git.oblomov.eu Git - linux-2.6/blob - net/sunrpc/xprtrdma/transport.c

   1 /*
   2  * Copyright (c) 2003-2007 Network Appliance, Inc. All rights reserved.
   3  *
   4  * This software is available to you under a choice of one of two
   5  * licenses.  You may choose to be licensed under the terms of the GNU
   6  * General Public License (GPL) Version 2, available from the file
   7  * COPYING in the main directory of this source tree, or the BSD-type
   8  * license below:
   9  *
  10  * Redistribution and use in source and binary forms, with or without
  11  * modification, are permitted provided that the following conditions
  12  * are met:
  13  *
  14  *      Redistributions of source code must retain the above copyright
  15  *      notice, this list of conditions and the following disclaimer.
  16  *
  17  *      Redistributions in binary form must reproduce the above
  18  *      copyright notice, this list of conditions and the following
  19  *      disclaimer in the documentation and/or other materials provided
  20  *      with the distribution.
  21  *
  22  *      Neither the name of the Network Appliance, Inc. nor the names of
  23  *      its contributors may be used to endorse or promote products
  24  *      derived from this software without specific prior written
  25  *      permission.
  26  *
  27  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  28  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  29  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
  30  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
  31  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
  32  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
  33  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  34  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  35  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  36  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  37  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  38  */
  39
  40 /*
  41  * transport.c
  42  *
  43  * This file contains the top-level implementation of an RPC RDMA
  44  * transport.
  45  *
  46  * Naming convention: functions beginning with xprt_ are part of the
  47  * transport switch. All others are RPC RDMA internal.
  48  */
  49
  50 #include <linux/module.h>
  51 #include <linux/init.h>
  52 #include <linux/seq_file.h>
  53
  54 #include "xprt_rdma.h"
  55
  56 #ifdef RPC_DEBUG
  57 # define RPCDBG_FACILITY        RPCDBG_TRANS
  58 #endif
  59
  60 MODULE_LICENSE("Dual BSD/GPL");
  61
  62 MODULE_DESCRIPTION("RPC/RDMA Transport for Linux kernel NFS");
  63 MODULE_AUTHOR("Network Appliance, Inc.");
  64
  65 /*
  66  * tunables
  67  */
  68
  69 static unsigned int xprt_rdma_slot_table_entries = RPCRDMA_DEF_SLOT_TABLE;
  70 static unsigned int xprt_rdma_max_inline_read = RPCRDMA_DEF_INLINE;
  71 static unsigned int xprt_rdma_max_inline_write = RPCRDMA_DEF_INLINE;
  72 static unsigned int xprt_rdma_inline_write_padding;
  73 static unsigned int xprt_rdma_memreg_strategy = RPCRDMA_FRMR;
  74
  75 #ifdef RPC_DEBUG
  76
  77 static unsigned int min_slot_table_size = RPCRDMA_MIN_SLOT_TABLE;
  78 static unsigned int max_slot_table_size = RPCRDMA_MAX_SLOT_TABLE;
  79 static unsigned int zero;
  80 static unsigned int max_padding = PAGE_SIZE;
  81 static unsigned int min_memreg = RPCRDMA_BOUNCEBUFFERS;
  82 static unsigned int max_memreg = RPCRDMA_LAST - 1;
  83
  84 static struct ctl_table_header *sunrpc_table_header;
  85
  86 static ctl_table xr_tunables_table[] = {
  87         {
  88                 .ctl_name       = CTL_UNNUMBERED,
  89                 .procname       = "rdma_slot_table_entries",
  90                 .data           = &xprt_rdma_slot_table_entries,
  91                 .maxlen         = sizeof(unsigned int),
  92                 .mode           = 0644,
  93                 .proc_handler   = &proc_dointvec_minmax,
  94                 .strategy       = &sysctl_intvec,
  95                 .extra1         = &min_slot_table_size,
  96                 .extra2         = &max_slot_table_size
  97         },
  98         {
  99                 .ctl_name       = CTL_UNNUMBERED,
 100                 .procname       = "rdma_max_inline_read",
 101                 .data           = &xprt_rdma_max_inline_read,
 102                 .maxlen         = sizeof(unsigned int),
 103                 .mode           = 0644,
 104                 .proc_handler   = &proc_dointvec,
 105                 .strategy       = &sysctl_intvec,
 106         },
 107         {
 108                 .ctl_name       = CTL_UNNUMBERED,
 109                 .procname       = "rdma_max_inline_write",
 110                 .data           = &xprt_rdma_max_inline_write,
 111                 .maxlen         = sizeof(unsigned int),
 112                 .mode           = 0644,
 113                 .proc_handler   = &proc_dointvec,
 114                 .strategy       = &sysctl_intvec,
 115         },
 116         {
 117                 .ctl_name       = CTL_UNNUMBERED,
 118                 .procname       = "rdma_inline_write_padding",
 119                 .data           = &xprt_rdma_inline_write_padding,
 120                 .maxlen         = sizeof(unsigned int),
 121                 .mode           = 0644,
 122                 .proc_handler   = &proc_dointvec_minmax,
 123                 .strategy       = &sysctl_intvec,
 124                 .extra1         = &zero,
 125                 .extra2         = &max_padding,
 126         },
 127         {
 128                 .ctl_name       = CTL_UNNUMBERED,
 129                 .procname       = "rdma_memreg_strategy",
 130                 .data           = &xprt_rdma_memreg_strategy,
 131                 .maxlen         = sizeof(unsigned int),
 132                 .mode           = 0644,
 133                 .proc_handler   = &proc_dointvec_minmax,
 134                 .strategy       = &sysctl_intvec,
 135                 .extra1         = &min_memreg,
 136                 .extra2         = &max_memreg,
 137         },
 138         {
 139                 .ctl_name = 0,
 140         },
 141 };
 142
 143 static ctl_table sunrpc_table[] = {
 144         {
 145                 .ctl_name       = CTL_SUNRPC,
 146                 .procname       = "sunrpc",
 147                 .mode           = 0555,
 148                 .child          = xr_tunables_table
 149         },
 150         {
 151                 .ctl_name = 0,
 152         },
 153 };
 154
 155 #endif
 156
 157 static struct rpc_xprt_ops xprt_rdma_procs;     /* forward reference */
 158
 159 static void
 160 xprt_rdma_format_addresses(struct rpc_xprt *xprt)
 161 {
 162         struct sockaddr_in *addr = (struct sockaddr_in *)
 163                                         &rpcx_to_rdmad(xprt).addr;
 164         char *buf;
 165
 166         buf = kzalloc(20, GFP_KERNEL);
 167         if (buf)
 168                 snprintf(buf, 20, NIPQUAD_FMT, NIPQUAD(addr->sin_addr.s_addr));
 169         xprt->address_strings[RPC_DISPLAY_ADDR] = buf;
 170
 171         buf = kzalloc(8, GFP_KERNEL);
 172         if (buf)
 173                 snprintf(buf, 8, "%u", ntohs(addr->sin_port));
 174         xprt->address_strings[RPC_DISPLAY_PORT] = buf;
 175
 176         xprt->address_strings[RPC_DISPLAY_PROTO] = "rdma";
 177
 178         buf = kzalloc(48, GFP_KERNEL);
 179         if (buf)
 180                 snprintf(buf, 48, "addr="NIPQUAD_FMT" port=%u proto=%s",
 181                         NIPQUAD(addr->sin_addr.s_addr),
 182                         ntohs(addr->sin_port), "rdma");
 183         xprt->address_strings[RPC_DISPLAY_ALL] = buf;
 184
 185         buf = kzalloc(10, GFP_KERNEL);
 186         if (buf)
 187                 snprintf(buf, 10, "%02x%02x%02x%02x",
 188                         NIPQUAD(addr->sin_addr.s_addr));
 189         xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = buf;
 190
 191         buf = kzalloc(8, GFP_KERNEL);
 192         if (buf)
 193                 snprintf(buf, 8, "%4hx", ntohs(addr->sin_port));
 194         xprt->address_strings[RPC_DISPLAY_HEX_PORT] = buf;
 195
 196         buf = kzalloc(30, GFP_KERNEL);
 197         if (buf)
 198                 snprintf(buf, 30, NIPQUAD_FMT".%u.%u",
