Cant install on Hp Vectra dual P90 w SCSI

jnriopel · 01-26-2005, 07:43 PM

Quote:

Originally posted by DaHammer
Yes, ideX=noprobe, where X = the device #, ie normally ide0 and/or ide1. So just add those to your append line or specify them at the prompt:

ide0=noprobe ide1=noprobe

That will prevent the kernel from using/trying to access the IDE interface. If you have the kernel source decompressed somewhere, then check out Documentation/ide.txt and/or Documentation/kernel-parameters.txt for all available parameters. The source may be available decompressed online somewhere, as well, I'm not sure.

BTW. Disabling something in the BIOS usually won't prevent the kernel from trying to access it. It sees it anyway, and tries, it just wouldn't work.

FWIW, My Sandra reporting tool under Win2K says I have a ATA disk controller "Silicon Image Inc (Was CMD Technology Inc) PCI-0640 EIDE Adapter (Single FIFO)" Rev A/3(2)

DaHammer · 01-26-2005, 11:20 PM

Sandra, wow been a while since I've heard that. Anyway, I guess the noprobe doesn't completely turn it off, else maybe it just reports that but doesn't actually try to access it. I'm out of ideas aside from playing around with the kernel at the moment though.

Edit: I did some googling and ran across some posts about interrupts but that's about it. Seems these PCs are norturious for IO related problems and sometimes you have to manually configure the IRQs. You can read Documentation/i386/IO-APIC.txt in the kernel source tree for more details on that.

I also ran across Mr Dave Madsen at Iowa State University. Apparently he tackled this beast some years ago and won, so you may want to contact him to see if he can be of assistance. You can find his contact info at http://www.cs.iastate.edu/~madsen/. Beyond that, you could try stripping hardware out of the box down to the bare minimum to see if you can isolate the problem that way. Anyway, keep us posted. Issues like this peak my interests and I have half a mind to see if I can locate one of these dinosours to play with myself.

uberNUT69 · 01-27-2005, 12:22 AM

Hi,

perhaps you might try adding noapic to your kernel options at the boot prompt

DaHammer · 01-27-2005, 01:07 AM

A little more info. There is some info at http://forums1.itrc.hp.com/service/f...readId=136143. The guy in the last post recommends using the tmscsim SCSI driver vs the AM one. Unfortunately, the only pre-compiled Slackware kernels I could find with this driver instead were in the scsi.s kernel for Slackware 4.0. It's modular in the others I looked at. Anyway, wouldn't hurt to try recompiling the Slackware 10 kernel, disabling CONFIG_SCSI_AM53C974 and enabling CONFIG_SCSI_DC390T.

jnriopel · 01-30-2005, 08:57 AM

hum... I get

Code:

system unavailable
The system is currently unavailable. Please try again later. 

wpa-pl-wpaframework-10000

at that URL...

Sent an e-mail to Mr. Madsen, let's see if he has ideas... Else I'm gonna have to resign myself to running Win2K on the HP. sigh.

DaHammer · 01-30-2005, 04:13 PM

Hmm, guess I messed the link up somehow, here it is again:
http://forums1.itrc.hp.com/service/f...hreadId=136143

And the post:

Quote:

Posted by Robert Gadsdon
I have installed Linux on several of these over the years, and they have all been 'fun'..

Things to watch out for:

Make sure the Distro is not trying to install the old 53c974 SCSI driver - this is known to be 'bad' and will produce many of the symptoms descibed in previous postings - use the (much better) TMSCSIM driver. The later Red Hat distros (7.3) seem to use the right driver by default (found on the extra 'drivers' floppy).

If you can, remove the SCSI disk and use a spare old IDE one instead (but remember that the bios will not correctly recognise any over 4MB, although Linux will). I have successfully used two old 4MB IDE disks in an XU/90, keeping the CDROM on the SCSI connection.

Use LILO with the LINEAR option, and put the 'boot' directory on a separate partition (the first one on the first disk). Make sure the first disk is not over 4MB (you need to use an _old_ one..) This should (hopefully) avoid the 'LI...' problems.

If you have both CPUs installed, make sure that the uniprocessor kernel is also available from LILO ("linux-up") in case of interrupt problems etc. when booting the SMP kernel.

As has been mentioned before, this mobo/scsi combination (using the old 'Neptune' chipset) is not exactly bug-free. _If_ you have the skills, I would also recommend compiling your own kernel, especially for reliable SMP support...