libevent/bufferevent_sock.c
Nick Mathewson 02fbf68770 Use finalization feature so bufferevents can avoid deadlocks
Since the bufferevents' events are now EV_FINALIZE (name pending),
they won't deadlock.  To clean up properly, though, we must use the
finalization feature.

This patch also split bufferevent deallocation into an "unlink" step
that happens fast, and a "destruct" step that happens after
finalization.

More work is needed: there needs to be a way to specify a finalizer
for the bufferevent's argument itself.  Also, this finalizer business
makes lots of the reference counting we were doing unnecessary.

Also, more testing is needed.
2013-04-26 12:18:07 -04:00

712 lines
18 KiB
C

/*
* Copyright (c) 2007-2012 Niels Provos and Nick Mathewson
* Copyright (c) 2002-2006 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. 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 "event2/event-config.h"
#include "evconfig-private.h"
#include <sys/types.h>
#ifdef EVENT__HAVE_SYS_TIME_H
#include <sys/time.h>
#endif
#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#ifdef EVENT__HAVE_STDARG_H
#include <stdarg.h>
#endif
#ifdef EVENT__HAVE_UNISTD_H
#include <unistd.h>
#endif
#ifdef _WIN32
#include <winsock2.h>
#include <ws2tcpip.h>
#endif
#ifdef EVENT__HAVE_SYS_SOCKET_H
#include <sys/socket.h>
#endif
#ifdef EVENT__HAVE_NETINET_IN_H
#include <netinet/in.h>
#endif
#ifdef EVENT__HAVE_NETINET_IN6_H
#include <netinet/in6.h>
#endif
#include "event2/util.h"
#include "event2/bufferevent.h"
#include "event2/buffer.h"
#include "event2/bufferevent_struct.h"
#include "event2/bufferevent_compat.h"
#include "event2/event.h"
#include "log-internal.h"
#include "mm-internal.h"
#include "bufferevent-internal.h"
#include "util-internal.h"
#ifdef _WIN32
#include "iocp-internal.h"
#endif
/* prototypes */
static int be_socket_enable(struct bufferevent *, short);
static int be_socket_disable(struct bufferevent *, short);
static void be_socket_destruct(struct bufferevent *);
static int be_socket_adj_timeouts(struct bufferevent *);
static int be_socket_flush(struct bufferevent *, short, enum bufferevent_flush_mode);
static int be_socket_ctrl(struct bufferevent *, enum bufferevent_ctrl_op, union bufferevent_ctrl_data *);
static void be_socket_setfd(struct bufferevent *, evutil_socket_t);
const struct bufferevent_ops bufferevent_ops_socket = {
"socket",
evutil_offsetof(struct bufferevent_private, bev),
be_socket_enable,
be_socket_disable,
NULL, /* unlink */
be_socket_destruct,
be_socket_adj_timeouts,
be_socket_flush,
be_socket_ctrl,
};
#define be_socket_add(ev, t) \
bufferevent_add_event_((ev), (t))
static void
bufferevent_socket_outbuf_cb(struct evbuffer *buf,
const struct evbuffer_cb_info *cbinfo,
void *arg)
{
struct bufferevent *bufev = arg;
struct bufferevent_private *bufev_p =
EVUTIL_UPCAST(bufev, struct bufferevent_private, bev);
if (cbinfo->n_added &&
(bufev->enabled & EV_WRITE) &&
!event_pending(&bufev->ev_write, EV_WRITE, NULL) &&
!bufev_p->write_suspended) {
/* Somebody added data to the buffer, and we would like to
* write, and we were not writing. So, start writing. */
if (be_socket_add(&bufev->ev_write, &bufev->timeout_write) == -1) {
/* Should we log this? */
}
}
}
static void
bufferevent_readcb(evutil_socket_t fd, short event, void *arg)
{
struct bufferevent *bufev = arg;
struct bufferevent_private *bufev_p =
EVUTIL_UPCAST(bufev, struct bufferevent_private, bev);
struct evbuffer *input;
int res = 0;
short what = BEV_EVENT_READING;
ev_ssize_t howmuch = -1, readmax=-1;
bufferevent_incref_and_lock_(bufev);
if (event == EV_TIMEOUT) {
/* Note that we only check for event==EV_TIMEOUT. If
* event==EV_TIMEOUT|EV_READ, we can safely ignore the
* timeout, since a read has occurred */
what |= BEV_EVENT_TIMEOUT;
goto error;
}
input = bufev->input;
/*
* If we have a high watermark configured then we don't want to
* read more data than would make us reach the watermark.
