Initramfs compression does not make much sense for the architectures
that support compressed kernel images because in this case the data
would be compressed twice. This will eventually result in a bigger
kernel image and time overhead when uncompressing it.
The only reason to use compressed initramfs is to reduce memory
usage when the kernel prepares rootfs, and both the unpacked
filesystem and initramfs.cpio are present in the memory.
Buildroot attempts to force GZIP compression for initramfs,
however it doesn't always work because initramfs compression mode
depends on RAM disk compression supported by the kernel.
Thus, CONFIG_INITRAMFS_COMPRESSION_GZIP depends on CONFIG_RD_GZIP.
If CONFIG_RD_GZIP is not set, setting GZIP initramfs compression
will have no effect.
Besides, the kernel also supports other compression methods,
like BZIP2, LZMA, XZ and LZO. Forcing the good old GZIP does not
really make much sense any more.
This removes initramfs compression settings from Buildroot,
so that the default value preset in the kernel config is used,
which is CONFIG_INITRAMFS_COMPRESSION_NONE.
If initramfs compression is still needed, it can be set
in the kernel config (using make linux-menuconfig)
Signed-off-by: Valentine Barshak <gvaxon@gmail.com>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
genext2fs is built only if the user selected an ext2 root filesystem.
However, some use-cases can't live with the full target/ dir on the
root filesystem, and requires separate partitions (eg. for /usr).
In this case, the user would not select an ext2 root fs in the
Buildrooot menu, and would only generate a tarball of the rootfs.
This tarball would then be used from a post-image script to build
the actual required FSes.
But then, genext2fs is not built, since the ext2 root FS was not
selected.
As for the other filesystem generators, provide a host variant of
genext2fs (genext2fs is already host-package aware, so only needs
adding a Kconfig entry).
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Turned out that setting a nil-UUID is no better than clearing it.
What currently happens is as follows:
- first, genext2fs does not generate a UUID
- then we tune2fs to upgrade the filesystem
- then we run fsck, which generates a random UUID
- then we re-run tune2fs to set a nil-UUID
So, on the target, if the file system is improperly unmounted (eg.
with a power failure), on next boot, fsck may be run, and a new
random UUID will be generated.
*However*, fsck improperly updates the filesystem when it adds the
UUID, and there are a few group descriptor checksum errors.
Those errors will go undetected until the next fsck, which will then
block for user input (bad on embedded systems, bad).
Fix that by systematically generating a random UUID _before_ we call
to fsck.
A random UUID is not so bad, after all, since there are already so
many sources of unpredictability in the filesystem: files date and
ordering, files content (date, paths...) which renders a fixed UUID
unneeded.
And it is still possible to set the UUID in a post-image script if
needed, anyway.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Peter Korsgaard <jacmet@uclibc.org>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
"tune2fs -U clear" creates an invalid filesystem, that fsck.ext2
whines about later:
$ make rootfs-ext2
[--SNIP--]
$ ./host/usr/sbin/fsck.ext4 images/rootfs.ext2
e2fsck 1.42.7 (21-Jan-2013)
Filesystem did not have a UUID; generating one.
images/rootfs.ext2: clean, 4616/5120 files, 53171/131072 blocks
$ ./host/usr/sbin/fsck.ext4 -f images/rootfs.ext2
e2fsck 1.42.7 (21-Jan-2013)
One or more block group descriptor checksums are invalid. Fix<y>? yes
Group descriptor 0 checksum is 0x4131, should be 0x8bdb. FIXED.
[--SNIP--]
So we set an explicitly NULL UUID instead.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Packages that install daemons may need those daemons to run as a non-root,
or an otherwise non-system (eg. 'daemon'), user.
Add infrastructure for packages to create users, by declaring the FOO_USERS
variable that contain a makedev-syntax-like description of the user(s) to
add.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Samuel Martin <s.martin49@gmail.com>
Cc: Cam Hutchison <camh@xdna.net>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Factorize all the "depends on" into "if...endif" blocks
All the UBIFS options use "depends on BR2_TARGET_ROOTFS_UBIFS" but
we can simplify the config file by enclosing them in an "if..endif"
block.
Signed-off-by: Mark Jackson <mpfj@newflow.co.uk>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Allow extra options to be passed to ubifs creation tools.
Signed-off-by: Mark Jackson <mpfj@newflow.co.uk>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Add support for LZO and XZ compression methods to cpio, ext2, tar and
ubifs filesystem targets.
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
A quick test with a ~3.5MB ext4 filesystem shows that 1081 blocks isn't
enough:
tune2fs 1.42.7 (21-Jan-2013)
Please run e2fsck on the filesystem.
Creating journal inode:
Journal size too big for filesystem.
So bump it a bit. Overestimating the journal size is probably not really
a big deal for the kind of systems using ext3/4 anyway.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Set the count- and time-based checks intervals to 0, thus effectively
disabling automatic checks at boot (after a suggestion by Arnout).
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Arnout Vandecappelle <arnout@mind.be>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Some bootloaders have a buggy ext2 support, and require ext2 rev1
instead of the traditional ext2 rev0 that genext2fs produces.
tune2fs accepts only one '-O list' at a time, so we need to construct
a list of -O options.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Arnout Vandecappelle <arnout@mind.be>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Use the host-e2fsprogs to tune2fs the generated rootfs.ext2 image,
and upgrade it to either one of ext2, ext3 or ext4.
Since calling tune2fs may require running e2fsck (tune2fs will warn
to do so when certain FS options are changed), we systematically call
e2fsck. This makes the code path simpler, and as a side-effect checks
that genext2fs did not generate garbage.
In turn, e2fsck will unconditionally add a UUID to the filesystem,
which is bad for reproducibility, so we call tune2fs again to remove
the UUID. This does not require checking the filesystem.
To ensure compatibility of Buildroot's .config, leave ext2 as the
default. Boards' .config can override this at will.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Arnout Vandecappelle <arnout@mind.be>
Cc: Peter Korsgaard <jacmet@uclibc.org>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Use a 'if...endif' construct instead of repeating the
'depends on' for each symbols.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Acked-by: Arnout Vandecappelle <arnout@mind.be>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Instead of explicitly listing the sub-architectures or architectures
that are big-endian, use BR2_ENDIAN directly.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Since ROOTFS_DEVICE_TABLES is a concatenation of two strings separated by
spaces, there will always be a space in it, which means it's never empty.
Therefore, when testing for empty, the condition never evaluate to false.
The following change fixes this problem; it runs qstrip on the overall
combination of the variables, causing the space to be removed if it's the
only thing left.
Tested-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com>
Signed-off-by: Thomas De Schampheleire <patrickdepinguin+buildroot@gmail.com>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
This makes things easier to understand and more consistent with the pkg-infra.
For example, it removes the need for '$$@' in the CMD variables of fs/*/*.mk.
It also makes it possible to update the variables later, e.g. in the package
override file.
It also makes sure that the date will be recorded correctly in Yann E. Morin's
patch that logs the MESSAGE macros to a file.
The fs/*/*.mk must be updated as well because the '$@' shouldn't be quoted
anymore in the CMD variables or the hooks.
The $(eval ...) for the dependencies is redundant, because the $(ROOTFS_TARGET)
variable is already eval'd. Note that it is only redundant if the evaluation of
the uses of the variable is also delayed.
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
A /dev/console node must be present in rootfs when the Linux kernel
boots otherwise the kernel will print the following warning:
"Warning: unable to open an initial console"
This is because when we use an initramfs the /dev directory is not
populated at this point. This can cause problems when a program
(e.g ldso with early debugging enabled) opens a standard file
descriptor for read/write before these descriptors are actually
created by the init process later on.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Tested-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
A very common mistake done by our users is that they use
output/target/ directory as their root filesystem. Even though this is
loudly documented in our Buildroot manual, people don't read
documentation, so it is not sufficient.
This patch adds a text file named
output/target/THIS_IS_NOT_YOUR_ROOT_FILESYSTEM which explains why
output/target isn't appropriate to use as the root filesystem. The
process is:
* At the beginning of the build, right after the skeleton has been
copied, support/misc/target-dir-warning.txt is copied to
output/target/THIS_IS_NOT_YOUR_ROOT_FILESYSTEM
* In the filesystem images creation code, this file is removed before
launching fakeroot, and restored right after that, so that this
file is not present in the generated root filesystem images.
Note that the file has not been added to the default skeleton for two
reasons:
* It would have annoying to have in our source tree a file named in
capital letters inside system/skeleton/
* The proposed way works even if the user uses a custom skeleton.
[Peter: fixed typo]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Juha Lumme <juha.lumme@gmail.com>
Acked-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
The ext2 initrd is pretty broken, because it requires an additional
root=/dev/ram0 command line parameter, and a /init to mount
devtmps that isn't there in out ext2 rootfs. So just use a cpio
instead.
Note that there is no check if the kernel supports initramfs or the
selected compression method.
Also removed a bit of dead code in iso9660.mk.
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Recent mails on the list show that it is not very clear how to create
an initial RAM fs with buildroot. So make this more explicit in the
cpio and initramfs help texts. Hopefully this will reduce the /init
debugging we have to do.
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
This directory groups the following elements:
* the default root filesystem skeleton
* the default device tables
* the Config.in options for system configuration (UART port for
getty, system hostname, etc.)
* the make rules to apply the system configuration options
Even though the skeleton and device tables could have lived in fs/, it
would have been strange to have the UART, system hostname and other
related options into fs/. A new system/ directory makes more sense.
As a consequence, this patch also removes target/Makefile.in, which
has become useless in the process.
[Peter: fixup TARGET_SKELETON settings / documentation to match]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Yann E. MORIN <yann.morin.1998@free.fr>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Defaulting to UTC (which is what uClibc will default to when /etc/TZ is
missing) seems more sensible than US Mountain Time Zone in the default
rootfs skeleton.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
echo -e is not a portable way to do this, better use printf.
Works with MacOS X.
Signed-off-by: Waldemar Brodkorb <mail@waldemar-brodkorb.de>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
The www-data user is defined in fs/skeleton/etc/passwd, with gid 33. But
gid 33 is not present in the group file.
Signed-off-by: Simon Dawson <spdawson@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
There used to be a mechanism using which packages could leave a
.fakeroot.<something> file which could contain commands to be executed
within the fakeroot environment. Since this mechanism is no longer
used by any package, remove it from the common infrastructure.
The latest user was nfs-utils, which used this mechanism to do the
"make install" as root, since doing otherwise was not supported. But
since 16e7b8255c, nfs-utils has been
upgraded and converted to the package infrastructure, and this hack is
no longer necessary. Another past user was the ltp-testsuite package,
for the same reason, and since
a72a670489, the fakeroot hack is no
longer needed.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Systemd introduced a new policy described here:
http://lists.freedesktop.org/archives/systemd-devel/2011-March/001823.html
Add the "lock" group to buildroot to allow systemd to set this in place
at boot time
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
As systemd checks if /etc/mtab is a symlink to /proc/mount or
/proc/self/mounts, we need to change it so that we can run systemd.
http://cgit.freedesktop.org/systemd/tree/src/main.c#n1082
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
There are a couple of Renesas SH devices with 8 serial ports used.
Signed-off-by: Phil Edworthy <phil.edworthy@renesas.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Add a way for packages to declare files they need instead of relying
only on device tables, which creates files no matter if the package is
indeed enabled, as we can see for busybox.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
This commit follows commit ad501b66. Start up of the busybox logging
daemons were moved to an init script but the shutdown were still
performed in inittab. This commit moves the shutdown policy to an
rcK script that calls the stop function of all the init scripts in
a reversed order.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
For custom projects, it is more maintainable to add custom profile settings
in a separate file, than directly in /etc/profile.
This patch modifies /etc/profile to read in *.sh files from /etc/profile.d/,
a technique commonly used in Linux distributions.
Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Cramfs still passes the device table to mkcramfs, which is irrelevant
now with fakeroot, remove this behaviour and the associated patch.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
An initramfs is in fact the same as a cpio archive, but embedded in
the kernel. So instead of duplicating the cpio infrastructure,
we can simply build images/rootfs.cpio and link that into the
kernel.
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Devtmpfs doesn't get automounted by the kernel when an initramfs is used.
Since cpio archives are used almost exclusively as an initramfs, the same
fix should be applied to it as for the initramfs.
Cfr. commit 424888e474 and
10a130f91e.
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
As the kernel doesn't automatically mount devtmpfs when an initramfs
is used, commit 424888e474 has
introduced a small wrapper script that mounts devtmpfs before starting
the real init.
Unfortunately, the problem is that in this case, the init process runs
without any 0, 1 and 2 file descriptors, so none of the
messages/errors printed by the various initialization scripts can be
seen. This is due to the fact the init process relies on 0, 1 and 2
being opened by the kernel before init is started. However, as
/dev/console isn't present on the filesystem at the time the kernel
tries to open the console to create the 0, 1 and 2 file descriptors,
the kernel fails on this and prints the famous "Warning: unable to
open an initial console".
The proposed workaround is to actually open 0, 1 and 2 to /dev/console
in the wrapper script, right after mounting the devtmpfs filesystem,
and before starting the real init. The "Warning" from the kernel is
still shown, but at least the messages from the init scripts are
visible.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
