Re: [PATCH 5.10 01/25] md: revert io stats accounting

[Greg Kroah-Hartman]

On Wed, Jan 26, 2022 at 11:09:46AM +0100, Jack Wang wrote:
> Hi,
> 
> Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> 于2022年1月14日周五 19:57写道：
> >
> > From: Guoqing Jiang <jgq516@xxxxxxxxx>
> >
> > commit ad3fc798800fb7ca04c1dfc439dba946818048d8 upstream.
> >
> > The commit 41d2d848e5c0 ("md: improve io stats accounting") could cause
> > double fault problem per the report [1], and also it is not correct to
> > change ->bi_end_io if md don't own it, so let's revert it.
> >
> > And io stats accounting will be replemented in later commits.
> >
> > [1]. https://lore.kernel.org/linux-raid/3bf04253-3fad-434a-63a7-20214e38cf26@xxxxxxxxx/T/#t
> >
> > Fixes: 41d2d848e5c0 ("md: improve io stats accounting")
> > Signed-off-by: Guoqing Jiang <jiangguoqing@xxxxxxxxxx>
> > Signed-off-by: Song Liu <song@xxxxxxxxxx>
> > [GM: backport to 5.10-stable]
> > Signed-off-by: Guillaume Morin <guillaume@xxxxxxxxxxx>
> > Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> > ---
> >  drivers/md/md.c |   57 +++++++++++---------------------------------------------
> >  drivers/md/md.h |    1
> >  2 files changed, 12 insertions(+), 46 deletions(-)
> >
> > --- a/drivers/md/md.c
> > +++ b/drivers/md/md.c
> > @@ -459,34 +459,12 @@ check_suspended:
> >  }
> >  EXPORT_SYMBOL(md_handle_request);
> >
> > -struct md_io {
> > -       struct mddev *mddev;
> > -       bio_end_io_t *orig_bi_end_io;
> > -       void *orig_bi_private;
> > -       unsigned long start_time;
> > -       struct hd_struct *part;
> > -};
> > -
> > -static void md_end_io(struct bio *bio)
> > -{
> > -       struct md_io *md_io = bio->bi_private;
> > -       struct mddev *mddev = md_io->mddev;
> > -
> > -       part_end_io_acct(md_io->part, bio, md_io->start_time);
> > -
> > -       bio->bi_end_io = md_io->orig_bi_end_io;
> > -       bio->bi_private = md_io->orig_bi_private;
> > -
> > -       mempool_free(md_io, &mddev->md_io_pool);
> > -
> > -       if (bio->bi_end_io)
> > -               bio->bi_end_io(bio);
> > -}
> > -
> >  static blk_qc_t md_submit_bio(struct bio *bio)
> >  {
> >         const int rw = bio_data_dir(bio);
> > +       const int sgrp = op_stat_group(bio_op(bio));
> >         struct mddev *mddev = bio->bi_disk->private_data;
> > +       unsigned int sectors;
> >
> >         if (mddev == NULL || mddev->pers == NULL) {
> >                 bio_io_error(bio);
> > @@ -507,26 +485,21 @@ static blk_qc_t md_submit_bio(struct bio
> >                 return BLK_QC_T_NONE;
> >         }
> >
> > -       if (bio->bi_end_io != md_end_io) {
> > -               struct md_io *md_io;
> > -
> > -               md_io = mempool_alloc(&mddev->md_io_pool, GFP_NOIO);
> > -               md_io->mddev = mddev;
> > -               md_io->orig_bi_end_io = bio->bi_end_io;
> > -               md_io->orig_bi_private = bio->bi_private;
> > -
> > -               bio->bi_end_io = md_end_io;
> > -               bio->bi_private = md_io;
> > -
> > -               md_io->start_time = part_start_io_acct(mddev->gendisk,
> > -                                                      &md_io->part, bio);
> > -       }
> > -
> > +       /*
> > +        * save the sectors now since our bio can
> > +        * go away inside make_request
> > +        */
> > +       sectors = bio_sectors(bio);
> This code snip is not inside the original patch, and it's not in
> latest upstream too.
> >         /* bio could be mergeable after passing to underlayer */
> >         bio->bi_opf &= ~REQ_NOMERGE;
> >
> >         md_handle_request(mddev, bio);
> >
> > +       part_stat_lock();
> > +       part_stat_inc(&mddev->gendisk->part0, ios[sgrp]);
> > +       part_stat_add(&mddev->gendisk->part0, sectors[sgrp], sectors);
> > +       part_stat_unlock();
> > +
> same here, this code snip is not inside the original patch, and it's
> not in latest upstream too.

Is it a problem?

> I think would be good keep it as the upstream version.

Can you send a revert of this commit (it is in 5.10.92), and a backport
of the correct fix?

thanks,

greg k-h