On Thu, May 2, 2013 at 10:55 AM, Jiri Svoboda <[email protected]>wrote:

> Hi Jana,
>
> I am afraid that you overestimate the completeness of the network stack in
> HelenOS slightly. We have a stub setsockopt() function, but it does not do
> anything (only returns ENOTSUP).
>

Thanks for clarification. In the last revision (mainline 1817),
setsockopt() seems to do some calls into HelenOS IPC subsystem which
confused me.

>
> I am not familiar with any existing DHCP implementation, so pardon my
> ignorance. What exactly do we need SO_REUSEADDR for?
>

SO_REUSEADDR should be used so that multiple sockets could bind to the same
UDP port. It probably isn't necessary, but every DHCP implementation I've
ever seen uses it.

>
> The bad news is that IP broadcasting is not implemented at all. In other
> words, if you try to send to an address such as 192.168.255.255, it will
> try to discover that literal address via ARP. If you try to send to
> 255.255.255.255, inet will most likely decide that no direct route to such
> address exists and will send the datagram to the default router only.
> Improvements to ethip and inet will be necessary.
>

That's the next thing I have been trying to figure out, thanks!

>
> Also, I believe DHCP needs special provisions in order to be able to send
> and receive UDP datagrams on an interface that has no address configured
> (bind to an interface as opposed to a subnet). Modifications to inet and
> udp will probably be necessary.
>

As far as I know, DHCP needs some special treatment, but I'm not sure I
understand you. Do you mean something like
setsockopt(SOL_SOCKET, SO_BINDTODEVICE, device) in Linux?

Cheers,
Jana
_______________________________________________
HelenOS-devel mailing list
[email protected]
http://lists.modry.cz/cgi-bin/listinfo/helenos-devel

Reply via email to