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 #include <linux/freezer.h>
  22
  23 #include <linux/sunrpc/clnt.h>
  24
  25 #include "sunrpc.h"
  26
  27 #ifdef RPC_DEBUG
  28 #define RPCDBG_FACILITY         RPCDBG_SCHED
  29 #endif
  30
  31 #define CREATE_TRACE_POINTS
  32 #include <trace/events/sunrpc.h>
  33
  34 /*
  35  * RPC slabs and memory pools
  36  */
  37 #define RPC_BUFFER_MAXSIZE      (2048)
  38 #define RPC_BUFFER_POOLSIZE     (8)
  39 #define RPC_TASK_POOLSIZE       (8)
  40 static struct kmem_cache        *rpc_task_slabp __read_mostly;
  41 static struct kmem_cache        *rpc_buffer_slabp __read_mostly;
  42 static mempool_t        *rpc_task_mempool __read_mostly;
  43 static mempool_t        *rpc_buffer_mempool __read_mostly;
  44
  45 static void                     rpc_async_schedule(struct work_struct *);
  46 static void                      rpc_release_task(struct rpc_task *task);
  47 static void __rpc_queue_timer_fn(unsigned long ptr);
  48
  49 /*
  50  * RPC tasks sit here while waiting for conditions to improve.
  51  */
  52 static struct rpc_wait_queue delay_queue;
  53
  54 /*
  55  * rpciod-related stuff
  56  */
  57 struct workqueue_struct *rpciod_workqueue;
  58
  59 /*
  60  * Disable the timer for a given RPC task. Should be called with
  61  * queue->lock and bh_disabled in order to avoid races within
  62  * rpc_run_timer().
  63  */
  64 static void
  65 __rpc_disable_timer(struct rpc_wait_queue *queue, struct rpc_task *task)
  66 {
  67         if (task->tk_timeout == 0)
  68                 return;
  69         dprintk("RPC: %5u disabling timer\n", task->tk_pid);
  70         task->tk_timeout = 0;
  71         list_del(&task->u.tk_wait.timer_list);
  72         if (list_empty(&queue->timer_list.list))
  73                 del_timer(&queue->timer_list.timer);
  74 }
  75
  76 static void
  77 rpc_set_queue_timer(struct rpc_wait_queue *queue, unsigned long expires)
  78 {
  79         queue->timer_list.expires = expires;
  80         mod_timer(&queue->timer_list.timer, expires);
  81 }
  82
  83 /*
  84  * Set up a timer for the current task.
  85  */
  86 static void
  87 __rpc_add_timer(struct rpc_wait_queue *queue, struct rpc_task *task)
  88 {
  89         if (!task->tk_timeout)
  90                 return;
  91
  92         dprintk("RPC: %5u setting alarm for %lu ms\n",
  93                         task->tk_pid, task->tk_timeout * 1000 / HZ);
  94
  95         task->u.tk_wait.expires = jiffies + task->tk_timeout;
  96         if (list_empty(&queue->timer_list.list) || time_before(task->u.tk_wait.expires, queue->timer_list.expires))
  97                 rpc_set_queue_timer(queue, task->u.tk_wait.expires);
  98         list_add(&task->u.tk_wait.timer_list, &queue->timer_list.list);
  99 }
 100
 101 /*
 102  * Add new request to a priority queue.
 103  */
 104 static void __rpc_add_wait_queue_priority(struct rpc_wait_queue *queue,
 105                 struct rpc_task *task,
 106                 unsigned char queue_priority)
 107 {
 108         struct list_head *q;
 109         struct rpc_task *t;
 110
 111         INIT_LIST_HEAD(&task->u.tk_wait.links);
 112         q = &queue->tasks[queue_priority];
 113         if (unlikely(queue_priority > queue->maxpriority))
 114                 q = &queue->tasks[queue->maxpriority];
 115         list_for_each_entry(t, q, u.tk_wait.list) {
 116                 if (t->tk_owner == task->tk_owner) {
 117                         list_add_tail(&task->u.tk_wait.list, &t->u.tk_wait.links);
 118                         return;
 119                 }
 120         }
 121         list_add_tail(&task->u.tk_wait.list, q);
 122 }
 123
 124 /*
 125  * Add new request to wait queue.
 126  *
 127  * Swapper tasks always get inserted at the head of the queue.
 128  * This should avoid many nasty memory deadlocks and hopefully
 129  * improve overall performance.
 130  * Everyone else gets appended to the queue to ensure proper FIFO behavior.
 131  */
 132 static void __rpc_add_wait_queue(struct rpc_wait_queue *queue,
 133                 struct rpc_task *task,
 134                 unsigned char queue_priority)
 135 {
 136         BUG_ON (RPC_IS_QUEUED(task));
 137
 138         if (RPC_IS_PRIORITY(queue))
 139                 __rpc_add_wait_queue_priority(queue, task, queue_priority);
 140         else if (RPC_IS_SWAPPER(task))
 141                 list_add(&task->u.tk_wait.list, &queue->tasks[0]);
 142         else
 143                 list_add_tail(&task->u.tk_wait.list, &queue->tasks[0]);
 144         task->tk_waitqueue = queue;
 145         queue->qlen++;
 146         rpc_set_queued(task);
 147
 148         dprintk("RPC: %5u added to queue %p \"%s\"\n",
 149                         task->tk_pid, queue, rpc_qname(queue));
 150 }
 151
 152 /*
 153  * Remove request from a priority queue.
