libeio/etp.c

/*
 * libetp implementation
 *
 * Copyright (c) 2007,2008,2009,2010,2011,2012,2013 Marc Alexander Lehmann <libetp@schmorp.de>
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without modifica-
 * tion, are permitted provided that the following conditions are met:
 *
 *   1.  Redistributions of source code must retain the above copyright notice,
 *       this list of conditions and the following disclaimer.
 *
 *   2.  Redistributions in binary form must reproduce the above copyright
 *       notice, this list of conditions and the following disclaimer in the
 *       documentation and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED
 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MER-
 * CHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO
 * EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPE-
 * CIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTH-
 * ERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
 * OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 * Alternatively, the contents of this file may be used under the terms of
 * the GNU General Public License ("GPL") version 2 or any later version,
 * in which case the provisions of the GPL are applicable instead of
 * the above. If you wish to allow the use of your version of this file
 * only under the terms of the GPL and not to allow others to use your
 * version of this file under the BSD license, indicate your decision
 * by deleting the provisions above and replace them with the notice
 * and other provisions required by the GPL. If you do not delete the
 * provisions above, a recipient may use your version of this file under
 * either the BSD or the GPL.
 */

#ifndef ETP_API_DECL
# define ETP_API_DECL static
#endif

#ifndef ETP_PRI_MIN
# define ETP_PRI_MIN 0
# define ETP_PRI_MAX 0
#endif

#ifndef ETP_TYPE_QUIT
# define ETP_TYPE_QUIT 0
#endif

#ifndef ETP_TYPE_GROUP
# define ETP_TYPE_GROUP 1
#endif

#define ETP_NUM_PRI (ETP_PRI_MAX - ETP_PRI_MIN + 1)

#define ETP_TICKS ((1000000 + 1023) >> 10)

enum {
  ETP_FLAG_GROUPADD = 0x04, /* some request was added to the group */
  ETP_FLAG_DELAYED  = 0x08, /* groiup request has been delayed */
};

/* calculate time difference in ~1/ETP_TICKS of a second */
ecb_inline int
etp_tvdiff (struct timeval *tv1, struct timeval *tv2)
{
  return  (tv2->tv_sec  - tv1->tv_sec ) * ETP_TICKS
       + ((tv2->tv_usec - tv1->tv_usec) >> 10);
}

static unsigned int started, idle, wanted = 4;

static void (*want_poll_cb) (void);
static void (*done_poll_cb) (void);
 
static unsigned int max_poll_time;     /* reslock */
static unsigned int max_poll_reqs;     /* reslock */

static unsigned int nreqs;    /* reqlock */
static unsigned int nready;   /* reqlock */
static unsigned int npending; /* reqlock */
static unsigned int max_idle = 4;      /* maximum number of threads that can idle indefinitely */
static unsigned int idle_timeout = 10; /* number of seconds after which an idle threads exit */

static xmutex_t wrklock;
static xmutex_t reslock;
static xmutex_t reqlock;
static xcond_t  reqwait;

struct etp_tmpbuf
{
  void *ptr;
  int len;
};

static void *
etp_tmpbuf_get (struct etp_tmpbuf *buf, int len)
{
  if (buf->len < len)
    {
      free (buf->ptr);
      buf->ptr = malloc (buf->len = len);
    }

  return buf->ptr;
}

typedef struct etp_worker
{
  struct etp_tmpbuf tmpbuf;

  /* locked by wrklock */
  struct etp_worker *prev, *next;

  xthread_t tid;

#ifdef ETP_WORKER_COMMON
  ETP_WORKER_COMMON
#endif
} etp_worker;

static etp_worker wrk_first; /* NOT etp */

#define ETP_WORKER_LOCK(wrk)   X_LOCK   (wrklock)
#define ETP_WORKER_UNLOCK(wrk) X_UNLOCK (wrklock)

/* worker threads management */

static void
etp_worker_clear (etp_worker *wrk)
{
}

static void ecb_cold
etp_worker_free (etp_worker *wrk)
{
  free (wrk->tmpbuf.ptr);

  wrk->next->prev = wrk->prev;
  wrk->prev->next = wrk->next;

  free (wrk);
}

ETP_API_DECL unsigned int
etp_nreqs (void)
{
  int retval;
  if (WORDACCESS_UNSAFE) X_LOCK   (reqlock);
  retval = nreqs;
  if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);
  return retval;
}

ETP_API_DECL unsigned int
etp_nready (void)
{
  unsigned int retval;

  if (WORDACCESS_UNSAFE) X_LOCK   (reqlock);
  retval = nready;
  if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);

  return retval;
}

ETP_API_DECL unsigned int
etp_npending (void)
{
  unsigned int retval;

  if (WORDACCESS_UNSAFE) X_LOCK   (reqlock);
  retval = npending;
  if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);

  return retval;
}

ETP_API_DECL unsigned int
etp_nthreads (void)
{
  unsigned int retval;

  if (WORDACCESS_UNSAFE) X_LOCK   (reqlock);
  retval = started;
  if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);

  return retval;
}

/*
 * a somewhat faster data structure might be nice, but
 * with 8 priorities this actually needs <20 insns
 * per shift, the most expensive operation.
 */
typedef struct {
  ETP_REQ *qs[ETP_NUM_PRI], *qe[ETP_NUM_PRI]; /* qstart, qend */
  int size;
} etp_reqq;

static etp_reqq req_queue;
static etp_reqq res_queue;

static void ecb_noinline ecb_cold
reqq_init (etp_reqq *q)
{
  int pri;

  for (pri = 0; pri < ETP_NUM_PRI; ++pri)
    q->qs[pri] = q->qe[pri] = 0;

  q->size = 0;
}

static int ecb_noinline
reqq_push (etp_reqq *q, ETP_REQ *req)
{
  int pri = req->pri;
  req->next = 0;

  if (q->qe[pri])
    {
      q->qe[pri]->next = req;
      q->qe[pri] = req;
    }
  else
    q->qe[pri] = q->qs[pri] = req;

  return q->size++;
}

static ETP_REQ * ecb_noinline
reqq_shift (etp_reqq *q)
{
  int pri;

  if (!q->size)
    return 0;

  --q->size;

  for (pri = ETP_NUM_PRI; pri--; )
    {
      ETP_REQ *req = q->qs[pri];

      if (req)
        {
          if (!(q->qs[pri] = (ETP_REQ *)req->next))
            q->qe[pri] = 0;

          return req;
        }
    }

  abort ();
}

ETP_API_DECL int ecb_cold
etp_init (void (*want_poll)(void), void (*done_poll)(void))
{
  X_MUTEX_CREATE (wrklock);
  X_MUTEX_CREATE (reslock);
  X_MUTEX_CREATE (reqlock);
  X_COND_CREATE  (reqwait);

  reqq_init (&req_queue);
  reqq_init (&res_queue);

  wrk_first.next =
  wrk_first.prev = &wrk_first;

  started  = 0;
  idle     = 0;
  nreqs    = 0;
  nready   = 0;
  npending = 0;

  want_poll_cb = want_poll;
  done_poll_cb = done_poll;

  return 0;
}

static void ecb_noinline ecb_cold
etp_proc_init (void)
{
#if HAVE_PRCTL_SET_NAME
  /* provide a more sensible "thread name" */
  char name[16 + 1];
  const int namelen = sizeof (name) - 1;
  int len;

  prctl (PR_GET_NAME, (unsigned long)name, 0, 0, 0);
  name [namelen] = 0;
  len = strlen (name);
  strcpy (name + (len <= namelen - 4 ? len : namelen - 4), "/eio");
  prctl (PR_SET_NAME, (unsigned long)name, 0, 0, 0);
#endif
}

X_THREAD_PROC (etp_proc)
{
  ETP_REQ *req;
  struct timespec ts;
  etp_worker *self = (etp_worker *)thr_arg;

  etp_proc_init ();

  /* try to distribute timeouts somewhat evenly */
  ts.tv_nsec = ((unsigned long)self & 1023UL) * (1000000000UL / 1024UL);

  for (;;)
    {
      ts.tv_sec = 0;

      X_LOCK (reqlock);

      for (;;)
        {
          req = reqq_shift (&req_queue);

          if (req)
            break;

          if (ts.tv_sec == 1) /* no request, but timeout detected, let's quit */
            {
              X_UNLOCK (reqlock);
              X_LOCK (wrklock);
              --started;
              X_UNLOCK (wrklock);
              goto quit;
            }

