Oh, I didn't know that the transport was supposed to stay open. Anyway, the
way the workaround in current mainline works is good enough for my purposes
- thanks!
JB
Dne 29.4.2015 19:23 "Jiri Svoboda" <[email protected]> napsal(a):

> Hi Jan,
>
> the problem you are describing is the subject of
> http://trac.helenos.org/ticket/626
>
> the problem is in udp/sockets which currently cannot handle two DHCP
> associations being bound to the same port albeit on two different
> interfaces. Your patch just works around that problem. You could also
> probably get DHCP failure on one of the links and it might only succeed on
> another attempt (since the links are configured in parallel, at least I
> think so).
>
> DHCP is supposed to keep the transport open all the time, since it needs
> to monitor link state and renew the lease after some time - not implemented
> yet.
>
> I am currently working towards fixing ticket 626. But since the problem is
> partly in the sockets implementation, which is 'legacy' code, I started by
> doing a cleanup of that area (see
> https://code.launchpad.net/~jsvoboda/helenos/nettlapi).  It will still
> take me some time to finish that (2-3 weeks? depending on how much time I
> can spare). I will address 626 as part of that work.
>
> I improved your patch so that the finalization of the transport is
> correctly acted upon (so no error message is printed) and commited to
> mainline. Hopefully it can help you until the time I can implement the
> proper solution.
>
> Cheers,
> Jiri
>
> ---------- Původní zpráva ----------
> Od: Jan Buchar <[email protected]>
> Komu: HelenOS development mailing list <[email protected]>
> Datum: 29. 4. 2015 14:25:46
> Předmět: [HelenOS-devel] DHCP on multiple interfaces
>
> Greetings,
>
> while working on my bachelor's thesis (HelenOS packet filter), I noticed
> that
> on a machine with two NICs DHCP fails to configure the second one. I found
> that this is caused by dhcp_transport_fini() not being called after
> finishing
> negotiation. Because the address is still in use, the second negotiation
> fails
> to bind the address.
>
> I attached a patch for this. However, this causes dhcp_recv_msg() to end
> with
> an error, because the socket it wants to read from is closed. The
> configuration seems to work fine, though. Is there a better place for the
> dhcp_transport_fini()? Or should I fix this some other way?
>
> JB_______________________________________________
> HelenOS-devel mailing list
> [email protected]
> http://lists.modry.cz/listinfo/helenos-devel
>
>
> _______________________________________________
> HelenOS-devel mailing list
> [email protected]
> http://lists.modry.cz/listinfo/helenos-devel
>
>
_______________________________________________
HelenOS-devel mailing list
[email protected]
http://lists.modry.cz/listinfo/helenos-devel

Reply via email to