 154  */
 155 static void __rpc_remove_wait_queue_priority(struct rpc_task *task)
 156 {
 157         struct rpc_task *t;
 158
 159         if (!list_empty(&task->u.tk_wait.links)) {
 160                 t = list_entry(task->u.tk_wait.links.next, struct rpc_task, u.tk_wait.list);
 161                 list_move(&t->u.tk_wait.list, &task->u.tk_wait.list);
 162                 list_splice_init(&task->u.tk_wait.links, &t->u.tk_wait.links);
 163         }
 164 }
 165
 166 /*
 167  * Remove request from queue.
 168  * Note: must be called with spin lock held.
 169  */
 170 static void __rpc_remove_wait_queue(struct rpc_wait_queue *queue, struct rpc_task *task)
 171 {
 172         __rpc_disable_timer(queue, task);
 173         if (RPC_IS_PRIORITY(queue))
 174                 __rpc_remove_wait_queue_priority(task);
 175         list_del(&task->u.tk_wait.list);
 176         queue->qlen--;
 177         dprintk("RPC: %5u removed from queue %p \"%s\"\n",
 178                         task->tk_pid, queue, rpc_qname(queue));
 179 }
 180
 181 static inline void rpc_set_waitqueue_priority(struct rpc_wait_queue *queue, int priority)
 182 {
 183         queue->priority = priority;
 184         queue->count = 1 << (priority * 2);
 185 }
 186
 187 static inline void rpc_set_waitqueue_owner(struct rpc_wait_queue *queue, pid_t pid)
 188 {
 189         queue->owner = pid;
 190         queue->nr = RPC_BATCH_COUNT;
 191 }
 192
 193 static inline void rpc_reset_waitqueue_priority(struct rpc_wait_queue *queue)
 194 {
 195         rpc_set_waitqueue_priority(queue, queue->maxpriority);
 196         rpc_set_waitqueue_owner(queue, 0);
 197 }
 198
 199 static void __rpc_init_priority_wait_queue(struct rpc_wait_queue *queue, const char *qname, unsigned char nr_queues)
 200 {
 201         int i;
 202
 203         spin_lock_init(&queue->lock);
 204         for (i = 0; i < ARRAY_SIZE(queue->tasks); i++)
 205                 INIT_LIST_HEAD(&queue->tasks[i]);
 206         queue->maxpriority = nr_queues - 1;
 207         rpc_reset_waitqueue_priority(queue);
 208         queue->qlen = 0;
 209         setup_timer(&queue->timer_list.timer, __rpc_queue_timer_fn, (unsigned long)queue);
 210         INIT_LIST_HEAD(&queue->timer_list.list);
 211         rpc_assign_waitqueue_name(queue, qname);
 212 }
 213
 214 void rpc_init_priority_wait_queue(struct rpc_wait_queue *queue, const char *qname)
 215 {
 216         __rpc_init_priority_wait_queue(queue, qname, RPC_NR_PRIORITY);
 217 }
 218 EXPORT_SYMBOL_GPL(rpc_init_priority_wait_queue);
 219
 220 void rpc_init_wait_queue(struct rpc_wait_queue *queue, const char *qname)
 221 {
 222         __rpc_init_priority_wait_queue(queue, qname, 1);
 223 }
 224 EXPORT_SYMBOL_GPL(rpc_init_wait_queue);
 225
 226 void rpc_destroy_wait_queue(struct rpc_wait_queue *queue)
 227 {
 228         del_timer_sync(&queue->timer_list.timer);
 229 }
 230 EXPORT_SYMBOL_GPL(rpc_destroy_wait_queue);
 231
 232 static int rpc_wait_bit_killable(void *word)
 233 {
 234         if (fatal_signal_pending(current))
 235                 return -ERESTARTSYS;
 236         freezable_schedule();
 237         return 0;
 238 }
 239
 240 #ifdef RPC_DEBUG
 241 static void rpc_task_set_debuginfo(struct rpc_task *task)
 242 {
 243         static atomic_t rpc_pid;
 244
 245         task->tk_pid = atomic_inc_return(&rpc_pid);
 246 }
 247 #else
 248 static inline void rpc_task_set_debuginfo(struct rpc_task *task)
 249 {
 250 }
 251 #endif
 252
 253 static void rpc_set_active(struct rpc_task *task)
 254 {
 255         trace_rpc_task_begin(task->tk_client, task, NULL);
 256
 257         rpc_task_set_debuginfo(task);
 258         set_bit(RPC_TASK_ACTIVE, &task->tk_runstate);
 259 }
 260
 261 /*
 262  * Mark an RPC call as having completed by clearing the 'active' bit
 263  * and then waking up all tasks that were sleeping.
