[Pkg-utopia-maintainers] Bug#475188: using NetworkManager /etc/network/if-up.d/openssh-server does not reload

[Colin Watson]

reassign 475188 network-manager
thanks

On Thu, Apr 10, 2008 at 08:44:13AM +0200, Jordi Pujol wrote:
> El Wednesday 09 April 2008 18:30:15 Colin Watson va escriure:
> > On Wed, Apr 09, 2008 at 05:44:46PM +0200, Jordi Pujol wrote:
> > > using NetworkManager, the script /etc/network/if-up.d/openssh-server does
> > > not reload, because it is called with the variable ADDRFAM equal
> > > to "NetworkManager"
> >
> > That makes no sense. In what way is NetworkManager an address family?
> 
> Both we are sure of that is not very coherent, but a Debian testing release 
> calls the programs in the directory if-up.d with this value when the 
> interfaces are started from KDE Network Manager.
> in this case the problem could be of Network Manager ?

Unless the network-manager maintainer says otherwise, yes, I think this
is a network-manager bug. It just doesn't make sense to set ADDRFAM like
this.

> > > I suggest comment those lines of code or change the condition like this:
> > >
> > > # OpenSSH only cares about inet and inet6. Get thee gone, strange people
> > > # still using ipx.
> > > if [ "${ADDRFAM}" = "ipx" ]; then
> > > 	exit 0
> > > fi
> >
> > IPX is just an example; there are no doubt others OpenSSH doesn't
> > support, and I don't think it's sensible to restart sshd every time one
> > of them appears.
> 
> I don't know the difference between reload and restart, 
> Is that reload scans for new interfaces without closing active connections, 
> and  restart finishes the daemon and starts again the program ?

In this case, reload causes sshd to re-exec itself, and restart forcibly
stops and starts it.

> In some daemons, to send a reload signal causes that immediately is executed 
> the process of verification, instead of wait for exhausting a certain 
> interval of time.
> 
> In the case that this daemon works in this way, to send the reload signal 
> should cause that the program will respond immediately to the changes of the 
> network. 

Indeed, it does, and that's why we use that signal. That still doesn't
mean it's sensible to send it willy-nilly, though.

Cheers,

-- 
Colin Watson                                       [cjwatson at debian.org]