[PATCH] Btrfs: use late_initcall instead of module_init

[PATCH] Btrfs: use late_initcall instead of module_init

[Filipe David Borba Manana]

It seems that when init_btrfs_fs() is called, crc32c/crc32c-intel might
not always be already initialized, which results in the call to crypto_alloc_shash()
returning -ENOENT, as experienced by Ahmet who reported this.

Therefore make sure init_btrfs_fs() is called after crc32c is initialized (which
is at initialization level 6, module_init), by using late_initcall (which is at
initialization level 7) instead of module_init for btrfs.

Reported-and-Tested-by: Ahmet Inan <ainan@mathematik.uni-freiburg.de>
Signed-off-by: Filipe David Borba Manana <fdmanana@gmail.com>
---
 fs/btrfs/super.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index c02f633..97cc241 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -1996,7 +1996,7 @@ static void __exit exit_btrfs_fs(void)
 	btrfs_hash_exit();
 }
 
-module_init(init_btrfs_fs)
+late_initcall(init_btrfs_fs);
 module_exit(exit_btrfs_fs)
 
 MODULE_LICENSE("GPL");
-- 
1.7.9.5

[PATCH] Btrfs: use late_initcall instead of module_init

[Tobias Klausmann]

Hey,

as discussed on IRC, this makes 3.14-rc1 able tomount brtfs
again. Key difference (aside from being (un)able to mount) is that
btrfs now actually shows up in /proc/filesystems, where with
vanilla 3.14-rc1 it does not.

Tested on an i5-3470T.

Tested-By: Tobias Klausmann <klausman@schwarzvogel.de>