inet_ntoa的實現

從一段代碼開始:

[cpp]
view plaincopyprint?

1. int test_ntoa() 
   
2. { 
3.     struct sockaddr_in recv = {0}; 
4.     struct sockaddr_in sa   = {0}; 
   
5.     recv.sin_addr.s_addr = -217732928;
   //192.168.5.243   
6.     sa.sin_addr.s_addr   = -939415360; //192.168.5.200 
7.     printf("ip of recv:%s ip of sa:%s\n", inet_ntoa(recv.sin_addr), inet_ntoa(sa.sin_addr)); 
8.     printf("ip of recv:%s\n", inet_ntoa(recv.sin_addr)); 
9.     printf("ip of sa  :%s\n", inet_ntoa(sa.sin_addr));  
10.     return 0;    
11. } 

int test_ntoa()<br />{<br />struct sockaddr_in recv = {0};<br />struct sockaddr_in sa = {0};<br />recv.sin_addr.s_addr = -217732928; //192.168.5.243<br />sa.sin_addr.s_addr = -939415360; //192.168.5.200<br />printf("ip of recv:%s ip of sa:%s\n", inet_ntoa(recv.sin_addr), inet_ntoa(sa.sin_addr));<br />printf("ip of recv:%s\n", inet_ntoa(recv.sin_addr));<br />printf("ip of sa :%s\n", inet_ntoa(sa.sin_addr));<br />return 0;<br />}

結果有點詫異:

[cpp]
view plaincopyprint?

1. ip of recv:192.168.5.243 ip of sa:192.168.5.243 
   
2. ip of recv:192.168.5.243 
3. ip of sa   :192.168.1.200 

ip of recv:192.168.5.243 ip of sa:192.168.5.243<br />ip of recv:192.168.5.243<br />ip of sa :192.168.1.200

C函數出現這類問題,多半是由於其實現使用了靜態記憶體所知.

google其源碼:

[cpp]
view plaincopyprint?

1. char * 
2. inet_ntoa (struct in_addr in) 
   
3. { 
4.   __libc_once_define (static, once); 
5.   char *buffer; 
   
6.   unsigned char *bytes; 
7.  
8.   /* If we have not yet initialized the buffer do it now.  */ 
9.   __libc_once (once, init); 
10.  
11.   if (static_buf != NULL) 
12.     buffer = static_buf; 
13.   else 
14.     { 
15.       /* We don't use the static buffer and so we have a key.  Use it
16.      to get the thread-specific buffer.  */ 
17.       buffer = __libc_getspecific (key); 
18.       if (buffer == NULL) 
19.     { 
20.       /* No buffer allocated so far.  */ 
21.       buffer = malloc (18); 
22.       if (buffer == NULL) 
23.         /* No more memory available.  We use the static buffer.  */ 
24.         buffer = local_buf; 
25.       else 
26.         __libc_setspecific (key, buffer); 
27.     } 
28.     } 
29.  
30.   bytes = (unsigned char *) ∈ 
    
31.   __snprintf (buffer, 18, "%d.%d.%d.%d", 
32.           bytes[0], bytes[1], bytes[2], bytes[3]); 
33.  
34.   return buffer; 
35. } 

char *<br />inet_ntoa (struct in_addr in)<br />{<br /> __libc_once_define (static, once);<br /> char *buffer;<br /> unsigned char *bytes;</p><p> /* If we have not yet initialized the buffer do it now. */<br /> __libc_once (once, init);</p><p> if (static_buf != NULL)<br /> buffer = static_buf;<br /> else<br /> {<br /> /* We don't use the static buffer and so we have a key. Use it<br /> to get the thread-specific buffer. */<br /> buffer = __libc_getspecific (key);<br /> if (buffer == NULL)<br />{<br /> /* No buffer allocated so far. */<br /> buffer = malloc (18);<br /> if (buffer == NULL)<br /> /* No more memory available. We use the static buffer. */<br /> buffer = local_buf;<br /> else<br /> __libc_setspecific (key, buffer);<br />}<br /> }</p><p> bytes = (unsigned char *) ∈<br /> __snprintf (buffer, 18, "%d.%d.%d.%d",<br /> bytes[0], bytes[1], bytes[2], bytes[3]);</p><p> return buffer;<br />}
這裡使用了靜態緩衝buffer,函數把轉換後的ip寫入緩衝buffer,然後返回buffer的首地址值.

看似無奈的選擇.glibc裡的某些底層函數都使用這類處理方式.

其實也可以選擇外部傳參的做法解決:

[cpp]
view plaincopyprint?

1. <pre name="code"class="cpp">char * 
2. inet_ntoa (struct in_addr in,
   char* buf, int buflen) 
3. { 
4.       ... ... 
5. } 

<pre name="code" class="cpp">char *<br />inet_ntoa (struct in_addr in, char* buf, int buflen)<br />{<br /> ... ...<br />}

這樣的處理方式,效率略低於第一種.或許這也是glibc選擇第一種方式的原因.