mpm_common.c

Apache HTTP Server 是一个功能强大的灵活的与HTTP/1.1相兼容的web服务器.这里给出的是Apache HTTP服务器的源码。

/* Copyright 2000-2005 The Apache Software Foundation or its licensors, as
 * applicable.
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

/* The purpose of this file is to store the code that MOST mpm's will need
 * this does not mean a function only goes into this file if every MPM needs
 * it.  It means that if a function is needed by more than one MPM, and
 * future maintenance would be served by making the code common, then the
 * function belongs here.
 *
 * This is going in src/main because it is not platform specific, it is
 * specific to multi-process servers, but NOT to Unix.  Which is why it
 * does not belong in src/os/unix
 */

#include "apr.h"
#include "apr_thread_proc.h"
#include "apr_signal.h"
#include "apr_strings.h"
#define APR_WANT_STRFUNC
#include "apr_want.h"
#include "apr_getopt.h"
#include "apr_optional.h"
#include "apr_allocator.h"

#include "httpd.h"
#include "http_config.h"
#include "http_log.h"
#include "http_main.h"
#include "mpm.h"
#include "mpm_common.h"
#include "ap_mpm.h"
#include "ap_listen.h"
#include "mpm_default.h"

#ifdef AP_MPM_WANT_SET_SCOREBOARD
#include "scoreboard.h"
#endif

#ifdef HAVE_PWD_H
#include <pwd.h>
#endif
#ifdef HAVE_GRP_H
#include <grp.h>
#endif
#if APR_HAVE_UNISTD_H
#include <unistd.h>
#endif

#ifdef AP_MPM_WANT_RECLAIM_CHILD_PROCESSES
void ap_reclaim_child_processes(int terminate)
{
    int i;
    long int waittime = 1024 * 16;      /* in usecs */
    apr_status_t waitret;
    int tries;
    int not_dead_yet;
    int max_daemons;

    ap_mpm_query(AP_MPMQ_MAX_DAEMON_USED, &max_daemons);

    for (tries = terminate ? 4 : 1; tries <= 9; ++tries) {
        /* don't want to hold up progress any more than
         * necessary, but we need to allow children a few moments to exit.
         * Set delay with an exponential backoff.
         */
        apr_sleep(waittime);
        waittime = waittime * 4;

        /* now see who is done */
        not_dead_yet = 0;
        for (i = 0; i < max_daemons; ++i) {
            pid_t pid = MPM_CHILD_PID(i);
            apr_proc_t proc;

            if (pid == 0)
                continue;

            proc.pid = pid;
            waitret = apr_proc_wait(&proc, NULL, NULL, APR_NOWAIT);
            if (waitret != APR_CHILD_NOTDONE) {
                MPM_NOTE_CHILD_KILLED(i);
                continue;
            }

            ++not_dead_yet;
            switch (tries) {
            case 1:     /*  16ms */
            case 2:     /*  82ms */
            case 3:     /* 344ms */
            case 4:     /*  16ms */
                break;

            case 5:     /*  82ms */
            case 6:     /* 344ms */
            case 7:     /* 1.4sec */
                /* ok, now it's being annoying */
                ap_log_error(APLOG_MARK, APLOG_WARNING,
                             0, ap_server_conf,
                             "child process %ld still did not exit, "
                             "sending a SIGTERM",
                             (long)pid);
                kill(pid, SIGTERM);
                break;

            case 8:     /*  6 sec */
                /* die child scum */
                ap_log_error(APLOG_MARK, APLOG_ERR,
                             0, ap_server_conf,
                             "child process %ld still did not exit, "
                             "sending a SIGKILL",
                             (long)pid);
#ifndef BEOS
                kill(pid, SIGKILL);
#else
                /* sending a SIGKILL kills the entire team on BeOS, and as
                 * httpd thread is part of that team it removes any chance
                 * of ever doing a restart.  To counter this I'm changing to
                 * use a kinder, gentler way of killing a specific thread
                 * that is just as effective.
                 */
                kill_thread(pid);
#endif
                break;

            case 9:     /* 14 sec */
                /* gave it our best shot, but alas...  If this really
                 * is a child we are trying to kill and it really hasn't
                 * exited, we will likely fail to bind to the port
                 * after the restart.
                 */
                ap_log_error(APLOG_MARK, APLOG_ERR,
                             0, ap_server_conf,
                             "could not make child process %ld exit, "
                             "attempting to continue anyway",
                             (long)pid);
                break;
            }
        }