*/
if (bufev->wm_read.high != 0) {
howmuch = bufev->wm_read.high - evbuffer_get_length(input);
/* we somehow lowered the watermark, stop reading */
if (howmuch <= 0) {
bufferevent_wm_suspend_read(bufev);
goto done;
}
}
readmax = bufferevent_get_read_max_(bufev_p);
if (howmuch < 0 || howmuch > readmax) /* The use of -1 for "unlimited"
* uglifies this code. XXXX */
howmuch = readmax;
if (bufev_p->read_suspended)
goto done;
evbuffer_unfreeze(input, 0);
res = evbuffer_read(input, fd, (int)howmuch); /* XXXX evbuffer_read would do better to take and return ev_ssize_t */
evbuffer_freeze(input, 0);
if (res == -1) {
int err = evutil_socket_geterror(fd);
if (EVUTIL_ERR_RW_RETRIABLE(err))
goto reschedule;
/* error case */
what |= BEV_EVENT_ERROR;
} else if (res == 0) {
/* eof case */
what |= BEV_EVENT_EOF;
}
if (res <= 0)
goto error;
bufferevent_decrement_read_buckets_(bufev_p, res);
/* Invoke the user callback - must always be called last */
if (evbuffer_get_length(input) >= bufev->wm_read.low)
bufferevent_run_readcb_(bufev);
goto done;
reschedule:
goto done;
error:
bufferevent_disable(bufev, EV_READ);
bufferevent_run_eventcb_(bufev, what);
done:
bufferevent_decref_and_unlock_(bufev);
}
static void
bufferevent_writecb(evutil_socket_t fd, short event, void *arg)
{
struct bufferevent *bufev = arg;
struct bufferevent_private *bufev_p =
EVUTIL_UPCAST(bufev, struct bufferevent_private, bev);
int res = 0;
short what = BEV_EVENT_WRITING;
int connected = 0;
ev_ssize_t atmost = -1;
bufferevent_incref_and_lock_(bufev);
if (event == EV_TIMEOUT) {
/* Note that we only check for event==EV_TIMEOUT. If
* event==EV_TIMEOUT|EV_WRITE, we can safely ignore the
* timeout, since a read has occurred */
what |= BEV_EVENT_TIMEOUT;
goto error;
}
if (bufev_p->connecting) {
int c = evutil_socket_finished_connecting_(fd);
/* we need to fake the error if the connection was refused
* immediately - usually connection to localhost on BSD */
if (bufev_p->connection_refused) {
bufev_p->connection_refused = 0;
c = -1;
}
if (c == 0)
goto done;
bufev_p->connecting = 0;
if (c < 0) {
event_del(&bufev->ev_write);
event_del(&bufev->ev_read);
bufferevent_run_eventcb_(bufev, BEV_EVENT_ERROR);
goto done;
} else {
connected = 1;
#ifdef _WIN32
if (BEV_IS_ASYNC(bufev)) {
event_del(&bufev->ev_write);
bufferevent_async_set_connected_(bufev);
bufferevent_run_eventcb_(bufev,
BEV_EVENT_CONNECTED);
goto done;
}
#endif
bufferevent_run_eventcb_(bufev,
BEV_EVENT_CONNECTED);
if (!(bufev->enabled & EV_WRITE) ||
bufev_p->write_suspended) {
event_del(&bufev->ev_write);
goto done;
}
}
}
atmost = bufferevent_get_write_max_(bufev_p);
if (bufev_p->write_suspended)
goto done;
if (evbuffer_get_length(bufev->output)) {
evbuffer_unfreeze(bufev->output, 1);
res = evbuffer_write_atmost(bufev->output, fd, atmost);
evbuffer_freeze(bufev->output, 1);
if (res == -1) {
int err = evutil_socket_geterror(fd);
if (EVUTIL_ERR_RW_RETRIABLE(err))
goto reschedule;
what |= BEV_EVENT_ERROR;
} else if (res == 0) {
/* eof case
XXXX Actually, a 0 on write doesn't indicate
an EOF. An ECONNRESET might be more typical.
*/
what |= BEV_EVENT_EOF;
}
if (res <= 0)
goto error;
bufferevent_decrement_write_buckets_(bufev_p, res);
}
if (evbuffer_get_length(bufev->output) == 0) {
event_del(&bufev->ev_write);
}
/*
* Invoke the user callback if our buffer is drained or below the
* low watermark.
*/
if ((res || !connected) &&
evbuffer_get_length(bufev->output) <= bufev->wm_write.low) {
bufferevent_run_writecb_(bufev);
}
goto done;
reschedule:
if (evbuffer_get_length(bufev->output) == 0) {
event_del(&bufev->ev_write);
}
goto done;
error:
bufferevent_disable(bufev, EV_WRITE);
bufferevent_run_eventcb_(bufev, what);
done:
bufferevent_decref_and_unlock_(bufev);
}
struct bufferevent *
bufferevent_socket_new(struct event_base *base, evutil_socket_t fd,
int options)
{
struct bufferevent_private *bufev_p;
struct bufferevent *bufev;
#ifdef _WIN32
if (base && event_base_get_iocp_(base))
return bufferevent_async_new_(base, fd, options);
#endif
if ((bufev_p = mm_calloc(1, sizeof(struct bufferevent_private)))== NULL)
return NULL;
if (bufferevent_init_common_(bufev_p, base, &bufferevent_ops_socket,
options) < 0) {
mm_free(bufev_p);
return NULL;
}
bufev = &bufev_p->bev;
evbuffer_set_flags(bufev->output, EVBUFFER_FLAG_DRAINS_TO_FD);
event_assign(&bufev->ev_read, bufev->ev_base, fd,
EV_READ|EV_PERSIST|EV_FINALIZE, bufferevent_readcb, bufev);
event_assign(&bufev->ev_write, bufev->ev_base, fd,
EV_WRITE|EV_PERSIST|EV_FINALIZE, bufferevent_writecb, bufev);
evbuffer_add_cb(bufev->output, bufferevent_socket_outbuf_cb, bufev);
evbuffer_freeze(bufev->input, 0);
evbuffer_freeze(bufev->output, 1);
return bufev;
}
int
bufferevent_socket_connect(struct bufferevent *bev,
struct sockaddr *sa, int socklen)
{
struct bufferevent_private *bufev_p =
EVUTIL_UPCAST(bev, struct bufferevent_private, bev);
evutil_socket_t fd;
int r = 0;
int result=-1;
int ownfd = 0;
bufferevent_incref_and_lock_(bev);
if (!bufev_p)
goto done;
fd = bufferevent_getfd(bev);
if (fd < 0) {
if (!sa)
goto done;
fd = evutil_socket_(sa->sa_family,
SOCK_STREAM|EVUTIL_SOCK_NONBLOCK, 0);
if (fd < 0)
goto done;
ownfd = 1;
}
if (sa) {
#ifdef _WIN32
if (bufferevent_async_can_connect_(bev)) {
bufferevent_setfd(bev, fd);
r = bufferevent_async_connect_(bev, fd, sa, socklen);
if (r < 0)
goto freesock;
bufev_p->connecting = 1;
result = 0;
goto done;
} else
#endif
r = evutil_socket_connect_(&fd, sa, socklen);
if (r < 0)
goto freesock;
}
#ifdef _WIN32
/* ConnectEx() isn't always around, even when IOCP is enabled.
* Here, we borrow the socket object's write handler to fall back
* on a non-blocking connect() when ConnectEx() is unavailable. */
if (BEV_IS_ASYNC(bev)) {
event_assign(&bev->ev_write, bev->ev_base, fd,
EV_WRITE|EV_PERSIST|EV_FINALIZE, bufferevent_writecb, bev);
}
#endif
bufferevent_setfd(bev, fd);
if (r == 0) {
if (! be_socket_enable(bev, EV_WRITE)) {
bufev_p->connecting = 1;
result = 0;
goto done;
}
} else if (r == 1) {
/* The connect succeeded already. How very BSD of it. */
result = 0;
bufev_p->connecting = 1;
event_active(&bev->ev_write, EV_WRITE, 1);
} else {
/* The connect failed already. How very BSD of it. */
bufev_p->connection_refused = 1;
bufev_p->connecting = 1;
result = 0;
event_active(&bev->ev_write, EV_WRITE, 1);
}
goto done;
freesock:
bufferevent_run_eventcb_(bev, BEV_EVENT_ERROR);
if (ownfd)
evutil_closesocket(fd);
/* do something about the error? */
done:
bufferevent_decref_and_unlock_(bev);
return result;
}
static void
bufferevent_connect_getaddrinfo_cb(int result, struct evutil_addrinfo *ai,
void *arg)
{
struct bufferevent *bev = arg;
struct bufferevent_private *bev_p =
EVUTIL_UPCAST(bev, struct bufferevent_private, bev);
int r;
BEV_LOCK(bev);
bufferevent_unsuspend_write_(bev, BEV_SUSPEND_LOOKUP);
bufferevent_unsuspend_read_(bev, BEV_SUSPEND_LOOKUP);
if (result != 0) {
bev_p->dns_error = result;
bufferevent_run_eventcb_(bev, BEV_EVENT_ERROR);
bufferevent_decref_and_unlock_(bev);
if (ai)
evutil_freeaddrinfo(ai);
return;
}
/* XXX use the other addrinfos? */
/* XXX use this return value */
r = bufferevent_socket_connect(bev, ai->ai_addr, (int)ai->ai_addrlen);
(void)r;
bufferevent_decref_and_unlock_(bev);
evutil_freeaddrinfo(ai);
}
int
bufferevent_socket_connect_hostname(struct bufferevent *bev,
struct evdns_base *evdns_base, int family, const char *hostname, int port)
{
char portbuf[10];
struct evutil_addrinfo hint;
int err;
struct bufferevent_private *bev_p =
EVUTIL_UPCAST(bev, struct bufferevent_private, bev);
if (family != AF_INET && family != AF_INET6 && family != AF_UNSPEC)
return -1;
if (port < 1 || port > 65535)
return -1;
BEV_LOCK(bev);
bev_p->dns_error = 0;
BEV_UNLOCK(bev);
evutil_snprintf(portbuf, sizeof(portbuf), "%d", port);
memset(&hint, 0, sizeof(hint));
hint.ai_family = family;
hint.ai_protocol = IPPROTO_TCP;
hint.ai_socktype = SOCK_STREAM;
bufferevent_suspend_write_(bev, BEV_SUSPEND_LOOKUP);
bufferevent_suspend_read_(bev, BEV_SUSPEND_LOOKUP);
bufferevent_incref_(bev);
err = evutil_getaddrinfo_async_(evdns_base, hostname, portbuf,
&hint, bufferevent_connect_getaddrinfo_cb, bev);
if (err == 0) {
return 0;
} else {
bufferevent_unsuspend_write_(bev, BEV_SUSPEND_LOOKUP);
bufferevent_unsuspend_read_(bev, BEV_SUSPEND_LOOKUP);
bufferevent_decref_(bev);
return -1;
}
}
int
bufferevent_socket_get_dns_error(struct bufferevent *bev)
{
int rv;
struct bufferevent_private *bev_p =
EVUTIL_UPCAST(bev, struct bufferevent_private, bev);
BEV_LOCK(bev);
rv = bev_p->dns_error;
BEV_UNLOCK(bev);
return rv;
}
/*
* Create a new buffered event object.
*
* The read callback is invoked whenever we read new data.
* The write callback is invoked whenever the output buffer is drained.
* The error callback is invoked on a write/read error or on EOF.
*
* Both read and write callbacks maybe NULL. The error callback is not
* allowed to be NULL and have to be provided always.
