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 * transport.c 44 * 45 * This file contains the top-level implementation of an RPC RDMA 46 * transport. 47 * 48 * Naming convention: functions beginning with xprt_ are part of the 49 * transport switch. All others are RPC RDMA internal. 50 */ 51 52 #include <linux/module.h> 53 #include <linux/slab.h> 54 #include <linux/seq_file.h> 55 #include <linux/smp.h> 56 57 #include <linux/sunrpc/addr.h> 58 #include <linux/sunrpc/svc_rdma.h> 59 60 #include "xprt_rdma.h" 61 #include <trace/events/rpcrdma.h> 62 63 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG) 64 # define RPCDBG_FACILITY RPCDBG_TRANS 65 #endif 66 67 /* 68 * tunables 69 */ 70 71 unsigned int xprt_rdma_slot_table_entries = RPCRDMA_DEF_SLOT_TABLE; 72 unsigned int xprt_rdma_max_inline_read = RPCRDMA_DEF_INLINE; 73 unsigned int xprt_rdma_max_inline_write = RPCRDMA_DEF_INLINE; 74 unsigned int xprt_rdma_memreg_strategy = RPCRDMA_FRWR; 75 int xprt_rdma_pad_optimize; 76 77 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG) 78 79 static unsigned int min_slot_table_size = RPCRDMA_MIN_SLOT_TABLE; 80 static unsigned int max_slot_table_size = RPCRDMA_MAX_SLOT_TABLE; 81 static unsigned int min_inline_size = RPCRDMA_MIN_INLINE; 82 static unsigned int max_inline_size = RPCRDMA_MAX_INLINE; 83 static unsigned int max_padding = PAGE_SIZE; 84 static unsigned int min_memreg = RPCRDMA_BOUNCEBUFFERS; 85 static unsigned int max_memreg = RPCRDMA_LAST - 1; 86 static unsigned int dummy; 87 88 static struct ctl_table_header *sunrpc_table_header; 89 90 static struct ctl_table xr_tunables_table[] = { 91 { 92 .procname = "rdma_slot_table_entries", 93 .data = &xprt_rdma_slot_table_entries, 94 .maxlen = sizeof(unsigned int), 95 .mode = 0644, 96 .proc_handler = proc_dointvec_minmax, 97 .extra1 = &min_slot_table_size, 98 .extra2 = &max_slot_table_size 99 }, 100 { 101 .procname = "rdma_max_inline_read", 102 .data = &xprt_rdma_max_inline_read, 103 .maxlen = sizeof(unsigned int), 104 .mode = 0644, 105 .proc_handler = proc_dointvec_minmax, 106 .extra1 = &min_inline_size, 107 .extra2 = &max_inline_size, 108 }, 109 { 110 .procname = "rdma_max_inline_write", 111 .data = &xprt_rdma_max_inline_write, 112 .maxlen = sizeof(unsigned int), 113 .mode = 0644, 114 .proc_handler = proc_dointvec_minmax, 115 .extra1 = &min_inline_size, 116 .extra2 = &max_inline_size, 117 }, 118 { 119 .procname = "rdma_inline_write_padding", 120 .data = &dummy, 121 .maxlen = sizeof(unsigned int), 122 .mode = 0644, 123 .proc_handler = proc_dointvec_minmax, 124 .extra1 = SYSCTL_ZERO, 125 .extra2 = &max_padding, 126 }, 127 { 128 .procname = "rdma_memreg_strategy", 129 .data = &xprt_rdma_memreg_strategy, 130 .maxlen = sizeof(unsigned int), 131 .mode = 0644, 132 .proc_handler = proc_dointvec_minmax, 133 .extra1 = &min_memreg, 134 .extra2 = &max_memreg, 135 }, 136 { 137 .procname = "rdma_pad_optimize", 138 .data = &xprt_rdma_pad_optimize, 139 .maxlen = sizeof(unsigned int), 140 .mode = 0644, 141 .proc_handler = proc_dointvec, 142 }, 143 { }, 144 }; 145 146 static struct ctl_table sunrpc_table[] = { 147 { 148 .procname = "sunrpc", 149 .mode = 0555, 150 .child = xr_tunables_table 151 }, 152 { }, 153 }; 154 155 #endif 156 157 static const struct rpc_xprt_ops xprt_rdma_procs; 158 159 static void 160 xprt_rdma_format_addresses4(struct rpc_xprt *xprt, struct sockaddr *sap) 161 { 162 struct sockaddr_in *sin = (struct sockaddr_in *)sap; 163 char buf[20]; 164 165 snprintf(buf, sizeof(buf), "%08x", ntohl(sin->sin_addr.s_addr)); 166 xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = kstrdup(buf, GFP_KERNEL); 167 168 xprt->address_strings[RPC_DISPLAY_NETID] = RPCBIND_NETID_RDMA; 169 } 170 171 static void 172 xprt_rdma_format_addresses6(struct rpc_xprt *xprt, struct sockaddr *sap) 173 { 174 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sap; 175 char buf[40]; 176 177 snprintf(buf, sizeof(buf), "%pi6", &sin6->sin6_addr); 178 xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = kstrdup(buf, GFP_KERNEL); 179 180 xprt->address_strings[RPC_DISPLAY_NETID] = RPCBIND_NETID_RDMA6; 181 } 182 183 void 184 xprt_rdma_format_addresses(struct rpc_xprt *xprt, struct sockaddr *sap) 185 { 186 char buf[128]; 187 188 switch (sap->sa_family) { 189 case AF_INET: 190 xprt_rdma_format_addresses4(xprt, sap); 191 break; 192 case AF_INET6: 193 xprt_rdma_format_addresses6(xprt, sap); 194 break; 195 default: 196 pr_err("rpcrdma: Unrecognized address family\n"); 197 return; 198 } 199 200 (void)rpc_ntop(sap, buf, sizeof(buf)); 201 xprt->address_strings[RPC_DISPLAY_ADDR] = kstrdup(buf, GFP_KERNEL); 202 203 snprintf(buf, sizeof(buf), "%u", rpc_get_port(sap)); 204 xprt->address_strings[RPC_DISPLAY_PORT] = kstrdup(buf, GFP_KERNEL); 205 206 snprintf(buf, sizeof(buf), "%4hx", rpc_get_port(sap)); 207 xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL); 208 209 xprt->address_strings[RPC_DISPLAY_PROTO] = "rdma"; 210 } 211 212 void 213 xprt_rdma_free_addresses(struct rpc_xprt *xprt) 214 { 215 unsigned int i; 216 217 for (i = 0; i < RPC_DISPLAY_MAX; i++) 218 switch (i) { 219 case RPC_DISPLAY_PROTO: 220 case RPC_DISPLAY_NETID: 221 continue; 222 default: 223 kfree(xprt->address_strings[i]); 224 } 225 } 226 227 /** 228 * xprt_rdma_connect_worker - establish connection in the background 229 * @work: worker thread context 230 * 231 * Requester holds the xprt's send lock to prevent activity on this 232 * transport while a fresh connection is being established. RPC tasks 233 * sleep on the xprt's pending queue waiting for connect to complete. 234 */ 235 static void 236 xprt_rdma_connect_worker(struct work_struct *work) 237 { 238 struct rpcrdma_xprt *r_xprt = container_of(work, struct rpcrdma_xprt, 239 rx_connect_worker.work); 240 struct rpc_xprt *xprt = &r_xprt->rx_xprt; 241 int rc; 242 243 rc = rpcrdma_ep_connect(&r_xprt->rx_ep, &r_xprt->rx_ia); 244 xprt_clear_connecting(xprt); 245 if (r_xprt->rx_ep.rep_connected > 0) { 246 if (!xprt_test_and_set_connected(xprt)) { 247 xprt->stat.connect_count++; 248 xprt->stat.connect_time += (long)jiffies - 249 xprt->stat.connect_start; 250 xprt_wake_pending_tasks(xprt, -EAGAIN); 251 } 252 } else { 253 if (xprt_test_and_clear_connected(xprt)) 254 xprt_wake_pending_tasks(xprt, rc); 255 } 256 } 257 258 /** 259 * xprt_rdma_inject_disconnect - inject a connection fault 260 * @xprt: transport context 261 * 262 * If @xprt is connected, disconnect it to simulate spurious connection 263 * loss. 264 */ 265 static void 266 xprt_rdma_inject_disconnect(struct rpc_xprt *xprt) 267 { 268 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); 269 270 trace_xprtrdma_op_inject_dsc(r_xprt); 271 rdma_disconnect(r_xprt->rx_ia.ri_id); 272 } 273 274 /** 275 * xprt_rdma_destroy - Full tear down of transport 276 * @xprt: doomed transport context 277 * 278 * Caller guarantees there will be no more calls to us with 279 * this @xprt. 280 */ 281 static void 282 xprt_rdma_destroy(struct rpc_xprt *xprt) 283 { 284 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); 285 286 trace_xprtrdma_op_destroy(r_xprt); 287 288 cancel_delayed_work_sync(&r_xprt->rx_connect_worker); 289 290 rpcrdma_ep_destroy(r_xprt); 291 rpcrdma_buffer_destroy(&r_xprt->rx_buf); 292 rpcrdma_ia_close(&r_xprt->rx_ia); 293 294 xprt_rdma_free_addresses(xprt); 295 xprt_free(xprt); 296 297 module_put(THIS_MODULE); 298 } 299 300 /* 60 second timeout, no retries */ 301 static const struct rpc_timeout xprt_rdma_default_timeout = { 302 .to_initval = 60 * HZ, 303 .to_maxval = 60 * HZ, 304 }; 305 306 /** 307 * xprt_setup_rdma - Set up transport to use RDMA 308 * 309 * @args: rpc transport arguments 310 */ 311 static struct rpc_xprt * 312 xprt_setup_rdma(struct xprt_create *args) 313 { 314 struct rpc_xprt *xprt; 315 struct rpcrdma_xprt *new_xprt; 316 struct sockaddr *sap; 317 int rc; 318 319 if (args->addrlen > sizeof(xprt->addr)) 320 return ERR_PTR(-EBADF); 321 322 xprt = xprt_alloc(args->net, sizeof(struct rpcrdma_xprt), 0, 0); 323 if (!xprt) 324 return ERR_PTR(-ENOMEM); 325 326 xprt->timeout = &xprt_rdma_default_timeout; 327 xprt->connect_timeout = xprt->timeout->to_initval; 328 xprt->max_reconnect_timeout = xprt->timeout->to_maxval; 329 xprt->bind_timeout = RPCRDMA_BIND_TO; 330 xprt->reestablish_timeout = RPCRDMA_INIT_REEST_TO; 331 xprt->idle_timeout = RPCRDMA_IDLE_DISC_TO; 332 333 xprt->resvport = 0; /* privileged port not needed */ 334 xprt->ops = &xprt_rdma_procs; 335 336 /* 337 * Set up RDMA-specific connect data. 338 */ 339 sap = args->dstaddr; 340 341 /* Ensure xprt->addr holds valid server TCP (not RDMA) 342 * address, for any side protocols which peek at it */ 343 xprt->prot = IPPROTO_TCP; 344 xprt->addrlen = args->addrlen; 345 memcpy(&xprt->addr, sap, xprt->addrlen); 346 347 if (rpc_get_port(sap)) 348 xprt_set_bound(xprt); 349 xprt_rdma_format_addresses(xprt, sap); 350 351 new_xprt = rpcx_to_rdmax(xprt); 352 rc = rpcrdma_ia_open(new_xprt); 353 if (rc) 354 goto out1; 355 356 