linux-xfs.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Dan Williams <dan.j.williams@intel.com>
To: Shiyang Ruan <ruansy.fnst@fujitsu.com>
Cc: "Darrick J. Wong" <djwong@kernel.org>,
	Christoph Hellwig <hch@lst.de>,
	linux-xfs <linux-xfs@vger.kernel.org>,
	david <david@fromorbit.com>,
	linux-fsdevel <linux-fsdevel@vger.kernel.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Linux NVDIMM <nvdimm@lists.linux.dev>,
	Goldwyn Rodrigues <rgoldwyn@suse.de>,
	Al Viro <viro@zeniv.linux.org.uk>,
	Matthew Wilcox <willy@infradead.org>
Subject: Re: [PATCH v7 7/8] fsdax: Introduce dax_iomap_ops for end of reflink
Date: Thu, 26 Aug 2021 22:04:16 -0700	[thread overview]
Message-ID: <CAPcyv4h801eipbvOpzSnw_GnUcuSxcm6eUfJdoHNW2ZmZgzW=Q@mail.gmail.com> (raw)
In-Reply-To: <32fa5333-b14e-2060-d659-d77f6c75ff16@fujitsu.com>

On Thu, Aug 26, 2021 at 8:30 PM Shiyang Ruan <ruansy.fnst@fujitsu.com> wrote:
>
>
>
> On 2021/8/20 23:18, Dan Williams wrote:
> > On Thu, Aug 19, 2021 at 11:13 PM ruansy.fnst <ruansy.fnst@fujitsu.com> wrote:
> >>
> >>
> >>
> >> On 2021/8/20 上午11:01, Dan Williams wrote:
> >>> On Sun, Aug 15, 2021 at 11:05 PM Shiyang Ruan <ruansy.fnst@fujitsu.com> wrote:
> >>>>
> >>>> After writing data, reflink requires end operations to remap those new
> >>>> allocated extents.  The current ->iomap_end() ignores the error code
> >>>> returned from ->actor(), so we introduce this dax_iomap_ops and change
> >>>> the dax_iomap_*() interfaces to do this job.
> >>>>
> >>>> - the dax_iomap_ops contains the original struct iomap_ops and fsdax
> >>>>       specific ->actor_end(), which is for the end operations of reflink
> >>>> - also introduce dax specific zero_range, truncate_page
> >>>> - create new dax_iomap_ops for ext2 and ext4
> >>>>
> >>>> Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
> >>>> ---
> >>>>    fs/dax.c               | 68 +++++++++++++++++++++++++++++++++++++-----
> >>>>    fs/ext2/ext2.h         |  3 ++
> >>>>    fs/ext2/file.c         |  6 ++--
> >>>>    fs/ext2/inode.c        | 11 +++++--
> >>>>    fs/ext4/ext4.h         |  3 ++
> >>>>    fs/ext4/file.c         |  6 ++--
> >>>>    fs/ext4/inode.c        | 13 ++++++--
> >>>>    fs/iomap/buffered-io.c |  3 +-
> >>>>    fs/xfs/xfs_bmap_util.c |  3 +-
> >>>>    fs/xfs/xfs_file.c      |  8 ++---
> >>>>    fs/xfs/xfs_iomap.c     | 36 +++++++++++++++++++++-
> >>>>    fs/xfs/xfs_iomap.h     | 33 ++++++++++++++++++++
> >>>>    fs/xfs/xfs_iops.c      |  7 ++---
> >>>>    fs/xfs/xfs_reflink.c   |  3 +-
> >>>>    include/linux/dax.h    | 21 ++++++++++---
> >>>>    include/linux/iomap.h  |  1 +
> >>>>    16 files changed, 189 insertions(+), 36 deletions(-)
> >>>>
> >>>> diff --git a/fs/dax.c b/fs/dax.c
> >>>> index 74dd918cff1f..0e0536765a7e 100644
> >>>> --- a/fs/dax.c
> >>>> +++ b/fs/dax.c
> >>>> @@ -1348,11 +1348,30 @@ static loff_t dax_iomap_iter(const struct iomap_iter *iomi,
> >>>>           return done ? done : ret;
> >>>>    }
> >>>>
> >>>> +static inline int
> >>>> +__dax_iomap_iter(struct iomap_iter *iter, const struct dax_iomap_ops *ops)
> >>>> +{
> >>>> +       int ret;
> >>>> +
> >>>> +       /*
> >>>> +        * Call dax_iomap_ops->actor_end() before iomap_ops->iomap_end() in
> >>>> +        * each iteration.
> >>>> +        */
> >>>> +       if (iter->iomap.length && ops->actor_end) {
> >>>> +               ret = ops->actor_end(iter->inode, iter->pos, iter->len,
> >>>> +                                    iter->processed);
> >>>> +               if (ret < 0)
> >>>> +                       return ret;
> >>>> +       }
> >>>> +
> >>>> +       return iomap_iter(iter, &ops->iomap_ops);
> >>>
> >>> This reorganization looks needlessly noisy. Why not require the
> >>> iomap_end operation to perform the actor_end work. I.e. why can't
> >>> xfs_dax_write_iomap_actor_end() just be the passed in iomap_end? I am
> >>> not seeing where the ->iomap_end() result is ignored?
> >>>
> >>
> >> The V6 patch[1] was did in this way.
> >> [1]https://lore.kernel.org/linux-xfs/20210526005159.GF202144@locust/T/#m79a66a928da2d089e2458c1a97c0516dbfde2f7f
> >>
> >> But Darrick reminded me that ->iomap_end() will always take zero or
> >> positive 'written' because iomap_apply() handles this argument.
> >>
> >> ```
> >>          if (ops->iomap_end) {
> >>                  ret = ops->iomap_end(inode, pos, length,
> >>                                       written > 0 ? written : 0,
> >>                                       flags, &iomap);
> >>          }
> >> ```
> >>
> >> So, we cannot get actual return code from CoW in ->actor(), and as a
> >> result, we cannot handle the xfs end_cow correctly in ->iomap_end().
> >> That's where the result of CoW was ignored.
> >
> > Ah, thank you for the explanation.
> >
> > However, this still seems like too much code thrash just to get back
> > to the original value of iter->processed. I notice you are talking
> > about iomap_apply(), but that routine is now gone in Darrick's latest
> > iomap-for-next branch. Instead iomap_iter() does this:
> >
> >          if (iter->iomap.length && ops->iomap_end) {
> >                  ret = ops->iomap_end(iter->inode, iter->pos, iomap_length(iter),
> >                                  iter->processed > 0 ? iter->processed : 0,
>
> As you can see, here is the same logic as the old iomap_apply(): the
> negative iter->processed won't be passed into ->iomap_end().
>
> >                                  iter->flags, &iter->iomap);
> >                  if (ret < 0 && !iter->processed)
> >                          return ret;
> >          }
> >
> >
> > I notice that the @iomap argument to ->iomap_end() is reliably coming
> > from @iter. So you could do the following in your iomap_end()
> > callback:
> >
> >          struct iomap_iter *iter = container_of(iomap, typeof(*iter), iomap);
> >          struct xfs_inode *ip = XFS_I(inode);
> >          ssize_t written = iter->processed;
>
> The written will be 0 or positive.  The original error code is ingnored.

Correct, but you can use container_of() to get back to the iter and
consider the raw untranslated value of iter->processed. As Christoph
mentioned this needs a comment explaining the layering violation, but
that's a cleaner change than the dax_iomap_ops approach.

  reply	other threads:[~2021-08-27  5:04 UTC|newest]

Thread overview: 28+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-08-16  6:03 [PATCH v7 0/8] fsdax,xfs: Add reflink&dedupe support for fsdax Shiyang Ruan
2021-08-16  6:03 ` [PATCH v7 1/8] fsdax: Output address in dax_iomap_pfn() and rename it Shiyang Ruan
2021-08-18 21:01   ` Dan Williams
2021-08-16  6:03 ` [PATCH v7 2/8] fsdax: Introduce dax_iomap_cow_copy() Shiyang Ruan
2021-08-19 22:35   ` Dan Williams
2021-08-16  6:03 ` [PATCH v7 3/8] fsdax: Replace mmap entry in case of CoW Shiyang Ruan
2021-08-19 22:54   ` Dan Williams
2021-08-23 12:57     ` Christoph Hellwig
2021-08-27  3:22       ` Shiyang Ruan
2021-08-27  5:00         ` Dan Williams
2021-08-27  5:26           ` Shiyang Ruan
2021-08-16  6:03 ` [PATCH v7 4/8] fsdax: Add dax_iomap_cow_copy() for dax_iomap_zero Shiyang Ruan
2021-08-20  2:39   ` Dan Williams
2021-08-27  3:23     ` Shiyang Ruan
2021-08-16  6:03 ` [PATCH v7 5/8] iomap: Introduce iomap_iter2 for two files Shiyang Ruan
2021-08-23 12:58   ` Christoph Hellwig
2021-08-16  6:03 ` [PATCH v7 6/8] fsdax: Dedup file range to use a compare function Shiyang Ruan
2021-08-23 13:16   ` Christoph Hellwig
2021-08-16  6:03 ` [PATCH v7 7/8] fsdax: Introduce dax_iomap_ops for end of reflink Shiyang Ruan
2021-08-20  3:01   ` Dan Williams
     [not found]     ` <c7e68dc8-5a43-f727-c262-58dcf244c711@fujitsu.com>
2021-08-20 15:18       ` Dan Williams
2021-08-23 13:02         ` Christoph Hellwig
2021-08-27  3:29         ` Shiyang Ruan
2021-08-27  5:04           ` Dan Williams [this message]
2021-08-27  5:27             ` Shiyang Ruan
2021-08-16  6:03 ` [PATCH v7 8/8] fs/xfs: Add dax dedupe support Shiyang Ruan
2021-08-20  3:08   ` Dan Williams
2021-08-27  3:36     ` Shiyang Ruan

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to='CAPcyv4h801eipbvOpzSnw_GnUcuSxcm6eUfJdoHNW2ZmZgzW=Q@mail.gmail.com' \
    --to=dan.j.williams@intel.com \
    --cc=david@fromorbit.com \
    --cc=djwong@kernel.org \
    --cc=hch@lst.de \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-xfs@vger.kernel.org \
    --cc=nvdimm@lists.linux.dev \
    --cc=rgoldwyn@suse.de \
    --cc=ruansy.fnst@fujitsu.com \
    --cc=viro@zeniv.linux.org.uk \
    --cc=willy@infradead.org \
    --subject='Re: [PATCH v7 7/8] fsdax: Introduce dax_iomap_ops for end of reflink' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).