lazytime for XFS V2

lazytime for XFS V2

[Christoph Hellwig]

Hi all,

this series implements the lazytime feature for XFS.  Unlike ext4 we don't
bother supporting the file system parsed text option as MS_LAZYTIME has
been supported by util-linux for a long time now.  The first patch is
needed to make sure XFS actually sees that a dirty inode was dirtied due
to lazy time, which is important for XFS as we don't otherwise use the
VFS inode dirty tracking for metadata.

Changes since V1:
 - fix i_version logging in xfs_vn_update_time,
 - document the racy I_DIRTY_TIME check

[PATCH 1/2] fs: don't clear I_DIRTY_TIME before calling mark_inode_dirty_sync

[Christoph Hellwig]

__mark_inode_dirty already takes care of that, and for the XFS lazytime
implementation we need to know that ->dirty_inode was called because
I_DIRTY_TIME was set.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Jan Kara <jack@suse.cz>
---
 fs/inode.c | 1 -
 fs/sync.c  | 6 +-----
 2 files changed, 1 insertion(+), 6 deletions(-)

diff --git a/fs/inode.c b/fs/inode.c
index ef362364d396..6295f1415761 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -1533,7 +1533,6 @@ void iput(struct inode *inode)
 	if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock)) {
 		if (inode->i_nlink && (inode->i_state & I_DIRTY_TIME)) {
 			atomic_inc(&inode->i_count);
-			inode->i_state &= ~I_DIRTY_TIME;
 			spin_unlock(&inode->i_lock);
 			trace_writeback_lazytime_iput(inode);
 			mark_inode_dirty_sync(inode);
diff --git a/fs/sync.c b/fs/sync.c
index 6e0a2cbaf6de..09f0259724ff 100644
--- a/fs/sync.c
+++ b/fs/sync.c
@@ -187,12 +187,8 @@ int vfs_fsync_range(struct file *file, loff_t start, loff_t end, int datasync)
 
 	if (!file->f_op->fsync)
 		return -EINVAL;
-	if (!datasync && (inode->i_state & I_DIRTY_TIME)) {
-		spin_lock(&inode->i_lock);
-		inode->i_state &= ~I_DIRTY_TIME;
-		spin_unlock(&inode->i_lock);
+	if (!datasync && (inode->i_state & I_DIRTY_TIME))
 		mark_inode_dirty_sync(inode);
-	}
 	return file->f_op->fsync(file, start, end, datasync);
 }
 EXPORT_SYMBOL(vfs_fsync_range);
-- 
2.14.2

[PATCH 2/2] xfs: implement the lazytime mount option

[Christoph Hellwig]

Use the VFS dirty inode tracking for lazytime inodes only, and just
log them in ->dirty_inode.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/xfs/xfs_iops.c        | 13 ++++++++++++-
 fs/xfs/xfs_super.c       | 23 +++++++++++++++++++++++
 fs/xfs/xfs_trans_inode.c | 14 ++++++++++++++
 3 files changed, 49 insertions(+), 1 deletion(-)

diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
index 56475fcd76f2..8567951eff10 100644
--- a/fs/xfs/xfs_iops.c
+++ b/fs/xfs/xfs_iops.c
@@ -46,6 +46,7 @@
 #include <linux/security.h>
 #include <linux/iomap.h>
 #include <linux/slab.h>
+#include <linux/iversion.h>
 
 /*
  * Directories have different lock order w.r.t. mmap_sem compared to regular
@@ -1052,11 +1053,21 @@ xfs_vn_update_time(
 {
 	struct xfs_inode	*ip = XFS_I(inode);
 	struct xfs_mount	*mp = ip->i_mount;
+	int			log_flags = XFS_ILOG_TIMESTAMP;
 	struct xfs_trans	*tp;
 	int			error;
 
 	trace_xfs_update_time(ip);
 
+	if (inode->i_sb->s_flags & SB_LAZYTIME) {
+		if (!((flags & S_VERSION) &&
+		      inode_maybe_inc_iversion(inode, false)))
+			return generic_update_time(inode, now, flags);
+
+		/* Capture the iversion update that just occurred */
+		log_flags |= XFS_ILOG_CORE;
+	}
+
 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_fsyncts, 0, 0, 0, &tp);
 	if (error)
 		return error;
@@ -1070,7 +1081,7 @@ xfs_vn_update_time(
 		inode->i_atime = *now;
 
 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
-	xfs_trans_log_inode(tp, ip, XFS_ILOG_TIMESTAMP);
+	xfs_trans_log_inode(tp, ip, log_flags);
 	return xfs_trans_commit(tp);
 }
 
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index 7aba628dc527..20d53f125c35 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -1007,6 +1007,28 @@ xfs_fs_destroy_inode(
 	xfs_inode_set_reclaim_tag(ip);
 }
 
+static void
+xfs_fs_dirty_inode(
+	struct inode			*inode,
+	int				flag)
+{
+	struct xfs_inode		*ip = XFS_I(inode);
+	struct xfs_mount		*mp = ip->i_mount;
+	struct xfs_trans		*tp;
+
+	if (!(inode->i_sb->s_flags & SB_LAZYTIME))
+		return;
+	if (flag != I_DIRTY_SYNC || !(inode->i_state & I_DIRTY_TIME))
+		return;
+
+	if (xfs_trans_alloc(mp, &M_RES(mp)->tr_fsyncts, 0, 0, 0, &tp))
+		return;
+	xfs_ilock(ip, XFS_ILOCK_EXCL);
+	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
+	xfs_trans_log_inode(tp, ip, XFS_ILOG_TIMESTAMP);
+	xfs_trans_commit(tp);
+}
+
 /*
  * Slab object creation initialisation for the XFS inode.
  * This covers only the idempotent fields in the XFS inode;
@@ -1787,6 +1809,7 @@ xfs_fs_free_cached_objects(
 static const struct super_operations xfs_super_operations = {
 	.alloc_inode		= xfs_fs_alloc_inode,
 	.destroy_inode		= xfs_fs_destroy_inode,
+	.dirty_inode		= xfs_fs_dirty_inode,
 	.drop_inode		= xfs_fs_drop_inode,
 	.put_super		= xfs_fs_put_super,
 	.sync_fs		= xfs_fs_sync_fs,
diff --git a/fs/xfs/xfs_trans_inode.c b/fs/xfs/xfs_trans_inode.c
index fddacf9575df..bbb596860bf8 100644
--- a/fs/xfs/xfs_trans_inode.c
+++ b/fs/xfs/xfs_trans_inode.c
@@ -98,9 +98,23 @@ xfs_trans_log_inode(
 	xfs_inode_t	*ip,
 	uint		flags)
 {
+	struct inode	*inode = VFS_I(ip);
+
 	ASSERT(ip->i_itemp != NULL);
 	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
 
+	/*
+	 * Don't bother with i_lock for the I_DIRTY_TIME check here, as races
+	 * don't matter - we either will need an extra transaction in 24 hours
+	 * to log the timestamps, or will clear already cleared fields in the
+	 * worst case.
+	 */
+	if (inode->i_state & (I_DIRTY_TIME | I_DIRTY_TIME_EXPIRED)) {
+		spin_lock(&inode->i_lock);
+		inode->i_state &= ~(I_DIRTY_TIME | I_DIRTY_TIME_EXPIRED);
+		spin_unlock(&inode->i_lock);
+	}
+
 	/*
 	 * First time we log the inode in a transaction, bump the inode change
 	 * counter if it is configured for this to occur. While we have the
-- 
2.14.2

Re: [PATCH 2/2] xfs: implement the lazytime mount option

[Dave Chinner]

On Wed, Feb 28, 2018 at 09:57:37AM -0800, Christoph Hellwig wrote:
> Use the VFS dirty inode tracking for lazytime inodes only, and just
> log them in ->dirty_inode.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>

Works fine. I can't think of a cleaner way to implement this, so it
looks good to me.

Reviewed-by: Dave Chinner <dchinner@redhat.com>
-- 
Dave Chinner
david@fromorbit.com