首先我们来看看nginx服务器的进程模型是怎样的
执行以下命令:
ps -eo pid,ppid,sid,tty,pgrp,comm,cmd | grep -E 'PID|nginx'
结果如图:
可以看出nginx服务器由一个master process进程与四个worker process进程组成(需要声明的是,Nginx默认情况下只会产生一个worker process进程,可以通过修改nginx配置文件来产生多少worker process进程)。
再仔细观察进程ID,父进程ID,session ID,与进程组ID能发现:
1. 所有worker进程都是master进程的子进程
2. master进程与worker进程同属同一个进程组,同一个session,且master进程是此session的leader。
3. 由tty可以看出master进程与woker都是脱离了终端的守护进程
4. 由CMD看出master与worker的进程名并不像一个可执行文件名,可见是由程序主动定义的。
带着这些现象,我们来看看nginx源码中是如何实现这些机制的。
首先从入口函数开始分析(nginx版本1.14.2),入口函数在 ./src/core/nginx.c 中,下面贴出入口函数:
int ngx_cdecl
main(int argc, char *const *argv)
{
ngx_buf_t *b;
ngx_log_t *log;
ngx_uint_t i;
ngx_cycle_t *cycle, init_cycle;
ngx_conf_dump_t *cd;
ngx_core_conf_t *ccf;
ngx_debug_init();
if (ngx_strerror_init() != NGX_OK) {
return 1;
}
if (ngx_get_options(argc, argv) != NGX_OK) {
return 1;
}
if (ngx_show_version) {
ngx_show_version_info();
if (!ngx_test_config) {
return 0;
}
}
/* TODO */ ngx_max_sockets = -1;
ngx_time_init();
#if (NGX_PCRE)
ngx_regex_init();
#endif
ngx_pid = ngx_getpid();
ngx_parent = ngx_getppid();
log = ngx_log_init(ngx_prefix);
if (log == NULL) {
return 1;
}
/* STUB */
#if (NGX_OPENSSL)
ngx_ssl_init(log);
#endif
/*
* init_cycle->log is required for signal handlers and
* ngx_process_options()
*/
ngx_memzero(&init_cycle, sizeof(ngx_cycle_t));
init_cycle.log = log;
ngx_cycle = &init_cycle;
init_cycle.pool = ngx_create_pool(1024, log);
if (init_cycle.pool == NULL) {
return 1;
}
if (ngx_save_argv(&init_cycle, argc, argv) != NGX_OK) {
return 1;
}
if (ngx_process_options(&init_cycle) != NGX_OK) {
return 1;
}
if (ngx_os_init(log) != NGX_OK) {
return 1;
}
/*
* ngx_crc32_table_init() requires ngx_cacheline_size set in ngx_os_init()
*/
if (ngx_crc32_table_init() != NGX_OK) {
return 1;
}
/*
* ngx_slab_sizes_init() requires ngx_pagesize set in ngx_os_init()
*/
ngx_slab_sizes_init();
if (ngx_add_inherited_sockets(&init_cycle) != NGX_OK) {
return 1;
}
if (ngx_preinit_modules() != NGX_OK) {
return 1;
}
cycle = ngx_init_cycle(&init_cycle);
if (cycle == NULL) {
if (ngx_test_config) {
ngx_log_stderr(0, "configuration file %s test failed",
init_cycle.conf_file.data);
}
return 1;
}
if (ngx_test_config) {
if (!ngx_quiet_mode) {
ngx_log_stderr(0, "configuration file %s test is successful",
cycle->conf_file.data);
}
if (ngx_dump_config) {
cd = cycle->config_dump.elts;
for (i = 0; i < cycle->config_dump.nelts; i++) {
ngx_write_stdout("# configuration file ");
