1 /*
2  * Copyright (c) 2014 Open Grid Computing, Inc. All rights reserved.
3  * Copyright (c) 2005-2007 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/svc_xprt.h>
44 #include <linux/sunrpc/debug.h>
45 #include <linux/sunrpc/rpc_rdma.h>
46 #include <linux/interrupt.h>
47 #include <linux/sched.h>
48 #include <linux/slab.h>
49 #include <linux/spinlock.h>
50 #include <linux/workqueue.h>
51 #include <rdma/ib_verbs.h>
52 #include <rdma/rdma_cm.h>
53 #include <linux/sunrpc/svc_rdma.h>
54 #include <linux/export.h>
55 #include "xprt_rdma.h"
56 
57 #define RPCDBG_FACILITY	RPCDBG_SVCXPRT
58 
59 static struct svcxprt_rdma *rdma_create_xprt(struct svc_serv *, int);
60 static struct svc_xprt *svc_rdma_create(struct svc_serv *serv,
61 					struct net *net,
62 					struct sockaddr *sa, int salen,
63 					int flags);
64 static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt);
65 static void svc_rdma_release_rqst(struct svc_rqst *);
66 static void svc_rdma_detach(struct svc_xprt *xprt);
67 static void svc_rdma_free(struct svc_xprt *xprt);
68 static int svc_rdma_has_wspace(struct svc_xprt *xprt);
69 static int svc_rdma_secure_port(struct svc_rqst *);
70 
71 static struct svc_xprt_ops svc_rdma_ops = {
72 	.xpo_create = svc_rdma_create,
73 	.xpo_recvfrom = svc_rdma_recvfrom,
74 	.xpo_sendto = svc_rdma_sendto,
75 	.xpo_release_rqst = svc_rdma_release_rqst,
76 	.xpo_detach = svc_rdma_detach,
77 	.xpo_free = svc_rdma_free,
78 	.xpo_prep_reply_hdr = svc_rdma_prep_reply_hdr,
79 	.xpo_has_wspace = svc_rdma_has_wspace,
80 	.xpo_accept = svc_rdma_accept,
81 	.xpo_secure_port = svc_rdma_secure_port,
82 };
83 
84 struct svc_xprt_class svc_rdma_class = {
85 	.xcl_name = "rdma",
86 	.xcl_owner = THIS_MODULE,
87 	.xcl_ops = &svc_rdma_ops,
88 	.xcl_max_payload = RPCSVC_MAXPAYLOAD_RDMA,
89 	.xcl_ident = XPRT_TRANSPORT_RDMA,
90 };
91 
92 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
93 static struct svc_xprt *svc_rdma_bc_create(struct svc_serv *, struct net *,
94 					   struct sockaddr *, int, int);
95 static void svc_rdma_bc_detach(struct svc_xprt *);
96 static void svc_rdma_bc_free(struct svc_xprt *);
97 
98 static struct svc_xprt_ops svc_rdma_bc_ops = {
99 	.xpo_create = svc_rdma_bc_create,
100 	.xpo_detach = svc_rdma_bc_detach,
101 	.xpo_free = svc_rdma_bc_free,
102 	.xpo_prep_reply_hdr = svc_rdma_prep_reply_hdr,
103 	.xpo_secure_port = svc_rdma_secure_port,
104 };
105 
106 struct svc_xprt_class svc_rdma_bc_class = {
107 	.xcl_name = "rdma-bc",
108 	.xcl_owner = THIS_MODULE,
109 	.xcl_ops = &svc_rdma_bc_ops,
110 	.xcl_max_payload = (1024 - RPCRDMA_HDRLEN_MIN)
111 };
112 
113 static struct svc_xprt *svc_rdma_bc_create(struct svc_serv *serv,
114 					   struct net *net,
115 					   struct sockaddr *sa, int salen,
116 					   int flags)
117 {
118 	struct svcxprt_rdma *cma_xprt;
119 	struct svc_xprt *xprt;
120 
121 	cma_xprt = rdma_create_xprt(serv, 0);
122 	if (!cma_xprt)
123 		return ERR_PTR(-ENOMEM);
124 	xprt = &cma_xprt->sc_xprt;
125 
126 	svc_xprt_init(net, &svc_rdma_bc_class, xprt, serv);
127 	serv->sv_bc_xprt = xprt;
128 
129 	dprintk("svcrdma: %s(%p)\n", __func__, xprt);
130 	return xprt;
131 }
132 
133 static void svc_rdma_bc_detach(struct svc_xprt *xprt)
134 {
135 	dprintk("svcrdma: %s(%p)\n", __func__, xprt);
136 }
137 
138 static void svc_rdma_bc_free(struct svc_xprt *xprt)
139 {
140 	struct svcxprt_rdma *rdma =
141 		container_of(xprt, struct svcxprt_rdma, sc_xprt);
142 
143 	dprintk("svcrdma: %s(%p)\n", __func__, xprt);
144 	if (xprt)
145 		kfree(rdma);
146 }
147 #endif	/* CONFIG_SUNRPC_BACKCHANNEL */
148 
149 static struct svc_rdma_op_ctxt *alloc_ctxt(struct svcxprt_rdma *xprt,
150 					   gfp_t flags)
151 {
152 	struct svc_rdma_op_ctxt *ctxt;
153 
154 	ctxt = kmalloc(sizeof(*ctxt), flags);
155 	if (ctxt) {
156 		ctxt->xprt = xprt;
157 		INIT_LIST_HEAD(&ctxt->free);
158 		INIT_LIST_HEAD(&ctxt->dto_q);
159 	}
160 	return ctxt;
161 }
162 
163 static bool svc_rdma_prealloc_ctxts(struct svcxprt_rdma *xprt)
164 {
165 	unsigned int i;
166 
167 	/* Each RPC/RDMA credit can consume a number of send
168 	 * and receive WQEs. One ctxt is allocated for each.
169 	 */
170 	i = xprt->sc_sq_depth + xprt->sc_rq_depth;
171 
172 	while (i--) {
173 		struct svc_rdma_op_ctxt *ctxt;
174 
175 		ctxt = alloc_ctxt(xprt, GFP_KERNEL);
176 		if (!ctxt) {
177 			dprintk("svcrdma: No memory for RDMA ctxt\n");
178 			return false;
179 		}
180 		list_add(&ctxt->free, &xprt->sc_ctxts);
181 	}
182 	return true;
183 }
184 
185 struct svc_rdma_op_ctxt *svc_rdma_get_context(struct svcxprt_rdma *xprt)
186 {
187 	struct svc_rdma_op_ctxt *ctxt = NULL;
188 
189 	spin_lock_bh(&xprt->sc_ctxt_lock);
190 	xprt->sc_ctxt_used++;
191 	if (list_empty(&xprt->sc_ctxts))
192 		goto out_empty;
193 
194 	ctxt = list_first_entry(&xprt->sc_ctxts,
195 				struct svc_rdma_op_ctxt, free);
196 	list_del_init(&ctxt->free);
197 	spin_unlock_bh(&xprt->sc_ctxt_lock);
198 
199 out:
200 	ctxt->count = 0;
201 	ctxt->frmr = NULL;
202 	return ctxt;
203 
204 out_empty:
205 	/* Either pre-allocation missed the mark, or send
206 	 * queue accounting is broken.
