[OE-core] [PATCH 0/9] kernel-yocto: consolidated pull request

[Bruce Ashfield]

On 08/22/2017 10:24 AM, Richard Purdie wrote:
> On Mon, 2017-08-21 at 12:12 -0400, Bruce Ashfield wrote:
>> On 2017-08-21 11:55 AM, Richard Purdie wrote:
>>>
>>> On Sun, 2017-08-20 at 22:58 -0400, Bruce Ashfield wrote:
>>>>
>>>> Here's the collected set of kernel changes that are ready for the
>>>> M3
>>>> builds.
>>>>
>>>> We have some kernel meta data changes:
>>>>
>>>>     kernel-yocto: configuration updates: x86 features
>>>>     linux-yocto/4.1: fix fsl-ls10xx sdhci
>>>>     linux-yocto: add usb-net configs by default
>>>>
>>>> Some bug fixes:
>>>>
>>>>     kernel-yocto: ensure that only valid BSPs are built
>>>>     linux-yocto/4.10: CVE & misc fixes
>>>>
>>>> And the important changes are the addition of the 4.12 kernel +
>>>> libc headers
>>>> and the removal of the 4.1 kernel.
>>>>
>>>> I also have some -stable updates queued for the active kernel
>>>> versions, but
>>>> they will come out once this series has made it into the tree,
>>>> since I'm
>>>> trying to isolate any 4.12 issues from other -stable updates.
>>>>
>>>> I built and booted all the qemu targets for 4.12, and have built
>>>> for
>>>> core-image-sato, core-image-kernel-dev, glibc and muslc for all
>>>> of the
>>>> architectures. My testing didn't pick up any regressions, but I
>>>> was battling
>>>> some other build errors (non kernel related) during my testing,
>>>> so I can
>>>> never be 100% sure.
>>>>
>>>> I've sent patches to poky to remove/bump any references to the
>>>> 4.1 kernel
>>>> .. if I've missed any, I'm sure people will shout.
>>> Thanks for the other fix, not got that tested yet as builds are
>>> still
>>> ongoing. However there appears to be a perf issue:
>>>
>>> https://autobuilder.yocto.io/builders/nightly-deb-non-deb/builds/42
>>> 9/steps/BuildImages/logs/stdio
>>> https://autobuilder.yocto.io/builders/nightly-packagemanagers/build
>>> s/77/steps/BuildImages/logs/stdio
>>> https://autobuilder.yocto.io/builders/nightly-packagemanagers/build
>>> s/77/steps/BuildImages_1/logs/stdio
>>> https://autobuilder.yocto.io/builders/nightly-packagemanagers/build
>>> s/77/steps/BuildImages_2/logs/stdio
>>> [and so on]
>>>
>> Aha. There's a difference in the clibrary that I was using. I'll have
>> a closer look once my switch over and build gets going.
> 
> Thanks, I think you got that one but the /bin/awk issue remained after
> your changes. Its easy to reproduce with "bitbake core-image-sato-sdk",
> or IMAGE_INSTALL_append = " kernel-devsrc". I ended up with this change
> to fix it:

Yah, I noticed all of those references as well. Saul had dealt with
the one that hit us before, but I wasn't able to trigger any installs
of the other parts in the images I built .. so they were being left
be for now, just to keep our footprint low.

> 
> http://git.yoctoproject.org/cgit.cgi/poky/commit/?h=master-next&id=a28acfa6c5890df4f5cb09962ef0c53837ac3b37
> 
> which I appreciate you won't like but it illustrates the problem. Any
> preference on what to do with that patch?

I can merge it directly into linux-yocto, since that is what we did
with Saul's change before, and at the same time, we can submit it
upstream and drop it next release.

> 
> We do also have a boot test issue on mips64:
> 
> https://autobuilder.yocto.io/builders/nightly-mips64/builds/423/steps/Running%20Sanity%20Tests/logs/stdio
> 
> and mips:
> 
> https://autobuilder.yocto.io/builders/nightly-mips/builds/425/steps/Running%20Sanity%20Tests/logs/stdio
> 
> I assume we just have to whitelist that error message?

Yep. It is something that has been around for a while and comes from
the general open firmware support. It is a warning only (obviously)
and can safely be ignored. We keep the OF support on for compatibility
so live with a few warnings like that.

Bruce

> 
> Cheers,
> 
> Richard
>