RE: slapd abort in (ITS#1655)

[hyc@highlandsun.com]

The suggested change looks good to me, I've committed it to HEAD.
No visible difference in the test suite, but we don't have a script
for this kind of activity.

  -- Howard Chu
  Chief Architect, Symas Corp.       Director, Highland Sun
  http://www.symas.com               http://highlandsun.com/hyc
  Symas: Premier OpenSource Development and Support 

> -----Original Message-----
> From: owner-openldap-bugs@OpenLDAP.org
> [mailto:owner-openldap-bugs@OpenLDAP.org]On Behalf Of akeller@inform.ch
> Sent: Wednesday, March 20, 2002 10:50 AM
> To: openldap-its@OpenLDAP.org
> Subject: slapd abort in (ITS#1655)
> 
> 
> Full_Name: Alfred Keller
> Version: 2.0.23
> OS: Win 2000
> URL: ftp://ftp.openldap.org/incoming/
> Submission from: (NULL) (195.134.131.162)
> 
> 
> When a slapd client opens and closes connections to slapd in 
> short intervals,
> then slapd frequently aborts at the following code line:
> 
> ber_sockbuf_ctrl( Sockbuf *sb, int opt, void *arg )
> {
> 	Sockbuf_IO_Desc		*p;
> 	int			ret = 0;
> 
> 	assert( sb != NULL );                <<<<<<<<<<<<<<< here
> 	assert( SOCKBUF_VALID( sb ) );
> --- snip ----
> 
> The calling function is:
> 
> static Connection* connection_get( ber_socket_t s )
> {
>   ....
> 
>   for(i=0; i<dtblsize; i++) {
>     ber_sockbuf_ctrl( connections[i].c_sb,            
> 			LBER_SB_OPT_GET_FD, &sd );     <<<<<<<< called here
> 	
>     if( connections[i].c_struct_state == SLAP_C_UNINITIALIZED ) {
>       assert( connections[i].c_conn_state == SLAP_C_INVALID );   
>       assert( connections[i].c_sb == 0 );               <<<<<<<< 
> look here !
>       break;
>     }
> ----- snip -----
> 
> Something is wrong here because at 'look here' you see that under some
> conditions connections[i].c_cb is expected to be 0, but in the call to
> ber_sockbuf_ctrl a few lines above the value of 
> connections[i].c_cb may never be
> 0??? (assert in ber_sockbuf_ctrl).
> 
> By the way, our application runs fine when we call 
> ber_sockbuf_ctrl *after* the
> 'if'-block in function connection_get. Be we can't say if this is 
> the correct
> way to fix the problem.