RE: [Xen-devel] [PATCH]x86:x2apic: Disable x2apic on x86-32 permanently

["Wei, Gang" <gang.wei@xxxxxxxxx>]

Jan Beulich wrote on 2011-02-09:
>>>> On 09.02.11 at 09:43, "Wei, Gang" <gang.wei@xxxxxxxxx> wrote:
>> Jan Beulich wrote on 2011-02-09:
>>>>>> On 09.02.11 at 07:57, "Wei, Gang" <gang.wei@xxxxxxxxx> wrote:
>>>> Jan Beulich wrote on 2011-01-31:
>>>>> How does this (namely the x2apic_enabled part) play together with
>>>>> the selection of the APIC driver, which in this case would have
>>>>> happened quite a bit earlier (from generic_apic_probe())?
>>>>> 
>>>>> I would therefore think that this change really belongs into
>>>>> check_x2apic_preenabled().
>>>> 
>>>> It is really a problem. But I do think we shall simply remove
>>>> below line from check_x2apic_preenabled() fn:
>>>> 
>>>>          genapic = apic_x2apic_probe(); The same line exists in
>>>> x2apic_bsp_setup() fn.
>>> 
>>> No, that would be exactly the wrong way round - we need genapic to
>>> be set early in case x2apic was pre-enabled (see -unstable c/s 22707).
>> 
>> I could not get the reason to set genapic as x2apic in case x2apic
>> was pre-enabled while referring to c/s 22707. Genapic->xxx was never
>> called before calling x2apic_bsp_setup(). Are you trying to keep
>> consistence (x2apic pre-enabled, so genapic have to be x2apic as
>> early as
> possible)?
> 
> Without setting genapic here, generic_apic_probe() would set it to
> &apic_default, which is obviously wrong (even if it later gets
> overridden) - just look at the various APIC driver related messages
> that can result on such a system (if you're trying to analyze a
> problem, seeing the APIC driver change perhaps multiple time in the log is 
> certainly not helpful).
> 
> Apart from that - what problem do you see with just moving the change
> you did into check_x2apic_preenabled()? We're using below patch on
> 4.0.x as replacement/extension to your -unstable c/s 22789 (i.e. a
> patch against current -unstable would need to revert some of what your change 
> did).

I just want to make it simple. Both your reason and patch are nice. So could 
you please send a formal patch for 4.1?

Jimmy



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel