[PATCH v3 0/4] fs, dax: lookup dax_device at mount time

[PATCH v3 0/4] fs, dax: lookup dax_device at mount time

[Dan Williams]

Changes since v2 [1]:
* Split fs_dax_get_by_bdev() to its own patch (Christoph)

* Push dax_device reference management into <fs>_{fill,put}_super()
  rather than the generic vfs. (Christoph)

[1]: https://lists.01.org/pipermail/linux-nvdimm/2017-August/012133.html

---

Christoph notes:

    I just noticed that we now do a fs_dax_get_by_host in every
    iomap_begin call for DAX. This function iterates a list, does a
    string compared and igrab. I really think we need to cache this in
    the superblock (possible even the fs superblock) similar to what we
    do for the block device.

This passes the libnvdimm unit tests.

---

Dan Williams (4):
      dax: introduce a fs_dax_get_by_bdev() helper
      xfs: perform dax_device lookup at mount
      ext2: perform dax_device lookup at mount
      ext4: perform dax_device lookup at mount


 drivers/dax/super.c |   10 ++++++++++
 fs/ext2/ext2.h      |    1 +
 fs/ext2/inode.c     |   11 +++--------
 fs/ext2/super.c     |    5 +++++
 fs/ext4/ext4.h      |    1 +
 fs/ext4/inode.c     |   11 +++--------
 fs/ext4/super.c     |    5 +++++
 fs/xfs/xfs_aops.c   |   13 +++++++++++++
 fs/xfs/xfs_aops.h   |    1 +
 fs/xfs/xfs_buf.c    |    4 +++-
 fs/xfs/xfs_buf.h    |    3 ++-
 fs/xfs/xfs_iomap.c  |   10 +---------
 fs/xfs/xfs_super.c  |   25 +++++++++++++++++++++----
 include/linux/dax.h |    6 ++++++
 14 files changed, 75 insertions(+), 31 deletions(-)

[PATCH v3 1/4] dax: introduce a fs_dax_get_by_bdev() helper

[Dan Williams]

Add a helper that can replace the following common pattern:

	if (blk_queue_dax(bdev->bd_queue))
		fs_dax_get_by_host(bdev->bd_disk->disk_name);

This will be used to move dax_device lookup from iomap-operation time to
fs-mount time.

Cc: Jan Kara <jack-AlSwsSmVLrQ@public.gmane.org>
Cc: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
Signed-off-by: Dan Williams <dan.j.williams-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/dax/super.c |   10 ++++++++++
 include/linux/dax.h |    6 ++++++
 2 files changed, 16 insertions(+)

diff --git a/drivers/dax/super.c b/drivers/dax/super.c
index 938eb4868f7f..b699aac268a6 100644
--- a/drivers/dax/super.c
+++ b/drivers/dax/super.c
@@ -46,6 +46,8 @@ void dax_read_unlock(int id)
 EXPORT_SYMBOL_GPL(dax_read_unlock);
 
 #ifdef CONFIG_BLOCK
+#include <linux/blkdev.h>
+
 int bdev_dax_pgoff(struct block_device *bdev, sector_t sector, size_t size,
 		pgoff_t *pgoff)
 {
@@ -59,6 +61,14 @@ int bdev_dax_pgoff(struct block_device *bdev, sector_t sector, size_t size,
 }
 EXPORT_SYMBOL(bdev_dax_pgoff);
 
+struct dax_device *fs_dax_get_by_bdev(struct block_device *bdev)
+{
+	if (!blk_queue_dax(bdev->bd_queue))
+		return NULL;
+	return fs_dax_get_by_host(bdev->bd_disk->disk_name);
+}
+EXPORT_SYMBOL_GPL(fs_dax_get_by_bdev);
+
 /**
  * __bdev_dax_supported() - Check if the device supports dax for filesystem
  * @sb: The superblock of the device
diff --git a/include/linux/dax.h b/include/linux/dax.h
index df97b7af7e2c..ac8afa18f707 100644
--- a/include/linux/dax.h
+++ b/include/linux/dax.h
@@ -57,6 +57,7 @@ static inline void fs_put_dax(struct dax_device *dax_dev)
 	put_dax(dax_dev);
 }
 
+struct dax_device *fs_dax_get_by_bdev(struct block_device *bdev);
 #else
 static inline int bdev_dax_supported(struct super_block *sb, int blocksize)
 {
@@ -71,6 +72,11 @@ static inline struct dax_device *fs_dax_get_by_host(const char *host)
 static inline void fs_put_dax(struct dax_device *dax_dev)
 {
 }
+
+static inline struct dax_device *fs_dax_get_by_bdev(struct block_device *bdev)
+{
+	return NULL;
+}
 #endif
 
 int dax_read_lock(void);

Re: [PATCH v3 1/4] dax: introduce a fs_dax_get_by_bdev() helper

[Darrick J. Wong]

On Wed, Aug 30, 2017 at 12:43:52PM -0700, Dan Williams wrote:
> Add a helper that can replace the following common pattern:
> 
> 	if (blk_queue_dax(bdev->bd_queue))
> 		fs_dax_get_by_host(bdev->bd_disk->disk_name);
> 
> This will be used to move dax_device lookup from iomap-operation time to
> fs-mount time.
> 
> Cc: Jan Kara <jack@suse.cz>
> Cc: Christoph Hellwig <hch@lst.de>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>

Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>

> ---
>  drivers/dax/super.c |   10 ++++++++++
>  include/linux/dax.h |    6 ++++++
>  2 files changed, 16 insertions(+)
> 
> diff --git a/drivers/dax/super.c b/drivers/dax/super.c
> index 938eb4868f7f..b699aac268a6 100644
> --- a/drivers/dax/super.c
> +++ b/drivers/dax/super.c
> @@ -46,6 +46,8 @@ void dax_read_unlock(int id)
>  EXPORT_SYMBOL_GPL(dax_read_unlock);
>  
>  #ifdef CONFIG_BLOCK
> +#include <linux/blkdev.h>
> +
>  int bdev_dax_pgoff(struct block_device *bdev, sector_t sector, size_t size,
>  		pgoff_t *pgoff)
>  {
> @@ -59,6 +61,14 @@ int bdev_dax_pgoff(struct block_device *bdev, sector_t sector, size_t size,
>  }
>  EXPORT_SYMBOL(bdev_dax_pgoff);
>  
> +struct dax_device *fs_dax_get_by_bdev(struct block_device *bdev)
> +{
> +	if (!blk_queue_dax(bdev->bd_queue))
> +		return NULL;
> +	return fs_dax_get_by_host(bdev->bd_disk->disk_name);
> +}
> +EXPORT_SYMBOL_GPL(fs_dax_get_by_bdev);
> +
>  /**
>   * __bdev_dax_supported() - Check if the device supports dax for filesystem
>   * @sb: The superblock of the device
> diff --git a/include/linux/dax.h b/include/linux/dax.h
> index df97b7af7e2c..ac8afa18f707 100644
> --- a/include/linux/dax.h
> +++ b/include/linux/dax.h
> @@ -57,6 +57,7 @@ static inline void fs_put_dax(struct dax_device *dax_dev)
>  	put_dax(dax_dev);
>  }
>  
> +struct dax_device *fs_dax_get_by_bdev(struct block_device *bdev);
>  #else
>  static inline int bdev_dax_supported(struct super_block *sb, int blocksize)
>  {
> @@ -71,6 +72,11 @@ static inline struct dax_device *fs_dax_get_by_host(const char *host)
>  static inline void fs_put_dax(struct dax_device *dax_dev)
>  {
>  }
> +
> +static inline struct dax_device *fs_dax_get_by_bdev(struct block_device *bdev)
> +{
> +	return NULL;
> +}
>  #endif
>  
>  int dax_read_lock(void);
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH v3 1/4] dax: introduce a fs_dax_get_by_bdev() helper

[Jan Kara]

On Wed 30-08-17 12:43:52, Dan Williams wrote:
> Add a helper that can replace the following common pattern:
> 
> 	if (blk_queue_dax(bdev->bd_queue))
> 		fs_dax_get_by_host(bdev->bd_disk->disk_name);
> 
> This will be used to move dax_device lookup from iomap-operation time to
> fs-mount time.
> 
> Cc: Jan Kara <jack-AlSwsSmVLrQ@public.gmane.org>
> Cc: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
> Signed-off-by: Dan Williams <dan.j.williams-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

Looks good to me. You can add:

Reviewed-by: Jan Kara <jack-AlSwsSmVLrQ@public.gmane.org>

								Honza
-- 
Jan Kara <jack-IBi9RG/b67k@public.gmane.org>
SUSE Labs, CR

Re: [PATCH v3 1/4] dax: introduce a fs_dax_get_by_bdev() helper

[Christoph Hellwig]

Looks fine,

Reviewed-by: Christoph Hellwig <hch@lst.de>

[PATCH v3 2/4] xfs: perform dax_device lookup at mount

[Dan Williams]

The ->iomap_begin() operation is a hot path, so cache the
fs_dax_get_by_host() result at mount time to avoid the incurring the
hash lookup overhead on a per-i/o basis.

Cc: "Darrick J. Wong" <darrick.wong-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
Reported-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
Signed-off-by: Dan Williams <dan.j.williams-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 fs/xfs/xfs_aops.c  |   13 +++++++++++++
 fs/xfs/xfs_aops.h  |    1 +
 fs/xfs/xfs_buf.c   |    4 +++-
 fs/xfs/xfs_buf.h   |    3 ++-
 fs/xfs/xfs_iomap.c |   10 +---------
 fs/xfs/xfs_super.c |   25 +++++++++++++++++++++----
 6 files changed, 41 insertions(+), 15 deletions(-)

diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c
index 6bf120bb1a17..78185f3b10b2 100644
--- a/fs/xfs/xfs_aops.c
+++ b/fs/xfs/xfs_aops.c
@@ -80,6 +80,19 @@ xfs_find_bdev_for_inode(
 		return mp->m_ddev_targp->bt_bdev;
 }
 
+struct dax_device *
+xfs_find_daxdev_for_inode(
+	struct inode		*inode)
+{
+	struct xfs_inode	*ip = XFS_I(inode);
+	struct xfs_mount	*mp = ip->i_mount;
+
+	if (XFS_IS_REALTIME_INODE(ip))
+		return mp->m_rtdev_targp->bt_daxdev;
+	else
+		return mp->m_ddev_targp->bt_daxdev;
+}
+
 /*
  * We're now finished for good with this page.  Update the page state via the
  * associated buffer_heads, paying attention to the start and end offsets that
diff --git a/fs/xfs/xfs_aops.h b/fs/xfs/xfs_aops.h
index cc174ec6c2fd..88c85ea63da0 100644
--- a/fs/xfs/xfs_aops.h
+++ b/fs/xfs/xfs_aops.h
@@ -59,5 +59,6 @@ int	xfs_setfilesize(struct xfs_inode *ip, xfs_off_t offset, size_t size);
 
 extern void xfs_count_page_state(struct page *, int *, int *);
 extern struct block_device *xfs_find_bdev_for_inode(struct inode *);
+extern struct dax_device *xfs_find_daxdev_for_inode(struct inode *);
 
