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); 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_secure_port(struct svc_rqst *); 77 static void svc_rdma_kill_temp_xprt(struct svc_xprt *); 78 79 static const struct svc_xprt_ops svc_rdma_ops = { 80 .xpo_create = svc_rdma_create, 81 .xpo_recvfrom = svc_rdma_recvfrom, 82 .xpo_sendto = svc_rdma_sendto, 83 .xpo_result_payload = svc_rdma_result_payload, 84 .xpo_release_rqst = svc_rdma_release_rqst, 85 .xpo_detach = svc_rdma_detach, 86 .xpo_free = svc_rdma_free, 87 .xpo_has_wspace = svc_rdma_has_wspace, 88 .xpo_accept = svc_rdma_accept, 89 .xpo_secure_port = svc_rdma_secure_port, 90 .xpo_kill_temp_xprt = svc_rdma_kill_temp_xprt, 91 }; 92 93 struct svc_xprt_class svc_rdma_class = { 94 .xcl_name = "rdma", 95 .xcl_owner = THIS_MODULE, 96 .xcl_ops = &svc_rdma_ops, 97 .xcl_max_payload = RPCSVC_MAXPAYLOAD_RDMA, 98 .xcl_ident = XPRT_TRANSPORT_RDMA, 99 }; 100 101 /* QP event handler */ 102 static void qp_event_handler(struct ib_event *event, void *context) 103 { 104 struct svc_xprt *xprt = context; 105 106 trace_svcrdma_qp_error(event, (struct sockaddr *)&xprt->xpt_remote); 107 switch (event->event) { 108 /* These are considered benign events */ 109 case IB_EVENT_PATH_MIG: 110 case IB_EVENT_COMM_EST: 111 case IB_EVENT_SQ_DRAINED: 112 case IB_EVENT_QP_LAST_WQE_REACHED: 113 break; 114 115 /* These are considered fatal events */ 116 case IB_EVENT_PATH_MIG_ERR: 117 case IB_EVENT_QP_FATAL: 118 case IB_EVENT_QP_REQ_ERR: 119 case IB_EVENT_QP_ACCESS_ERR: 120 case IB_EVENT_DEVICE_FATAL: 121 default: 122 svc_xprt_deferred_close(xprt); 123 break; 124 } 125 } 126 127 static struct svcxprt_rdma *svc_rdma_create_xprt(struct svc_serv *serv, 128 struct net *net) 129 { 130 struct svcxprt_rdma *cma_xprt = kzalloc(sizeof *cma_xprt, GFP_KERNEL); 131 132 if (!cma_xprt) { 133 dprintk("svcrdma: failed to create new transport\n"); 134 return NULL; 135 } 136 svc_xprt_init(net, &svc_rdma_class, &cma_xprt->sc_xprt, serv); 137 INIT_LIST_HEAD(&cma_xprt->sc_accept_q); 138 INIT_LIST_HEAD(&cma_xprt->sc_rq_dto_q); 139 INIT_LIST_HEAD(&cma_xprt->sc_send_ctxts); 140 init_llist_head(&cma_xprt->sc_recv_ctxts); 141 INIT_LIST_HEAD(&cma_xprt->sc_rw_ctxts); 142 init_waitqueue_head(&cma_xprt->sc_send_wait); 143 144 spin_lock_init(&cma_xprt->sc_lock); 145 spin_lock_init(&cma_xprt->sc_rq_dto_lock); 146 spin_lock_init(&cma_xprt->sc_send_lock); 147 spin_lock_init(&cma_xprt->sc_rw_ctxt_lock); 148 149 /* 150 * Note that this implies that the underlying transport support 151 * has some form of congestion control (see RFC 7530 section 3.1 152 * paragraph 2). For now, we assume that all supported RDMA 153 * transports are suitable here. 154 */ 155 set_bit(XPT_CONG_CTRL, &cma_xprt->sc_xprt.xpt_flags); 156 157 return cma_xprt; 158 } 159 160 static void 161 svc_rdma_parse_connect_private(struct svcxprt_rdma *newxprt, 162 struct rdma_conn_param *param) 163 { 164 const struct rpcrdma_connect_private *pmsg = param->private_data; 165 166 if (pmsg && 167 pmsg->cp_magic == rpcrdma_cmp_magic && 168 pmsg->cp_version == RPCRDMA_CMP_VERSION) { 169 newxprt->sc_snd_w_inv = pmsg->cp_flags & 170 RPCRDMA_CMP_F_SND_W_INV_OK; 171 172 dprintk("svcrdma: client send_size %u, recv_size %u " 173 "remote inv %ssupported\n", 174 rpcrdma_decode_buffer_size(pmsg->cp_send_size), 175 rpcrdma_decode_buffer_size(pmsg->cp_recv_size), 176 newxprt->sc_snd_w_inv ? "" : "un"); 177 } 178 } 179 180 /* 181 * This function handles the CONNECT_REQUEST event on a listening 182 * endpoint. It is passed the cma_id for the _new_ connection. The context in 183 * this cma_id is inherited from the listening cma_id and is the svc_xprt 184 * structure for the listening endpoint. 185 * 186 * This function creates a new xprt for the new connection and enqueues it on 187 * the accept queue for the listent xprt. When the listen thread is kicked, it 188 * will call the recvfrom method on the listen xprt which will accept the new 189 * connection. 190 */ 191 static void handle_connect_req(struct rdma_cm_id *new_cma_id, 192 struct rdma_conn_param *param) 193 { 194 struct svcxprt_rdma *listen_xprt = new_cma_id->context; 195 struct svcxprt_rdma *newxprt; 196 struct sockaddr *sa; 197 198 /* Create a new transport */ 199 newxprt = svc_rdma_create_xprt(listen_xprt->sc_xprt.xpt_server, 200 listen_xprt->sc_xprt.xpt_net); 201 if (!newxprt) 202 return; 203 newxprt->sc_cm_id = new_cma_id; 204 new_cma_id->context = newxprt; 205 svc_rdma_parse_connect_private(newxprt, param); 206 207 /* Save client advertised inbound read limit for use later in accept. */ 208 newxprt->sc_ord = param->initiator_depth; 209 210 sa = (struct sockaddr *)&newxprt->sc_cm_id->route.addr.dst_addr; 211 newxprt->sc_xprt.xpt_remotelen = svc_addr_len(sa); 212 memcpy(&newxprt->sc_xprt.xpt_remote, sa, 213 newxprt->sc_xprt.xpt_remotelen); 214 snprintf(newxprt->sc_xprt.xpt_remotebuf, 215 sizeof(newxprt->sc_xprt.xpt_remotebuf) - 1, "%pISc", sa); 216 217 /* The remote port is arbitrary and not under the control of the 218 * client ULP. Set it to a fixed value so that the DRC continues 219 * to be effective after a reconnect. 220 */ 221 rpc_set_port((struct sockaddr *)&newxprt->sc_xprt.xpt_remote, 0); 222 223 sa = (struct sockaddr *)&newxprt->sc_cm_id->route.addr.src_addr; 224 svc_xprt_set_local(&newxprt->sc_xprt, sa, svc_addr_len(sa)); 225 226 /* 227 * Enqueue the new transport on the accept queue of the listening 228 * transport 229 */ 230 spin_lock(&listen_xprt->sc_lock); 231 list_add_tail(&newxprt->sc_accept_q, &listen_xprt->sc_accept_q); 232 spin_unlock(&listen_xprt->sc_lock); 233 234 set_bit(XPT_CONN, &listen_xprt->sc_xprt.xpt_flags); 235 svc_xprt_enqueue(&listen_xprt->sc_xprt); 236 } 237 238 /** 239 * svc_rdma_listen_handler - Handle CM events generated on a listening endpoint 240 * @cma_id: the server's listener rdma_cm_id 241 * @event: details of the event 242 * 243 * Return values: 244 * %0: Do not destroy @cma_id 245 * %1: Destroy @cma_id (never returned here) 246 * 247 * NB: There is never a DEVICE_REMOVAL event for INADDR_ANY listeners. 