[3/3] xfs: introduce per-inode DAX enablement
diff mbox

Message ID 1451886892-15548-4-git-send-email-david@fromorbit.com
State New
Headers show

Commit Message

Dave Chinner Jan. 4, 2016, 5:54 a.m. UTC
From: Dave Chinner <dchinner@redhat.com>

Rather than just being able to turn DAX on and off via a mount
option, some applications may only want to enable DAX for certain
performance critical files in a filesystem.

This patch introduces a new inode flag to enable DAX in the v3 inode
di_flags2 field. It adds support for setting and clearing flags in
the di_flags2 field via the XFS_IOC_FSSETXATTR ioctl, and sets the
S_DAX inode flag appropriately when it is seen.

When this flag is set on a directory, it acts as an "inherit flag".
That is, inodes created in the directory will automatically inherit
the on-disk inode DAX flag, enabling administrators to set up
directory heirarchies that automatically use DAX. Setting this flag
on an empty root directory will make the entire filesystem use DAX
by default.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
 fs/xfs/libxfs/xfs_format.h |  9 +++++++++
 fs/xfs/xfs_inode.c         | 32 +++++++++++++++++++++++---------
 fs/xfs/xfs_ioctl.c         | 18 +++++++++++++++++-
 fs/xfs/xfs_iops.c          |  4 ++--
 include/uapi/linux/fs.h    |  1 +
 5 files changed, 52 insertions(+), 12 deletions(-)

Comments

Theodore Y. Ts'o Jan. 16, 2016, 9:22 p.m. UTC | #1
On Mon, Jan 04, 2016 at 04:54:52PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
> 
> Rather than just being able to turn DAX on and off via a mount
> option, some applications may only want to enable DAX for certain
> performance critical files in a filesystem.
> 
> This patch introduces a new inode flag to enable DAX in the v3 inode
> di_flags2 field. It adds support for setting and clearing flags in
> the di_flags2 field via the XFS_IOC_FSSETXATTR ioctl, and sets the
> S_DAX inode flag appropriately when it is seen.
> 
> When this flag is set on a directory, it acts as an "inherit flag".
> That is, inodes created in the directory will automatically inherit
> the on-disk inode DAX flag, enabling administrators to set up
> directory heirarchies that automatically use DAX. Setting this flag
> on an empty root directory will make the entire filesystem use DAX
> by default.
> 
> Signed-off-by: Dave Chinner <dchinner@redhat.com>

Looks good to me.

Reviewed-by: Theodore Ts'o <tytso@mit.edu>

							- Ted
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Dan Williams Jan. 21, 2016, 4:37 p.m. UTC | #2
On Sun, Jan 3, 2016 at 9:54 PM, Dave Chinner <david@fromorbit.com> wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Rather than just being able to turn DAX on and off via a mount
> option, some applications may only want to enable DAX for certain
> performance critical files in a filesystem.
>
> This patch introduces a new inode flag to enable DAX in the v3 inode
> di_flags2 field. It adds support for setting and clearing flags in
> the di_flags2 field via the XFS_IOC_FSSETXATTR ioctl, and sets the
> S_DAX inode flag appropriately when it is seen.
>
> When this flag is set on a directory, it acts as an "inherit flag".
> That is, inodes created in the directory will automatically inherit
> the on-disk inode DAX flag, enabling administrators to set up
> directory heirarchies that automatically use DAX. Setting this flag
> on an empty root directory will make the entire filesystem use DAX
> by default.

When switching from page-cache to DAX, don't we need to flush existing
page cache mappings and remap directly?  Or, is the thought that
userspace needs to comprehend the presence of mixed mappings after
changing S_DAX?
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Dave Chinner Jan. 21, 2016, 9:58 p.m. UTC | #3
On Thu, Jan 21, 2016 at 08:37:11AM -0800, Dan Williams wrote:
> On Sun, Jan 3, 2016 at 9:54 PM, Dave Chinner <david@fromorbit.com> wrote:
> > From: Dave Chinner <dchinner@redhat.com>
> >
> > Rather than just being able to turn DAX on and off via a mount
> > option, some applications may only want to enable DAX for certain
> > performance critical files in a filesystem.
> >
> > This patch introduces a new inode flag to enable DAX in the v3 inode
> > di_flags2 field. It adds support for setting and clearing flags in
> > the di_flags2 field via the XFS_IOC_FSSETXATTR ioctl, and sets the
> > S_DAX inode flag appropriately when it is seen.
> >
> > When this flag is set on a directory, it acts as an "inherit flag".
> > That is, inodes created in the directory will automatically inherit
> > the on-disk inode DAX flag, enabling administrators to set up
> > directory heirarchies that automatically use DAX. Setting this flag
> > on an empty root directory will make the entire filesystem use DAX
> > by default.
> 
> When switching from page-cache to DAX, don't we need to flush existing
> page cache mappings and remap directly?  Or, is the thought that
> userspace needs to comprehend the presence of mixed mappings after
> changing S_DAX?

