Release Candidates ?
David Zeuthen
david at fubar.dk
Tue Feb 20 07:52:24 PST 2007
On Tue, 2007-02-20 at 18:28 +0300, Andrey Borzenkov wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On Вторник 20 февраля 2007, David Zeuthen wrote:
> >
> > 7. Making umount(8) work with HAL for unprivileged users. This one
> > is described in detail here
> >
> > https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=188193
> >
> > and should be pretty easy to fix. It's a RH specific thing at the
> > moment I think....
>
> No it is not. It has been discussed on Manriva cooker list as well (an I
> believe there is a bug report for this)
(it would be useful to point to such discussions / bug reports)
Uh, so I just committed a few patches to HAL last night for doing this.
It works well. See
http://gitweb.freedesktop.org/?p=hal.git;a=commit;h=184afb9acb61816203c852ab3559a849248362d3
and later patches. Also, you might want to check out the util-linux
package from Fedora's Rawhide; as I stated I believe said patch for
util-linux is RH specific hence why you right now need to pass
--enable-umount-helper to HAL to make this work. Distros can pick up
this RH patch if and build HAL with this option if they want.
> Any real solution requires either making (u)mount depending on HAL/D-Bus or
> modifying kernel to accept proposed option (user or uhelper). Anything that
> relies on /etc/mtab hopelessly breaks if /etc/mtab -> /proc/mounts
Of course not. If /etc/mtab -> /proc/mounts then
1. I suspect mount(8) have enough brains not to pass the uhelper=hal
option to the kernel. Hence, only thing is that this breaks unmount
helpers, e.g. umount won't run /sbin/umount.hal. Big deal, back at
the situation we have without this.
2. Also, it should be possible for util-linux to store the associated
metadata (uhelper=hal was passed for /dev/sdc1) in *another* state
file say in /var/lib/util-linux/mtab. Then it works again.
> Well, it just shifts the problem further in the future when namespaces
> appear :)
How so? When, or rather if, name spaces become relevant the mount/umount
can just maintain state files a'la /etc/mtab per name-space somewhere else.
I seriously can't see what the problem is. Please clarify.
(Also, doing crap like making /etc/mtab -> /proc/mounts is kinda just
papering over problems but I suspect you knew that already :-) )
David
More information about the hal
mailing list