Re: [PATCH 2/5] btrfs: remove use of buffer_heads from superblock writeout

[Johannes Thumshirn <Johannes.Thumshirn@wdc.com>]

On 17/01/2020, 16:01, "linux-btrfs-owner@vger.kernel.org on behalf of Nikolay Borisov" <linux-btrfs-owner@vger.kernel.org on behalf of nborisov@suse.com> wrote: 
    
    On 17.01.20 г. 14:51 ч., Johannes Thumshirn wrote:
    > Similar to the superblock read path, change the write path to using BIOs
    > and pages instead of buffer_heads.
    >
    > This is based on a patch originally authored by Nikolay Borisov.
    >
    > Co-developed-by: Nikolay Borisov <nborisov@suse.com>
    > Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
    > ---
    >  fs/btrfs/disk-io.c | 107 ++++++++++++++++++++++++++-------------------
    >  1 file changed, 61 insertions(+), 46 deletions(-)
    >
    > diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
    > index 50c93ffe8d03..51e7b832c8fd 100644
    > --- a/fs/btrfs/disk-io.c
    > +++ b/fs/btrfs/disk-io.c
    > @@ -3353,25 +3353,33 @@ int __cold open_ctree(struct super_block *sb,
    >  }
    >  ALLOW_ERROR_INJECTION(open_ctree, ERRNO);
    >
    > -static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
    > +static void btrfs_end_super_write(struct bio *bio)
    >  {
    > -     if (uptodate) {
    > -             set_buffer_uptodate(bh);
    > -     } else {
    > -             struct btrfs_device *device = (struct btrfs_device *)
    > -                     bh->b_private;
    > -
    > -             btrfs_warn_rl_in_rcu(device->fs_info,
    > -                             "lost page write due to IO error on %s",
    > -                                       rcu_str_deref(device->name));
    > -             /* note, we don't set_buffer_write_io_error because we have
    > -              * our own ways of dealing with the IO errors
    > -              */
    > -             clear_buffer_uptodate(bh);
    > -             btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
    > +     struct btrfs_device *device = bio->bi_private;
    > +     struct bio_vec *bvec;
    > +     struct bvec_iter_all iter_all;
    > +     struct page *page;
    > +
    > +     bio_for_each_segment_all(bvec, bio, iter_all) {
    > +             page = bvec->bv_page;
    > +
    > +             if (blk_status_to_errno(bio->bi_status)) {
    > +                     btrfs_warn_rl_in_rcu(device->fs_info,
    > +                                          "lost page write due to IO error on %s",
    > +                                          rcu_str_deref(device->name));
    > +                     ClearPageUptodate(page);
    > +                     SetPageError(page);
    > +                     btrfs_dev_stat_inc_and_print(device,
    > +                                                  BTRFS_DEV_STAT_WRITE_ERRS);
    > +             } else {
    > +                     SetPageUptodate(page);
    > +             }
    > +
    > +             put_page(page);
    
    Isn't this extra put page? Perhahps it's not because that would be the
    reference from find_or_create_page in write_dev_supers. In any case I'd
    rather have it in that function.

No we can't do the put_page() in in write_dev_supers() as btrfs_end_super_write()
is the bio endio hook and it's called upon completion of the bio. This happens
asynchronous to write_dev_supers().