blob: d68941b2c8378b68bb8c62781e6420b7f5859ebd [file] [log] [blame]
/*
* AF_INET/AF_INET6 SOCK_STREAM protocol layer (tcp)
*
* Copyright 2000-2007 Willy Tarreau <w@1wt.eu>
*
* This program 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 of the License, or (at your option) any later version.
*
*/
#include <ctype.h>
#include <errno.h>
#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <time.h>
#include <sys/param.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <sys/un.h>
#include <common/compat.h>
#include <common/config.h>
#include <common/debug.h>
#include <common/errors.h>
#include <common/memory.h>
#include <common/mini-clist.h>
#include <common/standard.h>
#include <common/time.h>
#include <common/version.h>
#include <types/acl.h>
#include <types/client.h>
#include <types/global.h>
#include <types/polling.h>
#include <types/proxy.h>
#include <types/server.h>
#include <proto/acl.h>
#include <proto/backend.h>
#include <proto/buffers.h>
#include <proto/fd.h>
#include <proto/protocols.h>
#include <proto/proto_tcp.h>
#include <proto/queue.h>
#include <proto/senddata.h>
#include <proto/session.h>
#include <proto/stream_sock.h>
#include <proto/task.h>
static int tcp_bind_listeners(struct protocol *proto);
/* Note: must not be declared <const> as its list will be overwritten */
static struct protocol proto_tcpv4 = {
.name = "tcpv4",
.sock_domain = AF_INET,
.sock_type = SOCK_STREAM,
.sock_prot = IPPROTO_TCP,
.sock_family = AF_INET,
.sock_addrlen = sizeof(struct sockaddr_in),
.l3_addrlen = 32/8,
.read = &stream_sock_read,
.write = &stream_sock_write,
.bind_all = tcp_bind_listeners,
.unbind_all = unbind_all_listeners,
.enable_all = enable_all_listeners,
.listeners = LIST_HEAD_INIT(proto_tcpv4.listeners),
.nb_listeners = 0,
};
/* Note: must not be declared <const> as its list will be overwritten */
static struct protocol proto_tcpv6 = {
.name = "tcpv6",
.sock_domain = AF_INET6,
.sock_type = SOCK_STREAM,
.sock_prot = IPPROTO_TCP,
.sock_family = AF_INET6,
.sock_addrlen = sizeof(struct sockaddr_in6),
.l3_addrlen = 128/8,
.read = &stream_sock_read,
.write = &stream_sock_write,
.bind_all = tcp_bind_listeners,
.unbind_all = unbind_all_listeners,
.enable_all = enable_all_listeners,
.listeners = LIST_HEAD_INIT(proto_tcpv6.listeners),
.nb_listeners = 0,
};
/* This function tries to bind a TCPv4/v6 listener. It may return a warning or
* an error message in <err> if the message is at most <errlen> bytes long
* (including '\0'). The return value is composed from ERR_ABORT, ERR_WARN,
* ERR_ALERT, ERR_RETRYABLE and ERR_FATAL. ERR_NONE indicates that everything
* was alright and that no message was returned. ERR_RETRYABLE means that an
* error occurred but that it may vanish after a retry (eg: port in use), and
* ERR_FATAL indicates a non-fixable error.ERR_WARN and ERR_ALERT do not alter
* the meaning of the error, but just indicate that a message is present which
* should be displayed with the respective level. Last, ERR_ABORT indicates
* that it's pointless to try to start other listeners. No error message is
* returned if errlen is NULL.
*/
int tcp_bind_listener(struct listener *listener, char *errmsg, int errlen)
{
__label__ tcp_return, tcp_close_return;
int fd, err;
const char *msg = NULL;
/* ensure we never return garbage */
if (errmsg && errlen)
*errmsg = 0;
if (listener->state != LI_ASSIGNED)
return ERR_NONE; /* already bound */
err = ERR_NONE;
if ((fd = socket(listener->addr.ss_family, SOCK_STREAM, IPPROTO_TCP)) == -1) {
err |= ERR_RETRYABLE | ERR_ALERT;
msg = "cannot create listening socket";
goto tcp_return;
}
if (fd >= global.maxsock) {
err |= ERR_FATAL | ERR_ABORT | ERR_ALERT;
msg = "not enough free sockets (raise '-n' parameter)";
goto tcp_close_return;
}
if ((fcntl(fd, F_SETFL, O_NONBLOCK) == -1) ||
(setsockopt(fd, IPPROTO_TCP, TCP_NODELAY,
(char *) &one, sizeof(one)) == -1)) {
err |= ERR_FATAL | ERR_ALERT;
msg = "cannot make socket non-blocking";
goto tcp_close_return;
}
if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, (char *) &one, sizeof(one)) == -1) {
/* not fatal but should be reported */
msg = "cannot do so_reuseaddr";
err |= ERR_ALERT;
}
if (listener->options & LI_O_NOLINGER)
setsockopt(fd, SOL_SOCKET, SO_LINGER, (struct linger *) &nolinger, sizeof(struct linger));
#ifdef SO_REUSEPORT
/* OpenBSD supports this. As it's present in old libc versions of Linux,
* it might return an error that we will silently ignore.
*/
setsockopt(fd, SOL_SOCKET, SO_REUSEPORT, (char *) &one, sizeof(one));
#endif
#ifdef CONFIG_HAP_LINUX_TPROXY
if ((listener->options & LI_O_FOREIGN)
&& (setsockopt(fd, SOL_IP, IP_TRANSPARENT, (char *) &one, sizeof(one)) == -1)
&& (setsockopt(fd, SOL_IP, IP_FREEBIND, (char *) &one, sizeof(one)) == -1)) {
msg = "cannot make listening socket transparent";
err |= ERR_ALERT;
}
#endif
if (bind(fd, (struct sockaddr *)&listener->addr, listener->proto->sock_addrlen) == -1) {
err |= ERR_RETRYABLE | ERR_ALERT;
msg = "cannot bind socket";
goto tcp_close_return;
}
if (listen(fd, listener->backlog ? listener->backlog : listener->maxconn) == -1) {
err |= ERR_RETRYABLE | ERR_ALERT;
msg = "cannot listen to socket";
goto tcp_close_return;
}
/* the socket is ready */
listener->fd = fd;
listener->state = LI_LISTEN;
/* the function for the accept() event */
fd_insert(fd);
fdtab[fd].cb[DIR_RD].f = listener->accept;
fdtab[fd].cb[DIR_WR].f = NULL; /* never called */
fdtab[fd].cb[DIR_RD].b = fdtab[fd].cb[DIR_WR].b = NULL;
fdtab[fd].owner = (struct task *)listener; /* reference the listener instead of a task */
fdtab[fd].state = FD_STLISTEN;
fdtab[fd].peeraddr = NULL;
fdtab[fd].peerlen = 0;
fdtab[fd].listener = NULL;
fdtab[fd].ev = 0;
tcp_return:
if (msg && errlen)
strlcpy2(errmsg, msg, errlen);
return err;
tcp_close_return:
close(fd);
goto tcp_return;
}
/* This function creates all TCP sockets bound to the protocol entry <proto>.
* It is intended to be used as the protocol's bind_all() function.
* The sockets will be registered but not added to any fd_set, in order not to
* loose them across the fork(). A call to enable_all_listeners() is needed
* to complete initialization. The return value is composed from ERR_*.
*/
static int tcp_bind_listeners(struct protocol *proto)
{
struct listener *listener;
int err = ERR_NONE;
list_for_each_entry(listener, &proto->listeners, proto_list) {
err |= tcp_bind_listener(listener, NULL, 0);
if ((err & ERR_CODE) == ERR_ABORT)
break;
}
return err;
}
/* Add listener to the list of tcpv4 listeners. The listener's state
* is automatically updated from LI_INIT to LI_ASSIGNED. The number of
* listeners is updated. This is the function to use to add a new listener.
*/
void tcpv4_add_listener(struct listener *listener)
{
if (listener->state != LI_INIT)
return;
listener->state = LI_ASSIGNED;
listener->proto = &proto_tcpv4;
LIST_ADDQ(&proto_tcpv4.listeners, &listener->proto_list);
proto_tcpv4.nb_listeners++;
}
/* Add listener to the list of tcpv4 listeners. The listener's state
* is automatically updated from LI_INIT to LI_ASSIGNED. The number of
* listeners is updated. This is the function to use to add a new listener.
*/
void tcpv6_add_listener(struct listener *listener)
{
if (listener->state != LI_INIT)
return;
listener->state = LI_ASSIGNED;
listener->proto = &proto_tcpv6;
LIST_ADDQ(&proto_tcpv6.listeners, &listener->proto_list);
proto_tcpv6.nb_listeners++;
}
__attribute__((constructor))
static void __tcp_protocol_init(void)
{
protocol_register(&proto_tcpv4);
protocol_register(&proto_tcpv6);
}
/*
* Local variables:
* c-indent-level: 8
* c-basic-offset: 8
* End:
*/