struct rpcrdma_read_chunk *)&rmsgp->rm_body.rm_chunks[0];
313 	     ch->rc_discrim != 0; ch++, ch_no++) {
314 next_sge:
315 		if (!ctxt)
316 			ctxt = head;
317 		else {
318 			ctxt->next = svc_rdma_get_context(xprt);
319 			ctxt = ctxt->next;
320 		}
321 		ctxt->next = NULL;
322 		ctxt->direction = DMA_FROM_DEVICE;
323 		clear_bit(RDMACTXT_F_READ_DONE, &ctxt->flags);
324 		clear_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);
325 
326 		/* Prepare READ WR */
327 		memset(&read_wr, 0, sizeof read_wr);
328 		ctxt->wr_op = IB_WR_RDMA_READ;
329 		read_wr.wr_id = (unsigned long)ctxt;
330 		read_wr.opcode = IB_WR_RDMA_READ;
331 		read_wr.send_flags = IB_SEND_SIGNALED;
332 		read_wr.wr.rdma.rkey = ch->rc_target.rs_handle;
333 		read_wr.wr.rdma.remote_addr =
334 			get_unaligned(&(ch->rc_target.rs_offset)) +
335 			sgl_offset;
336 		read_wr.sg_list = &sge[ch_sge_ary[ch_no].start];
337 		read_wr.num_sge =
338 			rdma_read_max_sge(xprt, ch_sge_ary[ch_no].count);
339 		rdma_set_ctxt_sge(ctxt, &sge[ch_sge_ary[ch_no].start],
340 				  &sgl_offset,
341 				  read_wr.num_sge);
342 		if (((ch+1)->rc_discrim == 0) &&
343 		    (read_wr.num_sge == ch_sge_ary[ch_no].count)) {
344 			/*
345 			 * Mark the last RDMA_READ with a bit to
346 			 * indicate all RPC data has been fetched from
347 			 * the client and the RPC needs to be enqueued.
348 			 */
349 			set_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);
350 			ctxt->next = hdr_ctxt;
351 			hdr_ctxt->next = head;
352 		}
353 		/* Post the read */
354 		err = svc_rdma_send(xprt, &read_wr);
355 		if (err) {
356 			printk(KERN_ERR "svcrdma: Error posting send = %d\n",
357 			       err);
358 			/*
359 			 * Break the circular list so free knows when
360 			 * to stop if the error happened to occur on
361 			 * the last read
362 			 */
363 			ctxt->next = NULL;
364 			goto out;
365 		}
366 		atomic_inc(&rdma_stat_read);
367 
368 		if (read_wr.num_sge < ch_sge_ary[ch_no].count) {
369 			ch_sge_ary[ch_no].count -= read_wr.num_sge;
370 			ch_sge_ary[ch_no].start += read_wr.num_sge;
371 			goto next_sge;
372 		}
373 		sgl_offset = 0;
374 		err = 0;
375 	}
376 
377  out:
378 	svc_rdma_put_context(tmp_sge_ctxt, 0);
379 	svc_rdma_put_context(tmp_ch_ctxt, 0);
380 
381 	/* Detach arg pages. svc_recv will replenish them */
382 	for (ch_no = 0; &rqstp->rq_pages[ch_no] < rqstp->rq_respages; ch_no++)
383 		rqstp->rq_pages[ch_no] = NULL;
384 
385 	/*
386 	 * Detach res pages. svc_release must see a resused count of
387 	 * zero or it will attempt to put them.
388 	 */
389 	while (rqstp->rq_resused)
390 		rqstp->rq_respages[--rqstp->rq_resused] = NULL;
391 
392 	if (err) {
393 		printk(KERN_ERR "svcrdma : RDMA_READ error = %d\n", err);
394 		set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
395 		/* Free the linked list of read contexts */
396 		while (head != NULL) {
397 			ctxt = head->next;
398 			svc_rdma_put_context(head, 1);
399 			head = ctxt;
400 		}
401 		return 0;
402 	}
403 
404 	return 1;
405 }
406 
407 static int rdma_read_complete(struct svc_rqst *rqstp,
408 			      struct svc_rdma_op_ctxt *data)
409 {
410 	struct svc_rdma_op_ctxt *head = data->next;
411 	int page_no;
412 	int ret;
413 
414 	BUG_ON(!head);
415 
416 	/* Copy RPC pages */
417 	for (page_no = 0; page_no < head->count; page_no++) {
418 		put_page(rqstp->rq_pages[page_no]);
419 		rqstp->rq_pages[page_no] = head->pages[page_no];
420 	}
421 	/* Point rq_arg.pages past header */
422 	rqstp->rq_arg.pages = &rqstp->rq_pages[head->sge[0].length];
423 	rqstp->rq_arg.page_len = head->arg.page_len;
424 	rqstp->rq_arg.page_base = head->arg.page_base;
425 
426 	/* rq_respages starts after the last arg page */
427 	rqstp->rq_respages = &rqstp->rq_arg.pages[page_no];
428 	rqstp->rq_resused = 0;
429 
430 	/* Rebuild rq_arg head and tail. */
431 	rqstp->rq_arg.head[0] = head->arg.head[0];
432 	rqstp->rq_arg.tail[0] = head->arg.tail[0];
433 	rqstp->rq_arg.len = head->arg.len;
434 	rqstp->rq_arg.buflen = head->arg.buflen;
435 
436 	/* XXX: What should this be? */
437 	rqstp->rq_prot = IPPROTO_MAX;
438 
439 	/*
440 	 * Free the contexts we used to build the RDMA_READ. We have
441 	 * to be careful here because the context list uses the same
442 	 * next pointer used to chain the contexts associated with the
443 	 * RDMA_READ
444 	 */
445 	data->next = NULL;	/* terminate circular list */
446 	do {
447 		data = head->next;
448 		svc_rdma_put_context(head, 0);
449 		head = data;
450 	} while (head != NULL);
451 
452 	ret = rqstp->rq_arg.head[0].iov_len
453 		+ rqstp->rq_arg.page_len
454 		+ rqstp->rq_arg.tail[0].iov_len;
455 	dprintk("svcrdma: deferred read ret=%d, rq_arg.len =%d, "
456 		"rq_arg.head[0].iov_base=%p, rq_arg.head[0].iov_len = %zd\n",
457 		ret, rqstp->rq_arg.len,	rqstp->rq_arg.head[0].iov_base,
458 		rqstp->rq_arg.head[0].iov_len);
459 
460 	/* Indicate that we've consumed an RQ credit */
461 	rqstp->rq_xprt_ctxt = rqstp->rq_xprt;
462 	svc_xprt_received(rqstp->rq_xprt);
463 	return ret;
464 }
465 
466 /*
467  * Set up the rqstp thread context to point to the RQ buffer. If
468  * necessary, pull additional data from the client with an RDMA_READ
469  * request.
470  */
471 int svc_rdma_recvfrom(struct svc_rqst *rqstp)
472 {
473 	struct svc_xprt *xprt = rqstp->rq_xprt;
474 	struct svcxprt_rdma *rdma_xprt =
475 		container_of(xprt, struct svcxprt_rdma, sc_xprt);
476 	struct svc_rdma_op_ctxt *ctxt = NULL;
477 	struct rpcrdma_msg *rmsgp;
478 	int ret = 0;
479 	int len;
480 
481 	dprintk("svcrdma: rqstp=%p\n", rqstp);
482 
483 	/*
484 	 * The rq_xprt_ctxt indicates if we've consumed an RQ credit
485 	 * or not. It is used in the rdma xpo_release_rqst function to
486 	 * determine whether or not to return an RQ WQE to the RQ.
