Messages in this thread | | | Date | Mon, 14 Jul 2008 15:01:45 -0500 | From | Jack Steiner <> | Subject | Re: [PATCH] - GRU virtual -> physical translation |
| |
On Mon, Jul 14, 2008 at 02:50:18PM -0500, Robin Holt wrote: > On Mon, Jul 14, 2008 at 11:31:07AM -0500, Jack Steiner wrote: > > On Mon, Jul 14, 2008 at 09:24:51AM -0700, Andrew Morton wrote: > > > On Mon, 14 Jul 2008 09:52:55 -0500 Jack Steiner <steiner@sgi.com> wrote: > > > > > > > On Fri, Jul 11, 2008 at 12:17:36PM -0700, Andrew Morton wrote: > > > > > On Wed, 9 Jul 2008 14:14:39 -0500 Jack Steiner <steiner@sgi.com> wrote: > > > > > > > > > > > Open code the equivalent to follow_page(). This eliminates the > > > > > > requirement for an EXPORT of follow_page(). > > > > > > > > > > I'd prefer to export follow_page() - copying-n-pasting just to avoid > > > > > exporting the darn thing is silly. > > > > > > > > If follow_page() can be EXPORTed, I think that may make the most sense for > > > > now. > > > > > > What was Christoph's reason for objecting to the export? > > > > No clue. Just a NACK. > > > > Christoph??? > > Maybe I missed part of the discussion, but I thought follow_page() would > not work because you need this to function in the interrupt context and > locks would then need to be made irqsave/irqrestore.
Arggg. You are right. I forgot the issue with the pte locks.
Looks like I am back to the plan suggested by Nick. I'll add a get_user_pte_fast() function to gup.c.
The pte lookup function is very similar to the get_user_pages_fast() function.
Ignore the previous noise about follow_page().
--- jack
| |