This bug should be closed ("Karmic & Lucid (sparc) - Installer: "Fast Data Access MMU Miss" Ubuntu “silo” package Bug #476234). It is not a bug in silo

Asked by Mark Sitkowski

The 'bug' is actually a bug in the way the ISO is assembled. If the empty silo.conf is replaced with a valid file, and the ISO is rebuilt, the machine will boot, and the installer will run. However, all SPARC ISO's should be withdrawn immediately, as in every release except 6.06.2, the installer subsequently fails to find it's own CD, or any IDE disks. (Don't you guys actually test these ISO's?) I've rebuilt the ISO for every release from 8.04 upwards, and none of them will work - always for the same reason. I suspect the reason may lie in an inconsistent use of UUID's, but it also seems that there is no device driver loaded for cdrom, hda, hdb etc. If you start a shell from the installer, and mknod by hand, that, too, does nothing, as the installer probably never looks at /dev after it's started.
If you indicate where, in the installer source, I should look for the function that loads device drivers/checks cdrom and disks, I'll be quite happy to fix the bug, recompile and test the code, and submit it to you.

Question information

English Edit question
Ubuntu silo Edit question
No assignee Edit question
Last query:
Last reply:
All the details are in the bug report though as far as I can see.
Revision history for this message
Mark Sitkowski (xmarks) said :

Can we please have the real problem fixed as soon as possible?
Trying to bypass this by upgrading from 6.06.2 doesn't work either, since 6.06, ass installed from the ISO, has no command 'do-release-upgrade'.

Revision history for this message
Tim Penhey (thumper) said :

The bug doesn't indicate that the problem is in silo upstream, but in the silo package for ubuntu.

I'm retargetting the question to ubuntu/silo.

Revision history for this message
Mark Sitkowski (xmarks) said :

The only reason Sun machines give the message "Fast data access MMU miss", is that, with an empty silo.conf, silo doesn't know how big initrd is, and keeps loading data until OBP's buffer overflows (at around 10MB, for OBP 4.0). OBP then quits, and reverts to the 'ok' prompt.
If silo.conf contains valid data (the files in the ISO's for all recent releases are zero length), silo has no problem, loads the correct length of initrd, and the machine boots, then the installer runs - subsequently revealing a true bug, which is the inability to find the cdrom or the disks.
My point is, that there is no bug in silo, causing the "Fast data access MMU miss" - it happens because silo.conf contains no data.

Can you help with this problem?

Provide an answer of your own, or ask Mark Sitkowski for more information if necessary.

To post a message you must log in.