The configure script has already found IPT_LIB_DIR via pkgconfig, so
the configure-cross.patch really just introduces dead and broken
fallback code.
Broken, because the SYSROOT variable does not actually get set to a
sensible value - the argument $1 passed to the configure script when
invoked from the Makefile is KERNEL_INCLUDE, which we set to
${STAGING_INCDIR} in EXTRA_OEMAKE. Obviously that directory does not
have /lib or /usr subdirectories, so we're not really helping the
fallback logic in check_ipt_lib_dir() - in fact, we're more or less
guaranteeing that we won't find those .so files.
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Bedel, Alban [Tue, 23 Jul 2019 14:13:36 +0000 (14:13 +0000)]
kernel-uboot: remove useless special casing of arm64 Image
Since commit a725d188b5 (kernel-uboot: compress arm64 kernels)
arm64 Image files are handled exactly like in the default case, making
this special case pointless. Remove it to make the code simpler.
Signed-off-by: Alban Bedel <alban.bedel@aerq.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Bedel, Alban [Tue, 23 Jul 2019 14:11:46 +0000 (14:11 +0000)]
rng-tools: start rngd early in the boot process again
It do make sense to start rngd early in the boot process because
otherwise we would need to track every service that might need entropy
and explicitly configure it to start after rngd.
When used with systemd rngd blocked the shutdown process because it
simply missed the proper unit configuration. As the systemd
documentation explains, when using DefaultDependencies=no one also
have to explicitly configure the unit to properly stop at some point.
This is normaly achieved by having Before=shutdown.target and
Conflicts=shutdown.target set for the unit.
To have rngd started early again we reverte the changes done to
rngd.service in commit edf7606822 (rng-tools: fix rngd blocks system
shutdown). To have it properly stopped on shutdown we also add
Before=shutdown.target and Conflicts=shutdown.target.
Signed-off-by: Alban Bedel <alban.bedel@aerq.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
William Bourque [Mon, 22 Jul 2019 20:13:54 +0000 (16:13 -0400)]
meta/lib/oeqa: Test for bootimg-biosplusefi Source
Add unittests for bootimg-biosplusefi SourcePlugin in wic module.
First test check wic creation works correctly.
Second test uses qemu to boot image and checks that it has both
EFI and BIOS files in a single partition.
Signed-off-by: William Bourque <wbourque@gmail.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
William Bourque [Mon, 22 Jul 2019 20:11:29 +0000 (16:11 -0400)]
wic/plugins: Source that support both EFI and BIOS
Add a source plugin that support both EFI and legacy PC-Bios.
While using this plugin, both bootloaders configurations reside
in the same /boot partitions.
This plugin has very little code : to avoid code duplication,
we simply re-import bootimg-pcbios and bootmg-efi source and
call both their SourcePlugin methods.
Signed-off-by: William Bourque <wbourque@gmail.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
obstacks from GNUlib is used in some OE-Core packages e.g. elfutils
and other packages outside OE-Core, this recipe helps provide this
functionality standalone on musl systems, and helps in getting full
versions of dependent packages
Signed-off-by: Khem Raj <raj.khem@gmail.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
'secure_getenv' api is not uniformly implemented across all C libraries
therefore its good to include missing.h so it can use the alternative
implementation where its not awvailable
Fixes
../git/src/login/pam_systemd.c:344:13: error: implicit declaration of function 'secure_getenv' is invalid in C99 [-Werror,-Wimplicit-function-declaration]
v = secure_getenv(key);
Signed-off-by: Khem Raj <raj.khem@gmail.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Swap is a special filesystem that cannot be mounted, so do not try to,
otherwise we will have service that tries and fails to mount it with
the following error:
systemd[1]: Mounting /run/media/nvme0n1p3...
mount[1229]: mount: /run/media/nvme0n1p3: unknown filesystem type 'swap'.
Signed-off-by: Oleksandr Kravchuk <oleksandr.kravchuk@pelagicore.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Stefan Agner [Sat, 20 Jul 2019 00:12:37 +0000 (02:12 +0200)]
psplash: create psplash tmpfs mount directory in psplash-init
The psplash binary uses TMPDIR as directory to store the FIFO to
communicate with the psplash tools. This directory can be in any
location an init system determines to be suitable, psplash-init
uses /mnt/ for it. Rather than creating the mount directory in
the recipe, just create it in the init script itself. This allows
other init scripts to use a different location without having
an unnecessary .psplash directory in /mnt.
Signed-off-by: Stefan Agner <stefan.agner@toradex.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Ross Burton [Fri, 19 Jul 2019 20:19:57 +0000 (21:19 +0100)]
ffmpeg: don't use hardcoded lookup tables
ffmpeg can generate lookup tables at build time instead of runtime, but this is
no longer a recommended option. The size impact is significant (12% of the
total libavcodec size, nearly 2MB), the runtime impact of dynamic tables isn't
too costly, and only a few codecs actually use the pre-generated tables (MP3,
notably).
Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Ross Burton [Fri, 19 Jul 2019 20:33:19 +0000 (21:33 +0100)]
cve-update-db-native: clean up JSON fetching
Currently the code fetches the compressed JSON, writes it to a temporary file,
uncompresses that with gzip and passes the fake file object to update_db().
Instead, uncompress the gzip'd data in memory and pass the JSON directly to
update_db().
Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Ross Burton [Fri, 19 Jul 2019 20:33:18 +0000 (21:33 +0100)]
cve-update-db-native: improve metadata parsing
The metadata parser is fragile: first it coerces a bytes() to a str() (so the
string is b'LastModifiedDate:2019...'), assumes the first line is the date, and
then uses a regex to parse (which then includes the trailing quote as part of
the date).
Clean this up by parsing the bytes as UTF-8 (ASCII is probably fine, but this is
safer), iterate through the lines and split on colons to find the right
key/value pair.
Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Drop the following patches since the issues have been fixed upstream:
0001-file01.sh-Fix-in-was-not-recognized.patch
0001-lapi-Define-TST_ABI-32-64-to-detect-target-type.patch
0001-syscalls-setrlimit03.c-read-proc-sys-fs-nr_open-for-.patch
0007-fix-__WORDSIZE-undeclared-when-building-with-musl.patch
0009-fix-redefinition-of-struct-msgbuf-error-building-wit.patch
0021-Define-_GNU_SOURCE-for-MREMAP_MAYMOVE-definition.patch
0023-ptrace-Use-int-instead-of-enum-__ptrace_request.patch
0024-rt_sigaction-rt_sigprocmark-Define-_GNU_SOURCE.patch
0026-crash01-Define-_GNU_SOURCE.patch
0028-rt_sigaction.h-Use-sighandler_t-instead-of-__sighand.patch
0034-periodic_output.patch
0039-commands-ar01-Fix-for-test-in-deterministic-mode.patch
define-sigrtmin-and-sigrtmax-for-musl.patch
setregid01-security-string-formatting.patch
Refresh the following patches:
0004-build-Add-option-to-select-libc-implementation.patch
0005-kernel-controllers-Link-with-libfts-explicitly-on-mu.patch
0008-Check-if-__GLIBC_PREREQ-is-defined-before-using-it.patch
0018-guard-mallocopt-with-__GLIBC__.patch
0020-getdents-define-getdents-getdents64-only-for-glibc.patch
0035-fix-test_proc_kill-hang.patch
0036-testcases-network-nfsv4-acl-acl1.c-Security-fix-on-s.patch
0001-open_posix_testsuite-mmap24-2-Relax-condition-a-bit.patch
0001-shmctl01-don-t-use-hardcoded-index-0-for-SHM_STAT-te.patch
0001-diotest4-Let-kernel-pick-an-address-when-calling-mma.patch
0001-getrlimit03-adjust-a-bit-of-code-to-compatiable-with.patch
wic: Fix (again) partition files UIDs on multi rootfs images
Commit 450335ba5e73a375eb9932b4c4cf37979640dbfc copies the pseudo
database to the working directory in order to have ownership information
when the filesystem is generated.
Unfortunately this does not work anymore. The filenames on the database
are absolute and there is no information about the new directory.
Instead of fixing the database, we could redo a bit the way we patch the
fstab file. Now I am saving the old contents of fstab, modifying the
file and then reverting the changes on exit.
This is faster than the previous approach, although it can cause
indeterminism if the application is killed before finishing.
Signed-off-by: Ricardo Ribalda Delgado <ricardo@ribalda.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
cve-update-db-native: Remove hash column from database.
djb2 hash algorithm was found to do collisions, so the database was
sometime missing data. Remove this hash mechanism, clear and populate
elements from scratch in PRODUCTS table if the current year needs an
update.
Signed-off-by: Pierre Le Magourou <pierre.lemagourou@softbankrobotics.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The master branch's current tip commit as of this writing is [1], see the
squashfs-tool's repo at [0].
Because of commits [2]-[4] which are included in the master branch three
corresponding patches are dropped as they are not needed anymore. The single
remaining patch was rebased on top of [1] to apply cleanly.
Commits [5] & [6] introduced interesting features, namely zstd support and
reproducibility of created SquashFS images. They are reflected in two new
PACKAGECONFIG options now, but only the latter ("reproducible") is appended to
the default options as OE-core does not contain a recipe to build zstd at the
moment (a working zstd recipe can be found e.g. in meta-rauc, see [7]).
[0] https://github.com/plougher/squashfs-tools.git
[1] f95864afe883 ("unsquashfs-4: Add more sanity checks + fix CVE-2015-4645/6")
[2] 46bdc1726e5a ("mksquashfs: Make a load of functions static")
[3] b0ca8a5c98ff ("pseudo.c: add explicit <sys/stat.h> include")
[4] f95864afe883 ("unsquashfs-4: Add more sanity checks + fix CVE-2015-4645/6")
[5] 6113361316d5 ("squashfs-tools: Add zstd support")
[6] e0d74d07bb35 ("Add configuration and Mksquashfs build options for
reproducible builds")
[7] https://layers.openembedded.org/layerindex/recipe/79049/
Signed-off-by: Ulrich Ölmann <u.oelmann@pengutronix.de> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Kai Kang [Thu, 4 Jul 2019 13:45:19 +0000 (21:45 +0800)]
defaultsetup.conf: enable select init manager
Introduce a new variable INIT_MANAGER and create 4 init-manager-*.inc
files to configure init manager settings. Available values of
INIT_MANAGER are sysvinit, systemd, mdev-busybox and a default of none.
'none' provides backwards compatibility.
The settings of various VIRTUAL-RUNTIME variables are moved into these
files from the packagegroups.
[YOCTO #13031]
[Modifications by RP for backwards compatibility]
Signed-off-by: Kai Kang <kai.kang@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Richard Purdie [Thu, 18 Jul 2019 11:46:12 +0000 (12:46 +0100)]
staging: Drop clean_recipe_sysroot
With recent changes to runqueue, this fuction is unsafe as setscene tasks can run
at the same time as normal ones and doing things before do_fetch no longer
offers any guarantees.
There is other code which cleans out things from the sysroots as tasks rerun so
we should rely upon that instead.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Alistair Francis [Wed, 17 Jul 2019 20:51:49 +0000 (13:51 -0700)]
opensbi: Fix installed-vs-shipped warning
Fix the following warning by just deleting the files:
WARNING: opensbi-0.4-r0 do_package: QA Issue: opensbi: Files/directories were installed but not shipped in any package:
/lib
/lib/libsbiutils.a
/lib/libsbi.a
Signed-off-by: Alistair Francis <alistair.francis@wdc.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Error info:
getrlimit03.c:104: FAIL: __NR_prlimit64(0) had rlim_cur = ffffffffffffffff but __NR_getrlimit(0) had rlim_cur = 7fffffff
According to kernel code: [arch/mips/include/uapi/asm/resource.h]
RLIM_INFINITY is set to 0x7fffffffUL instead of ULONG_MAX on mips32.
/*
* SuS says limits have to be unsigned.
* Which makes a ton more sense anyway,
* but we keep the old value on MIPS32,
* for compatibility:
*/
#ifndef __mips64
# define RLIM_INFINITY 0x7fffffffUL
#endif
Adding conditional statement about mips to fix this.
Signed-off-by: Hongzhi.Song <hongzhi.song@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Ross Burton [Wed, 17 Jul 2019 10:45:37 +0000 (11:45 +0100)]
flex: set CVE_PRODUCT to include vendor
There are many projects called Flex and they have CVEs, so also set the vendor
to remove these false positives.
Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Ross Burton [Wed, 17 Jul 2019 10:45:36 +0000 (11:45 +0100)]
cve-check: allow comparison of Vendor as well as Product
Some product names are too vague to be searched without also matching the
vendor, for example Flex could be the parser compiler we ship, or Adobe Flex, or
Apache Flex, or IBM Flex.
