fs: push rcu_barrier() from deactivate_locked_super() to filesystems
authorKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Wed, 26 Sep 2012 01:33:07 +0000 (11:33 +1000)
committerAl Viro <viro@zeniv.linux.org.uk>
Wed, 3 Oct 2012 01:35:55 +0000 (21:35 -0400)
commit8c0a85377048b64c880e76ec7368904fe46d0b94
tree12e075f4159e88b09a89e35866017f82f80bcc7d
parent99621b44aa194eab594e1f17217231c02b519211
fs: push rcu_barrier() from deactivate_locked_super() to filesystems

There's no reason to call rcu_barrier() on every
deactivate_locked_super().  We only need to make sure that all delayed rcu
free inodes are flushed before we destroy related cache.

Removing rcu_barrier() from deactivate_locked_super() affects some fast
paths.  E.g.  on my machine exit_group() of a last process in IPC
namespace takes 0.07538s.  rcu_barrier() takes 0.05188s of that time.

Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
47 files changed:
fs/9p/v9fs.c
fs/adfs/super.c
fs/affs/super.c
fs/afs/super.c
fs/befs/linuxvfs.c
fs/bfs/inode.c
fs/btrfs/extent_io.c
fs/btrfs/inode.c
fs/ceph/super.c
fs/cifs/cifsfs.c
fs/coda/inode.c
fs/ecryptfs/main.c
fs/efs/super.c
fs/exofs/super.c
fs/ext2/super.c
fs/ext3/super.c
fs/ext4/super.c
fs/fat/inode.c
fs/freevxfs/vxfs_super.c
fs/fuse/inode.c
fs/hfs/super.c
fs/hfsplus/super.c
fs/hpfs/super.c
fs/hugetlbfs/inode.c
fs/isofs/inode.c
fs/jffs2/super.c
fs/jfs/super.c
fs/logfs/inode.c
fs/minix/inode.c
fs/ncpfs/inode.c
fs/nfs/inode.c
fs/nilfs2/super.c
fs/ntfs/super.c
fs/ocfs2/dlmfs/dlmfs.c
fs/ocfs2/super.c
fs/openpromfs/inode.c
fs/qnx4/inode.c
fs/qnx6/inode.c
fs/reiserfs/super.c
fs/romfs/super.c
fs/squashfs/super.c
fs/super.c
fs/sysv/inode.c
fs/ubifs/super.c
fs/udf/super.c
fs/ufs/super.c
fs/xfs/xfs_super.c