[PATCH V8 00/11] Enable per-file/per-directory DAX operations V8

[PATCH V8 00/11] Enable per-file/per-directory DAX operations V8

[ira.weiny]

From: Ira Weiny <ira.weiny@intel.com>

Changes from V7:
	Add DCACHE_DONTCACHE
	If mount override don't worry about inode caching
	Change mount flags to NEVER/ALWAYS
	Clean up xfs_inode_enable_dax()
	Clarify comments
	Update documentation

At LSF/MM'19 [1] [2] we discussed applications that overestimate memory
consumption due to their inability to detect whether the kernel will
instantiate page cache for a file, and cases where a global dax enable via a
mount option is too coarse.

The following patch series enables the use of DAX on individual files and/or
directories on xfs, and lays some groundwork to do so in ext4.  It further
enhances the dax mount option to be a tri-state of 'always', 'never', or
'iflag' (default).  Furthermore, it maintians '-o dax' to be equivalent to '-o
dax=always'.

The insight at LSF/MM was to separate the per-mount or per-file "physical"
(FS_XFLAG_DAX) capability switch from an "effective" (S_DAX) attribute for the
file.

At LSF/MM we discussed the difficulties of switching the DAX state of a file
with active mappings / page cache.  It was thought the races could be avoided
by limiting DAX state flips to 0-length files.

However, this turns out to not be true.[3][5] This is because address space
operations (a_ops) may be in use at any time the inode is referenced.

For this reason direct manipulation of the FS_XFLAG_DAX is allowed but the
operation of the file (S_DAX) is not immediately changed.

Details of when and how DAX state can be changed on a file is included in a
documentation patch.

It should be noted that FS_XFLAG_DAX inheritance is not shown in this patch set
as it was maintained from previous work on XFS.  FS_XFLAG_DAX and it's
inheritance will need to be added to other file systems for user control. 


[1] https://lwn.net/Articles/787973/
[2] https://lwn.net/Articles/787233/
[3] https://lkml.org/lkml/2019/10/20/96
[4] https://patchwork.kernel.org/patch/11310511/
[5] https://lore.kernel.org/lkml/20200405061945.GA94792@iweiny-DESK2.sc.intel.com/


To: linux-kernel@vger.kernel.org
Cc: "Darrick J. Wong" <darrick.wong@oracle.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: "Theodore Y. Ts'o" <tytso@mit.edu>
Cc: Jan Kara <jack@suse.cz>
Cc: linux-ext4@vger.kernel.org
Cc: linux-xfs@vger.kernel.org
Cc: linux-fsdevel@vger.kernel.org

Changes from V6:
	Incorporate feedback on patches
	Add ability to change FS_XFLAG_DAX on files at any time.
		Add a don't cache flag to the VFS layer
		Preserve internal XFS IDONTCACHE behavior for bulkstat
		operations

Changes from V5:
	* make dax mount option a tri-state
	* Reject changes to FS_XFLAG_DAX for regular files
		- Allow only on directories
	* Update documentation

Changes from V4:
	* Open code the aops lock rather than add it to the xfs_ilock()
	  subsystem (Darrick's comments were obsoleted by this change)
	* Fix lkp build suggestions and bugs

Changes from V3:
	* Remove global locking...  :-D
	* put back per inode locking and remove pre-mature optimizations
	* Fix issues with Directories having IS_DAX() set
	* Fix kernel crash issues reported by Jeff
	* Add some clean up patches
	* Consolidate diflags to iflags functions
	* Update/add documentation
	* Reorder/rename patches quite a bit

Changes from V2:

	* Move i_dax_sem to be a global percpu_rw_sem rather than per inode
		Internal discussions with Dan determined this would be easier,
		just as performant, and slightly less overhead that having it
		in the SB as suggested by Jan
	* Fix locking order in comments and throughout code
	* Change "mode" to "state" throughout commits
	* Add CONFIG_FS_DAX wrapper to disable inode_[un]lock_state() when not
		configured
	* Add static branch for which is activated by a device which supports
		DAX in XFS
	* Change "lock/unlock" to up/down read/write as appropriate
		Previous names were over simplified
	* Update comments/documentation

	* Remove the xfs specific lock to the vfs (global) layer.
	* Fix i_dax_sem locking order and comments

	* Move 'i_mapped' count from struct inode to struct address_space and
		rename it to mmap_count
	* Add inode_has_mappings() call

	* Fix build issues
	* Clean up syntax spacing and minor issues
	* Update man page text for STATX_ATTR_DAX
	* Add reviewed-by's
	* Rebase to 5.6

	Rename patch:
		from: fs/xfs: Add lock/unlock state to xfs
		to: fs/xfs: Add write DAX lock to xfs layer
	Add patch:
		fs/xfs: Clarify lockdep dependency for xfs_isilocked()
	Drop patch:
		fs/xfs: Fix truncate up

Ira Weiny (11):
  fs/xfs: Remove unnecessary initialization of i_rwsem
  fs: Remove unneeded IS_DAX() check in io_is_direct()
  fs/stat: Define DAX statx attribute
  fs/xfs: Change XFS_MOUNT_DAX to XFS_MOUNT_DAX_ALWAYS
  fs/xfs: Make DAX mount option a tri-state
  fs/xfs: Create function xfs_inode_enable_dax()
  fs/xfs: Combine xfs_diflags_to_linux() and xfs_diflags_to_iflags()
  fs: Define I_DONTCACNE in VFS layer
  fs: Introduce DCACHE_DONTCACHE
  fs/xfs: Change xfs_ioctl_setattr_dax_invalidate()
  Documentation/dax: Update Usage section

 Documentation/filesystems/dax.txt | 166 +++++++++++++++++++++++++++++-
 drivers/block/loop.c              |   6 +-
 fs/dcache.c                       |   4 +
 fs/inode.c                        |  15 +++
 fs/stat.c                         |   3 +
 fs/xfs/xfs_icache.c               |   4 +-
 fs/xfs/xfs_inode.h                |   1 +
 fs/xfs/xfs_ioctl.c                | 138 ++++---------------------
 fs/xfs/xfs_iops.c                 |  72 ++++++++-----
 fs/xfs/xfs_mount.h                |   3 +-
 fs/xfs/xfs_super.c                |  46 ++++++++-
 include/linux/dcache.h            |   2 +
 include/linux/fs.h                |  14 +--
 include/uapi/linux/stat.h         |   1 +
 14 files changed, 312 insertions(+), 163 deletions(-)

-- 
2.25.1

[PATCH V8 01/11] fs/xfs: Remove unnecessary initialization of i_rwsem

[ira.weiny]

From: Ira Weiny <ira.weiny@intel.com>

An earlier call of xfs_reinit_inode() from xfs_iget_cache_hit() already
handles initialization of i_rwsem.

Doing so again is unneeded.

Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Ira Weiny <ira.weiny@intel.com>

---
Changes from V4:
	Update commit message to make it clear the xfs_iget_cache_hit()
	is actually doing the initialization via xfs_reinit_inode()

New for V4:

NOTE: This was found while ensuring the new i_aops_sem was properly
handled.  It seems like this is a layering violation so I think it is
worth cleaning up so as to not confuse others.
---
 fs/xfs/xfs_icache.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c
index 8dc2e5414276..836a1f09be03 100644
--- a/fs/xfs/xfs_icache.c
+++ b/fs/xfs/xfs_icache.c
@@ -419,6 +419,7 @@ xfs_iget_cache_hit(
 		spin_unlock(&ip->i_flags_lock);
 		rcu_read_unlock();
 
+		ASSERT(!rwsem_is_locked(&inode->i_rwsem));
 		error = xfs_reinit_inode(mp, inode);
 		if (error) {
 			bool wake;
@@ -452,9 +453,6 @@ xfs_iget_cache_hit(
 		ip->i_sick = 0;
 		ip->i_checked = 0;
 
-		ASSERT(!rwsem_is_locked(&inode->i_rwsem));
-		init_rwsem(&inode->i_rwsem);
-
 		spin_unlock(&ip->i_flags_lock);
 		spin_unlock(&pag->pag_ici_lock);
 	} else {
-- 
2.25.1

[PATCH V8 02/11] fs: Remove unneeded IS_DAX() check in io_is_direct()

[ira.weiny]

From: Ira Weiny <ira.weiny@intel.com>

Remove the check because DAX now has it's own read/write methods and
file systems which support DAX check IS_DAX() prior to IOCB_DIRECT on
their own.  Therefore, it does not matter if the file state is DAX when
the iocb flags are created.

Also remove io_is_direct() as it is just a simple flag check.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ira Weiny <ira.weiny@intel.com>

---
Changes from v6:
	remove io_is_direct() as well.
	Remove Reviews since this is quite a bit different.

Changes from v3:
	Reword commit message.
	Reordered to be a 'pre-cleanup' patch
---
 drivers/block/loop.c | 6 +++---
 include/linux/fs.h   | 7 +------
 2 files changed, 4 insertions(+), 9 deletions(-)

diff --git a/drivers/block/loop.c b/drivers/block/loop.c
index 739b372a5112..9a9af78974ac 100644
--- a/drivers/block/loop.c
+++ b/drivers/block/loop.c
@@ -631,8 +631,8 @@ static int do_req_filebacked(struct loop_device *lo, struct request *rq)
 
 static inline void loop_update_dio(struct loop_device *lo)
 {
-	__loop_update_dio(lo, io_is_direct(lo->lo_backing_file) |
-			lo->use_dio);
+	__loop_update_dio(lo, (lo->lo_backing_file->f_flags & O_DIRECT) |
+				lo->use_dio);
 }
 
 static void loop_reread_partitions(struct loop_device *lo,
@@ -1006,7 +1006,7 @@ static int loop_set_fd(struct loop_device *lo, fmode_t mode,
 	if (!(lo_flags & LO_FLAGS_READ_ONLY) && file->f_op->fsync)
 		blk_queue_write_cache(lo->lo_queue, true, false);
 
-	if (io_is_direct(lo->lo_backing_file) && inode->i_sb->s_bdev) {
+	if ((lo->lo_backing_file->f_flags & O_DIRECT) && inode->i_sb->s_bdev) {
 		/* In case of direct I/O, match underlying block size */
 		unsigned short bsize = bdev_logical_block_size(
 			inode->i_sb->s_bdev);
diff --git a/include/linux/fs.h b/include/linux/fs.h
index abedbffe2c9e..a818ced22961 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -3387,11 +3387,6 @@ extern void setattr_copy(struct inode *inode, const struct iattr *attr);
 
 extern int file_update_time(struct file *file);
 
-static inline bool io_is_direct(struct file *filp)
-{
-	return (filp->f_flags & O_DIRECT) || IS_DAX(filp->f_mapping->host);
-}
-
 static inline bool vma_is_dax(struct vm_area_struct *vma)
 {
 	return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host);
@@ -3416,7 +3411,7 @@ static inline int iocb_flags(struct file *file)
 	int res = 0;
 	if (file->f_flags & O_APPEND)
 		res |= IOCB_APPEND;
-	if (io_is_direct(file))
+	if (file->f_flags & O_DIRECT)
 		res |= IOCB_DIRECT;
 	if ((file->f_flags & O_DSYNC) || IS_SYNC(file->f_mapping->host))
 		res |= IOCB_DSYNC;
-- 
2.25.1

[PATCH V8 03/11] fs/stat: Define DAX statx attribute

[ira.weiny]

From: Ira Weiny <ira.weiny@intel.com>

In order for users to determine if a file is currently operating in DAX
state (effective DAX).  Define a statx attribute value and set that
attribute if the effective DAX flag is set.

To go along with this we propose the following addition to the statx man
page:

STATX_ATTR_DAX

	The file is in the DAX (cpu direct access) state.  DAX state
	attempts to minimize software cache effects for both I/O and
	memory mappings of this file.  It requires a file system which
	has been configured to support DAX.

	DAX generally assumes all accesses are via cpu load / store
	instructions which can minimize overhead for small accesses, but
	may adversely affect cpu utilization for large transfers.

	File I/O is done directly to/from user-space buffers and memory
	mapped I/O may be performed with direct memory mappings that
	bypass kernel page cache.

	While the DAX property tends to result in data being transferred
	synchronously, it does not give the same guarantees of O_SYNC
	where data and the necessary metadata are transferred together.

	A DAX file may support being mapped with the MAP_SYNC flag,
	which enables a program to use CPU cache flush instructions to
	persist CPU store operations without an explicit fsync(2).  See
	mmap(2) for more information.

Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Ira Weiny <ira.weiny@intel.com>

---
Changes from V2:
	Update man page text with comments from Darrick, Jan, Dan, and
	Dave.
---
 fs/stat.c                 | 3 +++
 include/uapi/linux/stat.h | 1 +
 2 files changed, 4 insertions(+)

diff --git a/fs/stat.c b/fs/stat.c
index 030008796479..894699c74dde 100644
--- a/fs/stat.c
+++ b/fs/stat.c
@@ -79,6 +79,9 @@ int vfs_getattr_nosec(const struct path *path, struct kstat *stat,
 	if (IS_AUTOMOUNT(inode))
 		stat->attributes |= STATX_ATTR_AUTOMOUNT;
 
+	if (IS_DAX(inode))
+		stat->attributes |= STATX_ATTR_DAX;
+
 	if (inode->i_op->getattr)
 		return inode->i_op->getattr(path, stat, request_mask,
 					    query_flags);
diff --git a/include/uapi/linux/stat.h b/include/uapi/linux/stat.h
index ad80a5c885d5..e5f9d5517f6b 100644
--- a/include/uapi/linux/stat.h
+++ b/include/uapi/linux/stat.h
@@ -169,6 +169,7 @@ struct statx {
 #define STATX_ATTR_ENCRYPTED		0x00000800 /* [I] File requires key to decrypt in fs */
 #define STATX_ATTR_AUTOMOUNT		0x00001000 /* Dir: Automount trigger */
 #define STATX_ATTR_VERITY		0x00100000 /* [I] Verity protected file */
+#define STATX_ATTR_DAX			0x00002000 /* [I] File is DAX */
 
 
 #endif /* _UAPI_LINUX_STAT_H */
-- 
2.25.1

[PATCH V8 04/11] fs/xfs: Change XFS_MOUNT_DAX to XFS_MOUNT_DAX_ALWAYS

[ira.weiny]

From: Ira Weiny <ira.weiny@intel.com>

In prep for the new tri-state mount option which then introduces
XFS_MOUNT_DAX_NEVER.

Signed-off-by: Ira Weiny <ira.weiny@intel.com>
---
 fs/xfs/xfs_iops.c  | 2 +-
 fs/xfs/xfs_mount.h | 2 +-
 fs/xfs/xfs_super.c | 8 ++++----
 3 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
index 81f2f93caec0..a6e634631da8 100644
--- a/fs/xfs/xfs_iops.c
+++ b/fs/xfs/xfs_iops.c
@@ -1248,7 +1248,7 @@ xfs_inode_supports_dax(
 		return false;
 
 	/* DAX mount option or DAX iflag must be set. */
-	if (!(mp->m_flags & XFS_MOUNT_DAX) &&
+	if (!(mp->m_flags & XFS_MOUNT_DAX_ALWAYS) &&
 	    !(ip->i_d.di_flags2 & XFS_DIFLAG2_DAX))
 		return false;
 
diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index 88ab09ed29e7..54bd74088936 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -233,7 +233,7 @@ typedef struct xfs_mount {
 						   allocator */
 #define XFS_MOUNT_NOATTR2	(1ULL << 25)	/* disable use of attr2 format */
 
-#define XFS_MOUNT_DAX		(1ULL << 62)	/* TEST ONLY! */
+#define XFS_MOUNT_DAX_ALWAYS	(1ULL << 62)	/* TEST ONLY! */
 
 /*
  * Max and min values for mount-option defined I/O
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index 2094386af8ac..3863f41757d2 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -129,7 +129,7 @@ xfs_fs_show_options(
 		{ XFS_MOUNT_GRPID,		",grpid" },
 		{ XFS_MOUNT_DISCARD,		",discard" },
 		{ XFS_MOUNT_LARGEIO,		",largeio" },
-		{ XFS_MOUNT_DAX,		",dax" },
+		{ XFS_MOUNT_DAX_ALWAYS,		",dax" },
 		{ 0, NULL }
 	};
 	struct xfs_mount	*mp = XFS_M(root->d_sb);
@@ -1244,7 +1244,7 @@ xfs_fc_parse_param(
 		return 0;
 #ifdef CONFIG_FS_DAX
 	case Opt_dax:
-		mp->m_flags |= XFS_MOUNT_DAX;
+		mp->m_flags |= XFS_MOUNT_DAX_ALWAYS;
 		return 0;
 #endif
 	default:
@@ -1437,7 +1437,7 @@ xfs_fc_fill_super(
 	if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5)
 		sb->s_flags |= SB_I_VERSION;
 
-	if (mp->m_flags & XFS_MOUNT_DAX) {
+	if (mp->m_flags & XFS_MOUNT_DAX_ALWAYS) {
 		bool rtdev_is_dax = false, datadev_is_dax;
 
 		xfs_warn(mp,
@@ -1451,7 +1451,7 @@ xfs_fc_fill_super(
 		if (!rtdev_is_dax && !datadev_is_dax) {
 			xfs_alert(mp,
 			"DAX unsupported by block device. Turning off DAX.");
-			mp->m_flags &= ~XFS_MOUNT_DAX;
+			mp->m_flags &= ~XFS_MOUNT_DAX_ALWAYS;
 		}
 		if (xfs_sb_version_hasreflink(&mp->m_sb)) {
 			xfs_alert(mp,
-- 
2.25.1

[PATCH V8 05/11] fs/xfs: Make DAX mount option a tri-state

[ira.weiny]

From: Ira Weiny <ira.weiny@intel.com>

As agreed upon[1].  We make the dax mount option a tri-state.  '-o dax'
continues to operate the same.  We add 'always', 'never', and 'inode'
(default).

[1] https://lore.kernel.org/lkml/20200405061945.GA94792@iweiny-DESK2.sc.intel.com/

Signed-off-by: Ira Weiny <ira.weiny@intel.com>

---
Changes from v7:
	Change to XFS_MOUNT_DAX_NEVER

Changes from v6:
	Use 2 flag bits rather than a field.
	change iflag to inode

Changes from v5:
	New Patch
---
 fs/xfs/xfs_mount.h |  3 ++-
 fs/xfs/xfs_super.c | 44 ++++++++++++++++++++++++++++++++++++++++----
 2 files changed, 42 insertions(+), 5 deletions(-)

diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index 54bd74088936..2e88c30642e3 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -233,7 +233,8 @@ typedef struct xfs_mount {
 						   allocator */
 #define XFS_MOUNT_NOATTR2	(1ULL << 25)	/* disable use of attr2 format */
 
-#define XFS_MOUNT_DAX_ALWAYS	(1ULL << 62)	/* TEST ONLY! */
+#define XFS_MOUNT_DAX_ALWAYS	(1ULL << 62)
+#define XFS_MOUNT_DAX_NEVER	(1ULL << 63)
 
 /*
  * Max and min values for mount-option defined I/O
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index 3863f41757d2..142e5d03566f 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -47,6 +47,32 @@ static struct kset *xfs_kset;		/* top-level xfs sysfs dir */
 static struct xfs_kobj xfs_dbg_kobj;	/* global debug sysfs attrs */
 #endif
 
+enum {
+	XFS_DAX_INODE = 0,
+	XFS_DAX_ALWAYS = 1,
+	XFS_DAX_NEVER = 2,
+};
+
+static void xfs_mount_set_dax_mode(struct xfs_mount *mp, u32 val)
+{
+	if (val == XFS_DAX_INODE) {
+		mp->m_flags &= ~(XFS_MOUNT_DAX_ALWAYS | XFS_MOUNT_DAX_NEVER);
+	} else if (val == XFS_DAX_ALWAYS) {
+		mp->m_flags |= XFS_MOUNT_DAX_ALWAYS;
+		mp->m_flags &= ~XFS_MOUNT_DAX_NEVER;
+	} else if (val == XFS_DAX_NEVER) {
+		mp->m_flags |= XFS_MOUNT_DAX_NEVER;
+		mp->m_flags &= ~XFS_MOUNT_DAX_ALWAYS;
+	}
+}
+
+static const struct constant_table dax_param_enums[] = {
+	{"inode",	XFS_DAX_INODE },
+	{"always",	XFS_DAX_ALWAYS },
+	{"never",	XFS_DAX_NEVER },
+	{}
+};
+
 /*
  * Table driven mount option parser.
  */
@@ -59,7 +85,7 @@ enum {
 	Opt_filestreams, Opt_quota, Opt_noquota, Opt_usrquota, Opt_grpquota,
 	Opt_prjquota, Opt_uquota, Opt_gquota, Opt_pquota,
 	Opt_uqnoenforce, Opt_gqnoenforce, Opt_pqnoenforce, Opt_qnoenforce,
-	Opt_discard, Opt_nodiscard, Opt_dax,
+	Opt_discard, Opt_nodiscard, Opt_dax, Opt_dax_enum,
 };
 
 static const struct fs_parameter_spec xfs_fs_parameters[] = {
@@ -103,6 +129,7 @@ static const struct fs_parameter_spec xfs_fs_parameters[] = {
 	fsparam_flag("discard",		Opt_discard),
 	fsparam_flag("nodiscard",	Opt_nodiscard),
 	fsparam_flag("dax",		Opt_dax),
+	fsparam_enum("dax",		Opt_dax_enum, dax_param_enums),
 	{}
 };
 
@@ -129,7 +156,6 @@ xfs_fs_show_options(
 		{ XFS_MOUNT_GRPID,		",grpid" },
 		{ XFS_MOUNT_DISCARD,		",discard" },
 		{ XFS_MOUNT_LARGEIO,		",largeio" },
-		{ XFS_MOUNT_DAX_ALWAYS,		",dax" },
 		{ 0, NULL }
 	};
 	struct xfs_mount	*mp = XFS_M(root->d_sb);
@@ -185,6 +211,13 @@ xfs_fs_show_options(
 	if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
 		seq_puts(m, ",noquota");
 
+	if (mp->m_flags & XFS_MOUNT_DAX_ALWAYS)
+		seq_puts(m, ",dax=always");
+	else if (mp->m_flags & XFS_MOUNT_DAX_NEVER)
+		seq_puts(m, ",dax=never");
+	else
+		seq_puts(m, ",dax=inode");
+
 	return 0;
 }
 
@@ -1244,7 +1277,10 @@ xfs_fc_parse_param(
 		return 0;
 #ifdef CONFIG_FS_DAX
 	case Opt_dax:
-		mp->m_flags |= XFS_MOUNT_DAX_ALWAYS;
+		xfs_mount_set_dax_mode(mp, XFS_DAX_ALWAYS);
+		return 0;
+	case Opt_dax_enum:
+		xfs_mount_set_dax_mode(mp, result.uint_32);
 		return 0;
 #endif
 	default:
@@ -1451,7 +1487,7 @@ xfs_fc_fill_super(
 		if (!rtdev_is_dax && !datadev_is_dax) {
 			xfs_alert(mp,
 			"DAX unsupported by block device. Turning off DAX.");
-			mp->m_flags &= ~XFS_MOUNT_DAX_ALWAYS;
+			xfs_mount_set_dax_mode(mp, XFS_DAX_NEVER);
 		}
 		if (xfs_sb_version_hasreflink(&mp->m_sb)) {
 			xfs_alert(mp,
-- 
2.25.1

[PATCH V8 06/11] fs/xfs: Create function xfs_inode_enable_dax()

[ira.weiny]

From: Ira Weiny <ira.weiny@intel.com>

xfs_inode_supports_dax() should reflect if the inode can support DAX not
that it is enabled for DAX.

Change the use of xfs_inode_supports_dax() to reflect only if the inode
and underlying storage support dax.

Add a new function xfs_inode_enable_dax() which reflects if the inode
should be enabled for DAX.

Signed-off-by: Ira Weiny <ira.weiny@intel.com>

---
Changes from v7:
	Move S_ISREG check first
	use IS_ENABLED(CONFIG_FS_DAX) rather than duplicated function

Changes from v6:
	Change enable checks to be sequential logic.
	Update for 2 bit tri-state option.
	Make 'static' consistent.
	Don't set S_DAX if !CONFIG_FS_DAX

Changes from v5:
	Update to reflect the new tri-state mount option

Changes from v3:
	Update functions and names to be more clear
	Update commit message
	Merge with
		'fs/xfs: Clean up DAX support check'
		don't allow IS_DAX() on a directory
		use STATIC macro for static
		make xfs_inode_supports_dax() static
---
 fs/xfs/xfs_iops.c | 28 ++++++++++++++++++++++------
 1 file changed, 22 insertions(+), 6 deletions(-)

diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
index a6e634631da8..2ecc2b2050ab 100644
--- a/fs/xfs/xfs_iops.c
+++ b/fs/xfs/xfs_iops.c
@@ -1243,13 +1243,12 @@ xfs_inode_supports_dax(
 {
 	struct xfs_mount	*mp = ip->i_mount;
 
-	/* Only supported on non-reflinked files. */
-	if (!S_ISREG(VFS_I(ip)->i_mode) || xfs_is_reflink_inode(ip))
+	/* Only supported on regular files. */
+	if (!S_ISREG(VFS_I(ip)->i_mode))
 		return false;
 
-	/* DAX mount option or DAX iflag must be set. */
-	if (!(mp->m_flags & XFS_MOUNT_DAX_ALWAYS) &&
-	    !(ip->i_d.di_flags2 & XFS_DIFLAG2_DAX))
+	/* Only supported on non-reflinked files. */
+	if (xfs_is_reflink_inode(ip))
 		return false;
 
 	/* Block size must match page size */
@@ -1260,6 +1259,23 @@ xfs_inode_supports_dax(
 	return xfs_inode_buftarg(ip)->bt_daxdev != NULL;
 }
 
+static bool
+xfs_inode_enable_dax(
+	struct xfs_inode *ip)
+{
+	if (!IS_ENABLED(CONFIG_FS_DAX))
+		return false;
+	if (ip->i_mount->m_flags & XFS_MOUNT_DAX_NEVER)
+		return false;
+	if (!xfs_inode_supports_dax(ip))
+		return false;
+	if (ip->i_mount->m_flags & XFS_MOUNT_DAX_ALWAYS)
+		return true;
+	if (ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)
+		return true;
+	return false;
+}
+
 STATIC void
 xfs_diflags_to_iflags(
 	struct inode		*inode,
@@ -1278,7 +1294,7 @@ xfs_diflags_to_iflags(
 		inode->i_flags |= S_SYNC;
 	if (flags & XFS_DIFLAG_NOATIME)
 		inode->i_flags |= S_NOATIME;
-	if (xfs_inode_supports_dax(ip))
+	if (xfs_inode_enable_dax(ip))
 		inode->i_flags |= S_DAX;
 }
 
-- 
2.25.1

[PATCH V8 07/11] fs/xfs: Combine xfs_diflags_to_linux() and xfs_diflags_to_iflags()

[ira.weiny]

From: Ira Weiny <ira.weiny@intel.com>

The functionality in xfs_diflags_to_linux() and xfs_diflags_to_iflags() are
nearly identical.  The only difference is that *_to_linux() is called after
inode setup and disallows changing the DAX flag.

Combining them can be done with a flag which indicates if this is the initial
setup to allow the DAX flag to be properly set only at init time.

So remove xfs_diflags_to_linux() and call the modified xfs_diflags_to_iflags()
directly.

While we are here simplify xfs_diflags_to_iflags() to take struct xfs_inode and
use xfs_ip2xflags() to ensure future diflags are included correctly.

Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Ira Weiny <ira.weiny@intel.com>

---
Changes from V7:
	Clarify with a comment the reason for leaving S_DAX out of the
	mask

Changes from V6:
	Move unrelated hunk to previous patch.
	Change logic for better code generation.

Changes from V5:
	The functions are no longer identical so we can only combine
	them rather than deleting one completely.  This is reflected in
	the new init parameter.
---
 fs/xfs/xfs_inode.h |  1 +
 fs/xfs/xfs_ioctl.c | 33 +--------------------------------
 fs/xfs/xfs_iops.c  | 46 +++++++++++++++++++++++++++-------------------
 3 files changed, 29 insertions(+), 51 deletions(-)

diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h
index 492e53992fa9..e76ed9ca17f7 100644
--- a/fs/xfs/xfs_inode.h
+++ b/fs/xfs/xfs_inode.h
@@ -466,6 +466,7 @@ int	xfs_break_layouts(struct inode *inode, uint *iolock,
 /* from xfs_iops.c */
 extern void xfs_setup_inode(struct xfs_inode *ip);
 extern void xfs_setup_iops(struct xfs_inode *ip);
+extern void xfs_diflags_to_iflags(struct xfs_inode *ip, bool init);
 
 /*
  * When setting up a newly allocated inode, we need to call
diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
index d42de92cb283..c6cd92ef4a05 100644
--- a/fs/xfs/xfs_ioctl.c
+++ b/fs/xfs/xfs_ioctl.c
@@ -1100,37 +1100,6 @@ xfs_flags2diflags2(
 	return di_flags2;
 }
 
-STATIC void
-xfs_diflags_to_linux(
-	struct xfs_inode	*ip)
-{
-	struct inode		*inode = VFS_I(ip);
-	unsigned int		xflags = xfs_ip2xflags(ip);
-
-	if (xflags & FS_XFLAG_IMMUTABLE)
-		inode->i_flags |= S_IMMUTABLE;
-	else
-		inode->i_flags &= ~S_IMMUTABLE;
-	if (xflags & FS_XFLAG_APPEND)
-		inode->i_flags |= S_APPEND;
-	else
-		inode->i_flags &= ~S_APPEND;
-	if (xflags & FS_XFLAG_SYNC)
-		inode->i_flags |= S_SYNC;
-	else
-		inode->i_flags &= ~S_SYNC;
-	if (xflags & FS_XFLAG_NOATIME)
-		inode->i_flags |= S_NOATIME;
-	else
-		inode->i_flags &= ~S_NOATIME;
-#if 0	/* disabled until the flag switching races are sorted out */
-	if (xflags & FS_XFLAG_DAX)
-		inode->i_flags |= S_DAX;
-	else
-		inode->i_flags &= ~S_DAX;
-#endif
-}
-
 static int
 xfs_ioctl_setattr_xflags(
 	struct xfs_trans	*tp,
@@ -1168,7 +1137,7 @@ xfs_ioctl_setattr_xflags(
 	ip->i_d.di_flags = xfs_flags2diflags(ip, fa->fsx_xflags);
 	ip->i_d.di_flags2 = di_flags2;
 
-	xfs_diflags_to_linux(ip);
+	xfs_diflags_to_iflags(ip, false);
 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
 	XFS_STATS_INC(mp, xs_ig_attrchg);
diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
index 2ecc2b2050ab..d0c83e0fbf46 100644
--- a/fs/xfs/xfs_iops.c
+++ b/fs/xfs/xfs_iops.c
@@ -1276,26 +1276,34 @@ xfs_inode_enable_dax(
 	return false;
 }
 
-STATIC void
+void
 xfs_diflags_to_iflags(
-	struct inode		*inode,
-	struct xfs_inode	*ip)
+	struct xfs_inode	*ip,
+	bool init)
 {
-	uint16_t		flags = ip->i_d.di_flags;
-
-	inode->i_flags &= ~(S_IMMUTABLE | S_APPEND | S_SYNC |
-			    S_NOATIME | S_DAX);
-
-	if (flags & XFS_DIFLAG_IMMUTABLE)
-		inode->i_flags |= S_IMMUTABLE;
-	if (flags & XFS_DIFLAG_APPEND)
-		inode->i_flags |= S_APPEND;
-	if (flags & XFS_DIFLAG_SYNC)
-		inode->i_flags |= S_SYNC;
-	if (flags & XFS_DIFLAG_NOATIME)
-		inode->i_flags |= S_NOATIME;
-	if (xfs_inode_enable_dax(ip))
-		inode->i_flags |= S_DAX;
+	struct inode            *inode = VFS_I(ip);
+	unsigned int            xflags = xfs_ip2xflags(ip);
+	unsigned int            flags = 0;
+
+	ASSERT(!(IS_DAX(inode) && init));
+
+	if (xflags & FS_XFLAG_IMMUTABLE)
+		flags |= S_IMMUTABLE;
+	if (xflags & FS_XFLAG_APPEND)
+		flags |= S_APPEND;
+	if (xflags & FS_XFLAG_SYNC)
+		flags |= S_SYNC;
+	if (xflags & FS_XFLAG_NOATIME)
+		flags |= S_NOATIME;
+	if (init && xfs_inode_enable_dax(ip))
+		flags |= S_DAX;
+
+	/*
+	 * S_DAX can only be set during inode initialization and is never set by
+	 * the VFS, so we cannot mask off S_DAX in i_flags.
+	 */
+	inode->i_flags &= ~(S_IMMUTABLE | S_APPEND | S_SYNC | S_NOATIME);
+	inode->i_flags |= flags;
 }
 
 /*
@@ -1324,7 +1332,7 @@ xfs_setup_inode(
 	inode->i_gid    = xfs_gid_to_kgid(ip->i_d.di_gid);
 
 	i_size_write(inode, ip->i_d.di_size);
-	xfs_diflags_to_iflags(inode, ip);
+	xfs_diflags_to_iflags(ip, true);
 
 	if (S_ISDIR(inode->i_mode)) {
 		/*
-- 
2.25.1

[PATCH V8 08/11] fs: Define I_DONTCACNE in VFS layer

[ira.weiny]

From: Ira Weiny <ira.weiny@intel.com>

DAX effective mode changes (setting of S_DAX) require inode eviction.

Define a flag which can be set to inform the VFS layer that inodes
should not be cached.  This will expedite the eviction of those nodes
requiring reload.

Signed-off-by: Ira Weiny <ira.weiny@intel.com>
---
 include/linux/fs.h | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/include/linux/fs.h b/include/linux/fs.h
index a818ced22961..e2db71d150c3 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -2151,6 +2151,8 @@ static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
  *
  * I_CREATING		New object's inode in the middle of setting up.
  *
+ * I_DONTCACHE		Do not cache the inode
+ *
  * Q: What is the difference between I_WILL_FREE and I_FREEING?
  */
 #define I_DIRTY_SYNC		(1 << 0)
@@ -2173,6 +2175,7 @@ static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
 #define I_WB_SWITCH		(1 << 13)
 #define I_OVL_INUSE		(1 << 14)
 #define I_CREATING		(1 << 15)
+#define I_DONTCACHE		(1 << 16)
 
 #define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC)
 #define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES)
@@ -3042,7 +3045,8 @@ extern int inode_needs_sync(struct inode *inode);
 extern int generic_delete_inode(struct inode *inode);
 static inline int generic_drop_inode(struct inode *inode)
 {
-	return !inode->i_nlink || inode_unhashed(inode);
+	return !inode->i_nlink || inode_unhashed(inode) ||
+		(inode->i_state & I_DONTCACHE);
 }
 
 extern struct inode *ilookup5_nowait(struct super_block *sb,
-- 
2.25.1

[PATCH V8 09/11] fs: Introduce DCACHE_DONTCACHE

[ira.weiny]

From: Ira Weiny <ira.weiny@intel.com>

DCACHE_DONTCACHE indicates a dentry should not be cached on final
dput().

Also add a helper function which will flag I_DONTCACHE as well ad
DCACHE_DONTCACHE on all dentries point to a specified inode.

Signed-off-by: Ira Weiny <ira.weiny@intel.com>

---
Changes from V7:
	new patch
---
 fs/dcache.c            |  4 ++++
 fs/inode.c             | 15 +++++++++++++++
 include/linux/dcache.h |  2 ++
 include/linux/fs.h     |  1 +
 4 files changed, 22 insertions(+)

diff --git a/fs/dcache.c b/fs/dcache.c
index b280e07e162b..0030fabab2c4 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -647,6 +647,10 @@ static inline bool retain_dentry(struct dentry *dentry)
 		if (dentry->d_op->d_delete(dentry))
 			return false;
 	}
+
+	if (unlikely(dentry->d_flags & DCACHE_DONTCACHE))
+		return false;
+
 	/* retain; LRU fodder */
 	dentry->d_lockref.count--;
 	if (unlikely(!(dentry->d_flags & DCACHE_LRU_LIST)))
diff --git a/fs/inode.c b/fs/inode.c
index 93d9252a00ab..b8b1917a324e 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -1526,6 +1526,21 @@ int generic_delete_inode(struct inode *inode)
 }
 EXPORT_SYMBOL(generic_delete_inode);
 
+void flag_inode_dontcache(struct inode *inode)
+{
+	struct dentry *dent;
+
+	rcu_read_lock();
+	hlist_for_each_entry(dent, &inode->i_dentry, d_u.d_alias) {
+		spin_lock(&dent->d_lock);
+		dent->d_flags |= DCACHE_DONTCACHE;
+		spin_unlock(&dent->d_lock);
+	}
+	rcu_read_unlock();
+	inode->i_state |= I_DONTCACHE;
+}
+EXPORT_SYMBOL(flag_inode_dontcache);
+
 /*
  * Called when we're dropping the last reference
  * to an inode.
diff --git a/include/linux/dcache.h b/include/linux/dcache.h
index c1488cc84fd9..56b1482d9223 100644
--- a/include/linux/dcache.h
+++ b/include/linux/dcache.h
@@ -177,6 +177,8 @@ struct dentry_operations {
 
 #define DCACHE_REFERENCED		0x00000040 /* Recently used, don't discard. */
 
+#define DCACHE_DONTCACHE		0x00000080 /* don't cache on final dput() */
+
 #define DCACHE_CANT_MOUNT		0x00000100
 #define DCACHE_GENOCIDE			0x00000200
 #define DCACHE_SHRINK_LIST		0x00000400
diff --git a/include/linux/fs.h b/include/linux/fs.h
index e2db71d150c3..f2916c99616a 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -3048,6 +3048,7 @@ static inline int generic_drop_inode(struct inode *inode)
 	return !inode->i_nlink || inode_unhashed(inode) ||
 		(inode->i_state & I_DONTCACHE);
 }
+extern void flag_inode_dontcache(struct inode *inode);
 
 extern struct inode *ilookup5_nowait(struct super_block *sb,
 		unsigned long hashval, int (*test)(struct inode *, void *),
-- 
2.25.1

[PATCH V8 10/11] fs/xfs: Change xfs_ioctl_setattr_dax_invalidate()

[ira.weiny]

From: Ira Weiny <ira.weiny@intel.com>

We only support changing FS_XFLAG_DAX on directories.  Files get their
flag from the parent directory on creation only.  So no data
invalidation needs to happen.

Alter the xfs_ioctl_setattr_dax_invalidate() to be
xfs_ioctl_setattr_dax_validate().  xfs_ioctl_setattr_dax_validate() now
validates that any FS_XFLAG_DAX change is ok.

This also allows use to remove the join_flags logic.

Signed-off-by: Ira Weiny <ira.weiny@intel.com>

---
Changes from V7:
	Use new flag_inode_dontcache()
	Skip don't cache if mount over ride is active.

Changes from v6:
	Fix completely broken implementation and update commit message.
	Use the new VFS layer I_DONTCACHE to facilitate inode eviction
	and S_DAX changing on drop_caches

Changes from v5:
	New patch
---
 fs/xfs/xfs_ioctl.c | 105 +++++++++------------------------------------
 1 file changed, 20 insertions(+), 85 deletions(-)

diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
index c6cd92ef4a05..75d4a830ef38 100644
--- a/fs/xfs/xfs_ioctl.c
+++ b/fs/xfs/xfs_ioctl.c
@@ -1145,63 +1145,28 @@ xfs_ioctl_setattr_xflags(
 }
 
 /*
- * If we are changing DAX flags, we have to ensure the file is clean and any
- * cached objects in the address space are invalidated and removed. This
- * requires us to lock out other IO and page faults similar to a truncate
- * operation. The locks need to be held until the transaction has been committed
- * so that the cache invalidation is atomic with respect to the DAX flag
- * manipulation.
+ * Mark inodes with a changing FS_XFLAG_DAX, I_DONTCACHE
  */
-static int
+static void
 xfs_ioctl_setattr_dax_invalidate(
 	struct xfs_inode	*ip,
-	struct fsxattr		*fa,
-	int			*join_flags)
+	struct fsxattr		*fa)
 {
-	struct inode		*inode = VFS_I(ip);
-	struct super_block	*sb = inode->i_sb;
-	int			error;
-
-	*join_flags = 0;
-
-	/*
-	 * It is only valid to set the DAX flag on regular files and
-	 * directories on filesystems where the block size is equal to the page
-	 * size. On directories it serves as an inherited hint so we don't
-	 * have to check the device for dax support or flush pagecache.
-	 */
-	if (fa->fsx_xflags & FS_XFLAG_DAX) {
-		struct xfs_buftarg	*target = xfs_inode_buftarg(ip);
-
-		if (!bdev_dax_supported(target->bt_bdev, sb->s_blocksize))
-			return -EINVAL;
-	}
-
-	/* If the DAX state is not changing, we have nothing to do here. */
-	if ((fa->fsx_xflags & FS_XFLAG_DAX) && IS_DAX(inode))
-		return 0;
-	if (!(fa->fsx_xflags & FS_XFLAG_DAX) && !IS_DAX(inode))
-		return 0;
+	struct xfs_mount	*mp = ip->i_mount;
+	struct inode            *inode = VFS_I(ip);
 
 	if (S_ISDIR(inode->i_mode))
-		return 0;
-
-	/* lock, flush and invalidate mapping in preparation for flag change */
-	xfs_ilock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
-	error = filemap_write_and_wait(inode->i_mapping);
-	if (error)
-		goto out_unlock;
-	error = invalidate_inode_pages2(inode->i_mapping);
-	if (error)
-		goto out_unlock;
+		return;
 
-	*join_flags = XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL;
-	return 0;
-
-out_unlock:
-	xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
-	return error;
+	if (mp->m_flags & XFS_MOUNT_DAX_ALWAYS ||
+	    mp->m_flags & XFS_MOUNT_DAX_NEVER)
+		return;
 
+	if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
+	    !(ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)) ||
+	    (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
+	     (ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)))
+		flag_inode_dontcache(inode);
 }
 
 /*
@@ -1209,17 +1174,10 @@ xfs_ioctl_setattr_dax_invalidate(
  * have permission to do so. On success, return a clean transaction and the
  * inode locked exclusively ready for further operation specific checks. On
  * failure, return an error without modifying or locking the inode.
- *
- * The inode might already be IO locked on call. If this is the case, it is
- * indicated in @join_flags and we take full responsibility for ensuring they
- * are unlocked from now on. Hence if we have an error here, we still have to
- * unlock them. Otherwise, once they are joined to the transaction, they will
- * be unlocked on commit/cancel.
  */
 static struct xfs_trans *
 xfs_ioctl_setattr_get_trans(
-	struct xfs_inode	*ip,
-	int			join_flags)
+	struct xfs_inode	*ip)
 {
 	struct xfs_mount	*mp = ip->i_mount;
 	struct xfs_trans	*tp;
@@ -1236,8 +1194,7 @@ xfs_ioctl_setattr_get_trans(
 		goto out_unlock;
 
 	xfs_ilock(ip, XFS_ILOCK_EXCL);
-	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | join_flags);
-	join_flags = 0;
+	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
 
 	/*
 	 * CAP_FOWNER overrides the following restrictions:
@@ -1258,8 +1215,6 @@ xfs_ioctl_setattr_get_trans(
 out_cancel:
 	xfs_trans_cancel(tp);
 out_unlock:
-	if (join_flags)
-		xfs_iunlock(ip, join_flags);
 	return ERR_PTR(error);
 }
 
@@ -1386,7 +1341,6 @@ xfs_ioctl_setattr(
 	struct xfs_dquot	*pdqp = NULL;
 	struct xfs_dquot	*olddquot = NULL;
 	int			code;
-	int			join_flags = 0;
 
 	trace_xfs_ioctl_setattr(ip);
 
@@ -1410,18 +1364,9 @@ xfs_ioctl_setattr(
 			return code;
 	}
 
-	/*
-	 * Changing DAX config may require inode locking for mapping
-	 * invalidation. These need to be held all the way to transaction commit
-	 * or cancel time, so need to be passed through to
-	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
-	 * appropriately.
-	 */
-	code = xfs_ioctl_setattr_dax_invalidate(ip, fa, &join_flags);
-	if (code)
-		goto error_free_dquots;
+	xfs_ioctl_setattr_dax_invalidate(ip, fa);
 
-	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
+	tp = xfs_ioctl_setattr_get_trans(ip);
 	if (IS_ERR(tp)) {
 		code = PTR_ERR(tp);
 		goto error_free_dquots;
@@ -1552,7 +1497,6 @@ xfs_ioc_setxflags(
 	struct fsxattr		fa;
 	struct fsxattr		old_fa;
 	unsigned int		flags;
-	int			join_flags = 0;
 	int			error;
 
 	if (copy_from_user(&flags, arg, sizeof(flags)))
@@ -1569,18 +1513,9 @@ xfs_ioc_setxflags(
 	if (error)
 		return error;
 
-	/*
-	 * Changing DAX config may require inode locking for mapping
-	 * invalidation. These need to be held all the way to transaction commit
-	 * or cancel time, so need to be passed through to
-	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
-	 * appropriately.
-	 */
-	error = xfs_ioctl_setattr_dax_invalidate(ip, &fa, &join_flags);
-	if (error)
-		goto out_drop_write;
+	xfs_ioctl_setattr_dax_invalidate(ip, &fa);
 
-	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
+	tp = xfs_ioctl_setattr_get_trans(ip);
 	if (IS_ERR(tp)) {
 		error = PTR_ERR(tp);
 		goto out_drop_write;
-- 
2.25.1

[PATCH V8 11/11] Documentation/dax: Update Usage section

[ira.weiny]

From: Ira Weiny <ira.weiny@intel.com>

Update the Usage section to reflect the new individual dax selection
functionality.

Signed-off-by: Ira Weiny <ira.weiny@intel.com>

---
Changes from V7:
	Cleanups/clarifications from Darrick and Dan

Changes from V6:
	Update to allow setting FS_XFLAG_DAX any time.
	Update with list of behaviors from Darrick
	https://lore.kernel.org/lkml/20200409165927.GD6741@magnolia/

Changes from V5:
	Update to reflect the agreed upon semantics
	https://lore.kernel.org/lkml/20200405061945.GA94792@iweiny-DESK2.sc.intel.com/
---
 Documentation/filesystems/dax.txt | 166 +++++++++++++++++++++++++++++-
 1 file changed, 163 insertions(+), 3 deletions(-)

diff --git a/Documentation/filesystems/dax.txt b/Documentation/filesystems/dax.txt
index 679729442fd2..893820c53f49 100644
--- a/Documentation/filesystems/dax.txt
+++ b/Documentation/filesystems/dax.txt
@@ -17,11 +17,171 @@ For file mappings, the storage device is mapped directly into userspace.
 Usage
 -----
 
-If you have a block device which supports DAX, you can make a filesystem
+If you have a block device which supports DAX, you can make a file system
 on it as usual.  The DAX code currently only supports files with a block
 size equal to your kernel's PAGE_SIZE, so you may need to specify a block
-size when creating the filesystem.  When mounting it, use the "-o dax"
-option on the command line or add 'dax' to the options in /etc/fstab.
+size when creating the file system.
+
+Currently 3 filesystems support DAX, ext2, ext4 and xfs.  Enabling DAX on them
+is different.
+
+Enabling DAX on ext4 and ext2
+-----------------------------
+
+When mounting the filesystem, use the "-o dax" option on the command line or
+add 'dax' to the options in /etc/fstab.  This works to enable DAX on all files
+within the filesystem.  It is equivalent to the '-o dax=always' behavior below
+with the exception that the STATX_ATTR_DAX flag is not supported, nor needed,
+as it is always true.
+
+
+Enabling DAX on xfs
+-------------------
+
+Summary
+-------
+
+ 1. There exists an in-kernel file access mode flag S_DAX that corresponds to
+    the statx flag STATX_ATTR_DAX.  See the manpage for statx(2) for details
+    about this access mode.
+
+ 2. There exists an advisory file inode flag FS_XFLAG_DAX that is
+    inherited from the parent directory FS_XFLAG_DAX inode flag at file
+    creation time.  This advisory flag can be set or cleared at any
+    time, but doing so does not immediately affect the S_DAX state.
+
+    Unless overridden by mount options (see (3)), if FS_XFLAG_DAX is set
+    and the fs is on pmem then it will enable S_DAX at inode load time;
+    if FS_XFLAG_DAX is not set, it will not enable S_DAX.
+
+ 3. There exists a dax= mount option.
+
+    "-o dax=never"  means "never set S_DAX, ignore FS_XFLAG_DAX."
+
+    "-o dax=always" means "always set S_DAX (at least on pmem),
+                    and ignore FS_XFLAG_DAX."
+
+    "-o dax"        is an alias for "dax=always".
+
+    "-o dax=inode"  means "follow FS_XFLAG_DAX" and is the default.
+
+ 4. There exists an advisory directory inode flag FS_XFLAG_DAX that can
+    be set or cleared at any time.  The flag state is inherited by any files or
+    subdirectories when they are created within that directory.
+
+ 5. Programs that require a specific file access mode (DAX or not DAX)
+    can do one of the following:
+
+    (a) Create files in directories that the FS_XFLAG_DAX flag set as
+        needed; or
+
+    (b) Have the administrator set an override via mount option; or
+
+    (c) Set or clear the file's FS_XFLAG_DAX flag as needed.  Programs
+        must then cause the kernel to evict the inode from memory.  This
+        can be done by:
+
+        i>  Closing the file and re-opening the file and using statx to
+            see if the fs has changed the S_DAX flag; and
+
+        ii> If the file still does not have the desired S_DAX access
+            mode, either unmount and remount the filesystem, or close
+            the file and use drop_caches.
+
+ 6. It is expected that users who want to squeeze every last bit of performance
+    out of the particular rough and tumble bits of their storage will also be
+    exposed to the difficulties of what happens when the operating system can't
+    totally virtualize those hardware capabilities.  DAX is such a feature.
+
+
+Details
+-------
+
+There are 2 per-file dax flags.  One is a physical inode setting (FS_XFLAG_DAX)
+and the other a currently enabled state (S_DAX).
+
+FS_XFLAG_DAX is maintained, on disk, on individual inodes.  It is preserved
+within the file system.  This 'physical' config setting can be set using an
+ioctl and/or an application such as "xfs_io -c 'chattr [-+]x'".  Files and
+directories automatically inherit FS_XFLAG_DAX from their parent directory
+_when_ _created_.  Therefore, setting FS_XFLAG_DAX at directory creation time
+can be used to set a default behavior for an entire sub-tree.  (Doing so on the
+root directory acts to set a default for the entire file system.)
+
+To clarify inheritance here are 3 examples:
+
+Example A:
+
+mkdir -p a/b/c
+xfs_io 'chattr +x' a
+mkdir a/b/c/d
+mkdir a/e
+
+	dax: a,e
+	no dax: b,c,d
+
+Example B:
+
+mkdir a
+xfs_io 'chattr +x' a
+mkdir -p a/b/c/d
+
+	dax: a,b,c,d
+	no dax:
+
+Example C:
+
+mkdir -p a/b/c
+xfs_io 'chattr +x' c
+mkdir a/b/c/d
+
+	dax: c,d
+	no dax: a,b
+
+
+The current enabled state (S_DAX) is set when a file inode is _loaded_ based on
+the underlying media support, the value of FS_XFLAG_DAX, and the file systems
+dax mount option setting.  See below.
+
+statx can be used to query S_DAX.  NOTE that a directory will never have S_DAX
+set and therefore statx will never indicate that S_DAX is set on directories.
+
+NOTE: Setting the FS_XFLAG_DAX (specifically or through inheritance) occurs
+even if the underlying media does not support dax and/or the file system is
+overridden with a mount option.
+
+
+Overriding FS_XFLAG_DAX (dax= mount option)
+-------------------------------------------
+
+There exists a dax mount option.  Using the mount option does not change the
+physical configured state of individual files but overrides the S_DAX operating
+state when inodes are loaded.
+
+Given underlying media support, the dax mount option is a tri-state option
+(never, always, inode) with the following meanings:
+
+   "-o dax=never" means "never set S_DAX, ignore FS_XFLAG_DAX"
+   "-o dax=always" means "always set S_DAX, ignore FS_XFLAG_DAX"
+        "-o dax" by itself means "dax=always" to remain compatible with older
+	         kernels
+   "-o dax=inode" means "follow FS_XFLAG_DAX"
+
+The default state is 'inode'.  Given underlying media support, the following
+algorithm is used to determine the effective mode of the file S_DAX on a
+capable device.
+
+	S_DAX = FS_XFLAG_DAX;
+
+	if (dax_mount == "always")
+		S_DAX = true;
+	else if (dax_mount == "off"
+		S_DAX = false;
+
+To reiterate: Setting, and inheritance, continues to affect FS_XFLAG_DAX even
+while the file system is mounted with a dax override.  However, in-core inode
+state (S_DAX) will continue to be overridden until the filesystem is remounted
+with dax=inode and the inode is evicted."
 
 
 Implementation Tips for Block Driver Writers
-- 
2.25.1

Re: [PATCH V8 08/11] fs: Define I_DONTCACNE in VFS layer

[Jan Kara]

There's a typo in the subject - I_DONTCACNE.

On Tue 14-04-20 23:45:20, ira.weiny@intel.com wrote:
> From: Ira Weiny <ira.weiny@intel.com>
> 
> DAX effective mode changes (setting of S_DAX) require inode eviction.
> 
> Define a flag which can be set to inform the VFS layer that inodes
> should not be cached.  This will expedite the eviction of those nodes
> requiring reload.
> 
> Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> ---
>  include/linux/fs.h | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index a818ced22961..e2db71d150c3 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -2151,6 +2151,8 @@ static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
>   *
>   * I_CREATING		New object's inode in the middle of setting up.
>   *
> + * I_DONTCACHE		Do not cache the inode
> + *

Maybe, I'd be more specific here and write: "Evict inode as soon as it is
not used anymore"?

Otherwise the patch looks good to me so feel free to add:

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

Also it would be good to CC Al Viro on this one (and the dentry flag) I
guess.

								Honza

>   * Q: What is the difference between I_WILL_FREE and I_FREEING?
>   */
>  #define I_DIRTY_SYNC		(1 << 0)
> @@ -2173,6 +2175,7 @@ static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
>  #define I_WB_SWITCH		(1 << 13)
>  #define I_OVL_INUSE		(1 << 14)
>  #define I_CREATING		(1 << 15)
> +#define I_DONTCACHE		(1 << 16)
>  
>  #define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC)
>  #define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES)
> @@ -3042,7 +3045,8 @@ extern int inode_needs_sync(struct inode *inode);
>  extern int generic_delete_inode(struct inode *inode);
>  static inline int generic_drop_inode(struct inode *inode)
>  {
> -	return !inode->i_nlink || inode_unhashed(inode);
> +	return !inode->i_nlink || inode_unhashed(inode) ||
> +		(inode->i_state & I_DONTCACHE);
>  }
>  
>  extern struct inode *ilookup5_nowait(struct super_block *sb,
> -- 
> 2.25.1
> 
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

Re: [PATCH V8 09/11] fs: Introduce DCACHE_DONTCACHE

[Jan Kara]

On Tue 14-04-20 23:45:21, ira.weiny@intel.com wrote:
> From: Ira Weiny <ira.weiny@intel.com>
> 
> DCACHE_DONTCACHE indicates a dentry should not be cached on final
> dput().
> 
> Also add a helper function which will flag I_DONTCACHE as well ad
> DCACHE_DONTCACHE on all dentries point to a specified inode.

