net/sunrpc/xprt.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  *  linux/net/sunrpc/xprt.c
   4  *
   5  *  This is a generic RPC call interface supporting congestion avoidance,
   6  *  and asynchronous calls.
   7  *
   8  *  The interface works like this:
   9  *
  10  *  -   When a process places a call, it allocates a request slot if
  11  *      one is available. Otherwise, it sleeps on the backlog queue
  12  *      (xprt_reserve).
  13  *  -   Next, the caller puts together the RPC message, stuffs it into
  14  *      the request struct, and calls xprt_transmit().
  15  *  -   xprt_transmit sends the message and installs the caller on the
  16  *      transport's wait list. At the same time, if a reply is expected,
  17  *      it installs a timer that is run after the packet's timeout has
  18  *      expired.
  19  *  -   When a packet arrives, the data_ready handler walks the list of
  20  *      pending requests for that transport. If a matching XID is found, the
  21  *      caller is woken up, and the timer removed.
  22  *  -   When no reply arrives within the timeout interval, the timer is
  23  *      fired by the kernel and runs xprt_timer(). It either adjusts the
  24  *      timeout values (minor timeout) or wakes up the caller with a status
  25  *      of -ETIMEDOUT.
  26  *  -   When the caller receives a notification from RPC that a reply arrived,
  27  *      it should release the RPC slot, and process the reply.
  28  *      If the call timed out, it may choose to retry the operation by
  29  *      adjusting the initial timeout value, and simply calling rpc_call
  30  *      again.
  31  *
  32  *  Support for async RPC is done through a set of RPC-specific scheduling
  33  *  primitives that `transparently' work for processes as well as async
  34  *  tasks that rely on callbacks.
  35  *
  36  *  Copyright (C) 1995-1997, Olaf Kirch <okir@monad.swb.de>
  37  *
  38  *  Transport switch API copyright (C) 2005, Chuck Lever <cel@netapp.com>
  39  */
  40
  41 #include <linux/module.h>
  42
  43 #include <linux/types.h>
  44 #include <linux/interrupt.h>
  45 #include <linux/workqueue.h>
  46 #include <linux/net.h>
  47 #include <linux/ktime.h>
  48
  49 #include <linux/sunrpc/clnt.h>
  50 #include <linux/sunrpc/metrics.h>
  51 #include <linux/sunrpc/bc_xprt.h>
  52 #include <linux/rcupdate.h>
  53 #include <linux/sched/mm.h>
  54
  55 #include <trace/events/sunrpc.h>
  56
  57 #include "sunrpc.h"
  58
  59 /*
  60  * Local variables
  61  */
  62
  63 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
  64 # define RPCDBG_FACILITY        RPCDBG_XPRT
  65 #endif
  66
  67 /*
  68  * Local functions
  69  */
  70 static void      xprt_init(struct rpc_xprt *xprt, struct net *net);
  71 static __be32   xprt_alloc_xid(struct rpc_xprt *xprt);
  72 static void      xprt_destroy(struct rpc_xprt *xprt);
  73
  74 static DEFINE_SPINLOCK(xprt_list_lock);
  75 static LIST_HEAD(xprt_list);
  76
  77 static unsigned long xprt_request_timeout(const struct rpc_rqst *req)
  78 {
  79         unsigned long timeout = jiffies + req->rq_timeout;
  80
  81         if (time_before(timeout, req->rq_majortimeo))
  82                 return timeout;
  83         return req->rq_majortimeo;
  84 }
  85
  86 /**
  87  * xprt_register_transport - register a transport implementation
  88  * @transport: transport to register
  89  *
  90  * If a transport implementation is loaded as a kernel module, it can
  91  * call this interface to make itself known to the RPC client.
  92  *
  93  * Returns:
  94  * 0:           transport successfully registered
  95  * -EEXIST:     transport already registered
  96  * -EINVAL:     transport module being unloaded
  97  */
  98 int xprt_register_transport(struct xprt_class *transport)
  99 {
 100         struct xprt_class *t;
 101         int result;
 102
 103         result = -EEXIST;
 104         spin_lock(&xprt_list_lock);
 105         list_for_each_entry(t, &xprt_list, list) {
 106                 /* don't register the same transport class twice */
 107                 if (t->ident == transport->ident)
 108                         goto out;
 109         }
 110
 111         list_add_tail(&transport->list, &xprt_list);
 112         printk(KERN_INFO "RPC: Registered %s transport module.\n",
 113                transport->name);
 114         result = 0;
 115
 116 out:
 117         spin_unlock(&xprt_list_lock);
 118         return result;
 119 }
 120 EXPORT_SYMBOL_GPL(xprt_register_transport);
 121
 122 /**
 123  * xprt_unregister_transport - unregister a transport implementation
 124  * @transport: transport to unregister
 125  *
 126  * Returns:
 127  * 0:           transport successfully unregistered
 128  * -ENOENT:     transport never registered
 129  */
 130 int xprt_unregister_transport(struct xprt_class *transport)
 131 {
 132         struct xprt_class *t;
 133         int result;
 134
 135         result = 0;
 136         spin_lock(&xprt_list_lock);
 137         list_for_each_entry(t, &xprt_list, list) {
 138                 if (t == transport) {
 139                         printk(KERN_INFO
 140                                 "RPC: Unregistered %s transport module.\n",
 141                                 transport->name);
 142                         list_del_init(&transport->list);
 143                         goto out;
 144                 }
 145         }
 146         result = -ENOENT;
 147
 148 out:
 149         spin_unlock(&xprt_list_lock);
 150         return result;
 151 }
 152 EXPORT_SYMBOL_GPL(xprt_unregister_transport);
 153
 154 /**
 155  * xprt_load_transport - load a transport implementation
 156  * @transport_name: transport to load
 157  *
 158  * Returns:
 159  * 0:           transport successfully loaded
 160  * -ENOENT:     transport module not available
 161  */
 162 int xprt_load_transport(const char *transport_name)
 163 {
 164         struct xprt_class *t;
 165         int result;
 166
 167         result = 0;
 168         spin_lock(&xprt_list_lock);
 169         list_for_each_entry(t, &xprt_list, list) {
 170                 if (strcmp(t->name, transport_name) == 0) {
 171                         spin_unlock(&xprt_list_lock);
 172                         goto out;
 173                 }
 174         }
 175         spin_unlock(&xprt_list_lock);
 176         result = request_module("xprt%s", transport_name);
 177 out:
 178         return result;
 179 }
 180 EXPORT_SYMBOL_GPL(xprt_load_transport);
 181
 182 static void xprt_clear_locked(struct rpc_xprt *xprt)
 183 {
 184         xprt->snd_task = NULL;
 185         if (!test_bit(XPRT_CLOSE_WAIT, &xprt->state)) {
 186                 smp_mb__before_atomic();
 187                 clear_bit(XPRT_LOCKED, &xprt->state);
 188                 smp_mb__after_atomic();
 189         } else
 190                 queue_work(xprtiod_workqueue, &xprt->task_cleanup);
 191 }
 192
 193 /**
 194  * xprt_reserve_xprt - serialize write access to transports
 195  * @task: task that is requesting access to the transport
 196  * @xprt: pointer to the target transport
 197  *
 198  * This prevents mixing the payload of separate requests, and prevents
 199  * transport connects from colliding with writes.  No congestion control
 200  * is provided.
 201  */
 202 int xprt_reserve_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
 203 {
 204         struct rpc_rqst *req = task->tk_rqstp;
 205
 206         if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
 207                 if (task == xprt->snd_task)
 208                         return 1;
 209                 goto out_sleep;
 210         }
 211         if (test_bit(XPRT_WRITE_SPACE, &xprt->state))
 212                 goto out_unlock;
 213         xprt->snd_task = task;
 214
 215         return 1;
 216
 217 out_unlock:
 218         xprt_clear_locked(xprt);
 219 out_sleep:
 220         dprintk("RPC: %5u failed to lock transport %p\n",
 221                         task->tk_pid, xprt);
 222         task->tk_status = -EAGAIN;
 223         if  (RPC_IS_SOFT(task))
 224                 rpc_sleep_on_timeout(&xprt->sending, task, NULL,
 225                                 xprt_request_timeout(req));
 226         else
 227                 rpc_sleep_on(&xprt->sending, task, NULL);
 228         return 0;
 229 }
 230 EXPORT_SYMBOL_GPL(xprt_reserve_xprt);
 231
 232 static bool
 233 xprt_need_congestion_window_wait(struct rpc_xprt *xprt)
 234 {
 235         return test_bit(XPRT_CWND_WAIT, &xprt->state);
 236 }
 237
 238 static void
 239 xprt_set_congestion_window_wait(struct rpc_xprt *xprt)
 240 {
 241         if (!list_empty(&xprt->xmit_queue)) {
 242                 /* Peek at head of queue to see if it can make progress */
 243                 if (list_first_entry(&xprt->xmit_queue, struct rpc_rqst,
 244                                         rq_xmit)->rq_cong)
 245                         return;
 246         }
 247         set_bit(XPRT_CWND_WAIT, &xprt->state);
 248 }
 249
 250 static void
 251 xprt_test_and_clear_congestion_window_wait(struct rpc_xprt *xprt)
 252 {
 253         if (!RPCXPRT_CONGESTED(xprt))
 254                 clear_bit(XPRT_CWND_WAIT, &xprt->state);
 255 }
 256
 257 /*
 258  * xprt_reserve_xprt_cong - serialize write access to transports
 259  * @task: task that is requesting access to the transport
 260  *
 261  * Same as xprt_reserve_xprt, but Van Jacobson congestion control is
 262  * integrated into the decision of whether a request is allowed to be
 263  * woken up and given access to the transport.
 264  * Note that the lock is only granted if we know there are free slots.
 265  */
 266 int xprt_reserve_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
 267 {
 268         struct rpc_rqst *req = task->tk_rqstp;
 269
 270         if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
 271                 if (task == xprt->snd_task)
 272                         return 1;
 273                 goto out_sleep;
 274         }
 275         if (req == NULL) {
 276                 xprt->snd_task = task;
 277                 return 1;
 278         }
 279         if (test_bit(XPRT_WRITE_SPACE, &xprt->state))
 280                 goto out_unlock;
 281         if (!xprt_need_congestion_window_wait(xprt)) {
 282                 xprt->snd_task = task;
 283                 return 1;
 284         }
 285 out_unlock:
 286         xprt_clear_locked(xprt);
 287 out_sleep:
 288         dprintk("RPC: %5u failed to lock transport %p\n", task->tk_pid, xprt);
 289         task->tk_status = -EAGAIN;
 290         if (RPC_IS_SOFT(task))
 291                 rpc_sleep_on_timeout(&xprt->sending, task, NULL,
 292                                 xprt_request_timeout(req));
 293         else
 294                 rpc_sleep_on(&xprt->sending, task, NULL);
 295         return 0;
 296 }
 297 EXPORT_SYMBOL_GPL(xprt_reserve_xprt_cong);
 298
 299 static inline int xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
 300 {
 301         int retval;
 302
 303         if (test_bit(XPRT_LOCKED, &xprt->state) && xprt->snd_task == task)
 304                 return 1;
 305         spin_lock(&xprt->transport_lock);
 306         retval = xprt->ops->reserve_xprt(xprt, task);
 307         spin_unlock(&xprt->transport_lock);
 308         return retval;
 309 }
 310
 311 static bool __xprt_lock_write_func(struct rpc_task *task, void *data)
 312 {
 313         struct rpc_xprt *xprt = data;
 314
 315         xprt->snd_task = task;
 316         return true;
 317 }
 318
 319 static void __xprt_lock_write_next(struct rpc_xprt *xprt)
 320 {
 321         if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
 322                 return;
 323         if (test_bit(XPRT_WRITE_SPACE, &xprt->state))
 324                 goto out_unlock;
 325         if (rpc_wake_up_first_on_wq(xprtiod_workqueue, &xprt->sending,
 326                                 __xprt_lock_write_func, xprt))
 327                 return;
 328 out_unlock:
 329         xprt_clear_locked(xprt);
 330 }
 331
 332 static void __xprt_lock_write_next_cong(struct rpc_xprt *xprt)
 333 {
 334         if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
 335                 return;
 336         if (test_bit(XPRT_WRITE_SPACE, &xprt->state))
 337                 goto out_unlock;
 338         if (xprt_need_congestion_window_wait(xprt))
 339                 goto out_unlock;
 340         if (rpc_wake_up_first_on_wq(xprtiod_workqueue, &xprt->sending,
 341                                 __xprt_lock_write_func, xprt))
 342                 return;
 343 out_unlock:
 344         xprt_clear_locked(xprt);
 345 }
 346
 347 /**
 348  * xprt_release_xprt - allow other requests to use a transport
 349  * @xprt: transport with other tasks potentially waiting
 350  * @task: task that is releasing access to the transport
 351  *
 352  * Note that "task" can be NULL.  No congestion control is provided.
 353  */
 354 void xprt_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
 355 {
 356         if (xprt->snd_task == task) {
 357                 xprt_clear_locked(xprt);
 358                 __xprt_lock_write_next(xprt);
 359         }
 360 }
 361 EXPORT_SYMBOL_GPL(xprt_release_xprt);
 362
 363 /**
 364  * xprt_release_xprt_cong - allow other requests to use a transport
 365  * @xprt: transport with other tasks potentially waiting
 366  * @task: task that is releasing access to the transport
 367  *
 368  * Note that "task" can be NULL.  Another task is awoken to use the
 369  * transport if the transport's congestion window allows it.
 370  */
 371 void xprt_release_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
 372 {
 373         if (xprt->snd_task == task) {
 374                 xprt_clear_locked(xprt);
 375                 __xprt_lock_write_next_cong(xprt);
 376         }
 377 }
 378 EXPORT_SYMBOL_GPL(xprt_release_xprt_cong);
 379
 380 static inline void xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
 381 {
 382         if (xprt->snd_task != task)
 383                 return;
 384         spin_lock(&xprt->transport_lock);
 385         xprt->ops->release_xprt(xprt, task);
 386         spin_unlock(&xprt->transport_lock);
 387 }
 388
 389 /*
 390  * Van Jacobson congestion avoidance. Check if the congestion window
 391  * overflowed. Put the task to sleep if this is the case.
 392  */
 393 static int
 394 __xprt_get_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
 395 {
 396         if (req->rq_cong)
 397                 return 1;
 398         dprintk("RPC: %5u xprt_cwnd_limited cong = %lu cwnd = %lu\n",
 399                         req->rq_task->tk_pid, xprt->cong, xprt->cwnd);
 400         if (RPCXPRT_CONGESTED(xprt)) {
 401                 xprt_set_congestion_window_wait(xprt);
 402                 return 0;
 403         }
 404         req->rq_cong = 1;
 405         xprt->cong += RPC_CWNDSCALE;
 406         return 1;
 407 }
 408
 409 /*
 410  * Adjust the congestion window, and wake up the next task
 411  * that has been sleeping due to congestion
 412  */
 413 static void
 414 __xprt_put_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
 415 {
 416         if (!req->rq_cong)
 417                 return;
 418         req->rq_cong = 0;
 419         xprt->cong -= RPC_CWNDSCALE;
 420         xprt_test_and_clear_congestion_window_wait(xprt);
 421         __xprt_lock_write_next_cong(xprt);
 422 }
 423
 424 /**
 425  * xprt_request_get_cong - Request congestion control credits
 426  * @xprt: pointer to transport
 427  * @req: pointer to RPC request
 428  *
 429  * Useful for transports that require congestion control.
 430  */
 431 bool
 432 xprt_request_get_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
 433 {
 434         bool ret = false;
 435
 436         if (req->rq_cong)
 437                 return true;
 438         spin_lock(&xprt->transport_lock);
 439         ret = __xprt_get_cong(xprt, req) != 0;
 440         spin_unlock(&xprt->transport_lock);
 441         return ret;
 442 }
 443 EXPORT_SYMBOL_GPL(xprt_request_get_cong);
 444
 445 /**
 446  * xprt_release_rqst_cong - housekeeping when request is complete
 447  * @task: RPC request that recently completed
 448  *
 449  * Useful for transports that require congestion control.
 450  */
 451 void xprt_release_rqst_cong(struct rpc_task *task)
 452 {
 453         struct rpc_rqst *req = task->tk_rqstp;
 454
 455         __xprt_put_cong(req->rq_xprt, req);
 456 }
 457 EXPORT_SYMBOL_GPL(xprt_release_rqst_cong);
 458
 459 static void xprt_clear_congestion_window_wait_locked(struct rpc_xprt *xprt)
 460 {
 461         if (test_and_clear_bit(XPRT_CWND_WAIT, &xprt->state))
 462                 __xprt_lock_write_next_cong(xprt);
 463 }
 464
 465 /*
 466  * Clear the congestion window wait flag and wake up the next
 467  * entry on xprt->sending
 468  */
 469 static void
 470 xprt_clear_congestion_window_wait(struct rpc_xprt *xprt)
 471 {
 472         if (test_and_clear_bit(XPRT_CWND_WAIT, &xprt->state)) {
 473                 spin_lock(&xprt->transport_lock);
 474                 __xprt_lock_write_next_cong(xprt);
 475                 spin_unlock(&xprt->transport_lock);
 476         }
 477 }
 478
 479 /**
 480  * xprt_adjust_cwnd - adjust transport congestion window
 481  * @xprt: pointer to xprt
 482  * @task: recently completed RPC request used to adjust window
 483  * @result: result code of completed RPC request
 484  *
 485  * The transport code maintains an estimate on the maximum number of out-
 486  * standing RPC requests, using a smoothed version of the congestion
 487  * avoidance implemented in 44BSD. This is basically the Van Jacobson
 488  * congestion algorithm: If a retransmit occurs, the congestion window is
 489  * halved; otherwise, it is incremented by 1/cwnd when
 490  *
 491  *      -       a reply is received and
 492  *      -       a full number of requests are outstanding and
 493  *      -       the congestion window hasn't been updated recently.
 494  */
 495 void xprt_adjust_cwnd(struct rpc_xprt *xprt, struct rpc_task *task, int result)
 496 {
 497         struct rpc_rqst *req = task->tk_rqstp;
 498         unsigned long cwnd = xprt->cwnd;
 499
 500         if (result >= 0 && cwnd <= xprt->cong) {
 501                 /* The (cwnd >> 1) term makes sure
 502                  * the result gets rounded properly. */
 503                 cwnd += (RPC_CWNDSCALE * RPC_CWNDSCALE + (cwnd >> 1)) / cwnd;
 504                 if (cwnd > RPC_MAXCWND(xprt))
 505                         cwnd = RPC_MAXCWND(xprt);
 506                 __xprt_lock_write_next_cong(xprt);
 507         } else if (result == -ETIMEDOUT) {
 508                 cwnd >>= 1;
 509                 if (cwnd < RPC_CWNDSCALE)
 510                         cwnd = RPC_CWNDSCALE;
 511         }
 512         dprintk("RPC:       cong %ld, cwnd was %ld, now %ld\n",
 513                         xprt->cong, xprt->cwnd, cwnd);
 514         xprt->cwnd = cwnd;
 515         __xprt_put_cong(xprt, req);
 516 }
 517 EXPORT_SYMBOL_GPL(xprt_adjust_cwnd);
 518
 519 /**
 520  * xprt_wake_pending_tasks - wake all tasks on a transport's pending queue
 521  * @xprt: transport with waiting tasks
 522  * @status: result code to plant in each task before waking it
 523  *
 524  */
 525 void xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status)
 526 {
 527         if (status < 0)
 528                 rpc_wake_up_status(&xprt->pending, status);
 529         else
 530                 rpc_wake_up(&xprt->pending);
 531 }
 532 EXPORT_SYMBOL_GPL(xprt_wake_pending_tasks);
 533
 534 /**
 535  * xprt_wait_for_buffer_space - wait for transport output buffer to clear
 536  * @xprt: transport
 537  *
 538  * Note that we only set the timer for the case of RPC_IS_SOFT(), since
 539  * we don't in general want to force a socket disconnection due to
 540  * an incomplete RPC call transmission.