          ++idle;

          if (idle <= max_idle)
            /* we are allowed to idle, so do so without any timeout */
            X_COND_WAIT (reqwait, reqlock);
          else
            {
              /* initialise timeout once */
              if (!ts.tv_sec)
                ts.tv_sec = time (0) + idle_timeout;

              if (X_COND_TIMEDWAIT (reqwait, reqlock, ts) == ETIMEDOUT)
                ts.tv_sec = 1; /* assuming this is not a value computed above.,.. */
            }

          --idle;
        }

      --nready;

      X_UNLOCK (reqlock);
     
      if (req->type == ETP_TYPE_QUIT)
        goto quit;

      ETP_EXECUTE (self, req);

      X_LOCK (reslock);

      ++npending;

      if (!reqq_push (&res_queue, req) && want_poll_cb)
        want_poll_cb ();

      etp_worker_clear (self);

      X_UNLOCK (reslock);
    }

quit:
  free (req);

  X_LOCK (wrklock);
  etp_worker_free (self);
  X_UNLOCK (wrklock);

  return 0;
}

static void ecb_cold
etp_start_thread (void)
{
  etp_worker *wrk = calloc (1, sizeof (etp_worker));

  /*TODO*/
  assert (("unable to allocate worker thread data", wrk));

  X_LOCK (wrklock);

  if (xthread_create (&wrk->tid, etp_proc, (void *)wrk))
    {
      wrk->prev = &wrk_first;
      wrk->next = wrk_first.next;
      wrk_first.next->prev = wrk;
      wrk_first.next = wrk;
      ++started;
    }
  else
    free (wrk);

  X_UNLOCK (wrklock);
}

static void
etp_maybe_start_thread (void)
{
  if (ecb_expect_true (etp_nthreads () >= wanted))
    return;
  
  /* todo: maybe use idle here, but might be less exact */
  if (ecb_expect_true (0 <= (int)etp_nthreads () + (int)etp_npending () - (int)etp_nreqs ()))
    return;

  etp_start_thread ();
}

static void ecb_cold
etp_end_thread (void)
{
  ETP_REQ *req = calloc (1, sizeof (ETP_REQ)); /* will be freed by worker */

  req->type = ETP_TYPE_QUIT;
  req->pri  = ETP_PRI_MAX - ETP_PRI_MIN;

  X_LOCK (reqlock);
  reqq_push (&req_queue, req);
  X_COND_SIGNAL (reqwait);
  X_UNLOCK (reqlock);

  X_LOCK (wrklock);
  --started;
  X_UNLOCK (wrklock);
}

ETP_API_DECL int
etp_poll (void)
{
  unsigned int maxreqs;
  unsigned int maxtime;
  struct timeval tv_start, tv_now;

  X_LOCK (reslock);
  maxreqs = max_poll_reqs;
  maxtime = max_poll_time;
  X_UNLOCK (reslock);

  if (maxtime)
    gettimeofday (&tv_start, 0);

  for (;;)
    {
      ETP_REQ *req;

      etp_maybe_start_thread ();

      X_LOCK (reslock);
      req = reqq_shift (&res_queue);

      if (req)
        {
          --npending;

          if (!res_queue.size && done_poll_cb)
            done_poll_cb ();
        }

      X_UNLOCK (reslock);

      if (!req)
        return 0;

      X_LOCK (reqlock);
      --nreqs;
      X_UNLOCK (reqlock);

      if (ecb_expect_false (req->type == ETP_TYPE_GROUP && req->size))
        {
          req->flags |= ETP_FLAG_DELAYED; /* mark request as delayed */
          continue;
        }
      else
        {
          int res = ETP_FINISH (req);
          if (ecb_expect_false (res))
            return res;
        }

      if (ecb_expect_false (maxreqs && !--maxreqs))
        break;

      if (maxtime)
        {
          gettimeofday (&tv_now, 0);

          if (etp_tvdiff (&tv_start, &tv_now) >= maxtime)
            break;
        }
    }

  errno = EAGAIN;
  return -1;
}

ETP_API_DECL void
etp_grp_cancel (ETP_REQ *grp);

ETP_API_DECL void
etp_cancel (ETP_REQ *req)
{
  req->cancelled = 1;

  etp_grp_cancel (req);
}

ETP_API_DECL void
etp_grp_cancel (ETP_REQ *grp)
{
  for (grp = grp->grp_first; grp; grp = grp->grp_next)
    etp_cancel (grp);
}