 #endif /* __XFS_AOPS_H__ */
diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
index 72f038492ba8..6deb86c845d1 100644
--- a/fs/xfs/xfs_buf.c
+++ b/fs/xfs/xfs_buf.c
@@ -1802,7 +1802,8 @@ xfs_setsize_buftarg_early(
 xfs_buftarg_t *
 xfs_alloc_buftarg(
 	struct xfs_mount	*mp,
-	struct block_device	*bdev)
+	struct block_device	*bdev,
+	struct dax_device	*dax_dev)
 {
 	xfs_buftarg_t		*btp;
 
@@ -1811,6 +1812,7 @@ xfs_alloc_buftarg(
 	btp->bt_mount = mp;
 	btp->bt_dev =  bdev->bd_dev;
 	btp->bt_bdev = bdev;
+	btp->bt_daxdev = dax_dev;
 
 	if (xfs_setsize_buftarg_early(btp, bdev))
 		goto error;
diff --git a/fs/xfs/xfs_buf.h b/fs/xfs/xfs_buf.h
index 20721261dae5..bf71507ddb16 100644
--- a/fs/xfs/xfs_buf.h
+++ b/fs/xfs/xfs_buf.h
@@ -108,6 +108,7 @@ typedef unsigned int xfs_buf_flags_t;
 typedef struct xfs_buftarg {
 	dev_t			bt_dev;
 	struct block_device	*bt_bdev;
+	struct dax_device	*bt_daxdev;
 	struct xfs_mount	*bt_mount;
 	unsigned int		bt_meta_sectorsize;
 	size_t			bt_meta_sectormask;
@@ -385,7 +386,7 @@ xfs_buf_update_cksum(struct xfs_buf *bp, unsigned long cksum_offset)
  *	Handling of buftargs.
  */
 extern xfs_buftarg_t *xfs_alloc_buftarg(struct xfs_mount *,
-			struct block_device *);
+			struct block_device *, struct dax_device *);
 extern void xfs_free_buftarg(struct xfs_mount *, struct xfs_buftarg *);
 extern void xfs_wait_buftarg(xfs_buftarg_t *);
 extern int xfs_setsize_buftarg(xfs_buftarg_t *, unsigned int);
diff --git a/fs/xfs/xfs_iomap.c b/fs/xfs/xfs_iomap.c
index 813394c62849..7c934e407332 100644
--- a/fs/xfs/xfs_iomap.c
+++ b/fs/xfs/xfs_iomap.c
@@ -69,6 +69,7 @@ xfs_bmbt_to_iomap(
 	iomap->offset = XFS_FSB_TO_B(mp, imap->br_startoff);
 	iomap->length = XFS_FSB_TO_B(mp, imap->br_blockcount);
 	iomap->bdev = xfs_find_bdev_for_inode(VFS_I(ip));
+	iomap->dax_dev = xfs_find_daxdev_for_inode(VFS_I(ip));
 }
 
 xfs_extlen_t
@@ -976,7 +977,6 @@ xfs_file_iomap_begin(
 	int			nimaps = 1, error = 0;
 	bool			shared = false, trimmed = false;
 	unsigned		lockmode;
-	struct block_device	*bdev;
 
 	if (XFS_FORCED_SHUTDOWN(mp))
 		return -EIO;
@@ -1087,13 +1087,6 @@ xfs_file_iomap_begin(
 
 	xfs_bmbt_to_iomap(ip, iomap, &imap);
 
-	/* optionally associate a dax device with the iomap bdev */
-	bdev = iomap->bdev;
-	if (blk_queue_dax(bdev->bd_queue))
-		iomap->dax_dev = fs_dax_get_by_host(bdev->bd_disk->disk_name);
-	else
-		iomap->dax_dev = NULL;
-
 	if (shared)
 		iomap->flags |= IOMAP_F_SHARED;
 	return 0;
@@ -1171,7 +1164,6 @@ xfs_file_iomap_end(
 	unsigned		flags,
 	struct iomap		*iomap)
 {
-	fs_put_dax(iomap->dax_dev);
 	if ((flags & IOMAP_WRITE) && iomap->type == IOMAP_DELALLOC)
 		return xfs_file_iomap_end_delalloc(XFS_I(inode), offset,
 				length, written, iomap);
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index 38aaacdbb8b3..ee4225c65f0c 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -714,17 +714,26 @@ STATIC void
 xfs_close_devices(
 	struct xfs_mount	*mp)
 {
+	struct dax_device *dax_ddev = mp->m_ddev_targp->bt_daxdev;
+
 	if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
 		struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
+		struct dax_device *dax_logdev = mp->m_logdev_targp->bt_daxdev;
+
 		xfs_free_buftarg(mp, mp->m_logdev_targp);
 		xfs_blkdev_put(logdev);
+		fs_put_dax(dax_logdev);
 	}
 	if (mp->m_rtdev_targp) {
 		struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
+		struct dax_device *dax_rtdev = mp->m_rtdev_targp->bt_daxdev;
+
 		xfs_free_buftarg(mp, mp->m_rtdev_targp);
 		xfs_blkdev_put(rtdev);
+		fs_put_dax(dax_rtdev);
 	}
 	xfs_free_buftarg(mp, mp->m_ddev_targp);
+	fs_put_dax(dax_ddev);
 }
 
 /*
@@ -742,6 +751,8 @@ xfs_open_devices(
 	struct xfs_mount	*mp)
 {
 	struct block_device	*ddev = mp->m_super->s_bdev;
+	struct dax_device	*dax_ddev = fs_dax_get_by_bdev(ddev);
+	struct dax_device	*dax_logdev = NULL, *dax_rtdev = NULL;
 	struct block_device	*logdev = NULL, *rtdev = NULL;
 	int			error;
 
@@ -752,6 +763,7 @@ xfs_open_devices(
 		error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
 		if (error)
 			goto out;
+		dax_logdev = fs_dax_get_by_bdev(logdev);
 	}
 
 	if (mp->m_rtname) {
@@ -765,24 +777,25 @@ xfs_open_devices(
 			error = -EINVAL;
 			goto out_close_rtdev;
 		}
+		dax_rtdev = fs_dax_get_by_bdev(rtdev);
 	}
 
 	/*
 	 * Setup xfs_mount buffer target pointers
 	 */
 	error = -ENOMEM;
-	mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev);
+	mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev, dax_ddev);
 	if (!mp->m_ddev_targp)
 		goto out_close_rtdev;
 
 	if (rtdev) {
-		mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev);
+		mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev, dax_rtdev);
 		if (!mp->m_rtdev_targp)
 			goto out_free_ddev_targ;
 	}
 
 	if (logdev && logdev != ddev) {
-		mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev);
+		mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev, dax_logdev);
 		if (!mp->m_logdev_targp)
 			goto out_free_rtdev_targ;
 	} else {
@@ -798,10 +811,14 @@ xfs_open_devices(
 	xfs_free_buftarg(mp, mp->m_ddev_targp);
  out_close_rtdev:
 	xfs_blkdev_put(rtdev);
+	fs_put_dax(dax_rtdev);
  out_close_logdev:
-	if (logdev && logdev != ddev)
+	if (logdev && logdev != ddev) {
 		xfs_blkdev_put(logdev);
+		fs_put_dax(dax_logdev);
+	}
  out:
+	fs_put_dax(dax_ddev);
 	return error;
 }

Re: [PATCH v3 2/4] xfs: perform dax_device lookup at mount

[Darrick J. Wong]

On Wed, Aug 30, 2017 at 12:43:58PM -0700, Dan Williams wrote:
> The ->iomap_begin() operation is a hot path, so cache the
> fs_dax_get_by_host() result at mount time to avoid the incurring the
> hash lookup overhead on a per-i/o basis.
> 
> Cc: "Darrick J. Wong" <darrick.wong@oracle.com>
> Reported-by: Christoph Hellwig <hch@lst.de>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>

Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>

> ---
>  fs/xfs/xfs_aops.c  |   13 +++++++++++++
>  fs/xfs/xfs_aops.h  |    1 +
>  fs/xfs/xfs_buf.c   |    4 +++-
>  fs/xfs/xfs_buf.h   |    3 ++-
>  fs/xfs/xfs_iomap.c |   10 +---------
>  fs/xfs/xfs_super.c |   25 +++++++++++++++++++++----
>  6 files changed, 41 insertions(+), 15 deletions(-)
> 
> diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c
> index 6bf120bb1a17..78185f3b10b2 100644
> --- a/fs/xfs/xfs_aops.c
> +++ b/fs/xfs/xfs_aops.c
> @@ -80,6 +80,19 @@ xfs_find_bdev_for_inode(
>  		return mp->m_ddev_targp->bt_bdev;
>  }
>  
> +struct dax_device *
> +xfs_find_daxdev_for_inode(
> +	struct inode		*inode)
> +{
> +	struct xfs_inode	*ip = XFS_I(inode);
> +	struct xfs_mount	*mp = ip->i_mount;
> +
> +	if (XFS_IS_REALTIME_INODE(ip))
> +		return mp->m_rtdev_targp->bt_daxdev;
> +	else
> +		return mp->m_ddev_targp->bt_daxdev;
> +}
> +
>  /*
>   * We're now finished for good with this page.  Update the page state via the
>   * associated buffer_heads, paying attention to the start and end offsets that
> diff --git a/fs/xfs/xfs_aops.h b/fs/xfs/xfs_aops.h
> index cc174ec6c2fd..88c85ea63da0 100644
> --- a/fs/xfs/xfs_aops.h
> +++ b/fs/xfs/xfs_aops.h
> @@ -59,5 +59,6 @@ int	xfs_setfilesize(struct xfs_inode *ip, xfs_off_t offset, size_t size);
>  
>  extern void xfs_count_page_state(struct page *, int *, int *);
>  extern struct block_device *xfs_find_bdev_for_inode(struct inode *);
> +extern struct dax_device *xfs_find_daxdev_for_inode(struct inode *);
>  
>  #endif /* __XFS_AOPS_H__ */
> diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
> index 72f038492ba8..6deb86c845d1 100644
> --- a/fs/xfs/xfs_buf.c
> +++ b/fs/xfs/xfs_buf.c
> @@ -1802,7 +1802,8 @@ xfs_setsize_buftarg_early(
>  xfs_buftarg_t *
>  xfs_alloc_buftarg(
>  	struct xfs_mount	*mp,
> -	struct block_device	*bdev)
> +	struct block_device	*bdev,
> +	struct dax_device	*dax_dev)
>  {
>  	xfs_buftarg_t		*btp;
>  
> @@ -1811,6 +1812,7 @@ xfs_alloc_buftarg(
>  	btp->bt_mount = mp;
>  	btp->bt_dev =  bdev->bd_dev;
>  	btp->bt_bdev = bdev;
> +	btp->bt_daxdev = dax_dev;
>  
>  	if (xfs_setsize_buftarg_early(btp, bdev))
>  		goto error;
> diff --git a/fs/xfs/xfs_buf.h b/fs/xfs/xfs_buf.h
> index 20721261dae5..bf71507ddb16 100644
> --- a/fs/xfs/xfs_buf.h
> +++ b/fs/xfs/xfs_buf.h
> @@ -108,6 +108,7 @@ typedef unsigned int xfs_buf_flags_t;
>  typedef struct xfs_buftarg {
>  	dev_t			bt_dev;
>  	struct block_device	*bt_bdev;
> +	struct dax_device	*bt_daxdev;
>  	struct xfs_mount	*bt_mount;
>  	unsigned int		bt_meta_sectorsize;
>  	size_t			bt_meta_sectormask;
> @@ -385,7 +386,7 @@ xfs_buf_update_cksum(struct xfs_buf *bp, unsigned long cksum_offset)
>   *	Handling of buftargs.
>   */
>  extern xfs_buftarg_t *xfs_alloc_buftarg(struct xfs_mount *,
> -			struct block_device *);
> +			struct block_device *, struct dax_device *);
>  extern void xfs_free_buftarg(struct xfs_mount *, struct xfs_buftarg *);
>  extern void xfs_wait_buftarg(xfs_buftarg_t *);
>  extern int xfs_setsize_buftarg(xfs_buftarg_t *, unsigned int);
> diff --git a/fs/xfs/xfs_iomap.c b/fs/xfs/xfs_iomap.c
> index 813394c62849..7c934e407332 100644
> --- a/fs/xfs/xfs_iomap.c
> +++ b/fs/xfs/xfs_iomap.c
> @@ -69,6 +69,7 @@ xfs_bmbt_to_iomap(
>  	iomap->offset = XFS_FSB_TO_B(mp, imap->br_startoff);
>  	iomap->length = XFS_FSB_TO_B(mp, imap->br_blockcount);
>  	iomap->bdev = xfs_find_bdev_for_inode(VFS_I(ip));
> +	iomap->dax_dev = xfs_find_daxdev_for_inode(VFS_I(ip));
>  }
>  
>  xfs_extlen_t
> @@ -976,7 +977,6 @@ xfs_file_iomap_begin(
>  	int			nimaps = 1, error = 0;
>  	bool			shared = false, trimmed = false;
>  	unsigned		lockmode;
> -	struct block_device	*bdev;
>  
>  	if (XFS_FORCED_SHUTDOWN(mp))
>  		return -EIO;
> @@ -1087,13 +1087,6 @@ xfs_file_iomap_begin(
>  
>  	xfs_bmbt_to_iomap(ip, iomap, &imap);
>  
> -	/* optionally associate a dax device with the iomap bdev */
> -	bdev = iomap->bdev;
> -	if (blk_queue_dax(bdev->bd_queue))
> -		iomap->dax_dev = fs_dax_get_by_host(bdev->bd_disk->disk_name);
> -	else
> -		iomap->dax_dev = NULL;
> -
>  	if (shared)
>  		iomap->flags |= IOMAP_F_SHARED;
>  	return 0;
> @@ -1171,7 +1164,6 @@ xfs_file_iomap_end(
>  	unsigned		flags,
>  	struct iomap		*iomap)
>  {
> -	fs_put_dax(iomap->dax_dev);
>  	if ((flags & IOMAP_WRITE) && iomap->type == IOMAP_DELALLOC)
>  		return xfs_file_iomap_end_delalloc(XFS_I(inode), offset,
>  				length, written, iomap);
> diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
> index 38aaacdbb8b3..ee4225c65f0c 100644
> --- a/fs/xfs/xfs_super.c
> +++ b/fs/xfs/xfs_super.c
> @@ -714,17 +714,26 @@ STATIC void
>  xfs_close_devices(
>  	struct xfs_mount	*mp)
>  {
> +	struct dax_device *dax_ddev = mp->m_ddev_targp->bt_daxdev;
> +
>  	if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
>  		struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
> +		struct dax_device *dax_logdev = mp->m_logdev_targp->bt_daxdev;
> +
>  		xfs_free_buftarg(mp, mp->m_logdev_targp);
>  		xfs_blkdev_put(logdev);
> +		fs_put_dax(dax_logdev);
>  	}
>  	if (mp->m_rtdev_targp) {
>  		struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
> +		struct dax_device *dax_rtdev = mp->m_rtdev_targp->bt_daxdev;
> +
>  		xfs_free_buftarg(mp, mp->m_rtdev_targp);
>  		xfs_blkdev_put(rtdev);
> +		fs_put_dax(dax_rtdev);
>  	}
>  	xfs_free_buftarg(mp, mp->m_ddev_targp);
> +	fs_put_dax(dax_ddev);
>  }
>  
>  /*
> @@ -742,6 +751,8 @@ xfs_open_devices(
>  	struct xfs_mount	*mp)
>  {
>  	struct block_device	*ddev = mp->m_super->s_bdev;
> +	struct dax_device	*dax_ddev = fs_dax_get_by_bdev(ddev);
> +	struct dax_device	*dax_logdev = NULL, *dax_rtdev = NULL;
>  	struct block_device	*logdev = NULL, *rtdev = NULL;
>  	int			error;
>  
> @@ -752,6 +763,7 @@ xfs_open_devices(
>  		error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
>  		if (error)
>  			goto out;
> +		dax_logdev = fs_dax_get_by_bdev(logdev);
>  	}
>  
>  	if (mp->m_rtname) {
> @@ -765,24 +777,25 @@ xfs_open_devices(
>  			error = -EINVAL;
>  			goto out_close_rtdev;
>  		}
> +		dax_rtdev = fs_dax_get_by_bdev(rtdev);
>  	}
>  
>  	/*
>  	 * Setup xfs_mount buffer target pointers
>  	 */
>  	error = -ENOMEM;
> -	mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev);
> +	mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev, dax_ddev);
>  	if (!mp->m_ddev_targp)
>  		goto out_close_rtdev;
>  
>  	if (rtdev) {
> -		mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev);
> +		mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev, dax_rtdev);
>  		if (!mp->m_rtdev_targp)
>  			goto out_free_ddev_targ;
>  	}
>  
>  	if (logdev && logdev != ddev) {
> -		mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev);
> +		mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev, dax_logdev);
>  		if (!mp->m_logdev_targp)
>  			goto out_free_rtdev_targ;
>  	} else {
> @@ -798,10 +811,14 @@ xfs_open_devices(
>  	xfs_free_buftarg(mp, mp->m_ddev_targp);
>   out_close_rtdev:
>  	xfs_blkdev_put(rtdev);
> +	fs_put_dax(dax_rtdev);
>   out_close_logdev:
> -	if (logdev && logdev != ddev)
> +	if (logdev && logdev != ddev) {
>  		xfs_blkdev_put(logdev);
> +		fs_put_dax(dax_logdev);
> +	}
>   out:
> +	fs_put_dax(dax_ddev);
>  	return error;
>  }
>  
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH v3 2/4] xfs: perform dax_device lookup at mount

