On Wed, Dec 18, 2002 at 07:09:05PM +0100, jochen wrote:
> On Wed, Dec 18, 2002 at 11:58:22AM -0500, Jan Harkes wrote:
> > [ ... lots about binding specific interfaces or not ... ]
> To work around such problems, many servers bind every single IP instead
> of listening on 0.0.0.0 as the coda servers do. If the server binds
> every single IP, it is as reachable is if it had bound to 0.0.0.0, but
> the server can determin to which IP addr the packet was send and it can
> decide with which IP addr to answer.

But it also makes things a bit difficult. For instance, the server still
needs to bind to 0.0.0.0 or else it would miss packets on interfaces
that came up after the server was started. The server also has to deal
with interfaces that are brought down.

When you're not responding to a request, but initiating a transaction
(server->server during resolution, or server->client while setting up
the callback connection) you have to make a decision which socket to
use. So in a way the application suddenly has to know a whole lot about
how your network is routed etc.

> > createvol_rep should be run on the SCM, and 'volutil' without a hostname
> > connects to localhost. i.e. the server on SCM.
> localhost resolves here to 192.168.0.1 while he rpc is sending from
> 192.168.2.1... thus it fails

AFAIK localhost should be 127.0.0.1.

> > >   * cfs mkmount works but it segfaults and locks the whole system. After
> > >     rebooting, the mountpoint exists and everything is fine
> > 
> > Never had that happen to me. What linux kernel are you using?
>
> debian/sid, linux-2.4.0 + cryptoapi-0.1.0 + rsbac-1.2.1 (rsbac.org) +
> super-freeswan-1.99-kb2 (www.freeswan.ca/code/super-freeswan/). Note
> that this happens while running rsbac in softmode (i.e. the access
> restrictions don't apply) and the rsbac system isn't complaining about
> some security violations)

2.4.0 or 2.4.20? I didn't think anyone ever used 2.4.0, it was a bit buggy.

Jan