dhcpcd-discuss

On 07/08/2014 09:46 PM, Roy Marples wrote:
> Two subsequent patches, config specific:
> http://roy.marples.name/projects/dhcpcd/ci/13cee734cc603613d5c1975984ba4e0af69982f4?sbs=0
>
> http://roy.marples.name/projects/dhcpcd/ci/ec6da39906241bf4dcc992e864889356503dfef7?sbs=0

This all does seem to work. I'm still waiting for the first lease 
renewal to see how that goes.

I am just following the head of the archive and doing updates for this 
testing. Whatever was in the archive last night (7/16/2014) is 
"ia_pd_mix works for me and comcast".

Relevant Config (intX interfaces are brctl bridges, ext0 is the cable 
modem segment):

interface ext0
iaid 1
noipv4ll
ia_na 1
ia_pd 2/::/60 int0/3/64 int1/4/64
ia_pd_mix

# necessary plumbing, but don't want
# dhcpcd to step on the IPv4 NAT setup
# nor are there any IPv6 servers to answer
# on the intX interfaces.
interface int1
noipv4
noipv6

interface int0
noipv4
noipv6


===

Stupid System Trick:
I was having some issues with boot order.

If your clients are going to use the ia_pd and name a bridge as one of 
the target interfaces (e.g. "ia_pd ext0 ... br0/2/64" ) and that bridge 
is going to only stock wireless devices (a la hostapd) the dhcpcd stuff 
may try to configure the bridge before it has a non-zero MAC address. 
Hostapd can take a relatively long time to initialize the wireless card 
and add it to the bridge, all of which happens after the startup script 
returns. This may not work. ...

Solution: Add a dummy interface (e.g. "modprobe dummy; brctl addif br0 
dummy0" or equiv) to the bridge unconditionally at initial bridge setup. 
The MAC address will be random, but it will be there and be workable.

[Note that this has no bearing on previous testing done by me because I 
was invoking dhcpcd by hand in testing and long after boot when the 
bridge was nice and stable. It was also one of several targets. I 
mention this conidition here because once I did boot integration I was 
getting dhcpcd messages about skipping the interface that were not 
dhcpcd's fault.]

-- Rob.