c半同步半异步进程池模型之cgi服务器
程序员文章站
2022-03-26 22:43:30
对半同步半异步进程池模型垂涎已久，这次中秋放假撸了下代码，代码写了几个模块，分别是：
util：封装了套接字创建、unix族socket管道创建、中断信号、简单屏幕输出(可自行替...
对半同步半异步进程池模型垂涎已久，这次中秋放假撸了下代码，代码写了几个模块，分别是：
util：封装了套接字创建、unix族socket管道创建、中断信号、简单屏幕输出(可自行替换为日志文件输出)
epoll_wrapper：封装了epoll相关操作包括创建epfd、添加epoll监听事件、删除epoll监听事件
myhshappool(我的半同步半异步进程池 - -!…)：封装了进程池初始化、启动进程池进行事件监听
client_handle：进程池监听到客户事件、即调用client_handle封装的处理事件，这里封装的是执行cgi文件向客户端浏览器返回服务器时间(最近在看unix网络编程，里面都是时间获取的服务器，借鉴下拿来搞事，当然，嵌入式里拿来控制个灯泡开关想来特别带劲，用android做个网页app，板子接wifi模块接智能灯，cgi负责开关灯泡 。。)
cgisrv：入口，初始化进程池，启动进程池
代码快1k行，不知道一个博客文章能不能写下，不太会用github，况且这种玩具demo代码就不往github放了。代码中凑合写了注释(有时候不想切换中英文因此用了蹩脚的英文注释)，限(wo)于(tai)篇(lan)幅(le)没有写文件头注释和函数头注释。
util.h：
#ifndef _UTIL_H
#define _UTIL_H
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
//这里不用 #转字符串了，不方便管理等级
#ifdef LEVELNUMPRINT
#define LEVEL0 "LEVEL0"
#define LEVEL1 "LEVEL1"
#define LEVEL2 "LEVEL2"
#define LEVEL3 "LEVEL3"
#define LEVEL4 "LEVEL4"
#define LEVEL4 "LEVEL5"
#else
#define LEVEL0 "DEBUG"
#define LEVEL1 "INFO"
#define LEVEL2 "NOTICE"
#define LEVEL3 "WARN"
#define LEVEL4 "ERROR"
#define LEVEL5 "FATAL"
#endif
#define PRINTINFO( LEVEL, format, args... ) \
do { \
printf("[%s]:(pid:%d)/(file:%s)/(func:%s)/(line:%d)", \
LEVEL, getpid(), __FILE__, __func__, __LINE__); \
/*printf("\t" #LEVEL ":");*/ \
printf("--|--"); \
printf( format, ##args ); \
printf("\n"); \
} while( 0 )
#define PRINTINFO_ERR( LEVEL, format, args... ) \
do { \
printf("[%s]:(pid:%d)/(file:%s)/(func:%s)/(line:%d)", \
LEVEL, getpid(), __FILE__, __func__, __LINE__); \
/*printf("\t" #LEVEL ":");*/ \
printf("--|--"); \
printf( format, ##args ); \
printf("(errmsg:%s)", strerror(errno)); \
printf("\n"); \
} while( 0 )
void Add_sig( int sig, void (*handler)(int),
int restart_syscall );
void Socketpair( int *pairpipefd );
int Socket_create( char *ipaddr, int port );
void Setnonblocking( int fd );
#endif
util.c:
#include "util.h"
static int add_sig( int sig, void (*handler)(int),
int restart_syscall )
{
struct sigaction act;
bzero( &act, sizeof(act) );
act.sa_handler = handler;
act.sa_flags = 0;
//早期unix系统对于进程在执行一个低速系统调用(如ioctl、
//read、write、wait)而阻塞期间捕捉到一个信号，则系统
//调用被中断不再执行，该系统调用返回错误，设置errno为
//EINTR，随后的bsd系统引入了自动重启，即再次进行此系统
//调用。unix衍生系统默认的方式可能为可选、总是等，类
//unix系统的linux系统可能默认为不重启，因此添加重启标识
if ( restart_syscall ) {
act.sa_flags |= SA_RESTART;
}
//宏定义：
//#define sigfillset(*p) (*p) = ~(0,0)
sigfillset( &act.sa_mask );
if ( -1 == sigaction(sig, &act, NULL) ) {
PRINTINFO_ERR( LEVEL4, "sigaction error" );
return -1;
}
return 0;
}
void Add_sig( int sig, void (*handler)(int),
int restart_syscall )
{
if ( add_sig(sig, handler, restart_syscall) < 0 ) {
PRINTINFO( LEVEL5, "add_sig error" );
exit( 0 );
}
}
void Socketpair( int *pairpipefd )
{
int ret;
ret = socketpair( PF_UNIX, SOCK_STREAM,
0, pairpipefd );
