alpha system boot problem...

poly_poly-man · Posted: Tue Dec 30, 2008 4:34 pm Post subject: alpha system boot problem...

I just did an install from stage1 on my alpha... and configured the kernel as best I could (this is my first alpha system) - it's a digital personal workstation 600a (miata).

Anyway, my kernel boots up pretty for, and then something like this happens (copying by hand because my desktop system doesn't have a serial port for console):

klausman · Posted: Wed Dec 31, 2008 5:25 pm Post subject:

This looks like a hardware problem to me. My XP1000 has the very same Cypress controller amd it always worked ok (as good as an ISA IDE controller will get). You might try using a different cable ora different cdr drive to see if the problem goes away. I doubt it's a software problem; what kernel version are you using?

poly_poly-man · Posted: Wed Dec 31, 2008 6:27 pm Post subject:

klausman · Posted: Wed Dec 31, 2008 6:31 pm Post subject:

poly_poly-man · Posted: Wed Dec 31, 2008 6:35 pm Post subject:

klausman · Posted: Wed Dec 31, 2008 6:38 pm Post subject:

poly_poly-man · Posted: Wed Dec 31, 2008 6:47 pm Post subject:

my SRM is as up-to-date as it can be, I believe (haven't found a newer version)..

This is weird... totally unplugged the IDE cable, and it boots just fine. You were right, but it still doesn't explain why the livecd worked...

well, it boots but my keyboard doesn't work, at least (don't know if it's frozen..)

I hope that time I shorted out the fan inside the big mother<->daughterboard connector doesn't affect anything...
_________________
iVBORw0KGgoAAAANSUhEUgAAA

avatar: new version of logo - see topic 838248. Potentially still a WiP.

mattst88 · Developer Joined: 28 Oct 2004 Posts: 422

According to the Wikipedia article, there are two revisions of the Digital Personal Workstation. The latter used a different ATA controller.

You can easily tell if your PWS is first or second version, as the second version has USB ports while the first doesn't.

I wonder if this is a problem with the first version.

Also, good to have another Alpha user

You should join us on IRC in the #alpha and #gentoo-alpha IRC channels on irc.freenode.net. I'm mattst88 and klausman is Blackb|rd on IRC. Hope to see you there.
_________________
My Wiki page

mattst88 · Developer Joined: 28 Oct 2004 Posts: 422

And one more thing, please double check that you have the latest SRM [*], version 7.2-1.

[*] http://h18002.www1.hp.com/alphaserver/firmware/readmes/digitalpw.html
_________________
My Wiki page

poly_poly-man · Posted: Thu Jan 01, 2009 4:31 am Post subject:

Apparently, I have a second revision - USB ports for me.

I couldn't remember the version off-hand initially, but yep, that's the version I'm running.

I've already deduced that it both has to be a hardware problem (by the fact that it (mostly) works with the cable unplugged), and that it can't be a hardware problem (by the fact that the livecd boots perfectly..)

I'm doing a little more digging around in there... then I don't know what to do

_________________
iVBORw0KGgoAAAANSUhEUgAAA

avatar: new version of logo - see topic 838248. Potentially still a WiP.

poly_poly-man · Posted: Thu Jan 01, 2009 5:41 am Post subject:

okay, so booting without the cable in still doesn't take input, but it doesn't take input at all, so it seems to be a keyboard driver issue...

...oddly enough, serial console booting doesn't work right (kernel finishes, init never gets anywhere), so can't test that out.

Time to sleep anyway.... I'll hack more at it tomorrow.
_________________
iVBORw0KGgoAAAANSUhEUgAAA

avatar: new version of logo - see topic 838248. Potentially still a WiP.

mattst88 · Developer Joined: 28 Oct 2004 Posts: 422

I'll quote a very helpful person in a thread I started about configuring my old HPPA box.

I think this is probably the fix you need for your serial console to work.

poly_poly-man · Posted: Thu Jan 01, 2009 10:12 pm Post subject:

serial console was not working at all...

I just re-compiled the kernel using the livecd's .config (from /proc/config.gz)... used oldconfig and defaults for the things it asked, and my kernel boots fine, even with the IDE cable attached (and even lets me use the keyboard).

So....... WTF?

Would someone mind looking at my .config, because I am genuinely confused.
_________________
iVBORw0KGgoAAAANSUhEUgAAA

avatar: new version of logo - see topic 838248. Potentially still a WiP.

mattst88 · Developer Joined: 28 Oct 2004 Posts: 422

Now try to remove all the features you don't want.

Sometimes there's magic involved, and it's best not to mess with it if it's working properly.
_________________
My Wiki page

poly_poly-man · Posted: Fri Jan 02, 2009 10:12 pm Post subject:

Okay, I made two config files - one that works and one that doesn't. The only difference: the one that works has "generic" selected for the system. The one that doesn't has "miata" selected.

I can guarantee that this is a miata system - it even says miata on the motherboard.

What is causing this PITA?
_________________
iVBORw0KGgoAAAANSUhEUgAAA

avatar: new version of logo - see topic 838248. Potentially still a WiP.

NormalBoy · Posted: Thu Oct 08, 2009 4:46 am Post subject:

Ever find a resolution to this issue? I too have a Miata (500au) and have been unable to boot a new 2.6.30-gentoo-r6 kernel due to these 'lost interrupt' errors. However, I have a 2.6.27-gentoo-r8 kernel that runs perfectly. Even using the same config, the .30 kernel won't boot.

Dumping the IDE drive isn't an option, as I need an IDE drive to boot off of as SRM doesn't recognize my PCI-X LSI Fusion-MPT SCSI controller

mattst88 · Developer Joined: 28 Oct 2004 Posts: 422

NormalBoy · Posted: Sat Oct 10, 2009 5:42 am Post subject:

Any particular reason for the change from specifying the correct machine type?

I tried out a 2.6.29-gentoo-r6 kernel and it boots fine. It also seems to have resolved my "unable to handle kernel paging request at virtual address" errors as well. Since this kernel has ext4 support, is there any advantage to moving to 2.6.30?

Altivo · Posted: Fri Aug 21, 2015 12:17 pm Post subject:

Just thought I'd note here that this problem still persists in Gentoo 4.0.5 Jan 2015.

Plus at this point the LiveCD loses connection to the CDROM drive once it boots up. I had to make two copies of the CD, and put the second one on a USB attached CD drive. The LiveCD kernel can find that and continue, though the SRM BIOS cannot boot from USB at all. I believe the LiveCD is now using libata rather then the old IDE code.

Custom kernel compiled with Miata setting fails to find the IDE CDROM, which is on a Cypress controller, PCI bus, even with PATA-Cypress in the kernel. The internal sound card (ESS1888 if I recall correctly) is also invisible to the kernel and ALSA, though it works correctly on Debian 4.0 etch.

I guess I'll have to revert to the generic setting for the architecture. Looks like a kernel bug from what I see here.

(Hardware here is PWS 433au, and passes all hardware validation tests.)
_________________
Have patience with me, I'm only half-baked.
--Altivo

Altivo · Posted: Fri Aug 21, 2015 2:27 pm Post subject:

An additional thought:

The Miata systems actually came in two distinct flavors, with some hybrids during the transition. The later version is sometimes called MiataGL, and it has better DMA functionality and supposedly better bridging between the expansion slots. PWS models that have two USB ports on the back just below the two serial port connectors are usually MiataGL. The older Miata systems lacked the USB ports.

It is conceivable that the "Miata" designation in the Gentoo kernel does not take into account all of the chipset variations, hence the difficulties with some machines.

Later SRM releases for the Miata machines have the capability of recognizing "problem" PCI cards and disabling them. This was due to difficulties with the earlier DMA implementation. There is a software switch in SRM to override this feature and changing the setting may allow unrecognized cards to be used successfully. (Or not!) I believe this limitation applies only to the "long slot" or 64 bit PCI slots, and not to the shorter ones that share space with the ISA bus.
_________________
Have patience with me, I'm only half-baked.
--Altivo