[pkg-wpa-devel] Can't shutdown/reboot clean 'cause of nfs-mounts

kelrin at tpg.com.au kelrin at tpg.com.au
Thu May 18 02:12:46 UTC 2006


Quoting kelrin at tpg.com.au:

> Quoting Reinhard Tartler <siretart at tauware.de>:
> 
> > On Sat, May 13, 2006 at 07:56:10PM +0200, Elimar Riesebieter
> wrote:
> >  
> > > In my home-lan I am sharing files via nfs. On Debian-systems the
> > > umount of nfsfiles is via /etc/rc(0|6).d/S31umountnfs.sh which
> > indeed
> > > starts later than /etc/rc(0|6).d/S20sendsigs which kills the wpa
> > > daemon. So if you have nfs entries in mtab umountnfs.sh
> > > hangs|freezes the machine. Hardresets aren't the best for the
> > > filesystems.
> > > 
> > > I manually fixed S31umountnfs.sh to S19umountnfs.sh but isn't
> > really
> > > reasonable for me.
> > 
> > Thats really an interesting setup. the whole concept seems to
> assume
> > that no daemons are necessary to have a filesystem mounted. The
> > longer I
> > think about it, this is only solveable with a true dependency
> based
> > init
> > concept, which we currently don't have in debian.
> > 
> > > Has one similar experience? Should I file a bug against
> > initscripts?
> > 
> > I think having a bugnr to refer to this issue is useful. So filing
> a
> > bug
> > seems a good idea to me. I don't expect it to be solved soon,
> > though.
> > (well, maybe the problem can be worked around, like you did
> locally)
> > 
> 
> Maybe unmounting the nfs share via a pre-down line in /e/n/i would
> also
> be a valid workaround?
> 

Ah, of course not. Because the wpasupplicant daemon would have been sent
the TERM signal by then, right.

Kel.



More information about the Pkg-wpa-devel mailing list