On Wed, 08 Aug 2007 16:14:36 -0700 (PDT)
David Miller <[EMAIL PROTECTED]> wrote:

> From: "Luck, Tony" <[EMAIL PROTECTED]>
> Date: Wed, 8 Aug 2007 16:04:44 -0700
> 
> > This didn't raise any comment on LKML ... probably the
> > "ia64" in the Subject: deterred you all from reading it :-)
> > 
> > I don't think I understand either the migration code or the
> > virtual indexed cache issues well enough to figure this out.
> > 
> > Part 2of2 goes on to mess further with this code to cope with
> > the idiosyncracies on ia64 (keeping icache and dcache coherent).
> 
> set_pte_at() on sparc64 does all of the D-cache alias handling
> that would be necessary for page-cache pages.
> 
> Perhaps someone should see if IA64 can do similar instead of
> sprinkling the kernel with flush_whatever() calls at seemingly
> arbitrary places.
> 
> set_pte_at() is when mappings change, and you can see exactly how they
> are changing and for what address space and virutal address, therefore
> there is no reason that caching issues cannot be handled in it.
> 
> pte_clear() is implemented using set_pte_at() on sparc64, which allows
> it to cover all PTE changes made by the kernel in any context.
> 

Hmm...sprinking and messy.
Why I added sync_icache_dcache() as new macro is that the places where
sync_icache_dcache() is necessary are different from flush_(i)cache_page().
But I admit that "where sync_icache_dcache() should be inserted" is complexed.

In older version, I added flush_icache_page()/flush_cache_page() for ia64.
This can fix the problem without new macro, but nacked by Zoltan because
we have to sync icache-dcache too much.

In May,  [EMAIL PROTECTED] just moved lazy_mmu_prot_update() to proper place,
but Nacked because it just hides wrong macro definition, lazy_mmu_prot_update().
(This was 2 lines patch.)

Hmm...should I post flush_(i)cache_page() for ia64 again ?
It will be the neatest, maybe.

Thanks,
-Kame











-
To unsubscribe from this list: send the line "unsubscribe linux-arch" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to