The change should be transparent to userspace. In general, I don't
expect users to change the behaviour of files that are in active use
(why would you do that?). This patch is really just introducing the
flag, the userspace API and making it propagate correctly via the
on-disk format. We'll fix up whatever problems with switching it
on/off dynamically as we go, like we do with most experimental
features once the on-disk behaviour is sorted out.

i.e. I've already got a couple of fixes we need to add to this - the
DAX flag is only valid on CRC enabled filesystems, so we need to
check that in the ioctl (general problem with using di_flags2 field,
not DAX flag specific issue). Adding a code to sync and unmap when
changing the flag is probably also necessary in the ioctl - I don't
have code to do that yet, but I have been thinking about it...

Cheers,

Dave.
Dan Williams Jan. 21, 2016, 10:53 p.m. UTC | #4
On Thu, Jan 21, 2016 at 1:58 PM, Dave Chinner <david@fromorbit.com> wrote:
> On Thu, Jan 21, 2016 at 08:37:11AM -0800, Dan Williams wrote:
>> On Sun, Jan 3, 2016 at 9:54 PM, Dave Chinner <david@fromorbit.com> wrote:
>> > From: Dave Chinner <dchinner@redhat.com>
>> >
>> > Rather than just being able to turn DAX on and off via a mount
>> > option, some applications may only want to enable DAX for certain
>> > performance critical files in a filesystem.
>> >
>> > This patch introduces a new inode flag to enable DAX in the v3 inode
>> > di_flags2 field. It adds support for setting and clearing flags in
>> > the di_flags2 field via the XFS_IOC_FSSETXATTR ioctl, and sets the
>> > S_DAX inode flag appropriately when it is seen.
>> >
>> > When this flag is set on a directory, it acts as an "inherit flag".
>> > That is, inodes created in the directory will automatically inherit
>> > the on-disk inode DAX flag, enabling administrators to set up
>> > directory heirarchies that automatically use DAX. Setting this flag
>> > on an empty root directory will make the entire filesystem use DAX
>> > by default.
>>
>> When switching from page-cache to DAX, don't we need to flush existing
>> page cache mappings and remap directly?  Or, is the thought that
>> userspace needs to comprehend the presence of mixed mappings after
>> changing S_DAX?
>
> The change should be transparent to userspace. In general, I don't
> expect users to change the behaviour of files that are in active use
> (why would you do that?).

If by accident someone tries to dynamically change S_DAX while
existing mappings are established I think the kernel should just
return EBUSY.  I was not proposing we support it as a first-class
operation.

> This patch is really just introducing the
> flag, the userspace API and making it propagate correctly via the
> on-disk format. We'll fix up whatever problems with switching it
> on/off dynamically as we go, like we do with most experimental
> features once the on-disk behaviour is sorted out.

Ok.

> i.e. I've already got a couple of fixes we need to add to this - the
> DAX flag is only valid on CRC enabled filesystems,

I assume for torn-write protection?  The CRC limitation makes sense,
but we theoretically could get the same effect by using a separate
logdev that does not tear writes, right?

> so we need to
> check that in the ioctl (general problem with using di_flags2 field,
> not DAX flag specific issue). Adding a code to sync and unmap when
> changing the flag is probably also necessary in the ioctl - I don't
> have code to do that yet, but I have been thinking about it...