(void) ngx_write_fd(ngx_stdout, cd[i].name.data,
cd[i].name.len);
ngx_write_stdout(":" NGX_LINEFEED);
b = cd[i].buffer;
(void) ngx_write_fd(ngx_stdout, b->pos, b->last - b->pos);
ngx_write_stdout(NGX_LINEFEED);
}
}
return 0;
}
if (ngx_signal) {
return ngx_signal_process(cycle, ngx_signal);
}
ngx_os_status(cycle->log);
ngx_cycle = cycle;
ccf = (ngx_core_conf_t *) ngx_get_conf(cycle->conf_ctx, ngx_core_module);
if (ccf->master && ngx_process == NGX_PROCESS_SINGLE) {
ngx_process = NGX_PROCESS_MASTER;
}
#if !(NGX_WIN32)
if (ngx_init_signals(cycle->log) != NGX_OK) {
return 1;
}
if (!ngx_inherited && ccf->daemon) {
if (ngx_daemon(cycle->log) != NGX_OK) {
return 1;
}
ngx_daemonized = 1;
}
if (ngx_inherited) {
ngx_daemonized = 1;
}
#endif
if (ngx_create_pidfile(&ccf->pid, cycle->log) != NGX_OK) {
return 1;
}
if (ngx_log_redirect_stderr(cycle) != NGX_OK) {
return 1;
}
if (log->file->fd != ngx_stderr) {
if (ngx_close_file(log->file->fd) == NGX_FILE_ERROR) {
ngx_log_error(NGX_LOG_ALERT, cycle->log, ngx_errno,
ngx_close_file_n " built-in log failed");
}
}
ngx_use_stderr = 0;
if (ngx_process == NGX_PROCESS_SINGLE) {
ngx_single_process_cycle(cycle);
} else {
ngx_master_process_cycle(cycle);
}
return 0;
}
nginx初始化时做了很多的工作,我暂时就只带领大家分析上述提出的4点。
首先我们先注意到main函数中64行调用的ngx_save_argv函数:
static ngx_int_t
ngx_save_argv(ngx_cycle_t *cycle, int argc, char *const *argv)
{
#if (NGX_FREEBSD)
ngx_os_argv = (char **) argv;
ngx_argc = argc;
ngx_argv = (char **) argv;
#else
size_t len;
ngx_int_t i;
ngx_os_argv = (char **) argv;
ngx_argc = argc;
ngx_argv = ngx_alloc((argc + 1) * sizeof(char *), cycle->log);
if (ngx_argv == NULL) {
return NGX_ERROR;
}
for (i = 0; i < argc; i++) {
len = ngx_strlen(argv[i]) + 1;
ngx_argv[i] = ngx_alloc(len, cycle->log);
if (ngx_argv[i] == NULL) {
return NGX_ERROR;
}
(void) ngx_cpystrn((u_char *) ngx_argv[i], (u_char *) argv[i], len);
}
ngx_argv[i] = NULL;
#endif
ngx_os_environ = environ;
return NGX_OK;
}
从代码可以看到nginx拷贝了argv参数到一块分配出来的内存。
其次我们定位到main函数150行 ngx_init_signals 函数,该函数里nginx进行了信号的捕获操作:
ngx_int_t
ngx_init_signals(ngx_log_t *log)
{
ngx_signal_t *sig;
struct sigaction sa;
for (sig = signals; sig->signo != 0; sig++) {
ngx_memzero(&sa, sizeof(struct sigaction));
if (sig->handler) {
sa.sa_sigaction = sig->handler;
sa.sa_flags = SA_SIGINFO;
} else {
sa.sa_handler = SIG_IGN;
}
sigemptyset(&sa.sa_mask);
if (sigaction(sig->signo, &sa, NULL) == -1) {
#if (NGX_VALGRIND)
ngx_log_error(NGX_LOG_ALERT, log, ngx_errno,
"sigaction(%s) failed, ignored", sig->signame);
