dexter.c revision b0f20a4a26bcfa85724b1c2e5ec6a077f12ef44c
e609c337f729875bc20e01096c7e610f45356f54nilgun/* ====================================================================
e609c337f729875bc20e01096c7e610f45356f54nilgun * Copyright (c) 1995-1999 The Apache Group. All rights reserved.
fd9abdda70912b99b24e3bf1a38f26fde908a74cnd *
fd9abdda70912b99b24e3bf1a38f26fde908a74cnd * Redistribution and use in source and binary forms, with or without
fd9abdda70912b99b24e3bf1a38f26fde908a74cnd * modification, are permitted provided that the following conditions
e609c337f729875bc20e01096c7e610f45356f54nilgun * are met:
e609c337f729875bc20e01096c7e610f45356f54nilgun *
e609c337f729875bc20e01096c7e610f45356f54nilgun * 1. Redistributions of source code must retain the above copyright
e609c337f729875bc20e01096c7e610f45356f54nilgun * notice, this list of conditions and the following disclaimer.
96ad5d81ee4a2cc66a4ae19893efc8aa6d06fae7jailletc *
e609c337f729875bc20e01096c7e610f45356f54nilgun * 2. Redistributions in binary form must reproduce the above copyright
e609c337f729875bc20e01096c7e610f45356f54nilgun * notice, this list of conditions and the following disclaimer in
d29d9ab4614ff992b0e8de6e2b88d52b6f1f153erbowen * the documentation and/or other materials provided with the
2e545ce2450a9953665f701bb05350f0d3f26275nd * distribution.
d29d9ab4614ff992b0e8de6e2b88d52b6f1f153erbowen *
d29d9ab4614ff992b0e8de6e2b88d52b6f1f153erbowen * 3. All advertising materials mentioning features or use of this
e609c337f729875bc20e01096c7e610f45356f54nilgun * software must display the following acknowledgment:
e609c337f729875bc20e01096c7e610f45356f54nilgun * "This product includes software developed by the Apache Group
af33a4994ae2ff15bc67d19ff1a7feb906745bf8rbowen * for use in the Apache HTTP server project (http://www.apache.org/)."
3f08db06526d6901aa08c110b5bc7dde6bc39905nd *
e609c337f729875bc20e01096c7e610f45356f54nilgun * 4. The names "Apache Server" and "Apache Group" must not be used to
e609c337f729875bc20e01096c7e610f45356f54nilgun * endorse or promote products derived from this software without
e609c337f729875bc20e01096c7e610f45356f54nilgun * prior written permission. For written permission, please contact
3f08db06526d6901aa08c110b5bc7dde6bc39905nd * apache@apache.org.
e609c337f729875bc20e01096c7e610f45356f54nilgun *
e609c337f729875bc20e01096c7e610f45356f54nilgun * 5. Products derived from this software may not be called "Apache"
e609c337f729875bc20e01096c7e610f45356f54nilgun * nor may "Apache" appear in their names without prior written
27dcd8d81085fd60aadcd8a9bad35a607b26b758nilgun * permission of the Apache Group.
e609c337f729875bc20e01096c7e610f45356f54nilgun *
e609c337f729875bc20e01096c7e610f45356f54nilgun * 6. Redistributions of any form whatsoever must retain the following
f086b4b402fa9a2fefc7dda85de2a3cc1cd0a654rjung * acknowledgment:
e609c337f729875bc20e01096c7e610f45356f54nilgun * "This product includes software developed by the Apache Group
35ac4e1a6ec5aa3cfa1a34d8f20fe8a841cc46b7rbowen * for use in the Apache HTTP server project (http://www.apache.org/)."
e609c337f729875bc20e01096c7e610f45356f54nilgun *
e609c337f729875bc20e01096c7e610f45356f54nilgun * THIS SOFTWARE IS PROVIDED BY THE APACHE GROUP ``AS IS'' AND ANY
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE APACHE GROUP OR
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * OF THE POSSIBILITY OF SUCH DAMAGE.
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * ====================================================================
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun *
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * This software consists of voluntary contributions made by many
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * individuals on behalf of the Apache Group and was originally based
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * on public domain software written at the National Center for
e609c337f729875bc20e01096c7e610f45356f54nilgun * Supercomputing Applications, University of Illinois, Urbana-Champaign.
e609c337f729875bc20e01096c7e610f45356f54nilgun * For more information on the Apache Group and the Apache HTTP server
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * project, please see <http://www.apache.org/>.
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun *
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun */
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun
30471a4650391f57975f60bbb6e4a90be7b284bfhumbedooh#define CORE_PRIVATE
e609c337f729875bc20e01096c7e610f45356f54nilgun
e609c337f729875bc20e01096c7e610f45356f54nilgun#include "ap_config.h"
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#include "apr_portable.h"
e609c337f729875bc20e01096c7e610f45356f54nilgun#include "httpd.h"
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#include "http_main.h"
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#include "http_log.h"
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#include "http_config.h" /* for read_config */
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#include "http_core.h" /* for get_remote_host */
e609c337f729875bc20e01096c7e610f45356f54nilgun#include "http_connection.h"
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#include "ap_mpm.h"
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#include "unixd.h"
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#include "iol_socket.h"
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#include "ap_listen.h"
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#include "mpm_default.h"
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#include "dexter.h"
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#include "scoreboard.h"
e609c337f729875bc20e01096c7e610f45356f54nilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#include <poll.h>
e609c337f729875bc20e01096c7e610f45356f54nilgun#include <netinet/tcp.h>
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#include <pthread.h>
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#include <signal.h>
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun
e609c337f729875bc20e01096c7e610f45356f54nilgun/*
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * Actual definitions of config globals
e609c337f729875bc20e01096c7e610f45356f54nilgun */
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic int threads_to_start = 0; /* Worker threads per child */
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic int min_spare_threads = 0;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic int max_spare_threads = 0;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic int max_threads = 0;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic int max_requests_per_child = 0;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic char *ap_pid_fname=NULL;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic int num_daemons=0;
e609c337f729875bc20e01096c7e610f45356f54nilgunstatic int workers_may_exit = 0;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic int requests_this_child;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic int num_listenfds = 0;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic struct pollfd *listenfds;
e609c337f729875bc20e01096c7e610f45356f54nilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun/* Table of child status */
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#define SERVER_DEAD 0
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#define SERVER_DYING 1
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#define SERVER_ALIVE 2
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic struct {
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun pid_t pid;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun unsigned char status;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun} child_table[HARD_SERVER_LIMIT];
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun/*
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * The max child slot ever assigned, preserved across restarts. Necessary
e609c337f729875bc20e01096c7e610f45356f54nilgun * to deal with NumServers changes across SIGWINCH restarts. We use this
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * value to optimize routines that have to scan the entire child table.
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun *
e609c337f729875bc20e01096c7e610f45356f54nilgun * XXX - It might not be worth keeping this code in. There aren't very
e609c337f729875bc20e01096c7e610f45356f54nilgun * many child processes in this MPM.
