hal-system-storage-mount bug

[Jeffrey Stedfast]

Awesome news David, that rocks!

Jeff

On Sun, 2006-02-12 at 21:44 -0500, David Zeuthen wrote:
> On Fri, 2006-02-10 at 15:40 -0500, Jeffrey Stedfast wrote: 
> > I've been trying to add support for the iocharset= mount option and it
> > won't work because the '-' character is not allowed by the mount script.
> > 
> > changing to:
> > 
> > GIVEN_MOUNTOPTIONS=${GIVEN_MOUNTOPTIONS//[^a-zA-Z0-9_=-[:space:]]/_}
> > 
> > fixes the problem.
> 
> I've committed this change. I've also fixed up all the mount scripts to
> allow [:alnum:] and [:space:] in mount point name meaning mount points
> like "foobar der æøå サイトの" [1] should work too. Let's see how much
> this breaks :-)... Though GNOME seems to be rather happy with these
> entries in /proc/mounts, /etc/mtab and hal... 
> 
> It should be secure too, we use
> 
> GIVEN_MOUNTPOINT=${GIVEN_MOUNTPOINT//[^[:alnum:]_=[:space:]:.+-]/@}
> 
> in LANG=en_US.UTF-8 meaning attacks like passing `/bin/evil` as
> GIVEN_MOUNTPOINT (which we expand later in the script) or other fun
> tricks shouldn't work.
> 
> Also fixed is the way hal handles surprise removal, e.g. yanking a USB
> stick or media without unmounting it. We used to issue '/bin/umount -l',
> which could block (bad: hang in state D => kill hald entirely), now we
> ask hald-runner to do the heavy lifting.
> 
> Also now.. even if you don't use gnome-[mount|umount|eject], e.g. just
> do a  'umount /dev/sda1' as root we detect this and clean up the mount
> point that hal's Mount() method created.
> 
> Cheers,
> David
> 
> [1] : I have no idea what "サイトの" means - I just grabbed it from the
> yahoo.jp homepage :-)
> 
>