 541  */
 542 void xprt_wait_for_buffer_space(struct rpc_xprt *xprt)
 543 {
 544         set_bit(XPRT_WRITE_SPACE, &xprt->state);
 545 }
 546 EXPORT_SYMBOL_GPL(xprt_wait_for_buffer_space);
 547
 548 static bool
 549 xprt_clear_write_space_locked(struct rpc_xprt *xprt)
 550 {
 551         if (test_and_clear_bit(XPRT_WRITE_SPACE, &xprt->state)) {
 552                 __xprt_lock_write_next(xprt);
 553                 dprintk("RPC:       write space: waking waiting task on "
 554                                 "xprt %p\n", xprt);
 555                 return true;
 556         }
 557         return false;
 558 }
 559
 560 /**
 561  * xprt_write_space - wake the task waiting for transport output buffer space
 562  * @xprt: transport with waiting tasks
 563  *
 564  * Can be called in a soft IRQ context, so xprt_write_space never sleeps.
 565  */
 566 bool xprt_write_space(struct rpc_xprt *xprt)
 567 {
 568         bool ret;
 569
 570         if (!test_bit(XPRT_WRITE_SPACE, &xprt->state))
 571                 return false;
 572         spin_lock(&xprt->transport_lock);
 573         ret = xprt_clear_write_space_locked(xprt);
 574         spin_unlock(&xprt->transport_lock);
 575         return ret;
 576 }
 577 EXPORT_SYMBOL_GPL(xprt_write_space);
 578
 579 static unsigned long xprt_abs_ktime_to_jiffies(ktime_t abstime)
 580 {
 581         s64 delta = ktime_to_ns(ktime_get() - abstime);
 582         return likely(delta >= 0) ?
 583                 jiffies - nsecs_to_jiffies(delta) :
 584                 jiffies + nsecs_to_jiffies(-delta);
 585 }
 586
 587 static unsigned long xprt_calc_majortimeo(struct rpc_rqst *req)
 588 {
 589         const struct rpc_timeout *to = req->rq_task->tk_client->cl_timeout;
 590         unsigned long majortimeo = req->rq_timeout;
 591
 592         if (to->to_exponential)
 593                 majortimeo <<= to->to_retries;
 594         else
 595                 majortimeo += to->to_increment * to->to_retries;
 596         if (majortimeo > to->to_maxval || majortimeo == 0)
 597                 majortimeo = to->to_maxval;
 598         return majortimeo;
 599 }
 600
 601 static void xprt_reset_majortimeo(struct rpc_rqst *req)
 602 {
 603         req->rq_majortimeo += xprt_calc_majortimeo(req);
 604 }
 605
 606 static void xprt_init_majortimeo(struct rpc_task *task, struct rpc_rqst *req)
 607 {
 608         unsigned long time_init;
 609         struct rpc_xprt *xprt = req->rq_xprt;
 610
 611         if (likely(xprt && xprt_connected(xprt)))
 612                 time_init = jiffies;
 613         else
 614                 time_init = xprt_abs_ktime_to_jiffies(task->tk_start);
 615         req->rq_timeout = task->tk_client->cl_timeout->to_initval;
 616         req->rq_majortimeo = time_init + xprt_calc_majortimeo(req);
 617 }
 618
 619 /**
 620  * xprt_adjust_timeout - adjust timeout values for next retransmit
 621  * @req: RPC request containing parameters to use for the adjustment
 622  *
 623  */
 624 int xprt_adjust_timeout(struct rpc_rqst *req)
 625 {
 626         struct rpc_xprt *xprt = req->rq_xprt;
 627         const struct rpc_timeout *to = req->rq_task->tk_client->cl_timeout;
 628         int status = 0;
 629
 630         if (time_before(jiffies, req->rq_majortimeo)) {
 631                 if (to->to_exponential)
 632                         req->rq_timeout <<= 1;
 633                 else
 634                         req->rq_timeout += to->to_increment;
 635                 if (to->to_maxval && req->rq_timeout >= to->to_maxval)
 636                         req->rq_timeout = to->to_maxval;
 637                 req->rq_retries++;
 638         } else {
 639                 req->rq_timeout = to->to_initval;
 640                 req->rq_retries = 0;
 641                 xprt_reset_majortimeo(req);
 642                 /* Reset the RTT counters == "slow start" */
 643                 spin_lock(&xprt->transport_lock);
 644                 rpc_init_rtt(req->rq_task->tk_client->cl_rtt, to->to_initval);
 645                 spin_unlock(&xprt->transport_lock);
 646                 status = -ETIMEDOUT;
 647         }
 648
 649         if (req->rq_timeout == 0) {
 650                 printk(KERN_WARNING "xprt_adjust_timeout: rq_timeout = 0!\n");
 651                 req->rq_timeout = 5 * HZ;
 652         }
 653         return status;
 654 }
 655
 656 static void xprt_autoclose(struct work_struct *work)
 657 {
 658         struct rpc_xprt *xprt =
 659                 container_of(work, struct rpc_xprt, task_cleanup);
 660         unsigned int pflags = memalloc_nofs_save();
 661
 662         clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
 663         xprt->ops->close(xprt);
 664         xprt_release_write(xprt, NULL);
 665         wake_up_bit(&xprt->state, XPRT_LOCKED);
 666         memalloc_nofs_restore(pflags);
 667 }
 668
 669 /**
 670  * xprt_disconnect_done - mark a transport as disconnected
 671  * @xprt: transport to flag for disconnect
 672  *
 673  */
 674 void xprt_disconnect_done(struct rpc_xprt *xprt)
 675 {
 676         dprintk("RPC:       disconnected transport %p\n", xprt);
 677         spin_lock(&xprt->transport_lock);
 678         xprt_clear_connected(xprt);
 679         xprt_clear_write_space_locked(xprt);
 680         xprt_clear_congestion_window_wait_locked(xprt);
 681         xprt_wake_pending_tasks(xprt, -ENOTCONN);
 682         spin_unlock(&xprt->transport_lock);
 683 }
 684 EXPORT_SYMBOL_GPL(xprt_disconnect_done);
 685
 686 /**
 687  * xprt_force_disconnect - force a transport to disconnect
 688  * @xprt: transport to disconnect
 689  *
 690  */
 691 void xprt_force_disconnect(struct rpc_xprt *xprt)
 692 {
 693         /* Don't race with the test_bit() in xprt_clear_locked() */
 694         spin_lock(&xprt->transport_lock);
 695         set_bit(XPRT_CLOSE_WAIT, &xprt->state);
 696         /* Try to schedule an autoclose RPC call */
 697         if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
 698                 queue_work(xprtiod_workqueue, &xprt->task_cleanup);
 699         else if (xprt->snd_task)
 700                 rpc_wake_up_queued_task_set_status(&xprt->pending,
 701                                 xprt->snd_task, -ENOTCONN);
 702         spin_unlock(&xprt->transport_lock);
 703 }
 704 EXPORT_SYMBOL_GPL(xprt_force_disconnect);
 705
 706 static unsigned int
 707 xprt_connect_cookie(struct rpc_xprt *xprt)
 708 {
 709         return READ_ONCE(xprt->connect_cookie);
 710 }
 711
 712 static bool
 713 xprt_request_retransmit_after_disconnect(struct rpc_task *task)
 714 {
 715         struct rpc_rqst *req = task->tk_rqstp;
 716         struct rpc_xprt *xprt = req->rq_xprt;
 717
 718         return req->rq_connect_cookie != xprt_connect_cookie(xprt) ||
 719                 !xprt_connected(xprt);
 720 }
 721
 722 /**
 723  * xprt_conditional_disconnect - force a transport to disconnect
 724  * @xprt: transport to disconnect
 725  * @cookie: 'connection cookie'
 726  *
 727  * This attempts to break the connection if and only if 'cookie' matches
 728  * the current transport 'connection cookie'. It ensures that we don't
 729  * try to break the connection more than once when we need to retransmit
 730  * a batch of RPC requests.
