dhcpcd-discuss

On 07/01/2014 02:24 PM, Roy Marples wrote:
> This has now been implemented.
> http://roy.marples.name/projects/dhcpcd/ci/11451d70930a170b2ebaf36c28ca718f33d426e5?sbs=0
>
> http://roy.marples.name/projects/dhcpcd/ci/49bc4c095d5f5a7dc49da75ad941097bd5551d0e?sbs=0
>
>
> May not apply cleanly to dhcpcd-6.4.0, you may need to get a fossil
> snapshot.
>
> Could do with some testing. I tested it with isc-dhcpd-4.2.5-P1 and it
> seemed to work fine.
> Let me know how it works for you and ComCast!

It _almost_ works....

Comcast, it seems, honors the multiple requests in one packet, but it 
replies with one packet per IA, each with the same transaction ID etc.

I believe that the second and subsequent packets with the same xid are 
being filtered or ignored. Whichever IA request comes first in the 
config file is the one honored so I'm guessing the response packets are 
being generated stepwise by the server instead of cumulatively.

It looks like you "clear the status" (and free the data structure) after 
a successful reply is parsed. Some sort of keep/more flag in the 
ifo->ia[] might be a work around. You'd hang on to the status buffer and 
until all ia(s) were satisfied or until it was time to make a new 
request. The flag would then also act as a de-dupe for repeated replies 
for the same iaid in the same transaction.

A three-packet tcpdump is attached.

Rob "loves a good hole in a standard" White.

Attachment: tf.dump
Description: Binary data