Date:   Thu, 2 Dec 2021 12:57:56 -0800
From:   Eric Biggers <ebiggers@kernel.org>
To:     Jaegeuk Kim <jaegeuk@kernel.org>
Cc:     linux-kernel@vger.kernel.org,
        linux-f2fs-devel@lists.sourceforge.net
Subject: Re: [f2fs-dev] [PATCH 5/6] f2fs: implement iomap operations
Message-ID: <YakzVAkHAVMueJwe@sol.localdomain>
References: <20211116214510.2934905-1-jaegeuk@kernel.org>
 <20211116214510.2934905-5-jaegeuk@kernel.org>
 <b36e2d44-4834-3931-6a32-4fa52d1d7785@kernel.org>
 <YahIf3UlhuxJT1O4@sol.localdomain>
 <1ba0f002-42c7-b085-0c54-6071664ba79b@kernel.org>
 <YakPrVPz5+qQm3kQ@sol.localdomain>
 <YakXMV0UvBV7TVwe@google.com>
 <YakX38onzsW5Ef8B@google.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <YakX38onzsW5Ef8B@google.com>
Precedence: bulk

On Thu, Dec 02, 2021 at 11:00:47AM -0800, Jaegeuk Kim wrote:
> On 12/02, Jaegeuk Kim wrote:
> > On 12/02, Eric Biggers wrote:
> > > On Thu, Dec 02, 2021 at 10:04:11PM +0800, Chao Yu wrote:
> > > > On 2021/12/2 12:15, Eric Biggers wrote:
> > > > > On Thu, Dec 02, 2021 at 11:10:41AM +0800, Chao Yu wrote:
> > > > > > Why not relocating this check before f2fs_map_blocks()?
> > > > 
> > > > Wait, it supports DIO in multi-device image after commit 	71f2c8206202
> > > > ("f2fs: multidevice: support direct IO"), how about
> > > > checking with f2fs_allow_multi_device_dio()?
> > > > 
> > > > Thanks,
> > > > 
> > > 
> > > Okay, that was not the case when I sent this patch originally.  We'll need to
> > > update this to support multiple devices.
> > 
> > Chao/Eric, does this make sense?
> > 
> > --- a/fs/f2fs/data.c
> > +++ b/fs/f2fs/data.c
> > @@ -4070,11 +4070,10 @@ static int f2fs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
> >                 }
> >                 if (WARN_ON_ONCE(!__is_valid_data_blkaddr(map.m_pblk)))
> >                         return -EINVAL;
> > -               iomap->addr = blks_to_bytes(inode, map.m_pblk);
> > 
> > -               if (WARN_ON_ONCE(f2fs_is_multi_device(F2FS_I_SB(inode))))
> > -                       return -EINVAL;
> > -               iomap->bdev = inode->i_sb->s_bdev;
> > +               iomap->bdev = map->m_multidev_dio ? map.m_bdev :
> 
> correction:			map.m_multidev_dio
> 

I guess so, but why doesn't f2fs_map_blocks() just always set m_bdev to the
correct block device?  What is the point of m_multidev_dio?

- Eric