Karmic HP blade 460c generation 1 hp support pack hpacucli error

[William]

Linux Server 2.6.31-14-server #48-Ubuntu SMP Fri Oct 16 15:07:34 UTC 2009 x86_64 GNU/Linux

Ubuntu 2.6.31-14.48-server

Attached dmesg log when using the following commands:

        /usr/sbin/hpacucli ctrl slot=0 del forced
        /usr/sbin/hpacucli ctrl slot=0 create type=ld raid=0 aa=enable
        /usr/sbin/hpacucli ctrl slot=0 modify dwc=enable forced

When executing the second command it will result in a kernel error:
[ 838.125938] kobject (ffff8808070800d0): tried to init an initialized object, something is seriously wrong.
[ 838.126006] Pid: 5225, comm: .hpacucli Not tainted 2.6.31-14-server #48-Ubuntu

The hpacucli tool is from the HP support pack for jaunty.

This bug report is for HP certification program!

With kind regards

William

[Jerone Young (jerone)]

Should be reproducible on any HP server using the "cciss" driver

[dann frazier (dannf-hp)]

You appear to be accessing logical disk c0d0 (filesystem mounted), at the time you delete it. Can you confirm?

If this is the case, then this looks like expected behavior - akin to yanking out your boot disk at runtime.

[William (kc-cobradevil)]

Hello Dann

I tested again with latest kernel and made very sure the disk was not in use but still the same issue.

2.6.31-16-server

I boot an image in memory and access the tool from there.

So thats why i can test the tool without the disk being in use.

With kind regards

William

[dann frazier (dannf-hp)]

Thanks for that information, I can reproduce this using Ubuntu 9.10 in livecd mode.
I'll attach the kern.log, which contains a kernel backtrace.

[dann frazier (dannf-hp)]

I was able to reproduce using kernel.org 2.6.31, but not able to reproduce on latest git. I bisected the two and hit this commit:

commit 617e1344229d22ea9ecb6538e50808541618ed2b
Author: Stephen M. Cameron <email address hidden>
Date: Thu Sep 17 13:47:14 2009 -0500

    cciss: Dynamically allocate struct device for each logical drive as needed.

    Dynamically allocate struct device for each logical drive as needed
    instead of allocating the maximum we would ever need at driver init time.

    Signed-off-by: Stephen M. Cameron <email address hidden>
    Signed-off-by: Jens Axboe <email address hidden>

[Manoj Iyer (manjo)]

I will pull this patch and upload a test kernel on my people page, and submit an SRU for karmic.

[Manoj Iyer (manjo)]

Can you please test the karmic kernel in
http://people.canonical.com/~manjo/lp490290-karmic/

I had to pull in 24 or so patches, I am not sure if this will pass SRU policy. The last patch I applied is:

commit 9cef0d2f4f68a5a2c6ea0495f958a074d21fbd07
Author: Stephen M. Cameron <email address hidden>
Date: Thu Sep 17 13:48:31 2009 -0500

    cciss: Dynamically allocate the drive_info_struct for each logical drive.

This patch seems to modify some of the code in patch pointed to by dann.

But please test this kernel and post any test results to this here.

[William (kc-cobradevil)]

Hello Manoj

I confirm that the bug is gone with the kernel http://people.canonical.com/~manjo/lp490290-karmic/.

This feels better ;)

With kind regards

William

[Manoj Iyer (manjo)]

William,

Thanks a ton for verifying the kernel fixes, I am working with the Ubuntu stable maintainer to see if I can SRU these patches.

Cheers
manjo

[William (kc-cobradevil)]

Hello Manoj

no thanks, i'm happy when it works like expected.
And it's good to get ubuntu 9.10 hp certified for the upcoming lts version.

i'm happy when i can run 28 servers with the ram image to serve 4200 firefox sessions for internet ;)

With kind regards

William van de Velde

[Manoj Iyer (manjo)]

I am going to close the bug in Karmic (9.10). Lucid (release end of this month) 10.04 should have these patches and fixes, Lucid is our LTS release so you should be updating to Lucid anyways. If this bugs occurs on Lucid please reopen the bug on Lucid.

[Manoj Iyer (manjo)]

Making fixed released in Lucid.

[Manoj Iyer (manjo)]

Marking fix released in Lucid.

[Manoj Iyer (manjo)]

Marking as fix released in Lucid.