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_transmit().
  14  *  -   xprt_transmit sends the message and installs the caller on the
  15  *      transport's wait list. At the same time, if a reply is expected,
  16  *      it installs a timer that is run after the packet's timeout has
  17  *      expired.
  18  *  -   When a packet arrives, the data_ready handler walks the list of
  19  *      pending requests for that transport. If a matching XID is found, the
  20  *      caller is woken up, and the timer removed.
  21  *  -   When no reply arrives within the timeout interval, the timer is
  22  *      fired by the kernel and runs xprt_timer(). It either adjusts the
  23  *      timeout values (minor timeout) or wakes up the caller with a status
  24  *      of -ETIMEDOUT.
  25  *  -   When the caller receives a notification from RPC that a reply arrived,
  26  *      it should release the RPC slot, and process the reply.
  27  *      If the call timed out, it may choose to retry the operation by
  28  *      adjusting the initial timeout value, and simply calling rpc_call
  29  *      again.
  30  *
  31  *  Support for async RPC is done through a set of RPC-specific scheduling
  32  *  primitives that `transparently' work for processes as well as async
  33  *  tasks that rely on callbacks.
  34  *
  35  *  Copyright (C) 1995-1997, Olaf Kirch <okir@monad.swb.de>
  36  *
  37  *  Transport switch API copyright (C) 2005, Chuck Lever <cel@netapp.com>
  38  */
  39
  40 #include <linux/module.h>
  41
  42 #include <linux/types.h>
  43 #include <linux/interrupt.h>
  44 #include <linux/workqueue.h>
  45 #include <linux/net.h>
  46 #include <linux/ktime.h>
  47
  48 #include <linux/sunrpc/clnt.h>
  49 #include <linux/sunrpc/metrics.h>
  50 #include <linux/sunrpc/bc_xprt.h>
  51
  52 #include "sunrpc.h"
  53
  54 /*
  55  * Local variables
  56  */
  57
  58 #ifdef RPC_DEBUG
  59 # define RPCDBG_FACILITY        RPCDBG_XPRT
  60 #endif
  61
  62 /*
  63  * Local functions
  64  */
  65 static void     xprt_request_init(struct rpc_task *, struct rpc_xprt *);
  66 static void     xprt_connect_status(struct rpc_task *task);
  67 static int      __xprt_get_cong(struct rpc_xprt *, struct rpc_task *);
  68
  69 static DEFINE_SPINLOCK(xprt_list_lock);
  70 static LIST_HEAD(xprt_list);
  71
  72 /*
  73  * The transport code maintains an estimate on the maximum number of out-
  74  * standing RPC requests, using a smoothed version of the congestion
  75  * avoidance implemented in 44BSD. This is basically the Van Jacobson
  76  * congestion algorithm: If a retransmit occurs, the congestion window is
  77  * halved; otherwise, it is incremented by 1/cwnd when
  78  *
  79  *      -       a reply is received and
  80  *      -       a full number of requests are outstanding and
  81  *      -       the congestion window hasn't been updated recently.
  82  */
  83 #define RPC_CWNDSHIFT           (8U)
  84 #define RPC_CWNDSCALE           (1U << RPC_CWNDSHIFT)
  85 #define RPC_INITCWND            RPC_CWNDSCALE
  86 #define RPC_MAXCWND(xprt)       ((xprt)->max_reqs << RPC_CWNDSHIFT)
  87
  88 #define RPCXPRT_CONGESTED(xprt) ((xprt)->cong >= (xprt)->cwnd)
  89
  90 /**
  91  * xprt_register_transport - register a transport implementation
  92  * @transport: transport to register
  93  *
  94  * If a transport implementation is loaded as a kernel module, it can
  95  * call this interface to make itself known to the RPC client.
  96  *
  97  * Returns:
  98  * 0:           transport successfully registered
  99  * -EEXIST:     transport already registered
 100  * -EINVAL:     transport module being unloaded
 101  */
 102 int xprt_register_transport(struct xprt_class *transport)
 103 {
 104         struct xprt_class *t;
 105         int result;
 106
 107         result = -EEXIST;
 108         spin_lock(&xprt_list_lock);
 109         list_for_each_entry(t, &xprt_list, list) {
 110                 /* don't register the same transport class twice */
 111                 if (t->ident == transport->ident)
 112                         goto out;
 113         }
 114
 115         list_add_tail(&transport->list, &xprt_list);
 116         printk(KERN_INFO "RPC: Registered %s transport module.\n",
 117                transport->name);
 118         result = 0;
 119
 120 out:
 121         spin_unlock(&xprt_list_lock);
 122         return result;
 123 }
 124 EXPORT_SYMBOL_GPL(xprt_register_transport);
 125
 126 /**
 127  * xprt_unregister_transport - unregister a transport implementation
 128  * @transport: transport to unregister
 129  *
 130  * Returns:
 131  * 0:           transport successfully unregistered
 132  * -ENOENT:     transport never registered
 133  */
 134 int xprt_unregister_transport(struct xprt_class *transport)
 135 {
 136         struct xprt_class *t;
 137         int result;
 138
 139         result = 0;
 140         spin_lock(&xprt_list_lock);
 141         list_for_each_entry(t, &xprt_list, list) {
 142                 if (t == transport) {
 143                         printk(KERN_INFO
 144                                 "RPC: Unregistered %s transport module.\n",
 145                                 transport->name);
 146                         list_del_init(&transport->list);
 147                         goto out;
 148                 }
 149         }
 150         result = -ENOENT;
 151
 152 out:
 153         spin_unlock(&xprt_list_lock);
 154         return result;
 155 }
 156 EXPORT_SYMBOL_GPL(xprt_unregister_transport);
 157
 158 /**
 159  * xprt_load_transport - load a transport implementation
 160  * @transport_name: transport to load
 161  *
 162  * Returns:
 163  * 0:           transport successfully loaded
 164  * -ENOENT:     transport module not available
 165  */
 166 int xprt_load_transport(const char *transport_name)
 167 {
 168         struct xprt_class *t;
 169         int result;
 170
 171         result = 0;
 172         spin_lock(&xprt_list_lock);
 173         list_for_each_entry(t, &xprt_list, list) {
 174                 if (strcmp(t->name, transport_name) == 0) {
 175                         spin_unlock(&xprt_list_lock);
 176                         goto out;
 177                 }
 178         }
 179         spin_unlock(&xprt_list_lock);
 180         result = request_module("xprt%s", transport_name);
 181 out:
 182         return result;
 183 }
 184 EXPORT_SYMBOL_GPL(xprt_load_transport);
 185
 186 /**
 187  * xprt_reserve_xprt - serialize write access to transports
 188  * @task: task that is requesting access to the transport
 189  *
 190  * This prevents mixing the payload of separate requests, and prevents
 191  * transport connects from colliding with writes.  No congestion control
 192  * is provided.
 193  */
 194 int xprt_reserve_xprt(struct rpc_task *task)
 195 {
 196         struct rpc_rqst *req = task->tk_rqstp;
 197         struct rpc_xprt *xprt = req->rq_xprt;
 198
 199         if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
 200                 if (task == xprt->snd_task)
 201                         return 1;
 202                 if (task == NULL)
 203                         return 0;
 204                 goto out_sleep;
 205         }
 206         xprt->snd_task = task;
 207         if (req) {
 208                 req->rq_bytes_sent = 0;
 209                 req->rq_ntrans++;
 210         }
 211         return 1;
 212
 213 out_sleep:
 214         dprintk("RPC: %5u failed to lock transport %p\n",
 215                         task->tk_pid, xprt);
 216         task->tk_timeout = 0;
 217         task->tk_status = -EAGAIN;
 218         if (req && req->rq_ntrans)
 219                 rpc_sleep_on(&xprt->resend, task, NULL);
 220         else
 221                 rpc_sleep_on(&xprt->sending, task, NULL);
 222         return 0;
 223 }
 224 EXPORT_SYMBOL_GPL(xprt_reserve_xprt);
 225
 226 static void xprt_clear_locked(struct rpc_xprt *xprt)
 227 {
 228         xprt->snd_task = NULL;
 229         if (!test_bit(XPRT_CLOSE_WAIT, &xprt->state) || xprt->shutdown) {
 230                 smp_mb__before_clear_bit();
 231                 clear_bit(XPRT_LOCKED, &xprt->state);
 232                 smp_mb__after_clear_bit();
 233         } else
 234                 queue_work(rpciod_workqueue, &xprt->task_cleanup);
 235 }
 236
 237 /*
 238  * xprt_reserve_xprt_cong - serialize write access to transports
 239  * @task: task that is requesting access to the transport
 240  *
 241  * Same as xprt_reserve_xprt, but Van Jacobson congestion control is
 242  * integrated into the decision of whether a request is allowed to be
 243  * woken up and given access to the transport.
 244  */
 245 int xprt_reserve_xprt_cong(struct rpc_task *task)
 246 {
 247         struct rpc_xprt *xprt = task->tk_xprt;
 248         struct rpc_rqst *req = task->tk_rqstp;
 249
 250         if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
 251                 if (task == xprt->snd_task)
 252                         return 1;
 253                 goto out_sleep;
 254         }
 255         if (__xprt_get_cong(xprt, task)) {
 256                 xprt->snd_task = task;
 257                 if (req) {
 258                         req->rq_bytes_sent = 0;
 259                         req->rq_ntrans++;
 260                 }
 261                 return 1;
 262         }
 263         xprt_clear_locked(xprt);
 264 out_sleep:
 265         dprintk("RPC: %5u failed to lock transport %p\n", task->tk_pid, xprt);
 266         task->tk_timeout = 0;
 267         task->tk_status = -EAGAIN;
 268         if (req && req->rq_ntrans)
 269                 rpc_sleep_on(&xprt->resend, task, NULL);
 270         else
 271                 rpc_sleep_on(&xprt->sending, task, NULL);
 272         return 0;
 273 }
 274 EXPORT_SYMBOL_GPL(xprt_reserve_xprt_cong);
 275
 276 static inline int xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
 277 {
 278         int retval;
 279
 280         spin_lock_bh(&xprt->transport_lock);
 281         retval = xprt->ops->reserve_xprt(task);
 282         spin_unlock_bh(&xprt->transport_lock);
 283         return retval;
 284 }
 285
 286 static void __xprt_lock_write_next(struct rpc_xprt *xprt)
 287 {
 288         struct rpc_task *task;
 289         struct rpc_rqst *req;
 290
 291         if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
 292                 return;
 293
 294         task = rpc_wake_up_next(&xprt->resend);
 295         if (!task) {
 296                 task = rpc_wake_up_next(&xprt->sending);
 297                 if (!task)
 298                         goto out_unlock;
 299         }
 300
 301         req = task->tk_rqstp;
 302         xprt->snd_task = task;
 303         if (req) {
 304                 req->rq_bytes_sent = 0;
 305                 req->rq_ntrans++;
 306         }
 307         return;
 308
 309 out_unlock:
 310         xprt_clear_locked(xprt);
 311 }
 312
 313 static void __xprt_lock_write_next_cong(struct rpc_xprt *xprt)
 314 {
 315         struct rpc_task *task;
 316
 317         if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
 318                 return;
 319         if (RPCXPRT_CONGESTED(xprt))
 320                 goto out_unlock;
 321         task = rpc_wake_up_next(&xprt->resend);
 322         if (!task) {
 323                 task = rpc_wake_up_next(&xprt->sending);
 324                 if (!task)
 325                         goto out_unlock;
 326         }
 327         if (__xprt_get_cong(xprt, task)) {
 328                 struct rpc_rqst *req = task->tk_rqstp;
 329                 xprt->snd_task = task;
 330                 if (req) {
 331                         req->rq_bytes_sent = 0;
 332                         req->rq_ntrans++;
 333                 }
 334                 return;
 335         }
 336 out_unlock:
 337         xprt_clear_locked(xprt);
 338 }
 339
 340 /**
 341  * xprt_release_xprt - allow other requests to use a transport
 342  * @xprt: transport with other tasks potentially waiting
 343  * @task: task that is releasing access to the transport
 344  *
 345  * Note that "task" can be NULL.  No congestion control is provided.
 346  */
 347 void xprt_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
 348 {
 349         if (xprt->snd_task == task) {
 350                 xprt_clear_locked(xprt);
 351                 __xprt_lock_write_next(xprt);
 352         }
 353 }
 354 EXPORT_SYMBOL_GPL(xprt_release_xprt);
 355
 356 /**
 357  * xprt_release_xprt_cong - allow other requests to use a transport
 358  * @xprt: transport with other tasks potentially waiting
 359  * @task: task that is releasing access to the transport
 360  *
 361  * Note that "task" can be NULL.  Another task is awoken to use the
 362  * transport if the transport's congestion window allows it.
 363  */
 364 void xprt_release_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
 365 {
 366         if (xprt->snd_task == task) {
 367                 xprt_clear_locked(xprt);
 368                 __xprt_lock_write_next_cong(xprt);
 369         }
 370 }
 371 EXPORT_SYMBOL_GPL(xprt_release_xprt_cong);
 372
 373 static inline void xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
 374 {
 375         spin_lock_bh(&xprt->transport_lock);
 376         xprt->ops->release_xprt(xprt, task);
 377         spin_unlock_bh(&xprt->transport_lock);
 378 }
 379
 380 /*
 381  * Van Jacobson congestion avoidance. Check if the congestion window
 382  * overflowed. Put the task to sleep if this is the case.
 383  */
 384 static int
 385 __xprt_get_cong(struct rpc_xprt *xprt, struct rpc_task *task)
 386 {
 387         struct rpc_rqst *req = task->tk_rqstp;
 388
 389         if (req->rq_cong)
 390                 return 1;
 391         dprintk("RPC: %5u xprt_cwnd_limited cong = %lu cwnd = %lu\n",
 392                         task->tk_pid, xprt->cong, xprt->cwnd);
 393         if (RPCXPRT_CONGESTED(xprt))
 394                 return 0;
 395         req->rq_cong = 1;
 396         xprt->cong += RPC_CWNDSCALE;
 397         return 1;
 398 }
 399
 400 /*
 401  * Adjust the congestion window, and wake up the next task
 402  * that has been sleeping due to congestion
 403  */
 404 static void
 405 __xprt_put_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
 406 {
 407         if (!req->rq_cong)
 408                 return;
 409         req->rq_cong = 0;
 410         xprt->cong -= RPC_CWNDSCALE;
 411         __xprt_lock_write_next_cong(xprt);
 412 }
 413
 414 /**
 415  * xprt_release_rqst_cong - housekeeping when request is complete
 416  * @task: RPC request that recently completed
 417  *
 418  * Useful for transports that require congestion control.
 419  */
 420 void xprt_release_rqst_cong(struct rpc_task *task)
 421 {
 422         __xprt_put_cong(task->tk_xprt, task->tk_rqstp);
 423 }
 424 EXPORT_SYMBOL_GPL(xprt_release_rqst_cong);
 425
 426 /**
 427  * xprt_adjust_cwnd - adjust transport congestion window
 428  * @task: recently completed RPC request used to adjust window
 429  * @result: result code of completed RPC request
 430  *
 431  * We use a time-smoothed congestion estimator to avoid heavy oscillation.
 432  */
 433 void xprt_adjust_cwnd(struct rpc_task *task, int result)
 434 {
 435         struct rpc_rqst *req = task->tk_rqstp;
 436         struct rpc_xprt *xprt = task->tk_xprt;
 437         unsigned long cwnd = xprt->cwnd;
 438
 439         if (result >= 0 && cwnd <= xprt->cong) {
 440                 /* The (cwnd >> 1) term makes sure
 441                  * the result gets rounded properly. */
 442                 cwnd += (RPC_CWNDSCALE * RPC_CWNDSCALE + (cwnd >> 1)) / cwnd;
 443                 if (cwnd > RPC_MAXCWND(xprt))
 444                         cwnd = RPC_MAXCWND(xprt);
 445                 __xprt_lock_write_next_cong(xprt);
 446         } else if (result == -ETIMEDOUT) {
 447                 cwnd >>= 1;
 448                 if (cwnd < RPC_CWNDSCALE)
 449                         cwnd = RPC_CWNDSCALE;
 450         }
 451         dprintk("RPC:       cong %ld, cwnd was %ld, now %ld\n",
 452                         xprt->cong, xprt->cwnd, cwnd);
 453         xprt->cwnd = cwnd;
 454         __xprt_put_cong(xprt, req);
 455 }
 456 EXPORT_SYMBOL_GPL(xprt_adjust_cwnd);
 457
 458 /**
 459  * xprt_wake_pending_tasks - wake all tasks on a transport's pending queue
 460  * @xprt: transport with waiting tasks
 461  * @status: result code to plant in each task before waking it
 462  *
 463  */
 464 void xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status)
 465 {
 466         if (status < 0)
 467                 rpc_wake_up_status(&xprt->pending, status);
 468         else
 469                 rpc_wake_up(&xprt->pending);
 470 }
 471 EXPORT_SYMBOL_GPL(xprt_wake_pending_tasks);
 472
 473 /**
 474  * xprt_wait_for_buffer_space - wait for transport output buffer to clear
 475  * @task: task to be put to sleep
 476  * @action: function pointer to be executed after wait
 477  */
 478 void xprt_wait_for_buffer_space(struct rpc_task *task, rpc_action action)
 479 {
 480         struct rpc_rqst *req = task->tk_rqstp;
 481         struct rpc_xprt *xprt = req->rq_xprt;
 482
 483         task->tk_timeout = req->rq_timeout;
 484         rpc_sleep_on(&xprt->pending, task, action);
 485 }
 486 EXPORT_SYMBOL_GPL(xprt_wait_for_buffer_space);
 487
 488 /**
 489  * xprt_write_space - wake the task waiting for transport output buffer space
 490  * @xprt: transport with waiting tasks
 491  *
 492  * Can be called in a soft IRQ context, so xprt_write_space never sleeps.
