Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 8 Nov 2004 13:09:05 -0500
From:      John Baldwin <jhb@FreeBSD.org>
To:        Pascal Hofstee <caelian@gmail.com>
Cc:        cvs-all@FreeBSD.org
Subject:   Re: cvs commit: src/sys/kern sched_ule.c
Message-ID:  <200411081309.05747.jhb@FreeBSD.org>
In-Reply-To: <d8a0b76204110519223860ef33@mail.gmail.com>
References:  <200410300735.i9U7ZrL2031884@repoman.freebsd.org> <d8a0b76204110519223860ef33@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Friday 05 November 2004 10:22 pm, Pascal Hofstee wrote:
> On Sat, 30 Oct 2004 07:35:53 +0000 (UTC), Jeff Roberson
>
> <jeff@freebsd.org> wrote:
> > jeff        2004-10-30 07:35:53 UTC
> >
> >   FreeBSD src repository
> >
> >   Modified files:
> >     sys/kern             sched_ule.c
> >   Log:
> >    - In sched_prio() check to see if the kse is assigned to a runq as the
> >      check for TD_ON_RUNQ() no longer means the thread is really on a
> > run- queue.  I suspect this state should be re-evaluated as it must mean
> > something else now.  This fixes ULE+KSE+PREEMPTION on UP x86.
>
> Hi Jeff,
>
> I decided to take this commit as a good moment to give SCHED_ULE +
> PREEMPTION a shot again in the hope of triggering potential bugs to
> further improve SCHED_ULE. And i must say ... without
> INVARIANTS/WITNESS ... SCED_ULE + PREEMPTION == SOLID ... No Panics
> for well over 2 days during general desktop usage and compilation.
>
> I then decided to give it a real stress-test and decided to also
> enable WITNESS, INVARIANTS and FULL_PREEMPTION. This combination
> seemed just as stable at first ... untill i decided to shutdown
> Firefox .. which caused a system panic with the famous "Negative nice
> count" KASSERT from kern/sched_ule.c
>
> I did some additional testing and so far it seems that Every threaded
> application seems to exhibit this panic upon exit. I don't have an
> actual kerneldump at the moment ... but i should be able to generate
> one and provide additional information  and/or a backtrace if desired.
>
> To reproduce simply exit a multi-threaded application ... firefox,
> ogg123 and apache-worker all caused the same panic.
>
> This is on a UP P2-400 MHz running 6.0-CURRENT as of November 3rd or 4th,

I don't think multithreaded processes are ready for FULL_PREEMPTION yet on 
either scheduler, so I wouldn't use FULL_PREEMPTION for now.

-- 
John Baldwin <jhb@FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve"  =  http://www.FreeBSD.org



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200411081309.05747.jhb>