Re: [Xen-devel] Single step in HVM domU on Intel machine may see wrong DB6

["Zhang, Yang Z" <yang.z.zhang@xxxxxxxxx>]

Jan Beulich wrote on 2014-03-11:
>>>> On 11.03.14 at 03:10, "Zhang, Yang Z" <yang.z.zhang@xxxxxxxxx> wrote:
>> Jan Beulich wrote on 2014-03-07:
>>>>>> On 07.03.14 at 06:10, "Zhang, Yang Z" <yang.z.zhang@xxxxxxxxx>
> wrote:
>>>> @@ -2689,10 +2698,18 @@ void vmx_vmexit_handler(struct
> cpu_user_regs *regs)
>>>>               */
>>>>              __vmread(EXIT_QUALIFICATION, &exit_qualification);
>>>>              HVMTRACE_1D(TRAP_DEBUG, exit_qualification);
>>>> -            write_debugreg(6, exit_qualification | 0xffff0ff0); -
>>>>       if ( !v->domain->debugger_attached ||
> cpu_has_monitor_trap_flag )
>>>> -                goto exit_and_crash; -
>>>> domain_pause_for_debugger(); +            exit_qualification |=
>>>> 0xffff0ff0;
>>> 
>>> Is this really needed?
>> 
>> Yes. The reserved bits need to set to 1 in DB6. But it is cleared in
>> exit_qualification.
> 
> In which case I'd strongly suggest adding a respective #define to
> debugreg.h and using it here.

OK.

> 
>>>> +            if ( v->domain->debugger_attached )
>>>> +            {
>>>> +                write_debugreg(6, exit_qualification);
>>>> +                domain_pause_for_debugger();
>>>> +            }
>>>> +            else
>>>> +            {
>>>> +                __restore_debug_registers(v);
>>>> +                write_debugreg(6, exit_qualification |
>>> read_debugreg(6));
>>> 
>>> I still wonder whether it wouldn't be more efficient to simply or
>>> exit_qualification into v->arch.debugreg[6] before calling
>>> __restore_debug_registers().
>>> 
>> 
>> __restore_debug_registers() only copy the v->arch.debugreg[6] into
>> hardware DB6 when flag_dr_dirty is cleared. So as I mentioned before,
>> the hardware DB register will hold the latest value if flag_dr_dirty is
>> set and we should write hardware DB6 directly.
> 
> Except that, as per the earlier discussion, the intercept should not
> occur when flag_dr_dirty is set, as it ought to have got disabled
> when setting the flag while restoring debug registers.
> 

Yes, I see your point.

> Jan


Best regards,
Yang



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel