Re: [patch 1/5] x84_64, apic: Use probe routines to simplify apic selection

[Cyrill Gorcunov <gorcunov@openvz.org>]

On Mon, May 16, 2011 at 12:08 PM, Ingo Molnar <mingo@elte.hu> wrote:
>
> * Suresh Siddha <suresh.b.siddha@intel.com> wrote:
>
>> From: Suresh Siddha <suresh.b.siddha@intel.com>
>> Subject: x84_64, apic: Use probe routines to simplify apic selection
>>
>> Use the unused probe routine in the apic driver to finalize the apic model
>> selection. This cleans up the default_setup_apic_routing() and this probe
>> routine in future can also be used for doing any apic model specific
>> initialization.
>>
>> Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
>> Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
>> ---
>>  arch/x86/kernel/apic/apic_flat_64.c   |   10 +++++++++-
>>  arch/x86/kernel/apic/probe_64.c       |   22 +++++++---------------
>>  arch/x86/kernel/apic/x2apic_cluster.c |    7 ++++++-
>>  arch/x86/kernel/apic/x2apic_phys.c    |   10 +++++++++-
>>  arch/x86/kernel/apic/x2apic_uv_x.c    |    7 ++++++-
>>  5 files changed, 37 insertions(+), 19 deletions(-)
>
> Ok, looks like a step forward in the right direction.
>
> Wouldnt it be more self-contained if the probe function returned an apic driver
> pointer?
>
> We could add an __apicdriver section trick to 'know' about all APIC probing
> functions - and we'd call them one by one and use the first one that returns a
> non-NULL result.
>
> Then we'd have the generic fallback APIC drivers as well - not marked
> __apicdriver but used by the probe function directly.
>
> Or we could have them as __apicdriver as well, but then the .o link ordering
> matters to probing order.
>
> Thanks,
>
>        Ingo
>

  Yes,  this should be a cleaner step but it requires some more code churn that
is why I think Suresh' approach is pretty well done! Ingo, probably we could
do so on *top* of this changes as only time permits? The problem is that I
personally can't guarantee I continue working on the task this week, if Suresh
is able to -- this would be great, if not -- probably even such
minimal disturbing
patch worth merging, it definitley makes code better I think ;)