[Christoph Hellwig]

We don't really need the dax_device for the log device (yet), but
acquiring it seems harmless, so:

Reviewed-by: Christoph Hellwig <hch@lst.de>

[PATCH v3 3/4] ext2: perform dax_device lookup at mount

[Dan Williams]

The ->iomap_begin() operation is a hot path, so cache the
fs_dax_get_by_host() result at mount time to avoid the incurring the
hash lookup overhead on a per-i/o basis.

Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Andreas Dilger <adilger.kernel@dilger.ca>
Cc: Jan Kara <jack@suse.cz>
Reported-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 fs/ext2/ext2.h  |    1 +
 fs/ext2/inode.c |   11 +++--------
 fs/ext2/super.c |    5 +++++
 3 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/fs/ext2/ext2.h b/fs/ext2/ext2.h
index 23ebb92484c6..28de3edd4f4d 100644
--- a/fs/ext2/ext2.h
+++ b/fs/ext2/ext2.h
@@ -114,6 +114,7 @@ struct ext2_sb_info {
 	 */
 	spinlock_t s_lock;
 	struct mb_cache *s_ea_block_cache;
+	struct dax_device *s_daxdev;
 };
 
 static inline spinlock_t *
diff --git a/fs/ext2/inode.c b/fs/ext2/inode.c
index 30163d007b2f..4dca6f348714 100644
--- a/fs/ext2/inode.c
+++ b/fs/ext2/inode.c
@@ -800,10 +800,10 @@ int ext2_get_block(struct inode *inode, sector_t iblock,
 static int ext2_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 		unsigned flags, struct iomap *iomap)
 {
-	struct block_device *bdev;
 	unsigned int blkbits = inode->i_blkbits;
 	unsigned long first_block = offset >> blkbits;
 	unsigned long max_blocks = (length + (1 << blkbits) - 1) >> blkbits;
+	struct ext2_sb_info *sbi = EXT2_SB(inode->i_sb);
 	bool new = false, boundary = false;
 	u32 bno;
 	int ret;
@@ -814,13 +814,9 @@ static int ext2_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 		return ret;
 
 	iomap->flags = 0;
-	bdev = inode->i_sb->s_bdev;
-	iomap->bdev = bdev;
+	iomap->bdev = inode->i_sb->s_bdev;
 	iomap->offset = (u64)first_block << blkbits;
-	if (blk_queue_dax(bdev->bd_queue))
-		iomap->dax_dev = fs_dax_get_by_host(bdev->bd_disk->disk_name);
-	else
-		iomap->dax_dev = NULL;
+	iomap->dax_dev = sbi->s_daxdev;
 
 	if (ret == 0) {
 		iomap->type = IOMAP_HOLE;
@@ -842,7 +838,6 @@ static int
 ext2_iomap_end(struct inode *inode, loff_t offset, loff_t length,
 		ssize_t written, unsigned flags, struct iomap *iomap)
 {
-	fs_put_dax(iomap->dax_dev);
 	if (iomap->type == IOMAP_MAPPED &&
 	    written < length &&
 	    (flags & IOMAP_WRITE))
diff --git a/fs/ext2/super.c b/fs/ext2/super.c
index 7b1bc9059863..d9dd999568c2 100644
--- a/fs/ext2/super.c
+++ b/fs/ext2/super.c
@@ -144,6 +144,7 @@ static void ext2_put_super (struct super_block * sb)
 	int db_count;
 	int i;
 	struct ext2_sb_info *sbi = EXT2_SB(sb);
+	struct dax_device *dax_dev = sbi->s_daxdev;
 
 	ext2_quota_off_umount(sb);
 
