dhcpcd-discuss

On Wed, 25 May 2016 21:25:44 -0400
"Neal P. Murphy" <neal.p.murphy@xxxxxxxxxxxx> wrote:

> On Thu, 26 May 2016 00:28:12 +0100
> Roy Marples <roy@xxxxxxxxxxxx> wrote:
> 
> > On 2016-05-25 23:32, Neal P. Murphy wrote:
> > > I'm playing with getting Smoothwall Express to use IPv6. And not
> > > getting anywhere very quickly. But that's partly due to my newness to
> > > IPv6.
> > > 
> > > Something is weird.
> > > 
> > > With 6.10.3 and 6.11.0, any of these lines in the .conf:
> > > ----
> > > ia_pd 0b:17:f2:7d/::/60 eth0/1 eth2/3
> > > ia_pd 0b:17:f2:7d/::0/60 eth0/1 eth2/3
> > > ia_pd 0b:17:f2:7d//60 eth0/1 eth2/3
> > > ----
> > > 
> > > causes
> > > ----
> > > ...
> > > DUID 00:01:00:01:1b:ff:1d:50:00:90:0b:17:f2:7d
> > > eth0: IAID 0b:17:f2:7a
> > > eth2: IAID 0b:17:f2:7c
> > > eth3: IAID 0b:17:f2:7d
> > > eth3: delaying IPv6 router solicitation for 0.8 seconds
> > > eth3: reading lease `/var/db/dhcpcd-eth3.lease6'
> > > eth3: soliciting a DHCPv6 lease
> > > eth3: delaying SOLICIT6 (xid 0x32dd0d), next in 0.9 seconds
> > > eth3: soliciting an IPv6 router
> > > *** Error in `dhcpcd': malloc(): memory corruption: 0x09c8fb50 ***
> > > ----
> > > 
> > > And dhcpcd must be killed via 'kill -9 PID'.
> > > 
> > > Since others seem to use the syntax without trouble, is it
> > > possible--or probable--that I'm building dhcpcd wrong?
> > > 
> > > As to the coredump I mentioned in the subject, *now* I can't reproduce
> > > it. I can send the coredump and unstripped binary (for 6.10.3), if
> > > desired.
> > 
> > Yes please.
> > The core dump and the binary will tell us where the problem is.
> > 
> > dhcpcd-6.11.0 did fix some potential memory issues you may have found 
> > with dhcpcd-6.10.x which would explain why you cannot replicate it.
> 
> Trouble is, I couldn't reproduce the coredump on 6.10.3, either.
> 
> Attached is a tarball of several coredumps and unstripped dhcpcd from 6.10.3 and 6.11.0. I think the coredumps are all from 6.10.3. I included the 6.11.0 just in case.
> 
> I'm unable to provide coredumps for the memory corruption. I can't get the program to dump core ; neither <CTRL/\> nor SIGSEV work. Is there another way to obtain a process' core? ... Oh, duh. I might be able to gdb the program and get a stack trace that way.
> 
> N

I think I forgot to attach....

Attachment: dhcpcd-prog-core.tar.gz
Description: application/gzip