自己动手写 HTTP Server
-
作者:heiyeluren
-
时间:2008-06-22
【 原理 】 一般来说,HTTP Server 也是我们常说的Web服务器,大名鼎鼎的 Apache,还是微软的 IIS (Internet Information Server),开源领域的有 Lighttpd 和最近风头正劲的 Nginx 都是典型的Web服务器。最近想想能不能做一个Web服务器,就提供简单的功能,但是速度很快,能够作为一个专门处理 HTML/CSS/JS/Image 的Web服务器,那样能够让静态资源文件迅速的被访问到,如果有反向代理功能就更帅了,当然了,要是有Cache功能啥的,并且能够编写自定义插件(扩展)就很完美了。。。YY中。。。
基于这个思想,我就花十天时间使用标准C写了一个千行代码小型的HTTP Server,当然目前还不具有反向代理和扩展功能,只是能够简单的支持 HTML/CSS/JS/Image 静态资源文件的访问。HTTP Server 的名字叫做 tmhttpd - TieMa (Tiny&Mini) HTTP Server,小巧,代码少,执行速度快,目前具有的功能包括:
- Support GET/HEAD method
- The common MIME types.
- Support Self custom default index page
- Directory listings.
- Support access log
- Support Self custom port and max clients
- Use fork mode accept new conntion
- Support daemon type work
- more ..
目前已经发布了 tmhttpd-1.0.0_alpha 版本,包括for Unix/Linux (在 Ubuntu/Fedoar/FreeBSD 下编译通过) 和 在cygwin环境下编译的 for Windows 版本,下面地址有下载:
- Unix/Linux: http://heiyeluren.googlecode.com/files/tmhttpd-1.0.0_alpha.tar.gz
- Windows: http://heiyeluren.googlecode.com/files/tmhttpd-1.0.0_alpha-win32.zip
下面大致来聊聊怎么写一个 HTTP Server,先看看一个HTTP请求的流程:
大致就是 客户端的浏览器(IE、Firefox、Opera、Lynx、Curl、Robot ...) 访问到Web服务器的端口(一般是80),然后端口接受到请求后开始解析请求,如果请求不正确或者非法则直接返回指定的错误代码(可以参考 RFC 1945),如果请求合法,那么就查找相应用户请求的文件,把文件读取后发送给客户端,关闭连接,请求结束。
【 实现 】
贴部分tmhttpd核心代码来描述这个过程:
/**
* 初始化服务器端的Socket连接,等待连接,连接成功后fork新进程处理
*
*/
static void InitServerListen( unsigned int port, unsigned int max_client ){
int serversock, clientsock;
struct sockaddr_in server_addr, client_addr;
char currtime[32];
/* Create the TCP socket */
if ((serversock = socket(PF_INET, SOCK_STREAM, IPPROTO_TCP)) < 0){
die("Failed to create socket");
}
/* Construct the server sockaddr_in structure */
memset(&server_addr, 0, sizeof(server_addr)); /* Clear struct */
server_addr.sin_family = AF_INET; /* Internet/IP */
server_addr.sin_addr.s_addr = htonl(INADDR_ANY); /* Incoming addr */
server_addr.sin_port = htons(port); /* server port */
/* Bind the server socket */
if (bind(serversock, (struct sockaddr *) &server_addr, sizeof(server_addr)) < 0){
die("Failed to bind the server socket");
}
/* Listen on the server socket */
if (listen(serversock, max_client) < 0){
die("Failed to listen on server socket");
}
/* Print listening message */
getdate(currtime);
fprintf(stdout, "[%s] Start server listening at port %d .../n", currtime, port);
fprintf(stdout, "[%s] Waiting client connection .../n", currtime);
/* Run until cancelled */
while (1){
unsigned int clientlen = sizeof(client_addr);
memset(currtime, 0, sizeof(currtime));
getdate(currtime);
/* Wait for client connection */
if ((clientsock = accept(serversock, (struct sockaddr *) &client_addr, &clientlen)) < 0){
die("Failed to accept client connection");
}
/* Use child process new connection */
if ( fork() == 0 ){
HandleClient(clientsock, client_addr);
} else {
wait(NULL);
}
/* Not use close socket connection */
close(clientsock);
}
}
/**
* 获取一个连接,读取连接客户端发送的请求数据,把请求数据叫给请求解析函数进行解析
*
*/
static void HandleClient( int client_sock, struct sockaddr_in client_addr ){
char buf[REQUEST_MAX_SIZE];
if ( read(client_sock, buf, REQUEST_MAX_SIZE) < 0){
SendError( client_sock, 500, "Internal Server Error", "", "Client request not success." );
die("read sock");
}
ParseRequest( client_sock, client_addr, buf );
close(client_sock);
exit(0);
}
/**
* 解析一个请求,解析出GET/HEAD方法,需要请求的文件,协议版本等,构造成结构体提交给处理函数
*
*/
static int ParseRequest( int client_sock, struct sockaddr_in client_addr, char *req ){
char **buf, **method_buf, **query_buf, currtime[32], cwd[1024], tmp_path[1536], pathinfo[512], path[256], file[256], log[1024];
int line_total, method_total, query_total, i;
struct st_request_info request_info;
/* Split client request */
getdate(currtime);
explode(req, '/n', &buf, &line_total);
/* Print log message */
memset(log, 0, sizeof(log));
sprintf(log, "[%s] %s %s/n", currtime, inet_ntoa(client_addr.sin_addr), buf[0]);
WriteLog(log);
/* Check request is empty */
if (strcmp(buf[0], "/n") == 0 || strcmp(buf[0], "/r/n") == 0){
SendError( client_sock, 400, "Bad Request", "", "Can't parse request." );
}
/* Check method is implement */
explode(buf[0], ' ', &method_buf, &method_total);
if ( strcmp( strtolower(method_buf[0]), "get") != 0 && strcmp( strtolower(method_buf[0]), "head") != 0 ){
SendError( client_sock, 501, "Not Implemented", "", "That method is not implemented." );
}
explode(method_buf[1], '?', &query_buf, &query_total);
/* Make request data */
getcwd(cwd, sizeof(cwd));
strcpy(pathinfo, query_buf[0]);
substr(query_buf[0], 0, strrpos(pathinfo, '/')+1, path);
substr(query_buf[0], strrpos(pathinfo, '/')+1, 0, file);
/* Pad request struct */
memset(&request_info, 0, sizeof(request_info));
strcat(cwd, pathinfo);
request_info.method = method_buf[0];
request_info.pathinfo = pathinfo;
request_info.query = (query_total == 2 ? query_buf[1] : "");
request_info.protocal = strtolower(method_buf[2]);
request_info.path = path;
request_info.file = file;
request_info.physical_path = cwd;
/* Is a directory pad default index page */
memset(tmp_path, 0, sizeof(tmp_path));
strcpy(tmp_path, cwd);
if ( is_dir(tmp_path) ){
strcat(tmp_path, g_dir_index);
if ( file_exists(tmp_path) ){
request_info.physical_path = tmp_path;
}
}
/* Debug message */
if ( g_is_debug ){
fprintf(stderr, "[ Request ]/n");
for(i=0; i<line_total; i++){
fprintf(stderr, "%s/n", buf[i]);
}
}
/* Process client request */
ProcRequest( client_sock, client_addr, request_info );
return 0;
}
/**
* 处理函数按照解析出来的请求内容进行数据返回,返回文件/目录列表或者提示错误
*
*/
static int ProcRequest( int client_sock, struct sockaddr_in client_addr, struct st_request_info request_info ){
char buf[128];
/* File is exist or has access permission */
if ( !file_exists( request_info.physical_path ) ){
memset(buf, 0, sizeof(buf));
sprintf(buf, "File %s not found.", request_info.pathinfo);
SendError( client_sock, 404, "Not Found", "", buf);
}
if ( access(request_info.physical_path, R_OK) != 0 ){
memset(buf, 0, sizeof(buf));
sprintf(buf, "File %s is protected.", request_info.pathinfo);
SendError( client_sock, 403, "Forbidden", "", buf);
}
/* Check target is regular file or directory */
if ( is_file(request_info.physical_path) == 1 ){
SendFile( client_sock, request_info.physical_path, request_info.pathinfo );
} else if ( is_dir(request_info.physical_path) == 1 ){
/* Is a directory choose browse dir list */
if ( g_is_browse ){
SendDirectory( client_sock, request_info.physical_path, request_info.pathinfo );
} else {
memset(buf, 0, sizeof(buf));
sprintf(buf, "File %s is protected.", request_info.pathinfo);
SendError( client_sock, 403, "Forbidden", "", buf);
}
} else {
memset(buf, 0, sizeof(buf));
sprintf(buf, "File %s is protected.", request_info.pathinfo);
SendError( client_sock, 403, "Forbidden", "", buf);
}
return 0;
}
主要核心的函数就是这四个,如果要了解其他函数,包括字符串处理,HTTP头信息发送,错误发送,读取文件,遍历目录等等请下载源码回去研究。源码下载地址:http://heiyeluren.googlecode.com/files/tmhttpd-1.0.0_alpha.tar.gz
【 结束 】
其实还有很多功能能够增加的,比如性能,采用 select 或者 epool 肯定要比单纯的fork性能更好,另外增加处理动态内容,比如CGI,能够接受POST请求,能够作为一个反向代理存在,能够具有缓存功能,能够缓存请求过的文件到内存中,能够具有插件扩展功能等等,都是需要进一步提升的。
参考文档:
- RFC 1945: http://www.w3.org/Protocols/rfc1945/rfc1945
- RFC 2616: http://www.w3.org/Protocols/rfc2616/rfc2616.html