vfs: call rcu_barrier after ->kill_sb()
authorBoaz Harrosh <[email protected]>
Thu, 10 Feb 2011 23:01:20 +0000 (15:01 -0800)
committerLinus Torvalds <[email protected]>
Sat, 12 Feb 2011 00:12:19 +0000 (16:12 -0800)
commitd863b50ab01333659314c2034890cb76d9fdc3c7
tree88d27ed4717bb49ba696f2efb5ef024388412af3
parent2dab597441667d6c04451a7dcf215241ad4c74f6
vfs: call rcu_barrier after ->kill_sb()

In commit fa0d7e3de6d6 ("fs: icache RCU free inodes"), we use rcu free
inode instead of freeing the inode directly.  It causes a crash when we
rmmod immediately after we umount the volume[1].

So we need to call rcu_barrier after we kill_sb so that the inode is
freed before we do rmmod.  The idea is inspired by Aneesh Kumar.
rcu_barrier will wait for all callbacks to end before preceding.  The
original patch was done by Tao Ma, but synchronize_rcu() is not enough
here.

1. http://marc.info/?l=linux-fsdevel&m=129680863330185&w=2

Tested-by: Tao Ma <[email protected]>
Signed-off-by: Boaz Harrosh <[email protected]>
Cc: Nick Piggin <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Chris Mason <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
fs/super.c