Re: [mdadm PATCH] Kill: fix when zeroing busy devices.

[Neil Brown <neilb@suse.de>]

On Thu, 28 Jan 2010 17:31:57 +0100
Andre Noll <maan@systemlinux.org> wrote:

> On 16:48, Artur Wojcik wrote:
> > mdadm should not wait indefinitely when it trys to reset
> > superblock on device which is busy (cannot be opened).

Thanks.

> > 
> > Signed-off-by: Artur Wojcik <artur.wojcik@intel.com>
> > ---
> >  Kill.c |    2 +-
> >  1 files changed, 1 insertions(+), 1 deletions(-)
> > 
> > diff --git a/Kill.c b/Kill.c
> > index 63442a6..a65b865 100644
> > --- a/Kill.c
> > +++ b/Kill.c
> > @@ -46,7 +46,7 @@ int Kill(char *dev, struct supertype *st, int force, int quiet, int noexcl)
> >  			fprintf(stderr, Name ": Couldn't open %s for write - not zeroing\n",
> >  				dev);
> >  		close(fd);
> > -		return 1;
> > +		return 2;
> >  	}
> >  	if (st == NULL)
> >  		st = guess_super(fd);
> 
> Jup.
> 
> 	Acked-by: Andre Noll <maan@systemlinux.org>
> 
> BTW: Mind to kill the close(fd) as well (as fd is negative) while
> we're at it? A comment that explains the meaning of the return value
> would also be nice.

Done and done.

I really don't know why the loop in mdadm.c cared about the different
possible return values - I cannot see any justification for it.

Thanks,
NeilBrown