[PATCH] suggested fix for lingering stuff

[Marc Slemko <ma...@znep.com>]

I think that the below patch reflects what I would suggest doing to
fix up the lingering close stuff.

Untested, and we need someone with a SVR4 system (and a Unixware one
would be good... may also need to add USE_SHUTDOWN in the UW section
of conf.h) to do some testing and see if they can replicate the problem
with data being truncated (compiling without any of the three options)
and then see if USE_SHUTDOWN fixes it.

Comments?

Index: PORTING
===================================================================
RCS file: /home/marcs/archive/apache/cvs/apache/src/PORTING,v
retrieving revision 1.2
diff -c -r1.2 PORTING
*** PORTING	1996/12/17 23:19:15	1.2
--- PORTING	1997/01/19 19:20:58
***************
*** 205,210 ****
--- 205,225 ----
       USE_LONGJMP:
        use the longjmp() call instead of siglongjmp()
  
+      USE_LINGCLOSE:
+       Use Apache's soft "lingering close" feature to close
+       connections; designed to be a better behaved USE_SO_LINGER.  
+       Causes problems on some platforms.
+ 
+      USE_SHUTDOWN:
+       Call shutdown() before close()ing a connection.  Needed on
+       some broken (mainly SVR4) TCP stacks which don't send queued
+       data when a close() is done on a socket.
+      
+      USE_SO_LINGER:
+       Set the SO_LINGER option on sockets.  Possible alternative
+       to USE_LINGCLOSE, however some broken TCP stacks do not
+       implement a timeout, leaving hung connections around.
+ 
  --
  
    NO_*:
***************
*** 215,223 ****
        NO_SETSID:
        NO_USE_SIGACTION:
         Do not use the sigaction() call, even if we have it.
!       NO_LINGCLOSE:
!        Do not use Apache's soft, "lingering" close feature to
!        terminate connections.
  --
  
    MISC #DEFINES:
--- 230,236 ----
        NO_SETSID:
        NO_USE_SIGACTION:
         Do not use the sigaction() call, even if we have it.
! 
  --
  
    MISC #DEFINES:
Index: conf.h
===================================================================
RCS file: /home/marcs/archive/apache/cvs/apache/src/conf.h,v
retrieving revision 1.69
diff -c -r1.69 conf.h
*** conf.h	1997/01/10 13:23:31	1.69
--- conf.h	1997/01/19 19:13:37
***************
*** 285,291 ****
   * to 0x2000000 which is still PLENTY of space. I doubt if
   * even on heavy systems sbrk() would be called at all...
   */
- #define NO_LINGCLOSE
  
  #elif defined(SVR4)
  #define NO_KILLPG
--- 285,290 ----
***************
*** 298,303 ****
--- 297,303 ----
  #define getwd(d) getcwd(d,MAX_STRING_LEN)
  /* A lot of SVR4 systems need this */
  #define USE_FCNTL_SERIALIZED_ACCEPT
+ #define USE_SHUTDOWN
  
  #elif defined(UW)
  #define NO_KILLPG
Index: http_main.c
===================================================================
RCS file: /home/marcs/archive/apache/cvs/apache/src/http_main.c,v
retrieving revision 1.106
diff -c -r1.106 http_main.c
*** http_main.c	1997/01/12 20:38:12	1.106
--- http_main.c	1997/01/19 19:15:17
***************
*** 309,315 ****
   * before closing the connection.
   */
  
! #ifndef NO_LINGCLOSE
  static void lingering_close (int sd, server_rec *server_conf)
  {
      int dummybuf[512];
--- 309,315 ----
   * before closing the connection.
   */
  
! #ifdef USE_LINGCLOSE
  static void lingering_close (int sd, server_rec *server_conf)
  {
      int dummybuf[512];
***************
*** 360,366 ****
      shutdown (sd, 2);
      close (sd);
  }
! #endif
  
  void usage(char *bin)
  {
--- 360,366 ----
      shutdown (sd, 2);
      close (sd);
  }
! #endif /* USE_LINGCLOSE */
  
  void usage(char *bin)
  {
***************
*** 1667,1680 ****
  #endif		
  	bflush(conn_io);
  
! #ifdef NO_LINGCLOSE
  	bclose(conn_io);	/* just close it */
  #else
  	if (r)
  	    lingering_close (conn_io->fd, r->server);
  	else
  	    close (conn_io->fd);
! #endif	
      }    
  }
  
--- 1667,1686 ----
  #endif		
  	bflush(conn_io);
  
! #ifndef USE_LINGCLOSE
! #ifdef USE_SHUTDOWN
! 	/* broken (mostly SVR4) TCP stacks discard queued data if a
! 	 * close() is done before a shutdown().
! 	 */
! 	shutdown(conn_io->fd, 2); 
! #endif
  	bclose(conn_io);	/* just close it */
  #else
  	if (r)
  	    lingering_close (conn_io->fd, r->server);
  	else
  	    close (conn_io->fd);
! #endif
      }    
  }
	  

Re: [PATCH] suggested fix for lingering stuff

[Dean Gaudet <dg...@arctic.org>]

+1

Dean