 199                         NIPQUAD(addr->sin_addr.s_addr),
 200                         ntohs(addr->sin_port) >> 8,
 201                         ntohs(addr->sin_port) & 0xff);
 202         xprt->address_strings[RPC_DISPLAY_UNIVERSAL_ADDR] = buf;
 203
 204         /* netid */
 205         xprt->address_strings[RPC_DISPLAY_NETID] = "rdma";
 206 }
 207
 208 static void
 209 xprt_rdma_free_addresses(struct rpc_xprt *xprt)
 210 {
 211         unsigned int i;
 212
 213         for (i = 0; i < RPC_DISPLAY_MAX; i++)
 214                 switch (i) {
 215                 case RPC_DISPLAY_PROTO:
 216                 case RPC_DISPLAY_NETID:
 217                         continue;
 218                 default:
 219                         kfree(xprt->address_strings[i]);
 220                 }
 221 }
 222
 223 static void
 224 xprt_rdma_connect_worker(struct work_struct *work)
 225 {
 226         struct rpcrdma_xprt *r_xprt =
 227                 container_of(work, struct rpcrdma_xprt, rdma_connect.work);
 228         struct rpc_xprt *xprt = &r_xprt->xprt;
 229         int rc = 0;
 230
 231         if (!xprt->shutdown) {
 232                 xprt_clear_connected(xprt);
 233
 234                 dprintk("RPC:       %s: %sconnect\n", __func__,
 235                                 r_xprt->rx_ep.rep_connected != 0 ? "re" : "");
 236                 rc = rpcrdma_ep_connect(&r_xprt->rx_ep, &r_xprt->rx_ia);
 237                 if (rc)
 238                         goto out;
 239         }
 240         goto out_clear;
 241
 242 out:
 243         xprt_wake_pending_tasks(xprt, rc);
 244
 245 out_clear:
 246         dprintk("RPC:       %s: exit\n", __func__);
 247         xprt_clear_connecting(xprt);
 248 }
 249
 250 /*
 251  * xprt_rdma_destroy
 252  *
 253  * Destroy the xprt.
 254  * Free all memory associated with the object, including its own.
 255  * NOTE: none of the *destroy methods free memory for their top-level
 256  * objects, even though they may have allocated it (they do free
 257  * private memory). It's up to the caller to handle it. In this
 258  * case (RDMA transport), all structure memory is inlined with the
 259  * struct rpcrdma_xprt.
 260  */
 261 static void
 262 xprt_rdma_destroy(struct rpc_xprt *xprt)
 263 {
 264         struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
 265         int rc;
 266
 267         dprintk("RPC:       %s: called\n", __func__);
 268
 269         cancel_delayed_work(&r_xprt->rdma_connect);
 270         flush_scheduled_work();
 271
 272         xprt_clear_connected(xprt);
 273
 274         rpcrdma_buffer_destroy(&r_xprt->rx_buf);
 275         rc = rpcrdma_ep_destroy(&r_xprt->rx_ep, &r_xprt->rx_ia);
 276         if (rc)
 277                 dprintk("RPC:       %s: rpcrdma_ep_destroy returned %i\n",
 278                         __func__, rc);
 279         rpcrdma_ia_close(&r_xprt->rx_ia);
 280
 281         xprt_rdma_free_addresses(xprt);
 282
 283         kfree(xprt->slot);
 284         xprt->slot = NULL;
 285         kfree(xprt);
 286
 287         dprintk("RPC:       %s: returning\n", __func__);
 288
 289         module_put(THIS_MODULE);
 290 }
 291
 292 static const struct rpc_timeout xprt_rdma_default_timeout = {
 293         .to_initval = 60 * HZ,
 294         .to_maxval = 60 * HZ,
 295 };
 296
 297 /**
 298  * xprt_setup_rdma - Set up transport to use RDMA
 299  *
 300  * @args: rpc transport arguments
 301  */
 302 static struct rpc_xprt *
 303 xprt_setup_rdma(struct xprt_create *args)
 304 {
 305         struct rpcrdma_create_data_internal cdata;
 306         struct rpc_xprt *xprt;
 307         struct rpcrdma_xprt *new_xprt;
 308         struct rpcrdma_ep *new_ep;
 309         struct sockaddr_in *sin;
 310         int rc;
 311
 312         if (args->addrlen > sizeof(xprt->addr)) {
 313                 dprintk("RPC:       %s: address too large\n", __func__);
 314                 return ERR_PTR(-EBADF);
 315         }
 316
 317         xprt = kzalloc(sizeof(struct rpcrdma_xprt), GFP_KERNEL);
 318         if (xprt == NULL) {
 319                 dprintk("RPC:       %s: couldn't allocate rpcrdma_xprt\n",
 320                         __func__);
 321                 return ERR_PTR(-ENOMEM);
 322         }
 323
 324         xprt->max_reqs = xprt_rdma_slot_table_entries;
 325         xprt->slot = kcalloc(xprt->max_reqs,
 326                                 sizeof(struct rpc_rqst), GFP_KERNEL);
 327         if (xprt->slot == NULL) {
 328                 dprintk("RPC:       %s: couldn't allocate %d slots\n",
 329                         __func__, xprt->max_reqs);
 330                 kfree(xprt);
 331                 return ERR_PTR(-ENOMEM);
 332         }
 333
 334         /* 60 second timeout, no retries */
 335         xprt->timeout = &xprt_rdma_default_timeout;
 336         xprt->bind_timeout = (60U * HZ);
 337         xprt->connect_timeout = (60U * HZ);
 338         xprt->reestablish_timeout = (5U * HZ);
 339         xprt->idle_timeout = (5U * 60 * HZ);
 340
 341         xprt->resvport = 0;             /* privileged port not needed */
 342         xprt->tsh_size = 0;             /* RPC-RDMA handles framing */
 343         xprt->max_payload = RPCRDMA_MAX_DATA_SEGS * PAGE_SIZE;
 344         xprt->ops = &xprt_rdma_procs;
 345
 346         /*
 347          * Set up RDMA-specific connect data.
 348          */
 349
 350         /* Put server RDMA address in local cdata */
 351         memcpy(&cdata.addr, args->dstaddr, args->addrlen);
 352
 353         /* Ensure xprt->addr holds valid server TCP (not RDMA)
 354          * address, for any side protocols which peek at it */
 355         xprt->prot = IPPROTO_TCP;
 356         xprt->addrlen = args->addrlen;
 357         memcpy(&xprt->addr, &cdata.addr, xprt->addrlen);
 358
 359         sin = (struct sockaddr_in *)&cdata.addr;
 360         if (ntohs(sin->sin_port) != 0)
 361                 xprt_set_bound(xprt);
 362
 363         dprintk("RPC:       %s: %u.%u.%u.%u:%u\n", __func__,
 364                         NIPQUAD(sin->sin_addr.s_addr), ntohs(sin->sin_port));
 365
 366         /* Set max requests */
 367         cdata.max_requests = xprt->max_reqs;
 368
 369         /* Set some length limits */
 370         cdata.rsize = RPCRDMA_MAX_SEGS * PAGE_SIZE; /* RDMA write max */
 371         cdata.wsize = RPCRDMA_MAX_SEGS * PAGE_SIZE; /* RDMA read max */
 372
 373         cdata.inline_wsize = xprt_rdma_max_inline_write;
 374         if (cdata.inline_wsize > cdata.wsize)
 375                 cdata.inline_wsize = cdata.wsize;
 376
 377         cdata.inline_rsize = xprt_rdma_max_inline_read;
 378         if (cdata.inline_rsize > cdata.rsize)
 379                 cdata.inline_rsize = cdata.rsize;
 380
 381         cdata.padding = xprt_rdma_inline_write_padding;
 382
 383         /*
 384          * Create new transport instance, which includes initialized
 385          *  o ia
 386          *  o endpoint
 387          *  o buffers
 388          */
 389
 390         new_xprt = rpcx_to_rdmax(xprt);
 391
 392         rc = rpcrdma_ia_open(new_xprt, (struct sockaddr *) &cdata.addr,
 393                                 xprt_rdma_memreg_strategy);
 394         if (rc)
 395                 goto out1;
 396
 397         /*
 398          * initialize and create ep
 399          */
 400         new_xprt->rx_data = cdata;
 401         new_ep = &new_xprt->rx_ep;
 402         new_ep->rep_remote_addr = cdata.addr;
 403
 404         rc = rpcrdma_ep_create(&new_xprt->rx_ep,
 405                                 &new_xprt->rx_ia, &new_xprt->rx_data);
 406         if (rc)
 407                 goto out2;
 408
 409         /*
 410          * Allocate pre-registered send and receive buffers for headers and
 411          * any inline data. Also specify any padding which will be provided
 412          * from a preregistered zero buffer.
 413          */
 414         rc = rpcrdma_buffer_create(&new_xprt->rx_buf, new_ep, &new_xprt->rx_ia,
 415                                 &new_xprt->rx_data);
 416         if (rc)
 417                 goto out3;
 418
 419         /*
 420          * Register a callback for connection events. This is necessary because
 421          * connection loss notification is async. We also catch connection loss
 422          * when reaping receives.
 423          */
 424         INIT_DELAYED_WORK(&new_xprt->rdma_connect, xprt_rdma_connect_worker);
 425         new_ep->rep_func = rpcrdma_conn_func;
 426         new_ep->rep_xprt = xprt;
 427
 428         xprt_rdma_format_addresses(xprt);
 429
 430         if (!try_module_get(THIS_MODULE))
 431                 goto out4;
 432
 433         return xprt;
 434
 435 out4:
 436         xprt_rdma_free_addresses(xprt);
 437         rc = -EINVAL;
 438 out3:
 439         (void) rpcrdma_ep_destroy(new_ep, &new_xprt->rx_ia);
 440 out2:
 441         rpcrdma_ia_close(&new_xprt->rx_ia);
 442 out1:
 443         kfree(xprt->slot);
 444         kfree(xprt);
 445         return ERR_PTR(rc);
 446 }
 447
 448 /*
 449  * Close a connection, during shutdown or timeout/reconnect
 450  */
 451 static void
 452 xprt_rdma_close(struct rpc_xprt *xprt)
 453 {
 454         struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
 455
 456         dprintk("RPC:       %s: closing\n", __func__);
 457         xprt_disconnect_done(xprt);
 458         (void) rpcrdma_ep_disconnect(&r_xprt->rx_ep, &r_xprt->rx_ia);
 459 }
 460
 461 static void
 462 xprt_rdma_set_port(struct rpc_xprt *xprt, u16 port)
 463 {
 464         struct sockaddr_in *sap;
 465
 466         sap = (struct sockaddr_in *)&xprt->addr;
 467         sap->sin_port = htons(port);
 468         sap = (struct sockaddr_in *)&rpcx_to_rdmad(xprt).addr;
 469         sap->sin_port = htons(port);
 470         dprintk("RPC:       %s: %u\n", __func__, port);
 471 }
 472
 473 static void
 474 xprt_rdma_connect(struct rpc_task *task)
 475 {
 476         struct rpc_xprt *xprt = (struct rpc_xprt *)task->tk_xprt;
 477         struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
 478
 479         if (!xprt_test_and_set_connecting(xprt)) {
 480                 if (r_xprt->rx_ep.rep_connected != 0) {
 481                         /* Reconnect */
 482                         schedule_delayed_work(&r_xprt->rdma_connect,
 483                                 xprt->reestablish_timeout);
 484                 } else {
 485                         schedule_delayed_work(&r_xprt->rdma_connect, 0);
 486                         if (!RPC_IS_ASYNC(task))
 487                                 flush_scheduled_work();
 488                 }
 489         }
 490 }
 491
 492 static int
 493 xprt_rdma_reserve_xprt(struct rpc_task *task)
 494 {
 495         struct rpc_xprt *xprt = task->tk_xprt;
 496         struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
 497         int credits = atomic_read(&r_xprt->rx_buf.rb_credits);
 498
 499         /* == RPC_CWNDSCALE @ init, but *after* setup */
 500         if (r_xprt->rx_buf.rb_cwndscale == 0UL) {
 501                 r_xprt->rx_buf.rb_cwndscale = xprt->cwnd;
 502                 dprintk("RPC:       %s: cwndscale %lu\n", __func__,
 503                         r_xprt->rx_buf.rb_cwndscale);
 504                 BUG_ON(r_xprt->rx_buf.rb_cwndscale <= 0);
 505         }
 506         xprt->cwnd = credits * r_xprt->rx_buf.rb_cwndscale;
 507         return xprt_reserve_xprt_cong(task);
 508 }
 509
 510 /*
 511  * The RDMA allocate/free functions need the task structure as a place
 512  * to hide the struct rpcrdma_req, which is necessary for the actual send/recv
 513  * sequence. For this reason, the recv buffers are attached to send
 514  * buffers for portions of the RPC. Note that the RPC layer allocates
 515  * both send and receive buffers in the same call. We may register
 516  * the receive buffer portion when using reply chunks.
 517  */
 518 static void *
 519 xprt_rdma_allocate(struct rpc_task *task, size_t size)
 520 {
 521         struct rpc_xprt *xprt = task->tk_xprt;
 522         struct rpcrdma_req *req, *nreq;
 523
 524         req = rpcrdma_buffer_get(&rpcx_to_rdmax(xprt)->rx_buf);
 525         BUG_ON(NULL == req);
 526
 527         if (size > req->rl_size) {
 528                 dprintk("RPC:       %s: size %zd too large for buffer[%zd]: "
 529                         "prog %d vers %d proc %d\n",
 530                         __func__, size, req->rl_size,
 531                         task->tk_client->cl_prog, task->tk_client->cl_vers,
 532                         task->tk_msg.rpc_proc->p_proc);
 533                 /*
 534                  * Outgoing length shortage. Our inline write max must have
 535                  * been configured to perform direct i/o.
 536                  *
 537                  * This is therefore a large metadata operation, and the
 538                  * allocate call was made on the maximum possible message,
 539                  * e.g. containing long filename(s) or symlink data. In
 540                  * fact, while these metadata operations *might* carry
 541                  * large outgoing payloads, they rarely *do*. However, we
 542                  * have to commit to the request here, so reallocate and
 543                  * register it now. The data path will never require this
 544                  * reallocation.
 545                  *
 546                  * If the allocation or registration fails, the RPC framework
 547                  * will (doggedly) retry.
 548                  */
 549                 if (rpcx_to_rdmax(xprt)->rx_ia.ri_memreg_strategy ==
 550                                 RPCRDMA_BOUNCEBUFFERS) {
 551                         /* forced to "pure inline" */
 552                         dprintk("RPC:       %s: too much data (%zd) for inline "
 553                                         "(r/w max %d/%d)\n", __func__, size,
 554                                         rpcx_to_rdmad(xprt).inline_rsize,
 555                                         rpcx_to_rdmad(xprt).inline_wsize);
 556                         size = req->rl_size;
 557                         rpc_exit(task, -EIO);           /* fail the operation */
 558                         rpcx_to_rdmax(xprt)->rx_stats.failed_marshal_count++;
 559                         goto out;
 560                 }
 561                 if (task->tk_flags & RPC_TASK_SWAPPER)
 562                         nreq = kmalloc(sizeof *req + size, GFP_ATOMIC);
 563                 else
 564                         nreq = kmalloc(sizeof *req + size, GFP_NOFS);
 565                 if (nreq == NULL)
 566                         goto outfail;
 567
 568                 if (rpcrdma_register_internal(&rpcx_to_rdmax(xprt)->rx_ia,
 569                                 nreq->rl_base, size + sizeof(struct rpcrdma_req)
 570                                 - offsetof(struct rpcrdma_req, rl_base),
 571                                 &nreq->rl_handle, &nreq->rl_iov)) {
 572                         kfree(nreq);
 573                         goto outfail;
 574                 }
 575                 rpcx_to_rdmax(xprt)->rx_stats.hardway_register_count += size;
 576                 nreq->rl_size = size;
 577                 nreq->rl_niovs = 0;
 578                 nreq->rl_nchunks = 0;
 579                 nreq->rl_buffer = (struct rpcrdma_buffer *)req;
 580                 nreq->rl_reply = req->rl_reply;
 581                 memcpy(nreq->rl_segments,
 582                         req->rl_segments, sizeof nreq->rl_segments);
 583                 /* flag the swap with an unused field */
 584                 nreq->rl_iov.length = 0;
 585                 req->rl_reply = NULL;
 586                 req = nreq;
 587         }
 588         dprintk("RPC:       %s: size %zd, request 0x%p\n", __func__, size, req);
 589 out:
 590         return req->rl_xdr_buf;
 591
 592 outfail:
 593         rpcrdma_buffer_put(req);
 594         rpcx_to_rdmax(xprt)->rx_stats.failed_marshal_count++;
 595         return NULL;
 596 }
 597
 598 /*
 599  * This function returns all RDMA resources to the pool.
 600  */
 601 static void
 602 xprt_rdma_free(void *buffer)
 603 {
 604         struct rpcrdma_req *req;
 605         struct rpcrdma_xprt *r_xprt;
 606         struct rpcrdma_rep *rep;
 607         int i;
 608
 609         if (buffer == NULL)
 610                 return;
 611
 612         req = container_of(buffer, struct rpcrdma_req, rl_xdr_buf[0]);
 613         if (req->rl_iov.length == 0) {  /* see allocate above */
 614                 r_xprt = container_of(((struct rpcrdma_req *) req->rl_buffer)->rl_buffer,
 615                                       struct rpcrdma_xprt, rx_buf);
 616         } else
 617                 r_xprt = container_of(req->rl_buffer, struct rpcrdma_xprt, rx_buf);
 618         rep = req->rl_reply;
 619
 620         dprintk("RPC:       %s: called on 0x%p%s\n",
 621                 __func__, rep, (rep && rep->rr_func) ? " (with waiter)" : "");
 622
 623         /*
 624          * Finish the deregistration. When using mw bind, this was
 625          * begun in rpcrdma_reply_handler(). In all other modes, we
 626          * do it here, in thread context. The process is considered
 627          * complete when the rr_func vector becomes NULL - this
 628          * was put in place during rpcrdma_reply_handler() - the wait
 629          * call below will not block if the dereg is "done". If
 630          * interrupted, our framework will clean up.
 631          */
 632         for (i = 0; req->rl_nchunks;) {
 633                 --req->rl_nchunks;
 634                 i += rpcrdma_deregister_external(
 635                         &req->rl_segments[i], r_xprt, NULL);
 636         }
 637
 638         if (rep && wait_event_interruptible(rep->rr_unbind, !rep->rr_func)) {
 639                 rep->rr_func = NULL;    /* abandon the callback */
 640                 req->rl_reply = NULL;
 641         }
 642
 643         if (req->rl_iov.length == 0) {  /* see allocate above */
 644                 struct rpcrdma_req *oreq = (struct rpcrdma_req *)req->rl_buffer;
 645                 oreq->rl_reply = req->rl_reply;
 646                 (void) rpcrdma_deregister_internal(&r_xprt->rx_ia,
 647                                                    req->rl_handle,
 648                                                    &req->rl_iov);
 649                 kfree(req);
 650                 req = oreq;
 651         }
 652
 653         /* Put back request+reply buffers */
 654         rpcrdma_buffer_put(req);
 655 }
 656
 657 /*
 658  * send_request invokes the meat of RPC RDMA. It must do the following:
 659  *  1.  Marshal the RPC request into an RPC RDMA request, which means
 660  *      putting a header in front of data, and creating IOVs for RDMA
 661  *      from those in the request.
 662  *  2.  In marshaling, detect opportunities for RDMA, and use them.
 663  *  3.  Post a recv message to set up asynch completion, then send
 664  *      the request (rpcrdma_ep_post).
 665  *  4.  No partial sends are possible in the RPC-RDMA protocol (as in UDP).
 666  */
 667
 668 static int
 669 xprt_rdma_send_request(struct rpc_task *task)
 670 {
 671         struct rpc_rqst *rqst = task->tk_rqstp;
 672         struct rpc_xprt *xprt = task->tk_xprt;
 673         struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
 674         struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
 675
 676         /* marshal the send itself */
 677         if (req->rl_niovs == 0 && rpcrdma_marshal_req(rqst) != 0) {
 678                 r_xprt->rx_stats.failed_marshal_count++;
 679                 dprintk("RPC:       %s: rpcrdma_marshal_req failed\n",
 680                         __func__);
 681                 return -EIO;
 682         }
 683
 684         if (req->rl_reply == NULL)              /* e.g. reconnection */
 685                 rpcrdma_recv_buffer_get(req);
 686
 687         if (req->rl_reply) {
 688                 req->rl_reply->rr_func = rpcrdma_reply_handler;
 689                 /* this need only be done once, but... */
 690                 req->rl_reply->rr_xprt = xprt;
 691         }
 692
 693         if (rpcrdma_ep_post(&r_xprt->rx_ia, &r_xprt->rx_ep, req)) {
 694                 xprt_disconnect_done(xprt);
 695                 return -ENOTCONN;       /* implies disconnect */
 696         }
 697
 698         rqst->rq_bytes_sent = 0;
 699         return 0;
 700 }
 701
 702 static void xprt_rdma_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
 703 {
 704         struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
 705         long idle_time = 0;
 706
 707         if (xprt_connected(xprt))
 708                 idle_time = (long)(jiffies - xprt->last_used) / HZ;
 709
 710         seq_printf(seq,
 711           "\txprt:\trdma %u %lu %lu %lu %ld %lu %lu %lu %Lu %Lu "
 712           "%lu %lu %lu %Lu %Lu %Lu %Lu %lu %lu %lu\n",
 713
 714            0,   /* need a local port? */
 715            xprt->stat.bind_count,
 716            xprt->stat.connect_count,
 717            xprt->stat.connect_time,
 718            idle_time,
 719            xprt->stat.sends,
 720            xprt->stat.recvs,
 721            xprt->stat.bad_xids,
 722            xprt->stat.req_u,
 723            xprt->stat.bklog_u,
 724
 725            r_xprt->rx_stats.read_chunk_count,
 726            r_xprt->rx_stats.write_chunk_count,
 727            r_xprt->rx_stats.reply_chunk_count,
 728            r_xprt->rx_stats.total_rdma_request,
 729            r_xprt->rx_stats.total_rdma_reply,
 730            r_xprt->rx_stats.pullup_copy_count,
 731            r_xprt->rx_stats.fixup_copy_count,
 732            r_xprt->rx_stats.hardway_register_count,
 733            r_xprt->rx_stats.failed_marshal_count,
 734            r_xprt->rx_stats.bad_reply_count);
 735 }
 736
 737 /*
 738  * Plumbing for rpc transport switch and kernel module
 739  */
 740
 741 static struct rpc_xprt_ops xprt_rdma_procs = {
 742         .reserve_xprt           = xprt_rdma_reserve_xprt,
 743         .release_xprt           = xprt_release_xprt_cong, /* sunrpc/xprt.c */
 744         .release_request        = xprt_release_rqst_cong,       /* ditto */
 745         .set_retrans_timeout    = xprt_set_retrans_timeout_def, /* ditto */
 746         .rpcbind                = rpcb_getport_async,   /* sunrpc/rpcb_clnt.c */
 747         .set_port               = xprt_rdma_set_port,
 748         .connect                = xprt_rdma_connect,
 749         .buf_alloc              = xprt_rdma_allocate,
 750         .buf_free               = xprt_rdma_free,
 751         .send_request           = xprt_rdma_send_request,
 752         .close                  = xprt_rdma_close,
 753         .destroy                = xprt_rdma_destroy,
 754         .print_stats            = xprt_rdma_print_stats
 755 };
 756
 757 static struct xprt_class xprt_rdma = {
 758         .list                   = LIST_HEAD_INIT(xprt_rdma.list),
 759         .name                   = "rdma",
 760         .owner                  = THIS_MODULE,
 761         .ident                  = XPRT_TRANSPORT_RDMA,
 762         .setup                  = xprt_setup_rdma,
 763 };
 764
 765 static void __exit xprt_rdma_cleanup(void)
 766 {
 767         int rc;
 768
 769         dprintk("RPCRDMA Module Removed, deregister RPC RDMA transport\n");
 770 #ifdef RPC_DEBUG
 771         if (sunrpc_table_header) {
 772                 unregister_sysctl_table(sunrpc_table_header);
 773                 sunrpc_table_header = NULL;
 774         }
 775 #endif
 776         rc = xprt_unregister_transport(&xprt_rdma);
 777         if (rc)
 778                 dprintk("RPC:       %s: xprt_unregister returned %i\n",
 779                         __func__, rc);
 780 }
 781
 782 static int __init xprt_rdma_init(void)
 783 {
 784         int rc;
 785
 786         rc = xprt_register_transport(&xprt_rdma);
 787
 788         if (rc)
 789                 return rc;
 790
 791         dprintk(KERN_INFO "RPCRDMA Module Init, register RPC RDMA transport\n");
 792
 793         dprintk(KERN_INFO "Defaults:\n");
 794         dprintk(KERN_INFO "\tSlots %d\n"
 795                 "\tMaxInlineRead %d\n\tMaxInlineWrite %d\n",
 796                 xprt_rdma_slot_table_entries,
 797                 xprt_rdma_max_inline_read, xprt_rdma_max_inline_write);
 798         dprintk(KERN_INFO "\tPadding %d\n\tMemreg %d\n",
 799                 xprt_rdma_inline_write_padding, xprt_rdma_memreg_strategy);
 800
 801 #ifdef RPC_DEBUG
 802         if (!sunrpc_table_header)
 803                 sunrpc_table_header = register_sysctl_table(sunrpc_table);
 804 #endif
 805         return 0;
 806 }
 807
 808 module_init(xprt_rdma_init);
 809 module_exit(xprt_rdma_cleanup);