[systemd-devel] [PATCH 3/4] manager: stop start timeout when sysinit.target is reached

Zbigniew Jędrzejewski-Szmek zbyszek at in.waw.pl
Tue Oct 28 06:49:40 PDT 2014


On Tue, Oct 28, 2014 at 10:50:05AM +0100, Lennart Poettering wrote:
> On Tue, 28.10.14 03:52, Zbigniew Jędrzejewski-Szmek (zbyszek at in.waw.pl) wrote:
> 
> > On Tue, Oct 28, 2014 at 02:32:37AM +0100, Lennart Poettering wrote:
> > > On Mon, 27.10.14 23:07, Lennart Poettering (lennart at poettering.net) wrote:
> > > 
> > > > On Mon, 27.10.14 20:03, Zbigniew Jędrzejewski-Szmek (zbyszek at in.waw.pl) wrote:
> > > > 
> > > > > On Mon, Oct 27, 2014 at 06:51:08PM +0100, Lennart Poettering wrote:
> > > > > > On Sat, 25.10.14 21:43, Zbigniew Jędrzejewski-Szmek (zbyszek at in.waw.pl) wrote:
> > > > > > 
> > > > > > >  
> > > > > > > +        if (result == JOB_DONE && unit_has_name(u, SPECIAL_SYSINIT_TARGET))
> > > > > > > +                manager_cancel_start_timeout(u->manager);
> > > > > > > +
> > > > > > 
> > > > > > I'd really prefer to hook this up with manager_state(), so that we
> > > > > > don't have to hardcode the target name more than once, and so that we
> > > > > > know that the rescue/emergency mode is properly handled...
> > > > > Yeah, I thought about that too, but only after I sent the patch.
> > > > 
> > > > I'l look into hacking this up now, I really want to get the release
> > > > out of the door now.
> > > 
> > > I hacked this up now. Though after thinking forth and back quite
> > > differently than what we had before. I now removed the system-wide
> > > timeout, in favour of beefing up the existing JobTimeoutSec= stuff we
> > > have in units, so that we can set per-unit timeouts that trigger
> > > failure actions. With this in place I have now set default timeouts of
> > > 15min on basic.target and 30min on poweroff.target+reboot.target. 
> >
> > Good, that sounds more flexible (and simpler). I'll push the change
> > to avoid clobbering of password prompts. 
> 
> Hmm, did you see my comments about using BSD file locks for locking
> access to /dev/console, instead of explict checks for the ask-passwor
> stuff?
> 
> http://lists.freedesktop.org/archives/systemd-devel/2014-October/024582.html
Oops, I see it now. It does seem to be a better solution.

Zbyszek


More information about the systemd-devel mailing list