I think the problem lies here:
/*
* If the guest has used debug registers, at least dr7
* will be disabled while returning to the host.
* If we don't have active breakpoints in the host, we don't
* care about the messed up debug address registers. But if
* we have some of them active, restore the old state.
*/
if (hw_breakpoint_active()) {
hw_breakpoint_restore();
}
First we need to reload the DRs after we do hw_breakpoint_restore(), no?
vcpu->arch.switch_db_regs |= KVM_DEBUGREG_RELOAD;
Second, I am unsure whether hw_breakpoint_active() is the correct condition.
In my defense note that my tests did not his this case since I set affinity
of the VCPUs to physical cores, and prevented other processes from running.
Regards,
Nadav
Paolo Bonzini <pbonzini@xxxxxxxxxx> wrote:
>
>
> On 29/01/2016 23:21, Andrew Vagin wrote:
>> On Thu, Jan 28, 2016 at 02:42:25PM -0800, Andrey Wagin wrote:
>>> On Thu, Jan 28, 2016 at 10:33:28PM +0100, Paolo Bonzini wrote:
>>>> On 28/01/2016 09:31, Andrey Wagin wrote:
>>>>> I tried to print drX registers after a break-point. Looks like they
>>>>> are set correctly.
>>>>
>>>> Can you try this KVM patch?
>>>
>>> Looks like it fixes a case when reproducers are running only in VM.
>>
>> Actually Oleg's reproducer detects the bug with this patch when they are
>> rinning only in VM.
>
> That's actually a good thing, because the patch was a long shot and I
> had no clue _why_ it would have fixed the bug. Oleg's reproducer
> spanning host and a VM actually gives me an idea of what is going on,
> I'll try to reproduce this week.
>
> Paolo
> --
> To unsubscribe from this list: send the line "unsubscribe kvm" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
