lib/workqueue.c

/* 
 * Quagga Work Queue Support.
 *
 * Copyright (C) 2005 Sun Microsystems, Inc.
 *
 * This file is part of GNU Zebra.
 *
 * Quagga is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License as published by the
 * Free Software Foundation; either version 2, or (at your option) any
 * later version.
 *
 * Quagga is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with Quagga; see the file COPYING.  If not, write to the Free
 * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
 * 02111-1307, USA.  
 */

#include <zebra.h>
#include "thread.h"
#include "memory.h"
#include "workqueue.h"
#include "linklist.h"
#include "command.h"
#include "log.h"

/* master list of work_queues */
static struct list _work_queues;
/* pointer primarily to avoid an otherwise harmless warning on
 * ALL_LIST_ELEMENTS_RO 
 */
static struct list *work_queues = &_work_queues;

#define WORK_QUEUE_MIN_GRANULARITY 1

static struct work_queue_item *
work_queue_item_new (struct work_queue *wq)
{
  struct work_queue_item *item;
  assert (wq);

  item = XCALLOC (MTYPE_WORK_QUEUE_ITEM, 
                  sizeof (struct work_queue_item));
  
  return item;
}

static void
work_queue_item_free (struct work_queue_item *item)
{
  XFREE (MTYPE_WORK_QUEUE_ITEM, item);
  return;
}

/* create new work queue */
struct work_queue *
work_queue_new (struct thread_master *m, const char *queue_name)
{
  struct work_queue *new;
  
  new = XCALLOC (MTYPE_WORK_QUEUE, sizeof (struct work_queue));

  if (new == NULL)
    return new;
  
  new->name = XSTRDUP (MTYPE_WORK_QUEUE_NAME, queue_name);
  new->master = m;
  SET_FLAG (new->flags, WQ_UNPLUGGED);
  
  if ( (new->items = list_new ()) == NULL)
    {
      XFREE (MTYPE_WORK_QUEUE_NAME, new->name);
      XFREE (MTYPE_WORK_QUEUE, new);
      
      return NULL;
    }
  
  new->items->del = (void (*)(void *)) work_queue_item_free;  
  
  listnode_add (work_queues, new);
  
  new->cycles.granularity = WORK_QUEUE_MIN_GRANULARITY;
  new->cycles.worst = UINT_MAX;
  
  /* Default values, can be overriden by caller */
  new->spec.hold = WORK_QUEUE_DEFAULT_HOLD;
    
  return new;
}

void
work_queue_free (struct work_queue *wq)
{
  if (wq->thread != NULL)
    thread_cancel(wq->thread);
  
  /* list_delete frees items via callback */
  list_delete (wq->items);
  listnode_delete (work_queues, wq);
  
  XFREE (MTYPE_WORK_QUEUE_NAME, wq->name);
  XFREE (MTYPE_WORK_QUEUE, wq);
  return;
}

bool
work_queue_is_scheduled (struct work_queue *wq)
{
  return (wq->thread != NULL);
}

static int
work_queue_schedule (struct work_queue *wq, unsigned int delay)
{
  /* if appropriate, schedule work queue thread */
  if ( CHECK_FLAG (wq->flags, WQ_UNPLUGGED)
       && (wq->thread == NULL)
       && (listcount (wq->items) > 0) )
    {
      wq->thread = thread_add_background (wq->master, work_queue_run, 
                                          wq, delay);
      return 1;
    }
  else
    return 0;
}
  
void
work_queue_add (struct work_queue *wq, void *data)
{
  struct work_queue_item *item;
  
  assert (wq);

  if (!(item = work_queue_item_new (wq)))
    {
      zlog_err ("%s: unable to get new queue item", __func__);
      return;
    }
  
  item->data = data;
  listnode_add (wq->items, item);
  
  work_queue_schedule (wq, wq->spec.hold);
  
  return;
}

static void
work_queue_item_remove (struct work_queue *wq, struct listnode *ln)
{
  struct work_queue_item *item = listgetdata (ln);

  assert (item && item->data);

  /* call private data deletion callback if needed */  
  if (wq->spec.del_item_data)
    wq->spec.del_item_data (wq, item->data);

  list_delete_node (wq->items, ln);
  work_queue_item_free (item);
  
  return;
}

static void
work_queue_item_requeue (struct work_queue *wq, struct listnode *ln)
{
  LISTNODE_DETACH (wq->items, ln);
  LISTNODE_ATTACH (wq->items, ln); /* attach to end of list */
}

DEFUN(show_work_queues,
      show_work_queues_cmd,
      "show work-queues",
      SHOW_STR
      "Work Queue information\n")
{
  struct listnode *node;
  struct work_queue *wq;
  
  vty_out (vty, 
           "%c %8s %5s %8s %21s %6s %5s%s",
           ' ', "List","(ms) ","Q. Runs","Cycle Counts   ",
           " ","Worst",
           VTY_NEWLINE);
  vty_out (vty,
           "%c %8s %5s %8s %7s %6s %6s %6s %5s %s%s",
           'P',
           "Items",
           "Hold",
           "Total",
           "Best","Worst","Gran.","Avg.", "Lat.",
           "Name", 
           VTY_NEWLINE);
 
  for (ALL_LIST_ELEMENTS_RO (work_queues, node, wq))
    {
      vty_out (vty,"%c %8u %5u %8lu %7u %6u %6u %6u %5lu %s%s",
               (CHECK_FLAG (wq->flags, WQ_UNPLUGGED) ? ' ' : 'P'),
               listcount (wq->items),
               wq->spec.hold,
               wq->runs,
               wq->cycles.best, 
               MIN(wq->cycles.best, wq->cycles.worst),
               wq->cycles.granularity,
                 (wq->runs) ? 
                   (unsigned int) (wq->cycles.total / wq->runs) : 0,
               wq->worst_usec,
               wq->name,
               VTY_NEWLINE);
    }
    
  return CMD_SUCCESS;
}

/* 'plug' a queue: Stop it from being scheduled,
 * ie: prevent the queue from draining.
 */
void
work_queue_plug (struct work_queue *wq)
{
  if (wq->thread)
    thread_cancel (wq->thread);
  
  wq->thread = NULL;
  
  UNSET_FLAG (wq->flags, WQ_UNPLUGGED);
}

/* unplug queue, schedule it again, if appropriate
 * Ie: Allow the queue to be drained again
 */
void
work_queue_unplug (struct work_queue *wq)
{
  SET_FLAG (wq->flags, WQ_UNPLUGGED);

  /* if thread isnt already waiting, add one */
  work_queue_schedule (wq, wq->spec.hold);
}

/* timer thread to process a work queue
 * will reschedule itself if required,
 * otherwise work_queue_item_add 
 */
int
work_queue_run (struct thread *thread)
{
  struct work_queue *wq;
  struct work_queue_item *item;
  unsigned long took;
  wq_item_status ret;
  unsigned int cycles = 0;
  struct listnode *node, *nnode;
  char yielded = 0;

  wq = THREAD_ARG (thread);
  wq->thread = NULL;

  assert (wq && wq->items);

  /* calculate cycle granularity:
   * list iteration == 1 cycle
   * granularity == # cycles between checks whether we should yield.
   *
   * granularity should be > 0, and can increase slowly after each run to
   * provide some hysteris, but not past cycles.best or 2*cycles.
   *
   * Best: starts low, can only increase
   *
   * Worst: starts at MAX, can only decrease.
   *
   * Granularity: starts at WORK_QUEUE_MIN_GRANULARITY, can be decreased 
   *              if we run to end of time slot, can increase otherwise 
   *              by a small factor.
   *
   * We could use just the average and save some work, however we want to be
   * able to adjust quickly to CPU pressure. Average wont shift much if
   * daemon has been running a long time.
   */
   if (wq->cycles.granularity == 0)
     wq->cycles.granularity = WORK_QUEUE_MIN_GRANULARITY;

  for (ALL_LIST_ELEMENTS (wq->items, node, nnode, item))
  {
    assert (item && item->data);
    
    /* dont run items which are past their allowed retries */
    if (item->ran > wq->spec.max_retries)
      {
        /* run error handler, if any */
	if (wq->spec.errorfunc)
	  wq->spec.errorfunc (wq, item->data);
	work_queue_item_remove (wq, node);
	continue;
      }

    /* run and take care of items that want to be retried immediately */
    do
      {
        ret = wq->spec.workfunc (wq, item->data);
        item->ran++;
      }
    while ((ret == WQ_RETRY_NOW) 
           && (item->ran < wq->spec.max_retries));

    switch (ret)
      {
      case WQ_QUEUE_BLOCKED:
        {
          /* decrement item->ran again, cause this isn't an item
           * specific error, and fall through to WQ_RETRY_LATER
           */
          item->ran--;
        }
      case WQ_RETRY_LATER:
	{
	  goto stats;
	}
      case WQ_REQUEUE:
	{
	  item->ran--;
	  work_queue_item_requeue (wq, node);
	  break;
	}
      case WQ_RETRY_NOW:
        /* a RETRY_NOW that gets here has exceeded max_tries, same as ERROR */
      case WQ_ERROR:
	{
	  if (wq->spec.errorfunc)
	    wq->spec.errorfunc (wq, item);
	}
	/* fall through here is deliberate */
      case WQ_SUCCESS:
      default:
	{
	  work_queue_item_remove (wq, node);
	  break;
	}
      }

    /* completed cycle */
    cycles++;

    /* test if we should yield */
    if ( !(cycles % wq->cycles.granularity) 
        && (took = thread_should_yield (thread)))
      {
        yielded = 1;
        goto stats;
      }
  }

stats:

#define WQ_HYSTERESIS_FACTOR 4

  if (cycles > wq->cycles.best)
    wq->cycles.best = cycles;
  
  if (took > wq->worst_usec)
    wq->worst_usec = took;
    
  /* we yielded, check whether granularity should be reduced */
  if (yielded && (cycles < wq->cycles.granularity))
    {
      wq->cycles.granularity = ((cycles > 0) ? cycles 
                                             : WORK_QUEUE_MIN_GRANULARITY);
      if (cycles < wq->cycles.worst)
        wq->cycles.worst = cycles;
    }
  /* otherwise, should granularity increase? */
  else if (cycles >= (wq->cycles.granularity))
    {
      /* along with yielded check, provides hysteresis for granularity */      
      if (cycles > (wq->cycles.granularity * WQ_HYSTERESIS_FACTOR
                                           * WQ_HYSTERESIS_FACTOR))
        wq->cycles.granularity *= WQ_HYSTERESIS_FACTOR; /* quick ramp-up */
      else if (cycles > (wq->cycles.granularity * WQ_HYSTERESIS_FACTOR))
        wq->cycles.granularity += WQ_HYSTERESIS_FACTOR;
        
      /* clamp granularity down to the worst yielded cycle count */
      wq->cycles.granularity = MIN(wq->cycles.granularity, wq->cycles.worst);
    }
#undef WQ_HYSTERIS_FACTOR
  
  wq->runs++;
  wq->cycles.total += cycles;

#if 0
  printf ("%s: cycles %d, new: best %d, worst %d\n",
            __func__, cycles, wq->cycles.best, wq->cycles.granularity);
#endif
  
  /* Is the queue done yet? If it is, call the completion callback. */
  if (listcount (wq->items) > 0)
    work_queue_schedule (wq, 0);
  else if (wq->spec.completion_func)
    wq->spec.completion_func (wq);
  
  return 0;
}