This patch applies the upstream glib-2-88 stable backport chain for
CVE-2026-58015. The issue is in the D-Bus SHA-1 authentication
mechanism, where a malicious peer could provide an unchecked cookie
context and cause the client to access unintended files while resolving
the cookie challenge.
Backport the upstream GLib fix chain from the glib-2-88 stable branch:
- db9c8fae398b validates cookie_context before keyring lookup. This is
the primary security fix for CVE-2026-58015 [1].
- c0531125344b tightens cookie ID parsing so empty, negative, and
out-of-range values are rejected. This hardens the same SHA-1 cookie
challenge parser and is covered by the upstream regression test [2].
- 060aea67de75 exposes the private client reject-reason vfunc. This is
test-support plumbing required by the upstream regression test [3].
- 091930196229 adds the upstream regression test for SHA-1 cookie
challenge parsing [4].
Add dbus-native to PACKAGECONFIG[tests] so Meson can find dbus-daemon
when building the new installed D-Bus regression test for ptest. This is
kept as a native-only test dependency to avoid adding a target dbus
dependency to glib.
[1] db9c8fae39
[2] c053112534
[3] 060aea67de
[4] 0919301962
[5] https://nvd.nist.gov/vuln/detail/CVE-2026-58015
(From OE-Core rev: b6b82e3c1442b658bd4c1689e09792c9a3a96947)
Signed-off-by: Deepak Rathore <deeratho@cisco.com>
Signed-off-by: Fabien Thomas <fabien.thomas@smile.fr>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Bump b-a SRCREV again to pull in docs changes.
(From OE-Core rev: 70dc15941dd33270a92d1001174efb3093e79bdf)
Signed-off-by: Paul Barker <paul@pbarker.dev>
A flaw was found in GLib. A state confusion issue exists in
g_dbus_node_info_new_for_xml() in the gio/gdbusintrospection.c file when
processing malformed D-Bus introspection XML, specifically with a <node>
element nested within other elements like <method>, <signal>, <property>
or <arg>. This issue can cause an unsigned integer overflow and lead to an
out-of-bounds read, resulting in a denial of service.
The CVE NVD entry is wrong, it indicates that the CVE is fixed in 2.88.1
but the fix was realized in 2.89.0, see [1]. The fix is not present in 2.88.2.
[1] c9da977c17
(From OE-Core rev: b7bb8b7eace676556d50d130a08a1363088c98d5)
Signed-off-by: Benjamin Robin (Schneider Electric) <benjamin.robin@bootlin.com>
Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit d52f4d582cc71ada3c8ebe54be1a5b70278ea1ca)
[YC: re-added the removed Signed-off-bys from the patches]
Signed-off-by: Yoann Congal <yoann.congal@smile.fr>
Signed-off-by: Paul Barker <paul@pbarker.dev>
This patch applies the upstream stable/v2.41 backport for
CVE-2026-13595. The upstream fix merge or commit is referenced in [1],
and the public CVE advisory is referenced in [2]. The individual
backported commit links are recorded in the embedded patch headers
when the fix expands to multiple commits.
[1] 132d9c8aa1
[2] https://access.redhat.com/security/cve/CVE-2026-13595
(From OE-Core rev: 9f4a0789961def89303e6e8a211c944e41c29cab)
Signed-off-by: Deepak Rathore <deeratho@cisco.com>
Signed-off-by: Yoann Congal <yoann.congal@smile.fr>
Signed-off-by: Paul Barker <paul@pbarker.dev>
git log --oneline ce65d944e38a20cb70af2a48a4b8aa5d8fabe1cc..be1e627cd72db31161a3b4ce1c8114674f0895eb
be1e627cd7 Linux: Only define OPEN_TREE_* macros in <sys/mount.h> if undefined (bug 33921)
98bc06a361 include: isolate __O_CLOEXEC flag for sys/mount.h and fcntl.h
3e13579841 Use pending character state in IBM1390, IBM1399 character sets (CVE-2026-4046)
0dc95ae109 elf: parse /proc/self/maps as the last resort to find the gap for tst-link-map-contiguous-ldso
9344c796f7 resolv: Check hostname for validity (CVE-2026-4438)
5663ab0b83 resolv: Count records correctly (CVE-2026-4437)
c53cd6e738 posix: Run tst-wordexp-reuse-mem test
2760e4c5ed iconvdata: Fix invalid pointer arithmetic in ANSI_X3.110 module
ba29a36aa3 posix: Fix invalid flags test for p{write,read}v2
60b039bf6a socket: Add new test for shutdown
Testing Results:
Before After Diff
PASS 4892 4896 +4
XPASS 4 4 0
FAIL 371 372 +1
XFAIL 16 16 0
UNSUPPORTED 224 224 0
Changes in testcases:
testcase-name before after
posix/tst-wordexp-reuse-mem(new) - PASS (native)
[Note: posix/tst-wordexp-reuse-mem is a new test added by this uplift
(c53cd6e738). It fails under QEMU user-mode because the test-wrapper
cannot support LD_PRELOAD and MALLOC_TRACE needed for mtrace. Running
natively with LD_PRELOAD=libc_malloc_debug.so confirms the test passes
with no memory leaks.
nptl/tst-getpid3 is a flaky test under QEMU user-mode (passes 7/10
re-runs). No nptl code was changed in this uplift.]
(From OE-Core rev: 2afc207fb9d2b720912a5eedb0f368e40f5a236e)
Signed-off-by: Jaipaul Cheernam <jaipaul.cheernam@est.tech>
Signed-off-by: Yoann Congal <yoann.congal@smile.fr>
Signed-off-by: Paul Barker <paul@pbarker.dev>
When moving the updated CVE database file to the downloads directory,
ensure that it has a different inode number to the previous version of
this file.
We have seen "sqlite3.DatabaseError: database disk image is malformed"
exceptions on our autobuilder when trying to read the CVE database in
do_cve_check tasks. The context here is that the downloads directory
(where the updated database file is copied to) is shared between workers
as an NFS mount. Different autobuilder workers were seeing different
checksums for the database file, which indicates that a mix of both new
and stale data was being read. Forcing each new version of the database
file to have a different inode number will prevent stale data from being
read from local caches.
This should fix [YOCTO #16086].
(From OE-Core rev: 8aca19cff468c5f15c919c973c46be58e020af46)
Signed-off-by: Paul Barker <paul@pbarker.dev>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit f63622bbec1cfaca6d0b3e05e11466e4c10fa86e)
[YC: removed cve-update-db-native part, file was removed in
17eb0788514 (cve-update-db-native: remove, 2023-06-23)]
Signed-off-by: Yoann Congal <yoann.congal@smile.fr>
Signed-off-by: Paul Barker <paul@pbarker.dev>
After the gawk dependency change, we need to change PR/hashequiv version
to replace the corrupted sstate/hashequiv data.
(From OE-Core rev: a455b21f9170b3f2d74763b5bf99625dbda81ff9)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit f0f7632595792a73ea0a935b924e8bdf9954ec7b)
Signed-off-by: Yoann Congal <yoann.congal@smile.fr>
Signed-off-by: Paul Barker <paul@pbarker.dev>
Different versions of gawk can produce different output,
so depending on which version is installed on the build host,
reproducibility issues can occur:
https://bugzilla.yoctoproject.org/show_bug.cgi?id=16072
So far only glibc and grub have been identified to have
the issue; probably more fixes of similar nature will be
required going forward.
Adjust the gawk recipe to apply target-only tweaks
(particularly the removal of awk symlink to allow for alternatives)
to only target and nativesdk variants, so that native installs
both awk and gawk executables.
[YOCTO #16072]
(From OE-Core rev: 288ecfd7d9cd24222cc0f1277105c15cf0889718)
Signed-off-by: Alexander Kanavin <alex@linutronix.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit c5bbf0a60b1d63e68f849a63e5d3872954e7cd3f)
Signed-off-by: Yoann Congal <yoann.congal@smile.fr>
Signed-off-by: Paul Barker <paul@pbarker.dev>
It doesn't apply after udev-builtin-path_id.c was also updated by
CVE-2026-40225-01.patch in
https://git.openembedded.org/openembedded-core/commit/?h=scarthgap&id=fc2d33dbb2d5180b77c10865156db342f9d582da
causing do_patch failure for scarthgap builds with musl:
Hunk #1 FAILED at 39.
1 out of 1 hunk FAILED -- rejects in file src/udev/udev-builtin-net_id.c
patching file src/udev/udev-builtin-path_id.c
patching file src/udev/udev-event.c
patching file src/udev/udev-rules.c
Patch 0003-src-basic-missing.h-check-for-missing-strndupa.patch does not apply (enforce with -f)
stderr: ')
ERROR: Logfile of failure stored in: /OE/build/oe-core/tmp-musl/work/core2-64-oe-linux-musl/systemd/255.21/temp/log.do_patch.215528
ERROR: Task (/OE/build/oe-core/openembedded-core/meta/recipes-core/systemd/systemd_255.21.bb:do_patch) failed with exit code '1'
(From OE-Core rev: 0e66eb22a34e17939cfdaf5cdad84361b7f18e6e)
Signed-off-by: Martin Jansa <martin.jansa@gmail.com>
Signed-off-by: Yoann Congal <yoann.congal@smile.fr>
Signed-off-by: Paul Barker <paul@pbarker.dev>
Fixes [YOCTO #16265]
The glibc recipe is supposed to be building with
--enable-stack-protector=strong, but some CACHED_CONFIGVARS values are
actually breaking this, causing glibc to be built with no stack
protector at all.
Remove these CACHED_CONFIGVARS values so that stack protector support is
detected properly in do_configure and then enabled properly during
do_compile.
