NetBSD-Bugs archive

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

Re: kern/28612 ([scheduler activation] any local user can lock the system)



On Mon, Jun 23, 2008 at 09:26:25PM +0900, ITOH Yasufumi wrote:
> wrstuden%NetBSD.org@localhost writes:
> > State-Changed-From-To: analyzed->feedback
> > State-Changed-By: wrstuden%NetBSD.org@localhost
> > State-Changed-When: Mon, 23 Jun 2008 04:24:35 +0000
> > State-Changed-Why:
> > Bug is believed to have been fixed in 2006, in rev 1.73 of kern/kern_sa.c
> 
> Yes, the commit message is correct.
> The process still loops but at least it is killable.

Ok, so this at least drops it into the class of apps that go out of 
control and gobble up a lot of CPU. So it isn't as strong a security issue 
as it used to be.

My instinct is that if we run out of stacks here, we should just kill the 
app. libpthread installed 16 stacks in NetBSD 2.0, and we have never 
strongly supported concurrency. So "real" pthread apps should always have 
a lot of upcall stacks per VP. Thus we really shouldn't run out.

So my thought is add a sysctl that root can set/clear to control killing 
apps here. Set the default to "yes."

Thoughts?

Take care,

Bill

Attachment: pgpcFTDH9ZC2N.pgp
Description: PGP signature



Home | Main Index | Thread Index | Old Index