haproxy/src/proto_udp.c
Willy Tarreau 91ed52976c MINOR: dgram: allow to set rcv/sndbuf for dgram sockets as well
tune.rcvbuf.client and tune.rcvbuf.server are not suitable for shared
dgram sockets because they're per connection so their units are not the
same. However, QUIC's listener and log servers are not connected and
take per-thread or per-process traffic where a socket log buffer might
be too small, causing undesirable packet losses and retransmits in the
case of QUIC. This essentially manifests in listener mode with new
connections taking a lot of time to set up under heavy traffic due to
the small queues causing delays. Let's add a few new settings allowing
to set these shared socket sizes on the frontend and backend side (which
reminds that these are per-front/back and not per client/server hence
not per connection).
2023-10-18 17:01:19 +02:00

248 lines
7.5 KiB
C

/*
* UDP protocol layer on top of AF_INET/AF_INET6
*
* Copyright 2019 HAProxy Technologies, Frederic Lecaille <flecaille@haproxy.com>
*
* Partial merge by Emeric Brun <ebrun@haproxy.com>
*
* 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 <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <time.h>
#include <sys/param.h>
#include <sys/socket.h>
#include <sys/types.h>
#include <netinet/udp.h>
#include <netinet/in.h>
#include <haproxy/fd.h>
#include <haproxy/listener.h>
#include <haproxy/log.h>
#include <haproxy/namespace.h>
#include <haproxy/port_range.h>
#include <haproxy/protocol.h>
#include <haproxy/proto_udp.h>
#include <haproxy/proxy.h>
#include <haproxy/server.h>
#include <haproxy/sock.h>
#include <haproxy/sock_inet.h>
#include <haproxy/task.h>
#include <haproxy/tools.h>
static int udp_bind_listener(struct listener *listener, char *errmsg, int errlen);
static void udp_enable_listener(struct listener *listener);
static void udp_disable_listener(struct listener *listener);
/* Note: must not be declared <const> as its list will be overwritten */
struct protocol proto_udp4 = {
.name = "udp4",
/* connection layer */
.xprt_type = PROTO_TYPE_DGRAM,
.listen = udp_bind_listener,
.enable = udp_enable_listener,
.disable = udp_disable_listener,
.add = default_add_listener,
.unbind = default_unbind_listener,
.suspend = default_suspend_listener,
.resume = default_resume_listener,
/* binding layer */
.rx_suspend = udp_suspend_receiver,
.rx_resume = udp_resume_receiver,
/* address family */
.fam = &proto_fam_inet4,
/* socket layer */
.proto_type = PROTO_TYPE_DGRAM,
.sock_type = SOCK_DGRAM,
.sock_prot = IPPROTO_UDP,
.rx_enable = sock_enable,
.rx_disable = sock_disable,
.rx_unbind = sock_unbind,
.receivers = LIST_HEAD_INIT(proto_udp4.receivers),
.nb_receivers = 0,
#ifdef SO_REUSEPORT
.flags = PROTO_F_REUSEPORT_SUPPORTED,
#endif
};
INITCALL1(STG_REGISTER, protocol_register, &proto_udp4);
/* Note: must not be declared <const> as its list will be overwritten */
struct protocol proto_udp6 = {
.name = "udp6",
/* connection layer */
.xprt_type = PROTO_TYPE_DGRAM,
.listen = udp_bind_listener,
.enable = udp_enable_listener,
.disable = udp_disable_listener,
.add = default_add_listener,
.unbind = default_unbind_listener,
.suspend = default_suspend_listener,
.resume = default_resume_listener,
/* binding layer */
.rx_suspend = udp_suspend_receiver,
.rx_resume = udp_resume_receiver,
/* address family */
.fam = &proto_fam_inet6,
/* socket layer */
.proto_type = PROTO_TYPE_DGRAM,
.sock_type = SOCK_DGRAM,
.sock_prot = IPPROTO_UDP,
.rx_enable = sock_enable,
.rx_disable = sock_disable,
.rx_unbind = sock_unbind,
.receivers = LIST_HEAD_INIT(proto_udp6.receivers),
.nb_receivers = 0,
#ifdef SO_REUSEPORT
.flags = PROTO_F_REUSEPORT_SUPPORTED,
#endif
};
INITCALL1(STG_REGISTER, protocol_register, &proto_udp6);
/* This function tries to bind a UDPv4/v6 listener. It may return a warning or
* an error message in <errmsg> if the message is at most <errlen> bytes long
* (including '\0'). Note that <errmsg> may be NULL if <errlen> is also zero.
* 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 udp_bind_listener(struct listener *listener, char *errmsg, int errlen)
{
int err = ERR_NONE;
char *msg = NULL;
/* ensure we never return garbage */
if (errlen)
*errmsg = 0;
if (listener->state != LI_ASSIGNED)
return ERR_NONE; /* already bound */
if (!(listener->rx.flags & RX_F_BOUND)) {
msg = "receiving socket not bound";
goto udp_return;
}
/* we may want to adjust the output buffer (tune.sndbuf.backend) */
if (global.tune.frontend_rcvbuf)
setsockopt(listener->rx.fd, SOL_SOCKET, SO_RCVBUF, &global.tune.frontend_rcvbuf, sizeof(global.tune.frontend_rcvbuf));
if (global.tune.frontend_sndbuf)
setsockopt(listener->rx.fd, SOL_SOCKET, SO_SNDBUF, &global.tune.frontend_sndbuf, sizeof(global.tune.frontend_sndbuf));
listener_set_state(listener, LI_LISTEN);
udp_return:
if (msg && errlen) {
char pn[INET6_ADDRSTRLEN];
addr_to_str(&listener->rx.addr, pn, sizeof(pn));
snprintf(errmsg, errlen, "%s for [%s:%d]", msg, pn, get_host_port(&listener->rx.addr));
}
return err;
}
/* Enable receipt of incoming connections for listener <l>. The receiver must
* still be valid.
*/
static void udp_enable_listener(struct listener *l)
{
fd_want_recv_safe(l->rx.fd);
}
/* Disable receipt of incoming connections for listener <l>. The receiver must
* still be valid.
*/
static void udp_disable_listener(struct listener *l)
{
fd_stop_recv(l->rx.fd);
}
/* Suspend a receiver. Returns < 0 in case of failure, 0 if the receiver
* was totally stopped, or > 0 if correctly suspended.
* The principle is a bit ugly but works well, at least on Linux: in order to
* suspend the receiver, we want it to stop receiving traffic, which means that
* the socket must be unhashed from the kernel's socket table. The simple way
* to do this is to connect to any address that is reachable and will not be
* used by regular traffic, and a great one is reconnecting to self. Note that
* inherited FDs are neither suspended nor resumed, we only enable/disable
* polling on them.
*/
int udp_suspend_receiver(struct receiver *rx)
{
struct sockaddr_storage ss;
socklen_t len = sizeof(ss);
if (rx->fd < 0)
return 0;
/* we never do that with a shared FD otherwise we'd break it in the
* parent process and any possible subsequent worker inheriting it.
*/
if (rx->flags & RX_F_INHERITED)
goto done;
if (getsockname(rx->fd, (struct sockaddr *)&ss, &len) < 0)
return -1;
if (connect(rx->fd, (struct sockaddr *)&ss, len) < 0)
return -1;
done:
/* not necessary but may make debugging clearer */
fd_stop_recv(rx->fd);
return 1;
}
/* Resume a receiver. Returns < 0 in case of failure, 0 if the receiver
* was totally stopped, or > 0 if correctly suspended.
* The principle is to reverse the change above, we'll break the connection by
* connecting to AF_UNSPEC. The association breaks and the socket starts to
* receive from everywhere again. Note that inherited FDs are neither suspended
* nor resumed, we only enable/disable polling on them.
*/
int udp_resume_receiver(struct receiver *rx)
{
const struct sockaddr sa = { .sa_family = AF_UNSPEC };
if (rx->fd < 0)
return 0;
if (!(rx->flags & RX_F_INHERITED) && connect(rx->fd, &sa, sizeof(sa)) < 0)
return -1;
fd_want_recv(rx->fd);
return 1;
}
/*
* Local variables:
* c-indent-level: 8
* c-basic-offset: 8
* End:
*/