 731  *
 732  */
 733 void xprt_conditional_disconnect(struct rpc_xprt *xprt, unsigned int cookie)
 734 {
 735         /* Don't race with the test_bit() in xprt_clear_locked() */
 736         spin_lock(&xprt->transport_lock);
 737         if (cookie != xprt->connect_cookie)
 738                 goto out;
 739         if (test_bit(XPRT_CLOSING, &xprt->state))
 740                 goto out;
 741         set_bit(XPRT_CLOSE_WAIT, &xprt->state);
 742         /* Try to schedule an autoclose RPC call */
 743         if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
 744                 queue_work(xprtiod_workqueue, &xprt->task_cleanup);
 745         xprt_wake_pending_tasks(xprt, -EAGAIN);
 746 out:
 747         spin_unlock(&xprt->transport_lock);
 748 }
 749
 750 static bool
 751 xprt_has_timer(const struct rpc_xprt *xprt)
 752 {
 753         return xprt->idle_timeout != 0;
 754 }
 755
 756 static void
 757 xprt_schedule_autodisconnect(struct rpc_xprt *xprt)
 758         __must_hold(&xprt->transport_lock)
 759 {
 760         xprt->last_used = jiffies;
 761         if (RB_EMPTY_ROOT(&xprt->recv_queue) && xprt_has_timer(xprt))
 762                 mod_timer(&xprt->timer, xprt->last_used + xprt->idle_timeout);
 763 }
 764
 765 static void
 766 xprt_init_autodisconnect(struct timer_list *t)
 767 {
 768         struct rpc_xprt *xprt = from_timer(xprt, t, timer);
 769
 770         if (!RB_EMPTY_ROOT(&xprt->recv_queue))
 771                 return;
 772         /* Reset xprt->last_used to avoid connect/autodisconnect cycling */
 773         xprt->last_used = jiffies;
 774         if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
 775                 return;
 776         queue_work(xprtiod_workqueue, &xprt->task_cleanup);
 777 }
 778
 779 bool xprt_lock_connect(struct rpc_xprt *xprt,
 780                 struct rpc_task *task,
 781                 void *cookie)
 782 {
 783         bool ret = false;
 784
 785         spin_lock(&xprt->transport_lock);
 786         if (!test_bit(XPRT_LOCKED, &xprt->state))
 787                 goto out;
 788         if (xprt->snd_task != task)
 789                 goto out;
 790         xprt->snd_task = cookie;
 791         ret = true;
 792 out:
 793         spin_unlock(&xprt->transport_lock);
 794         return ret;
 795 }
 796
 797 void xprt_unlock_connect(struct rpc_xprt *xprt, void *cookie)
 798 {
 799         spin_lock(&xprt->transport_lock);
 800         if (xprt->snd_task != cookie)
 801                 goto out;
 802         if (!test_bit(XPRT_LOCKED, &xprt->state))
 803                 goto out;
 804         xprt->snd_task =NULL;
 805         xprt->ops->release_xprt(xprt, NULL);
 806         xprt_schedule_autodisconnect(xprt);
 807 out:
 808         spin_unlock(&xprt->transport_lock);
 809         wake_up_bit(&xprt->state, XPRT_LOCKED);
 810 }
 811
 812 /**
 813  * xprt_connect - schedule a transport connect operation
 814  * @task: RPC task that is requesting the connect
 815  *
 816  */
 817 void xprt_connect(struct rpc_task *task)
 818 {
 819         struct rpc_xprt *xprt = task->tk_rqstp->rq_xprt;
 820
 821         dprintk("RPC: %5u xprt_connect xprt %p %s connected\n", task->tk_pid,
 822                         xprt, (xprt_connected(xprt) ? "is" : "is not"));
 823
 824         if (!xprt_bound(xprt)) {
 825                 task->tk_status = -EAGAIN;
 826                 return;
 827         }
 828         if (!xprt_lock_write(xprt, task))
 829                 return;
 830
 831         if (test_and_clear_bit(XPRT_CLOSE_WAIT, &xprt->state))
 832                 xprt->ops->close(xprt);
 833
 834         if (!xprt_connected(xprt)) {
 835                 task->tk_rqstp->rq_connect_cookie = xprt->connect_cookie;
 836                 rpc_sleep_on_timeout(&xprt->pending, task, NULL,
 837                                 xprt_request_timeout(task->tk_rqstp));
 838
 839                 if (test_bit(XPRT_CLOSING, &xprt->state))
 840                         return;
 841                 if (xprt_test_and_set_connecting(xprt))
 842                         return;
 843                 /* Race breaker */
 844                 if (!xprt_connected(xprt)) {
 845                         xprt->stat.connect_start = jiffies;
 846                         xprt->ops->connect(xprt, task);
 847                 } else {
 848                         xprt_clear_connecting(xprt);
 849                         task->tk_status = 0;
 850                         rpc_wake_up_queued_task(&xprt->pending, task);
 851                 }
 852         }
 853         xprt_release_write(xprt, task);
 854 }
 855
 856 /**
 857  * xprt_reconnect_delay - compute the wait before scheduling a connect
 858  * @xprt: transport instance
 859  *
 860  */
 861 unsigned long xprt_reconnect_delay(const struct rpc_xprt *xprt)
 862 {
 863         unsigned long start, now = jiffies;
 864
 865         start = xprt->stat.connect_start + xprt->reestablish_timeout;
 866         if (time_after(start, now))
 867                 return start - now;
 868         return 0;
 869 }
 870 EXPORT_SYMBOL_GPL(xprt_reconnect_delay);
 871
 872 /**
 873  * xprt_reconnect_backoff - compute the new re-establish timeout
 874  * @xprt: transport instance
 875  * @init_to: initial reestablish timeout
 876  *
 877  */
 878 void xprt_reconnect_backoff(struct rpc_xprt *xprt, unsigned long init_to)
 879 {
 880         xprt->reestablish_timeout <<= 1;
 881         if (xprt->reestablish_timeout > xprt->max_reconnect_timeout)
 882                 xprt->reestablish_timeout = xprt->max_reconnect_timeout;
 883         if (xprt->reestablish_timeout < init_to)
 884                 xprt->reestablish_timeout = init_to;
 885 }
 886 EXPORT_SYMBOL_GPL(xprt_reconnect_backoff);
 887
 888 enum xprt_xid_rb_cmp {
 889         XID_RB_EQUAL,
 890         XID_RB_LEFT,
 891         XID_RB_RIGHT,
 892 };
 893 static enum xprt_xid_rb_cmp
 894 xprt_xid_cmp(__be32 xid1, __be32 xid2)
 895 {
 896         if (xid1 == xid2)
 897                 return XID_RB_EQUAL;
 898         if ((__force u32)xid1 < (__force u32)xid2)
 899                 return XID_RB_LEFT;
 900         return XID_RB_RIGHT;
 901 }
 902
 903 static struct rpc_rqst *
 904 xprt_request_rb_find(struct rpc_xprt *xprt, __be32 xid)
 905 {
 906         struct rb_node *n = xprt->recv_queue.rb_node;
 907         struct rpc_rqst *req;
 908
 909         while (n != NULL) {
 910                 req = rb_entry(n, struct rpc_rqst, rq_recv);
 911                 switch (xprt_xid_cmp(xid, req->rq_xid)) {
 912                 case XID_RB_LEFT:
 913                         n = n->rb_left;
 914                         break;
 915                 case XID_RB_RIGHT:
 916                         n = n->rb_right;
 917                         break;
 918                 case XID_RB_EQUAL:
 919                         return req;
 920                 }
 921         }
 922         return NULL;
 923 }
 924
 925 static void
 926 xprt_request_rb_insert(struct rpc_xprt *xprt, struct rpc_rqst *new)
 927 {
 928         struct rb_node **p = &xprt->recv_queue.rb_node;
 929         struct rb_node *n = NULL;
 930         struct rpc_rqst *req;
 931
 932         while (*p != NULL) {
 933                 n = *p;
 934                 req = rb_entry(n, struct rpc_rqst, rq_recv);
 935                 switch(xprt_xid_cmp(new->rq_xid, req->rq_xid)) {
 936                 case XID_RB_LEFT:
 937                         p = &n->rb_left;
 938                         break;
 939                 case XID_RB_RIGHT:
 940                         p = &n->rb_right;
 941                         break;
 942                 case XID_RB_EQUAL:
 943                         WARN_ON_ONCE(new != req);
 944                         return;
 945                 }
 946         }
 947         rb_link_node(&new->rq_recv, n, p);
 948         rb_insert_color(&new->rq_recv, &xprt->recv_queue);
 949 }
 950
 951 static void
 952 xprt_request_rb_remove(struct rpc_xprt *xprt, struct rpc_rqst *req)
 953 {
 954         rb_erase(&req->rq_recv, &xprt->recv_queue);
 955 }
 956
 957 /**
 958  * xprt_lookup_rqst - find an RPC request corresponding to an XID
 959  * @xprt: transport on which the original request was transmitted
 960  * @xid: RPC XID of incoming reply
 961  *
 962  * Caller holds xprt->queue_lock.
 963  */
 964 struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, __be32 xid)
 965 {
 966         struct rpc_rqst *entry;
 967
 968         entry = xprt_request_rb_find(xprt, xid);
 969         if (entry != NULL) {
 970                 trace_xprt_lookup_rqst(xprt, xid, 0);
 971                 entry->rq_rtt = ktime_sub(ktime_get(), entry->rq_xtime);
 972                 return entry;
 973         }
 974
 975         dprintk("RPC:       xprt_lookup_rqst did not find xid %08x\n",
 976                         ntohl(xid));
 977         trace_xprt_lookup_rqst(xprt, xid, -ENOENT);
 978         xprt->stat.bad_xids++;
 979         return NULL;
 980 }
 981 EXPORT_SYMBOL_GPL(xprt_lookup_rqst);
 982
 983 static bool
 984 xprt_is_pinned_rqst(struct rpc_rqst *req)
 985 {
 986         return atomic_read(&req->rq_pin) != 0;
 987 }
 988
 989 /**
 990  * xprt_pin_rqst - Pin a request on the transport receive list
 991  * @req: Request to pin
 992  *
 993  * Caller must ensure this is atomic with the call to xprt_lookup_rqst()
 994  * so should be holding xprt->queue_lock.
 995  */
 996 void xprt_pin_rqst(struct rpc_rqst *req)
 997 {
 998         atomic_inc(&req->rq_pin);
 999 }
1000 EXPORT_SYMBOL_GPL(xprt_pin_rqst);
1001
1002 /**
1003  * xprt_unpin_rqst - Unpin a request on the transport receive list
1004  * @req: Request to pin
1005  *
1006  * Caller should be holding xprt->queue_lock.