 493  */
 494 void xprt_write_space(struct rpc_xprt *xprt)
 495 {
 496         if (unlikely(xprt->shutdown))
 497                 return;
 498
 499         spin_lock_bh(&xprt->transport_lock);
 500         if (xprt->snd_task) {
 501                 dprintk("RPC:       write space: waking waiting task on "
 502                                 "xprt %p\n", xprt);
 503                 rpc_wake_up_queued_task(&xprt->pending, xprt->snd_task);
 504         }
 505         spin_unlock_bh(&xprt->transport_lock);
 506 }
 507 EXPORT_SYMBOL_GPL(xprt_write_space);
 508
 509 /**
 510  * xprt_set_retrans_timeout_def - set a request's retransmit timeout
 511  * @task: task whose timeout is to be set
 512  *
 513  * Set a request's retransmit timeout based on the transport's
 514  * default timeout parameters.  Used by transports that don't adjust
 515  * the retransmit timeout based on round-trip time estimation.
 516  */
 517 void xprt_set_retrans_timeout_def(struct rpc_task *task)
 518 {
 519         task->tk_timeout = task->tk_rqstp->rq_timeout;
 520 }
 521 EXPORT_SYMBOL_GPL(xprt_set_retrans_timeout_def);
 522
 523 /*
 524  * xprt_set_retrans_timeout_rtt - set a request's retransmit timeout
 525  * @task: task whose timeout is to be set
 526  *
 527  * Set a request's retransmit timeout using the RTT estimator.
 528  */
 529 void xprt_set_retrans_timeout_rtt(struct rpc_task *task)
 530 {
 531         int timer = task->tk_msg.rpc_proc->p_timer;
 532         struct rpc_clnt *clnt = task->tk_client;
 533         struct rpc_rtt *rtt = clnt->cl_rtt;
 534         struct rpc_rqst *req = task->tk_rqstp;
 535         unsigned long max_timeout = clnt->cl_timeout->to_maxval;
 536
 537         task->tk_timeout = rpc_calc_rto(rtt, timer);
 538         task->tk_timeout <<= rpc_ntimeo(rtt, timer) + req->rq_retries;
 539         if (task->tk_timeout > max_timeout || task->tk_timeout == 0)
 540                 task->tk_timeout = max_timeout;
 541 }
 542 EXPORT_SYMBOL_GPL(xprt_set_retrans_timeout_rtt);
 543
 544 static void xprt_reset_majortimeo(struct rpc_rqst *req)
 545 {
 546         const struct rpc_timeout *to = req->rq_task->tk_client->cl_timeout;
 547
 548         req->rq_majortimeo = req->rq_timeout;
 549         if (to->to_exponential)
 550                 req->rq_majortimeo <<= to->to_retries;
 551         else
 552                 req->rq_majortimeo += to->to_increment * to->to_retries;
 553         if (req->rq_majortimeo > to->to_maxval || req->rq_majortimeo == 0)
 554                 req->rq_majortimeo = to->to_maxval;
 555         req->rq_majortimeo += jiffies;
 556 }
 557
 558 /**
 559  * xprt_adjust_timeout - adjust timeout values for next retransmit
 560  * @req: RPC request containing parameters to use for the adjustment
 561  *
 562  */
 563 int xprt_adjust_timeout(struct rpc_rqst *req)
 564 {
 565         struct rpc_xprt *xprt = req->rq_xprt;
 566         const struct rpc_timeout *to = req->rq_task->tk_client->cl_timeout;
 567         int status = 0;
 568
 569         if (time_before(jiffies, req->rq_majortimeo)) {
 570                 if (to->to_exponential)
 571                         req->rq_timeout <<= 1;
 572                 else
 573                         req->rq_timeout += to->to_increment;
 574                 if (to->to_maxval && req->rq_timeout >= to->to_maxval)
 575                         req->rq_timeout = to->to_maxval;
 576                 req->rq_retries++;
 577         } else {
 578                 req->rq_timeout = to->to_initval;
 579                 req->rq_retries = 0;
 580                 xprt_reset_majortimeo(req);
 581                 /* Reset the RTT counters == "slow start" */
 582                 spin_lock_bh(&xprt->transport_lock);
 583                 rpc_init_rtt(req->rq_task->tk_client->cl_rtt, to->to_initval);
 584                 spin_unlock_bh(&xprt->transport_lock);
 585                 status = -ETIMEDOUT;
 586         }
 587
 588         if (req->rq_timeout == 0) {
 589                 printk(KERN_WARNING "xprt_adjust_timeout: rq_timeout = 0!\n");
 590                 req->rq_timeout = 5 * HZ;
 591         }
 592         return status;
 593 }
 594
 595 static void xprt_autoclose(struct work_struct *work)
 596 {
 597         struct rpc_xprt *xprt =
 598                 container_of(work, struct rpc_xprt, task_cleanup);
 599
 600         xprt->ops->close(xprt);
 601         clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
 602         xprt_release_write(xprt, NULL);
 603 }
 604
 605 /**
 606  * xprt_disconnect_done - mark a transport as disconnected
 607  * @xprt: transport to flag for disconnect
 608  *
 609  */
 610 void xprt_disconnect_done(struct rpc_xprt *xprt)
 611 {
 612         dprintk("RPC:       disconnected transport %p\n", xprt);
 613         spin_lock_bh(&xprt->transport_lock);
 614         xprt_clear_connected(xprt);
 615         xprt_wake_pending_tasks(xprt, -EAGAIN);