#else
ngx_log_error(NGX_LOG_EMERG, log, ngx_errno,
"sigaction(%s) failed", sig->signame);
return NGX_ERROR;
#endif
}
}
return NGX_OK;
}
接着往下走,在main函数155行调用了ngx_daemon函数,此函数即为创建守护进程的函数:
ngx_int_t
ngx_daemon(ngx_log_t *log)
{
int fd;
switch (fork()) {
case -1:
ngx_log_error(NGX_LOG_EMERG, log, ngx_errno, "fork() failed");
return NGX_ERROR;
case 0:
break;
default:
exit(0);
}
ngx_parent = ngx_pid;
ngx_pid = ngx_getpid();
if (setsid() == -1) {
ngx_log_error(NGX_LOG_EMERG, log, ngx_errno, "setsid() failed");
return NGX_ERROR;
}
umask(0);
fd = open("/dev/null", O_RDWR);
if (fd == -1) {
ngx_log_error(NGX_LOG_EMERG, log, ngx_errno,
"open(\"/dev/null\") failed");
return NGX_ERROR;
}
if (dup2(fd, STDIN_FILENO) == -1) {
ngx_log_error(NGX_LOG_EMERG, log, ngx_errno, "dup2(STDIN) failed");
return NGX_ERROR;
}
if (dup2(fd, STDOUT_FILENO) == -1) {
ngx_log_error(NGX_LOG_EMERG, log, ngx_errno, "dup2(STDOUT) failed");
return NGX_ERROR;
}
#if 0
if (dup2(fd, STDERR_FILENO) == -1) {
ngx_log_error(NGX_LOG_EMERG, log, ngx_errno, "dup2(STDERR) failed");
return NGX_ERROR;
}
#endif
if (fd > STDERR_FILENO) {
if (close(fd) == -1) {
ngx_log_error(NGX_LOG_EMERG, log, ngx_errno, "close() failed");
return NGX_ERROR;
}
}
return NGX_OK;
}
其中可看出fork后,直接exit(0)了父进程,而子进程进行了 setsid(),umask(),dup2() 等创建守护进程的操作。
接着回到main函数最后,第189行 ngx_master_process_cycle 函数:
void
ngx_master_process_cycle(ngx_cycle_t *cycle)
{
char *title;
u_char *p;
size_t size;
ngx_int_t i;
ngx_uint_t n, sigio;
sigset_t set;
struct itimerval itv;
ngx_uint_t live;
ngx_msec_t delay;
ngx_listening_t *ls;
ngx_core_conf_t *ccf;
sigemptyset(&set);
sigaddset(&set, SIGCHLD);
sigaddset(&set, SIGALRM);
sigaddset(&set, SIGIO);
sigaddset(&set, SIGINT);
sigaddset(&set, ngx_signal_value(NGX_RECONFIGURE_SIGNAL));
sigaddset(&set, ngx_signal_value(NGX_REOPEN_SIGNAL));
sigaddset(&set, ngx_signal_value(NGX_NOACCEPT_SIGNAL));
sigaddset(&set, ngx_signal_value(NGX_TERMINATE_SIGNAL));
sigaddset(&set, ngx_signal_value(NGX_SHUTDOWN_SIGNAL));
sigaddset(&set, ngx_signal_value(NGX_CHANGEBIN_SIGNAL));
if (sigprocmask(SIG_BLOCK, &set, NULL) == -1) {
ngx_log_error(NGX_LOG_ALERT, cycle->log, ngx_errno,
"sigprocmask() failed");
}
sigemptyset(&set);
size = sizeof(master_process);
for (i = 0; i < ngx_argc; i++) {
size += ngx_strlen(ngx_argv[i]) + 1;
}
title = ngx_pnalloc(cycle->pool, size);
if (title == NULL) {
/* fatal */
exit(2);
}
p = ngx_cpymem(title, master_process, sizeof(master_process) - 1);
for (i = 0; i < ngx_argc; i++) {
*p++ = ' ';
p = ngx_cpystrn(p, (u_char *) ngx_argv[i], size);
}
ngx_setproctitle(title);
ccf = (ngx_core_conf_t *) ngx_get_conf(cycle->conf_ctx, ngx_core_module);
ngx_start_worker_processes(cycle, ccf->worker_processes,
NGX_PROCESS_RESPAWN);
ngx_start_cache_manager_processes(cycle, 0);
ngx_new_binary = 0;
delay = 0;
sigio = 0;
live = 1;
for ( ;; ) {
if (delay) {
if (ngx_sigalrm) {
sigio = 0;
delay *= 2;
ngx_sigalrm = 0;
}
ngx_log_debug1(NGX_LOG_DEBUG_EVENT, cycle->log, 0,
"termination cycle: %M", delay);
itv.it_interval.tv_sec = 0;
itv.it_interval.tv_usec = 0;
itv.it_value.tv_sec = delay / 1000;
itv.it_value.tv_usec = (delay % 1000 ) * 1000;
if (setitimer(ITIMER_REAL, &itv, NULL) == -1) {
ngx_log_error(NGX_LOG_ALERT, cycle->log, ngx_errno,
"setitimer() failed");
}
}
ngx_log_debug0(NGX_LOG_DEBUG_EVENT, cycle->log, 0, "sigsuspend");
sigsuspend(&set);
ngx_time_update();
ngx_log_debug1(NGX_LOG_DEBUG_EVENT, cycle->log, 0,
"wake up, sigio %i", sigio);
if (ngx_reap) {
ngx_reap = 0;
ngx_log_debug0(NGX_LOG_DEBUG_EVENT, cycle->log, 0, "reap children");
live = ngx_reap_children(cycle);
}
if (!live && (ngx_terminate || ngx_quit)) {
ngx_master_process_exit(cycle);
}
if (ngx_terminate) {
if (delay == 0) {
delay = 50;
}
if (sigio) {
sigio--;
continue;
}
sigio = ccf->worker_processes + 2 /* cache processes */;
if (delay > 1000) {
ngx_signal_worker_processes(cycle, SIGKILL);
} else {
ngx_signal_worker_processes(cycle,
ngx_signal_value(NGX_TERMINATE_SIGNAL));
}
continue;
}
if (ngx_quit) {
ngx_signal_worker_processes(cycle,
ngx_signal_value(NGX_SHUTDOWN_SIGNAL));
ls = cycle->listening.elts;
for (n = 0; n < cycle->listening.nelts; n++) {
if (ngx_close_socket(ls[n].fd) == -1) {
ngx_log_error(NGX_LOG_EMERG, cycle->log, ngx_socket_errno,
ngx_close_socket_n " %V failed",
&ls[n].addr_text);
}
}
cycle->listening.nelts = 0;
continue;
}
if (ngx_reconfigure) {
ngx_reconfigure = 0;
if (ngx_new_binary) {
ngx_start_worker_processes(cycle, ccf->worker_processes,
NGX_PROCESS_RESPAWN);
ngx_start_cache_manager_processes(cycle, 0);
ngx_noaccepting = 0;
continue;
}
ngx_log_error(NGX_LOG_NOTICE, cycle->log, 0, "reconfiguring");
cycle = ngx_init_cycle(cycle);
if (cycle == NULL) {
cycle = (ngx_cycle_t *) ngx_cycle;
continue;
}
ngx_cycle = cycle;
ccf = (ngx_core_conf_t *) ngx_get_conf(cycle->conf_ctx,
ngx_core_module);
ngx_start_worker_processes(cycle, ccf->worker_processes,
NGX_PROCESS_JUST_RESPAWN);
