Re: [PATCH 00/10] kernel-yocto: consolidated pull request

[Bruce Ashfield <bruce.ashfield@gmail.com>]

In message: Re: [PATCH 00/10] kernel-yocto: consolidated pull request
on 02/08/2024 Richard Purdie wrote:

> Hi Bruce,
> 
> On Thu, 2024-08-01 at 11:08 -0400, bruce.ashfield@gmail.com wrote:
> > Here's the next batch of kernel updates. Obviously the -stable
> > updates to 6.1 are routine, but the larger changes are the
> > introduction of the 6.10 kernel as a new / validated reference
> > version.
> > 
> > Some notes:
> > 
> > � - I've included the poky/poky-tiny preferred version bumps
> > ��� in this pull request for simplicity. Yes, I know they go
> > ��� to meta-yocto, but it is easier to make this available for
> > ��� AB testing when they are all together.
> > 
> > � - I'm watching the status of 6.10, if for some reason it
> > ��� EOLs before we release, we'll jump to 6.11, which I have
> > ��� under test and will push as linux-yocto-dev shortly.
> > 
> > � - There's one known build issue with 6.10, it is zfs in
> > ��� meta-oe. The upstream hasn't updated to support 6.10 yet,
> > ��� so I'm watching and waiting on that .. but there's no
> > ��� reason to hold this request for more testing on that.
> > 
> > � - I've tested 6.10 for all architectures on core-image-minimal,
> > ��� core-image-sato and core-image-kernel-dev. I've also
> > ��� tested against musl and poky-altconfig (6.1 with the
> > ��� 6.10 headers)
> > 
> > There will surely be AB issues with this, but it is now
> > to the point where I need additional coverage and eyes
> > on the changes.
> 
> Thanks for this, I put it through testing after tweaking it to remove a
> rouge inherit. There were some issues:
> 
> a) genericarm64 version issue:
> https://autobuilder.yoctoproject.org/typhoon/#/builders/166/builds/687
> 
> Warnings since genericarm64 isn't available as 6.10. Everything else
> seemed to miss the preferred_version and build 6.6

Unlike the other hardware references, genericarm64 doesn't set
a preferred version. It isn't something that I typically test,
since it is one of the hardware platform.

Until it is tested against 6.10, I could send an update to set
it explicity to 6.6.

Would that be acceptable to everyone ?

> 
> b) crypto-dev failures:
> https://autobuilder.yoctoproject.org/typhoon/#/builders/64/builds/9329
> https://autobuilder.yoctoproject.org/typhoon/#/builders/45/builds/9340
> https://autobuilder.yoctoproject.org/typhoon/#/builders/40/builds/9313
> https://autobuilder.yoctoproject.org/typhoon/#/builders/52/builds/9161/steps/11/logs/stdio
> https://autobuilder.yoctoproject.org/typhoon/#/builders/117/builds/5141/steps/12/logs/stdio
> (the latter taking out the reproducibility build so that wasn't tested)

Hopefully your patch addressed it, let me know if I need to
look into it more.

> 
> c) meta-virt xen issues:
> https://autobuilder.yoctoproject.org/typhoon/#/builders/128/builds/3455

I pushed a new .inc to meta-virt

> 
> d) meta-agl kernel config fragment warnings:
> https://autobuilder.yoctoproject.org/typhoon/#/builders/120/builds/4556
> 
> e) meta-arm kernel config fragment warnings:
> https://autobuilder.yoctoproject.org/typhoon/#/builders/113/builds/5810

d/e I'll leave for the layer owners.

> 
> f) qemumips64 parselogs issue:
> https://autobuilder.yoctoproject.org/typhoon/#/builders/74/builds/9240/steps/14/logs/stdio
> "failed when register_sysctl_sz sched_fair_sysctls to kernel"
> "failed when register_sysctl_sz sched_core_sysctls to kernel"
> "pci 0000:00:00.0: [Firmware Bug]: BAR 5: invalid; can't size"
> (and other BARs)

One of the format's changed and the other is harmless, I have
a patch to update and add the other to what we ignore.

> 
> g) qemumips parselogs issue:
> https://autobuilder.yoctoproject.org/typhoon/#/builders/60/builds/9256/steps/15/logs/stdio
> Similar to qemumips64

Same fix.

> 
> h) qemuarmv5 config fragment warning:
> https://autobuilder.yoctoproject.org/typhoon/#/builders/160/builds/989

Fixed.

I'll send v2 later in the day on Tuesday.

Bruce

> 
> I've copied agl/arm people.
> 
> Probably nothing too surprising. I'll drop some of the patches whilst
> we resolve these but see if we can keep the headers upgrade and the
> point release bumps.
> 
> I'll try and test the crypto-dev upgrade on the list as that might help
> that issue.
> 
> Cheers,
> 
> Richard
> 
>