 264  */
 265 static int rpc_complete_task(struct rpc_task *task)
 266 {
 267         void *m = &task->tk_runstate;
 268         wait_queue_head_t *wq = bit_waitqueue(m, RPC_TASK_ACTIVE);
 269         struct wait_bit_key k = __WAIT_BIT_KEY_INITIALIZER(m, RPC_TASK_ACTIVE);
 270         unsigned long flags;
 271         int ret;
 272
 273         trace_rpc_task_complete(task->tk_client, task, NULL);
 274
 275         spin_lock_irqsave(&wq->lock, flags);
 276         clear_bit(RPC_TASK_ACTIVE, &task->tk_runstate);
 277         ret = atomic_dec_and_test(&task->tk_count);
 278         if (waitqueue_active(wq))
 279                 __wake_up_locked_key(wq, TASK_NORMAL, &k);
 280         spin_unlock_irqrestore(&wq->lock, flags);
 281         return ret;
 282 }
 283
 284 /*
 285  * Allow callers to wait for completion of an RPC call
 286  *
 287  * Note the use of out_of_line_wait_on_bit() rather than wait_on_bit()
 288  * to enforce taking of the wq->lock and hence avoid races with
 289  * rpc_complete_task().
 290  */
 291 int __rpc_wait_for_completion_task(struct rpc_task *task, int (*action)(void *))
 292 {
 293         if (action == NULL)
 294                 action = rpc_wait_bit_killable;
 295         return out_of_line_wait_on_bit(&task->tk_runstate, RPC_TASK_ACTIVE,
 296                         action, TASK_KILLABLE);
 297 }
 298 EXPORT_SYMBOL_GPL(__rpc_wait_for_completion_task);
 299
 300 /*
 301  * Make an RPC task runnable.
 302  *
 303  * Note: If the task is ASYNC, this must be called with
 304  * the spinlock held to protect the wait queue operation.
 305  */
 306 static void rpc_make_runnable(struct rpc_task *task)
 307 {
 308         rpc_clear_queued(task);
 309         if (rpc_test_and_set_running(task))
 310                 return;
 311         if (RPC_IS_ASYNC(task)) {
 312                 INIT_WORK(&task->u.tk_work, rpc_async_schedule);
 313                 queue_work(rpciod_workqueue, &task->u.tk_work);
 314         } else
 315                 wake_up_bit(&task->tk_runstate, RPC_TASK_QUEUED);
 316 }
 317
 318 /*
 319  * Prepare for sleeping on a wait queue.
 320  * By always appending tasks to the list we ensure FIFO behavior.
 321  * NB: An RPC task will only receive interrupt-driven events as long
 322  * as it's on a wait queue.
 323  */
 324 static void __rpc_sleep_on_priority(struct rpc_wait_queue *q,
 325                 struct rpc_task *task,
 326                 rpc_action action,
 327                 unsigned char queue_priority)
 328 {
 329         dprintk("RPC: %5u sleep_on(queue \"%s\" time %lu)\n",
 330                         task->tk_pid, rpc_qname(q), jiffies);
 331
 332         trace_rpc_task_sleep(task->tk_client, task, q);
 333
 334         __rpc_add_wait_queue(q, task, queue_priority);
 335
 336         BUG_ON(task->tk_callback != NULL);
 337         task->tk_callback = action;
 338         __rpc_add_timer(q, task);
 339 }
 340
 341 void rpc_sleep_on(struct rpc_wait_queue *q, struct rpc_task *task,
 342                                 rpc_action action)
 343 {
 344         /* We shouldn't ever put an inactive task to sleep */
 345         BUG_ON(!RPC_IS_ACTIVATED(task));
 346
 347         /*
 348          * Protect the queue operations.
 349          */
 350         spin_lock_bh(&q->lock);
 351         __rpc_sleep_on_priority(q, task, action, task->tk_priority);
 352         spin_unlock_bh(&q->lock);
 353 }
 354 EXPORT_SYMBOL_GPL(rpc_sleep_on);
 355
 356 void rpc_sleep_on_priority(struct rpc_wait_queue *q, struct rpc_task *task,
 357                 rpc_action action, int priority)
 358 {
 359         /* We shouldn't ever put an inactive task to sleep */
 360         BUG_ON(!RPC_IS_ACTIVATED(task));
 361
 362         /*
 363          * Protect the queue operations.
 364          */
 365         spin_lock_bh(&q->lock);
 366         __rpc_sleep_on_priority(q, task, action, priority - RPC_PRIORITY_LOW);
 367         spin_unlock_bh(&q->lock);
 368 }
 369
 370 /**
 371  * __rpc_do_wake_up_task - wake up a single rpc_task
 372  * @queue: wait queue
 373  * @task: task to be woken up
 374  *
 375  * Caller must hold queue->lock, and have cleared the task queued flag.