rc = rpcrdma_ep_create(new_xprt); 357 if (rc) 358 goto out2; 359 360 rc = rpcrdma_buffer_create(new_xprt); 361 if (rc) 362 goto out3; 363 364 INIT_DELAYED_WORK(&new_xprt->rx_connect_worker, 365 xprt_rdma_connect_worker); 366 367 xprt->max_payload = frwr_maxpages(new_xprt); 368 if (xprt->max_payload == 0) 369 goto out4; 370 xprt->max_payload <<= PAGE_SHIFT; 371 dprintk("RPC: %s: transport data payload maximum: %zu bytes\n", 372 __func__, xprt->max_payload); 373 374 if (!try_module_get(THIS_MODULE)) 375 goto out4; 376 377 dprintk("RPC: %s: %s:%s\n", __func__, 378 xprt->address_strings[RPC_DISPLAY_ADDR], 379 xprt->address_strings[RPC_DISPLAY_PORT]); 380 trace_xprtrdma_create(new_xprt); 381 return xprt; 382 383 out4: 384 rpcrdma_buffer_destroy(&new_xprt->rx_buf); 385 rc = -ENODEV; 386 out3: 387 rpcrdma_ep_destroy(new_xprt); 388 out2: 389 rpcrdma_ia_close(&new_xprt->rx_ia); 390 out1: 391 trace_xprtrdma_op_destroy(new_xprt); 392 xprt_rdma_free_addresses(xprt); 393 xprt_free(xprt); 394 return ERR_PTR(rc); 395 } 396 397 /** 398 * xprt_rdma_close - close a transport connection 399 * @xprt: transport context 400 * 401 * Called during autoclose or device removal. 402 * 403 * Caller holds @xprt's send lock to prevent activity on this 404 * transport while the connection is torn down. 405 */ 406 void xprt_rdma_close(struct rpc_xprt *xprt) 407 { 408 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); 409 struct rpcrdma_ep *ep = &r_xprt->rx_ep; 410 struct rpcrdma_ia *ia = &r_xprt->rx_ia; 411 412 might_sleep(); 413 414 trace_xprtrdma_op_close(r_xprt); 415 416 /* Prevent marshaling and sending of new requests */ 417 xprt_clear_connected(xprt); 418 419 if (test_and_clear_bit(RPCRDMA_IAF_REMOVING, &ia->ri_flags)) { 420 rpcrdma_ia_remove(ia); 421 goto out; 422 } 423 424 if (ep->rep_connected == -ENODEV) 425 return; 426 if (ep->rep_connected > 0) 427 xprt->reestablish_timeout = 0; 428 rpcrdma_ep_disconnect(ep, ia); 429 430 /* Prepare @xprt for the next connection by reinitializing 431 * its credit grant to one (see RFC 8166, Section 3.3.3). 432 */ 433 r_xprt->rx_buf.rb_credits = 1; 434 xprt->cwnd = RPC_CWNDSHIFT; 435 436 out: 437 ++xprt->connect_cookie; 438 xprt_disconnect_done(xprt); 439 } 440 441 /** 442 * xprt_rdma_set_port - update server port with rpcbind result 443 * @xprt: controlling RPC transport 444 * @port: new port value 445 * 446 * Transport connect status is unchanged. 447 */ 448 static void 449 xprt_rdma_set_port(struct rpc_xprt *xprt, u16 port) 450 { 451 struct sockaddr *sap = (struct sockaddr *)&xprt->addr; 452 char buf[8]; 453 454 dprintk("RPC: %s: setting port for xprt %p (%s:%s) to %u\n", 455 __func__, xprt, 456 xprt->address_strings[RPC_DISPLAY_ADDR], 457 xprt->address_strings[RPC_DISPLAY_PORT], 458 port); 459 460 rpc_set_port(sap, port); 461 462 kfree(xprt->address_strings[RPC_DISPLAY_PORT]); 463 snprintf(buf, sizeof(buf), "%u", port); 464 xprt->address_strings[RPC_DISPLAY_PORT] = kstrdup(buf, GFP_KERNEL); 465 466 kfree(xprt->address_strings[RPC_DISPLAY_HEX_PORT]); 467 snprintf(buf, sizeof(buf), "%4hx", port); 468 xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL); 469 } 470 471 /** 472 * xprt_rdma_timer - invoked when an RPC times out 473 * @xprt: controlling RPC transport 474 * @task: RPC task that timed out 475 * 476 * Invoked when the transport is still connected, but an RPC 477 * retransmit timeout occurs. 