dhcpcd-discuss

On 14/11/2017 20:34, Neal P. Murphy wrote:
> On Tue, 14 Nov 2017 19:11:54 +0000
> Roy Marples <roy@xxxxxxxxxxxx> wrote:
>
> > Hi Neal
> >
> > On 14/11/2017 19:02, Neal P. Murphy wrote:
> > > Here are the results from this morning.
> > >     - eth3-fail.log		output of dhcpcd with the output of my rc.update6red script
> > >     - eth3-fail.tcpdump		tcpdump pcap file
> > >     - ip-addr_1-pre.log		addrs before attempt
> > >     - ip-addr_2-up.log		addrs after dhcpcd started
> > >     - ip-addr_3-down.log		addrs after dhcpcd stopped
> > >     - dhcpcd6.conf		config file I used
> > >     - eth3-fail_my-debug.log	detailed ENV data from rc.update6red (might be helpful)
> > >
> > > I've noticed that if I start with plain '::/60' in the .conf, it seems to work; but if I use the last delegated /60, it doesn't. Could it be that comcrash doesn't like customers requesting addresses for expired leases? (Which is probably why I had that one-year lease in the conf before, thinking it would get comcrash to keep the lease alive longer.)
> >
> > Yes, that makes sense.
> > Well, it doesn'r make sense to me, but it can explain what you are
> > seeing I guess. Still, poor form from Comcast.
> > Have you considered asking their support about it?
> >
> > > I also noticed that with and without 'persistent', eth3's /64 addr is not removed, but the /128 is removed as are the /64s from the other two active IFs.
> >
> > This should be fixed in the git HEAD.
> > Are you still seeing this issue there?
>
> I did a git clone to build the prog I'm running. First commit from git log is bdbe98f8fc6e4e8188c1c3a0da66abba93a1cefd. The md5sums of the built (stripped) binary matches the one I'm using. ... Stupid eejit; I neglected to comment out '/usr/sbin/dhcpcd ...', with './dhcpcd ...' on the next line; script was still using RC3 and not HEAD. I'll try again. I really *am* acting like a noob; you'd think after 35 years I'd be able to do something *that* simple right. Sigh.
>
> OK, now using HEAD (well, three commits shy of HEAD).

Simple mistakes, but as long as it's corrected :)

> First run, I saw individual DELEGATED6 calls for eth0 and eth2. Next runs, I see only DELEGATED6 for eth2. Hmmm. Odd. Sometimes I see one for each eth0 and eth2, sometimes only for eth2, sometimes none. And never for eth3, even though it does get a delegated prefix.
>
> HEAD doesn't remove the /64 from eth3, either. But it does remove the /128.

I'll try and look into this tomorrow, by replicating your env somehow.

> Is there a way to build with 'extra instrumentation' so you can see what dhcpcd's doing?

There is only the debug flag.

Roy