 376  */
 377 static void __rpc_do_wake_up_task(struct rpc_wait_queue *queue, struct rpc_task *task)
 378 {
 379         dprintk("RPC: %5u __rpc_wake_up_task (now %lu)\n",
 380                         task->tk_pid, jiffies);
 381
 382         /* Has the task been executed yet? If not, we cannot wake it up! */
 383         if (!RPC_IS_ACTIVATED(task)) {
 384                 printk(KERN_ERR "RPC: Inactive task (%p) being woken up!\n", task);
 385                 return;
 386         }
 387
 388         trace_rpc_task_wakeup(task->tk_client, task, queue);
 389
 390         __rpc_remove_wait_queue(queue, task);
 391
 392         rpc_make_runnable(task);
 393
 394         dprintk("RPC:       __rpc_wake_up_task done\n");
 395 }
 396
 397 /*
 398  * Wake up a queued task while the queue lock is being held
 399  */
 400 static void rpc_wake_up_task_queue_locked(struct rpc_wait_queue *queue, struct rpc_task *task)
 401 {
 402         if (RPC_IS_QUEUED(task) && task->tk_waitqueue == queue)
 403                 __rpc_do_wake_up_task(queue, task);
 404 }
 405
 406 /*
 407  * Tests whether rpc queue is empty
 408  */
 409 int rpc_queue_empty(struct rpc_wait_queue *queue)
 410 {
 411         int res;
 412
 413         spin_lock_bh(&queue->lock);
 414         res = queue->qlen;
 415         spin_unlock_bh(&queue->lock);
 416         return res == 0;
 417 }
 418 EXPORT_SYMBOL_GPL(rpc_queue_empty);
 419
 420 /*
 421  * Wake up a task on a specific queue
 422  */
 423 void rpc_wake_up_queued_task(struct rpc_wait_queue *queue, struct rpc_task *task)
 424 {
 425         spin_lock_bh(&queue->lock);
 426         rpc_wake_up_task_queue_locked(queue, task);
 427         spin_unlock_bh(&queue->lock);
 428 }
 429 EXPORT_SYMBOL_GPL(rpc_wake_up_queued_task);
 430
 431 /*
 432  * Wake up the next task on a priority queue.
 433  */
 434 static struct rpc_task *__rpc_find_next_queued_priority(struct rpc_wait_queue *queue)
 435 {
 436         struct list_head *q;
 437         struct rpc_task *task;
 438
 439         /*
 440          * Service a batch of tasks from a single owner.
 441          */
 442         q = &queue->tasks[queue->priority];
 443         if (!list_empty(q)) {
 444                 task = list_entry(q->next, struct rpc_task, u.tk_wait.list);
 445                 if (queue->owner == task->tk_owner) {
 446                         if (--queue->nr)
 447                                 goto out;
 448                         list_move_tail(&task->u.tk_wait.list, q);
 449                 }
 450                 /*
 451                  * Check if we need to switch queues.
 452                  */
 453                 if (--queue->count)
 454                         goto new_owner;
 455         }
 456
 457         /*
 458          * Service the next queue.
 459          */
 460         do {
 461                 if (q == &queue->tasks[0])
 462                         q = &queue->tasks[queue->maxpriority];
 463                 else
 464                         q = q - 1;
 465                 if (!list_empty(q)) {
 466                         task = list_entry(q->next, struct rpc_task, u.tk_wait.list);
 467                         goto new_queue;
 468                 }
 469         } while (q != &queue->tasks[queue->priority]);
 470
 471         rpc_reset_waitqueue_priority(queue);
 472         return NULL;
 473
 474 new_queue:
 475         rpc_set_waitqueue_priority(queue, (unsigned int)(q - &queue->tasks[0]));
 476 new_owner:
 477         rpc_set_waitqueue_owner(queue, task->tk_owner);
 478 out:
 479         return task;
 480 }
 481
 482 static struct rpc_task *__rpc_find_next_queued(struct rpc_wait_queue *queue)
 483 {
 484         if (RPC_IS_PRIORITY(queue))
 485                 return __rpc_find_next_queued_priority(queue);
 486         if (!list_empty(&queue->tasks[0]))
 487                 return list_first_entry(&queue->tasks[0], struct rpc_task, u.tk_wait.list);
 488         return NULL;
 489 }
 490
 491 /*
 492  * Wake up the first task on the wait queue.
 493  */
 494 struct rpc_task *rpc_wake_up_first(struct rpc_wait_queue *queue,
 495                 bool (*func)(struct rpc_task *, void *), void *data)
 496 {
 497         struct rpc_task *task = NULL;
 498
 499         dprintk("RPC:       wake_up_first(%p \"%s\")\n",
 500                         queue, rpc_qname(queue));
 501         spin_lock_bh(&queue->lock);
 502         task = __rpc_find_next_queued(queue);
 503         if (task != NULL) {
 504                 if (func(task, data))
 505                         rpc_wake_up_task_queue_locked(queue, task);
 506                 else
 507                         task = NULL;
 508         }
 509         spin_unlock_bh(&queue->lock);
 510
 511         return task;
 512 }
 513 EXPORT_SYMBOL_GPL(rpc_wake_up_first);
 514
 515 static bool rpc_wake_up_next_func(struct rpc_task *task, void *data)
 516 {
 517         return true;
 518 }
 519
 520 /*
 521  * Wake up the next task on the wait queue.
 522 */
 523 struct rpc_task *rpc_wake_up_next(struct rpc_wait_queue *queue)
 524 {
 525         return rpc_wake_up_first(queue, rpc_wake_up_next_func, NULL);
 526 }
 527 EXPORT_SYMBOL_GPL(rpc_wake_up_next);
 528
 529 /**
 530  * rpc_wake_up - wake up all rpc_tasks
 531  * @queue: rpc_wait_queue on which the tasks are sleeping
 532  *
 533  * Grabs queue->lock
 534  */
 535 void rpc_wake_up(struct rpc_wait_queue *queue)
 536 {
 537         struct list_head *head;
 538
 539         spin_lock_bh(&queue->lock);
 540         head = &queue->tasks[queue->maxpriority];
 541         for (;;) {
 542                 while (!list_empty(head)) {
 543                         struct rpc_task *task;
 544                         task = list_first_entry(head,
 545                                         struct rpc_task,
 546                                         u.tk_wait.list);
 547                         rpc_wake_up_task_queue_locked(queue, task);
 548                 }
 549                 if (head == &queue->tasks[0])
 550                         break;
 551                 head--;
 552         }
 553         spin_unlock_bh(&queue->lock);
 554 }
 555 EXPORT_SYMBOL_GPL(rpc_wake_up);
 556
 557 /**
 558  * rpc_wake_up_status - wake up all rpc_tasks and set their status value.
 559  * @queue: rpc_wait_queue on which the tasks are sleeping
 560  * @status: status value to set
 561  *
 562  * Grabs queue->lock
 563  */
 564 void rpc_wake_up_status(struct rpc_wait_queue *queue, int status)
 565 {
 566         struct list_head *head;
 567
 568         spin_lock_bh(&queue->lock);
 569         head = &queue->tasks[queue->maxpriority];
 570         for (;;) {
 571                 while (!list_empty(head)) {
 572                         struct rpc_task *task;
 573                         task = list_first_entry(head,
 574                                         struct rpc_task,
 575                                         u.tk_wait.list);
 576                         task->tk_status = status;
 577                         rpc_wake_up_task_queue_locked(queue, task);
 578                 }
 579                 if (head == &queue->tasks[0])
 580                         break;
 581                 head--;
 582         }
 583         spin_unlock_bh(&queue->lock);
 584 }
 585 EXPORT_SYMBOL_GPL(rpc_wake_up_status);
 586
 587 static void __rpc_queue_timer_fn(unsigned long ptr)
 588 {
 589         struct rpc_wait_queue *queue = (struct rpc_wait_queue *)ptr;
 590         struct rpc_task *task, *n;
 591         unsigned long expires, now, timeo;
 592
 593         spin_lock(&queue->lock);
 594         expires = now = jiffies;
 595         list_for_each_entry_safe(task, n, &queue->timer_list.list, u.tk_wait.timer_list) {
 596                 timeo = task->u.tk_wait.expires;
 597                 if (time_after_eq(now, timeo)) {
 598                         dprintk("RPC: %5u timeout\n", task->tk_pid);
 599                         task->tk_status = -ETIMEDOUT;
 600                         rpc_wake_up_task_queue_locked(queue, task);
 601                         continue;
 602                 }
 603                 if (expires == now || time_after(expires, timeo))
 604                         expires = timeo;
 605         }
 606         if (!list_empty(&queue->timer_list.list))
 607                 rpc_set_queue_timer(queue, expires);
 608         spin_unlock(&queue->lock);
 609 }
 610
 611 static void __rpc_atrun(struct rpc_task *task)
 612 {
 613         task->tk_status = 0;
 614 }
 615
 616 /*
 617  * Run a task at a later time
 618  */
 619 void rpc_delay(struct rpc_task *task, unsigned long delay)
 620 {
 621         task->tk_timeout = delay;
 622         rpc_sleep_on(&delay_queue, task, __rpc_atrun);
 623 }
 624 EXPORT_SYMBOL_GPL(rpc_delay);
 625
 626 /*
 627  * Helper to call task->tk_ops->rpc_call_prepare
 628  */
 629 void rpc_prepare_task(struct rpc_task *task)
 630 {
 631         task->tk_ops->rpc_call_prepare(task, task->tk_calldata);
 632 }
 633
 634 static void
 635 rpc_init_task_statistics(struct rpc_task *task)
 636 {
 637         /* Initialize retry counters */
 638         task->tk_garb_retry = 2;
 639         task->tk_cred_retry = 2;
 640         task->tk_rebind_retry = 2;
 641
 642         /* starting timestamp */
 643         task->tk_start = ktime_get();
 644 }
 645
 646 static void
 647 rpc_reset_task_statistics(struct rpc_task *task)
 648 {
 649         task->tk_timeouts = 0;
 650         task->tk_flags &= ~(RPC_CALL_MAJORSEEN|RPC_TASK_KILLED|RPC_TASK_SENT);
 651
 652         rpc_init_task_statistics(task);
 653 }
 654
 655 /*
 656  * Helper that calls task->tk_ops->rpc_call_done if it exists
 657  */
 658 void rpc_exit_task(struct rpc_task *task)
 659 {
 660         task->tk_action = NULL;
 661         if (task->tk_ops->rpc_call_done != NULL) {
 662                 task->tk_ops->rpc_call_done(task, task->tk_calldata);
 663                 if (task->tk_action != NULL) {
 664                         WARN_ON(RPC_ASSASSINATED(task));
 665                         /* Always release the RPC slot and buffer memory */
 666                         xprt_release(task);
 667                         rpc_reset_task_statistics(task);
 668                 }
 669         }
 670 }
 671
 672 void rpc_exit(struct rpc_task *task, int status)
 673 {
 674         task->tk_status = status;
 675         task->tk_action = rpc_exit_task;
 676         if (RPC_IS_QUEUED(task))
 677                 rpc_wake_up_queued_task(task->tk_waitqueue, task);
 678 }
 679 EXPORT_SYMBOL_GPL(rpc_exit);
 680
 681 void rpc_release_calldata(const struct rpc_call_ops *ops, void *calldata)
 682 {
 683         if (ops->rpc_release != NULL)
 684                 ops->rpc_release(calldata);
 685 }
 686
 687 /*
 688  * This is the RPC `scheduler' (or rather, the finite state machine).
