1 /*
2  * Copyright (c) 2014 Open Grid Computing, Inc. All rights reserved.
3  * Copyright (c) 2005-2006 Network Appliance, Inc. All rights reserved.
4  *
5  * This software is available to you under a choice of one of two
6  * licenses.  You may choose to be licensed under the terms of the GNU
7  * General Public License (GPL) Version 2, available from the file
8  * COPYING in the main directory of this source tree, or the BSD-type
9  * license below:
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  *
15  *      Redistributions of source code must retain the above copyright
16  *      notice, this list of conditions and the following disclaimer.
17  *
18  *      Redistributions in binary form must reproduce the above
19  *      copyright notice, this list of conditions and the following
20  *      disclaimer in the documentation and/or other materials provided
21  *      with the distribution.
22  *
23  *      Neither the name of the Network Appliance, Inc. nor the names of
24  *      its contributors may be used to endorse or promote products
25  *      derived from this software without specific prior written
26  *      permission.
27  *
28  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
29  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
30  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
31  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
32  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
33  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
34  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
35  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
36  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
37  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
38  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
39  *
40  * Author: Tom Tucker <tom@opengridcomputing.com>
41  */
42 
43 #include <linux/sunrpc/debug.h>
44 #include <linux/sunrpc/rpc_rdma.h>
45 #include <linux/spinlock.h>
46 #include <asm/unaligned.h>
47 #include <rdma/ib_verbs.h>
48 #include <rdma/rdma_cm.h>
49 #include <linux/sunrpc/svc_rdma.h>
50 
51 #define RPCDBG_FACILITY	RPCDBG_SVCXPRT
52 
53 /*
54  * Replace the pages in the rq_argpages array with the pages from the SGE in
55  * the RDMA_RECV completion. The SGL should contain full pages up until the
56  * last one.
57  */
rdma_build_arg_xdr(struct svc_rqst * rqstp,struct svc_rdma_op_ctxt * ctxt,u32 byte_count)58 static void rdma_build_arg_xdr(struct svc_rqst *rqstp,
59 			       struct svc_rdma_op_ctxt *ctxt,
60 			       u32 byte_count)
61 {
62 	struct rpcrdma_msg *rmsgp;
63 	struct page *page;
64 	u32 bc;
65 	int sge_no;
66 
67 	/* Swap the page in the SGE with the page in argpages */
68 	page = ctxt->pages[0];
69 	put_page(rqstp->rq_pages[0]);
70 	rqstp->rq_pages[0] = page;
71 
72 	/* Set up the XDR head */
73 	rqstp->rq_arg.head[0].iov_base = page_address(page);
74 	rqstp->rq_arg.head[0].iov_len =
75 		min_t(size_t, byte_count, ctxt->sge[0].length);
76 	rqstp->rq_arg.len = byte_count;
77 	rqstp->rq_arg.buflen = byte_count;
78 
79 	/* Compute bytes past head in the SGL */
80 	bc = byte_count - rqstp->rq_arg.head[0].iov_len;
81 
82 	/* If data remains, store it in the pagelist */
83 	rqstp->rq_arg.page_len = bc;
84 	rqstp->rq_arg.page_base = 0;
85 
86 	/* RDMA_NOMSG: RDMA READ data should land just after RDMA RECV data */
87 	rmsgp = (struct rpcrdma_msg *)rqstp->rq_arg.head[0].iov_base;
88 	if (be32_to_cpu(rmsgp->rm_type) == RDMA_NOMSG)
89 		rqstp->rq_arg.pages = &rqstp->rq_pages[0];
90 	else
91 		rqstp->rq_arg.pages = &rqstp->rq_pages[1];
92 
93 	sge_no = 1;
94 	while (bc && sge_no < ctxt->count) {
95 		page = ctxt->pages[sge_no];
96 		put_page(rqstp->rq_pages[sge_no]);
97 		rqstp->rq_pages[sge_no] = page;
98 		bc -= min_t(u32, bc, ctxt->sge[sge_no].length);
99 		rqstp->rq_arg.buflen += ctxt->sge[sge_no].length;
100 		sge_no++;
101 	}
102 	rqstp->rq_respages = &rqstp->rq_pages[sge_no];
103 	rqstp->rq_next_page = rqstp->rq_respages + 1;
104 
105 	/* If not all pages were used from the SGL, free the remaining ones */
106 	bc = sge_no;
107 	while (sge_no < ctxt->count) {
108 		page = ctxt->pages[sge_no++];
109 		put_page(page);
110 	}
111 	ctxt->count = bc;
112 
113 	/* Set up tail */
114 	rqstp->rq_arg.tail[0].iov_base = NULL;
115 	rqstp->rq_arg.tail[0].iov_len = 0;
116 }
117 
rdma_read_max_sge(struct svcxprt_rdma * xprt,int sge_count)118 static int rdma_read_max_sge(struct svcxprt_rdma *xprt, int sge_count)
119 {
120 	if (rdma_node_get_transport(xprt->sc_cm_id->device->node_type) ==
121 	     RDMA_TRANSPORT_IWARP)
122 		return 1;
123 	else
124 		return min_t(int, sge_count, xprt->sc_max_sge);
125 }
126 
127 /* Issue an RDMA_READ using the local lkey to map the data sink */
rdma_read_chunk_lcl(struct svcxprt_rdma * xprt,struct svc_rqst * rqstp,struct svc_rdma_op_ctxt * head,int * page_no,u32 * page_offset,u32 rs_handle,u32 rs_length,u64 rs_offset,bool last)128 int rdma_read_chunk_lcl(struct svcxprt_rdma *xprt,
129 			struct svc_rqst *rqstp,
130 			struct svc_rdma_op_ctxt *head,
131 			int *page_no,
132 			u32 *page_offset,
133 			u32 rs_handle,
134 			u32 rs_length,
135 			u64 rs_offset,
136 			bool last)
137 {
138 	struct ib_send_wr read_wr;
139 	int pages_needed = PAGE_ALIGN(*page_offset + rs_length) >> PAGE_SHIFT;
140 	struct svc_rdma_op_ctxt *ctxt = svc_rdma_get_context(xprt);
141 	int ret, read, pno;
142 	u32 pg_off = *page_offset;
143 	u32 pg_no = *page_no;
144 
145 	ctxt->direction = DMA_FROM_DEVICE;
146 	ctxt->read_hdr = head;
147 	pages_needed =
148 		min_t(int, pages_needed, rdma_read_max_sge(xprt, pages_needed));
149 	read = min_t(int, (pages_needed << PAGE_SHIFT) - *page_offset,
150 		     rs_length);
151 
152 	for (pno = 0; pno < pages_needed; pno++) {
153 		int len = min_t(int, rs_length, PAGE_SIZE - pg_off);
154 
155 		head->arg.pages[pg_no] = rqstp->rq_arg.pages[pg_no];
156 		head->arg.page_len += len;
157 		head->arg.len += len;
158 		if (!pg_off)
159 			head->count++;
160 		rqstp->rq_respages = &rqstp->rq_arg.pages[pg_no+1];
161 		rqstp->rq_next_page = rqstp->rq_respages + 1;
162 		ctxt->sge[pno].addr =
163 			ib_dma_map_page(xprt->sc_cm_id->device,
164 					head->arg.pages[pg_no], pg_off,
165 					PAGE_SIZE - pg_off,
166 					DMA_FROM_DEVICE);
167 		ret = ib_dma_mapping_error(xprt->sc_cm_id->device,
168 					   ctxt->sge[pno].addr);
169 		if (ret)
170 			goto err;
171 		atomic_inc(&xprt->sc_dma_used);
172 
173 		/* The lkey here is either a local dma lkey or a dma_mr lkey */
174 		ctxt->sge[pno].lkey = xprt->sc_dma_lkey;
175 		ctxt->sge[pno].length = len;
176 		ctxt->count++;
177 
178 		/* adjust offset and wrap to next page if needed */
179 		pg_off += len;
180 		if (pg_off == PAGE_SIZE) {
181 			pg_off = 0;
182 			pg_no++;
183 		}
184 		rs_length -= len;
185 	}
186 
187 	if (last && rs_length == 0)
188 		set_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);
189 	else
190 		clear_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);
191 
192 	memset(&read_wr, 0, sizeof(read_wr));
193 	read_wr.wr_id = (unsigned long)ctxt;
194 	read_wr.opcode = IB_WR_RDMA_READ;
195 	ctxt->wr_op = read_wr.opcode;
196 	read_wr.send_flags = IB_SEND_SIGNALED;
197 	read_wr.wr.rdma.rkey = rs_handle;
198 	read_wr.wr.rdma.remote_addr = rs_offset;
199 	read_wr.sg_list = ctxt->sge;
200 	read_wr.num_sge = pages_needed;
201 
202 	ret = svc_rdma_send(xprt, &read_wr);
203 	if (ret) {
204 		pr_err("svcrdma: Error %d posting RDMA_READ\n", ret);
205 		set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
206 		goto err;
207 	}
208 
209 	/* return current location in page array */
210 	*page_no = pg_no;
211 	*page_offset = pg_off;
212 	ret = read;
213 	atomic_inc(&rdma_stat_read);
214 	return ret;
215  err:
216 	svc_rdma_unmap_dma(ctxt);
217 	svc_rdma_put_context(ctxt, 0);
218 	return ret;
219 }
220 
221 /* Issue an RDMA_READ using an FRMR to map the data sink */
rdma_read_chunk_frmr(struct svcxprt_rdma * xprt,struct svc_rqst * rqstp,struct svc_rdma_op_ctxt * head,int * page_no,u32 * page_offset,u32 rs_handle,u32 rs_length,u64 rs_offset,bool last)222 int rdma_read_chunk_frmr(struct svcxprt_rdma *xprt,
223 			 struct svc_rqst *rqstp,
224 			 struct svc_rdma_op_ctxt *head,
225 			 int *page_no,
226 			 u32 *page_offset,
227 			 u32 rs_handle,
228 			 u32 rs_length,
229 			 u64 rs_offset,
230 			 bool last)
231 {
232 	struct ib_send_wr read_wr;
233 	struct ib_send_wr inv_wr;
234 	struct ib_send_wr fastreg_wr;
235 	u8 key;
236 	int pages_needed = PAGE_ALIGN(*page_offset + rs_length) >> PAGE_SHIFT;
237 	struct svc_rdma_op_ctxt *ctxt = svc_rdma_get_context(xprt);
238 	struct svc_rdma_fastreg_mr *frmr = svc_rdma_get_frmr(xprt);
239 	int ret, read, pno;
240 	u32 pg_off = *page_offset;
241 	u32 pg_no = *page_no;
242 
243 	if (IS_ERR(frmr))
244 		return -ENOMEM;
245 
246 	ctxt->direction = DMA_FROM_DEVICE;
247 	ctxt->frmr = frmr;
248 	pages_needed = min_t(int, pages_needed, xprt->sc_frmr_pg_list_len);
249 	read = min_t(int, (pages_needed << PAGE_SHIFT) - *page_offset,
250 		     rs_length);
251 
252 	frmr->kva = page_address(rqstp->rq_arg.pages[pg_no]);
253 	frmr->direction = DMA_FROM_DEVICE;
254 	frmr->access_flags = (IB_ACCESS_LOCAL_WRITE|IB_ACCESS_REMOTE_WRITE);
255 	frmr->map_len = pages_needed << PAGE_SHIFT;
256 	frmr->page_list_len = pages_needed;
257 
258 	for (pno = 0; pno < pages_needed; pno++) {
259 		int len = min_t(int, rs_length, PAGE_SIZE - pg_off);
260 
261 		head->arg.pages[pg_no] = rqstp->rq_arg.pages[pg_no];
262 		head->arg.page_len += len;
263 		head->arg.len += len;
264 		if (!pg_off)
265 			head->count++;
266 		rqstp->rq_respages = &rqstp->rq_arg.pages[pg_no+1];
267 		rqstp->rq_next_page = rqstp->rq_respages + 1;
268 		frmr->page_list->page_list[pno] =
269 			ib_dma_map_page(xprt->sc_cm_id->device,
270 					head->arg.pages[pg_no], 0,
271 					PAGE_SIZE, DMA_FROM_DEVICE);
272 		ret = ib_dma_mapping_error(xprt->sc_cm_id->device,
273 					   frmr->page_list->page_list[pno]);
274 		if (ret)
275 			goto err;
276 		atomic_inc(&xprt->sc_dma_used);
277 
278 		/* adjust offset and wrap to next page if needed */
279 		pg_off += len;
280 		if (pg_off == PAGE_SIZE) {
281 			pg_off = 0;
282 			pg_no++;
283 		}
284 		rs_length -= len;
285 	}
286 
287 	if (last && rs_length == 0)
288 		set_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);
289 	else
290 		clear_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);
291 
292 	/* Bump the key */
293 	key = (u8)(frmr->mr->lkey & 0x000000FF);
294 	ib_update_fast_reg_key(frmr->mr, ++key);
295 
296 	ctxt->sge[0].addr = (unsigned long)frmr->kva + *page_offset;
297 	ctxt->sge[0].lkey = frmr->mr->lkey;
298 	ctxt->sge[0].length = read;
299 	ctxt->count = 1;
300 	ctxt->read_hdr = head;
301 
302 	/* Prepare FASTREG WR */
303 	memset(&fastreg_wr, 0, sizeof(fastreg_wr));
304 	fastreg_wr.opcode = IB_WR_FAST_REG_MR;
305 	fastreg_wr.send_flags = IB_SEND_SIGNALED;
306 	fastreg_wr.wr.fast_reg.iova_start = (unsigned long)frmr->kva;
307 	fastreg_wr.wr.fast_reg.page_list = frmr->page_list;
308 	fastreg_wr.wr.fast_reg.page_list_len = frmr->page_list_len;
309 	fastreg_wr.wr.fast_reg.page_shift = PAGE_SHIFT;
310 	fastreg_wr.wr.fast_reg.length = frmr->map_len;
311 	fastreg_wr.wr.fast_reg.access_flags = frmr->access_flags;
312 	fastreg_wr.wr.fast_reg.rkey = frmr->mr->lkey;
313 	fastreg_wr.next = &read_wr;
314 
315 	/* Prepare RDMA_READ */
316 	memset(&read_wr, 0, sizeof(read_wr));
317 	read_wr.send_flags = IB_SEND_SIGNALED;
318 	read_wr.wr.rdma.rkey = rs_handle;
319 	read_wr.wr.rdma.remote_addr = rs_offset;
320 	read_wr.sg_list = ctxt->sge;
321 	read_wr.num_sge = 1;
322 	if (xprt->sc_dev_caps & SVCRDMA_DEVCAP_READ_W_INV) {
323 		read_wr.opcode = IB_WR_RDMA_READ_WITH_INV;
324 		read_wr.wr_id = (unsigned long)ctxt;
325 		read_wr.ex.invalidate_rkey = ctxt->frmr->mr->lkey;
326 	} else {
327 		read_wr.opcode = IB_WR_RDMA_READ;
328 		read_wr.next = &inv_wr;
329 		/* Prepare invalidate */
330 		memset(&inv_wr, 0, sizeof(inv_wr));
331 		inv_wr.wr_id = (unsigned long)ctxt;
332 		inv_wr.opcode = IB_WR_LOCAL_INV;
333 		inv_wr.send_flags = IB_SEND_SIGNALED | IB_SEND_FENCE;
334 		inv_wr.ex.invalidate_rkey = frmr->mr->lkey;
335 	}
336 	ctxt->wr_op = read_wr.opcode;
337 
338 	/* Post the chain */
339 	ret = svc_rdma_send(xprt, &fastreg_wr);
340 	if (ret) {
341 		pr_err("svcrdma: Error %d posting RDMA_READ\n", ret);
342 		set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
343 		goto err;
344 	}
345 
346 	/* return current location in page array */
347 	*page_no = pg_no;
348 	*page_offset = pg_off;
349 	ret = read;
350 	atomic_inc(&rdma_stat_read);
351 	return ret;
352  err:
353 	svc_rdma_unmap_dma(ctxt);
354 	svc_rdma_put_context(ctxt, 0);
355 	svc_rdma_put_frmr(xprt, frmr);
356 	return ret;
357 }
358 
359 static unsigned int
rdma_rcl_chunk_count(struct rpcrdma_read_chunk * ch)360 rdma_rcl_chunk_count(struct rpcrdma_read_chunk *ch)
361 {
362 	unsigned int count;
363 
364 	for (count = 0; ch->rc_discrim != xdr_zero; ch++)
365 		count++;
366 	return count;
367 }
368 
369 /* If there was additional inline content, append it to the end of arg.pages.
370  * Tail copy has to be done after the reader function has determined how many
371  * pages are needed for RDMA READ.
372  */
373 static int
rdma_copy_tail(struct svc_rqst * rqstp,struct svc_rdma_op_ctxt * head,u32 position,u32 byte_count,u32 page_offset,int page_no)374 rdma_copy_tail(struct svc_rqst *rqstp, struct svc_rdma_op_ctxt *head,
375 	       u32 position, u32 byte_count, u32 page_offset, int page_no)
376 {
377 	char *srcp, *destp;
378 	int ret;
379 
380 	ret = 0;
381 	srcp = head->arg.head[0].iov_base + position;
382 	byte_count = head->arg.head[0].iov_len - position;
383 	if (byte_count > PAGE_SIZE) {
384 		dprintk("svcrdma: large tail unsupported\n");
385 		return 0;
386 	}
387 
388 	/* Fit as much of the tail on the current page as possible */
389 	if (page_offset != PAGE_SIZE) {
390 		destp = page_address(rqstp->rq_arg.pages[page_no]);
391 		destp += page_offset;
392 		while (byte_count--) {
393 			*destp++ = *srcp++;
394 			page_offset++;
395 			if (page_offset == PAGE_SIZE && byte_count)
396 				goto more;
397 		}
398 		goto done;
399 	}
400 
401 more:
402 	/* Fit the rest on the next page */
403 	page_no++;
404 	destp = page_address(rqstp->rq_arg.pages[page_no]);
405 	while (byte_count--)
406 		*destp++ = *srcp++;
407 
408 	rqstp->rq_respages = &rqstp->rq_arg.pages[page_no+1];
409 	rqstp->rq_next_page = rqstp->rq_respages + 1;
410 
411 done:
412 	byte_count = head->arg.head[0].iov_len - position;
413 	head->arg.page_len += byte_count;
414 	head->arg.len += byte_count;
415 	head->arg.buflen += byte_count;
416 	return 1;
417 }
418 
rdma_read_chunks(struct svcxprt_rdma * xprt,struct rpcrdma_msg * rmsgp,struct svc_rqst * rqstp,struct svc_rdma_op_ctxt * head)419 static int rdma_read_chunks(struct svcxprt_rdma *xprt,
420 			    struct rpcrdma_msg *rmsgp,
421 			    struct svc_rqst *rqstp,
422 			    struct svc_rdma_op_ctxt *head)
423 {
424 	int page_no, ret;
425 	struct rpcrdma_read_chunk *ch;
426 	u32 handle, page_offset, byte_count;
427 	u32 position;
428 	u64 rs_offset;
429 	bool last;
430 
431 	/* If no read list is present, return 0 */
432 	ch = svc_rdma_get_read_chunk(rmsgp);
433 	if (!ch)
434 		return 0;
435 
436 	if (rdma_rcl_chunk_count(ch) > RPCSVC_MAXPAGES)
437 		return -EINVAL;
438 
439 	/* The request is completed when the RDMA_READs complete. The
440 	 * head context keeps all the pages that comprise the
441 	 * request.
442 	 */
443 	head->arg.head[0] = rqstp->rq_arg.head[0];
444 	head->arg.tail[0] = rqstp->rq_arg.tail[0];
445 	head->hdr_count = head->count;
446 	head->arg.page_base = 0;
447 	head->arg.page_len = 0;
448 	head->arg.len = rqstp->rq_arg.len;
449 	head->arg.buflen = rqstp->rq_arg.buflen;
450 
451 	ch = (struct rpcrdma_read_chunk *)&rmsgp->rm_body.rm_chunks[0];
452 	position = be32_to_cpu(ch->rc_position);
453 
454 	/* RDMA_NOMSG: RDMA READ data should land just after RDMA RECV data */
455 	if (position == 0) {
456 		head->arg.pages = &head->pages[0];
457 		page_offset = head->byte_len;
458 	} else {
459 		head->arg.pages = &head->pages[head->count];
460 		page_offset = 0;
461 	}
462 
463 	ret = 0;
464 	page_no = 0;
465 	for (; ch->rc_discrim != xdr_zero; ch++) {
466 		if (be32_to_cpu(ch->rc_position) != position)
467 			goto err;
468 
469 		handle = be32_to_cpu(ch->rc_target.rs_handle),
470 		byte_count = be32_to_cpu(ch->rc_target.rs_length);
471 		xdr_decode_hyper((__be32 *)&ch->rc_target.rs_offset,
472 				 &rs_offset);
473 
474 		while (byte_count > 0) {
475 			last = (ch + 1)->rc_discrim == xdr_zero;
476 			ret = xprt->sc_reader(xprt, rqstp, head,
477 					      &page_no, &page_offset,
478 					      handle, byte_count,
479 					      rs_offset, last);
480 			if (ret < 0)
481 				goto err;
482 			byte_count -= ret;
483 			rs_offset += ret;
484 			head->arg.buflen += ret;
485 		}
486 	}
487 
488 	/* Read list may need XDR round-up (see RFC 5666, s. 3.7) */
489 	if (page_offset & 3) {
490 		u32 pad = 4 - (page_offset & 3);
491 
492 		head->arg.page_len += pad;
493 		head->arg.len += pad;
494 		head->arg.buflen += pad;
495 		page_offset += pad;
496 	}
497 
498 	ret = 1;
499 	if (position && position < head->arg.head[0].iov_len)
500 		ret = rdma_copy_tail(rqstp, head, position,
501 				     byte_count, page_offset, page_no);
502 	head->arg.head[0].iov_len = position;
503 	head->position = position;
504 
505  err:
506 	/* Detach arg pages. svc_recv will replenish them */
507 	for (page_no = 0;
508 	     &rqstp->rq_pages[page_no] < rqstp->rq_respages; page_no++)
509 		rqstp->rq_pages[page_no] = NULL;
510 
511 	return ret;
512 }
513 
rdma_read_complete(struct svc_rqst * rqstp,struct svc_rdma_op_ctxt * head)514 static int rdma_read_complete(struct svc_rqst *rqstp,
515 			      struct svc_rdma_op_ctxt *head)
516 {
517 	int page_no;
518 	int ret;
519 
520 	/* Copy RPC pages */
521 	for (page_no = 0; page_no < head->count; page_no++) {
522 		put_page(rqstp->rq_pages[page_no]);
523 		rqstp->rq_pages[page_no] = head->pages[page_no];
524 	}
525 
526 	/* Adjustments made for RDMA_NOMSG type requests */
527 	if (head->position == 0) {
528 		if (head->arg.len <= head->sge[0].length) {
529 			head->arg.head[0].iov_len = head->arg.len -
530 							head->byte_len;
531 			head->arg.page_len = 0;
532 		} else {
533 			head->arg.head[0].iov_len = head->sge[0].length -
534 								head->byte_len;
535 			head->arg.page_len = head->arg.len -
536 						head->sge[0].length;
537 		}
538 	}
539 
540 	/* Point rq_arg.pages past header */
541 	rqstp->rq_arg.pages = &rqstp->rq_pages[head->hdr_count];
542 	rqstp->rq_arg.page_len = head->arg.page_len;
543 	rqstp->rq_arg.page_base = head->arg.page_base;
544 
545 	/* rq_respages starts after the last arg page */
546 	rqstp->rq_respages = &rqstp->rq_arg.pages[page_no];
547 	rqstp->rq_next_page = rqstp->rq_respages + 1;
548 
549 	/* Rebuild rq_arg head and tail. */
550 	rqstp->rq_arg.head[0] = head->arg.head[0];
551 	rqstp->rq_arg.tail[0] = head->arg.tail[0];
552 	rqstp->rq_arg.len = head->arg.len;
553 	rqstp->rq_arg.buflen = head->arg.buflen;
554 
555 	/* Free the context */
556 	svc_rdma_put_context(head, 0);
557 
558 	/* XXX: What should this be? */
559 	rqstp->rq_prot = IPPROTO_MAX;
560 	svc_xprt_copy_addrs(rqstp, rqstp->rq_xprt);
561 
562 	ret = rqstp->rq_arg.head[0].iov_len
563 		+ rqstp->rq_arg.page_len
564 		+ rqstp->rq_arg.tail[0].iov_len;
565 	dprintk("svcrdma: deferred read ret=%d, rq_arg.len=%u, "
566 		"rq_arg.head[0].iov_base=%p, rq_arg.head[0].iov_len=%zu\n",
567 		ret, rqstp->rq_arg.len,	rqstp->rq_arg.head[0].iov_base,
568 		rqstp->rq_arg.head[0].iov_len);
569 
570 	return ret;
571 }
572 
573 /*
574  * Set up the rqstp thread context to point to the RQ buffer. If
575  * necessary, pull additional data from the client with an RDMA_READ
576  * request.
577  */
svc_rdma_recvfrom(struct svc_rqst * rqstp)578 int svc_rdma_recvfrom(struct svc_rqst *rqstp)
579 {
580 	struct svc_xprt *xprt = rqstp->rq_xprt;
581 	struct svcxprt_rdma *rdma_xprt =
582 		container_of(xprt, struct svcxprt_rdma, sc_xprt);
583 	struct svc_rdma_op_ctxt *ctxt = NULL;
584 	struct rpcrdma_msg *rmsgp;
585 	int ret = 0;
586 	int len;
587 
588 	dprintk("svcrdma: rqstp=%p\n", rqstp);
589 
590 	spin_lock_bh(&rdma_xprt->sc_rq_dto_lock);
591 	if (!list_empty(&rdma_xprt->sc_read_complete_q)) {
592 		ctxt = list_entry(rdma_xprt->sc_read_complete_q.next,
593 				  struct svc_rdma_op_ctxt,
594 				  dto_q);
595 		list_del_init(&ctxt->dto_q);
596 		spin_unlock_bh(&rdma_xprt->sc_rq_dto_lock);
597 		return rdma_read_complete(rqstp, ctxt);
598 	} else if (!list_empty(&rdma_xprt->sc_rq_dto_q)) {
599 		ctxt = list_entry(rdma_xprt->sc_rq_dto_q.next,
600 				  struct svc_rdma_op_ctxt,
601 				  dto_q);
602 		list_del_init(&ctxt->dto_q);
603 	} else {
604 		atomic_inc(&rdma_stat_rq_starve);
605 		clear_bit(XPT_DATA, &xprt->xpt_flags);
606 		ctxt = NULL;
607 	}
608 	spin_unlock_bh(&rdma_xprt->sc_rq_dto_lock);
609 	if (!ctxt) {
610 		/* This is the EAGAIN path. The svc_recv routine will
611 		 * return -EAGAIN, the nfsd thread will go to call into
612 		 * svc_recv again and we shouldn't be on the active
613 		 * transport list
614 		 */
615 		if (test_bit(XPT_CLOSE, &xprt->xpt_flags))
616 			goto close_out;
617 
618 		goto out;
619 	}
620 	dprintk("svcrdma: processing ctxt=%p on xprt=%p, rqstp=%p, status=%d\n",
621 		ctxt, rdma_xprt, rqstp, ctxt->wc_status);
622 	atomic_inc(&rdma_stat_recv);
623 
624 	/* Build up the XDR from the receive buffers. */
625 	rdma_build_arg_xdr(rqstp, ctxt, ctxt->byte_len);
626 
627 	/* Decode the RDMA header. */
628 	len = svc_rdma_xdr_decode_req(&rmsgp, rqstp);
629 	rqstp->rq_xprt_hlen = len;
630 
631 	/* If the request is invalid, reply with an error */
632 	if (len < 0) {
633 		if (len == -ENOSYS)
634 			svc_rdma_send_error(rdma_xprt, rmsgp, ERR_VERS);
635 		goto close_out;
636 	}
637 
638 	/* Read read-list data. */
639 	ret = rdma_read_chunks(rdma_xprt, rmsgp, rqstp, ctxt);
640 	if (ret > 0) {
641 		/* read-list posted, defer until data received from client. */
642 		goto defer;
643 	} else if (ret < 0) {
644 		/* Post of read-list failed, free context. */
645 		svc_rdma_put_context(ctxt, 1);
646 		return 0;
647 	}
648 
649 	ret = rqstp->rq_arg.head[0].iov_len
650 		+ rqstp->rq_arg.page_len
651 		+ rqstp->rq_arg.tail[0].iov_len;
652 	svc_rdma_put_context(ctxt, 0);
653  out:
654 	dprintk("svcrdma: ret=%d, rq_arg.len=%u, "
655 		"rq_arg.head[0].iov_base=%p, rq_arg.head[0].iov_len=%zd\n",
656 		ret, rqstp->rq_arg.len,
657 		rqstp->rq_arg.head[0].iov_base,
658 		rqstp->rq_arg.head[0].iov_len);
659 	rqstp->rq_prot = IPPROTO_MAX;
660 	svc_xprt_copy_addrs(rqstp, xprt);
661 	return ret;
662 
663  close_out:
664 	if (ctxt)
665 		svc_rdma_put_context(ctxt, 1);
666 	dprintk("svcrdma: transport %p is closing\n", xprt);
667 	/*
668 	 * Set the close bit and enqueue it. svc_recv will see the
669 	 * close bit and call svc_xprt_delete
670 	 */
671 	set_bit(XPT_CLOSE, &xprt->xpt_flags);
672 defer:
673 	return 0;
674 }
675