On Wed, Sep 28, 2022 at 10:09 AM Zack Rusin <zackr@xxxxxxxxxx> wrote:
>
> On Tue, 2022-09-27 at 02:38 +0900, Masahiro Yamada wrote:
> > ⚠ External Email
> >
> > On Fri, Sep 23, 2022 at 3:45 AM Zack Rusin <zack@xxxxxxx> wrote:
> > >
> > > From: Zack Rusin <zackr@xxxxxxxxxx>
> > >
> > > The debug vmlinux takes up the vast majority of space in the built
> > > rpm's. While having it enabled by default is a good idea because it
> > > makes debugging easier, having an option to skip it is highly valuable
> > > for CI/CD systems where small packages are a lot easier to deal with
> > > e.g. kernel rpm built using binrpm-pkg on Fedora 36 default 5.19.8 kernel
> > > config and localmodconfig goes from 255MB to 65MB which is an almost
> > > 4x difference.
> > >
> > > To skip adding vmlinux.bz2 to the built rpm add SKIP_RPM_VMLINUX
> > > environment variable which when set to "y", e.g. via
> > > "SKIP_RPM_VMLINUX=y make binrpm-pkg" won't include vmlinux.bz2 in the
> > > built rpm.
> > >
> > > Signed-off-by: Zack Rusin <zackr@xxxxxxxxxx>
> > > Cc: Masahiro Yamada <masahiroy@xxxxxxxxxx>
> > > Cc: Michal Marek <michal.lkml@xxxxxxxxxxx>
> > > Cc: Nick Desaulniers <ndesaulniers@xxxxxxxxxx>
> > > Cc: linux-kbuild@xxxxxxxxxxxxxxx
> > > Cc: linux-kernel@xxxxxxxxxxxxxxx
> > > ---
> > > scripts/package/mkspec | 10 ++++++++--
> > > 1 file changed, 8 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/scripts/package/mkspec b/scripts/package/mkspec
> > > index 7c477ca7dc98..5a71fc0852b0 100755
> > > --- a/scripts/package/mkspec
> > > +++ b/scripts/package/mkspec
> > > @@ -23,6 +23,12 @@ else
> > > M=DEL
> > > fi
> > >
> > > +if [ "$RPM_SKIP_VMLINUX" = y ]; then
> > > + D=DEL
> > > +else
> > > + D=
> > > +fi
> > > +
> > > if grep -q CONFIG_DRM=y .config; then
> > > PROVIDES=kernel-drm
> > > fi
> > > @@ -94,8 +100,8 @@ $M $MAKE %{?_smp_mflags} INSTALL_MOD_PATH=%{buildroot} modules_install
> > > $MAKE %{?_smp_mflags} INSTALL_HDR_PATH=%{buildroot}/usr headers_install
> > > cp System.map %{buildroot}/boot/System.map-$KERNELRELEASE
> > > cp .config %{buildroot}/boot/config-$KERNELRELEASE
> > > - bzip2 -9 --keep vmlinux
> > > - mv vmlinux.bz2 %{buildroot}/boot/vmlinux-$KERNELRELEASE.bz2
> > > +$D bzip2 -9 --keep vmlinux
> > > +$D mv vmlinux.bz2 %{buildroot}/boot/vmlinux-$KERNELRELEASE.bz2
> > > $S$M rm -f %{buildroot}/lib/modules/$KERNELRELEASE/build
> > > $S$M rm -f %{buildroot}/lib/modules/$KERNELRELEASE/source
> > > $S$M mkdir -p %{buildroot}/usr/src/kernels/$KERNELRELEASE
> > > --
> > > 2.34.1
> > >
> >
> >
> >
> >
> > This came from fc370ecfdb37b853bd8e2118c7ad9f99fa9ac5cd
> > I do not know how useful or annoying it is.
> > Presumably, it was a cheesy work, and rather annoying than useful.
> >
> >
> > In debian (scripts/package/mkdebian), this kind of stuff is
> > a separate debug package, and only built when CONFIG_DEBUG_INFO=y.
> >
> >
> > Take more time in case somebody may come up with a better idea.
>
> I'd be happy to rework it in whatever way would be more convenient. Because rpm
> builds already depend on environment vars due to RPMOPTS I thought this approach was
> fitting but I'm not particularly attached to it. The important thing is the the end
> result of not including vmlinux.bz2 in the rpm itself.
>
> I think the other reasonable question to ask is: is anyone still using vmlinux.bz2
> from rpms? Because maybe just removing that code completely is the better option
> here.
Yes, I like it.
--
Best Regards
Masahiro Yamada