 689  */
 690 static void __rpc_execute(struct rpc_task *task)
 691 {
 692         struct rpc_wait_queue *queue;
 693         int task_is_async = RPC_IS_ASYNC(task);
 694         int status = 0;
 695
 696         dprintk("RPC: %5u __rpc_execute flags=0x%x\n",
 697                         task->tk_pid, task->tk_flags);
 698
 699         BUG_ON(RPC_IS_QUEUED(task));
 700
 701         for (;;) {
 702                 void (*do_action)(struct rpc_task *);
 703
 704                 /*
 705                  * Execute any pending callback first.
 706                  */
 707                 do_action = task->tk_callback;
 708                 task->tk_callback = NULL;
 709                 if (do_action == NULL) {
 710                         /*
 711                          * Perform the next FSM step.
 712                          * tk_action may be NULL if the task has been killed.
 713                          * In particular, note that rpc_killall_tasks may
 714                          * do this at any time, so beware when dereferencing.
 715                          */
 716                         do_action = task->tk_action;
 717                         if (do_action == NULL)
 718                                 break;
 719                 }
 720                 trace_rpc_task_run_action(task->tk_client, task, task->tk_action);
 721                 do_action(task);
 722
 723                 /*
 724                  * Lockless check for whether task is sleeping or not.
 725                  */
 726                 if (!RPC_IS_QUEUED(task))
 727                         continue;
 728                 /*
 729                  * The queue->lock protects against races with
 730                  * rpc_make_runnable().
 731                  *
 732                  * Note that once we clear RPC_TASK_RUNNING on an asynchronous
 733                  * rpc_task, rpc_make_runnable() can assign it to a
 734                  * different workqueue. We therefore cannot assume that the
 735                  * rpc_task pointer may still be dereferenced.
 736                  */
 737                 queue = task->tk_waitqueue;
 738                 spin_lock_bh(&queue->lock);
 739                 if (!RPC_IS_QUEUED(task)) {
 740                         spin_unlock_bh(&queue->lock);
 741                         continue;
 742                 }
 743                 rpc_clear_running(task);
 744                 spin_unlock_bh(&queue->lock);
 745                 if (task_is_async)
 746                         return;
 747
 748                 /* sync task: sleep here */
 749                 dprintk("RPC: %5u sync task going to sleep\n", task->tk_pid);
 750                 status = out_of_line_wait_on_bit(&task->tk_runstate,
 751                                 RPC_TASK_QUEUED, rpc_wait_bit_killable,
 752                                 TASK_KILLABLE);
 753                 if (status == -ERESTARTSYS) {
 754                         /*
 755                          * When a sync task receives a signal, it exits with
 756                          * -ERESTARTSYS. In order to catch any callbacks that
 757                          * clean up after sleeping on some queue, we don't
 758                          * break the loop here, but go around once more.
 759                          */
 760                         dprintk("RPC: %5u got signal\n", task->tk_pid);
 761                         task->tk_flags |= RPC_TASK_KILLED;
 762                         rpc_exit(task, -ERESTARTSYS);
 763                 }
 764                 rpc_set_running(task);
 765                 dprintk("RPC: %5u sync task resuming\n", task->tk_pid);
 766         }
 767
 768         dprintk("RPC: %5u return %d, status %d\n", task->tk_pid, status,
 769                         task->tk_status);
 770         /* Release all resources associated with the task */
 771         rpc_release_task(task);
 772 }
 773
 774 /*
 775  * User-visible entry point to the scheduler.
 776  *
 777  * This may be called recursively if e.g. an async NFS task updates
 778  * the attributes and finds that dirty pages must be flushed.
 779  * NOTE: Upon exit of this function the task is guaranteed to be
 780  *       released. In particular note that tk_release() will have
 781  *       been called, so your task memory may have been freed.
 782  */
 783 void rpc_execute(struct rpc_task *task)
 784 {
 785         rpc_set_active(task);
 786         rpc_make_runnable(task);
 787         if (!RPC_IS_ASYNC(task))
 788                 __rpc_execute(task);
 789 }
 790
 791 static void rpc_async_schedule(struct work_struct *work)
 792 {
 793         __rpc_execute(container_of(work, struct rpc_task, u.tk_work));
 794 }
 795
 796 /**
 797  * rpc_malloc - allocate an RPC buffer
 798  * @task: RPC task that will use this buffer
 799  * @size: requested byte size
 800  *
 801  * To prevent rpciod from hanging, this allocator never sleeps,
 802  * returning NULL if the request cannot be serviced immediately.
 803  * The caller can arrange to sleep in a way that is safe for rpciod.
 804  *
 805  * Most requests are 'small' (under 2KiB) and can be serviced from a
 806  * mempool, ensuring that NFS reads and writes can always proceed,
 807  * and that there is good locality of reference for these buffers.
