CGI简介
CGI 是Web 服务器运行时外部程序的规范,按CGI 编写的程序可以扩展服务器功能。CGI 应用程序能与浏览器进行交互,还可通过数据库API 与数据库服务器等外部数据源进行通信,从数据库服务器中获取数据。格式化为HTML文档后,发送给浏览器,也可以将从浏览器获得的数据放到数据库中。几乎所有服务器都支持CGI,可用任何语言编写CGI,包括流行的C、C ++、VB 和Delphi 等。CGI 分为标准CGI 和间接CGI两种。标准CGI 使用命令行参数或环境变量表示服务器的详细请求,服务器与浏览器通信采用标准输入输出方式。间接CGI 又称缓冲CGI,在CGI 程序和CGI 接口之间插入一个缓冲程序,缓冲程序与CGI 接口间用标准输入输出进行通信。
简而言之:CGI就是web服务器可以调用的其他程序,该程序将输出重定向,在该程序中通过printf生成html,通过管道将数据发给web服务器,完成交互。
boa CGI实现
1.boa获取到数据时,执行process_requests()-->读取http头read_header()--->process_header_end();根据头部信息,获取到该请求为CGI
2.执行init_cgi()
int init_cgi(request * req)
{
int child_pid;
int pipes[2];
int use_pipes = 0;
SQUASH_KA(req);
if (req->is_cgi) {
if (complete_env(req) == 0) { //设置环境变量
return 0;
}
}
#ifdef FASCIST_LOGGING
{
int i;
for (i = 0; i < req->cgi_env_index; ++i)
fprintf(stderr, "%s - environment variable for cgi: \"%s\"\n",
__FILE__, req->cgi_env[i]);
}
#endif
if (req->is_cgi == CGI || 1) {
use_pipes = 1; //使用管道
if (pipe(pipes) == -1) {
log_error_time();
perror("pipe");
return 0;
}
/* set the read end of the socket to non-blocking */
if (set_nonblock_fd(pipes[0]) == -1) { //设置读管道为非阻塞
log_error_time();
perror("cgi-fcntl");
close(pipes[0]);
close(pipes[1]);
return 0;
}
}
child_pid = fork(); //创建子进程,用于启动cgi程序,程序执行完成后退出boa
switch(child_pid) {
case -1:
/* fork unsuccessful */
log_error_time();
perror("fork");
if (use_pipes) {
close(pipes[0]);
close(pipes[1]);
}
send_r_error(req);
/* FIXME: There is aproblem here. send_r_error would work
for NPH and CGI, but not for GUNZIP. Fix that. */
/* i'd like to send_r_error, but.... */
return 0;
break;
case 0:
/* child */
if (req->is_cgi == CGI || req->is_cgi == NPH) {
char *foo = strdup(req->pathname);
char *c;
if (!foo) {
WARN("unable to strdup pathname for req->pathname");
_exit(1);
}
c = strrchr(foo, '/');
if (c) {
++c;
*c = '\0';
} else {
/* we have a serious problem */
log_error_time();
perror("chdir");
if (use_pipes)
close(pipes[1]);
_exit(1);
}
if (chdir(foo) != 0) {
log_error_time();
perror("chdir");
if (use_pipes)
close(pipes[1]);
_exit(1);
}
}
if (use_pipes) {
close(pipes[0]);
/* tie cgi's STDOUT to it's write end of pipe */
if (dup2(pipes[1], STDOUT_FILENO) == -1) { //将写管道重定向到标准输出,在cgi中使用printf,数据就会被发送到管道中
log_error_time();
perror("dup2 - pipes");
close(pipes[1]);
_exit(1);
}
close(pipes[1]);
if (set_block_fd(STDOUT_FILENO) == -1) { //设置管道为阻塞
log_error_time();
perror("cgi-fcntl");
_exit(1);
}
} else {
/* tie stdout to socket */
if (dup2(req->fd, STDOUT_FILENO) == -1) {
log_error_time();
perror("dup2 - fd");
_exit(1);
}
/* Switch socket flags back to blocking */
if (set_block_fd(req->fd) == -1) {
log_error_time();
perror("cgi-fcntl");
_exit(1);
}
}
/* tie post_data_fd to POST stdin */
if (req->method == M_POST) { /* tie stdin to file */
lseek(req->post_data_fd, SEEK_SET, 0);
dup2(req->post_data_fd, STDIN_FILENO);
close(req->post_data_fd);
}
/* Close access log, so CGI program can't scribble
* where it shouldn't
*/
close_access_log();
/*
* tie STDERR to cgi_log_fd
* cgi_log_fd will automatically close, close-on-exec rocks!
* if we don't tied STDERR (current log_error) to cgi_log_fd,
* then we ought to close it.
*/
if (!cgi_log_fd)
dup2(devnullfd, STDERR_FILENO);
else
dup2(cgi_log_fd, STDERR_FILENO);
if (req->is_cgi) {
char *aargv[CGI_ARGC_MAX + 1];
create_argv(req, aargv);
execve(req->pathname, aargv, req->cgi_env); //启动cgi程序
} else {
if (req->pathname[strlen(req->pathname) - 1] == '/')
execl(dirmaker, dirmaker, req->pathname, req->request_uri,
NULL);
#ifdef GUNZIP
else
execl(GUNZIP, GUNZIP, "--stdout", "--decompress",
req->pathname, NULL);
#endif
}
/* execve failed */
WARN(req->pathname);
_exit(1);
break;
default:
/* parent */
/* if here, fork was successful */
if (verbose_cgi_logs) {
log_error_time();
fprintf(stderr, "Forked child \"%s\" pid %d\n",
req->pathname, child_pid);
}
if (req->method == M_POST) {
close(req->post_data_fd); /* child closed it too */
req->post_data_fd = 0;
}
/* NPH, GUNZIP, etc... all go straight to the fd */
if (!use_pipes)
return 0;
close(pipes[1]);
req->data_fd = pipes[0]; //保存读管道的fd
req->status = PIPE_READ; //状态置为PIPE_READ
if (req->is_cgi == CGI) {
req->cgi_status = CGI_PARSE; /* got to parse cgi header */
/* for cgi_header... I get half the buffer! */
req->header_line = req->header_end =
(req->buffer + BUFFER_SIZE / 2);
} else {
req->cgi_status = CGI_BUFFER;
/* I get all the buffer! */
req->header_line = req->header_end = req->buffer;
}
/* reset req->filepos for logging (it's used in pipe.c) */
/* still don't know why req->filesize might be reset though */
req->filepos = 0;
break;
}
return 1; //more to do
}
3.下一次轮询中,程序回到process_request,状态为
PIPE_READ,准备读取管道中的数据。读取后,继续轮询,直至全部读取完毕,状态置为
PIPE_WRITE,将数据发送给客户端。(正是由于boa的这种轮询机制,使得它可以处理多个请求而不会卡住)