@@ -172,6 +173,7 @@ static void ext2_put_super (struct super_block * sb)
 	sb->s_fs_info = NULL;
 	kfree(sbi->s_blockgroup_lock);
 	kfree(sbi);
+	fs_put_dax(dax_dev);
 }
 
 static struct kmem_cache * ext2_inode_cachep;
@@ -813,6 +815,7 @@ static unsigned long descriptor_loc(struct super_block *sb,
 
 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 {
+	struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
 	struct buffer_head * bh;
 	struct ext2_sb_info * sbi;
 	struct ext2_super_block * es;
@@ -842,6 +845,7 @@ static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 	}
 	sb->s_fs_info = sbi;
 	sbi->s_sb_block = sb_block;
+	sbi->s_daxdev = dax_dev;
 
 	spin_lock_init(&sbi->s_lock);
 
@@ -1200,6 +1204,7 @@ static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 	kfree(sbi->s_blockgroup_lock);
 	kfree(sbi);
 failed:
+	fs_put_dax(dax_dev);
 	return ret;
 }
 

Re: [PATCH v3 3/4] ext2: perform dax_device lookup at mount

[Jan Kara]

On Wed 30-08-17 12:44:03, Dan Williams wrote:
> The ->iomap_begin() operation is a hot path, so cache the
> fs_dax_get_by_host() result at mount time to avoid the incurring the
> hash lookup overhead on a per-i/o basis.
> 
> Cc: "Theodore Ts'o" <tytso@mit.edu>
> Cc: Andreas Dilger <adilger.kernel@dilger.ca>
> Cc: Jan Kara <jack@suse.cz>
> Reported-by: Christoph Hellwig <hch@lst.de>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>

Looks good to me (and it looks even cleaner than your previous approach).
Just one nit below. You can add:

Reviewed-by: Jan Kara <jack@suse.cz>

> diff --git a/fs/ext2/super.c b/fs/ext2/super.c
> index 7b1bc9059863..d9dd999568c2 100644
> --- a/fs/ext2/super.c
> +++ b/fs/ext2/super.c
> @@ -144,6 +144,7 @@ static void ext2_put_super (struct super_block * sb)
>  	int db_count;
>  	int i;
>  	struct ext2_sb_info *sbi = EXT2_SB(sb);
> +	struct dax_device *dax_dev = sbi->s_daxdev;
>  
>  	ext2_quota_off_umount(sb);
>  
> @@ -172,6 +173,7 @@ static void ext2_put_super (struct super_block * sb)
>  	sb->s_fs_info = NULL;
>  	kfree(sbi->s_blockgroup_lock);
>  	kfree(sbi);
> +	fs_put_dax(dax_dev);

The local variable looks superfluous here. I'd just do
	fs_put_dax(sbi->s_daxdev);

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

Re: [PATCH v3 3/4] ext2: perform dax_device lookup at mount

[Dan Williams]

On Thu, Aug 31, 2017 at 1:22 AM, Jan Kara <jack@suse.cz> wrote:
> On Wed 30-08-17 12:44:03, Dan Williams wrote:
>> The ->iomap_begin() operation is a hot path, so cache the
>> fs_dax_get_by_host() result at mount time to avoid the incurring the
>> hash lookup overhead on a per-i/o basis.
>>
>> Cc: "Theodore Ts'o" <tytso@mit.edu>
>> Cc: Andreas Dilger <adilger.kernel@dilger.ca>
>> Cc: Jan Kara <jack@suse.cz>
>> Reported-by: Christoph Hellwig <hch@lst.de>
>> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
>
> Looks good to me (and it looks even cleaner than your previous approach).
> Just one nit below. You can add:
>
> Reviewed-by: Jan Kara <jack@suse.cz>
>
>> diff --git a/fs/ext2/super.c b/fs/ext2/super.c
>> index 7b1bc9059863..d9dd999568c2 100644
>> --- a/fs/ext2/super.c
>> +++ b/fs/ext2/super.c
>> @@ -144,6 +144,7 @@ static void ext2_put_super (struct super_block * sb)
>>       int db_count;
>>       int i;
>>       struct ext2_sb_info *sbi = EXT2_SB(sb);
>> +     struct dax_device *dax_dev = sbi->s_daxdev;
>>
>>       ext2_quota_off_umount(sb);
>>
>> @@ -172,6 +173,7 @@ static void ext2_put_super (struct super_block * sb)
>>       sb->s_fs_info = NULL;
>>       kfree(sbi->s_blockgroup_lock);
>>       kfree(sbi);
>> +     fs_put_dax(dax_dev);
>
> The local variable looks superfluous here. I'd just do
>         fs_put_dax(sbi->s_daxdev);

Ok... and move that before the kfree(sbi).

[PATCH v3 4/4] ext4: perform dax_device lookup at mount

[Dan Williams]

The ->iomap_begin() operation is a hot path, so cache the
fs_dax_get_by_host() result at mount time to avoid the incurring the
hash lookup overhead on a per-i/o basis.

Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Andreas Dilger <adilger.kernel@dilger.ca>
Cc: Jan Kara <jack@suse.cz>
Reported-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 fs/ext4/ext4.h  |    1 +
 fs/ext4/inode.c |   11 +++--------
 fs/ext4/super.c |    5 +++++
 3 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index a2bb7d2870e4..194e622dc3dd 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -1528,6 +1528,7 @@ struct ext4_sb_info {
 
 	/* Barrier between changing inodes' journal flags and writepages ops. */
 	struct percpu_rw_semaphore s_journal_flag_rwsem;
+	struct dax_device *s_daxdev;
 };
 
 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index c774bdc22759..16424b5c4e88 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -3404,7 +3404,7 @@ static int ext4_releasepage(struct page *page, gfp_t wait)
 static int ext4_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 			    unsigned flags, struct iomap *iomap)
 {
-	struct block_device *bdev;
+	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
 	unsigned int blkbits = inode->i_blkbits;
 	unsigned long first_block = offset >> blkbits;
 	unsigned long last_block = (offset + length - 1) >> blkbits;
@@ -3473,12 +3473,8 @@ static int ext4_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 	}
 
 	iomap->flags = 0;
-	bdev = inode->i_sb->s_bdev;
-	iomap->bdev = bdev;
-	if (blk_queue_dax(bdev->bd_queue))
-		iomap->dax_dev = fs_dax_get_by_host(bdev->bd_disk->disk_name);
-	else
-		iomap->dax_dev = NULL;
+	iomap->bdev = inode->i_sb->s_bdev;
+	iomap->dax_dev = sbi->s_daxdev;
 	iomap->offset = first_block << blkbits;
 
 	if (ret == 0) {
@@ -3511,7 +3507,6 @@ static int ext4_iomap_end(struct inode *inode, loff_t offset, loff_t length,
 	int blkbits = inode->i_blkbits;
 	bool truncate = false;
 
-	fs_put_dax(iomap->dax_dev);
 	if (!(flags & IOMAP_WRITE) || (flags & IOMAP_FAULT))
 		return 0;
 
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index d61a70e2193a..0b3d94e25efb 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -864,6 +864,7 @@ static inline void ext4_quota_off_umount(struct super_block *sb)
 static void ext4_put_super(struct super_block *sb)
 {
 	struct ext4_sb_info *sbi = EXT4_SB(sb);
+	struct dax_device *dax_dev = sbi->s_daxdev;
 	struct ext4_super_block *es = sbi->s_es;
 	int aborted = 0;
 	int i, err;
@@ -952,6 +953,7 @@ static void ext4_put_super(struct super_block *sb)
 		crypto_free_shash(sbi->s_chksum_driver);
 	kfree(sbi->s_blockgroup_lock);
 	kfree(sbi);
+	fs_put_dax(dax_dev);
 }
 
 static struct kmem_cache *ext4_inode_cachep;
@@ -3377,6 +3379,7 @@ static void ext4_set_resv_clusters(struct super_block *sb)
 
 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 {
+	struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
 	char *orig_data = kstrdup(data, GFP_KERNEL);
 	struct buffer_head *bh;
 	struct ext4_super_block *es = NULL;
@@ -3399,6 +3402,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
 	ext4_group_t first_not_zeroed;
 
+	sbi->s_daxdev = dax_dev;
 	if ((data && !orig_data) || !sbi)
 		goto out_free_base;
 
@@ -4378,6 +4382,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 out_free_base:
 	kfree(sbi);
 	kfree(orig_data);
+	fs_put_dax(dax_dev);
 	return err ? err : ret;
 }
 

Re: [PATCH v3 4/4] ext4: perform dax_device lookup at mount

[Jan Kara]

On Wed 30-08-17 12:44:08, Dan Williams wrote:
> The ->iomap_begin() operation is a hot path, so cache the
> fs_dax_get_by_host() result at mount time to avoid the incurring the
> hash lookup overhead on a per-i/o basis.
> 
> Cc: "Theodore Ts'o" <tytso@mit.edu>
> Cc: Andreas Dilger <adilger.kernel@dilger.ca>
> Cc: Jan Kara <jack@suse.cz>
> Reported-by: Christoph Hellwig <hch@lst.de>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>

Looks good to me. You can add:

Reviewed-by: Jan Kara <jack@suse.cz>

With the same nit as for ext2:

> diff --git a/fs/ext4/super.c b/fs/ext4/super.c
> index d61a70e2193a..0b3d94e25efb 100644
> --- a/fs/ext4/super.c
> +++ b/fs/ext4/super.c
> @@ -864,6 +864,7 @@ static inline void ext4_quota_off_umount(struct super_block *sb)
>  static void ext4_put_super(struct super_block *sb)
>  {
>  	struct ext4_sb_info *sbi = EXT4_SB(sb);
> +	struct dax_device *dax_dev = sbi->s_daxdev;
>  	struct ext4_super_block *es = sbi->s_es;
>  	int aborted = 0;
>  	int i, err;
> @@ -952,6 +953,7 @@ static void ext4_put_super(struct super_block *sb)
>  		crypto_free_shash(sbi->s_chksum_driver);
>  	kfree(sbi->s_blockgroup_lock);
>  	kfree(sbi);
> +	fs_put_dax(dax_dev);

Just remove dax_dev local variable...

								Honza

-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR