From: Jeff Dike <jdike@ka...> - 2003-02-01 00:11:47
> On Tue, Jan 28, 2003 at 06:01:27PM -0500, James Neal wrote:
> > Hmm.. How would you implement that? A hook in daemon_read and
> > daemon_write that calls a disconnect_to_switch (which would need to be
> > written) followed by a connect_to_switch?
> To be honest, i didn't think much over it yet ;-)
That's something I've been wanting to do for a while. It shouldn't be too
hard to separate the interface (eth0) from the underlying hardware (uml_switch
in this case) so that you can bring them up and down independently of each
If something tries sending out traffic when the interface isn't connected to
anything on the host, then it'll just sleep. I don't know if any traffic
is generated from interrupt context. If it is, then that would be a problem.
> But again, i didn't looked much yet at the existing code. [
> interesting, i rapidly saw that there is something already configured
> as SOCK_DGRAM ]
All data is sent over SOCK_DGRAM sockets. Doing it over stream sockets is
asking for trouble because you end up layering TCP on TCP, and having
interesting times when they independently start timing out and retransmitting.
From: Henrik Nordstrom <hno@ma...> - 2003-02-01 11:50:13
On Fri, 31 Jan 2003, Jeff Dike wrote:
> If something tries sending out traffic when the interface isn't connected to
> anything on the host, then it'll just sleep. I don't know if any traffic
> is generated from interrupt context. If it is, then that would be a problem.
It should behave like most normal network drivers.. if there is no link
reject the send request.