This is the mail archive of the ecos-discuss@sources.redhat.com mailing list for the eCos project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: RedBoot sequencenumber generation


On Dienstag, 16. März 2004 18:05, Mark Salter wrote:
> >>>>> Roland =?iso-8859-15?q?Ca=DFebohm?= writes:
> >
> > Hello Mark,
> > thanks for helping.
> >
> > In most cases it works, but if the initial sequencenumber, which
> > is now a pseudo random number, is anyhow in the window of the
> > TIME_WAIT connections of the host I get the old problem.
> > The SYN packet don't get a responce and will be retransmitted
> > on and on with the same sequencenumber.
>
> Retransmitting the same seqnum with the SYN is okay. Interestingly,
> the RFCs state that while in the TIME_WAIT state, the server refuses
> connection requests using the same tcp socket pair. In the tcpdump
> you sent, the second connection uses a different socket pair (port
> 7800 vs 7801). Some BSD stacks are more stringent and refuse to reuse
> the local port number while in TIME_WAIT. It looks like this is the
> problem you are running into.
>
> However, there is hope. Most BSD implementations allow a new
> connection in TIME_WAIT iff the new seqnum is greater than the
> final seqnum of the connection in TIME_WAIT. So, here's the same
> patch, but with an extra manipulation of initial_seqnum when an
> active close FIN is received from the server.
>

Going back I think the not working connection was after a reboot
of RedBoot. I tried multiple times and it worked and then I reseted
RedBoot. The next try wasn't working. Then I think it is the same
socket pair of a connection in TIME_WAIT used before rebooting.

So at least after a reboot it is possible to come in this state,
there must be a way out (I hope).

Roland

> --Mark
>
>
> Index: redboot/current/src/net/tcp.c
> ===================================================================
> RCS file: /cvs/ecos/ecos/packages/redboot/current/src/net/tcp.c,v
> retrieving revision 1.11
> diff -u -p -5 -r1.11 tcp.c
> --- redboot/current/src/net/tcp.c	21 Dec 2003 13:17:52 -0000	1.11
> +++ redboot/current/src/net/tcp.c	16 Mar 2004 17:02:20 -0000
> @@ -58,10 +58,12 @@
>  #include <cyg/hal/hal_if.h>
>
>  #define MAX_TCP_SEGMENT (ETH_MAX_PKTLEN - (sizeof(eth_header_t) +
> sizeof(ip_header_t))) #define MAX_TCP_DATA    (MAX_TCP_SEGMENT -
> sizeof(tcp_header_t))
>
> +/* setting this to 1 is technically wrong, but BSD does it too... */
> +static int initial_seqnum = 1;
>
>  /* sequence number comparison macros */
>  #define SEQ_LT(a,b) ((int)((a)-(b)) < 0)
>  #define SEQ_LE(a,b) ((int)((a)-(b)) <= 0)
>  #define SEQ_GT(a,b) ((int)((a)-(b)) > 0)
> @@ -479,10 +481,11 @@ __tcp_handler(pktbuf_t *pkt, ip_route_t
>                        return;
>                    }
>                    s->state = _ESTABLISHED;
>                    s->ack = ntohl(tcp->seqnum) + 1;
>                    s->seq = ntohl(tcp->acknum);
> +                  initial_seqnum += 64000;
>  		  __timer_cancel(&s->timer);
>                    send_ack(s);
>  		break;
>
>  	      case _LISTEN:
> @@ -523,10 +526,11 @@ __tcp_handler(pktbuf_t *pkt, ip_route_t
>  		} else if ((tcp->flags & TCP_FLAG_ACK) &&
>  			   ntohl(tcp->acknum) == (s->seq + 1)) {
>  		    /* we've established the connection */
>  		    s->state = _ESTABLISHED;
>  		    s->seq++;
> +                    initial_seqnum += 64000;
>
>  		    BSPLOG(bsp_log("ACK received - connection established\n"));
>  		}
>  		break;
>
> @@ -543,10 +547,12 @@ __tcp_handler(pktbuf_t *pkt, ip_route_t
>
>  		    BSPLOG(bsp_log("FIN received - going to _CLOSE_WAIT\n"));
>
>  		    s->ack++;
>  		    s->state = _CLOSE_WAIT;
> +
> +                    initial_seqnum = ntohl(tcp->seqnum) + 64000;
>  		}
>  		/*
>  		 * Send an ack if neccessary.
>  		 */
>  		if (s->ack != ack || pkt->pkt_bytes > (tcp->hdr_len << 2))
> @@ -643,10 +649,13 @@ void
>  __tcp_poll(void)
>  {
>      __enet_poll();
>      MS_TICKS_DELAY();
>      __timer_poll();
> +
> +    /* rfc793 says this should be incremented approx. once per 4ms */
> +    initial_seqnum += 1000 / 4;
>  }
>
>
>  int
>  __tcp_listen(tcp_socket_t *s, word port)
> @@ -890,11 +899,11 @@ __tcp_open(tcp_socket_t *s, struct socka
>      s->his_port = host->sin_port;
>      s->pkt.buf = (word *)s->pktbuf;
>      s->pkt.bufsize = ETH_MAX_PKTLEN;
>      s->pkt.ip_hdr  = (ip_header_t *)s->pkt.buf;
>      s->pkt.tcp_hdr = (tcp_header_t *)(s->pkt.ip_hdr + 1);
> -    s->seq = (port << 16) | 0xDE77;
> +    s->seq = initial_seqnum;
>      s->ack = 0;
>      if (__arp_lookup((ip_addr_t *)&host->sin_addr, &s->his_addr) < 0) {
>          diag_printf("%s: Can't find address of server\n", __FUNCTION__);
>          return -1;
>      }

-- 

___________________________________________________

VS Vision Systems GmbH, Industrial Image Processing
Dipl.-Ing. Roland Caßebohm
Aspelohe 27A, D-22848 Norderstedt, Germany
Mail: roland.cassebohm@visionsystems.de
http://www.visionsystems.de
___________________________________________________


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


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]