webbench源码分析

webench是运行在linux操作系统下的一款web压力测试工具，其源码也只有几百行。本文在webbench 1.5版本的基础，对源码进行了详细的注释。
webbench.c

#include "socket.c"
#include <unistd.h>
#include <sys/param.h>
#include <rpc/types.h>
#include <getopt.h>
#include <strings.h>
#include <time.h>
#include <signal.h>

/* values */
volatile int timerexpired=0;
int speed=0;
int failed=0;
int bytes=0;
/* globals */
int http10=1; /* 0 - http/0.9, 1 - http/1.0, 2 - http/1.1 */
/* Allow: GET, HEAD, OPTIONS, TRACE */
#define METHOD_GET 0
#define METHOD_HEAD 1
#define METHOD_OPTIONS 2
#define METHOD_TRACE 3
#define PROGRAM_VERSION "1.5"
int method=METHOD_GET;
int clients=1;
int force=0;
int force_reload=0;
int proxyport=80;
char *proxyhost=NULL;
int benchtime=30;
/* internal */
int mypipe[2];
char host[MAXHOSTNAMELEN];
#define REQUEST_SIZE 2048
char request[REQUEST_SIZE];

static const struct option long_options[]=
{
 {"force",no_argument,&force,1},
 {"reload",no_argument,&force_reload,1},
 {"time",required_argument,NULL,'t'},
 {"help",no_argument,NULL,'?'},
 {"http09",no_argument,NULL,'9'},
 {"http10",no_argument,NULL,'1'},
 {"http11",no_argument,NULL,'2'},
 {"get",no_argument,&method,METHOD_GET},
 {"head",no_argument,&method,METHOD_HEAD},
 {"options",no_argument,&method,METHOD_OPTIONS},
 {"trace",no_argument,&method,METHOD_TRACE},
 {"version",no_argument,NULL,'V'},
 {"proxy",required_argument,NULL,'p'},
 {"clients",required_argument,NULL,'c'},
 {NULL,0,NULL,0}
};

/* prototypes */
static void benchcore(const char* host,const int port, const char *request);
static int bench(void);
static void build_request(const char *url);

static void alarm_handler(int signal)
{
   timerexpired=1;
}   

static void usage(void)
{
   fprintf(stderr,
    "webbench [option]... URL\n"
    "  -f|--force               Don't wait for reply from server.\n"
    "  -r|--reload              Send reload request - Pragma: no-cache.\n"
    "  -t|--time <sec>          Run benchmark for <sec> seconds. Default 30.\n"
    "  -p|--proxy <server:port> Use proxy server for request.\n"
    "  -c|--clients <n>         Run <n> HTTP clients at once. Default one.\n"
    "  -9|--http09              Use HTTP/0.9 style requests.\n"
    "  -1|--http10              Use HTTP/1.0 protocol.\n"
    "  -2|--http11              Use HTTP/1.1 protocol.\n"
    "  --get                    Use GET request method.\n"
    "  --head                   Use HEAD request method.\n"
    "  --options                Use OPTIONS request method.\n"
    "  --trace                  Use TRACE request method.\n"
    "  -?|-h|--help             This information.\n"
    "  -V|--version             Display program version.\n"
    );
};
int main(int argc, char *argv[])
{
 int opt=0;
 int options_index=0;
 char *tmp=NULL;

 if(argc==1)
 {
      usage();
          return 2;
 } 

 //解析出所有的命令行参数
 while((opt=getopt_long(argc,argv,"912Vfrt:p:c:?h",long_options,&options_index))!=EOF )
 {
  switch(opt)
  {
   case  0 : break;
   case 'f': force=1;break;
   case 'r': force_reload=1;break; 
   case '9': http10=0;break;
   case '1': http10=1;break;
   case '2': http10=2;break;
   case 'V': printf(PROGRAM_VERSION"\n");exit(0);
   case 't': benchtime=atoi(optarg);break;//一个客户端运行时间        
   case 'p': 
         /* proxy server parsing server:port */
         tmp=strrchr(optarg,':');//使用代理服务器
         proxyhost=optarg;
         if(tmp==NULL)
         {
             break;
         }
         if(tmp==optarg)
         {
             fprintf(stderr,"Error in option --proxy %s: Missing hostname.\n",optarg);
             return 2;
         }
         if(tmp==optarg+strlen(optarg)-1)
         {
             fprintf(stderr,"Error in option --proxy %s Port number is missing.\n",optarg);
             return 2;
         }
         *tmp='\0';
         proxyport=atoi(tmp+1);break;//代理的端口
   case ':':
   case 'h':
   case '?': usage();return 2;break;//打印出应用程序使用帮助信息
   case 'c': clients=atoi(optarg);break;//模拟多少个客户端同时进行请求
  }
 }

 //参数中没有URL
 if(optind==argc) {
                      fprintf(stderr,"webbench: Missing URL!\n");
              usage();
              return 2;
                    }

 if(clients==0) clients=1;
 if(benchtime==0) benchtime=60;
 /* Copyright */
 fprintf(stderr,"Webbench - Simple Web Benchmark "PROGRAM_VERSION"\n"
     "Copyright (c) Radim Kolar 1997-2004, GPL Open Source Software.\n"
     );
 /** 生成http请求的字符串 ***/
 build_request(argv[optind]);
 /* print bench info */
 printf("\nBenchmarking: ");
 switch(method)
 {
     case METHOD_GET:
     default:
         printf("GET");break;
     case METHOD_OPTIONS:
         printf("OPTIONS");break;
     case METHOD_HEAD:
         printf("HEAD");break;
     case METHOD_TRACE:
         printf("TRACE");break;
 }
 printf(" %s",argv[optind]);
 switch(http10)
 {
     case 0: printf(" (using HTTP/0.9)");break;
     case 2: printf(" (using HTTP/1.1)");break;
 }
 printf("\n");
 if(clients==1) printf("1 client");
 else
   printf("%d clients",clients);

 printf(", running %d sec", benchtime);
 if(force) printf(", early socket close");
 if(proxyhost!=NULL) printf(", via proxy server %s:%d",proxyhost,proxyport);
 if(force_reload) printf(", forcing reload");
 printf(".\n");
 return bench();
}

void build_request(const char *url)
{
  char tmp[10];
  int i;

  bzero(host,MAXHOSTNAMELEN);
  bzero(request,REQUEST_SIZE);

  if(force_reload && proxyhost!=NULL && http10<1) http10=1;
  if(method==METHOD_HEAD && http10<1) http10=1;
  if(method==METHOD_OPTIONS && http10<2) http10=2;
  if(method==METHOD_TRACE && http10<2) http10=2;

 /** 设置请求的方法 **/
  switch(method)
  {
      default:
      case METHOD_GET: strcpy(request,"GET");break;
      case METHOD_HEAD: strcpy(request,"HEAD");break;
      case METHOD_OPTIONS: strcpy(request,"OPTIONS");break;
      case METHOD_TRACE: strcpy(request,"TRACE");break;
  }

  strcat(request," ");

    //找到http://或https://中的://
  if(NULL==strstr(url,"://"))//如果没有则说明请求的URL不合法
  {
      fprintf(stderr, "\n%s: is not a valid URL.\n",url);
      exit(2);
  }
  if(strlen(url)>1500)//请求的url字符串太长。
  {
         fprintf(stderr,"URL is too long.\n");
     exit(2);
  }
  if(proxyhost==NULL)
       if (0!=strncasecmp("http://",url,7)) //除了开头的http://如果还有http://报错URL不合法
       { fprintf(stderr,"\nOnly HTTP protocol is directly supported, set --proxy for others.\n");
             exit(2);
           }
  /* protocol/host delimiter */
  i=strstr(url,"://")-url+3;//找到://后面第一个字符在URL字符串所在下标值，也就是域名或ip地址的第一个字符
  /* printf("%d\n",i); */

  if(strchr(url+i,'/')==NULL) {//://后面没有/不正确的URL
                                fprintf(stderr,"\nInvalid URL syntax - hostname don't ends with '/'.\n");
                                exit(2);
                              }
  if(proxyhost==NULL)
  {
      //URL地址中带有端口号
   if(index(url+i,':')!=NULL &&
      index(url+i,':')<index(url+i,'/'))
   {
       //解析出域名或IP地址，放到host中。
       strncpy(host,url+i,strchr(url+i,':')-url-i);
       bzero(tmp,10);
       //解析出URL中的端口
       strncpy(tmp,index(url+i,':')+1,strchr(url+i,'/')-index(url+i,':')-1);
       proxyport=atoi(tmp);
       if(proxyport==0) proxyport=80;
   } else
   {
       //URL中没有端口，直接解析出主机域名或IP地址放到host中。
     strncpy(host,url+i,strcspn(url+i,"/"));
   }
   //将要请求的URI放到GET或HEAE等后面
   strcat(request+strlen(request),url+i+strcspn(url+i,"/"));
  } else
  {
      //使用代理，直接将整个输入的URL都做为get或head后面的参数
   strcat(request,url);
  }
  if(http10==1)
      strcat(request," HTTP/1.0");
  else if (http10==2)
      strcat(request," HTTP/1.1");
  strcat(request,"\r\n");
  if(http10>0)
      strcat(request,"User-Agent: WebBench "PROGRAM_VERSION"\r\n");
  if(proxyhost==NULL && http10>0)
  {
      strcat(request,"Host: ");
      strcat(request,host);//不使用代理，在http请求中需要设置Host
      strcat(request,"\r\n");
  }
  if(force_reload && proxyhost!=NULL)
  {
      strcat(request,"Pragma: no-cache\r\n");
  }
  if(http10>1)
      strcat(request,"Connection: close\r\n");//不使用长连接
  //http语法，http请求头结束需要一个空行\r\n
  if(http10>0) strcat(request,"\r\n"); 
  // printf("Req=%s\n",request);
}

/* vraci system rc error kod */
static int bench(void)
{
  int i,j,k;    
  pid_t pid=0;
  FILE *f;

  /* check avaibility of target server */
  //创建一个连接到服务器的socket标识
  i=Socket(proxyhost==NULL?host:proxyhost,proxyport);
  if(i<0) { 
       fprintf(stderr,"\nConnect to server failed. Aborting benchmark.\n");
           return 1;
         }
  close(i);
  /* create pipe */
  if(pipe(mypipe))//创建管道,mypipe[0]为管道里的读取端,mypipe[1]为管道里的写入端
  {
      perror("pipe failed.");
      return 3;
  }

  /* fork childs */
  for(i=0;i<clients;i++)
  {
       pid=fork();//使用fork创建clients个进程，用于模拟clients个客户端用于请求
       if(pid <= (pid_t) 0)
       {
           /* child process or error*/
               sleep(1); /* make childs faster */
           break;//如果返回的是子进程，或者子进程创建失败，退出循环，否则子进程会再创建进程。
       }
  }

  if( pid< (pid_t) 0)
  {
          fprintf(stderr,"problems forking worker no. %d\n",i);
      perror("fork failed.");
      return 3;
  }

  if(pid== (pid_t) 0)
  {
    //pid=0是子进程，进行http请求
    if(proxyhost==NULL)
      benchcore(host,proxyport,request);
         else
      benchcore(proxyhost,proxyport,request);

         /* write results to pipe */
     f=fdopen(mypipe[1],"w");//mypipe[1]进行写入
     if(f==NULL)
     {
         perror("open pipe for writing failed.");
         return 3;
     }
     /* fprintf(stderr,"Child - %d %d\n",speed,failed); */
     fprintf(f,"%d %d %d\n",speed,failed,bytes);//向管道里写成功数，失败数，即服务器返回的字节数
     fclose(f);
     return 0;
  } else
  {
      //返回到父进程中
      f=fdopen(mypipe[0],"r");//mypipe[0]用于读取
      if(f==NULL) 
      {
          perror("open pipe for reading failed.");
          return 3;
      }
      setvbuf(f,NULL,_IONBF,0);
      speed=0;
          failed=0;
          bytes=0;
        /** 统计clients个进程的数据信息 ***/
      while(1)
      {//需要进行循环，否则的话，由于子进程的创建，很有可能会返回父进程。
          pid=fscanf(f,"%d %d %d",&i,&j,&k);//阻塞进行读取数据，等所有的子进程都返回再退出循环
          if(pid<2)
                  {//有进程成了僵尸进程，退出循环，不再进行统计
                       fprintf(stderr,"Some of our childrens died.\n");
                       break;
                  }
          speed+=i;//统计成功请求数
          failed+=j;//统计失败请求数
          bytes+=k;//统计服务器总共返回的字节数
          /* fprintf(stderr,"*Knock* %d %d read=%d\n",speed,failed,pid); */
          if(--clients==0) break;//成功读取一个客户端成功执行返回
      }
      fclose(f);

  printf("\nSpeed=%d pages/min, %d bytes/sec.\nRequests: %d susceed, %d failed.\n",
          (int)((speed+failed)/(benchtime/60.0f)),//每分钟的请求数
          (int)(bytes/(float)benchtime),//每秒服务器返回的字节数
          speed,//成功请求数
          failed);//失败请求数
  }
  return i;
}

void benchcore(const char *host,const int port,const char *req)
{
 int rlen;
 char buf[1500];
 int s,i;
 struct sigaction sa;

 /* setup alarm signal handler */
 sa.sa_handler=alarm_handler;
 sa.sa_flags=0;
 if(sigaction(SIGALRM,&sa,NULL))
    exit(3);
 alarm(benchtime);

 rlen=strlen(req);
 nexttry:while(1)
 {
    if(timerexpired)
    {//客户端执行时间到
       if(failed>0)
       {
          /* fprintf(stderr,"Correcting failed by signal\n"); */
          failed--;
       }
       return;//返回到调用benchcore()函数处
    }
    //创建连接到服务器的socket标识
    s=Socket(host,port);                          
    if(s<0) { failed++;continue;} 
    //向服务器发送rlen个http的请求字符
    if(rlen!=write(s,req,rlen)) {failed++;close(s);continue;}
    if(http10==0) 
        if(shutdown(s,1)) { failed++;close(s);continue;}
    if(force==0) 
    {
            /* read all available data from socket */
        while(1)
        {
              if(timerexpired) break; 
          //读取从服务器返回的数据
          i=read(s,buf,1500);
              /* fprintf(stderr,"%d\n",i); */
          if(i<0) 
              { //读取失败
                 failed++;
                 close(s);
                 goto nexttry;
              }
           else
               if(i==0) break;//读取完毕
               else
                   bytes+=i;//计算返回字节数
        }
    }
    if(close(s)) {failed++;continue;}
    speed++;
 }
}

socket.c

#include <sys/types.h>
#include <sys/socket.h>
#include <fcntl.h>
#include <netinet/in.h>
#include <arpa/inet.h>
#include <netdb.h>
#include <sys/time.h>
#include <string.h>
#include <unistd.h>
#include <stdio.h>
#include <stdlib.h>
#include <stdarg.h>

int Socket(const char *host, int clientPort)
{
    int sock;
    unsigned long inaddr;
    struct sockaddr_in ad;
    struct hostent *hp;

    memset(&ad, 0, sizeof(ad));
    ad.sin_family = AF_INET;

    inaddr = inet_addr(host);
    if (inaddr != INADDR_NONE)
        memcpy(&ad.sin_addr, &inaddr, sizeof(inaddr));
    else
    {
        hp = gethostbyname(host);
        if (hp == NULL)
            return -1;
        memcpy(&ad.sin_addr, hp->h_addr, hp->h_length);
    }
    ad.sin_port = htons(clientPort);

    sock = socket(AF_INET, SOCK_STREAM, 0);
    if (sock < 0)
        return sock;
    if (connect(sock, (struct sockaddr *)&ad, sizeof(ad)) < 0)
        return -1;
    return sock;
}

下面是使用webbench运行的一个测试