Re: Xen 3.2.1-rc1: FATAL PAGE FAULT

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

On 3/4/08 15:04, "Christopher S. Aker" <caker@theshore.net> wrote:

>>> Maybe the symbols would help just a little bit?  In any case, here are
>>> the files:
>>> 
>>> http://theshore.net/~caker/xen/BUGfatal_page_fault/
>> 
>> I will take a look. It might help narrow down the possibilities a bit.
>> 

My analysis is that the hypervisor crashed because one of the entries in a
dying guest's third-level page directory has the present bit (bit 0) set,
yet the physical address mapped by that entry is 0xb72dedde51000. That is a
rather large and obviously bogus number! It causes us to access way off the
end of an array indexed by physical address, resulting in a fatal page
fault.

Obviously the question is: Where did the bogus address come from?

That's going to be rather hard to answer without finding a more reliable
repro of the bug, and then adding some hypervisor tracing.

 -- Keir