spmt_os2.c

Apache V2.0.15 Alpha For Linuxhttpd-2_0_15-alpha.tar.Z

    r->connection->keepalive = 0;
    thread_control[THREAD_GLOBAL(thread_num)].deferred_die = 1;
}



int ap_graceful_stop_signalled(void)
{
    if (thread_control[THREAD_GLOBAL(thread_num)].deferred_die ||
	ap_scoreboard_image->global.running_generation != thread_control[THREAD_GLOBAL(thread_num)].generation) {
	return 1;
    }
    return 0;
}



int ap_stop_signalled(void)
{
    if (shutdown_pending || restart_pending ||
        thread_control[THREAD_GLOBAL(thread_num)].deferred_die ||
	ap_scoreboard_image->global.running_generation != thread_control[THREAD_GLOBAL(thread_num)].generation) {
	return 1;
    }
    return 0;
}



static int setup_listen_poll(apr_pool_t *pchild, apr_pollfd_t **listen_poll)
{
    ap_listen_rec *lr;
    int numfds = 0;

    for (lr = ap_listeners; lr; lr = lr->next) {
        numfds++;
    }

    apr_poll_setup(listen_poll, numfds, pchild);

    for (lr = ap_listeners; lr; lr = lr->next) {
	apr_poll_socket_add(*listen_poll, lr->sd, APR_POLLIN);
    }
    return 0;
}



static void thread_main(void *thread_num_arg)
{
    ap_listen_rec *lr = NULL;
    ap_listen_rec *first_lr = NULL;
    apr_pool_t *ptrans;
    conn_rec *current_conn;
    apr_pool_t *pchild;
    int requests_this_child = 0;
    apr_pollfd_t *listen_poll;
    apr_socket_t *csd = NULL;
    int nsds, rv;

    /* Disable the restart signal handlers and enable the just_die stuff.
     * Note that since restart() just notes that a restart has been
     * requested there's no race condition here.
     */

    set_signals(); /* signals aren't inherrited by child threads */
    signal(SIGHUP, just_die);
    signal(SIGUSR1, just_die);
    signal(SIGTERM, just_die);

    /* Get a sub pool for global allocations in this child, so that
     * we can have cleanups occur when the child exits.
     */
    apr_pool_create(&pchild, pconf);
    *ppthread_globals = (struct thread_globals *)apr_palloc(pchild, sizeof(struct thread_globals));
    THREAD_GLOBAL(thread_num) = (int)thread_num_arg;
    THREAD_GLOBAL(pchild) = pchild;
    thread_control[THREAD_GLOBAL(thread_num)].generation = ap_scoreboard_image->global.running_generation;
    apr_pool_create(&ptrans, pchild);

    if (setup_listen_poll(pchild, &listen_poll)) {
	clean_child_exit(1);
    }

    /* needs to be done before we switch UIDs so we have permissions */
    SAFE_ACCEPT(accept_mutex_child_init(pchild));

    ap_run_child_init(pchild, ap_server_conf);

    (void) ap_update_child_status(0, THREAD_GLOBAL(thread_num), SERVER_READY, (request_rec *) NULL);
    

    signal(SIGHUP, just_die);
    signal(SIGTERM, just_die);

    while (!ap_stop_signalled()) {
        int srv;
        apr_socket_t *sd;

	/* Prepare to receive a SIGUSR1 due to graceful restart so that
	 * we can exit cleanly.
	 */
	THREAD_GLOBAL(usr1_just_die) = 1;
	signal(SIGUSR1, usr1_handler);

	/*
	 * (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(0, THREAD_GLOBAL(thread_num), SERVER_READY, (request_rec *) NULL);

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

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

        if (ap_stop_signalled()) {
            clean_child_exit(0);
        }

	for (;;) {
	    if (ap_listeners->next) {
		/* more than one socket */
                srv = apr_poll(listen_poll, &nsds, -1);

		if (srv != APR_SUCCESS) {
		    /* 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, errno, ap_server_conf, "select: (listen)");
		    clean_child_exit(1);
		}

		/* we remember the last_lr we searched last time around so that
		   we don't end up starving any particular listening socket */
		if (first_lr == NULL) {
		    first_lr = ap_listeners;
		}
		
                lr = first_lr;
		
                do {
                    apr_int16_t event;

		    if (!lr) {
			lr = ap_listeners;
		    }

                    apr_poll_revents_get(&event, lr->sd, listen_poll);

		    if (event == APR_POLLIN) {
                        first_lr = lr->next;
		        break;
		    }
		    lr = lr->next;
		} while (lr != first_lr);
		
		if (lr == first_lr) {
		    continue;
		}
		sd = lr->sd;
	    }
	    else {
		/* only one socket, just pretend we did the other stuff */
		sd = ap_listeners->sd;
	    }

	    /* if we accept() something we don't want to die, so we have to
	     * defer the exit
	     */
            THREAD_GLOBAL(usr1_just_die) = 0;
            rv = apr_accept(&csd, sd, ptrans);

            if (APR_STATUS_IS_SUCCESS(rv)) {
		break;		/* We have a socket ready for reading */
            }
            else if (APR_STATUS_IS_ECONNABORTED(rv) 
                  || APR_STATUS_IS_ECONNRESET(rv)
                  || APR_STATUS_IS_ETIMEDOUT(rv)
                  || APR_STATUS_IS_EHOSTUNREACH(rv)
                  || APR_STATUS_IS_ENETUNREACH(rv)) {

		/* Our old behaviour here was to continue after accept()
		 * errors.  But this leads us into lots of troubles
		 * because most of the errors are quite fatal.  For
		 * example, EMFILE can be caused by slow descriptor
		 * leaks (say in a 3rd party module, or libc).  It's
		 * foolish for us to continue after an EMFILE.  We also
		 * seem to tickle kernel bugs on some platforms which
		 * lead to never-ending loops here.  So it seems best
		 * to just exit in most cases.
		 */

                /* Linux generates most of these, other tcp
                 * stacks (i.e. bsd) tend to hide them behind
                 * getsockopt() interfaces.  They occur when
                 * the net goes sour or the client disconnects
                 * after the three-way handshake has been done
                 * in the kernel but before userland has picked
                 * up the socket.
                 */
                 break;
            }
            else if (APR_STATUS_IS_EINTR(rv)) {
                /* We only get hit by an EINTR if the parent is
                 * killing us off
                 */
                clean_child_exit(0);
            }
            else {
                ap_log_error(APLOG_MARK, APLOG_ERR, rv, ap_server_conf,
                             "accept: (client socket)");
                clean_child_exit(1);
	    }

	    if (ap_stop_signalled()) {
		clean_child_exit(0);
	    }
	    THREAD_GLOBAL(usr1_just_die) = 1;
	}

	SAFE_ACCEPT(accept_mutex_off());	/* unlock after "accept" */

	/* We've got a socket, let's at least process one request off the
	 * socket before we accept a graceful restart request.  We set
	 * the signal to ignore because we don't want to disturb any
	 * third party code.
	 */
	signal(SIGUSR1, SIG_IGN);

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

	ap_sock_disable_nagle(csd);

	current_conn = ap_new_connection(ptrans, ap_server_conf, csd,
                                         THREAD_GLOBAL(thread_num));

        if (current_conn) {
            ap_process_connection(current_conn);
            ap_lingering_close(current_conn);
        }
    }

    clean_child_exit(0);
}


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

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

    if (one_process) {
        struct thread_globals *parent_globals = *ppthread_globals;
	signal(SIGHUP, just_die);
	signal(SIGINT, just_die);
#ifdef SIGQUIT
	signal(SIGQUIT, SIG_DFL);
#endif
	signal(SIGTERM, just_die);
        thread_main((void *)slot);
        *ppthread_globals = parent_globals;
    }

    ap_update_child_status(0, slot, SERVER_STARTING, (request_rec *) NULL);

    if ((tid = _beginthread(thread_main, NULL,  256*1024, (void *)slot)) == -1) {
	ap_log_error(APLOG_MARK, APLOG_ERR|APLOG_NOERRNO, 0, s, "_beginthread: Unable to create new thread");

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

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

	return -1;
    }

    ap_scoreboard_image->servers[0][slot].tid = tid;
    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[0][i].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(void)
{
    int i;
    int to_kill;
    int idle_count;
    short_score *ss;
    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 >= max_daemons_limit && free_length == idle_spawn_rate)
	    break;
	ss = &ap_scoreboard_image->servers[0][i];
	status = ss->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;
	}
    }
    max_daemons_limit = last_non_dead + 1;
    if (idle_count > ap_daemons_max_free) {
	/* kill off one child... we use SIGUSR1 because that'll cause it to
	 * shut down gracefully, in case it happened to pick up a request
	 * while we were counting
	 */
	thread_control[to_kill].deferred_die = 1;
	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_NOERRNO|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_NOERRNO|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) {
		make_child(ap_server_conf, free_slots[i]);
	    }
	    /* 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 {