[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#485961: installation-reports: Installer can't locate CD-ROM drive (sparc)

Mark Morgan Lloyd wrote:
[1] http://bugs.debian.org/485961

Two questions that might possibly help. The first is that I suspect that the internal SCSI bus termination isn't particularly robust, I think that putting an external (68-pin, SE) terminator on the external SCSI connector improves reliability of hardware detection and subsequent driver load from initrd. So:

* Is the kernel doing something which affects the electrical interface, e.g. glitching auto-termination state during initialisation?

The second thing is that Splackware (Bobware 10.2) is much more reliable at detecting disc drives on the internal bus than Debian (Etch, 4.0). One particular failure mode I see with Debian and at least some disk types is that when it scans for the CD-ROM drive it does /something/ which causes the hard disc to appear at multiple SCSI IDs including the one that the CD is at- which obviously screws things totally. So:

* Is the installer doing something which affects the electrical interface, e.g. a badly-implemented SCAM probe?

I've managed to duplicate some of these problems on another system, specifically an Ultra-1. I'm fairly sure it's related in some way to termination, and it appears to be crucially sensitive to the make and model of disc.

I found that a pair of drives supplied with an E4500 were invisible to the Debian "Etch" installation CD in a U1, these were IBM and Fujitsu OEM rebranded as Sun 9G and were visible to Slackware. Using Seagates (OEM, actually branded Compaq) I was able to install Debian, but they still weren't reliable in the E4500.

In all cases Slackware is more reliable, but there's also a possibility that there's some OpenPROM command that preconditions the internal SCSI bus to handle discs that would otherwise be ignored or that would cause repeated bus resets.

Mark Morgan Lloyd
markMLl .AT. telemetry.co .DOT. uk

[Opinions above are the author's, not those of his employers or colleagues]

Reply to: