FreeBSD-7.0-Beta2 (i386) on Opteron?

[Mark Staudinger]

On 11/14/07, Kris Kennaway <kris at freebsd.org > wrote:
>
> Mark Staudinger wrote:
> > I wasn't sure if -questions or -current was the right group, but I
> searched
> > both before posting.
> >
> > I'm trying to begin testing FreeBSD-7.0-Beta2, and I have a mixture of
> Intel
> > and AMD-based machines that I work with.
> > I'm using the i386 release.  I've previously used FreeBSD-4.9R and 6.2Rfor
> > i386 on these machines, but cannot get
> > 7.0-Beta2 to work on the AMD machines.  These are 100- and 1000-series
> > dual-core Opteron CPUs.
> >
> > I've tried:
> > Updating using source from 6.2-Release
> > Installing from 7.0-Beta2 CD
> > Installing from 7.0-CURRENT-200710-i386 snapshot CD
> >
> > I see a variety of errors, including:
> >
> > 1) BTX Halted
> >
> > 2) errors when running /bin/sh from /etc/rc
> >        /libexec/ld-elf.so.1  Undefined symbol "opterr" referenced from
> COPY
> > relocation in sh
> >        /libexec/ld-elf.so.1 Invalid file format (on multiple binaries)
> >        various other "undefined symbol" messages
> >        signal 10, signal 11 on multiple binaries
>
> Sounds like you might have a corrupted installation.  Did you verify the
> MD5 checksum on the ISO images?
>
> > Has anyone successfully used 7.0-Beta2 for i386 on AMD Opteron-based
> > platform?  Any suggestions for tracking
> > down the problem?
>
> Yes, it works on mine.
>
> Kris
>
>
I just compared the md5 sum and they match.  Considering that I used 2
different CD sources,
as well as a source update, all with similar results, it would seem unlikely
that a corrupted file/image
could be at fault.  And of course, the same hard CD can be used to install
to an Intel-based machine
with no failures.

I've tried 3 different MB/ CPU combinations.  All are Supermicro-branded,
and all have the HT1000
ServerWorks chipset.  I've just found Jeff Robertson's post:
http://docs.freebsd.org/cgi/mid.cgi?20071109114636.Y639

Perhaps he can confirm the behavior.  In the meantime I will try the patch
posted by John Baldwin
yesterday on -current, and test that with both IDE and MMIO modes.

Thanks,
-ms