Date:	Wed, 26 Nov 2014 05:23:52 -0500
From:	Theodore Ts'o <tytso@....edu>
To:	Linux Filesystem Development List <linux-fsdevel@...r.kernel.org>
Cc:	Ext4 Developers List <linux-ext4@...r.kernel.org>,
	Linux btrfs Developers List <linux-btrfs@...r.kernel.org>,
	XFS Developers <xfs@....sgi.com>, Theodore Ts'o <tytso@....edu>
Subject: [PATCH-v4 2/7] vfs: add support for a lazytime mount option

Add a new mount option which enables a new "lazytime" mode.  This mode
causes atime, mtime, and ctime updates to only be made to the
in-memory version of the inode.  The on-disk times will only get
updated when (a) if the inode needs to be updated for some non-time
related change, (b) if userspace calls fsync(), syncfs() or sync(), or
(c) just before an undeleted inode is evicted from memory.

This is OK according to POSIX because there are no guarantees after a
crash unless userspace explicitly requests via a fsync(2) call.

For workloads which feature a large number of random write to a
preallocated file, the lazytime mount option significantly reduces
writes to the inode table.  The repeated 4k writes to a single block
will result in undesirable stress on flash devices and SMR disk
drives.  Even on conventional HDD's, the repeated writes to the inode
table block will trigger Adjacent Track Interference (ATI) remediation
latencies, which very negatively impact 99.9 percentile latencies ---
which is a very big deal for web serving tiers (for example).

Google-Bug-Id: 18297052

Signed-off-by: Theodore Ts'o <tytso@....edu>
---
 fs/fs-writeback.c           | 55 ++++++++++++++++++++++++++++++++++++++++-----
 fs/inode.c                  | 43 ++++++++++++++++++++++++++++++-----
 fs/proc_namespace.c         |  1 +
 fs/sync.c                   |  8 +++++++
 include/linux/backing-dev.h |  1 +
 include/linux/fs.h          | 11 +++++++--
 include/uapi/linux/fs.h     |  1 +
 mm/backing-dev.c            |  9 ++++++--
 8 files changed, 113 insertions(+), 16 deletions(-)

diff --git a/fs/fs-writeback.c b/fs/fs-writeback.c
index ef9bef1..ef8c5d8 100644
--- a/fs/fs-writeback.c
+++ b/fs/fs-writeback.c
@@ -397,7 +397,7 @@ static void requeue_inode(struct inode *inode, struct bdi_writeback *wb,
 	 * shot. If still dirty, it will be redirty_tail()'ed below.  Update
 	 * the dirty time to prevent enqueue and sync it again.
 	 */
-	if ((inode->i_state & I_DIRTY) &&
+	if ((inode->i_state & I_DIRTY_WB) &&
 	    (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages))
 		inode->dirtied_when = jiffies;
 
@@ -428,13 +428,15 @@ static void requeue_inode(struct inode *inode, struct bdi_writeback *wb,
 			 */
 			redirty_tail(inode, wb);
 		}
-	} else if (inode->i_state & I_DIRTY) {
+	} else if (inode->i_state & I_DIRTY_WB) {
 		/*
 		 * Filesystems can dirty the inode during writeback operations,
 		 * such as delayed allocation during submission or metadata
 		 * updates after data IO completion.
 		 */
 		redirty_tail(inode, wb);
+	} else if (inode->i_state & I_DIRTY_TIME) {
+		list_move(&inode->i_wb_list, &wb->b_dirty_time);
 	} else {
 		/* The inode is clean. Remove from writeback lists. */
 		list_del_init(&inode->i_wb_list);
@@ -482,11 +484,11 @@ __writeback_single_inode(struct inode *inode, struct writeback_control *wbc)
 	/* Clear I_DIRTY_PAGES if we've written out all dirty pages */
 	if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
 		inode->i_state &= ~I_DIRTY_PAGES;
-	dirty = inode->i_state & I_DIRTY;
-	inode->i_state &= ~(I_DIRTY_SYNC | I_DIRTY_DATASYNC);
+	dirty = inode->i_state & I_DIRTY_INODE;
+	inode->i_state &= ~I_DIRTY_INODE;
 	spin_unlock(&inode->i_lock);
 	/* Don't write the inode if only I_DIRTY_PAGES was set */
-	if (dirty & (I_DIRTY_SYNC | I_DIRTY_DATASYNC)) {
+	if (dirty) {
 		int err = write_inode(inode, wbc);
 		if (ret == 0)
 			ret = err;
@@ -1162,7 +1164,7 @@ void __mark_inode_dirty(struct inode *inode, int flags)
 
 	spin_lock(&inode->i_lock);
 	if ((inode->i_state & flags) != flags) {
-		const int was_dirty = inode->i_state & I_DIRTY;
+		const int was_dirty = inode->i_state & I_DIRTY_WB;
 
 		inode->i_state |= flags;
 
@@ -1224,6 +1226,24 @@ out_unlock_inode:
 }
 EXPORT_SYMBOL(__mark_inode_dirty);
 
+void inode_requeue_dirtytime(struct inode *inode)
+{
+	struct backing_dev_info *bdi = inode_to_bdi(inode);
+
+	spin_lock(&bdi->wb.list_lock);
+	spin_lock(&inode->i_lock);
+	if ((inode->i_state & I_DIRTY_WB) == 0) {
+		if (inode->i_state & I_DIRTY_TIME)
+			list_move(&inode->i_wb_list, &bdi->wb.b_dirty_time);
+		else
+			list_del_init(&inode->i_wb_list);
+	}
+	spin_unlock(&inode->i_lock);
+	spin_unlock(&bdi->wb.list_lock);
+
+}
+EXPORT_SYMBOL(inode_requeue_dirtytime);
+
 static void wait_sb_inodes(struct super_block *sb)
 {
 	struct inode *inode, *old_inode = NULL;
@@ -1277,6 +1297,28 @@ static void wait_sb_inodes(struct super_block *sb)
 	iput(old_inode);
 }
 
+/*
+ * Take all of the indoes on the dirty_time list, and mark them as
+ * dirty, so they will be written out.
+ */
+static void flush_sb_dirty_time(struct super_block *sb)
+{
+	struct bdi_writeback *wb = &sb->s_bdi->wb;
+	LIST_HEAD(tmp);
+
+	spin_lock(&wb->list_lock);
+	list_cut_position(&tmp, &wb->b_dirty_time, wb->b_dirty_time.prev);
+	while (!list_empty(&tmp)) {
+		struct inode *inode = wb_inode(tmp.prev);
+
+		list_del_init(&inode->i_wb_list);
+		spin_unlock(&wb->list_lock);
+		mark_inode_dirty_sync(inode);
+		spin_lock(&wb->list_lock);
+	}
+	spin_unlock(&wb->list_lock);
+}
+
 /**
  * writeback_inodes_sb_nr -	writeback dirty inodes from given super_block
  * @sb: the superblock
@@ -1388,6 +1430,7 @@ void sync_inodes_sb(struct super_block *sb)
 		return;
 	WARN_ON(!rwsem_is_locked(&sb->s_umount));
 
+	flush_sb_dirty_time(sb);
 	bdi_queue_work(sb->s_bdi, &work);
 	wait_for_completion(&done);
 
diff --git a/fs/inode.c b/fs/inode.c
index 8f5c4b5..9e464cc 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -30,7 +30,7 @@
  * inode_sb_list_lock protects:
  *   sb->s_inodes, inode->i_sb_list
  * bdi->wb.list_lock protects:
- *   bdi->wb.b_{dirty,io,more_io}, inode->i_wb_list
+ *   bdi->wb.b_{dirty,io,more_io,dirty_time}, inode->i_wb_list
  * inode_hash_lock protects:
  *   inode_hashtable, inode->i_hash
  *
@@ -1430,11 +1430,22 @@ static void iput_final(struct inode *inode)
  */
 void iput(struct inode *inode)
 {
-	if (inode) {
-		BUG_ON(inode->i_state & I_CLEAR);
-
-		if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock))
-			iput_final(inode);
+	if (!inode)
+		return;
+	BUG_ON(inode->i_state & I_CLEAR);
+retry:
+	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);
+			if (inode->i_op->write_time)
+				inode->i_op->write_time(inode);
+			else if (inode->i_sb->s_op->write_inode)
+				mark_inode_dirty_sync(inode);
+			goto retry;
+		}
+		iput_final(inode);
 	}
 }
 EXPORT_SYMBOL(iput);
@@ -1515,6 +1526,26 @@ static int update_time(struct inode *inode, struct timespec *time, int flags)
 		if (flags & S_MTIME)
 			inode->i_mtime = *time;
 	}
+	if ((inode->i_sb->s_flags & MS_LAZYTIME) &&
+	    !(flags & S_VERSION) &&
+	    !(inode->i_state & (I_DIRTY_SYNC | I_DIRTY_DATASYNC))) {
+		if (inode->i_state & I_DIRTY_TIME)
+			return 0;
+		spin_lock(&inode->i_lock);
+		if (inode->i_state & (I_DIRTY_SYNC | I_DIRTY_DATASYNC)) {
+			spin_unlock(&inode->i_lock);
+			goto force_dirty;
+		}
+		if (inode->i_state & I_DIRTY_TIME) {
+			spin_unlock(&inode->i_lock);
+			return 0;
+		}
+		inode->i_state |= I_DIRTY_TIME;
+		spin_unlock(&inode->i_lock);
+		inode_requeue_dirtytime(inode);
+		return 0;
+	}
+force_dirty:
 	if (inode->i_op->write_time)
 		return inode->i_op->write_time(inode);
 	mark_inode_dirty_sync(inode);
diff --git a/fs/proc_namespace.c b/fs/proc_namespace.c
index 73ca174..f98234a 100644
--- a/fs/proc_namespace.c
+++ b/fs/proc_namespace.c
@@ -44,6 +44,7 @@ static int show_sb_opts(struct seq_file *m, struct super_block *sb)
 		{ MS_SYNCHRONOUS, ",sync" },
 		{ MS_DIRSYNC, ",dirsync" },
 		{ MS_MANDLOCK, ",mand" },
+		{ MS_LAZYTIME, ",lazytime" },
 		{ 0, NULL }
 	};
 	const struct proc_fs_info *fs_infop;
diff --git a/fs/sync.c b/fs/sync.c
index bdc729d..6ac7bf0 100644
--- a/fs/sync.c
+++ b/fs/sync.c
@@ -177,8 +177,16 @@ SYSCALL_DEFINE1(syncfs, int, fd)
  */
 int vfs_fsync_range(struct file *file, loff_t start, loff_t end, int datasync)
 {
+	struct inode *inode = file->f_mapping->host;
+
 	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);
+		mark_inode_dirty_sync(inode);
+	}
 	return file->f_op->fsync(file, start, end, datasync);
 }
 EXPORT_SYMBOL(vfs_fsync_range);
diff --git a/include/linux/backing-dev.h b/include/linux/backing-dev.h
index 5da6012..4cdf733 100644
--- a/include/linux/backing-dev.h
+++ b/include/linux/backing-dev.h
@@ -55,6 +55,7 @@ struct bdi_writeback {
 	struct list_head b_dirty;	/* dirty inodes */
 	struct list_head b_io;		/* parked for writeback */
 	struct list_head b_more_io;	/* parked for more writeback */
+	struct list_head b_dirty_time;	/* time stamps are dirty */
 	spinlock_t list_lock;		/* protects the b_* lists */
 };
 
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 3633239..55cf34d 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1721,19 +1721,26 @@ struct super_operations {
 #define __I_DIO_WAKEUP		9
 #define I_DIO_WAKEUP		(1 << I_DIO_WAKEUP)
 #define I_LINKABLE		(1 << 10)
+#define I_DIRTY_TIME		(1 << 11)
 
-#define I_DIRTY (I_DIRTY_SYNC | I_DIRTY_DATASYNC | I_DIRTY_PAGES)
+/* Inode should be on the b_dirty/b_io/b_more_io lists */
+#define I_DIRTY_WB (I_DIRTY_SYNC | I_DIRTY_DATASYNC | I_DIRTY_PAGES)
+/* Inode should be on the b_dirty/b_io/b_more_io/b_dirty_time lists */
+#define I_DIRTY (I_DIRTY_SYNC | I_DIRTY_DATASYNC | I_DIRTY_PAGES | I_DIRTY_TIME)
+/* The inode itself is dirty  */
+#define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC | I_DIRTY_TIME)
 
 extern void __mark_inode_dirty(struct inode *, int);
 static inline void mark_inode_dirty(struct inode *inode)
 {
-	__mark_inode_dirty(inode, I_DIRTY);
+	__mark_inode_dirty(inode, I_DIRTY_WB);
 }
 
 static inline void mark_inode_dirty_sync(struct inode *inode)
 {
 	__mark_inode_dirty(inode, I_DIRTY_SYNC);
 }
+extern void inode_requeue_dirtytime(struct inode *);
 
 extern void inc_nlink(struct inode *inode);
 extern void drop_nlink(struct inode *inode);
diff --git a/include/uapi/linux/fs.h b/include/uapi/linux/fs.h
index 3735fa0..cc9713a 100644
--- a/include/uapi/linux/fs.h
+++ b/include/uapi/linux/fs.h
@@ -90,6 +90,7 @@ struct inodes_stat_t {
 #define MS_KERNMOUNT	(1<<22) /* this is a kern_mount call */
 #define MS_I_VERSION	(1<<23) /* Update inode I_version field */
 #define MS_STRICTATIME	(1<<24) /* Always perform atime updates */
+#define MS_LAZYTIME	(1<<25) /* Update the on-disk [acm]times lazily */
 
 /* These sb flags are internal to the kernel */
 #define MS_NOSEC	(1<<28)
diff --git a/mm/backing-dev.c b/mm/backing-dev.c
index 0ae0df5..14851fe 100644
--- a/mm/backing-dev.c
+++ b/mm/backing-dev.c
@@ -69,10 +69,10 @@ static int bdi_debug_stats_show(struct seq_file *m, void *v)
 	unsigned long background_thresh;
 	unsigned long dirty_thresh;
 	unsigned long bdi_thresh;
-	unsigned long nr_dirty, nr_io, nr_more_io;
+	unsigned long nr_dirty, nr_io, nr_more_io, nr_dirty_time;
 	struct inode *inode;
 
-	nr_dirty = nr_io = nr_more_io = 0;
+	nr_dirty = nr_io = nr_more_io = nr_dirty_time = 0;
 	spin_lock(&wb->list_lock);
 	list_for_each_entry(inode, &wb->b_dirty, i_wb_list)
 		nr_dirty++;
@@ -80,6 +80,8 @@ static int bdi_debug_stats_show(struct seq_file *m, void *v)
 		nr_io++;
 	list_for_each_entry(inode, &wb->b_more_io, i_wb_list)
 		nr_more_io++;
+	list_for_each_entry(inode, &wb->b_dirty_time, i_wb_list)
+		nr_dirty_time++;
 	spin_unlock(&wb->list_lock);
 
 	global_dirty_limits(&background_thresh, &dirty_thresh);
@@ -98,6 +100,7 @@ static int bdi_debug_stats_show(struct seq_file *m, void *v)
 		   "b_dirty:            %10lu\n"
 		   "b_io:               %10lu\n"
 		   "b_more_io:          %10lu\n"
+		   "b_dirty_time:       %10lu\n"
 		   "bdi_list:           %10u\n"
 		   "state:              %10lx\n",
 		   (unsigned long) K(bdi_stat(bdi, BDI_WRITEBACK)),
@@ -111,6 +114,7 @@ static int bdi_debug_stats_show(struct seq_file *m, void *v)
 		   nr_dirty,
 		   nr_io,
 		   nr_more_io,
+		   nr_dirty_time,
 		   !list_empty(&bdi->bdi_list), bdi->state);
 #undef K
 
@@ -418,6 +422,7 @@ static void bdi_wb_init(struct bdi_writeback *wb, struct backing_dev_info *bdi)
 	INIT_LIST_HEAD(&wb->b_dirty);
 	INIT_LIST_HEAD(&wb->b_io);
 	INIT_LIST_HEAD(&wb->b_more_io);
+	INIT_LIST_HEAD(&wb->b_dirty_time);
 	spin_lock_init(&wb->list_lock);
 	INIT_DELAYED_WORK(&wb->dwork, bdi_writeback_workfn);
 }
-- 
2.1.0

--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Powered by blists - more mailing lists