support for Linux eventpoll mechanism

svn:r42

3 files changed, 366 insertions, 0 deletions

diff --git a/epoll.c b/epoll.c
new file mode 100644
index 00000000..95dab907
--- /dev/null
+++ b/epoll.c
@@ -0,0 +1,334 @@
+/*	$OpenBSD: kqueue.c,v 1.5 2002/07/10 14:41:31 art Exp $	*/
+
+/*
+ * Copyright 2000-2002 Niels Provos <provos@citi.umich.edu>
+ * All rights reserved.
+ *
+ * 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. All advertising materials mentioning features or use of this software
+ *    must display the following acknowledgement:
+ *      This product includes software developed by Niels Provos.
+ * 4. 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.
+ */
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+#include <stdint.h>
+#include <sys/types.h>
+#ifdef HAVE_SYS_TIME_H
+#include <sys/time.h>
+#else
+#include <sys/_time.h>
+#endif
+#include <sys/queue.h>
+#include <sys/epoll.h>
+#include <signal.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <errno.h>
+#include <err.h>
+
+#ifdef USE_LOG
+#include "log.h"
+#else
+#define LOG_DBG(x)
+#define log_error	warn
+#endif
+
+#include "event.h"
+
+extern struct event_list eventqueue;
+
+extern volatile sig_atomic_t evsignal_caught;
+
+/* due to limitations in the epoll interface, we need to keep track of 
+ * all file descriptors outself.
+ */
+struct evepoll {
+	struct event *evread;
+	struct event *evwrite;
+};
+
+struct epollop {
+	struct evepoll *fds;
+	int nfds;
+	struct epoll_event *events;
+	int nevents;
+	int epfd;
+	sigset_t evsigmask;
+} epollop;
+
+void evsignal_init(sigset_t *);
+void evsignal_process(void);
+int evsignal_recalc(sigset_t *);
+int evsignal_deliver(void);
+int evsignal_add(sigset_t *, struct event *);
+int evsignal_del(sigset_t *, struct event *);
+
+void *epoll_init	(void);
+int epoll_add	(void *, struct event *);
+int epoll_del	(void *, struct event *);
+int epoll_recalc	(void *, int);
+int epoll_dispatch	(void *, struct timeval *);
+
+struct eventop epollops = {
+	"epoll",
+	epoll_init,
+	epoll_add,
+	epoll_del,
+	epoll_recalc,
+	epoll_dispatch
+};
+
+#define NEVENT	64
+
+void *
+epoll_init(void)
+{
+	int epfd;
+
+	/* Disable epollueue when this environment variable is set */
+	if (getenv("EVENT_NOEPOLL"))
+		return (NULL);
+
+	memset(&epollop, 0, sizeof(epollop));
+
+	/* Initalize the kernel queue */
+	
+	if ((epfd = epoll_create(NEVENT)) == -1) {
+		log_error("epoll_create");
+		return (NULL);
+	}
+
+	epollop.epfd = epfd;
+
+	/* Initalize fields */
+	epollop.events = malloc(NEVENT * sizeof(struct epoll_event));
+	if (epollop.events == NULL)
+		return (NULL);
+	epollop.nevents = NEVENT;
+
+	epollop.fds = calloc(NEVENT, sizeof(struct evepoll));
+	if (epollop.fds == NULL) {
+		free(epollop.events);
+		return (NULL);
+	}
+	epollop.nfds = NEVENT;
+
+	evsignal_init(&epollop.evsigmask);
+
+	return (&epollop);
+}
+
+int
+epoll_recalc(void *arg, int max)
+{
+	struct epollop *epollop = arg;
+
+	if (max > epollop->nfds) {
+		struct evepoll *fds;
+		int nfds;
+
+		nfds = 2*epollop->nfds;
+		if (nfds < max)
+			nfds = max;
+
+		fds = realloc(epollop->fds, nfds * sizeof(struct evepoll));
+		if (fds == NULL) {
+			log_error("realloc");
+			return (-1);
+		}
+		epollop->fds = fds;
+		memset(fds + epollop->nfds, 0,
+		    (nfds - epollop->nfds) * sizeof(struct evepoll));
+		epollop->nfds = nfds;
+	}
+
+	return (evsignal_recalc(&epollop->evsigmask));
+}
+
+int
+epoll_dispatch(void *arg, struct timeval *tv)
+{
+	struct epollop *epollop = arg;
+	struct epoll_event *events = epollop->events;
+	struct evepoll *evep;
+	struct event *ev;
+	int i, res, timeout;
+
+	if (evsignal_deliver() == -1)
+		return (-1);
+
+	timeout = tv->tv_sec * 1000 + tv->tv_usec / 1000;
+	res = epoll_wait(epollop->epfd, events, epollop->nevents, timeout);
+
+	if (evsignal_recalc(&epollop->evsigmask) == -1)
+		return (-1);
+
+	if (res == -1) {
+		if (errno != EINTR) {
+			log_error("epoll_wait");
+			return (-1);
+		}
+
+		evsignal_process();
+		return (0);
