SAFE public projects git trees. - safe/jmp/linux-2.6/blob - net/sunrpc/xprt.c

   1 /*
   2  *  linux/net/sunrpc/xprt.c
   3  *
   4  *  This is a generic RPC call interface supporting congestion avoidance,
   5  *  and asynchronous calls.
   6  *
   7  *  The interface works like this:
   8  *
   9  *  -   When a process places a call, it allocates a request slot if
  10  *      one is available. Otherwise, it sleeps on the backlog queue
  11  *      (xprt_reserve).
  12  *  -   Next, the caller puts together the RPC message, stuffs it into
  13  *      the request struct, and calls xprt_call().
  14  *  -   xprt_call transmits the message and installs the caller on the
  15  *      socket's wait list. At the same time, it installs a timer that
  16  *      is run after the packet's timeout has expired.
  17  *  -   When a packet arrives, the data_ready handler walks the list of
  18  *      pending requests for that socket. If a matching XID is found, the
  19  *      caller is woken up, and the timer removed.
  20  *  -   When no reply arrives within the timeout interval, the timer is
  21  *      fired by the kernel and runs xprt_timer(). It either adjusts the
  22  *      timeout values (minor timeout) or wakes up the caller with a status
  23  *      of -ETIMEDOUT.
  24  *  -   When the caller receives a notification from RPC that a reply arrived,
  25  *      it should release the RPC slot, and process the reply.
  26  *      If the call timed out, it may choose to retry the operation by
  27  *      adjusting the initial timeout value, and simply calling rpc_call
  28  *      again.
  29  *
  30  *  Support for async RPC is done through a set of RPC-specific scheduling
  31  *  primitives that `transparently' work for processes as well as async
  32  *  tasks that rely on callbacks.
  33  *
  34  *  Copyright (C) 1995-1997, Olaf Kirch <okir@monad.swb.de>
  35  */
  36
  37 #include <linux/module.h>
  38
  39 #include <linux/types.h>
  40 #include <linux/interrupt.h>
  41 #include <linux/workqueue.h>
  42 #include <linux/random.h>
  43
  44 #include <linux/sunrpc/clnt.h>
  45
  46 /*
  47  * Local variables
  48  */
  49
  50 #ifdef RPC_DEBUG
  51 # undef  RPC_DEBUG_DATA
  52 # define RPCDBG_FACILITY        RPCDBG_XPRT
  53 #endif
  54
  55 #define XPRT_MAX_BACKOFF        (8)
  56
  57 /*
  58  * Local functions
  59  */
  60 static void     xprt_request_init(struct rpc_task *, struct rpc_xprt *);
  61 static inline void      do_xprt_reserve(struct rpc_task *);
  62 static void     xprt_connect_status(struct rpc_task *task);
  63 static int      __xprt_get_cong(struct rpc_xprt *, struct rpc_task *);
  64
  65 static int      xprt_clear_backlog(struct rpc_xprt *xprt);
  66
  67 /*
  68  * Serialize write access to sockets, in order to prevent different
  69  * requests from interfering with each other.
  70  * Also prevents TCP socket connects from colliding with writes.
  71  */
  72 static int
  73 __xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
  74 {
  75         struct rpc_rqst *req = task->tk_rqstp;
  76
  77         if (test_and_set_bit(XPRT_LOCKED, &xprt->sockstate)) {
  78                 if (task == xprt->snd_task)
  79                         return 1;
  80                 goto out_sleep;
  81         }
  82         if (xprt->nocong || __xprt_get_cong(xprt, task)) {
  83                 xprt->snd_task = task;
  84                 if (req) {
  85                         req->rq_bytes_sent = 0;
  86                         req->rq_ntrans++;
  87                 }
  88                 return 1;
  89         }
  90         smp_mb__before_clear_bit();
  91         clear_bit(XPRT_LOCKED, &xprt->sockstate);
  92         smp_mb__after_clear_bit();
  93 out_sleep:
  94         dprintk("RPC: %4d failed to lock socket %p\n", task->tk_pid, xprt);
  95         task->tk_timeout = 0;
  96         task->tk_status = -EAGAIN;
  97         if (req && req->rq_ntrans)
  98                 rpc_sleep_on(&xprt->resend, task, NULL, NULL);
  99         else
 100                 rpc_sleep_on(&xprt->sending, task, NULL, NULL);
 101         return 0;
 102 }
 103
 104 static inline int
 105 xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
 106 {
 107         int retval;
 108
 109         spin_lock_bh(&xprt->transport_lock);
 110         retval = __xprt_lock_write(xprt, task);
 111         spin_unlock_bh(&xprt->transport_lock);
 112         return retval;
 113 }
 114
 115
 116 static void
 117 __xprt_lock_write_next(struct rpc_xprt *xprt)
 118 {
 119         struct rpc_task *task;
 120
 121         if (test_and_set_bit(XPRT_LOCKED, &xprt->sockstate))
 122                 return;
 123         if (!xprt->nocong && RPCXPRT_CONGESTED(xprt))
 124                 goto out_unlock;
 125         task = rpc_wake_up_next(&xprt->resend);
 126         if (!task) {
 127                 task = rpc_wake_up_next(&xprt->sending);
 128                 if (!task)
 129                         goto out_unlock;
 130         }
 131         if (xprt->nocong || __xprt_get_cong(xprt, task)) {
 132                 struct rpc_rqst *req = task->tk_rqstp;
 133                 xprt->snd_task = task;
 134                 if (req) {
 135                         req->rq_bytes_sent = 0;
 136                         req->rq_ntrans++;
 137                 }
 138                 return;
 139         }
 140 out_unlock:
 141         smp_mb__before_clear_bit();
 142         clear_bit(XPRT_LOCKED, &xprt->sockstate);
 143         smp_mb__after_clear_bit();
 144 }
 145
 146 /*
 147  * Releases the socket for use by other requests.
 148  */
 149 static void
 150 __xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
 151 {
 152         if (xprt->snd_task == task) {
 153                 xprt->snd_task = NULL;
 154                 smp_mb__before_clear_bit();
 155                 clear_bit(XPRT_LOCKED, &xprt->sockstate);
 156                 smp_mb__after_clear_bit();
 157                 __xprt_lock_write_next(xprt);
 158         }
 159 }
 160
 161 static inline void
 162 xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
 163 {
 164         spin_lock_bh(&xprt->transport_lock);
 165         __xprt_release_write(xprt, task);
 166         spin_unlock_bh(&xprt->transport_lock);
 167 }
 168
 169 /*
 170  * Van Jacobson congestion avoidance. Check if the congestion window
 171  * overflowed. Put the task to sleep if this is the case.
 172  */
 173 static int
 174 __xprt_get_cong(struct rpc_xprt *xprt, struct rpc_task *task)
 175 {
 176         struct rpc_rqst *req = task->tk_rqstp;
 177
 178         if (req->rq_cong)
 179                 return 1;
 180         dprintk("RPC: %4d xprt_cwnd_limited cong = %ld cwnd = %ld\n",
 181                         task->tk_pid, xprt->cong, xprt->cwnd);
 182         if (RPCXPRT_CONGESTED(xprt))
 183                 return 0;
 184         req->rq_cong = 1;
 185         xprt->cong += RPC_CWNDSCALE;
 186         return 1;
 187 }
 188
 189 /*
 190  * Adjust the congestion window, and wake up the next task
 191  * that has been sleeping due to congestion
 192  */
 193 static void
 194 __xprt_put_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
 195 {
 196         if (!req->rq_cong)
 197                 return;
 198         req->rq_cong = 0;
 199         xprt->cong -= RPC_CWNDSCALE;
 200         __xprt_lock_write_next(xprt);
 201 }
 202
 203 /*
 204  * Adjust RPC congestion window
 205  * We use a time-smoothed congestion estimator to avoid heavy oscillation.
 206  */
 207 static void
 208 xprt_adjust_cwnd(struct rpc_xprt *xprt, int result)
 209 {
 210         unsigned long   cwnd;
 211
 212         cwnd = xprt->cwnd;
 213         if (result >= 0 && cwnd <= xprt->cong) {
 214                 /* The (cwnd >> 1) term makes sure
 215                  * the result gets rounded properly. */
 216                 cwnd += (RPC_CWNDSCALE * RPC_CWNDSCALE + (cwnd >> 1)) / cwnd;
 217                 if (cwnd > RPC_MAXCWND(xprt))
 218                         cwnd = RPC_MAXCWND(xprt);
 219                 __xprt_lock_write_next(xprt);
 220         } else if (result == -ETIMEDOUT) {
 221                 cwnd >>= 1;
 222                 if (cwnd < RPC_CWNDSCALE)
 223                         cwnd = RPC_CWNDSCALE;
 224         }
 225         dprintk("RPC:      cong %ld, cwnd was %ld, now %ld\n",
 226                         xprt->cong, xprt->cwnd, cwnd);
 227         xprt->cwnd = cwnd;
 228 }
 229
 230 static void xprt_reset_majortimeo(struct rpc_rqst *req)
 231 {
 232         struct rpc_timeout *to = &req->rq_xprt->timeout;
 233
 234         req->rq_majortimeo = req->rq_timeout;
 235         if (to->to_exponential)
 236                 req->rq_majortimeo <<= to->to_retries;
 237         else
 238                 req->rq_majortimeo += to->to_increment * to->to_retries;
 239         if (req->rq_majortimeo > to->to_maxval || req->rq_majortimeo == 0)
 240                 req->rq_majortimeo = to->to_maxval;
 241         req->rq_majortimeo += jiffies;
 242 }
 243
 244 /**
 245  * xprt_adjust_timeout - adjust timeout values for next retransmit
 246  * @req: RPC request containing parameters to use for the adjustment
 247  *
 248  */
 249 int xprt_adjust_timeout(struct rpc_rqst *req)
 250 {
 251         struct rpc_xprt *xprt = req->rq_xprt;
 252         struct rpc_timeout *to = &xprt->timeout;
 253         int status = 0;
 254
 255         if (time_before(jiffies, req->rq_majortimeo)) {
 256                 if (to->to_exponential)
 257                         req->rq_timeout <<= 1;
 258                 else
 259                         req->rq_timeout += to->to_increment;
 260                 if (to->to_maxval && req->rq_timeout >= to->to_maxval)
 261                         req->rq_timeout = to->to_maxval;
 262                 req->rq_retries++;
 263                 pprintk("RPC: %lu retrans\n", jiffies);
 264         } else {
 265                 req->rq_timeout = to->to_initval;
 266                 req->rq_retries = 0;
 267                 xprt_reset_majortimeo(req);
 268                 /* Reset the RTT counters == "slow start" */
 269                 spin_lock_bh(&xprt->transport_lock);
 270                 rpc_init_rtt(req->rq_task->tk_client->cl_rtt, to->to_initval);
 271                 spin_unlock_bh(&xprt->transport_lock);
 272                 pprintk("RPC: %lu timeout\n", jiffies);
 273                 status = -ETIMEDOUT;
 274         }
 275
 276         if (req->rq_timeout == 0) {
 277                 printk(KERN_WARNING "xprt_adjust_timeout: rq_timeout = 0!\n");
 278                 req->rq_timeout = 5 * HZ;
 279         }
 280         return status;
 281 }
 282
 283 static void
 284 xprt_socket_autoclose(void *args)
 285 {
 286         struct rpc_xprt *xprt = (struct rpc_xprt *)args;
 287
 288         xprt_disconnect(xprt);
 289         xprt->ops->close(xprt);
 290         xprt_release_write(xprt, NULL);
 291 }
 292
 293 /**
 294  * xprt_disconnect - mark a transport as disconnected
 295  * @xprt: transport to flag for disconnect
 296  *
 297  */
 298 void xprt_disconnect(struct rpc_xprt *xprt)
 299 {
 300         dprintk("RPC:      disconnected transport %p\n", xprt);
 301         spin_lock_bh(&xprt->transport_lock);
 302         xprt_clear_connected(xprt);
 303         rpc_wake_up_status(&xprt->pending, -ENOTCONN);
 304         spin_unlock_bh(&xprt->transport_lock);
 305 }
 306
 307 static void
 308 xprt_init_autodisconnect(unsigned long data)
 309 {
 310         struct rpc_xprt *xprt = (struct rpc_xprt *)data;
 311
 312         spin_lock(&xprt->transport_lock);
 313         if (!list_empty(&xprt->recv) || xprt->shutdown)
 314                 goto out_abort;
 315         if (test_and_set_bit(XPRT_LOCKED, &xprt->sockstate))
 316                 goto out_abort;
 317         spin_unlock(&xprt->transport_lock);
 318         /* Let keventd close the socket */
 319         if (test_bit(XPRT_CONNECTING, &xprt->sockstate) != 0)
 320                 xprt_release_write(xprt, NULL);
 321         else
 322                 schedule_work(&xprt->task_cleanup);
 323         return;
 324 out_abort:
 325         spin_unlock(&xprt->transport_lock);
 326 }
 327
 328 /**
 329  * xprt_connect - schedule a transport connect operation
 330  * @task: RPC task that is requesting the connect
 331  *
 332  */
 333 void xprt_connect(struct rpc_task *task)
 334 {
 335         struct rpc_xprt *xprt = task->tk_xprt;
 336
 337         dprintk("RPC: %4d xprt_connect xprt %p %s connected\n", task->tk_pid,
 338                         xprt, (xprt_connected(xprt) ? "is" : "is not"));
 339
 340         if (xprt->shutdown) {
 341                 task->tk_status = -EIO;
 342                 return;
 343         }
 344         if (!xprt->addr.sin_port) {
 345                 task->tk_status = -EIO;
 346                 return;
 347         }
 348         if (!xprt_lock_write(xprt, task))
 349                 return;
 350         if (xprt_connected(xprt))
 351                 xprt_release_write(xprt, task);
 352         else {
 353                 if (task->tk_rqstp)
 354                         task->tk_rqstp->rq_bytes_sent = 0;
 355
 356                 task->tk_timeout = RPC_CONNECT_TIMEOUT;
 357                 rpc_sleep_on(&xprt->pending, task, xprt_connect_status, NULL);
 358                 xprt->ops->connect(task);
 359         }
 360         return;
 361 }
 362
 363 static void xprt_connect_status(struct rpc_task *task)
 364 {
 365         struct rpc_xprt *xprt = task->tk_xprt;
 366
 367         if (task->tk_status >= 0) {
 368                 dprintk("RPC: %4d xprt_connect_status: connection established\n",
 369                                 task->tk_pid);
 370                 return;
 371         }
 372
 373         switch (task->tk_status) {
 374         case -ECONNREFUSED:
 375         case -ECONNRESET:
 376                 dprintk("RPC: %4d xprt_connect_status: server %s refused connection\n",
 377                                 task->tk_pid, task->tk_client->cl_server);
 378                 break;
 379         case -ENOTCONN:
 380                 dprintk("RPC: %4d xprt_connect_status: connection broken\n",
 381                                 task->tk_pid);
 382                 break;
 383         case -ETIMEDOUT:
 384                 dprintk("RPC: %4d xprt_connect_status: connect attempt timed out\n",
 385                                 task->tk_pid);
 386                 break;
 387         default:
 388                 dprintk("RPC: %4d xprt_connect_status: error %d connecting to server %s\n",
 389                                 task->tk_pid, -task->tk_status, task->tk_client->cl_server);
 390                 xprt_release_write(xprt, task);
 391                 task->tk_status = -EIO;
 392                 return;
 393         }
 394
 395         /* if soft mounted, just cause this RPC to fail */
 396         if (RPC_IS_SOFT(task)) {
 397                 xprt_release_write(xprt, task);
 398                 task->tk_status = -EIO;
 399         }
 400 }
 401
 402 /**
 403  * xprt_lookup_rqst - find an RPC request corresponding to an XID
 404  * @xprt: transport on which the original request was transmitted
 405  * @xid: RPC XID of incoming reply
 406  *
 407  */
 408 struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid)
 409 {
 410         struct list_head *pos;
 411         struct rpc_rqst *req = NULL;
 412
 413         list_for_each(pos, &xprt->recv) {
 414                 struct rpc_rqst *entry = list_entry(pos, struct rpc_rqst, rq_list);
 415                 if (entry->rq_xid == xid) {
 416                         req = entry;
 417                         break;
 418                 }
 419         }
 420         return req;
 421 }
 422
 423 /**
 424  * xprt_complete_rqst - called when reply processing is complete
 425  * @xprt: controlling transport
 426  * @req: RPC request that just completed
 427  * @copied: actual number of bytes received from the transport
 428  *
 429  */
 430 void xprt_complete_rqst(struct rpc_xprt *xprt, struct rpc_rqst *req, int copied)
 431 {
 432         struct rpc_task *task = req->rq_task;
 433         struct rpc_clnt *clnt = task->tk_client;
 434
 435         /* Adjust congestion window */
 436         if (!xprt->nocong) {
 437                 unsigned timer = task->tk_msg.rpc_proc->p_timer;
 438                 xprt_adjust_cwnd(xprt, copied);
 439                 __xprt_put_cong(xprt, req);
 440                 if (timer) {
 441                         if (req->rq_ntrans == 1)
 442                                 rpc_update_rtt(clnt->cl_rtt, timer,
 443                                                 (long)jiffies - req->rq_xtime);
 444                         rpc_set_timeo(clnt->cl_rtt, timer, req->rq_ntrans - 1);
 445                 }
 446         }
 447
 448 #ifdef RPC_PROFILE
 449         /* Profile only reads for now */
 450         if (copied > 1024) {
 451                 static unsigned long    nextstat;
 452                 static unsigned long    pkt_rtt, pkt_len, pkt_cnt;
 453
 454                 pkt_cnt++;
 455                 pkt_len += req->rq_slen + copied;
 456                 pkt_rtt += jiffies - req->rq_xtime;
 457                 if (time_before(nextstat, jiffies)) {
 458                         printk("RPC: %lu %ld cwnd\n", jiffies, xprt->cwnd);
 459                         printk("RPC: %ld %ld %ld %ld stat\n",
 460                                         jiffies, pkt_cnt, pkt_len, pkt_rtt);
 461                         pkt_rtt = pkt_len = pkt_cnt = 0;
 462                         nextstat = jiffies + 5 * HZ;
 463                 }
 464         }
 465 #endif
 466
 467         dprintk("RPC: %4d has input (%d bytes)\n", task->tk_pid, copied);
 468         list_del_init(&req->rq_list);
 469         req->rq_received = req->rq_private_buf.len = copied;
 470
 471         /* ... and wake up the process. */
 472         rpc_wake_up_task(task);
 473         return;
 474 }
 475
 476 /*
 477  * RPC receive timeout handler.
 478  */
 479 static void
 480 xprt_timer(struct rpc_task *task)
 481 {
 482         struct rpc_rqst *req = task->tk_rqstp;
 483         struct rpc_xprt *xprt = req->rq_xprt;
 484
 485         spin_lock(&xprt->transport_lock);
 486         if (req->rq_received)
 487                 goto out;
 488
 489         xprt_adjust_cwnd(req->rq_xprt, -ETIMEDOUT);
 490         __xprt_put_cong(xprt, req);
 491
 492         dprintk("RPC: %4d xprt_timer (%s request)\n",
 493                 task->tk_pid, req ? "pending" : "backlogged");
 494
 495         task->tk_status  = -ETIMEDOUT;
 496 out:
 497         task->tk_timeout = 0;
 498         rpc_wake_up_task(task);
 499         spin_unlock(&xprt->transport_lock);
 500 }
 501
 502 /**
 503  * xprt_prepare_transmit - reserve the transport before sending a request
 504  * @task: RPC task about to send a request
 505  *
 506  */
 507 int xprt_prepare_transmit(struct rpc_task *task)
 508 {
 509         struct rpc_rqst *req = task->tk_rqstp;
 510         struct rpc_xprt *xprt = req->rq_xprt;
 511         int err = 0;
 512
 513         dprintk("RPC: %4d xprt_prepare_transmit\n", task->tk_pid);
 514
 515         if (xprt->shutdown)
 516                 return -EIO;
 517
 518         spin_lock_bh(&xprt->transport_lock);
 519         if (req->rq_received && !req->rq_bytes_sent) {
 520                 err = req->rq_received;
 521                 goto out_unlock;
 522         }
 523         if (!__xprt_lock_write(xprt, task)) {
 524                 err = -EAGAIN;
 525                 goto out_unlock;
 526         }
 527
 528         if (!xprt_connected(xprt)) {
 529                 err = -ENOTCONN;
 530                 goto out_unlock;
 531         }
 532 out_unlock:
 533         spin_unlock_bh(&xprt->transport_lock);
 534         return err;
 535 }
 536
 537 /**
 538  * xprt_transmit - send an RPC request on a transport
 539  * @task: controlling RPC task
 540  *
 541  * We have to copy the iovec because sendmsg fiddles with its contents.
 542  */
 543 void xprt_transmit(struct rpc_task *task)
 544 {
 545         struct rpc_clnt *clnt = task->tk_client;
 546         struct rpc_rqst *req = task->tk_rqstp;
 547         struct rpc_xprt *xprt = req->rq_xprt;
 548         int status;
 549
 550         dprintk("RPC: %4d xprt_transmit(%u)\n", task->tk_pid, req->rq_slen);
 551
 552         smp_rmb();
 553         if (!req->rq_received) {
 554                 if (list_empty(&req->rq_list)) {
 555                         spin_lock_bh(&xprt->transport_lock);
 556                         /* Update the softirq receive buffer */
 557                         memcpy(&req->rq_private_buf, &req->rq_rcv_buf,
 558                                         sizeof(req->rq_private_buf));
 559                         /* Add request to the receive list */
 560                         list_add_tail(&req->rq_list, &xprt->recv);
 561                         spin_unlock_bh(&xprt->transport_lock);
 562                         xprt_reset_majortimeo(req);
 563                         /* Turn off autodisconnect */
 564                         del_singleshot_timer_sync(&xprt->timer);
 565                 }
 566         } else if (!req->rq_bytes_sent)
 567                 return;
 568
 569         status = xprt->ops->send_request(task);
 570         if (!status)
 571                 goto out_receive;
 572
 573         /* Note: at this point, task->tk_sleeping has not yet been set,
 574          *       hence there is no danger of the waking up task being put on
 575          *       schedq, and being picked up by a parallel run of rpciod().
 576          */
 577         task->tk_status = status;
 578
 579         switch (status) {
 580         case -ECONNREFUSED:
 581                 task->tk_timeout = RPC_REESTABLISH_TIMEOUT;
 582                 rpc_sleep_on(&xprt->sending, task, NULL, NULL);
 583         case -EAGAIN:
 584         case -ENOTCONN:
 585                 return;
 586         default:
 587                 if (xprt->stream)
 588                         xprt_disconnect(xprt);
 589         }
 590         xprt_release_write(xprt, task);
 591         return;
 592  out_receive:
 593         dprintk("RPC: %4d xmit complete\n", task->tk_pid);
 594         /* Set the task's receive timeout value */
 595         spin_lock_bh(&xprt->transport_lock);
 596         if (!xprt->nocong) {
 597                 int timer = task->tk_msg.rpc_proc->p_timer;
 598                 task->tk_timeout = rpc_calc_rto(clnt->cl_rtt, timer);
 599                 task->tk_timeout <<= rpc_ntimeo(clnt->cl_rtt, timer) + req->rq_retries;
 600                 if (task->tk_timeout > xprt->timeout.to_maxval || task->tk_timeout == 0)
 601                         task->tk_timeout = xprt->timeout.to_maxval;
 602         } else
 603                 task->tk_timeout = req->rq_timeout;
 604         /* Don't race with disconnect */
 605         if (!xprt_connected(xprt))
 606                 task->tk_status = -ENOTCONN;
 607         else if (!req->rq_received)
 608                 rpc_sleep_on(&xprt->pending, task, NULL, xprt_timer);
 609         __xprt_release_write(xprt, task);
 610         spin_unlock_bh(&xprt->transport_lock);
 611 }
 612
 613 static inline void do_xprt_reserve(struct rpc_task *task)
 614 {
 615         struct rpc_xprt *xprt = task->tk_xprt;
 616
 617         task->tk_status = 0;
 618         if (task->tk_rqstp)
 619                 return;
 620         if (!list_empty(&xprt->free)) {
 621                 struct rpc_rqst *req = list_entry(xprt->free.next, struct rpc_rqst, rq_list);
 622                 list_del_init(&req->rq_list);
 623                 task->tk_rqstp = req;
 624                 xprt_request_init(task, xprt);
 625                 return;
 626         }
 627         dprintk("RPC:      waiting for request slot\n");
 628         task->tk_status = -EAGAIN;
 629         task->tk_timeout = 0;
 630         rpc_sleep_on(&xprt->backlog, task, NULL, NULL);
 631 }
 632
 633 /**
 634  * xprt_reserve - allocate an RPC request slot
 635  * @task: RPC task requesting a slot allocation
 636  *
 637  * If no more slots are available, place the task on the transport's
 638  * backlog queue.
