Patch reversion needed too for -ac and 2.5.x

Hi David :)

I understand why you suggested the reversion, and, unfortunately,
the MM code is plagued by bad usage of the PAGE_ALIGN macro. This macro fails for aligning a *size* when this size is larger than SIZE_MAX-PAGE_SIZE. Too bad.

Anyway, if you want the patch reverted, tell too to Alan Cox and
Linus, because their branches have this patch applied and well, although I think that the patch does no harm, really the solution is to write a new macro for aligning sizes, not addresses, and this is a very big change that surely won't be accepted.

Tell Alan and Linus for having the patch reverted for their trees
if you feel like that. Do you think that a new macro will solve the corner case without doing nothing on 'big TASK_SIZE' archs? If you think so, I can provide with a macro and have it checked by you :)

Don't take it bad, Dave ;) It's just that althoug my solution was
not perfect, certainly is best than what we have now ;)

Raúl