1 // SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause 2 /* 3 * Copyright (c) 2014-2017 Oracle. All rights reserved. 4 * Copyright (c) 2003-2007 Network Appliance, Inc. All rights reserved. 5 * 6 * This software is available to you under a choice of one of two 7 * licenses. You may choose to be licensed under the terms of the GNU 8 * General Public License (GPL) Version 2, available from the file 9 * COPYING in the main directory of this source tree, or the BSD-type 10 * license below: 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 16 * Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 19 * Redistributions in binary form must reproduce the above 20 * copyright notice, this list of conditions and the following 21 * disclaimer in the documentation and/or other materials provided 22 * with the distribution. 23 * 24 * Neither the name of the Network Appliance, Inc. nor the names of 25 * its contributors may be used to endorse or promote products 26 * derived from this software without specific prior written 27 * permission. 28 * 29 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 30 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 31 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 32 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 33 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 34 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 35 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 36 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 37 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 38 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 39 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 40 */ 41 42 /* 43 * verbs.c 44 * 45 * Encapsulates the major functions managing: 46 * o adapters 47 * o endpoints 48 * o connections 49 * o buffer memory 50 */ 51 52 #include <linux/interrupt.h> 53 #include <linux/slab.h> 54 #include <linux/sunrpc/addr.h> 55 #include <linux/sunrpc/svc_rdma.h> 56 #include <linux/log2.h> 57 58 #include <asm-generic/barrier.h> 59 #include <asm/bitops.h> 60 61 #include <rdma/ib_cm.h> 62 63 #include "xprt_rdma.h" 64 #include <trace/events/rpcrdma.h> 65 66 /* 67 * Globals/Macros 68 */ 69 70 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG) 71 # define RPCDBG_FACILITY RPCDBG_TRANS 72 #endif 73 74 /* 75 * internal functions 76 */ 77 static int rpcrdma_sendctxs_create(struct rpcrdma_xprt *r_xprt); 78 static void rpcrdma_sendctxs_destroy(struct rpcrdma_xprt *r_xprt); 79 static void rpcrdma_sendctx_put_locked(struct rpcrdma_xprt *r_xprt, 80 struct rpcrdma_sendctx *sc); 81 static int rpcrdma_reqs_setup(struct rpcrdma_xprt *r_xprt); 82 static void rpcrdma_reqs_reset(struct rpcrdma_xprt *r_xprt); 83 static void rpcrdma_rep_destroy(struct rpcrdma_rep *rep); 84 static void rpcrdma_reps_unmap(struct rpcrdma_xprt *r_xprt); 85 static void rpcrdma_mrs_create(struct rpcrdma_xprt *r_xprt); 86 static void rpcrdma_mrs_destroy(struct rpcrdma_xprt *r_xprt); 87 static void rpcrdma_ep_get(struct rpcrdma_ep *ep); 88 static int rpcrdma_ep_put(struct rpcrdma_ep *ep); 89 static struct rpcrdma_regbuf * 90 rpcrdma_regbuf_alloc(size_t size, enum dma_data_direction direction, 91 gfp_t flags); 92 static void rpcrdma_regbuf_dma_unmap(struct rpcrdma_regbuf *rb); 93 static void rpcrdma_regbuf_free(struct rpcrdma_regbuf *rb); 94 95 /* Wait for outstanding transport work to finish. ib_drain_qp 96 * handles the drains in the wrong order for us, so open code 97 * them here. 98 */ 99 static void rpcrdma_xprt_drain(struct rpcrdma_xprt *r_xprt) 100 { 101 struct rpcrdma_ep *ep = r_xprt->rx_ep; 102 struct rdma_cm_id *id = ep->re_id; 103 104 /* Wait for rpcrdma_post_recvs() to leave its critical 105 * section. 106 */ 107 if (atomic_inc_return(&ep->re_receiving) > 1) 108 wait_for_completion(&ep->re_done); 109 110 /* Flush Receives, then wait for deferred Reply work 111 * to complete. 112 */ 113 ib_drain_rq(id->qp); 114 115 /* Deferred Reply processing might have scheduled 116 * local invalidations. 117 */ 118 ib_drain_sq(id->qp); 119 120 rpcrdma_ep_put(ep); 121 } 122 123 /* Ensure xprt_force_disconnect() is invoked exactly once when a 124 * connection is closed or lost. (The important thing is it needs 125 * to be invoked "at least" once). 126 */ 127 void rpcrdma_force_disconnect(struct rpcrdma_ep *ep) 128 { 129 if (atomic_add_unless(&ep->re_force_disconnect, 1, 1)) 130 xprt_force_disconnect(ep->re_xprt); 131 } 132 133 /** 134 * rpcrdma_flush_disconnect - Disconnect on flushed completion 135 * @r_xprt: transport to disconnect 136 * @wc: work completion entry 137 * 138 * Must be called in process context. 139 */ 140 void rpcrdma_flush_disconnect(struct rpcrdma_xprt *r_xprt, struct ib_wc *wc) 141 { 142 if (wc->status != IB_WC_SUCCESS) 143 rpcrdma_force_disconnect(r_xprt->rx_ep); 144 } 145 146 /** 147 * rpcrdma_wc_send - Invoked by RDMA provider for each polled Send WC 148 * @cq: completion queue 149 * @wc: WCE for a completed Send WR 150 * 151 */ 152 static void rpcrdma_wc_send(struct ib_cq *cq, struct ib_wc *wc) 153 { 154 struct ib_cqe *cqe = wc->wr_cqe; 155 struct rpcrdma_sendctx *sc = 156 container_of(cqe, struct rpcrdma_sendctx, sc_cqe); 157 struct rpcrdma_xprt *r_xprt = cq->cq_context; 158 159 /* WARNING: Only wr_cqe and status are reliable at this point */ 160 trace_xprtrdma_wc_send(wc, &sc->sc_cid); 161 rpcrdma_sendctx_put_locked(r_xprt, sc); 162 rpcrdma_flush_disconnect(r_xprt, wc); 163 } 164 165 /** 166 * rpcrdma_wc_receive - Invoked by RDMA provider for each polled Receive WC 167 * @cq: completion queue 168 * @wc: WCE for a completed Receive WR 169 * 170 */ 171 static void rpcrdma_wc_receive(struct ib_cq *cq, struct ib_wc *wc) 172 { 173 struct ib_cqe *cqe = wc->wr_cqe; 174 struct rpcrdma_rep *rep = container_of(cqe, struct rpcrdma_rep, 175 rr_cqe); 176 struct rpcrdma_xprt *r_xprt = cq->cq_context; 177 178 /* WARNING: Only wr_cqe and status are reliable at this point */ 179 trace_xprtrdma_wc_receive(wc, &rep->rr_cid); 180 --r_xprt->rx_ep->re_receive_count; 181 if (wc->status != IB_WC_SUCCESS) 182 goto out_flushed; 183 184 /* status == SUCCESS means all fields in wc are trustworthy */ 185 rpcrdma_set_xdrlen(&rep->rr_hdrbuf, wc->byte_len); 186 rep->rr_wc_flags = wc->wc_flags; 187 rep->rr_inv_rkey = wc->ex.invalidate_rkey; 188 189 ib_dma_sync_single_for_cpu(rdmab_device(rep->rr_rdmabuf), 190 rdmab_addr(rep->rr_rdmabuf), 191 wc->byte_len, DMA_FROM_DEVICE); 192 193 rpcrdma_reply_handler(rep); 194 return; 195 196 out_flushed: 197 rpcrdma_flush_disconnect(r_xprt, wc); 198 rpcrdma_rep_put(&r_xprt->rx_buf, rep); 199 } 200 201 static void rpcrdma_update_cm_private(struct rpcrdma_ep *ep, 202 struct rdma_conn_param *param) 203 { 204 const struct rpcrdma_connect_private *pmsg = param->private_data; 205 unsigned int rsize, wsize; 206 207 /* Default settings for RPC-over-RDMA Version One */ 208 ep->re_implicit_roundup = xprt_rdma_pad_optimize; 209 rsize = RPCRDMA_V1_DEF_INLINE_SIZE; 210 wsize = RPCRDMA_V1_DEF_INLINE_SIZE; 211 212 if (pmsg && 213 pmsg->cp_magic == rpcrdma_cmp_magic && 214 pmsg->cp_version == RPCRDMA_CMP_VERSION) { 215 ep->re_implicit_roundup = true; 216 rsize = rpcrdma_decode_buffer_size(pmsg->cp_send_size); 217 wsize = rpcrdma_decode_buffer_size(pmsg->cp_recv_size); 218 } 219 220 if (rsize < ep->re_inline_recv) 221 ep->re_inline_recv = rsize; 222 if (wsize < ep->re_inline_send) 223 ep->re_inline_send = wsize; 224 225 rpcrdma_set_max_header_sizes(ep); 226 } 227 228 /** 229 * rpcrdma_cm_event_handler - Handle RDMA CM events 230 * @id: rdma_cm_id on which an event has occurred 231 * @event: details of the event 232 * 233 * Called with @id's mutex held. Returns 1 if caller should 234 * destroy @id, otherwise 0. 235 */ 236 static int 237 rpcrdma_cm_event_handler(struct rdma_cm_id *id, struct rdma_cm_event *event) 238 { 239 struct sockaddr *sap = (struct sockaddr *)&id->route.addr.dst_addr; 240 struct rpcrdma_ep *ep = id->context; 241 242 might_sleep(); 243 244 switch (event->event) { 245 case RDMA_CM_EVENT_ADDR_RESOLVED: 246 case RDMA_CM_EVENT_ROUTE_RESOLVED: 247 ep->re_async_rc = 0; 248 complete(&ep->re_done); 249 return 0; 250 case RDMA_CM_EVENT_ADDR_ERROR: 251 ep->re_async_rc = -EPROTO; 252 complete(&ep->re_done); 253 return 0; 254 case RDMA_CM_EVENT_ROUTE_ERROR: 255 ep->re_async_rc = -ENETUNREACH; 256 complete(&ep->re_done); 257 return 0; 258 case RDMA_CM_EVENT_DEVICE_REMOVAL: 259 pr_info("rpcrdma: removing device %s for %pISpc\n", 260 ep->re_id->device->name, sap); 261 fallthrough; 262 case RDMA_CM_EVENT_ADDR_CHANGE: 263 ep->re_connect_status = -ENODEV; 264 goto disconnected; 265 case RDMA_CM_EVENT_ESTABLISHED: 266 rpcrdma_ep_get(ep); 267 ep->re_connect_status = 1; 268 rpcrdma_update_cm_private(ep, &event->param.conn); 269 trace_xprtrdma_inline_thresh(ep); 270 wake_up_all(&ep->re_connect_wait); 271 break; 272 case RDMA_CM_EVENT_CONNECT_ERROR: 273 ep->re_connect_status = -ENOTCONN; 274 goto wake_connect_worker; 275 case RDMA_CM_EVENT_UNREACHABLE: 276 ep->re_connect_status = -ENETUNREACH; 277 goto wake_connect_worker; 278 case RDMA_CM_EVENT_REJECTED: 279 dprintk("rpcrdma: connection to %pISpc rejected: %s\n", 280 sap, rdma_reject_msg(id, event->status)); 281 ep->re_connect_status = -ECONNREFUSED; 282 if (event->status == IB_CM_REJ_STALE_CONN) 283 ep->re_connect_status = -ENOTCONN; 284 wake_connect_worker: 285 wake_up_all(&ep->re_connect_wait); 286 return 0; 287 case RDMA_CM_EVENT_DISCONNECTED: 288 ep->re_connect_status = -ECONNABORTED; 289 disconnected: 290 rpcrdma_force_disconnect(ep); 291 return rpcrdma_ep_put(ep); 292 default: 293 break; 294 } 295 296 dprintk("RPC: %s: %pISpc on %s/frwr: %s\n", __func__, sap, 297 ep->re_id->device->name, rdma_event_msg(event->event)); 298 return 0; 299 } 300 301 static struct rdma_cm_id *rpcrdma_create_id(struct rpcrdma_xprt *r_xprt, 302 struct rpcrdma_ep *ep) 303 { 304 unsigned long wtimeout = msecs_to_jiffies(RDMA_RESOLVE_TIMEOUT) + 1; 305 struct rpc_xprt *xprt = &r_xprt->rx_xprt; 306 struct rdma_cm_id *id; 307 int rc; 308 309 init_completion(&ep->re_done); 310 311 id = rdma_create_id(xprt->xprt_net, rpcrdma_cm_event_handler, ep, 312 RDMA_PS_TCP, IB_QPT_RC); 313 if (IS_ERR(id)) 314 return id; 315 316 ep->re_async_rc = -ETIMEDOUT; 317 rc = rdma_resolve_addr(id, NULL, (struct sockaddr *)&xprt->addr, 318 RDMA_RESOLVE_TIMEOUT); 319 if (rc) 320 goto out; 321 rc = wait_for_completion_interruptible_timeout(&ep->re_done, wtimeout); 322 if (rc < 0) 323 goto out; 324 325 rc = ep->re_async_rc; 326 if (rc) 327 goto out; 328 329 ep->re_async_rc = -ETIMEDOUT; 330 rc = rdma_resolve_route(id, RDMA_RESOLVE_TIMEOUT); 331 if (rc) 332 goto out; 333 rc = wait_for_completion_interruptible_timeout(&ep->re_done, wtimeout); 334 if (rc < 0) 335 goto out; 336 rc = ep->re_async_rc; 337 if (rc) 338 goto out; 339 340 return id; 341 342 out: 343 rdma_destroy_id(id); 344 return ERR_PTR(rc); 345 } 346 347 static void rpcrdma_ep_destroy(struct kref *kref) 348 { 349 struct rpcrdma_ep *ep = container_of(kref, struct rpcrdma_ep, re_kref); 350 351 if (ep->re_id->qp) { 352 rdma_destroy_qp(ep->re_id); 353 ep->re_id->qp = NULL; 354 } 355 356 if (ep->re_attr.recv_cq) 357 ib_free_cq(ep->re_attr.recv_cq); 358 ep->re_attr.recv_cq = NULL; 359 if (ep->re_attr.send_cq) 360 ib_free_cq(ep->re_attr.send_cq); 361 ep->re_attr.send_cq = NULL; 362 363 if (ep->re_pd) 364 ib_dealloc_pd(ep->re_pd); 365 ep->re_pd = NULL; 366 367 kfree(ep); 368 module_put(THIS_MODULE); 369 } 370 371 static noinline void rpcrdma_ep_get(struct rpcrdma_ep *ep) 372 { 373 kref_get(&ep->re_kref); 374 } 375 376 /* Returns: 377 * %0 if @ep still has a positive kref count, or 378 * %1 if @ep was destroyed successfully. 379 */ 380 static noinline int rpcrdma_ep_put(struct rpcrdma_ep *ep) 381 { 382 return kref_put(&ep->re_kref, rpcrdma_ep_destroy); 383 } 384 385 static int rpcrdma_ep_create(struct rpcrdma_xprt *r_xprt) 386 { 387 struct rpcrdma_connect_private *pmsg; 388 struct ib_device *device; 389 struct rdma_cm_id *id; 390 struct rpcrdma_ep *ep; 391 int rc; 392 393 ep = kzalloc(sizeof(*ep), GFP_NOFS); 394 if (!ep) 395 return -ENOTCONN; 396 ep->re_xprt = &r_xprt->rx_xprt; 397 kref_init(&ep->re_kref); 398 399 id = rpcrdma_create_id(r_xprt, ep); 400 if (IS_ERR(id)) { 401 kfree(ep); 402 return PTR_ERR(id); 403 } 404 __module_get(THIS_MODULE); 405 device = id->device; 406 ep->re_id = id; 407 reinit_completion(&ep->re_done); 408 409 ep->re_max_requests = r_xprt->rx_xprt.max_reqs; 410 ep->re_inline_send = xprt_rdma_max_inline_write; 411 ep->re_inline_recv = xprt_rdma_max_inline_read; 412 rc = frwr_query_device(ep, device); 413 if (rc) 414 goto out_destroy; 415 416 r_xprt->rx_buf.rb_max_requests = cpu_to_be32(ep->re_max_requests); 417 418 ep->re_attr.srq = NULL; 419 ep->re_attr.cap.max_inline_data = 0; 420 ep->re_attr.sq_sig_type = IB_SIGNAL_REQ_WR; 421 ep->re_attr.qp_type = IB_QPT_RC; 422 ep->re_attr.port_num = ~0; 423 424 dprintk("RPC: %s: requested max: dtos: send %d recv %d; " 425 "iovs: send %d recv %d\n", 426 __func__, 427 ep->re_attr.cap.max_send_wr, 428 ep->re_attr.cap.max_recv_wr, 429 ep->re_attr.cap.max_send_sge, 430 ep->re_attr.cap.max_recv_sge); 431 432 ep->re_send_batch = ep->re_max_requests >> 3; 433 ep->re_send_count = ep->re_send_batch; 434 init_waitqueue_head(&ep->re_connect_wait); 435 436 ep->re_attr.send_cq = ib_alloc_cq_any(device, r_xprt, 437 ep->re_attr.cap.max_send_wr, 438 IB_POLL_WORKQUEUE); 439 if (IS_ERR(ep->re_attr.send_cq)) { 440 rc = PTR_ERR(ep->re_attr.send_cq); 441 goto out_destroy; 442 } 443 444 ep->re_attr.recv_cq = ib_alloc_cq_any(device, r_xprt, 445 ep->re_attr.cap.max_recv_wr, 446 IB_POLL_WORKQUEUE); 447 if (IS_ERR(ep->re_attr.recv_cq)) { 448 rc = PTR_ERR(ep->re_attr.recv_cq); 449 goto out_destroy; 450 } 451 ep->re_receive_count = 0; 452 453 /* Initialize cma parameters */ 454 memset(&ep->re_remote_cma, 0, sizeof(ep->re_remote_cma)); 455 456 /* Prepare RDMA-CM private message */ 457 pmsg = &ep->re_cm_private; 458 pmsg->cp_magic = rpcrdma_cmp_magic; 459 pmsg->cp_version = RPCRDMA_CMP_VERSION; 460 pmsg->cp_flags |= RPCRDMA_CMP_F_SND_W_INV_OK; 461 pmsg->cp_send_size = rpcrdma_encode_buffer_size(ep->re_inline_send); 462 pmsg->cp_recv_size = rpcrdma_encode_buffer_size(ep->re_inline_recv); 463 ep->re_remote_cma.private_data = pmsg; 464 ep->re_remote_cma.private_data_len = sizeof(*pmsg); 465 466 /* Client offers RDMA Read but does not initiate */ 467 ep->re_remote_cma.initiator_depth = 0; 468 ep->re_remote_cma.responder_resources = 469 min_t(int, U8_MAX, device->attrs.max_qp_rd_atom); 470 471 /* Limit transport retries so client can detect server 472 * GID changes quickly. RPC layer handles re-establishing 473 * transport connection and retransmission. 474 */ 475 ep->re_remote_cma.retry_count = 6; 476 477 /* RPC-over-RDMA handles its own flow control. In addition, 478 * make all RNR NAKs visible so we know that RPC-over-RDMA 479 * flow control is working correctly (no NAKs should be seen). 480 */ 481 ep->re_remote_cma.flow_control = 0; 482 ep->re_remote_cma.rnr_retry_count = 0; 483 484 ep->re_pd = ib_alloc_pd(device, 0); 485 if (IS_ERR(ep->re_pd)) { 486 rc = PTR_ERR(ep->re_pd); 487 goto out_destroy; 488 } 489 490 rc = rdma_create_qp(id, ep->re_pd, &ep->re_attr); 491 if (rc) 492 goto out_destroy; 493 494 r_xprt->rx_ep = ep; 495 return 0; 496 497 out_destroy: 498 rpcrdma_ep_put(ep); 499 rdma_destroy_id(id); 500 return rc; 501 } 502 503 /** 504 * rpcrdma_xprt_connect - Connect an unconnected transport 505 * @r_xprt: controlling transport instance 506 * 507 * Returns 0 on success or a negative errno. 