Date: Thu, 6 Jan 2005 17:52:28 -0500 From: John Baldwin <jhb@FreeBSD.org> To: Stephan Uphoff <ups@tree.com> Cc: Tony Arcieri <tarcieri@atmos.colostate.edu> Subject: Re: cvs commit: src/sys/kern sched_ule.c (fwd) Message-ID: <200501061752.28411.jhb@FreeBSD.org> In-Reply-To: <1105050997.27981.26.camel@palm.tree.com> References: <20041214222444.GA9668@flash.atmos.colostate.edu> <41DDAEA9.90401@elischer.org> <1105050997.27981.26.camel@palm.tree.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thursday 06 January 2005 05:36 pm, Stephan Uphoff wrote:
> On Thu, 2005-01-06 at 16:33, Julian Elischer wrote:
> > John Baldwin wrote:
> > >On Wednesday 15 December 2004 05:27 pm, Julian Elischer wrote:
> > >>Tony Arcieri wrote:
> > >>>On Wed, Dec 15, 2004 at 04:40:50PM -0500, David Schultz wrote:
> > >>>>On Wed, Dec 15, 2004, Tony Arcieri wrote:
> > >>>>>And am I correct that the UMA implementation in RELENG_5 has
> > >>>>> rendered proc_fini() obsolete and thus it won't ever be called?
> > >>>>
> > >>>>This has very little to do with either UMA or ULE. Yes, it's
> > >>>>unused, but it's still there as a reminder that it *ought* to be
> > >>>>used. Unless there are still races I don't know about, it's
> > >>>>probably safe to start using it again.
> > >>>
> > >>>Well, I'm going by the comments and implementation from kern_proc.c in
> > >>>HEAD:
> > >>>
> > >>>/*
> > >>>* UMA should ensure that this function is never called.
> > >>>* Freeing a proc structure would violate type stability.
> > >>>*/
> > >>>static void
> > >>>proc_fini(void *mem, int size)
> > >>>{
> > >>>
> > >>> panic("proc reclaimed");
> > >>>}
> > >>>
> > >>>The implementation in RELENG_5 invokes a scheduler function which is
> > >>> no longer present in HEAD.
> > >>
> > >>when we declare teh zone for processes we tell UMA that it must never
> > >> free a proc back to system memory. thus the 'fini' routine, that would
> > >> be called is a page of that zone were to be returned to the system,
> > >> should never be called.
> > >
> > >Why are struct procs forced to be type-stable?
> >
> > I have forgotten.. but they did..
> > Peter also knew at one stage and he too has forgotten :-)
>
> kern/62890 ?
> Guess this one is mine now :-(
Hmm. pidp sounds ok to me. Some other users of fork1() use RFSTOPPED and
then do a setrunnable() of the first thread after they have finished their
localized setup.
--
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?200501061752.28411.jhb>