On Mon, Jun 20, 2016 at 05:38:33PM +1000, Brian May wrote:
> According to Bastian Blank the the fix for XSA 173 breaks i386
> support. The HVM domains die immediately due to a triple fault. His
> working theory is that invalid bits slip into the initial page table of
> the domain. He is recommending that we drop support for i386 in order
> that we can continue patching Xen in wheezy with the latest security
> updates.
I forgot the provide further findings.
Each HVM domain immediately dies with a triple fault:
| (XEN) hvm.c:1134:d1 Triple fault on VCPU0 - invoking HVM system reset.
| (XEN) *** Dumping Dom1 vcpu#0 state: ***
| (XEN) ----[ Xen-4.1.6.1 x86_32p debug=n Not tainted ]----
| (XEN) CPU: 1
| (XEN) EIP: 0000:[<00101520>]
| (XEN) EFLAGS: 00000002 CONTEXT: hvm guest
| (XEN) eax: 00000000 ebx: 00000000 ecx: 00000000 edx: 00000000
| (XEN) esi: 00000000 edi: 00000000 ebp: 00000000 esp: 00000000
| (XEN) cr0: 00000011 cr4: 00000000 cr3: 00000000 cr2: 00101520
| (XEN) ds: 0000 es: 0000 fs: 0000 gs: 0000 ss: 0000 cs: 0000
CR0 shows the system is in a pretty early state: it is already in
protected mode but nothing else. It dies during access of the the page
fault handler specified in CR2.
A domain in this state can't be dumped.
Regards,
Bastian
--
Lots of people drink from the wrong bottle sometimes.
-- Edith Keeler, "The City on the Edge of Forever",
stardate unknown
