From: Yann E. MORIN <hidden> Date: 2014-12-06 12:29:42
Hello All!
This series enhances the fs/ext2 generation.
Not much to say beside the individual commit logs... ;-)
Changes v2 -> v3:
- fix quoting for labels with a space in them (K?roly)
Changes v1 -> v2:
- move our genext2fs.sh wrapper and make it a proper host package,
with extensive tweaks to the code (Thomas)
Regards,
Yann E. MORIN.
The following changes since commit c5f0e93d2cc6107cce249e1f11158aeccb4a3449:
rp-pppoe: install default conf file (2014-12-03 23:26:21 +0100)
are available in the git repository at:
git://git.busybox.net/~ymorin/git/buildroot yem/fs-vol-id
for you to fetch changes up to 0244e2c2f1c3933e110df8478544efafe66c7e5c:
fs/ext2: add option to specify a filesystem label (2014-12-06 13:26:58 +0100)
----------------------------------------------------------------
Yann E. MORIN (3):
package/mke2img: new package
fs/ext2: use the new mke2img host package
fs/ext2: add option to specify a filesystem label
fs/ext2/Config.in | 6 +-
fs/ext2/ext2.mk | 19 ++++--
fs/ext2/genext2fs.sh | 121 ---------------------------------
package/Config.in.host | 1 +
package/mke2img/Config.in.host | 9 +++
package/mke2img/mke2img | 148 +++++++++++++++++++++++++++++++++++++++++
package/mke2img/mke2img.mk | 14 ++++
7 files changed, 188 insertions(+), 130 deletions(-)
delete mode 100755 fs/ext2/genext2fs.sh
create mode 100644 package/mke2img/Config.in.host
create mode 100755 package/mke2img/mke2img
create mode 100644 package/mke2img/mke2img.mk
--
.-----------------.--------------------.------------------.--------------------.
| Yann E. MORIN | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software Designer | \ / CAMPAIGN | ___ |
| +33 223 225 172 `------------.-------: X AGAINST | \e/ There is no |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL | v conspiracy. |
'------------------------------^-------^------------------^--------------------'
From: Yann E. MORIN <hidden> Date: 2014-12-06 12:29:44
Currently, we are using a shell script called genext2fs, that
impersonates the real genext2fs. But that script does much more than
just call genextfs: it also calls tune2fs and e2fsck.
Because it impersonates genext2fs, we can not easily add new options,
and are constrained by the genext2fs options.
But it turns out that, of all the options supported by the real
genext2fs, we only really care for a subset, namely:
- number of blocks
- number of inodes
- percentage of blocks reeserved to root
- the root directory which to generate the image from
So, we introduce a new host package, mke2img, that is intended to
eventually replace genext2fs.sh.
This new script is highly modeled from the existing genext2fs.sh, but
was slightly refreshed, and a new, supposedly sane set of options has
been choosen for the features we need (see above), and some new options
were added, too, rather than relying on the arguments order or
environment variables:
-b <nb-blocks> number of blocks in the filesystem
-i <nb-inodes> number of inodes in the filesystem
-r <pc-reserved> percentage of reserved blocks
* -d <root-dir> directory containing the root of the filesystem
* -o <img-file> output image file
* -G <ext-gen> extfs generation: 2, 3, or 4
* -R <ext-rev> ext2 revision: 0 or 1 (1 is assumed for ext3 and ext4)
-l <label> filesystem label
-u <uid> filesystem UUID; if not specified, a random one is used
* Mandatory options
Since the upstream e2fsprogs are expected to release a new mke2fs that
will be able to generate a filesystem image from a directory, we then
will be able to replace all the logic in mke2img, to use mke2fs instead
of the (relatively fragile) combination of the three tools we currently
use.
An entry is added for it in the "Host utilities" menu, so it can be
selected for use by post-{build,image} scripts. The ext2 filesystem
selection is changed to select that now.
