0
0
Fork 0
mirror of https://github.com/matrix-construct/construct synced 2024-11-26 00:32:35 +01:00
construct/ircd/ircd.cc

487 lines
12 KiB
C++

/*
* charybdis: A slightly useful ircd.
* ircd.c: Starts up and runs the ircd.
*
* Copyright (C) 1990 Jarkko Oikarinen and University of Oulu, Co Center
* Copyright (C) 1996-2002 Hybrid Development Team
* Copyright (C) 2002-2008 ircd-ratbox development team
* Copyright (C) 2005-2013 charybdis development team
*
* 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.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
* USA
*/
#include <boost/asio.hpp>
#include <ircd/ctx_ctx.h>
namespace ircd
{
bool debugmode; // set by command line
boost::asio::io_service *ios; // user's io service
std::shared_ptr<client> me; // That's me
main_exit_cb main_exit_func; // Called when main context exits
bool main_exited; // Set when IRCd is finished
void seed_random();
void init_system();
void main_exiting() noexcept;
void handle_sigusr2();
void handle_sigusr1();
void handle_sigterm();
void handle_sigquit();
void main() noexcept;
}
//
// Sets up the IRCd, handlers (main context), and then returns without blocking.
// Pass your io_service instance, it will share it with the rest of your program.
// An exception will be thrown on error.
//
void
ircd::init(boost::asio::io_service &io_service,
const std::string &configfile,
main_exit_cb main_exit_func)
{
ircd::ios = &io_service;
init_system();
// The log is available, but it is console-only until conf opens files.
log::init();
log::mark("log started");
// The configuration is parsed for grammatical errors only. It is not evaluated.
log::info("parsing your configuration");
conf::parse(configfile);
// The caller has no other way to know IRCd has halted as everyone shares the
// same io_service, which may not return if the caller has their own tasks.
at_main_exit(std::move(main_exit_func));
// The master of ceremonies runs the show after this function returns and ios.run().
// The SELF_DESTRUCT flag indicates it will clean itself up and cannot join here.
log::debug("spawning main context");
context(8_MiB, ircd::main, ctx::DEFER_POST | ctx::SELF_DESTRUCT);
log::debug("IRCd initialization completed.");
}
//
// Main context; Main program loop.
// This function is spawned by init(). Do not call this function.
//
void
ircd::main()
noexcept try
{
log::debug("IRCd entered main context.");
const scope main_exit(&main_exiting); // The user is notified when this function ends
// These objects are the init()'s and fini()'s for each subsystem. Appearing here ties them
// to the main context. Initialization can also occur in ircd::init() if static initialization
// and destruction is not possible, but there is no complementary destruction up there.
mods::init _mods_;
// Create IRCd's agency
ircd::me = add_client();
// This is where the configuration is finally evalulated. This must occur
// in a context which can block. If execute() does not transact successfully
// IRCd never reaches the main loop and aborts.
log::info("executing configuration");
conf::execute();
log::debug("setting up signals");
boost::asio::signal_set sigfd(*ios);
sigfd.add(SIGTERM);
sigfd.add(SIGQUIT);
sigfd.add(SIGUSR1);
sigfd.add(SIGUSR2);
// This is the main program loop. All it does is handle signals.
// The configuration will have spawned other loops.
log::notice("IRCd ready");
do switch(sigfd.async_wait(yield(continuation())))
{
case SIGTERM: handle_sigterm(); return;
case SIGQUIT: handle_sigquit(); return;
case SIGUSR1: handle_sigusr1(); continue;
case SIGUSR2: handle_sigusr2(); continue;
}
while(1);
}
catch(const std::exception &e)
{
log::error("IRCd finished: %s", e.what());
}
void
ircd::handle_sigterm()
{
using namespace ircd;
log::notice("IRCd finished: SIGTERM");
}
void
ircd::handle_sigquit()
{
using namespace ircd;
log::notice("SIGQUIT. Quitting...");
}
void
ircd::handle_sigusr1()
{
using namespace ircd;
log::notice("SIGUSR1. Rehashing configuration...");
// dorehash (no longer hangup!)
}
void
ircd::handle_sigusr2()
{
using namespace ircd;
log::notice("SIGUSR2. Flushing caches...");
// dorehashbans
// doremotd
}
//
// Cleanup function for the main context.
//
void
ircd::main_exiting()
noexcept try
{
if(main_exit_func)
{
// This function will notify the user of IRCd shutdown. The notification is
// posted to the io_service ensuring THERE IS NO CONTINUATION ON THIS STACK by
// the user, who is then free to destruct all elements of IRCd.
log::debug("Notifying user of IRCd completion");
ios->post(main_exit_func);
}
main_exited = true;
}
catch(const std::exception &e)
{
log::critical("main context exit: %s", e.what());
std::terminate();
}
void
ircd::at_main_exit(main_exit_cb main_exit_func)
{
ircd::main_exit_func = std::move(main_exit_func);
}
void
ircd::init_system()
{
/*
* Setup corefile size immediately after boot -kre
*/
#ifdef HAVE_SYS_RESOURCE_H
struct rlimit rlim; /* resource limits */
/* Set corefilesize to maximum */
if(!getrlimit(RLIMIT_CORE, &rlim))
{
rlim.rlim_cur = rlim.rlim_max;
setrlimit(RLIMIT_CORE, &rlim);
}
#endif
rb_set_time();
rb_init_prng(NULL, RB_PRNG_DEFAULT);
seed_random();
}
static bool
seed_with_urandom()
{
unsigned int seed;
int fd;
fd = open("/dev/urandom", O_RDONLY);
if(fd >= 0)
{
if(read(fd, &seed, sizeof(seed)) == sizeof(seed))
{
close(fd);
srand(seed);
return true;
}
close(fd);
}
return false;
}
static void
seed_with_clock()
{
const struct timeval *tv;
rb_set_time();
tv = rb_current_time_tv();
srand(tv->tv_sec ^ (tv->tv_usec | (getpid() << 20)));
}
void
ircd::seed_random()
{
unsigned int seed;
if(rb_get_random(&seed, sizeof(seed)) == -1)
{
if(!seed_with_urandom())
seed_with_clock();
return;
}
srand(seed);
}
// namespace ircd {
/* /quote set variables */
/*
struct SetOptions GlobalSetOptions;
struct Counter Count;
struct ServerStatistics ServerStats;
struct ev_entry *check_splitmode_ev;
int maxconnections;
//struct LocalUser meLocalUser; // That's also part of me
rb_dlink_list global_client_list;
*/
/* unknown/client pointer lists */
rb_dlink_list unknown_list; /* unknown clients ON this server only */
rb_dlink_list lclient_list; /* local clients only ON this server */
rb_dlink_list serv_list; /* local servers to this server ONLY */
rb_dlink_list global_serv_list; /* global servers on the network */
rb_dlink_list local_oper_list; /* our opers, duplicated in lclient_list */
rb_dlink_list oper_list; /* network opers */
/*
char * const *myargv;
volatile sig_atomic_t dorehash = false;
volatile sig_atomic_t dorehashbans = false;
volatile sig_atomic_t doremotd = false;
bool kline_queued = false;
bool server_state_foreground = true;
bool opers_see_all_users = false;
bool ircd_ssl_ok = false;
bool ircd_zlib_ok = true;
int testing_conf = 0;
//time_t startup_time;
int default_server_capabs;
int splitmode;
int splitchecking;
int split_users;
int split_servers;
int eob_count;
*/
// }
/*
static void
check_rehash(void *unused)
{
if(dorehash)
{
rehash(true);
dorehash = false;
}
if(dorehashbans)
{
rehash_bans();
dorehashbans = false;
}
if(doremotd)
{
sendto_realops_snomask(sno::GENERAL, L_ALL,
"Got signal SIGUSR1, reloading ircd motd file");
cache::motd::cache_user();
doremotd = false;
}
}
*/
/*
* initalialize_global_set_options
*
* inputs - none
* output - none
* side effects - This sets all global set options needed
*/
static void
initialize_global_set_options(void)
{
/*
memset(&GlobalSetOptions, 0, sizeof(GlobalSetOptions));
// memset( &ConfigFileEntry, 0, sizeof(ConfigFileEntry));
GlobalSetOptions.maxclients = ServerInfo.default_max_clients;
if(GlobalSetOptions.maxclients > (maxconnections - MAX_BUFFER) || (GlobalSetOptions.maxclients <= 0))
GlobalSetOptions.maxclients = maxconnections - MAX_BUFFER;
GlobalSetOptions.autoconn = 1;
GlobalSetOptions.spam_time = MIN_JOIN_LEAVE_TIME;
GlobalSetOptions.spam_num = MAX_JOIN_LEAVE_COUNT;
GlobalSetOptions.floodcount = ConfigFileEntry.default_floodcount;
split_servers = ConfigChannel.default_split_server_count;
split_users = ConfigChannel.default_split_user_count;
if(split_users && split_servers
&& (ConfigChannel.no_create_on_split || ConfigChannel.no_join_on_split))
{
splitmode = 1;
splitchecking = 1;
}
GlobalSetOptions.ident_timeout = ConfigFileEntry.default_ident_timeout;
rb_strlcpy(GlobalSetOptions.operstring,
ConfigFileEntry.default_operstring,
sizeof(GlobalSetOptions.operstring));
rb_strlcpy(GlobalSetOptions.adminstring,
ConfigFileEntry.default_adminstring,
sizeof(GlobalSetOptions.adminstring));
*/
}
// initialise operhash fairly early.
//init_operhash();
//me.localClient = &meLocalUser;
//rb_dlinkAddTail(&me, &me.node, &global_client_list);
//init_builtin_capabs();
//default_server_capabs = CAP_MASK;
//newconf_init();
//init_s_conf();
//init_s_newconf();
//init_hash();
//init_host_hash();
//client::init();
//init_hook();
//chan::init();
//initclass();
//whowas::init();
//init_reject();
//cache::init();
//init_monitor();
//chan::mode::init();
//init_authd(); /* Start up authd. */
//init_dns(); /* Start up DNS query system */
//privilegeset_set_new("default", "", 0);
//read_conf_files(true); /* cold start init conf files */
//mods::autoload();
//supported::init();
//init_bandb();
//init_ssld();
//init_wsockd();
//rehash_bans();
//default_server_capabs &= ~CAP_ZIP; // Set up default_server_capabs
//initialize_global_set_options();
/*
if(ServerInfo.name == NULL)
throw error("no server name specified in serverinfo block.");
rb_strlcpy(me.name, ServerInfo.name, sizeof(me.name));
if(ServerInfo.sid[0] == '\0')
throw error("no server sid specified in serverinfo block.");
rb_strlcpy(me.id, ServerInfo.sid, sizeof(me.id));
//client::init_uid();
// serverinfo{} description must exist. If not, error out.
if(ServerInfo.description == NULL)
throw error("no server description specified in serverinfo block.");
rb_strlcpy(me.info, ServerInfo.description, sizeof(me.info));
*/
/*
if(ServerInfo.ssl_cert != NULL)
{
// just do the rb_setup_ssl_server to validate the config
if(!rb_setup_ssl_server(ServerInfo.ssl_cert, ServerInfo.ssl_private_key, ServerInfo.ssl_dh_params, ServerInfo.ssl_cipher_list))
{
ilog(L_MAIN, "WARNING: Unable to setup SSL.");
ircd_ssl_ok = false;
}
else
ircd_ssl_ok = true;
}
*/
//me.from = &me;
//me.servptr = &me;
//set_me(me);
//make_serv(me);
//add_to_client_hash(me.name, &me);
//add_to_id_hash(me.id, &me);
//nameinfo(serv(me)) = cache::serv::connect(me.name, me.info);
//rb_dlinkAddAlloc(&me, &global_serv_list);
//check_class();
//cache::help::load();
//configure_authd();
/* We want try_connections to be called as soon as possible now! -- adrian */
/* No, 'cause after a restart it would cause all sorts of nick collides */
/* um. by waiting even longer, that just means we have even *more*
* nick collisions. what a stupid idea. set an event for the IO loop --fl
*/
// rb_event_addish("try_connections", try_connections, NULL, STARTUP_CONNECTIONS_TIME);
// rb_event_addonce("try_connections_startup", try_connections, NULL, 2);
// rb_event_add("check_rehash", check_rehash, NULL, 3);
// rb_event_addish("reseed_srand", seed_random, NULL, 300); /* reseed every 10 minutes */
// if(splitmode)
// check_splitmode_ev = rb_event_add("check_splitmode", chan::check_splitmode, NULL, 5);