Re: [OE-core] [PATCH 0/21] linux-yocto: combined and consolidated pull request

[Mikko Rapeli <mikko.rapeli@linaro.org>]

Hi,

On Tue, Dec 23, 2025 at 09:31:06AM +0200, Mikko Rapeli via lists.openembedded.org wrote:
> On Mon, Dec 22, 2025 at 10:34:14PM -0500, Bruce Ashfield via lists.openembedded.org wrote:
> > On Mon, Dec 22, 2025 at 9:13 AM Richard Purdie
> > <richard.purdie@linuxfoundation.org> wrote:
> > >
> > > Hi Bruce,
> > >
> > > On Thu, 2025-12-18 at 15:22 -0500, bruce.ashfield@gmail.com wrote:
> > > > As discussed in the weekly engineering call, I'm sending this as
> > > > single series even though it crossed boundaries of the repositories.
> > > >
> > > > I don't have access to meta-yocto-contrib yet, so I haven't pushed
> > > > the component parts to contrib branches yet, but will do that for
> > > > future pull requests.
> > > >
> > > > I'm using this as a first run through some things I've cooked up
> > > > to coordinate the patches across the repos. The formatting is my
> > > > own, so apologies if it isn't clear .. and we'll see if it all
> > > >
> > > > Now to the series for OE-core:
> > > >
> > > >  - We have stable updates to 6.12
> > > >  - A kern-tools fix
> > > >  - Updates to 6.17
> > > >  - Removal of 6.17
> > > >  - Introduction of 6.18
> > > >  - Bump to libc-headers to 6.18
> > > >
> > > > I've built and booted what I can locally, and I know that the
> > > > major architectures work, and are functional with respect to
> > > > core features.
> > > >
> > > > I have more libc-headers testing running against meta-oe and
> > > > muslc, so there may be some breakage there and I'll help deal
> > > > with that if it happens.
> > > >
> > > > The update and then removal of 6.17 is on purpose. In case any
> > > > one was using it, they should get the latest tested before it
> > > > starts to be removed.
> > > >
> > > > For meta-yocto:
> > > >
> > > >   - removal of any remaining 6.17 references
> > > >   - introduction of 6.18
> > > >
> > > > The meta-yocto default changes are obviously REALLY RFC/RFT
> > > > there will be issues, but I've provided them anyway to make it
> > > > clear that we are going to 6.18 as the new default for all the
> > > > variants.
> > > >
> > > > Once we get 6.18 fully green and the h/w references udpated,
> > > > I'll remove 6.12 and 6.16 from master. That isn't in this series
> > > > by design.
> > >
> > > I ran this through some testing and it isn't straight forward.
> > >
> > 
> > Back online after a 10 hour drive!
> > 
> > > We need a newer strace with the newer kernel so I queued that patch
> > > from Robert. That caused ptest failures so I disabled the failing bpf
> > > strace ptests.
> > 
> > ok. I'll ignore strace and assume that bump will handle it.
> > 
> > >
> > > ltp failed to compile so I backported a patch to fix that.
> > >
> > 
> > ack'd
> > 
> > > We still have a cryptodev-module failure:
> > > https://autobuilder.yoctoproject.org/valkyrie/#/builders/6/builds/2922
> > > https://autobuilder.yoctoproject.org/valkyrie/#/builders/25/builds/2897
> > > (all world builds I think)
> > 
> > I can handle this one, it'll be.a day or so, but I also haven't checked
> > the  autoupdate list yet to see if this is in it. I'll check the lists before
> > I start and see if anyone else has an update in flight.
> > 
> > >
> > > and an initramfs module space problem:
> > > https://autobuilder.yoctoproject.org/valkyrie/#/builders/22/builds/2931
> > 
> > And a config warning (easy for me to fix).'
> > 
> > For the space. Is it only genericarm64 that we do this test on ?  It
> > could either e normal kernel size increases, or it could be all the work
> > that has been done in genericarm64's config that has causes this.
> > 
> > If it is normal kernel, then I can have a look for something obvious,
> > and then we can increase the size. If it is only genericarm64, then
> > we should get Mikko's opinion on what might be options we could
> > tweak to reduce the size.
> 
> This is normal. More drivers as modules, dependencies to firmware and slight
> increase in kernel size hitting the limit again. I think increasing the limit
> is the way to go now. I've been hitting this limit locally with 6.17 kernel too
> after some more arm64 drivers are enabled.

I see Richard added a patch in meta-yocto master-next to update genericarm64
INITRAMFS_MAXSIZE to 280 Mb. This is fine for me.

A lot of the kernel drivers and firmware blobs are strictly not needed in initramfs
to mount the rootfs, but we can't separate them at the moment without full manual
control of the initramfs package contents.

Cheers,

-Mikko