[fm-discuss] Fmadm faulty not working

[Michael Shapiro]

> On 2/9/07, Michael Shapiro <mws at zion.eng.sun.com> wrote:
> 
> > The memory errors there are pre-FMA messages: there is no FMA support
> > for UltraSPARC-II.  So you're seeing the old pre-FMA code.  Meantime,
> > your faulty memory is being accessed for DMA by your I/O path, and that
> > is producing FMA error reports, but our I/O diagnosis code knows that
> > they are secondary effects, so it's ignoring them and not diagnosing anything.
> >
> > So if you had FMA for US-II, you'd see a DIMM diagnosis.  Anyway,
> > this is why we did FMA :)  But I'm sorry it's not retroactively
> > available for the really old UltraSPARCS ...
> 
> Thanks Mike and Cynthia for the feedback. Are the FMA diagnosis
> engines only aware of AMD CPUs and the UltraSPARC III and above? Out
> of curiosity, is there a reason that diagnosis engines weren't
> developed for the UltraSPARC II CPUs (other than the CPUs being older
> citizens in the SPARC community)?
> 
> Thanks again for the feedback,
> - Ryan

Correct, for SPARC US-III and all its variants and above, and for x86
we support AMD now (with Intel now on the roadmap).  Basically it was
just a question of investment: Sun's view was that it was too costly
to go back and do the work for pre-US-III systems, and that these were
likely to be running older releases anyway and less likely to upgrade.
This is, however, something the community could do and contribute.

-Mike

-- 
Mike Shapiro, Solaris Kernel Development. blogs.sun.com/mws/