Re: [Bug 15906] serious performance regression in "umount" on ext4 over LVM

[Jens Axboe <jens.axboe@oracle.com>]

On Wed, May 05 2010, Dmitry Monakhov wrote:
> bugzilla-daemon@bugzilla.kernel.org writes:
> 
> Hi Jens,
> 
> Just FYI, we have found a regression which was caused by your famous
> writeback patch 03ba3782e8dcc5b0e1efe440d33084f066e38cae
> I'm not allowed to add you to CC in BZ, that's why i wrote this mail.

You need to use axboe@kernel.dk as that is what I use there.

> Before the patch __sync_filesystem() called writeback_single_inode()
> directly, but now it is called indirectly (from flush-X:X task)
> which require a super_block in question to be pinned.
> But this is impossible to pin this SB on umount because we already
> hold s_umount sem for write, so effectively we already pinned that SB.
> So my proposal is to treat umount similar to WB_SYNC_ALL, and skip
> pining stage.

Hmm I see, yes that is a bug. How about adding a WB_SYNC_NONE_PINNED or
something to that effect, which acts like WB_SYNC_NONE but the caller is
required to hold the s_umount already (like WB_SYNC_ALL).

Something like the below. It compiles, but not tested.

diff --git a/fs/fs-writeback.c b/fs/fs-writeback.c
index 4b37f7c..4327465 100644
--- a/fs/fs-writeback.c
+++ b/fs/fs-writeback.c
@@ -245,19 +245,20 @@ static void bdi_sync_writeback(struct backing_dev_info *bdi,
  * @bdi: the backing device to write from
  * @sb: write inodes from this super_block
  * @nr_pages: the number of pages to write
+ * @sb_locked: caller already holds sb umount sem.
  *
  * Description:
  *   This does WB_SYNC_NONE opportunistic writeback. The IO is only
  *   started when this function returns, we make no guarentees on
- *   completion. Caller need not hold sb s_umount semaphore.
+ *   completion. Caller specifies whether sb umount sem is held already or not.
  *
  */
 void bdi_start_writeback(struct backing_dev_info *bdi, struct super_block *sb,
-			 long nr_pages)
+			 long nr_pages, int sb_locked)
 {
 	struct wb_writeback_args args = {
 		.sb		= sb,
-		.sync_mode	= WB_SYNC_NONE,
+		.sync_mode	= sb_locked ? WB_SYNC_NONE_PIN : WB_SYNC_NONE,
 		.nr_pages	= nr_pages,
 		.range_cyclic	= 1,
 	};
@@ -577,7 +578,8 @@ static enum sb_pin_state pin_sb_for_writeback(struct writeback_control *wbc,
 	/*
 	 * Caller must already hold the ref for this
 	 */
-	if (wbc->sync_mode == WB_SYNC_ALL) {
+	if (wbc->sync_mode == WB_SYNC_ALL ||
+	    wbc->sync_mode == WB_SYNC_NONE_PIN) {
 		WARN_ON(!rwsem_is_locked(&sb->s_umount));
 		return SB_NOT_PINNED;
 	}
@@ -1183,6 +1185,18 @@ static void wait_sb_inodes(struct super_block *sb)
 	iput(old_inode);
 }
 
+static void __writeback_inodes_sb(struct super_block *sb, int sb_locked)
+{
+	unsigned long nr_dirty = global_page_state(NR_FILE_DIRTY);
+	unsigned long nr_unstable = global_page_state(NR_UNSTABLE_NFS);
+	long nr_to_write;
+
+	nr_to_write = nr_dirty + nr_unstable +
+			(inodes_stat.nr_inodes - inodes_stat.nr_unused);
+
+	bdi_start_writeback(sb->s_bdi, sb, nr_to_write, sb_locked);
+}
+
 /**
  * writeback_inodes_sb	-	writeback dirty inodes from given super_block
  * @sb: the superblock
@@ -1194,18 +1208,23 @@ static void wait_sb_inodes(struct super_block *sb)
  */
 void writeback_inodes_sb(struct super_block *sb)
 {
-	unsigned long nr_dirty = global_page_state(NR_FILE_DIRTY);
-	unsigned long nr_unstable = global_page_state(NR_UNSTABLE_NFS);
-	long nr_to_write;
-
-	nr_to_write = nr_dirty + nr_unstable +
-			(inodes_stat.nr_inodes - inodes_stat.nr_unused);
-
-	bdi_start_writeback(sb->s_bdi, sb, nr_to_write);
+	__writeback_inodes_sb(sb, 0);
 }
 EXPORT_SYMBOL(writeback_inodes_sb);
 
 /**
+ * writeback_inodes_sb_locked	- writeback dirty inodes from given super_block
+ * @sb: the superblock
+ *
+ * Like writeback_inodes_sb(), except the caller already holds the
+ * sb umount sem.
+ */
+void writeback_inodes_sb_locked(struct super_block *sb)
+{
+	__writeback_inodes_sb(sb, 1);
+}
+
+/**
  * writeback_inodes_sb_if_idle	-	start writeback if none underway
  * @sb: the superblock
  *
diff --git a/fs/sync.c b/fs/sync.c
index 92b2281..de6a441 100644
--- a/fs/sync.c
+++ b/fs/sync.c
@@ -42,7 +42,7 @@ static int __sync_filesystem(struct super_block *sb, int wait)
 	if (wait)
 		sync_inodes_sb(sb);
 	else
-		writeback_inodes_sb(sb);
+		writeback_inodes_sb_locked(sb);
 
 	if (sb->s_op->sync_fs)
 		sb->s_op->sync_fs(sb, wait);
diff --git a/include/linux/backing-dev.h b/include/linux/backing-dev.h
index bd0e3c6..90e677a 100644
--- a/include/linux/backing-dev.h
+++ b/include/linux/backing-dev.h
@@ -103,7 +103,7 @@ int bdi_register_dev(struct backing_dev_info *bdi, dev_t dev);
 void bdi_unregister(struct backing_dev_info *bdi);
 int bdi_setup_and_register(struct backing_dev_info *, char *, unsigned int);
 void bdi_start_writeback(struct backing_dev_info *bdi, struct super_block *sb,
-				long nr_pages);
+				long nr_pages, int sb_locked);
 int bdi_writeback_task(struct bdi_writeback *wb);
 int bdi_has_dirty_io(struct backing_dev_info *bdi);
 
diff --git a/include/linux/writeback.h b/include/linux/writeback.h
index 36520de..3cd39b0 100644
--- a/include/linux/writeback.h
+++ b/include/linux/writeback.h
@@ -19,6 +19,7 @@ extern struct list_head inode_unused;
 enum writeback_sync_modes {
 	WB_SYNC_NONE,	/* Don't wait on anything */
 	WB_SYNC_ALL,	/* Wait on every mapping */
+	WB_SYNC_NONE_PIN,	/* Like WB_SYNC_NONE, but s_umount held */
 };
 
 /*
@@ -73,6 +74,7 @@ struct writeback_control {
 struct bdi_writeback;
 int inode_wait(void *);
 void writeback_inodes_sb(struct super_block *);
+void writeback_inodes_sb_locked(struct super_block *);
 int writeback_inodes_sb_if_idle(struct super_block *);
 void sync_inodes_sb(struct super_block *);
 void writeback_inodes_wbc(struct writeback_control *wbc);
diff --git a/mm/page-writeback.c b/mm/page-writeback.c
index 0b19943..49d3508 100644
--- a/mm/page-writeback.c
+++ b/mm/page-writeback.c
@@ -597,7 +597,7 @@ static void balance_dirty_pages(struct address_space *mapping,
 	    (!laptop_mode && ((global_page_state(NR_FILE_DIRTY)
 			       + global_page_state(NR_UNSTABLE_NFS))
 					  > background_thresh)))
-		bdi_start_writeback(bdi, NULL, 0);
+		bdi_start_writeback(bdi, NULL, 0, 0);
 }
 
 void set_page_dirty_balance(struct page *page, int page_mkwrite)

-- 
Jens Axboe