248 */ 249 static int svc_rdma_listen_handler(struct rdma_cm_id *cma_id, 250 struct rdma_cm_event *event) 251 { 252 switch (event->event) { 253 case RDMA_CM_EVENT_CONNECT_REQUEST: 254 handle_connect_req(cma_id, &event->param.conn); 255 break; 256 default: 257 break; 258 } 259 return 0; 260 } 261 262 /** 263 * svc_rdma_cma_handler - Handle CM events on client connections 264 * @cma_id: the server's listener rdma_cm_id 265 * @event: details of the event 266 * 267 * Return values: 268 * %0: Do not destroy @cma_id 269 * %1: Destroy @cma_id (never returned here) 270 */ 271 static int svc_rdma_cma_handler(struct rdma_cm_id *cma_id, 272 struct rdma_cm_event *event) 273 { 274 struct svcxprt_rdma *rdma = cma_id->context; 275 struct svc_xprt *xprt = &rdma->sc_xprt; 276 277 switch (event->event) { 278 case RDMA_CM_EVENT_ESTABLISHED: 279 clear_bit(RDMAXPRT_CONN_PENDING, &rdma->sc_flags); 280 281 /* Handle any requests that were received while 282 * CONN_PENDING was set. */ 283 svc_xprt_enqueue(xprt); 284 break; 285 case RDMA_CM_EVENT_DISCONNECTED: 286 case RDMA_CM_EVENT_DEVICE_REMOVAL: 287 svc_xprt_deferred_close(xprt); 288 break; 289 default: 290 break; 291 } 292 return 0; 293 } 294 295 /* 296 * Create a listening RDMA service endpoint. 297 */ 298 static struct svc_xprt *svc_rdma_create(struct svc_serv *serv, 299 struct net *net, 300 struct sockaddr *sa, int salen, 301 int flags) 302 { 303 struct rdma_cm_id *listen_id; 304 struct svcxprt_rdma *cma_xprt; 305 int ret; 306 307 if (sa->sa_family != AF_INET && sa->sa_family != AF_INET6) 308 return ERR_PTR(-EAFNOSUPPORT); 309 cma_xprt = svc_rdma_create_xprt(serv, net); 310 if (!cma_xprt) 311 return ERR_PTR(-ENOMEM); 312 set_bit(XPT_LISTENER, &cma_xprt->sc_xprt.xpt_flags); 313 strcpy(cma_xprt->sc_xprt.xpt_remotebuf, "listener"); 314 315 listen_id = rdma_create_id(net, svc_rdma_listen_handler, cma_xprt, 316 RDMA_PS_TCP, IB_QPT_RC); 317 if (IS_ERR(listen_id)) { 318 ret = PTR_ERR(listen_id); 319 goto err0; 320 } 321 322 /* Allow both IPv4 and IPv6 sockets to bind a single port 323 * at the same time. 324 */ 325 #if IS_ENABLED(CONFIG_IPV6) 326 ret = rdma_set_afonly(listen_id, 1); 327 if (ret) 328 goto err1; 329 #endif 330 ret = rdma_bind_addr(listen_id, sa); 331 if (ret) 332 goto err1; 333 cma_xprt->sc_cm_id = listen_id; 334 335 ret = rdma_listen(listen_id, RPCRDMA_LISTEN_BACKLOG); 336 if (ret) 337 goto err1; 338 339 /* 340 * We need to use the address from the cm_id in case the 341 * caller specified 0 for the port number. 342 */ 343 sa = (struct sockaddr *)&cma_xprt->sc_cm_id->route.addr.src_addr; 344 svc_xprt_set_local(&cma_xprt->sc_xprt, sa, salen); 345 346 return &cma_xprt->sc_xprt; 347 348 err1: 349 rdma_destroy_id(listen_id); 350 err0: 351 kfree(cma_xprt); 352 return ERR_PTR(ret); 353 } 354 355 /* 356 * This is the xpo_recvfrom function for listening endpoints. Its 357 * purpose is to accept incoming connections. The CMA callback handler 358 * has already created a new transport and attached it to the new CMA 359 * ID. 360 * 361 * There is a queue of pending connections hung on the listening 362 * transport. This queue contains the new svc_xprt structure. This 363 * function takes svc_xprt structures off the accept_q and completes 364 * the connection. 365 */ 366 static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt) 367 { 368 struct svcxprt_rdma *listen_rdma; 369 struct svcxprt_rdma *newxprt = NULL; 370 struct rdma_conn_param conn_param; 371 struct rpcrdma_connect_private pmsg; 372 struct ib_qp_init_attr qp_attr; 373 unsigned int ctxts, rq_depth; 374 struct ib_device *dev; 375 int ret = 0; 376 RPC_IFDEBUG(struct sockaddr *sap); 377 378 listen_rdma = container_of(xprt, struct svcxprt_rdma, sc_xprt); 379 clear_bit(XPT_CONN, &xprt->xpt_flags); 380 /* Get the next entry off the accept list */ 381 spin_lock(&listen_rdma->sc_lock); 382 if (!list_empty(&listen_rdma->sc_accept_q)) { 383 newxprt = list_entry(listen_rdma->sc_accept_q.next, 384 struct svcxprt_rdma, sc_accept_q); 385 list_del_init(&newxprt->sc_accept_q); 386 } 387 if (!list_empty(&listen_rdma->sc_accept_q)) 388 set_bit(XPT_CONN, &listen_rdma->sc_xprt.xpt_flags); 389 spin_unlock(&listen_rdma->sc_lock); 390 if (!newxprt) 391 return NULL; 392 393 dev = newxprt->sc_cm_id->device; 394 newxprt->sc_port_num = newxprt->sc_cm_id->port_num; 395 396 /* Qualify the transport resource defaults with the 397 * capabilities of this particular device */ 398 /* Transport header, head iovec, tail iovec */ 399 newxprt->sc_max_send_sges = 3; 400 /* Add one SGE per page list entry */ 401 newxprt->sc_max_send_sges += (svcrdma_max_req_size / PAGE_SIZE) + 1; 402 if (newxprt->sc_max_send_sges > dev->attrs.max_send_sge) 403 newxprt->sc_max_send_sges = dev->attrs.max_send_sge; 404 newxprt->sc_max_req_size = svcrdma_max_req_size; 405 newxprt->sc_max_requests = svcrdma_max_requests; 406 newxprt->sc_max_bc_requests = svcrdma_max_bc_requests; 407 newxprt->sc_recv_batch = RPCRDMA_MAX_RECV_BATCH; 408 rq_depth = newxprt->sc_max_requests + newxprt->sc_max_bc_requests + 409 newxprt->sc_recv_batch; 410 if (rq_depth > dev->attrs.max_qp_wr) { 411 pr_warn("svcrdma: reducing receive depth to %d\n", 412 dev->attrs.max_qp_wr); 413 rq_depth = dev->attrs.max_qp_wr; 414 newxprt->sc_recv_batch = 1; 415 newxprt->sc_max_requests = rq_depth - 2; 416 newxprt->sc_max_bc_requests = 2; 417 } 418 newxprt->sc_fc_credits = cpu_to_be32(newxprt->sc_max_requests); 419 ctxts = rdma_rw_mr_factor(dev, newxprt->sc_port_num, RPCSVC_MAXPAGES); 420 ctxts *= newxprt->sc_max_requests; 421 newxprt->sc_sq_depth = rq_depth + ctxts; 422 if (newxprt->sc_sq_depth > dev->attrs.max_qp_wr) { 423 pr_warn("svcrdma: reducing send depth to %d\n", 424 dev->attrs.max_qp_wr); 425 newxprt->sc_sq_depth = dev->attrs.max_qp_wr; 426 } 427 atomic_set(&newxprt->sc_sq_avail, newxprt->sc_sq_depth); 428 429 newxprt->sc_pd = ib_alloc_pd(dev, 0); 430 if (IS_ERR(newxprt->sc_pd)) { 431 trace_svcrdma_pd_err(newxprt, PTR_ERR(newxprt->sc_pd)); 432 goto errout; 433 } 434 newxprt->sc_sq_cq = ib_alloc_cq_any(dev, newxprt, newxprt->sc_sq_depth, 435 IB_POLL_WORKQUEUE); 436 if (IS_ERR(newxprt->sc_sq_cq)) 437 goto errout; 438 newxprt->sc_rq_cq = 439 ib_alloc_cq_any(dev, newxprt, rq_depth, IB_POLL_WORKQUEUE); 440 if (IS_ERR(newxprt->sc_rq_cq)) 441 goto errout; 442 443 memset(&qp_attr, 0, sizeof qp_attr); 444 qp_attr.event_handler = qp_event_handler; 445 qp_attr.qp_context = &newxprt->sc_xprt; 446 qp_attr.port_num = newxprt->sc_port_num; 447 qp_attr.cap.max_rdma_ctxs = ctxts; 448 qp_attr.cap.max_send_wr = newxprt->sc_sq_depth - ctxts; 449 qp_attr.cap.max_recv_wr = rq_depth; 450 qp_attr.cap.max_send_sge = newxprt->sc_max_send_sges; 451 qp_attr.cap.max_recv_sge = 1; 452 qp_attr.sq_sig_type = IB_SIGNAL_REQ_WR; 453 qp_attr.qp_type = IB_QPT_RC; 454 qp_attr.send_cq = newxprt->sc_sq_cq; 455 qp_attr.recv_cq = newxprt->sc_rq_cq; 456 dprintk("svcrdma: newxprt->sc_cm_id=%p, newxprt->sc_pd=%p\n", 457 newxprt->sc_cm_id, newxprt->sc_pd); 458 dprintk(" cap.max_send_wr = %d, cap.max_recv_wr = %d\n", 459 qp_attr.cap.max_send_wr, qp_attr.cap.max_recv_wr); 460 dprintk(" cap.max_send_sge = %d, cap.max_recv_sge = %d\n", 461 qp_attr.cap.max_send_sge, qp_attr.cap.max_recv_sge); 462 463 ret = rdma_create_qp(newxprt->sc_cm_id, newxprt->sc_pd, &qp_attr); 464 if (ret) { 465 trace_svcrdma_qp_err(newxprt, ret); 466 goto errout; 467 } 468 newxprt->sc_qp = newxprt->sc_cm_id->qp; 469 470 if (!(dev->attrs.device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS)) 471 newxprt->sc_snd_w_inv = false; 472 if (!rdma_protocol_iwarp(dev, newxprt->sc_port_num) && 473 !rdma_ib_or_roce(dev, newxprt->sc_port_num)) { 474 trace_svcrdma_fabric_err(newxprt, -EINVAL); 475 goto errout; 476 } 477 478 if (!svc_rdma_post_recvs(newxprt)) 479 goto errout; 480 481 /* Construct RDMA-CM private message */ 482 pmsg.cp_magic = rpcrdma_cmp_magic; 483 pmsg.cp_version = RPCRDMA_CMP_VERSION; 484 pmsg.cp_flags = 0; 485 pmsg.cp_send_size = pmsg.cp_recv_size = 486 rpcrdma_encode_buffer_size(newxprt->sc_max_req_size); 487 488 /* Accept Connection */ 489 set_bit(RDMAXPRT_CONN_PENDING, &newxprt->sc_flags); 490 memset(&conn_param, 0, sizeof conn_param); 491 conn_param.responder_resources = 0; 492 conn_param.initiator_depth = min_t(int, newxprt->sc_ord, 493 dev->attrs.max_qp_init_rd_atom); 494 if (!conn_param.initiator_depth) { 495 ret = -EINVAL; 496 trace_svcrdma_initdepth_err(newxprt, ret); 497 goto errout; 498 } 499 conn_param.private_data = &pmsg; 500 conn_param.private_data_len = sizeof(pmsg); 501 rdma_lock_handler(newxprt->sc_cm_id); 502 newxprt->sc_cm_id->event_handler = svc_rdma_cma_handler; 503 ret = rdma_accept(newxprt->sc_cm_id, &conn_param); 504 rdma_unlock_handler(newxprt->sc_cm_id); 505 if (ret) { 506 trace_svcrdma_accept_err(newxprt, ret); 507 goto errout; 508 } 509 510 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG) 511 dprintk("svcrdma: new connection %p accepted:\n", newxprt); 512 sap = (struct sockaddr *)&newxprt->sc_cm_id->route.addr.src_addr; 513 dprintk(" local address : %pIS:%u\n", sap, rpc_get_port(sap)); 514 sap = (struct sockaddr *)&newxprt->sc_cm_id->route.addr.dst_addr; 515 dprintk(" remote address : %pIS:%u\n", sap, rpc_get_port(sap)); 516 dprintk(" max_sge : %d\n", newxprt->sc_max_send_sges); 517 dprintk(" sq_depth : %d\n", newxprt->sc_sq_depth); 518 dprintk(" rdma_rw_ctxs : %d\n", ctxts); 519 dprintk(" max_requests : %d\n", newxprt->sc_max_requests); 520 dprintk(" ord : %d\n", conn_param.initiator_depth); 521 #endif 522 523 return &newxprt->sc_xprt; 524 525 errout: 526 /* Take a reference in case the DTO handler runs */ 527 svc_xprt_get(&newxprt->sc_xprt); 528 if (newxprt->sc_qp && !IS_ERR(newxprt->sc_qp)) 529 ib_destroy_qp(newxprt->sc_qp); 530 rdma_destroy_id(newxprt->sc_cm_id); 531 /* This call to put will destroy the transport */ 532 svc_xprt_put(&newxprt->sc_xprt); 533 return NULL; 534 } 535 536 static void svc_rdma_detach(struct svc_xprt *xprt) 537 { 538 struct svcxprt_rdma *rdma = 539 container_of(xprt, struct svcxprt_rdma, sc_xprt); 540 541 rdma_disconnect(rdma->sc_cm_id); 542 } 543 544 static void __svc_rdma_free(struct work_struct *work) 545 { 546 struct svcxprt_rdma *rdma = 547 container_of(work, struct svcxprt_rdma, sc_work); 548 struct svc_xprt *xprt = &rdma->sc_xprt; 549 550 /* This blocks until the Completion Queues are empty */ 551 if (rdma->sc_qp && !IS_ERR(rdma->sc_qp)) 552 ib_drain_qp(rdma->sc_qp); 553 554 svc_rdma_flush_recv_queues(rdma); 555 556 /* Final put of backchannel client transport */ 557 if (xprt->xpt_bc_xprt) { 558 xprt_put(xprt->xpt_bc_xprt); 559 xprt->xpt_bc_xprt = NULL; 560 } 561 562 svc_rdma_destroy_rw_ctxts(rdma); 563 svc_rdma_send_ctxts_destroy(rdma); 564 svc_rdma_recv_ctxts_destroy(rdma); 565 566 /* Destroy the QP if present (not a listener) */ 567 if (rdma->sc_qp && !IS_ERR(rdma->sc_qp)) 568 ib_destroy_qp(rdma->sc_qp); 569 570 if (rdma->sc_sq_cq && !IS_ERR(rdma->sc_sq_cq)) 571 ib_free_cq(rdma->sc_sq_cq); 572 573 if (rdma->sc_rq_cq && !IS_ERR(rdma->sc_rq_cq)) 574 ib_free_cq(rdma->sc_rq_cq); 575 576 if (rdma->sc_pd && !IS_ERR(rdma->sc_pd)) 577 ib_dealloc_pd(rdma->sc_pd); 578 579 /* Destroy the CM ID */ 580 rdma_destroy_id(rdma->sc_cm_id); 581 582 kfree(rdma); 583 } 584 585 static void svc_rdma_free(struct svc_xprt *xprt) 586 { 587 struct svcxprt_rdma *rdma = 588 container_of(xprt, struct svcxprt_rdma, sc_xprt); 589 590 INIT_WORK(&rdma->sc_work, __svc_rdma_free); 591 schedule_work(&rdma->sc_work); 592 } 593 594 static int svc_rdma_has_wspace(struct svc_xprt *xprt) 595 { 596 struct svcxprt_rdma *rdma = 597 container_of(xprt, struct svcxprt_rdma, sc_xprt); 598 599 /* 600 * If there are already waiters on the SQ, 601 * return false. 602 */ 603 if (waitqueue_active(&rdma->sc_send_wait)) 604 return 0; 605 606 /* Otherwise return true. */ 607 return 1; 608 } 609 610 static void svc_rdma_secure_port(struct svc_rqst *rqstp) 611 { 612 set_bit(RQ_SECURE, &rqstp->rq_flags); 613 } 614 615 static void svc_rdma_kill_temp_xprt(struct svc_xprt *xprt) 616 { 617 } 618