[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: kill (0, SIGSTOP) is freezing the system (was: /hurd/init and /hurd/

From: Justus Winter
Subject: Re: kill (0, SIGSTOP) is freezing the system (was: /hurd/init and /hurd/proc)
Date: Wed, 31 Jul 2013 11:50:55 +0200
User-agent: alot/0.3.4

Quoting Samuel Thibault (2013-07-29 13:36:33)
> Justus Winter, le Fri 26 Jul 2013 13:24:49 +0200, a écrit :
> > > >   I'm not sure about how important it is not to freeze anyone of them,
> > > >   but at least procfs must not be stopped b/c killall5 wants to
> > > >   iterate over /proc.
> > > 
> > > And /proc might not even be started yet, so exec need to be unfrozen at
> > > least, to start it. Perhaps killall5.c could have a Hurd-specific
> > > function to use libps to identify procfs and let it continue.
> > > 
> > > > But it would surely be nice not to freeze filesystem translators, and
> > > > probably the pagers and the term translators.
> > > 
> > > So should perhaps translators be marked as such and avoided?
> > 
> > So you are saying that any translator started by root should be
> > considered essential?
> For instance, yes.
> > A more explicit variant would be to let translators set themself as
> > essential, would you prefer that?
> That should work too.  I'd tend to prefer that.

Cool, will do.

> That said, we don't want to mix up "essential" as it happens to be now
> (whenever it crashes, init crashes the whole system) with "essential" as
> we want it to have (do not kill it on killall -1).

Good point. I've been consulting thesaur{i,uses} to find a synonym for
"essential". The best ones I found were

* crucial
* vital
* important (being a little weaker than essential, which might be
* something more different like untouchable (protected, shielded,

Suggestions? Opinions?

Somewhat related, I was thinking not to deliver fatal signals to
*essential* (in the traditional /hurd/init way) servers, at least not
until there is a recovery/reincartation strategy for these servers. As
I understand POSIX this is okay, e. g. killing pid 1 is denied on most
systems (I think Linux just silently does nothing). Thoughts?


reply via email to

[Prev in Thread] Current Thread [Next in Thread]