Write dirty inodes on close

When the property atime=on is set operations which only access
and inode do cause an atime update.  However, it turns out that
dirty inodes with updated atimes are only written to disk when
the inodes get evicted from the cache.  Somewhat surprisingly
the source suggests that this isn't a ZoL specific issue.

This behavior may in part explain why zfs's reclaim logic has
been observed to be slow.  When reclaiming inodes its likely
that they have a dirty atime which will force a write to disk.

Obviously we don't want to force a write to disk for every
atime update, these needs to be batched.  The right way to
do this is to fully implement the .dirty_inode and .write_inode
callbacks.  However, to do that right requires proper unification
of some fields in the znode/inode.  Then we could just mark the
inode dirty and leave it to the VFS to call .write_inode
periodically.

Until that work gets done we have to settle for some middle
ground.  The simplest and safest thing we can do for now is
to write the dirty inode on last close.  This should prevent
the majority of inodes in the cache from having dirty atimes
and not drastically increase the number of writes.

Some rudimentally testing to show how long it takes to drop
500,000 inodes from the cache shows promising results.  This
is as expected because we're no longer do lots of IO as part
of the eviction, it was done earlier during the close.

w/out patch: ~30s to drop 500,000 inodes with drop_caches.
with patch:  ~3s to drop 500,000 inodes with drop_caches.

Signed-off-by: Brian Behlendorf <[email protected]>

module/zfs/zpl_file.c

@@ -52,6 +52,9 @@ zpl_release(struct inode *ip, struct file *filp)
 	cred_t *cr = CRED();
 	int error;
 
+	if (ITOZ(ip)->z_atime_dirty)
+		mark_inode_dirty(ip);
+
 	crhold(cr);
 	error = -zfs_close(ip, filp->f_flags, cr);
 	crfree(cr);