Re: A big-ish machine, cannot boot

On Friday, November 19, 2010 1:20:53 pm Jung-uk Kim wrote:
> On Friday 19 November 2010 12:04 pm, Garrett Cooper wrote:
> > On Fri, Nov 19, 2010 at 8:59 AM, Andriy Gapon <avg_at_freebsd.org> 
> wrote:
> > > on 19/11/2010 18:57 Ivan Voras said the following:
> > >> On 19 November 2010 17:13, Gary Jennejohn 
> <gljennjohn_at_googlemail.com> wrote:
> > >>> On Fri, 19 Nov 2010 17:51:22 +0200
> > >>>
> > >>> Andriy Gapon <avg_at_freebsd.org> wrote:
> > >>>> on 19/11/2010 17:41 Ivan Voras said the following:
> > >>>>> Fujitsu TX300
> > >>>>
> > >>>> [Thunderbird 3 sometimes fails quoting while replying - blame
> > >>>> it]
> > >>>>
> > >>>> Perhaps I am wrong, but isn't the "twirly" shown by the
> > >>>> loader, still? Not sure if the kernel does that.
> > >>>
> > >>> Yup, that's the boot loader. �The kernel spits out printfs.
> > >>
> > >> Good news, of sorts - I left while I went for dinner and
> > >> apparently it did boot in the meantime. So it's not a complete
> > >> hang, it just takes unexpectedly long (10+ minutes?)
> > >>
> > >> I'm currently running "make -j24 buildworld" and once it boots
> > >> it looks very fast!
> > >
> > > You ought to determine a cause of the long boot, though.
> > > No compromises or excuses! :-)
> >
> > A similar issue occurred with an HP box recently (in the last 3-4
> > months?). I'd check the archives for more details.
> 
> I bet these are "legacy free" machines, right?  I recently noticed 
> that recent Intel chipsets cause incredibly long delays when 
> non-existent ISA ports are accessed, most notably AT keyboard ports.  
> (My gut tells me it is going in and out of SMM repeatedly for 
> nothing.)  Back in the old days, when we had real ISA bus, it used to 
> delay very short and fixed amount time.  Those days, this behaviour 
> was even (ab)used as a delay function where a real timer is not 
> available yet. ;-)
> 
> Try getting rid of all unnecessary device drivers from your kernel 
> configuration.

Someone had some patches to the atkbd early probe to work around long delays 
during the keyboard probe for this reason.

-- 
John Baldwin