Bug#758902: systemd: Please make ^C interrupt systemd-fsck

[Brian Potkin]

On Tue 23 Dec 2014 at 13:19:08 +0100, alberto fuentes wrote:

> On Mon, 22 Dec 2014 21:30:38 -0500 Stefan Monnier
> <monnier at iro.umontreal.ca> wrote:
> 
> > I think this is a very nasty regression that will hit relatively few
> > people but will really piss them off.
> 
> Not sure why you think they will be few people affected. I think almost
> everybody is going to be hit by this at least once... probably more than
> once. Computers have hard drives attached. They are checked at
> semi-random intervals. If the hard drives is relatively large, it can
> take from 20min to more than 1h to check.

File systems created with mk2fs on fresh Wheezy and Jessie installs have
mount-count and time based file checking disabled. That's quite a few
people who are never liable to experience a surprise fsck.

> I was biten by this once a month until one time I had to do a
> presentation and the computer refused to bootup without checking the
> disks, then decided to write the bug and disable checking on the disks
> risking corruption

Disabling perodic checking (which upstream doesn't regard as conferring
any particular benefit) eliminates the surprise factor. Without elements
of surprise and inconvenience to combat there is no need to overstrive
for a solution. A fsck can always be run by a user as part of regular
maintenance.

Wishlist is about right for this bug; there isn't a lower severity
level.

> > In my experience, systemd has introduced various different *new* ways in
> > which the boot can get stuck in some long-running "operation" (which
> > might just be waiting for a timeout). If we don't want to alienate
> > users, we should raise the severity of this bug (and broaden its scope).
> 
> While I agree and something generic and well implemented should be put
> in place, i would be well off with a hack to interrupt it.
> 
> It would take less than 15 min to implement it to the systemd people to
> avoid this 10+ years regresion. Calling a pre-script,or pre-service or
> however systemd works
> 
>  "echo 'press ctrl - C to interrupt'; sleep 3;systemd-f*ck"
> 
> I dont know the details of how it would go in systemd but I dont think
> this hack cant be put in place with the mighty systemd, can it?

Changes to ways of working are not unknown as software develops. It
could take much less than 15 minutes for a user to implement a solution
which took the changes into account.

> The major problem is that the computer refuses to bootup without doing
> the preventive check. Not even a reboot because you missed the ctrl - c
> screen will help and you are forced to wait like is 1999

The first statement is misleading; new Wheezy and Jessie users would
disagree with it. They also realise that the consequences of enabling
mount-count or time based file checking falls on them, not someone else.

Regards,

Brian.