prefork.c

Apache 2.0.63 is the current stable version of the 2.0 series, and is recommended over any previous

    int curr_pollfd, last_pollfd = 0;
    apr_pollfd_t *pollset;
    int offset;
    void *csd;
    ap_sb_handle_t *sbh;
    apr_status_t rv;
    apr_bucket_alloc_t *bucket_alloc;

    mpm_state = AP_MPMQ_STARTING; /* for benefit of any hooks that run as this
                                  * child initializes
                                  */
    
    my_child_num = child_num_arg;
    ap_my_pid = getpid();
    csd = NULL;
    requests_this_child = 0;

    ap_fatal_signal_child_setup(ap_server_conf);

    /* Get a sub context for global allocations in this child, so that
     * we can have cleanups occur when the child exits.
     */
    apr_allocator_create(&allocator);
    apr_allocator_max_free_set(allocator, ap_max_mem_free);
    apr_pool_create_ex(&pchild, pconf, NULL, allocator);
    apr_allocator_owner_set(allocator, pchild);

    apr_pool_create(&ptrans, pchild);
    apr_pool_tag(ptrans, "transaction");

    /* needs to be done before we switch UIDs so we have permissions */
    ap_reopen_scoreboard(pchild, NULL, 0);
    rv = apr_proc_mutex_child_init(&accept_mutex, ap_lock_fname, pchild);
    if (rv != APR_SUCCESS) {
        ap_log_error(APLOG_MARK, APLOG_EMERG, rv, ap_server_conf,
                     "Couldn't initialize cross-process lock in child");
        clean_child_exit(APEXIT_CHILDFATAL);
    }

    if (unixd_setup_child()) {
	clean_child_exit(APEXIT_CHILDFATAL);
    }

    ap_run_child_init(pchild, ap_server_conf);

    ap_create_sb_handle(&sbh, pchild, my_child_num, 0);

    (void) ap_update_child_status(sbh, SERVER_READY, (request_rec *) NULL);

    /* Set up the pollfd array */
    listensocks = apr_pcalloc(pchild,
                            sizeof(*listensocks) * (num_listensocks));
    for (lr = ap_listeners, i = 0; i < num_listensocks; lr = lr->next, i++) {
        listensocks[i].accept_func = lr->accept_func;
        listensocks[i].sd = lr->sd;
    }

    pollset = apr_palloc(pchild, sizeof(*pollset) * num_listensocks);
    pollset[0].p = pchild;
    for (i = 0; i < num_listensocks; i++) {
        pollset[i].desc.s = listensocks[i].sd;
        pollset[i].desc_type = APR_POLL_SOCKET;
        pollset[i].reqevents = APR_POLLIN;
    }

    mpm_state = AP_MPMQ_RUNNING;
    
    bucket_alloc = apr_bucket_alloc_create(pchild);

    while (!die_now) {
	/*
	 * (Re)initialize this child to a pre-connection state.
	 */

	current_conn = NULL;

	apr_pool_clear(ptrans);

	if ((ap_max_requests_per_child > 0
	     && requests_this_child++ >= ap_max_requests_per_child)) {
	    clean_child_exit(0);
	}

	(void) ap_update_child_status(sbh, SERVER_READY, (request_rec *) NULL);

	/*
	 * Wait for an acceptable connection to arrive.
	 */

	/* Lock around "accept", if necessary */
	SAFE_ACCEPT(accept_mutex_on());

        if (num_listensocks == 1) {
            offset = 0;
        }
        else {
            /* multiple listening sockets - need to poll */
	    for (;;) {
                apr_status_t ret;
                apr_int32_t n;

                ret = apr_poll(pollset, num_listensocks, &n, -1);
                if (ret != APR_SUCCESS) {
                    if (APR_STATUS_IS_EINTR(ret)) {
                        continue;
                    }
    	            /* Single Unix documents select as returning errnos
    	             * EBADF, EINTR, and EINVAL... and in none of those
    	             * cases does it make sense to continue.  In fact
    	             * on Linux 2.0.x we seem to end up with EFAULT
    	             * occasionally, and we'd loop forever due to it.
    	             */
    	            ap_log_error(APLOG_MARK, APLOG_ERR, ret, ap_server_conf,
                             "apr_poll: (listen)");
    	            clean_child_exit(1);
                }
                /* find a listener */
                curr_pollfd = last_pollfd;
                do {
                    curr_pollfd++;
                    if (curr_pollfd >= num_listensocks) {
                        curr_pollfd = 0;
                    }
                    /* XXX: Should we check for POLLERR? */
                    if (pollset[curr_pollfd].rtnevents & APR_POLLIN) {
                        last_pollfd = curr_pollfd;
                        offset = curr_pollfd;
                        goto got_fd;
                    }
                } while (curr_pollfd != last_pollfd);

                continue;
            }
        }
    got_fd:
	/* if we accept() something we don't want to die, so we have to
	 * defer the exit
	 */
        status = listensocks[offset].accept_func(&csd, 
                                                 &listensocks[offset], ptrans);
        SAFE_ACCEPT(accept_mutex_off());	/* unlock after "accept" */

        if (status == APR_EGENERAL) {
            /* resource shortage or should-not-occur occured */
            clean_child_exit(1);
        }
        else if (status != APR_SUCCESS) {
            continue;
        }

	/*
	 * We now have a connection, so set it up with the appropriate
	 * socket options, file descriptors, and read/write buffers.
	 */

	current_conn = ap_run_create_connection(ptrans, ap_server_conf, csd, my_child_num, sbh, bucket_alloc);
        if (current_conn) {
            ap_process_connection(current_conn, csd);
            ap_lingering_close(current_conn);
        }
        
        /* Check the pod and the generation number after processing a
         * connection so that we'll go away if a graceful restart occurred
         * while we were processing the connection or we are the lucky
         * idle server process that gets to die.
         */
        if (ap_mpm_pod_check(pod) == APR_SUCCESS) { /* selected as idle? */
            die_now = 1;
        }
        else if (ap_my_generation !=
                 ap_scoreboard_image->global->running_generation) { /* restart? */
            /* yeah, this could be non-graceful restart, in which case the
             * parent will kill us soon enough, but why bother checking?
             */
            die_now = 1;
        }
    }
    clean_child_exit(0);
}


static int make_child(server_rec *s, int slot)
{
    int pid;

    if (slot + 1 > ap_max_daemons_limit) {
	ap_max_daemons_limit = slot + 1;
    }

    if (one_process) {
	apr_signal(SIGHUP, just_die);
        /* Don't catch AP_SIG_GRACEFUL in ONE_PROCESS mode :) */
	apr_signal(SIGINT, just_die);
#ifdef SIGQUIT
	apr_signal(SIGQUIT, SIG_DFL);
#endif
	apr_signal(SIGTERM, just_die);
	child_main(slot);
    }

    (void) ap_update_child_status_from_indexes(slot, 0, SERVER_STARTING,
                                               (request_rec *) NULL);


#ifdef _OSD_POSIX
    /* BS2000 requires a "special" version of fork() before a setuid() call */
    if ((pid = os_fork(unixd_config.user_name)) == -1) {
#elif defined(TPF)
    if ((pid = os_fork(s, slot)) == -1) {
#else
    if ((pid = fork()) == -1) {
#endif
	ap_log_error(APLOG_MARK, APLOG_ERR, errno, s, "fork: Unable to fork new process");

	/* fork didn't succeed. Fix the scoreboard or else
	 * it will say SERVER_STARTING forever and ever
	 */
	(void) ap_update_child_status_from_indexes(slot, 0, SERVER_DEAD,
                                                   (request_rec *) NULL);

	/* In case system resources are maxxed out, we don't want
	   Apache running away with the CPU trying to fork over and
	   over and over again. */
	sleep(10);

	return -1;
    }

    if (!pid) {
#ifdef HAVE_BINDPROCESSOR
        /* by default AIX binds to a single processor
         * this bit unbinds children which will then bind to another cpu
         */
	int status = bindprocessor(BINDPROCESS, (int)getpid(), 
				   PROCESSOR_CLASS_ANY);
	if (status != OK) {
	    ap_log_error(APLOG_MARK, APLOG_WARNING, errno, 
                         ap_server_conf, "processor unbind failed %d", status);
	}
#endif
	RAISE_SIGSTOP(MAKE_CHILD);
        AP_MONCONTROL(1);
        /* Disable the parent's signal handlers and set up proper handling in
         * the child.
	 */
	apr_signal(SIGHUP, just_die);
	apr_signal(SIGTERM, just_die);
        /* The child process doesn't do anything for AP_SIG_GRACEFUL.  
         * Instead, the pod is used for signalling graceful restart.
         */
        apr_signal(AP_SIG_GRACEFUL, SIG_IGN);
	child_main(slot);
    }

    ap_scoreboard_image->parent[slot].pid = pid;

    return 0;
}


/* start up a bunch of children */
static void startup_children(int number_to_start)
{
    int i;

    for (i = 0; number_to_start && i < ap_daemons_limit; ++i) {
	if (ap_scoreboard_image->servers[i][0].status != SERVER_DEAD) {
	    continue;
	}
	if (make_child(ap_server_conf, i) < 0) {
	    break;
	}
	--number_to_start;
    }
}


/*
 * idle_spawn_rate is the number of children that will be spawned on the
 * next maintenance cycle if there aren't enough idle servers.  It is
 * doubled up to MAX_SPAWN_RATE, and reset only when a cycle goes by
 * without the need to spawn.
 */
static int idle_spawn_rate = 1;
#ifndef MAX_SPAWN_RATE
#define MAX_SPAWN_RATE	(32)
#endif
static int hold_off_on_exponential_spawning;

static void perform_idle_server_maintenance(apr_pool_t *p)
{
    int i;
    int to_kill;
    int idle_count;
    worker_score *ws;
    int free_length;
    int free_slots[MAX_SPAWN_RATE];
    int last_non_dead;
    int total_non_dead;

    /* initialize the free_list */
    free_length = 0;

    to_kill = -1;
    idle_count = 0;
    last_non_dead = -1;
    total_non_dead = 0;

    for (i = 0; i < ap_daemons_limit; ++i) {
	int status;

	if (i >= ap_max_daemons_limit && free_length == idle_spawn_rate)
	    break;
	ws = &ap_scoreboard_image->servers[i][0];
	status = ws->status;
	if (status == SERVER_DEAD) {
	    /* try to keep children numbers as low as possible */
	    if (free_length < idle_spawn_rate) {
		free_slots[free_length] = i;
		++free_length;
	    }
	}
	else {
	    /* We consider a starting server as idle because we started it
	     * at least a cycle ago, and if it still hasn't finished starting
	     * then we're just going to swamp things worse by forking more.
	     * So we hopefully won't need to fork more if we count it.
	     * This depends on the ordering of SERVER_READY and SERVER_STARTING.
	     */
	    if (status <= SERVER_READY) {
		++ idle_count;
		/* always kill the highest numbered child if we have to...
		 * no really well thought out reason ... other than observing
		 * the server behaviour under linux where lower numbered children
		 * tend to service more hits (and hence are more likely to have
		 * their data in cpu caches).
		 */
		to_kill = i;
	    }

	    ++total_non_dead;
	    last_non_dead = i;
	}
    }
    ap_max_daemons_limit = last_non_dead + 1;
    if (idle_count > ap_daemons_max_free) {
	/* kill off one child... we use the pod because that'll cause it to
	 * shut down gracefully, in case it happened to pick up a request
	 * while we were counting
	 */
	ap_mpm_pod_signal(pod);
	idle_spawn_rate = 1;
    }
    else if (idle_count < ap_daemons_min_free) {
	/* terminate the free list */
	if (free_length == 0) {
	    /* only report this condition once */
	    static int reported = 0;

	    if (!reported) {
		ap_log_error(APLOG_MARK, APLOG_ERR, 0, ap_server_conf,
			    "server reached MaxClients setting, consider"
			    " raising the MaxClients setting");
		reported = 1;
	    }
	    idle_spawn_rate = 1;
	}
	else {
	    if (idle_spawn_rate >= 8) {
		ap_log_error(APLOG_MARK, APLOG_INFO, 0, ap_server_conf,
		    "server seems busy, (you may need "
		    "to increase StartServers, or Min/MaxSpareServers), "
		    "spawning %d children, there are %d idle, and "
		    "%d total children", idle_spawn_rate,
		    idle_count, total_non_dead);
	    }
	    for (i = 0; i < free_length; ++i) {
#ifdef TPF
        if (make_child(ap_server_conf, free_slots[i]) == -1) {
            if(free_length == 1) {
                shutdown_pending = 1;
                ap_log_error(APLOG_MARK, APLOG_EMERG, 0, ap_server_conf,
                "No active child processes: shutting down");
            }
        }
#else
		make_child(ap_server_conf, free_slots[i]);
#endif /* TPF */
	    }
	    /* the next time around we want to spawn twice as many if this
	     * wasn't good enough, but not if we've just done a graceful
	     */
	    if (hold_off_on_exponential_spawning) {
		--hold_off_on_exponential_spawning;
	    }
	    else if (idle_spawn_rate < MAX_SPAWN_RATE) {
		idle_spawn_rate *= 2;
	    }
	}
    }
    else {
	idle_spawn_rate = 1;
    }
}

/*****************************************************************
 * Executive routines.
 */

int ap_mpm_run(apr_pool_t *_pconf, apr_pool_t *plog, server_rec *s)
{
    int index;
    int remaining_children_to_start;
    apr_status_t rv;

    ap_log_pid(pconf, ap_pid_fname);

    first_server_limit = server_limit;
    if (changed_limit_at_restart) {
        ap_log_error(APLOG_MARK, APLOG_WARNING, 0, s,
                     "WARNING: Attempt to change ServerLimit "
                     "ignored during restart");
        changed_limit_at_restart = 0;
    }

    /* Initialize cross-process accept lock */
    ap_lock_fname = apr_psprintf(_pconf, "%s.%" APR_PID_T_FMT,
                                 ap_server_root_relative(_pconf, ap_lock_fname),
                                 ap_my_pid);

    rv = apr_proc_mutex_create(&accept_mutex, ap_lock_fname, 
                               ap_accept_lock_mech, _pconf);
    if (rv != APR_SUCCESS) {
        ap_log_error(APLOG_MARK, APLOG_EMERG, rv, s,
                     "Couldn't create accept lock");
        mpm_state = AP_MPMQ_STOPPING;
        return 1;
    }

#if APR_USE_SYSVSEM_SERIALIZE
    if (ap_accept_lock_mech == APR_LOCK_DEFAULT || 
        ap_accept_lock_mech == APR_LOCK_SYSVSEM) {
#else
    if (ap_accept_lock_mech == APR_LOCK_SYSVSEM) {
#endif
        rv = unixd_set_proc_mutex_perms(accept_mutex);
        if (rv != APR_SUCCESS) {
            ap_log_error(APLOG_MARK, APLOG_EMERG, rv, s,
                         "Couldn't set permissions on cross-process lock; "
                         "check User and Group directives");
            mpm_state = AP_MPMQ_STOPPING;
            return 1;