ETP_API_DECL void
etp_submit (ETP_REQ *req)
{
  req->pri -= ETP_PRI_MIN;

  if (ecb_expect_false (req->pri < ETP_PRI_MIN - ETP_PRI_MIN)) req->pri = ETP_PRI_MIN - ETP_PRI_MIN;
  if (ecb_expect_false (req->pri > ETP_PRI_MAX - ETP_PRI_MIN)) req->pri = ETP_PRI_MAX - ETP_PRI_MIN;

  if (ecb_expect_false (req->type == ETP_TYPE_GROUP))
    {
      /* I hope this is worth it :/ */
      X_LOCK (reqlock);
      ++nreqs;
      X_UNLOCK (reqlock);

      X_LOCK (reslock);

      ++npending;

      if (!reqq_push (&res_queue, req) && want_poll_cb)
        want_poll_cb ();

      X_UNLOCK (reslock);
    }
  else
    {
      X_LOCK (reqlock);
      ++nreqs;
      ++nready;
      reqq_push (&req_queue, req);
      X_COND_SIGNAL (reqwait);
      X_UNLOCK (reqlock);

      etp_maybe_start_thread ();
    }
}

ETP_API_DECL void ecb_cold
etp_set_max_poll_time (double nseconds)
{
  if (WORDACCESS_UNSAFE) X_LOCK   (reslock);
  max_poll_time = nseconds * ETP_TICKS;
  if (WORDACCESS_UNSAFE) X_UNLOCK (reslock);
}

ETP_API_DECL void ecb_cold
etp_set_max_poll_reqs (unsigned int maxreqs)
{
  if (WORDACCESS_UNSAFE) X_LOCK   (reslock);
  max_poll_reqs = maxreqs;
  if (WORDACCESS_UNSAFE) X_UNLOCK (reslock);
}

ETP_API_DECL void ecb_cold
etp_set_max_idle (unsigned int nthreads)
{
  if (WORDACCESS_UNSAFE) X_LOCK   (reqlock);
  max_idle = nthreads;
  if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);
}

ETP_API_DECL void ecb_cold
etp_set_idle_timeout (unsigned int seconds)
{
  if (WORDACCESS_UNSAFE) X_LOCK   (reqlock);
  idle_timeout = seconds;
  if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);
}

ETP_API_DECL void ecb_cold
etp_set_min_parallel (unsigned int nthreads)
{
  if (wanted < nthreads)
    wanted = nthreads;
}

ETP_API_DECL void ecb_cold
etp_set_max_parallel (unsigned int nthreads)
{
  if (wanted > nthreads)
    wanted = nthreads;

  while (started > wanted)
    etp_end_thread ();
}

Revision:	1.4
Committed:	Thu Jun 25 17:05:07 2015 UTC (8 years, 11 months ago) by root
Content type:	text/plain
Branch:	MAIN
Changes since 1.3:	+104 -3 lines
Log Message:	* empty log message *
#	User	Rev	Content
1	root	1.1	/*
2			* libetp implementation
3			*
4			* Copyright (c) 2007,2008,2009,2010,2011,2012,2013 Marc Alexander Lehmann <libetp@schmorp.de>
5			* All rights reserved.
6			*
7			* Redistribution and use in source and binary forms, with or without modifica-
8			* tion, are permitted provided that the following conditions are met:
9			*
10			* 1. Redistributions of source code must retain the above copyright notice,
11			* this list of conditions and the following disclaimer.
12			*
13			* 2. Redistributions in binary form must reproduce the above copyright
14			* notice, this list of conditions and the following disclaimer in the
15			* documentation and/or other materials provided with the distribution.
16			*
17			* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED
18			* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MER-
19			* CHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO
20			* EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPE-
21			* CIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
22			* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
23			* OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
24			* WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTH-
25			* ERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
26			* OF THE POSSIBILITY OF SUCH DAMAGE.
27			*
28			* Alternatively, the contents of this file may be used under the terms of
29			* the GNU General Public License ("GPL") version 2 or any later version,
30			* in which case the provisions of the GPL are applicable instead of
31			* the above. If you wish to allow the use of your version of this file
32			* only under the terms of the GPL and not to allow others to use your
33			* version of this file under the BSD license, indicate your decision
34			* by deleting the provisions above and replace them with the notice
35			* and other provisions required by the GPL. If you do not delete the
36			* provisions above, a recipient may use your version of this file under
37			* either the BSD or the GPL.
38			*/
39
40			#ifndef ETP_API_DECL
41			# define ETP_API_DECL static
42			#endif
43
44			#ifndef ETP_PRI_MIN
45			# define ETP_PRI_MIN 0
46			# define ETP_PRI_MAX 0
47			#endif
48
49			#ifndef ETP_TYPE_QUIT
50			# define ETP_TYPE_QUIT 0
51			#endif
52
53			#ifndef ETP_TYPE_GROUP
54			# define ETP_TYPE_GROUP 1
55			#endif
56
57			#define ETP_NUM_PRI (ETP_PRI_MAX - ETP_PRI_MIN + 1)
58
59			#define ETP_TICKS ((1000000 + 1023) >> 10)
60
61	root	1.4	enum {
62			ETP_FLAG_GROUPADD = 0x04, /* some request was added to the group */
63			ETP_FLAG_DELAYED = 0x08, /* groiup request has been delayed */
64			};
65
66	root	1.1	/* calculate time difference in ~1/ETP_TICKS of a second */
67			ecb_inline int
68			etp_tvdiff (struct timeval tv1, struct timeval tv2)
69			{
70			return (tv2->tv_sec - tv1->tv_sec ) * ETP_TICKS
71			+ ((tv2->tv_usec - tv1->tv_usec) >> 10);
72			}
73
74			static unsigned int started, idle, wanted = 4;
75
76			static void (*want_poll_cb) (void);
77			static void (*done_poll_cb) (void);
78
79			static unsigned int max_poll_time; /* reslock */
80			static unsigned int max_poll_reqs; /* reslock */
81
82			static unsigned int nreqs; /* reqlock */
83			static unsigned int nready; /* reqlock */
84			static unsigned int npending; /* reqlock */
85			static unsigned int max_idle = 4; /* maximum number of threads that can idle indefinitely */
86			static unsigned int idle_timeout = 10; /* number of seconds after which an idle threads exit */
87
88			static xmutex_t wrklock;
89			static xmutex_t reslock;
90			static xmutex_t reqlock;
91			static xcond_t reqwait;
92
93	root	1.3	struct etp_tmpbuf
94			{
95			void *ptr;
96			int len;
97			};
98
99			static void *
100			etp_tmpbuf_get (struct etp_tmpbuf *buf, int len)
101			{
102			if (buf->len < len)
103			{
104			free (buf->ptr);
105			buf->ptr = malloc (buf->len = len);
106			}
107
108			return buf->ptr;
109			}
110
111	root	1.1	typedef struct etp_worker
112			{
113	root	1.3	struct etp_tmpbuf tmpbuf;
114	root	1.1
115			/* locked by wrklock */
116			struct etp_worker prev, next;
117
118			xthread_t tid;
119
120			#ifdef ETP_WORKER_COMMON
121			ETP_WORKER_COMMON
122			#endif
123			} etp_worker;
124
125			static etp_worker wrk_first; /* NOT etp */
126
127			#define ETP_WORKER_LOCK(wrk) X_LOCK (wrklock)
128			#define ETP_WORKER_UNLOCK(wrk) X_UNLOCK (wrklock)
129
130			/* worker threads management */
131
132			static void
133			etp_worker_clear (etp_worker *wrk)
134			{
135			}
136
137			static void ecb_cold
138			etp_worker_free (etp_worker *wrk)
139			{
140			free (wrk->tmpbuf.ptr);
141
142			wrk->next->prev = wrk->prev;
143			wrk->prev->next = wrk->next;
144
145			free (wrk);
146			}
147
148			ETP_API_DECL unsigned int
149			etp_nreqs (void)
150			{
151			int retval;
152			if (WORDACCESS_UNSAFE) X_LOCK (reqlock);
153			retval = nreqs;
154			if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);
155			return retval;
156			}
157
158			ETP_API_DECL unsigned int
159			etp_nready (void)
160			{
161			unsigned int retval;
162
163			if (WORDACCESS_UNSAFE) X_LOCK (reqlock);
164			retval = nready;
165			if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);
166
167			return retval;
168			}
169
170			ETP_API_DECL unsigned int
171			etp_npending (void)
172			{
173			unsigned int retval;
174
175			if (WORDACCESS_UNSAFE) X_LOCK (reqlock);
176			retval = npending;
177			if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);
178
179			return retval;
180			}
181
182			ETP_API_DECL unsigned int
183			etp_nthreads (void)
184			{
185			unsigned int retval;
186
187			if (WORDACCESS_UNSAFE) X_LOCK (reqlock);
188			retval = started;
189			if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);
190
191			return retval;
192			}
193
194			/*
195			* a somewhat faster data structure might be nice, but
196			* with 8 priorities this actually needs <20 insns
197			* per shift, the most expensive operation.
198			*/
199			typedef struct {
200			ETP_REQ qs[ETP_NUM_PRI], qe[ETP_NUM_PRI]; /* qstart, qend */
201			int size;
202			} etp_reqq;
203
204			static etp_reqq req_queue;
205			static etp_reqq res_queue;
206
207			static void ecb_noinline ecb_cold
208			reqq_init (etp_reqq *q)
209			{
210			int pri;
211
212			for (pri = 0; pri < ETP_NUM_PRI; ++pri)
213			q->qs[pri] = q->qe[pri] = 0;
214
215			q->size = 0;
216			}
217
218			static int ecb_noinline
219			reqq_push (etp_reqq q, ETP_REQ req)
220			{
221			int pri = req->pri;
222			req->next = 0;
223
224			if (q->qe[pri])
225			{
226			q->qe[pri]->next = req;
227			q->qe[pri] = req;
228			}
229			else
230			q->qe[pri] = q->qs[pri] = req;
231
232			return q->size++;
233			}
234
235			static ETP_REQ * ecb_noinline
236			reqq_shift (etp_reqq *q)
237			{
238			int pri;
239
240			if (!q->size)
241			return 0;
242
243			--q->size;
244
245			for (pri = ETP_NUM_PRI; pri--; )
246			{
247	root	1.2	ETP_REQ *req = q->qs[pri];
248	root	1.1
249			if (req)
250			{
251	root	1.2	if (!(q->qs[pri] = (ETP_REQ *)req->next))
252	root	1.1	q->qe[pri] = 0;
253
254			return req;
255			}
256			}
257
258			abort ();
259			}
260
261			ETP_API_DECL int ecb_cold
262			etp_init (void (want_poll)(void), void (done_poll)(void))
263			{
264			X_MUTEX_CREATE (wrklock);
265			X_MUTEX_CREATE (reslock);
266			X_MUTEX_CREATE (reqlock);
267			X_COND_CREATE (reqwait);
268
269			reqq_init (&req_queue);
270			reqq_init (&res_queue);
271
272			wrk_first.next =
273			wrk_first.prev = &wrk_first;
274
275			started = 0;
276			idle = 0;
277			nreqs = 0;
278			nready = 0;
279			npending = 0;
280
281			want_poll_cb = want_poll;
282			done_poll_cb = done_poll;
283
284			return 0;
285			}
286
287	root	1.4	static void ecb_noinline ecb_cold
288			etp_proc_init (void)
289			{
290			#if HAVE_PRCTL_SET_NAME
291			/* provide a more sensible "thread name" */
292			char name[16 + 1];
293			const int namelen = sizeof (name) - 1;
294			int len;
295
296			prctl (PR_GET_NAME, (unsigned long)name, 0, 0, 0);
297			name [namelen] = 0;
298			len = strlen (name);
299			strcpy (name + (len <= namelen - 4 ? len : namelen - 4), "/eio");
300			prctl (PR_SET_NAME, (unsigned long)name, 0, 0, 0);
301			#endif
302			}
303
304			X_THREAD_PROC (etp_proc)
305			{
306			ETP_REQ *req;
307			struct timespec ts;
308			etp_worker self = (etp_worker )thr_arg;
309
310			etp_proc_init ();
311
312			/* try to distribute timeouts somewhat evenly */
313			ts.tv_nsec = ((unsigned long)self & 1023UL) * (1000000000UL / 1024UL);
314
315			for (;;)
316			{
317			ts.tv_sec = 0;
318
319			X_LOCK (reqlock);
320
321			for (;;)
322			{
323			req = reqq_shift (&req_queue);
324
325			if (req)
326			break;
327
328			if (ts.tv_sec == 1) /* no request, but timeout detected, let's quit */
329			{
330			X_UNLOCK (reqlock);
331			X_LOCK (wrklock);
332			--started;
333			X_UNLOCK (wrklock);
334			goto quit;
335			}
336
337			++idle;
338
339			if (idle <= max_idle)
340			/* we are allowed to idle, so do so without any timeout */
341			X_COND_WAIT (reqwait, reqlock);
342			else
343			{
344			/* initialise timeout once */
345			if (!ts.tv_sec)
346			ts.tv_sec = time (0) + idle_timeout;
347
348			if (X_COND_TIMEDWAIT (reqwait, reqlock, ts) == ETIMEDOUT)
349			ts.tv_sec = 1; /* assuming this is not a value computed above.,.. */
350			}
351
352			--idle;
353			}
354
355			--nready;
356
357			X_UNLOCK (reqlock);
358
359			if (req->type == ETP_TYPE_QUIT)
360			goto quit;
361
362			ETP_EXECUTE (self, req);
363
364			X_LOCK (reslock);
365
366			++npending;
367
368			if (!reqq_push (&res_queue, req) && want_poll_cb)
369			want_poll_cb ();
370
371			etp_worker_clear (self);
372
373			X_UNLOCK (reslock);
374			}
375
376			quit:
377			free (req);
378
379			X_LOCK (wrklock);
380			etp_worker_free (self);
381			X_UNLOCK (wrklock);
382
383			return 0;
384			}
385	root	1.1
386			static void ecb_cold
387			etp_start_thread (void)
388			{
389			etp_worker *wrk = calloc (1, sizeof (etp_worker));
390
391			/TODO/
392			assert (("unable to allocate worker thread data", wrk));
393
394			X_LOCK (wrklock);
395
396			if (xthread_create (&wrk->tid, etp_proc, (void *)wrk))
397			{
398			wrk->prev = &wrk_first;
399			wrk->next = wrk_first.next;
400			wrk_first.next->prev = wrk;
401			wrk_first.next = wrk;
402			++started;
403			}
404			else
405			free (wrk);
406
407			X_UNLOCK (wrklock);
408			}
409
410			static void
411			etp_maybe_start_thread (void)
412			{
413			if (ecb_expect_true (etp_nthreads () >= wanted))
414			return;
415
416			/* todo: maybe use idle here, but might be less exact */
417			if (ecb_expect_true (0 <= (int)etp_nthreads () + (int)etp_npending () - (int)etp_nreqs ()))
418			return;
419
420			etp_start_thread ();
421			}
422
423			static void ecb_cold
424			etp_end_thread (void)
425			{
426	root	1.2	ETP_REQ req = calloc (1, sizeof (ETP_REQ)); / will be freed by worker */
427	root	1.1
428			req->type = ETP_TYPE_QUIT;
429			req->pri = ETP_PRI_MAX - ETP_PRI_MIN;
430
431			X_LOCK (reqlock);
432			reqq_push (&req_queue, req);
433			X_COND_SIGNAL (reqwait);
434			X_UNLOCK (reqlock);
435
436			X_LOCK (wrklock);
437			--started;
438			X_UNLOCK (wrklock);
439			}
440
441			ETP_API_DECL int
442			etp_poll (void)
443			{
444			unsigned int maxreqs;
445			unsigned int maxtime;
446			struct timeval tv_start, tv_now;
447
448			X_LOCK (reslock);
449			maxreqs = max_poll_reqs;
450			maxtime = max_poll_time;
451			X_UNLOCK (reslock);
452
453			if (maxtime)
454			gettimeofday (&tv_start, 0);
455
456			for (;;)
457			{
458			ETP_REQ *req;
459
460			etp_maybe_start_thread ();
461
462			X_LOCK (reslock);
463			req = reqq_shift (&res_queue);
464
465			if (req)
466			{
467			--npending;
468
469			if (!res_queue.size && done_poll_cb)
470			done_poll_cb ();
471			}
472
473			X_UNLOCK (reslock);
474
475			if (!req)
476			return 0;
477
478			X_LOCK (reqlock);
479			--nreqs;
480			X_UNLOCK (reqlock);
481
482			if (ecb_expect_false (req->type == ETP_TYPE_GROUP && req->size))
483			{
484	root	1.4	req->flags \|= ETP_FLAG_DELAYED; /* mark request as delayed */
485	root	1.1	continue;
486			}
487			else
488			{
489			int res = ETP_FINISH (req);
490			if (ecb_expect_false (res))
491			return res;
492			}
493
494			if (ecb_expect_false (maxreqs && !--maxreqs))
495			break;
496
497			if (maxtime)
498			{
499			gettimeofday (&tv_now, 0);
500
501			if (etp_tvdiff (&tv_start, &tv_now) >= maxtime)
502			break;
503			}
504			}
505
506			errno = EAGAIN;
507			return -1;
508			}
509
510			ETP_API_DECL void
511			etp_grp_cancel (ETP_REQ *grp);
512
513			ETP_API_DECL void
514			etp_cancel (ETP_REQ *req)
515			{
516			req->cancelled = 1;
517
518			etp_grp_cancel (req);
519			}
520
521			ETP_API_DECL void
522			etp_grp_cancel (ETP_REQ *grp)
523			{
524			for (grp = grp->grp_first; grp; grp = grp->grp_next)
525			etp_cancel (grp);
526			}
527
528			ETP_API_DECL void
529			etp_submit (ETP_REQ *req)
530			{
531			req->pri -= ETP_PRI_MIN;
532
533			if (ecb_expect_false (req->pri < ETP_PRI_MIN - ETP_PRI_MIN)) req->pri = ETP_PRI_MIN - ETP_PRI_MIN;
534			if (ecb_expect_false (req->pri > ETP_PRI_MAX - ETP_PRI_MIN)) req->pri = ETP_PRI_MAX - ETP_PRI_MIN;
535
536			if (ecb_expect_false (req->type == ETP_TYPE_GROUP))
537			{
538			/* I hope this is worth it :/ */
539			X_LOCK (reqlock);
540			++nreqs;
541			X_UNLOCK (reqlock);
542
543			X_LOCK (reslock);
544
545			++npending;
546
547			if (!reqq_push (&res_queue, req) && want_poll_cb)
548			want_poll_cb ();
549
550			X_UNLOCK (reslock);
551			}
552			else
553			{
554			X_LOCK (reqlock);
555			++nreqs;
556			++nready;
557			reqq_push (&req_queue, req);
558			X_COND_SIGNAL (reqwait);
559			X_UNLOCK (reqlock);
560
561			etp_maybe_start_thread ();
562			}
563			}
564
565			ETP_API_DECL void ecb_cold
566			etp_set_max_poll_time (double nseconds)
567			{
568			if (WORDACCESS_UNSAFE) X_LOCK (reslock);
569			max_poll_time = nseconds * ETP_TICKS;
570			if (WORDACCESS_UNSAFE) X_UNLOCK (reslock);
571			}
572
573			ETP_API_DECL void ecb_cold
574			etp_set_max_poll_reqs (unsigned int maxreqs)
575			{
576			if (WORDACCESS_UNSAFE) X_LOCK (reslock);
577			max_poll_reqs = maxreqs;
578			if (WORDACCESS_UNSAFE) X_UNLOCK (reslock);
579			}
580
581			ETP_API_DECL void ecb_cold
582			etp_set_max_idle (unsigned int nthreads)
583			{
584			if (WORDACCESS_UNSAFE) X_LOCK (reqlock);
585			max_idle = nthreads;
586			if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);
587			}
588
589			ETP_API_DECL void ecb_cold
590			etp_set_idle_timeout (unsigned int seconds)
591			{
592			if (WORDACCESS_UNSAFE) X_LOCK (reqlock);
593			idle_timeout = seconds;
594			if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);
595			}
596
597			ETP_API_DECL void ecb_cold
598			etp_set_min_parallel (unsigned int nthreads)
599			{
600			if (wanted < nthreads)
601			wanted = nthreads;
602			}
603
604			ETP_API_DECL void ecb_cold
605			etp_set_max_parallel (unsigned int nthreads)
606			{
607			if (wanted > nthreads)
608			wanted = nthreads;
609
610			while (started > wanted)
611			etp_end_thread ();
612			}
613