Btrfs: Handle SGID bit when creating inodes
authorChris Ball <cjb@laptop.org>
Wed, 4 Feb 2009 14:29:54 +0000 (09:29 -0500)
committerChris Mason <chris.mason@oracle.com>
Wed, 4 Feb 2009 14:29:54 +0000 (09:29 -0500)
Before this patch, new files/dirs would ignore the SGID bit on their
parent directory and always be owned by the creating user's uid/gid.

Signed-off-by: Chris Ball <cjb@laptop.org>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
fs/btrfs/inode.c

index 95ea58c..5792816 100644 (file)
@@ -3472,7 +3472,14 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
                root->highest_inode = objectid;
 
        inode->i_uid = current_fsuid();
-       inode->i_gid = current_fsgid();
+
+       if (dir->i_mode & S_ISGID) {
+               inode->i_gid = dir->i_gid;
+               if (S_ISDIR(mode))
+                       mode |= S_ISGID;
+       } else
+               inode->i_gid = current_fsgid();
+
        inode->i_mode = mode;
        inode->i_ino = objectid;
        inode_set_bytes(inode, 0);