[systemd-devel] Masking local configuration system unit fails on file collision

Vito Caputo vcaputo at pengaru.com
Sat Mar 23 09:15:32 UTC 2019


On Sat, Mar 23, 2019 at 11:59:57AM +0300, Andrei Borzenkov wrote:
> 22.03.2019 8:03, Vito Caputo пишет:
> > On Fri, Mar 22, 2019 at 06:46:21AM +0200, Mantas Mikulėnas wrote:
> >> Masking *is* local configuration. It's there so that the admin could
> >> simulate deleting a unit from /usr, which is considered more-or-less
> >> read-only; same idea as overriding /usr units in general.
> >>
> >> When you've created your own unit in /etc, there's no point in masking it
> >> if you can just move the unit file away.
> > 
> > From the perspective of the user running `systemctl mask $foo` where the
> > service file happens to be is completely out of context.  It should work
> > uniformly, not break spuriously because the config file is in
> > /etc/systemd/system.
> > 
> > Not to mention it's absurd to require the user to have to move the
> > service file somewhere else to accomodate the mask then have ot remember
> > what convention they used to either rename it to something like
> > foo.service~ or what directory they stowed the thing in when the time
> > comes to unmask it.
> > 
> > I'm surprised this requires any debate, nice top post btw.
> > 
> 
> I do not debate that current behavior is inconsistent. But as
> implemented currently there is no easy way to support uniform masking of
> units.
> 
> So you would need to start with proposal (and I suspect it will need to
> include more than "use different directory") that can actually be
> discussed.

There's no point in investing in any kind of elaborate proposal if I'm
the only relevant person who thinks this needs fixing.

BTW I've created an issue on github:
https://github.com/systemd/systemd/issues/12084

Regards,
Vito Caputo


More information about the systemd-devel mailing list