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

[Linus Torvalds <torvalds@transmeta.com>]

On Sun, 7 Jul 2002, Andrew Morton wrote:
>
> Probably the biggest offenders are generic_file_read/write.  In
> generic_file_write() we're already faulting in the user page(s)
> beforehand (somewhat racily, btw).  We could formalise that into
> a pin_user_page_range() or whatever and use an atomic kmap
> in there.

I'd really prefer not to. We're talking of a difference between one
single-cycle instruction (the address should be in the TLB 99% of all
times), and a long slow TLB walk with various locks etc.

Anyway, it couldn't be an atomic kmap in file_send_actor anyway, since the
write itself may need to block for other reasons (ie socket buffer full
etc). THAT is the one that can get misused - the others are not a big
deal, I think.

So kmap_atomic definitely doesn't work there.

		Linus

--
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/