 616         spin_unlock_bh(&xprt->transport_lock);
 617 }
 618 EXPORT_SYMBOL_GPL(xprt_disconnect_done);
 619
 620 /**
 621  * xprt_force_disconnect - force a transport to disconnect
 622  * @xprt: transport to disconnect
 623  *
 624  */
 625 void xprt_force_disconnect(struct rpc_xprt *xprt)
 626 {
 627         /* Don't race with the test_bit() in xprt_clear_locked() */
 628         spin_lock_bh(&xprt->transport_lock);
 629         set_bit(XPRT_CLOSE_WAIT, &xprt->state);
 630         /* Try to schedule an autoclose RPC call */
 631         if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
 632                 queue_work(rpciod_workqueue, &xprt->task_cleanup);
 633         xprt_wake_pending_tasks(xprt, -EAGAIN);
 634         spin_unlock_bh(&xprt->transport_lock);
 635 }
 636
 637 /**
 638  * xprt_conditional_disconnect - force a transport to disconnect
 639  * @xprt: transport to disconnect
 640  * @cookie: 'connection cookie'
 641  *
 642  * This attempts to break the connection if and only if 'cookie' matches
 643  * the current transport 'connection cookie'. It ensures that we don't
 644  * try to break the connection more than once when we need to retransmit
 645  * a batch of RPC requests.
 646  *
 647  */
 648 void xprt_conditional_disconnect(struct rpc_xprt *xprt, unsigned int cookie)
 649 {
 650         /* Don't race with the test_bit() in xprt_clear_locked() */
 651         spin_lock_bh(&xprt->transport_lock);
 652         if (cookie != xprt->connect_cookie)
 653                 goto out;
 654         if (test_bit(XPRT_CLOSING, &xprt->state) || !xprt_connected(xprt))
 655                 goto out;
 656         set_bit(XPRT_CLOSE_WAIT, &xprt->state);
 657         /* Try to schedule an autoclose RPC call */
 658         if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
 659                 queue_work(rpciod_workqueue, &xprt->task_cleanup);
 660         xprt_wake_pending_tasks(xprt, -EAGAIN);
 661 out:
 662         spin_unlock_bh(&xprt->transport_lock);
 663 }
 664
 665 static void
 666 xprt_init_autodisconnect(unsigned long data)
 667 {
 668         struct rpc_xprt *xprt = (struct rpc_xprt *)data;
 669
 670         spin_lock(&xprt->transport_lock);
 671         if (!list_empty(&xprt->recv) || xprt->shutdown)
 672                 goto out_abort;
 673         if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
 674                 goto out_abort;
 675         spin_unlock(&xprt->transport_lock);
 676         set_bit(XPRT_CONNECTION_CLOSE, &xprt->state);
 677         queue_work(rpciod_workqueue, &xprt->task_cleanup);
 678         return;
 679 out_abort:
 680         spin_unlock(&xprt->transport_lock);
 681 }
 682
 683 /**
 684  * xprt_connect - schedule a transport connect operation
 685  * @task: RPC task that is requesting the connect
 686  *
 687  */
 688 void xprt_connect(struct rpc_task *task)
 689 {
 690         struct rpc_xprt *xprt = task->tk_xprt;
 691
 692         dprintk("RPC: %5u xprt_connect xprt %p %s connected\n", task->tk_pid,
 693                         xprt, (xprt_connected(xprt) ? "is" : "is not"));
 694
 695         if (!xprt_bound(xprt)) {
 696                 task->tk_status = -EAGAIN;
 697                 return;
 698         }
 699         if (!xprt_lock_write(xprt, task))
 700                 return;
 701
 702         if (test_and_clear_bit(XPRT_CLOSE_WAIT, &xprt->state))
 703                 xprt->ops->close(xprt);
 704
 705         if (xprt_connected(xprt))
 706                 xprt_release_write(xprt, task);
 707         else {
 708                 if (task->tk_rqstp)
 709                         task->tk_rqstp->rq_bytes_sent = 0;
 710
 711                 task->tk_timeout = task->tk_rqstp->rq_timeout;
 712                 rpc_sleep_on(&xprt->pending, task, xprt_connect_status);
 713
 714                 if (test_bit(XPRT_CLOSING, &xprt->state))
 715                         return;
 716                 if (xprt_test_and_set_connecting(xprt))
 717                         return;
 718                 xprt->stat.connect_start = jiffies;
 719                 xprt->ops->connect(task);
 720         }
 721 }
 722
 723 static void xprt_connect_status(struct rpc_task *task)
 724 {
 725         struct rpc_xprt *xprt = task->tk_xprt;
 726
 727         if (task->tk_status == 0) {
 728                 xprt->stat.connect_count++;
 729                 xprt->stat.connect_time += (long)jiffies - xprt->stat.connect_start;
 730                 dprintk("RPC: %5u xprt_connect_status: connection established\n",
 731                                 task->tk_pid);
 732                 return;
 733         }
 734
 735         switch (task->tk_status) {
 736         case -EAGAIN:
 737                 dprintk("RPC: %5u xprt_connect_status: retrying\n", task->tk_pid);
 738                 break;
 739         case -ETIMEDOUT:
 740                 dprintk("RPC: %5u xprt_connect_status: connect attempt timed "