 808  *
 809  * In order to avoid memory starvation triggering more writebacks of
 810  * NFS requests, we avoid using GFP_KERNEL.
 811  */
 812 void *rpc_malloc(struct rpc_task *task, size_t size)
 813 {
 814         struct rpc_buffer *buf;
 815         gfp_t gfp = RPC_IS_SWAPPER(task) ? GFP_ATOMIC : GFP_NOWAIT;
 816
 817         size += sizeof(struct rpc_buffer);
 818         if (size <= RPC_BUFFER_MAXSIZE)
 819                 buf = mempool_alloc(rpc_buffer_mempool, gfp);
 820         else
 821                 buf = kmalloc(size, gfp);
 822
 823         if (!buf)
 824                 return NULL;
 825
 826         buf->len = size;
 827         dprintk("RPC: %5u allocated buffer of size %zu at %p\n",
 828                         task->tk_pid, size, buf);
 829         return &buf->data;
 830 }
 831 EXPORT_SYMBOL_GPL(rpc_malloc);
 832
 833 /**
 834  * rpc_free - free buffer allocated via rpc_malloc
 835  * @buffer: buffer to free
 836  *
 837  */
 838 void rpc_free(void *buffer)
 839 {
 840         size_t size;
 841         struct rpc_buffer *buf;
 842
 843         if (!buffer)
 844                 return;
 845
 846         buf = container_of(buffer, struct rpc_buffer, data);
 847         size = buf->len;
 848
 849         dprintk("RPC:       freeing buffer of size %zu at %p\n",
 850                         size, buf);
 851
 852         if (size <= RPC_BUFFER_MAXSIZE)
 853                 mempool_free(buf, rpc_buffer_mempool);
 854         else
 855                 kfree(buf);
 856 }
 857 EXPORT_SYMBOL_GPL(rpc_free);
 858
 859 /*
 860  * Creation and deletion of RPC task structures
 861  */
 862 static void rpc_init_task(struct rpc_task *task, const struct rpc_task_setup *task_setup_data)
 863 {
 864         memset(task, 0, sizeof(*task));
 865         atomic_set(&task->tk_count, 1);
 866         task->tk_flags  = task_setup_data->flags;
 867         task->tk_ops = task_setup_data->callback_ops;
 868         task->tk_calldata = task_setup_data->callback_data;
 869         INIT_LIST_HEAD(&task->tk_task);
 870
 871         task->tk_priority = task_setup_data->priority - RPC_PRIORITY_LOW;
 872         task->tk_owner = current->tgid;
 873
 874         /* Initialize workqueue for async tasks */
 875         task->tk_workqueue = task_setup_data->workqueue;
 876
 877         if (task->tk_ops->rpc_call_prepare != NULL)
 878                 task->tk_action = rpc_prepare_task;
 879
 880         rpc_init_task_statistics(task);
 881
 882         dprintk("RPC:       new task initialized, procpid %u\n",
 883                                 task_pid_nr(current));
 884 }
 885
 886 static struct rpc_task *
 887 rpc_alloc_task(void)
 888 {
 889         return (struct rpc_task *)mempool_alloc(rpc_task_mempool, GFP_NOFS);
 890 }
 891
 892 /*
 893  * Create a new task for the specified client.
 894  */
 895 struct rpc_task *rpc_new_task(const struct rpc_task_setup *setup_data)
 896 {
 897         struct rpc_task *task = setup_data->task;
 898         unsigned short flags = 0;
 899
 900         if (task == NULL) {
 901                 task = rpc_alloc_task();
 902                 if (task == NULL) {
 903                         rpc_release_calldata(setup_data->callback_ops,
 904                                         setup_data->callback_data);
 905                         return ERR_PTR(-ENOMEM);
 906                 }
 907                 flags = RPC_TASK_DYNAMIC;
 908         }
 909
 910         rpc_init_task(task, setup_data);
 911         task->tk_flags |= flags;
 912         dprintk("RPC:       allocated task %p\n", task);
 913         return task;
 914 }
 915
 916 static void rpc_free_task(struct rpc_task *task)
 917 {
 918         const struct rpc_call_ops *tk_ops = task->tk_ops;
 919         void *calldata = task->tk_calldata;
 920
 921         if (task->tk_flags & RPC_TASK_DYNAMIC) {
 922                 dprintk("RPC: %5u freeing task\n", task->tk_pid);
 923                 mempool_free(task, rpc_task_mempool);
 924         }
 925         rpc_release_calldata(tk_ops, calldata);
 926 }
 927
 928 static void rpc_async_release(struct work_struct *work)
 929 {
 930         rpc_free_task(container_of(work, struct rpc_task, u.tk_work));
 931 }
 932
 933 static void rpc_release_resources_task(struct rpc_task *task)
 934 {
 935         if (task->tk_rqstp)
 936                 xprt_release(task);
 937         if (task->tk_msg.rpc_cred) {
 938                 put_rpccred(task->tk_msg.rpc_cred);
 939                 task->tk_msg.rpc_cred = NULL;
 940         }
 941         rpc_task_release_client(task);
 942 }
 943
 944 static void rpc_final_put_task(struct rpc_task *task,
 945                 struct workqueue_struct *q)
 946 {
 947         if (q != NULL) {
 948                 INIT_WORK(&task->u.tk_work, rpc_async_release);
 949                 queue_work(q, &task->u.tk_work);
 950         } else
 951                 rpc_free_task(task);
 952 }
 953
 954 static void rpc_do_put_task(struct rpc_task *task, struct workqueue_struct *q)
 955 {
 956         if (atomic_dec_and_test(&task->tk_count)) {
 957                 rpc_release_resources_task(task);
 958                 rpc_final_put_task(task, q);
 959         }
 960 }
 961
 962 void rpc_put_task(struct rpc_task *task)
 963 {
 964         rpc_do_put_task(task, NULL);
 965 }
 966 EXPORT_SYMBOL_GPL(rpc_put_task);
 967
 968 void rpc_put_task_async(struct rpc_task *task)
 969 {
 970         rpc_do_put_task(task, task->tk_workqueue);
 971 }
 972 EXPORT_SYMBOL_GPL(rpc_put_task_async);
 973
 974 static void rpc_release_task(struct rpc_task *task)
 975 {
 976         dprintk("RPC: %5u release task\n", task->tk_pid);
 977
 978         BUG_ON (RPC_IS_QUEUED(task));
 979
 980         rpc_release_resources_task(task);
 981
 982         /*
 983          * Note: at this point we have been removed from rpc_clnt->cl_tasks,
 984          * so it should be safe to use task->tk_count as a test for whether
 985          * or not any other processes still hold references to our rpc_task.