Full details are here:
https://bugzilla.yoctoproject.org/show_bug.cgi?id=16265
(From OE-Core rev: 7952d214393b6c5230ba115f63b6f6d245a728bc)
Signed-off-by: Ivan Nestlerode <ivan.nestlerode@sonos.com>
Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 43f0602ede37428f3c35cf665bba934b84355240)
Signed-off-by: Yoann Congal <yoann.congal@smile.fr>
Signed-off-by: Paul Barker <paul@pbarker.dev>
This removes rust uutils coreutils CVEs from reports.
Comparing sbom-cve-check shows that only
CVE-2026-35338..CVE-2026-35381 are removed and all of them contained
reference to uutils.
(From OE-Core rev: 348391ccf91ac474252f75a5679fc42505faa54d)
Signed-off-by: Peter Marko <peter.marko@siemens.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(From OE-Core rev: 5c39687f62e5864ea783cbed497c2eb5387dcf96)
Signed-off-by: Peter Marko <peter.marko@siemens.com>
Signed-off-by: Yoann Congal <yoann.congal@smile.fr>
Signed-off-by: Paul Barker <paul@pbarker.dev>
According to [1],
EDK2 contains a vulnerability in BIOS where an attacker may cause “Exposure of
Sensitive Information to an Unauthorized Actor” by local access. Successful
exploitation of this vulnerability will lead to possible information disclosure
or escalation of privilege and impact Confidentiality.
Backport a patch [2] from upstream to fix CVE-2024-38798
[1] https://nvd.nist.gov/vuln/detail/CVE-2024-38798
[2] 0cad130cb4
(From OE-Core rev: ed444adf325d3a985ed8f9ae0a009ecbaf67c3fd)
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Fabien Thomas <fabien.thomas@smile.fr>
Signed-off-by: Paul Barker <paul@pbarker.dev>
According to [1], EDK2 contains a vulnerability in BIOS where an attacker may
cause “ Improper Input Validation” by local access. Successful exploitation of
this vulnerability could alter control flow in unexpected ways, potentially
allowing arbitrary command execution and impacting Confidentiality, Integrity,
and Availability.
Backport patches from upstream [2] to fix CVE-2025-2296
Note: backport 0001-AmdSev-Halt-on-failed-blob-allocation.patch to apply
the CVE patches without confliction
[1] https://nvd.nist.gov/vuln/detail/CVE-2025-2296
[2] https://github.com/tianocore/edk2/pull/10628
(From OE-Core rev: 09be6658833e7ac4143eeb26bdaf67c6c94e260a)
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Fabien Thomas <fabien.thomas@smile.fr>
Signed-off-by: Paul Barker <paul@pbarker.dev>
On scarthgap images built without systemd in DISTRO_FEATURES, dbus
still shipped dbus.socket and dbus.service under
${systemd_user_unitdir} (/usr/lib/systemd/user), because the
'user-session' PACKAGECONFIG was always enabled and passed
--enable-user-session --with-systemduserunitdir=... to configure.
In dbus-1.14.10 the user-session autoconf option (configure.ac and
bus/Makefile.am 'if DBUS_ENABLE_USER_SESSION') only installs systemd
user units; it has no non-systemd effect. Enabling it on a sysvinit
image has no benefit and produces the stale unit files.
Make user-session a systemd-gated PACKAGECONFIG by using
bb.utils.contains in the default, so it is enabled when systemd is
in DISTRO_FEATURES and disabled otherwise. No changes to the
PACKAGECONFIG[user-session] or PACKAGECONFIG[systemd] entries are
needed: --disable-user-session is passed on sysvinit builds, which
prevents the configure/Makefile machinery from ever setting up the
user-unit install step.
This is the scarthgap equivalent of master commit a296b0623eb2
("dbus: use the systemd class to handle the unit files"), adapted
to the autotools 1.14.10 recipe. The master fix was broader because
the meson 1.16.2 build handles unit-file install differently, which
let that commit drop the manual do_install unit block, the
systemctl mask postinst, and PACKAGE_WRITE_DEPS. On 1.14.10 those
pieces are still needed; the minimal correct gate here is the
user-session default.
Fixes [YOCTO #15779]
(From OE-Core rev: 5550d6eadb2fea1ecb13e035a04a57450510441f)
Signed-off-by: Jhonata Poma-Hansen <jhonata.poma@gmail.com>
Signed-off-by: Fabien Thomas <fabien.thomas@smile.fr>
Signed-off-by: Paul Barker <paul@pbarker.dev>
Following up on commit 139102a73d41 ("recipes: Default to https git protocol where possible"),
> The recommendation from server maintainers is that the https protocol
> is both faster and more reliable than the dedicated git protocol at this point.
> Switch to it where possible.
(cherry picked from commit 9bab238d26a3317a6212dc711427f4917eaac50e)
(From OE-Core rev: 6cfdf7fc1a63a3c2ee8093462fedc80d6358c54c)
Signed-off-by: Yoann Congal <yoann.congal@smile.fr>
Signed-off-by: Paul Barker <paul@pbarker.dev>
