[karmic] linux-image 2.6.31-1.13 crashes on boot

[Roland Dreier]

On my Lenovo ThinkPad X200, the recently built 2.6.31-1.13 kernel crashes during boot with a NULL dereference in acpi_get_pci_dev(). This is a bug in 2.6.31-rc1 that has already been fixed in the upstream commit 412af978 ("ACPI: video: prevent NULL deref in acpi_get_pci_dev()"). It would be good to get a new 2.6.31-rc1 build with at least this commit cherry-picked, or possibly just to update to the latest upstream; the changelog for this commit mentions that the same failure occurs on Dell Latitude E6400, Dell XPS M1330 and Lenovo ThinkPad X61, so this is likely to hit a lot of people if linux-meta is updated to the current 2.6.31 kernel.

ProblemType: Bug
Architecture: amd64
Date: Fri Jun 26 14:08:49 2009
DistroRelease: Ubuntu 9.10
HibernationDevice: RESUME=UUID=a4f94a8f-1f9c-4611-9090-37d2072dc9a7
MachineType: LENOVO 7458B18
Package: linux-image-2.6.30-10-generic 2.6.30-10.12
ProcCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.30-10-generic root=UUID=aa241de1-af8e-490b-ad56-ad5be9aac3e5 ro
ProcEnviron:
 LANG=en_US.UTF-8
 SHELL=/bin/bash
ProcVersionSignature: Ubuntu 2.6.30-10.12-generic
RelatedPackageVersions: linux-backports-modules-2.6.30-10-generic N/A
SourcePackage: linux
Uname: Linux 2.6.30-10-generic x86_64
dmi.bios.date: 05/18/2009
dmi.bios.vendor: LENOVO
dmi.bios.version: 6DET53WW (3.03 )
dmi.board.name: 7458B18
dmi.board.vendor: LENOVO
dmi.board.version: Not Available
dmi.chassis.asset.tag: No Asset Information
dmi.chassis.type: 10
dmi.chassis.vendor: LENOVO
dmi.chassis.version: Not Available
dmi.modalias: dmi:bvnLENOVO:bvr6DET53WW(3.03):bd05/18/2009:svnLENOVO:pn7458B18:pvrThinkPadX200:rvnLENOVO:rn7458B18:rvrNotAvailable:cvnLENOVO:ct10:cvrNotAvailable:
dmi.product.name: 7458B18
dmi.product.version: ThinkPad X200
dmi.sys.vendor: LENOVO

[Roland Dreier (roland.dreier)]

I just pulled the upstream tree up to commit 4075ea8c into the karmit kernel tree and rebuilt it, and that boots fine on my X200.

[Roland Dreier (roland.dreier)]

The 2.6.31-rc1+(git fix for boot crash) kernel seems to fix a *lot* of problems seen with 2.6.30 on my ThinkPad X200 laptop: the intel X driver gets the right screen size to avoid things cut off on the right, suspend/resume works with KMS, and compiz works with KMS. So getting a fixed 2.6.31-rc1+git build and a linux-meta update out soon would probably be a good idea.

[Robert Hooker (sarvatt)]

There is no linux-meta for 2.6.31 yet and considering the sheer number of things that do not build against it yet (anything block device/i2c/agpgart/or using the old net_device ops that builds modules such as nvidia fglrx bcmwl and vbox) I have a feeling the meta update might end up waiting a bit :)

[Roland Dreier (roland.dreier)]

Hmm, sounds unfortunate that closed/non-free or out-of-tree things block 2.6.31 from reaching testers when (as I mentioned in my previous comment) so many things are fixed for Intel graphics in the new kernel. Oh well, I can continue to use my own builds...

[christoph sturm (christoph-sturm)]

can you put your kernel package into a ppa?

[Roland Dreier (roland.dreier)]

The Ubuntu kernel build system is hard enough that I have no idea how to get source packages I could upload to a ppa -- just building packages locally is pretty much the limit of my understanding of building Ubuntu kernels. However if you get the latest daily upstream kernel build (which is available from a kernel team ppa) then it should have all the fixes required.

[Roland Dreier (roland.dreier)]

What went wrong here? It seems linux-meta 2.6.31.1.11 was uploaded on June 29, 3 days after I reported that kernel 2.6.31-1.13 would not boot on a large class of machines, and giving the exact upstream changelog required to fix the bug. I understand that using versions of Ubuntu that are in the early alpha stage is likely to run into breakage, but how can we fix the process to avoid updating testers to a known-broken kernel with a known fix? Unfortunately a lot of testers' time is going to be wasted by this bug, and also I'm sure a lot of developer time is going to be wasted marking other bugs as dupes of this one.

[Jeffrey Baker (jwbaker)]

Awesome. Uploading an unbootable kernel to the archive is pretty dumb, even for an Alpha. Possibly someone should take ownership and set the importance....

[DSHR (s-heuer)]

I can confirm this on a thinkpad X60 with kernel 2.6.31-1

[Roland Dreier (roland.dreier)]

Thanks, but I don't think further confirmations are needed at this point. The bug is already well-confirmed, and the fix is known. It's just a matter of getting a kernel with the fix uploaded to the archive.

[Roland Dreier (roland.dreier)]

I see from the karmic-changes list that linux 2.6.31-1.14 has been uploaded with the fix for this bug; however the ubuntu-karmic.git tree on kernel.ubuntu.com does not seem to include the change in question. Is it typical not to update the git tree with everything in the kernel package actually used in the archive?

[Christophe Dumez (hydr0g3n)]

Finally, I found this bug report. I cannot boot on my DELL XPS M1330 since I updated to Karmic.

I know this problem is already confirmed and you don't need any more confirmation but I would like to say that recent versions still do not boot on my computer.

I get a lot of input/output errors on boot (startup is stuck, typing "sudo startx" already returns an input/output error).
I'm using kernel 2.6.31-2.16.

[Christophe Dumez (hydr0g3n)]

Here is the output with kernel 2.6.31-2.16. I no longer have a crash/backtrace but I get a lot of Input/output errors. I'll post dmesg too.

[Christophe Dumez (hydr0g3n)]

Oh, I do get a backtrace when I boot in safe mode.

[Brian Rogers (brian-rogers)]

Christophe,

I can see from that backtrace that the crash you are seeing is unrelated to this bug. So please file a new bug report and attach the same information in the new report. Thanks.

[Andy Whitcroft (apw)]

The panic in acpi_get_pci_dev() was fixed in 2.6.31-1.14. If your panic is not occuring in acpi_get_pci_dev() then you have a different problem and need to file a new bug. Closing Fix Released.