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun */
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunint max_daemons_limit = -1;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun
e609c337f729875bc20e01096c7e610f45356f54nilgunstatic char ap_coredump_dir[MAX_STRING_LEN];
e609c337f729875bc20e01096c7e610f45356f54nilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic int pipe_of_death[2];
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic pthread_mutex_t pipe_of_death_mutex;
e609c337f729875bc20e01096c7e610f45356f54nilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun/* *Non*-shared http_main globals... */
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun
e609c337f729875bc20e01096c7e610f45356f54nilgunstatic server_rec *server_conf;
e609c337f729875bc20e01096c7e610f45356f54nilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun/* one_process --- debugging mode variable; can be set from the command line
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * with the -X flag. If set, this gets you the child_main loop running
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * in the process which originally started up (no detach, no make_child),
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * which is a pretty nice debugging environment. (You'll get a SIGHUP
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * early in standalone_main; just continue through. This is the server
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * trying to kill off any child processes which it might have lying
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * around --- Apache doesn't keep track of their pids, it just sends
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * SIGHUP to the process group, ignoring it in the root process.
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * Continue through and you'll be fine.).
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun */
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic int one_process = 0;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#ifdef DEBUG_SIGSTOP
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunint raise_sigstop_flags;
e609c337f729875bc20e01096c7e610f45356f54nilgun#endif
e609c337f729875bc20e01096c7e610f45356f54nilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#ifdef HAS_OTHER_CHILD
e609c337f729875bc20e01096c7e610f45356f54nilgun/* used to maintain list of children which aren't part of the child table */
e05e9b1990428410383f42a260dfaf9b8067f24dnilguntypedef struct other_child_rec other_child_rec;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstruct other_child_rec {
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun other_child_rec *next;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun int pid;
e609c337f729875bc20e01096c7e610f45356f54nilgun void (*maintenance) (int, void *, ap_wait_t);
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun void *data;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun int write_fd;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun};
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic other_child_rec *other_children;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#endif
e609c337f729875bc20e01096c7e610f45356f54nilgun
e609c337f729875bc20e01096c7e610f45356f54nilgunstatic ap_context_t *pconf; /* Pool for config stuff */
e609c337f729875bc20e01096c7e610f45356f54nilgunstatic ap_context_t *pchild; /* Pool for httpd child stuff */
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic ap_context_t *thread_pool_parent; /* Parent of per-thread pools */
e609c337f729875bc20e01096c7e610f45356f54nilgunstatic pthread_mutex_t thread_pool_parent_mutex;
e609c337f729875bc20e01096c7e610f45356f54nilgun
e609c337f729875bc20e01096c7e610f45356f54nilgunstatic int child_num;
e609c337f729875bc20e01096c7e610f45356f54nilgunstatic int my_pid; /* Linux getpid() doesn't work except in main thread. Use
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun this instead */
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun/* Keep track of the number of worker threads currently active */
e609c337f729875bc20e01096c7e610f45356f54nilgunstatic int worker_thread_count;
e609c337f729875bc20e01096c7e610f45356f54nilgunstatic pthread_mutex_t worker_thread_count_mutex;
e609c337f729875bc20e01096c7e610f45356f54nilgunstatic int worker_thread_free_ids[HARD_THREAD_LIMIT];
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic pthread_attr_t worker_thread_attr;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun/* Keep track of the number of idle worker threads */
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic int idle_thread_count;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic pthread_mutex_t idle_thread_count_mutex;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun/* Locks for accept serialization */
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#ifdef NO_SERIALIZED_ACCEPT
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#define SAFE_ACCEPT(stmt) APR_SUCCESS
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#else
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#define SAFE_ACCEPT(stmt) (stmt)
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic ap_lock_t *process_accept_mutex;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#endif /* NO_SERIALIZED_ACCEPT */
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic char *lock_fname;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic pthread_mutex_t thread_accept_mutex = PTHREAD_MUTEX_INITIALIZER;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun/* Global, alas, so http_core can talk to us */
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunenum server_token_type ap_server_tokens = SrvTk_FULL;
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunAPI_EXPORT(const server_rec *) ap_get_server_conf(void)
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun{
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun return (server_conf);
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun}
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun/* a clean exit from a child with proper cleanup */
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunstatic void clean_child_exit(int code)
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun{
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun if (pchild) {
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun ap_destroy_pool(pchild);
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun }
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun exit(code);
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun}
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun/*****************************************************************
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun * dealing with other children
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun */
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun
e05e9b1990428410383f42a260dfaf9b8067f24dnilgun#ifdef HAS_OTHER_CHILD
e05e9b1990428410383f42a260dfaf9b8067f24dnilgunAPI_EXPORT(void) ap_register_other_child(int pid,
e609c337f729875bc20e01096c7e610f45356f54nilgun void (*maintenance) (int reason, void *, ap_wait_t status),
e609c337f729875bc20e01096c7e610f45356f54nilgun void *data, int write_fd)
e609c337f729875bc20e01096c7e610f45356f54nilgun{
27dcd8d81085fd60aadcd8a9bad35a607b26b758nilgun other_child_rec *ocr;
e609c337f729875bc20e01096c7e610f45356f54nilgun
e609c337f729875bc20e01096c7e610f45356f54nilgun ocr = ap_palloc(pconf, sizeof(*ocr));
f086b4b402fa9a2fefc7dda85de2a3cc1cd0a654rjung ocr->pid = pid;
727872d18412fc021f03969b8641810d8896820bhumbedooh ocr->maintenance = maintenance;
0d0ba3a410038e179b695446bb149cce6264e0abnd ocr->data = data;
727872d18412fc021f03969b8641810d8896820bhumbedooh ocr->write_fd = write_fd;
cc7e1025de9ac63bd4db6fe7f71c158b2cf09fe4humbedooh ocr->next = other_children;
0d0ba3a410038e179b695446bb149cce6264e0abnd other_children = ocr;
cc7e1025de9ac63bd4db6fe7f71c158b2cf09fe4humbedooh}
727872d18412fc021f03969b8641810d8896820bhumbedooh
0d0ba3a410038e179b695446bb149cce6264e0abnd/* note that since this can be called by a maintenance function while we're
0d0ba3a410038e179b695446bb149cce6264e0abnd * scanning the other_children list, all scanners should protect themself
0d0ba3a410038e179b695446bb149cce6264e0abnd * by loading ocr->next before calling any maintenance function.
ac082aefa89416cbdc9a1836eaf3bed9698201c8humbedooh */
0d0ba3a410038e179b695446bb149cce6264e0abndAPI_EXPORT(void) ap_unregister_other_child(void *data)
0d0ba3a410038e179b695446bb149cce6264e0abnd{
0d0ba3a410038e179b695446bb149cce6264e0abnd other_child_rec **pocr, *nocr;
727872d18412fc021f03969b8641810d8896820bhumbedooh
0d0ba3a410038e179b695446bb149cce6264e0abnd for (pocr = &other_children; *pocr; pocr = &(*pocr)->next) {
0d0ba3a410038e179b695446bb149cce6264e0abnd if ((*pocr)->data == data) {
30471a4650391f57975f60bbb6e4a90be7b284bfhumbedooh nocr = (*pocr)->next;
205f749042ed530040a4f0080dbcb47ceae8a374rjung (*(*pocr)->maintenance) (OC_REASON_UNREGISTER, (*pocr)->data, -1);
af33a4994ae2ff15bc67d19ff1a7feb906745bf8rbowen *pocr = nocr;
0d0ba3a410038e179b695446bb149cce6264e0abnd /* XXX: um, well we've just wasted some space in pconf ? */
7fec19672a491661b2fe4b29f685bc7f4efa64d4nd return;
7fec19672a491661b2fe4b29f685bc7f4efa64d4nd }
7fec19672a491661b2fe4b29f685bc7f4efa64d4nd }
e609c337f729875bc20e01096c7e610f45356f54nilgun}
/* test to ensure that the write_fds are all still writable, otherwise
* invoke the maintenance functions as appropriate */
static void probe_writable_fds(void)
{
return;
#if 0
fd_set writable_fds;
int fd_max;
other_child_rec *ocr, *nocr;
struct timeval tv;
int rc;
if (other_children == NULL)
return;
fd_max = 0;
FD_ZERO(&writable_fds);
do {
for (ocr = other_children; ocr; ocr = ocr->next) {
if (ocr->write_fd == -1)
continue;
FD_SET(ocr->write_fd, &writable_fds);
if (ocr->write_fd > fd_max) {
fd_max = ocr->write_fd;
}
}
if (fd_max == 0)
return;
tv.tv_sec = 0;
tv.tv_usec = 0;
rc = ap_select(fd_max + 1, NULL, &writable_fds, NULL, &tv);
} while (rc == -1 && errno == EINTR);
if (rc == -1) {
/* XXX: uhh this could be really bad, we could have a bad file
* descriptor due to a bug in one of the maintenance routines */
ap_log_unixerr("probe_writable_fds", "select",
"could not probe writable fds", server_conf);
return;
}
if (rc == 0)
return;
for (ocr = other_children; ocr; ocr = nocr) {
nocr = ocr->next;
if (ocr->write_fd == -1)
continue;
if (FD_ISSET(ocr->write_fd, &writable_fds))
continue;
(*ocr->maintenance) (OC_REASON_UNWRITABLE, ocr->data, -1);
}
#endif
}
/* possibly reap an other_child, return 0 if yes, -1 if not */
static int reap_other_child(int pid, ap_wait_t status)
{
other_child_rec *ocr, *nocr;
for (ocr = other_children; ocr; ocr = nocr) {
nocr = ocr->next;
if (ocr->pid != pid)
continue;
ocr->pid = -1;
(*ocr->maintenance) (OC_REASON_DEATH, ocr->data, status);
return 0;
}
return -1;
}
#endif
static void reclaim_child_processes(int terminate)
{
int i, status;
long int waittime = 1024 * 16; /* in usecs */
struct timeval tv;
int waitret, tries;
int not_dead_yet;
#ifdef HAS_OTHER_CHILD
other_child_rec *ocr, *nocr;
#endif
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.
*/
tv.tv_sec = waittime / 1000000;
tv.tv_usec = waittime % 1000000;
waittime = waittime * 4;
ap_select(0, NULL, NULL, NULL, &tv);
/* now see who is done */
not_dead_yet = 0;
for (i = 0; i < max_daemons_limit; ++i) {
int pid;
if (child_table[i].status == SERVER_DEAD)
continue;
pid = child_table[i].pid;
waitret = waitpid(pid, &status, WNOHANG);
if (waitret == pid || waitret == -1) {
child_table[i].status = SERVER_DEAD;
continue;
}
++not_dead_yet;
switch (tries) {
case 1: /* 16ms */
case 2: /* 82ms */
break;
case 3: /* 344ms */
case 4: /* 16ms */
case 5: /* 82ms */
case 6: /* 344ms */
case 7: /* 1.4sec */
/* ok, now it's being annoying */
ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_WARNING, errno,
server_conf,
"child process %d still did not exit, sending a SIGTERM",
pid);
kill(pid, SIGTERM);
break;
case 8: /* 6 sec */
/* die child scum */
ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_ERR, errno,
server_conf,
"child process %d still did not exit, sending a SIGKILL",
pid);
kill(pid, SIGKILL);
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_NOERRNO|APLOG_ERR, errno,
server_conf,
"could not make child process %d exit, "
"attempting to continue anyway", pid);
break;
}
}
#ifdef HAS_OTHER_CHILD
for (ocr = other_children; ocr; ocr = nocr) {
nocr = ocr->next;
if (ocr->pid == -1)
continue;
waitret = waitpid(ocr->pid, &status, WNOHANG);
if (waitret == ocr->pid) {
ocr->pid = -1;
(*ocr->maintenance) (OC_REASON_DEATH, ocr->data, status);
}
else if (waitret == 0) {
(*ocr->maintenance) (OC_REASON_RESTART, ocr->data, -1);
++not_dead_yet;
}
else if (waitret == -1) {
/* uh what the heck? they didn't call unregister? */
ocr->pid = -1;
(*ocr->maintenance) (OC_REASON_LOST, ocr->data, -1);
}
}
#endif
if (!not_dead_yet) {
/* nothing left to wait for */
break;
}
}
}
/* Finally, this routine is used by the caretaker process to wait for
* a while...
*/
/* 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;
static int wait_or_timeout(ap_wait_t *status)
{
struct timeval tv;
int ret;
++wait_or_timeout_counter;
if (wait_or_timeout_counter == INTERVAL_OF_WRITABLE_PROBES) {
wait_or_timeout_counter = 0;
#ifdef HAS_OTHER_CHILD
probe_writable_fds();
#endif
}
ret = waitpid(-1, status, WNOHANG);
if (ret == -1 && errno == EINTR) {
return -1;
}
if (ret > 0) {
return ret;
}
tv.tv_sec = SCOREBOARD_MAINTENANCE_INTERVAL / 1000000;
tv.tv_usec = SCOREBOARD_MAINTENANCE_INTERVAL % 1000000;
ap_select(0, NULL, NULL, NULL, &tv);
return -1;
}
/* handle all varieties of core dumping signals */
static void sig_coredump(int sig)
{
chdir(ap_coredump_dir);
signal(sig, SIG_DFL);
kill(getpid(), sig);
/* At this point we've got sig blocked, because we're still inside
* the signal handler. When we leave the signal handler it will
* be unblocked, and we'll take the signal... and coredump or whatever
* is appropriate for this particular Unix. In addition the parent
* will see the real signal we received -- whereas if we called
* abort() here, the parent would only see SIGABRT.
*/
}
static void just_die(int sig)
{
clean_child_exit(0);
}
/*****************************************************************
* Connection structures and accounting...
*/
/* volatile just in case */
static int volatile shutdown_pending;
static int volatile restart_pending;
static int volatile is_graceful;
/*
* ap_start_shutdown() and ap_start_restart(), below, are a first stab at
* functions to initiate shutdown or restart without relying on signals.
* Previously this was initiated in sig_term() and restart() signal handlers,
* but we want to be able to start a shutdown/restart from other sources --
* e.g. on Win32, from the service manager. Now the service manager can
* call ap_start_shutdown() or ap_start_restart() as appropiate. Note that
* these functions can also be called by the child processes, since global
* variables are no longer used to pass on the required action to the parent.
*
* These should only be called from the parent process itself, since the
* parent process will use the shutdown_pending and restart_pending variables
* to determine whether to shutdown or restart. The child process should
* call signal_parent() directly to tell the parent to die -- this will
* cause neither of those variable to be set, which the parent will
* assume means something serious is wrong (which it will be, for the
* child to force an exit) and so do an exit anyway.
*/
void ap_start_shutdown(void)
{
if (shutdown_pending == 1) {
/* Um, is this _probably_ not an error, if the user has
* tried to do a shutdown twice quickly, so we won't
* worry about reporting it.
*/
return;
}
shutdown_pending = 1;
}
/* do a graceful restart if graceful == 1 */
void ap_start_restart(int graceful)
{
if (restart_pending == 1) {
/* Probably not an error - don't bother reporting it */
return;
}
restart_pending = 1;
is_graceful = graceful;
}
static void sig_term(int sig)
{
ap_start_shutdown();
}
static void restart(int sig)
{
#ifndef WIN32
ap_start_restart(sig == SIGWINCH);
#else
ap_start_restart(1);
#endif
}
static void set_signals(void)
{
#ifndef NO_USE_SIGACTION
struct sigaction sa;
sigemptyset(&sa.sa_mask);
sa.sa_flags = 0;
if (!one_process) {
sa.sa_handler = sig_coredump;
#if defined(SA_ONESHOT)
sa.sa_flags = SA_ONESHOT;
#elif defined(SA_RESETHAND)
sa.sa_flags = SA_RESETHAND;
#endif
if (sigaction(SIGSEGV, &sa, NULL) < 0)
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, server_conf, "sigaction(SIGSEGV)");
#ifdef SIGBUS
if (sigaction(SIGBUS, &sa, NULL) < 0)
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, server_conf, "sigaction(SIGBUS)");
#endif
#ifdef SIGABORT
if (sigaction(SIGABORT, &sa, NULL) < 0)
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, server_conf, "sigaction(SIGABORT)");
#endif
#ifdef SIGABRT
if (sigaction(SIGABRT, &sa, NULL) < 0)
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, server_conf, "sigaction(SIGABRT)");
#endif
#ifdef SIGILL
if (sigaction(SIGILL, &sa, NULL) < 0)
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, server_conf, "sigaction(SIGILL)");
#endif
sa.sa_flags = 0;
}
sa.sa_handler = sig_term;
if (sigaction(SIGTERM, &sa, NULL) < 0)
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, server_conf, "sigaction(SIGTERM)");
#ifdef SIGINT
if (sigaction(SIGINT, &sa, NULL) < 0)
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, server_conf, "sigaction(SIGINT)");
#endif
#ifdef SIGXCPU
sa.sa_handler = SIG_DFL;
if (sigaction(SIGXCPU, &sa, NULL) < 0)
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, server_conf, "sigaction(SIGXCPU)");
#endif
#ifdef SIGXFSZ
sa.sa_handler = SIG_DFL;
if (sigaction(SIGXFSZ, &sa, NULL) < 0)
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, server_conf, "sigaction(SIGXFSZ)");
#endif
#ifdef SIGPIPE
sa.sa_handler = SIG_IGN;
if (sigaction(SIGPIPE, &sa, NULL) < 0)
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, server_conf, "sigaction(SIGPIPE)");
#endif
/* we want to ignore HUPs and WINCH while we're busy processing one */
sigaddset(&sa.sa_mask, SIGHUP);
sigaddset(&sa.sa_mask, SIGWINCH);
sa.sa_handler = restart;
if (sigaction(SIGHUP, &sa, NULL) < 0)
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, server_conf, "sigaction(SIGHUP)");
if (sigaction(SIGWINCH, &sa, NULL) < 0)
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, server_conf, "sigaction(SIGWINCH)");
#else
if (!one_process) {
signal(SIGSEGV, sig_coredump);
#ifdef SIGBUS
signal(SIGBUS, sig_coredump);
#endif /* SIGBUS */
#ifdef SIGABORT
signal(SIGABORT, sig_coredump);
#endif /* SIGABORT */
#ifdef SIGABRT
signal(SIGABRT, sig_coredump);
#endif /* SIGABRT */
#ifdef SIGILL
signal(SIGILL, sig_coredump);
#endif /* SIGILL */
#ifdef SIGXCPU
signal(SIGXCPU, SIG_DFL);
#endif /* SIGXCPU */
#ifdef SIGXFSZ
signal(SIGXFSZ, SIG_DFL);
#endif /* SIGXFSZ */
}
signal(SIGTERM, sig_term);
#ifdef SIGHUP
signal(SIGHUP, restart);
#endif /* SIGHUP */
#ifdef SIGWINCH
signal(SIGWINCH, restart);
#endif /* SIGWINCH */
#ifdef SIGPIPE
signal(SIGPIPE, SIG_IGN);
#endif /* SIGPIPE */
#endif
}
static void process_child_status(int pid, ap_wait_t status)
{
/* Child died... if it died due to a fatal error,
* we should simply bail out.
*/
if ((WIFEXITED(status)) &&
WEXITSTATUS(status) == APEXIT_CHILDFATAL) {
ap_log_error(APLOG_MARK, APLOG_ALERT|APLOG_NOERRNO, errno, server_conf,
"Child %d returned a Fatal error... \n"
"Apache is exiting!",
pid);
exit(APEXIT_CHILDFATAL);
}
if (WIFSIGNALED(status)) {
switch (WTERMSIG(status)) {
case SIGTERM:
case SIGHUP:
case SIGUSR1:
case SIGKILL:
break;
default:
#ifdef SYS_SIGLIST
#ifdef WCOREDUMP
if (WCOREDUMP(status)) {
ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_NOTICE, errno,
server_conf,
"child pid %d exit signal %s (%d), "
"possible coredump in %s",
pid, (WTERMSIG(status) >= NumSIG) ? "" :
SYS_SIGLIST[WTERMSIG(status)], WTERMSIG(status),
ap_coredump_dir);
}
else {
#endif
ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_NOTICE, errno,
server_conf,
"child pid %d exit signal %s (%d)", pid,
SYS_SIGLIST[WTERMSIG(status)], WTERMSIG(status));
#ifdef WCOREDUMP
}
#endif
#else
ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_NOTICE, errno,
server_conf,
"child pid %d exit signal %d",
pid, WTERMSIG(status));
#endif
}
}
}
static int setup_listeners(server_rec *s)
{
ap_listen_rec *lr;
int num_listeners = 0;
if (ap_listen_open(s->process, s->port)) {
return 0;
}
for (lr = ap_listeners; lr; lr = lr->next) {
num_listeners++;
}
return num_listeners;
}
/*****************************************************************
* Here follows a long bunch of generic server bookkeeping stuff...
*/
#if defined(TCP_NODELAY) && !defined(MPE) && !defined(TPF)
static void sock_disable_nagle(int 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.
*/
int just_say_no = 1;
if (setsockopt(s, IPPROTO_TCP, TCP_NODELAY, (char *) &just_say_no,
sizeof(int)) < 0) {
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, server_conf,
"setsockopt: (TCP_NODELAY)");
}
}
#else
#define sock_disable_nagle(s) /* NOOP */
#endif
int ap_graceful_stop_signalled(void)
{
/* XXX - Does this really work? - Manoj */
return is_graceful;
}
/*****************************************************************
* Child process main loop.
*/
static void process_socket(ap_context_t *p, ap_socket_t *sock, long conn_id)
{
BUFF *conn_io;
conn_rec *current_conn;
ap_iol *iol;
int csd;
ap_status_t rv;
if ((rv = ap_get_os_sock(&csd, sock)) != APR_SUCCESS) {
ap_log_error(APLOG_MARK, APLOG_ERR, rv, NULL, "ap_get_os_sock");
}
sock_disable_nagle(csd);
iol = unix_attach_socket(sock);
if (iol == NULL) {
if (errno == EBADF) {
ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_WARNING, errno, NULL,
"filedescriptor (%u) larger than FD_SETSIZE (%u) "
"found, you probably need to rebuild Apache with a "
"larger FD_SETSIZE", csd, FD_SETSIZE);
}
else {
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, NULL,
"error attaching to socket");
}
ap_close_socket(sock);
return;
}
conn_io = ap_bcreate(p, B_RDWR);
ap_bpush_iol(conn_io, iol);
current_conn = ap_new_apr_connection(p, server_conf, conn_io, sock,
conn_id);
ap_process_connection(current_conn);
}
static void *worker_thread(void *);
/* Starts a thread as long as we're below max_threads */
static int start_thread(void)
{
pthread_t thread;
pthread_mutex_lock(&worker_thread_count_mutex);
if (worker_thread_count < max_threads) {
if (pthread_create(&thread, &worker_thread_attr, worker_thread,
&worker_thread_free_ids[worker_thread_count])) {
ap_log_error(APLOG_MARK, APLOG_ALERT, errno, server_conf,
"pthread_create: unable to create worker thread");
/* 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 if we exit. */
sleep(10);
workers_may_exit = 1;
pthread_mutex_unlock(&worker_thread_count_mutex);
return 0;
}
else {
worker_thread_count++;
}
}
else {
static int reported = 0;
if (!reported) {
ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_ERR, errno, server_conf,
"server reached MaxThreadsPerChild setting, consider raising the"
" MaxThreadsPerChild or NumServers settings");
reported = 1;
}
pthread_mutex_unlock(&worker_thread_count_mutex);
return 0;
}
pthread_mutex_unlock(&worker_thread_count_mutex);
return 1;
}
/* Sets workers_may_exit if we received a character on the pipe_of_death */
static void check_pipe_of_death(void)
{
pthread_mutex_lock(&pipe_of_death_mutex);
if (!workers_may_exit) {
int ret;
char pipe_read_char;
ret = read(listenfds[0].fd, &pipe_read_char, 1);
if (ret == -1 && errno == EAGAIN) {
/* It lost the lottery. It must continue to suffer
* through a life of servitude. */
}
else {
/* It won the lottery (or something else is very
* wrong). Embrace death with open arms. */
workers_may_exit = 1;
}
}
pthread_mutex_unlock(&pipe_of_death_mutex);
}
/* idle_thread_count should be incremented before starting a worker_thread */
static void *worker_thread(void *arg)
{
ap_socket_t *csd = NULL;
ap_context_t *tpool; /* Pool for this thread */
ap_context_t *ptrans; /* Pool for per-transaction stuff */
ap_socket_t *sd = NULL;
int srv;
int curr_pollfd, last_pollfd = 0;
int thread_just_started = 1;
int thread_num = *((int *) arg);
long conn_id = child_num * HARD_THREAD_LIMIT + thread_num;
ap_status_t rv;
pthread_mutex_lock(&thread_pool_parent_mutex);
ap_create_context(&tpool, thread_pool_parent);
pthread_mutex_unlock(&thread_pool_parent_mutex);
ap_create_context(&ptrans, tpool);
while (!workers_may_exit) {
workers_may_exit |= (max_requests_per_child != 0) && (requests_this_child <= 0);
if (workers_may_exit) break;
if (!thread_just_started) {
pthread_mutex_lock(&idle_thread_count_mutex);
if (idle_thread_count < max_spare_threads) {
idle_thread_count++;
pthread_mutex_unlock(&idle_thread_count_mutex);
}
else {
pthread_mutex_unlock(&idle_thread_count_mutex);
break;
}
}
else {
thread_just_started = 0;
}
pthread_mutex_lock(&thread_accept_mutex);
if (workers_may_exit) {
pthread_mutex_unlock(&thread_accept_mutex);
break;
}
if ((rv = SAFE_ACCEPT(ap_lock(process_accept_mutex)))
!= APR_SUCCESS) {
ap_log_error(APLOG_MARK, APLOG_EMERG, rv, server_conf,
"ap_lock failed. Attempting to shutdown "
"process gracefully.");
workers_may_exit = 1;
}
while (!workers_may_exit) {
srv = poll(listenfds, num_listenfds + 1, -1);
if (srv < 0) {
if (errno == EINTR) {
continue;
}
/* poll() will only return errors in catastrophic
* circumstances. Let's try exiting gracefully, for now. */
ap_log_error(APLOG_MARK, APLOG_ERR, errno, (const server_rec *)
ap_get_server_conf(), "poll: (listen)");
workers_may_exit = 1;
}
if (workers_may_exit) break;
if (listenfds[0].revents & POLLIN) {
/* A process got a signal on the shutdown pipe. Check if we're
* the lucky process to die. */
check_pipe_of_death();
continue;
}
if (num_listenfds == 1) {
sd = ap_listeners->sd;
goto got_fd;
}
else {
/* find a listener */
curr_pollfd = last_pollfd;
do {
curr_pollfd++;
if (curr_pollfd > num_listenfds) {
curr_pollfd = 1;
}
/* XXX: Should we check for POLLERR? */
if (listenfds[curr_pollfd].revents & POLLIN) {
last_pollfd = curr_pollfd;
sd = NULL;
ap_put_os_sock(&sd, &listenfds[curr_pollfd].fd, ptrans);
goto got_fd;
}
} while (curr_pollfd != last_pollfd);
}
}
got_fd:
if (!workers_may_exit) {
if ((rv = ap_accept(&csd, sd, ptrans)) != APR_SUCCESS) {
ap_log_error(APLOG_MARK, APLOG_ERR, rv, NULL, "ap_accept");
}
if ((rv = SAFE_ACCEPT(ap_unlock(process_accept_mutex)))
!= APR_SUCCESS) {
ap_log_error(APLOG_MARK, APLOG_EMERG, rv, server_conf,
"ap_unlock failed. Attempting to shutdown "
"process gracefully.");
workers_may_exit = 1;
}
pthread_mutex_unlock(&thread_accept_mutex);
pthread_mutex_lock(&idle_thread_count_mutex);
if (idle_thread_count > min_spare_threads) {
idle_thread_count--;
}
else {
if (!start_thread()) {
idle_thread_count--;
}
}
pthread_mutex_unlock(&idle_thread_count_mutex);
process_socket(ptrans, csd, conn_id);
requests_this_child--;
} else {
if ((rv = SAFE_ACCEPT(ap_unlock(process_accept_mutex)))
!= APR_SUCCESS) {
ap_log_error(APLOG_MARK, APLOG_EMERG, rv, server_conf,
"ap_unlock failed. Attempting to shutdown "
"process gracefully.");
workers_may_exit = 1;
}
pthread_mutex_unlock(&thread_accept_mutex);
pthread_mutex_lock(&idle_thread_count_mutex);
idle_thread_count--;
pthread_mutex_unlock(&idle_thread_count_mutex);
break;
}
ap_clear_pool(ptrans);
}
pthread_mutex_lock(&thread_pool_parent_mutex);
ap_destroy_context(tpool);
pthread_mutex_unlock(&thread_pool_parent_mutex);
pthread_mutex_lock(&worker_thread_count_mutex);
worker_thread_count--;
worker_thread_free_ids[worker_thread_count] = thread_num;
if (worker_thread_count == 0) {
/* All the threads have exited, now finish the shutdown process
* by signalling the sigwait thread */
kill(my_pid, SIGTERM);
}
pthread_mutex_unlock(&worker_thread_count_mutex);
return NULL;
}
static void child_main(int child_num_arg)
{
sigset_t sig_mask;
int signal_received;
int i;
ap_listen_rec *lr;
ap_status_t rv;
my_pid = getpid();
child_num = child_num_arg;
ap_create_context(&pchild, pconf);
/*stuff to do before we switch id's, so we have permissions.*/
rv = SAFE_ACCEPT(ap_child_init_lock(&process_accept_mutex, lock_fname,
pchild));
if (rv != APR_SUCCESS) {
ap_log_error(APLOG_MARK, APLOG_EMERG, rv, 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_child_init_hook(pchild, server_conf);
/*done with init critical section */
/* All threads should mask signals out, accoring to sigwait(2) man page */
sigfillset(&sig_mask);
if (pthread_sigmask(SIG_SETMASK, &sig_mask, NULL) != 0) {
ap_log_error(APLOG_MARK, APLOG_ALERT, errno, server_conf,
"pthread_sigmask");
}
requests_this_child = max_requests_per_child;
/* Set up the pollfd array */
listenfds = ap_palloc(pchild, sizeof(struct pollfd) * (num_listenfds + 1));
listenfds[0].fd = pipe_of_death[0];
listenfds[0].events = POLLIN;
listenfds[0].revents = 0;
for (lr = ap_listeners, i = 1; i <= num_listenfds; lr = lr->next, ++i) {
ap_get_os_sock(&listenfds[i].fd, lr->sd);
listenfds[i].events = POLLIN; /* should we add POLLPRI ?*/
listenfds[i].revents = 0;
}
/* Setup worker threads */
if (threads_to_start > max_threads) {
threads_to_start = max_threads;
}
idle_thread_count = threads_to_start;
worker_thread_count = 0;
for (i = 0; i < max_threads; i++) {
worker_thread_free_ids[i] = i;
}
ap_create_context(&thread_pool_parent, pchild);
pthread_mutex_init(&thread_pool_parent_mutex, NULL);
pthread_mutex_init(&idle_thread_count_mutex, NULL);
pthread_mutex_init(&worker_thread_count_mutex, NULL);
pthread_mutex_init(&pipe_of_death_mutex, NULL);
pthread_attr_init(&worker_thread_attr);
pthread_attr_setdetachstate(&worker_thread_attr, PTHREAD_CREATE_DETACHED);
/* We are creating worker threads right now */
for (i=0; i < threads_to_start; i++) {
/* start_thread shouldn't fail here */
if (!start_thread()) {
break;
}
}
/* This thread will be the one responsible for handling signals */
sigemptyset(&sig_mask);
sigaddset(&sig_mask, SIGTERM);
sigaddset(&sig_mask, SIGINT);
sigwait(&sig_mask, &signal_received);
switch (signal_received) {
case SIGTERM:
case SIGINT:
just_die(signal_received);
break;
default:
ap_log_error(APLOG_MARK, APLOG_ALERT, errno, server_conf,
"received impossible signal: %d", signal_received);
just_die(SIGTERM);
}
}
static int make_child(server_rec *s, int slot, time_t now)
{
int pid;
if (slot + 1 > max_daemons_limit) {
max_daemons_limit = slot + 1;
}
if (one_process) {
set_signals();
child_table[slot].pid = getpid();
child_table[slot].status = SERVER_ALIVE;
child_main(slot);
}
if ((pid = fork()) == -1) {
ap_log_error(APLOG_MARK, APLOG_ERR, errno, s,
"fork: Unable to fork new process");
/* 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 AIX_BIND_PROCESSOR
/* By default, AIX binds to a single processor. This bit unbinds
children which will then bind to another CPU.
*/
#include <sys/processor.h>
int status = bindprocessor(BINDPROCESS, (int)getpid(),
PROCESSOR_CLASS_ANY);
if (status != OK)
ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_WARNING, errno,
server_conf, "processor unbind failed %d", status);
#endif
RAISE_SIGSTOP(MAKE_CHILD);
/* XXX - For an unthreaded server, a signal handler will be necessary
signal(SIGTERM, just_die);
*/
child_main(slot);
return 0;
}
/* else */
child_table[slot].pid = pid;
child_table[slot].status = SERVER_ALIVE;
return 0;
}
/* start up a bunch of children */
static int startup_children(int number_to_start)
{
int i;
for (i = 0; number_to_start && i < num_daemons; ++i) {
if (child_table[i].status != SERVER_DEAD) {
continue;
}
if (make_child(server_conf, i, 0) < 0) {
break;
}
--number_to_start;
}
return number_to_start;
}
/*
* spawn_rate is the number of children that will be spawned on the
* next maintenance cycle if there aren't enough servers. It is
* doubled up to MAX_SPAWN_RATE, and reset only when a cycle goes by
* without the need to spawn.
*/
static int spawn_rate = 1;
#ifndef MAX_SPAWN_RATE
#define MAX_SPAWN_RATE (32)
#endif
static int hold_off_on_exponential_spawning;
static void perform_child_maintenance(void)
{
int i;
time_t now = 0;
int free_length;
int free_slots[MAX_SPAWN_RATE];
int last_non_dead = -1;
/* initialize the free_list */
free_length = 0;
for (i = 0; i < num_daemons; ++i) {
if (child_table[i].status == SERVER_DEAD) {
if (free_length < spawn_rate) {
free_slots[free_length] = i;
++free_length;
}
}
else {
last_non_dead = i;
}
if (i >= max_daemons_limit && free_length >= spawn_rate) {
break;
}
}
max_daemons_limit = last_non_dead + 1;
if (free_length > 0) {
for (i = 0; i < free_length; ++i) {
make_child(server_conf, free_slots[i], now);
}
/* 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 (spawn_rate < MAX_SPAWN_RATE) {
spawn_rate *= 2;
}
}
else {
spawn_rate = 1;
}
}
static void server_main_loop(int remaining_children_to_start)
{
int child_slot;
ap_wait_t status;
int pid;
int i;
while (!restart_pending && !shutdown_pending) {
pid = wait_or_timeout(&status);
if (pid >= 0) {
process_child_status(pid, status);
/* non-fatal death... note that it's gone in the child table and
* clean out the status table. */
child_slot = -1;
for (i = 0; i < max_daemons_limit; ++i) {
if (child_table[i].pid == pid) {
int j;
child_slot = i;
for (j = 0; j < HARD_THREAD_LIMIT; j++) {
ap_dexter_force_reset_connection_status(i * HARD_THREAD_LIMIT + j);
}
break;
}
}
if (child_slot >= 0) {
child_table[child_slot].status = SERVER_DEAD;
if (remaining_children_to_start
&& child_slot < num_daemons) {
/* we're still doing a 1-for-1 replacement of dead
* children with new children
*/
make_child(server_conf, child_slot, time(NULL));
--remaining_children_to_start;
}
#ifdef HAS_OTHER_CHILD
}
else if (reap_other_child(pid, status) == 0) {
/* handled */
#endif
}
else if (is_graceful) {
/* Great, we've probably just lost a slot in the
* child table. Somehow we don't know about this
* child.
*/
ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_WARNING, errno,
server_conf,
"long lost child came home! (pid %d)", pid);
}
/* Don't perform idle maintenance when a child dies,
* only do it when there's a timeout. Remember only a
* finite number of children can die, and it's pretty
* pathological for a lot to die suddenly.
*/
continue;
}
else if (remaining_children_to_start) {
/* we hit a 1 second timeout in which none of the previous
* generation of children needed to be reaped... so assume
* they're all done, and pick up the slack if any is left.
*/
remaining_children_to_start = \
startup_children(remaining_children_to_start);
/* In any event we really shouldn't do the code below because
* few of the servers we just started are in the IDLE state
* yet, so we'd mistakenly create an extra server.
