Merge lp:~colin-king/ecryptfs/lp-1098584 into lp:ecryptfs

[Colin Ian King]

Fix bug LP: #1098584. It appears that btrfs lies about how much free space is available which is doubly problematic with small btrfs partitions of a few GB. To prevent us hitting this, I've increased the amount of slop available to ensure we don't run out of space. Also turn of btrfs COW as this helps and also helps other tests that may cause lots of seek + writes.

[Tyler Hicks (tyhicks)]

Looks good. Thanks, Colin!

review: Approve

=== modified file 'tests/lib/etl_funcs.sh'
--- tests/lib/etl_funcs.sh	2012-11-02 23:20:18 +0000
+++ tests/lib/etl_funcs.sh	2013-01-14 14:18:26 +0000
@@ -31,6 +31,7 @@
 default_lmount_opts="rw,relatime"
 default_ext2_opts="user_xattr,acl"
 default_ext3_opts="user_xattr,acl,commit=600,barrier=1,data=ordered"
+default_btrfs_opts="nodatacow"
 default_mount_opts="rw,relatime,ecryptfs_cipher=aes,ecryptfs_key_bytes=16,ecryptfs_sig=\${ETL_FEKEK_SIG}"
 default_fne_mount_opts="${default_mount_opts},ecryptfs_fnek_sig=\${ETL_FNEK_SIG}"
 
@@ -334,6 +335,9 @@
 	ext3|ext4)
 		lmount_opts=${default_lmount_opts},${default_ext3_opts}
 		;;
+	btrfs)
+		lmount_opts=${default_lmount_opts},${default_btrfs_opts}
+		;;
 	*)
 		lmount_opts=$default_lmount_opts
 		;;
@@ -394,10 +398,11 @@
 	btrfs)
 		# btrfs is a pain, since there is a big difference between the
 		# amount of free space it reports and the maximum size of a file
-		# one can produce before filling up the partition. So instead
-		# we divide by 2 to ensure we have more than enough free space. 
+		# one can produce before filling up the partition, especially
+		# with small partitions. So instead we divide by 4 to ensure
+		# we have more than enough free space. 
 		#
-		blks=$((blks / 2))
+		blks=$((blks / 4))
 		;;
 	xfs)
 		# xfs misbehaves on small file systems when we truncate, according to