Re: [patch 2/2] lockless get_user_pages

[Nick Piggin <npiggin@suse.de>]

On Tue, May 27, 2008 at 11:46:27AM +0900, KOSAKI Motohiro wrote:
> > Aw, nobody likes fast_gup? ;)
> 
> Ah, I misunderstood your intention.
> I thought you disklike fast_gup..
> 
> I don't dislike it :()
> 
> > 
> > Technically get_user_pages_lockless is wrong: the implementation may
> > not be lockless so one cannot assume it will not take mmap sem and
> > ptls.
> 
> agreed.
> 
> 
> > But I do like to make it clear that it is related to get_user_pages.
> > get_current_user_pages(), maybe? Hmm, that's harder to grep for
> > both then I guess. get_user_pages_current?
> 
> Yeah, good name.

Hmm, although now that I think about it more, fast_gup is not _quite_
just a get_user_pages for "current". In particular it requires a bit
of thought as to whether the pages are likely to require page faults
or not... so I've called it get_user_pages_fast()