*/
continue;
}
perform_child_maintenance();
}
}
static ap_status_t cleanup_fd(void *fdptr)
{
if (close(*((int *) fdptr)) < 0) {
return APR_EBADF;
}
return APR_SUCCESS;
}
int ap_mpm_run(ap_context_t *_pconf, ap_context_t *plog, server_rec *s)
{
int remaining_children_to_start;
int i;
ap_status_t rv;
pconf = _pconf;
server_conf = s;
if (pipe(pipe_of_death) == -1) {
ap_log_error(APLOG_MARK, APLOG_ERR, errno,
(const server_rec*) server_conf,
"pipe: (pipe_of_death)");
exit(1);
}
ap_register_cleanup(pconf, &pipe_of_death[0], cleanup_fd, cleanup_fd);
ap_register_cleanup(pconf, &pipe_of_death[1], cleanup_fd, cleanup_fd);
if (fcntl(pipe_of_death[0], F_SETFD, O_NONBLOCK) == -1) {
ap_log_error(APLOG_MARK, APLOG_ERR, errno,
(const server_rec*) server_conf,
"fcntl: O_NONBLOCKing (pipe_of_death)");
exit(1);
}
server_conf = s;
if ((num_listenfds = setup_listeners(server_conf)) < 1) {
/* XXX: hey, what's the right way for the mpm to indicate a fatal error? */
ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_ALERT, errno, s,
"no listening sockets available, shutting down");
return 1;
}
ap_log_pid(pconf, ap_pid_fname);
/* Initialize cross-process accept lock */
lock_fname = ap_psprintf(_pconf, "%s.%lu",
ap_server_root_relative(_pconf, lock_fname),
my_pid);
rv = SAFE_ACCEPT(ap_create_lock(&process_accept_mutex, APR_MUTEX,
APR_CROSS_PROCESS, lock_fname, _pconf));
if (rv != APR_SUCCESS) {
ap_log_error(APLOG_MARK, APLOG_EMERG, rv, s,
"Couldn't create cross-process lock");
return 1;
}
if (!is_graceful) {
reinit_scoreboard(pconf);
}
/* Initialize the child table */
if (!is_graceful) {
for (i = 0; i < HARD_SERVER_LIMIT; i++) {
child_table[i].status = SERVER_DEAD;
}
}
set_signals();
/* If we're doing a graceful_restart then we're going to see a lot
* of children exiting immediately when we get into the main loop
* below (because we just sent them SIGWINCH). This happens pretty
* rapidly... and for each one that exits we'll start a new one until
* we reach at least daemons_min_free. But we may be permitted to
* start more than that, so we'll just keep track of how many we're
* supposed to start up without the 1 second penalty between each fork.
*/
remaining_children_to_start = num_daemons;
if (!is_graceful) {
remaining_children_to_start = \
startup_children(remaining_children_to_start);
}
else {
/* give the system some time to recover before kicking into
* exponential mode */
hold_off_on_exponential_spawning = 10;
}
ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_NOTICE, errno, server_conf,
"%s configured -- resuming normal operations",
ap_get_server_version());
ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_INFO, errno, server_conf,
"Server built: %s", ap_get_server_built());
restart_pending = shutdown_pending = 0;
server_main_loop(remaining_children_to_start);
if (shutdown_pending) {
/* Time to gracefully shut down:
* Kill child processes, tell them to call child_exit, etc...
*/
if (unixd_killpg(getpgrp(), SIGTERM) < 0) {
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, server_conf,
"killpg SIGTERM");
}
reclaim_child_processes(1); /* Start with SIGTERM */
/* cleanup pid file on normal shutdown */
{
const char *pidfile = NULL;
pidfile = ap_server_root_relative (pconf, ap_pid_fname);
if ( pidfile != NULL && unlink(pidfile) == 0)
ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_INFO, errno,
server_conf,
"removed PID file %s (pid=%ld)",
pidfile, (long)getpid());
}
ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_NOTICE, errno,
server_conf, "caught SIGTERM, shutting down");
return 1;
}
/* we've been told to restart */
signal(SIGHUP, SIG_IGN);
if (one_process) {
/* not worth thinking about */
return 1;
}
if (is_graceful) {
char char_of_death = '!';
ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_NOTICE, errno, server_conf,
"SIGWINCH received. Doing graceful restart");
/* This is mostly for debugging... so that we know what is still
* gracefully dealing with existing request.
*/
for (i = 0; i < num_daemons; ++i) {
if (child_table[i].status != SERVER_DEAD) {
child_table[i].status = SERVER_DYING;
}
}
/* give the children the signal to die */
for (i = 0; i < num_daemons;) {
if (write(pipe_of_death[1], &char_of_death, 1) == -1) {
if (errno == EINTR) continue;
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, server_conf,
"write pipe_of_death");
}
i++;
}
}
else {
/* Kill 'em all. Since the child acts the same on the parents SIGTERM
* and a SIGHUP, we may as well use the same signal, because some user
* pthreads are stealing signals from us left and right.
*/
if (unixd_killpg(getpgrp(), SIGTERM) < 0) {
ap_log_error(APLOG_MARK, APLOG_WARNING, errno, server_conf,
"killpg SIGTERM");
}
reclaim_child_processes(1); /* Start with SIGTERM */
ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_NOTICE, errno,
server_conf, "SIGHUP received. Attempting to restart");
}
return 0;
}
static void dexter_pre_config(ap_context_t *p, ap_context_t *plog, ap_context_t *ptemp)
{
static int restart_num = 0;
one_process = !!getenv("ONE_PROCESS");
/* sigh, want this only the second time around */
if (restart_num++ == 1) {
is_graceful = 0;
if (!one_process) {
unixd_detach();
}
my_pid = getpid();
}
unixd_pre_config();
ap_listen_pre_config();
num_daemons = DEFAULT_NUM_DAEMON;
threads_to_start = DEFAULT_START_THREAD;
min_spare_threads = DEFAULT_MIN_SPARE_THREAD;
max_spare_threads = DEFAULT_MAX_SPARE_THREAD;
max_threads = HARD_THREAD_LIMIT;
ap_pid_fname = DEFAULT_PIDLOG;
lock_fname = DEFAULT_LOCKFILE;
max_requests_per_child = DEFAULT_MAX_REQUESTS_PER_CHILD;
ap_dexter_set_maintain_connection_status(1);
ap_cpystrn(ap_coredump_dir, ap_server_root, sizeof(ap_coredump_dir));
}
static void dexter_hooks(void)
{
ap_hook_pre_config(dexter_pre_config, NULL, NULL, HOOK_MIDDLE);
INIT_SIGLIST()
one_process = 0;
}
static const char *set_pidfile(cmd_parms *cmd, void *dummy, char *arg)
{
const char *err = ap_check_cmd_context(cmd, GLOBAL_ONLY);
if (err != NULL) {
return err;
}
if (cmd->server->is_virtual) {
return "PidFile directive not allowed in <VirtualHost>";
}
ap_pid_fname = arg;
return NULL;
}
static const char *set_lockfile(cmd_parms *cmd, void *dummy, char *arg)
{
const char *err = ap_check_cmd_context(cmd, GLOBAL_ONLY);
if (err != NULL) {
return err;
}
lock_fname = arg;
return NULL;
}
static const char *set_num_daemons (cmd_parms *cmd, void *dummy, char *arg)
{
const char *err = ap_check_cmd_context(cmd, GLOBAL_ONLY);
if (err != NULL) {
return err;
}
num_daemons = atoi(arg);
if (num_daemons > HARD_SERVER_LIMIT) {
ap_log_error(APLOG_MARK, APLOG_STARTUP | APLOG_NOERRNO, 0, NULL,
"WARNING: NumServers of %d exceeds compile time limit "
"of %d servers,", num_daemons, HARD_SERVER_LIMIT);
ap_log_error(APLOG_MARK, APLOG_STARTUP | APLOG_NOERRNO, 0, NULL,
" lowering NumServers to %d. To increase, please "
"see the", HARD_SERVER_LIMIT);
ap_log_error(APLOG_MARK, APLOG_STARTUP | APLOG_NOERRNO, 0, NULL,
" HARD_SERVER_LIMIT define in src/include/httpd.h.");
num_daemons = HARD_SERVER_LIMIT;
}
else if (num_daemons < 1) {
ap_log_error(APLOG_MARK, APLOG_STARTUP | APLOG_NOERRNO, 0, NULL,
"WARNING: Require NumServers > 0, setting to 1");
num_daemons = 1;
}
return NULL;
}
static const char *set_threads_to_start (cmd_parms *cmd, void *dummy, char *arg)
{
const char *err = ap_check_cmd_context(cmd, GLOBAL_ONLY);
if (err != NULL) {
return err;
}
threads_to_start = atoi(arg);
if (threads_to_start > HARD_THREAD_LIMIT) {
ap_log_error(APLOG_MARK, APLOG_STARTUP | APLOG_NOERRNO, 0, NULL,
"WARNING: StartThreads of %d exceeds compile time"
"limit of %d threads,", threads_to_start,
HARD_THREAD_LIMIT);
ap_log_error(APLOG_MARK, APLOG_STARTUP | APLOG_NOERRNO, 0, NULL,
"lowering StartThreads to %d. To increase, please"
"see the", HARD_THREAD_LIMIT);
ap_log_error(APLOG_MARK, APLOG_STARTUP | APLOG_NOERRNO, 0, NULL,
"HARD_THREAD_LIMIT define in src/include/httpd.h.");
}
else if (threads_to_start < 1) {
ap_log_error(APLOG_MARK, APLOG_STARTUP | APLOG_NOERRNO, 0, NULL,
"WARNING: Require StartThreads > 0, setting to 1");
threads_to_start = 1;
}
return NULL;
}
static const char *set_min_spare_threads(cmd_parms *cmd, void *dummy, char *arg)
{
const char *err = ap_check_cmd_context(cmd, GLOBAL_ONLY);
if (err != NULL) {
return err;
}
min_spare_threads = atoi(arg);
if (min_spare_threads <= 0) {
ap_log_error(APLOG_MARK, APLOG_STARTUP | APLOG_NOERRNO, 0, NULL,
"WARNING: detected MinSpareThreads set to non-positive.");
ap_log_error(APLOG_MARK, APLOG_STARTUP | APLOG_NOERRNO, 0, NULL,
"Resetting to 1 to avoid almost certain Apache failure.");
ap_log_error(APLOG_MARK, APLOG_STARTUP | APLOG_NOERRNO, 0, NULL,
"Please read the documentation.");
min_spare_threads = 1;
}
return NULL;
}
static const char *set_max_spare_threads(cmd_parms *cmd, void *dummy, char *arg)
{
const char *err = ap_check_cmd_context(cmd, GLOBAL_ONLY);
if (err != NULL) {
return err;
}
max_spare_threads = atoi(arg);
if (max_spare_threads >= HARD_THREAD_LIMIT) {
ap_log_error(APLOG_MARK, APLOG_STARTUP | APLOG_NOERRNO, 0, NULL,
"WARNING: detected MinSpareThreads set higher than");
ap_log_error(APLOG_MARK, APLOG_STARTUP | APLOG_NOERRNO, 0, NULL,
"HARD_THREAD_LIMIT. Resetting to %d", HARD_THREAD_LIMIT);
max_spare_threads = HARD_THREAD_LIMIT;
}
return NULL;
}
static const char *set_max_threads(cmd_parms *cmd, void *dummy, char *arg)
{
const char *err = ap_check_cmd_context(cmd, GLOBAL_ONLY);
if (err != NULL) {
return err;
}
max_threads = atoi(arg);
if (max_threads > HARD_THREAD_LIMIT) {
ap_log_error(APLOG_MARK, APLOG_STARTUP | APLOG_NOERRNO, 0, NULL,
"WARNING: detected MaxThreadsPerChild set higher than");
ap_log_error(APLOG_MARK, APLOG_STARTUP | APLOG_NOERRNO, 0, NULL,
"HARD_THREAD_LIMIT. Resetting to %d", HARD_THREAD_LIMIT);
max_threads = HARD_THREAD_LIMIT;
}
return NULL;
}
static const char *set_max_requests(cmd_parms *cmd, void *dummy, char *arg)
{
const char *err = ap_check_cmd_context(cmd, GLOBAL_ONLY);
if (err != NULL) {
return err;
}
max_requests_per_child = atoi(arg);
return NULL;
}
static const char *set_maintain_connection_status(cmd_parms *cmd,
core_dir_config *d, int arg)
{
const char *err = ap_check_cmd_context(cmd, GLOBAL_ONLY);
if (err != NULL) {
return err;
}
ap_dexter_set_maintain_connection_status(arg != 0);
return NULL;
}
static const char *set_coredumpdir (cmd_parms *cmd, void *dummy, char *arg)
{
struct stat finfo;
const char *fname;
const char *err = ap_check_cmd_context(cmd, GLOBAL_ONLY);
if (err != NULL) {
return err;
}
fname = ap_server_root_relative(cmd->pool, arg);
if ((stat(fname, &finfo) == -1) || !S_ISDIR(finfo.st_mode)) {
return ap_pstrcat(cmd->pool, "CoreDumpDirectory ", fname,
" does not exist or is not a directory", NULL);
}
ap_cpystrn(ap_coredump_dir, fname, sizeof(ap_coredump_dir));
return NULL;
}
static const command_rec dexter_cmds[] = {
UNIX_DAEMON_COMMANDS
LISTEN_COMMANDS
{ "PidFile", set_pidfile, NULL, RSRC_CONF, TAKE1,
"A file for logging the server process ID"},
{ "LockFile", set_lockfile, NULL, RSRC_CONF, TAKE1,
"The lockfile used when Apache needs to lock the accept() call"},
{ "NumServers", set_num_daemons, NULL, RSRC_CONF, TAKE1,
"Number of children alive at the same time" },
{ "StartThreads", set_threads_to_start, NULL, RSRC_CONF, TAKE1,
"Number of threads each child creates" },
{ "MinSpareThreads", set_min_spare_threads, NULL, RSRC_CONF, TAKE1,
"Minimum number of idle threads per child, to handle request spikes" },
{ "MaxSpareThreads", set_max_spare_threads, NULL, RSRC_CONF, TAKE1,
"Maximum number of idle threads per child" },
{ "MaxThreadsPerChild", set_max_threads, NULL, RSRC_CONF, TAKE1,
"Maximum number of threads per child" },
{ "MaxRequestsPerChild", set_max_requests, NULL, RSRC_CONF, TAKE1,
"Maximum number of requests a particular child serves before dying." },
{ "ConnectionStatus", set_maintain_connection_status, NULL, RSRC_CONF, FLAG,
"Whether or not to maintain status information on current connections"},
{ "CoreDumpDirectory", set_coredumpdir, NULL, RSRC_CONF, TAKE1,
"The location of the directory Apache changes to before dumping core" },
{ NULL }
};
module MODULE_VAR_EXPORT mpm_dexter_module = {
STANDARD20_MODULE_STUFF,
NULL, /* create per-directory config structure */
NULL, /* merge per-directory config structures */
NULL, /* create per-server config structure */
NULL, /* merge per-server config structures */
dexter_cmds, /* command ap_table_t */
NULL, /* handlers */
dexter_hooks /* register_hooks */
};