*/
struct bufferevent *
bufferevent_new(evutil_socket_t fd,
bufferevent_data_cb readcb, bufferevent_data_cb writecb,
bufferevent_event_cb eventcb, void *cbarg)
{
struct bufferevent *bufev;
if (!(bufev = bufferevent_socket_new(NULL, fd, 0)))
return NULL;
bufferevent_setcb(bufev, readcb, writecb, eventcb, cbarg);
return bufev;
}
static int
be_socket_enable(struct bufferevent *bufev, short event)
{
if (event & EV_READ) {
if (be_socket_add(&bufev->ev_read,&bufev->timeout_read) == -1)
return -1;
}
if (event & EV_WRITE) {
if (be_socket_add(&bufev->ev_write,&bufev->timeout_write) == -1)
return -1;
}
return 0;
}
static int
be_socket_disable(struct bufferevent *bufev, short event)
{
struct bufferevent_private *bufev_p =
EVUTIL_UPCAST(bufev, struct bufferevent_private, bev);
if (event & EV_READ) {
if (event_del(&bufev->ev_read) == -1)
return -1;
}
/* Don't actually disable the write if we are trying to connect. */
if ((event & EV_WRITE) && ! bufev_p->connecting) {
if (event_del(&bufev->ev_write) == -1)
return -1;
}
return 0;
}
static void
be_socket_destruct(struct bufferevent *bufev)
{
struct bufferevent_private *bufev_p =
EVUTIL_UPCAST(bufev, struct bufferevent_private, bev);
evutil_socket_t fd;
EVUTIL_ASSERT(bufev->be_ops == &bufferevent_ops_socket);
fd = event_get_fd(&bufev->ev_read);
if ((bufev_p->options & BEV_OPT_CLOSE_ON_FREE) && fd >= 0)
EVUTIL_CLOSESOCKET(fd);
}
static int
be_socket_adj_timeouts(struct bufferevent *bufev)
{
int r = 0;
if (event_pending(&bufev->ev_read, EV_READ, NULL)) {
if (evutil_timerisset(&bufev->timeout_read)) {
if (be_socket_add(&bufev->ev_read, &bufev->timeout_read) < 0)
r = -1;
} else {
event_remove_timer(&bufev->ev_read);
}
}
if (event_pending(&bufev->ev_write, EV_WRITE, NULL)) {
if (evutil_timerisset(&bufev->timeout_write)) {
if (be_socket_add(&bufev->ev_write, &bufev->timeout_write) < 0)
r = -1;
} else {
event_remove_timer(&bufev->ev_write);
}
}
return r;
}
static int
be_socket_flush(struct bufferevent *bev, short iotype,
enum bufferevent_flush_mode mode)
{
return 0;
}
static void
be_socket_setfd(struct bufferevent *bufev, evutil_socket_t fd)
{
BEV_LOCK(bufev);
EVUTIL_ASSERT(bufev->be_ops == &bufferevent_ops_socket);
event_del(&bufev->ev_read);
event_del(&bufev->ev_write);
event_assign(&bufev->ev_read, bufev->ev_base, fd,
EV_READ|EV_PERSIST|EV_FINALIZE, bufferevent_readcb, bufev);
event_assign(&bufev->ev_write, bufev->ev_base, fd,
EV_WRITE|EV_PERSIST|EV_FINALIZE, bufferevent_writecb, bufev);
if (fd >= 0)
bufferevent_enable(bufev, bufev->enabled);
BEV_UNLOCK(bufev);
}
/* XXXX Should non-socket bufferevents support this? */
int
bufferevent_priority_set(struct bufferevent *bufev, int priority)
{
int r = -1;
struct bufferevent_private *bufev_p =
EVUTIL_UPCAST(bufev, struct bufferevent_private, bev);
BEV_LOCK(bufev);
if (bufev->be_ops != &bufferevent_ops_socket)
goto done;
if (event_priority_set(&bufev->ev_read, priority) == -1)
goto done;
if (event_priority_set(&bufev->ev_write, priority) == -1)
goto done;
event_deferred_cb_set_priority_(&bufev_p->deferred, priority);
r = 0;
done:
BEV_UNLOCK(bufev);
return r;
}
/* XXXX Should non-socket bufferevents support this? */
int
bufferevent_base_set(struct event_base *base, struct bufferevent *bufev)
{
int res = -1;
BEV_LOCK(bufev);
if (bufev->be_ops != &bufferevent_ops_socket)
goto done;
bufev->ev_base = base;
res = event_base_set(base, &bufev->ev_read);
if (res == -1)
goto done;
res = event_base_set(base, &bufev->ev_write);
done:
BEV_UNLOCK(bufev);
return res;
}
static int
be_socket_ctrl(struct bufferevent *bev, enum bufferevent_ctrl_op op,
union bufferevent_ctrl_data *data)
{
switch (op) {
case BEV_CTRL_SET_FD:
be_socket_setfd(bev, data->fd);
return 0;
case BEV_CTRL_GET_FD:
data->fd = event_get_fd(&bev->ev_read);
return 0;
case BEV_CTRL_GET_UNDERLYING:
case BEV_CTRL_CANCEL_ALL:
default:
return -1;
}
}