ATF in version 1.2 fails to build with:
./build/juno/release/bl1/context_mgmt.o: In function `cm_prepare_el3_exit':
context_mgmt.c:(.text.cm_prepare_el3_exit+0x54): undefined reference to `cm_set_next_context'
context_mgmt.c:(.text.cm_prepare_el3_exit+0x54): relocation truncated to fit: R_AARCH64_JUMP26 against undefined symbol `cm_set_next_context'
This has been fixed in ATF v1.3. Even though there are even newer
versions of ATF available, we take a conservative approach, and bump
to the first version that has the build issue fixed.
Fixes:
https://gitlab.com/buildroot.org/buildroot/-/jobs/88314771
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
512B is not a correct size to express "512 bytes", and causes a
genimage failure:
ERROR: Invalid size suffix 'B' in '512B'
To express "512 bytes", using just "512" is sufficient. With this
commit, genimage works fine, and we indeed have a 512 bytes unused
partition:
$ fdisk -l output/images/sdcard.img
Disk output/images/sdcard.img: 60 MiB, 62915584 bytes, 122882 sectors
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disklabel type: dos
Disk identifier: 0x00000000
Device Boot Start End Sectors Size Id Type
output/images/sdcard.img1 1 1 1 512B 0 Empty
output/images/sdcard.img2 2 524289 524288 256M 83 Linux
Fixes:
https://gitlab.com/buildroot.org/buildroot/-/jobs/88314963
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
In commit 2cdfa6c849 ("synopsys/axs10x:
Update /etc/inittab by post-build"), two Synopsys platforms were
changed to use a post-build script adding an extra getty in the
inittab instead of a rootfs overlay containing a custom
inittab. However, in this commit, configs/snps_archs38_vdk_defconfig
was not changed, even though it was using the same rootfs overlay.
This commit therefore adjusts configs/snps_archs38_vdk_defconfig to
also use the newly introduced post-build script, fixing the build of
configs/snps_archs38_vdk_defconfig.
Fixes:
https://gitlab.com/buildroot.org/buildroot/-/jobs/88314952
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
The default size of the ext4 filesystem is no longer sufficient to
hold all the kernel modules built by the RPi Linux kernel
configuration. Let's increase to 120 MB, like we did for
raspberrypi3_defconfig.
Fixes:
https://gitlab.com/buildroot.org/buildroot/-/jobs/88314938
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
With some "old" toolchains (glibc, uclibc in version 4.9.4, 5.3, 5.4,
5.5 ...), the following error is raised by the compiler:
../src/screen.cxx:60:29: required from here
/usr/lfs/v0/rc-buildroot-test/scripts/instance-1/output/host/opt/ext-toolchain/mips-linux-gnu/include/c++/5.3.0/ext/new_allocator.h:120:4:
error: no matching function for call to 'std::pair<const screen_functions* const, std::unique_ptr<Page> >::pair(const screen_functions*, Page*)'
[...]
/usr/lfs/v0/rc-buildroot-test/scripts/instance-1/output/host/opt/ext-toolchain/mips-linux-gnu/include/c++/5.3.0/bits/stl_pair.h:112:26:
note: candidate: constexpr std::pair<_T1, _T2>::pair(const _T1&, const _T2&) [with _T1 = const screen_functions* const; _T2 = std::unique_ptr<Page>]
_GLIBCXX_CONSTEXPR pair(const _T1& __a, const _T2& __b)
^
/usr/lfs/v0/rc-buildroot-test/scripts/instance-1/output/host/opt/ext-toolchain/mips-linux-gnu/include/c++/5.3.0/bits/stl_pair.h:112:26:
note: no known conversion for argument 2 from 'Page*' to 'const std::unique_ptr<Page>&'
This is due to the fact that init function of screen_functions
structure returns Page* but PageMap wants a std::unique_ptr<Page>
To fix this, cast raw pointer into a unique_ptr with an explicit cast
Fixes:
- http://autobuild.buildroot.net/results/d8a7339d8bdd5cdc6bd1716585d4bcf15a2e8015
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
As noted by Arnout in [1], the logic in mesa3d-headers.mk generates a
bogus dri.pc file, which looks like this:
prefix=/usr
exec_prefix=/usr
libdir=/lib
includedir=/include
dridriverdir=/dri
Indeed, the ${...} are expanded as shell variables when the sed
command is executed, while the intention is that those ${...} should
go in the .pc file. By escaping those using $${...}, we get the
expected .pc file:
prefix=/usr
exec_prefix=/usr
libdir=${exec_prefix}/lib
includedir=${prefix}/include
dridriverdir=${libdir}/dri
This was detected by the not yet committed check-package improvement
from Ricardo that detects bogus ${...} usage to reference make
variables.
