failing to boot pass mdadm monitor

[Peter M. Petrakis]

On 03/13/2012 02:47 PM, Asif Iqbal wrote:
> On Tue, Mar 13, 2012 at 12:35 AM, Peter M. Petrakis
> <peter.petrakis at canonical.com>  wrote:
>>
>>
>> On 03/13/2012 12:09 AM, Asif Iqbal wrote:
>>>
>>> On Mon, Mar 12, 2012 at 4:13 PM, Asif Iqbal<vadud3 at gmail.com>    wrote:
>>>>
>>>> I am failing to boot this server x4270 pass the mdadm --monitor.
>>>> Installed lucid amd64. This is a first install.
>>>>
>>>> details: http://paste.ubuntu.com/880895/
>>>>
>>>> It boots all the way in recovery mode. what gives?
>>>
>>>
>>> ok. I reinstalled lucid 64bit. No more GPT error. But it hangs right
>>> after mdadm --monitor
>>
>>
>> At least that's an improvement.
>>
>>>
>>> http://paste.ubuntu.com/881409/
>>
>>
>> ...
>>
>> mdadm: sending ioctl 1261 to a partition!
>>
>> (I confused syscall with storage ioctl earlier, whoops :)
>>
>> Now that we're just down to this problem, a little google...
>>
>> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=656899:
>> https://bugzilla.redhat.com/show_bug.cgi?id=783823
>> https://bugzilla.redhat.com/show_bug.cgi?id=783955
>>
>> http://www.serverphorums.com/read.php?12,436266
>>
>> and ... I think there's a fix:
>> https://lkml.org/lkml/2012/1/24/136
>>
>> Probably worth filing a bug at this point, also please try and
>> reproduce this issue with oneiric, and the latest mainline
>> kernel  [1] if possible. Thanks.
>
> installed oneiric this time. failed to mount /var, /home, /opt and
> /usr/local LVs.
> I had to run vgchange -a y to make those LVs available and then
> booted all the way. details: http://paste.ubuntu.com/882168/
>
> And I choose skip.. it does boot all the way but not so useful w/o /var
> as shown below.

So if you read the Debian bug report, you would find that this issue
is a result of recent security update that tightened ioctl access
to disks, virtual machines were able to write to backing stores. It
appears to require some tuning which is the basis for the fix I posted earlier.

At this point you have to follow through with the rest of my original
advice, and also consider going back one or two kernel revs where
this security fix doesn't exist. That should alleviate the 'ioctl 1261'
issue, and the bug you're supposed to report will get the requisite patch
into the next version of the distro kernel so you can move forward.

So in lucid:
https://launchpad.net/ubuntu/+source/linux/2.6.32-39.86
   * block: fail SCSI passthrough ioctls on partition devices
     - LP: #926321

and oneiric:
https://launchpad.net/ubuntu/+source/linux/3.0.0-16.29
   * block: fail SCSI passthrough ioctls on partition devices
     - LP: #922799

I think that's the CVE that's the cause of your problems. So rewind kernels
until it's not there and verify that your problem is resolved. I think you
can take the rest from here.

Peter


>
>>
>> Peter
>>
>> 1. https://wiki.ubuntu.com/Kernel/MainlineBuilds
>>    http://kernel.ubuntu.com/~kernel-ppa/mainline/
>>
>>
>>
>>>
>>>>
>>>> --
>>>> Asif Iqbal
>>>> PGP Key: 0xE62693C5 KeyServer: pgp.mit.edu
>>>> A: Because it messes up the order in which people normally read text.
>>>> Q: Why is top-posting such a bad thing?
>>>
>>>
>>>
>>>
>>
>> --
>> ubuntu-server mailing list
>> ubuntu-server at lists.ubuntu.com
>> https://lists.ubuntu.com/mailman/listinfo/ubuntu-server
>> More info: https://wiki.ubuntu.com/ServerTeam
>
>
>