update Documentation/filesystems/Locking
authorChristoph Hellwig <hch@lst.de>
Fri, 19 Jun 2009 18:22:37 +0000 (20:22 +0200)
committerAl Viro <viro@zeniv.linux.org.uk>
Wed, 24 Jun 2009 12:15:25 +0000 (08:15 -0400)
The rules for locking in many superblock operations has changed
significantly, so update the documentation for it.  Also correct some
older updates and ommissions.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Documentation/filesystems/Locking

index 229d7b7c50a350053170e5fba11ac9a0168c0e70..18b9d0ca0630e281bb20dc5c5991e4b6f06b4643 100644 (file)
@@ -109,27 +109,28 @@ prototypes:
 
 locking rules:
        All may block.
-                       BKL     s_lock  s_umount
-alloc_inode:           no      no      no
-destroy_inode:         no
-dirty_inode:           no                              (must not sleep)
-write_inode:           no
-drop_inode:            no                              !!!inode_lock!!!
-delete_inode:          no
-put_super:             yes     yes     no
-write_super:           no      yes     read
-sync_fs:               no      no      read
-freeze_fs:             ?
-unfreeze_fs:           ?
-statfs:                        no      no      no
-remount_fs:            yes     yes     maybe           (see below)
-clear_inode:           no
-umount_begin:          yes     no      no
-show_options:          no                              (vfsmount->sem)
-quota_read:            no      no      no              (see below)
-quota_write:           no      no      no              (see below)
-
-->remount_fs() will have the s_umount lock if it's already mounted.
+       None have BKL
+                       s_umount
+alloc_inode:
+destroy_inode:
+dirty_inode:                           (must not sleep)
+write_inode:
+drop_inode:                            !!!inode_lock!!!
+delete_inode:
+put_super:             write
+write_super:           read
+sync_fs:               read
+freeze_fs:             read
+unfreeze_fs:           read
+statfs:                        no
+remount_fs:            maybe           (see below)
+clear_inode:
+umount_begin:          no
+show_options:          no              (namespace_sem)
+quota_read:            no              (see below)
+quota_write:           no              (see below)
+
+->remount_fs() will have the s_umount exclusive lock if it's already mounted.
 When called from get_sb_single, it does NOT have the s_umount lock.
 ->quota_read() and ->quota_write() functions are both guaranteed to
 be the only ones operating on the quota file by the quota code (via