Btrfs: avoid syncing log in the fast fsync path when not necessary

Commit 3a8b36f3 ("Btrfs: fix data loss in the fast fsync path") added
a performance regression for that causes an unnecessary sync of the log
trees (fs/subvol and root log trees) when 2 consecutive fsyncs are done
against a file, without no writes or any metadata updates to the inode in
between them and if a transaction is committed before the second fsync is
called.

Huang Ying reported this to lkml (https://lkml.org/lkml/2015/3/18/99)
after a test sysbench test that measured a -62% decrease of file io
requests per second for that tests' workload.

The test is:

  echo performance > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
  echo performance > /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor
  echo performance > /sys/devices/system/cpu/cpu2/cpufreq/scaling_governor
  echo performance > /sys/devices/system/cpu/cpu3/cpufreq/scaling_governor
  mkfs -t btrfs /dev/sda2
  mount -t btrfs /dev/sda2 /fs/sda2
  cd /fs/sda2
  for ((i = 0; i < 1024; i++)); do fallocate -l 67108864 testfile.$i; done
  sysbench --test=fileio --max-requests=0 --num-threads=4 --max-time=600 \
    --file-test-mode=rndwr --file-total-size=68719476736 --file-io-mode=sync \
    --file-num=1024 run

A test on kvm guest, running a debug kernel gave me the following results:

Without 3a8b36f3:             16.01 reqs/sec
With 3a8b36f3:                 3.39 reqs/sec
With 3a8b36f3 and this patch: 16.04 reqs/sec
Reported-by: Huang Ying <ying.huang@intel.com>
Tested-by: Huang, Ying <ying.huang@intel.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Chris Mason <clm@fb.com>

fs/btrfs/file.c

@@ -1868,6 +1868,7 @@ int btrfs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
 	struct btrfs_log_ctx ctx;
 	int ret = 0;
 	bool full_sync = 0;
+	const u64 len = end - start + 1;
 
 	trace_btrfs_sync_file(file, datasync);
 
@@ -1896,7 +1897,7 @@ int btrfs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
 		 * all extents are persisted and the respective file extent
 		 * items are in the fs/subvol btree.
 		 */
-		ret = btrfs_wait_ordered_range(inode, start, end - start + 1);
+		ret = btrfs_wait_ordered_range(inode, start, len);
 	} else {
 		/*
 		 * Start any new ordered operations before starting to log the
@@ -1968,8 +1969,10 @@ int btrfs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
 	 */
 	smp_mb();
 	if (btrfs_inode_in_log(inode, root->fs_info->generation) ||
-	    (full_sync && BTRFS_I(inode)->last_trans <=
-	     root->fs_info->last_trans_committed)) {
+	    (BTRFS_I(inode)->last_trans <=
+	     root->fs_info->last_trans_committed &&
+	     (full_sync ||
+	      !btrfs_have_ordered_extents_in_range(inode, start, len)))) {
 		/*
 		 * We'v had everything committed since the last time we were
 		 * modified so clear this flag in case it was set for whatever

fs/btrfs/ordered-data.c

@@ -837,6 +837,20 @@ struct btrfs_ordered_extent *btrfs_lookup_ordered_range(struct inode *inode,
 	return entry;
 }
 
+bool btrfs_have_ordered_extents_in_range(struct inode *inode,
+					 u64 file_offset,
+					 u64 len)
+{
+	struct btrfs_ordered_extent *oe;
+
+	oe = btrfs_lookup_ordered_range(inode, file_offset, len);
+	if (oe) {
+		btrfs_put_ordered_extent(oe);
+		return true;
+	}
+	return false;
+}
+
 /*
  * lookup and return any extent before 'file_offset'.  NULL is returned
  * if none is found

fs/btrfs/ordered-data.h

@@ -188,6 +188,9 @@ btrfs_lookup_first_ordered_extent(struct inode * inode, u64 file_offset);
 struct btrfs_ordered_extent *btrfs_lookup_ordered_range(struct inode *inode,
 							u64 file_offset,
 							u64 len);
+bool btrfs_have_ordered_extents_in_range(struct inode *inode,
+					 u64 file_offset,
+					 u64 len);
 int btrfs_ordered_update_i_size(struct inode *inode, u64 offset,
 				struct btrfs_ordered_extent *ordered);
 int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u64 disk_bytenr,