On Thu, 2026-10-01 at 09:50 +0200, Jaipaul Cheernam wrote: > NEWS: https://github.com/systemd/systemd/blob/v262-stable/NEWS#L3-L877 > > - Refresh patches for v262 > - Drop deprecated meson build options [1] > - Drop 0001-include-fix-limits.h-override.patch [2] > - Add rules: install 60-persistent-media-controller.rules [3] > - Gate systemd-coredumpd.service on kernel >= 6.19 > > 262's systemd-coredumpd.service needs PIDFD_INFO_COREDUMP_SIGNAL (Linux > 6.19); on older kernels its ExecCondition fails and the unit flaps into > start-limit-hit, failing test_systemd_failed on qemu*-alt (6.18). Add a > ConditionKernelVersion=>=6.19 drop-in so it is skipped on 6.18 and active > on 7.2. Upstream report: [4] > > [1] https://github.com/systemd/systemd/blob/v262-stable/NEWS#L41 > [2] > https://github.com/systemd/systemd/commit/6a131b32e0fe948e845ae4bf41f8fed003ec30ce > [3] https://github.com/systemd/systemd/commit/d153b49473 > [4] https://github.com/systemd/systemd/issues/43963 > > Signed-off-by: Jaipaul Cheernam <[email protected]>
> diff --git a/meta/recipes-core/systemd/systemd/coredumpd-kernel-version.conf > b/meta/recipes-core/systemd/systemd/coredumpd-kernel-version.conf > new file mode 100644 > index 00000000000..dcc23825f42 > --- /dev/null > +++ b/meta/recipes-core/systemd/systemd/coredumpd-kernel-version.conf > @@ -0,0 +1,14 @@ > +# systemd 262's kernel coredump socket manager (systemd-coredumpd.service) > +# requires the PIDFD_INFO_COREDUMP_SIGNAL pidfd flag, added in Linux 6.19 > +# (PIDFD_INFO_COREDUMP itself is from 6.16). Its ExecCondition > +# (systemd-coredump --check-requirements) fails on older kernels, and > combined > +# with Restart=on-failure and the Upholds= relationship the unit flaps into > +# start-limit-hit instead of being skipped, which fails the systemd runtime > +# tests. OE-Core ships both linux-yocto 6.18 (< 6.19) and 7.2 (>= 6.19), so > +# gate activation on the kernel version at runtime: this keeps coredumpd > active > +# on 7.2 while cleanly skipping it on 6.18. A ConditionKernelVersion= > failure is > +# a clean skip and does not trigger the restart/start-limit escalation. > +# The per-coredump [email protected] path remains available on all > kernels. > +# Upstream report: https://github.com/systemd/systemd/issues/43963 > +[Unit] > +ConditionKernelVersion=>=6.19 Hi Jaipaul, Thanks for the patch! The comment here needs a little re-work, it's too long and verbose. It also references specific kernel recipe versions which will change regularly, and this patch might not be looked at again until after the comment becomes out-of-date. We should also keep this conf file even when OE-core stops shipping Linux 6.18 because BSP layers may continue using older kernel versions for a while longer. I think just keeping the first sentence of this comment is probably enough detail. Best regards, -- Paul Barker
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#247276): https://lists.openembedded.org/g/openembedded-core/message/247276 Mute This Topic: https://lists.openembedded.org/mt/121523296/21656 Group Owner: [email protected] Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
