[Olsr-dev] 0.5.6-rc5 is out

[John Hay]

Ok, I have not found the bug yet, but here are 2 patches to help the
debugging a bit. The first is to make "HNA Network" timer message
work and the second is to fix the printf in prefix_to_netmask().

If anyone is interrested, I have left the output of olsrd -nofork -d 3
at: ftp://ftp2.za.freebsd.org/pub/local/screenlog.0.bz2

It lost a few HNA entries, but not all yet. One to look for is:
2001:4200:7000:4

> > >
> > > it would be interesting to get some datapoints on the timer
> > > manipulations for all HNA Network entries.
> > >
> > > i have added a patch to http://gredler.at/hg/olsrd/
> > > that adds the string "HNA Network" to all HNA Network related
> > > timers.
> 
> I pulled the tip.tar.gz from there and it extracts as olsrd-f3301101db38.
> Doing a grep for "HNA Network" I only found one match in hna_set.c
> 
> > > can you compile a DEBUG image (see Makefile.inc) and run it into
> > > your network and grep for "HNA Network" timer manipulations.
> 
> I have compiled it with DEBUG on and am running it with
> "olsrd -nofork -d 3", but do not see any lines with "HNA Network" in
> it, not even with "Network" in it. Lots lines that start with TIMER:
> and a lot of them "fire unknown timer".
> 
> > > >> We just found a bug in the windows code path of 0.5.6rc5, so windows
> > > >> testers should use the recent mercurial version
> > > >>
> > > >> http://gredler.at/hg/olsrd/rev/fddb68abd9b7
> > > >
> > > > Ok, I have tried this on a Wrap (x86) board with FreeBSD 7.0-stable
> > > > and IPv6, like always. :-)
> > > >
> > > > What I see is that everything looks ok after I start olsrd. It pick
> > > > up the routes and populate the kernel routing table. And then over
> > > > a period of several hours all the HNA routes will be removed from
> > > > the kernel. I wrote a little script to monitor it. Sometimes only
> > > > one route is removed and sometimes a few at a time. Sometimes it
> > > > will be hours where no HNA entry will be removed. I have the dot_
> > > > draw plugin loaded and when I connect to it, all the routes are
> > > > there. :-) It does not happen to the other routes. I have quite a
> > > > few other boxes that run 0.5.5 and have never noticed it on them.

John
-- 
John Hay -- (spam-protected) / (spam-protected)


--- src/hna_set.c.orig	2008-06-13 20:32:42.000000000 +0200
+++ src/hna_set.c	2008-06-14 19:53:19.000000000 +0200
@@ -277,7 +277,7 @@
   olsr_set_timer(&net_entry->hna_net_timer, vtime,
                  OLSR_HNA_NET_JITTER, OLSR_TIMER_ONESHOT,
                  &olsr_expire_hna_net_entry, net_entry,
-                 hna_net_timer_cookie);
+                 hna_net_timer_cookie->ci_id);
 }
 
 

--- src/ipcalc.c.orig	2008-06-13 20:32:42.000000000 +0200
+++ src/ipcalc.c	2008-06-14 13:00:52.000000000 +0200
@@ -45,6 +45,7 @@
 {
 #if !defined(NODEBUG) && defined(DEBUG)
   struct ipaddr_str buf;
+  const olsr_u8_t *a_start = a;
 #endif
   int p;
   const olsr_u8_t *a_end;
@@ -62,7 +63,7 @@
   }
 
 #ifdef DEBUG
-  OLSR_PRINTF(3, "Prefix %d = Netmask: %s\n", prefixlen, inet_ntop(olsr_cnf->ip_version, a, buf.buf, sizeof(buf.buf)));
+  OLSR_PRINTF(3, "Prefix %d = Netmask: %s\n", prefixlen, inet_ntop(olsr_cnf->ip_version, a_start, buf.buf, sizeof(buf.buf)));
 #endif
   return 1;
 }