Matthew and I have also talked about a modification of mincore(2) to
interrogate the effective mapping mode.  It seems we'll need that or
something like it given the growing list of caveats with setting up a
DAX mapping.
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Darrick J. Wong Jan. 22, 2016, 12:45 a.m. UTC | #5
On Mon, Jan 04, 2016 at 04:54:52PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
> 
> Rather than just being able to turn DAX on and off via a mount
> option, some applications may only want to enable DAX for certain
> performance critical files in a filesystem.
> 
> This patch introduces a new inode flag to enable DAX in the v3 inode
> di_flags2 field. It adds support for setting and clearing flags in
> the di_flags2 field via the XFS_IOC_FSSETXATTR ioctl, and sets the
> S_DAX inode flag appropriately when it is seen.
> 
> When this flag is set on a directory, it acts as an "inherit flag".
> That is, inodes created in the directory will automatically inherit
> the on-disk inode DAX flag, enabling administrators to set up
> directory heirarchies that automatically use DAX. Setting this flag
> on an empty root directory will make the entire filesystem use DAX
> by default.
> 
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
>  fs/xfs/libxfs/xfs_format.h |  9 +++++++++
>  fs/xfs/xfs_inode.c         | 32 +++++++++++++++++++++++---------
>  fs/xfs/xfs_ioctl.c         | 18 +++++++++++++++++-
>  fs/xfs/xfs_iops.c          |  4 ++--
>  include/uapi/linux/fs.h    |  1 +
>  5 files changed, 52 insertions(+), 12 deletions(-)
> 
> diff --git a/fs/xfs/libxfs/xfs_format.h b/fs/xfs/libxfs/xfs_format.h
> index f28eeab..b4ae7ce 100644
> --- a/fs/xfs/libxfs/xfs_format.h
> +++ b/fs/xfs/libxfs/xfs_format.h
> @@ -1024,6 +1024,15 @@ static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev)
>  	 XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM)
>  
>  /*
> + * Values for di_flags2 These start by being exposed to userspace in the upper

"Values for di_flags2.  These are exposed to userspace in the..."

> + * 16 bits of the XFS_XFLAG_s range.
> + */
> +#define XFS_DIFLAG2_DAX_BIT	0	/* use DAX for this inode */
> +#define XFS_DIFLAG2_DAX		(1 << XFS_DIFLAG2_DAX_BIT)
> +
> +#define XFS_DIFLAG2_ANY		(XFS_DIFLAG2_DAX)

Heh, I was gonna use bit 0 for reflink and bit 1 for cowextszhint.  Well, I'll
move them up by one.

"cowextszhint"... in which language is that a word? ;)

