quit on init with QuickSilver (G4)

iMike · Last edited by iMike on Fri Jan 12, 2007 2:31 pm; edited 2 times in total

I'm having trouble installing Gentoo on a QuickSilver PowerMac. When I boot into the Kernel it gets as far as:

JoseJX · Retired Dev Joined: 28 Apr 2002 Posts: 2774

I didn't see anything wrong with the config (I just skimmed it), but you could try make pmac32_defconfig before compiling the kernel. This configuration should work on all Power Macs.

If that doesn't make a difference, can you try booting with init=/bin/bash appended to your kernel boot parameters? This should boot into bash instead of trying to start init. From here, we can try manually starting init and see what exactly is causing the problem.
_________________
Gentoo PPC FAQ: http://www.gentoo.org/doc/en/gentoo-ppc-faq.xml

iMike · boot: kernel-2.6.18-gentoo-r3 init=/bin/bash

JoseJX,

I tried something very close to pmac32_defconfig (started with pmac32_defconfig and only added reiserfs, and lvm2 support), but still no go.

So, going with Plan B, do I add the init=/bin/bash after the kernel name, just as one would with a Gentoo install CD:

JoseJX · Retired Dev Joined: 28 Apr 2002 Posts: 2774

I'm not so sure now since the 2006.1 CD doesn't work either. Does 2006.0 work? Does using the kernel from the last working CD work with the new installation?
_________________
Gentoo PPC FAQ: http://www.gentoo.org/doc/en/gentoo-ppc-faq.xml

iMike · 2006.0 minimal CD (2006-03-11):

Home now, and after trying things out I can see that booting the Gentoo 2006.1 CD makes it through to different places before crashing, and last time I tried it said "Loading modules....". I have now tried my entire arsenal of Gentoo CDs (back to 2005) with the following results, from newest to oldest CD:

l2006.1 live cd experimental (downloaded 2006-12-3):

JoseJX · Retired Dev Joined: 28 Apr 2002 Posts: 2774

Okay, this is starting to look more like a kernel issue rather than a configuration issue. 2005.1 shipped with a 2.6.12 kernel (ugh). First, try with 2.6.20_rc3 to see if it's been fixed in a recent kernel. If it hasn't, you can get older versions of the kernel from kernel.org, but start with the last version of 2.6.14 (2006.0 shipped with 2.6.15), then 2.6.13, then 2.6.12. You're trying to find the last kernel that worked. :) After this, we can try a git bisect to see if we can figure out exactly what broke.
_________________
Gentoo PPC FAQ: http://www.gentoo.org/doc/en/gentoo-ppc-faq.xml

iMike · Posted: Wed Jan 03, 2007 12:35 pm Post subject:

I didn't see 2.6.20_r3 in portage, so I assume you mean go to www.kernel.org and download it myself.

Being somewhat lazy, I downloaded the latest ~ppc vanilla from portage, which was 2.6.19.1. It compiled OK, but when I rebooted, there was still a problem. I did this remotely, so when I'm home, I'll check and see what the console messages are.

I'm wondering about the likelihood that this really is a kernel problem. The QuickSilver has been out since around 2002. Surely someone else would have noticed a problem with it before. It there anyway to get more messages from the kernel so I can tell what is going on at init? (As mentioned, using init=/bin/bash did not change the messages I received when booting with the 2006.1 minimal CD, which in fact occaisionally did make it through "Loading Modules" and hung on the getting CD image from /dev/hda part.)

This is currently the fastest machine I have at home, so it would be a pitty if I couldn't get it going with Gentoo. I will ultimately try 2.6.20_r3, but if THAT fails too, (and the same disk boots in another G4), then I'll probably have to try some other not-as-nice distro like FC6 or openSUSE. Now that would be sad.

/Mike

iMike · Posted: Wed Jan 03, 2007 7:54 pm Post subject:

...and the result from trying a 2.6.19.1 vanilla kernel is exactly the same

. I took the disk with said 2.6.19.1 kernel and moved it to another G4. It booted without problem. I then replaced the graphics card and the memory from the QuickSilver with those from the working G4, still in hopes that it is basically a hardware problem. This also did not solve the problem.

I'm not sure I'm gonig to be able to keep to my promise of not bathing until I get some resolution to this. :twisted:

iMike · patch -p0 < patch-2.6.20-rc3

Could anyone advise me on exactly how I make the 2.6.20-r3 out of the 2.6.19.1 and the patch from www.kernel.org? I downloaded linux-2.6.19.1.tar.bz2 and patch-2.6.20-rc3.bz2 from a kernel.org mirror, then--with both linux-2.6.19.1.tar.bz2 bunziped and in its original form--I tried

JoseJX · Retired Dev Joined: 28 Apr 2002 Posts: 2774

The _rc patches are against the previous 2.6.x release. So you need to patch against 2.6.19, not 2.6.19.1. Usually, there's a version in vanilla-sources, but I guess someone's falling behind. :p I'll look into it.
_________________
Gentoo PPC FAQ: http://www.gentoo.org/doc/en/gentoo-ppc-faq.xml

iMike · Posted: Sat Jan 06, 2007 1:56 pm Post subject:

No worries about getting 2.6.19 in portage. I downloaded it directly from a kernel.org mirror. I then successfully applied the patch for 2.6.20-r3. The result was at least different. It gives an oops at the point of the trouble earlier (i.e., init), but then hangs on checking the root file system. I double checked that there is nothing wrong with root, so I think the problem there is it never successfully comes out of the kernel oops. :cry:

So, it was back to the future. I tried to get gentoo-sources for 2.6.12-gentoo-r6, the same kernel on the 2005.1 CD that successfully boots. Unfortunately, that is not in portage. I went ahead and downloaded 2.6.12.6 from kernels.org and compiled that. To my extremely pleasure, it booted perfectly.

Now I can skip the drudgery of booting from CD and manually mounting all those LVM volumes by hand just to check a kernel! Hooray!

I don't think it's quite right to mark this post as SOLVED, but perhaps it is for now. The real problem is of course: when did the kernel stop supporting this particular PowerPC model, and what should one do about that? I don't have a ton of time to debug kernel problems. Are there good tools in Gentoo portage for trying that? Any other advice?

Best regards,
/iMike

JoseJX · Retired Dev Joined: 28 Apr 2002 Posts: 2774

Well, the best thing to do would be to determine the kernel which actually screws up. Try all of the kernels between the known working and the known not working. The easiest way to "debug" this is actually to determine which patch broke it, otherwise, there's too many possibilities to easily determine the issue. Once you have found the kernel version that broke it, you can perform a git bisect to determine which patch is the culprit. Yes, it's time consuming, but it's really the easiest way.
_________________
Gentoo PPC FAQ: http://www.gentoo.org/doc/en/gentoo-ppc-faq.xml

iMike · Posted: Fri Jan 12, 2007 2:51 pm Post subject:

OK, JoseJX, I have now been busy compiling kernels for the machine and can report that the problem lies between vanilla-2.6.14.7 (the last of the 2.6.14 line) and vanilla-2.6.15.1 (the first of the 2.6.15 line). This took a fair amount of time to determine, so I'd like to see the information used somehow. But what now? Is this something I take up with the kernel.org folks? Are there some tools in gentoo to help? Where do I find out about the "git bisection" you mention?

Best regards,
/iMike

BTW: The following vanilla kernels were tested and found to work:

JoseJX · Retired Dev Joined: 28 Apr 2002 Posts: 2774

Okay, that's a good first step. (no groaning!) :)

Now that you know where the problem is occurring, we need to figure out which patch was the patch that actually broke things. To do this, we'll use the git bisect procedure as outlined here:
http://www.kernel.org/pub/software/scm/git/docs/howto/isolate-bugs-with-bisect.txt

Basically, you'll want to start like this:

iMike · Posted: Thu Jan 18, 2007 3:13 pm Post subject:

Thanks for the further information. I am now trying it, but have run into another snag. My current idea for a work around is to disable sound and try again. Here is exactly what happened. Note that the starting point for the following commands is in directory /root:

JoseJX · Retired Dev Joined: 28 Apr 2002 Posts: 2774

I wouldn't worry about it since it's just modules, and like you said, you're not using any. :)
_________________
Gentoo PPC FAQ: http://www.gentoo.org/doc/en/gentoo-ppc-faq.xml

iMike · Posted: Sat Jan 20, 2007 4:36 pm Post subject:

I've been through a few iteration of git bisect now (tested 5 kernels), but unfortunately, hit a (real) problem now. After the original:

JoseJX · Retired Dev Joined: 28 Apr 2002 Posts: 2774

To be honest, it looks like you could just disable EHCI (USB 2.0) support and it should still compile fine. OCHI USB 1.1 should still compile fine and it looks like you could keep going.
_________________
Gentoo PPC FAQ: http://www.gentoo.org/doc/en/gentoo-ppc-faq.xml

iMike · Posted: Tue Jan 30, 2007 8:47 pm Post subject:

Your advice was right on the money. I was able to proceed through several iterations more until git bisect gives the final answer as:

JoseJX · Retired Dev Joined: 28 Apr 2002 Posts: 2774

No worries, just start again with git bisect start. Since you already know the result of the tests for the working/broken kernels for the previous tests, there's no need to do them again as long as you remember each one. :) There's probably also a way to start with a specific patch set, but I'm not sure how, I'd have to check the manual.
_________________
Gentoo PPC FAQ: http://www.gentoo.org/doc/en/gentoo-ppc-faq.xml