207 	 */
208 	spin_unlock_bh(&xprt->sc_ctxt_lock);
209 
210 	ctxt = alloc_ctxt(xprt, GFP_NOIO);
211 	if (ctxt)
212 		goto out;
213 
214 	spin_lock_bh(&xprt->sc_ctxt_lock);
215 	xprt->sc_ctxt_used--;
216 	spin_unlock_bh(&xprt->sc_ctxt_lock);
217 	WARN_ONCE(1, "svcrdma: empty RDMA ctxt list?\n");
218 	return NULL;
219 }
220 
221 void svc_rdma_unmap_dma(struct svc_rdma_op_ctxt *ctxt)
222 {
223 	struct svcxprt_rdma *xprt = ctxt->xprt;
224 	int i;
225 	for (i = 0; i < ctxt->count && ctxt->sge[i].length; i++) {
226 		/*
227 		 * Unmap the DMA addr in the SGE if the lkey matches
228 		 * the local_dma_lkey, otherwise, ignore it since it is
229 		 * an FRMR lkey and will be unmapped later when the
230 		 * last WR that uses it completes.
231 		 */
232 		if (ctxt->sge[i].lkey == xprt->sc_pd->local_dma_lkey) {
233 			atomic_dec(&xprt->sc_dma_used);
234 			ib_dma_unmap_page(xprt->sc_cm_id->device,
235 					    ctxt->sge[i].addr,
236 					    ctxt->sge[i].length,
237 					    ctxt->direction);
238 		}
239 	}
240 }
241 
242 void svc_rdma_put_context(struct svc_rdma_op_ctxt *ctxt, int free_pages)
243 {
244 	struct svcxprt_rdma *xprt = ctxt->xprt;
245 	int i;
246 
247 	if (free_pages)
248 		for (i = 0; i < ctxt->count; i++)
249 			put_page(ctxt->pages[i]);
250 
251 	spin_lock_bh(&xprt->sc_ctxt_lock);
252 	xprt->sc_ctxt_used--;
253 	list_add(&ctxt->free, &xprt->sc_ctxts);
254 	spin_unlock_bh(&xprt->sc_ctxt_lock);
255 }
256 
257 static void svc_rdma_destroy_ctxts(struct svcxprt_rdma *xprt)
258 {
259 	while (!list_empty(&xprt->sc_ctxts)) {
260 		struct svc_rdma_op_ctxt *ctxt;
261 
262 		ctxt = list_first_entry(&xprt->sc_ctxts,
263 					struct svc_rdma_op_ctxt, free);
264 		list_del(&ctxt->free);
265 		kfree(ctxt);
266 	}
267 }
268 
269 static struct svc_rdma_req_map *alloc_req_map(gfp_t flags)
270 {
271 	struct svc_rdma_req_map *map;
272 
273 	map = kmalloc(sizeof(*map), flags);
274 	if (map)
275 		INIT_LIST_HEAD(&map->free);
276 	return map;
277 }
278 
279 static bool svc_rdma_prealloc_maps(struct svcxprt_rdma *xprt)
280 {
281 	unsigned int i;
282 
283 	/* One for each receive buffer on this connection. */
284 	i = xprt->sc_max_requests;
285 
286 	while (i--) {
287 		struct svc_rdma_req_map *map;
288 
289 		map = alloc_req_map(GFP_KERNEL);
290 		if (!map) {
291 			dprintk("svcrdma: No memory for request map\n");
292 			return false;
293 		}
294 		list_add(&map->free, &xprt->sc_maps);
295 	}
296 	return true;
297 }
298 
299 struct svc_rdma_req_map *svc_rdma_get_req_map(struct svcxprt_rdma *xprt)
300 {
301 	struct svc_rdma_req_map *map = NULL;
302 
303 	spin_lock(&xprt->sc_map_lock);
304 	if (list_empty(&xprt->sc_maps))
305 		goto out_empty;
306 
307 	map = list_first_entry(&xprt->sc_maps,
308 			       struct svc_rdma_req_map, free);
309 	list_del_init(&map->free);
310 	spin_unlock(&xprt->sc_map_lock);
311 
312 out:
313 	map->count = 0;
314 	return map;
315 
316 out_empty:
317 	spin_unlock(&xprt->sc_map_lock);
318 
319 	/* Pre-allocation amount was incorrect */
320 	map = alloc_req_map(GFP_NOIO);
321 	if (map)
322 		goto out;
323 
324 	WARN_ONCE(1, "svcrdma: empty request map list?\n");
325 	return NULL;
326 }
327 
328 void svc_rdma_put_req_map(struct svcxprt_rdma *xprt,
329 			  struct svc_rdma_req_map *map)
330 {
331 	spin_lock(&xprt->sc_map_lock);
332 	list_add(&map->free, &xprt->sc_maps);
333 	spin_unlock(&xprt->sc_map_lock);
334 }
335 
336 static void svc_rdma_destroy_maps(struct svcxprt_rdma *xprt)
337 {
338 	while (!list_empty(&xprt->sc_maps)) {
339 		struct svc_rdma_req_map *map;
340 
341 		map = list_first_entry(&xprt->sc_maps,
342 				       struct svc_rdma_req_map, free);
343 		list_del(&map->free);
344 		kfree(map);
345 	}
346 }
347 
348 /* QP event handler */
349 static void qp_event_handler(struct ib_event *event, void *context)
350 {
351 	struct svc_xprt *xprt = context;
352 
353 	switch (event->event) {
354 	/* These are considered benign events */
355 	case IB_EVENT_PATH_MIG:
356 	case IB_EVENT_COMM_EST:
357 	case IB_EVENT_SQ_DRAINED:
358 	case IB_EVENT_QP_LAST_WQE_REACHED:
359 		dprintk("svcrdma: QP event %s (%d) received for QP=%p\n",
360 			ib_event_msg(event->event), event->event,
361 			event->element.qp);
362 		break;
363 	/* These are considered fatal events */
364 	case IB_EVENT_PATH_MIG_ERR:
365 	case IB_EVENT_QP_FATAL:
366 	case IB_EVENT_QP_REQ_ERR:
367 	case IB_EVENT_QP_ACCESS_ERR:
368 	case IB_EVENT_DEVICE_FATAL:
369 	default:
370 		dprintk("svcrdma: QP ERROR event %s (%d) received for QP=%p, "
371 			"closing transport\n",
372 			ib_event_msg(event->event), event->event,
373 			event->element.qp);
374 		set_bit(XPT_CLOSE, &xprt->xpt_flags);
375 		break;
376 	}
377 }
378 
379 /**
380  * svc_rdma_wc_receive - Invoked by RDMA provider for each polled Receive WC
381  * @cq:        completion queue
382  * @wc:        completed WR
383  *
384  */
385 static void svc_rdma_wc_receive(struct ib_cq *cq, struct ib_wc *wc)
386 {
387 	struct svcxprt_rdma *xprt = cq->cq_context;
388 	struct ib_cqe *cqe = wc->wr_cqe;
389 	struct svc_rdma_op_ctxt *ctxt;
390 
391 	/* WARNING: Only wc->wr_cqe and wc->status are reliable */
392 	ctxt = container_of(cqe, struct svc_rdma_op_ctxt, cqe);
393 	ctxt->wc_status = wc->status;
394 	svc_rdma_unmap_dma(ctxt);
395 
396 	if (wc->status != IB_WC_SUCCESS)
397 		goto flushed;
398 
399 	/* All wc fields are now known to be valid */
400 	ctxt->byte_len = wc->byte_len;
401 	spin_lock(&xprt->sc_rq_dto_lock);
402 	list_add_tail(&ctxt->dto_q, &xprt->sc_rq_dto_q);
403 	spin_unlock(&xprt->sc_rq_dto_lock);
404 
405 	set_bit(XPT_DATA, &xprt->sc_xprt.xpt_flags);
406 	if (test_bit(RDMAXPRT_CONN_PENDING, &xprt->sc_flags))
407 		goto out;
408 	svc_xprt_enqueue(&xprt->sc_xprt);
409 	goto out;
410 
411 flushed:
412 	if (wc->status != IB_WC_WR_FLUSH_ERR)
413 		pr_warn("svcrdma: receive: %s (%u/0x%x)\n",
414 			ib_wc_status_msg(wc->status),
415 			wc->status, wc->vendor_err);
416 	set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
417 	svc_rdma_put_context(ctxt, 1);
418 
419 out:
420 	svc_xprt_put(&xprt->sc_xprt);
421 }
422 
423 static void svc_rdma_send_wc_common(struct svcxprt_rdma *xprt,
424 				    struct ib_wc *wc,
425 				    const char *opname)
426 {
427 	if (wc->status != IB_WC_SUCCESS)
428 		goto err;
429 
430 out:
431 	atomic_dec(&xprt->sc_sq_count);
432 	wake_up(&xprt->sc_send_wait);
433 	return;
434 
435 err:
436 	set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
437 	if (wc->status != IB_WC_WR_FLUSH_ERR)
438 		pr_err("svcrdma: %s: %s (%u/0x%x)\n",
439 		       opname, ib_wc_status_msg(wc->status),
440 		       wc->status, wc->vendor_err);
441 	goto out;
442 }
443 
444 static void svc_rdma_send_wc_common_put(struct ib_cq *cq, struct ib_wc *wc,
445 					const char *opname)
446 {
447 	struct svcxprt_rdma *xprt = cq->cq_context;
448 
449 	svc_rdma_send_wc_common(xprt, wc, opname);
450 	svc_xprt_put(&xprt->sc_xprt);
451 }
452 
453 /**
454  * svc_rdma_wc_send - Invoked by RDMA provider for each polled Send WC
455  * @cq:        completion queue
456  * @wc:        completed WR
457  *
458  */
459 void svc_rdma_wc_send(struct ib_cq *cq, struct ib_wc *wc)
460 {
461 	struct ib_cqe *cqe = wc->wr_cqe;
462 	struct svc_rdma_op_ctxt *ctxt;
463 
464 	svc_rdma_send_wc_common_put(cq, wc, "send");
465 
466 	ctxt = container_of(cqe, struct svc_rdma_op_ctxt, cqe);
467 	svc_rdma_unmap_dma(ctxt);
468 	svc_rdma_put_context(ctxt, 1);
469 }
470 
471 /**
472  * svc_rdma_wc_write - Invoked by RDMA provider for each polled Write WC
473  * @cq:        completion queue
474  * @wc:        completed WR
475  *
476  */
477 void svc_rdma_wc_write(struct ib_cq *cq, struct ib_wc *wc)
478 {
479 	struct ib_cqe *cqe = wc->wr_cqe;
480 	struct svc_rdma_op_ctxt *ctxt;
481 
482 	svc_rdma_send_wc_common_put(cq, wc, "write");
483 
484 	ctxt = container_of(cqe, struct svc_rdma_op_ctxt, cqe);
485 	svc_rdma_unmap_dma(ctxt);
486 	svc_rdma_put_context(ctxt, 0);
487 }
488 
489 /**
490  * svc_rdma_wc_reg - Invoked by RDMA provider for each polled FASTREG WC
491  * @cq:        completion queue
492  * @wc:        completed WR
493  *
494  */
495 void svc_rdma_wc_reg(struct ib_cq *cq, struct ib_wc *wc)
496 {
497 	svc_rdma_send_wc_common_put(cq, wc, "fastreg");
498 }
499 
500 /**
501  * svc_rdma_wc_read - Invoked by RDMA provider for each polled Read WC
502  * @cq:        completion queue
503  * @wc:        completed WR
504  *
505  */
506 void svc_rdma_wc_read(struct ib_cq *cq, struct ib_wc *wc)
507 {
508 	struct svcxprt_rdma *xprt = cq->cq_context;
509 	struct ib_cqe *cqe = wc->wr_cqe;
510 	struct svc_rdma_op_ctxt *ctxt;
511 
512 	svc_rdma_send_wc_common(xprt, wc, "read");
513 
514 	ctxt = container_of(cqe, struct svc_rdma_op_ctxt, cqe);
515 	svc_rdma_unmap_dma(ctxt);
516 	svc_rdma_put_frmr(xprt, ctxt->frmr);
517 
518 	if (test_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags)) {
519 		struct svc_rdma_op_ctxt *read_hdr;
520 
521 		read_hdr = ctxt->read_hdr;
522 		spin_lock(&xprt->sc_rq_dto_lock);
523 		list_add_tail(&read_hdr->dto_q,
524 			      &xprt->sc_read_complete_q);
525 		spin_unlock(&xprt->sc_rq_dto_lock);
526 
527 		set_bit(XPT_DATA, &xprt->sc_xprt.xpt_flags);
528 		svc_xprt_enqueue(&xprt->sc_xprt);
529 	}
530 
531 	svc_rdma_put_context(ctxt, 0);
532 	svc_xprt_put(&xprt->sc_xprt);
533 }
534 
535 /**
536  * svc_rdma_wc_inv - Invoked by RDMA provider for each polled LOCAL_INV WC
537  * @cq:        completion queue
538  * @wc:        completed WR
539  *
540  */
541 void svc_rdma_wc_inv(struct ib_cq *cq, struct ib_wc *wc)
542 {
543 	svc_rdma_send_wc_common_put(cq, wc, "localInv");
544 }
545 
546 static struct svcxprt_rdma *rdma_create_xprt(struct svc_serv *serv,
547 					     int listener)
548 {
549 	struct svcxprt_rdma *cma_xprt = kzalloc(sizeof *cma_xprt, GFP_KERNEL);
550 
551 	if (!cma_xprt)
552 		return NULL;
553 	svc_xprt_init(&init_net, &svc_rdma_class, &cma_xprt->sc_xprt, serv);
