[PATCH 2/6] drm: ci: Force db410c to host mode

[Rob Clark]

On Fri, Aug 25, 2023 at 6:56 AM Jani Nikula <jani.nikula at linux.intel.com> wrote:
>
> On Fri, 25 Aug 2023, Vignesh Raman <vignesh.raman at collabora.com> wrote:
> > Force db410c to host mode to fix network issue which results in failure
> > to mount root fs via NFS.
> > See https://gitlab.freedesktop.org/gfx-ci/linux/-/commit/cb72a629b8c15c80a54dda510743cefd1c4b65b8
> >
> > Since this fix is not sent upstream, add it to build.sh script
> > before building the kernel and dts. Better approach would be
> > to use devicetree overlays.
> >
> > Signed-off-by: Vignesh Raman <vignesh.raman at collabora.com>
> > ---
> >  drivers/gpu/drm/ci/build.sh | 4 ++++
> >  1 file changed, 4 insertions(+)
> >
> > diff --git a/drivers/gpu/drm/ci/build.sh b/drivers/gpu/drm/ci/build.sh
> > index 7b014287a041..c39834bd6bd7 100644
> > --- a/drivers/gpu/drm/ci/build.sh
> > +++ b/drivers/gpu/drm/ci/build.sh
> > @@ -70,6 +70,10 @@ if [ -z "$CI_MERGE_REQUEST_PROJECT_PATH" ]; then
> >      fi
> >  fi
> >
> > +# Force db410c to host mode to fix network issue which results in failure to mount root fs via NFS.
> > +# See https://gitlab.freedesktop.org/gfx-ci/linux/-/commit/cb72a629b8c15c80a54dda510743cefd1c4b65b8
> > +sed -i '/&usb {/,/status = "okay";/s/status = "okay";/&\n\tdr_mode = "host";/' arch/arm64/boot/dts/qcom/apq8016-sbc.dts
> > +
>
> It seems like a really bad idea to me to have the CI build modify the
> source tree before building.
>
> The kernel being built will have a dirty git repo, and the localversion
> will have -dirty in it.
>
> I think it would be better to do out-of-tree builds and assume the
> source is read-only.

We have the ${target_branch}-external-fixes mechanism to merge
necessary changes before building the kernel for CI.  Which is
necessary for a couple of reasons:

1) patches like this which aren't appropriate upstream but necessary
due to the CI lab setup
2) target branch if often based on an early -rc, and it isn't unheard
of to need some fix for some board or another which isn't appropriate
to land via drm-next

We should use the -external-fixes branch mechanism for patches like this one.

BR,
-R

> >  for opt in $ENABLE_KCONFIGS; do
> >    echo CONFIG_$opt=y >> drivers/gpu/drm/ci/${KERNEL_ARCH}.config
> >  done
>
> Ditto for the config changes in the context here. Those are files in
> git, don't change them.
>
> Shouldn't this use something like 'scripts/config --enable' or
> 'scripts/config --disable' on the .config file to be used for building
> instead?
>
>
> BR,
> Jani.
>
>
> --
> Jani Nikula, Intel Open Source Graphics Center