508 */ 509 int rpcrdma_xprt_connect(struct rpcrdma_xprt *r_xprt) 510 { 511 struct rpc_xprt *xprt = &r_xprt->rx_xprt; 512 struct rpcrdma_ep *ep; 513 int rc; 514 515 rc = rpcrdma_ep_create(r_xprt); 516 if (rc) 517 return rc; 518 ep = r_xprt->rx_ep; 519 520 xprt_clear_connected(xprt); 521 rpcrdma_reset_cwnd(r_xprt); 522 523 /* Bump the ep's reference count while there are 524 * outstanding Receives. 525 */ 526 rpcrdma_ep_get(ep); 527 rpcrdma_post_recvs(r_xprt, 1, true); 528 529 rc = rdma_connect(ep->re_id, &ep->re_remote_cma); 530 if (rc) 531 goto out; 532 533 if (xprt->reestablish_timeout < RPCRDMA_INIT_REEST_TO) 534 xprt->reestablish_timeout = RPCRDMA_INIT_REEST_TO; 535 wait_event_interruptible(ep->re_connect_wait, 536 ep->re_connect_status != 0); 537 if (ep->re_connect_status <= 0) { 538 rc = ep->re_connect_status; 539 goto out; 540 } 541 542 rc = rpcrdma_sendctxs_create(r_xprt); 543 if (rc) { 544 rc = -ENOTCONN; 545 goto out; 546 } 547 548 rc = rpcrdma_reqs_setup(r_xprt); 549 if (rc) { 550 rc = -ENOTCONN; 551 goto out; 552 } 553 rpcrdma_mrs_create(r_xprt); 554 555 out: 556 trace_xprtrdma_connect(r_xprt, rc); 557 return rc; 558 } 559 560 /** 561 * rpcrdma_xprt_disconnect - Disconnect underlying transport 562 * @r_xprt: controlling transport instance 563 * 564 * Caller serializes. Either the transport send lock is held, 565 * or we're being called to destroy the transport. 566 * 567 * On return, @r_xprt is completely divested of all hardware 568 * resources and prepared for the next ->connect operation. 569 */ 570 void rpcrdma_xprt_disconnect(struct rpcrdma_xprt *r_xprt) 571 { 572 struct rpcrdma_ep *ep = r_xprt->rx_ep; 573 struct rdma_cm_id *id; 574 int rc; 575 576 if (!ep) 577 return; 578 579 id = ep->re_id; 580 rc = rdma_disconnect(id); 581 trace_xprtrdma_disconnect(r_xprt, rc); 582 583 rpcrdma_xprt_drain(r_xprt); 584 rpcrdma_reps_unmap(r_xprt); 585 rpcrdma_reqs_reset(r_xprt); 586 rpcrdma_mrs_destroy(r_xprt); 587 rpcrdma_sendctxs_destroy(r_xprt); 588 589 if (rpcrdma_ep_put(ep)) 590 rdma_destroy_id(id); 591 592 r_xprt->rx_ep = NULL; 593 } 594 595 /* Fixed-size circular FIFO queue. This implementation is wait-free and 596 * lock-free. 597 * 598 * Consumer is the code path that posts Sends. This path dequeues a 599 * sendctx for use by a Send operation. Multiple consumer threads 600 * are serialized by the RPC transport lock, which allows only one 601 * ->send_request call at a time. 602 * 603 * Producer is the code path that handles Send completions. This path 604 * enqueues a sendctx that has been completed. Multiple producer 605 * threads are serialized by the ib_poll_cq() function. 606 */ 607 608 /* rpcrdma_sendctxs_destroy() assumes caller has already quiesced 609 * queue activity, and rpcrdma_xprt_drain has flushed all remaining 610 * Send requests. 611 */ 612 static void rpcrdma_sendctxs_destroy(struct rpcrdma_xprt *r_xprt) 613 { 614 struct rpcrdma_buffer *buf = &r_xprt->rx_buf; 615 unsigned long i; 616 617 if (!buf->rb_sc_ctxs) 618 return; 619 for (i = 0; i <= buf->rb_sc_last; i++) 620 kfree(buf->rb_sc_ctxs[i]); 621 kfree(buf->rb_sc_ctxs); 622 buf->rb_sc_ctxs = NULL; 623 } 624 625 static struct rpcrdma_sendctx *rpcrdma_sendctx_create(struct rpcrdma_ep *ep) 626 { 627 struct rpcrdma_sendctx *sc; 628 629 sc = kzalloc(struct_size(sc, sc_sges, ep->re_attr.cap.max_send_sge), 630 GFP_KERNEL); 631 if (!sc) 632 return NULL; 633 634 sc->sc_cqe.done = rpcrdma_wc_send; 635 sc->sc_cid.ci_queue_id = ep->re_attr.send_cq->res.id; 636 sc->sc_cid.ci_completion_id = 637 atomic_inc_return(&ep->re_completion_ids); 638 return sc; 639 } 640 641 static int rpcrdma_sendctxs_create(struct rpcrdma_xprt *r_xprt) 642 { 643 struct rpcrdma_buffer *buf = &r_xprt->rx_buf; 644 struct rpcrdma_sendctx *sc; 645 unsigned long i; 646 647 /* Maximum number of concurrent outstanding Send WRs. Capping 648 * the circular queue size stops Send Queue overflow by causing 649 * the ->send_request call to fail temporarily before too many 650 * Sends are posted. 651 */ 652 i = r_xprt->rx_ep->re_max_requests + RPCRDMA_MAX_BC_REQUESTS; 653 buf->rb_sc_ctxs = kcalloc(i, sizeof(sc), GFP_KERNEL); 654 if (!buf->rb_sc_ctxs) 655 return -ENOMEM; 656 657 buf->rb_sc_last = i - 1; 658 for (i = 0; i <= buf->rb_sc_last; i++) { 659 sc = rpcrdma_sendctx_create(r_xprt->rx_ep); 660 if (!sc) 661 return -ENOMEM; 662 663 buf->rb_sc_ctxs[i] = sc; 664 } 665 666 buf->rb_sc_head = 0; 667 buf->rb_sc_tail = 0; 668 return 0; 669 } 670 671 /* The sendctx queue is not guaranteed to have a size that is a 672 * power of two, thus the helpers in circ_buf.h cannot be used. 673 * The other option is to use modulus (%), which can be expensive. 674 */ 675 static unsigned long rpcrdma_sendctx_next(struct rpcrdma_buffer *buf, 676 unsigned long item) 677 { 678 return likely(item < buf->rb_sc_last) ? item + 1 : 0; 679 } 680 681 /** 682 * rpcrdma_sendctx_get_locked - Acquire a send context 683 * @r_xprt: controlling transport instance 684 * 685 * Returns pointer to a free send completion context; or NULL if 686 * the queue is empty. 687 * 688 * Usage: Called to acquire an SGE array before preparing a Send WR. 689 * 690 * The caller serializes calls to this function (per transport), and 691 * provides an effective memory barrier that flushes the new value 692 * of rb_sc_head. 693 */ 694 struct rpcrdma_sendctx *rpcrdma_sendctx_get_locked(struct rpcrdma_xprt *r_xprt) 695 { 696 struct rpcrdma_buffer *buf = &r_xprt->rx_buf; 697 struct rpcrdma_sendctx *sc; 698 unsigned long next_head; 699 700 next_head = rpcrdma_sendctx_next(buf, buf->rb_sc_head); 701 702 if (next_head == READ_ONCE(buf->rb_sc_tail)) 703 goto out_emptyq; 704 705 /* ORDER: item must be accessed _before_ head is updated */ 706 sc = buf->rb_sc_ctxs[next_head]; 707 708 /* Releasing the lock in the caller acts as a memory 709 * barrier that flushes rb_sc_head. 710 */ 711 buf->rb_sc_head = next_head; 712 713 return sc; 714 715 out_emptyq: 716 /* The queue is "empty" if there have not been enough Send 717 * completions recently. This is a sign the Send Queue is 718 * backing up. Cause the caller to pause and try again. 719 */ 720 xprt_wait_for_buffer_space(&r_xprt->rx_xprt); 721 r_xprt->rx_stats.empty_sendctx_q++; 722 return NULL; 723 } 724 725 /** 726 * rpcrdma_sendctx_put_locked - Release a send context 727 * @r_xprt: controlling transport instance 728 * @sc: send context to release 729 * 730 * Usage: Called from Send completion to return a sendctxt 731 * to the queue. 732 * 733 * The caller serializes calls to this function (per transport). 734 */ 735 static void rpcrdma_sendctx_put_locked(struct rpcrdma_xprt *r_xprt, 736 struct rpcrdma_sendctx *sc) 737 { 738 struct rpcrdma_buffer *buf = &r_xprt->rx_buf; 739 unsigned long next_tail; 740 741 /* Unmap SGEs of previously completed but unsignaled 742 * Sends by walking up the queue until @sc is found. 743 */ 744 next_tail = buf->rb_sc_tail; 745 do { 746 next_tail = rpcrdma_sendctx_next(buf, next_tail); 747 748 /* ORDER: item must be accessed _before_ tail is updated */ 749 rpcrdma_sendctx_unmap(buf->rb_sc_ctxs[next_tail]); 750 751 } while (buf->rb_sc_ctxs[next_tail] != sc); 752 753 /* Paired with READ_ONCE */ 754 smp_store_release(&buf->rb_sc_tail, next_tail); 755 756 xprt_write_space(&r_xprt->rx_xprt); 757 } 758 759 static void 760 rpcrdma_mrs_create(struct rpcrdma_xprt *r_xprt) 761 { 762 struct rpcrdma_buffer *buf = &r_xprt->rx_buf; 763 struct rpcrdma_ep *ep = r_xprt->rx_ep; 764 unsigned int count; 765 766 for (count = 0; count < ep->re_max_rdma_segs; count++) { 767 struct rpcrdma_mr *mr; 768 int rc; 769 770 mr = kzalloc(sizeof(*mr), GFP_NOFS); 771 if (!mr) 772 break; 773 774 rc = frwr_mr_init(r_xprt, mr); 775 if (rc) { 776 kfree(mr); 777 break; 778 } 779 780 spin_lock(&buf->rb_lock); 781 rpcrdma_mr_push(mr, &buf->rb_mrs); 782 list_add(&mr->mr_all, &buf->rb_all_mrs); 783 spin_unlock(&buf->rb_lock); 784 } 785 786 r_xprt->rx_stats.mrs_allocated += count; 787 trace_xprtrdma_createmrs(r_xprt, count); 788 } 789 790 static void 791 rpcrdma_mr_refresh_worker(struct work_struct *work) 792 { 793 struct rpcrdma_buffer *buf = container_of(work, struct rpcrdma_buffer, 794 rb_refresh_worker); 795 struct rpcrdma_xprt *r_xprt = container_of(buf, struct rpcrdma_xprt, 796 rx_buf); 797 798 rpcrdma_mrs_create(r_xprt); 799 xprt_write_space(&r_xprt->rx_xprt); 800 } 801 802 /** 803 * rpcrdma_mrs_refresh - Wake the MR refresh worker 804 * @r_xprt: controlling transport instance 805 * 806 */ 807 void rpcrdma_mrs_refresh(struct rpcrdma_xprt *r_xprt) 808 { 809 struct rpcrdma_buffer *buf = &r_xprt->rx_buf; 810 struct rpcrdma_ep *ep = r_xprt->rx_ep; 811 812 /* If there is no underlying connection, it's no use 813 * to wake the refresh worker. 814 */ 815 if (ep->re_connect_status == 1) { 816 /* The work is scheduled on a WQ_MEM_RECLAIM 817 * workqueue in order to prevent MR allocation 818 * from recursing into NFS during direct reclaim. 819 */ 820 queue_work(xprtiod_workqueue, &buf->rb_refresh_worker); 821 } 822 } 823 824 /** 825 * rpcrdma_req_create - Allocate an rpcrdma_req object 826 * @r_xprt: controlling r_xprt 827 * @size: initial size, in bytes, of send and receive buffers 828 * @flags: GFP flags passed to memory allocators 829 * 830 * Returns an allocated and fully initialized rpcrdma_req or NULL. 831 */ 832 struct rpcrdma_req *rpcrdma_req_create(struct rpcrdma_xprt *r_xprt, size_t size, 833 gfp_t flags) 834 { 835 struct rpcrdma_buffer *buffer = &r_xprt->rx_buf; 836 struct rpcrdma_req *req; 837 838 req = kzalloc(sizeof(*req), flags); 839 if (req == NULL) 840 goto out1; 841 842 req->rl_sendbuf = rpcrdma_regbuf_alloc(size, DMA_TO_DEVICE, flags); 843 if (!req->rl_sendbuf) 844 goto out2; 845 846 req->rl_recvbuf = rpcrdma_regbuf_alloc(size, DMA_NONE, flags); 847 if (!req->rl_recvbuf) 848 goto out3; 849 850 INIT_LIST_HEAD(&req->rl_free_mrs); 851 INIT_LIST_HEAD(&req->rl_registered); 852 spin_lock(&buffer->rb_lock); 853 list_add(&req->rl_all, &buffer->rb_allreqs); 854 spin_unlock(&buffer->rb_lock); 855 return req; 856 857 out3: 858 kfree(req->rl_sendbuf); 859 out2: 860 kfree(req); 861 out1: 862 return NULL; 863 } 864 865 /** 866 * rpcrdma_req_setup - Per-connection instance setup of an rpcrdma_req object 867 * @r_xprt: controlling transport instance 868 * @req: rpcrdma_req object to set up 869 * 870 * Returns zero on success, and a negative errno on failure. 871 */ 872 int rpcrdma_req_setup(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req) 873 { 874 struct rpcrdma_regbuf *rb; 875 size_t maxhdrsize; 876 877 /* Compute maximum header buffer size in bytes */ 878 maxhdrsize = rpcrdma_fixed_maxsz + 3 + 879 r_xprt->rx_ep->re_max_rdma_segs * rpcrdma_readchunk_maxsz; 880 maxhdrsize *= sizeof(__be32); 881 rb = rpcrdma_regbuf_alloc(__roundup_pow_of_two(maxhdrsize), 882 DMA_TO_DEVICE, GFP_KERNEL); 883 if (!rb) 884 goto out; 885 886 if (!__rpcrdma_regbuf_dma_map(r_xprt, rb)) 887 goto out_free; 888 889 req->rl_rdmabuf = rb; 890 xdr_buf_init(&req->rl_hdrbuf, rdmab_data(rb), rdmab_length(rb)); 891 return 0; 892 893 out_free: 894 rpcrdma_regbuf_free(rb); 895 out: 896 return -ENOMEM; 897 } 898 899 /* ASSUMPTION: the rb_allreqs list is stable for the duration, 900 * and thus can be walked without holding rb_lock. Eg. the 901 * caller is holding the transport send lock to exclude 902 * device removal or disconnection. 903 */ 904 static int rpcrdma_reqs_setup(struct rpcrdma_xprt *r_xprt) 905 { 906 struct rpcrdma_buffer *buf = &r_xprt->rx_buf; 907 struct rpcrdma_req *req; 908 int rc; 909 910 list_for_each_entry(req, &buf->rb_allreqs, rl_all) { 911 rc = rpcrdma_req_setup(r_xprt, req); 912 if (rc) 913 return rc; 914 } 915 return 0; 916 } 917 918 static void rpcrdma_req_reset(struct rpcrdma_req *req) 919 { 920 /* Credits are valid for only one connection */ 921 req->rl_slot.rq_cong = 0; 922 923 rpcrdma_regbuf_free(req->rl_rdmabuf); 924 req->rl_rdmabuf = NULL; 925 926 rpcrdma_regbuf_dma_unmap(req->rl_sendbuf); 927 rpcrdma_regbuf_dma_unmap(req->rl_recvbuf); 928 929 frwr_reset(req); 930 } 931 932 /* ASSUMPTION: the rb_allreqs list is stable for the duration, 933 * and thus can be walked without holding rb_lock. Eg. the 934 * caller is holding the transport send lock to exclude 935 * device removal or disconnection. 936 */ 937 static void rpcrdma_reqs_reset(struct rpcrdma_xprt *r_xprt) 938 { 939 struct rpcrdma_buffer *buf = &r_xprt->rx_buf; 940 struct rpcrdma_req *req; 941 942 list_for_each_entry(req, &buf->rb_allreqs, rl_all) 943 rpcrdma_req_reset(req); 944 } 945 946 static noinline 947 struct rpcrdma_rep *rpcrdma_rep_create(struct rpcrdma_xprt *r_xprt, 948 bool temp) 949 { 950 struct rpcrdma_buffer *buf = &r_xprt->rx_buf; 951 struct rpcrdma_rep *rep; 952 953 rep = kzalloc(sizeof(*rep), GFP_KERNEL); 954 if (rep == NULL) 955 goto out; 956 957 rep->rr_rdmabuf = rpcrdma_regbuf_alloc(r_xprt->rx_ep->re_inline_recv, 958 DMA_FROM_DEVICE, GFP_KERNEL); 959 if (!rep->rr_rdmabuf) 960 goto out_free; 961 962 if (!rpcrdma_regbuf_dma_map(r_xprt, rep->rr_rdmabuf)) 963 goto out_free_regbuf; 964 965 rep->rr_cid.ci_completion_id = 966 atomic_inc_return(&r_xprt->rx_ep->re_completion_ids); 967 968 xdr_buf_init(&rep->rr_hdrbuf, rdmab_data(rep->rr_rdmabuf), 969 rdmab_length(rep->rr_rdmabuf)); 970 rep->rr_cqe.done = rpcrdma_wc_receive; 971 rep->rr_rxprt = r_xprt; 972 rep->rr_recv_wr.next = NULL; 973 rep->rr_recv_wr.wr_cqe = &rep->rr_cqe; 974 rep->rr_recv_wr.sg_list = &rep->rr_rdmabuf->rg_iov; 975 rep->rr_recv_wr.num_sge = 1; 976 rep->rr_temp = temp; 977 978 spin_lock(&buf->rb_lock); 979 list_add(&rep->rr_all, &buf->rb_all_reps); 980 spin_unlock(&buf->rb_lock); 981 return rep; 982 983 out_free_regbuf: 984 rpcrdma_regbuf_free(rep->rr_rdmabuf); 985 out_free: 986 kfree(rep); 987 out: 988 return NULL; 989 } 990 991 static void rpcrdma_rep_free(struct rpcrdma_rep *rep) 992 { 993 rpcrdma_regbuf_free(rep->rr_rdmabuf); 994 kfree(rep); 995 } 996 997 static void rpcrdma_rep_destroy(struct rpcrdma_rep *rep) 998 { 999 struct rpcrdma_buffer *buf = &rep->rr_rxprt->rx_buf; 1000 1001 spin_lock(&buf->rb_lock); 1002 list_del(&rep->rr_all); 1003 spin_unlock(&buf->rb_lock); 1004 1005 rpcrdma_rep_free(rep); 1006 } 1007 1008 static struct rpcrdma_rep *rpcrdma_rep_get_locked(struct rpcrdma_buffer *buf) 1009 { 1010 struct llist_node *node; 1011 1012 /* Calls to llist_del_first are required to be serialized */ 1013 node = llist_del_first(&buf->rb_free_reps); 1014 if (!node) 1015 return NULL; 1016 return llist_entry(node, struct rpcrdma_rep, rr_node); 1017 } 1018 1019 /** 1020 * rpcrdma_rep_put - Release rpcrdma_rep back to free list 1021 * @buf: buffer pool 1022 * @rep: rep to release 1023 * 1024 */ 1025 void rpcrdma_rep_put(struct rpcrdma_buffer *buf, struct rpcrdma_rep *rep) 1026 { 1027 llist_add(&rep->rr_node, &buf->rb_free_reps); 1028 } 1029 1030 /* Caller must ensure the QP is quiescent (RQ is drained) before 1031 * invoking this function, to guarantee rb_all_reps is not 1032 * changing. 1033 */ 1034 static void rpcrdma_reps_unmap(struct rpcrdma_xprt *r_xprt) 1035 { 1036 struct rpcrdma_buffer *buf = &r_xprt->rx_buf; 1037 struct rpcrdma_rep *rep; 1038 1039 list_for_each_entry(rep, &buf->rb_all_reps, rr_all) { 1040 rpcrdma_regbuf_dma_unmap(rep->rr_rdmabuf); 1041 rep->rr_temp = true; /* Mark this rep for destruction */ 1042 } 1043 } 1044 1045 static void rpcrdma_reps_destroy(struct rpcrdma_buffer *buf) 1046 { 1047 struct rpcrdma_rep *rep; 1048 1049 spin_lock(&buf->rb_lock); 1050 while ((rep = list_first_entry_or_null(&buf->rb_all_reps, 1051 struct rpcrdma_rep, 1052 rr_all)) != NULL) { 1053 list_del(&rep->rr_all); 1054 spin_unlock(&buf->rb_lock); 1055 1056 rpcrdma_rep_free(rep); 1057 1058 spin_lock(&buf->rb_lock); 1059 } 1060 spin_unlock(&buf->rb_lock); 1061 } 1062 1063 /** 1064 * rpcrdma_buffer_create - Create initial set of req/rep objects 1065 * @r_xprt: transport instance to (re)initialize 1066 * 1067 * Returns zero on success, otherwise a negative errno. 1068 */ 1069 int rpcrdma_buffer_create(struct rpcrdma_xprt *r_xprt) 1070 { 1071 struct rpcrdma_buffer *buf = &r_xprt->rx_buf; 1072 int i, rc; 1073 1074 buf->rb_bc_srv_max_requests = 0; 1075 spin_lock_init(&buf->rb_lock); 1076 INIT_LIST_HEAD(&buf->rb_mrs); 1077 INIT_LIST_HEAD(&buf->rb_all_mrs); 1078 INIT_WORK(&buf->rb_refresh_worker, rpcrdma_mr_refresh_worker); 1079 1080 INIT_LIST_HEAD(&buf->rb_send_bufs); 1081 INIT_LIST_HEAD(&buf->rb_allreqs); 1082 INIT_LIST_HEAD(&buf->rb_all_reps); 1083 1084 rc = -ENOMEM; 1085 for (i = 0; i < r_xprt->rx_xprt.max_reqs; i++) { 1086 struct rpcrdma_req *req; 1087 1088 req = rpcrdma_req_create(r_xprt, RPCRDMA_V1_DEF_INLINE_SIZE * 2, 1089 GFP_KERNEL); 1090 if (!req) 1091 goto out; 1092 list_add(&req->rl_list, &buf->rb_send_bufs); 1093 } 1094 1095 init_llist_head(&buf->rb_free_reps); 1096 1097 return 0; 1098 out: 1099 rpcrdma_buffer_destroy(buf); 1100 return rc; 1101 } 1102 1103 /** 1104 * rpcrdma_req_destroy - Destroy an rpcrdma_req object 1105 * @req: unused object to be destroyed 1106 * 1107 * Relies on caller holding the transport send lock to protect 1108 * removing req->rl_all from buf->rb_all_reqs safely. 1109 */ 1110 void rpcrdma_req_destroy(struct rpcrdma_req *req) 1111 { 1112 struct rpcrdma_mr *mr; 1113 1114 list_del(&req->rl_all); 1115 1116 while ((mr = rpcrdma_mr_pop(&req->rl_free_mrs))) { 1117 struct rpcrdma_buffer *buf = &mr->mr_xprt->rx_buf; 1118 1119 spin_lock(&buf->rb_lock); 1120 list_del(&mr->mr_all); 1121 spin_unlock(&buf->rb_lock); 1122 1123 frwr_mr_release(mr); 1124 } 1125 1126 rpcrdma_regbuf_free(req->rl_recvbuf); 1127 rpcrdma_regbuf_free(req->rl_sendbuf); 1128 rpcrdma_regbuf_free(req->rl_rdmabuf); 1129 kfree(req); 1130 } 1131 1132 /** 1133 * rpcrdma_mrs_destroy - Release all of a transport's MRs 1134 * @r_xprt: controlling transport instance 1135 * 1136 * Relies on caller holding the transport send lock to protect 1137 * removing mr->mr_list from req->rl_free_mrs safely. 1138 */ 1139 static void rpcrdma_mrs_destroy(struct rpcrdma_xprt *r_xprt) 1140 { 1141 struct rpcrdma_buffer *buf = &r_xprt->rx_buf; 1142 struct rpcrdma_mr *mr; 1143 1144 cancel_work_sync(&buf->rb_refresh_worker); 1145 1146 spin_lock(&buf->rb_lock); 1147 while ((mr = list_first_entry_or_null(&buf->rb_all_mrs, 1148 struct rpcrdma_mr, 1149 mr_all)) != NULL) { 1150 list_del(&mr->mr_list); 1151 list_del(&mr->mr_all); 1152 spin_unlock(&buf->rb_lock); 1153 1154 frwr_mr_release(mr); 1155 1156 spin_lock(&buf->rb_lock); 1157 } 1158 spin_unlock(&buf->rb_lock); 1159 } 1160 1161 /** 1162 * rpcrdma_buffer_destroy - Release all hw resources 1163 * @buf: root control block for resources 1164 * 1165 * ORDERING: relies on a prior rpcrdma_xprt_drain : 1166 * - No more Send or Receive completions can occur 1167 * - All MRs, reps, and reqs are returned to their free lists 1168 */ 1169 void 1170 rpcrdma_buffer_destroy(struct rpcrdma_buffer *buf) 1171 { 1172 rpcrdma_reps_destroy(buf); 1173 1174 while (!list_empty(&buf->rb_send_bufs)) { 1175 struct rpcrdma_req *req; 1176 1177 req = list_first_entry(&buf->rb_send_bufs, 1178 struct rpcrdma_req, rl_list); 1179 list_del(&req->rl_list); 1180 rpcrdma_req_destroy(req); 1181 } 1182 } 1183 1184 /** 1185 * rpcrdma_mr_get - Allocate an rpcrdma_mr object 1186 * @r_xprt: controlling transport 1187 * 1188 * Returns an initialized rpcrdma_mr or NULL if no free 1189 * rpcrdma_mr objects are available. 1190 */ 1191 struct rpcrdma_mr * 1192 rpcrdma_mr_get(struct rpcrdma_xprt *r_xprt) 1193 { 1194 struct rpcrdma_buffer *buf = &r_xprt->rx_buf; 1195 struct rpcrdma_mr *mr; 1196 1197 spin_lock(&buf->rb_lock); 1198 mr = rpcrdma_mr_pop(&buf->rb_mrs); 1199 spin_unlock(&buf->rb_lock); 1200 return mr; 1201 } 1202 1203 /** 1204 * rpcrdma_reply_put - Put reply buffers back into pool 1205 * @buffers: buffer pool 1206 * @req: object to return 1207 * 1208 */ 1209 void rpcrdma_reply_put(struct rpcrdma_buffer *buffers, struct rpcrdma_req *req) 1210 { 1211 if (req->rl_reply) { 1212 rpcrdma_rep_put(buffers, req->rl_reply); 1213 req->rl_reply = NULL; 1214 } 1215 } 1216 1217 /** 1218 * rpcrdma_buffer_get - Get a request buffer 1219 * @buffers: Buffer pool from which to obtain a buffer 1220 * 1221 * Returns a fresh rpcrdma_req, or NULL if none are available. 1222 */ 1223 struct rpcrdma_req * 1224 rpcrdma_buffer_get(struct rpcrdma_buffer *buffers) 1225 { 1226 struct rpcrdma_req *req; 1227 1228 spin_lock(&buffers->rb_lock); 1229 req = list_first_entry_or_null(&buffers->rb_send_bufs, 1230 struct rpcrdma_req, rl_list); 1231 if (req) 1232 list_del_init(&req->rl_list); 1233 spin_unlock(&buffers->rb_lock); 1234 return req; 1235 } 1236 1237 /** 1238 * rpcrdma_buffer_put - Put request/reply buffers back into pool 1239 * @buffers: buffer pool 1240 * @req: object to return 1241 * 1242 */ 1243 void rpcrdma_buffer_put(struct rpcrdma_buffer *buffers, struct rpcrdma_req *req) 1244 { 1245 rpcrdma_reply_put(buffers, req); 1246 1247 spin_lock(&buffers->rb_lock); 1248 list_add(&req->rl_list, &buffers->rb_send_bufs); 1249 spin_unlock(&buffers->rb_lock); 1250 } 1251 1252 /* Returns a pointer to a rpcrdma_regbuf object, or NULL. 1253 * 1254 * xprtrdma uses a regbuf for posting an outgoing RDMA SEND, or for 1255 * receiving the payload of RDMA RECV operations. During Long Calls 1256 * or Replies they may be registered externally via frwr_map. 1257 */ 1258 static struct rpcrdma_regbuf * 1259 rpcrdma_regbuf_alloc(size_t size, enum dma_data_direction direction, 1260 gfp_t flags) 1261 { 1262 struct rpcrdma_regbuf *rb; 1263 1264 rb = kmalloc(sizeof(*rb), flags); 1265 if (!rb) 1266 return NULL; 1267 rb->rg_data = kmalloc(size, flags); 1268 if (!rb->rg_data) { 1269 kfree(rb); 1270 return NULL; 1271 } 1272 1273 rb->rg_device = NULL; 1274 rb->rg_direction = direction; 1275 rb->rg_iov.length = size; 1276 return rb; 1277 } 1278 1279 /** 1280 * rpcrdma_regbuf_realloc - re-allocate a SEND/RECV buffer 1281 * @rb: regbuf to reallocate 1282 * @size: size of buffer to be allocated, in bytes 1283 * @flags: GFP flags 1284 * 1285 * Returns true if reallocation was successful. If false is 1286 * returned, @rb is left untouched. 1287 */ 1288 bool rpcrdma_regbuf_realloc(struct rpcrdma_regbuf *rb, size_t size, gfp_t flags) 1289 { 1290 void *buf; 1291 1292 buf = kmalloc(size, flags); 1293 if (!buf) 1294 return false; 1295 1296 rpcrdma_regbuf_dma_unmap(rb); 1297 kfree(rb->rg_data); 1298 1299 rb->rg_data = buf; 1300 rb->rg_iov.length = size; 1301 return true; 1302 } 1303 1304 /** 1305 * __rpcrdma_regbuf_dma_map - DMA-map a regbuf 1306 * @r_xprt: controlling transport instance 1307 * @rb: regbuf to be mapped 1308 * 1309 * Returns true if the buffer is now DMA mapped to @r_xprt's device 1310 */ 1311 bool __rpcrdma_regbuf_dma_map(struct rpcrdma_xprt *r_xprt, 1312 struct rpcrdma_regbuf *rb) 1313 { 1314 struct ib_device *device = r_xprt->rx_ep->re_id->device; 1315 1316 if (rb->rg_direction == DMA_NONE) 1317 return false; 1318 1319 rb->rg_iov.addr = ib_dma_map_single(device, rdmab_data(rb), 1320 rdmab_length(rb), rb->rg_direction); 1321 if (ib_dma_mapping_error(device, rdmab_addr(rb))) { 1322 trace_xprtrdma_dma_maperr(rdmab_addr(rb)); 1323 return false; 1324 } 1325 1326 rb->rg_device = device; 1327 rb->rg_iov.lkey = r_xprt->rx_ep->re_pd->local_dma_lkey; 1328 return true; 1329 } 1330 1331 static void rpcrdma_regbuf_dma_unmap(struct rpcrdma_regbuf *rb) 1332 { 1333 if (!rb) 1334 return; 1335 1336 if (!rpcrdma_regbuf_is_mapped(rb)) 1337 return; 1338 1339 ib_dma_unmap_single(rb->rg_device, rdmab_addr(rb), rdmab_length(rb), 1340 rb->rg_direction); 1341 rb->rg_device = NULL; 1342 } 1343 1344 static void rpcrdma_regbuf_free(struct rpcrdma_regbuf *rb) 1345 { 1346 rpcrdma_regbuf_dma_unmap(rb); 1347 if (rb) 1348 kfree(rb->rg_data); 1349 kfree(rb); 1350 } 1351 1352 /** 1353 * rpcrdma_post_recvs - Refill the Receive Queue 1354 * @r_xprt: controlling transport instance 1355 * @needed: current credit grant 1356 * @temp: mark Receive buffers to be deleted after one use 1357 * 1358 */ 1359 void rpcrdma_post_recvs(struct rpcrdma_xprt *r_xprt, int needed, bool temp) 1360 { 1361 struct rpcrdma_buffer *buf = &r_xprt->rx_buf; 1362 struct rpcrdma_ep *ep = r_xprt->rx_ep; 1363 struct ib_recv_wr *wr, *bad_wr; 1364 struct rpcrdma_rep *rep; 1365 int count, rc; 1366 1367 rc = 0; 1368 count = 0; 1369 1370 if (likely(ep->re_receive_count > needed)) 1371 goto out; 1372 needed -= ep->re_receive_count; 1373 if (!temp) 1374 needed += RPCRDMA_MAX_RECV_BATCH; 1375 1376 if (atomic_inc_return(&ep->re_receiving) > 1) 1377 goto out; 1378 1379 /* fast path: all needed reps can be found on the free list */ 1380 wr = NULL; 1381 while (needed) { 1382 rep = rpcrdma_rep_get_locked(buf); 1383 if (rep && rep->rr_temp) { 1384 rpcrdma_rep_destroy(rep); 1385 continue; 1386 } 1387 if (!rep) 1388 rep = rpcrdma_rep_create(r_xprt, temp); 1389 if (!rep) 1390 break; 1391 1392 rep->rr_cid.ci_queue_id = ep->re_attr.recv_cq->res.id; 1393 trace_xprtrdma_post_recv(rep); 1394 rep->rr_recv_wr.next = wr; 1395 wr = &rep->rr_recv_wr; 1396 --needed; 1397 ++count; 1398 } 1399 if (!wr) 1400 goto out; 1401 1402 rc = ib_post_recv(ep->re_id->qp, wr, 1403 (const struct ib_recv_wr **)&bad_wr); 1404 if (rc) { 1405 trace_xprtrdma_post_recvs_err(r_xprt, rc); 1406 for (wr = bad_wr; wr;) { 1407 struct rpcrdma_rep *rep; 1408 1409 rep = container_of(wr, struct rpcrdma_rep, rr_recv_wr); 1410 wr = wr->next; 1411 rpcrdma_rep_put(buf, rep); 1412 --count; 1413 } 1414 } 1415 if (atomic_dec_return(&ep->re_receiving) > 0) 1416 complete(&ep->re_done); 1417 1418 out: 1419 trace_xprtrdma_post_recvs(r_xprt, count); 1420 ep->re_receive_count += count; 1421 return; 1422 } 1423