On 19-06-27 00:21:27, Chaitanya Kulkarni wrote: > > > On 6/26/19, 1:52 PM, "linux-block-owner@xxxxxxxxxxxxxxx on behalf of Minwoo Im" <linux-block-owner@xxxxxxxxxxxxxxx on behalf of minwoo.im.dev@xxxxxxxxx> wrote: > > On 19-06-26 15:49:21, Christoph Hellwig wrote: > > A lot of callers of bio_release_pages also want to mark the released > > pages as dirty. Add a mark_dirty parameter to avoid a second > > relatively expensive bio_for_each_segment_all loop. > > > > Signed-off-by: Christoph Hellwig <hch@xxxxxx> > > --- > > block/bio.c | 12 +++++++----- > > include/linux/bio.h | 2 +- > > 2 files changed, 8 insertions(+), 6 deletions(-) > > > > diff --git a/block/bio.c b/block/bio.c > > index 9bc7d28ae997..7f3920b6baca 100644 > > --- a/block/bio.c > > +++ b/block/bio.c > > @@ -845,7 +845,7 @@ static void bio_get_pages(struct bio *bio) > > get_page(bvec->bv_page); > > } > > > > -void bio_release_pages(struct bio *bio) > > +void bio_release_pages(struct bio *bio, bool mark_dirty) > > { > > struct bvec_iter_all iter_all; > > struct bio_vec *bvec; > > @@ -853,8 +853,11 @@ void bio_release_pages(struct bio *bio) > > if (bio_flagged(bio, BIO_NO_PAGE_REF)) > > return; > > > > - bio_for_each_segment_all(bvec, bio, iter_all) > > + bio_for_each_segment_all(bvec, bio, iter_all) { > > + if (mark_dirty && !PageCompound(bvec->bv_page)) > > + set_page_dirty_lock(bvec->bv_page); > > Christoph, > > Could you please explain a bit why we should not make it dirty in case > of compound page? > > Maybe because of [PATCH 7/9] block_dev: use bio_release_pages in bio_unmap_user :- > > @@ -259,13 +258,7 @@ __blkdev_direct_IO_simple(struct kiocb *iocb, struct iov_iter *iter, > } > __set_current_state(TASK_RUNNING); > - bio_for_each_segment_all(bvec, &bio, iter_all) { > - if (should_dirty && !PageCompound(bvec->bv_page)) > - set_page_dirty_lock(bvec->bv_page); > - if (!bio_flagged(&bio, BIO_NO_PAGE_REF)) > - put_page(bvec->bv_page); > - } > - > + bio_release_pages(&bio, should_dirty); > > I'll let Christoph confirm that. Chaitanya, Thanks for sharing this. Actually I have seen that commit log but didn't catch up why the compound pages are so. Now I understood what Christoph has mentioned. Thanks for your help, again.