lvm2 test for snapshots on a 4K drive
RFC PATCH 1/1 lvm2 test for snapshots on a 4K drive
This patch adds a generic way to establish a scsi_debug-based 4K drive for use by an LVM2 test (rather than using the traditional loop device).
t-4k-drive.sh shows how this new 'prepare_scsi_debug_dev' function can be used repeatedly (within a script) to test LVM2 ontop of a ramdisk-based SCSI device w/ arbitrary scsi_debug features.
prepare_scsi_debug_dev currently assumes exclussive access to the scsi_debug module. Any script will be skipped if it tries to use prepare_scsi_debug_dev when scsi_debug is unavailable or already loaded into the kernel.
---
test/t-4k-drive.sh new file +98 -0
@@ -0,0 +1,98 @@ +# Copyright (C) 2009 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 + +which awk || exit 200 +which mkfs.ext3 || exit 200 + +. ./test-utils.sh + +check_logical_block_size() +{ + local DEV_=$1 + local LOGICAL_BS=$2 + # Verify logical_block_size - requires Linux >= 2.6.31 + SYSFS_LOGICAL_BLOCK_SIZE=`echo /sys/block/$(basename $DEV_)/queue/logical_block_size` + if [ -f "$SYSFS_LOGICAL_BLOCK_SIZE" ] ; then + ACTUAL_LOGICAL_BLOCK_SIZE=`cat $SYSFS_LOGICAL_BLOCK_SIZE` + test $ACTUAL_LOGICAL_BLOCK_SIZE = $LOGICAL_BS + fi +} + +lvdev_() +{ + echo "$G_dev_/$1/$2" +} + +test_snapshot_mount() +{ + lvcreate -L 16M -n $lv1 $vg $dev1 + mkfs.ext3 $(lvdev_ $vg $lv1) + mkdir test_mnt + mount $(lvdev_ $vg $lv1) test_mnt + lvcreate -L 16M -n $lv2 -s $vg/$lv1 + umount test_mnt + # mount the origin + mount $(lvdev_ $vg $lv1) test_mnt + umount test_mnt + # mount the snapshot + mount $(lvdev_ $vg $lv2) test_mnt + umount test_mnt + rm -r test_mnt + #sleep 1 # needed so 'vgchange -an' works reliably within KVM (cache=off) + vgchange -an $vg + #sleep 1 # needed so 'lvremove -f' works reliably within KVM (cache=off) + lvremove -f $vg/$lv2 + lvremove -f $vg/$lv1 +} + + +NUM_DEVS=1 +PER_DEV_SIZE=33 +DEV_SIZE=$(($NUM_DEVS*$PER_DEV_SIZE)) + +# --------------------------------------------- +# Create "desktop-class" 4K drive +# (logical_block_size=512, physical_block_size=4096, alignment_offset=0): +LOGICAL_BLOCK_SIZE=512 +prepare_scsi_debug_dev $DEV_SIZE \ + sector_size=$LOGICAL_BLOCK_SIZE physblk_exp=3 +check_logical_block_size $SCSI_DEBUG_DEV $LOGICAL_BLOCK_SIZE + +aux prepare_vg $NUM_DEVS $PER_DEV_SIZE +test_snapshot_mount +vgremove $vg + +cleanup_scsi_debug_dev + +# --------------------------------------------- +# Create "desktop-class" 4K drive w/ 63-sector DOS partition compensation +# (logical_block_size=512, physical_block_size=4096, alignment_offset=3584): +LOGICAL_BLOCK_SIZE=512 +prepare_scsi_debug_dev $DEV_SIZE \ + sector_size=$LOGICAL_BLOCK_SIZE physblk_exp=3 lowest_aligned=7 +check_logical_block_size $SCSI_DEBUG_DEV $LOGICAL_BLOCK_SIZE + +aux prepare_vg $NUM_DEVS $PER_DEV_SIZE +test_snapshot_mount +vgremove $vg + +cleanup_scsi_debug_dev + +# --------------------------------------------- +# Create "enterprise-class" 4K drive +# (logical_block_size=4096, physical_block_size=4096, alignment_offset=0): +LOGICAL_BLOCK_SIZE=4096 +prepare_scsi_debug_dev $DEV_SIZE \ + sector_size=$LOGICAL_BLOCK_SIZE +check_logical_block_size $SCSI_DEBUG_DEV $LOGICAL_BLOCK_SIZE + +prepare_vg $NUM_DEVS $PER_DEV_SIZE +test_snapshot_mount +vgremove $vg
test/test-utils.sh +66 -2
80 lines not on this page · raw message
Quoted 11 lines · Mike Snitzer
This patch adds a generic way to establish a scsi_debug-based 4K drive for use by an LVM2 test (rather than using the traditional loop device).
t-4k-drive.shshows how this new 'prepare_scsi_debug_dev' function can be used repeatedly (within a script) to test LVM2 ontop of a ramdisk-based SCSI device w/ arbitraryscsi_debugfeatures.
prepare_scsi_debug_devcurrently assumes exclussive access to thescsi_debugmodule. Any script will be skipped if it tries to useprepare_scsi_debug_devwhenscsi_debugis unavailable or already loaded into the kernel.
And here is alternative way to allow a script to override the type of underlying device that is used for all PVs in a test.
It is a bit more rigid to do this way (as compared to the original RFC patch and its multiple calls to 'prepare_scsi_debug_dev'). Exposing the 'prepare_dev_' hook just enables a script a different way to formalize the use of an alternative underlying device for LVM2 testing.
Not sure this is of any value but figured I'd share in the hopes of clarifying the original RFC patch a bit more.
Mike
---
test/test-utils.sh +13 -0
@@ -158,6 +158,16 @@ cleanup_scsi_debug_dev() unset SCSI_DEBUG_DEV } +# This is a stub function that is run via prepare_devs(). +# Override it with a per-test function, e.g., to setup more +# sophisticated physical volumes (e.g. with scsi_debug). +# Any test that provides this override _must_ set $LOOP to the +# full path of the device that was created under $G_dev_ +# - Setting $LOOP provides the means to short-circuit +# prepare_devs()'s use of prepare_loop() +# - TODO: rename $LOOP to $LVMTEST_DEVICE? +prepare_dev_() { :; } + prepare_devs() { local n="$1" test -z "$n" && n=3
@@ -166,6 +176,9 @@ prepare_devs() { local pvname="$3" test -z "$pvname" && pvname="pv" + # prepare_dev_ is a user-overridable that prepares the test's + # underlying device (it _must_ set $LOOP to be effective) + prepare_dev_ $(($n*$devsize)) prepare_loop $(($n*$devsize)) PREFIX="LVMTEST$$"
test/t-4k-drive-prime.sh new file +71 -0
71 lines not on this page · raw message