Signed-off-by: "Yann E. MORIN" <redacted>
Cc: Thomas Petazzoni <redacted>
Tested-by: Karoly Kasza <redacted>
Reviewed-by: Karoly Kasza <redacted>
---
fs/ext2/Config.in | 3 +-
package/Config.in.host | 1 +
package/mke2img/Config.in.host | 9 +++
package/mke2img/mke2img | 148 +++++++++++++++++++++++++++++++++++++++++
package/mke2img/mke2img.mk | 14 ++++
5 files changed, 173 insertions(+), 2 deletions(-)
create mode 100644 package/mke2img/Config.in.host
create mode 100755 package/mke2img/mke2img
create mode 100644 package/mke2img/mke2img.mk
@@ -0,0 +1,148 @@+#!/bin/bash++# Buildroot wrapper to the collection of ext2/3/4 filesystem tools:+# - genext2fs, to generate ext2 filesystem images+# - tune2fs, to modify an ext2/3/4 filesystem (possibly in an image file)+# - e2fsck, to check and fix an ext2/3/4 filesystem (possibly in an image file)++set -e++main() {+ local OPT OPTARG+ local nb_blocks nb_inodes nb_res_blocks root_dir image gen rev label uuid+ local -a genext2fs_opts+ local -a tune2fs_opts+ local tune2fs_O_opts++ while getopts :b:i:r:d:o:G:R:l:u: OPT; do+ case "${OPT}" in+ b) nb_blocks=${OPTARG};;+ i) nb_inodes=${OPTARG};;+ r) nb_res_blocks=${OPTARG};;+ d) root_dir="${OPTARG}";;+ o) image="${OPTARG}";;+ G) gen=${OPTARG};;+ R) rev=${OPTARG};;+ l) label="${OPTARG}";;+ u) uuid="${OPTARG}";;+ :) error "option '%s' expects a mandatory argument\n" "${OPTARG}";;+ \?) error "unknown option '%s'\n" "${OPTARG}";;+ esac+ done++ # Sanity checks+ if [ -z "${root_dir}" ]; then+ error "you must specify a root directory with '-d'\n"+ fi+ if [ -z "${image}" ]; then+ error "you must specify an output image file with '-o'\n"+ fi+ case "${gen}:${rev}" in+ 2:0|2:1|3:1|4:1) ;;+ *) error "unknown ext generation '%s' and/or revision '%s'\n" \+ "${gen}" "${rev}"+ ;;+ esac++ # calculate needed inodes+ if [ -z "${nb_inodes}" ]; then+ nb_inodes=$(find "${root_dir}" | wc -l)+ nb_inodes=$((nb_inodes+400))+ fi++ # calculate needed blocks+ if [ -z "${nb_blocks}" ]; then+ # size ~= superblock, block+inode bitmaps, inodes (8 per block),+ # blocks; we scale inodes / blocks with 10% to compensate for+ # bitmaps size + slack+ nb_blocks=$(du -s -k "${root_dir}" |sed -r -e 's/[[:space:]]+.*$//')+ nb_blocks=$((500+(nb_blocks+nb_inodes/8)*11/10))+ fi++ # Upgrade to rev1 if needed+ if [ ${rev} -ge 1 ]; then+ tune2fs_O_opts+=",filetype"+ fi++ # Add a journal for ext3 and above+ if [ ${gen} -ge 3 ]; then+ tune2fs_opts+=( -j -J size=1 )+ # we add 1300 blocks (a bit more than 1 MiB, assuming 1KiB blocks)+ # for the journal+ # Note: I came to 1300 blocks after trial-and-error checks. YMMV.+ nb_blocks=$((nb_blocks+1300))+ fi++ # Add ext4 specific features+ if [ ${gen} -ge 4 ]; then+ tune2fs_O_opts+=",extents,uninit_bg,dir_index"+ fi++ # Add our -O options (there will be at most one leading comma, remove it)+ if [ -n "${tune2fs_O_opts}" ]; then+ tune2fs_opts+=( -O "${tune2fs_O_opts#,}" )+ fi++ # Add the label if specified+ if [ -n "${label}" ]; then+ tune2fs_opts+=( -L "${label}" )+ fi++ # Generate the filesystem+ genext2fs_opts=( -b ${nb_blocks} -N ${nb_inodes} -d "${root_dir}" )+ if [ -n "${nb_res_blocks}" ]; then+ genext2fs_opts+=( -m ${nb_res_blocks} )+ fi+ genext2fs "${genext2fs_opts[@]}" "${image}"++ # genext2fs does not generate a UUID, but fsck will whine if one is+ # is missing, so we need to add a UUID.+ # Of course, this has to happend _before_ we run fsck.+ # Also, some ext4 metadata are based on the UUID, so we must+ # set it before we can convert the filesystem to ext4.+ # If the user did not specify a UUID, we generate a random one.+ # Although a random UUID may seem bad for reproducibility, there+ # already are so many things that are not reproducible in a+ # filesystem: file dates, file ordering, content of the files...+ tune2fs -U "${uuid:-random}" "${image}"++ # Upgrade the filesystem+ if [ ${#tune2fs_opts[@]} -ne 0 ]; then+ tune2fs "${tune2fs_opts[@]}" "${image}"+ fi++ # After changing filesystem options, running fsck is required+ # (see: man tune2fs). Running e2fsck in other cases will ensure+ # coherency of the filesystem, although it is not required.+ # 'e2fsck -pDf' means:+ # - automatically repair+ # - optimise and check for duplicate entries+ # - force checking+ # Sending output to oblivion, as e2fsck can be *very* verbose,+ # especially with filesystems generated by genext2fs.+ # Exit codes 1 & 2 are OK, it means fs errors were successfully+ # corrected, hence our little trick with $ret.+ ret=0+ e2fsck -pDf "${image}" >/dev/null || ret=$?+ case ${ret} in+ 0|1|2) ;;+ *) errorN ${ret} "failed to run e2fsck on '%s' (ext%d)\n" \+ "${image}" ${gen}+ esac+ printf "\n"+ trace "e2fsck was successfully run on '%s' (ext%d)\n" "${image}" ${gen}+ printf "\n"++ # Remove count- and time-based checks, they are not welcome+ # on embedded devices, where they can cause serious boot-time+ # issues by tremendously slowing down the boot.+ tune2fs -c 0 -i 0 "${image}"+}++trace() { local msg="${1}"; shift; printf "%s: ${msg}" "${my_name}" "${@}"; }+warn() { trace "${@}" >&2; }+errorN() { local ret="${1}"; shift; warn "${@}"; exit ${ret}; }+error() { errorN 1 "${@}"; }++my_name="${0##*/}"+main "$@"
@@ -1,121 +0,0 @@-#!/bin/sh-# genext2fs wrapper calculating needed blocks/inodes values if not specified-set -e--export LC_ALL=C--CALC_BLOCKS=1-CALC_INODES=1-EXT_OPTS=-EXT_OPTS_O=--while getopts x:d:D:b:i:N:m:g:e:zfqUPhVv f-do- case $f in- b) CALC_BLOCKS=0 ;;- N) CALC_INODES=0; INODES=$OPTARG ;;- d) TARGET_DIR=$OPTARG ;;- esac-done-eval IMG="\"\${${OPTIND}}\""--# calculate needed inodes-if [ $CALC_INODES -eq 1 ];-then- INODES=$(find $TARGET_DIR | wc -l)- INODES=$(expr $INODES + 400)- set -- $@ -N $INODES-fi--# calculate needed blocks-if [ $CALC_BLOCKS -eq 1 ];-then- # size ~= superblock, block+inode bitmaps, inodes (8 per block), blocks- # we scale inodes / blocks with 10% to compensate for bitmaps size + slack- BLOCKS=$(du -s -c -k $TARGET_DIR | grep total | sed -e "s/total//")- BLOCKS=$(expr 500 + \( $BLOCKS + $INODES / 8 \) \* 11 / 10)- # we add 1300 blocks (a bit more than 1 MiB, assuming 1KiB blocks) for- # the journal if ext3/4- # Note: I came to 1300 blocks after trial-and-error checks. YMMV.- if [ ${GEN} -ge 3 ]; then- BLOCKS=$(expr 1300 + $BLOCKS )- fi- set -- $@ -b $BLOCKS-fi--e2tunefsck() {- # genext2fs does not generate a UUID, but fsck will whine if one is- # is missing, so we need to add a UUID.- # Of course, this has to happend _before_ we run fsck.- # Also, some ext4 metadata are based on the UUID, so we must- # set it before we can convert the filesystem to ext4.- # Although a random UUID may seem bad for reproducibility, there- # already are so many things that are not reproducible in a- # filesystem: file dates, file ordering, content of the files...- tune2fs -U random "${IMG}"-- # Upgrade the filesystem- if [ $# -ne 0 ]; then- tune2fs "$@" "${IMG}"- fi-- # After changing filesystem options, running fsck is required- # (see: man tune2fs). Running e2fsck in other cases will ensure- # coherency of the filesystem, although it is not required.- # 'e2fsck -pDf' means:- # - automatically repair- # - optimise and check for duplicate entries- # - force checking- # Sending output to oblivion, as e2fsck can be *very* verbose,- # especially with filesystems generated by genext2fs.- # Exit codes 1 & 2 are OK, it means fs errors were successfully- # corrected, hence our little trick with $ret.- ret=0- e2fsck -pDf "${IMG}" >/dev/null || ret=$?- case ${ret} in- 0|1|2) ;;- *) exit ${ret};;- esac- printf "\ne2fsck was successfully run on '%s' (ext%d)\n\n" \- "${IMG##*/}" "${GEN}"-- # Remove count- and time-based checks, they are not welcome- # on embedded devices, where they can cause serious boot-time- # issues by tremendously slowing down the boot.- tune2fs -c 0 -i 0 "${IMG}"-}--# Check we know what generation to generate-case "${GEN}:${REV}" in- 2:0|2:1|3:1|4:1)- ;;- *)- printf "%s: unknown ext generation '%s' and/or revision '%s'\n" \- "${0##*/}" "${GEN}" "${REV}" >&2- exit 1- ;;-esac--# Upgrade to rev1 if needed-if [ ${REV} -ge 1 ]; then- EXT_OPTS_O="${EXT_OPTS_O},filetype"-fi--# Add a journal for ext3 and above-if [ ${GEN} -ge 3 ]; then- EXT_OPTS="${EXT_OPTS} -j -J size=1"-fi--# Add ext4 specific features-if [ ${GEN} -ge 4 ]; then- EXT_OPTS_O="${EXT_OPTS_O},extents,uninit_bg,dir_index"-fi--# Add our -O options (there will be at most one leading comma, remove it)-if [ -n "${EXT_OPTS_O}" ]; then- EXT_OPTS="${EXT_OPTS} -O ${EXT_OPTS_O#,}"-fi--# Generate and upgrade the filesystem-genext2fs "$@"-e2tunefsck ${EXT_OPTS}
From: Yann E. MORIN <hidden> Date: 2014-12-06 12:29:46
Filesystems of the ext familly can carry a filesystem label.
Add an option for the user to specify such a label.
Signed-off-by: "Yann E. MORIN" <redacted>
Cc: Thomas Petazzoni <redacted>
Cc: K?roly Kasza <redacted>
---
Changes v2 -> v3:
- fix quoting for when the label contains a space (K?roly)
---
fs/ext2/Config.in | 3 +++
fs/ext2/ext2.mk | 8 ++++++++
2 files changed, 11 insertions(+)
@@ -41,6 +41,9 @@ config BR2_TARGET_ROOTFS_EXT2_REVdefault0ifBR2_TARGET_ROOTFS_EXT2_2r0default1if!BR2_TARGET_ROOTFS_EXT2_2r0+configBR2_TARGET_ROOTFS_EXT2_LABEL+string"filesystem label"+configBR2_TARGET_ROOTFS_EXT2_BLOCKSint"size in blocks (leave at 0 for auto calculation)"default0
@@ -18,6 +18,14 @@ ifneq ($(strip $(BR2_TARGET_ROOTFS_EXT2_RESBLKS)),0)EXT2_OPTS+=-r$(BR2_TARGET_ROOTFS_EXT2_RESBLKS)endif+# Not qstrip-ing the variable, because it may contain spaces, but we must+# qstrip it when checking. Furthermore, we need to further quote it, so+# that the quotes do not get eaten by the echo statement when creating the+# fakeroot script+ifneq ($(call qstrip,$(BR2_TARGET_ROOTFS_EXT2_LABEL)),)+EXT2_OPTS+=-l'$(BR2_TARGET_ROOTFS_EXT2_LABEL)'+endif+ROOTFS_EXT2_DEPENDENCIES=host-mke2imgdefine ROOTFS_EXT2_CMD
From: Károly Kasza <hidden> Date: 2014-12-06 14:15:59
Hi Yann, Thomas, list,
Signed-off-by: "Yann E. MORIN" <redacted>
Cc: Thomas Petazzoni <redacted>
Cc: K?roly Kasza <redacted>
Tested-by: Karoly Kasza <redacted>
Reviewed-by: Karoly Kasza <redacted>
Tested the whole series of 3 together (v3), successfully built different
rootfs.ext images and booted them.
Tested with master branch, x86_64 arch in QEMU, internal toolchain GCC
4.9.2 w/ uclibc, Linux 3.17.4.
Regards,
Karoly
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20141206/dd6f609b/attachment.html>
From: Thomas Petazzoni <hidden> Date: 2014-12-06 18:20:32
Dear Yann E. MORIN,
On Sat, 6 Dec 2014 13:29:44 +0100, Yann E. MORIN wrote:
This new script is highly modeled from the existing genext2fs.sh, but
was slightly refreshed, and a new, supposedly sane set of options has
been choosen for the features we need (see above), and some new options
were added, too, rather than relying on the arguments order or
environment variables:
-b <nb-blocks> number of blocks in the filesystem
-i <nb-inodes> number of inodes in the filesystem
-r <pc-reserved> percentage of reserved blocks
* -d <root-dir> directory containing the root of the filesystem
* -o <img-file> output image file
* -G <ext-gen> extfs generation: 2, 3, or 4
* -R <ext-rev> ext2 revision: 0 or 1 (1 is assumed for ext3 and ext4)
-l <label> filesystem label
-u <uid> filesystem UUID; if not specified, a random one is used
* Mandatory options
What about putting this in a "help" of the mke2img tool, or at least in
the comment header at the beginning of the script?
+ # genext2fs does not generate a UUID, but fsck will whine if one is
+ # is missing, so we need to add a UUID.
duplicate "is".
+ # Of course, this has to happend _before_ we run fsck.
happend -> happen
quoted hunk
+ # Also, some ext4 metadata are based on the UUID, so we must+ # set it before we can convert the filesystem to ext4.+ # If the user did not specify a UUID, we generate a random one.+ # Although a random UUID may seem bad for reproducibility, there+ # already are so many things that are not reproducible in a
s/already are/are already/
Other than that, looks good.
Tested-by: Thomas Petazzoni <redacted>
(on x86-64 in qemu, tested an ext4 filesystem with only the ext4
kernel driver compiled in, and then tested an ext2 filesystem with
only the ext2 kernel driver compiled in).
Thanks!
Thomas
--
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com
From: Thomas Petazzoni <hidden> Date: 2014-12-06 18:21:52
Dear Yann E. MORIN,
On Sat, 6 Dec 2014 13:29:45 +0100, Yann E. MORIN wrote:
And get rid of the old genext2fs.sh wrapper.
Signed-off-by: "Yann E. MORIN" <redacted>
Cc: Thomas Petazzoni <redacted>
Tested-by: Karoly Kasza <redacted>
Reviewed-by: Karoly Kasza <redacted>
---
fs/ext2/ext2.mk | 13 +++---
fs/ext2/genext2fs.sh | 121 ---------------------------------------------------
2 files changed, 5 insertions(+), 129 deletions(-)
delete mode 100755 fs/ext2/genext2fs.sh
Reviewed-by: Thomas Petazzoni <redacted>
Tested-by: Thomas Petazzoni <redacted>
(on x86-64 in qemu, tested an ext4 filesystem with only the ext4
kernel driver compiled in, and then tested an ext2 filesystem with
only the ext2 kernel driver compiled in).
Thanks,
Thomas
--
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com
From: Thomas Petazzoni <hidden> Date: 2014-12-06 18:23:25
Dear Yann E. MORIN,
On Sat, 6 Dec 2014 13:29:46 +0100, Yann E. MORIN wrote:
Filesystems of the ext familly can carry a filesystem label.
Add an option for the user to specify such a label.
Signed-off-by: "Yann E. MORIN" <redacted>
Cc: Thomas Petazzoni <redacted>
Cc: K?roly Kasza <redacted>
Reviewed-by: Thomas Petazzoni <redacted>
Tested-by: Thomas Petazzoni <redacted>
(on x86-64 in qemu, tested an ext4 filesystem with only the ext4
kernel driver compiled in, and then tested an ext2 filesystem with
only the ext2 kernel driver compiled in).
Thanks,
Thomas
--
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com
From: Yann E. MORIN <hidden> Date: 2014-12-06 19:26:20
Thomas, All,
On 2014-12-06 19:20 +0100, Thomas Petazzoni spake thusly:
On Sat, 6 Dec 2014 13:29:44 +0100, Yann E. MORIN wrote:
quoted
This new script is highly modeled from the existing genext2fs.sh, but
was slightly refreshed, and a new, supposedly sane set of options has
been choosen for the features we need (see above), and some new options
were added, too, rather than relying on the arguments order or
environment variables:
-b <nb-blocks> number of blocks in the filesystem
-i <nb-inodes> number of inodes in the filesystem
-r <pc-reserved> percentage of reserved blocks
* -d <root-dir> directory containing the root of the filesystem
* -o <img-file> output image file
* -G <ext-gen> extfs generation: 2, 3, or 4
* -R <ext-rev> ext2 revision: 0 or 1 (1 is assumed for ext3 and ext4)
-l <label> filesystem label
-u <uid> filesystem UUID; if not specified, a random one is used
* Mandatory options
What about putting this in a "help" of the mke2img tool, or at least in
the comment header at the beginning of the script?
Ack.
quoted
+ # genext2fs does not generate a UUID, but fsck will whine if one is
+ # is missing, so we need to add a UUID.
duplicate "is".
Ack.
quoted
+ # Of course, this has to happend _before_ we run fsck.
happend -> happen
Ack.
quoted
+ # Also, some ext4 metadata are based on the UUID, so we must+ # set it before we can convert the filesystem to ext4.+ # If the user did not specify a UUID, we generate a random one.+ # Although a random UUID may seem bad for reproducibility, there+ # already are so many things that are not reproducible in a