Re: scalable kmap (was Re: vm lock contention reduction)

[Benjamin LaHaise <bcrl@redhat.com>]

On Mon, Jul 08, 2002 at 01:39:04PM -0700, Andrew Morton wrote:
> I think I'll just go for pinning the damn page.  It's a spinlock and
> maybe three cachelines but the kernel is about to do a 4k memcpy
> anyway.  And get_user_pages() doesn't show up much on O_DIRECT
> profiles and it'll be a net win and we need to do SOMETHING, dammit.

Pinning the page costs too much (remember, it's only a win with a 
reduced copy of more that 512 bytes).  The right way of doing it is 
letting copy_*_user fail on a page fault for places like this where 
we need to drop locks before going into the page fault handler.

		-ben
-- 
"You will be reincarnated as a toad; and you will be much happier."
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/