Re: New release candidate for Xen 4.0.0 (RC9)

[Keir Fraser <keir.fraser@eu.citrix.com>]

On 31/03/2010 22:07, "Konrad Rzeszutek Wilk" <konrad.wilk@oracle.com> wrote:

>> I'm only interested in fixing regressions right now for 4.0.0. If this bug
>> has existed since at least 3.4 then the fix (when we have one!) can wait a
>> bit longer, for 4.0.1. My guess is that the MTRR interface is getting little
> 
> Yeah. I get the same failure with 3.4 from xen-3.4-testing:

I didn't reproduce this with my setup, but the crash below is because
spin_unlock() is finding that the lock is not locked on entry. This is the
set_atomicity_lock as acquired/released by prepare_set/post_set. Looking at
the code they seem to be used in matched pairs and I can't see any memory
corruption issues or anything. If you want to get this fixed then you'll
have to add some tracing to find out what's going on (e.g, dump the lock
state at the end of prepare_set and start of post_set). If you can reproduce
this deterministically with your setup then should be pretty easy to nail
this bug.

 -- Keir

> (XEN)    [<ffff828c8011bd14>] _spin_unlock+0xc/0x22
> (XEN)    [<ffff828c80184265>] post_set+0x70/0x72
> (XEN)    [<ffff828c801843d7>] generic_set_mtrr+0xd9/0xf3
> (XEN)    [<0000000000000002>] ???
> (XEN)    
> (XEN) 
> (XEN) ****************************************
> (XEN) Panic on CPU 0:
> (XEN) Assertion '((lock)->lock <= 0)' failed at
> /home/konrad/git/tip/nex-VI/xen/xen/include/asm/sp:18
> (XEN) ****************************************
> (XEN)