 986          */
 987         if (atomic_read(&task->tk_count) != 1 + !RPC_IS_ASYNC(task)) {
 988                 /* Wake up anyone who may be waiting for task completion */
 989                 if (!rpc_complete_task(task))
 990                         return;
 991         } else {
 992                 if (!atomic_dec_and_test(&task->tk_count))
 993                         return;
 994         }
 995         rpc_final_put_task(task, task->tk_workqueue);
 996 }
 997
 998 int rpciod_up(void)
 999 {
1000         return try_module_get(THIS_MODULE) ? 0 : -EINVAL;
1001 }
1002
1003 void rpciod_down(void)
1004 {
1005         module_put(THIS_MODULE);
1006 }
1007
1008 /*
1009  * Start up the rpciod workqueue.
1010  */
1011 static int rpciod_start(void)
1012 {
1013         struct workqueue_struct *wq;
1014
1015         /*
1016          * Create the rpciod thread and wait for it to start.
1017          */
1018         dprintk("RPC:       creating workqueue rpciod\n");
1019         wq = alloc_workqueue("rpciod", WQ_MEM_RECLAIM, 0);
1020         rpciod_workqueue = wq;
1021         return rpciod_workqueue != NULL;
1022 }
1023
1024 static void rpciod_stop(void)
1025 {
1026         struct workqueue_struct *wq = NULL;
1027
1028         if (rpciod_workqueue == NULL)
1029                 return;
1030         dprintk("RPC:       destroying workqueue rpciod\n");
1031
1032         wq = rpciod_workqueue;
1033         rpciod_workqueue = NULL;
1034         destroy_workqueue(wq);
1035 }
1036
1037 void
1038 rpc_destroy_mempool(void)
1039 {
1040         rpciod_stop();
1041         if (rpc_buffer_mempool)
1042                 mempool_destroy(rpc_buffer_mempool);
1043         if (rpc_task_mempool)
1044                 mempool_destroy(rpc_task_mempool);
1045         if (rpc_task_slabp)
1046                 kmem_cache_destroy(rpc_task_slabp);
1047         if (rpc_buffer_slabp)
1048                 kmem_cache_destroy(rpc_buffer_slabp);
1049         rpc_destroy_wait_queue(&delay_queue);
1050 }
1051
1052 int
1053 rpc_init_mempool(void)
1054 {
1055         /*
1056          * The following is not strictly a mempool initialisation,
1057          * but there is no harm in doing it here
1058          */
1059         rpc_init_wait_queue(&delay_queue, "delayq");
1060         if (!rpciod_start())
1061                 goto err_nomem;
1062
1063         rpc_task_slabp = kmem_cache_create("rpc_tasks",
1064                                              sizeof(struct rpc_task),
1065                                              0, SLAB_HWCACHE_ALIGN,
1066                                              NULL);
1067         if (!rpc_task_slabp)
1068                 goto err_nomem;
1069         rpc_buffer_slabp = kmem_cache_create("rpc_buffers",
1070                                              RPC_BUFFER_MAXSIZE,
1071                                              0, SLAB_HWCACHE_ALIGN,
1072                                              NULL);
1073         if (!rpc_buffer_slabp)
1074                 goto err_nomem;
1075         rpc_task_mempool = mempool_create_slab_pool(RPC_TASK_POOLSIZE,
1076                                                     rpc_task_slabp);
1077         if (!rpc_task_mempool)
1078                 goto err_nomem;
1079         rpc_buffer_mempool = mempool_create_slab_pool(RPC_BUFFER_POOLSIZE,
1080                                                       rpc_buffer_slabp);
1081         if (!rpc_buffer_mempool)
1082                 goto err_nomem;
1083         return 0;
1084 err_nomem:
1085         rpc_destroy_mempool();
1086         return -ENOMEM;
1087 }