Broken CONFIG_HIGHPTE

[Kumar Gala]

On Apr 1, 2009, at 12:38 AM, Benjamin Herrenschmidt wrote:

> Hi !
>
> Just noticed... And I'll be travelling for a while so with no much
> time to look at it.
>
> Your changes to use the PTE page free batch via RCU on ppc32 broke
> CONFIG_HIGHPTE for hash32...
>
> You rely on passing a page_address() of the PTE page down to the
> freeing layer. That doesn't work for highmem pages. That needs to
> be changes to a struct page * on 32-bit (not on 64-bit where we
> are using slabs though)

I can look into this, how does this show up?

- k