1007  */
1008 void xprt_unpin_rqst(struct rpc_rqst *req)
1009 {
1010         if (!test_bit(RPC_TASK_MSG_PIN_WAIT, &req->rq_task->tk_runstate)) {
1011                 atomic_dec(&req->rq_pin);
1012                 return;
1013         }
1014         if (atomic_dec_and_test(&req->rq_pin))
1015                 wake_up_var(&req->rq_pin);
1016 }
1017 EXPORT_SYMBOL_GPL(xprt_unpin_rqst);
1018
1019 static void xprt_wait_on_pinned_rqst(struct rpc_rqst *req)
1020 {
1021         wait_var_event(&req->rq_pin, !xprt_is_pinned_rqst(req));
1022 }
1023
1024 static bool
1025 xprt_request_data_received(struct rpc_task *task)
1026 {
1027         return !test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate) &&
1028                 READ_ONCE(task->tk_rqstp->rq_reply_bytes_recvd) != 0;
1029 }
1030
1031 static bool
1032 xprt_request_need_enqueue_receive(struct rpc_task *task, struct rpc_rqst *req)
1033 {
1034         return !test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate) &&
1035                 READ_ONCE(task->tk_rqstp->rq_reply_bytes_recvd) == 0;
1036 }
1037
1038 /**
1039  * xprt_request_enqueue_receive - Add an request to the receive queue
1040  * @task: RPC task
1041  *
1042  */
1043 void
1044 xprt_request_enqueue_receive(struct rpc_task *task)
1045 {
1046         struct rpc_rqst *req = task->tk_rqstp;
1047         struct rpc_xprt *xprt = req->rq_xprt;
1048
1049         if (!xprt_request_need_enqueue_receive(task, req))
1050                 return;
1051
1052         xprt_request_prepare(task->tk_rqstp);
1053         spin_lock(&xprt->queue_lock);
1054
1055         /* Update the softirq receive buffer */
1056         memcpy(&req->rq_private_buf, &req->rq_rcv_buf,
1057                         sizeof(req->rq_private_buf));
1058
1059         /* Add request to the receive list */
1060         xprt_request_rb_insert(xprt, req);
1061         set_bit(RPC_TASK_NEED_RECV, &task->tk_runstate);
1062         spin_unlock(&xprt->queue_lock);
1063
1064         /* Turn off autodisconnect */
1065         del_singleshot_timer_sync(&xprt->timer);
1066 }
1067
1068 /**
1069  * xprt_request_dequeue_receive_locked - Remove a request from the receive queue
1070  * @task: RPC task
1071  *
1072  * Caller must hold xprt->queue_lock.
1073  */
1074 static void
1075 xprt_request_dequeue_receive_locked(struct rpc_task *task)
1076 {
1077         struct rpc_rqst *req = task->tk_rqstp;
1078
1079         if (test_and_clear_bit(RPC_TASK_NEED_RECV, &task->tk_runstate))
1080                 xprt_request_rb_remove(req->rq_xprt, req);
1081 }
1082
1083 /**
1084  * xprt_update_rtt - Update RPC RTT statistics
1085  * @task: RPC request that recently completed
1086  *
1087  * Caller holds xprt->queue_lock.
1088  */
1089 void xprt_update_rtt(struct rpc_task *task)
1090 {
1091         struct rpc_rqst *req = task->tk_rqstp;
1092         struct rpc_rtt *rtt = task->tk_client->cl_rtt;
1093         unsigned int timer = task->tk_msg.rpc_proc->p_timer;
1094         long m = usecs_to_jiffies(ktime_to_us(req->rq_rtt));
1095
1096         if (timer) {
1097                 if (req->rq_ntrans == 1)
1098                         rpc_update_rtt(rtt, timer, m);
1099                 rpc_set_timeo(rtt, timer, req->rq_ntrans - 1);
1100         }
1101 }
1102 EXPORT_SYMBOL_GPL(xprt_update_rtt);
1103
1104 /**
1105  * xprt_complete_rqst - called when reply processing is complete
1106  * @task: RPC request that recently completed
1107  * @copied: actual number of bytes received from the transport
1108  *
1109  * Caller holds xprt->queue_lock.
1110  */
1111 void xprt_complete_rqst(struct rpc_task *task, int copied)
1112 {
1113         struct rpc_rqst *req = task->tk_rqstp;
1114         struct rpc_xprt *xprt = req->rq_xprt;
1115
1116         dprintk("RPC: %5u xid %08x complete (%d bytes received)\n",
1117                         task->tk_pid, ntohl(req->rq_xid), copied);
1118         trace_xprt_complete_rqst(xprt, req->rq_xid, copied);
1119
1120         xprt->stat.recvs++;
1121
1122         req->rq_private_buf.len = copied;
1123         /* Ensure all writes are done before we update */
1124         /* req->rq_reply_bytes_recvd */
1125         smp_wmb();
1126         req->rq_reply_bytes_recvd = copied;
1127         xprt_request_dequeue_receive_locked(task);
1128         rpc_wake_up_queued_task(&xprt->pending, task);
1129 }
1130 EXPORT_SYMBOL_GPL(xprt_complete_rqst);
1131
1132 static void xprt_timer(struct rpc_task *task)
1133 {
1134         struct rpc_rqst *req = task->tk_rqstp;
1135         struct rpc_xprt *xprt = req->rq_xprt;
1136
1137         if (task->tk_status != -ETIMEDOUT)
1138                 return;
1139
1140         trace_xprt_timer(xprt, req->rq_xid, task->tk_status);
1141         if (!req->rq_reply_bytes_recvd) {
1142                 if (xprt->ops->timer)
1143                         xprt->ops->timer(xprt, task);
1144         } else
1145                 task->tk_status = 0;
1146 }
1147
1148 /**
1149  * xprt_wait_for_reply_request_def - wait for reply
1150  * @task: pointer to rpc_task
1151  *
1152  * Set a request's retransmit timeout based on the transport's
1153  * default timeout parameters.  Used by transports that don't adjust
1154  * the retransmit timeout based on round-trip time estimation,
1155  * and put the task to sleep on the pending queue.
1156  */
1157 void xprt_wait_for_reply_request_def(struct rpc_task *task)
1158 {
1159         struct rpc_rqst *req = task->tk_rqstp;
1160
1161         rpc_sleep_on_timeout(&req->rq_xprt->pending, task, xprt_timer,
1162                         xprt_request_timeout(req));
1163 }
1164 EXPORT_SYMBOL_GPL(xprt_wait_for_reply_request_def);
1165
1166 /**
1167  * xprt_wait_for_reply_request_rtt - wait for reply using RTT estimator
1168  * @task: pointer to rpc_task
1169  *
1170  * Set a request's retransmit timeout using the RTT estimator,
1171  * and put the task to sleep on the pending queue.
1172  */
1173 void xprt_wait_for_reply_request_rtt(struct rpc_task *task)
1174 {
1175         int timer = task->tk_msg.rpc_proc->p_timer;
1176         struct rpc_clnt *clnt = task->tk_client;
1177         struct rpc_rtt *rtt = clnt->cl_rtt;
1178         struct rpc_rqst *req = task->tk_rqstp;
1179         unsigned long max_timeout = clnt->cl_timeout->to_maxval;
1180         unsigned long timeout;
1181
1182         timeout = rpc_calc_rto(rtt, timer);
1183         timeout <<= rpc_ntimeo(rtt, timer) + req->rq_retries;
1184         if (timeout > max_timeout || timeout == 0)
1185                 timeout = max_timeout;
1186         rpc_sleep_on_timeout(&req->rq_xprt->pending, task, xprt_timer,
1187                         jiffies + timeout);
1188 }
1189 EXPORT_SYMBOL_GPL(xprt_wait_for_reply_request_rtt);
1190
1191 /**
1192  * xprt_request_wait_receive - wait for the reply to an RPC request
1193  * @task: RPC task about to send a request
1194  *
1195  */
1196 void xprt_request_wait_receive(struct rpc_task *task)
1197 {
1198         struct rpc_rqst *req = task->tk_rqstp;
1199         struct rpc_xprt *xprt = req->rq_xprt;
1200
1201         if (!test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate))
1202                 return;
1203         /*
1204          * Sleep on the pending queue if we're expecting a reply.
1205          * The spinlock ensures atomicity between the test of
1206          * req->rq_reply_bytes_recvd, and the call to rpc_sleep_on().
1207          */
1208         spin_lock(&xprt->queue_lock);
1209         if (test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate)) {
1210                 xprt->ops->wait_for_reply_request(task);
1211                 /*
1212                  * Send an extra queue wakeup call if the
1213                  * connection was dropped in case the call to
1214                  * rpc_sleep_on() raced.
1215                  */
1216                 if (xprt_request_retransmit_after_disconnect(task))
1217                         rpc_wake_up_queued_task_set_status(&xprt->pending,
1218                                         task, -ENOTCONN);
1219         }
1220         spin_unlock(&xprt->queue_lock);
1221 }
1222
1223 static bool
1224 xprt_request_need_enqueue_transmit(struct rpc_task *task, struct rpc_rqst *req)
1225 {
1226         return !test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate);
1227 }
1228
1229 /**
1230  * xprt_request_enqueue_transmit - queue a task for transmission
1231  * @task: pointer to rpc_task
1232  *
1233  * Add a task to the transmission queue.
1234  */
1235 void
1236 xprt_request_enqueue_transmit(struct rpc_task *task)
1237 {
1238         struct rpc_rqst *pos, *req = task->tk_rqstp;
1239         struct rpc_xprt *xprt = req->rq_xprt;
1240
1241         if (xprt_request_need_enqueue_transmit(task, req)) {
1242                 req->rq_bytes_sent = 0;
1243                 spin_lock(&xprt->queue_lock);
1244                 /*
1245                  * Requests that carry congestion control credits are added
1246                  * to the head of the list to avoid starvation issues.
1247                  */
1248                 if (req->rq_cong) {
1249                         xprt_clear_congestion_window_wait(xprt);
1250                         list_for_each_entry(pos, &xprt->xmit_queue, rq_xmit) {
1251                                 if (pos->rq_cong)
1252                                         continue;
1253                                 /* Note: req is added _before_ pos */
1254                                 list_add_tail(&req->rq_xmit, &pos->rq_xmit);
1255                                 INIT_LIST_HEAD(&req->rq_xmit2);
1256                                 trace_xprt_enq_xmit(task, 1);
1257                                 goto out;
1258                         }
1259                 } else if (RPC_IS_SWAPPER(task)) {
1260                         list_for_each_entry(pos, &xprt->xmit_queue, rq_xmit) {
1261                                 if (pos->rq_cong || pos->rq_bytes_sent)
1262                                         continue;
1263                                 if (RPC_IS_SWAPPER(pos->rq_task))
1264                                         continue;
1265                                 /* Note: req is added _before_ pos */
1266                                 list_add_tail(&req->rq_xmit, &pos->rq_xmit);
1267                                 INIT_LIST_HEAD(&req->rq_xmit2);
1268                                 trace_xprt_enq_xmit(task, 2);
1269                                 goto out;
1270                         }
1271                 } else if (!req->rq_seqno) {
1272                         list_for_each_entry(pos, &xprt->xmit_queue, rq_xmit) {
1273                                 if (pos->rq_task->tk_owner != task->tk_owner)
1274                                         continue;
1275                                 list_add_tail(&req->rq_xmit2, &pos->rq_xmit2);
1276                                 INIT_LIST_HEAD(&req->rq_xmit);
1277                                 trace_xprt_enq_xmit(task, 3);
1278                                 goto out;
1279                         }
1280                 }
1281                 list_add_tail(&req->rq_xmit, &xprt->xmit_queue);
1282                 INIT_LIST_HEAD(&req->rq_xmit2);
1283                 trace_xprt_enq_xmit(task, 4);
1284 out:
1285                 set_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate);
1286                 spin_unlock(&xprt->queue_lock);
1287         }
1288 }
1289
1290 /**
1291  * xprt_request_dequeue_transmit_locked - remove a task from the transmission queue
1292  * @task: pointer to rpc_task
1293  *
1294  * Remove a task from the transmission queue
1295  * Caller must hold xprt->queue_lock
1296  */
1297 static void
1298 xprt_request_dequeue_transmit_locked(struct rpc_task *task)
1299 {
1300         struct rpc_rqst *req = task->tk_rqstp;
1301
1302         if (!test_and_clear_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate))
1303                 return;
1304         if (!list_empty(&req->rq_xmit)) {
1305                 list_del(&req->rq_xmit);
1306                 if (!list_empty(&req->rq_xmit2)) {
1307                         struct rpc_rqst *next = list_first_entry(&req->rq_xmit2,
1308                                         struct rpc_rqst, rq_xmit2);
1309                         list_del(&req->rq_xmit2);
1310                         list_add_tail(&next->rq_xmit, &next->rq_xprt->xmit_queue);
1311                 }
1312         } else
1313                 list_del(&req->rq_xmit2);
1314 }
1315
1316 /**
1317  * xprt_request_dequeue_transmit - remove a task from the transmission queue
1318  * @task: pointer to rpc_task
1319  *
1320  * Remove a task from the transmission queue
1321  */
1322 static void
1323 xprt_request_dequeue_transmit(struct rpc_task *task)
1324 {
1325         struct rpc_rqst *req = task->tk_rqstp;
1326         struct rpc_xprt *xprt = req->rq_xprt;
1327
1328         spin_lock(&xprt->queue_lock);
1329         xprt_request_dequeue_transmit_locked(task);
1330         spin_unlock(&xprt->queue_lock);
1331 }
1332
1333 /**
1334  * xprt_request_dequeue_xprt - remove a task from the transmit+receive queue
1335  * @task: pointer to rpc_task
1336  *
1337  * Remove a task from the transmit and receive queues, and ensure that
1338  * it is not pinned by the receive work item.
1339  */
1340 void
1341 xprt_request_dequeue_xprt(struct rpc_task *task)
1342 {
1343         struct rpc_rqst *req = task->tk_rqstp;
1344         struct rpc_xprt *xprt = req->rq_xprt;
1345
1346         if (test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate) ||
1347             test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate) ||
1348             xprt_is_pinned_rqst(req)) {
1349                 spin_lock(&xprt->queue_lock);
1350                 xprt_request_dequeue_transmit_locked(task);
1351                 xprt_request_dequeue_receive_locked(task);
1352                 while (xprt_is_pinned_rqst(req)) {
1353                         set_bit(RPC_TASK_MSG_PIN_WAIT, &task->tk_runstate);
1354                         spin_unlock(&xprt->queue_lock);
1355                         xprt_wait_on_pinned_rqst(req);
1356                         spin_lock(&xprt->queue_lock);
1357                         clear_bit(RPC_TASK_MSG_PIN_WAIT, &task->tk_runstate);
1358                 }
1359                 spin_unlock(&xprt->queue_lock);
1360         }
1361 }
1362
1363 /**
1364  * xprt_request_prepare - prepare an encoded request for transport
1365  * @req: pointer to rpc_rqst
1366  *
1367  * Calls into the transport layer to do whatever is needed to prepare
1368  * the request for transmission or receive.
1369  */
1370 void
1371 xprt_request_prepare(struct rpc_rqst *req)
1372 {
1373         struct rpc_xprt *xprt = req->rq_xprt;
1374
1375         if (xprt->ops->prepare_request)
1376                 xprt->ops->prepare_request(req);
1377 }
1378
1379 /**
1380  * xprt_request_need_retransmit - Test if a task needs retransmission
1381  * @task: pointer to rpc_task
1382  *
1383  * Test for whether a connection breakage requires the task to retransmit
1384  */
1385 bool
1386 xprt_request_need_retransmit(struct rpc_task *task)
1387 {
1388         return xprt_request_retransmit_after_disconnect(task);
1389 }
1390
1391 /**
1392  * xprt_prepare_transmit - reserve the transport before sending a request
1393  * @task: RPC task about to send a request
1394  *
1395  */
1396 bool xprt_prepare_transmit(struct rpc_task *task)
1397 {
1398         struct rpc_rqst *req = task->tk_rqstp;
1399         struct rpc_xprt *xprt = req->rq_xprt;
1400
1401         dprintk("RPC: %5u xprt_prepare_transmit\n", task->tk_pid);
1402
1403         if (!xprt_lock_write(xprt, task)) {
1404                 /* Race breaker: someone may have transmitted us */
1405                 if (!test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate))
1406                         rpc_wake_up_queued_task_set_status(&xprt->sending,
1407                                         task, 0);
1408                 return false;
1409
1410         }
1411         return true;
1412 }
1413
1414 void xprt_end_transmit(struct rpc_task *task)
1415 {
1416         xprt_release_write(task->tk_rqstp->rq_xprt, task);
1417 }
1418
1419 /**
1420  * xprt_request_transmit - send an RPC request on a transport
1421  * @req: pointer to request to transmit
1422  * @snd_task: RPC task that owns the transport lock
1423  *
1424  * This performs the transmission of a single request.
1425  * Note that if the request is not the same as snd_task, then it
1426  * does need to be pinned.
1427  * Returns '0' on success.
1428  */
1429 static int
1430 xprt_request_transmit(struct rpc_rqst *req, struct rpc_task *snd_task)
1431 {
1432         struct rpc_xprt *xprt = req->rq_xprt;
1433         struct rpc_task *task = req->rq_task;
1434         unsigned int connect_cookie;
1435         int is_retrans = RPC_WAS_SENT(task);
1436         int status;
1437
1438         if (!req->rq_bytes_sent) {
1439                 if (xprt_request_data_received(task)) {
1440                         status = 0;
1441                         goto out_dequeue;
1442                 }
1443                 /* Verify that our message lies in the RPCSEC_GSS window */
1444                 if (rpcauth_xmit_need_reencode(task)) {
1445                         status = -EBADMSG;
1446                         goto out_dequeue;
1447                 }
1448                 if (RPC_SIGNALLED(task)) {
1449                         status = -ERESTARTSYS;
1450                         goto out_dequeue;
1451                 }
1452         }
1453
1454         /*
1455          * Update req->rq_ntrans before transmitting to avoid races with
1456          * xprt_update_rtt(), which needs to know that it is recording a
1457          * reply to the first transmission.
1458          */
1459         req->rq_ntrans++;
1460
1461         connect_cookie = xprt->connect_cookie;
1462         status = xprt->ops->send_request(req);
1463         if (status != 0) {
1464                 req->rq_ntrans--;
1465                 trace_xprt_transmit(req, status);
1466                 return status;
1467         }
1468
1469         if (is_retrans)
1470                 task->tk_client->cl_stats->rpcretrans++;
1471
1472         xprt_inject_disconnect(xprt);
1473
1474         task->tk_flags |= RPC_TASK_SENT;
1475         spin_lock(&xprt->transport_lock);
1476
1477         xprt->stat.sends++;
1478         xprt->stat.req_u += xprt->stat.sends - xprt->stat.recvs;
1479         xprt->stat.bklog_u += xprt->backlog.qlen;
1480         xprt->stat.sending_u += xprt->sending.qlen;
1481         xprt->stat.pending_u += xprt->pending.qlen;
1482         spin_unlock(&xprt->transport_lock);
1483
1484         req->rq_connect_cookie = connect_cookie;
1485 out_dequeue:
1486         trace_xprt_transmit(req, status);
1487         xprt_request_dequeue_transmit(task);
1488         rpc_wake_up_queued_task_set_status(&xprt->sending, task, status);
1489         return status;
1490 }
1491
1492 /**
1493  * xprt_transmit - send an RPC request on a transport
1494  * @task: controlling RPC task
1495  *
1496  * Attempts to drain the transmit queue. On exit, either the transport
1497  * signalled an error that needs to be handled before transmission can
1498  * resume, or @task finished transmitting, and detected that it already
1499  * received a reply.
