xref: /linux/net/sunrpc/xprtrdma/svc_rdma_transport.c (revision 36a1818f5a1e50b805317ba13f827067d50f6970)
1 // SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause
2 /*
3  * Copyright (c) 2015-2018 Oracle. All rights reserved.
4  * Copyright (c) 2014 Open Grid Computing, Inc. All rights reserved.
5  * Copyright (c) 2005-2007 Network Appliance, Inc. All rights reserved.
6  *
7  * This software is available to you under a choice of one of two
8  * licenses.  You may choose to be licensed under the terms of the GNU
9  * General Public License (GPL) Version 2, available from the file
10  * COPYING in the main directory of this source tree, or the BSD-type
11  * license below:
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  *
17  *      Redistributions of source code must retain the above copyright
18  *      notice, this list of conditions and the following disclaimer.
19  *
20  *      Redistributions in binary form must reproduce the above
21  *      copyright notice, this list of conditions and the following
22  *      disclaimer in the documentation and/or other materials provided
23  *      with the distribution.
24  *
25  *      Neither the name of the Network Appliance, Inc. nor the names of
26  *      its contributors may be used to endorse or promote products
27  *      derived from this software without specific prior written
28  *      permission.
29  *
30  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
31  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
32  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
33  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
34  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
35  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
36  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
37  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
38  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
39  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
40  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
41  *
42  * Author: Tom Tucker <tom@opengridcomputing.com>
43  */
44 
45 #include <linux/interrupt.h>
46 #include <linux/sched.h>
47 #include <linux/slab.h>
48 #include <linux/spinlock.h>
49 #include <linux/workqueue.h>
50 #include <linux/export.h>
51 
52 #include <rdma/ib_verbs.h>
53 #include <rdma/rdma_cm.h>
54 #include <rdma/rw.h>
55 
56 #include <linux/sunrpc/addr.h>
57 #include <linux/sunrpc/debug.h>
58 #include <linux/sunrpc/svc_xprt.h>
59 #include <linux/sunrpc/svc_rdma.h>
60 
61 #include "xprt_rdma.h"
62 #include <trace/events/rpcrdma.h>
63 
64 #define RPCDBG_FACILITY	RPCDBG_SVCXPRT
65 
66 static struct svcxprt_rdma *svc_rdma_create_xprt(struct svc_serv *serv,
67 						 struct net *net, int node);
68 static struct svc_xprt *svc_rdma_create(struct svc_serv *serv,
69 					struct net *net,
70 					struct sockaddr *sa, int salen,
71 					int flags);
72 static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt);
73 static void svc_rdma_detach(struct svc_xprt *xprt);
74 static void svc_rdma_free(struct svc_xprt *xprt);
75 static int svc_rdma_has_wspace(struct svc_xprt *xprt);
76 static void svc_rdma_kill_temp_xprt(struct svc_xprt *);
77 
78 static const struct svc_xprt_ops svc_rdma_ops = {
79 	.xpo_create = svc_rdma_create,
80 	.xpo_recvfrom = svc_rdma_recvfrom,
81 	.xpo_sendto = svc_rdma_sendto,
82 	.xpo_result_payload = svc_rdma_result_payload,
83 	.xpo_release_ctxt = svc_rdma_release_ctxt,
84 	.xpo_detach = svc_rdma_detach,
85 	.xpo_free = svc_rdma_free,
86 	.xpo_has_wspace = svc_rdma_has_wspace,
87 	.xpo_accept = svc_rdma_accept,
88 	.xpo_kill_temp_xprt = svc_rdma_kill_temp_xprt,
89 };
90 
91 struct svc_xprt_class svc_rdma_class = {
92 	.xcl_name = "rdma",
93 	.xcl_owner = THIS_MODULE,
94 	.xcl_ops = &svc_rdma_ops,
95 	.xcl_max_payload = RPCSVC_MAXPAYLOAD_RDMA,
96 	.xcl_ident = XPRT_TRANSPORT_RDMA,
97 };
98 
99 /* QP event handler */
100 static void qp_event_handler(struct ib_event *event, void *context)
101 {
102 	struct svc_xprt *xprt = context;
103 
104 	trace_svcrdma_qp_error(event, (struct sockaddr *)&xprt->xpt_remote);
105 	switch (event->event) {
106 	/* These are considered benign events */
107 	case IB_EVENT_PATH_MIG:
108 	case IB_EVENT_COMM_EST:
109 	case IB_EVENT_SQ_DRAINED:
110 	case IB_EVENT_QP_LAST_WQE_REACHED:
111 		break;
112 
113 	/* These are considered fatal events */
114 	case IB_EVENT_PATH_MIG_ERR:
115 	case IB_EVENT_QP_FATAL:
116 	case IB_EVENT_QP_REQ_ERR:
117 	case IB_EVENT_QP_ACCESS_ERR:
118 	case IB_EVENT_DEVICE_FATAL:
119 	default:
120 		svc_xprt_deferred_close(xprt);
121 		break;
122 	}
123 }
124 
125 static struct svcxprt_rdma *svc_rdma_create_xprt(struct svc_serv *serv,
126 						 struct net *net, int node)
127 {
128 	static struct lock_class_key svcrdma_rwctx_lock;
129 	static struct lock_class_key svcrdma_sctx_lock;
130 	static struct lock_class_key svcrdma_dto_lock;
131 	struct svcxprt_rdma *cma_xprt;
132 
133 	cma_xprt = kzalloc_node(sizeof(*cma_xprt), GFP_KERNEL, node);
134 	if (!cma_xprt)
135 		return NULL;
136 
137 	svc_xprt_init(net, &svc_rdma_class, &cma_xprt->sc_xprt, serv);
138 	INIT_LIST_HEAD(&cma_xprt->sc_accept_q);
139 	INIT_LIST_HEAD(&cma_xprt->sc_rq_dto_q);
140 	INIT_LIST_HEAD(&cma_xprt->sc_read_complete_q);
141 	init_llist_head(&cma_xprt->sc_send_ctxts);
142 	init_llist_head(&cma_xprt->sc_recv_ctxts);
143 	init_llist_head(&cma_xprt->sc_rw_ctxts);
144 	init_waitqueue_head(&cma_xprt->sc_send_wait);
145 
146 	spin_lock_init(&cma_xprt->sc_lock);
147 	spin_lock_init(&cma_xprt->sc_rq_dto_lock);
148 	lockdep_set_class(&cma_xprt->sc_rq_dto_lock, &svcrdma_dto_lock);
149 	spin_lock_init(&cma_xprt->sc_send_lock);
150 	lockdep_set_class(&cma_xprt->sc_send_lock, &svcrdma_sctx_lock);
151 	spin_lock_init(&cma_xprt->sc_rw_ctxt_lock);
152 	lockdep_set_class(&cma_xprt->sc_rw_ctxt_lock, &svcrdma_rwctx_lock);
153 
154 	/*
155 	 * Note that this implies that the underlying transport support
156 	 * has some form of congestion control (see RFC 7530 section 3.1
157 	 * paragraph 2). For now, we assume that all supported RDMA
158 	 * transports are suitable here.
159 	 */
160 	set_bit(XPT_CONG_CTRL, &cma_xprt->sc_xprt.xpt_flags);
161 
162 	return cma_xprt;
163 }
164 
165 static void
166 svc_rdma_parse_connect_private(struct svcxprt_rdma *newxprt,
167 			       struct rdma_conn_param *param)
168 {
169 	const struct rpcrdma_connect_private *pmsg = param->private_data;
170 
171 	if (pmsg &&
172 	    pmsg->cp_magic == rpcrdma_cmp_magic &&
173 	    pmsg->cp_version == RPCRDMA_CMP_VERSION) {
174 		newxprt->sc_snd_w_inv = pmsg->cp_flags &
175 					RPCRDMA_CMP_F_SND_W_INV_OK;
176 
177 		dprintk("svcrdma: client send_size %u, recv_size %u "
178 			"remote inv %ssupported\n",
179 			rpcrdma_decode_buffer_size(pmsg->cp_send_size),
180 			rpcrdma_decode_buffer_size(pmsg->cp_recv_size),
181 			newxprt->sc_snd_w_inv ? "" : "un");
182 	}
183 }
184 
185 /*
186  * This function handles the CONNECT_REQUEST event on a listening
187  * endpoint. It is passed the cma_id for the _new_ connection. The context in
188  * this cma_id is inherited from the listening cma_id and is the svc_xprt
189  * structure for the listening endpoint.
190  *
191  * This function creates a new xprt for the new connection and enqueues it on
192  * the accept queue for the listent xprt. When the listen thread is kicked, it
193  * will call the recvfrom method on the listen xprt which will accept the new
194  * connection.
195  */
196 static void handle_connect_req(struct rdma_cm_id *new_cma_id,
197 			       struct rdma_conn_param *param)
198 {
199 	struct svcxprt_rdma *listen_xprt = new_cma_id->context;
200 	struct svcxprt_rdma *newxprt;
201 	struct sockaddr *sa;
202 
203 	newxprt = svc_rdma_create_xprt(listen_xprt->sc_xprt.xpt_server,
204 				       listen_xprt->sc_xprt.xpt_net,
205 				       ibdev_to_node(new_cma_id->device));
206 	if (!newxprt)
207 		return;
208 	newxprt->sc_cm_id = new_cma_id;
209 	new_cma_id->context = newxprt;
210 	svc_rdma_parse_connect_private(newxprt, param);
211 
212 	/* Save client advertised inbound read limit for use later in accept. */
213 	newxprt->sc_ord = param->initiator_depth;
214 
215 	sa = (struct sockaddr *)&newxprt->sc_cm_id->route.addr.dst_addr;
216 	newxprt->sc_xprt.xpt_remotelen = svc_addr_len(sa);
217 	memcpy(&newxprt->sc_xprt.xpt_remote, sa,
218 	       newxprt->sc_xprt.xpt_remotelen);
219 	snprintf(newxprt->sc_xprt.xpt_remotebuf,
220 		 sizeof(newxprt->sc_xprt.xpt_remotebuf) - 1, "%pISc", sa);
221 
222 	/* The remote port is arbitrary and not under the control of the
223 	 * client ULP. Set it to a fixed value so that the DRC continues
224 	 * to be effective after a reconnect.
225 	 */
226 	rpc_set_port((struct sockaddr *)&newxprt->sc_xprt.xpt_remote, 0);
227 
228 	sa = (struct sockaddr *)&newxprt->sc_cm_id->route.addr.src_addr;
229 	svc_xprt_set_local(&newxprt->sc_xprt, sa, svc_addr_len(sa));
230 
231 	/*
232 	 * Enqueue the new transport on the accept queue of the listening
233 	 * transport
234 	 */
235 	spin_lock(&listen_xprt->sc_lock);
236 	list_add_tail(&newxprt->sc_accept_q, &listen_xprt->sc_accept_q);
237 	spin_unlock(&listen_xprt->sc_lock);
238 
239 	set_bit(XPT_CONN, &listen_xprt->sc_xprt.xpt_flags);
240 	svc_xprt_enqueue(&listen_xprt->sc_xprt);
241 }
242 
243 /**
244  * svc_rdma_listen_handler - Handle CM events generated on a listening endpoint
245  * @cma_id: the server's listener rdma_cm_id
246  * @event: details of the event
247  *
248  * Return values:
249  *     %0: Do not destroy @cma_id
250  *     %1: Destroy @cma_id (never returned here)
251  *
252  * NB: There is never a DEVICE_REMOVAL event for INADDR_ANY listeners.
253  */
254 static int svc_rdma_listen_handler(struct rdma_cm_id *cma_id,
255 				   struct rdma_cm_event *event)
256 {
257 	switch (event->event) {
258 	case RDMA_CM_EVENT_CONNECT_REQUEST:
259 		handle_connect_req(cma_id, &event->param.conn);
260 		break;
261 	default:
262 		break;
263 	}
264 	return 0;
265 }
266 
267 /**
268  * svc_rdma_cma_handler - Handle CM events on client connections
269  * @cma_id: the server's listener rdma_cm_id
270  * @event: details of the event
271  *
272  * Return values:
273  *     %0: Do not destroy @cma_id
274  *     %1: Destroy @cma_id (never returned here)
275  */
276 static int svc_rdma_cma_handler(struct rdma_cm_id *cma_id,
277 				struct rdma_cm_event *event)
278 {
279 	struct svcxprt_rdma *rdma = cma_id->context;
280 	struct svc_xprt *xprt = &rdma->sc_xprt;
281 
282 	switch (event->event) {
283 	case RDMA_CM_EVENT_ESTABLISHED:
284 		clear_bit(RDMAXPRT_CONN_PENDING, &rdma->sc_flags);
285 
286 		/* Handle any requests that were received while
287 		 * CONN_PENDING was set. */
288 		svc_xprt_enqueue(xprt);
289 		break;
290 	case RDMA_CM_EVENT_DISCONNECTED:
291 	case RDMA_CM_EVENT_DEVICE_REMOVAL:
292 		svc_xprt_deferred_close(xprt);
293 		break;
294 	default:
295 		break;
296 	}
297 	return 0;
298 }
299 
300 /*
301  * Create a listening RDMA service endpoint.
302  */
303 static struct svc_xprt *svc_rdma_create(struct svc_serv *serv,
304 					struct net *net,
305 					struct sockaddr *sa, int salen,
306 					int flags)
307 {
308 	struct rdma_cm_id *listen_id;
309 	struct svcxprt_rdma *cma_xprt;
310 	int ret;
311 
312 	if (sa->sa_family != AF_INET && sa->sa_family != AF_INET6)
313 		return ERR_PTR(-EAFNOSUPPORT);
314 	cma_xprt = svc_rdma_create_xprt(serv, net, NUMA_NO_NODE);
315 	if (!cma_xprt)
316 		return ERR_PTR(-ENOMEM);
317 	set_bit(XPT_LISTENER, &cma_xprt->sc_xprt.xpt_flags);
318 	strcpy(cma_xprt->sc_xprt.xpt_remotebuf, "listener");
319 
320 	listen_id = rdma_create_id(net, svc_rdma_listen_handler, cma_xprt,
321 				   RDMA_PS_TCP, IB_QPT_RC);
322 	if (IS_ERR(listen_id)) {
323 		ret = PTR_ERR(listen_id);
324 		goto err0;
325 	}
326 
327 	/* Allow both IPv4 and IPv6 sockets to bind a single port
328 	 * at the same time.
329 	 */
330 #if IS_ENABLED(CONFIG_IPV6)
331 	ret = rdma_set_afonly(listen_id, 1);
332 	if (ret)
333 		goto err1;
334 #endif
335 	ret = rdma_bind_addr(listen_id, sa);
336 	if (ret)
337 		goto err1;
338 	cma_xprt->sc_cm_id = listen_id;
339 
340 	ret = rdma_listen(listen_id, RPCRDMA_LISTEN_BACKLOG);
341 	if (ret)
342 		goto err1;
343 
344 	/*
345 	 * We need to use the address from the cm_id in case the
346 	 * caller specified 0 for the port number.
347 	 */
348 	sa = (struct sockaddr *)&cma_xprt->sc_cm_id->route.addr.src_addr;
349 	svc_xprt_set_local(&cma_xprt->sc_xprt, sa, salen);
350 
351 	return &cma_xprt->sc_xprt;
352 
353  err1:
354 	rdma_destroy_id(listen_id);
355  err0:
356 	kfree(cma_xprt);
357 	return ERR_PTR(ret);
358 }
359 
360 /*
361  * This is the xpo_recvfrom function for listening endpoints. Its
362  * purpose is to accept incoming connections. The CMA callback handler
363  * has already created a new transport and attached it to the new CMA
364  * ID.
365  *
366  * There is a queue of pending connections hung on the listening
367  * transport. This queue contains the new svc_xprt structure. This
368  * function takes svc_xprt structures off the accept_q and completes
369  * the connection.
370  */
371 static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt)
372 {
373 	struct svcxprt_rdma *listen_rdma;
374 	struct svcxprt_rdma *newxprt = NULL;
375 	struct rdma_conn_param conn_param;
376 	struct rpcrdma_connect_private pmsg;
377 	struct ib_qp_init_attr qp_attr;
378 	unsigned int ctxts, rq_depth;
379 	struct ib_device *dev;
380 	int ret = 0;
381 	RPC_IFDEBUG(struct sockaddr *sap);
382 
383 	listen_rdma = container_of(xprt, struct svcxprt_rdma, sc_xprt);
384 	clear_bit(XPT_CONN, &xprt->xpt_flags);
385 	/* Get the next entry off the accept list */
386 	spin_lock(&listen_rdma->sc_lock);
387 	if (!list_empty(&listen_rdma->sc_accept_q)) {
388 		newxprt = list_entry(listen_rdma->sc_accept_q.next,
389 				     struct svcxprt_rdma, sc_accept_q);
390 		list_del_init(&newxprt->sc_accept_q);
391 	}
392 	if (!list_empty(&listen_rdma->sc_accept_q))
393 		set_bit(XPT_CONN, &listen_rdma->sc_xprt.xpt_flags);
394 	spin_unlock(&listen_rdma->sc_lock);
395 	if (!newxprt)
396 		return NULL;
397 
398 	dev = newxprt->sc_cm_id->device;
399 	newxprt->sc_port_num = newxprt->sc_cm_id->port_num;
400 
401 	newxprt->sc_max_req_size = svcrdma_max_req_size;
402 	newxprt->sc_max_requests = svcrdma_max_requests;
403 	newxprt->sc_max_bc_requests = svcrdma_max_bc_requests;
404 	newxprt->sc_recv_batch = RPCRDMA_MAX_RECV_BATCH;
405 	newxprt->sc_fc_credits = cpu_to_be32(newxprt->sc_max_requests);
406 
407 	/* Qualify the transport's resource defaults with the
408 	 * capabilities of this particular device.
409 	 */
410 
411 	/* Transport header, head iovec, tail iovec */
412 	newxprt->sc_max_send_sges = 3;
413 	/* Add one SGE per page list entry */
414 	newxprt->sc_max_send_sges += (svcrdma_max_req_size / PAGE_SIZE) + 1;
415 	if (newxprt->sc_max_send_sges > dev->attrs.max_send_sge)
416 		newxprt->sc_max_send_sges = dev->attrs.max_send_sge;
417 	rq_depth = newxprt->sc_max_requests + newxprt->sc_max_bc_requests +
418 		   newxprt->sc_recv_batch + 1 /* drain */;
419 	if (rq_depth > dev->attrs.max_qp_wr) {
420 		rq_depth = dev->attrs.max_qp_wr;
421 		newxprt->sc_recv_batch = 1;
422 		newxprt->sc_max_requests = rq_depth - 2;
423 		newxprt->sc_max_bc_requests = 2;
424 	}
425 
426 	/* Arbitrarily estimate the number of rw_ctxs needed for
427 	 * this transport. This is enough rw_ctxs to make forward
428 	 * progress even if the client is using one rkey per page
429 	 * in each Read chunk.
430 	 */
431 	ctxts = 3 * RPCSVC_MAXPAGES;
432 	newxprt->sc_sq_depth = rq_depth + ctxts;
433 	if (newxprt->sc_sq_depth > dev->attrs.max_qp_wr)
434 		newxprt->sc_sq_depth = dev->attrs.max_qp_wr;
435 	atomic_set(&newxprt->sc_sq_avail, newxprt->sc_sq_depth);
436 
437 	newxprt->sc_pd = ib_alloc_pd(dev, 0);
438 	if (IS_ERR(newxprt->sc_pd)) {
439 		trace_svcrdma_pd_err(newxprt, PTR_ERR(newxprt->sc_pd));
440 		goto errout;
441 	}
442 	newxprt->sc_sq_cq = ib_alloc_cq_any(dev, newxprt, newxprt->sc_sq_depth,
443 					    IB_POLL_WORKQUEUE);
444 	if (IS_ERR(newxprt->sc_sq_cq))
445 		goto errout;
446 	newxprt->sc_rq_cq =
447 		ib_alloc_cq_any(dev, newxprt, rq_depth, IB_POLL_WORKQUEUE);
448 	if (IS_ERR(newxprt->sc_rq_cq))
449 		goto errout;
450 
451 	memset(&qp_attr, 0, sizeof qp_attr);
452 	qp_attr.event_handler = qp_event_handler;
453 	qp_attr.qp_context = &newxprt->sc_xprt;
454 	qp_attr.port_num = newxprt->sc_port_num;
455 	qp_attr.cap.max_rdma_ctxs = ctxts;
456 	qp_attr.cap.max_send_wr = newxprt->sc_sq_depth - ctxts;
457 	qp_attr.cap.max_recv_wr = rq_depth;
458 	qp_attr.cap.max_send_sge = newxprt->sc_max_send_sges;
459 	qp_attr.cap.max_recv_sge = 1;
460 	qp_attr.sq_sig_type = IB_SIGNAL_REQ_WR;
461 	qp_attr.qp_type = IB_QPT_RC;
462 	qp_attr.send_cq = newxprt->sc_sq_cq;
463 	qp_attr.recv_cq = newxprt->sc_rq_cq;
464 	dprintk("    cap.max_send_wr = %d, cap.max_recv_wr = %d\n",
465 		qp_attr.cap.max_send_wr, qp_attr.cap.max_recv_wr);
466 	dprintk("    cap.max_send_sge = %d, cap.max_recv_sge = %d\n",
467 		qp_attr.cap.max_send_sge, qp_attr.cap.max_recv_sge);
468 	dprintk("    send CQ depth = %u, recv CQ depth = %u\n",
469 		newxprt->sc_sq_depth, rq_depth);
470 	ret = rdma_create_qp(newxprt->sc_cm_id, newxprt->sc_pd, &qp_attr);
471 	if (ret) {
472 		trace_svcrdma_qp_err(newxprt, ret);
473 		goto errout;
474 	}
475 	newxprt->sc_max_send_sges = qp_attr.cap.max_send_sge;
476 	newxprt->sc_qp = newxprt->sc_cm_id->qp;
477 
478 	if (!(dev->attrs.device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS))
479 		newxprt->sc_snd_w_inv = false;
480 	if (!rdma_protocol_iwarp(dev, newxprt->sc_port_num) &&
481 	    !rdma_ib_or_roce(dev, newxprt->sc_port_num)) {
482 		trace_svcrdma_fabric_err(newxprt, -EINVAL);
483 		goto errout;
484 	}
485 
486 	if (!svc_rdma_post_recvs(newxprt))
487 		goto errout;
488 
489 	/* Construct RDMA-CM private message */
490 	pmsg.cp_magic = rpcrdma_cmp_magic;
491 	pmsg.cp_version = RPCRDMA_CMP_VERSION;
492 	pmsg.cp_flags = 0;
493 	pmsg.cp_send_size = pmsg.cp_recv_size =
494 		rpcrdma_encode_buffer_size(newxprt->sc_max_req_size);
495 
496 	/* Accept Connection */
497 	set_bit(RDMAXPRT_CONN_PENDING, &newxprt->sc_flags);
498 	memset(&conn_param, 0, sizeof conn_param);
499 	conn_param.responder_resources = 0;
500 	conn_param.initiator_depth = min_t(int, newxprt->sc_ord,
501 					   dev->attrs.max_qp_init_rd_atom);
502 	if (!conn_param.initiator_depth) {
503 		ret = -EINVAL;
504 		trace_svcrdma_initdepth_err(newxprt, ret);
505 		goto errout;
506 	}
507 	conn_param.private_data = &pmsg;
508 	conn_param.private_data_len = sizeof(pmsg);
509 	rdma_lock_handler(newxprt->sc_cm_id);
510 	newxprt->sc_cm_id->event_handler = svc_rdma_cma_handler;
511 	ret = rdma_accept(newxprt->sc_cm_id, &conn_param);
512 	rdma_unlock_handler(newxprt->sc_cm_id);
513 	if (ret) {
514 		trace_svcrdma_accept_err(newxprt, ret);
515 		goto errout;
516 	}
517 
518 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
519 	dprintk("svcrdma: new connection accepted on device %s:\n", dev->name);
520 	sap = (struct sockaddr *)&newxprt->sc_cm_id->route.addr.src_addr;
521 	dprintk("    local address   : %pIS:%u\n", sap, rpc_get_port(sap));
522 	sap = (struct sockaddr *)&newxprt->sc_cm_id->route.addr.dst_addr;
523 	dprintk("    remote address  : %pIS:%u\n", sap, rpc_get_port(sap));
524 	dprintk("    max_sge         : %d\n", newxprt->sc_max_send_sges);
525 	dprintk("    sq_depth        : %d\n", newxprt->sc_sq_depth);
526 	dprintk("    rdma_rw_ctxs    : %d\n", ctxts);
527 	dprintk("    max_requests    : %d\n", newxprt->sc_max_requests);
528 	dprintk("    ord             : %d\n", conn_param.initiator_depth);
529 #endif
530 
531 	return &newxprt->sc_xprt;
532 
533  errout:
534 	/* Take a reference in case the DTO handler runs */
535 	svc_xprt_get(&newxprt->sc_xprt);
536 	if (newxprt->sc_qp && !IS_ERR(newxprt->sc_qp))
537 		ib_destroy_qp(newxprt->sc_qp);
538 	rdma_destroy_id(newxprt->sc_cm_id);
539 	/* This call to put will destroy the transport */
540 	svc_xprt_put(&newxprt->sc_xprt);
541 	return NULL;
542 }
543 
544 static void svc_rdma_detach(struct svc_xprt *xprt)
545 {
546 	struct svcxprt_rdma *rdma =
547 		container_of(xprt, struct svcxprt_rdma, sc_xprt);
548 
549 	rdma_disconnect(rdma->sc_cm_id);
550 }
551 
552 static void __svc_rdma_free(struct work_struct *work)
553 {
554 	struct svcxprt_rdma *rdma =
555 		container_of(work, struct svcxprt_rdma, sc_work);
556 
557 	/* This blocks until the Completion Queues are empty */
558 	if (rdma->sc_qp && !IS_ERR(rdma->sc_qp))
559 		ib_drain_qp(rdma->sc_qp);
560 	flush_workqueue(svcrdma_wq);
561 
562 	svc_rdma_flush_recv_queues(rdma);
563 
564 	svc_rdma_destroy_rw_ctxts(rdma);
565 	svc_rdma_send_ctxts_destroy(rdma);
566 	svc_rdma_recv_ctxts_destroy(rdma);
567 
568 	/* Destroy the QP if present (not a listener) */
569 	if (rdma->sc_qp && !IS_ERR(rdma->sc_qp))
570 		ib_destroy_qp(rdma->sc_qp);
571 
572 	if (rdma->sc_sq_cq && !IS_ERR(rdma->sc_sq_cq))
573 		ib_free_cq(rdma->sc_sq_cq);
574 
575 	if (rdma->sc_rq_cq && !IS_ERR(rdma->sc_rq_cq))
576 		ib_free_cq(rdma->sc_rq_cq);
577 
578 	if (rdma->sc_pd && !IS_ERR(rdma->sc_pd))
579 		ib_dealloc_pd(rdma->sc_pd);
580 
581 	/* Destroy the CM ID */
582 	rdma_destroy_id(rdma->sc_cm_id);
583 
584 	kfree(rdma);
585 }
586 
587 static void svc_rdma_free(struct svc_xprt *xprt)
588 {
589 	struct svcxprt_rdma *rdma =
590 		container_of(xprt, struct svcxprt_rdma, sc_xprt);
591 
592 	INIT_WORK(&rdma->sc_work, __svc_rdma_free);
593 	schedule_work(&rdma->sc_work);
594 }
595 
596 static int svc_rdma_has_wspace(struct svc_xprt *xprt)
597 {
598 	struct svcxprt_rdma *rdma =
599 		container_of(xprt, struct svcxprt_rdma, sc_xprt);
600 
601 	/*
602 	 * If there are already waiters on the SQ,
603 	 * return false.
604 	 */
605 	if (waitqueue_active(&rdma->sc_send_wait))
606 		return 0;
607 
608 	/* Otherwise return true. */
609 	return 1;
610 }
611 
612 static void svc_rdma_kill_temp_xprt(struct svc_xprt *xprt)
613 {
614 }
615