From b7c15c31519dc44c1f691e0466badd556ffe9423 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Sun, 7 Apr 2024 18:18:56 +0200 Subject: Adding upstream version 3.7.10. Signed-off-by: Daniel Baumann --- src/master/event_server.c | 968 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 968 insertions(+) create mode 100644 src/master/event_server.c (limited to 'src/master/event_server.c') diff --git a/src/master/event_server.c b/src/master/event_server.c new file mode 100644 index 0000000..9802bdf --- /dev/null +++ b/src/master/event_server.c @@ -0,0 +1,968 @@ +/*++ +/* NAME +/* event_server 3 +/* SUMMARY +/* skeleton multi-threaded mail subsystem +/* SYNOPSIS +/* #include +/* +/* NORETURN event_server_main(argc, argv, service, key, value, ...) +/* int argc; +/* char **argv; +/* void (*service)(VSTREAM *stream, char *service_name, char **argv); +/* int key; +/* +/* void event_server_disconnect(fd) +/* VSTREAM *stream; +/* +/* void event_server_drain() +/* DESCRIPTION +/* This module implements a skeleton for event-driven +/* mail subsystems: mail subsystem programs that service multiple +/* clients at the same time. The resulting program expects to be run +/* from the \fBmaster\fR process. +/* +/* event_server_main() is the skeleton entry point. It should be +/* called from the application main program. The skeleton does all +/* the generic command-line processing, initialization of +/* configurable parameters, and connection management. +/* Unlike multi_server, this skeleton does not attempt to manage +/* all the events on a client connection. +/* The skeleton never returns. +/* +/* Arguments: +/* .IP "void (*service)(VSTREAM *stream, char *service_name, char **argv)" +/* A pointer to a function that is called by the skeleton each +/* time a client connects to the program's service port. The +/* function is run after the program has optionally dropped +/* its privileges. The application is responsible for managing +/* subsequent I/O events on the stream, and is responsible for +/* calling event_server_disconnect() when the stream is closed. +/* The stream initial state is non-blocking mode. +/* Optional connection attributes are provided as a hash that +/* is attached as stream context. NOTE: the attributes are +/* destroyed after this function is called. The service +/* name argument corresponds to the service name in the master.cf +/* file. The argv argument specifies command-line arguments +/* left over after options processing. +/* .PP +/* Optional arguments are specified as a null-terminated list +/* with macros that have zero or more arguments: +/* .IP "CA_MAIL_SERVER_REQ_INT_TABLE(CONFIG_INT_TABLE *)" +/* A table with configurable parameters, to be loaded from the +/* global Postfix configuration file. Tables are loaded in the +/* order as specified, and multiple instances of the same type +/* are allowed. +/* .IP "CA_MAIL_SERVER_REQ_LONG_TABLE(CONFIG_LONG_TABLE *)" +/* A table with configurable parameters, to be loaded from the +/* global Postfix configuration file. Tables are loaded in the +/* order as specified, and multiple instances of the same type +/* are allowed. +/* .IP "CA_MAIL_SERVER_REQ_STR_TABLE(CONFIG_STR_TABLE *)" +/* A table with configurable parameters, to be loaded from the +/* global Postfix configuration file. Tables are loaded in the +/* order as specified, and multiple instances of the same type +/* are allowed. +/* .IP "CA_MAIL_SERVER_REQ_BOOL_TABLE(CONFIG_BOOL_TABLE *)" +/* A table with configurable parameters, to be loaded from the +/* global Postfix configuration file. Tables are loaded in the +/* order as specified, and multiple instances of the same type +/* are allowed. +/* .IP "CA_MAIL_SERVER_REQ_TIME_TABLE(CONFIG_TIME_TABLE *)" +/* A table with configurable parameters, to be loaded from the +/* global Postfix configuration file. Tables are loaded in the +/* order as specified, and multiple instances of the same type +/* are allowed. +/* .IP "CA_MAIL_SERVER_REQ_RAW_TABLE(CONFIG_RAW_TABLE *)" +/* A table with configurable parameters, to be loaded from the +/* global Postfix configuration file. Tables are loaded in the +/* order as specified, and multiple instances of the same type +/* are allowed. Raw parameters are not subjected to $name +/* evaluation. +/* .IP "CA_MAIL_SERVER_REQ_NINT_TABLE(CONFIG_NINT_TABLE *)" +/* A table with configurable parameters, to be loaded from the +/* global Postfix configuration file. Tables are loaded in the +/* order as specified, and multiple instances of the same type +/* are allowed. +/* .IP "CA_MAIL_SERVER_REQ_NBOOL_TABLE(CONFIG_NBOOL_TABLE *)" +/* A table with configurable parameters, to be loaded from the +/* global Postfix configuration file. Tables are loaded in the +/* order as specified, and multiple instances of the same type +/* are allowed. +/* .IP "CA_MAIL_SERVER_REQ_PRE_INIT(void *(char *service_name, char **argv))" +/* A pointer to a function that is called once +/* by the skeleton after it has read the global configuration file +/* and after it has processed command-line arguments, but before +/* the skeleton has optionally relinquished the process privileges. +/* .sp +/* Only the last instance of this parameter type is remembered. +/* .IP "CA_MAIL_SERVER_REQ_POST_INIT(void *(char *service_name, char **argv))" +/* A pointer to a function that is called once +/* by the skeleton after it has optionally relinquished the process +/* privileges, but before servicing client connection requests. +/* .sp +/* Only the last instance of this parameter type is remembered. +/* .IP "CA_MAIL_SERVER_REQ_LOOP(int *(char *service_name, char **argv))" +/* A pointer to function that is executed from +/* within the event loop, whenever an I/O or timer event has happened, +/* or whenever nothing has happened for a specified amount of time. +/* The result value of the function specifies how long to wait until +/* the next event. Specify -1 to wait for "as long as it takes". +/* .sp +/* Only the last instance of this parameter type is remembered. +/* .IP "CA_MAIL_SERVER_EXIT(void *(char *service_name, char **argv))" +/* A pointer to function that is executed immediately before normal +/* process termination. +/* .IP "CA_MAIL_SERVER_PRE_ACCEPT(void *(char *service_name, char **argv))" +/* Function to be executed prior to accepting a new connection. +/* .sp +/* Only the last instance of this parameter type is remembered. +/* .IP "CA_MAIL_SERVER_PRE_DISCONN(VSTREAM *, char *service_name, char **argv)" +/* A pointer to a function that is called +/* by the event_server_disconnect() function (see below). +/* .sp +/* Only the last instance of this parameter type is remembered. +/* .IP CA_MAIL_SERVER_IN_FLOW_DELAY +/* Pause $in_flow_delay seconds when no "mail flow control token" +/* is available. A token is consumed for each connection request. +/* .IP CA_MAIL_SERVER_SOLITARY +/* This service must be configured with process limit of 1. +/* .IP CA_MAIL_SERVER_UNLIMITED +/* This service must be configured with process limit of 0. +/* .IP CA_MAIL_SERVER_PRIVILEGED +/* This service must be configured as privileged. +/* .IP "CA_MAIL_SERVER_SLOW_EXIT(void *(char *service_name, char **argv))" +/* A pointer to a function that is called after "postfix reload" +/* or "master exit". The application can call event_server_drain() +/* (see below) to finish ongoing activities in the background. +/* .IP "CA_MAIL_SERVER_WATCHDOG(int *)" +/* Override the default 1000s watchdog timeout. The value is +/* used after command-line and main.cf file processing. +/* .IP "CA_MAIL_SERVER_BOUNCE_INIT(const char *, const char **)" +/* Initialize the DSN filter for the bounce/defer service +/* clients with the specified map source and map names. +/* .IP "CA_MAIL_SERVER_RETIRE_ME" +/* Prevent a process from being reused indefinitely. After +/* (var_max_use * var_max_idle) seconds or some sane constant, +/* stop accepting new connections and terminate voluntarily +/* when the process becomes idle. +/* .PP +/* event_server_disconnect() should be called by the application +/* to close a client connection. +/* +/* event_server_drain() should be called when the application +/* no longer wishes to accept new client connections. Existing +/* clients are handled in a background process, and the process +/* terminates when the last client is disconnected. A non-zero +/* result means this call should be tried again later. +/* +/* The var_use_limit variable limits the number of clients +/* that a server can service before it commits suicide. This +/* value is taken from the global \fBmain.cf\fR configuration +/* file. Setting \fBvar_use_limit\fR to zero disables the +/* client limit. +/* +/* The var_idle_limit variable limits the time that a service +/* receives no client connection requests before it commits +/* suicide. This value is taken from the global \fBmain.cf\fR +/* configuration file. Setting \fBvar_idle_limit\fR to zero +/* disables the idle limit. +/* DIAGNOSTICS +/* Problems and transactions are logged to \fBsyslogd\fR(8) +/* or \fBpostlogd\fR(8). +/* SEE ALSO +/* master(8), master process +/* postlogd(8), Postfix logging +/* syslogd(8), system logging +/* LICENSE +/* .ad +/* .fi +/* The Secure Mailer license must be distributed with this software. +/* AUTHOR(S) +/* Wietse Venema +/* IBM T.J. Watson Research +/* P.O. Box 704 +/* Yorktown Heights, NY 10598, USA +/* +/* Wietse Venema +/* Google, Inc. +/* 111 8th Avenue +/* New York, NY 10011, USA +/*--*/ + +/* System library. */ + +#include +#include +#include /* select() */ +#include +#include +#include +#include +#include +#include +#include +#include +#ifdef STRCASECMP_IN_STRINGS_H +#include +#endif +#include + +#ifdef USE_SYS_SELECT_H +#include /* select() */ +#endif + +/* Utility library. */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +/* Global library. */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +/* Process manager. */ + +#include "master_proto.h" + +/* Application-specific */ + +#include "mail_server.h" + + /* + * Global state. + */ +static int client_count; +static int use_count; +static int socket_count = 1; + +static void (*event_server_service) (VSTREAM *, char *, char **); +static char *event_server_name; +static char **event_server_argv; +static void (*event_server_accept) (int, void *); +static void (*event_server_onexit) (char *, char **); +static void (*event_server_pre_accept) (char *, char **); +static VSTREAM *event_server_lock; +static int event_server_in_flow_delay; +static unsigned event_server_generation; +static void (*event_server_pre_disconn) (VSTREAM *, char *, char **); +static void (*event_server_slow_exit) (char *, char **); +static int event_server_watchdog = 1000; + +/* event_server_exit - normal termination */ + +static NORETURN event_server_exit(void) +{ + if (event_server_onexit) + event_server_onexit(event_server_name, event_server_argv); + exit(0); +} + +/* event_server_retire - retire when idle */ + +static void event_server_retire(int unused_event, void *unused_context) +{ + if (msg_verbose) + msg_info("time to retire -- %s", event_server_slow_exit ? + "draining" : "exiting"); + event_disable_readwrite(MASTER_STATUS_FD); + if (event_server_slow_exit) + event_server_slow_exit(event_server_name, event_server_argv); + else + event_server_exit(); +} + +/* event_server_abort - terminate after abnormal master exit */ + +static void event_server_abort(int unused_event, void *unused_context) +{ + if (msg_verbose) + msg_info("master disconnect -- %s", event_server_slow_exit ? + "draining" : "exiting"); + event_disable_readwrite(MASTER_STATUS_FD); + if (event_server_slow_exit) + event_server_slow_exit(event_server_name, event_server_argv); + else + event_server_exit(); +} + +/* event_server_timeout - idle time exceeded */ + +static void event_server_timeout(int unused_event, void *unused_context) +{ + if (msg_verbose) + msg_info("idle timeout -- exiting"); + event_server_exit(); +} + +/* event_server_drain - stop accepting new clients */ + +int event_server_drain(void) +{ + const char *myname = "event_server_drain"; + int fd; + + switch (fork()) { + /* Try again later. */ + case -1: + return (-1); + /* Finish existing clients in the background, then terminate. */ + case 0: + (void) msg_cleanup((MSG_CLEANUP_FN) 0); + event_fork(); + for (fd = MASTER_LISTEN_FD; fd < MASTER_LISTEN_FD + socket_count; fd++) { + event_disable_readwrite(fd); + (void) close(fd); + /* Play safe - don't reuse this file number. */ + if (DUP2(STDIN_FILENO, fd) < 0) + msg_warn("%s: dup2(%d, %d): %m", myname, STDIN_FILENO, fd); + } + var_use_limit = 1; + return (0); + /* Let the master start a new process. */ + default: + exit(0); + } +} + +/* event_server_disconnect - terminate client session */ + +void event_server_disconnect(VSTREAM *stream) +{ + if (msg_verbose) + msg_info("connection closed fd %d", vstream_fileno(stream)); + if (event_server_pre_disconn) + event_server_pre_disconn(stream, event_server_name, event_server_argv); + (void) vstream_fclose(stream); + client_count--; + /* Avoid integer wrap-around in a persistent process. */ + if (use_count < INT_MAX) + use_count++; + if (client_count == 0 && var_idle_limit > 0) + event_request_timer(event_server_timeout, (void *) 0, var_idle_limit); +} + +/* event_server_execute - in case (char *) != (struct *) */ + +static void event_server_execute(int unused_event, void *context) +{ + VSTREAM *stream = (VSTREAM *) context; + HTABLE *attr = (HTABLE *) vstream_context(stream); + + if (event_server_lock != 0 + && myflock(vstream_fileno(event_server_lock), INTERNAL_LOCK, + MYFLOCK_OP_NONE) < 0) + msg_fatal("select unlock: %m"); + + /* + * Do bother the application when the client disconnected. Don't drop the + * already accepted client request after "postfix reload"; that would be + * rude. + */ + if (master_notify(var_pid, event_server_generation, MASTER_STAT_TAKEN) < 0) + /* void */ ; + event_server_service(stream, event_server_name, event_server_argv); + if (master_notify(var_pid, event_server_generation, MASTER_STAT_AVAIL) < 0) + event_server_abort(EVENT_NULL_TYPE, EVENT_NULL_CONTEXT); + if (attr) + htable_free(attr, myfree); +} + +/* event_server_wakeup - wake up application */ + +static void event_server_wakeup(int fd, HTABLE *attr) +{ + VSTREAM *stream; + char *tmp; + +#if defined(F_DUPFD) && (EVENTS_STYLE != EVENTS_STYLE_SELECT) +#ifndef THRESHOLD_FD_WORKAROUND +#define THRESHOLD_FD_WORKAROUND 128 +#endif + int new_fd; + + /* + * Leave some handles < FD_SETSIZE for DBMS libraries, in the unlikely + * case of a multi-server with a thousand clients. + */ + if (fd < THRESHOLD_FD_WORKAROUND) { + if ((new_fd = fcntl(fd, F_DUPFD, THRESHOLD_FD_WORKAROUND)) < 0) + msg_fatal("fcntl F_DUPFD: %m"); + (void) close(fd); + fd = new_fd; + } +#endif + if (msg_verbose) + msg_info("connection established fd %d", fd); + non_blocking(fd, BLOCKING); + close_on_exec(fd, CLOSE_ON_EXEC); + client_count++; + stream = vstream_fdopen(fd, O_RDWR); + tmp = concatenate(event_server_name, " socket", (char *) 0); + vstream_control(stream, + CA_VSTREAM_CTL_PATH(tmp), + CA_VSTREAM_CTL_CONTEXT((void *) attr), + CA_VSTREAM_CTL_END); + myfree(tmp); + timed_ipc_setup(stream); + if (event_server_in_flow_delay && mail_flow_get(1) < 0) + event_request_timer(event_server_execute, (void *) stream, + var_in_flow_delay); + else + event_server_execute(0, (void *) stream); +} + +/* event_server_accept_local - accept client connection request */ + +static void event_server_accept_local(int unused_event, void *context) +{ + int listen_fd = CAST_ANY_PTR_TO_INT(context); + int time_left = -1; + int fd; + + /* + * Be prepared for accept() to fail because some other process already + * got the connection (the number of processes competing for clients is + * kept small, so this is not a "thundering herd" problem). If the + * accept() succeeds, be sure to disable non-blocking I/O, in order to + * minimize confusion. + */ + if (client_count == 0 && var_idle_limit > 0) + time_left = event_cancel_timer(event_server_timeout, (void *) 0); + + if (event_server_pre_accept) + event_server_pre_accept(event_server_name, event_server_argv); + fd = LOCAL_ACCEPT(listen_fd); + if (event_server_lock != 0 + && myflock(vstream_fileno(event_server_lock), INTERNAL_LOCK, + MYFLOCK_OP_NONE) < 0) + msg_fatal("select unlock: %m"); + if (fd < 0) { + if (errno != EAGAIN) + msg_error("accept connection: %m"); + if (time_left >= 0) + event_request_timer(event_server_timeout, (void *) 0, time_left); + return; + } + event_server_wakeup(fd, (HTABLE *) 0); +} + +#ifdef MASTER_XPORT_NAME_PASS + +/* event_server_accept_pass - accept descriptor */ + +static void event_server_accept_pass(int unused_event, void *context) +{ + int listen_fd = CAST_ANY_PTR_TO_INT(context); + int time_left = -1; + int fd; + HTABLE *attr = 0; + + /* + * Be prepared for accept() to fail because some other process already + * got the connection (the number of processes competing for clients is + * kept small, so this is not a "thundering herd" problem). If the + * accept() succeeds, be sure to disable non-blocking I/O, in order to + * minimize confusion. + */ + if (client_count == 0 && var_idle_limit > 0) + time_left = event_cancel_timer(event_server_timeout, (void *) 0); + + if (event_server_pre_accept) + event_server_pre_accept(event_server_name, event_server_argv); + fd = pass_accept_attr(listen_fd, &attr); + if (event_server_lock != 0 + && myflock(vstream_fileno(event_server_lock), INTERNAL_LOCK, + MYFLOCK_OP_NONE) < 0) + msg_fatal("select unlock: %m"); + if (fd < 0) { + if (errno != EAGAIN) + msg_error("accept connection: %m"); + if (time_left >= 0) + event_request_timer(event_server_timeout, (void *) 0, time_left); + return; + } + event_server_wakeup(fd, attr); +} + +#endif + +/* event_server_accept_inet - accept client connection request */ + +static void event_server_accept_inet(int unused_event, void *context) +{ + int listen_fd = CAST_ANY_PTR_TO_INT(context); + int time_left = -1; + int fd; + + /* + * Be prepared for accept() to fail because some other process already + * got the connection (the number of processes competing for clients is + * kept small, so this is not a "thundering herd" problem). If the + * accept() succeeds, be sure to disable non-blocking I/O, in order to + * minimize confusion. + */ + if (client_count == 0 && var_idle_limit > 0) + time_left = event_cancel_timer(event_server_timeout, (void *) 0); + + if (event_server_pre_accept) + event_server_pre_accept(event_server_name, event_server_argv); + fd = inet_accept(listen_fd); + if (event_server_lock != 0 + && myflock(vstream_fileno(event_server_lock), INTERNAL_LOCK, + MYFLOCK_OP_NONE) < 0) + msg_fatal("select unlock: %m"); + if (fd < 0) { + if (errno != EAGAIN) + msg_error("accept connection: %m"); + if (time_left >= 0) + event_request_timer(event_server_timeout, (void *) 0, time_left); + return; + } + event_server_wakeup(fd, (HTABLE *) 0); +} + +/* event_server_main - the real main program */ + +NORETURN event_server_main(int argc, char **argv, MULTI_SERVER_FN service,...) +{ + const char *myname = "event_server_main"; + VSTREAM *stream = 0; + char *root_dir = 0; + char *user_name = 0; + int debug_me = 0; + int daemon_mode = 1; + char *service_name = basename(argv[0]); + int delay; + int c; + int fd; + va_list ap; + MAIL_SERVER_INIT_FN pre_init = 0; + MAIL_SERVER_INIT_FN post_init = 0; + MAIL_SERVER_LOOP_FN loop = 0; + int key; + char *transport = 0; + +#if 0 + char *lock_path; + VSTRING *why; + +#endif + int alone = 0; + int zerolimit = 0; + WATCHDOG *watchdog; + char *oname_val; + char *oname; + char *oval; + const char *err; + char *generation; + int msg_vstream_needed = 0; + const char *dsn_filter_title; + const char **dsn_filter_maps; + int retire_me_from_flags = 0; + int retire_me = 0; + + /* + * Process environment options as early as we can. + */ + if (getenv(CONF_ENV_VERB)) + msg_verbose = 1; + if (getenv(CONF_ENV_DEBUG)) + debug_me = 1; + + /* + * Don't die when a process goes away unexpectedly. + */ + signal(SIGPIPE, SIG_IGN); + + /* + * Don't die for frivolous reasons. + */ +#ifdef SIGXFSZ + signal(SIGXFSZ, SIG_IGN); +#endif + + /* + * May need this every now and then. + */ + var_procname = mystrdup(basename(argv[0])); + set_mail_conf_str(VAR_PROCNAME, var_procname); + + /* + * Initialize logging and exit handler. Do the syslog first, so that its + * initialization completes before we enter the optional chroot jail. + */ + maillog_client_init(mail_task(var_procname), MAILLOG_CLIENT_FLAG_NONE); + if (msg_verbose) + msg_info("daemon started"); + + /* + * Check the Postfix library version as soon as we enable logging. + */ + MAIL_VERSION_CHECK; + + /* + * Initialize from the configuration file. Allow command-line options to + * override compiled-in defaults or configured parameter values. + */ + mail_conf_suck(); + + /* + * After database open error, continue execution with reduced + * functionality. + */ + dict_allow_surrogate = 1; + + /* + * Pick up policy settings from master process. Shut up error messages to + * stderr, because no-one is going to see them. + */ + opterr = 0; + while ((c = GETOPT(argc, argv, "cdDi:lm:n:o:r:s:St:uvVz")) > 0) { + switch (c) { + case 'c': + root_dir = "setme"; + break; + case 'd': + daemon_mode = 0; + break; + case 'D': + debug_me = 1; + break; + case 'i': + mail_conf_update(VAR_MAX_IDLE, optarg); + break; + case 'l': + alone = 1; + break; + case 'm': + mail_conf_update(VAR_MAX_USE, optarg); + break; + case 'n': + service_name = optarg; + break; + case 'o': + oname_val = mystrdup(optarg); + if ((err = split_nameval(oname_val, &oname, &oval)) != 0) + msg_fatal("invalid \"-o %s\" option value: %s", optarg, err); + mail_conf_update(oname, oval); + myfree(oname_val); + break; + case 'r': + if ((retire_me_from_flags = atoi(optarg)) <= 0) + msg_fatal("invalid retirement time: %s", optarg); + break; + case 's': + if ((socket_count = atoi(optarg)) <= 0) + msg_fatal("invalid socket_count: %s", optarg); + break; + case 'S': + stream = VSTREAM_IN; + break; + case 'u': + user_name = "setme"; + break; + case 't': + transport = optarg; + break; + case 'v': + msg_verbose++; + break; + case 'V': + if (++msg_vstream_needed == 1) + msg_vstream_init(mail_task(var_procname), VSTREAM_ERR); + break; + case 'z': + zerolimit = 1; + break; + default: + msg_fatal("invalid option: %c", optopt); + break; + } + } + set_mail_conf_str(VAR_SERVNAME, service_name); + + /* + * Initialize generic parameters and re-initialize logging in case of a + * non-default program name or logging destination. + */ + mail_params_init(); + maillog_client_init(mail_task(var_procname), MAILLOG_CLIENT_FLAG_NONE); + + /* + * Register higher-level dictionaries and initialize the support for + * dynamically-loaded dictionaries. + */ + mail_dict_init(); + + /* + * If not connected to stdin, stdin must not be a terminal. + */ + if (daemon_mode && stream == 0 && isatty(STDIN_FILENO)) { + msg_vstream_init(var_procname, VSTREAM_ERR); + msg_fatal("do not run this command by hand"); + } + + /* + * Application-specific initialization. + */ + va_start(ap, service); + while ((key = va_arg(ap, int)) != 0) { + switch (key) { + case MAIL_SERVER_INT_TABLE: + get_mail_conf_int_table(va_arg(ap, CONFIG_INT_TABLE *)); + break; + case MAIL_SERVER_LONG_TABLE: + get_mail_conf_long_table(va_arg(ap, CONFIG_LONG_TABLE *)); + break; + case MAIL_SERVER_STR_TABLE: + get_mail_conf_str_table(va_arg(ap, CONFIG_STR_TABLE *)); + break; + case MAIL_SERVER_BOOL_TABLE: + get_mail_conf_bool_table(va_arg(ap, CONFIG_BOOL_TABLE *)); + break; + case MAIL_SERVER_TIME_TABLE: + get_mail_conf_time_table(va_arg(ap, CONFIG_TIME_TABLE *)); + break; + case MAIL_SERVER_RAW_TABLE: + get_mail_conf_raw_table(va_arg(ap, CONFIG_RAW_TABLE *)); + break; + case MAIL_SERVER_NINT_TABLE: + get_mail_conf_nint_table(va_arg(ap, CONFIG_NINT_TABLE *)); + break; + case MAIL_SERVER_NBOOL_TABLE: + get_mail_conf_nbool_table(va_arg(ap, CONFIG_NBOOL_TABLE *)); + break; + case MAIL_SERVER_PRE_INIT: + pre_init = va_arg(ap, MAIL_SERVER_INIT_FN); + break; + case MAIL_SERVER_POST_INIT: + post_init = va_arg(ap, MAIL_SERVER_INIT_FN); + break; + case MAIL_SERVER_LOOP: + loop = va_arg(ap, MAIL_SERVER_LOOP_FN); + break; + case MAIL_SERVER_EXIT: + event_server_onexit = va_arg(ap, MAIL_SERVER_EXIT_FN); + break; + case MAIL_SERVER_PRE_ACCEPT: + event_server_pre_accept = va_arg(ap, MAIL_SERVER_ACCEPT_FN); + break; + case MAIL_SERVER_PRE_DISCONN: + event_server_pre_disconn = va_arg(ap, MAIL_SERVER_DISCONN_FN); + break; + case MAIL_SERVER_IN_FLOW_DELAY: + event_server_in_flow_delay = 1; + break; + case MAIL_SERVER_SOLITARY: + if (stream == 0 && !alone) + msg_fatal("service %s requires a process limit of 1", + service_name); + break; + case MAIL_SERVER_UNLIMITED: + if (stream == 0 && !zerolimit) + msg_fatal("service %s requires a process limit of 0", + service_name); + break; + case MAIL_SERVER_PRIVILEGED: + if (user_name) + msg_fatal("service %s requires privileged operation", + service_name); + break; + case MAIL_SERVER_WATCHDOG: + event_server_watchdog = *va_arg(ap, int *); + break; + case MAIL_SERVER_SLOW_EXIT: + event_server_slow_exit = va_arg(ap, MAIL_SERVER_SLOW_EXIT_FN); + break; + case MAIL_SERVER_BOUNCE_INIT: + dsn_filter_title = va_arg(ap, const char *); + dsn_filter_maps = va_arg(ap, const char **); + bounce_client_init(dsn_filter_title, *dsn_filter_maps); + break; + case MAIL_SERVER_RETIRE_ME: + if (retire_me_from_flags > 0) + retire_me = retire_me_from_flags; + else if (var_idle_limit == 0 || var_use_limit == 0 + || var_idle_limit > 18000 / var_use_limit) + retire_me = 18000; + else + retire_me = var_idle_limit * var_use_limit; + break; + default: + msg_panic("%s: unknown argument type: %d", myname, key); + } + } + va_end(ap); + + if (root_dir) + root_dir = var_queue_dir; + if (user_name) + user_name = var_mail_owner; + + /* + * Can options be required? + */ + if (stream == 0) { + if (transport == 0) + msg_fatal("no transport type specified"); + if (strcasecmp(transport, MASTER_XPORT_NAME_INET) == 0) + event_server_accept = event_server_accept_inet; + else if (strcasecmp(transport, MASTER_XPORT_NAME_UNIX) == 0) + event_server_accept = event_server_accept_local; +#ifdef MASTER_XPORT_NAME_PASS + else if (strcasecmp(transport, MASTER_XPORT_NAME_PASS) == 0) + event_server_accept = event_server_accept_pass; +#endif + else + msg_fatal("unsupported transport type: %s", transport); + } + + /* + * Retrieve process generation from environment. + */ + if ((generation = getenv(MASTER_GEN_NAME)) != 0) { + if (!alldig(generation)) + msg_fatal("bad generation: %s", generation); + OCTAL_TO_UNSIGNED(event_server_generation, generation); + if (msg_verbose) + msg_info("process generation: %s (%o)", + generation, event_server_generation); + } + + /* + * Optionally start the debugger on ourself. + */ + if (debug_me) + debug_process(); + + /* + * Traditionally, BSD select() can't handle multiple processes selecting + * on the same socket, and wakes up every process in select(). See TCP/IP + * Illustrated volume 2 page 532. We avoid select() collisions with an + * external lock file. + */ + + /* + * XXX Can't compete for exclusive access to the listen socket because we + * also have to monitor existing client connections for service requests. + */ +#if 0 + if (stream == 0 && !alone) { + lock_path = concatenate(DEF_PID_DIR, "/", transport, + ".", service_name, (char *) 0); + why = vstring_alloc(1); + if ((event_server_lock = safe_open(lock_path, O_CREAT | O_RDWR, 0600, + (struct stat *) 0, -1, -1, why)) == 0) + msg_fatal("open lock file %s: %s", lock_path, vstring_str(why)); + close_on_exec(vstream_fileno(event_server_lock), CLOSE_ON_EXEC); + myfree(lock_path); + vstring_free(why); + } +#endif + + /* + * Set up call-back info. + */ + event_server_service = service; + event_server_name = service_name; + event_server_argv = argv + optind; + + /* + * Run pre-jail initialization. + */ + if (chdir(var_queue_dir) < 0) + msg_fatal("chdir(\"%s\"): %m", var_queue_dir); + if (pre_init) + pre_init(event_server_name, event_server_argv); + + /* + * Optionally, restrict the damage that this process can do. + */ + resolve_local_init(); + tzset(); + chroot_uid(root_dir, user_name); + + /* + * Run post-jail initialization. + */ + if (post_init) + post_init(event_server_name, event_server_argv); + + /* + * Are we running as a one-shot server with the client connection on + * standard input? If so, make sure the output is written to stdout so as + * to satisfy common expectation. + */ + if (stream != 0) { + vstream_control(stream, + CA_VSTREAM_CTL_DOUBLE, + CA_VSTREAM_CTL_WRITE_FD(STDOUT_FILENO), + CA_VSTREAM_CTL_END); + service(stream, event_server_name, event_server_argv); + vstream_fflush(stream); + event_server_exit(); + } + + /* + * Running as a semi-resident server. Service connection requests. + * Terminate when we have serviced a sufficient number of clients, when + * no-one has been talking to us for a configurable amount of time, or + * when the master process terminated abnormally. + */ + if (var_idle_limit > 0) + event_request_timer(event_server_timeout, (void *) 0, var_idle_limit); + if (retire_me) + event_request_timer(event_server_retire, (void *) 0, retire_me); + for (fd = MASTER_LISTEN_FD; fd < MASTER_LISTEN_FD + socket_count; fd++) { + event_enable_read(fd, event_server_accept, CAST_INT_TO_VOID_PTR(fd)); + close_on_exec(fd, CLOSE_ON_EXEC); + } + event_enable_read(MASTER_STATUS_FD, event_server_abort, (void *) 0); + close_on_exec(MASTER_STATUS_FD, CLOSE_ON_EXEC); + close_on_exec(MASTER_FLOW_READ, CLOSE_ON_EXEC); + close_on_exec(MASTER_FLOW_WRITE, CLOSE_ON_EXEC); + watchdog = watchdog_create(event_server_watchdog, + (WATCHDOG_FN) 0, (void *) 0); + + /* + * The event loop, at last. + */ + while (var_use_limit == 0 || use_count < var_use_limit || client_count > 0) { + if (event_server_lock != 0) { + watchdog_stop(watchdog); + if (myflock(vstream_fileno(event_server_lock), INTERNAL_LOCK, + MYFLOCK_OP_EXCLUSIVE) < 0) + msg_fatal("select lock: %m"); + } + watchdog_start(watchdog); + delay = loop ? loop(event_server_name, event_server_argv) : -1; + event_loop(delay); + } + event_server_exit(); +} -- cgit v1.2.3