554 	INIT_LIST_HEAD(&cma_xprt->sc_accept_q);
555 	INIT_LIST_HEAD(&cma_xprt->sc_dto_q);
556 	INIT_LIST_HEAD(&cma_xprt->sc_rq_dto_q);
557 	INIT_LIST_HEAD(&cma_xprt->sc_read_complete_q);
558 	INIT_LIST_HEAD(&cma_xprt->sc_frmr_q);
559 	INIT_LIST_HEAD(&cma_xprt->sc_ctxts);
560 	INIT_LIST_HEAD(&cma_xprt->sc_maps);
561 	init_waitqueue_head(&cma_xprt->sc_send_wait);
562 
563 	spin_lock_init(&cma_xprt->sc_lock);
564 	spin_lock_init(&cma_xprt->sc_rq_dto_lock);
565 	spin_lock_init(&cma_xprt->sc_frmr_q_lock);
566 	spin_lock_init(&cma_xprt->sc_ctxt_lock);
567 	spin_lock_init(&cma_xprt->sc_map_lock);
568 
569 	if (listener)
570 		set_bit(XPT_LISTENER, &cma_xprt->sc_xprt.xpt_flags);
571 
572 	return cma_xprt;
573 }
574 
575 int svc_rdma_post_recv(struct svcxprt_rdma *xprt, gfp_t flags)
576 {
577 	struct ib_recv_wr recv_wr, *bad_recv_wr;
578 	struct svc_rdma_op_ctxt *ctxt;
579 	struct page *page;
580 	dma_addr_t pa;
581 	int sge_no;
582 	int buflen;
583 	int ret;
584 
585 	ctxt = svc_rdma_get_context(xprt);
586 	buflen = 0;
587 	ctxt->direction = DMA_FROM_DEVICE;
588 	ctxt->cqe.done = svc_rdma_wc_receive;
589 	for (sge_no = 0; buflen < xprt->sc_max_req_size; sge_no++) {
590 		if (sge_no >= xprt->sc_max_sge) {
591 			pr_err("svcrdma: Too many sges (%d)\n", sge_no);
592 			goto err_put_ctxt;
593 		}
594 		page = alloc_page(flags);
595 		if (!page)
596 			goto err_put_ctxt;
597 		ctxt->pages[sge_no] = page;
598 		pa = ib_dma_map_page(xprt->sc_cm_id->device,
599 				     page, 0, PAGE_SIZE,
600 				     DMA_FROM_DEVICE);
601 		if (ib_dma_mapping_error(xprt->sc_cm_id->device, pa))
602 			goto err_put_ctxt;
603 		atomic_inc(&xprt->sc_dma_used);
604 		ctxt->sge[sge_no].addr = pa;
605 		ctxt->sge[sge_no].length = PAGE_SIZE;
606 		ctxt->sge[sge_no].lkey = xprt->sc_pd->local_dma_lkey;
607 		ctxt->count = sge_no + 1;
608 		buflen += PAGE_SIZE;
609 	}
610 	recv_wr.next = NULL;
611 	recv_wr.sg_list = &ctxt->sge[0];
612 	recv_wr.num_sge = ctxt->count;
613 	recv_wr.wr_cqe = &ctxt->cqe;
614 
615 	svc_xprt_get(&xprt->sc_xprt);
616 	ret = ib_post_recv(xprt->sc_qp, &recv_wr, &bad_recv_wr);
617 	if (ret) {
618 		svc_rdma_unmap_dma(ctxt);
619 		svc_rdma_put_context(ctxt, 1);
620 		svc_xprt_put(&xprt->sc_xprt);
621 	}
622 	return ret;
623 
624  err_put_ctxt:
625 	svc_rdma_unmap_dma(ctxt);
626 	svc_rdma_put_context(ctxt, 1);
627 	return -ENOMEM;
628 }
629 
630 int svc_rdma_repost_recv(struct svcxprt_rdma *xprt, gfp_t flags)
631 {
632 	int ret = 0;
633 
634 	ret = svc_rdma_post_recv(xprt, flags);
635 	if (ret) {
636 		pr_err("svcrdma: could not post a receive buffer, err=%d.\n",
637 		       ret);
638 		pr_err("svcrdma: closing transport %p.\n", xprt);
639 		set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
640 		ret = -ENOTCONN;
641 	}
642 	return ret;
643 }
644 
645 /*
646  * This function handles the CONNECT_REQUEST event on a listening
647  * endpoint. It is passed the cma_id for the _new_ connection. The context in
648  * this cma_id is inherited from the listening cma_id and is the svc_xprt
649  * structure for the listening endpoint.
650  *
651  * This function creates a new xprt for the new connection and enqueues it on
652  * the accept queue for the listent xprt. When the listen thread is kicked, it
653  * will call the recvfrom method on the listen xprt which will accept the new
654  * connection.
655  */
656 static void handle_connect_req(struct rdma_cm_id *new_cma_id, size_t client_ird)
657 {
658 	struct svcxprt_rdma *listen_xprt = new_cma_id->context;
659 	struct svcxprt_rdma *newxprt;
660 	struct sockaddr *sa;
661 
662 	/* Create a new transport */
663 	newxprt = rdma_create_xprt(listen_xprt->sc_xprt.xpt_server, 0);
664 	if (!newxprt) {
665 		dprintk("svcrdma: failed to create new transport\n");
666 		return;
667 	}
668 	newxprt->sc_cm_id = new_cma_id;
669 	new_cma_id->context = newxprt;
670 	dprintk("svcrdma: Creating newxprt=%p, cm_id=%p, listenxprt=%p\n",
671 		newxprt, newxprt->sc_cm_id, listen_xprt);
672 
673 	/* Save client advertised inbound read limit for use later in accept. */
674 	newxprt->sc_ord = client_ird;
675 
676 	/* Set the local and remote addresses in the transport */
677 	sa = (struct sockaddr *)&newxprt->sc_cm_id->route.addr.dst_addr;
678 	svc_xprt_set_remote(&newxprt->sc_xprt, sa, svc_addr_len(sa));
679 	sa = (struct sockaddr *)&newxprt->sc_cm_id->route.addr.src_addr;
680 	svc_xprt_set_local(&newxprt->sc_xprt, sa, svc_addr_len(sa));
681 
682 	/*
683 	 * Enqueue the new transport on the accept queue of the listening
684 	 * transport
685 	 */
686 	spin_lock_bh(&listen_xprt->sc_lock);
687 	list_add_tail(&newxprt->sc_accept_q, &listen_xprt->sc_accept_q);
688 	spin_unlock_bh(&listen_xprt->sc_lock);
689 
690 	set_bit(XPT_CONN, &listen_xprt->sc_xprt.xpt_flags);
691 	svc_xprt_enqueue(&listen_xprt->sc_xprt);
692 }
693 
694 /*
695  * Handles events generated on the listening endpoint. These events will be
696  * either be incoming connect requests or adapter removal  events.
697  */
698 static int rdma_listen_handler(struct rdma_cm_id *cma_id,
699 			       struct rdma_cm_event *event)
700 {
701 	struct svcxprt_rdma *xprt = cma_id->context;
702 	int ret = 0;
703 
704 	switch (event->event) {
705 	case RDMA_CM_EVENT_CONNECT_REQUEST:
706 		dprintk("svcrdma: Connect request on cma_id=%p, xprt = %p, "
707 			"event = %s (%d)\n", cma_id, cma_id->context,
708 			rdma_event_msg(event->event), event->event);
709 		handle_connect_req(cma_id,
710 				   event->param.conn.initiator_depth);
711 		break;
712 
713 	case RDMA_CM_EVENT_ESTABLISHED:
714 		/* Accept complete */
715 		dprintk("svcrdma: Connection completed on LISTEN xprt=%p, "
716 			"cm_id=%p\n", xprt, cma_id);
717 		break;
718 
719 	case RDMA_CM_EVENT_DEVICE_REMOVAL:
720 		dprintk("svcrdma: Device removal xprt=%p, cm_id=%p\n",
721 			xprt, cma_id);
722 		if (xprt)
723 			set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
724 		break;
725 
726 	default:
727 		dprintk("svcrdma: Unexpected event on listening endpoint %p, "
728 			"event = %s (%d)\n", cma_id,
729 			rdma_event_msg(event->event), event->event);
730 		break;
731 	}
732 
733 	return ret;
734 }
735 
736 static int rdma_cma_handler(struct rdma_cm_id *cma_id,
737 			    struct rdma_cm_event *event)
738 {
739 	struct svc_xprt *xprt = cma_id->context;
740 	struct svcxprt_rdma *rdma =
741 		container_of(xprt, struct svcxprt_rdma, sc_xprt);
742 	switch (event->event) {
743 	case RDMA_CM_EVENT_ESTABLISHED:
744 		/* Accept complete */
745 		svc_xprt_get(xprt);
746 		dprintk("svcrdma: Connection completed on DTO xprt=%p, "
747 			"cm_id=%p\n", xprt, cma_id);
748 		clear_bit(RDMAXPRT_CONN_PENDING, &rdma->sc_flags);
749 		svc_xprt_enqueue(xprt);
750 		break;
751 	case RDMA_CM_EVENT_DISCONNECTED:
752 		dprintk("svcrdma: Disconnect on DTO xprt=%p, cm_id=%p\n",
753 			xprt, cma_id);
754 		if (xprt) {
755 			set_bit(XPT_CLOSE, &xprt->xpt_flags);
756 			svc_xprt_enqueue(xprt);
757 			svc_xprt_put(xprt);
758 		}
759 		break;
760 	case RDMA_CM_EVENT_DEVICE_REMOVAL:
761 		dprintk("svcrdma: Device removal cma_id=%p, xprt = %p, "
762 			"event = %s (%d)\n", cma_id, xprt,
763 			rdma_event_msg(event->event), event->event);
764 		if (xprt) {
765 			set_bit(XPT_CLOSE, &xprt->xpt_flags);
766 			svc_xprt_enqueue(xprt);
767 			svc_xprt_put(xprt);
768 		}
769 		break;
770 	default:
771 		dprintk("svcrdma: Unexpected event on DTO endpoint %p, "
772 			"event = %s (%d)\n", cma_id,
773 			rdma_event_msg(event->event), event->event);
774 		break;
775 	}
776 	return 0;
777 }
778 
779 /*
780  * Create a listening RDMA service endpoint.
781  */
782 static struct svc_xprt *svc_rdma_create(struct svc_serv *serv,
783 					struct net *net,
784 					struct sockaddr *sa, int salen,
785 					int flags)
786 {
787 	struct rdma_cm_id *listen_id;
788 	struct svcxprt_rdma *cma_xprt;
789 	int ret;
790 
791 	dprintk("svcrdma: Creating RDMA socket\n");
792 	if (sa->sa_family != AF_INET) {
793 		dprintk("svcrdma: Address family %d is not supported.\n", sa->sa_family);
794 		return ERR_PTR(-EAFNOSUPPORT);
795 	}
796 	cma_xprt = rdma_create_xprt(serv, 1);
797 	if (!cma_xprt)
798 		return ERR_PTR(-ENOMEM);
799 
800 	listen_id = rdma_create_id(&init_net, rdma_listen_handler, cma_xprt,
801 				   RDMA_PS_TCP, IB_QPT_RC);
802 	if (IS_ERR(listen_id)) {
803 		ret = PTR_ERR(listen_id);
804 		dprintk("svcrdma: rdma_create_id failed = %d\n", ret);
805 		goto err0;
806 	}
807 
808 	ret = rdma_bind_addr(listen_id, sa);
809 	if (ret) {
810 		dprintk("svcrdma: rdma_bind_addr failed = %d\n", ret);
811 		goto err1;
812 	}
813 	cma_xprt->sc_cm_id = listen_id;
814 
815 	ret = rdma_listen(listen_id, RPCRDMA_LISTEN_BACKLOG);
816 	if (ret) {
817 		dprintk("svcrdma: rdma_listen failed = %d\n", ret);
818 		goto err1;
819 	}
820 
821 	/*
822 	 * We need to use the address from the cm_id in case the
823 	 * caller specified 0 for the port number.
824 	 */
825 	sa = (struct sockaddr *)&cma_xprt->sc_cm_id->route.addr.src_addr;
826 	svc_xprt_set_local(&cma_xprt->sc_xprt, sa, salen);
827 
828 	return &cma_xprt->sc_xprt;
829 
830  err1:
831 	rdma_destroy_id(listen_id);
832  err0:
833 	kfree(cma_xprt);
834 	return ERR_PTR(ret);
835 }
836 
837 static struct svc_rdma_fastreg_mr *rdma_alloc_frmr(struct svcxprt_rdma *xprt)
838 {
839 	struct ib_mr *mr;
840 	struct scatterlist *sg;
841 	struct svc_rdma_fastreg_mr *frmr;
842 	u32 num_sg;
843 
844 	frmr = kmalloc(sizeof(*frmr), GFP_KERNEL);
845 	if (!frmr)
846 		goto err;
847 
848 	num_sg = min_t(u32, RPCSVC_MAXPAGES, xprt->sc_frmr_pg_list_len);
849 	mr = ib_alloc_mr(xprt->sc_pd, IB_MR_TYPE_MEM_REG, num_sg);
850 	if (IS_ERR(mr))
851 		goto err_free_frmr;
852 
853 	sg = kcalloc(RPCSVC_MAXPAGES, sizeof(*sg), GFP_KERNEL);
854 	if (!sg)
855 		goto err_free_mr;
856 
857 	sg_init_table(sg, RPCSVC_MAXPAGES);
858 
859 	frmr->mr = mr;
860 	frmr->sg = sg;
861 	INIT_LIST_HEAD(&frmr->frmr_list);
862 	return frmr;
863 
864  err_free_mr:
865 	ib_dereg_mr(mr);
866  err_free_frmr:
867 	kfree(frmr);
868  err:
869 	return ERR_PTR(-ENOMEM);
870 }
871 
872 static void rdma_dealloc_frmr_q(struct svcxprt_rdma *xprt)
873 {
874 	struct svc_rdma_fastreg_mr *frmr;
875 
876 	while (!list_empty(&xprt->sc_frmr_q)) {
877 		frmr = list_entry(xprt->sc_frmr_q.next,
878 				  struct svc_rdma_fastreg_mr, frmr_list);
879 		list_del_init(&frmr->frmr_list);
880 		kfree(frmr->sg);
881 		ib_dereg_mr(frmr->mr);
882 		kfree(frmr);
883 	}
884 }
885 
886 struct svc_rdma_fastreg_mr *svc_rdma_get_frmr(struct svcxprt_rdma *rdma)
887 {
888 	struct svc_rdma_fastreg_mr *frmr = NULL;
889 
890 	spin_lock_bh(&rdma->sc_frmr_q_lock);
891 	if (!list_empty(&rdma->sc_frmr_q)) {
892 		frmr = list_entry(rdma->sc_frmr_q.next,
893 				  struct svc_rdma_fastreg_mr, frmr_list);
894 		list_del_init(&frmr->frmr_list);
895 		frmr->sg_nents = 0;
896 	}
897 	spin_unlock_bh(&rdma->sc_frmr_q_lock);
898 	if (frmr)
899 		return frmr;
900 
901 	return rdma_alloc_frmr(rdma);
902 }
903 
904 void svc_rdma_put_frmr(struct svcxprt_rdma *rdma,
905 		       struct svc_rdma_fastreg_mr *frmr)
906 {
907 	if (frmr) {
908 		ib_dma_unmap_sg(rdma->sc_cm_id->device,
909 				frmr->sg, frmr->sg_nents, frmr->direction);
910 		atomic_dec(&rdma->sc_dma_used);
911 		spin_lock_bh(&rdma->sc_frmr_q_lock);
912 		WARN_ON_ONCE(!list_empty(&frmr->frmr_list));
913 		list_add(&frmr->frmr_list, &rdma->sc_frmr_q);
914 		spin_unlock_bh(&rdma->sc_frmr_q_lock);
915 	}
916 }
917 
918 /*
919  * This is the xpo_recvfrom function for listening endpoints. Its
920  * purpose is to accept incoming connections. The CMA callback handler
921  * has already created a new transport and attached it to the new CMA
922  * ID.
923  *
924  * There is a queue of pending connections hung on the listening
925  * transport. This queue contains the new svc_xprt structure. This
926  * function takes svc_xprt structures off the accept_q and completes
927  * the connection.
928  */
929 static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt)
930 {
931 	struct svcxprt_rdma *listen_rdma;
932 	struct svcxprt_rdma *newxprt = NULL;
933 	struct rdma_conn_param conn_param;
934 	struct ib_qp_init_attr qp_attr;
935 	struct ib_device *dev;
936 	unsigned int i;
937 	int ret = 0;
938 
939 	listen_rdma = container_of(xprt, struct svcxprt_rdma, sc_xprt);
940 	clear_bit(XPT_CONN, &xprt->xpt_flags);
941 	/* Get the next entry off the accept list */
942 	spin_lock_bh(&listen_rdma->sc_lock);
943 	if (!list_empty(&listen_rdma->sc_accept_q)) {
944 		newxprt = list_entry(listen_rdma->sc_accept_q.next,
945 				     struct svcxprt_rdma, sc_accept_q);
946 		list_del_init(&newxprt->sc_accept_q);
947 	}
948 	if (!list_empty(&listen_rdma->sc_accept_q))
949 		set_bit(XPT_CONN, &listen_rdma->sc_xprt.xpt_flags);
950 	spin_unlock_bh(&listen_rdma->sc_lock);
951 	if (!newxprt)
952 		return NULL;
953 
954 	dprintk("svcrdma: newxprt from accept queue = %p, cm_id=%p\n",
955 		newxprt, newxprt->sc_cm_id);
956 
957 	dev = newxprt->sc_cm_id->device;
958 
959 	/* Qualify the transport resource defaults with the
960 	 * capabilities of this particular device */
961 	newxprt->sc_max_sge = min((size_t)dev->attrs.max_sge,
962 				  (size_t)RPCSVC_MAXPAGES);
963 	newxprt->sc_max_sge_rd = min_t(size_t, dev->attrs.max_sge_rd,
964 				       RPCSVC_MAXPAGES);
965 	newxprt->sc_max_req_size = svcrdma_max_req_size;
966 	newxprt->sc_max_requests = min_t(u32, dev->attrs.max_qp_wr,
967 					 svcrdma_max_requests);
968 	newxprt->sc_max_bc_requests = min_t(u32, dev->attrs.max_qp_wr,
969 					    svcrdma_max_bc_requests);
970 	newxprt->sc_rq_depth = newxprt->sc_max_requests +
971 			       newxprt->sc_max_bc_requests;
972 	newxprt->sc_sq_depth = RPCRDMA_SQ_DEPTH_MULT * newxprt->sc_rq_depth;
973 
974 	if (!svc_rdma_prealloc_ctxts(newxprt))
975 		goto errout;
976 	if (!svc_rdma_prealloc_maps(newxprt))
977 		goto errout;
978 
979 	/*
980 	 * Limit ORD based on client limit, local device limit, and
981 	 * configured svcrdma limit.
982 	 */
983 	newxprt->sc_ord = min_t(size_t, dev->attrs.max_qp_rd_atom, newxprt->sc_ord);
984 	newxprt->sc_ord = min_t(size_t,	svcrdma_ord, newxprt->sc_ord);
985 
986 	newxprt->sc_pd = ib_alloc_pd(dev);
987 	if (IS_ERR(newxprt->sc_pd)) {
988 		dprintk("svcrdma: error creating PD for connect request\n");
989 		goto errout;
990 	}
991 	newxprt->sc_sq_cq = ib_alloc_cq(dev, newxprt, newxprt->sc_sq_depth,
992 					0, IB_POLL_SOFTIRQ);
993 	if (IS_ERR(newxprt->sc_sq_cq)) {
994 		dprintk("svcrdma: error creating SQ CQ for connect request\n");
995 		goto errout;
996 	}
997 	newxprt->sc_rq_cq = ib_alloc_cq(dev, newxprt, newxprt->sc_rq_depth,
998 					0, IB_POLL_SOFTIRQ);
999 	if (IS_ERR(newxprt->sc_rq_cq)) {
1000 		dprintk("svcrdma: error creating RQ CQ for connect request\n");
1001 		goto errout;
1002 	}
1003 
1004 	memset(&qp_attr, 0, sizeof qp_attr);
1005 	qp_attr.event_handler = qp_event_handler;
1006 	qp_attr.qp_context = &newxprt->sc_xprt;
1007 	qp_attr.cap.max_send_wr = newxprt->sc_sq_depth;
1008 	qp_attr.cap.max_recv_wr = newxprt->sc_rq_depth;
1009 	qp_attr.cap.max_send_sge = newxprt->sc_max_sge;
1010 	qp_attr.cap.max_recv_sge = newxprt->sc_max_sge;
1011 	qp_attr.sq_sig_type = IB_SIGNAL_REQ_WR;
1012 	qp_attr.qp_type = IB_QPT_RC;
1013 	qp_attr.send_cq = newxprt->sc_sq_cq;
1014 	qp_attr.recv_cq = newxprt->sc_rq_cq;
1015 	dprintk("svcrdma: newxprt->sc_cm_id=%p, newxprt->sc_pd=%p\n"
1016 		"    cm_id->device=%p, sc_pd->device=%p\n"
1017 		"    cap.max_send_wr = %d\n"
1018 		"    cap.max_recv_wr = %d\n"
1019 		"    cap.max_send_sge = %d\n"
1020 		"    cap.max_recv_sge = %d\n",
1021 		newxprt->sc_cm_id, newxprt->sc_pd,
1022 		dev, newxprt->sc_pd->device,
1023 		qp_attr.cap.max_send_wr,
1024 		qp_attr.cap.max_recv_wr,
1025 		qp_attr.cap.max_send_sge,
1026 		qp_attr.cap.max_recv_sge);
1027 
1028 	ret = rdma_create_qp(newxprt->sc_cm_id, newxprt->sc_pd, &qp_attr);
1029 	if (ret) {
1030 		dprintk("svcrdma: failed to create QP, ret=%d\n", ret);
1031 		goto errout;
1032 	}
1033 	newxprt->sc_qp = newxprt->sc_cm_id->qp;
1034 
1035 	/*
1036 	 * Use the most secure set of MR resources based on the
1037 	 * transport type and available memory management features in
1038 	 * the device. Here's the table implemented below:
1039 	 *
1040 	 *		Fast	Global	DMA	Remote WR
1041 	 *		Reg	LKEY	MR	Access
1042 	 *		Sup'd	Sup'd	Needed	Needed
1043 	 *
1044 	 * IWARP	N	N	Y	Y
1045 	 *		N	Y	Y	Y
1046 	 *		Y	N	Y	N
1047 	 *		Y	Y	N	-
1048 	 *
1049 	 * IB		N	N	Y	N
1050 	 *		N	Y	N	-
1051 	 *		Y	N	Y	N
1052 	 *		Y	Y	N	-
1053 	 *
1054 	 * NB:	iWARP requires remote write access for the data sink
1055 	 *	of an RDMA_READ. IB does not.
1056 	 */
1057 	newxprt->sc_reader = rdma_read_chunk_lcl;
1058 	if (dev->attrs.device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS) {
1059 		newxprt->sc_frmr_pg_list_len =
1060 			dev->attrs.max_fast_reg_page_list_len;
1061 		newxprt->sc_dev_caps |= SVCRDMA_DEVCAP_FAST_REG;
1062 		newxprt->sc_reader = rdma_read_chunk_frmr;
1063 	}
1064 
1065 	/*
1066 	 * Determine if a DMA MR is required and if so, what privs are required
1067 	 */
1068 	if (!rdma_protocol_iwarp(dev, newxprt->sc_cm_id->port_num) &&
1069 	    !rdma_ib_or_roce(dev, newxprt->sc_cm_id->port_num))
1070 		goto errout;
1071 
1072 	if (rdma_protocol_iwarp(dev, newxprt->sc_cm_id->port_num))
1073 		newxprt->sc_dev_caps |= SVCRDMA_DEVCAP_READ_W_INV;
1074 
1075 	/* Post receive buffers */
1076 	for (i = 0; i < newxprt->sc_rq_depth; i++) {
1077 		ret = svc_rdma_post_recv(newxprt, GFP_KERNEL);
1078 		if (ret) {
1079 			dprintk("svcrdma: failure posting receive buffers\n");
1080 			goto errout;
1081 		}
1082 	}
1083 
1084 	/* Swap out the handler */
1085 	newxprt->sc_cm_id->event_handler = rdma_cma_handler;
1086 
1087 	/* Accept Connection */
1088 	set_bit(RDMAXPRT_CONN_PENDING, &newxprt->sc_flags);
1089 	memset(&conn_param, 0, sizeof conn_param);
1090 	conn_param.responder_resources = 0;
1091 	conn_param.initiator_depth = newxprt->sc_ord;
1092 	ret = rdma_accept(newxprt->sc_cm_id, &conn_param);
1093 	if (ret) {
1094 		dprintk("svcrdma: failed to accept new connection, ret=%d\n",
1095 		       ret);
1096 		goto errout;
1097 	}
1098 
1099 	dprintk("svcrdma: new connection %p accepted with the following "
1100 		"attributes:\n"
1101 		"    local_ip        : %pI4\n"
1102 		"    local_port	     : %d\n"
1103 		"    remote_ip       : %pI4\n"
1104 		"    remote_port     : %d\n"
1105 		"    max_sge         : %d\n"
1106 		"    max_sge_rd      : %d\n"
1107 		"    sq_depth        : %d\n"
1108 		"    max_requests    : %d\n"
1109 		"    ord             : %d\n",
1110 		newxprt,
1111 		&((struct sockaddr_in *)&newxprt->sc_cm_id->
1112 			 route.addr.src_addr)->sin_addr.s_addr,
1113 		ntohs(((struct sockaddr_in *)&newxprt->sc_cm_id->
1114 		       route.addr.src_addr)->sin_port),
1115 		&((struct sockaddr_in *)&newxprt->sc_cm_id->
1116 			 route.addr.dst_addr)->sin_addr.s_addr,
1117 		ntohs(((struct sockaddr_in *)&newxprt->sc_cm_id->
1118 		       route.addr.dst_addr)->sin_port),
1119 		newxprt->sc_max_sge,
1120 		newxprt->sc_max_sge_rd,
1121 		newxprt->sc_sq_depth,
1122 		newxprt->sc_max_requests,
1123 		newxprt->sc_ord);
1124 
1125 	return &newxprt->sc_xprt;
1126 
1127  errout:
1128 	dprintk("svcrdma: failure accepting new connection rc=%d.\n", ret);
1129 	/* Take a reference in case the DTO handler runs */
1130 	svc_xprt_get(&newxprt->sc_xprt);
1131 	if (newxprt->sc_qp && !IS_ERR(newxprt->sc_qp))
1132 		ib_destroy_qp(newxprt->sc_qp);
1133 	rdma_destroy_id(newxprt->sc_cm_id);
1134 	/* This call to put will destroy the transport */
1135 	svc_xprt_put(&newxprt->sc_xprt);
1136 	return NULL;
1137 }
1138 
1139 static void svc_rdma_release_rqst(struct svc_rqst *rqstp)
1140 {
1141 }
1142 
1143 /*
1144  * When connected, an svc_xprt has at least two references:
1145  *
1146  * - A reference held by the cm_id between the ESTABLISHED and
1147  *   DISCONNECTED events. If the remote peer disconnected first, this
1148  *   reference could be gone.
1149  *
1150  * - A reference held by the svc_recv code that called this function
1151  *   as part of close processing.
1152  *
1153  * At a minimum one references should still be held.
