[PATCH 1/3] Signal semantics for /sbin/init

Oleg Nesterov oleg at tv-sign.ru
Mon Sep 3 09:45:48 PDT 2007


On 09/03, sukadev at us.ibm.com wrote:
>
> Oleg Nesterov [oleg at tv-sign.ru] wrote:
> | On 08/31, sukadev at us.ibm.com wrote:
> | > 
> | > -static int sig_ignored(struct task_struct *t, int sig)
> | > +	// Currently this check is a bit racy with exec(),
> | > +	// we can _simplify_ de_thread and close the race.
> | > +	if (likely(!is_container_init(tsk->group_leader)))
> | > +		return 0;
> | > +
> | > +	if (!in_interrupt())
> | > +		return 0;
> | 
> | I don't understand why you are trying to mix this patch with pid_ns changes.
> | 
> | We don't need in_interrupt() check unless we use current do decide if the
> | signal goes from the parent namespace.
> | 
> | And in fact, I'd personally prefer to use "is_global_init()" for this patch,
> | because it hopefully can fix the problems we have even without namespaces.
> | This also matches the current check in get_signal_to_deliver().
> 
> Sorry. I wasn't paying enough attenttion to this patch and including it
> only for reference. Was planning to replace this with your final patch.
> Or do you want me to fix the two bugs and resend ?

Sorry! I didn't have any time for the kernel hacking last days. There are
some other minor (and not related) problems with the blocked signals which
I'd like to check before doing the final patch.

Please feel free to do what you think right. I am going to KS this night,
and I will be completely offline till september 10. Any chance we could
delay this a bit? In any case, patches 2-3 should not depend on any further
possible changes in this patch.

Oleg.



More information about the Containers mailing list