 741                                 "out\n", task->tk_pid);
 742                 break;
 743         default:
 744                 dprintk("RPC: %5u xprt_connect_status: error %d connecting to "
 745                                 "server %s\n", task->tk_pid, -task->tk_status,
 746                                 task->tk_client->cl_server);
 747                 xprt_release_write(xprt, task);
 748                 task->tk_status = -EIO;
 749         }
 750 }
 751
 752 /**
 753  * xprt_lookup_rqst - find an RPC request corresponding to an XID
 754  * @xprt: transport on which the original request was transmitted
 755  * @xid: RPC XID of incoming reply
 756  *
 757  */
 758 struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, __be32 xid)
 759 {
 760         struct list_head *pos;
 761
 762         list_for_each(pos, &xprt->recv) {
 763                 struct rpc_rqst *entry = list_entry(pos, struct rpc_rqst, rq_list);
 764                 if (entry->rq_xid == xid)
 765                         return entry;
 766         }
 767
 768         dprintk("RPC:       xprt_lookup_rqst did not find xid %08x\n",
 769                         ntohl(xid));
 770         xprt->stat.bad_xids++;
 771         return NULL;
 772 }
 773 EXPORT_SYMBOL_GPL(xprt_lookup_rqst);
 774
 775 static void xprt_update_rtt(struct rpc_task *task)
 776 {
 777         struct rpc_rqst *req = task->tk_rqstp;
 778         struct rpc_rtt *rtt = task->tk_client->cl_rtt;
 779         unsigned timer = task->tk_msg.rpc_proc->p_timer;
 780         long m = usecs_to_jiffies(ktime_to_us(req->rq_rtt));
 781
 782         if (timer) {
 783                 if (req->rq_ntrans == 1)
 784                         rpc_update_rtt(rtt, timer, m);
 785                 rpc_set_timeo(rtt, timer, req->rq_ntrans - 1);
 786         }
 787 }
 788
 789 /**
 790  * xprt_complete_rqst - called when reply processing is complete
 791  * @task: RPC request that recently completed
 792  * @copied: actual number of bytes received from the transport
 793  *
 794  * Caller holds transport lock.
 795  */
 796 void xprt_complete_rqst(struct rpc_task *task, int copied)
 797 {
 798         struct rpc_rqst *req = task->tk_rqstp;
 799         struct rpc_xprt *xprt = req->rq_xprt;
 800
 801         dprintk("RPC: %5u xid %08x complete (%d bytes received)\n",
 802                         task->tk_pid, ntohl(req->rq_xid), copied);
 803
 804         xprt->stat.recvs++;
 805         req->rq_rtt = ktime_sub(ktime_get(), req->rq_xtime);
 806         if (xprt->ops->timer != NULL)
 807                 xprt_update_rtt(task);
 808
 809         list_del_init(&req->rq_list);
 810         req->rq_private_buf.len = copied;
 811         /* Ensure all writes are done before we update */
 812         /* req->rq_reply_bytes_recvd */
 813         smp_wmb();
 814         req->rq_reply_bytes_recvd = copied;
 815         rpc_wake_up_queued_task(&xprt->pending, task);
 816 }
 817 EXPORT_SYMBOL_GPL(xprt_complete_rqst);
 818
 819 static void xprt_timer(struct rpc_task *task)
 820 {
 821         struct rpc_rqst *req = task->tk_rqstp;
 822         struct rpc_xprt *xprt = req->rq_xprt;
 823
 824         if (task->tk_status != -ETIMEDOUT)
 825                 return;
 826         dprintk("RPC: %5u xprt_timer\n", task->tk_pid);
 827
 828         spin_lock_bh(&xprt->transport_lock);
 829         if (!req->rq_reply_bytes_recvd) {
 830                 if (xprt->ops->timer)
 831                         xprt->ops->timer(task);
 832         } else
 833                 task->tk_status = 0;
 834         spin_unlock_bh(&xprt->transport_lock);
 835 }
 836
 837 static inline int xprt_has_timer(struct rpc_xprt *xprt)
 838 {
 839         return xprt->idle_timeout != 0;
 840 }
 841
 842 /**
 843  * xprt_prepare_transmit - reserve the transport before sending a request
 844  * @task: RPC task about to send a request
 845  *
 846  */
 847 int xprt_prepare_transmit(struct rpc_task *task)
 848 {
 849         struct rpc_rqst *req = task->tk_rqstp;
 850         struct rpc_xprt *xprt = req->rq_xprt;
 851         int err = 0;
 852
 853         dprintk("RPC: %5u xprt_prepare_transmit\n", task->tk_pid);
 854
 855         spin_lock_bh(&xprt->transport_lock);
 856         if (req->rq_reply_bytes_recvd && !req->rq_bytes_sent) {
 857                 err = req->rq_reply_bytes_recvd;
 858                 goto out_unlock;
 859         }
 860         if (!xprt->ops->reserve_xprt(task))
 861                 err = -EAGAIN;
 862 out_unlock:
 863         spin_unlock_bh(&xprt->transport_lock);
 864         return err;
 865 }
 866
 867 void xprt_end_transmit(struct rpc_task *task)
 868 {
 869         xprt_release_write(task->tk_rqstp->rq_xprt, task);
 870 }
 871
 872 /**
 873  * xprt_transmit - send an RPC request on a transport
 874  * @task: controlling RPC task
 875  *
 876  * We have to copy the iovec because sendmsg fiddles with its contents.
 877  */
 878 void xprt_transmit(struct rpc_task *task)
 879 {
 880         struct rpc_rqst *req = task->tk_rqstp;
 881         struct rpc_xprt *xprt = req->rq_xprt;
 882         int status;
 883
 884         dprintk("RPC: %5u xprt_transmit(%u)\n", task->tk_pid, req->rq_slen);
 885
 886         if (!req->rq_reply_bytes_recvd) {
 887                 if (list_empty(&req->rq_list) && rpc_reply_expected(task)) {
 888                         /*
 889                          * Add to the list only if we're expecting a reply
 890                          */
 891                         spin_lock_bh(&xprt->transport_lock);
 892                         /* Update the softirq receive buffer */
 893                         memcpy(&req->rq_private_buf, &req->rq_rcv_buf,
 894                                         sizeof(req->rq_private_buf));
 895                         /* Add request to the receive list */
 896                         list_add_tail(&req->rq_list, &xprt->recv);
 897                         spin_unlock_bh(&xprt->transport_lock);
 898                         xprt_reset_majortimeo(req);
 899                         /* Turn off autodisconnect */
 900                         del_singleshot_timer_sync(&xprt->timer);
 901                 }
 902         } else if (!req->rq_bytes_sent)
 903                 return;
 904
 905         req->rq_connect_cookie = xprt->connect_cookie;
 906         req->rq_xtime = ktime_get();
 907         status = xprt->ops->send_request(task);
 908         if (status != 0) {
 909                 task->tk_status = status;
 910                 return;
 911         }
 912
 913         dprintk("RPC: %5u xmit complete\n", task->tk_pid);
 914         spin_lock_bh(&xprt->transport_lock);
 915
 916         xprt->ops->set_retrans_timeout(task);
 917
 918         xprt->stat.sends++;
 919         xprt->stat.req_u += xprt->stat.sends - xprt->stat.recvs;
 920         xprt->stat.bklog_u += xprt->backlog.qlen;
 921
 922         /* Don't race with disconnect */
 923         if (!xprt_connected(xprt))
 924                 task->tk_status = -ENOTCONN;
 925         else if (!req->rq_reply_bytes_recvd && rpc_reply_expected(task)) {
 926                 /*
 927                  * Sleep on the pending queue since
 928                  * we're expecting a reply.
 929                  */
 930                 rpc_sleep_on(&xprt->pending, task, xprt_timer);
 931         }
 932         spin_unlock_bh(&xprt->transport_lock);
 933 }
 934
 935 static void xprt_alloc_slot(struct rpc_task *task)
 936 {
 937         struct rpc_xprt *xprt = task->tk_xprt;
 938
 939         task->tk_status = 0;
 940         if (task->tk_rqstp)
 941                 return;
 942         if (!list_empty(&xprt->free)) {
 943                 struct rpc_rqst *req = list_entry(xprt->free.next, struct rpc_rqst, rq_list);
 944                 list_del_init(&req->rq_list);
 945                 task->tk_rqstp = req;
 946                 xprt_request_init(task, xprt);
 947                 return;
 948         }
 949         dprintk("RPC:       waiting for request slot\n");
 950         task->tk_status = -EAGAIN;
 951         task->tk_timeout = 0;
 952         rpc_sleep_on(&xprt->backlog, task, NULL);
 953 }
 954
 955 static void xprt_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
 956 {
 957         memset(req, 0, sizeof(*req));   /* mark unused */
 958
 959         spin_lock(&xprt->reserve_lock);
 960         list_add(&req->rq_list, &xprt->free);
 961         rpc_wake_up_next(&xprt->backlog);
 962         spin_unlock(&xprt->reserve_lock);
 963 }
 964
 965 /**
 966  * xprt_reserve - allocate an RPC request slot
 967  * @task: RPC task requesting a slot allocation
 968  *
 969  * If no more slots are available, place the task on the transport's
 970  * backlog queue.
 971  */
 972 void xprt_reserve(struct rpc_task *task)
 973 {
 974         struct rpc_xprt *xprt = task->tk_xprt;
 975
 976         task->tk_status = -EIO;
 977         spin_lock(&xprt->reserve_lock);
 978         xprt_alloc_slot(task);
 979         spin_unlock(&xprt->reserve_lock);
 980 }
 981
 982 static inline __be32 xprt_alloc_xid(struct rpc_xprt *xprt)
 983 {
 984         return (__force __be32)xprt->xid++;
 985 }
 986
 987 static inline void xprt_init_xid(struct rpc_xprt *xprt)
 988 {
 989         xprt->xid = net_random();
 990 }
 991
 992 static void xprt_request_init(struct rpc_task *task, struct rpc_xprt *xprt)
 993 {
 994         struct rpc_rqst *req = task->tk_rqstp;
 995
 996         req->rq_timeout = task->tk_client->cl_timeout->to_initval;
 997         req->rq_task    = task;
 998         req->rq_xprt    = xprt;
 999         req->rq_buffer  = NULL;
1000         req->rq_xid     = xprt_alloc_xid(xprt);
1001         req->rq_release_snd_buf = NULL;
1002         xprt_reset_majortimeo(req);
1003         dprintk("RPC: %5u reserved req %p xid %08x\n", task->tk_pid,
1004                         req, ntohl(req->rq_xid));
1005 }
1006
1007 /**
1008  * xprt_release - release an RPC request slot
1009  * @task: task which is finished with the slot
1010  *
1011  */
1012 void xprt_release(struct rpc_task *task)
1013 {
1014         struct rpc_xprt *xprt;
1015         struct rpc_rqst *req;
1016
