[Top] [All Lists]

Re: [PATCH 5/8] xfs: add DAX file operations support

To: Boaz Harrosh <boaz@xxxxxxxxxxxxx>
Subject: Re: [PATCH 5/8] xfs: add DAX file operations support
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Wed, 25 Mar 2015 08:17:40 +1100
Cc: xfs@xxxxxxxxxxx, linux-fsdevel@xxxxxxxxxxxxxxx, willy@xxxxxxxxxxxxxxx, jack@xxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <551153C4.6000808@xxxxxxxxxxxxx>
References: <1427194266-2885-1-git-send-email-david@xxxxxxxxxxxxx> <1427194266-2885-6-git-send-email-david@xxxxxxxxxxxxx> <551153C4.6000808@xxxxxxxxxxxxx>
User-agent: Mutt/1.5.21 (2010-09-15)
On Tue, Mar 24, 2015 at 02:08:36PM +0200, Boaz Harrosh wrote:
> On 03/24/2015 12:51 PM, Dave Chinner wrote:
> > From: Dave Chinner <dchinner@xxxxxxxxxx>
> > 
> > Add the initial support for DAX file operations to XFS. This
> > includes the necessary block allocation and mmap page fault hooks
> > for DAX to function.
> > 
> > Note: we specifically have to disable splice_read/write from
> It looks from below code that splice_read is perfectly supportable
> through the call to default_file_splice_read() so you might
> want to change the comment here.

Ah, I forgot to update the description.

> Regarding splice_write:
> It looks to me like you left the vector at iter_file_splice_write().
> If I understand correctly I think you need to call default_file_splice_write()
> that uses memcpy, just as it was at the previous patches when you left
> the vector NULL. So I think you need the same switch for write
> as you do below for read.

iter_file_splice_write() uses vfs_iter_write, which goes through
->write_iter, which we punt to the direct IO path for DAX.


Dave Chinner

<Prev in Thread] Current Thread [Next in Thread>