Re: page fault scalability patch V9: [7/7] atomic pte operatiosn for s390

[Christoph Lameter <clameter@sgi.com>]

On Tue, 28 Sep 2004, Martin Schwidefsky wrote:

> operation that needs to get implemented by use of the ipte. You'll
> find a new patch that adds an implementation of ptep_xchg_flush for
> s390 below, the old patch is broken. Please replace.

Ok. Thanks. Added to my patch.


> There are some more things that need improvement:
> * You should introduce a #define for each of the new primitives
>   (__HAVE_ARCH_PTEP_CMPXCHG, __HAVE_ARCH_PTEP_XCHG_FLUSH, etc) so
>   that the architectures can redefine the primitives separatly.
>   For s390 I need to define a ptep_xchg_flush but want to use the
>   default implementation for all the other new primitives.

Ok. done.

> * In the generic implementations of the new primitives please use
>   the mm from the vma to do the locking. To rely on the fact that
>   the memory structure in the functions that use the macros are always
>   called "mm" is quite a hack.

I did not notice that. The vma was passed to the macro but the mm
was used for locking... Amazingly it worked. Thanks...

> * The ptep_get_clear_flush primitive isn't used anywhere and should
>   be removed.

Done.

> * The flush_tlb_page in the generic ptep_cmpxchg implementation
>   shouldn't be there, the macro is used in code sequences where it
>   isn't needed (the old code didn't flush).

Removed.

Thanks for all the feedback and sorry for the late response. I was busy
with other stuff and there was a deadline upon us.

Hopefully I will get V10 out soon.

Greetings
	Christoph