1017         if (!(req = task->tk_rqstp))
1018                 return;
1019
1020         xprt = req->rq_xprt;
1021         rpc_count_iostats(task);
1022         spin_lock_bh(&xprt->transport_lock);
1023         xprt->ops->release_xprt(xprt, task);
1024         if (xprt->ops->release_request)
1025                 xprt->ops->release_request(task);
1026         if (!list_empty(&req->rq_list))
1027                 list_del(&req->rq_list);
1028         xprt->last_used = jiffies;
1029         if (list_empty(&xprt->recv) && xprt_has_timer(xprt))
1030                 mod_timer(&xprt->timer,
1031                                 xprt->last_used + xprt->idle_timeout);
1032         spin_unlock_bh(&xprt->transport_lock);
1033         if (req->rq_buffer)
1034                 xprt->ops->buf_free(req->rq_buffer);
1035         if (req->rq_cred != NULL)
1036                 put_rpccred(req->rq_cred);
1037         task->tk_rqstp = NULL;
1038         if (req->rq_release_snd_buf)
1039                 req->rq_release_snd_buf(req);
1040
1041         dprintk("RPC: %5u release request %p\n", task->tk_pid, req);
1042         if (likely(!bc_prealloc(req)))
1043                 xprt_free_slot(xprt, req);
1044         else
1045                 xprt_free_bc_request(req);
1046 }
1047
1048 /**
1049  * xprt_create_transport - create an RPC transport
1050  * @args: rpc transport creation arguments
1051  *
1052  */
1053 struct rpc_xprt *xprt_create_transport(struct xprt_create *args)
1054 {
1055         struct rpc_xprt *xprt;
1056         struct rpc_rqst *req;
1057         struct xprt_class *t;
1058
1059         spin_lock(&xprt_list_lock);
1060         list_for_each_entry(t, &xprt_list, list) {
1061                 if (t->ident == args->ident) {
1062                         spin_unlock(&xprt_list_lock);
1063                         goto found;
1064                 }
1065         }
1066         spin_unlock(&xprt_list_lock);
1067         printk(KERN_ERR "RPC: transport (%d) not supported\n", args->ident);
1068         return ERR_PTR(-EIO);
1069
1070 found:
1071         xprt = t->setup(args);
1072         if (IS_ERR(xprt)) {
1073                 dprintk("RPC:       xprt_create_transport: failed, %ld\n",
1074                                 -PTR_ERR(xprt));
1075                 return xprt;
1076         }
1077
1078         kref_init(&xprt->kref);
1079         spin_lock_init(&xprt->transport_lock);
1080         spin_lock_init(&xprt->reserve_lock);
1081
1082         INIT_LIST_HEAD(&xprt->free);
1083         INIT_LIST_HEAD(&xprt->recv);
1084 #if defined(CONFIG_NFS_V4_1)
1085         spin_lock_init(&xprt->bc_pa_lock);
1086         INIT_LIST_HEAD(&xprt->bc_pa_list);
1087 #endif /* CONFIG_NFS_V4_1 */
1088
1089         INIT_WORK(&xprt->task_cleanup, xprt_autoclose);
1090         if (xprt_has_timer(xprt))
1091                 setup_timer(&xprt->timer, xprt_init_autodisconnect,
1092                             (unsigned long)xprt);
1093         else
1094                 init_timer(&xprt->timer);
1095         xprt->last_used = jiffies;
1096         xprt->cwnd = RPC_INITCWND;
1097         xprt->bind_index = 0;
1098
1099         rpc_init_wait_queue(&xprt->binding, "xprt_binding");
1100         rpc_init_wait_queue(&xprt->pending, "xprt_pending");
1101         rpc_init_wait_queue(&xprt->sending, "xprt_sending");
1102         rpc_init_wait_queue(&xprt->resend, "xprt_resend");
1103         rpc_init_priority_wait_queue(&xprt->backlog, "xprt_backlog");
1104
1105         /* initialize free list */
1106         for (req = &xprt->slot[xprt->max_reqs-1]; req >= &xprt->slot[0]; req--)
1107                 list_add(&req->rq_list, &xprt->free);
1108
1109         xprt_init_xid(xprt);
1110
1111         dprintk("RPC:       created transport %p with %u slots\n", xprt,
1112                         xprt->max_reqs);
1113         return xprt;
1114 }
1115
1116 /**
1117  * xprt_destroy - destroy an RPC transport, killing off all requests.
1118  * @kref: kref for the transport to destroy
1119  *
1120  */
1121 static void xprt_destroy(struct kref *kref)
1122 {
1123         struct rpc_xprt *xprt = container_of(kref, struct rpc_xprt, kref);
1124
1125         dprintk("RPC:       destroying transport %p\n", xprt);
1126         xprt->shutdown = 1;
1127         del_timer_sync(&xprt->timer);
1128
1129         rpc_destroy_wait_queue(&xprt->binding);
1130         rpc_destroy_wait_queue(&xprt->pending);
1131         rpc_destroy_wait_queue(&xprt->sending);
1132         rpc_destroy_wait_queue(&xprt->resend);
1133         rpc_destroy_wait_queue(&xprt->backlog);
1134         cancel_work_sync(&xprt->task_cleanup);
1135         /*
1136          * Tear down transport state and free the rpc_xprt
1137          */
1138         xprt->ops->destroy(xprt);
1139 }
1140
1141 /**
1142  * xprt_put - release a reference to an RPC transport.
1143  * @xprt: pointer to the transport
1144  *
1145  */
1146 void xprt_put(struct rpc_xprt *xprt)
1147 {
1148         kref_put(&xprt->kref, xprt_destroy);
1149 }
1150
1151 /**
1152  * xprt_get - return a reference to an RPC transport.
1153  * @xprt: pointer to the transport
1154  *
1155  */
1156 struct rpc_xprt *xprt_get(struct rpc_xprt *xprt)
1157 {
1158         kref_get(&xprt->kref);
1159         return xprt;
1160 }