 639  */
 640 void xprt_reserve(struct rpc_task *task)
 641 {
 642         struct rpc_xprt *xprt = task->tk_xprt;
 643
 644         task->tk_status = -EIO;
 645         if (!xprt->shutdown) {
 646                 spin_lock(&xprt->reserve_lock);
 647                 do_xprt_reserve(task);
 648                 spin_unlock(&xprt->reserve_lock);
 649         }
 650 }
 651
 652 static inline u32 xprt_alloc_xid(struct rpc_xprt *xprt)
 653 {
 654         return xprt->xid++;
 655 }
 656
 657 static inline void xprt_init_xid(struct rpc_xprt *xprt)
 658 {
 659         get_random_bytes(&xprt->xid, sizeof(xprt->xid));
 660 }
 661
 662 static void xprt_request_init(struct rpc_task *task, struct rpc_xprt *xprt)
 663 {
 664         struct rpc_rqst *req = task->tk_rqstp;
 665
 666         req->rq_timeout = xprt->timeout.to_initval;
 667         req->rq_task    = task;
 668         req->rq_xprt    = xprt;
 669         req->rq_xid     = xprt_alloc_xid(xprt);
 670         dprintk("RPC: %4d reserved req %p xid %08x\n", task->tk_pid,
 671                         req, ntohl(req->rq_xid));
 672 }
 673
 674 /**
 675  * xprt_release - release an RPC request slot
 676  * @task: task which is finished with the slot
 677  *
 678  */
 679 void xprt_release(struct rpc_task *task)
 680 {
 681         struct rpc_xprt *xprt = task->tk_xprt;
 682         struct rpc_rqst *req;
 683
 684         if (!(req = task->tk_rqstp))
 685                 return;
 686         spin_lock_bh(&xprt->transport_lock);
 687         __xprt_release_write(xprt, task);
 688         __xprt_put_cong(xprt, req);
 689         if (!list_empty(&req->rq_list))
 690                 list_del(&req->rq_list);
 691         xprt->last_used = jiffies;
 692         if (list_empty(&xprt->recv) && !xprt->shutdown)
 693                 mod_timer(&xprt->timer,
 694                                 xprt->last_used + RPC_IDLE_DISCONNECT_TIMEOUT);
 695         spin_unlock_bh(&xprt->transport_lock);
 696         task->tk_rqstp = NULL;
 697         memset(req, 0, sizeof(*req));   /* mark unused */
 698
 699         dprintk("RPC: %4d release request %p\n", task->tk_pid, req);
 700
 701         spin_lock(&xprt->reserve_lock);
 702         list_add(&req->rq_list, &xprt->free);
 703         xprt_clear_backlog(xprt);
 704         spin_unlock(&xprt->reserve_lock);
 705 }
 706
 707 /**
 708  * xprt_set_timeout - set constant RPC timeout
 709  * @to: RPC timeout parameters to set up
 710  * @retr: number of retries
 711  * @incr: amount of increase after each retry
 712  *
 713  */
 714 void xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long incr)
 715 {
 716         to->to_initval   =
 717         to->to_increment = incr;
 718         to->to_maxval    = to->to_initval + (incr * retr);
 719         to->to_retries   = retr;
 720         to->to_exponential = 0;
 721 }
 722
 723 static struct rpc_xprt *xprt_setup(int proto, struct sockaddr_in *ap, struct rpc_timeout *to)
 724 {
 725         int result;
 726         struct rpc_xprt *xprt;
 727         struct rpc_rqst *req;
 728
 729         if ((xprt = kmalloc(sizeof(struct rpc_xprt), GFP_KERNEL)) == NULL)
 730                 return ERR_PTR(-ENOMEM);
 731         memset(xprt, 0, sizeof(*xprt)); /* Nnnngh! */
 732
 733         xprt->addr = *ap;
 734
 735         switch (proto) {
 736         case IPPROTO_UDP:
 737                 result = xs_setup_udp(xprt, to);
 738                 break;
 739         case IPPROTO_TCP:
 740                 result = xs_setup_tcp(xprt, to);
 741                 break;
 742         default:
 743                 printk(KERN_ERR "RPC: unrecognized transport protocol: %d\n",
 744                                 proto);
 745                 result = -EIO;
 746                 break;
 747         }
 748         if (result) {
 749                 kfree(xprt);
 750                 return ERR_PTR(result);
 751         }
 752
 753         spin_lock_init(&xprt->transport_lock);
 754         spin_lock_init(&xprt->reserve_lock);
 755         init_waitqueue_head(&xprt->cong_wait);
 756
 757         INIT_LIST_HEAD(&xprt->free);
 758         INIT_LIST_HEAD(&xprt->recv);
 759         INIT_WORK(&xprt->task_cleanup, xprt_socket_autoclose, xprt);
 760         init_timer(&xprt->timer);
 761         xprt->timer.function = xprt_init_autodisconnect;
 762         xprt->timer.data = (unsigned long) xprt;
 763         xprt->last_used = jiffies;
 764
 765         rpc_init_wait_queue(&xprt->pending, "xprt_pending");
 766         rpc_init_wait_queue(&xprt->sending, "xprt_sending");
 767         rpc_init_wait_queue(&xprt->resend, "xprt_resend");
 768         rpc_init_priority_wait_queue(&xprt->backlog, "xprt_backlog");
 769
 770         /* initialize free list */
 771         for (req = &xprt->slot[xprt->max_reqs-1]; req >= &xprt->slot[0]; req--)
 772                 list_add(&req->rq_list, &xprt->free);
 773
 774         xprt_init_xid(xprt);
 775
 776         dprintk("RPC:      created transport %p with %u slots\n", xprt,
 777                         xprt->max_reqs);
 778
 779         return xprt;
 780 }
 781
 782 /**
 783  * xprt_create_proto - create an RPC client transport
 784  * @proto: requested transport protocol
 785  * @sap: remote peer's address
 786  * @to: timeout parameters for new transport
 787  *
 788  */
 789 struct rpc_xprt *xprt_create_proto(int proto, struct sockaddr_in *sap, struct rpc_timeout *to)
 790 {
 791         struct rpc_xprt *xprt;
 792
 793         xprt = xprt_setup(proto, sap, to);
 794         if (IS_ERR(xprt))
 795                 dprintk("RPC:      xprt_create_proto failed\n");
 796         else
 797                 dprintk("RPC:      xprt_create_proto created xprt %p\n", xprt);
 798         return xprt;
 799 }
 800
 801 static void xprt_shutdown(struct rpc_xprt *xprt)
 802 {
 803         xprt->shutdown = 1;
 804         rpc_wake_up(&xprt->sending);
 805         rpc_wake_up(&xprt->resend);
 806         rpc_wake_up(&xprt->pending);
 807         rpc_wake_up(&xprt->backlog);
 808         wake_up(&xprt->cong_wait);
 809         del_timer_sync(&xprt->timer);
 810 }
 811
 812 static int xprt_clear_backlog(struct rpc_xprt *xprt) {
 813         rpc_wake_up_next(&xprt->backlog);
 814         wake_up(&xprt->cong_wait);
 815         return 1;
 816 }
 817
 818 /**
 819  * xprt_destroy - destroy an RPC transport, killing off all requests.
 820  * @xprt: transport to destroy
 821  *
 822  */
 823 int xprt_destroy(struct rpc_xprt *xprt)
 824 {
 825         dprintk("RPC:      destroying transport %p\n", xprt);
 826         xprt_shutdown(xprt);
 827         xprt->ops->destroy(xprt);
 828         kfree(xprt);
 829
 830         return 0;
 831 }