Hi Bruce, OK, no problem, I will send an updated patch Richard

Hi Bruce, Thanks. For LTS, I think we said the strategy should be a bit different: - keep the original versions and update them. - keep the same default version - add/update other recipe upon

merged. Bruce In message: [meta-virtualization][PATCH] libvmi: Fix out of box build failure with xen disabled on 13/04/2022 He Zhe wrote:

merged. Bruce In message: [meta-virtualization] [hardknott][PATCH] libvirt: fix CVE-2022-0897 on 13/04/2022 Changqing Li wrote:

I'd actually prefer that the list not be sorted, at the same time as we are adding to it. Can you re-submit with just the new kernel module recommendations added. Bruce -- - Thou shalt not follow

FYI: I'm not ignoring these, and I will make sure they get into the upcoming release branch and are in place for the LTS release. I'm just waiting until Christopher gets a chance to review and see if

vbd was added for xen based VMs since bdee00fac9b4 ("Disk reading interface for Xen based VMs and example") and should be built only when xen is enabled, otherwise there would not be necessary xen

Hi Bruce, Could you please also backport this patch to the honister branch? Thanks! Richard

K3s (and Kubernetes) supports load balancing via IPVS, and by default rep= orts errors when IPVS kernel modules cannot be loaded. This patch adds the missing reported kernel modules to the k3s

From: Changqing Li <changqing.li@...> Signed-off-by: Changqing Li <changqing.li@...> --- .../libvirt/libvirt/CVE-2022-0897.patch | 57 +++++++++++++++++++

Just for the record, there is${nonarch_libdir} that expands to “/usr/lib” regardless of whether multilib is used or not. It is typically intended for cases like this. //Peter

Hi Qi You are right. My build configuration is also using multilib and this issue appeared.

Hmm. Indeed. In that case, I'll go with a variant of the original patch, with a tweaked commit log. I'll take care of that now. Bruce -- - Thou shalt not follow the NULL pointer, for chaos and

The only package I found which uses /usr/lib/sysctl.d is systemd. It packages ${exec_prefix}/lib/sysctl.d. And it has its reason. It treats /usr/lib as an arch-independent vendor dir, and $libdir is

I've done multiple executions of that exact set of steps, and I've never seen the error you are encountering. Clearly there's something different in the various build environments. We are also getting

That's a larger issue, and it breaks other parts of the system. We shouldn't have anything with hard coded /usr/lib. How are other packages dealing with this ? In that case, I'll do individual scripts

See the patch that I sent as well. There's no reason to have /usr/lib anywhere in the packaging at all. We should just rely on ${libdir}, if libvirt is installing outside of the definition of

Hi Bruce, Looking at the issue a little further, I found we need to use the hardcoded ‘/usr/lib’. This is because sysctl is hardcoding that. Some codes in sysctl.c from procps: static

The libvirtd package has already packaged ${sysconfdir}/sysctl.d, and also the /usr/lib/sysctl.d/60-libvirtd.conf, it would be reasonable for it to package other sysctl conf files, otherwise we will

I've pushed my WIP patch here: https://git.yoctoproject.org/meta-virtualization/commit/?h=master-next Bruce -- - Thou shalt not follow the NULL pointer, for chaos and madness await thee at its end -