1500  */
1501 void
1502 xprt_transmit(struct rpc_task *task)
1503 {
1504         struct rpc_rqst *next, *req = task->tk_rqstp;
1505         struct rpc_xprt *xprt = req->rq_xprt;
1506         int status;
1507
1508         spin_lock(&xprt->queue_lock);
1509         while (!list_empty(&xprt->xmit_queue)) {
1510                 next = list_first_entry(&xprt->xmit_queue,
1511                                 struct rpc_rqst, rq_xmit);
1512                 xprt_pin_rqst(next);
1513                 spin_unlock(&xprt->queue_lock);
1514                 status = xprt_request_transmit(next, task);
1515                 if (status == -EBADMSG && next != req)
1516                         status = 0;
1517                 cond_resched();
1518                 spin_lock(&xprt->queue_lock);
1519                 xprt_unpin_rqst(next);
1520                 if (status == 0) {
1521                         if (!xprt_request_data_received(task) ||
1522                             test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate))
1523                                 continue;
1524                 } else if (test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate))
1525                         task->tk_status = status;
1526                 break;
1527         }
1528         spin_unlock(&xprt->queue_lock);
1529 }
1530
1531 static void xprt_add_backlog(struct rpc_xprt *xprt, struct rpc_task *task)
1532 {
1533         set_bit(XPRT_CONGESTED, &xprt->state);
1534         rpc_sleep_on(&xprt->backlog, task, NULL);
1535 }
1536
1537 static void xprt_wake_up_backlog(struct rpc_xprt *xprt)
1538 {
1539         if (rpc_wake_up_next(&xprt->backlog) == NULL)
1540                 clear_bit(XPRT_CONGESTED, &xprt->state);
1541 }
1542
1543 static bool xprt_throttle_congested(struct rpc_xprt *xprt, struct rpc_task *task)
1544 {
1545         bool ret = false;
1546
1547         if (!test_bit(XPRT_CONGESTED, &xprt->state))
1548                 goto out;
1549         spin_lock(&xprt->reserve_lock);
1550         if (test_bit(XPRT_CONGESTED, &xprt->state)) {
1551                 rpc_sleep_on(&xprt->backlog, task, NULL);
1552                 ret = true;
1553         }
1554         spin_unlock(&xprt->reserve_lock);
1555 out:
1556         return ret;
1557 }
1558
1559 static struct rpc_rqst *xprt_dynamic_alloc_slot(struct rpc_xprt *xprt)
1560 {
1561         struct rpc_rqst *req = ERR_PTR(-EAGAIN);
1562
1563         if (xprt->num_reqs >= xprt->max_reqs)
1564                 goto out;
1565         ++xprt->num_reqs;
1566         spin_unlock(&xprt->reserve_lock);
1567         req = kzalloc(sizeof(struct rpc_rqst), GFP_NOFS);
1568         spin_lock(&xprt->reserve_lock);
1569         if (req != NULL)
1570                 goto out;
1571         --xprt->num_reqs;
1572         req = ERR_PTR(-ENOMEM);
1573 out:
1574         return req;
1575 }
1576
1577 static bool xprt_dynamic_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
1578 {
1579         if (xprt->num_reqs > xprt->min_reqs) {
1580                 --xprt->num_reqs;
1581                 kfree(req);
1582                 return true;
1583         }
1584         return false;
1585 }
1586
1587 void xprt_alloc_slot(struct rpc_xprt *xprt, struct rpc_task *task)
1588 {
1589         struct rpc_rqst *req;
1590
1591         spin_lock(&xprt->reserve_lock);
1592         if (!list_empty(&xprt->free)) {
1593                 req = list_entry(xprt->free.next, struct rpc_rqst, rq_list);
1594                 list_del(&req->rq_list);
1595                 goto out_init_req;
1596         }
1597         req = xprt_dynamic_alloc_slot(xprt);
1598         if (!IS_ERR(req))
1599                 goto out_init_req;
1600         switch (PTR_ERR(req)) {
1601         case -ENOMEM:
1602                 dprintk("RPC:       dynamic allocation of request slot "
1603                                 "failed! Retrying\n");
1604                 task->tk_status = -ENOMEM;
1605                 break;
1606         case -EAGAIN:
1607                 xprt_add_backlog(xprt, task);
1608                 dprintk("RPC:       waiting for request slot\n");
1609                 /* fall through */
1610         default:
1611                 task->tk_status = -EAGAIN;
1612         }
1613         spin_unlock(&xprt->reserve_lock);
1614         return;
1615 out_init_req:
1616         xprt->stat.max_slots = max_t(unsigned int, xprt->stat.max_slots,
1617                                      xprt->num_reqs);
1618         spin_unlock(&xprt->reserve_lock);
1619
1620         task->tk_status = 0;
1621         task->tk_rqstp = req;
1622 }
1623 EXPORT_SYMBOL_GPL(xprt_alloc_slot);
1624
1625 void xprt_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
1626 {
1627         spin_lock(&xprt->reserve_lock);
1628         if (!xprt_dynamic_free_slot(xprt, req)) {
1629                 memset(req, 0, sizeof(*req));   /* mark unused */
1630                 list_add(&req->rq_list, &xprt->free);
1631         }
1632         xprt_wake_up_backlog(xprt);
1633         spin_unlock(&xprt->reserve_lock);
1634 }
1635 EXPORT_SYMBOL_GPL(xprt_free_slot);
1636
1637 static void xprt_free_all_slots(struct rpc_xprt *xprt)
1638 {
1639         struct rpc_rqst *req;
1640         while (!list_empty(&xprt->free)) {
1641                 req = list_first_entry(&xprt->free, struct rpc_rqst, rq_list);
1642                 list_del(&req->rq_list);
1643                 kfree(req);
1644         }
1645 }
1646
1647 struct rpc_xprt *xprt_alloc(struct net *net, size_t size,
1648                 unsigned int num_prealloc,
1649                 unsigned int max_alloc)
1650 {
1651         struct rpc_xprt *xprt;
1652         struct rpc_rqst *req;
1653         int i;
1654
1655         xprt = kzalloc(size, GFP_KERNEL);
1656         if (xprt == NULL)
1657                 goto out;
1658
1659         xprt_init(xprt, net);
1660
1661         for (i = 0; i < num_prealloc; i++) {
1662                 req = kzalloc(sizeof(struct rpc_rqst), GFP_KERNEL);
1663                 if (!req)
1664                         goto out_free;
1665                 list_add(&req->rq_list, &xprt->free);
1666         }
1667         if (max_alloc > num_prealloc)
1668                 xprt->max_reqs = max_alloc;
1669         else
1670                 xprt->max_reqs = num_prealloc;
1671         xprt->min_reqs = num_prealloc;
1672         xprt->num_reqs = num_prealloc;
1673
1674         return xprt;
1675
1676 out_free:
1677         xprt_free(xprt);
1678 out:
1679         return NULL;
1680 }
1681 EXPORT_SYMBOL_GPL(xprt_alloc);
1682
1683 void xprt_free(struct rpc_xprt *xprt)
1684 {
1685         put_net(xprt->xprt_net);
1686         xprt_free_all_slots(xprt);
1687         kfree_rcu(xprt, rcu);
1688 }
1689 EXPORT_SYMBOL_GPL(xprt_free);
1690
1691 static void
1692 xprt_init_connect_cookie(struct rpc_rqst *req, struct rpc_xprt *xprt)
1693 {
1694         req->rq_connect_cookie = xprt_connect_cookie(xprt) - 1;
1695 }
1696
1697 static __be32
1698 xprt_alloc_xid(struct rpc_xprt *xprt)
1699 {
1700         __be32 xid;
1701
1702         spin_lock(&xprt->reserve_lock);
1703         xid = (__force __be32)xprt->xid++;
1704         spin_unlock(&xprt->reserve_lock);
1705         return xid;
1706 }
1707
1708 static void
1709 xprt_init_xid(struct rpc_xprt *xprt)
1710 {
1711         xprt->xid = prandom_u32();
1712 }
1713
1714 static void
1715 xprt_request_init(struct rpc_task *task)
1716 {
1717         struct rpc_xprt *xprt = task->tk_xprt;
1718         struct rpc_rqst *req = task->tk_rqstp;
1719
1720         req->rq_task    = task;
1721         req->rq_xprt    = xprt;
1722         req->rq_buffer  = NULL;
1723         req->rq_xid     = xprt_alloc_xid(xprt);
1724         xprt_init_connect_cookie(req, xprt);
1725         req->rq_snd_buf.len = 0;
1726         req->rq_snd_buf.buflen = 0;
1727         req->rq_rcv_buf.len = 0;
1728         req->rq_rcv_buf.buflen = 0;
1729         req->rq_snd_buf.bvec = NULL;
1730         req->rq_rcv_buf.bvec = NULL;
1731         req->rq_release_snd_buf = NULL;
1732         xprt_init_majortimeo(task, req);
1733         dprintk("RPC: %5u reserved req %p xid %08x\n", task->tk_pid,
1734                         req, ntohl(req->rq_xid));
1735 }
1736
1737 static void
1738 xprt_do_reserve(struct rpc_xprt *xprt, struct rpc_task *task)
1739 {
1740         xprt->ops->alloc_slot(xprt, task);
1741         if (task->tk_rqstp != NULL)
1742                 xprt_request_init(task);
1743 }
1744
1745 /**
1746  * xprt_reserve - allocate an RPC request slot
1747  * @task: RPC task requesting a slot allocation
1748  *
1749  * If the transport is marked as being congested, or if no more
1750  * slots are available, place the task on the transport's
1751  * backlog queue.
1752  */
1753 void xprt_reserve(struct rpc_task *task)
1754 {
1755         struct rpc_xprt *xprt = task->tk_xprt;
1756
1757         task->tk_status = 0;
1758         if (task->tk_rqstp != NULL)
1759                 return;
1760
1761         task->tk_status = -EAGAIN;
1762         if (!xprt_throttle_congested(xprt, task))
1763                 xprt_do_reserve(xprt, task);
1764 }
1765
1766 /**
1767  * xprt_retry_reserve - allocate an RPC request slot
1768  * @task: RPC task requesting a slot allocation
1769  *
1770  * If no more slots are available, place the task on the transport's
1771  * backlog queue.
1772  * Note that the only difference with xprt_reserve is that we now
1773  * ignore the value of the XPRT_CONGESTED flag.
1774  */
1775 void xprt_retry_reserve(struct rpc_task *task)
1776 {
1777         struct rpc_xprt *xprt = task->tk_xprt;
1778
1779         task->tk_status = 0;
1780         if (task->tk_rqstp != NULL)
1781                 return;
1782
1783         task->tk_status = -EAGAIN;
1784         xprt_do_reserve(xprt, task);
1785 }
1786
1787 /**
1788  * xprt_release - release an RPC request slot
1789  * @task: task which is finished with the slot
1790  *
1791  */
1792 void xprt_release(struct rpc_task *task)
1793 {
1794         struct rpc_xprt *xprt;
1795         struct rpc_rqst *req = task->tk_rqstp;
1796
1797         if (req == NULL) {
1798                 if (task->tk_client) {
1799                         xprt = task->tk_xprt;
1800                         xprt_release_write(xprt, task);
1801                 }
1802                 return;
1803         }
1804
1805         xprt = req->rq_xprt;
1806         xprt_request_dequeue_xprt(task);
1807         spin_lock(&xprt->transport_lock);
1808         xprt->ops->release_xprt(xprt, task);
1809         if (xprt->ops->release_request)
1810                 xprt->ops->release_request(task);
1811         xprt_schedule_autodisconnect(xprt);
1812         spin_unlock(&xprt->transport_lock);
1813         if (req->rq_buffer)
1814                 xprt->ops->buf_free(task);
1815         xprt_inject_disconnect(xprt);
1816         xdr_free_bvec(&req->rq_rcv_buf);
1817         xdr_free_bvec(&req->rq_snd_buf);
1818         if (req->rq_cred != NULL)
1819                 put_rpccred(req->rq_cred);
1820         task->tk_rqstp = NULL;
1821         if (req->rq_release_snd_buf)
1822                 req->rq_release_snd_buf(req);
1823
1824         dprintk("RPC: %5u release request %p\n", task->tk_pid, req);
1825         if (likely(!bc_prealloc(req)))
1826                 xprt->ops->free_slot(xprt, req);
1827         else
1828                 xprt_free_bc_request(req);
1829 }
1830
1831 #ifdef CONFIG_SUNRPC_BACKCHANNEL
1832 void
1833 xprt_init_bc_request(struct rpc_rqst *req, struct rpc_task *task)
1834 {
1835         struct xdr_buf *xbufp = &req->rq_snd_buf;
1836
1837         task->tk_rqstp = req;
1838         req->rq_task = task;
1839         xprt_init_connect_cookie(req, req->rq_xprt);
1840         /*
1841          * Set up the xdr_buf length.
1842          * This also indicates that the buffer is XDR encoded already.
1843          */
1844         xbufp->len = xbufp->head[0].iov_len + xbufp->page_len +
1845                 xbufp->tail[0].iov_len;
1846 }
1847 #endif
1848
1849 static void xprt_init(struct rpc_xprt *xprt, struct net *net)
1850 {
1851         kref_init(&xprt->kref);
1852
1853         spin_lock_init(&xprt->transport_lock);
1854         spin_lock_init(&xprt->reserve_lock);
1855         spin_lock_init(&xprt->queue_lock);
1856
1857         INIT_LIST_HEAD(&xprt->free);
1858         xprt->recv_queue = RB_ROOT;
1859         INIT_LIST_HEAD(&xprt->xmit_queue);
1860 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
1861         spin_lock_init(&xprt->bc_pa_lock);
1862         INIT_LIST_HEAD(&xprt->bc_pa_list);
1863 #endif /* CONFIG_SUNRPC_BACKCHANNEL */
1864         INIT_LIST_HEAD(&xprt->xprt_switch);
1865
1866         xprt->last_used = jiffies;
1867         xprt->cwnd = RPC_INITCWND;
1868         xprt->bind_index = 0;
1869
1870         rpc_init_wait_queue(&xprt->binding, "xprt_binding");
1871         rpc_init_wait_queue(&xprt->pending, "xprt_pending");
1872         rpc_init_wait_queue(&xprt->sending, "xprt_sending");
1873         rpc_init_priority_wait_queue(&xprt->backlog, "xprt_backlog");
1874
1875         xprt_init_xid(xprt);
1876
1877         xprt->xprt_net = get_net(net);
1878 }
1879
1880 /**
1881  * xprt_create_transport - create an RPC transport
1882  * @args: rpc transport creation arguments
1883  *
1884  */
1885 struct rpc_xprt *xprt_create_transport(struct xprt_create *args)
1886 {
1887         struct rpc_xprt *xprt;
1888         struct xprt_class *t;
1889
1890         spin_lock(&xprt_list_lock);
1891         list_for_each_entry(t, &xprt_list, list) {
1892                 if (t->ident == args->ident) {
1893                         spin_unlock(&xprt_list_lock);
1894                         goto found;
1895                 }
1896         }
1897         spin_unlock(&xprt_list_lock);
1898         dprintk("RPC: transport (%d) not supported\n", args->ident);
1899         return ERR_PTR(-EIO);
1900
1901 found:
1902         xprt = t->setup(args);
1903         if (IS_ERR(xprt)) {
1904                 dprintk("RPC:       xprt_create_transport: failed, %ld\n",
1905                                 -PTR_ERR(xprt));
1906                 goto out;
1907         }
1908         if (args->flags & XPRT_CREATE_NO_IDLE_TIMEOUT)
1909                 xprt->idle_timeout = 0;
1910         INIT_WORK(&xprt->task_cleanup, xprt_autoclose);
1911         if (xprt_has_timer(xprt))
1912                 timer_setup(&xprt->timer, xprt_init_autodisconnect, 0);
1913         else
1914                 timer_setup(&xprt->timer, NULL, 0);
1915
1916         if (strlen(args->servername) > RPC_MAXNETNAMELEN) {
1917                 xprt_destroy(xprt);
1918                 return ERR_PTR(-EINVAL);
1919         }
1920         xprt->servername = kstrdup(args->servername, GFP_KERNEL);
1921         if (xprt->servername == NULL) {
1922                 xprt_destroy(xprt);
1923                 return ERR_PTR(-ENOMEM);
1924         }
1925
1926         rpc_xprt_debugfs_register(xprt);
1927
1928         dprintk("RPC:       created transport %p with %u slots\n", xprt,
1929                         xprt->max_reqs);
1930 out:
1931         return xprt;
1932 }
1933
1934 static void xprt_destroy_cb(struct work_struct *work)
1935 {
1936         struct rpc_xprt *xprt =
1937                 container_of(work, struct rpc_xprt, task_cleanup);
1938
1939         rpc_xprt_debugfs_unregister(xprt);
1940         rpc_destroy_wait_queue(&xprt->binding);
1941         rpc_destroy_wait_queue(&xprt->pending);
1942         rpc_destroy_wait_queue(&xprt->sending);
1943         rpc_destroy_wait_queue(&xprt->backlog);
1944         kfree(xprt->servername);
1945         /*
1946          * Destroy any existing back channel
1947          */
1948         xprt_destroy_backchannel(xprt, UINT_MAX);
1949
1950         /*
1951          * Tear down transport state and free the rpc_xprt
1952          */
1953         xprt->ops->destroy(xprt);
1954 }
1955
1956 /**
1957  * xprt_destroy - destroy an RPC transport, killing off all requests.
1958  * @xprt: transport to destroy
1959  *
1960  */
1961 static void xprt_destroy(struct rpc_xprt *xprt)
1962 {
1963         dprintk("RPC:       destroying transport %p\n", xprt);
1964
1965         /*
1966          * Exclude transport connect/disconnect handlers and autoclose
1967          */
1968         wait_on_bit_lock(&xprt->state, XPRT_LOCKED, TASK_UNINTERRUPTIBLE);
1969
1970         del_timer_sync(&xprt->timer);
1971
1972         /*
1973          * Destroy sockets etc from the system workqueue so they can
1974          * safely flush receive work running on rpciod.
1975          */
1976         INIT_WORK(&xprt->task_cleanup, xprt_destroy_cb);
1977         schedule_work(&xprt->task_cleanup);
1978 }
1979
1980 static void xprt_destroy_kref(struct kref *kref)
1981 {
1982         xprt_destroy(container_of(kref, struct rpc_xprt, kref));
1983 }
1984
1985 /**
1986  * xprt_get - return a reference to an RPC transport.
1987  * @xprt: pointer to the transport
1988  *
1989  */
1990 struct rpc_xprt *xprt_get(struct rpc_xprt *xprt)
1991 {
1992         if (xprt != NULL && kref_get_unless_zero(&xprt->kref))
1993                 return xprt;
1994         return NULL;
1995 }
1996 EXPORT_SYMBOL_GPL(xprt_get);
1997
1998 /**
1999  * xprt_put - release a reference to an RPC transport.
2000  * @xprt: pointer to the transport
2001  *
2002  */
2003 void xprt_put(struct rpc_xprt *xprt)
2004 {
2005         if (xprt != NULL)
2006                 kref_put(&xprt->kref, xprt_destroy_kref);
2007 }
2008 EXPORT_SYMBOL_GPL(xprt_put);