If entries in CVE_PRODUCT contain a colon then split it as vendor:product to improve the search.
Also don't use .format() to construct SQL as that can lead to security
issues. Instead, use ? placeholders and lets sqlite3 handle the escaping.
Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Andrej Valek [Wed, 17 Jul 2019 13:25:44 +0000 (15:25 +0200)]
oe/copy_buildsystem: move layer into layers directory
Layers could be located outside from poky but inside the build directory.
This case should be covered in eSDK.
meta-abc
meta-def/meta-ghi
meta-def/poky
meta-def/meta-oe/meta-oe
...
It should take all enabled layers and put them into 'layers' dir during
build-time with respecting new relative path to poky.
layers/meta-abc
layers/meta-ghi
layers/poky
layers/meta-oe/meta-oe
...
Signed-off-by: Andrej Valek <andrej.valek@siemens.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Mikko Rapeli [Wed, 17 Jul 2019 09:08:36 +0000 (12:08 +0300)]
busybox: enable unicode support
While creating and deleting files with unicode or other
encodings works, it's annoying when ls and other core utils
show questionmarks instead of the unicode characters.
In 2019, it's quite common that users of embedded devices
based on yocto need unicode support. Debugging a box with
unicode encoded file names is a bit annoying when core utils
from busybox don't support them.
The unicode config fragment has the same config as Debian in their
deb and udeb builds of version 1:1.30.1-4.
If developers do not want this or other default yocto features in busybox,
or optimize the configuration for size, then they likely run a completely
custom configuration. Thus I think it's safe to enable unicode support
by default.
Signed-off-by: Mikko Rapeli <mikko.rapeli@bmw.de> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
u-boot.bin with dtb & signature should be placed in ${B} so that
it can be deployed by u-boot as expected. Otherwise, the version
without signature is installed.
Signed-off-by: Jun Nie <jun.nie@linaro.org> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Chee Yang Lee [Wed, 17 Jul 2019 02:02:42 +0000 (10:02 +0800)]
wic: add support for kernel with initramfs bundled
when INITRAMFS_IMAGE_BUNDLE and INITRAMFS_IMAGE are set, wic should
look for kernel with initramfs image bundled.
Include required variable MACHINE, INITRAMFS_IMAGE_BUNDLE,
INITRAMFS_IMAGE, INITRAMFS_LINK_NAME and KERNEL_IMAGETYPE in WICVARS.
No longer require default value for variable kernel as KERNEL_IMAGETYPE
is not optional variable and included in WICVARS.
image_types_wic to inherit kernel-artifact-names to obtain default
INITRAMFS_LINK_NAME when INITRAMFS_IMAGE_BUNDLE are set.
update wic.Wic2.test_image_env test case to filter optional
variable INITRAMFS_LINK_NAME, INITRAMFS_IMAGE and INITRAMFS_IMAGE_BUNDLE.
Signed-off-by: Chee Yang Lee <chee.yang.lee@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Some distributions for various reasons (like for example mounting a
tmpfs over /var at runtime) can't use /var/lib to store the opkg
metadata, so a different path is required to have a functioning
package manager.
${localstatedir} can't be modified to something other than the
hardcoded value in bitbake.conf because other recipes depending on it
will fail to install.
So the only recourse, which is also the least invasive, is to allow
distros to overwrite the OPKGLIBDIR variable just like they are also
allowed to overwrite OPKGBUILDCMD.
Signed-off-by: Adrian Ratiu <adrian.ratiu@collabora.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Ross Burton [Tue, 16 Jul 2019 12:47:21 +0000 (13:47 +0100)]
glibc: exclude child recipes from CVE scanning
As glibc will be scanned for CVEs, we don't need to scan glibc-locale,
glibc-mtrace, and glibc-scripts which are all separate recipes for technical
reasons.
Exclude the recipes by setting CVE_PRODUCT in the recipe, instead of using the
global whitelist.
Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
CVE-2014-2524 is a readline CVE that was fixed in 6.3patch3 onwards, but the
tooling wasn't able to detect this version. As we now ship readline 8 we don't
need to manually whitelist it, and if we did then the whitelisting should be in
the readline recipe.
Signed-off-by: Ross Burton <ross.burton@intel.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