+	} else if (evsignal_caught)
+		evsignal_process();
+
+	LOG_DBG((LOG_MISC, 80, "%s: epoll_wait reports %d", __func__, res));
+
+	for (i = 0; i < res; i++) {
+		int which = 0, what;
+
+		evep = (struct evepoll *)events[i].data.ptr;
+		what = events[i].events;
+		if (what & EPOLLIN) {
+			which |= EV_READ;
+		}
+
+		if (what & EPOLLOUT) {
+			which |= EV_WRITE;
+		}
+
+		if (!which)
+			continue;
+
+		if (which & EV_READ) {
+			ev = evep->evread;
+			if (!(ev->ev_events & EV_PERSIST))
+				event_del(ev);
+			event_active(ev, EV_READ, 1);
+		}
+		if (which & EV_WRITE) {
+			ev = evep->evwrite;
+			if (!(ev->ev_events & EV_PERSIST))
+				event_del(ev);
+			event_active(ev, EV_WRITE, 1);
+		}
+	}
+
+	return (0);
+}
+
+
+int
+epoll_add(void *arg, struct event *ev)
+{
+	struct epollop *epollop = arg;
+	struct epoll_event epev;
+	struct evepoll *evep;
+	int fd, op, events;
+
+	if (ev->ev_events & EV_SIGNAL)
+		return (evsignal_add(&epollop->evsigmask, ev));
+
+	fd = ev->ev_fd;
+	if (fd >= epollop->nfds) {
+		/* Extent the file descriptor array as necessary */
+		if (epoll_recalc(epollop, fd) == -1)
+			return (-1);
+	}
+	evep = &epollop->fds[fd];
+	op = EPOLL_CTL_ADD;
+	events = 0;
+	if (evep->evread != NULL) {
+		events |= EPOLLIN;
+		op = EPOLL_CTL_MOD;
+	}
+	if (evep->evwrite != NULL) {
+		events |= EPOLLOUT;
+		op = EPOLL_CTL_MOD;
+	}
+
+	if (ev->ev_events & EV_READ)
+		events |= EPOLLIN;
+	if (ev->ev_events & EV_WRITE)
+		events |= EPOLLOUT;
+
+	epev.data.ptr = evep;
+	epev.events = events;
+	if (epoll_ctl(epollop->epfd, op, ev->ev_fd, &epev) == -1)
+			return (-1);
+
+	/* Update events responsible */
+	if (ev->ev_events & EV_READ)
+		evep->evread = ev;
+	if (ev->ev_events & EV_WRITE)
+		evep->evwrite = ev;
+
+	return (0);
+}
+
+int
+epoll_del(void *arg, struct event *ev)
+{
+	struct epollop *epollop = arg;
+	struct epoll_event epev;
+	struct evepoll *evep;
+	int fd, events, op;
+	int needwritedelete = 1, needreaddelete = 1;
+
+	if (ev->ev_events & EV_SIGNAL)
+		return (evsignal_del(&epollop->evsigmask, ev));
+
+	fd = ev->ev_fd;
+	if (fd >= epollop->nfds)
+		return (0);
+	evep = &epollop->fds[fd];
+
+	op = EPOLL_CTL_DEL;
+	events = 0;
+
+	if (ev->ev_events & EV_READ)
+		events |= EPOLLIN;
+	if (ev->ev_events & EV_WRITE)
+		events |= EPOLLOUT;
+
+	if ((events & (EPOLLIN|EPOLLOUT)) != (EPOLLIN|EPOLLOUT)) {
+		if ((events & EPOLLIN) && evep->evwrite != NULL) {
+			needwritedelete = 0;
+			events = EPOLLOUT;
+			op = EPOLL_CTL_MOD;
+		} else if ((events & EPOLLOUT) && evep->evread != NULL) {
+			needreaddelete = 0;
+			events = EPOLLIN;
+			op = EPOLL_CTL_MOD;
+		}
+	}
+
+	epev.events = events;
+	epev.data.ptr = evep;
+
+	if (epoll_ctl(epollop->epfd, op, ev->ev_fd, &epev) == -1)
+		return (-1);
+
+	if (needreaddelete)
+		evep->evread = NULL;
+	if (needwritedelete)
+		evep->evwrite = NULL;
+	
+	return (0);
+}
diff --git a/epoll_sub.c b/epoll_sub.c
new file mode 100644
index 00000000..598681e7
--- /dev/null
+++ b/epoll_sub.c
@@ -0,0 +1,26 @@
+#include <stdint.h>
+
+#include <sys/param.h>
+#include <sys/types.h>
+#include <sys/syscall.h>
+#include <sys/epoll.h>
+#include <unistd.h>
+
+int
+epoll_create(int size)
+{
+	return (syscall(__NR_epoll_create, size));
+}
+
+int
+epoll_ctl(int epfd, int op, int fd, struct epoll_event *event)
+{
+
+	return (syscall(__NR_epoll_ctl, epfd, op, fd, event));
+}
+
+int
+epoll_wait(int epfd, struct epoll_event *events, int maxevents, int timeout)
+{
+	return (syscall(__NR_epoll_wait, epfd, events, maxevents, timeout));
+}
diff --git a/event.c b/event.c
index 48073c79..1bf545c4 100644
--- a/event.c
+++ b/event.c
@@ -64,6 +64,9 @@ extern struct eventop selectops;
 #ifdef HAVE_POLL
 extern struct eventop pollops;
 #endif
+#ifdef HAVE_EPOLL
+extern struct eventop epollops;
+#endif
 #ifdef HAVE_WORKING_KQUEUE
 extern struct eventop kqops;
 #endif
@@ -73,6 +76,9 @@ struct eventop *eventops[] = {
 #ifdef HAVE_WORKING_KQUEUE
 	&kqops,
 #endif
+#ifdef HAVE_EPOLL
+	&epollops,
+#endif
 #ifdef HAVE_POLL
 	&pollops,
 #endif