scrub/e2scrub.in

   1 #!/bin/bash
   2
   3 #  Copyright (C) 2018 Oracle.  All Rights Reserved.
   4 #
   5 #  Author: Darrick J. Wong <darrick.wong@oracle.com>
   6 #
   7 #  This program is free software; you can redistribute it and/or
   8 #  modify it under the terms of the GNU General Public License
   9 #  as published by the Free Software Foundation; either version 2
  10 #  of the License, or (at your option) any later version.
  11 #
  12 #  This program is distributed in the hope that it would be useful,
  13 #  but WITHOUT ANY WARRANTY; without even the implied warranty of
  14 #  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15 #  GNU General Public License for more details.
  16 #
  17 #  You should have received a copy of the GNU General Public License
  18 #  along with this program; if not, write the Free Software Foundation,
  19 #  Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
  20
  21 # Automatically check an LVM-managed filesystem online.
  22 # We use lvm snapshots to do this, which means that we can only
  23 # check filesystems in VGs that have at least 256MB (or so) of
  24 # free space.
  25
  26 PATH=/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin
  27
  28 if (( $EUID != 0 )); then
  29     echo "e2scrub must be run as root"
  30     exit 1
  31 fi
  32
  33 snap_size_mb=256
  34 fstrim=0
  35 reap=0
  36 e2fsck_opts=""
  37 conffile="@root_sysconfdir@/e2scrub.conf"
  38
  39 test -f "${conffile}" && . "${conffile}"
  40
  41 print_help() {
  42         echo "Usage: $0 [OPTIONS] mountpoint | device"
  43         echo
  44         echo "mountpoint must be on an LVM-managed block device"
  45         echo "-n: Show what commands e2scrub would execute."
  46         echo "-r: Remove e2scrub snapshot and exit, do not check anything."
  47         echo "-t: Run fstrim if successful."
  48         echo "-V: Print version information and exit."
  49 }
  50
  51 print_version() {
  52         echo "e2scrub @E2FSPROGS_VERSION@ (@E2FSPROGS_DATE@)"
  53 }
  54
  55 exitcode() {
  56         ret="$1"
  57
  58         # If we're being run as a service, the return code must fit the LSB
  59         # init script action error guidelines, which is to say that we
  60         # compress all errors to 1 ("generic or unspecified error", LSB 5.0
  61         # section 22.2) and hope the admin will scan the log for what
  62         # actually happened.
  63
  64         # We have to sleep 2 seconds here because journald uses the pid to
  65         # connect our log messages to the systemd service.  This is critical
  66         # for capturing all the log messages if the scrub fails, because the
  67         # fail service uses the service name to gather log messages for the
  68         # error report.
  69         if [ -n "${SERVICE_MODE}" ]; then
  70                 test "${ret}" -ne 0 && ret=1
  71                 sleep 2
  72         fi
  73
  74         exit "${ret}"
  75 }
  76
  77 while getopts "nrtV" opt; do
  78     case "${opt}" in
  79         "n") DBG="echo Would execute: " ;;
  80         "r") reap=1;;
  81         "t") fstrim=1;;
  82         "V") print_version; exitcode 0;;
  83         *) print_help; exitcode 2;;
  84         esac
  85 done
  86 shift "$((OPTIND - 1))"
  87
  88 arg="$1"
  89 if [ -z "${arg}" ]; then
  90         print_help
  91         exitcode 1
  92 fi
  93
  94 if ! type lsblk >& /dev/null ; then
  95     echo "e2scrub: can't find lsblk --- is util-linux installed?"
  96     exitcode 1
  97 fi
  98
  99 if ! type lvcreate >& /dev/null ; then
 100     echo "e2scrub: can't find lvcreate --- is lvm2 installed?"
 101     exitcode 1
 102 fi
 103
 104 # Find the device for a given mountpoint
 105 dev_from_mount() {
 106         local mountpt="$(realpath "$1")"
 107
 108         lsblk -o NAME,FSTYPE,MOUNTPOINT -p -P -n 2> /dev/null | while read vars; do
 109                 eval "${vars}"
 110                 if [ "${mountpt}" != "${MOUNTPOINT}" ]; then
 111                         continue
 112                 fi
 113                 case "${FSTYPE}" in
 114                 ext[234])
 115                         echo "${NAME}"
 116                         return 0
 117                         ;;
 118                 esac
 119         done
 120         return 1
 121 }
 122
 123 # Check a device argument
 124 dev_from_arg() {
 125         local dev="$1"
 126         local fstype="$(lsblk -o FSTYPE -n "${dev}" 2> /dev/null)"
 127
 128         case "${fstype}" in
 129         ext[234])
 130                 echo "${dev}"
 131                 return 0
 132                 ;;
 133         esac
 134         return 1
 135 }
 136
 137 mnt_from_dev() {
 138         local dev="$1"
 139
 140         if [ -n "${dev}" ]; then
 141                 lsblk -o MOUNTPOINT -n "${dev}"
 142         fi
 143 }
 144
 145 # Construct block device path and mountpoint from argument
 146 if [ -b "${arg}" ]; then
 147         dev="$(dev_from_arg "${arg}")"
 148         mnt="$(mnt_from_dev "${dev}")"
 149 else
 150         dev="$(dev_from_mount "${arg}")"
 151         mnt="${arg}"
 152 fi
 153 if [ ! -e "${dev}" ]; then
 154         echo "${arg}: Not an ext[234] filesystem."
 155         print_help
 156         exitcode 16
 157 fi
 158
 159 # Make sure this is an LVM device we can snapshot
 160 lvm_vars="$(lvs --nameprefixes -o name,vgname,lv_role --noheadings "${dev}" 2> /dev/null)"
 161 eval "${lvm_vars}"
 162 if [ -z "${LVM2_VG_NAME}" ] || [ -z "${LVM2_LV_NAME}" ] ||
 163    echo "${LVM2_LV_ROLE}" | grep -q "snapshot"; then
 164         echo "${arg}: Not connnected to an LVM logical volume."
 165         print_help
 166         exitcode 16
 167 fi
 168 start_time="$(date +'%Y%m%d%H%M%S')"
 169 snap="${LVM2_LV_NAME}.e2scrub"
 170 snap_dev="/dev/${LVM2_VG_NAME}/${snap}"
 171
 172 teardown() {
 173         # Remove and wait for removal to succeed.
 174         ${DBG} lvremove -f "${LVM2_VG_NAME}/${snap}" 3>&-
 175         while [ -e "${snap_dev}" ] && [ "$?" -eq "5" ]; do
 176                 sleep 0.5
 177                 ${DBG} lvremove -f "${LVM2_VG_NAME}/${snap}" 3>&-
 178         done
 179 }
 180
 181 check() {
 182         # First we recover the journal, then we see if e2fsck tries any
 183         # non-optimization repairs.  If either of these two returns a
 184         # non-zero status (errors fixed or remaining) then this fs is bad.
 185         E2FSCK_FIXES_ONLY=1
 186         export E2FSCK_FIXES_ONLY
 187         ${DBG} "@root_sbindir@/e2fsck" -E journal_only -p ${e2fsck_opts} "${snap_dev}" || return $?
 188         ${DBG} "@root_sbindir@/e2fsck" -f -y ${e2fsck_opts} "${snap_dev}"
 189 }
 190
 191 mark_clean() {
 192         ${DBG} "@root_sbindir@/tune2fs" -C 0 -T "${start_time}" "${dev}"
 193 }
 194
 195 mark_corrupt() {
 196         ${DBG} "@root_sbindir@/tune2fs" -E force_fsck "${dev}"
 197 }
 198
 199 setup() {
 200         # Try to remove snapshot for 30s, bail out if we can't remove it.
 201         lveremove_deadline="$(( $(date "+%s") + 30))"
 202         ${DBG} lvremove -f "${LVM2_VG_NAME}/${snap}" 3>&- 2>/dev/null
 203         while [ -e "${snap_dev}" ] && [ "$?" -eq "5" ] &&
 204               [ "$(date "+%s")" -lt "${lvremove_deadline}" ]; do
 205                 sleep 0.5
 206                 ${DBG} lvremove -f "${LVM2_VG_NAME}/${snap}" 3>&-
 207         done
 208         if [ -e "${snap_dev}" ]; then
 209                 echo "${arg}: e2scrub snapshot is in use, cannot check!"
 210                 return 1
 211         fi
 212         # Create the snapshot, wait for device to appear.
 213         ${DBG} lvcreate -s -L "${snap_size_mb}m" -n "${snap}" "${LVM2_VG_NAME}/${LVM2_LV_NAME}" 3>&-
 214         if [ $? -ne 0 ]; then
 215                 echo "${arg}: e2scrub snapshot FAILED, will not check!"
 216                 return 1
 217         fi
 218         ${DBG} udevadm settle 2> /dev/null
 219         return 0
 220 }
 221
 222 if [ "${reap}" -gt 0 ]; then
 223         if [ -e "${snap_dev}" ]; then
 224                 teardown 2> /dev/null
 225         fi
 226         exit 0
 227 fi
 228 if ! setup; then
 229         exitcode 8
 230 fi
 231 trap "teardown; exit 1" EXIT INT QUIT TERM
 232
 233 # Check and react
 234 check
 235 case "$?" in
 236 "0")
 237         # Clean check!
 238         echo "${arg}: Scrub succeeded."
 239         mark_clean
 240         teardown
 241         trap '' EXIT
 242
 243         # Trim the free space, which requires the snapshot be deleted.
 244         if [ "${fstrim}" -eq 1 ] && [ -d "${mnt}" ] && type fstrim > /dev/null 2>&1; then
 245                 echo "${arg}: Trimming free space."
 246                 fstrim -v "${mnt}"
 247         fi
 248
 249         ret=0
 250         ;;
 251 "8")
 252         # Operational error, what now?
 253         echo "${arg}: e2fsck operational error."
 254         teardown
 255         trap '' EXIT
 256         ret=8
 257         ;;
 258 *)
 259         # fsck failed.  Check if the snapshot is invalid; if so, make a
 260         # note of that at the end of the log.  This isn't necessarily a
 261         # failure because the mounted fs could have overflowed the
 262         # snapshot with regular disk writes /or/ our repair process
 263         # could have done it by repairing too much.
 264         #
 265         # If it's really corrupt we ought to fsck at next boot.
 266         is_invalid="$(lvs -o lv_snapshot_invalid --noheadings "${snap_dev}" | awk '{print $1}')"
 267         if [ -n "${is_invalid}" ]; then
 268                 echo "${arg}: Scrub FAILED due to invalid snapshot."
 269                 ret=8
 270         else
 271                 echo "${arg}: Scrub FAILED due to corruption!  Unmount and run e2fsck -y."
 272                 mark_corrupt
 273                 ret=6
 274         fi
 275         teardown
 276         trap '' EXIT
 277         ;;
 278 esac
 279
 280 exitcode "${ret}"