[Xorp-users] rip on freebsd

[John Hay]

On Tue, May 12, 2009 at 08:26:06AM +0100, Bruce Simpson wrote:
> Bruce Simpson wrote:
> >I will dig up this script and modify it to use ktrace instead of gdb; it 
> >is essential that the problem gets pinned down to a particular set of 
> >socket syscalls.
> >  
> 
> Here's that script. This is intended for trapping the creation of a XORP 
> child process and hooking GDB up to it, I think I wrote it to debug a 
> DoS in the XRL input path, it is probably in Bugzilla somewhere also. As 
> written here, it is intended for use under Gentoo Linux.
...
> 
> Using 'kdump -E -r -f /tmp/xorp_fea.ktrace' should shed more light on 
> what syscalls the FEA is invoking when the multicast socket failure is 
> encountered. kdump knows how to decode some setsockopt() calls in a 
> human-readable format, but not all.
> 
> Alternatively, you could just ktrace xorp_rtrmgr and all children; this 
> will not race xorp_fea's creation, but will produce LOTS of output. You 
> could also hack the FEA's template file to run xorp_fea under ktrace 
> directly, which is how I usually ran valgrind to find various QA issues 
> in XORP.

I tried an easier (for me) way. I moved xorp_fea to xorp_fea-bin and
replaced it with a script that just did:
exec ktrace -f /tmp/xorp_fea.ktrace /usr/local/fea/xorp_fea-bin $*

I have put xorp.conf, xorp_fea.ktrace.gz, xorp_fea.kdump.gz and
xorp_rtrmgr.log on http://people.freebsd.org/~jhay/xorp/

BTW I found that I do not need a default route for rip to work,
"route add 224.0.0.0/4 -interface lo0" is enough.

John
-- 
John Hay -- John.Hay at meraka.csir.co.za / jhay at FreeBSD.org