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

[nptl] Re: http://bugme.osdl.org/show_bug.cgi?id=3770



> >> Hi again, Nick
> >>
> >> I got the static version to work when I linked it to a vanilla glibc,
> >> not the one from Fedora...
> >>
> >> I've attached it to this mail.
> >>
> >> With this release I still get the same issue:
> >>
> >>
> >
> > OK, I can reproduce it here! Thanks.
> >
> > Give me a few days to work on it and I'll get back to you if I come up
> > with anything.
> >
> 
> Ah, the problem is that when the driver thread has a higher priority 
> than the
> worker threads, so when the driver goes into an infinite loop waiting, the
> workers which are on the same CPU don't get to schedule at all. Those 
> that are
> on the other CPU are able to schedule, however.
> 
> The driver really needs to sleep, use a mutex, use a lower priority, or
> something in order for it to work.
> 

Well, actually, I think the kernel has to migrate those lower priority
threads on the 'free' CPU, as the SCHED_RR scheduling policy is clear on
this point: the same-priority threads are scheduled equally on the free
resource...

This was why this testcase was written at first; an it looks like the
kernel does not obey this rule. 

What is your opinion?

Thanks, 
Seb.



-- 
-------------------------------
Sebastien DECUGIS
NPTL Test & Trace Project
http://nptl.bullopensource.org/

"You may fail if you try.
You -will- fail if you don't."

Liste de diffusion nptl
Pour se désinscrire : mailto:nptl_request@bullopensource.org?subject=unsubscribe