Re: [RFC PATCH v1 11/18] xfs: add async buffered write support

["Darrick J. Wong" <djwong@kernel.org>]

On Tue, May 10, 2022 at 09:24:25AM +1000, Dave Chinner wrote:
> On Mon, May 09, 2022 at 12:32:59PM -0700, Stefan Roesch wrote:
> > On 5/6/22 2:29 AM, Dave Chinner wrote:
> > > On Mon, May 02, 2022 at 02:21:17PM -0700, Stefan Roesch wrote:
> > >> On 4/28/22 2:54 PM, Dave Chinner wrote:
> > >>> On Thu, Apr 28, 2022 at 12:58:59PM -0700, Stefan Roesch wrote:
> > >> - replace the pointer to iocb with pointer to xfs_inode in the function xfs_ilock_iocb()
> > >>   and also pass in the flags value as a parameter.
> > >> or
> > >> - create function xfs_ilock_inode(), which xfs_ilock_iocb() calls. The existing
> > >>   calls will not need to change, only the xfs_ilock in xfs_file_buffered_write()
> > >>   will use xfs_ilock_inode().
> > > 
> > > You're making this way more complex than it needs to be. As I said:
> > > 
> > >>> Regardless, if this is a problem, then just pass the XFS inode to
> > >>> xfs_ilock_iocb() and this is a moot point.
> > > 
> > 
> > The function xfs_ilock_iocb() is expecting a pointer to the data structure kiocb, not
> > a pointer to xfs_inode. I don't see how that's possible without changing the signature
> > of xfs_ilock_iocb().
> 
> For the *third time*: pass the xfs_inode to xfs_ilock_iocb() and
> update all the callers to do the same thing.

I still don't understand why /any/ of this is necessary.  When does
iocb->ki_filp->f_inode != iocb->ki_filp->f_mapping->host? 

--D

> -Dave.
> -- 
> Dave Chinner
> david@fromorbit.com