#if APR_HAS_OTHER_CHILD
        apr_proc_other_child_check();
#endif

        if (!not_dead_yet) {
            /* nothing left to wait for */
            break;
        }
    }
}
#endif /* AP_MPM_WANT_RECLAIM_CHILD_PROCESSES */

#ifdef AP_MPM_WANT_WAIT_OR_TIMEOUT

/* number of calls to wait_or_timeout between writable probes */
#ifndef INTERVAL_OF_WRITABLE_PROBES
#define INTERVAL_OF_WRITABLE_PROBES 10
#endif
static int wait_or_timeout_counter;

void ap_wait_or_timeout(apr_exit_why_e *status, int *exitcode, apr_proc_t *ret,
                        apr_pool_t *p)
{
    apr_status_t rv;

    ++wait_or_timeout_counter;
    if (wait_or_timeout_counter == INTERVAL_OF_WRITABLE_PROBES) {
        wait_or_timeout_counter = 0;
    }

    rv = apr_proc_wait_all_procs(ret, exitcode, status, APR_NOWAIT, p);
    if (APR_STATUS_IS_EINTR(rv)) {
        ret->pid = -1;
        return;
    }

    if (APR_STATUS_IS_CHILD_DONE(rv)) {
        return;
    }

#ifdef NEED_WAITPID
    if ((ret = reap_children(exitcode, status)) > 0) {
        return;
    }
#endif

    apr_sleep(SCOREBOARD_MAINTENANCE_INTERVAL);
    ret->pid = -1;
    return;
}
#endif /* AP_MPM_WANT_WAIT_OR_TIMEOUT */

#ifdef AP_MPM_WANT_PROCESS_CHILD_STATUS
int ap_process_child_status(apr_proc_t *pid, apr_exit_why_e why, int status)
{
    int signum = status;
    const char *sigdesc = apr_signal_description_get(signum);

    /* Child died... if it died due to a fatal error,
     * we should simply bail out.  The caller needs to
     * check for bad rc from us and exit, running any
     * appropriate cleanups.
     *
     * If the child died due to a resource shortage, 
     * the parent should limit the rate of forking
     */
    if (APR_PROC_CHECK_EXIT(why)) {
        if (status == APEXIT_CHILDSICK) {
            return status;
        }

        if (status == APEXIT_CHILDFATAL) {
            ap_log_error(APLOG_MARK, APLOG_ALERT,
                         0, ap_server_conf,
                         "Child %" APR_PID_T_FMT
                         " returned a Fatal error..." APR_EOL_STR
                         "Apache is exiting!",
                         pid->pid);
            return APEXIT_CHILDFATAL;
        }

        return 0;
    }

    if (APR_PROC_CHECK_SIGNALED(why)) {
        switch (signum) {
        case SIGTERM:
        case SIGHUP:
        case AP_SIG_GRACEFUL:
        case SIGKILL:
            break;

        default:
            if (APR_PROC_CHECK_CORE_DUMP(why)) {
                ap_log_error(APLOG_MARK, APLOG_NOTICE,
                             0, ap_server_conf,
                             "child pid %ld exit signal %s (%d), "
                             "possible coredump in %s",
                             (long)pid->pid, sigdesc, signum,
                             ap_coredump_dir);
            }
            else {
                ap_log_error(APLOG_MARK, APLOG_NOTICE,
                             0, ap_server_conf,
                             "child pid %ld exit signal %s (%d)",
                             (long)pid->pid, sigdesc, signum);
            }
        }
    }
    return 0;
}
#endif /* AP_MPM_WANT_PROCESS_CHILD_STATUS */

#if defined(TCP_NODELAY) && !defined(MPE) && !defined(TPF)
void ap_sock_disable_nagle(apr_socket_t *s)
{
    /* The Nagle algorithm says that we should delay sending partial
     * packets in hopes of getting more data.  We don't want to do
     * this; we are not telnet.  There are bad interactions between
     * persistent connections and Nagle's algorithm that have very severe
     * performance penalties.  (Failing to disable Nagle is not much of a
     * problem with simple HTTP.)
     *
     * In spite of these problems, failure here is not a shooting offense.
     */
    apr_status_t status = apr_socket_opt_set(s, APR_TCP_NODELAY, 1);

    if (status != APR_SUCCESS) {
        ap_log_error(APLOG_MARK, APLOG_WARNING, status, ap_server_conf,
                     "apr_socket_opt_set: (TCP_NODELAY)");
    }
}
#endif

#ifdef HAVE_GETPWNAM
AP_DECLARE(uid_t) ap_uname2id(const char *name)
{
    struct passwd *ent;

    if (name[0] == '#')
        return (atoi(&name[1]));

    if (!(ent = getpwnam(name))) {
        ap_log_error(APLOG_MARK, APLOG_STARTUP, 0, NULL,
                     "%s: bad user name %s", ap_server_argv0, name);
        exit(1);
    }

    return (ent->pw_uid);
}
#endif

#ifdef HAVE_GETGRNAM
AP_DECLARE(gid_t) ap_gname2id(const char *name)
{
    struct group *ent;

    if (name[0] == '#')
        return (atoi(&name[1]));

    if (!(ent = getgrnam(name))) {
        ap_log_error(APLOG_MARK, APLOG_STARTUP, 0, NULL,
                     "%s: bad group name %s", ap_server_argv0, name);
        exit(1);
    }

    return (ent->gr_gid);
}
#endif

#ifndef HAVE_INITGROUPS
int initgroups(const char *name, gid_t basegid)
{
#if defined(QNX) || defined(MPE) || defined(BEOS) || defined(_OSD_POSIX) || defined(TPF) || defined(__TANDEM) || defined(OS2) || defined(WIN32) || defined(NETWARE)
/* QNX, MPE and BeOS do not appear to support supplementary groups. */
    return 0;
#else /* ndef QNX */
    gid_t groups[NGROUPS_MAX];
    struct group *g;
    int index = 0;

    setgrent();

    groups[index++] = basegid;

    while (index < NGROUPS_MAX && ((g = getgrent()) != NULL)) {
        if (g->gr_gid != basegid) {
            char **names;

            for (names = g->gr_mem; *names != NULL; ++names) {
                if (!strcmp(*names, name))
                    groups[index++] = g->gr_gid;
            }
        }
    }

    endgrent();

    return setgroups(index, groups);
#endif /* def QNX */
}
#endif /* def NEED_INITGROUPS */

#ifdef AP_MPM_USES_POD

AP_DECLARE(apr_status_t) ap_mpm_pod_open(apr_pool_t *p, ap_pod_t **pod)
{
    apr_status_t rv;

    *pod = apr_palloc(p, sizeof(**pod));
    rv = apr_file_pipe_create(&((*pod)->pod_in), &((*pod)->pod_out), p);
    if (rv != APR_SUCCESS) {
        return rv;
    }

    apr_file_pipe_timeout_set((*pod)->pod_in, 0);
    (*pod)->p = p;

    apr_sockaddr_info_get(&(*pod)->sa, ap_listeners->bind_addr->hostname,
                          APR_UNSPEC, ap_listeners->bind_addr->port, 0, p);

    /* close these before exec. */
    apr_file_unset_inherit((*pod)->pod_in);
    apr_file_unset_inherit((*pod)->pod_out);

    return APR_SUCCESS;
}

AP_DECLARE(apr_status_t) ap_mpm_pod_check(ap_pod_t *pod)
{
    char c;
    apr_size_t len = 1;
    apr_status_t rv;

    rv = apr_file_read(pod->pod_in, &c, &len);

    if ((rv == APR_SUCCESS) && (len == 1)) {
        return APR_SUCCESS;
    }

    if (rv != APR_SUCCESS) {
        return rv;
    }

    return AP_NORESTART;
}

AP_DECLARE(apr_status_t) ap_mpm_pod_close(ap_pod_t *pod)
{
    apr_status_t rv;

    rv = apr_file_close(pod->pod_out);
    if (rv != APR_SUCCESS) {
        return rv;
    }

    rv = apr_file_close(pod->pod_in);
    if (rv != APR_SUCCESS) {
        return rv;
    }

    return APR_SUCCESS;
}

static apr_status_t pod_signal_internal(ap_pod_t *pod)
{
    apr_status_t rv;
    char char_of_death = '!';
    apr_size_t one = 1;

    rv = apr_file_write(pod->pod_out, &char_of_death, &one);
    if (rv != APR_SUCCESS) {
        ap_log_error(APLOG_MARK, APLOG_WARNING, rv, ap_server_conf,
                     "write pipe_of_death");
    }

    return rv;
}

/* This function connects to the server, then immediately closes the connection.
 * This permits the MPM to skip the poll when there is only one listening
 * socket, because it provides a alternate way to unblock an accept() when
 * the pod is used.
 */
static apr_status_t dummy_connection(ap_pod_t *pod)
{
    apr_status_t rv;
    apr_socket_t *sock;
    apr_pool_t *p;

    /* create a temporary pool for the socket.  pconf stays around too long */
    rv = apr_pool_create(&p, pod->p);
    if (rv != APR_SUCCESS) {
        return rv;
    }

    rv = apr_socket_create(&sock, pod->sa->family, SOCK_STREAM, p);
    if (rv != APR_SUCCESS) {
        ap_log_error(APLOG_MARK, APLOG_WARNING, rv, ap_server_conf,
                     "get socket to connect to listener");
        return rv;
    }

    /* on some platforms (e.g., FreeBSD), the kernel won't accept many
     * queued connections before it starts blocking local connects...
     * we need to keep from blocking too long and instead return an error,
     * because the MPM won't want to hold up a graceful restart for a
     * long time
     */
    rv = apr_socket_timeout_set(sock, apr_time_from_sec(3));
    if (rv != APR_SUCCESS) {
        ap_log_error(APLOG_MARK, APLOG_WARNING, rv, ap_server_conf,
                     "set timeout on socket to connect to listener");
        apr_socket_close(sock);
        return rv;
    }

    rv = apr_connect(sock, pod->sa);
    if (rv != APR_SUCCESS) {
        int log_level = APLOG_WARNING;

        if (APR_STATUS_IS_TIMEUP(rv)) {
            /* probably some server processes bailed out already and there
             * is nobody around to call accept and clear out the kernel
             * connection queue; usually this is not worth logging
             */
            log_level = APLOG_DEBUG;
        }

        ap_log_error(APLOG_MARK, log_level, rv, ap_server_conf,
                     "connect to listener");
    }

    apr_socket_close(sock);
    apr_pool_destroy(p);

    return rv;
}

AP_DECLARE(apr_status_t) ap_mpm_pod_signal(ap_pod_t *pod)
{
    apr_status_t rv;

    rv = pod_signal_internal(pod);
    if (rv != APR_SUCCESS) {
        return rv;
    }

    return dummy_connection(pod);
}

void ap_mpm_pod_killpg(ap_pod_t *pod, int num)
{
    int i;
    apr_status_t rv = APR_SUCCESS;

    /* we don't write anything to the pod here...  we assume
     * that the would-be reader of the pod has another way to
     * see that it is time to die once we wake it up
     *
     * writing lots of things to the pod at once is very
     * problematic... we can fill the kernel pipe buffer and
     * be blocked until somebody consumes some bytes or
     * we hit a timeout...  if we hit a timeout we can't just