Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220111161937.56272-1-pankaj.gupta.linux@gmail.com>
 <20220111161937.56272-3-pankaj.gupta.linux@gmail.com> <CAPcyv4gM99M8Waw9uEZefvpK0BsTkjGznLxUOMcMkGpk6SuHyA@mail.gmail.com>
 <CAM9Jb+iYXn+Diq-vou+_hXdxXLR9rEXm6GOsd2tZpAg9zXn1Fw@mail.gmail.com>
In-Reply-To: <CAM9Jb+iYXn+Diq-vou+_hXdxXLR9rEXm6GOsd2tZpAg9zXn1Fw@mail.gmail.com>
From:   Dan Williams <dan.j.williams@intel.com>
Date:   Wed, 16 Feb 2022 08:25:48 -0800
Message-ID: <CAPcyv4iPhtbhAfpjCtbt9RGFOXuGCj-q3Gm_y7zaNk44Z7uq9Q@mail.gmail.com>
Subject: Re: [RFC v3 2/2] pmem: enable pmem_submit_bio for asynchronous flush
To:     Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Cc:     Linux NVDIMM <nvdimm@lists.linux.dev>,
        virtualization@lists.linux-foundation.org,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        jmoyer <jmoyer@redhat.com>,
        Stefan Hajnoczi <stefanha@redhat.com>,
        David Hildenbrand <david@redhat.com>,
        "Michael S. Tsirkin" <mst@redhat.com>,
        Cornelia Huck <cohuck@redhat.com>,
        Vishal L Verma <vishal.l.verma@intel.com>,
        Dave Jiang <dave.jiang@intel.com>,
        "Weiny, Ira" <ira.weiny@intel.com>,
        Pankaj Gupta <pankaj.gupta@ionos.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, Feb 16, 2022 at 12:39 AM Pankaj Gupta
<pankaj.gupta.linux@gmail.com> wrote:
>
> > >
> > > Return from "pmem_submit_bio" when asynchronous flush is
> > > still in progress in other context.
> > >
> > > Signed-off-by: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
> > > ---
> > >  drivers/nvdimm/pmem.c        | 15 ++++++++++++---
> > >  drivers/nvdimm/region_devs.c |  4 +++-
> > >  2 files changed, 15 insertions(+), 4 deletions(-)
> > >
> > > diff --git a/drivers/nvdimm/pmem.c b/drivers/nvdimm/pmem.c
> > > index fe7ece1534e1..f20e30277a68 100644
> > > --- a/drivers/nvdimm/pmem.c
> > > +++ b/drivers/nvdimm/pmem.c
> > > @@ -201,8 +201,12 @@ static void pmem_submit_bio(struct bio *bio)
> > >         struct pmem_device *pmem = bio->bi_bdev->bd_disk->private_data;
> > >         struct nd_region *nd_region = to_region(pmem);
> > >
> > > -       if (bio->bi_opf & REQ_PREFLUSH)
> > > +       if (bio->bi_opf & REQ_PREFLUSH) {
> > >                 ret = nvdimm_flush(nd_region, bio);
> > > +               /* asynchronous flush completes in other context */
> >
> > I think a negative error code is a confusing way to capture the case
> > of "bio successfully coalesced to previously pending flush request.
> > Perhaps reserve negative codes for failure, 0 for synchronously
> > completed, and > 0 for coalesced flush request.
>
> Yes. I implemented this way previously, will revert it to. Thanks!
>
> >
> > > +               if (ret == -EINPROGRESS)
> > > +                       return;
> > > +       }
> > >
> > >         do_acct = blk_queue_io_stat(bio->bi_bdev->bd_disk->queue);
> > >         if (do_acct)
> > > @@ -222,13 +226,18 @@ static void pmem_submit_bio(struct bio *bio)
> > >         if (do_acct)
> > >                 bio_end_io_acct(bio, start);
> > >
> > > -       if (bio->bi_opf & REQ_FUA)
> > > +       if (bio->bi_opf & REQ_FUA) {
> > >                 ret = nvdimm_flush(nd_region, bio);
> > > +               /* asynchronous flush completes in other context */
> > > +               if (ret == -EINPROGRESS)
> > > +                       return;
> > > +       }
> > >
> > >         if (ret)
> > >                 bio->bi_status = errno_to_blk_status(ret);
> > >
> > > -       bio_endio(bio);
> > > +       if (bio)
> > > +               bio_endio(bio);
> > >  }
> > >
> > >  static int pmem_rw_page(struct block_device *bdev, sector_t sector,
> > > diff --git a/drivers/nvdimm/region_devs.c b/drivers/nvdimm/region_devs.c
> > > index 9ccf3d608799..8512d2eaed4e 100644
> > > --- a/drivers/nvdimm/region_devs.c
> > > +++ b/drivers/nvdimm/region_devs.c
> > > @@ -1190,7 +1190,9 @@ int nvdimm_flush(struct nd_region *nd_region, struct bio *bio)
> > >         if (!nd_region->flush)
> > >                 rc = generic_nvdimm_flush(nd_region);
> > >         else {
> > > -               if (nd_region->flush(nd_region, bio))
> > > +               rc = nd_region->flush(nd_region, bio);
> > > +               /* ongoing flush in other context */
> > > +               if (rc && rc != -EINPROGRESS)
> > >                         rc = -EIO;
> >
> > Why change this to -EIO vs just let the error code through untranslated?
>
> The reason was to be generic error code instead of returning host side
> return codes to guest?

Ok, maybe a comment to indicate the need to avoid exposing these error
codes toa guest so someone does not ask the same question in the
future?