On Wed, 08 Aug 2007 18:01:36 -0700 (PDT) David Miller <[EMAIL PROTECTED]> wrote:
> From: KAMEZAWA Hiroyuki <[EMAIL PROTECTED]> > Date: Thu, 9 Aug 2007 09:50:42 +0900 > > > Hmm...should I post flush_(i)cache_page() for ia64 again ? > > It will be the neatest, maybe. > > The problem with the flush_*() interfaces is that it tells the > platform "what to do" instead of giving it information like "this just > happened". > > That's why I encourage approaches that use set_pte_at() because it's > an event that gives information. Hmm, I see. Trying to write a patch will not be much work. Sigh, I didn't think this bugfix took one month... > > Different platforms need to flush when certain circumstances are > true. Therefore, the better we annotate the events in the memory > management, the easier it will be to write correct optimized flushing > code for a platform. > > But I cannot state enough that set_pte_at() and pte_clear() very > likely serves your needs completely if you try to use it. Every place > I've seen IA64 add a hook has been right before or right after a > set_pte_at() call, and this is no coincidence! > All lazy_mmu_prot_update() which were used after set_pte() have to be removed. It's buggy (my patch removes them) All cache flusing hook should be inserted before set_pte(). 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