478 * 479 * Since RDMA connections don't have a keep-alive, forcibly 480 * disconnect and retry to connect. This drives full 481 * detection of the network path, and retransmissions of 482 * all pending RPCs. 483 */ 484 static void 485 xprt_rdma_timer(struct rpc_xprt *xprt, struct rpc_task *task) 486 { 487 xprt_force_disconnect(xprt); 488 } 489 490 /** 491 * xprt_rdma_set_connect_timeout - set timeouts for establishing a connection 492 * @xprt: controlling transport instance 493 * @connect_timeout: reconnect timeout after client disconnects 494 * @reconnect_timeout: reconnect timeout after server disconnects 495 * 496 */ 497 static void xprt_rdma_tcp_set_connect_timeout(struct rpc_xprt *xprt, 498 unsigned long connect_timeout, 499 unsigned long reconnect_timeout) 500 { 501 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); 502 503 trace_xprtrdma_op_set_cto(r_xprt, connect_timeout, reconnect_timeout); 504 505 spin_lock(&xprt->transport_lock); 506 507 if (connect_timeout < xprt->connect_timeout) { 508 struct rpc_timeout to; 509 unsigned long initval; 510 511 to = *xprt->timeout; 512 initval = connect_timeout; 513 if (initval < RPCRDMA_INIT_REEST_TO << 1) 514 initval = RPCRDMA_INIT_REEST_TO << 1; 515 to.to_initval = initval; 516 to.to_maxval = initval; 517 r_xprt->rx_timeout = to; 518 xprt->timeout = &r_xprt->rx_timeout; 519 xprt->connect_timeout = connect_timeout; 520 } 521 522 if (reconnect_timeout < xprt->max_reconnect_timeout) 523 xprt->max_reconnect_timeout = reconnect_timeout; 524 525 spin_unlock(&xprt->transport_lock); 526 } 527 528 /** 529 * xprt_rdma_connect - schedule an attempt to reconnect 530 * @xprt: transport state 531 * @task: RPC scheduler context (unused) 532 * 533 */ 534 static void 535 xprt_rdma_connect(struct rpc_xprt *xprt, struct rpc_task *task) 536 { 537 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); 538 unsigned long delay; 539 540 trace_xprtrdma_op_connect(r_xprt); 541 542 delay = 0; 543 if (r_xprt->rx_ep.rep_connected != 0) { 544 delay = xprt_reconnect_delay(xprt); 545 xprt_reconnect_backoff(xprt, RPCRDMA_INIT_REEST_TO); 546 } 547 queue_delayed_work(xprtiod_workqueue, &r_xprt->rx_connect_worker, 548 delay); 549 } 550 551 /** 552 * xprt_rdma_alloc_slot - allocate an rpc_rqst 553 * @xprt: controlling RPC transport 554 * @task: RPC task requesting a fresh rpc_rqst 555 * 556 * tk_status values: 557 * %0 if task->tk_rqstp points to a fresh rpc_rqst 558 * %-EAGAIN if no rpc_rqst is available; queued on backlog 559 */ 560 static void 561 xprt_rdma_alloc_slot(struct rpc_xprt *xprt, struct rpc_task *task) 562 { 563 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); 564 struct rpcrdma_req *req; 565 566 req = rpcrdma_buffer_get(&r_xprt->rx_buf); 567 if (!req) 568 goto out_sleep; 569 task->tk_rqstp = &req->rl_slot; 570 task->tk_status = 0; 571 return; 572 573 out_sleep: 574 rpc_sleep_on(&xprt->backlog, task, NULL); 575 task->tk_status = -EAGAIN; 576 } 577 578 /** 579 * xprt_rdma_free_slot - release an rpc_rqst 580 * @xprt: controlling RPC transport 581 * @rqst: rpc_rqst to release 582 * 583 */ 584 static void 585 xprt_rdma_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *rqst) 586 { 587 struct rpcrdma_xprt *r_xprt = 588 container_of(xprt, struct rpcrdma_xprt, rx_xprt); 589 590 memset(rqst, 0, sizeof(*rqst)); 591 rpcrdma_buffer_put(&r_xprt->rx_buf, rpcr_to_rdmar(rqst)); 592 rpc_wake_up_next(&xprt->backlog); 593 } 594 595 static bool rpcrdma_check_regbuf(struct rpcrdma_xprt *r_xprt, 596 struct rpcrdma_regbuf *rb, size_t size, 597 gfp_t flags) 598 { 599 if (unlikely(rdmab_length(rb) < size)) { 600 if (!rpcrdma_regbuf_realloc(rb, size, flags)) 601 return false; 602 r_xprt->rx_stats.hardway_register_count += size; 603 } 604 return true; 605 } 606 607 /** 608 * xprt_rdma_allocate - allocate transport resources for an RPC 609 * @task: RPC task 610 * 611 * Return values: 612 * 0: Success; rq_buffer points to RPC buffer to use 613 * ENOMEM: Out of memory, call again later 614 * EIO: A permanent error occurred, do not retry 615 */ 616 static int 617 xprt_rdma_allocate(struct rpc_task *task) 618 { 619 struct rpc_rqst *rqst = task->tk_rqstp; 620 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(rqst->rq_xprt); 621 struct rpcrdma_req *req = rpcr_to_rdmar(rqst); 622 gfp_t flags; 623 624 flags = RPCRDMA_DEF_GFP; 625 if (RPC_IS_SWAPPER(task)) 626 flags = __GFP_MEMALLOC | GFP_NOWAIT | __GFP_NOWARN; 627 628 if (!rpcrdma_check_regbuf(r_xprt, req->rl_sendbuf, rqst->rq_callsize, 629 flags)) 630 goto out_fail; 631 if (!rpcrdma_check_regbuf(r_xprt, req->rl_recvbuf, rqst->rq_rcvsize, 632 flags)) 633 goto out_fail; 634 635 rqst->rq_buffer = rdmab_data(req->rl_sendbuf); 636 rqst->rq_rbuffer = rdmab_data(req->rl_recvbuf); 637 trace_xprtrdma_op_allocate(task, req); 638 return 0; 639 640 out_fail: 641 trace_xprtrdma_op_allocate(task, NULL); 642 return -ENOMEM; 643 } 644 645 /** 646 * xprt_rdma_free - release resources allocated by xprt_rdma_allocate 647 * @task: RPC task 648 * 649 * Caller guarantees rqst->rq_buffer is non-NULL. 650 */ 651 static void 652 xprt_rdma_free(struct rpc_task *task) 653 { 654 struct rpc_rqst *rqst = task->tk_rqstp; 655 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(rqst->rq_xprt); 656 struct rpcrdma_req *req = rpcr_to_rdmar(rqst); 657 658 trace_xprtrdma_op_free(task, req); 659 660 if (!list_empty(&req->rl_registered)) 661 frwr_unmap_sync(r_xprt, req); 662 663 /* XXX: If the RPC is completing because of a signal and 664 * not because a reply was received, we ought to ensure 665 * that the Send completion has fired, so that memory 666 * involved with the Send is not still visible to the NIC. 667 */ 668 } 669 670 /** 671 * xprt_rdma_send_request - marshal and send an RPC request 672 * @rqst: RPC message in rq_snd_buf 673 * 674 * Caller holds the transport's write lock. 675 * 676 * Returns: 677 * %0 if the RPC message has been sent 678 * %-ENOTCONN if the caller should reconnect and call again 679 * %-EAGAIN if the caller should call again 680 * %-ENOBUFS if the caller should call again after a delay 681 * %-EMSGSIZE if encoding ran out of buffer space. The request 682 * was not sent. Do not try to send this message again. 683 * %-EIO if an I/O error occurred. The request was not sent. 684 * Do not try to send this message again. 685 */ 686 static int 687 xprt_rdma_send_request(struct rpc_rqst *rqst) 688 { 689 struct rpc_xprt *xprt = rqst->rq_xprt; 690 struct rpcrdma_req *req = rpcr_to_rdmar(rqst); 691 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); 692 int rc = 0; 693 694 #if defined(CONFIG_SUNRPC_BACKCHANNEL) 695 if (unlikely(!rqst->rq_buffer)) 696 return xprt_rdma_bc_send_reply(rqst); 697 #endif /* CONFIG_SUNRPC_BACKCHANNEL */ 698 699 if (!xprt_connected(xprt)) 700 return -ENOTCONN; 701 702 if (!xprt_request_get_cong(xprt, rqst)) 703 return -EBADSLT; 704 705 rc = rpcrdma_marshal_req(r_xprt, rqst); 706 if (rc < 0) 707 goto failed_marshal; 708 709 /* Must suppress retransmit to maintain credits */ 710 if (rqst->rq_connect_cookie == xprt->connect_cookie) 711 goto drop_connection; 712 rqst->rq_xtime = ktime_get(); 713 714 if (rpcrdma_ep_post(&r_xprt->rx_ia, &r_xprt->rx_ep, req)) 715 goto drop_connection; 716 717 rqst->rq_xmit_bytes_sent += rqst->rq_snd_buf.len; 718 719 /* An RPC with no reply will throw off credit accounting, 720 * so drop the connection to reset the credit grant. 721 */ 722 if (!rpc_reply_expected(rqst->rq_task)) 723 goto drop_connection; 724 return 0; 725 726 failed_marshal: 727 if (rc != -ENOTCONN) 728 return rc; 729 drop_connection: 730 xprt_rdma_close(xprt); 731 return -ENOTCONN; 732 } 733 734 void xprt_rdma_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) 735 { 736 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); 737 long idle_time = 0; 738 739 if (xprt_connected(xprt)) 740 idle_time = (long)(jiffies - xprt->last_used) / HZ; 741 742 seq_puts(seq, "\txprt:\trdma "); 743 seq_printf(seq, "%u %lu %lu %lu %ld %lu %lu %lu %llu %llu ", 744 0, /* need a local port? */ 745 xprt->stat.bind_count, 746 xprt->stat.connect_count, 747 xprt->stat.connect_time / HZ, 748 idle_time, 749 xprt->stat.sends, 750 xprt->stat.recvs, 751 xprt->stat.bad_xids, 752 xprt->stat.req_u, 753 xprt->stat.bklog_u); 754 seq_printf(seq, "%lu %lu %lu %llu %llu %llu %llu %lu %lu %lu %lu ", 755 r_xprt->rx_stats.read_chunk_count, 756 r_xprt->rx_stats.write_chunk_count, 757 r_xprt->rx_stats.reply_chunk_count, 758 r_xprt->rx_stats.total_rdma_request, 759 r_xprt->rx_stats.total_rdma_reply, 760 r_xprt->rx_stats.pullup_copy_count, 761 r_xprt->rx_stats.fixup_copy_count, 762 r_xprt->rx_stats.hardway_register_count, 763 r_xprt->rx_stats.failed_marshal_count, 764 r_xprt->rx_stats.bad_reply_count, 765 r_xprt->rx_stats.nomsg_call_count); 766 seq_printf(seq, "%lu %lu %lu %lu %lu %lu\n", 767 r_xprt->rx_stats.mrs_recycled, 768 r_xprt->rx_stats.mrs_orphaned, 769 r_xprt->rx_stats.mrs_allocated, 770 r_xprt->rx_stats.local_inv_needed, 771 r_xprt->rx_stats.empty_sendctx_q, 772 r_xprt->rx_stats.reply_waits_for_send); 773 } 774 775 static int 776 xprt_rdma_enable_swap(struct rpc_xprt *xprt) 777 { 778 return 0; 779 } 780 781 static void 782 xprt_rdma_disable_swap(struct rpc_xprt *xprt) 783 { 784 } 785 786 /* 787 * Plumbing for rpc transport switch and kernel module 788 */ 789 790 static const struct rpc_xprt_ops xprt_rdma_procs = { 791 .reserve_xprt = xprt_reserve_xprt_cong, 792 .release_xprt = xprt_release_xprt_cong, /* sunrpc/xprt.c */ 793 .alloc_slot = xprt_rdma_alloc_slot, 794 .free_slot = xprt_rdma_free_slot, 795 .release_request = xprt_release_rqst_cong, /* ditto */ 796 .wait_for_reply_request = xprt_wait_for_reply_request_def, /* ditto */ 797 .timer = xprt_rdma_timer, 798 .rpcbind = rpcb_getport_async, /* sunrpc/rpcb_clnt.c */ 799 .set_port = xprt_rdma_set_port, 800 .connect = xprt_rdma_connect, 801 .buf_alloc = xprt_rdma_allocate, 802 .buf_free = xprt_rdma_free, 803 .send_request = xprt_rdma_send_request, 804 .close = xprt_rdma_close, 805 .destroy = xprt_rdma_destroy, 806 .set_connect_timeout = xprt_rdma_tcp_set_connect_timeout, 807 .print_stats = xprt_rdma_print_stats, 808 .enable_swap = xprt_rdma_enable_swap, 809 .disable_swap = xprt_rdma_disable_swap, 810 .inject_disconnect = xprt_rdma_inject_disconnect, 811 #if defined(CONFIG_SUNRPC_BACKCHANNEL) 812 .bc_setup = xprt_rdma_bc_setup, 813 .bc_maxpayload = xprt_rdma_bc_maxpayload, 814 .bc_num_slots = xprt_rdma_bc_max_slots, 815 .bc_free_rqst = xprt_rdma_bc_free_rqst, 816 .bc_destroy = xprt_rdma_bc_destroy, 817 #endif 818 }; 819 820 static struct xprt_class xprt_rdma = { 821 .list = LIST_HEAD_INIT(xprt_rdma.list), 822 .name = "rdma", 823 .owner = THIS_MODULE, 824 .ident = XPRT_TRANSPORT_RDMA, 825 .setup = xprt_setup_rdma, 826 }; 827 828 void xprt_rdma_cleanup(void) 829 { 830 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG) 831 if (sunrpc_table_header) { 832 unregister_sysctl_table(sunrpc_table_header); 833 sunrpc_table_header = NULL; 834 } 835 #endif 836 837 xprt_unregister_transport(&xprt_rdma); 838 xprt_unregister_transport(&xprt_rdma_bc); 839 } 840 841 int xprt_rdma_init(void) 842 { 843 int rc; 844 845 rc = xprt_register_transport(&xprt_rdma); 846 if (rc) 847 return rc; 848 849 rc = xprt_register_transport(&xprt_rdma_bc); 850 if (rc) { 851 xprt_unregister_transport(&xprt_rdma); 852 return rc; 853 } 854 855 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG) 856 if (!sunrpc_table_header) 857 sunrpc_table_header = register_sysctl_table(sunrpc_table); 858 #endif 859 return 0; 860 } 861