net/sunrpc/sched.c

   1 /*
   2  * linux/net/sunrpc/sched.c
   3  *
   4  * Scheduling for synchronous and asynchronous RPC requests.
   5  *
   6  * Copyright (C) 1996 Olaf Kirch, <okir@monad.swb.de>
   7  *
   8  * TCP NFS related read + write fixes
   9  * (C) 1999 Dave Airlie, University of Limerick, Ireland <airlied@linux.ie>
  10  */
  11
  12 #include <linux/module.h>
  13
  14 #include <linux/sched.h>
  15 #include <linux/interrupt.h>
  16 #include <linux/slab.h>
  17 #include <linux/mempool.h>
  18 #include <linux/smp.h>
  19 #include <linux/spinlock.h>
  20 #include <linux/mutex.h>
  21
  22 #include <linux/sunrpc/clnt.h>
  23
  24 #include "sunrpc.h"
  25
  26 #ifdef RPC_DEBUG
  27 #define RPCDBG_FACILITY         RPCDBG_SCHED
  28 #endif
  29
  30 /*
  31  * RPC slabs and memory pools
  32  */
  33 #define RPC_BUFFER_MAXSIZE      (2048)
  34 #define RPC_BUFFER_POOLSIZE     (8)
  35 #define RPC_TASK_POOLSIZE       (8)
  36 static struct kmem_cache        *rpc_task_slabp __read_mostly;
  37 static struct kmem_cache        *rpc_buffer_slabp __read_mostly;
  38 static mempool_t        *rpc_task_mempool __read_mostly;
  39 static mempool_t        *rpc_buffer_mempool __read_mostly;
  40
  41 static void                     rpc_async_schedule(struct work_struct *);
  42 static void                      rpc_release_task(struct rpc_task *task);
  43 static void __rpc_queue_timer_fn(unsigned long ptr);
  44
  45 /*
  46  * RPC tasks sit here while waiting for conditions to improve.
  47  */
  48 static struct rpc_wait_queue delay_queue;
  49
  50 /*
  51  * rpciod-related stuff
  52  */
  53 struct workqueue_struct *rpciod_workqueue;
  54
  55 /*
  56  * Disable the timer for a given RPC task. Should be called with
  57  * queue->lock and bh_disabled in order to avoid races within
  58  * rpc_run_timer().
  59  */
  60 static void
  61 __rpc_disable_timer(struct rpc_wait_queue *queue, struct rpc_task *task)
  62 {
  63         if (task->tk_timeout == 0)
  64                 return;
  65         dprintk("RPC: %5u disabling timer\n", task->tk_pid);
  66         task->tk_timeout = 0;
  67         list_del(&task->u.tk_wait.timer_list);
  68         if (list_empty(&queue->timer_list.list))
  69                 del_timer(&queue->timer_list.timer);
  70 }
  71
  72 static void
  73 rpc_set_queue_timer(struct rpc_wait_queue *queue, unsigned long expires)
  74 {
  75         queue->timer_list.expires = expires;
  76         mod_timer(&queue->timer_list.timer, expires);
  77 }
  78
  79 /*
  80  * Set up a timer for the current task.
  81  */
  82 static void
  83 __rpc_add_timer(struct rpc_wait_queue *queue, struct rpc_task *task)
  84 {
  85         if (!task->tk_timeout)
  86                 return;
  87
  88         dprintk("RPC: %5u setting alarm for %lu ms\n",
  89                         task->tk_pid, task->tk_timeout * 1000 / HZ);
  90
  91         task->u.tk_wait.expires = jiffies + task->tk_timeout;
  92         if (list_empty(&queue->timer_list.list) || time_before(task->u.tk_wait.expires, queue->timer_list.expires))
  93                 rpc_set_queue_timer(queue, task->u.tk_wait.expires);
  94         list_add(&task->u.tk_wait.timer_list, &queue->timer_list.list);
  95 }
  96
  97 /*
  98  * Add new request to a priority queue.
  99  */
 100 static void __rpc_add_wait_queue_priority(struct rpc_wait_queue *queue, struct rpc_task *task)
 101 {
 102         struct list_head *q;
 103         struct rpc_task *t;
 104
 105         INIT_LIST_HEAD(&task->u.tk_wait.links);
 106         q = &queue->tasks[task->tk_priority];
 107         if (unlikely(task->tk_priority > queue->maxpriority))
 108                 q = &queue->tasks[queue->maxpriority];
 109         list_for_each_entry(t, q, u.tk_wait.list) {
 110                 if (t->tk_owner == task->tk_owner) {
 111                         list_add_tail(&task->u.tk_wait.list, &t->u.tk_wait.links);
 112                         return;
 113                 }
 114         }
 115         list_add_tail(&task->u.tk_wait.list, q);
 116 }
 117
 118 /*
 119  * Add new request to wait queue.
