ceph/qa/run_xfstests-obsolete.sh

459 lines
12 KiB
Bash

#!/bin/bash
# Copyright (C) 2012 Dreamhost, LLC
#
# This is free software; see the source for copying conditions.
# There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR
# A PARTICULAR PURPOSE.
#
# This is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as
# published by the Free Software Foundation version 2.
# Usage:
# run_xfs_tests -t /dev/<testdev> -s /dev/<scratchdev> -f <fstype> <tests>
# - test device and scratch device will both get trashed
# - fstypes can be xfs, ext4, or btrfs (xfs default)
# - tests can be listed individually or in ranges: 1 3-5 8
# tests can also be specified by group: -g quick
#
# Exit status:
# 0: success
# 1: usage error
# 2: other runtime error
# 99: argument count error (programming error)
# 100: getopt error (internal error)
# Alex Elder <elder@dreamhost.com>
# April 13, 2012
set -e
PROGNAME=$(basename $0)
# xfstests is downloaded from this git repository and then built.
# XFSTESTS_REPO="git://oss.sgi.com/xfs/cmds/xfstests.git"
XFSTESTS_REPO="git://ceph.com/git/xfstests.git"
# Default command line option values
COUNT="1"
FS_TYPE="xfs"
SCRATCH_DEV="" # MUST BE SPECIFIED
TEST_DEV="" # MUST BE SPECIFIED
TESTS="-g auto" # The "auto" group is supposed to be "known good"
# rbd presents geometry information that causes mkfs.xfs to
# issue a warning. This option avoids this class of problems.
XFS_MKFS_OPTIONS="-l su=32k"
# Override the default test list with a list of tests known to pass
# until we can work through getting them all passing reliably.
TESTS="1-7 9 11-15 17 19-21 26-29 31-34 41 46-48 50-54 56 61 63-67 69-70 74-76"
TESTS="${TESTS} 78 79 84-89 91-92 100 103 105 108 110 116-121 124 126"
TESTS="${TESTS} 129-135 137-141 164-167 182 184 187-190 192 194"
TESTS="${TESTS} 196 199 201 203 214-216 220-227 234 236-238 241 243-249"
TESTS="${TESTS} 253 257-259 261 262 269 273 275 277 278 280 285 286"
# 275 was the highest available test as of 4/10/12.
# 289 was the highest available test as of 11/15/12.
######
# Some explanation of why tests have been excluded above:
#
# Test 008 was pulled because it contained a race condition leading to
# spurious failures.
#
# Test 049 was pulled because it caused a kernel fault.
# http://tracker.newdream.net/issues/2260
# Test 232 was pulled because it caused an XFS error
# http://tracker.newdream.net/issues/2302
#
# This test passes but takes a LONG time (1+ hours): 127
#
# These were not run for one (anticipated) reason or another:
# 010 016 030 035 040 044 057 058-060 072 077 090 093-095 097-099 104
# 112 113 122 123 125 128 142 147-163 168 175-178 180 185 191 193
# 195 197 198 207-213 217 228 230-233 235 239 240 252 254 255 264-266
# 270-272 276 278-279 281-284 288 289
#
# These tests all failed (produced output different from golden):
# 042 073 083 096 109 169 170 200 202 204-206 218 229 240 242 250
# 263 276 277 279 287
#
# The rest were not part of the "auto" group:
# 018 022 023 024 025 036 037 038 039 043 055 071 080 081 082 101
# 102 106 107 111 114 115 136 171 172 173 251 267 268
######
# print an error message and quit with non-zero status
function err() {
if [ $# -gt 0 ]; then
echo "" >&2
echo "${PROGNAME}: ${FUNCNAME[1]}: $@" >&2
fi
exit 2
}
# routine used to validate argument counts to all shell functions
function arg_count() {
local func
local want
local got
if [ $# -eq 2 ]; then
func="${FUNCNAME[1]}" # calling function
want=$1
got=$2
else
func="${FUNCNAME[0]}" # i.e., arg_count
want=2
got=$#
fi
[ "${want}" -eq "${got}" ] && return 0
echo "${PROGNAME}: ${func}: arg count bad (want ${want} got ${got})" >&2
exit 99
}
# validation function for repeat count argument
function count_valid() {
arg_count 1 $#
test "$1" -gt 0 # 0 is pointless; negative is wrong
}
# validation function for filesystem type argument
function fs_type_valid() {
arg_count 1 $#
case "$1" in
xfs|ext4|btrfs) return 0 ;;
*) return 1 ;;
esac
}
# validation function for device arguments
function device_valid() {
arg_count 1 $#
# Very simple testing--really should try to be more careful...
test -b "$1"
}
# print a usage message and quit
#
# if a message is supplied, print that first, and then exit
# with non-zero status
function usage() {
if [ $# -gt 0 ]; then
echo "" >&2
echo "$@" >&2
fi
echo "" >&2
echo "Usage: ${PROGNAME} <options> <tests>" >&2
echo "" >&2
echo " options:" >&2
echo " -h or --help" >&2
echo " show this message" >&2
echo " -c or --count" >&2
echo " iteration count (1 or more)" >&2
echo " -f or --fs-type" >&2
echo " one of: xfs, ext4, btrfs" >&2
echo " (default fs-type: xfs)" >&2
echo " -s or --scratch-dev (REQUIRED)" >&2
echo " name of device used for scratch filesystem" >&2
echo " -t or --test-dev (REQUIRED)" >&2
echo " name of device used for test filesystem" >&2
echo " tests:" >&2
echo " list of test numbers or ranges, e.g.:" >&2
echo " 1-9 11-15 17 19-21 26-28 31-34 41" >&2
echo " or possibly an xfstests test group, e.g.:" >&2
echo " -g quick" >&2
echo " (default tests: -g auto)" >&2
echo "" >&2
[ $# -gt 0 ] && exit 1
exit 0 # This is used for a --help
}
# parse command line arguments
function parseargs() {
# Short option flags
SHORT_OPTS=""
SHORT_OPTS="${SHORT_OPTS},h"
SHORT_OPTS="${SHORT_OPTS},c:"
SHORT_OPTS="${SHORT_OPTS},f:"
SHORT_OPTS="${SHORT_OPTS},s:"
SHORT_OPTS="${SHORT_OPTS},t:"
# Short option flags
LONG_OPTS=""
LONG_OPTS="${LONG_OPTS},help"
LONG_OPTS="${LONG_OPTS},count:"
LONG_OPTS="${LONG_OPTS},fs-type:"
LONG_OPTS="${LONG_OPTS},scratch-dev:"
LONG_OPTS="${LONG_OPTS},test-dev:"
TEMP=$(getopt --name "${PROGNAME}" \
--options "${SHORT_OPTS}" \
--longoptions "${LONG_OPTS}" \
-- "$@")
eval set -- "$TEMP"
while [ "$1" != "--" ]; do
case "$1" in
-h|--help)
usage
;;
-c|--count)
count_valid "$2" ||
usage "invalid count '$2'"
COUNT="$2"
shift
;;
-f|--fs-type)
fs_type_valid "$2" ||
usage "invalid fs_type '$2'"
FS_TYPE="$2"
shift
;;
-s|--scratch-dev)
device_valid "$2" ||
usage "invalid scratch-dev '$2'"
SCRATCH_DEV="$2"
shift
;;
-t|--test-dev)
device_valid "$2" ||
usage "invalid test-dev '$2'"
TEST_DEV="$2"
shift
;;
*)
exit 100 # Internal error
;;
esac
shift
done
shift
[ -n "${TEST_DEV}" ] || usage "test-dev must be supplied"
[ -n "${SCRATCH_DEV}" ] || usage "scratch-dev must be supplied"
[ $# -eq 0 ] || TESTS="$@"
}
################################################################
[ -z "$TESTDIR" ] && export TESTDIR="/tmp/cephtest"
# Set up some environment for normal teuthology test setup.
# This really should not be necessary but I found it was.
export CEPH_ARGS="--conf ${TESTDIR}/ceph.conf"
export CEPH_ARGS="${CEPH_ARGS} --keyring ${TESTDIR}/data/client.0.keyring"
export CEPH_ARGS="${CEPH_ARGS} --name client.0"
export LD_LIBRARY_PATH="${TESTDIR}/binary/usr/local/lib:${LD_LIBRARY_PATH}"
export PATH="${TESTDIR}/binary/usr/local/bin:${PATH}"
export PATH="${TESTDIR}/binary/usr/local/sbin:${PATH}"
################################################################
# Filesystem-specific mkfs options--set if not supplied
export XFS_MKFS_OPTIONS="${XFS_MKFS_OPTIONS:--f -l su=65536}"
export EXT4_MKFS_OPTIONS="${EXT4_MKFS_OPTIONS:--F}"
export BTRFS_MKFS_OPTION # No defaults
XFSTESTS_DIR="/var/lib/xfstests" # Where the tests live
# download, build, and install xfstests
function install_xfstests() {
arg_count 0 $#
local multiple=""
local ncpu
pushd "${TESTDIR}"
git clone "${XFSTESTS_REPO}"
cd xfstests
# FIXME: use an older version before the tests were rearranged!
git reset --hard e5f1a13792f20cfac097fef98007610b422f2cac
ncpu=$(getconf _NPROCESSORS_ONLN 2>&1)
[ -n "${ncpu}" -a "${ncpu}" -gt 1 ] && multiple="-j ${ncpu}"
make realclean
make ${multiple}
make -k install
popd
}
# remove previously-installed xfstests files
function remove_xfstests() {
arg_count 0 $#
rm -rf "${TESTDIR}/xfstests"
rm -rf "${XFSTESTS_DIR}"
}
# create a host options file that uses the specified devices
function setup_host_options() {
arg_count 0 $#
# Create mount points for the test and scratch filesystems
local test_dir="$(mktemp -d ${TESTDIR}/test_dir.XXXXXXXXXX)"
local scratch_dir="$(mktemp -d ${TESTDIR}/scratch_mnt.XXXXXXXXXX)"
# Write a host options file that uses these devices.
# xfstests uses the file defined by HOST_OPTIONS as the
# place to get configuration variables for its run, and
# all (or most) of the variables set here are required.
export HOST_OPTIONS="$(mktemp ${TESTDIR}/host_options.XXXXXXXXXX)"
cat > "${HOST_OPTIONS}" <<-!
# Created by ${PROGNAME} on $(date)
# HOST_OPTIONS="${HOST_OPTIONS}"
TEST_DEV="${TEST_DEV}"
SCRATCH_DEV="${SCRATCH_DEV}"
TEST_DIR="${test_dir}"
SCRATCH_MNT="${scratch_dir}"
FSTYP="${FS_TYPE}"
export TEST_DEV SCRATCH_DEV TEST_DIR SCRATCH_MNT FSTYP
#
export XFS_MKFS_OPTIONS="${XFS_MKFS_OPTIONS}"
!
# Now ensure we are using the same values
. "${HOST_OPTIONS}"
}
# remove the host options file, plus the directories it refers to
function cleanup_host_options() {
arg_count 0 $#
rm -rf "${TEST_DIR}" "${SCRATCH_MNT}"
rm -f "${HOST_OPTIONS}"
}
# run mkfs on the given device using the specified filesystem type
function do_mkfs() {
arg_count 1 $#
local dev="${1}"
local options
case "${FSTYP}" in
xfs) options="${XFS_MKFS_OPTIONS}" ;;
ext4) options="${EXT4_MKFS_OPTIONS}" ;;
btrfs) options="${BTRFS_MKFS_OPTIONS}" ;;
esac
"mkfs.${FSTYP}" ${options} "${dev}" ||
err "unable to make ${FSTYP} file system on device \"${dev}\""
}
# mount the given device on the given mount point
function do_mount() {
arg_count 2 $#
local dev="${1}"
local dir="${2}"
mount "${dev}" "${dir}" ||
err "unable to mount file system \"${dev}\" on \"${dir}\""
}
# unmount a previously-mounted device
function do_umount() {
arg_count 1 $#
local dev="${1}"
if mount | grep "${dev}" > /dev/null; then
if ! umount "${dev}"; then
err "unable to unmount device \"${dev}\""
fi
else
# Report it but don't error out
echo "device \"${dev}\" was not mounted" >&2
fi
}
# do basic xfstests setup--make and mount the test and scratch filesystems
function setup_xfstests() {
arg_count 0 $#
# TEST_DEV can persist across test runs, but for now we
# don't bother. I believe xfstests prefers its devices to
# have been already been formatted for the desired
# filesystem type--it uses blkid to identify things or
# something. So we mkfs both here for a fresh start.
do_mkfs "${TEST_DEV}"
do_mkfs "${SCRATCH_DEV}"
# I believe the test device is expected to be mounted; the
# scratch doesn't need to be (but it doesn't hurt).
do_mount "${TEST_DEV}" "${TEST_DIR}"
do_mount "${SCRATCH_DEV}" "${SCRATCH_MNT}"
}
# clean up changes made by setup_xfstests
function cleanup_xfstests() {
arg_count 0 $#
# Unmount these in case a test left them mounted (plus
# the corresponding setup function mounted them...)
do_umount "${TEST_DEV}"
do_umount "${SCRATCH_DEV}"
}
# top-level setup routine
function setup() {
arg_count 0 $#
setup_host_options
install_xfstests
setup_xfstests
}
# top-level (final) cleanup routine
function cleanup() {
arg_count 0 $#
cd /
cleanup_xfstests
remove_xfstests
cleanup_host_options
}
trap cleanup EXIT ERR HUP INT QUIT
# ################################################################
start_date="$(date)"
parseargs "$@"
setup
pushd "${XFSTESTS_DIR}"
for (( i = 1 ; i <= "${COUNT}" ; i++ )); do
[ "${COUNT}" -gt 1 ] && echo "=== Iteration "$i" starting at: $(date)"
./check ${TESTS} # Here we actually run the tests
status=$?
[ "${COUNT}" -gt 1 ] && echo "=== Iteration "$i" complete at: $(date)"
done
popd
# cleanup is called via the trap call, above
echo "This xfstests run started at: ${start_date}"
echo "xfstests run completed at: $(date)"
[ "${COUNT}" -gt 1 ] && echo "xfstests run consisted of ${COUNT} iterations"
exit "${status}"