if ( ret < 0 ) {
PRINTINFO_ERR( LEVEL5, "socketpair error!!" );
exit( 0 );
}
}
static int socket_create( char *ipaddr, int port,
int backlog )
{
int sockfd;
int ret;
sockfd = socket( AF_INET, SOCK_STREAM, 0 );
if ( sockfd < 0 ) {
PRINTINFO_ERR( LEVEL4, "socket error!!!" );
return -1;
}
struct sockaddr_in addr;
bzero( &addr, sizeof(addr) ) ;
addr.sin_family = AF_INET;
addr.sin_port = htons( port );
inet_pton( AF_INET, ipaddr, &addr.sin_addr );
int reuseaddr = 1;
setsockopt( sockfd, SOL_SOCKET, SO_REUSEADDR,
&reuseaddr, sizeof(int) );
ret = bind( sockfd, (struct sockaddr *)&addr, (socklen_t)sizeof(addr) ) ;
if ( ret < 0 ) {
PRINTINFO_ERR( LEVEL4, "bind error!!!" );
return -1;
}
ret = listen( sockfd, backlog );
if ( ret < 0 ) {
PRINTINFO_ERR( LEVEL4, "listen error!!!" );
return -1;
}
return sockfd;
}
int Socket_create( char *ipaddr, int port )
{
int ret;
ret = socket_create( ipaddr, port, 5 );
if ( ret < 0 ) {
PRINTINFO( LEVEL5, "socket_creaet error!!!" );
exit( 0 );
}
return ret;
}
static int setnonblocking( int fd )
{
int old_opt = fcntl( fd, F_GETFL );
int new_opt = old_opt | O_NONBLOCK;
fcntl( fd, F_SETFL, new_opt );
return old_opt;
}
void Setnonblocking( int fd )
{
setnonblocking( fd );
}
int Send ( int socket_fd, const unsigned char * send_buf,
int buf_size, int flag )
{
int snd_bytes = 0;
int snd_total_bytes = 0;
int snd_count = 3;
while ( snd_count -- ) {
snd_bytes = send( socket_fd, send_buf, buf_size, flag );
if ( snd_bytes <= 0 ) {
if ( EAGAIN == errno || EINTR == errno
|| EWOULDBLOCK == errno ) { //暂时发送失败，需要重复发送
usleep( 50 );
continue;
}else { //连接不正常，返回-1交由上层清理此套接字
PRINTINFO_ERR( LEVEL4, "send return error!!!" );
return -1;
}
}
snd_total_bytes += snd_bytes;
if ( snd_total_bytes >= buf_size ) {
break;
}
}
if ( !snd_count ) {
PRINTINFO( LEVEL4, "send timeout!!!" );
return -1;
}
return snd_total_bytes;
}
#if 0
int main()
{
PRINTINFO( LEVEL0, "likun:%d", 123 );
PRINTINFO( LEVEL1, "likun:" );
//PRINTINFO( likun, "likun:" );
return 0;
}
#endif
epoll_wrapper.h:
#ifndef _EPOLL_WRAPPER_H
#define _EPOLL_WRAPPER_H
#include
#include
int Epoll_create( int size );
int Epoll_wait( int epfd, struct epoll_event *events,
int maxevents, int timeout );
void Epoll_add_fd( int epfd, int fd );
void Epoll_del_fd( int epfd, int fd );
#endif
epoll_wrapper.c:
#include "epoll_wrapper.h"
#include "util.h"
static int epoll_create0( int size )
{
int ret;
ret = epoll_create( size );
if ( ret <= 0 ) {
PRINTINFO_ERR( LEVEL3, "epoll_create error!!!" );
return -1;
}
return ret;
}
int Epoll_create( int size )
{
int ret;
if ( (ret = epoll_create0(size)) < 0 ) {
PRINTINFO( LEVEL5, "epoll_create0 error!!!" );
exit( 0 );
}
return ret;
}
int Epoll_wait( int epfd, struct epoll_event *events,
int maxevents, int timeout )
{
return epoll_wait( epfd, events, maxevents, timeout );
}
static int epoll_add_fd( int epfd, int fd )
{
struct epoll_event event;
event.data.fd = fd;
event.events = EPOLLIN | EPOLLET;
epoll_ctl( epfd, EPOLL_CTL_ADD, fd, &event );
Setnonblocking( fd );
return 0;
}
void Epoll_add_fd( int epfd, int fd )
{
epoll_add_fd( epfd, fd );
}
void Epoll_del_fd( int epfd, int fd )
{
epoll_ctl( epfd, EPOLL_CTL_DEL, fd, NULL );
}
#if 0
int main()
{}
#endif
myhshappool.h:
#ifndef _MY_HS_HA_P_POOL_H
#define _MY_HS_HA_P_POOL_H
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
typedef struct process {
//当前进程id号
pid_t pid;
//与父进程通信用的管道
//0端父进程写
//1端子进程读
int pipefd[2];
} process;
typedef struct processpool {
//进程池最大进程数
int max_process_num;
//每个进程最大处理客户数量
int max_user_num;
//epoll最多处理的事件数
int max_epoll_event;
//当前进程池进程总数
int cur_process_num;
//当前子进程在进程池序号，从0开始
int index;
//每个进程一个epoll内核事件表
int epollfd;
//监听socket
int listenfd;
//停止线程池
int stop;
//进程池子进程管理
struct process *sub_process;
} processpool;
int init_process_pool( processpool *ppool, int maxpnum,
int maxunum, int maxeevent, int curpnum, int listenfd );
void run( processpool *ppool );
#endif
myhshappool.c:
#include "myhshappool.h"
#include "client_handle.h"
#include "util.h"
//用于信号中断时主进程通信，
//统一处理事件，即将客户端连接
//事件、信号事件都统一用epoll
//监听处理，0端信号处理函数写，
//1端进程读
static int sig_pipefd[2];
static void sig_handler( int sig )
{
//保存旧的errno，对后续的send不
//进行错误判定，但send假如返回
//失败会设置errno，信号中断调用
//结束后影响进程其它模块判断
int old_errno = errno;
char signo = (char)sig;
send( sig_pipefd[0], (char *)&signo, 1, 0 );
errno = old_errno;
}
static void init_signal( processpool *ppool )
{
Socketpair( sig_pipefd );
Setnonblocking( sig_pipefd[0] );
//Epoll_add_fd( ppool->epollfd, sig_pipefd[1] );
Add_sig( SIGCHLD, sig_handler, 1 );
Add_sig( SIGTERM, sig_handler, 1 );
Add_sig( SIGINT, sig_handler, 1 );
Add_sig( SIGPIPE, SIG_IGN, 1 );
}
int init_process_pool( processpool *ppool, int maxpnum,
int maxunum, int maxeevent, int curpnum, int listenfd )
{
if ( !ppool ) {
PRINTINFO( LEVEL4, "ppool is null!!!" );
return -1;
}
ppool->max_process_num = maxpnum;
ppool->max_user_num = maxunum;
ppool->max_epoll_event = maxeevent;
ppool->cur_process_num = curpnum;
ppool->listenfd = listenfd;
// ppool->epollfd = Epoll_create( 5 );
ppool->index = -1;
ppool->stop = 0;
ppool->sub_process =
(process *)calloc( sizeof(process), curpnum );
if ( !ppool->sub_process ) {
PRINTINFO( LEVEL4, "sub_process calloc error!!!" );
return -1;
}
int i = 0;
int pid;
//先模拟一下进程池创建之前的情况，假设终端
//bash shell进程id为1000，运行此程序进程id
//为1001，其父进程为1000，fork之后主进程不
//变，子进程id为1002，其父进程为1001，因此
//明白fork的过程，下面可以走一下进程池创建
//的流程（条件均为以上假设）：
//第一次fork:创建亲缘进程的管道，父进程1001
//，子进程1002，其父进程为1001，子进程不再
//执行for循环，且主进程与1002子进程有单独通
//信的管道
//第二次fork:创建亲缘进程的管道，父进程1001
//，子进程1003，其父进程为1001，子进程不再
//执行for循环，且主进程与1003子进程有单独通
//信的管道
//第三次fork .....1004.....
// ........
//通过以上过程，可以看到for循环次数为创建的
//子进程数量，且每个子进程可以单独与父进程
//通信
//
//这里进程创建，没有脱离当前终端的会话，
//我觉得可以setsid()来摆脱终端影响
for ( ; i < curpnum; i++ ) {
Socketpair( ppool->sub_process[i].pipefd );
pid = fork();
if ( pid > 0 ) { //parent fork
close( ppool->sub_process[i].pipefd[1] );
ppool->sub_process[i].pid = pid;
Setnonblocking( ppool->sub_process[i].pipefd[0] );
continue;
} else if ( pid == 0 ) { //child
ppool->index = i;
PRINTINFO( LEVEL0, "child(%d):%d\tparent:%d", i + 1, getpid(), getppid() );
close( ppool->sub_process[i].pipefd[0] );
//每次只由父进程去创建进程
break;
}
else {
PRINTINFO( LEVEL5, "fork error!!!" );
exit( 0 );
}
}
}
static int client_signal_handle( processpool *ppool,
char *signals, int signals_num )
{
int i = 0;
for ( ; i < signals_num; i++ ) {
switch( signals[i] ) {
case SIGCHLD:
{
PRINTINFO( LEVEL0, "child receive a SIGCHLD signal" );
pid_t pid;
int stat;
//catch SIGCHLD
while ( (pid = waitpid(-1, &stat, WNOHANG)) > 0 ) {
continue;
}
break;
}
case SIGTERM:
{
PRINTINFO( LEVEL0, "child receive a SIGTERM signal" );
ppool->stop = 1;
break;
}
case SIGINT:
{
PRINTINFO( LEVEL0, "child receive a SIGINT signal" );
ppool->stop = 1;
break;
}
default:
{
break;
}
}
}
}
static void run_child( processpool *ppool )
{
if ( !ppool ) {
PRINTINFO( LEVEL4, "ppool is null!!!" );
return;
}
init_signal( ppool );
ppool->epollfd = Epoll_create( 5 );
PRINTINFO( LEVEL0, "child cur process:%d", ppool->cur_process_num );
int pipefd = ppool->sub_process[ppool->index].pipefd[1];
Epoll_add_fd( ppool->epollfd, pipefd );
Epoll_add_fd( ppool->epollfd, sig_pipefd[1] );
struct epoll_event *events = (struct epoll_event *)
calloc( sizeof(struct epoll_event), ppool->max_epoll_event );
if ( !events ) {
PRINTINFO( LEVEL5, "calloc error!!!" );
goto _free_source;
}
struct client_param *cparam = NULL;
cparam = (struct client_param *)
calloc( sizeof(struct client_param), ppool->max_user_num );
if ( !cparam ) {
PRINTINFO( LEVEL5, "calloc error!!!" );
goto _free_source;
}
int event_num, event_fd;
int i, j, ret, onebyte;
while ( !ppool->stop ) {
event_num = Epoll_wait( ppool->epollfd, events, ppool->max_epoll_event, -1);
//PRINTINFO( LEVEL0, "child event num:%d", event_num );
if ( (event_num < 0) && (errno != EINTR) ) {
PRINTINFO_ERR( LEVEL4, "Epoll_wait error!!!" );
ppool->stop = 1;
break;
}
for ( i = 0; i < event_num; i++ ) {
event_fd = events[i].data.fd;
//parent process notify that there is a new client connect to.