 120  *
 121  * Swapper tasks always get inserted at the head of the queue.
 122  * This should avoid many nasty memory deadlocks and hopefully
 123  * improve overall performance.
 124  * Everyone else gets appended to the queue to ensure proper FIFO behavior.
 125  */
 126 static void __rpc_add_wait_queue(struct rpc_wait_queue *queue, struct rpc_task *task)
 127 {
 128         BUG_ON (RPC_IS_QUEUED(task));
 129
 130         if (RPC_IS_PRIORITY(queue))
 131                 __rpc_add_wait_queue_priority(queue, task);
 132         else if (RPC_IS_SWAPPER(task))
 133                 list_add(&task->u.tk_wait.list, &queue->tasks[0]);
 134         else
 135                 list_add_tail(&task->u.tk_wait.list, &queue->tasks[0]);
 136         task->tk_waitqueue = queue;
 137         queue->qlen++;
 138         rpc_set_queued(task);
 139
 140         dprintk("RPC: %5u added to queue %p \"%s\"\n",
 141                         task->tk_pid, queue, rpc_qname(queue));
 142 }
 143
 144 /*
 145  * Remove request from a priority queue.
 146  */
 147 static void __rpc_remove_wait_queue_priority(struct rpc_task *task)
 148 {
 149         struct rpc_task *t;
 150
 151         if (!list_empty(&task->u.tk_wait.links)) {
 152                 t = list_entry(task->u.tk_wait.links.next, struct rpc_task, u.tk_wait.list);
 153                 list_move(&t->u.tk_wait.list, &task->u.tk_wait.list);
 154                 list_splice_init(&task->u.tk_wait.links, &t->u.tk_wait.links);
 155         }
 156 }
 157
 158 /*
 159  * Remove request from queue.
 160  * Note: must be called with spin lock held.
 161  */
 162 static void __rpc_remove_wait_queue(struct rpc_wait_queue *queue, struct rpc_task *task)
 163 {
 164         __rpc_disable_timer(queue, task);
 165         if (RPC_IS_PRIORITY(queue))
 166                 __rpc_remove_wait_queue_priority(task);
 167         list_del(&task->u.tk_wait.list);
 168         queue->qlen--;
 169         dprintk("RPC: %5u removed from queue %p \"%s\"\n",
 170                         task->tk_pid, queue, rpc_qname(queue));
 171 }
 172
 173 static inline void rpc_set_waitqueue_priority(struct rpc_wait_queue *queue, int priority)
 174 {
 175         queue->priority = priority;
 176         queue->count = 1 << (priority * 2);
 177 }
 178
 179 static inline void rpc_set_waitqueue_owner(struct rpc_wait_queue *queue, pid_t pid)
 180 {
 181         queue->owner = pid;
 182         queue->nr = RPC_BATCH_COUNT;
 183 }
 184
 185 static inline void rpc_reset_waitqueue_priority(struct rpc_wait_queue *queue)
 186 {
 187         rpc_set_waitqueue_priority(queue, queue->maxpriority);
 188         rpc_set_waitqueue_owner(queue, 0);
 189 }
 190
 191 static void __rpc_init_priority_wait_queue(struct rpc_wait_queue *queue, const char *qname, unsigned char nr_queues)
 192 {
 193         int i;
 194
 195         spin_lock_init(&queue->lock);
 196         for (i = 0; i < ARRAY_SIZE(queue->tasks); i++)
 197                 INIT_LIST_HEAD(&queue->tasks[i]);
 198         queue->maxpriority = nr_queues - 1;
 199         rpc_reset_waitqueue_priority(queue);
 200         queue->qlen = 0;
 201         setup_timer(&queue->timer_list.timer, __rpc_queue_timer_fn, (unsigned long)queue);
 202         INIT_LIST_HEAD(&queue->timer_list.list);
 203 #ifdef RPC_DEBUG
 204         queue->name = qname;
 205 #endif
 206 }
 207
 208 void rpc_init_priority_wait_queue(struct rpc_wait_queue *queue, const char *qname)
 209 {
 210         __rpc_init_priority_wait_queue(queue, qname, RPC_NR_PRIORITY);
 211 }
 212 EXPORT_SYMBOL_GPL(rpc_init_priority_wait_queue);
 213
 214 void rpc_init_wait_queue(struct rpc_wait_queue *queue, const char *qname)
 215 {
 216         __rpc_init_priority_wait_queue(queue, qname, 1);
 217 }
 218 EXPORT_SYMBOL_GPL(rpc_init_wait_queue);
 219
 220 void rpc_destroy_wait_queue(struct rpc_wait_queue *queue)
 221 {
 222         del_timer_sync(&queue->timer_list.timer);
 223 }
