[ECOS] maybe a flaw in tcp_output() in ecos

[张 亮 <johnsonest at hotmail dot com>]

Hi,
  I feel another strange point(maybe a flaw) in tcp_output() function:
when we send a SYN message with MSS option,we will call tcp_mss()to 
calculate mss of this connection,in this function,tp->maxopt will be set as 
ifp->mtu-sizeof(struct tcpiphdr):(in tcp_input.c and tcp_mss())
              mss = rt->rt_rmx.rmx_mtu - sizeof(struct tcpiphdr);
              ....
              tp->t_maxopd = mss;
after call of tcp_mss(),and in tcp_output(),we will calculate length of 
data which will be sent like this:
              if (len > tp->t_maxopd - optlen) {
		len = tp->t_maxopd - optlen;
                  ...
the FLAW is that:
              if in ip_output(),we insert IP options into packet,whole 
packet length will larger than MTU of selected  interface,so fragment will 
be needed in IP layer in itself host,but as we know,tcp always trys to 
avoid frangmenting!
I am afraid that it's a flaw,maybe two familar functions should  be writen 
to meet tcp_output() and tcp_input(),right?


BRS/johnsonest      
 




_________________________________________________________________
与联机的朋友进行交流，请使用 MSN Messenger:  http://messenger.msn.com/cn  


--
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss