Re: The infamous "Fast Data Access MMU Miss"
I think I've made some progress here.
Mark Morgan Lloyd wrote:
> Since the machines are of unknown history the first thing I did, after
> selecting the best from the heap, was update OBP etc. using the latest flash
> from Sun. When I tried booting Sarge (r0a, R3, R5) from CD I got "Fast Data
> Access MMU Miss", sundry problems also trying to boot R5 over TFTP. RH6.1
> and Solaris 8 booted to early-installation stage from CD without problems.
>
> After tinkering for a few hours I selected another machine from the heap,
> grafted memory etc., and tried booting Sarge- it worked. Then I updated the
> flash using 104881-09. Bad move: trying to boot Sarge I now got "Fast Data
> Access MMU Miss" again.
> So to summarise: some of the more recent OBP patches might be responsible
> for putting some models (e.g. in this case Ultra-1 non-Enterprise) into a
> state where SILO (1.4.9?) and possibly other loaders can't run reliably.
> This is not in itself a Linux problem.
I was able to install Solaris on one of these machines and copied the 104881-06
flash-update set into /. Booting this from the OBP prompt I was able to walk the
firmware back to OBP 3.11.1 and POST 3.10.6 which will boot the 2.4 kernel from
Sarge r0a or r5 CDs but hit bad magic while trying to load initrd. Booting from
a previously-configured and up-to-date hard disc was fine, system came up
without any problems.
Applying the same procedure to the second machine fixed that one as well. In
other words at least some Sarge booting problems depend on the state of the
firmware, and it might be possible to work around them by flashing to a
known-good version; I've not tried using Linux to write the patch file to / of a
UFS filesystem and seeing if OBP can handle it.
Apologies for still being on 2.4 and not doing my bit testing 2.6 :-/
--
Mark Morgan Lloyd
markMLl .AT. telemetry.co .DOT. uk
[Opinions above are the author's, not those of his employers or colleagues]
Reply to: