On Mon, 2004-11-29 at 20:53 -0600, Eric Sandeen wrote:
> Henning Rohde wrote:
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA1
> >
> > Hi everybody,
> >
> > today I've stumbled over some very astonishing problem:
> >
> > When trying to copy a clean xfs-filesystem with xfs_copy, everything seems
> > fine until 100% - even until "All copies completed." is announced...
> >
> > But then I get the following line: "Killed".
>
> This is normal; at the end of xfs_copy main() it does:
>
> check_errors();
> killall();
> pthread_exit(NULL);
> /*NOTREACHED*/
> return 0;
>
> (I'm not quite sure why, perhaps Russell can chime in on this...)
This is how the original irix code was cleaning up it's children,
by sending a kill signal.
I suppose it would be better to create a exit var and have the children
break their loop and then pthread_exit.
>
> > $? is 137.
> > /var/tmp/xfs_copy.log.* is empty.
>
> Hm, not sure about that.
>
> >
> > But one of my filesystems to backup was about 17GB in size, filled at about
> > the halve:
> > same phenomenon, but mount failed afterwards, xfs_repair reported severe
> > corruptions.
> > (Logs of xfs_repair are to be reproduced, I'll send them once requested.)
>
> This is more troublesome; if you can post the logs somewhere
> (http://oss.sgi.com/bugzilla would be a fine place) we can take a look.
>
> -Eric
>
|