[1] http://lists.busybox.net/pipermail/buildroot/2018-July/225402.html
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Fixes configure warning:
checking for SPEEX... yes
checking for SPEEXDSP... no
configure: WARNING: Package speexdsp was not found in the pkg-config search path.
Perhaps you should add the directory containing `speexdsp.pc'
to the PKG_CONFIG_PATH environment variable
Package 'speexdsp', required by 'world', not found.
Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Currently makedevs silently ignores extended attributes with leading
whitespace, for example those added to a <PACKAGE>_PERMISSIONS following
the recommended style from check-package.
Makedevs already ignores leading whitespace for normal entries (file
permission changes and device files creation). Do the same for extended
attributes.
Fixes: #11191.
Reported-by: Jean-pierre Cartal <jpcartal@free.fr>
Signed-off-by: Ricardo Martincoski <ricardo.martincoski@gmail.com>
Cc: Arnout Vandecappelle <arnout@mind.be>
Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Since commit b35ad5d0b4 (ncurses: make host-ncurses use host terminfo), we
are now pointing host-ncurses to the host terminfo (typically) located in
/usr/share/terminfo.
With this change we are reusing the existing host terminfo database, so
there is no point in trying to install our own on top. The user running
buildroot typically will have no write access to /usr/share/terminfo, but
tic in that case falls back to writing the database to $HOME/.terminfo.
Neither of which are desirable.
In case $HOME/.terminfo also isn't writable, tic fails, breaking the install
step for host-ncurses:
** Building terminfo database, please wait...
Running sh ./shlib tic to install /usr/share/terminfo ...
You may see messages regarding extended capabilities, e.g., AX.
These are extended terminal capabilities which are compiled
using
tic -x
If you have ncurses 4.2 applications, you should read the INSTALL
document, and install the terminfo without the -x option.
"terminfo.tmp", line 21272, terminal 'v3220': /home/peko/.terminfo: permission denied (errno 30)
To fix all of this, simply disable the terminfo database install.
Suggested-by: Arnout Vandecappelle <arnout@mind.be>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Acked-by: Hollis Blanchard <hollis_blanchard@mentor.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
chrony calls getrandom() at startup if available, so it needs a workaround
for the blocking behaviour on recent (4.14.39+), similar to what was done
for util-linux in commit c4d86707cd (util-linux: add two upstream patches
to fix blocking on getrandom() with recent kernels).
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
This commit reformats
0002-patch-source3__libads__kerberos_keytab.c.patch as a Git-formatted
patch.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
This patch bumps the CIP Linux kernel to version v4.4.138-cip25
Signed-off-by: Angelo Compagnucci <angelo@amarulasolutions.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Fixes CVE-2018-10887 and CVE-2018-10888: out-of-bounds reads when
reading objects from a packfile.
Also fixes out-of-bounds reads when processing smart-protocol "ng"
packets (no known CVE yet).
Drop upstream patch.
Cc: Nicolas Cavallari <nicolas.cavallari@green-communications.fr>
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Reviewed-By: Nicolas Cavallari <nicolas.cavallari@green-communications.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Add patches for the following security issues:
CVE-2017-14501 - An out-of-bounds read flaw exists in parse_file_info in
archive_read_support_format_iso9660.c when extracting a specially
crafted iso9660 iso file.
CVE-2017-14502 - Off-by-one error for UTF-16 names in RAR archives,
leading to an out-of-bounds read in archive_read_format_rar_read_header.
CVE-2017-14503 - Out-of-bounds read within lha_read_data_none() in
archive_read_support_format_lha.c when extracting a specially crafted
lha archive.
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Current version of mono doesn't compile on mips due to a bug, so
disabling the mips target architecture until the bug is fixed upstream.
Fixes: http://autobuild.buildroot.net/results/b402af8cb7f00dc2e69c219926b474a8e0c8d372
and many others.
Signed-off-by: Angelo Compagnucci <angelo@amarulasolutions.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
linuxptp missing.h header implements a replacement clock_nanosleep()
function, which was used when the thread implementation was not NPTL,
because uClibc failed to provide clock_nanosleep() in such
configurations.
However, uClibc-ng has fixed this problem upstream, and has backported
this change in Buildroot in patch
package/uclibc/0002-librt-declare-clock_nanosleep-independent-of-thread-.patch
(the code is upstream and will be part of uClibc-ng 1.0.31). Due to
this, there is now a conflicting definition of clock_nanosleep()
between the C library and the linuxptp missing.h code, which manifests
itself by the following build failure:
missing.h:117:19: error: static declaration of 'clock_nanosleep' follows non-static declaration
static inline int clock_nanosleep(clockid_t clock_id, int flags,
^~~~~~~~~~~~~~~
In file included from clockadj.h:24:0,
from clockadj.c:24:
This commit fixes that by adding a patch that removes the replacement
clock_nanosleep() implementation from the linuxptp code base.
Fixes:
http://autobuild.buildroot.net/results/bf400095a853f5beb28c77a50fcffefe52c3d769/
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Static build of the wiringPi package is disabled by the tool author
(see wiringPi/noMoreStatic for details) since the version bump
to 2.46.
Fixes [1]:
arm-linux-gcc.br_real: error: ../wiringPi/libwiringPi.a: No such file or directory
[1] http://autobuild.buildroot.net/results/a1ca953247475a5f31fe9283ade05c9cba26853b
Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
As for sh4-r2d (little-endian) restore the old sh-sci driver behaviour
for sh4eb-r2d.
Tested with qemu_sh4eb_r2d_defconfig.
Signed-off-by: Romain Naour <romain.naour@gmail.com>
Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
This reverts commit 18e8cf159177100e69d528293f8cf6875c0b1bca (kernel)
The last Qemu kernel update [1] introduced a regresion in sh4 SCIF
serial device. Some keyboard presses are very slow to be taken into
account, perhaps not even taken into account at all. This would
explain why our test infrastructure doesn’t manage to login as root
[2][3][4].
git bisect reported a kernel patch from 4.11, increasing RX FIFO
trigger defaults value for sh-sci (H)SCIF. The kernel patch itself
looks good but the Qemu emulation is not ready to handle this new
setting.
>From Qemu (2.12.0): target/sh4/README.sh4
"Configuration of the second serial port (SCIF) is supported. FIFO
handling infrastructure has been started but is not completed yet."
We can't use the first serial port (ttySC0) because it's the second
SH UART that's emulated by Qemu.
In order to be able to test sh4 architecture with newer kernel,
revert to the old behaviour.
[1] 03fb00f217
[2] https://gitlab.com/free-electrons/toolchains-builder/-/jobs/72006425
[3] https://gitlab.com/free-electrons/toolchains-builder/-/jobs/72006427
[4] https://gitlab.com/free-electrons/toolchains-builder/-/jobs/72006426
Signed-off-by: Romain Naour <romain.naour@gmail.com>
Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Prevent creating a dangling symlink when vim is not present on the host
machine. With BR2_ROOTFS_MERGED_USR, just link to "vim", since they are
on the same directory, otherwise link to "../usr/bin/vim".
Signed-off-by: Carlos Santos <casantos@datacom.com.br>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Artem Panfilov <apanfilov@spectracom.com>
[Thomas: add explicit --without-pam when linux-pam is not enabled.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
gnuradio needs host-python2 however there is no way to enforce this so
add a dependency on !BR2_PACKAGE_PYTHON3.
Indeed, if BR2_PACKAGE_PYTHON3 is selected, then buildroot will only
build host-python3.
This issue was not raised in the previous version of gnuradio as
host-python-cheetah was the only dependency and host-python-cheetah
enforces python2 through HOST_PYTHON_CHEETAH_NEEDS_HOST_PYTHON however
we now have host-python-six and host-python-mako dependencies and it
does not seem right to enforces python2 on those packages
Fixes:
- http://autobuild.buildroot.net/results/a502fef3e2a40f38ca63332440b0b6bd2ed11d5a
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
host-python-cheetah has been replaced by host-python-mako:
1b2bedc868
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
The oldest gcc that is known to work with current Buildroot is that of
RHEL 6, version 4.4.
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