487 	 */
488 	rqstp->rq_xprt_ctxt = NULL;
489 
490 	spin_lock_bh(&rdma_xprt->sc_read_complete_lock);
491 	if (!list_empty(&rdma_xprt->sc_read_complete_q)) {
492 		ctxt = list_entry(rdma_xprt->sc_read_complete_q.next,
493 				  struct svc_rdma_op_ctxt,
494 				  dto_q);
495 		list_del_init(&ctxt->dto_q);
496 	}
497 	spin_unlock_bh(&rdma_xprt->sc_read_complete_lock);
498 	if (ctxt)
499 		return rdma_read_complete(rqstp, ctxt);
500 
501 	spin_lock_bh(&rdma_xprt->sc_rq_dto_lock);
502 	if (!list_empty(&rdma_xprt->sc_rq_dto_q)) {
503 		ctxt = list_entry(rdma_xprt->sc_rq_dto_q.next,
504 				  struct svc_rdma_op_ctxt,
505 				  dto_q);
506 		list_del_init(&ctxt->dto_q);
507 	} else {
508 		atomic_inc(&rdma_stat_rq_starve);
509 		clear_bit(XPT_DATA, &xprt->xpt_flags);
510 		ctxt = NULL;
511 	}
512 	spin_unlock_bh(&rdma_xprt->sc_rq_dto_lock);
513 	if (!ctxt) {
514 		/* This is the EAGAIN path. The svc_recv routine will
515 		 * return -EAGAIN, the nfsd thread will go to call into
516 		 * svc_recv again and we shouldn't be on the active
517 		 * transport list
518 		 */
519 		if (test_bit(XPT_CLOSE, &xprt->xpt_flags))
520 			goto close_out;
521 
522 		BUG_ON(ret);
523 		goto out;
524 	}
525 	dprintk("svcrdma: processing ctxt=%p on xprt=%p, rqstp=%p, status=%d\n",
526 		ctxt, rdma_xprt, rqstp, ctxt->wc_status);
527 	BUG_ON(ctxt->wc_status != IB_WC_SUCCESS);
528 	atomic_inc(&rdma_stat_recv);
529 
530 	/* Build up the XDR from the receive buffers. */
531 	rdma_build_arg_xdr(rqstp, ctxt, ctxt->byte_len);
532 
533 	/* Decode the RDMA header. */
534 	len = svc_rdma_xdr_decode_req(&rmsgp, rqstp);
535 	rqstp->rq_xprt_hlen = len;
536 
537 	/* If the request is invalid, reply with an error */
538 	if (len < 0) {
539 		if (len == -ENOSYS)
540 			(void)svc_rdma_send_error(rdma_xprt, rmsgp, ERR_VERS);
541 		goto close_out;
542 	}
543 
544 	/* Read read-list data. If we would need to wait, defer
545 	 * it. Not that in this case, we don't return the RQ credit
546 	 * until after the read completes.
547 	 */
548 	if (rdma_read_xdr(rdma_xprt, rmsgp, rqstp, ctxt)) {
549 		svc_xprt_received(xprt);
550 		return 0;
551 	}
552 
553 	/* Indicate we've consumed an RQ credit */
554 	rqstp->rq_xprt_ctxt = rqstp->rq_xprt;
555 
556 	ret = rqstp->rq_arg.head[0].iov_len
557 		+ rqstp->rq_arg.page_len
558 		+ rqstp->rq_arg.tail[0].iov_len;
559 	svc_rdma_put_context(ctxt, 0);
560  out:
561 	dprintk("svcrdma: ret = %d, rq_arg.len =%d, "
562 		"rq_arg.head[0].iov_base=%p, rq_arg.head[0].iov_len = %zd\n",
563 		ret, rqstp->rq_arg.len,
564 		rqstp->rq_arg.head[0].iov_base,
565 		rqstp->rq_arg.head[0].iov_len);
566 	rqstp->rq_prot = IPPROTO_MAX;
567 	svc_xprt_copy_addrs(rqstp, xprt);
568 	svc_xprt_received(xprt);
569 	return ret;
570 
571  close_out:
572 	if (ctxt) {
573 		svc_rdma_put_context(ctxt, 1);
574 		/* Indicate we've consumed an RQ credit */
575 		rqstp->rq_xprt_ctxt = rqstp->rq_xprt;
576 	}
577 	dprintk("svcrdma: transport %p is closing\n", xprt);
578 	/*
579 	 * Set the close bit and enqueue it. svc_recv will see the
580 	 * close bit and call svc_xprt_delete
581 	 */
582 	set_bit(XPT_CLOSE, &xprt->xpt_flags);
583 	svc_xprt_received(xprt);
584 	return 0;
585 }
586