MP Build... FYI

[Andres Salomon]

On Sun, 04 Nov 2007 10:55:05 -0500
Jim Gettys <jg at laptop.org> wrote:

> I'm still confused/concerned....  In particular, I have memories of a
> change for manufacturing test to cause the systems to wake up on
> multi-cast, to enable the mass suspend/resume testing of units.

It went into 624 and 625.

The DCON power cycle thing is in all builds, but the proper register
init path is only in 625.

> 
> Andres, did this ever go by chance into the 62x series of builds (as
> opposed to being a separate kernel RPM?
> 
> If so, at which build did it go in?
> 
> My understanding of the situation is:
> 	623 - last fully tested MP build
> 	624 - first attempt at DCON power cycle hammer code, 
> 		missing a case.
> 	625 - most recent build, only intended for manufacturing test
> 		and burn-in, with working DCON power cycle hammer
> code.
> 
>                                   - Jim
> 
> 
> 
> On Sun, 2007-11-04 at 02:27 -0500, Mary Lou Jepsen wrote:
> > thanks arnold.
> > - Mary Lou
> > 
> > On Sun, 2007-11-04 at 14:40 +0800, Arnold.Kao at quantatw.com wrote:
> > > Actually, I had asked our test engineer in CSMC to prepare 625
> > > for the factory suspend/resume test tomorrow. They are doing this
> > > right now.  It means we should have both 624 and 625 test image
> > > to be ready tomorrow.  We can make decision which image we want
> > > to use in the factory suspend/resume test after we confirm the
> > > test result of the 4 machines that were failed with DCON problem
> > > before.  
> > > 
> > >  
> > > Best Regards,
> > > Arnold Kao
> > > Quanta Computer Inc. 
> > > Tel : 886-3-327-2345 EXT:18958
> > > Fax : 886-3-328-9780
> > > 
> > > 
> > > -----Original Message-----
> > > From: Richard Smith [mailto:smithbone at gmail.com] On Behalf Of
> > > Richard A. Smith Sent: Sunday, November 04, 2007 1:46 PM
> > > To: John Watlington
> > > Cc: Kim Quirk; Jim Gettys; Arnold Kao (高顯宗); Mary Lou Jepsen;
> > > devel; Andres Salomon Subject: Re: MP Build... FYI
> > > 
> > > John Watlington wrote:
> > > 
> > > > Quanta wants assurance that the software workaround which was
> > > > broken in #4479 is fixed.
> > > > Richard's testing is necessary to confirm this.   It is also
> > > > essential that the kernel fix
> > > > which is theoretically the only difference between 624 and 625
> > > > be part of the production
> > > > test code to further confirm this.   If Quanta sees this
> > > > problem AT ALL in the production
> > > > testing, there will be pressure to halt until further
> > > > hardware/software fixes are found.
> > > 
> > > I'm getting a slightly different story.  I was trying to meet
> > > with people yesterday to discuss the whole mfg testing procedure
> > > so I knew what was going on and what I would need to do to get
> > > the new kernel into the Run-In image.
> > > 
> > > Arnold tells me that he thinks its  too late to get the DCON
> > > workaround kernel into the Run-In image.  He suggestion was that
> > > if we see a DCON problem that we pull the machine out of the
> > > rack, update the kernel and then re-add it back into the testing.
> > > 
> >