I think this sentence needs more work :). Like: Also add a helper function
which will mark the inode with I_DONTCACHE flag and also mark all dentries
pointing to a specified inode as DCACHE_DONTCACHE.

> 
> Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> 
> ---
> Changes from V7:
> 	new patch
> ---
>  fs/dcache.c            |  4 ++++
>  fs/inode.c             | 15 +++++++++++++++
>  include/linux/dcache.h |  2 ++
>  include/linux/fs.h     |  1 +
>  4 files changed, 22 insertions(+)

...

> diff --git a/fs/inode.c b/fs/inode.c
> index 93d9252a00ab..b8b1917a324e 100644
> --- a/fs/inode.c
> +++ b/fs/inode.c
> @@ -1526,6 +1526,21 @@ int generic_delete_inode(struct inode *inode)
>  }
>  EXPORT_SYMBOL(generic_delete_inode);
>  
> +void flag_inode_dontcache(struct inode *inode)

mark_inode_dontcache?

> +{
> +	struct dentry *dent;

This is really nitpicking but dentry variables are usually called 'de' or
'dentry' :)

> +
> +	rcu_read_lock();

I don't think this list is safe to traverse under RCU. E.g.
dentry_unlink_inode() does hlist_del_init(&dentry->d_u.d_alias). Usually,
we traverse this list under inode->i_lock protection AFAICS.

								Honza

> +	hlist_for_each_entry(dent, &inode->i_dentry, d_u.d_alias) {
> +		spin_lock(&dent->d_lock);
> +		dent->d_flags |= DCACHE_DONTCACHE;
> +		spin_unlock(&dent->d_lock);
> +	}
> +	rcu_read_unlock();
> +	inode->i_state |= I_DONTCACHE;
> +}
> +EXPORT_SYMBOL(flag_inode_dontcache);
> +
>  /*
>   * Called when we're dropping the last reference
>   * to an inode.
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

Re: [PATCH V8 04/11] fs/xfs: Change XFS_MOUNT_DAX to XFS_MOUNT_DAX_ALWAYS

[Darrick J. Wong]

On Tue, Apr 14, 2020 at 11:45:16PM -0700, ira.weiny@intel.com wrote:
> From: Ira Weiny <ira.weiny@intel.com>
> 
> In prep for the new tri-state mount option which then introduces
> XFS_MOUNT_DAX_NEVER.
> 
> Signed-off-by: Ira Weiny <ira.weiny@intel.com>

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

--D

> ---
>  fs/xfs/xfs_iops.c  | 2 +-
>  fs/xfs/xfs_mount.h | 2 +-
>  fs/xfs/xfs_super.c | 8 ++++----
>  3 files changed, 6 insertions(+), 6 deletions(-)
> 
> diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
> index 81f2f93caec0..a6e634631da8 100644
> --- a/fs/xfs/xfs_iops.c
> +++ b/fs/xfs/xfs_iops.c
> @@ -1248,7 +1248,7 @@ xfs_inode_supports_dax(
>  		return false;
>  
>  	/* DAX mount option or DAX iflag must be set. */
> -	if (!(mp->m_flags & XFS_MOUNT_DAX) &&
> +	if (!(mp->m_flags & XFS_MOUNT_DAX_ALWAYS) &&
>  	    !(ip->i_d.di_flags2 & XFS_DIFLAG2_DAX))
>  		return false;
>  
> diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
> index 88ab09ed29e7..54bd74088936 100644
> --- a/fs/xfs/xfs_mount.h
> +++ b/fs/xfs/xfs_mount.h
> @@ -233,7 +233,7 @@ typedef struct xfs_mount {
>  						   allocator */
>  #define XFS_MOUNT_NOATTR2	(1ULL << 25)	/* disable use of attr2 format */
>  
> -#define XFS_MOUNT_DAX		(1ULL << 62)	/* TEST ONLY! */
> +#define XFS_MOUNT_DAX_ALWAYS	(1ULL << 62)	/* TEST ONLY! */
>  
>  /*
>   * Max and min values for mount-option defined I/O
> diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
> index 2094386af8ac..3863f41757d2 100644
> --- a/fs/xfs/xfs_super.c
> +++ b/fs/xfs/xfs_super.c
> @@ -129,7 +129,7 @@ xfs_fs_show_options(
>  		{ XFS_MOUNT_GRPID,		",grpid" },
>  		{ XFS_MOUNT_DISCARD,		",discard" },
>  		{ XFS_MOUNT_LARGEIO,		",largeio" },
> -		{ XFS_MOUNT_DAX,		",dax" },
> +		{ XFS_MOUNT_DAX_ALWAYS,		",dax" },
>  		{ 0, NULL }
>  	};
>  	struct xfs_mount	*mp = XFS_M(root->d_sb);
> @@ -1244,7 +1244,7 @@ xfs_fc_parse_param(
>  		return 0;
>  #ifdef CONFIG_FS_DAX
>  	case Opt_dax:
> -		mp->m_flags |= XFS_MOUNT_DAX;
> +		mp->m_flags |= XFS_MOUNT_DAX_ALWAYS;
>  		return 0;
>  #endif
>  	default:
> @@ -1437,7 +1437,7 @@ xfs_fc_fill_super(
>  	if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5)
>  		sb->s_flags |= SB_I_VERSION;
>  
> -	if (mp->m_flags & XFS_MOUNT_DAX) {
> +	if (mp->m_flags & XFS_MOUNT_DAX_ALWAYS) {
>  		bool rtdev_is_dax = false, datadev_is_dax;
>  
>  		xfs_warn(mp,
> @@ -1451,7 +1451,7 @@ xfs_fc_fill_super(
>  		if (!rtdev_is_dax && !datadev_is_dax) {
>  			xfs_alert(mp,
>  			"DAX unsupported by block device. Turning off DAX.");
> -			mp->m_flags &= ~XFS_MOUNT_DAX;
> +			mp->m_flags &= ~XFS_MOUNT_DAX_ALWAYS;
>  		}
>  		if (xfs_sb_version_hasreflink(&mp->m_sb)) {
>  			xfs_alert(mp,
> -- 
> 2.25.1
> 

Re: [PATCH V8 01/11] fs/xfs: Remove unnecessary initialization of i_rwsem

[Darrick J. Wong]

On Tue, Apr 14, 2020 at 11:45:13PM -0700, ira.weiny@intel.com wrote:
> From: Ira Weiny <ira.weiny@intel.com>
> 
> An earlier call of xfs_reinit_inode() from xfs_iget_cache_hit() already
> handles initialization of i_rwsem.
> 
> Doing so again is unneeded.
> 
> Reviewed-by: Dave Chinner <dchinner@redhat.com>
> Signed-off-by: Ira Weiny <ira.weiny@intel.com>

Still looks ok,
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>

--D

> 
> ---
> Changes from V4:
> 	Update commit message to make it clear the xfs_iget_cache_hit()
> 	is actually doing the initialization via xfs_reinit_inode()
> 
> New for V4:
> 
> NOTE: This was found while ensuring the new i_aops_sem was properly
> handled.  It seems like this is a layering violation so I think it is
> worth cleaning up so as to not confuse others.
> ---
>  fs/xfs/xfs_icache.c | 4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
> 
> diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c
> index 8dc2e5414276..836a1f09be03 100644
> --- a/fs/xfs/xfs_icache.c
> +++ b/fs/xfs/xfs_icache.c
> @@ -419,6 +419,7 @@ xfs_iget_cache_hit(
>  		spin_unlock(&ip->i_flags_lock);
>  		rcu_read_unlock();
>  
> +		ASSERT(!rwsem_is_locked(&inode->i_rwsem));
>  		error = xfs_reinit_inode(mp, inode);
>  		if (error) {
>  			bool wake;
> @@ -452,9 +453,6 @@ xfs_iget_cache_hit(
>  		ip->i_sick = 0;
>  		ip->i_checked = 0;
>  
> -		ASSERT(!rwsem_is_locked(&inode->i_rwsem));
> -		init_rwsem(&inode->i_rwsem);
> -
>  		spin_unlock(&ip->i_flags_lock);
>  		spin_unlock(&pag->pag_ici_lock);
>  	} else {
> -- 
> 2.25.1
> 

Re: [PATCH V8 05/11] fs/xfs: Make DAX mount option a tri-state

[Darrick J. Wong]

On Tue, Apr 14, 2020 at 11:45:17PM -0700, ira.weiny@intel.com wrote:
> From: Ira Weiny <ira.weiny@intel.com>
> 
> As agreed upon[1].  We make the dax mount option a tri-state.  '-o dax'
> continues to operate the same.  We add 'always', 'never', and 'inode'
> (default).
> 
> [1] https://lore.kernel.org/lkml/20200405061945.GA94792@iweiny-DESK2.sc.intel.com/
> 
> Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> 
> ---
> Changes from v7:
> 	Change to XFS_MOUNT_DAX_NEVER
> 
> Changes from v6:
> 	Use 2 flag bits rather than a field.
> 	change iflag to inode
> 
> Changes from v5:
> 	New Patch
> ---
>  fs/xfs/xfs_mount.h |  3 ++-
>  fs/xfs/xfs_super.c | 44 ++++++++++++++++++++++++++++++++++++++++----
>  2 files changed, 42 insertions(+), 5 deletions(-)
> 
> diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
> index 54bd74088936..2e88c30642e3 100644
> --- a/fs/xfs/xfs_mount.h
> +++ b/fs/xfs/xfs_mount.h
> @@ -233,7 +233,8 @@ typedef struct xfs_mount {
>  						   allocator */
>  #define XFS_MOUNT_NOATTR2	(1ULL << 25)	/* disable use of attr2 format */
>  
> -#define XFS_MOUNT_DAX_ALWAYS	(1ULL << 62)	/* TEST ONLY! */
> +#define XFS_MOUNT_DAX_ALWAYS	(1ULL << 62)
> +#define XFS_MOUNT_DAX_NEVER	(1ULL << 63)
>  
>  /*
>   * Max and min values for mount-option defined I/O
> diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
> index 3863f41757d2..142e5d03566f 100644
> --- a/fs/xfs/xfs_super.c
> +++ b/fs/xfs/xfs_super.c
> @@ -47,6 +47,32 @@ static struct kset *xfs_kset;		/* top-level xfs sysfs dir */
>  static struct xfs_kobj xfs_dbg_kobj;	/* global debug sysfs attrs */
>  #endif
>  
> +enum {

enum xfs_dax_mode {  for the reasons given below?

> +	XFS_DAX_INODE = 0,
> +	XFS_DAX_ALWAYS = 1,
> +	XFS_DAX_NEVER = 2,
> +};
> +
> +static void xfs_mount_set_dax_mode(struct xfs_mount *mp, u32 val)

xfs style, please:

static void
xfs_mount_set_dax_mode(
	struct xfs_mount	*mp,
	u32			val)

or if you give a name to the enum above, you can enforce some type
safety too:

	enum xfs_dax_mode	val)

> +{
> +	if (val == XFS_DAX_INODE) {

and this probably could have been a "switch (val) {", in which case if
the enum ever gets expanded then gcc will whine about missing switch
cases.

The rest of the patch looks good.

--D

> +		mp->m_flags &= ~(XFS_MOUNT_DAX_ALWAYS | XFS_MOUNT_DAX_NEVER);
> +	} else if (val == XFS_DAX_ALWAYS) {
> +		mp->m_flags |= XFS_MOUNT_DAX_ALWAYS;
> +		mp->m_flags &= ~XFS_MOUNT_DAX_NEVER;
> +	} else if (val == XFS_DAX_NEVER) {
> +		mp->m_flags |= XFS_MOUNT_DAX_NEVER;
> +		mp->m_flags &= ~XFS_MOUNT_DAX_ALWAYS;
> +	}
> +}
> +
> +static const struct constant_table dax_param_enums[] = {
> +	{"inode",	XFS_DAX_INODE },
> +	{"always",	XFS_DAX_ALWAYS },
> +	{"never",	XFS_DAX_NEVER },
> +	{}
> +};
> +
>  /*
>   * Table driven mount option parser.
>   */
> @@ -59,7 +85,7 @@ enum {
>  	Opt_filestreams, Opt_quota, Opt_noquota, Opt_usrquota, Opt_grpquota,
>  	Opt_prjquota, Opt_uquota, Opt_gquota, Opt_pquota,
>  	Opt_uqnoenforce, Opt_gqnoenforce, Opt_pqnoenforce, Opt_qnoenforce,
> -	Opt_discard, Opt_nodiscard, Opt_dax,
> +	Opt_discard, Opt_nodiscard, Opt_dax, Opt_dax_enum,
>  };
>  
>  static const struct fs_parameter_spec xfs_fs_parameters[] = {
> @@ -103,6 +129,7 @@ static const struct fs_parameter_spec xfs_fs_parameters[] = {
>  	fsparam_flag("discard",		Opt_discard),
>  	fsparam_flag("nodiscard",	Opt_nodiscard),
>  	fsparam_flag("dax",		Opt_dax),
> +	fsparam_enum("dax",		Opt_dax_enum, dax_param_enums),
>  	{}
>  };
>  
> @@ -129,7 +156,6 @@ xfs_fs_show_options(
>  		{ XFS_MOUNT_GRPID,		",grpid" },
>  		{ XFS_MOUNT_DISCARD,		",discard" },
>  		{ XFS_MOUNT_LARGEIO,		",largeio" },
> -		{ XFS_MOUNT_DAX_ALWAYS,		",dax" },
>  		{ 0, NULL }
>  	};
>  	struct xfs_mount	*mp = XFS_M(root->d_sb);
> @@ -185,6 +211,13 @@ xfs_fs_show_options(
>  	if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
>  		seq_puts(m, ",noquota");
>  
> +	if (mp->m_flags & XFS_MOUNT_DAX_ALWAYS)
> +		seq_puts(m, ",dax=always");
> +	else if (mp->m_flags & XFS_MOUNT_DAX_NEVER)
> +		seq_puts(m, ",dax=never");
> +	else
> +		seq_puts(m, ",dax=inode");
> +
>  	return 0;
>  }
>  
> @@ -1244,7 +1277,10 @@ xfs_fc_parse_param(
>  		return 0;
>  #ifdef CONFIG_FS_DAX
>  	case Opt_dax:
> -		mp->m_flags |= XFS_MOUNT_DAX_ALWAYS;
> +		xfs_mount_set_dax_mode(mp, XFS_DAX_ALWAYS);
> +		return 0;
> +	case Opt_dax_enum:
> +		xfs_mount_set_dax_mode(mp, result.uint_32);
>  		return 0;
>  #endif
>  	default:
> @@ -1451,7 +1487,7 @@ xfs_fc_fill_super(
>  		if (!rtdev_is_dax && !datadev_is_dax) {
>  			xfs_alert(mp,
>  			"DAX unsupported by block device. Turning off DAX.");
> -			mp->m_flags &= ~XFS_MOUNT_DAX_ALWAYS;
> +			xfs_mount_set_dax_mode(mp, XFS_DAX_NEVER);
>  		}
>  		if (xfs_sb_version_hasreflink(&mp->m_sb)) {
>  			xfs_alert(mp,
> -- 
> 2.25.1
> 

Re: [PATCH V8 06/11] fs/xfs: Create function xfs_inode_enable_dax()

[Darrick J. Wong]

On Tue, Apr 14, 2020 at 11:45:18PM -0700, ira.weiny@intel.com wrote:
> From: Ira Weiny <ira.weiny@intel.com>
> 
> xfs_inode_supports_dax() should reflect if the inode can support DAX not
> that it is enabled for DAX.
> 
> Change the use of xfs_inode_supports_dax() to reflect only if the inode
> and underlying storage support dax.
> 
> Add a new function xfs_inode_enable_dax() which reflects if the inode
> should be enabled for DAX.
> 
> Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> 
> ---
> Changes from v7:
> 	Move S_ISREG check first
> 	use IS_ENABLED(CONFIG_FS_DAX) rather than duplicated function
> 
> Changes from v6:
> 	Change enable checks to be sequential logic.
> 	Update for 2 bit tri-state option.
> 	Make 'static' consistent.
> 	Don't set S_DAX if !CONFIG_FS_DAX
> 
> Changes from v5:
> 	Update to reflect the new tri-state mount option
> 
> Changes from v3:
> 	Update functions and names to be more clear
> 	Update commit message
> 	Merge with
> 		'fs/xfs: Clean up DAX support check'
> 		don't allow IS_DAX() on a directory
> 		use STATIC macro for static
> 		make xfs_inode_supports_dax() static
> ---
>  fs/xfs/xfs_iops.c | 28 ++++++++++++++++++++++------
>  1 file changed, 22 insertions(+), 6 deletions(-)
> 
> diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
> index a6e634631da8..2ecc2b2050ab 100644
> --- a/fs/xfs/xfs_iops.c
> +++ b/fs/xfs/xfs_iops.c
> @@ -1243,13 +1243,12 @@ xfs_inode_supports_dax(
>  {
>  	struct xfs_mount	*mp = ip->i_mount;
>  
> -	/* Only supported on non-reflinked files. */
> -	if (!S_ISREG(VFS_I(ip)->i_mode) || xfs_is_reflink_inode(ip))
> +	/* Only supported on regular files. */
> +	if (!S_ISREG(VFS_I(ip)->i_mode))
>  		return false;
>  
> -	/* DAX mount option or DAX iflag must be set. */
> -	if (!(mp->m_flags & XFS_MOUNT_DAX_ALWAYS) &&
> -	    !(ip->i_d.di_flags2 & XFS_DIFLAG2_DAX))
> +	/* Only supported on non-reflinked files. */
> +	if (xfs_is_reflink_inode(ip))
>  		return false;
>  
>  	/* Block size must match page size */
> @@ -1260,6 +1259,23 @@ xfs_inode_supports_dax(
>  	return xfs_inode_buftarg(ip)->bt_daxdev != NULL;
>  }
>  
> +static bool
> +xfs_inode_enable_dax(
> +	struct xfs_inode *ip)
> +{
> +	if (!IS_ENABLED(CONFIG_FS_DAX))
> +		return false;
> +	if (ip->i_mount->m_flags & XFS_MOUNT_DAX_NEVER)
> +		return false;
> +	if (!xfs_inode_supports_dax(ip))
> +		return false;
> +	if (ip->i_mount->m_flags & XFS_MOUNT_DAX_ALWAYS)
> +		return true;
> +	if (ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)
> +		return true;
> +	return false;

You could simplify the last two lines by making them read:

return ip->i_d.di_flags2 & XFS_DIFLAG2_DAX;

But it's probably not worth respinning the whole thing just for that.

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

--D

> +}
> +
>  STATIC void
>  xfs_diflags_to_iflags(
>  	struct inode		*inode,
> @@ -1278,7 +1294,7 @@ xfs_diflags_to_iflags(
>  		inode->i_flags |= S_SYNC;
>  	if (flags & XFS_DIFLAG_NOATIME)
>  		inode->i_flags |= S_NOATIME;
> -	if (xfs_inode_supports_dax(ip))
> +	if (xfs_inode_enable_dax(ip))
>  		inode->i_flags |= S_DAX;
>  }
>  
> -- 
> 2.25.1
> 

Re: [PATCH V8 08/11] fs: Define I_DONTCACNE in VFS layer

[Darrick J. Wong]

On Wed, Apr 15, 2020 at 10:52:16AM +0200, Jan Kara wrote:
> There's a typo in the subject - I_DONTCACNE.
> 
> On Tue 14-04-20 23:45:20, ira.weiny@intel.com wrote:
> > From: Ira Weiny <ira.weiny@intel.com>
> > 
> > DAX effective mode changes (setting of S_DAX) require inode eviction.
> > 
> > Define a flag which can be set to inform the VFS layer that inodes
> > should not be cached.  This will expedite the eviction of those nodes
> > requiring reload.
> > 
> > Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> > ---
> >  include/linux/fs.h | 6 +++++-
> >  1 file changed, 5 insertions(+), 1 deletion(-)
> > 
> > diff --git a/include/linux/fs.h b/include/linux/fs.h
> > index a818ced22961..e2db71d150c3 100644
> > --- a/include/linux/fs.h
> > +++ b/include/linux/fs.h
> > @@ -2151,6 +2151,8 @@ static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
> >   *
> >   * I_CREATING		New object's inode in the middle of setting up.
> >   *
> > + * I_DONTCACHE		Do not cache the inode
> > + *
> 
> Maybe, I'd be more specific here and write: "Evict inode as soon as it is
> not used anymore"?

I had the same two comments about the V7 version of this patch...

--D

> Otherwise the patch looks good to me so feel free to add:
> 
> Reviewed-by: Jan Kara <jack@suse.cz>
> 
> Also it would be good to CC Al Viro on this one (and the dentry flag) I
> guess.
> 
> 								Honza
> 
> >   * Q: What is the difference between I_WILL_FREE and I_FREEING?
> >   */
> >  #define I_DIRTY_SYNC		(1 << 0)
> > @@ -2173,6 +2175,7 @@ static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
> >  #define I_WB_SWITCH		(1 << 13)
> >  #define I_OVL_INUSE		(1 << 14)
> >  #define I_CREATING		(1 << 15)
> > +#define I_DONTCACHE		(1 << 16)
> >  
> >  #define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC)
> >  #define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES)
> > @@ -3042,7 +3045,8 @@ extern int inode_needs_sync(struct inode *inode);
> >  extern int generic_delete_inode(struct inode *inode);
> >  static inline int generic_drop_inode(struct inode *inode)
> >  {
> > -	return !inode->i_nlink || inode_unhashed(inode);
> > +	return !inode->i_nlink || inode_unhashed(inode) ||
> > +		(inode->i_state & I_DONTCACHE);
> >  }
> >  
> >  extern struct inode *ilookup5_nowait(struct super_block *sb,
> > -- 
> > 2.25.1
> > 
> -- 
> Jan Kara <jack@suse.com>
> SUSE Labs, CR

Re: [PATCH V8 10/11] fs/xfs: Change xfs_ioctl_setattr_dax_invalidate()

[Darrick J. Wong]

On Tue, Apr 14, 2020 at 11:45:22PM -0700, ira.weiny@intel.com wrote:
> From: Ira Weiny <ira.weiny@intel.com>
> 
> We only support changing FS_XFLAG_DAX on directories.  Files get their
> flag from the parent directory on creation only.  So no data
> invalidation needs to happen.
> 
> Alter the xfs_ioctl_setattr_dax_invalidate() to be
> xfs_ioctl_setattr_dax_validate().  xfs_ioctl_setattr_dax_validate() now
> validates that any FS_XFLAG_DAX change is ok.
> 
> This also allows use to remove the join_flags logic.
> 
> Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> 
> ---
> Changes from V7:
> 	Use new flag_inode_dontcache()
> 	Skip don't cache if mount over ride is active.
> 
> Changes from v6:
> 	Fix completely broken implementation and update commit message.
> 	Use the new VFS layer I_DONTCACHE to facilitate inode eviction
> 	and S_DAX changing on drop_caches
> 
> Changes from v5:
> 	New patch
> ---
>  fs/xfs/xfs_ioctl.c | 105 +++++++++------------------------------------
>  1 file changed, 20 insertions(+), 85 deletions(-)
> 
> diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
> index c6cd92ef4a05..75d4a830ef38 100644
> --- a/fs/xfs/xfs_ioctl.c
> +++ b/fs/xfs/xfs_ioctl.c
> @@ -1145,63 +1145,28 @@ xfs_ioctl_setattr_xflags(
>  }
>  
>  /*
> - * If we are changing DAX flags, we have to ensure the file is clean and any
> - * cached objects in the address space are invalidated and removed. This
> - * requires us to lock out other IO and page faults similar to a truncate
> - * operation. The locks need to be held until the transaction has been committed
> - * so that the cache invalidation is atomic with respect to the DAX flag
> - * manipulation.
> + * Mark inodes with a changing FS_XFLAG_DAX, I_DONTCACHE
>   */
> -static int
> +static void
>  xfs_ioctl_setattr_dax_invalidate(
>  	struct xfs_inode	*ip,
> -	struct fsxattr		*fa,
> -	int			*join_flags)
> +	struct fsxattr		*fa)
>  {
> -	struct inode		*inode = VFS_I(ip);
> -	struct super_block	*sb = inode->i_sb;
> -	int			error;
> -
> -	*join_flags = 0;
> -
> -	/*
> -	 * It is only valid to set the DAX flag on regular files and
> -	 * directories on filesystems where the block size is equal to the page
> -	 * size. On directories it serves as an inherited hint so we don't
> -	 * have to check the device for dax support or flush pagecache.
> -	 */
> -	if (fa->fsx_xflags & FS_XFLAG_DAX) {
> -		struct xfs_buftarg	*target = xfs_inode_buftarg(ip);
> -
> -		if (!bdev_dax_supported(target->bt_bdev, sb->s_blocksize))
> -			return -EINVAL;
> -	}
> -
> -	/* If the DAX state is not changing, we have nothing to do here. */
> -	if ((fa->fsx_xflags & FS_XFLAG_DAX) && IS_DAX(inode))
> -		return 0;
> -	if (!(fa->fsx_xflags & FS_XFLAG_DAX) && !IS_DAX(inode))
> -		return 0;
> +	struct xfs_mount	*mp = ip->i_mount;
> +	struct inode            *inode = VFS_I(ip);
>  
>  	if (S_ISDIR(inode->i_mode))
> -		return 0;
> -
> -	/* lock, flush and invalidate mapping in preparation for flag change */
> -	xfs_ilock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
> -	error = filemap_write_and_wait(inode->i_mapping);
> -	if (error)
> -		goto out_unlock;
> -	error = invalidate_inode_pages2(inode->i_mapping);
> -	if (error)
> -		goto out_unlock;
> +		return;
>  
> -	*join_flags = XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL;
> -	return 0;
> -
> -out_unlock:
> -	xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
> -	return error;
> +	if (mp->m_flags & XFS_MOUNT_DAX_ALWAYS ||
> +	    mp->m_flags & XFS_MOUNT_DAX_NEVER)

This ought to have parentheses around the two bit manipulation
operations to avoid complaints from picky/stupid compilers.

--D

> +		return;
>  
> +	if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
> +	    !(ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)) ||
> +	    (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
> +	     (ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)))
> +		flag_inode_dontcache(inode);
>  }
>  
>  /*
> @@ -1209,17 +1174,10 @@ xfs_ioctl_setattr_dax_invalidate(
>   * have permission to do so. On success, return a clean transaction and the
>   * inode locked exclusively ready for further operation specific checks. On
>   * failure, return an error without modifying or locking the inode.
> - *
> - * The inode might already be IO locked on call. If this is the case, it is
> - * indicated in @join_flags and we take full responsibility for ensuring they
> - * are unlocked from now on. Hence if we have an error here, we still have to
> - * unlock them. Otherwise, once they are joined to the transaction, they will
> - * be unlocked on commit/cancel.
>   */
>  static struct xfs_trans *
>  xfs_ioctl_setattr_get_trans(
> -	struct xfs_inode	*ip,
> -	int			join_flags)
> +	struct xfs_inode	*ip)
>  {
>  	struct xfs_mount	*mp = ip->i_mount;
>  	struct xfs_trans	*tp;
> @@ -1236,8 +1194,7 @@ xfs_ioctl_setattr_get_trans(
>  		goto out_unlock;
>  
>  	xfs_ilock(ip, XFS_ILOCK_EXCL);
> -	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | join_flags);
> -	join_flags = 0;
> +	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
>  
>  	/*
>  	 * CAP_FOWNER overrides the following restrictions:
> @@ -1258,8 +1215,6 @@ xfs_ioctl_setattr_get_trans(
>  out_cancel:
>  	xfs_trans_cancel(tp);
>  out_unlock:
> -	if (join_flags)
> -		xfs_iunlock(ip, join_flags);
>  	return ERR_PTR(error);
>  }
>  
> @@ -1386,7 +1341,6 @@ xfs_ioctl_setattr(
>  	struct xfs_dquot	*pdqp = NULL;
>  	struct xfs_dquot	*olddquot = NULL;
>  	int			code;
> -	int			join_flags = 0;
>  
>  	trace_xfs_ioctl_setattr(ip);
>  
> @@ -1410,18 +1364,9 @@ xfs_ioctl_setattr(
>  			return code;
>  	}
>  
> -	/*
> -	 * Changing DAX config may require inode locking for mapping
> -	 * invalidation. These need to be held all the way to transaction commit
> -	 * or cancel time, so need to be passed through to
> -	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
> -	 * appropriately.
> -	 */
> -	code = xfs_ioctl_setattr_dax_invalidate(ip, fa, &join_flags);
> -	if (code)
> -		goto error_free_dquots;
> +	xfs_ioctl_setattr_dax_invalidate(ip, fa);
>  
> -	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
> +	tp = xfs_ioctl_setattr_get_trans(ip);
>  	if (IS_ERR(tp)) {
>  		code = PTR_ERR(tp);
>  		goto error_free_dquots;
> @@ -1552,7 +1497,6 @@ xfs_ioc_setxflags(
>  	struct fsxattr		fa;
>  	struct fsxattr		old_fa;
>  	unsigned int		flags;
> -	int			join_flags = 0;
>  	int			error;
>  
>  	if (copy_from_user(&flags, arg, sizeof(flags)))
> @@ -1569,18 +1513,9 @@ xfs_ioc_setxflags(
>  	if (error)
>  		return error;
>  
> -	/*
> -	 * Changing DAX config may require inode locking for mapping
> -	 * invalidation. These need to be held all the way to transaction commit
> -	 * or cancel time, so need to be passed through to
> -	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
> -	 * appropriately.
> -	 */
> -	error = xfs_ioctl_setattr_dax_invalidate(ip, &fa, &join_flags);
> -	if (error)
> -		goto out_drop_write;
> +	xfs_ioctl_setattr_dax_invalidate(ip, &fa);
>  
> -	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
> +	tp = xfs_ioctl_setattr_get_trans(ip);
>  	if (IS_ERR(tp)) {
>  		error = PTR_ERR(tp);
>  		goto out_drop_write;
> -- 
> 2.25.1
> 

Re: [PATCH V8 11/11] Documentation/dax: Update Usage section

[Darrick J. Wong]

On Tue, Apr 14, 2020 at 11:45:23PM -0700, ira.weiny@intel.com wrote:
> From: Ira Weiny <ira.weiny@intel.com>
> 
> Update the Usage section to reflect the new individual dax selection
> functionality.
> 
> Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> 
> ---
> Changes from V7:
> 	Cleanups/clarifications from Darrick and Dan
> 
> Changes from V6:
> 	Update to allow setting FS_XFLAG_DAX any time.
> 	Update with list of behaviors from Darrick
> 	https://lore.kernel.org/lkml/20200409165927.GD6741@magnolia/
> 
> Changes from V5:
> 	Update to reflect the agreed upon semantics
> 	https://lore.kernel.org/lkml/20200405061945.GA94792@iweiny-DESK2.sc.intel.com/
> ---
>  Documentation/filesystems/dax.txt | 166 +++++++++++++++++++++++++++++-
>  1 file changed, 163 insertions(+), 3 deletions(-)
> 
> diff --git a/Documentation/filesystems/dax.txt b/Documentation/filesystems/dax.txt
> index 679729442fd2..893820c53f49 100644
> --- a/Documentation/filesystems/dax.txt
> +++ b/Documentation/filesystems/dax.txt
> @@ -17,11 +17,171 @@ For file mappings, the storage device is mapped directly into userspace.
>  Usage
>  -----
>  
> -If you have a block device which supports DAX, you can make a filesystem
> +If you have a block device which supports DAX, you can make a file system
>  on it as usual.  The DAX code currently only supports files with a block
>  size equal to your kernel's PAGE_SIZE, so you may need to specify a block
> -size when creating the filesystem.  When mounting it, use the "-o dax"
> -option on the command line or add 'dax' to the options in /etc/fstab.
> +size when creating the file system.
> +
> +Currently 3 filesystems support DAX, ext2, ext4 and xfs.  Enabling DAX on them

"...support DAX: ext2, ext4..."

Please put a colon after "DAX" since it's not part of the list.

> +is different.
> +
> +Enabling DAX on ext4 and ext2
> +-----------------------------
> +
> +When mounting the filesystem, use the "-o dax" option on the command line or
> +add 'dax' to the options in /etc/fstab.  This works to enable DAX on all files
> +within the filesystem.  It is equivalent to the '-o dax=always' behavior below
> +with the exception that the STATX_ATTR_DAX flag is not supported, nor needed,
> +as it is always true.

STATX_ATTR_DAX isn't supported?  I thought ext[24] set S_DAX, so the
statx flag should work the same as it does on xfs?

I also wonder if it's worth mentioning that in the long run ext4 will
match the xfs semantics, but maybe that's better left for the ext4 rfc
series.

> +
> +
> +Enabling DAX on xfs
> +-------------------
> +
> +Summary
> +-------
> +
> + 1. There exists an in-kernel file access mode flag S_DAX that corresponds to
> +    the statx flag STATX_ATTR_DAX.  See the manpage for statx(2) for details
> +    about this access mode.
> +
> + 2. There exists an advisory file inode flag FS_XFLAG_DAX that is
> +    inherited from the parent directory FS_XFLAG_DAX inode flag at file
> +    creation time.  This advisory flag can be set or cleared at any
> +    time, but doing so does not immediately affect the S_DAX state.
> +
> +    Unless overridden by mount options (see (3)), if FS_XFLAG_DAX is set
> +    and the fs is on pmem then it will enable S_DAX at inode load time;
> +    if FS_XFLAG_DAX is not set, it will not enable S_DAX.
> +
> + 3. There exists a dax= mount option.
> +
> +    "-o dax=never"  means "never set S_DAX, ignore FS_XFLAG_DAX."
> +
> +    "-o dax=always" means "always set S_DAX (at least on pmem),
> +                    and ignore FS_XFLAG_DAX."
> +
> +    "-o dax"        is an alias for "dax=always".
> +
> +    "-o dax=inode"  means "follow FS_XFLAG_DAX" and is the default.
> +
> + 4. There exists an advisory directory inode flag FS_XFLAG_DAX that can
> +    be set or cleared at any time.  The flag state is inherited by any files or
> +    subdirectories when they are created within that directory.
> +
> + 5. Programs that require a specific file access mode (DAX or not DAX)
> +    can do one of the following:
> +
> +    (a) Create files in directories that the FS_XFLAG_DAX flag set as
> +        needed; or
> +
> +    (b) Have the administrator set an override via mount option; or
> +
> +    (c) Set or clear the file's FS_XFLAG_DAX flag as needed.  Programs
> +        must then cause the kernel to evict the inode from memory.  This
> +        can be done by:
> +
> +        i>  Closing the file and re-opening the file and using statx to
> +            see if the fs has changed the S_DAX flag; and
> +
> +        ii> If the file still does not have the desired S_DAX access
> +            mode, either unmount and remount the filesystem, or close
> +            the file and use drop_caches.
> +
> + 6. It is expected that users who want to squeeze every last bit of performance
> +    out of the particular rough and tumble bits of their storage will also be
> +    exposed to the difficulties of what happens when the operating system can't
> +    totally virtualize those hardware capabilities.  DAX is such a feature.
> +
> +
> +Details
> +-------
> +
> +There are 2 per-file dax flags.  One is a physical inode setting (FS_XFLAG_DAX)
> +and the other a currently enabled state (S_DAX).
> +
> +FS_XFLAG_DAX is maintained, on disk, on individual inodes.  It is preserved
> +within the file system.  This 'physical' config setting can be set using an
> +ioctl and/or an application such as "xfs_io -c 'chattr [-+]x'".  Files and
> +directories automatically inherit FS_XFLAG_DAX from their parent directory
> +_when_ _created_.  Therefore, setting FS_XFLAG_DAX at directory creation time
> +can be used to set a default behavior for an entire sub-tree.  (Doing so on the
> +root directory acts to set a default for the entire file system.)

Urk, I guess I need to push that patch to make mkfs.xfs do this. ;)

> +To clarify inheritance here are 3 examples:
> +
> +Example A:
> +
> +mkdir -p a/b/c
> +xfs_io 'chattr +x' a
> +mkdir a/b/c/d
> +mkdir a/e
> +
> +	dax: a,e
> +	no dax: b,c,d
> +
> +Example B:
> +
> +mkdir a
> +xfs_io 'chattr +x' a
> +mkdir -p a/b/c/d
> +
> +	dax: a,b,c,d
> +	no dax:
> +
> +Example C:
> +
> +mkdir -p a/b/c
> +xfs_io 'chattr +x' c
> +mkdir a/b/c/d
> +
> +	dax: c,d
> +	no dax: a,b
> +
> +
> +The current enabled state (S_DAX) is set when a file inode is _loaded_ based on
> +the underlying media support, the value of FS_XFLAG_DAX, and the file systems
> +dax mount option setting.  See below.
> +
> +statx can be used to query S_DAX.  NOTE that a directory will never have S_DAX
> +set and therefore statx will never indicate that S_DAX is set on directories.
> +
> +NOTE: Setting the FS_XFLAG_DAX (specifically or through inheritance) occurs
> +even if the underlying media does not support dax and/or the file system is
> +overridden with a mount option.
> +
> +
> +Overriding FS_XFLAG_DAX (dax= mount option)
> +-------------------------------------------
> +
> +There exists a dax mount option.  Using the mount option does not change the
> +physical configured state of individual files but overrides the S_DAX operating
> +state when inodes are loaded.
> +
> +Given underlying media support, the dax mount option is a tri-state option
> +(never, always, inode) with the following meanings:
> +
> +   "-o dax=never" means "never set S_DAX, ignore FS_XFLAG_DAX"
> +   "-o dax=always" means "always set S_DAX, ignore FS_XFLAG_DAX"
> +        "-o dax" by itself means "dax=always" to remain compatible with older
> +	         kernels
> +   "-o dax=inode" means "follow FS_XFLAG_DAX"
> +
> +The default state is 'inode'.  Given underlying media support, the following
> +algorithm is used to determine the effective mode of the file S_DAX on a
> +capable device.
> +
> +	S_DAX = FS_XFLAG_DAX;
> +
> +	if (dax_mount == "always")
> +		S_DAX = true;
> +	else if (dax_mount == "off"
> +		S_DAX = false;

The logic in this pseudocode doesn't match the order that's in
xfs_inode_enable_dax.  I think the outcome is the same, but it's easier
to verify that if the statements are in roughly the same order.

	if dax=never:
		S_DAX = false
	elif the file system and media don't both support DAX:
		S_DAX = false
	elif dax=always:
		S_DAX = true
	else:
		S_DAX = inode flag status

--D

> +
> +To reiterate: Setting, and inheritance, continues to affect FS_XFLAG_DAX even
> +while the file system is mounted with a dax override.  However, in-core inode
> +state (S_DAX) will continue to be overridden until the filesystem is remounted
> +with dax=inode and the inode is evicted."
>  
>  
>  Implementation Tips for Block Driver Writers
> -- 
> 2.25.1
> 

Re: [PATCH V8 05/11] fs/xfs: Make DAX mount option a tri-state

[Ira Weiny]

On Wed, Apr 15, 2020 at 08:16:13AM -0700, Darrick J. Wong wrote:
> On Tue, Apr 14, 2020 at 11:45:17PM -0700, ira.weiny@intel.com wrote:
> > From: Ira Weiny <ira.weiny@intel.com>
> > 
> > As agreed upon[1].  We make the dax mount option a tri-state.  '-o dax'
> > continues to operate the same.  We add 'always', 'never', and 'inode'
> > (default).
> > 
> > [1] https://lore.kernel.org/lkml/20200405061945.GA94792@iweiny-DESK2.sc.intel.com/
> > 
> > Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> > 
> > ---
> > Changes from v7:
> > 	Change to XFS_MOUNT_DAX_NEVER
> > 
> > Changes from v6:
> > 	Use 2 flag bits rather than a field.
> > 	change iflag to inode
> > 
> > Changes from v5:
> > 	New Patch
> > ---
> >  fs/xfs/xfs_mount.h |  3 ++-
> >  fs/xfs/xfs_super.c | 44 ++++++++++++++++++++++++++++++++++++++++----
> >  2 files changed, 42 insertions(+), 5 deletions(-)
> > 
> > diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
> > index 54bd74088936..2e88c30642e3 100644
> > --- a/fs/xfs/xfs_mount.h
> > +++ b/fs/xfs/xfs_mount.h
> > @@ -233,7 +233,8 @@ typedef struct xfs_mount {
> >  						   allocator */
> >  #define XFS_MOUNT_NOATTR2	(1ULL << 25)	/* disable use of attr2 format */
> >  
> > -#define XFS_MOUNT_DAX_ALWAYS	(1ULL << 62)	/* TEST ONLY! */
> > +#define XFS_MOUNT_DAX_ALWAYS	(1ULL << 62)
> > +#define XFS_MOUNT_DAX_NEVER	(1ULL << 63)
> >  
> >  /*
> >   * Max and min values for mount-option defined I/O
> > diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
> > index 3863f41757d2..142e5d03566f 100644
> > --- a/fs/xfs/xfs_super.c
> > +++ b/fs/xfs/xfs_super.c
> > @@ -47,6 +47,32 @@ static struct kset *xfs_kset;		/* top-level xfs sysfs dir */
> >  static struct xfs_kobj xfs_dbg_kobj;	/* global debug sysfs attrs */
> >  #endif
> >  
> > +enum {
> 
> enum xfs_dax_mode {  for the reasons given below?
> 
> > +	XFS_DAX_INODE = 0,
> > +	XFS_DAX_ALWAYS = 1,
> > +	XFS_DAX_NEVER = 2,
> > +};
> > +
> > +static void xfs_mount_set_dax_mode(struct xfs_mount *mp, u32 val)
> 
> xfs style, please:
> 
> static void
> xfs_mount_set_dax_mode(
> 	struct xfs_mount	*mp,
> 	u32			val)

NP changed.

> 
> or if you give a name to the enum above, you can enforce some type
> safety too:
> 
> 	enum xfs_dax_mode	val)

The problem is that the option parsing returns result.uint_32 type.  Generally
I agree with using more specific types but here I'm not sure it matters.

> 
> > +{
> > +	if (val == XFS_DAX_INODE) {
> 
> and this probably could have been a "switch (val) {", in which case if
> the enum ever gets expanded then gcc will whine about missing switch
> cases.

Sure...  that does require the enum to be defined and used in the signature...

Ok I'll use the enum!  :-D

done.

Thanks,
Ira


> 
> The rest of the patch looks good.
> 
> --D
> 
> > +		mp->m_flags &= ~(XFS_MOUNT_DAX_ALWAYS | XFS_MOUNT_DAX_NEVER);
> > +	} else if (val == XFS_DAX_ALWAYS) {
> > +		mp->m_flags |= XFS_MOUNT_DAX_ALWAYS;
> > +		mp->m_flags &= ~XFS_MOUNT_DAX_NEVER;
> > +	} else if (val == XFS_DAX_NEVER) {
> > +		mp->m_flags |= XFS_MOUNT_DAX_NEVER;
> > +		mp->m_flags &= ~XFS_MOUNT_DAX_ALWAYS;
> > +	}
> > +}
> > +
> > +static const struct constant_table dax_param_enums[] = {
> > +	{"inode",	XFS_DAX_INODE },
> > +	{"always",	XFS_DAX_ALWAYS },
> > +	{"never",	XFS_DAX_NEVER },
> > +	{}
> > +};
> > +
> >  /*
> >   * Table driven mount option parser.
> >   */
> > @@ -59,7 +85,7 @@ enum {
> >  	Opt_filestreams, Opt_quota, Opt_noquota, Opt_usrquota, Opt_grpquota,
> >  	Opt_prjquota, Opt_uquota, Opt_gquota, Opt_pquota,
> >  	Opt_uqnoenforce, Opt_gqnoenforce, Opt_pqnoenforce, Opt_qnoenforce,
> > -	Opt_discard, Opt_nodiscard, Opt_dax,
> > +	Opt_discard, Opt_nodiscard, Opt_dax, Opt_dax_enum,
> >  };
> >  
> >  static const struct fs_parameter_spec xfs_fs_parameters[] = {
> > @@ -103,6 +129,7 @@ static const struct fs_parameter_spec xfs_fs_parameters[] = {
> >  	fsparam_flag("discard",		Opt_discard),
> >  	fsparam_flag("nodiscard",	Opt_nodiscard),
> >  	fsparam_flag("dax",		Opt_dax),
> > +	fsparam_enum("dax",		Opt_dax_enum, dax_param_enums),
> >  	{}
> >  };
> >  
> > @@ -129,7 +156,6 @@ xfs_fs_show_options(
> >  		{ XFS_MOUNT_GRPID,		",grpid" },
> >  		{ XFS_MOUNT_DISCARD,		",discard" },
> >  		{ XFS_MOUNT_LARGEIO,		",largeio" },
> > -		{ XFS_MOUNT_DAX_ALWAYS,		",dax" },
> >  		{ 0, NULL }
> >  	};
> >  	struct xfs_mount	*mp = XFS_M(root->d_sb);
> > @@ -185,6 +211,13 @@ xfs_fs_show_options(
> >  	if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
> >  		seq_puts(m, ",noquota");
> >  
> > +	if (mp->m_flags & XFS_MOUNT_DAX_ALWAYS)
> > +		seq_puts(m, ",dax=always");
> > +	else if (mp->m_flags & XFS_MOUNT_DAX_NEVER)
> > +		seq_puts(m, ",dax=never");
> > +	else
> > +		seq_puts(m, ",dax=inode");
> > +
> >  	return 0;
> >  }
> >  
> > @@ -1244,7 +1277,10 @@ xfs_fc_parse_param(
> >  		return 0;
> >  #ifdef CONFIG_FS_DAX
> >  	case Opt_dax:
> > -		mp->m_flags |= XFS_MOUNT_DAX_ALWAYS;
> > +		xfs_mount_set_dax_mode(mp, XFS_DAX_ALWAYS);
> > +		return 0;
> > +	case Opt_dax_enum:
> > +		xfs_mount_set_dax_mode(mp, result.uint_32);
> >  		return 0;
> >  #endif
> >  	default:
> > @@ -1451,7 +1487,7 @@ xfs_fc_fill_super(
> >  		if (!rtdev_is_dax && !datadev_is_dax) {
> >  			xfs_alert(mp,
> >  			"DAX unsupported by block device. Turning off DAX.");
> > -			mp->m_flags &= ~XFS_MOUNT_DAX_ALWAYS;
> > +			xfs_mount_set_dax_mode(mp, XFS_DAX_NEVER);
> >  		}
> >  		if (xfs_sb_version_hasreflink(&mp->m_sb)) {
> >  			xfs_alert(mp,
> > -- 
> > 2.25.1
> > 

Re: [PATCH V8 08/11] fs: Define I_DONTCACNE in VFS layer

[Ira Weiny]

On Wed, Apr 15, 2020 at 10:52:16AM +0200, Jan Kara wrote:
> There's a typo in the subject - I_DONTCACNE.
> 
> On Tue 14-04-20 23:45:20, ira.weiny@intel.com wrote:
> > From: Ira Weiny <ira.weiny@intel.com>
> > 
> > DAX effective mode changes (setting of S_DAX) require inode eviction.
> > 
> > Define a flag which can be set to inform the VFS layer that inodes
> > should not be cached.  This will expedite the eviction of those nodes
> > requiring reload.
> > 
> > Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> > ---
> >  include/linux/fs.h | 6 +++++-
> >  1 file changed, 5 insertions(+), 1 deletion(-)
> > 
> > diff --git a/include/linux/fs.h b/include/linux/fs.h
> > index a818ced22961..e2db71d150c3 100644
> > --- a/include/linux/fs.h
> > +++ b/include/linux/fs.h
> > @@ -2151,6 +2151,8 @@ static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
> >   *
> >   * I_CREATING		New object's inode in the middle of setting up.
> >   *
> > + * I_DONTCACHE		Do not cache the inode
> > + *
> 
> Maybe, I'd be more specific here and write: "Evict inode as soon as it is
> not used anymore"?

Thanks!  done

> 
> Otherwise the patch looks good to me so feel free to add:
> 
> Reviewed-by: Jan Kara <jack@suse.cz>
> 
> Also it would be good to CC Al Viro on this one (and the dentry flag) I
> guess.

Thanks I'll add him on those,

Thanks,
Ira

> 
> 								Honza
> 
> >   * Q: What is the difference between I_WILL_FREE and I_FREEING?
> >   */
> >  #define I_DIRTY_SYNC		(1 << 0)
> > @@ -2173,6 +2175,7 @@ static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
> >  #define I_WB_SWITCH		(1 << 13)
> >  #define I_OVL_INUSE		(1 << 14)
> >  #define I_CREATING		(1 << 15)
> > +#define I_DONTCACHE		(1 << 16)
> >  
> >  #define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC)
> >  #define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES)
> > @@ -3042,7 +3045,8 @@ extern int inode_needs_sync(struct inode *inode);
> >  extern int generic_delete_inode(struct inode *inode);
> >  static inline int generic_drop_inode(struct inode *inode)
> >  {
> > -	return !inode->i_nlink || inode_unhashed(inode);
> > +	return !inode->i_nlink || inode_unhashed(inode) ||
> > +		(inode->i_state & I_DONTCACHE);
> >  }
> >  
> >  extern struct inode *ilookup5_nowait(struct super_block *sb,
> > -- 
> > 2.25.1
> > 
> -- 
> Jan Kara <jack@suse.com>
> SUSE Labs, CR

Re: [PATCH V8 08/11] fs: Define I_DONTCACNE in VFS layer

[Ira Weiny]

On Wed, Apr 15, 2020 at 08:18:32AM -0700, Darrick J. Wong wrote:
> On Wed, Apr 15, 2020 at 10:52:16AM +0200, Jan Kara wrote:
> > There's a typo in the subject - I_DONTCACNE.
> > 
> > On Tue 14-04-20 23:45:20, ira.weiny@intel.com wrote:
> > > From: Ira Weiny <ira.weiny@intel.com>
> > > 
> > > DAX effective mode changes (setting of S_DAX) require inode eviction.
> > > 
> > > Define a flag which can be set to inform the VFS layer that inodes
> > > should not be cached.  This will expedite the eviction of those nodes
> > > requiring reload.
> > > 
> > > Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> > > ---
> > >  include/linux/fs.h | 6 +++++-
> > >  1 file changed, 5 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/include/linux/fs.h b/include/linux/fs.h
> > > index a818ced22961..e2db71d150c3 100644
> > > --- a/include/linux/fs.h
> > > +++ b/include/linux/fs.h
> > > @@ -2151,6 +2151,8 @@ static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
> > >   *
> > >   * I_CREATING		New object's inode in the middle of setting up.
> > >   *
> > > + * I_DONTCACHE		Do not cache the inode
> > > + *
> > 
> > Maybe, I'd be more specific here and write: "Evict inode as soon as it is
> > not used anymore"?
> 
> I had the same two comments about the V7 version of this patch...

Sorry I must have missed it.

Done for V9,
Ira

> 
> --D
> 
> > Otherwise the patch looks good to me so feel free to add:
> > 
> > Reviewed-by: Jan Kara <jack@suse.cz>
> > 
> > Also it would be good to CC Al Viro on this one (and the dentry flag) I
> > guess.
> > 
> > 								Honza
> > 
> > >   * Q: What is the difference between I_WILL_FREE and I_FREEING?
> > >   */
> > >  #define I_DIRTY_SYNC		(1 << 0)
> > > @@ -2173,6 +2175,7 @@ static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
> > >  #define I_WB_SWITCH		(1 << 13)
> > >  #define I_OVL_INUSE		(1 << 14)
> > >  #define I_CREATING		(1 << 15)
> > > +#define I_DONTCACHE		(1 << 16)
> > >  
> > >  #define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC)
> > >  #define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES)
> > > @@ -3042,7 +3045,8 @@ extern int inode_needs_sync(struct inode *inode);
> > >  extern int generic_delete_inode(struct inode *inode);
> > >  static inline int generic_drop_inode(struct inode *inode)
> > >  {
> > > -	return !inode->i_nlink || inode_unhashed(inode);
> > > +	return !inode->i_nlink || inode_unhashed(inode) ||
> > > +		(inode->i_state & I_DONTCACHE);
> > >  }
> > >  
> > >  extern struct inode *ilookup5_nowait(struct super_block *sb,
> > > -- 
> > > 2.25.1
> > > 
> > -- 
> > Jan Kara <jack@suse.com>
> > SUSE Labs, CR

Re: [PATCH V8 09/11] fs: Introduce DCACHE_DONTCACHE

[Ira Weiny]

On Wed, Apr 15, 2020 at 11:01:53AM +0200, Jan Kara wrote:
> On Tue 14-04-20 23:45:21, ira.weiny@intel.com wrote:
> > From: Ira Weiny <ira.weiny@intel.com>
> > 
> > DCACHE_DONTCACHE indicates a dentry should not be cached on final
> > dput().
> > 
> > Also add a helper function which will flag I_DONTCACHE as well ad
> > DCACHE_DONTCACHE on all dentries point to a specified inode.
> 
> I think this sentence needs more work :). Like: Also add a helper function
> which will mark the inode with I_DONTCACHE flag and also mark all dentries
> pointing to a specified inode as DCACHE_DONTCACHE.
> 
> > 
> > Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> > 
> > ---
> > Changes from V7:
> > 	new patch
> > ---
> >  fs/dcache.c            |  4 ++++
> >  fs/inode.c             | 15 +++++++++++++++
> >  include/linux/dcache.h |  2 ++
> >  include/linux/fs.h     |  1 +
> >  4 files changed, 22 insertions(+)
> 
> ...
> 
> > diff --git a/fs/inode.c b/fs/inode.c
> > index 93d9252a00ab..b8b1917a324e 100644
> > --- a/fs/inode.c
> > +++ b/fs/inode.c
> > @@ -1526,6 +1526,21 @@ int generic_delete_inode(struct inode *inode)
> >  }
> >  EXPORT_SYMBOL(generic_delete_inode);
> >  
> > +void flag_inode_dontcache(struct inode *inode)
> 
> mark_inode_dontcache?

That works.

> 
> > +{
> > +	struct dentry *dent;
> 
> This is really nitpicking but dentry variables are usually called 'de' or
> 'dentry' :)

Easy change.  done.

> 
> > +
> > +	rcu_read_lock();
> 
> I don't think this list is safe to traverse under RCU. E.g.
> dentry_unlink_inode() does hlist_del_init(&dentry->d_u.d_alias). Usually,
> we traverse this list under inode->i_lock protection AFAICS.

Ah...  not sure where I got that.  I thought I found this locked with rcu
somewhere but I obviously got confused.  And I did not even use the rcu version
of the list iterator...  :-(

I'll clean it up, thanks for the review,
Ira

> 
> 								Honza
> 
> > +	hlist_for_each_entry(dent, &inode->i_dentry, d_u.d_alias) {
> > +		spin_lock(&dent->d_lock);
> > +		dent->d_flags |= DCACHE_DONTCACHE;
> > +		spin_unlock(&dent->d_lock);
> > +	}
> > +	rcu_read_unlock();
> > +	inode->i_state |= I_DONTCACHE;
> > +}
> > +EXPORT_SYMBOL(flag_inode_dontcache);
> > +
> >  /*
> >   * Called when we're dropping the last reference
> >   * to an inode.
> -- 
> Jan Kara <jack@suse.com>
> SUSE Labs, CR

Re: [PATCH V8 10/11] fs/xfs: Change xfs_ioctl_setattr_dax_invalidate()

[Ira Weiny]

On Wed, Apr 15, 2020 at 08:21:12AM -0700, Darrick J. Wong wrote:
> On Tue, Apr 14, 2020 at 11:45:22PM -0700, ira.weiny@intel.com wrote:
> > From: Ira Weiny <ira.weiny@intel.com>
> > 
> > We only support changing FS_XFLAG_DAX on directories.  Files get their
> > flag from the parent directory on creation only.  So no data
> > invalidation needs to happen.
> > 
> > Alter the xfs_ioctl_setattr_dax_invalidate() to be
> > xfs_ioctl_setattr_dax_validate().  xfs_ioctl_setattr_dax_validate() now
> > validates that any FS_XFLAG_DAX change is ok.
> > 
> > This also allows use to remove the join_flags logic.
> > 
> > Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> > 
> > ---
> > Changes from V7:
> > 	Use new flag_inode_dontcache()
> > 	Skip don't cache if mount over ride is active.
> > 
> > Changes from v6:
> > 	Fix completely broken implementation and update commit message.
> > 	Use the new VFS layer I_DONTCACHE to facilitate inode eviction
> > 	and S_DAX changing on drop_caches
> > 
> > Changes from v5:
> > 	New patch
> > ---
> >  fs/xfs/xfs_ioctl.c | 105 +++++++++------------------------------------
> >  1 file changed, 20 insertions(+), 85 deletions(-)
> > 
> > diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
> > index c6cd92ef4a05..75d4a830ef38 100644
> > --- a/fs/xfs/xfs_ioctl.c
> > +++ b/fs/xfs/xfs_ioctl.c
> > @@ -1145,63 +1145,28 @@ xfs_ioctl_setattr_xflags(
> >  }
> >  
> >  /*
> > - * If we are changing DAX flags, we have to ensure the file is clean and any
> > - * cached objects in the address space are invalidated and removed. This
> > - * requires us to lock out other IO and page faults similar to a truncate
> > - * operation. The locks need to be held until the transaction has been committed
> > - * so that the cache invalidation is atomic with respect to the DAX flag
> > - * manipulation.
> > + * Mark inodes with a changing FS_XFLAG_DAX, I_DONTCACHE
> >   */
> > -static int
> > +static void
> >  xfs_ioctl_setattr_dax_invalidate(
> >  	struct xfs_inode	*ip,
> > -	struct fsxattr		*fa,
> > -	int			*join_flags)
> > +	struct fsxattr		*fa)
> >  {
> > -	struct inode		*inode = VFS_I(ip);
> > -	struct super_block	*sb = inode->i_sb;
> > -	int			error;
> > -
> > -	*join_flags = 0;
> > -
> > -	/*
> > -	 * It is only valid to set the DAX flag on regular files and
> > -	 * directories on filesystems where the block size is equal to the page
> > -	 * size. On directories it serves as an inherited hint so we don't
> > -	 * have to check the device for dax support or flush pagecache.
> > -	 */
> > -	if (fa->fsx_xflags & FS_XFLAG_DAX) {
> > -		struct xfs_buftarg	*target = xfs_inode_buftarg(ip);
> > -
> > -		if (!bdev_dax_supported(target->bt_bdev, sb->s_blocksize))
> > -			return -EINVAL;
> > -	}
> > -
> > -	/* If the DAX state is not changing, we have nothing to do here. */
> > -	if ((fa->fsx_xflags & FS_XFLAG_DAX) && IS_DAX(inode))
> > -		return 0;
> > -	if (!(fa->fsx_xflags & FS_XFLAG_DAX) && !IS_DAX(inode))
> > -		return 0;
> > +	struct xfs_mount	*mp = ip->i_mount;
> > +	struct inode            *inode = VFS_I(ip);
> >  
> >  	if (S_ISDIR(inode->i_mode))
> > -		return 0;
> > -
> > -	/* lock, flush and invalidate mapping in preparation for flag change */
> > -	xfs_ilock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
> > -	error = filemap_write_and_wait(inode->i_mapping);
> > -	if (error)
> > -		goto out_unlock;
> > -	error = invalidate_inode_pages2(inode->i_mapping);
> > -	if (error)
> > -		goto out_unlock;
> > +		return;
> >  
> > -	*join_flags = XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL;
> > -	return 0;
> > -
> > -out_unlock:
> > -	xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
> > -	return error;
> > +	if (mp->m_flags & XFS_MOUNT_DAX_ALWAYS ||
> > +	    mp->m_flags & XFS_MOUNT_DAX_NEVER)
> 
> This ought to have parentheses around the two bit manipulation
> operations to avoid complaints from picky/stupid compilers.

Fixed,
Ira

> 
> --D
> 
> > +		return;
> >  
> > +	if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
> > +	    !(ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)) ||
> > +	    (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
> > +	     (ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)))
> > +		flag_inode_dontcache(inode);
> >  }
> >  
> >  /*
> > @@ -1209,17 +1174,10 @@ xfs_ioctl_setattr_dax_invalidate(
> >   * have permission to do so. On success, return a clean transaction and the
> >   * inode locked exclusively ready for further operation specific checks. On
> >   * failure, return an error without modifying or locking the inode.
> > - *
> > - * The inode might already be IO locked on call. If this is the case, it is
> > - * indicated in @join_flags and we take full responsibility for ensuring they
> > - * are unlocked from now on. Hence if we have an error here, we still have to
> > - * unlock them. Otherwise, once they are joined to the transaction, they will
> > - * be unlocked on commit/cancel.
> >   */
> >  static struct xfs_trans *
> >  xfs_ioctl_setattr_get_trans(
> > -	struct xfs_inode	*ip,
> > -	int			join_flags)
> > +	struct xfs_inode	*ip)
> >  {
> >  	struct xfs_mount	*mp = ip->i_mount;
> >  	struct xfs_trans	*tp;
> > @@ -1236,8 +1194,7 @@ xfs_ioctl_setattr_get_trans(
> >  		goto out_unlock;
> >  
> >  	xfs_ilock(ip, XFS_ILOCK_EXCL);
> > -	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | join_flags);
> > -	join_flags = 0;
> > +	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
> >  
> >  	/*
> >  	 * CAP_FOWNER overrides the following restrictions:
> > @@ -1258,8 +1215,6 @@ xfs_ioctl_setattr_get_trans(
> >  out_cancel:
> >  	xfs_trans_cancel(tp);
> >  out_unlock:
> > -	if (join_flags)
> > -		xfs_iunlock(ip, join_flags);
> >  	return ERR_PTR(error);
> >  }
> >  
> > @@ -1386,7 +1341,6 @@ xfs_ioctl_setattr(
> >  	struct xfs_dquot	*pdqp = NULL;
> >  	struct xfs_dquot	*olddquot = NULL;
> >  	int			code;
> > -	int			join_flags = 0;
> >  
> >  	trace_xfs_ioctl_setattr(ip);
> >  
> > @@ -1410,18 +1364,9 @@ xfs_ioctl_setattr(
> >  			return code;
> >  	}
> >  
> > -	/*
> > -	 * Changing DAX config may require inode locking for mapping
> > -	 * invalidation. These need to be held all the way to transaction commit
> > -	 * or cancel time, so need to be passed through to
> > -	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
> > -	 * appropriately.
> > -	 */
> > -	code = xfs_ioctl_setattr_dax_invalidate(ip, fa, &join_flags);
> > -	if (code)
> > -		goto error_free_dquots;
> > +	xfs_ioctl_setattr_dax_invalidate(ip, fa);
> >  
> > -	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
> > +	tp = xfs_ioctl_setattr_get_trans(ip);
> >  	if (IS_ERR(tp)) {
> >  		code = PTR_ERR(tp);
> >  		goto error_free_dquots;
> > @@ -1552,7 +1497,6 @@ xfs_ioc_setxflags(
> >  	struct fsxattr		fa;
> >  	struct fsxattr		old_fa;
> >  	unsigned int		flags;
> > -	int			join_flags = 0;
> >  	int			error;
> >  
> >  	if (copy_from_user(&flags, arg, sizeof(flags)))
> > @@ -1569,18 +1513,9 @@ xfs_ioc_setxflags(
> >  	if (error)
> >  		return error;
> >  
> > -	/*
> > -	 * Changing DAX config may require inode locking for mapping
> > -	 * invalidation. These need to be held all the way to transaction commit
> > -	 * or cancel time, so need to be passed through to
> > -	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
> > -	 * appropriately.
> > -	 */
> > -	error = xfs_ioctl_setattr_dax_invalidate(ip, &fa, &join_flags);
> > -	if (error)
> > -		goto out_drop_write;
> > +	xfs_ioctl_setattr_dax_invalidate(ip, &fa);
> >  
> > -	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
> > +	tp = xfs_ioctl_setattr_get_trans(ip);
> >  	if (IS_ERR(tp)) {
> >  		error = PTR_ERR(tp);
> >  		goto out_drop_write;
> > -- 
> > 2.25.1
> > 

Re: [PATCH V8 11/11] Documentation/dax: Update Usage section

[Ira Weiny]

On Wed, Apr 15, 2020 at 08:29:42AM -0700, Darrick J. Wong wrote:
> On Tue, Apr 14, 2020 at 11:45:23PM -0700, ira.weiny@intel.com wrote:
> > From: Ira Weiny <ira.weiny@intel.com>
> > 
> > Update the Usage section to reflect the new individual dax selection
> > functionality.
> > 
> > Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> > 
> > ---
> > Changes from V7:
> > 	Cleanups/clarifications from Darrick and Dan
> > 
> > Changes from V6:
> > 	Update to allow setting FS_XFLAG_DAX any time.
> > 	Update with list of behaviors from Darrick
> > 	https://lore.kernel.org/lkml/20200409165927.GD6741@magnolia/
> > 
> > Changes from V5:
> > 	Update to reflect the agreed upon semantics
> > 	https://lore.kernel.org/lkml/20200405061945.GA94792@iweiny-DESK2.sc.intel.com/
> > ---
> >  Documentation/filesystems/dax.txt | 166 +++++++++++++++++++++++++++++-
> >  1 file changed, 163 insertions(+), 3 deletions(-)
> > 
> > diff --git a/Documentation/filesystems/dax.txt b/Documentation/filesystems/dax.txt
> > index 679729442fd2..893820c53f49 100644
> > --- a/Documentation/filesystems/dax.txt
> > +++ b/Documentation/filesystems/dax.txt
> > @@ -17,11 +17,171 @@ For file mappings, the storage device is mapped directly into userspace.
> >  Usage
> >  -----
> >  
> > -If you have a block device which supports DAX, you can make a filesystem
> > +If you have a block device which supports DAX, you can make a file system
> >  on it as usual.  The DAX code currently only supports files with a block
> >  size equal to your kernel's PAGE_SIZE, so you may need to specify a block
> > -size when creating the filesystem.  When mounting it, use the "-o dax"
> > -option on the command line or add 'dax' to the options in /etc/fstab.
> > +size when creating the file system.
> > +
> > +Currently 3 filesystems support DAX, ext2, ext4 and xfs.  Enabling DAX on them
> 
> "...support DAX: ext2, ext4..."
> 
> Please put a colon after "DAX" since it's not part of the list.
> 
> > +is different.
> > +
> > +Enabling DAX on ext4 and ext2
> > +-----------------------------
> > +
> > +When mounting the filesystem, use the "-o dax" option on the command line or
> > +add 'dax' to the options in /etc/fstab.  This works to enable DAX on all files
> > +within the filesystem.  It is equivalent to the '-o dax=always' behavior below
> > +with the exception that the STATX_ATTR_DAX flag is not supported, nor needed,
> > +as it is always true.
> 
> STATX_ATTR_DAX isn't supported?  I thought ext[24] set S_DAX, so the
> statx flag should work the same as it does on xfs?

yea it will work...

> 
> I also wonder if it's worth mentioning that in the long run ext4 will
> match the xfs semantics, but maybe that's better left for the ext4 rfc
> series.

That was my thought.  Leave ext4 for now.

> 
> > +
> > +
> > +Enabling DAX on xfs
> > +-------------------
> > +
> > +Summary
> > +-------
> > +
> > + 1. There exists an in-kernel file access mode flag S_DAX that corresponds to
> > +    the statx flag STATX_ATTR_DAX.  See the manpage for statx(2) for details
> > +    about this access mode.
> > +
> > + 2. There exists an advisory file inode flag FS_XFLAG_DAX that is
> > +    inherited from the parent directory FS_XFLAG_DAX inode flag at file
> > +    creation time.  This advisory flag can be set or cleared at any
> > +    time, but doing so does not immediately affect the S_DAX state.
> > +
> > +    Unless overridden by mount options (see (3)), if FS_XFLAG_DAX is set
> > +    and the fs is on pmem then it will enable S_DAX at inode load time;
> > +    if FS_XFLAG_DAX is not set, it will not enable S_DAX.
> > +
> > + 3. There exists a dax= mount option.
> > +
> > +    "-o dax=never"  means "never set S_DAX, ignore FS_XFLAG_DAX."
> > +
> > +    "-o dax=always" means "always set S_DAX (at least on pmem),
> > +                    and ignore FS_XFLAG_DAX."
> > +
> > +    "-o dax"        is an alias for "dax=always".
> > +
> > +    "-o dax=inode"  means "follow FS_XFLAG_DAX" and is the default.
> > +
> > + 4. There exists an advisory directory inode flag FS_XFLAG_DAX that can
> > +    be set or cleared at any time.  The flag state is inherited by any files or
> > +    subdirectories when they are created within that directory.
> > +
> > + 5. Programs that require a specific file access mode (DAX or not DAX)
> > +    can do one of the following:
> > +
> > +    (a) Create files in directories that the FS_XFLAG_DAX flag set as
> > +        needed; or
> > +
> > +    (b) Have the administrator set an override via mount option; or
> > +
> > +    (c) Set or clear the file's FS_XFLAG_DAX flag as needed.  Programs
> > +        must then cause the kernel to evict the inode from memory.  This
> > +        can be done by:
> > +
> > +        i>  Closing the file and re-opening the file and using statx to
> > +            see if the fs has changed the S_DAX flag; and
> > +
> > +        ii> If the file still does not have the desired S_DAX access
> > +            mode, either unmount and remount the filesystem, or close
> > +            the file and use drop_caches.
> > +
> > + 6. It is expected that users who want to squeeze every last bit of performance
> > +    out of the particular rough and tumble bits of their storage will also be
> > +    exposed to the difficulties of what happens when the operating system can't
> > +    totally virtualize those hardware capabilities.  DAX is such a feature.
> > +
> > +
> > +Details
> > +-------
> > +
> > +There are 2 per-file dax flags.  One is a physical inode setting (FS_XFLAG_DAX)
> > +and the other a currently enabled state (S_DAX).
> > +
> > +FS_XFLAG_DAX is maintained, on disk, on individual inodes.  It is preserved
> > +within the file system.  This 'physical' config setting can be set using an
> > +ioctl and/or an application such as "xfs_io -c 'chattr [-+]x'".  Files and
> > +directories automatically inherit FS_XFLAG_DAX from their parent directory
> > +_when_ _created_.  Therefore, setting FS_XFLAG_DAX at directory creation time
> > +can be used to set a default behavior for an entire sub-tree.  (Doing so on the
> > +root directory acts to set a default for the entire file system.)
> 
> Urk, I guess I need to push that patch to make mkfs.xfs do this. ;)

I think you should be able to set FS_XFLAG_DAX after mkfs.xfs.  But yea some
people might like the option...  :-D

> 
> > +To clarify inheritance here are 3 examples:
> > +
> > +Example A:
> > +
> > +mkdir -p a/b/c
> > +xfs_io 'chattr +x' a
> > +mkdir a/b/c/d
> > +mkdir a/e
> > +
> > +	dax: a,e
> > +	no dax: b,c,d
> > +
> > +Example B:
> > +
> > +mkdir a
> > +xfs_io 'chattr +x' a
> > +mkdir -p a/b/c/d
> > +
> > +	dax: a,b,c,d
> > +	no dax:
> > +
> > +Example C:
> > +
> > +mkdir -p a/b/c
> > +xfs_io 'chattr +x' c
> > +mkdir a/b/c/d
> > +
> > +	dax: c,d
> > +	no dax: a,b
> > +
> > +
> > +The current enabled state (S_DAX) is set when a file inode is _loaded_ based on
> > +the underlying media support, the value of FS_XFLAG_DAX, and the file systems
> > +dax mount option setting.  See below.
> > +
> > +statx can be used to query S_DAX.  NOTE that a directory will never have S_DAX
> > +set and therefore statx will never indicate that S_DAX is set on directories.
> > +
> > +NOTE: Setting the FS_XFLAG_DAX (specifically or through inheritance) occurs
> > +even if the underlying media does not support dax and/or the file system is
> > +overridden with a mount option.
> > +
> > +
> > +Overriding FS_XFLAG_DAX (dax= mount option)
> > +-------------------------------------------
> > +
> > +There exists a dax mount option.  Using the mount option does not change the
> > +physical configured state of individual files but overrides the S_DAX operating
> > +state when inodes are loaded.
> > +
> > +Given underlying media support, the dax mount option is a tri-state option
> > +(never, always, inode) with the following meanings:
> > +
> > +   "-o dax=never" means "never set S_DAX, ignore FS_XFLAG_DAX"
> > +   "-o dax=always" means "always set S_DAX, ignore FS_XFLAG_DAX"
> > +        "-o dax" by itself means "dax=always" to remain compatible with older
> > +	         kernels
> > +   "-o dax=inode" means "follow FS_XFLAG_DAX"
> > +
> > +The default state is 'inode'.  Given underlying media support, the following
> > +algorithm is used to determine the effective mode of the file S_DAX on a
> > +capable device.
> > +
> > +	S_DAX = FS_XFLAG_DAX;
> > +
> > +	if (dax_mount == "always")
> > +		S_DAX = true;
> > +	else if (dax_mount == "off"
> > +		S_DAX = false;
> 
> The logic in this pseudocode doesn't match the order that's in
> xfs_inode_enable_dax.

I struggled with this a bit.

> I think the outcome is the same, but it's easier
> to verify that if the statements are in roughly the same order.
> 
> 	if dax=never:
> 		S_DAX = false
> 	elif the file system and media don't both support DAX:

The above text does say "Given underlying media support".  Because I wanted to
make that separate.

> 		S_DAX = false
> 	elif dax=always:
> 		S_DAX = true
> 	else:
> 		S_DAX = inode flag status

The above text is trying to convey the "override" nature of the flags.  Where
the code is trying to be a bit more efficient.

Ira

> 
> --D
> 
> > +
> > +To reiterate: Setting, and inheritance, continues to affect FS_XFLAG_DAX even
> > +while the file system is mounted with a dax override.  However, in-core inode
> > +state (S_DAX) will continue to be overridden until the filesystem is remounted
> > +with dax=inode and the inode is evicted."
> >  
> >  
> >  Implementation Tips for Block Driver Writers
> > -- 
> > 2.25.1
> > 

Re: [PATCH V8 04/11] fs/xfs: Change XFS_MOUNT_DAX to XFS_MOUNT_DAX_ALWAYS

[Dave Chinner]

On Tue, Apr 14, 2020 at 11:45:16PM -0700, ira.weiny@intel.com wrote:
> From: Ira Weiny <ira.weiny@intel.com>
> 
> In prep for the new tri-state mount option which then introduces
> XFS_MOUNT_DAX_NEVER.
> 
> Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> ---
>  fs/xfs/xfs_iops.c  | 2 +-
>  fs/xfs/xfs_mount.h | 2 +-
>  fs/xfs/xfs_super.c | 8 ++++----
>  3 files changed, 6 insertions(+), 6 deletions(-)
> 
> diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
> index 81f2f93caec0..a6e634631da8 100644
> --- a/fs/xfs/xfs_iops.c
> +++ b/fs/xfs/xfs_iops.c
> @@ -1248,7 +1248,7 @@ xfs_inode_supports_dax(
>  		return false;
>  
>  	/* DAX mount option or DAX iflag must be set. */
> -	if (!(mp->m_flags & XFS_MOUNT_DAX) &&
> +	if (!(mp->m_flags & XFS_MOUNT_DAX_ALWAYS) &&
>  	    !(ip->i_d.di_flags2 & XFS_DIFLAG2_DAX))
>  		return false;
>  
> diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
> index 88ab09ed29e7..54bd74088936 100644
> --- a/fs/xfs/xfs_mount.h
> +++ b/fs/xfs/xfs_mount.h
> @@ -233,7 +233,7 @@ typedef struct xfs_mount {
>  						   allocator */
>  #define XFS_MOUNT_NOATTR2	(1ULL << 25)	/* disable use of attr2 format */
>  
> -#define XFS_MOUNT_DAX		(1ULL << 62)	/* TEST ONLY! */
> +#define XFS_MOUNT_DAX_ALWAYS	(1ULL << 62)	/* TEST ONLY! */

As this is going to be permanent, please remove the "Test only"
comment and renumber the bits used down to 26. - the high bit was
used only to keep it out of the ranges that permanent mount option
flags used...

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

Re: [PATCH V8 10/11] fs/xfs: Change xfs_ioctl_setattr_dax_invalidate()

[Dave Chinner]

On Tue, Apr 14, 2020 at 11:45:22PM -0700, ira.weiny@intel.com wrote:
> From: Ira Weiny <ira.weiny@intel.com>
> 
> We only support changing FS_XFLAG_DAX on directories.  Files get their
> flag from the parent directory on creation only.  So no data
> invalidation needs to happen.
> 
> Alter the xfs_ioctl_setattr_dax_invalidate() to be
> xfs_ioctl_setattr_dax_validate().  xfs_ioctl_setattr_dax_validate() now
> validates that any FS_XFLAG_DAX change is ok.
> 
> This also allows use to remove the join_flags logic.
> 
> Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> 
> ---
> Changes from V7:
> 	Use new flag_inode_dontcache()
> 	Skip don't cache if mount over ride is active.
> 
> Changes from v6:
> 	Fix completely broken implementation and update commit message.
> 	Use the new VFS layer I_DONTCACHE to facilitate inode eviction
> 	and S_DAX changing on drop_caches
> 
> Changes from v5:
> 	New patch
> ---
>  fs/xfs/xfs_ioctl.c | 105 +++++++++------------------------------------
>  1 file changed, 20 insertions(+), 85 deletions(-)
> 
> diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
> index c6cd92ef4a05..75d4a830ef38 100644
> --- a/fs/xfs/xfs_ioctl.c
> +++ b/fs/xfs/xfs_ioctl.c
> @@ -1145,63 +1145,28 @@ xfs_ioctl_setattr_xflags(
>  }
>  
>  /*
> - * If we are changing DAX flags, we have to ensure the file is clean and any
> - * cached objects in the address space are invalidated and removed. This
> - * requires us to lock out other IO and page faults similar to a truncate
> - * operation. The locks need to be held until the transaction has been committed
> - * so that the cache invalidation is atomic with respect to the DAX flag
> - * manipulation.
> + * Mark inodes with a changing FS_XFLAG_DAX, I_DONTCACHE

That describes what the code is doing, not why.

>   */
> -static int
> +static void
>  xfs_ioctl_setattr_dax_invalidate(
>  	struct xfs_inode	*ip,
> -	struct fsxattr		*fa,
> -	int			*join_flags)
> +	struct fsxattr		*fa)

It's not an invalidation function anymore, so needs a name change.

> -	if (!(fa->fsx_xflags & FS_XFLAG_DAX) && !IS_DAX(inode))
> -		return 0;
> +	struct xfs_mount	*mp = ip->i_mount;
> +	struct inode            *inode = VFS_I(ip);
>  
>  	if (S_ISDIR(inode->i_mode))
> -		return 0;
> -
> -	/* lock, flush and invalidate mapping in preparation for flag change */
> -	xfs_ilock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
> -	error = filemap_write_and_wait(inode->i_mapping);
> -	if (error)
> -		goto out_unlock;
> -	error = invalidate_inode_pages2(inode->i_mapping);
> -	if (error)
> -		goto out_unlock;
> +		return;
>  
> -	*join_flags = XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL;
> -	return 0;
> -
> -out_unlock:
> -	xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
> -	return error;
> +	if (mp->m_flags & XFS_MOUNT_DAX_ALWAYS ||
> +	    mp->m_flags & XFS_MOUNT_DAX_NEVER)
> +		return;

	if (mp->m_flags & (XFS_MOUNT_DAX_ALWAYS | XFS_MOUNT_DAX_NEVER))
		return;
> +	if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
> +	    !(ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)) ||
> +	    (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
> +	     (ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)))
> +		flag_inode_dontcache(inode);

This doesn't set the XFS inode's "don't cache" flag, despite it
having one that serves exactly the same purpose. IOWs, if the
XFS_IDONTCACHE flag is now redundant, please replace it's current
usage with this new flag and get rid of the XFS inode flag. i.e.
the only place we set XFS_IDONTCACHE can be replaced with a call to
this mark_inode_dontcache() call...

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

Re: [PATCH V8 04/11] fs/xfs: Change XFS_MOUNT_DAX to XFS_MOUNT_DAX_ALWAYS

[Ira Weiny]

On Mon, Apr 20, 2020 at 12:15:55PM +1000, Dave Chinner wrote:
> On Tue, Apr 14, 2020 at 11:45:16PM -0700, ira.weiny@intel.com wrote:
> > From: Ira Weiny <ira.weiny@intel.com>
> > 
> > In prep for the new tri-state mount option which then introduces
> > XFS_MOUNT_DAX_NEVER.
> > 
> > Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> > ---
> >  fs/xfs/xfs_iops.c  | 2 +-
> >  fs/xfs/xfs_mount.h | 2 +-
> >  fs/xfs/xfs_super.c | 8 ++++----
> >  3 files changed, 6 insertions(+), 6 deletions(-)
> > 
> > diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
> > index 81f2f93caec0..a6e634631da8 100644
> > --- a/fs/xfs/xfs_iops.c
> > +++ b/fs/xfs/xfs_iops.c
> > @@ -1248,7 +1248,7 @@ xfs_inode_supports_dax(
> >  		return false;
> >  
> >  	/* DAX mount option or DAX iflag must be set. */
> > -	if (!(mp->m_flags & XFS_MOUNT_DAX) &&
> > +	if (!(mp->m_flags & XFS_MOUNT_DAX_ALWAYS) &&
> >  	    !(ip->i_d.di_flags2 & XFS_DIFLAG2_DAX))
> >  		return false;
> >  
> > diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
> > index 88ab09ed29e7..54bd74088936 100644
> > --- a/fs/xfs/xfs_mount.h
> > +++ b/fs/xfs/xfs_mount.h
> > @@ -233,7 +233,7 @@ typedef struct xfs_mount {
> >  						   allocator */
> >  #define XFS_MOUNT_NOATTR2	(1ULL << 25)	/* disable use of attr2 format */
> >  
> > -#define XFS_MOUNT_DAX		(1ULL << 62)	/* TEST ONLY! */
> > +#define XFS_MOUNT_DAX_ALWAYS	(1ULL << 62)	/* TEST ONLY! */
> 
> As this is going to be permanent, please remove the "Test only"

I did that in the next patch...  But I'll move it here.

> comment and renumber the bits used down to 26. - the high bit was
> used only to keep it out of the ranges that permanent mount option
> flags used...

Ok, done for v9

Ira

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

Re: [PATCH V8 10/11] fs/xfs: Change xfs_ioctl_setattr_dax_invalidate()

[Ira Weiny]

On Mon, Apr 20, 2020 at 12:31:31PM +1000, Dave Chinner wrote:
> On Tue, Apr 14, 2020 at 11:45:22PM -0700, ira.weiny@intel.com wrote:
> > From: Ira Weiny <ira.weiny@intel.com>
> > 
> > We only support changing FS_XFLAG_DAX on directories.  Files get their
> > flag from the parent directory on creation only.  So no data
> > invalidation needs to happen.
> > 
> > Alter the xfs_ioctl_setattr_dax_invalidate() to be
> > xfs_ioctl_setattr_dax_validate().  xfs_ioctl_setattr_dax_validate() now
> > validates that any FS_XFLAG_DAX change is ok.
> > 
> > This also allows use to remove the join_flags logic.
> > 
> > Signed-off-by: Ira Weiny <ira.weiny@intel.com>
> > 
> > ---
> > Changes from V7:
> > 	Use new flag_inode_dontcache()
> > 	Skip don't cache if mount over ride is active.
> > 
> > Changes from v6:
> > 	Fix completely broken implementation and update commit message.
> > 	Use the new VFS layer I_DONTCACHE to facilitate inode eviction
> > 	and S_DAX changing on drop_caches
> > 
> > Changes from v5:
> > 	New patch
> > ---
> >  fs/xfs/xfs_ioctl.c | 105 +++++++++------------------------------------
> >  1 file changed, 20 insertions(+), 85 deletions(-)
> > 
> > diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
> > index c6cd92ef4a05..75d4a830ef38 100644
> > --- a/fs/xfs/xfs_ioctl.c
> > +++ b/fs/xfs/xfs_ioctl.c
> > @@ -1145,63 +1145,28 @@ xfs_ioctl_setattr_xflags(
> >  }
> >  
> >  /*
> > - * If we are changing DAX flags, we have to ensure the file is clean and any
> > - * cached objects in the address space are invalidated and removed. This
> > - * requires us to lock out other IO and page faults similar to a truncate
> > - * operation. The locks need to be held until the transaction has been committed
> > - * so that the cache invalidation is atomic with respect to the DAX flag
> > - * manipulation.
> > + * Mark inodes with a changing FS_XFLAG_DAX, I_DONTCACHE
> 
> That describes what the code is doing, not why.

I'll remove the comment.

> 
> >   */
> > -static int
> > +static void
> >  xfs_ioctl_setattr_dax_invalidate(
> >  	struct xfs_inode	*ip,
> > -	struct fsxattr		*fa,
> > -	int			*join_flags)
> > +	struct fsxattr		*fa)
> 
> It's not an invalidation function anymore, so needs a name change.

Done.

> 
> > -	if (!(fa->fsx_xflags & FS_XFLAG_DAX) && !IS_DAX(inode))
> > -		return 0;
> > +	struct xfs_mount	*mp = ip->i_mount;
> > +	struct inode            *inode = VFS_I(ip);
> >  
> >  	if (S_ISDIR(inode->i_mode))
> > -		return 0;
> > -
> > -	/* lock, flush and invalidate mapping in preparation for flag change */
> > -	xfs_ilock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
> > -	error = filemap_write_and_wait(inode->i_mapping);
> > -	if (error)
> > -		goto out_unlock;
> > -	error = invalidate_inode_pages2(inode->i_mapping);
> > -	if (error)
> > -		goto out_unlock;
> > +		return;
> >  
> > -	*join_flags = XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL;
> > -	return 0;
> > -
> > -out_unlock:
> > -	xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
> > -	return error;
> > +	if (mp->m_flags & XFS_MOUNT_DAX_ALWAYS ||
> > +	    mp->m_flags & XFS_MOUNT_DAX_NEVER)
> > +		return;
> 
> 	if (mp->m_flags & (XFS_MOUNT_DAX_ALWAYS | XFS_MOUNT_DAX_NEVER))
> 		return;
> > +	if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
> > +	    !(ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)) ||
> > +	    (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
> > +	     (ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)))
> > +		flag_inode_dontcache(inode);
> 
> This doesn't set the XFS inode's "don't cache" flag, despite it
> having one that serves exactly the same purpose.  IOWs, if the XFS_IDONTCACHE
> flag is now redundant, please replace it's current usage with this new flag
> and get rid of the XFS inode flag. i.e.  the only place we set XFS_IDONTCACHE
> can be replaced with a call to this mark_inode_dontcache() call...

I agree, and I would have removed XFS_IDONTCACHE, except I was not convinced
that XFS_IDONTCACHE was redundant.

Currently XFS_IDONTCACHE can be cleared if the inode is found in the cache and
I was unable to convince myself that it would be ok to remove it.  I mentioned
this to Darrick in V7.

https://lore.kernel.org/lkml/20200413194432.GD1649878@iweiny-DESK2.sc.intel.com/

What am I missing with this code?

xfs_iget_cache_hit():
...
        if (!(flags & XFS_IGET_INCORE))
		xfs_iflags_clear(ip, XFS_ISTALE | XFS_IDONTCACHE);
...

Why is XFS_IDONTCACHE not 'sticky'?  And why does xfs_iget_cache_hit() clear it
rather than fail when XFS_IDONTCACHE is set?

Ira

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

Re: [PATCH V8 10/11] fs/xfs: Change xfs_ioctl_setattr_dax_invalidate()

[Dave Chinner]

On Mon, Apr 20, 2020 at 11:36:17AM -0700, Ira Weiny wrote:
> On Mon, Apr 20, 2020 at 12:31:31PM +1000, Dave Chinner wrote:
> > On Tue, Apr 14, 2020 at 11:45:22PM -0700, ira.weiny@intel.com wrote:
> > > From: Ira Weiny <ira.weiny@intel.com>
> > > -out_unlock:
> > > -	xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
> > > -	return error;
> > > +	if (mp->m_flags & XFS_MOUNT_DAX_ALWAYS ||
> > > +	    mp->m_flags & XFS_MOUNT_DAX_NEVER)
> > > +		return;
> > 
> > 	if (mp->m_flags & (XFS_MOUNT_DAX_ALWAYS | XFS_MOUNT_DAX_NEVER))
> > 		return;
> > > +	if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
> > > +	    !(ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)) ||
> > > +	    (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
> > > +	     (ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)))
> > > +		flag_inode_dontcache(inode);
> > 
> > This doesn't set the XFS inode's "don't cache" flag, despite it
> > having one that serves exactly the same purpose.  IOWs, if the XFS_IDONTCACHE
> > flag is now redundant, please replace it's current usage with this new flag
> > and get rid of the XFS inode flag. i.e.  the only place we set XFS_IDONTCACHE
> > can be replaced with a call to this mark_inode_dontcache() call...
> 
> I agree, and I would have removed XFS_IDONTCACHE, except I was not convinced
> that XFS_IDONTCACHE was redundant.
> 
> Currently XFS_IDONTCACHE can be cleared if the inode is found in the cache and
> I was unable to convince myself that it would be ok to remove it.  I mentioned
> this to Darrick in V7.
> 
> https://lore.kernel.org/lkml/20200413194432.GD1649878@iweiny-DESK2.sc.intel.com/
> 
> What am I missing with this code?
> 
> xfs_iget_cache_hit():
> ...
>         if (!(flags & XFS_IGET_INCORE))
> 		xfs_iflags_clear(ip, XFS_ISTALE | XFS_IDONTCACHE);
> ...
> 
> Why is XFS_IDONTCACHE not 'sticky'?
> And why does xfs_iget_cache_hit() clear it

Because it was designed to do exactly what bulkstat required, and
nothing else.  xfs_iget() is an internal filesystem interface, not a
VFS level interface. Hence we can make up whatever semantics we
want. And if we get a cache hit, we have multiple references to the
inode so we probably should cache it regardless of whether the
original lookup said "I'm a one-shot wonder, so don't cache me".

IOWs, it's a classic "don't cache unless a second reference comes
along during the current life cycle" algorithm.

This isn't actually a frequently travelled path - bulkstat is a
pretty rare thing to be doing - so the behaviour is "be nice to the
cache because we can do it easily", not a hard requirement.

> rather than fail when XFS_IDONTCACHE is set?

Because then it would be impossible to access an inode that has
IDONTCACHE set on it. e.g. bulkstat an inode, now you can't open()
it because it has XFS_IDONTCACHE set and VFS pathwalk lookups fail
trying to resolve the inode number to a struct inode....

Same goes for I_DONTCACHE - this does not prevent new lookups from
taking references to the inode while it is still referenced. i.e.
the reference count can still go up after the flag is set. The flag
only takes effect when the reference count goes to zero.

Hence the only difference between XFS_IDONTCACHE and I_DONTCACHE is
the behaviour when cache hits on existing XFS_IDONTCACHE inodes
occur. It's not going to make a significant difference to cache
residency if we leave the I_DONTCACHE flag in place, because the
vast majority of inodes with that flag (from bulkstat) are still
one-shot wonders and hence the reclaim decision is still the
overwhelmingly correct decision to be making...

And, realistically, we have a second layer of inode caching in XFS
(the cluster buffers) and so it's likely if we evict and reclaim an
inode just before it gets re-used, then we'll hit the buffer cache
anyway. i.e. we still avoid the IO to read the inode back into
memory, we just burn a little more CPU re-instantiating it from the
buffer....

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com