[rescue] Netra NFS/Enterprise 150 question

[Mike Meredith]

On Sat, 7 Feb 2009 14:24:18 -0500 (EST), Carl R. Friend wrote:
>     On Sat, 7 Feb 2009, Bill Bradford wrote:
> 
> > Unfortunately they *completely* dropped drivers for other parts of
> > the U1 as well - he'd only be able to run really early revisions
> > of the OpenSolaris code.
> 
>     There's also the "issue" that it's apparently possible to wedge
> the early Ultra chips when running in 64-bit mode.  I know that I
> had to invoke some magick to get Solaris 8 to boot in 64-bit mode

Yes it's possible to wedge an Ultra-1 cpu in the same sense as it is
possible to find two files that give the same md5 checksum. Or in other
words you're not going to find it unless you hunt for it. I ran such a
machine in production running a 64-bit kernel for years without a
problem. Ok I wouldn't have done if students were logging in and
compiling code(*), but for a dedicated name server it was fine.

Of course later versions of Solaris 10 (I _think_ the early versions
just printed an error that amounted to saying you can't run this on
ancient hardware) may have Done Something Dumb to exercise that cpu
bug. Or possibly the code that displays the error has become very
broken.

*: In practise the biggest problem of allowing students to login to a
multiuser machine and run code is that they were taught at $work to
turn off the optimiser (because "it changes your code") and were under
the impression that an infinite loop that didn't do anything would
eventually exit. And would try it again on another "terminal" just to
check. Fortunately a V880 running Solaris 9 can run a very large number
of infinite loops that don't do anything.

-- 
Mike Meredith (http://zonky.org/)
 Power corrupts; Powerpoint corrupts absolutely.
 -- Vint Cerf