--D

> +
> +/*
>   * Inode number format:
>   * low inopblog bits - offset in block
>   * next agblklog bits - block number in ag
> diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
> index ca9ca5a..8929908 100644
> --- a/fs/xfs/xfs_inode.c
> +++ b/fs/xfs/xfs_inode.c
> @@ -610,7 +610,9 @@ __xfs_iflock(
>  
>  STATIC uint
>  _xfs_dic2xflags(
> -	__uint16_t		di_flags)
> +	__uint16_t		di_flags,
> +	uint64_t		di_flags2,
> +	bool			has_attr)
>  {
>  	uint			flags = 0;
>  
> @@ -645,25 +647,32 @@ _xfs_dic2xflags(
>  			flags |= FS_XFLAG_FILESTREAM;
>  	}
>  
> +	if (di_flags2 & XFS_DIFLAG2_ANY) {
> +		if (di_flags2 & XFS_DIFLAG2_DAX)
> +			flags |= FS_XFLAG_DAX;
> +	}
> +
> +	if (has_attr)
> +		flags |= FS_XFLAG_HASATTR;
> +
>  	return flags;
>  }
>  
>  uint
>  xfs_ip2xflags(
> -	xfs_inode_t		*ip)
> +	struct xfs_inode	*ip)
>  {
> -	xfs_icdinode_t		*dic = &ip->i_d;
> +	struct xfs_icdinode	*dic = &ip->i_d;
>  
> -	return _xfs_dic2xflags(dic->di_flags) |
> -				(XFS_IFORK_Q(ip) ? FS_XFLAG_HASATTR : 0);
> +	return _xfs_dic2xflags(dic->di_flags, dic->di_flags2, XFS_IFORK_Q(ip));
>  }
>  
>  uint
>  xfs_dic2xflags(
> -	xfs_dinode_t		*dip)
> +	struct xfs_dinode	*dip)
>  {
> -	return _xfs_dic2xflags(be16_to_cpu(dip->di_flags)) |
> -				(XFS_DFORK_Q(dip) ? FS_XFLAG_HASATTR : 0);
> +	return _xfs_dic2xflags(be16_to_cpu(dip->di_flags),
> +				be64_to_cpu(dip->di_flags2), XFS_DFORK_Q(dip));
>  }
>  
>  /*
> @@ -862,7 +871,8 @@ xfs_ialloc(
>  	case S_IFREG:
>  	case S_IFDIR:
>  		if (pip && (pip->i_d.di_flags & XFS_DIFLAG_ANY)) {
> -			uint	di_flags = 0;
> +			uint64_t	di_flags2 = 0;
> +			uint		di_flags = 0;
>  
>  			if (S_ISDIR(mode)) {
>  				if (pip->i_d.di_flags & XFS_DIFLAG_RTINHERIT)
> @@ -898,7 +908,11 @@ xfs_ialloc(
>  				di_flags |= XFS_DIFLAG_NODEFRAG;
>  			if (pip->i_d.di_flags & XFS_DIFLAG_FILESTREAM)
>  				di_flags |= XFS_DIFLAG_FILESTREAM;
> +			if (pip->i_d.di_flags2 & XFS_DIFLAG2_DAX)
> +				di_flags2 |= XFS_DIFLAG2_DAX;
> +
>  			ip->i_d.di_flags |= di_flags;
> +			ip->i_d.di_flags2 |= di_flags2;
>  		}
>  		/* FALLTHROUGH */
>  	case S_IFLNK:
> diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
> index 94b35eb3..478d04e 100644
> --- a/fs/xfs/xfs_ioctl.c
> +++ b/fs/xfs/xfs_ioctl.c
> @@ -945,6 +945,7 @@ xfs_set_diflags(
>  	unsigned int		xflags)
>  {
>  	unsigned int		di_flags;
> +	uint64_t		di_flags2;
>  
>  	/* can't set PREALLOC this way, just preserve it */
>  	di_flags = (ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
> @@ -977,8 +978,18 @@ xfs_set_diflags(
>  		if (xflags & FS_XFLAG_EXTSIZE)
>  			di_flags |= XFS_DIFLAG_EXTSIZE;
>  	}
> -
>  	ip->i_d.di_flags = di_flags;
> +
> +	/* diflags2 only valid for v3 inodes. */
> +	if (ip->i_d.di_version < 3)
> +		return;
> +
> +	di_flags2 = 0;
> +	if (xflags & FS_XFLAG_DAX)
> +		di_flags2 |= XFS_DIFLAG2_DAX;
> +
> +	ip->i_d.di_flags2 = di_flags2;
> +
>  }
>  
>  STATIC void
> @@ -1004,6 +1015,11 @@ xfs_diflags_to_linux(
>  		inode->i_flags |= S_NOATIME;
>  	else
>  		inode->i_flags &= ~S_NOATIME;
> +	if (xflags & FS_XFLAG_DAX)
> +		inode->i_flags |= S_DAX;
> +	else
> +		inode->i_flags &= ~S_DAX;
> +
>  }
>  
>  static int
> diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
> index 245268a..a1b8af1 100644
> --- a/fs/xfs/xfs_iops.c
> +++ b/fs/xfs/xfs_iops.c
> @@ -1201,8 +1201,8 @@ xfs_diflags_to_iflags(
>  		inode->i_flags |= S_SYNC;
>  	if (flags & XFS_DIFLAG_NOATIME)
>  		inode->i_flags |= S_NOATIME;
> -	/* XXX: Also needs an on-disk per inode flag! */
> -	if (ip->i_mount->m_flags & XFS_MOUNT_DAX)
> +	if (ip->i_mount->m_flags & XFS_MOUNT_DAX ||
> +	    ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)
>  		inode->i_flags |= S_DAX;
>  }
>  
> diff --git a/include/uapi/linux/fs.h b/include/uapi/linux/fs.h
> index df175dd..4cad4c8 100644
> --- a/include/uapi/linux/fs.h
> +++ b/include/uapi/linux/fs.h
> @@ -138,6 +138,7 @@ struct fsxattr {
>  #define FS_XFLAG_EXTSZINHERIT	0x00001000	/* inherit inode extent size */
>  #define FS_XFLAG_NODEFRAG	0x00002000	/* do not defragment */
>  #define FS_XFLAG_FILESTREAM	0x00004000	/* use filestream allocator */
> +#define FS_XFLAG_DAX		0x00008000	/* use DAX for IO */
>  #define FS_XFLAG_HASATTR	0x80000000	/* no DIFLAG for this	*/
>  
>  /* the read-only stuff doesn't really belong here, but any other place is
> -- 
> 2.5.0
> 
> _______________________________________________
> xfs mailing list
> xfs@oss.sgi.com
> http://oss.sgi.com/mailman/listinfo/xfs
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Dave Chinner Jan. 29, 2016, 10:20 p.m. UTC | #6
On Thu, Jan 21, 2016 at 02:53:06PM -0800, Dan Williams wrote:
> On Thu, Jan 21, 2016 at 1:58 PM, Dave Chinner <david@fromorbit.com> wrote:
> > On Thu, Jan 21, 2016 at 08:37:11AM -0800, Dan Williams wrote:
> >> On Sun, Jan 3, 2016 at 9:54 PM, Dave Chinner <david@fromorbit.com> wrote:
> > i.e. I've already got a couple of fixes we need to add to this - the
> > DAX flag is only valid on CRC enabled filesystems,
> 
> I assume for torn-write protection?  The CRC limitation makes sense,
> but we theoretically could get the same effect by using a separate
> logdev that does not tear writes, right?

No, it requires the CRC enabled format because there isn't any
flag space left in the v2 inode and so we need v3 inodes to store
the flag.

Cheers,

Dave.

Patch
diff mbox

diff --git a/fs/xfs/libxfs/xfs_format.h b/fs/xfs/libxfs/xfs_format.h
index f28eeab..b4ae7ce 100644
--- a/fs/xfs/libxfs/xfs_format.h
+++ b/fs/xfs/libxfs/xfs_format.h
@@ -1024,6 +1024,15 @@  static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev)
 	 XFS_DIFLAG_EXTSZINHERIT | XFS_DIFLAG_NODEFRAG | XFS_DIFLAG_FILESTREAM)
 
 /*
+ * Values for di_flags2 These start by being exposed to userspace in the upper
+ * 16 bits of the XFS_XFLAG_s range.
+ */
+#define XFS_DIFLAG2_DAX_BIT	0	/* use DAX for this inode */
+#define XFS_DIFLAG2_DAX		(1 << XFS_DIFLAG2_DAX_BIT)
+
+#define XFS_DIFLAG2_ANY		(XFS_DIFLAG2_DAX)
+
+/*
  * Inode number format:
  * low inopblog bits - offset in block
  * next agblklog bits - block number in ag
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index ca9ca5a..8929908 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -610,7 +610,9 @@  __xfs_iflock(
 
 STATIC uint
 _xfs_dic2xflags(
-	__uint16_t		di_flags)
+	__uint16_t		di_flags,
+	uint64_t		di_flags2,
+	bool			has_attr)
 {
 	uint			flags = 0;
 
@@ -645,25 +647,32 @@  _xfs_dic2xflags(
 			flags |= FS_XFLAG_FILESTREAM;
 	}
 
+	if (di_flags2 & XFS_DIFLAG2_ANY) {
+		if (di_flags2 & XFS_DIFLAG2_DAX)
+			flags |= FS_XFLAG_DAX;
+	}
+
+	if (has_attr)
+		flags |= FS_XFLAG_HASATTR;
+
 	return flags;
 }
 
 uint
 xfs_ip2xflags(
-	xfs_inode_t		*ip)
+	struct xfs_inode	*ip)
 {
-	xfs_icdinode_t		*dic = &ip->i_d;
+	struct xfs_icdinode	*dic = &ip->i_d;
 
-	return _xfs_dic2xflags(dic->di_flags) |
-				(XFS_IFORK_Q(ip) ? FS_XFLAG_HASATTR : 0);
+	return _xfs_dic2xflags(dic->di_flags, dic->di_flags2, XFS_IFORK_Q(ip));
 }
 
 uint
 xfs_dic2xflags(
-	xfs_dinode_t		*dip)
+	struct xfs_dinode	*dip)
 {
-	return _xfs_dic2xflags(be16_to_cpu(dip->di_flags)) |
-				(XFS_DFORK_Q(dip) ? FS_XFLAG_HASATTR : 0);
+	return _xfs_dic2xflags(be16_to_cpu(dip->di_flags),
+				be64_to_cpu(dip->di_flags2), XFS_DFORK_Q(dip));
 }
 
 /*
@@ -862,7 +871,8 @@  xfs_ialloc(
 	case S_IFREG:
 	case S_IFDIR:
 		if (pip && (pip->i_d.di_flags & XFS_DIFLAG_ANY)) {
-			uint	di_flags = 0;
+			uint64_t	di_flags2 = 0;
+			uint		di_flags = 0;
 
 			if (S_ISDIR(mode)) {
 				if (pip->i_d.di_flags & XFS_DIFLAG_RTINHERIT)
@@ -898,7 +908,11 @@  xfs_ialloc(
 				di_flags |= XFS_DIFLAG_NODEFRAG;
 			if (pip->i_d.di_flags & XFS_DIFLAG_FILESTREAM)
 				di_flags |= XFS_DIFLAG_FILESTREAM;
+			if (pip->i_d.di_flags2 & XFS_DIFLAG2_DAX)
+				di_flags2 |= XFS_DIFLAG2_DAX;
+
 			ip->i_d.di_flags |= di_flags;
+			ip->i_d.di_flags2 |= di_flags2;
 		}
 		/* FALLTHROUGH */
 	case S_IFLNK:
diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
index 94b35eb3..478d04e 100644
--- a/fs/xfs/xfs_ioctl.c
+++ b/fs/xfs/xfs_ioctl.c
@@ -945,6 +945,7 @@  xfs_set_diflags(
 	unsigned int		xflags)
 {
 	unsigned int		di_flags;
+	uint64_t		di_flags2;
 
 	/* can't set PREALLOC this way, just preserve it */
 	di_flags = (ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
@@ -977,8 +978,18 @@  xfs_set_diflags(
 		if (xflags & FS_XFLAG_EXTSIZE)
 			di_flags |= XFS_DIFLAG_EXTSIZE;
 	}
-
 	ip->i_d.di_flags = di_flags;
+
+	/* diflags2 only valid for v3 inodes. */
+	if (ip->i_d.di_version < 3)
+		return;
+
+	di_flags2 = 0;
+	if (xflags & FS_XFLAG_DAX)
+		di_flags2 |= XFS_DIFLAG2_DAX;
+
+	ip->i_d.di_flags2 = di_flags2;
+
 }
 
 STATIC void
@@ -1004,6 +1015,11 @@  xfs_diflags_to_linux(
 		inode->i_flags |= S_NOATIME;
 	else
 		inode->i_flags &= ~S_NOATIME;
+	if (xflags & FS_XFLAG_DAX)
+		inode->i_flags |= S_DAX;
+	else
+		inode->i_flags &= ~S_DAX;
+
 }
 
 static int
diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
index 245268a..a1b8af1 100644
--- a/fs/xfs/xfs_iops.c
+++ b/fs/xfs/xfs_iops.c
@@ -1201,8 +1201,8 @@  xfs_diflags_to_iflags(
 		inode->i_flags |= S_SYNC;
 	if (flags & XFS_DIFLAG_NOATIME)
 		inode->i_flags |= S_NOATIME;
-	/* XXX: Also needs an on-disk per inode flag! */
-	if (ip->i_mount->m_flags & XFS_MOUNT_DAX)
+	if (ip->i_mount->m_flags & XFS_MOUNT_DAX ||
+	    ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)
 		inode->i_flags |= S_DAX;
 }
 
diff --git a/include/uapi/linux/fs.h b/include/uapi/linux/fs.h
index df175dd..4cad4c8 100644
--- a/include/uapi/linux/fs.h
+++ b/include/uapi/linux/fs.h
@@ -138,6 +138,7 @@  struct fsxattr {
 #define FS_XFLAG_EXTSZINHERIT	0x00001000	/* inherit inode extent size */
 #define FS_XFLAG_NODEFRAG	0x00002000	/* do not defragment */
 #define FS_XFLAG_FILESTREAM	0x00004000	/* use filestream allocator */
+#define FS_XFLAG_DAX		0x00008000	/* use DAX for IO */
 #define FS_XFLAG_HASATTR	0x80000000	/* no DIFLAG for this	*/
 
 /* the read-only stuff doesn't really belong here, but any other place is