 224 EXPORT_SYMBOL_GPL(rpc_destroy_wait_queue);
 225
 226 static int rpc_wait_bit_killable(void *word)
 227 {
 228         if (fatal_signal_pending(current))
 229                 return -ERESTARTSYS;
 230         schedule();
 231         return 0;
 232 }
 233
 234 #ifdef RPC_DEBUG
 235 static void rpc_task_set_debuginfo(struct rpc_task *task)
 236 {
 237         static atomic_t rpc_pid;
 238
 239         task->tk_pid = atomic_inc_return(&rpc_pid);
 240 }
 241 #else
 242 static inline void rpc_task_set_debuginfo(struct rpc_task *task)
 243 {
 244 }
 245 #endif
 246
 247 static void rpc_set_active(struct rpc_task *task)
 248 {
 249         rpc_task_set_debuginfo(task);
 250         set_bit(RPC_TASK_ACTIVE, &task->tk_runstate);
 251 }
 252
 253 /*
 254  * Mark an RPC call as having completed by clearing the 'active' bit
 255  */
 256 static void rpc_mark_complete_task(struct rpc_task *task)
 257 {
 258         smp_mb__before_clear_bit();
 259         clear_bit(RPC_TASK_ACTIVE, &task->tk_runstate);
 260         smp_mb__after_clear_bit();
 261         wake_up_bit(&task->tk_runstate, RPC_TASK_ACTIVE);
 262 }
 263
 264 /*
 265  * Allow callers to wait for completion of an RPC call
 266  */
 267 int __rpc_wait_for_completion_task(struct rpc_task *task, int (*action)(void *))
 268 {
 269         if (action == NULL)
 270                 action = rpc_wait_bit_killable;
 271         return wait_on_bit(&task->tk_runstate, RPC_TASK_ACTIVE,
 272                         action, TASK_KILLABLE);
 273 }
 274 EXPORT_SYMBOL_GPL(__rpc_wait_for_completion_task);
 275
 276 /*
 277  * Make an RPC task runnable.
 278  *
 279  * Note: If the task is ASYNC, this must be called with
 280  * the spinlock held to protect the wait queue operation.
 281  */
 282 static void rpc_make_runnable(struct rpc_task *task)
 283 {
 284         rpc_clear_queued(task);
 285         if (rpc_test_and_set_running(task))
 286                 return;
 287         if (RPC_IS_ASYNC(task)) {
 288                 int status;
 289
 290                 INIT_WORK(&task->u.tk_work, rpc_async_schedule);
 291                 status = queue_work(rpciod_workqueue, &task->u.tk_work);
 292                 if (status < 0) {
 293                         printk(KERN_WARNING "RPC: failed to add task to queue: error: %d!\n", status);
 294                         task->tk_status = status;
 295                         return;
 296                 }
 297         } else
 298                 wake_up_bit(&task->tk_runstate, RPC_TASK_QUEUED);
 299 }
 300
 301 /*
 302  * Prepare for sleeping on a wait queue.
 303  * By always appending tasks to the list we ensure FIFO behavior.
 304  * NB: An RPC task will only receive interrupt-driven events as long
 305  * as it's on a wait queue.
 306  */
 307 static void __rpc_sleep_on(struct rpc_wait_queue *q, struct rpc_task *task,
 308                         rpc_action action)
 309 {
 310         dprintk("RPC: %5u sleep_on(queue \"%s\" time %lu)\n",
 311                         task->tk_pid, rpc_qname(q), jiffies);
 312
 313         __rpc_add_wait_queue(q, task);
 314
 315         BUG_ON(task->tk_callback != NULL);
 316         task->tk_callback = action;
 317         __rpc_add_timer(q, task);
 318 }
 319
 320 void rpc_sleep_on(struct rpc_wait_queue *q, struct rpc_task *task,
 321                                 rpc_action action)
 322 {
 323         /* We shouldn't ever put an inactive task to sleep */
 324         BUG_ON(!RPC_IS_ACTIVATED(task));
 325
 326         /*
 327          * Protect the queue operations.
 328          */
 329         spin_lock_bh(&q->lock);
 330         __rpc_sleep_on(q, task, action);
 331         spin_unlock_bh(&q->lock);
 332 }
 333 EXPORT_SYMBOL_GPL(rpc_sleep_on);
 334
 335 /**
 336  * __rpc_do_wake_up_task - wake up a single rpc_task
 337  * @queue: wait queue
 338  * @task: task to be woken up
 339  *
 340  * Caller must hold queue->lock, and have cleared the task queued flag.
 341  */
 342 static void __rpc_do_wake_up_task(struct rpc_wait_queue *queue, struct rpc_task *task)
 343 {
 344         dprintk("RPC: %5u __rpc_wake_up_task (now %lu)\n",
 345                         task->tk_pid, jiffies);
 346
 347         /* Has the task been executed yet? If not, we cannot wake it up! */
 348         if (!RPC_IS_ACTIVATED(task)) {
 349                 printk(KERN_ERR "RPC: Inactive task (%p) being woken up!\n", task);
 350                 return;
 351         }
 352
 353         __rpc_remove_wait_queue(queue, task);
 354
 355         rpc_make_runnable(task);
 356
 357         dprintk("RPC:       __rpc_wake_up_task done\n");
 358 }
 359
 360 /*
 361  * Wake up a queued task while the queue lock is being held
 362  */
 363 static void rpc_wake_up_task_queue_locked(struct rpc_wait_queue *queue, struct rpc_task *task)
 364 {
 365         if (RPC_IS_QUEUED(task) && task->tk_waitqueue == queue)
 366                 __rpc_do_wake_up_task(queue, task);
 367 }
 368
 369 /*
 370  * Tests whether rpc queue is empty
 371  */
 372 int rpc_queue_empty(struct rpc_wait_queue *queue)
 373 {
 374         int res;
 375
 376         spin_lock_bh(&queue->lock);
 377         res = queue->qlen;
 378         spin_unlock_bh(&queue->lock);
 379         return res == 0;
 380 }
 381 EXPORT_SYMBOL_GPL(rpc_queue_empty);
 382
 383 /*
 384  * Wake up a task on a specific queue
 385  */
 386 void rpc_wake_up_queued_task(struct rpc_wait_queue *queue, struct rpc_task *task)
 387 {
 388         spin_lock_bh(&queue->lock);
 389         rpc_wake_up_task_queue_locked(queue, task);
 390         spin_unlock_bh(&queue->lock);
 391 }
 392 EXPORT_SYMBOL_GPL(rpc_wake_up_queued_task);
 393
 394 /*
 395  * Wake up the next task on a priority queue.
 396  */
 397 static struct rpc_task * __rpc_wake_up_next_priority(struct rpc_wait_queue *queue)
 398 {
 399         struct list_head *q;
 400         struct rpc_task *task;
 401
 402         /*
 403          * Service a batch of tasks from a single owner.
 404          */
 405         q = &queue->tasks[queue->priority];
 406         if (!list_empty(q)) {
 407                 task = list_entry(q->next, struct rpc_task, u.tk_wait.list);
 408                 if (queue->owner == task->tk_owner) {
 409                         if (--queue->nr)
 410                                 goto out;
 411                         list_move_tail(&task->u.tk_wait.list, q);
 412                 }
 413                 /*
 414                  * Check if we need to switch queues.
 415                  */
 416                 if (--queue->count)
 417                         goto new_owner;
 418         }
 419
 420         /*
 421          * Service the next queue.
 422          */
 423         do {
 424                 if (q == &queue->tasks[0])
 425                         q = &queue->tasks[queue->maxpriority];
 426                 else
 427                         q = q - 1;
 428                 if (!list_empty(q)) {
 429                         task = list_entry(q->next, struct rpc_task, u.tk_wait.list);
 430                         goto new_queue;
 431                 }
 432         } while (q != &queue->tasks[queue->priority]);
 433
 434         rpc_reset_waitqueue_priority(queue);
 435         return NULL;
 436
 437 new_queue:
 438         rpc_set_waitqueue_priority(queue, (unsigned int)(q - &queue->tasks[0]));
 439 new_owner:
 440         rpc_set_waitqueue_owner(queue, task->tk_owner);
 441 out:
 442         rpc_wake_up_task_queue_locked(queue, task);
 443         return task;
 444 }
 445
 446 /*
 447  * Wake up the next task on the wait queue.
 448  */
 449 struct rpc_task * rpc_wake_up_next(struct rpc_wait_queue *queue)
 450 {
 451         struct rpc_task *task = NULL;
 452
 453         dprintk("RPC:       wake_up_next(%p \"%s\")\n",
 454                         queue, rpc_qname(queue));
 455         spin_lock_bh(&queue->lock);
 456         if (RPC_IS_PRIORITY(queue))
 457                 task = __rpc_wake_up_next_priority(queue);
 458         else {
 459                 task_for_first(task, &queue->tasks[0])
 460                         rpc_wake_up_task_queue_locked(queue, task);
 461         }
 462         spin_unlock_bh(&queue->lock);
 463
 464         return task;
 465 }
 466 EXPORT_SYMBOL_GPL(rpc_wake_up_next);
 467
 468 /**
 469  * rpc_wake_up - wake up all rpc_tasks
 470  * @queue: rpc_wait_queue on which the tasks are sleeping
 471  *
 472  * Grabs queue->lock
 473  */
 474 void rpc_wake_up(struct rpc_wait_queue *queue)
 475 {
 476         struct rpc_task *task, *next;
 477         struct list_head *head;
 478
 479         spin_lock_bh(&queue->lock);
 480         head = &queue->tasks[queue->maxpriority];
 481         for (;;) {
 482                 list_for_each_entry_safe(task, next, head, u.tk_wait.list)
 483                         rpc_wake_up_task_queue_locked(queue, task);
 484                 if (head == &queue->tasks[0])
 485                         break;
 486                 head--;
 487         }
 488         spin_unlock_bh(&queue->lock);
 489 }
 490 EXPORT_SYMBOL_GPL(rpc_wake_up);
 491
 492 /**
 493  * rpc_wake_up_status - wake up all rpc_tasks and set their status value.
 494  * @queue: rpc_wait_queue on which the tasks are sleeping
 495  * @status: status value to set
 496  *
 497  * Grabs queue->lock
 498  */
 499 void rpc_wake_up_status(struct rpc_wait_queue *queue, int status)
 500 {
 501         struct rpc_task *task, *next;
 502         struct list_head *head;
 503
 504         spin_lock_bh(&queue->lock);
 505         head = &queue->tasks[queue->maxpriority];
 506         for (;;) {
 507                 list_for_each_entry_safe(task, next, head, u.tk_wait.list) {
 508                         task->tk_status = status;
 509                         rpc_wake_up_task_queue_locked(queue, task);
 510                 }
 511                 if (head == &queue->tasks[0])
 512                         break;
 513                 head--;
 514         }
 515         spin_unlock_bh(&queue->lock);
 516 }
 517 EXPORT_SYMBOL_GPL(rpc_wake_up_status);
 518
 519 static void __rpc_queue_timer_fn(unsigned long ptr)
 520 {
 521         struct rpc_wait_queue *queue = (struct rpc_wait_queue *)ptr;
 522         struct rpc_task *task, *n;
 523         unsigned long expires, now, timeo;
 524
 525         spin_lock(&queue->lock);
 526         expires = now = jiffies;
 527         list_for_each_entry_safe(task, n, &queue->timer_list.list, u.tk_wait.timer_list) {
 528                 timeo = task->u.tk_wait.expires;
 529                 if (time_after_eq(now, timeo)) {
 530                         dprintk("RPC: %5u timeout\n", task->tk_pid);
 531                         task->tk_status = -ETIMEDOUT;
 532                         rpc_wake_up_task_queue_locked(queue, task);
 533                         continue;
 534                 }
 535                 if (expires == now || time_after(expires, timeo))
 536                         expires = timeo;
 537         }
 538         if (!list_empty(&queue->timer_list.list))
 539                 rpc_set_queue_timer(queue, expires);
 540         spin_unlock(&queue->lock);
 541 }
 542
 543 static void __rpc_atrun(struct rpc_task *task)
 544 {
 545         task->tk_status = 0;
 546 }
 547
 548 /*
 549  * Run a task at a later time
 550  */
 551 void rpc_delay(struct rpc_task *task, unsigned long delay)
 552 {
 553         task->tk_timeout = delay;
 554         rpc_sleep_on(&delay_queue, task, __rpc_atrun);
 555 }
 556 EXPORT_SYMBOL_GPL(rpc_delay);
 557
 558 /*
 559  * Helper to call task->tk_ops->rpc_call_prepare
 560  */
 561 void rpc_prepare_task(struct rpc_task *task)
 562 {
 563         task->tk_ops->rpc_call_prepare(task, task->tk_calldata);
 564 }
 565
 566 /*
 567  * Helper that calls task->tk_ops->rpc_call_done if it exists
 568  */
 569 void rpc_exit_task(struct rpc_task *task)
 570 {
 571         task->tk_action = NULL;
 572         if (task->tk_ops->rpc_call_done != NULL) {
 573                 task->tk_ops->rpc_call_done(task, task->tk_calldata);
 574                 if (task->tk_action != NULL) {
 575                         WARN_ON(RPC_ASSASSINATED(task));
 576                         /* Always release the RPC slot and buffer memory */
 577                         xprt_release(task);
 578                 }
 579         }
 580 }
 581
 582 void rpc_exit(struct rpc_task *task, int status)
 583 {
 584         task->tk_status = status;
 585         task->tk_action = rpc_exit_task;
 586         if (RPC_IS_QUEUED(task))
 587                 rpc_wake_up_queued_task(task->tk_waitqueue, task);
 588 }
 589 EXPORT_SYMBOL_GPL(rpc_exit);
 590
 591 void rpc_release_calldata(const struct rpc_call_ops *ops, void *calldata)
 592 {
 593         if (ops->rpc_release != NULL)
 594                 ops->rpc_release(calldata);
 595 }
 596
 597 /*
 598  * This is the RPC `scheduler' (or rather, the finite state machine).
 599  */
 600 static void __rpc_execute(struct rpc_task *task)
 601 {
 602         struct rpc_wait_queue *queue;
 603         int task_is_async = RPC_IS_ASYNC(task);
 604         int status = 0;
 605
 606         dprintk("RPC: %5u __rpc_execute flags=0x%x\n",
 607                         task->tk_pid, task->tk_flags);
 608
 609         BUG_ON(RPC_IS_QUEUED(task));
 610
 611         for (;;) {
 612
 613                 /*
 614                  * Execute any pending callback.
 615                  */
 616                 if (task->tk_callback) {
 617                         void (*save_callback)(struct rpc_task *);
 618
 619                         /*
 620                          * We set tk_callback to NULL before calling it,
 621                          * in case it sets the tk_callback field itself:
 622                          */
 623                         save_callback = task->tk_callback;
 624                         task->tk_callback = NULL;
 625                         save_callback(task);
 626                 }
 627
 628                 /*
 629                  * Perform the next FSM step.
 630                  * tk_action may be NULL when the task has been killed
 631                  * by someone else.
 632                  */
 633                 if (!RPC_IS_QUEUED(task)) {
 634                         if (task->tk_action == NULL)
 635                                 break;
 636                         task->tk_action(task);
 637                 }
 638
 639                 /*
 640                  * Lockless check for whether task is sleeping or not.
 641                  */
 642                 if (!RPC_IS_QUEUED(task))
 643                         continue;
 644                 /*
 645                  * The queue->lock protects against races with
 646                  * rpc_make_runnable().
 647                  *
 648                  * Note that once we clear RPC_TASK_RUNNING on an asynchronous
 649                  * rpc_task, rpc_make_runnable() can assign it to a
 650                  * different workqueue. We therefore cannot assume that the
 651                  * rpc_task pointer may still be dereferenced.
 652                  */
 653                 queue = task->tk_waitqueue;
 654                 spin_lock_bh(&queue->lock);
 655                 if (!RPC_IS_QUEUED(task)) {
 656                         spin_unlock_bh(&queue->lock);
 657                         continue;
 658                 }
 659                 rpc_clear_running(task);
 660                 spin_unlock_bh(&queue->lock);
 661                 if (task_is_async)
 662                         return;
 663
 664                 /* sync task: sleep here */
 665                 dprintk("RPC: %5u sync task going to sleep\n", task->tk_pid);
 666                 status = out_of_line_wait_on_bit(&task->tk_runstate,
 667                                 RPC_TASK_QUEUED, rpc_wait_bit_killable,
 668                                 TASK_KILLABLE);
 669                 if (status == -ERESTARTSYS) {
 670                         /*
 671                          * When a sync task receives a signal, it exits with
 672                          * -ERESTARTSYS. In order to catch any callbacks that
 673                          * clean up after sleeping on some queue, we don't
 674                          * break the loop here, but go around once more.
 675                          */
 676                         dprintk("RPC: %5u got signal\n", task->tk_pid);
 677                         task->tk_flags |= RPC_TASK_KILLED;
 678                         rpc_exit(task, -ERESTARTSYS);
 679                 }
 680                 rpc_set_running(task);
 681                 dprintk("RPC: %5u sync task resuming\n", task->tk_pid);
 682         }
 683
 684         dprintk("RPC: %5u return %d, status %d\n", task->tk_pid, status,
 685                         task->tk_status);
 686         /* Release all resources associated with the task */
 687         rpc_release_task(task);
 688 }
 689
 690 /*
 691  * User-visible entry point to the scheduler.
 692  *
 693  * This may be called recursively if e.g. an async NFS task updates
 694  * the attributes and finds that dirty pages must be flushed.
 695  * NOTE: Upon exit of this function the task is guaranteed to be
 696  *       released. In particular note that tk_release() will have
 697  *       been called, so your task memory may have been freed.
 698  */
 699 void rpc_execute(struct rpc_task *task)
 700 {
 701         rpc_set_active(task);
 702         rpc_make_runnable(task);
 703         if (!RPC_IS_ASYNC(task))
 704                 __rpc_execute(task);
 705 }
 706
 707 static void rpc_async_schedule(struct work_struct *work)
 708 {
 709         __rpc_execute(container_of(work, struct rpc_task, u.tk_work));
 710 }
 711
 712 /**
 713  * rpc_malloc - allocate an RPC buffer
 714  * @task: RPC task that will use this buffer
 715  * @size: requested byte size
 716  *
 717  * To prevent rpciod from hanging, this allocator never sleeps,
 718  * returning NULL if the request cannot be serviced immediately.
 719  * The caller can arrange to sleep in a way that is safe for rpciod.
 720  *
 721  * Most requests are 'small' (under 2KiB) and can be serviced from a
 722  * mempool, ensuring that NFS reads and writes can always proceed,
 723  * and that there is good locality of reference for these buffers.
 724  *
 725  * In order to avoid memory starvation triggering more writebacks of
 726  * NFS requests, we avoid using GFP_KERNEL.
 727  */
 728 void *rpc_malloc(struct rpc_task *task, size_t size)
 729 {
 730         struct rpc_buffer *buf;
 731         gfp_t gfp = RPC_IS_SWAPPER(task) ? GFP_ATOMIC : GFP_NOWAIT;
 732
 733         size += sizeof(struct rpc_buffer);
 734         if (size <= RPC_BUFFER_MAXSIZE)
 735                 buf = mempool_alloc(rpc_buffer_mempool, gfp);
 736         else
 737                 buf = kmalloc(size, gfp);
 738
 739         if (!buf)
 740                 return NULL;
 741
 742         buf->len = size;
 743         dprintk("RPC: %5u allocated buffer of size %zu at %p\n",
 744                         task->tk_pid, size, buf);
 745         return &buf->data;
 746 }
 747 EXPORT_SYMBOL_GPL(rpc_malloc);
 748
 749 /**
 750  * rpc_free - free buffer allocated via rpc_malloc
 751  * @buffer: buffer to free
 752  *
 753  */
 754 void rpc_free(void *buffer)
 755 {
 756         size_t size;
 757         struct rpc_buffer *buf;
 758
 759         if (!buffer)
 760                 return;
 761
 762         buf = container_of(buffer, struct rpc_buffer, data);
 763         size = buf->len;
 764
 765         dprintk("RPC:       freeing buffer of size %zu at %p\n",
 766                         size, buf);
 767
 768         if (size <= RPC_BUFFER_MAXSIZE)
 769                 mempool_free(buf, rpc_buffer_mempool);
 770         else
 771                 kfree(buf);
 772 }
 773 EXPORT_SYMBOL_GPL(rpc_free);
 774
 775 /*
 776  * Creation and deletion of RPC task structures
 777  */
 778 static void rpc_init_task(struct rpc_task *task, const struct rpc_task_setup *task_setup_data)
 779 {
 780         memset(task, 0, sizeof(*task));
 781         atomic_set(&task->tk_count, 1);
 782         task->tk_flags  = task_setup_data->flags;
 783         task->tk_ops = task_setup_data->callback_ops;
 784         task->tk_calldata = task_setup_data->callback_data;
 785         INIT_LIST_HEAD(&task->tk_task);
 786
 787         /* Initialize retry counters */
 788         task->tk_garb_retry = 2;
 789         task->tk_cred_retry = 2;
 790
 791         task->tk_priority = task_setup_data->priority - RPC_PRIORITY_LOW;
 792         task->tk_owner = current->tgid;
 793
 794         /* Initialize workqueue for async tasks */
 795         task->tk_workqueue = task_setup_data->workqueue;
 796
 797         if (task->tk_ops->rpc_call_prepare != NULL)
 798                 task->tk_action = rpc_prepare_task;
 799
 800         /* starting timestamp */
 801         task->tk_start = ktime_get();
 802
 803         dprintk("RPC:       new task initialized, procpid %u\n",
 804                                 task_pid_nr(current));
 805 }
 806
 807 static struct rpc_task *
 808 rpc_alloc_task(void)
 809 {
 810         return (struct rpc_task *)mempool_alloc(rpc_task_mempool, GFP_NOFS);
 811 }
 812
 813 /*
 814  * Create a new task for the specified client.
 815  */
 816 struct rpc_task *rpc_new_task(const struct rpc_task_setup *setup_data)
 817 {
 818         struct rpc_task *task = setup_data->task;
 819         unsigned short flags = 0;
 820
 821         if (task == NULL) {
 822                 task = rpc_alloc_task();
 823                 if (task == NULL) {
 824                         rpc_release_calldata(setup_data->callback_ops,
 825                                         setup_data->callback_data);
 826                         return ERR_PTR(-ENOMEM);
 827                 }
 828                 flags = RPC_TASK_DYNAMIC;
 829         }
 830
 831         rpc_init_task(task, setup_data);
 832         if (task->tk_status < 0) {
 833                 int err = task->tk_status;
 834                 rpc_put_task(task);
 835                 return ERR_PTR(err);
 836         }
 837
 838         task->tk_flags |= flags;
 839         dprintk("RPC:       allocated task %p\n", task);
 840         return task;
 841 }
 842
 843 static void rpc_free_task(struct rpc_task *task)
 844 {
 845         const struct rpc_call_ops *tk_ops = task->tk_ops;
 846         void *calldata = task->tk_calldata;
 847
 848         if (task->tk_flags & RPC_TASK_DYNAMIC) {
 849                 dprintk("RPC: %5u freeing task\n", task->tk_pid);
 850                 mempool_free(task, rpc_task_mempool);
 851         }
 852         rpc_release_calldata(tk_ops, calldata);
 853 }
 854
 855 static void rpc_async_release(struct work_struct *work)
 856 {
 857         rpc_free_task(container_of(work, struct rpc_task, u.tk_work));
 858 }
 859
 860 void rpc_put_task(struct rpc_task *task)
 861 {
 862         if (!atomic_dec_and_test(&task->tk_count))
 863                 return;
 864         /* Release resources */
 865         if (task->tk_rqstp)
 866                 xprt_release(task);
 867         if (task->tk_msg.rpc_cred)
 868                 put_rpccred(task->tk_msg.rpc_cred);
 869         rpc_task_release_client(task);
 870         if (task->tk_workqueue != NULL) {
 871                 INIT_WORK(&task->u.tk_work, rpc_async_release);
 872                 queue_work(task->tk_workqueue, &task->u.tk_work);
 873         } else
 874                 rpc_free_task(task);
 875 }
 876 EXPORT_SYMBOL_GPL(rpc_put_task);
 877
 878 static void rpc_release_task(struct rpc_task *task)
 879 {
 880         dprintk("RPC: %5u release task\n", task->tk_pid);
 881
 882         BUG_ON (RPC_IS_QUEUED(task));
 883
 884         /* Wake up anyone who is waiting for task completion */
 885         rpc_mark_complete_task(task);
 886
 887         rpc_put_task(task);
 888 }
 889
 890 int rpciod_up(void)
 891 {
 892         return try_module_get(THIS_MODULE) ? 0 : -EINVAL;
 893 }
 894
 895 void rpciod_down(void)
 896 {
 897         module_put(THIS_MODULE);
 898 }
 899
 900 /*
 901  * Start up the rpciod workqueue.
 902  */
 903 static int rpciod_start(void)
 904 {
 905         struct workqueue_struct *wq;
 906
 907         /*
 908          * Create the rpciod thread and wait for it to start.
 909          */
 910         dprintk("RPC:       creating workqueue rpciod\n");
 911         wq = alloc_workqueue("rpciod", WQ_RESCUER, 0);
 912         rpciod_workqueue = wq;
 913         return rpciod_workqueue != NULL;
 914 }
 915
 916 static void rpciod_stop(void)
 917 {
 918         struct workqueue_struct *wq = NULL;
 919
 920         if (rpciod_workqueue == NULL)
 921                 return;
 922         dprintk("RPC:       destroying workqueue rpciod\n");
 923
 924         wq = rpciod_workqueue;
 925         rpciod_workqueue = NULL;
 926         destroy_workqueue(wq);
 927 }
 928
 929 void
 930 rpc_destroy_mempool(void)
 931 {
 932         rpciod_stop();
 933         if (rpc_buffer_mempool)
 934                 mempool_destroy(rpc_buffer_mempool);
 935         if (rpc_task_mempool)
 936                 mempool_destroy(rpc_task_mempool);
 937         if (rpc_task_slabp)
 938                 kmem_cache_destroy(rpc_task_slabp);
 939         if (rpc_buffer_slabp)
 940                 kmem_cache_destroy(rpc_buffer_slabp);
 941         rpc_destroy_wait_queue(&delay_queue);
 942 }
 943
 944 int
 945 rpc_init_mempool(void)
 946 {
 947         /*
 948          * The following is not strictly a mempool initialisation,
 949          * but there is no harm in doing it here
 950          */
 951         rpc_init_wait_queue(&delay_queue, "delayq");
 952         if (!rpciod_start())
 953                 goto err_nomem;
 954
 955         rpc_task_slabp = kmem_cache_create("rpc_tasks",
 956                                              sizeof(struct rpc_task),
 957                                              0, SLAB_HWCACHE_ALIGN,
 958                                              NULL);
 959         if (!rpc_task_slabp)
 960                 goto err_nomem;
 961         rpc_buffer_slabp = kmem_cache_create("rpc_buffers",
 962                                              RPC_BUFFER_MAXSIZE,
 963                                              0, SLAB_HWCACHE_ALIGN,
 964                                              NULL);
 965         if (!rpc_buffer_slabp)
 966                 goto err_nomem;
 967         rpc_task_mempool = mempool_create_slab_pool(RPC_TASK_POOLSIZE,
 968                                                     rpc_task_slabp);
 969         if (!rpc_task_mempool)
 970                 goto err_nomem;
 971         rpc_buffer_mempool = mempool_create_slab_pool(RPC_BUFFER_POOLSIZE,
 972                                                       rpc_buffer_slabp);
 973         if (!rpc_buffer_mempool)
 974                 goto err_nomem;
 975         return 0;
 976 err_nomem:
 977         rpc_destroy_mempool();
 978         return -ENOMEM;
 979 }