import CS rear-2.6-12.el8

This commit is contained in:
eabdullin 2024-03-27 20:25:12 +00:00
parent 19a56f1c28
commit 31af81df28
6 changed files with 878 additions and 1 deletions

View File

@ -0,0 +1,32 @@
From 89b61793d80bc2cb2abe47a7d0549466fb087d16 Mon Sep 17 00:00:00 2001
From: Johannes Meixner <jsmeix@suse.com>
Date: Fri, 12 Jan 2024 08:04:40 +0100
Subject: [PATCH] Make initrd accessible only by root (#3123)
In pack/GNU/Linux/900_create_initramfs.sh call
chmod 0600 "$TMP_DIR/$REAR_INITRD_FILENAME"
to let only 'root' access the ReaR initrd because
the ReaR recovery system in the initrd can contain secrets
(not by default but when certain things are explicitly
configured by the user like SSH keys without passphrase)
see https://github.com/rear/rear/issues/3122
and https://bugzilla.opensuse.org/show_bug.cgi?id=1218728
---
usr/share/rear/pack/GNU/Linux/900_create_initramfs.sh | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/usr/share/rear/pack/GNU/Linux/900_create_initramfs.sh b/usr/share/rear/pack/GNU/Linux/900_create_initramfs.sh
index 1e0c11039..12be718ed 100644
--- a/usr/share/rear/pack/GNU/Linux/900_create_initramfs.sh
+++ b/usr/share/rear/pack/GNU/Linux/900_create_initramfs.sh
@@ -125,4 +125,10 @@ case "$REAR_INITRD_COMPRESSION" in
fi
;;
esac
+
+# Only root should be allowed to access the initrd
+# because the ReaR recovery system can contain secrets
+# cf. https://github.com/rear/rear/issues/3122
+test -s "$TMP_DIR/$REAR_INITRD_FILENAME" && chmod 0600 "$TMP_DIR/$REAR_INITRD_FILENAME"
+
popd >/dev/null

View File

@ -0,0 +1,569 @@
diff --git a/usr/share/rear/finalize/Linux-i386/630_install_grub.sh b/usr/share/rear/finalize/Linux-i386/630_install_grub.sh
index f3d9a8204..a0e87e1db 100644
--- a/usr/share/rear/finalize/Linux-i386/630_install_grub.sh
+++ b/usr/share/rear/finalize/Linux-i386/630_install_grub.sh
@@ -1,22 +1,18 @@
# This script is an improvement over the default grub-install '(hd0)'
#
-# However the following issues still exist:
+# However the following issue still exists:
#
# * We don't know what the first disk will be, so we cannot be sure the MBR
-# is written to the correct disk(s). That's why we make all disks bootable.
-#
-# * There is no guarantee that GRUB was the boot loader used originally.
-# One possible attempt would be to save and restore the MBR for each disk,
-# but this does not guarantee a correct boot order,
-# or even a working boot loader config
-# (eg. GRUB stage2 might not be at the exact same location).
+# is written to the correct disk(s). That's why we make all suitable disks bootable.
# Skip if another boot loader is already installed
# (then $NOBOOTLOADER is not a true value cf. finalize/default/010_prepare_checks.sh):
is_true $NOBOOTLOADER || return 0
-# For UEFI systems with grub legacy with should use efibootmgr instead:
-is_true $USING_UEFI_BOOTLOADER && return
+# For UEFI systems with grub legacy with should use efibootmgr instead,
+# but if BOOTLOADER is explicitly set to GRUB, we are on a hybrid (BIOS/UEFI)
+# boot system and we need to install GRUB to MBR as well.
+# Therefore, we don't test $USING_UEFI_BOOTLOADER.
# If the BOOTLOADER variable (read by finalize/default/010_prepare_checks.sh)
# is not "GRUB" (which means GRUB Legacy) skip this script (which is only for GRUB Legacy)
@@ -25,31 +21,27 @@ is_true $USING_UEFI_BOOTLOADER && return
test "GRUB" = "$BOOTLOADER" || return 0
# If the BOOTLOADER variable is "GRUB" (which means GRUB Legacy)
-# do not unconditionally trust that because https://github.com/rear/rear/pull/589
-# reads (excerpt):
-# Problems found:
-# The ..._install_grub.sh checked for GRUB2 which is not part
-# of the first 2048 bytes of a disk - only GRUB was present -
-# thus the check for grub-probe/grub2-probe
-# and https://github.com/rear/rear/commit/079de45b3ad8edcf0e3df54ded53fe955abded3b
-# reads (excerpt):
-# replace grub-install by grub-probe
-# as grub-install also exist in legacy grub
-# so that it seems there are cases where actually GRUB 2 is used
-# but wrongly detected as "GRUB" so that another test is needed
-# to detected if actually GRUB 2 is used and that test is to
-# check if grub-probe or grub2-probe is installed because
-# grub-probe or grub2-probe is only installed in case of GRUB 2
-# and when GRUB 2 is installed we assume GRUB 2 is used as boot loader
-# so that then we skip this script (which is only for GRUB Legacy)
-# because finalize/Linux-i386/660_install_grub2.sh is for installing GRUB 2:
-if type -p grub-probe >&2 || type -p grub2-probe >&2 ; then
- LogPrint "Skip installing GRUB Legacy boot loader because GRUB 2 is installed (grub-probe or grub2-probe exist)."
+# we could in principle trust that and continue because
+# layout/save/default/445_guess_bootloader.sh (where the value has been set)
+# is now able to distinguish between GRUB Legacy and GRUB 2.
+# But, as this code used to support the value "GRUB" for GRUB 2,
+# the user can have BOOTLOADER=GRUB set explicitly in the configuration file
+# and then it overrides the autodetection in layout/save/default/445_guess_bootloader.sh .
+# The user expects this setting to work with GRUB 2, thus for backward compatibility
+# we need to take into accout the possibility that GRUB actually means GRUB 2.
+if is_grub2_installed ; then
+ LogPrint "Skip installing GRUB Legacy boot loader because GRUB 2 is installed."
+ # We have the ErrorIfDeprecated function, but it aborts ReaR by default,
+ # which is not a good thing to do during recovery.
+ # Therefore it better to log a warning and continue.
+ LogPrintError "WARNING: setting BOOTLOADER=GRUB for GRUB 2 is deprecated, set BOOTLOADER=GRUB2 if setting BOOTLOADER explicitly"
return
fi
# The actual work:
LogPrint "Installing GRUB Legacy boot loader:"
+# See above for the reasoning why not to use ErrorIfDeprecated
+LogPrintError "WARNING: support for GRUB Legacy is deprecated"
# Installing GRUB Legacy boot loader requires an executable "grub":
type -p grub >&2 || Error "Cannot install GRUB Legacy boot loader because there is no 'grub' program."
@@ -79,8 +71,10 @@ if [[ -r "$LAYOUT_FILE" && -r "$LAYOUT_DEPS" ]] ; then
for disk in $disks ; do
# Installing grub on an LVM PV will wipe the metadata so we skip those
- # function is_disk_a_pv returns with 1 if disk is a PV
- is_disk_a_pv "$disk" || continue
+ # function is_disk_a_pv returns true if disk is a PV
+ is_disk_a_pv "$disk" && continue
+ # Is the disk suitable for GRUB installation at all?
+ is_disk_grub_candidate "$disk" || continue
# Use first boot partition by default
part=$( echo $bootparts | cut -d' ' -f1 )
diff --git a/usr/share/rear/finalize/Linux-i386/660_install_grub2.sh b/usr/share/rear/finalize/Linux-i386/660_install_grub2.sh
index 58163d622..f42b5bfbe 100644
--- a/usr/share/rear/finalize/Linux-i386/660_install_grub2.sh
+++ b/usr/share/rear/finalize/Linux-i386/660_install_grub2.sh
@@ -38,6 +38,37 @@
# so that after "rear recover" finished he can manually install the bootloader
# as appropriate for his particular system.
+local grub_name
+local grub2_install_failed grub2_install_device
+local source_disk target_disk junk
+local grub2_installed_disks
+local part bootparts
+local disk disks bootdisk
+
+function bios_grub_install ()
+{
+ local grub2_install_device="$1"
+
+ if is_true $USING_UEFI_BOOTLOADER ; then
+ # If running under UEFI, we need to specify the target explicitly, otherwise grub-install thinks
+ # that we are installing the EFI bootloader.
+ if ! chroot $TARGET_FS_ROOT /bin/bash --login -c "$grub_name-install --target=i386-pc $grub2_install_device" ; then
+ LogPrintError "Failed to install GRUB2 for BIOS boot (target i386-pc) on $bootdisk"
+ # purely informational test that may help to explain the reason for the error
+ if ! test -d "$TARGET_FS_ROOT/boot/$grub_name/i386-pc" ; then
+ LogPrintError "GRUB2 module dir for BIOS boot (boot/$grub_name/i386-pc in $TARGET_FS_ROOT) does not exist, is GRUB2 for BIOS (target i386-pc) installed?"
+ fi
+ return 1
+ fi
+ else
+ if ! chroot $TARGET_FS_ROOT /bin/bash --login -c "$grub_name-install $grub2_install_device" ; then
+ LogPrintError "Failed to install GRUB2 on $grub2_install_device"
+ return 1
+ fi
+ fi
+ return 0
+}
+
# Skip if another bootloader was already installed:
# In this case NOBOOTLOADER is not true,
# cf. finalize/default/050_prepare_checks.sh
@@ -45,12 +76,16 @@ is_true $NOBOOTLOADER || return 0
# For UEFI systems with grub2 we should use efibootmgr instead,
# cf. finalize/Linux-i386/670_run_efibootmgr.sh
-is_true $USING_UEFI_BOOTLOADER && return
+# but if BOOTLOADER is explicitly set to GRUB2, we are on a hybrid (BIOS/UEFI)
+# boot system and we need to install GRUB to MBR as well
+if is_true $USING_UEFI_BOOTLOADER && [ "GRUB2" != "$BOOTLOADER" ] ; then
+ return 0
+fi
# Only for GRUB2 - GRUB Legacy will be handled by its own script.
# GRUB2 is detected by testing for grub-probe or grub2-probe which does not exist in GRUB Legacy.
# If neither grub-probe nor grub2-probe is there assume GRUB2 is not there:
-type -p grub-probe || type -p grub2-probe || return 0
+is_grub2_installed || return 0
LogPrint "Installing GRUB2 boot loader..."
@@ -94,7 +129,7 @@ if test "$GRUB2_INSTALL_DEVICES" ; then
else
LogPrint "Installing GRUB2 on $grub2_install_device (specified in GRUB2_INSTALL_DEVICES)"
fi
- if ! chroot $TARGET_FS_ROOT /bin/bash --login -c "$grub_name-install $grub2_install_device" ; then
+ if ! bios_grub_install "$grub2_install_device" ; then
LogPrintError "Failed to install GRUB2 on $grub2_install_device"
grub2_install_failed="yes"
fi
@@ -138,8 +173,8 @@ fi
grub2_installed_disks=()
for disk in $disks ; do
# Installing GRUB2 on an LVM PV will wipe the metadata so we skip those:
- # function is_disk_a_pv returns with 1 if disk is a PV
- is_disk_a_pv "$disk" || continue
+ # function is_disk_a_pv returns true if disk is a PV
+ is_disk_a_pv "$disk" && continue
# Use first boot partition by default:
part=$( echo $bootparts | cut -d' ' -f1 )
@@ -158,6 +193,8 @@ for disk in $disks ; do
# Install GRUB2 on the boot disk if one was found:
if test "$bootdisk" ; then
+ # Is the disk suitable for GRUB installation at all?
+ is_disk_grub_candidate "$bootdisk" || continue
# Continue with the next possible boot disk when GRUB2 was already installed on the current one.
# When there are more disks like /dev/sda and /dev/sdb it can happen that
# for /dev/sda bootdisk=/dev/sda and GRUB2 gets installed on /dev/sda and
@@ -165,7 +202,7 @@ for disk in $disks ; do
# so we avoid that GRUB2 gets needlessly installed two times on the same device:
IsInArray "$bootdisk" "${grub2_installed_disks[@]}" && continue
LogPrint "Found possible boot disk $bootdisk - installing GRUB2 there"
- if chroot $TARGET_FS_ROOT /bin/bash --login -c "$grub_name-install $bootdisk" ; then
+ if bios_grub_install "$bootdisk" ; then
grub2_installed_disks+=( "$bootdisk" )
# In contrast to the above behaviour when GRUB2_INSTALL_DEVICES is specified
# consider it here as a successful bootloader installation when GRUB2
@@ -174,11 +211,14 @@ for disk in $disks ; do
# Continue with the next possible boot disk:
continue
fi
- LogPrintError "Failed to install GRUB2 on possible boot disk $bootdisk"
fi
done
is_true $NOBOOTLOADER || return 0
-LogPrintError "Failed to install GRUB2 - you may have to manually install it"
+if is_true $USING_UEFI_BOOTLOADER ; then
+ LogPrintError "Failed to install GRUB2 for BIOS boot - you may have to manually install it to preserve the hybrid BIOS/UEFI boot support, otherwise only UEFI boot will work"
+else
+ LogPrintError "Failed to install GRUB2 - you may have to manually install it"
+fi
return 1
diff --git a/usr/share/rear/finalize/default/050_prepare_checks.sh b/usr/share/rear/finalize/default/050_prepare_checks.sh
index 1679c9a41..57b44bca4 100644
--- a/usr/share/rear/finalize/default/050_prepare_checks.sh
+++ b/usr/share/rear/finalize/default/050_prepare_checks.sh
@@ -10,10 +10,18 @@
NOBOOTLOADER=1
# Try to read the BOOTLOADER value if /var/lib/rear/recovery/bootloader is not empty.
-# Currently (June 2016) the used BOOTLOADER values (grep for '$BOOTLOADER') are:
+# Currently (February 2024) the used BOOTLOADER values (grep for '$BOOTLOADER') are:
# GRUB for GRUB Legacy
# GRUB2 for GRUB 2
# ELILO for elilo
+# LILO for lilo
+# GRUB2-EFI for GRUB 2, EFI version
+# EFI for any EFI bootloader, dummy value
+# ARM for ARM devices, dummy value
+# ARM-ALLWINNER for Allwinner devices
+# ZIPL for zIPL, on IBM Z (s390x)
+# PPC for any bootloader in the PReP boot partition (can be LILO, YABOOT, GRUB2)
+
local bootloader_file="$VAR_DIR/recovery/bootloader"
# The output is stored in an artificial bash array so that $BOOTLOADER is the first word:
test -s $bootloader_file && BOOTLOADER=( $( grep -v '^[[:space:]]*#' $bootloader_file ) )
diff --git a/usr/share/rear/layout/save/default/445_guess_bootloader.sh b/usr/share/rear/layout/save/default/445_guess_bootloader.sh
index fd5267dcf..b9e636a02 100644
--- a/usr/share/rear/layout/save/default/445_guess_bootloader.sh
+++ b/usr/share/rear/layout/save/default/445_guess_bootloader.sh
@@ -1,7 +1,15 @@
# Determine or guess the used bootloader if not specified by the user
# and save this information into /var/lib/rear/recovery/bootloader
-bootloader_file="$VAR_DIR/recovery/bootloader"
+local bootloader_file="$VAR_DIR/recovery/bootloader"
+
+local sysconfig_bootloader
+local block_device
+local blockd
+local disk_device
+local bootloader_area_strings_file
+local block_size
+local known_bootloader
# When BOOTLOADER is specified use that:
if test "$BOOTLOADER" ; then
@@ -57,39 +65,31 @@ for block_device in /sys/block/* ; do
# Continue guessing the used bootloader by inspecting the first bytes on the next disk:
continue
fi
- # 'Hah!IdontNeedEFI' is the ASCII representation of the official GUID number
- # for a GPT BIOS boot partition which is 21686148-6449-6E6F-744E-656564454649
- # see https://en.wikipedia.org/wiki/BIOS_boot_partition (issue #1752).
- # Use single quotes for 'Hah!IdontNeedEFI' to be on the safe side
- # because with double quotes the ! would cause history expansion if that is enabled
- # (non-interactive shells do not perform history expansion by default but better safe than sorry):
- if grep -q 'Hah!IdontNeedEFI' $bootloader_area_strings_file ; then
- # Because 'Hah!IdontNeedEFI' contains the known bootloader 'EFI'
- # the default code below would falsely guess that 'EFI' is used
- # but actually another non-EFI bootloader is used here
- # cf. https://github.com/rear/rear/issues/1752#issue-303856221
- # so that in the 'Hah!IdontNeedEFI' case only non-EFI bootloaders are tested.
- # IBM Z (s390) uses zipl boot loader for RHEL and Ubuntu
- # cf. https://github.com/rear/rear/issues/2137
- for known_bootloader in GRUB2 GRUB ELILO LILO ZIPL ; do
- if grep -q -i "$known_bootloader" $bootloader_area_strings_file ; then
- LogPrint "Using guessed bootloader '$known_bootloader' (found in first bytes on $disk_device with GPT BIOS boot partition)"
- echo "$known_bootloader" >$bootloader_file
- return
- fi
- done
- # When in the 'Hah!IdontNeedEFI' case no known non-EFI bootloader is found
- # continue guessing the used bootloader by inspecting the first bytes on the next disk
- # because otherwise the default code below would falsely guess that 'EFI' is used
- # cf. https://github.com/rear/rear/pull/1754#issuecomment-383531597
- continue
- fi
# Check the default cases of known bootloaders.
# IBM Z (s390) uses zipl boot loader for RHEL and Ubuntu
# cf. https://github.com/rear/rear/issues/2137
- for known_bootloader in GRUB2-EFI EFI GRUB2 GRUB ELILO LILO ZIPL ; do
+ for known_bootloader in GRUB2 GRUB LILO ZIPL ; do
if grep -q -i "$known_bootloader" $bootloader_area_strings_file ; then
+ # If we find "GRUB" (which means GRUB Legacy)
+ # do not unconditionally trust that because https://github.com/rear/rear/pull/589
+ # reads (excerpt):
+ # Problems found:
+ # The ..._install_grub.sh checked for GRUB2 which is not part
+ # of the first 2048 bytes of a disk - only GRUB was present -
+ # thus the check for grub-probe/grub2-probe
+ # and https://github.com/rear/rear/commit/079de45b3ad8edcf0e3df54ded53fe955abded3b
+ # reads (excerpt):
+ # replace grub-install by grub-probe
+ # as grub-install also exist in legacy grub
+ # so that if actually GRUB 2 is used, the string in the bootloader area
+ # is "GRUB" so that another test is needed to detect if actually GRUB 2 is used.
+ # When GRUB 2 is installed we assume GRUB 2 is used as boot loader.
+ if [ "$known_bootloader" = "GRUB" ] && is_grub2_installed ; then
+ known_bootloader=GRUB2
+ LogPrint "GRUB found in first bytes on $disk_device and GRUB 2 is installed, using GRUB2 as a guessed bootloader for 'rear recover'"
+ else
LogPrint "Using guessed bootloader '$known_bootloader' (found in first bytes on $disk_device)"
+ fi
echo "$known_bootloader" >$bootloader_file
return
fi
@@ -103,6 +103,26 @@ for block_device in /sys/block/* ; do
Log "End of strings in the first bytes on $disk_device"
done
+# No bootloader detected, but we are using UEFI - there is probably an EFI bootloader
+if is_true $USING_UEFI_BOOTLOADER ; then
+ if is_grub2_installed ; then
+ echo "GRUB2-EFI" >$bootloader_file
+ elif test -f /sbin/elilo ; then
+ echo "ELILO" >$bootloader_file
+ else
+ # There is an EFI bootloader, we don't know which one exactly.
+ # The value "EFI" is a bit redundant with USING_UEFI_BOOTLOADER=1,
+ # which already indicates that there is an EFI bootloader. We use it as a placeholder
+ # to not leave $bootloader_file empty.
+ # Note that it is legal to have USING_UEFI_BOOTLOADER=1 and e.g. known_bootloader=GRUB2
+ # (i.e. a non=EFI bootloader). This will happen in BIOS/UEFI hybrid boot scenarios.
+ # known_bootloader=GRUB2 indicates that there is a BIOS bootloader and USING_UEFI_BOOTLOADER=1
+ # indicates that there is also an EFI bootloader. Only the EFI one is being used at this
+ # time, but both will need to be restored.
+ echo "EFI" >$bootloader_file
+ fi
+ return 0
+fi
# Error out when no bootloader was specified or could be autodetected:
Error "Cannot autodetect what is used as bootloader, see default.conf about 'BOOTLOADER'"
diff --git a/usr/share/rear/lib/bootloader-functions.sh b/usr/share/rear/lib/bootloader-functions.sh
index 5402f1da0..7aa40a589 100644
--- a/usr/share/rear/lib/bootloader-functions.sh
+++ b/usr/share/rear/lib/bootloader-functions.sh
@@ -491,6 +491,53 @@ function get_root_disk_UUID {
echo $(mount | grep ' on / ' | awk '{print $1}' | xargs blkid -s UUID -o value)
}
+# Detect whether actually GRUB 2 is installed and that test is to
+# check if grub-probe or grub2-probe is installed because
+# grub-probe or grub2-probe is only installed in case of GRUB 2.
+# Needed because one can't tell the GRUB version by looking at the MBR
+# (both GRUB 2 and GRUB Legacy have the string "GRUB" in their MBR).
+function is_grub2_installed () {
+ if type -p grub-probe >&2 || type -p grub2-probe >&2 ; then
+ Log "GRUB 2 is installed (grub-probe or grub2-probe exist)."
+ return 0
+ else
+ return 1
+ fi
+}
+
+# Determine whether a disk is worth detecting or installing GRUB on
+function is_disk_grub_candidate () {
+ local disk="$1"
+ local disk_partitions part
+ local label flags
+
+ # ToDo : validate $disk (does it even exist? Isn't it write-protected?)
+
+ # Installing grub on an LVM PV will wipe the metadata so we skip those
+ is_disk_a_pv "$disk" && return 1
+
+ label="$( get_disklabel_type "$disk" )" || return 1
+ # We don't care about the SUSE-specific 'gpt_sync_mbr' partition scheme
+ # anymore: https://github.com/rear/rear/pull/3145#discussion_r1481388431
+ if [ "$label" == gpt ] ; then
+ # GPT needs a special BIOS boot partition to store GRUB (BIOS version).
+ # Let's try to find it. It can be recognized as having the bios_grub flag.
+ disk_partitions=( $( get_child_components "$disk" "part" ) )
+ for part in "${disk_partitions[@]}" ; do
+ flags=( $( get_partition_flags "$part" ) )
+ IsInArray bios_grub "${flags[@]}" && return 0 # found!
+ done
+ # If a given GPT-partitioned disk does not contain a BIOS boot partition,
+ # GRUB for BIOS booting can not be installed into its MBR (grub-install errors out).
+ return 1
+ else
+ # Other disklabel types don't need anything special to install GRUB.
+ # The test for the PReP boot partition (finalize/Linux-ppc64le/660_install_grub2.sh)
+ # is a bit similar, but operates on the partition itself, not on the uderlying disk.
+ return 0
+ fi
+}
+
# Create configuration grub
function create_grub2_cfg {
root_uuid=$(get_root_disk_UUID)
diff --git a/usr/share/rear/lib/checklayout-workflow.sh b/usr/share/rear/lib/checklayout-workflow.sh
index 94b70fc06..744ca0be1 100644
--- a/usr/share/rear/lib/checklayout-workflow.sh
+++ b/usr/share/rear/lib/checklayout-workflow.sh
@@ -15,6 +15,10 @@ function WORKFLOW_checklayout () {
SourceStage "layout/precompare"
+ # layout code needs to know whether we are using UEFI (USING_UEFI_BOOTLOADER)
+ # as it also detects the bootloader in use ( layout/save/default/445_guess_bootloader.sh )
+ Source $SHARE_DIR/prep/default/320_include_uefi_env.sh
+
# In case of e.g. BACKUP_URL=file:///mybackup/ automatically exclude the matching component 'fs:/mybackup'
# otherwise 'rear checklayout' would always detect a changed layout with BACKUP_URL=file:///...
# because during 'rear mkrescue/mkbackup' such a component was automatically excluded this way
diff --git a/usr/share/rear/lib/layout-functions.sh b/usr/share/rear/lib/layout-functions.sh
index 4f5b8f6f8..f5fc7538e 100644
--- a/usr/share/rear/lib/layout-functions.sh
+++ b/usr/share/rear/lib/layout-functions.sh
@@ -520,6 +520,33 @@ get_component_type() {
grep -E "^[^ ]+ $1 " $LAYOUT_TODO | cut -d " " -f 3
}
+# Get the disklabel (partition table) type of the disk $1 from the layout file
+# (NOT from the actual disk, so layout file must exist before calling this,
+# and it is useful during recovery even before the disk layout has been recreated)
+function get_disklabel_type () {
+ # from create_disk() in layout/prepare/GNU/Linux/100_include_partition_code.sh
+ local component disk size label junk
+
+ disk=''
+
+ read component disk size label junk < <(grep "^disk $1 " "$LAYOUT_FILE")
+ test $disk || return 1
+
+ echo $label
+}
+
+# Get partition flags from layout (space-separated) of partition given as $1
+function get_partition_flags () {
+ local part disk size pstart name flags partition junk
+
+ while read part disk size pstart name flags partition junk; do
+ if [ "$partition" == "$1" ] ; then
+ echo "$flags" | tr ',' ' '
+ return 0
+ fi
+ done < <(grep "^part " $LAYOUT_FILE)
+}
+
# Function returns 0 when v1 is greater or equal than v2
version_newer() {
local v1list=( ${1//[-.]/ } )
@@ -794,17 +821,17 @@ blkid_label_of_device() {
echo "$label"
}
-# Returns 1 if the device is an LVM physical volume
-# Returns 0 otherwise or if the device doesn't exists
+# Returns true if the device is an LVM physical volume
+# Returns false otherwise or if the device doesn't exists
is_disk_a_pv() {
disk=$1
# Using awk, select the 'lvmdev' line for which $disk is the device (column 3),
# cf. https://github.com/rear/rear/pull/1897
# If exit == 1, then there is such line (so $disk is a PV),
- # otherwise exit with default value '0', which falls through to 'return 0' below.
- awk "\$1 == \"lvmdev\" && \$3 == \"${disk}\" { exit 1 }" "$LAYOUT_FILE" >/dev/null || return 1
- return 0
+ # otherwise exit with default value '0', which falls through to 'return 1' below.
+ awk "\$1 == \"lvmdev\" && \$3 == \"${disk}\" { exit 1 }" "$LAYOUT_FILE" >/dev/null || return 0
+ return 1
}
function is_multipath_path {
diff --git a/usr/share/rear/lib/savelayout-workflow.sh b/usr/share/rear/lib/savelayout-workflow.sh
index 69cda58e9..27bb0a1ad 100644
--- a/usr/share/rear/lib/savelayout-workflow.sh
+++ b/usr/share/rear/lib/savelayout-workflow.sh
@@ -10,6 +10,10 @@ if [[ "$VERBOSE" ]]; then
fi
WORKFLOWS+=( savelayout )
WORKFLOW_savelayout () {
+ # layout code needs to know whether we are using UEFI (USING_UEFI_BOOTLOADER)
+ # as it also detects the bootloader in use ( layout/save/default/445_guess_bootloader.sh )
+ Source $SHARE_DIR/prep/default/320_include_uefi_env.sh
+
#DISKLAYOUT_FILE=$VAR_DIR/layout/disklayout.conf # defined in default.conf now (issue #678)
SourceStage "layout/save"
}
diff --git a/usr/share/rear/prep/GNU/Linux/300_include_grub_tools.sh b/usr/share/rear/prep/GNU/Linux/300_include_grub_tools.sh
index fcf0a5ff6..7d494281a 100644
--- a/usr/share/rear/prep/GNU/Linux/300_include_grub_tools.sh
+++ b/usr/share/rear/prep/GNU/Linux/300_include_grub_tools.sh
@@ -1,8 +1,6 @@
#
# GRUB2 has much more commands than the legacy grub command, including modules
-test -d $VAR_DIR/recovery || mkdir -p $VAR_DIR/recovery
-
# cf. https://github.com/rear/rear/issues/2137
# s390 zlinux does not use grub
# *********************************************************************************
@@ -11,19 +9,8 @@ test -d $VAR_DIR/recovery || mkdir -p $VAR_DIR/recovery
# *********************************************************************************
[ "$ARCH" == "Linux-s390" ] && return 0
-# Because usr/sbin/rear sets 'shopt -s nullglob' the 'echo -n' command
-# outputs nothing if nothing matches the bash globbing pattern '/boot/grub*'
-local grubdir="$( echo -n /boot/grub* )"
-# Use '/boot/grub' as fallback if nothing matches '/boot/grub*'
-test -d "$grubdir" || grubdir='/boot/grub'
-
-# Check if we're using grub or grub2 before doing something.
-if has_binary grub-probe ; then
- grub-probe -t device $grubdir >$VAR_DIR/recovery/bootdisk 2>/dev/null || return 0
-elif has_binary grub2-probe ; then
- grub2-probe -t device $grubdir >$VAR_DIR/recovery/bootdisk 2>/dev/null || return 0
-fi
-
+# It is safe to assume that we are using GRUB and try to add these files to the rescue image
+# even if the assumption is wrong.
# Missing programs in the PROGS array are ignored:
PROGS+=( grub-bios-setup grub2-bios-setup
grub-install grub2-install
diff --git a/usr/share/rear/prep/Linux-s390/305_include_s390_tools.sh b/usr/share/rear/prep/Linux-s390/305_include_s390_tools.sh
index 084ea3104..4451f53d0 100644
--- a/usr/share/rear/prep/Linux-s390/305_include_s390_tools.sh
+++ b/usr/share/rear/prep/Linux-s390/305_include_s390_tools.sh
@@ -1,10 +1,13 @@
#
# s390 zIPL boot loader and grubby for configuring boot loader`
-test -d $VAR_DIR/recovery || mkdir -p $VAR_DIR/recovery
-
-local bootdir="$( echo -n /boot/ )"
-test -d "$bootdir" || $bootdir='/boot/'
+# See the code in prep/GNU/Linux/300_include_grub_tools.sh
+# that sets grubdir via
+# local grubdir="$( echo -n /boot/grub* )"
+# where 'shopt -s nullglob' results nothing when nothing matches
+# but that is not needed here to set a fixed bootdir="/boot"
+# cf. https://github.com/rear/rear/issues/1040#issuecomment-1034890880
+local bootdir="/boot/"
# cf. https://github.com/rear/rear/issues/2137
# findmnt is used the same as grub-probe to find the device where /boot is mounted
@@ -16,7 +19,7 @@ test -d "$bootdir" || $bootdir='/boot/'
# findmnt returns --> /dev/dasda3[/@/.snapshots/1/snapshot]
# use 300_include_grub_tools.sh instead of this file (grub2-probe)
if has_binary findmnt ; then
- findmnt -no SOURCE --target $bootdir >$VAR_DIR/recovery/bootdisk || return 0
+ findmnt -no SOURCE --target $bootdir > /dev/null || return 0
fi
# Missing programs in the PROGS array are ignored:
diff --git a/usr/share/rear/prep/default/320_include_uefi_env.sh b/usr/share/rear/prep/default/320_include_uefi_env.sh
index ea86af4ca..93e59eae5 100644
--- a/usr/share/rear/prep/default/320_include_uefi_env.sh
+++ b/usr/share/rear/prep/default/320_include_uefi_env.sh
@@ -87,7 +87,3 @@ fi
DebugPrint "Found EFI system partition ${esp_proc_mounts_line[0]} on ${esp_proc_mounts_line[1]} type ${esp_proc_mounts_line[2]}"
USING_UEFI_BOOTLOADER=1
LogPrint "Using UEFI Boot Loader for Linux (USING_UEFI_BOOTLOADER=1)"
-
-# Remember the ESP device node in VAR_DIR/recovery/bootdisk:
-echo "${esp_proc_mounts_line[0]}" >$VAR_DIR/recovery/bootdisk
-

View File

@ -0,0 +1,102 @@
From e7b84271536782fbc8673ef4573e155e1dfa850e Mon Sep 17 00:00:00 2001
From: pcahyna <pcahyna@users.noreply.github.com>
Date: Wed, 1 Nov 2023 12:53:33 +0100
Subject: [PATCH] Merge pull request #3061 from
pcahyna/save-lvm-poolmetadatasize
Save LVM pool metadata volume size in disk layout
---
.../layout/save/GNU/Linux/220_lvm_layout.sh | 39 ++++++++++++-------
1 file changed, 24 insertions(+), 15 deletions(-)
diff --git a/usr/share/rear/layout/save/GNU/Linux/220_lvm_layout.sh b/usr/share/rear/layout/save/GNU/Linux/220_lvm_layout.sh
index f21845df9..42f0e4126 100644
--- a/usr/share/rear/layout/save/GNU/Linux/220_lvm_layout.sh
+++ b/usr/share/rear/layout/save/GNU/Linux/220_lvm_layout.sh
@@ -18,7 +18,7 @@ local already_processed_lvs=()
local lv_layout_supported lvs_fields
local origin lv vg
local layout modules
-local thinpool chunksize stripes stripesize segmentsize
+local thinpool chunksize stripes stripesize segmentsize poolmetadatasize
local kval infokval
local lvs_exit_code
@@ -130,7 +130,7 @@ local lvs_exit_code
echo "# Skipping PV $pdev that is not part of a valid VG (VG '$vgrp' empty or more than one word):"
contains_visible_char "$vgrp" || vgrp='<missing_VG>'
echo "# lvmdev /dev/$vgrp $pdev $uuid $size"
- # Continue with the next line in the output of "lvm pvdisplay -c"
+ # Continue with the next line in the output of "lvm pvdisplay -C"
continue
fi
# With the above example the output is:
@@ -138,10 +138,10 @@ local lvs_exit_code
echo "lvmdev /dev/$vgrp $pdev $uuid $size"
done
- # Check the exit code of "lvm pvdisplay -c"
- # in the "lvm pvdisplay -c | while read line ; do ... done" pipe:
+ # Check the exit code of "lvm pvdisplay -C"
+ # in the "lvm pvdisplay -C ... | while read line ; do ... done" pipe:
pvdisplay_exit_code=${PIPESTATUS[0]}
- test $pvdisplay_exit_code -eq 0 || Error "LVM command 'lvm pvdisplay -c' failed with exit code $pvdisplay_exit_code"
+ test $pvdisplay_exit_code -eq 0 || Error "LVM command 'lvm pvdisplay -C ... -o pv_name,vg_name,pv_size,pv_uuid' failed with exit code $pvdisplay_exit_code"
# Get the volume group configuration:
# Format: lvmgrp <volume_group> <extentsize> [<size(extents)>] [<size(bytes)>]
@@ -200,17 +200,17 @@ local lvs_exit_code
# Specify the fields for the lvs command depending on whether or not the 'lv_layout' field is supported:
if is_true $lv_layout_supported ; then
- lvs_fields="origin,lv_name,vg_name,lv_size,lv_layout,pool_lv,chunk_size,stripes,stripe_size,seg_size"
+ lvs_fields="origin,lv_name,vg_name,lv_size,lv_layout,pool_lv,chunk_size,stripes,stripe_size,seg_size,lv_metadata_size"
else
# Use the 'modules' field as fallback replacement when the 'lv_layout' field is not supported:
- lvs_fields="origin,lv_name,vg_name,lv_size,modules,pool_lv,chunk_size,stripes,stripe_size,seg_size"
+ lvs_fields="origin,lv_name,vg_name,lv_size,modules,pool_lv,chunk_size,stripes,stripe_size,seg_size,lv_metadata_size"
fi
# Example output of "lvs --separator=':' --noheadings --units b --nosuffix -o $lvs_fields"
- # with lvs_fields="origin,lv_name,vg_name,lv_size,lv_layout,pool_lv,chunk_size,stripes,stripe_size,seg_size"
+ # with lvs_fields="origin,lv_name,vg_name,lv_size,lv_layout,pool_lv,chunk_size,stripes,stripe_size,seg_size,lv_metadata_size"
# i.e. when the 'lv_layout' field is supported:
- # :root:system:19927138304:linear::0:1:0:19927138304
- # :swap:system:1535115264:linear::0:1:0:1535115264
+ # :root:system:19927138304:linear::0:1:0:19927138304:
+ # :swap:system:1535115264:linear::0:1:0:1535115264:
# There are two leading blanks in the output (at least on SLES12-SP4 with LVM 2.02.180).
lvm lvs --separator=':' --noheadings --units b --nosuffix -o $lvs_fields | while read line ; do
@@ -261,14 +261,23 @@ local lvs_exit_code
# With the above example segmentsize=19927138304 and segmentsize=1535115264
segmentsize="$( echo "$line" | awk -F ':' '{ print $10 }' )"
- # TODO: Explain what that code is meant to do.
- # In particular a more explanatory variable name than 'kval' might help.
- # In 110_include_lvm_code.sh there is a comment what 'kval' means there
- # # kval: "key:value" pairs, separated by spaces
- # so probably 'kval' means the same here, but what is 'infokval'?
+ # With the above example poolmetadatasize=""
+ poolmetadatasize="$( echo "$line" | awk -F ':' '{ print $11 }' )"
+
+ # kval is a string of space-separated key:value pairs. Key names are chosen to represent
+ # long options to lvcreate, and value will be the parameter for each long option.
+ # e.g. "chunksize:${chunksize}b" will eventually become a --chunksize=${chunksize}b
+ # argument to lvcreate.
+ # This way 110_include_lvm_code.sh which constructs the arguments to lvcreate
+ # can be kept generic and does not need to be updated every time an argument is added,
+ # as long as the argument can follow this generic scheme.
+ # infokval are key:value pairs that are not used when restoring the layout
+ # and are kept in disklayout.conf only as comments for information
+ # (because the setting is not easy or desirable to preserve).
kval=""
infokval=""
[ -z "$thinpool" ] || kval="${kval:+$kval }thinpool:$thinpool"
+ [ -z "$poolmetadatasize" ] || kval="${kval:+$kval }poolmetadatasize:${poolmetadatasize}b"
[ $chunksize -eq 0 ] || kval="${kval:+$kval }chunksize:${chunksize}b"
[ $stripesize -eq 0 ] || kval="${kval:+$kval }stripesize:${stripesize}b"
[ $segmentsize -eq $size ] || infokval="${infokval:+$infokval }segmentsize:${segmentsize}b"
--
2.43.0

View File

@ -0,0 +1,60 @@
diff --git a/usr/share/rear/layout/save/GNU/Linux/200_partition_layout.sh b/usr/share/rear/layout/save/GNU/Linux/200_partition_layout.sh
index da6ce64c4..ab14ec83f 100644
--- a/usr/share/rear/layout/save/GNU/Linux/200_partition_layout.sh
+++ b/usr/share/rear/layout/save/GNU/Linux/200_partition_layout.sh
@@ -395,6 +395,10 @@ Log "Saving disk partitions."
elif [[ ! ($blockd = *rpmb || $blockd = *[0-9]boot[0-9]) ]]; then # Silently skip Replay Protected Memory Blocks and others
devname=$(get_device_name $disk)
devsize=$(get_disk_size ${disk#/sys/block/})
+ if ! validation_error=$(is_disk_valid $devname) ; then
+ LogPrintError "Ignoring $blockd: $validation_error"
+ continue
+ fi
disktype=$(parted -s $devname print | grep -E "Partition Table|Disk label" | cut -d ":" -f "2" | tr -d " ")
if [ "$disktype" != "dasd" ]; then
echo "# Disk $devname"
diff --git a/usr/share/rear/lib/layout-functions.sh b/usr/share/rear/lib/layout-functions.sh
index f5fc7538e..90b16cb20 100644
--- a/usr/share/rear/lib/layout-functions.sh
+++ b/usr/share/rear/lib/layout-functions.sh
@@ -834,6 +834,40 @@ is_disk_a_pv() {
return 1
}
+# Check whether disk is suitable for being added to layout
+# Can be used to skip obviously unsuitable/broken devices
+# (missing device node, zero size, device can't be opened).
+# Should not be used to skip potential mapping targets before layout restoration
+# - an invalid disk may become valid later, for example if it is a DASD that needs
+# low-level formatting (see 090_include_dasd_code.sh and 360_generate_dasd_format_code.sh),
+# unformatted DASDs show zero size.
+# Returns 0 if the device is ok
+# Returns nonzero code if it should be skipped, and a text describing the error
+# (without the device name) on stdout
+# usage example:
+# local err
+# if ! err=$(is_disk_valid /dev/sda); then
+
+function is_disk_valid {
+ local disk="$1"
+ local size
+
+ if ! test -b "$disk" ; then
+ echo "$disk is not a block device"
+ return 1
+ fi
+ # capture stdout in a variable and redirect stderr to stdout - the error message
+ # will be our output
+ if { size=$(blockdev --getsize64 "$disk") ; } 2>&1 ; then
+ if ! test "$size" -gt 0 2>/dev/null ; then
+ echo "$disk has invalid size $size"
+ return 1
+ fi
+ else
+ return 1
+ fi
+}
+
function is_multipath_path {
# Return 'false' if there is no device as argument:
test "$1" || return 1

View File

@ -0,0 +1,85 @@
diff --git a/usr/share/rear/layout/prepare/GNU/Linux/133_include_mount_filesystem_code.sh b/usr/share/rear/layout/prepare/GNU/Linux/133_include_mount_filesystem_code.sh
index d57077791..87ab5d691 100644
--- a/usr/share/rear/layout/prepare/GNU/Linux/133_include_mount_filesystem_code.sh
+++ b/usr/share/rear/layout/prepare/GNU/Linux/133_include_mount_filesystem_code.sh
@@ -29,6 +29,7 @@ mount_fs() {
case $name in
(options)
# Do not mount nodev, as chrooting later on would fail:
+ # FIXME: naive approach, will replace any "nodev" inside longer options/values
value=${value//nodev/dev}
# btrfs mount options like subvolid=259 or subvol=/@/.snapshots/1/snapshot
# from the old system cannot work here for recovery because btrfs subvolumes
@@ -147,6 +148,27 @@ mount_fs() {
echo "mount $mountopts,remount,user_xattr $device $TARGET_FS_ROOT$mountpoint"
) >> "$LAYOUT_CODE"
;;
+ (xfs)
+ # remove logbsize=... mount option. It is a purely performance/memory usage optimization option,
+ # which can lead to mount failures, because it must be an integer multiple of the log stripe unit
+ # and the log stripe unit can be different in the recreated filesystem from the original filesystem
+ # (for example when using MKFS_XFS_OPTIONS, or in some exotic situations involving an old filesystem,
+ # see GitHub issue #2777 ).
+ # If logbsize is not an integer multiple of the log stripe unit, mount fails with the warning
+ # "XFS (...): logbuf size must be greater than or equal to log stripe size"
+ # in the kernel log
+ # (and a confusing error message
+ # "mount: ...: wrong fs type, bad option, bad superblock on ..., missing codepage or helper program, or other error."
+ # from the mount command), causing the layout restoration in the recovery process to fail.
+ # Wrong sunit/swidth can cause mount to fail as well, with this in the kernel log:
+ # "kernel: XFS (...): alignment check failed: sunit/swidth vs. agsize",
+ # so remove the sunit=.../swidth=... mount options as well.
+ mountopts="$( remove_mount_options_values "$mountopts" logbsize sunit swidth )"
+ (
+ echo "mkdir -p $TARGET_FS_ROOT$mountpoint"
+ echo "mount $mountopts $device $TARGET_FS_ROOT$mountpoint"
+ ) >> "$LAYOUT_CODE"
+ ;;
(*)
(
echo "mkdir -p $TARGET_FS_ROOT$mountpoint"
diff --git a/usr/share/rear/lib/filesystems-functions.sh b/usr/share/rear/lib/filesystems-functions.sh
index afdd3f24c..658d757f4 100644
--- a/usr/share/rear/lib/filesystems-functions.sh
+++ b/usr/share/rear/lib/filesystems-functions.sh
@@ -239,3 +239,40 @@ function xfs_parse
# Output xfs options for further use
echo "$xfs_opts"
}
+
+
+# $1 is a mount command argument (string containing comma-separated
+# mount options). The remaining arguments to the function ($2 ... )
+# specify the mount options to remove from $1, together with a trailing "="
+# and any value that follows each option.
+# For example, the call
+# "remove_mount_options_values nodev,uid=1,rw,gid=1 uid gid"
+# returns "nodev,rw".
+# There is no support for removing a mount option without a value and "=",
+# so "remove_mount_options_values nodev,uid=1,rw,gid=1 rw" will not work.
+# The function will return the modified string on stdout.
+
+function remove_mount_options_values () {
+ local str="$1"
+
+ shift
+ # First add a comma at the end so that it is easier to remove a mount option at the end:
+ str="${str/%/,}"
+ for i in "$@" ; do
+ # FIXME this also removes trailing strings at the end of longer words
+ # For example if one wants to remove any id=... option,
+ # the function will also replace "uid=1" by "u" by removing
+ # the trailing "id=1", which is not intended.
+ # Not easy to fix because $str can contain prefixes which are not
+ # mount options but arguments to the mount command itself
+ # (in particluar, "-o ").
+ # FIXME this simple approach would fail in case of mount options
+ # containing commas, for example the "context" option values,
+ # see mount(8)
+
+ # the extglob shell option is enabled in rear
+ str="${str//$i=*([^,]),/}"
+ done
+ # Remove all commas at the end:
+ echo "${str/%,/}"
+}

View File

@ -3,7 +3,7 @@
Summary: Relax-and-Recover is a Linux disaster recovery and system migration tool
Name: rear
Version: 2.6
Release: 10%{?dist}
Release: 12%{?dist}
License: GPLv3
Group: Applications/File
URL: http://relax-and-recover.org/
@ -39,6 +39,20 @@ Patch59: rear-usb-uefi-part-size-bz2228402.patch
Patch60: rear-luks-key-bz2228779.patch
Patch61: rear-uefi-usb-secureboot-bz2196445.patch
Patch62: rear-vg-command-not-found-bz2121476.patch
Patch64: rear-save-lvm-poolmetadatasize-RHEL-6984.patch
Patch65: rear-skip-useless-xfs-mount-options-RHEL-10478.patch
# make initrd accessible only by root
# https://github.com/rear/rear/commit/89b61793d80bc2cb2abe47a7d0549466fb087d16
Patch111: rear-CVE-2024-23301.patch
# Support saving and restoring hybrid BIOS/UEFI bootloader setup and clean up bootloader detection
# https://github.com/rear/rear/pull/3145
Patch113: rear-restore-hybrid-bootloader-RHEL-16864.patch
# Skip invalid disk drives (zero sized, no media) when saving layout
# https://github.com/rear/rear/commit/808b15a677191aac62faadd1bc71885484091316
Patch115: rear-skip-invalid-drives-RHEL-22863.patch
### Dependencies on all distributions
BuildRequires: asciidoc
@ -171,6 +185,11 @@ fi
%patch60 -p1
%patch61 -p1
%patch62 -p1
%patch64 -p1
%patch65 -p1
%patch111 -p1
%patch113 -p1
%patch115 -p1
echo "30 1 * * * root test -f /var/lib/rear/layout/disklayout.conf && /usr/sbin/rear checklayout || /usr/sbin/rear mkrescue" >rear.cron
@ -204,6 +223,16 @@ TZ=UTC %{__make} -C doc
%{_sbindir}/rear
%changelog
* Wed Feb 21 2024 Pavel Cahyna <pcahyna@redhat.com> - 2.6-12
- Skip invalid disk drives when saving layout PR 3047
- Support saving and restoring hybrid BIOS/UEFI bootloader, PRs 3145 3136
- make initrd accessible only by root (CVE-2024-23301), PR 3123
- Backport PR 3061 to save LVM pool metadata volume size in disk layout
and restore it
- Backport PR 3058 to skip useless xfs mount options when mounting
during recovery, prevents mount errors like "logbuf size must be greater
than or equal to log stripe size"
* Tue Aug 22 2023 Pavel Cahyna <pcahyna@redhat.com> - 2.6-10
- Apply PR 3027 to ensure correct creation of the rescue environment
when a file is shrinking while being read