if ( event_fd == pipefd && events[i].events & EPOLLIN ) {
PRINTINFO( LEVEL0, "receive signal from parent there is a new client connection" );
ret = recv( event_fd, (char *)&onebyte, 1, 0 );
if ( ret <= 0 ) {
continue;
} else {
struct sockaddr_in clientaddr;
socklen_t addrlen = sizeof(clientaddr);
bzero( &clientaddr, addrlen );
int connfd = accept( ppool->listenfd,
(struct sockaddr *)&clientaddr, &addrlen );
if ( connfd < 0 ) {
PRINTINFO_ERR( LEVEL3, "accept a new client error!!!" );
continue;
}
PRINTINFO( LEVEL0, "one client conntect(fd:%d)", connfd );
Epoll_add_fd( ppool->epollfd, connfd );
client_param_init( &cparam[connfd], connfd, &clientaddr );
}
}
//process catch a signal
else if ( event_fd == sig_pipefd[1] && events[i].events & EPOLLIN ) {
int sig;
char signals[1024] = {0};
ret = recv( sig_pipefd[1], signals, sizeof(signals), 0 );
if ( ret <= 0 ) {
continue;
}
client_signal_handle( ppool, signals, ret );
}
//client socket fd has readable event,maybe a
//request
else if ( events[i].events & EPOLLIN ) {
client_handle( &cparam[event_fd] );
}
else {
continue;
}
}
}
_free_source:
free( events );
events = NULL;
free( cparam );
cparam = NULL;
close( pipefd );
close( ppool->epollfd );
}
static int parent_signal_handle( processpool *ppool,
char *signals, int signals_num )
{
int i = 0, j = 0;
for ( ; i < signals_num; i++ ) {
switch( signals[i] ) {
case SIGCHLD:
{
PRINTINFO( LEVEL0, "parent receive SIGCHLD signal" );
pid_t pid;
int stat;
while ( (pid = waitpid(-1, &stat, WNOHANG)) > 0 ) {
PRINTINFO( LEVEL0, "parent receive SIGCHLD signal(pid:%d)", pid );
for ( j = 0; j < ppool->cur_process_num; j++ ) {
if ( ppool->sub_process[j].pid == pid ) {
PRINTINFO( LEVEL3, "child process(%d) exit.", pid );
close( ppool->sub_process[j].pipefd[1] );
ppool->sub_process[j].pid = -1;
}
}
}
ppool->stop = 1;
for ( j = 0; j < ppool->cur_process_num; j++ ) {
PRINTINFO( LEVEL0, "pid:%d", ppool->sub_process[j].pid );
if ( ppool->sub_process[j].pid != -1 ) {
ppool->stop = 0;
break;
}
}
break;
}
case SIGTERM:
case SIGINT:
{
PRINTINFO( LEVEL2, "recv SIGINT/SIGTERM, kill all child process now." );
//PRINTINFO( LEVEL0, "cur_process_num:%d", ppool->cur_process_num );
for ( i = 0; i < ppool->cur_process_num; i++ ) {
int pid = ppool->sub_process[i].pid;
if ( pid != -1 ) {
PRINTINFO( LEVEL0, "kill process:%d", pid );
ppool->sub_process[i].pid = -1;
kill( pid, SIGTERM );
}
}
ppool->stop = 1;
break;
}
default:
{
break;
}
}
}
}
static void run_parent( processpool *ppool )
{
if ( !ppool ) {
PRINTINFO( LEVEL4, "ppool is null!!!" );
return;
}
init_signal( ppool );
ppool->epollfd = Epoll_create( 5 );
Epoll_add_fd( ppool->epollfd, ppool->listenfd );
Epoll_add_fd( ppool->epollfd, sig_pipefd[1] );
struct epoll_event *events = NULL;
events = (struct epoll_event *)
calloc( sizeof(struct epoll_event), ppool->max_epoll_event );
if ( !events ) {
PRINTINFO( LEVEL5, "calloc error!!!" );
char sig = SIGINT;
parent_signal_handle( ppool, &sig, 1 );
goto _free_source;
}
int event_num;
int i, onebyte = 1, ret, j;
//p_idx specifies current dispatched child process.
//roll_index specifies the next child process.
int p_idx , roll_index = 0;
//PRINTINFO( LEVEL0, "epollfd:%d", ppool->epollfd );
while ( !ppool->stop ) {
//Specifying a timeout of -1 causes epoll_wait()
//to block indefinitely.
event_num =
Epoll_wait( ppool->epollfd, events, ppool->max_epoll_event, -1 );
if ( event_num < 0 && errno != EINTR ) {
PRINTINFO_ERR( LEVEL5, "Epoll_wait error!!!" );
ppool->stop = 1;
break;
}
//PRINTINFO( LEVEL0, "event_num:%d", event_num );
for ( i = 0; i < event_num; i++ ) {
int event_fd = events[i].data.fd;
//listenfd,there is a new client connection.
//notify child process to accept
if ( event_fd == ppool->listenfd ) {
PRINTINFO( LEVEL0, "event:parent listenfd" );
//round robin dispatch
//easily roll polling
p_idx = roll_index;
do {
if ( ppool->sub_process[p_idx].pid != -1 ) {
break;
}
p_idx = ( p_idx + 1 ) % ppool->cur_process_num;
} while ( p_idx != roll_index );
//roll polling all the child process,but they are
//all run error.so p_idx equals to roll_index.
if ( ppool->sub_process[p_idx].pid < 0 ) {
ppool->stop = 1;
break;
}
roll_index = ( p_idx + 1 ) % ppool->cur_process_num;
if ( Send( ppool->sub_process[p_idx].pipefd[0],
(char *)&onebyte, 1, 0 ) < 0 ) {
PRINTINFO( LEVEL5, "Send error!!!" );
ppool->stop = 1;
break;
}
}
//receive signal from signal handler.
else if ( (event_fd == sig_pipefd[1])
&& (events[i].events & EPOLLIN) ) {
PRINTINFO( LEVEL0, "event:parent receive signal" );
int sig;
char signals[1024];
ret = recv( sig_pipefd[1], signals, sizeof(signals), 0 );
if ( ret <= 0 ) {
continue;
} else {
parent_signal_handle( ppool, signals, ret );
}
}
}
}
_free_source:
free( events );
events = NULL;
close( ppool->epollfd );
}
void run( processpool *ppool )
{
if ( ppool->index != -1 ) {
run_child( ppool );
return;
}
run_parent( ppool );
}
client_handle.h:
#ifndef _CLIENT_HANDLE_H
#define _CLIENT_HANDLE_H
#include
#include
#include
#include
#include
#include
#include
#include
#define MAXRECVBUF 1024
typedef struct client_param {
//用于recv返回出错移除sockfd
int epollfd;
int sockfd;
struct sockaddr_in addr;
char buf[ MAXRECVBUF ];
} client_param;
void client_param_init( client_param *cparam,
int connfd, struct sockaddr_in *addr );
void client_handle( client_param *cparam );
#endif
client_handle.c:
#include "client_handle.h"
#include "util.h"
void client_param_init( client_param *cparam,
int connfd, struct sockaddr_in *addr )
{
if ( !cparam ) {
PRINTINFO( LEVEL4, "cparam is null!!!" );
return;
}
cparam->sockfd = connfd;
memcpy( &cparam->addr, addr, sizeof(struct sockaddr_in) );
memset( cparam->buf, 0, sizeof(cparam->buf) );
}
void client_handle( client_param *cparam )
{
int i, ret;
while ( 1 ) {
memset( cparam->buf, 0, sizeof(cparam->buf) );
ret = recv( cparam->sockfd, cparam->buf, sizeof(cparam->buf), 0 );
if ( ret < 0 ) {
if ( errno != EAGAIN && errno != EWOULDBLOCK
&& errno != EINTR ) {
Epoll_del_fd( cparam->epollfd, cparam->sockfd );
}
close( cparam->sockfd );
break;
}
else if ( 0 == ret ) {
Epoll_del_fd( cparam->epollfd, cparam->sockfd );
close( cparam->sockfd );
break;
}
else {
if ( ret < 15 ) {
close( cparam->sockfd );
break;
}
//PRINTINFO( LEVEL0, "child receive buf:\n%s", cparam->buf );
fflush(stdout);
char *p_get = strstr( cparam->buf, "GET" );
if ( !p_get ) {
close( cparam->sockfd );
break;
}
char *p_http = strstr( cparam->buf, "HTTP" );
if ( !p_http ) {
close( cparam->sockfd );
break;
}
cparam->buf[ret] = '\0';
//GET filename HTTP/1.1 .....
char file_name[20] = {0};
int file_name_len = p_http - p_get - 6;
memcpy( file_name, p_get + 5, file_name_len );
if ( access( file_name, F_OK ) == -1 ) {
PRINTINFO( LEVEL3, "file:(%s) dosen't exist!!", file_name );
Epoll_del_fd( cparam->epollfd, cparam->sockfd );
close( cparam->sockfd );
break;
}
PRINTINFO( LEVEL0, "file name:%s--", file_name );
ret = fork();
if ( ret == -1 ) {
Epoll_del_fd( cparam->epollfd, cparam->sockfd );
close( cparam->sockfd );
break;
}
else if ( ret > 0 ) {
Epoll_del_fd( cparam->epollfd, cparam->sockfd );
close( cparam->sockfd );
break;
}
else {
close( STDOUT_FILENO );
//relocate the stdou to sockfd
PRINTINFO( LEVEL0, "sockfd:%d", cparam->sockfd );
dup( cparam->sockfd );
//printf("likun\n");
execl( file_name, file_name, NULL );
fflush(stdout);
Epoll_del_fd( cparam->epollfd, cparam->sockfd );
close( cparam->sockfd );
exit( 0 );
}
}
}
}
cgisrv.c:
#include "myhshappool.h"
#include "client_handle.h"
#include "util.h"
#include "epoll_wrapper.h"
//进程池最大数量
#define MAXPROCESSNUMBER 16
//每个进程支持客户连接任务
#define USERPERPROCESS 65535
//epool最大支持的监听事件数
#define MAXEPOLLEVENT 10000
processpool ppool;
int main()
{
int listenfd = Socket_create( "192.168.1.250", 8888 );
init_process_pool( &ppool, MAXPROCESSNUMBER,
USERPERPROCESS, MAXEPOLLEVENT, 5, listenfd );
run( &ppool );
close( listenfd );
return 0;
}
贴一下makefile：
CC = gcc
ROOTDIR = $(shell pwd)
OBJ = util.o myhshappool.o epoll_wrapper.o \
client_handle.o cgisrc.o
BIN = cgisrv.bin
CFLAG = -Wall -O2 -I./
LDFLAG += -c
$(BIN):${OBJ}
$(CC) $(CFLAG) -o $@ $^
%:%.c
$(CC) $(CFLAG) -o $@ $< $(LDFLAG)
.PHONY:clean
clean:
rm $(OBJ) $(BIN) -rf
还有cgi执行程序：
#include
#include
#include
/*..
*/
int main( int argc, char **argv )
{
time_t tt = time(NULL);
printf("");
printf("");
printf("");
printf("");
printf("");
printf("");
printf("");
printf("");
printf("当前服务器时间:%s", ctime(&tt));
printf("");
printf("");
printf("");
return 0;
}
测试：代码编译了，即可执行cgisrv.bin，主进程处于监听客户端连接情况，打开浏览器输入: http://xxx.xxx.xxx.xxx:8888/now.cgi，可以看到出现一行加下划线的绿字：当前服务器时间:Sat Sep 17 21:49:07 2016。
ip、端口、进程池数、最大epoll监听事件数等在入口模块(cgisrv)可以改。
模型的简单介绍：主进程创建几个子进程作为工作进程，主进程监听客户端connect事件，一旦有连接事件，即通过round robin(简单轮询)选取一个子进程， 通过父子进程间的通信管道通知子进程有连接事件，子进程epoll监听到管道通信时间，即知道有客户端连接，因此进程accept，并将客户端连接套接字加入epoll监听事件，客户端浏览器发送get请求，客户端监听到请求事件，即调用封装好的客户端事件进行处理，本例子的客户端处理为recv客户的请求，从中提取文件名(now.cgi)，并重定向客户端连接套接字到stdout，然后执行对应cgi文件，cgi文件打印html页面字符串，因为重定向的缘故，打印的html字符串发送给客户端，客户端浏览器即显示了html页面。
本例子写完，调试了几个地方，运行几个客户端发送get请求就没有做测试了，因此也不知道高并发是否会跑出bug（epoll模型+进程池处理上百上千的并发应该不是问题，主要是怕我代码有bug），其实应当弄个压力测试来跑一下的，不过也没时间弄了，还要学习半同步半反应堆的模型。
下一篇： js 闭包原理