On Wed, Jan 20, 2021 at 8:19 AM Jason Gunthorpe <jgg@xxxxxxxx> wrote: > > On Tue, Jan 19, 2021 at 08:43:31PM -0500, Pavel Tatashin wrote: > > When pages are longterm pinned, we must migrated them out of movable zone. > > The function that migrates them has a hidden loop with goto. The loop is > > to retry on isolation failures, and after successful migration. > > > > Make this code better by moving this loop to the caller. > > > > Signed-off-by: Pavel Tatashin <pasha.tatashin@xxxxxxxxxx> > > mm/gup.c | 88 +++++++++++++++++++++++--------------------------------- > > 1 file changed, 36 insertions(+), 52 deletions(-) > > This looks OK, it is better > > Reviewed-by: Jason Gunthorpe <jgg@xxxxxxxxxx> > > I really dislike we always have to go over the page list twice in pin > mode I agree, I also dislike that we have to loop twice. > > The is_pinnable_page() and LRU isolation should really be done inside > __get_user_pages_locked() as each page is added to the output list > > But that is more of a larger issue than this series We could also think about adding some optimization flags, i.e. clients could tell gup that the pinned memory content can be discarded FOLL_DISCARD. That way we could always allocate new pages in the right zones as we do with this series and free existing translations. No migration check would be necessary with such a flag. > > Jason
