F14 AMI does not work with High-Memory instance sizes

[Matt Solnit]

Hi everyone.  Unfortunately, when I try to launch the 64-bit Fedora 14 AMI (ami-e291668b)
as a m2.xlarge, m2.2xlarge, or m2.4xlarge instance, the instance is unresponsive.  Amazon
has confirmed that it is due to a kernel panic (please see
https://forums.aws.amazon.com/message.jspa?messageID=212820#212820 for details).

Could anyone take a look at this?  Let me know if there is any further information I can
provide.

-- Matt

[Marek Goldmann]

Hi Matt,

I can confirm this. I pasted the console log here: https://gist.github.com/757040 .
Unfortunately I cannot see cause of this. Anyone?

--Marek

On 2010-12-28, at 02:55, Matt Solnit wrote:

>  Hi everyone.  Unfortunately, when I try to launch the 64-bit Fedora
> 14 AMI (ami-e291668b) as a m2.xlarge, m2.2xlarge, or m2.4xlarge instance, the instance is
> unresponsive.  Amazon has confirmed that it is due to a kernel panic (please see
> https://forums.aws.amazon.com/message.jspa?messageID=212820#212820 for details).
>  
>  Could anyone take a look at this?  Let me know if there is any further information I can
> provide.
>  
>  -- Matt
>  _______________________________________________
>  cloud mailing list
>  cloud(a)lists.fedoraproject.org
>  https://admin.fedoraproject.org/mailman/listinfo/cloud

[Steven Dake]

On 12/28/2010 01:39 AM, Marek Goldmann wrote:
>  Hi Matt,
>  
>  I can confirm this. I pasted the console log here: https://gist.github.com/757040 .
> Unfortunately I cannot see cause of this. Anyone?
>  
>  --Marek
>  
>  On 2010-12-28, at 02:55, Matt Solnit wrote:
>  
> > Hi everyone.  Unfortunately, when I try to launch the 64-bit Fedora 14 AMI
> (ami-e291668b) as a m2.xlarge, m2.2xlarge, or m2.4xlarge instance, the instance is
> unresponsive.  Amazon has confirmed that it is due to a kernel panic (please see
> https://forums.aws.amazon.com/message.jspa?messageID=212820#212820 for details).
> >
> > Could anyone take a look at this?  Let me know if there is any further information I
> can provide.
> >
> > -- Matt
> > _______________________________________________
> > cloud mailing list
> > cloud(a)lists.fedoraproject.org
> > https://admin.fedoraproject.org/mailman/listinfo/cloud
>  
>  _______________________________________________
>  cloud mailing list
>  cloud(a)lists.fedoraproject.org
>  https://admin.fedoraproject.org/mailman/listinfo/cloud 
Keep in mind its been 5 years since I've done any real heavy kernel
development but... (and boy alot has changed!)

Looking at the stack backtrace, we see multiple traps that occur as a
result of invalid instructions being executed:

[1121923.344484] [<ffffffff8100b053>] do_invalid_op+0x98/0xa1
[1121923.344489] [<ffffffff81264975>] ? intel_idle+0xc6/0x139

and again

1121923.344525] [<ffffffff8100a95b>] invalid_op+0x1b/0x20
[1121923.344531] [<ffffffff81264975>] ? intel_idle+0xc6/0x139

instruction at address c6 in function intel_idle in the AMI is
apparently not known to the physical hardware (or xen..)

I am on PTO until Jan 4 and having just moved pretty busy, but I'll try
to decode the instruction that is invalid (this is simple with gdb and
the vmlinux image using gdb).

Could be incompatibility  between xen on the host and xen in the VM.

Regards
-steve