Re: [PATCH 0/5] x86/paravirt: clean up paravirt.h and optimise FPU context switch

[Ingo Molnar <mingo@elte.hu>]

* Jeremy Fitzhardinge <jeremy@goop.org> wrote:

> Hi Ingo,
> 
> This series:
>  - splits asm/paravirt.h into paravirt.h and paravirt_types.h
>  - splits out the context switch core of math_state_restore into __math_state_restore
>  - rearranges 32 and 64 bit context switch to be the same and to properly
>    batch FPU context switches when running paravirtualized
> 
> There should be no performance side-effects of these changes.  They amount
> to a small rearrangement of the existing code (and bringing the 64-bit
> code a bit closer to 32-bit, countering some genetic drift).
> 
> I think these are probably safe to go into this merge window, but I
> don't think I've posted these before, so I understand if you're hesitant
> (I've had them applied locally for some time with no ill-effects).
> The paravirt cleanup is pure movement between headers, so I don't think
> there's much risk there at all.
> 
>  arch/x86/include/asm/i387.h           |    1 +
>  arch/x86/include/asm/paravirt.h       |  711 +--------------------------------
>  arch/x86/include/asm/paravirt_types.h |  720 +++++++++++++++++++++++++++++++++
>  arch/x86/kernel/process_32.c          |   27 +-
>  arch/x86/kernel/process_64.c          |   33 +-
>  arch/x86/kernel/traps.c               |   33 +-
>  6 files changed, 782 insertions(+), 743 deletions(-)

Looks nice!

I havent checked all headers but do we now use the principle that we 
only include paravirt_types.h into other headers - and paravirt.h 
only into .c code that makes use of methods?

I'll try to find some time to pull + test + push it - it might miss 
the merge window, no promises.

	Ingo