Kernel Oops with Maverick in ahci_stop_engine

Probably related to
https://bugzilla.kernel.org/show_bug.cgi?id=16228

I'd built a test kernel for a different bug but it includes the patches from the upstream bugzilla report you noted here.

https://bugs.edge.launchpad.net/ubuntu/+source/linux/+bug/647043/comments/14 .

Give it a test and let us know your results. Thanks.

Dell T3500

worked just fine with 10.04

upgraded to 10.10 using update manager, and it won't boot. I get white text on black background, not even sure what I am looking at... dropped me into some kind of shell?

tried booting from the 10.10 install CD (which I burned to install on another machine) also won't boot off the install CD

get the Ubuntu logo, with dots underneath, then it again drops me into a shell.

back to booting from disk, but this time choosing the older kernel on the boot menu, and it booted up just fine. and seems to work fine.

not an expert, not sure what's wrong, and how to report this in more details.

Try the above kernel and then update your BIOS. I am going to do this
tomorrow. I can tell you then how it worked.
It seems to be a BIOS problem.

Thank you for your help.

I believe that I have the latest BIOS. No I don't, upgrading to A08 dated 9/17/10 right now.

I am not sure how to change kernels, do I use the update manager?

I have only two kernels in the boot choice (plus memtest and the old Windows partition), and only the older kernel is bootable.

I am afraid that if I add another kernel, the older (working one) will be deleted, and I'll end up with two "bad" kernels

Is there a way to keep more than two kernels?

-avi

FYI:

found this work around in another thread: https://bugs.launchpad.net/ubuntu/+bug/659149

Following this link : http://lkml.org/lkml/2010/6/16/323 I found a temporary solution : by inserting pci=nocrs in the kernel line in grub

Ok, I tested two things.

1. The kernel from comment #2 works.
2. Upgrading to the latest BIOS provided by DELL (A08, from A03) doesn't work with the old kernel.

So:
Kernel from #2 works with BIOS A03 and A08, stock Maverick kernel doesn't work with both BIOSs.

Should I
a) communicate this upstream,
b) file a bug at DELL?

Thanks for the realtime solution. :)

g

@Avi:

Download the deb packages from #2, reboot your old kernel, install new packages by "sudo aptitude install *.deb" and don't worry about removing old kernels, everything is preserved. Boot into your new shiny 2.6.25 kernel that actually boots. :)
Well, at least it worked for me on my T3500.

g

@Leann: Your kernel is installed as linux-image-2.6.35-22-generic.
Does this means it's going to be overwritten by the next round of
kernel updates when running update-manager?

g

as mentioned earlier, the A08 BIOS update did not resolve the problem.

pci=nocrs does work with the 10.10 Ubuntu kernel and A08 BIOS (did not try the older kernel with A08)

did not yet try the test patched kernel from #2, since I can boot with pci=nocrs and I need to get some real work done...

-avi

I just ran into this problem. And I can confirm post #6, adding the pci=nocrs to grub.conf solves the problem. Another workaround is to boot an older kernel version, some 2.6.34.xxx.

https://bugzilla.kernel.org/show_bug.cgi?id=16228 tells of a patch for fixing this is available.

And there they tell also, that there is no BIOS for the Dell T3500 available resolving the problem.

So IMHO for ubuntu either the pci=nocrs should go into the grub.conf (for Dell T3500) or - preferred - the mentioned patch should be included ... ?

@Gonzhauser, thanks for trying the test kernel noted in comment #2 and confirming it fixes the issue. The next round of kernel updates would supersede the test kernel. However, it appears the next official Maverick kernel update will be an ABI bump, so you would have the option of selecting the older test kernel or the newer updated kernel.

Getting these patches accepted as a Maverick stable release update will be a bit difficult (ie it's not exactly appropriate to qualify for SRU). Even Linus himself has refused to accept these into the upstream 2.6.36 kernel as noted in https://patchwork.kernel.org/patch/189182/

"I definitely am not going to pull this series for 2.6.36.

I could possibly take this first one that only prepares for the real
change and doesn't actually change anything in itself, but switching
around the order of allocations after -rc5 would be crazy. Yes, it may
help some people, but we have absolutely no idea who it could hurt. So
the whole thing is definitely something for the merge window (and
preferably pretty early there too)"

I have to agree that this is a very risky change with a high risk of regression which is always something that one tries to avoid when introducing changes/fixes to a released kernel. I still need to discuss with the rest of the Ubuntu Kernel SRU team, but we may just have to wait for this to resolve itself in Natty.

On Wed, Oct 13, 2010 at 6:29 PM, Leann Ogasawara
<email address hidden> wrote:
> @Gonzhauser, thanks for trying the test kernel noted in comment #2 and
> confirming it fixes the issue.  The next round of kernel updates would
> supersede the test kernel.  However, it appears the next official
> Maverick kernel update will be an ABI bump, so you would have the option
> of selecting the older test kernel or the newer updated kernel.

I installed your kernel again after running upgrade-manager and it
still works but I am waiting for the day when I will forget... :(
I am willing to bisect this but you/Red Hat/Dell/Linux
Foundation/whoever should seriously consider adding a T3500 to your
test configurations. ;)

g

For the record: I found the "Lock version" in synaptic. Everything
should be fine.

This is a duplicate of bug #653238, not of bug #647043.

Bug #653238 was marked a duplicate of bug #647043, but that is incorrect.