evmap.c

Branch :
Show log
Commit
Author : Nick Mathewson
Date : 2010-04-09 19:16:49
Hash : 819f949f
Message : Limit the maximum number of events on each socket to 65535 This lets us use less RAM for the evmap_io structure, which in turn can let us have fewer cache misses for evmap operations.
evmap.c
/*
 * Copyright (c) 2007-2010 Niels Provos and Nick Mathewson
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, 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.
 * 3. The name of the author may not be used to endorse or promote products
 *    derived from this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
 * INCIDENTAL, SPECIAL, 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 OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */
#include "event-config.h"

#ifdef WIN32
#include <winsock2.h>
#define WIN32_LEAN_AND_MEAN
#include <windows.h>
#undef WIN32_LEAN_AND_MEAN
#endif
#include <sys/types.h>
#if !defined(WIN32) && defined(_EVENT_HAVE_SYS_TIME_H)
#include <sys/time.h>
#endif
#include <sys/queue.h>
#include <stdio.h>
#include <stdlib.h>
#ifndef WIN32
#include <unistd.h>
#endif
#include <errno.h>
#include <signal.h>
#include <string.h>
#include <time.h>

#include "event-internal.h"
#include "evmap-internal.h"
#include "mm-internal.h"
#include "changelist-internal.h"

/** An entry for an evmap_io list: notes all the events that want to read or
	write on a given fd, and the number of each.
  */
struct evmap_io {
	struct event_list events;
	ev_uint16_t nread;
	ev_uint16_t nwrite;
};

/* An entry for an evmap_signal list: notes all the events that want to know
   when a signal triggers. */
struct evmap_signal {
	struct event_list events;
};

/* On some platforms, fds start at 0 and increment by 1 as they are
   allocated, and old numbers get used.  For these platforms, we
   implement io maps just like signal maps: as an array of pointers to
   struct evmap_io.  But on other platforms (windows), sockets are not
   0-indexed, not necessarily consecutive, and not necessarily reused.
   There, we use a hashtable to implement evmap_io.
*/
#ifdef EVMAP_USE_HT
struct event_map_entry {
	HT_ENTRY(event_map_entry) map_node;
	evutil_socket_t fd;
	union { /* This is a union in case we need to make more things that can
			   be in the hashtable. */
		struct evmap_io evmap_io;
	} ent;
};

/* Helper used by the event_io_map hashtable code; tries to return a good hash
 * of the fd in e->fd. */
static inline unsigned
hashsocket(struct event_map_entry *e)
{
	/* On win32, in practice, the low 2-3 bits of a SOCKET seem not to
	 * matter.  Our hashtable implementation really likes low-order bits,
	 * though, so let's do the rotate-and-add trick. */
	unsigned h = (unsigned) e->fd;
	h += (h >> 2) | (h << 30);
	return h;
}

/* Helper used by the event_io_map hashtable code; returns true iff e1 and e2
 * have the same e->fd. */
static inline int
eqsocket(struct event_map_entry *e1, struct event_map_entry *e2)
{
	return e1->fd == e2->fd;
}

HT_PROTOTYPE(event_io_map, event_map_entry, map_node, hashsocket, eqsocket);
HT_GENERATE(event_io_map, event_map_entry, map_node, hashsocket, eqsocket,
			0.5, mm_malloc, mm_realloc, mm_free);

#define GET_IO_SLOT(x, map, slot, type)					\
	do {								\
		struct event_map_entry _key, *_ent;			\
		_key.fd = slot;						\
		_ent = HT_FIND(event_io_map, map, &_key);		\
		(x) = _ent ? &_ent->ent.type : NULL;			\
	} while (0);

#define GET_IO_SLOT_AND_CTOR(x, map, slot, type, ctor, fdinfo_len)	\
	do {								\
		struct event_map_entry _key, *_ent;			\
		_key.fd = slot;						\
		_HT_FIND_OR_INSERT(event_io_map, map_node, hashsocket, map, \
		    event_map_entry, &_key, ptr,			\
		    {							\
			    _ent = *ptr;				\
		    },							\
		    {							\
			    _ent = mm_calloc(1,sizeof(struct event_map_entry)+fdinfo_len); \
			    EVUTIL_ASSERT(_ent);				\
			    _ent->fd = slot;				\
			    (ctor)(&_ent->ent.type);			\
			    _HT_FOI_INSERT(map_node, map, &_key, _ent, ptr) \
				});					\
		(x) = &_ent->ent.type;					\
	} while (0)

void evmap_io_initmap(struct event_io_map *ctx)
{
	HT_INIT(event_io_map, ctx);
}

void evmap_io_clear(struct event_io_map *ctx)
{
	struct event_map_entry **ent, **next, *this;
	for (ent = HT_START(event_io_map, ctx); ent; ent = next) {
		this = *ent;
		next = HT_NEXT_RMV(event_io_map, ctx, ent);
		mm_free(this);
	}
}
#endif

/* Set the variable 'x' to the field in event_map 'map' with fields of type
   'struct type *' corresponding to the fd or signal 'slot'.  Set 'x' to NULL
   if there are no entries for 'slot'.  Does no bounds-checking. */
#define GET_SIGNAL_SLOT(x, map, slot, type)			\
	(x) = (struct type *)((map)->entries[slot])
/* As GET_SLOT, but construct the entry for 'slot' if it is not present,
   by allocating enough memory for a 'struct type', and initializing the new
   value by calling the function 'ctor' on it.
 */
#define GET_SIGNAL_SLOT_AND_CTOR(x, map, slot, type, ctor, fdinfo_len)	\
	do {								\
		if ((map)->entries[slot] == NULL) {			\
			EVUTIL_ASSERT(ctor != NULL);				\
			(map)->entries[slot] =				\
			    mm_calloc(1,sizeof(struct type)+fdinfo_len); \
			EVUTIL_ASSERT((map)->entries[slot] != NULL);		\
			(ctor)((struct type *)(map)->entries[slot]);	\
		}							\
		(x) = (struct type *)((map)->entries[slot]);		\
	} while (0)

/* If we aren't using hashtables, then define the IO_SLOT macros and functions
   as thin aliases over the SIGNAL_SLOT versions. */
#ifndef EVMAP_USE_HT
#define GET_IO_SLOT(x,map,slot,type) GET_SIGNAL_SLOT(x,map,slot,type)
#define GET_IO_SLOT_AND_CTOR(x,map,slot,type,ctor,fdinfo_len)	\
	GET_SIGNAL_SLOT_AND_CTOR(x,map,slot,type,ctor,fdinfo_len)
#define FDINFO_OFFSET sizeof(struct evmap_io)
void
evmap_io_initmap(struct event_io_map* ctx)
{
	evmap_signal_initmap(ctx);
}
void
evmap_io_clear(struct event_io_map* ctx)
{
	evmap_signal_clear(ctx);
}
#endif


/** Expand 'map' with new entries of width 'msize' until it is big enough
	to store a value in 'slot'.
 */
static int
evmap_make_space(struct event_signal_map *map, int slot, int msize)
{
	if (map->nentries <= slot) {
		int nentries = map->nentries ? map->nentries : 32;
		void **tmp;

		while (nentries <= slot)
			nentries <<= 1;

		tmp = (void **)mm_realloc(map->entries, nentries * msize);
		if (tmp == NULL)
			return (-1);

		memset(&tmp[map->nentries], 0,
		    (nentries - map->nentries) * msize);

		map->nentries = nentries;
		map->entries = tmp;
	}

	return (0);
}

void
evmap_signal_initmap(struct event_signal_map *ctx)
{
	ctx->nentries = 0;
	ctx->entries = NULL;
}

void
evmap_signal_clear(struct event_signal_map *ctx)
{
	if (ctx->entries != NULL) {
		int i;
		for (i = 0; i < ctx->nentries; ++i) {
			if (ctx->entries[i] != NULL)
				mm_free(ctx->entries[i]);
		}
		mm_free(ctx->entries);
		ctx->entries = NULL;
	}
	ctx->nentries = 0;
}


/* code specific to file descriptors */

/** Constructor for struct evmap_io */
static void
evmap_io_init(struct evmap_io *entry)
{
	TAILQ_INIT(&entry->events);
	entry->nread = 0;
	entry->nwrite = 0;
}


/* return -1 on error, 0 on success if nothing changed in the event backend,
 * and 1 on success if something did. */
int
evmap_io_add(struct event_base *base, evutil_socket_t fd, struct event *ev)
{
	const struct eventop *evsel = base->evsel;
	struct event_io_map *io = &base->io;
	struct evmap_io *ctx = NULL;
	int nread, nwrite, retval = 0;
	short res = 0, old = 0;

	EVUTIL_ASSERT(fd == ev->ev_fd);

	if (fd < 0)
		return 0;

#ifndef EVMAP_USE_HT
	if (fd >= io->nentries) {
		if (evmap_make_space(io, fd, sizeof(struct evmap_io *)) == -1)
			return (-1);
	}
#endif
	GET_IO_SLOT_AND_CTOR(ctx, io, fd, evmap_io, evmap_io_init,
						 evsel->fdinfo_len);

	nread = ctx->nread;
	nwrite = ctx->nwrite;

	if (nread)
		old |= EV_READ;
	if (nwrite)
		old |= EV_WRITE;

	if (ev->ev_events & EV_READ) {
		if (++nread == 1)
			res |= EV_READ;
	}
	if (ev->ev_events & EV_WRITE) {
		if (++nwrite == 1)
			res |= EV_WRITE;
	}
	if (EVUTIL_UNLIKELY(nread > 0xffff || nwrite > 0xffff)) {
		event_warnx("Too many events reading or writing on fd %d",
		    (int)fd);
		return -1;
	}

	if (res) {
		void *extra = ((char*)ctx) + sizeof(struct evmap_io);
		/* XXX(niels): we cannot mix edge-triggered and
		 * level-triggered, we should probably assert on
		 * this. */
		if (evsel->add(base, ev->ev_fd,
			old, (ev->ev_events & EV_ET) | res, extra) == -1)
			return (-1);
		retval = 1;
	}

	ctx->nread = (ev_uint16_t) nread;
	ctx->nwrite = (ev_uint16_t) nwrite;
	TAILQ_INSERT_TAIL(&ctx->events, ev, ev_io_next);

	return (retval);
}

/* return -1 on error, 0 on success if nothing changed in the event backend,
 * and 1 on success if something did. */
int
evmap_io_del(struct event_base *base, evutil_socket_t fd, struct event *ev)
{
	const struct eventop *evsel = base->evsel;
	struct event_io_map *io = &base->io;
	struct evmap_io *ctx;
	int nread, nwrite, retval = 0;
	short res = 0, old = 0;

	if (fd < 0)
		return 0;

	EVUTIL_ASSERT(fd == ev->ev_fd);

#ifndef EVMAP_USE_HT
	if (fd >= io->nentries)
		return (-1);
#endif

	GET_IO_SLOT(ctx, io, fd, evmap_io);

	nread = ctx->nread;
	nwrite = ctx->nwrite;

	if (nread)
		old |= EV_READ;
	if (nwrite)
		old |= EV_WRITE;

	if (ev->ev_events & EV_READ) {
		if (--nread == 0)
			res |= EV_READ;
		EVUTIL_ASSERT(nread >= 0);
	}
	if (ev->ev_events & EV_WRITE) {
		if (--nwrite == 0)
			res |= EV_WRITE;
		EVUTIL_ASSERT(nwrite >= 0);
	}

	if (res) {
		void *extra = ((char*)ctx) + sizeof(struct evmap_io);
		if (evsel->del(base, ev->ev_fd, old, res, extra) == -1)
			return (-1);
		retval = 1;
	}

	ctx->nread = nread;
	ctx->nwrite = nwrite;
	TAILQ_REMOVE(&ctx->events, ev, ev_io_next);

	return (retval);
}

void
evmap_io_active(struct event_base *base, evutil_socket_t fd, short events)
{
	struct event_io_map *io = &base->io;
	struct evmap_io *ctx;
	struct event *ev;

#ifndef EVMAP_USE_HT
	EVUTIL_ASSERT(fd < io->nentries);
#endif
	GET_IO_SLOT(ctx, io, fd, evmap_io);

	EVUTIL_ASSERT(ctx);
	TAILQ_FOREACH(ev, &ctx->events, ev_io_next) {
		if (ev->ev_events & events)
			event_active_nolock(ev, ev->ev_events & events, 1);
	}
}

/* code specific to signals */

static void
evmap_signal_init(struct evmap_signal *entry)
{
	TAILQ_INIT(&entry->events);
}


int
evmap_signal_add(struct event_base *base, int sig, struct event *ev)
{
	const struct eventop *evsel = base->evsigsel;
	struct event_signal_map *map = &base->sigmap;
	struct evmap_signal *ctx = NULL;

	if (sig >= map->nentries) {
		if (evmap_make_space(
			map, sig, sizeof(struct evmap_signal *)) == -1)
			return (-1);
	}
	GET_SIGNAL_SLOT_AND_CTOR(ctx, map, sig, evmap_signal, evmap_signal_init,
	    base->evsigsel->fdinfo_len);

	if (TAILQ_EMPTY(&ctx->events)) {
		if (evsel->add(base, ev->ev_fd, 0, EV_SIGNAL, NULL)
		    == -1)
			return (-1);
	}

	TAILQ_INSERT_TAIL(&ctx->events, ev, ev_signal_next);

	return (1);
}

int
evmap_signal_del(struct event_base *base, int sig, struct event *ev)
{
	const struct eventop *evsel = base->evsigsel;
	struct event_signal_map *map = &base->sigmap;
	struct evmap_signal *ctx;

	if (sig >= map->nentries)
		return (-1);

	GET_SIGNAL_SLOT(ctx, map, sig, evmap_signal);

	if (TAILQ_FIRST(&ctx->events) == TAILQ_LAST(&ctx->events, event_list)) {
		if (evsel->del(base, ev->ev_fd, 0, EV_SIGNAL, NULL) == -1)
			return (-1);
	}

	TAILQ_REMOVE(&ctx->events, ev, ev_signal_next);

	return (1);
}

void
evmap_signal_active(struct event_base *base, evutil_socket_t sig, int ncalls)
{
	struct event_signal_map *map = &base->sigmap;
	struct evmap_signal *ctx;
	struct event *ev;

	EVUTIL_ASSERT(sig < map->nentries);
	GET_SIGNAL_SLOT(ctx, map, sig, evmap_signal);

	TAILQ_FOREACH(ev, &ctx->events, ev_signal_next)
		event_active_nolock(ev, EV_SIGNAL, ncalls);
}

void *
evmap_io_get_fdinfo(struct event_io_map *map, evutil_socket_t fd)
{
	struct evmap_io *ctx;
	GET_IO_SLOT(ctx, map, fd, evmap_io);
	if (ctx)
		return ((char*)ctx) + sizeof(struct evmap_io);
	else
		return NULL;
}

/** Per-fd structure for use with changelists.  It keeps track, for each fd or
 * signal using the changelist, of where its entry in the changelist is.
 */
struct event_changelist_fdinfo {
	int idxplus1; /* this is the index +1, so that memset(0) will make it
		       * a no-such-element */
};

void
event_changelist_init(struct event_changelist *changelist)
{
	changelist->changes = NULL;
	changelist->changes_size = 0;
	changelist->n_changes = 0;
}

/** Helper: return the changelist_fdinfo corresponding to a given change. */
static inline struct event_changelist_fdinfo *
event_change_get_fdinfo(struct event_base *base,
    const struct event_change *change)
{
	char *ptr;
	if (change->read_change & EV_CHANGE_SIGNAL) {
		struct evmap_signal *ctx;
		GET_SIGNAL_SLOT(ctx, &base->sigmap, change->fd, evmap_signal);
		ptr = ((char*)ctx) + sizeof(struct evmap_signal);
	} else {
		struct evmap_io *ctx;
		GET_IO_SLOT(ctx, &base->io, change->fd, evmap_io);
		ptr = ((char*)ctx) + sizeof(struct evmap_io);
	}
	return (void*)ptr;
}

#ifdef DEBUG_CHANGELIST
/** Make sure that the changelist is consistent with the evmap structures. */
static void
event_changelist_check(struct event_base *base)
{
	int i;
	struct event_changelist *changelist = &base->changelist;

	EVUTIL_ASSERT(changelist->changes_size >= changelist->n_changes);
	for (i = 0; i < changelist->n_changes; ++i) {
		struct event_change *c = &changelist->changes[i];
		struct event_changelist_fdinfo *f;
		EVUTIL_ASSERT(c->fd >= 0);
		f = event_change_get_fdinfo(base, c);
		EVUTIL_ASSERT(f);
		EVUTIL_ASSERT(f->idxplus1 == i + 1);
	}

	for (i = 0; i < base->io.nentries; ++i) {
		struct evmap_io *io = base->io.entries[i];
		struct event_changelist_fdinfo *f;
		if (!io)
			continue;
		f = (void*)
		    ( ((char*)io) + sizeof(struct evmap_io) );
		if (f->idxplus1) {
			struct event_change *c = &changelist->changes[f->idxplus1 - 1];
			EVUTIL_ASSERT(c->fd == i);
		}
	}
}
#else
#define event_changelist_check(base)  ((void)0)
#endif

void
event_changelist_remove_all(struct event_changelist *changelist,
    struct event_base *base)
{
	int i;

	event_changelist_check(base);

	for (i = 0; i < changelist->n_changes; ++i) {
		struct event_change *ch = &changelist->changes[i];
		struct event_changelist_fdinfo *fdinfo =
		    event_change_get_fdinfo(base, ch);
		EVUTIL_ASSERT(fdinfo->idxplus1 == i + 1);
		fdinfo->idxplus1 = 0;
	}

	changelist->n_changes = 0;

	event_changelist_check(base);
}

void
event_changelist_freemem(struct event_changelist *changelist)
{
	if (changelist->changes)
		mm_free(changelist->changes);
	event_changelist_init(changelist); /* zero it all out. */
}

/** Increase the size of 'changelist' to hold more changes. */
static int
event_changelist_grow(struct event_changelist *changelist)
{
	int new_size;
	struct event_change *new_changes;
	if (changelist->changes_size < 64)
		new_size = 64;
	else
		new_size = changelist->changes_size * 2;

	new_changes = mm_realloc(changelist->changes,
	    new_size * sizeof(struct event_change));

	if (EVUTIL_UNLIKELY(new_changes == NULL))
		return (-1);

	changelist->changes = new_changes;
	changelist->changes_size = new_size;

	return (0);
}

/** Return a pointer to the changelist entry for the file descriptor or signal
 * 'fd', whose fdinfo is 'fdinfo'.  If none exists, construct it, setting its
 * old_events field to old_events.
 */
static struct event_change *
event_changelist_get_or_construct(struct event_changelist *changelist,
    evutil_socket_t fd,
    short old_events,
    struct event_changelist_fdinfo *fdinfo)
{
	struct event_change *change;

	if (fdinfo->idxplus1 == 0) {
		int idx;
		EVUTIL_ASSERT(changelist->n_changes <= changelist->changes_size);

		if (changelist->n_changes == changelist->changes_size) {
			if (event_changelist_grow(changelist) < 0)
				return NULL;
		}

		idx = changelist->n_changes++;
		change = &changelist->changes[idx];
		fdinfo->idxplus1 = idx + 1;

		memset(change, 0, sizeof(struct event_change));
		change->fd = fd;
		change->old_events = old_events;
	} else {
		change = &changelist->changes[fdinfo->idxplus1 - 1];
		EVUTIL_ASSERT(change->fd == fd);
	}
	return change;
}

int
event_changelist_add(struct event_base *base, evutil_socket_t fd, short old, short events,
    void *p)
{
	struct event_changelist *changelist = &base->changelist;
	struct event_changelist_fdinfo *fdinfo = p;
	struct event_change *change;

	event_changelist_check(base);

	change = event_changelist_get_or_construct(changelist, fd, old, fdinfo);
	if (!change)
		return -1;

	/* An add replaces any previous delete, but doesn't result in a no-op,
	 * since the delete might fail (because the fd had been closed since
	 * the last add, for instance. */

	if (events & (EV_READ|EV_SIGNAL)) {
		change->read_change = EV_CHANGE_ADD |
		    (events & (EV_ET|EV_PERSIST|EV_SIGNAL));
	}
	if (events & EV_WRITE) {
		change->write_change = EV_CHANGE_ADD |
		    (events & (EV_ET|EV_PERSIST|EV_SIGNAL));
	}

	event_changelist_check(base);
	return (0);
}

int
event_changelist_del(struct event_base *base, evutil_socket_t fd, short old, short events,
    void *p)
{
	struct event_changelist *changelist = &base->changelist;
	struct event_changelist_fdinfo *fdinfo = p;
	struct event_change *change;

	event_changelist_check(base);
	change = event_changelist_get_or_construct(changelist, fd, old, fdinfo);
	event_changelist_check(base);
	if (!change)
		return -1;

	/* A delete removes any previous add, rather than replacing it:
	   on those platforms where "add, delete, dispatch" is not the same
	   as "no-op, dispatch", we want the no-op behavior.

	   If we have a no-op item, we could remove it it from the list
	   entirely, but really there's not much point: skipping the no-op
	   change when we do the dispatch later is far cheaper than rejuggling
	   the array now.
	 */

	if (events & (EV_READ|EV_SIGNAL)) {
		if (change->read_change & EV_CHANGE_ADD)
			change->read_change = 0;
		else
			change->read_change = EV_CHANGE_DEL;
	}
	if (events & EV_WRITE) {
		if (change->write_change & EV_CHANGE_ADD)
			change->write_change = 0;
		else
			change->write_change = EV_CHANGE_DEL;
	}

	event_changelist_check(base);
	return (0);
}
kc3-lang/libevent/evmap.c

Commit

kc3-lang/libevent /evmap.c
