[lvm-devel] master - tests: data correctness after thin-pool resize

[Zdenek Kabelac]

Gitweb:        http://git.fedorahosted.org/git/?p=lvm2.git;a=commitdiff;h=a220939d9ef77db63a1f6af5f3798573f047ab03
Commit:        a220939d9ef77db63a1f6af5f3798573f047ab03
Parent:        9243877ea10f46163506dc64d073d1f24446ab24
Author:        Zdenek Kabelac <zkabelac at redhat.com>
AuthorDate:    Fri Nov 20 22:28:21 2015 +0100
Committer:     Zdenek Kabelac <zkabelac at redhat.com>
CommitterDate: Tue Nov 24 11:29:27 2015 +0100

tests: data correctness after thin-pool resize

---
 test/shell/thin-resize-match.sh |   84 +++++++++++++++++++++++++++++++++++++++
 1 files changed, 84 insertions(+), 0 deletions(-)

diff --git a/test/shell/thin-resize-match.sh b/test/shell/thin-resize-match.sh
new file mode 100644
index 0000000..d5481e2
--- /dev/null
+++ b/test/shell/thin-resize-match.sh
@@ -0,0 +1,84 @@
+#!/bin/bash
+# Copyright (C) 2015 Red Hat, Inc. All rights reserved.
+#
+# This copyrighted material is made available to anyone wishing to use,
+# modify, copy, or redistribute it subject to the terms and conditions
+# of the GNU General Public License v.2.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation,
+# Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+
+# ensure there is no data loss during thin-pool resize
+
+export LVM_TEST_THIN_REPAIR_CMD=${LVM_TEST_THIN_REPAIR_CMD-/bin/false}
+
+. lib/inittest
+
+test -e LOCAL_LVMPOLLD && skip
+
+which md5sum || skip
+
+aux have_thin 1 0 0 || skip
+
+aux prepare_pvs 2 20
+
+vgcreate -s 512K $vg $(< DEVICES)
+
+lvcreate -L1M -V2M -n $lv1 -T $vg/pool
+
+# just ensure we check what we need to check
+check lv_field $vg/pool size "1.00m"
+check lv_field $vg/$lv1 size "2.00m"
+
+# prepare 2097152  file content
+seq 0 315465 > 2M
+md5sum 2M | cut -f 1 -d ' ' | tee MD5
+dd if=2M of="$DM_DEV_DIR/mapper/$vg-$lv1" bs=512K conv=fdatasync 2>&1 >log &
+#dd if=2M of="$DM_DEV_DIR/mapper/$vg-$lv1" bs=2M oflag=direct &
+
+# give it some time to fill thin-volume
+# eventually loop to wait for 100% full pool...
+sleep .1
+lvs -a $vg
+
+# this must not 'block & wait' on suspending flush
+# if it waits on thin-pool's target timeout
+# it will harm queued data
+lvextend -L+512k $vg/pool
+lvextend -L+512k $vg/pool
+
+# collect 'dd' result
+wait
+cat log
+
+lvs -a $vg
+
+dd if="$DM_DEV_DIR/mapper/$vg-$lv1" of=2M-2 iflag=direct
+md5sum 2M-2 | cut -f 1 -d ' '  | tee MD5-2
+
+# these 2 are supposed to match
+diff MD5  MD5-2
+
+
+# Do not want to see Live & Inactive table entry
+echo $(dm_info attr,name) | not grep "LI-.*${PREFIX}" || {
+        dmsetup table --inactive | grep ${PREFIX}
+	die "Found device with Inactive table"
+}
+
+# Check wrapping active thin-pool linear mapping has matching size
+POOLSZ=$(dmsetup table ${vg}-pool-tpool | cut -d ' ' -f 2)
+WRAPSZ=$(dmsetup table ${vg}-pool | cut -d ' ' -f 2)
+
+#
+# FIXME: currently requires to update 2 dependent targets in one 'preload'
+#        lvm2 cannot handle this and would need one extra --refresh pass.
+#        Once resolved - enabled this test.
+#        Maybe other solution without fake linear mapping could be found.
+#        Eventually strictly map just single sector as it has no real use?
+#
+#should test "${POOLSZ}" = "${WRAPSZ}" || \
+#        die "Wrapping pool device size does not match real pool size"
+
+vgremove -f $vg