ngx_start_cache_manager_processes(cycle, 1);
/* allow new processes to start */
ngx_msleep(100);
live = 1;
ngx_signal_worker_processes(cycle,
ngx_signal_value(NGX_SHUTDOWN_SIGNAL));
}
if (ngx_restart) {
ngx_restart = 0;
ngx_start_worker_processes(cycle, ccf->worker_processes,
NGX_PROCESS_RESPAWN);
ngx_start_cache_manager_processes(cycle, 0);
live = 1;
}
if (ngx_reopen) {
ngx_reopen = 0;
ngx_log_error(NGX_LOG_NOTICE, cycle->log, 0, "reopening logs");
ngx_reopen_files(cycle, ccf->user);
ngx_signal_worker_processes(cycle,
ngx_signal_value(NGX_REOPEN_SIGNAL));
}
if (ngx_change_binary) {
ngx_change_binary = 0;
ngx_log_error(NGX_LOG_NOTICE, cycle->log, 0, "changing binary");
ngx_new_binary = ngx_exec_new_binary(cycle, ngx_argv);
}
if (ngx_noaccept) {
ngx_noaccept = 0;
ngx_noaccepting = 1;
ngx_signal_worker_processes(cycle,
ngx_signal_value(NGX_SHUTDOWN_SIGNAL));
}
}
}
我们注意到 ngx_master_process_cycle 第54行调用了 ngx_setproctitle(title); 函数,该函数的作用就是实现我们在nginx进程模型中观察到的第4点,修改进程的进程名。
再看第59行,调用了 ngx_start_worker_processes(cycle, ccf>worker_processes, NGX_PROCESS_RESPAWN); 函数,此函数的作用就是在master进程中fork出进程模型中的worker进程。
至此我们就已经从源码分析完了nginx服务器进程模型的原理。但是有点还需要注意的是当worker子进程被杀死,nginx是如何处理僵尸进程的?这里我们定位到nginx捕获信号时设置的信号处理函数 ngx_signal_handler :
static void
ngx_signal_handler(int signo, siginfo_t *siginfo, void *ucontext)
{
char *action;
ngx_int_t ignore;
ngx_err_t err;
ngx_signal_t *sig;
ignore = 0;
err = ngx_errno;
for (sig = signals; sig->signo != 0; sig++) {
if (sig->signo == signo) {
break;
}
}
ngx_time_sigsafe_update();
action = "";
switch (ngx_process) {
case NGX_PROCESS_MASTER:
case NGX_PROCESS_SINGLE:
switch (signo) {
case ngx_signal_value(NGX_SHUTDOWN_SIGNAL):
ngx_quit = 1;
action = ", shutting down";
break;
case ngx_signal_value(NGX_TERMINATE_SIGNAL):
case SIGINT:
ngx_terminate = 1;
action = ", exiting";
break;
case ngx_signal_value(NGX_NOACCEPT_SIGNAL):
if (ngx_daemonized) {
ngx_noaccept = 1;
action = ", stop accepting connections";
}
break;
case ngx_signal_value(NGX_RECONFIGURE_SIGNAL):
ngx_reconfigure = 1;
action = ", reconfiguring";
break;
case ngx_signal_value(NGX_REOPEN_SIGNAL):
ngx_reopen = 1;
action = ", reopening logs";
break;
case ngx_signal_value(NGX_CHANGEBIN_SIGNAL):
if (ngx_getppid() == ngx_parent || ngx_new_binary > 0) {
/*
* Ignore the signal in the new binary if its parent is
* not changed, i.e. the old binary's process is still
* running. Or ignore the signal in the old binary's
* process if the new binary's process is already running.
*/
action = ", ignoring";
ignore = 1;
break;
}
ngx_change_binary = 1;
action = ", changing binary";
break;
case SIGALRM:
ngx_sigalrm = 1;
break;
case SIGIO:
ngx_sigio = 1;
break;
case SIGCHLD:
ngx_reap = 1;
break;
}
break;
case NGX_PROCESS_WORKER:
case NGX_PROCESS_HELPER:
switch (signo) {
case ngx_signal_value(NGX_NOACCEPT_SIGNAL):
if (!ngx_daemonized) {
break;
}
ngx_debug_quit = 1;
/* fall through */
case ngx_signal_value(NGX_SHUTDOWN_SIGNAL):
ngx_quit = 1;
action = ", shutting down";
break;
case ngx_signal_value(NGX_TERMINATE_SIGNAL):
case SIGINT:
ngx_terminate = 1;
action = ", exiting";
break;
case ngx_signal_value(NGX_REOPEN_SIGNAL):
ngx_reopen = 1;
action = ", reopening logs";
break;
case ngx_signal_value(NGX_RECONFIGURE_SIGNAL):
case ngx_signal_value(NGX_CHANGEBIN_SIGNAL):
case SIGIO:
action = ", ignoring";
break;
}
break;
}
if (siginfo && siginfo->si_pid) {
ngx_log_error(NGX_LOG_NOTICE, ngx_cycle->log, 0,
"signal %d (%s) received from %P%s",
signo, sig->signame, siginfo->si_pid, action);
} else {
ngx_log_error(NGX_LOG_NOTICE, ngx_cycle->log, 0,
"signal %d (%s) received%s",
signo, sig->signame, action);
}
if (ignore) {
ngx_log_error(NGX_LOG_CRIT, ngx_cycle->log, 0,
"the changing binary signal is ignored: "
"you should shutdown or terminate "
"before either old or new binary's process");
}
if (signo == SIGCHLD) {
ngx_process_get_status();
}
ngx_set_errno(err);
}
我们定位到函数最后第145行,当 if (signo == SIGCHLD) 即父进程收到了子进程被杀死的信号时,调用了 ngx_process_get_status(); 函数,这个函数就是处理僵尸进程的:
static void
ngx_process_get_status(void)
{
int status;
char *process;
ngx_pid_t pid;
ngx_err_t err;
ngx_int_t i;
ngx_uint_t one;
one = 0;
for ( ;; ) {
pid = waitpid(-1, &status, WNOHANG);
if (pid == 0) {
return;
}
if (pid == -1) {
err = ngx_errno;
if (err == NGX_EINTR) {
continue;
}
if (err == NGX_ECHILD && one) {
return;
}
/*
* Solaris always calls the signal handler for each exited process
* despite waitpid() may be already called for this process.
*
* When several processes exit at the same time FreeBSD may
* erroneously call the signal handler for exited process
* despite waitpid() may be already called for this process.
*/
if (err == NGX_ECHILD) {
ngx_log_error(NGX_LOG_INFO, ngx_cycle->log, err,
"waitpid() failed");
return;
}
ngx_log_error(NGX_LOG_ALERT, ngx_cycle->log, err,
"waitpid() failed");
return;
}
one = 1;
process = "unknown process";
for (i = 0; i < ngx_last_process; i++) {
if (ngx_processes[i].pid == pid) {
ngx_processes[i].status = status;
ngx_processes[i].exited = 1;
process = ngx_processes[i].name;
break;
}
}
if (WTERMSIG(status)) {
#ifdef WCOREDUMP
ngx_log_error(NGX_LOG_ALERT, ngx_cycle->log, 0,
"%s %P exited on signal %d%s",
process, pid, WTERMSIG(status),
WCOREDUMP(status) ? " (core dumped)" : "");
#else
ngx_log_error(NGX_LOG_ALERT, ngx_cycle->log, 0,
"%s %P exited on signal %d",
process, pid, WTERMSIG(status));
#endif
} else {
ngx_log_error(NGX_LOG_NOTICE, ngx_cycle->log, 0,
"%s %P exited with code %d",
process, pid, WEXITSTATUS(status));
}
if (WEXITSTATUS(status) == 2 && ngx_processes[i].respawn) {
ngx_log_error(NGX_LOG_ALERT, ngx_cycle->log, 0,
"%s %P exited with fatal code %d "
"and cannot be respawned",
process, pid, WEXITSTATUS(status));
ngx_processes[i].respawn = 0;
}
ngx_unlock_mutexes(pid);
}
}
最后,我们就已经分析完了nginx服务器进程模型的原理,nginx服务器非常的深奥,更多精髓的东西,我们后续继续分析。