1154  */
1155 static void svc_rdma_detach(struct svc_xprt *xprt)
1156 {
1157 	struct svcxprt_rdma *rdma =
1158 		container_of(xprt, struct svcxprt_rdma, sc_xprt);
1159 	dprintk("svc: svc_rdma_detach(%p)\n", xprt);
1160 
1161 	/* Disconnect and flush posted WQE */
1162 	rdma_disconnect(rdma->sc_cm_id);
1163 }
1164 
1165 static void __svc_rdma_free(struct work_struct *work)
1166 {
1167 	struct svcxprt_rdma *rdma =
1168 		container_of(work, struct svcxprt_rdma, sc_work);
1169 	struct svc_xprt *xprt = &rdma->sc_xprt;
1170 
1171 	dprintk("svcrdma: %s(%p)\n", __func__, rdma);
1172 
1173 	/* We should only be called from kref_put */
1174 	if (atomic_read(&xprt->xpt_ref.refcount) != 0)
1175 		pr_err("svcrdma: sc_xprt still in use? (%d)\n",
1176 		       atomic_read(&xprt->xpt_ref.refcount));
1177 
1178 	/*
1179 	 * Destroy queued, but not processed read completions. Note
1180 	 * that this cleanup has to be done before destroying the
1181 	 * cm_id because the device ptr is needed to unmap the dma in
1182 	 * svc_rdma_put_context.
1183 	 */
1184 	while (!list_empty(&rdma->sc_read_complete_q)) {
1185 		struct svc_rdma_op_ctxt *ctxt;
1186 		ctxt = list_entry(rdma->sc_read_complete_q.next,
1187 				  struct svc_rdma_op_ctxt,
1188 				  dto_q);
1189 		list_del_init(&ctxt->dto_q);
1190 		svc_rdma_put_context(ctxt, 1);
1191 	}
1192 
1193 	/* Destroy queued, but not processed recv completions */
1194 	while (!list_empty(&rdma->sc_rq_dto_q)) {
1195 		struct svc_rdma_op_ctxt *ctxt;
1196 		ctxt = list_entry(rdma->sc_rq_dto_q.next,
1197 				  struct svc_rdma_op_ctxt,
1198 				  dto_q);
1199 		list_del_init(&ctxt->dto_q);
1200 		svc_rdma_put_context(ctxt, 1);
1201 	}
1202 
1203 	/* Warn if we leaked a resource or under-referenced */
1204 	if (rdma->sc_ctxt_used != 0)
1205 		pr_err("svcrdma: ctxt still in use? (%d)\n",
1206 		       rdma->sc_ctxt_used);
1207 	if (atomic_read(&rdma->sc_dma_used) != 0)
1208 		pr_err("svcrdma: dma still in use? (%d)\n",
1209 		       atomic_read(&rdma->sc_dma_used));
1210 
1211 	/* Final put of backchannel client transport */
1212 	if (xprt->xpt_bc_xprt) {
1213 		xprt_put(xprt->xpt_bc_xprt);
1214 		xprt->xpt_bc_xprt = NULL;
1215 	}
1216 
1217 	rdma_dealloc_frmr_q(rdma);
1218 	svc_rdma_destroy_ctxts(rdma);
1219 	svc_rdma_destroy_maps(rdma);
1220 
1221 	/* Destroy the QP if present (not a listener) */
1222 	if (rdma->sc_qp && !IS_ERR(rdma->sc_qp))
1223 		ib_destroy_qp(rdma->sc_qp);
1224 
1225 	if (rdma->sc_sq_cq && !IS_ERR(rdma->sc_sq_cq))
1226 		ib_free_cq(rdma->sc_sq_cq);
1227 
1228 	if (rdma->sc_rq_cq && !IS_ERR(rdma->sc_rq_cq))
1229 		ib_free_cq(rdma->sc_rq_cq);
1230 
1231 	if (rdma->sc_pd && !IS_ERR(rdma->sc_pd))
1232 		ib_dealloc_pd(rdma->sc_pd);
1233 
1234 	/* Destroy the CM ID */
1235 	rdma_destroy_id(rdma->sc_cm_id);
1236 
1237 	kfree(rdma);
1238 }
1239 
1240 static void svc_rdma_free(struct svc_xprt *xprt)
1241 {
1242 	struct svcxprt_rdma *rdma =
1243 		container_of(xprt, struct svcxprt_rdma, sc_xprt);
1244 	INIT_WORK(&rdma->sc_work, __svc_rdma_free);
1245 	queue_work(svc_rdma_wq, &rdma->sc_work);
1246 }
1247 
1248 static int svc_rdma_has_wspace(struct svc_xprt *xprt)
1249 {
1250 	struct svcxprt_rdma *rdma =
1251 		container_of(xprt, struct svcxprt_rdma, sc_xprt);
1252 
1253 	/*
1254 	 * If there are already waiters on the SQ,
1255 	 * return false.
1256 	 */
1257 	if (waitqueue_active(&rdma->sc_send_wait))
1258 		return 0;
1259 
1260 	/* Otherwise return true. */
1261 	return 1;
1262 }
1263 
1264 static int svc_rdma_secure_port(struct svc_rqst *rqstp)
1265 {
1266 	return 1;
1267 }
1268 
1269 int svc_rdma_send(struct svcxprt_rdma *xprt, struct ib_send_wr *wr)
1270 {
1271 	struct ib_send_wr *bad_wr, *n_wr;
1272 	int wr_count;
1273 	int i;
1274 	int ret;
1275 
1276 	if (test_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags))
1277 		return -ENOTCONN;
1278 
1279 	wr_count = 1;
1280 	for (n_wr = wr->next; n_wr; n_wr = n_wr->next)
1281 		wr_count++;
1282 
1283 	/* If the SQ is full, wait until an SQ entry is available */
1284 	while (1) {
1285 		spin_lock_bh(&xprt->sc_lock);
1286 		if (xprt->sc_sq_depth < atomic_read(&xprt->sc_sq_count) + wr_count) {
1287 			spin_unlock_bh(&xprt->sc_lock);
1288 			atomic_inc(&rdma_stat_sq_starve);
1289 
1290 			/* Wait until SQ WR available if SQ still full */
1291 			wait_event(xprt->sc_send_wait,
1292 				   atomic_read(&xprt->sc_sq_count) <
1293 				   xprt->sc_sq_depth);
1294 			if (test_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags))
1295 				return -ENOTCONN;
1296 			continue;
1297 		}
1298 		/* Take a transport ref for each WR posted */
1299 		for (i = 0; i < wr_count; i++)
1300 			svc_xprt_get(&xprt->sc_xprt);
1301 
1302 		/* Bump used SQ WR count and post */
1303 		atomic_add(wr_count, &xprt->sc_sq_count);
1304 		ret = ib_post_send(xprt->sc_qp, wr, &bad_wr);
1305 		if (ret) {
1306 			set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
1307 			atomic_sub(wr_count, &xprt->sc_sq_count);
1308 			for (i = 0; i < wr_count; i ++)
1309 				svc_xprt_put(&xprt->sc_xprt);
1310 			dprintk("svcrdma: failed to post SQ WR rc=%d, "
1311 			       "sc_sq_count=%d, sc_sq_depth=%d\n",
1312 			       ret, atomic_read(&xprt->sc_sq_count),
1313 			       xprt->sc_sq_depth);
1314 		}
1315 		spin_unlock_bh(&xprt->sc_lock);
1316 		if (ret)
1317 			wake_up(&xprt->sc_send_wait);
1318 		break;
1319 	}
1320 	return ret;
1321 }
1322