Re: [PATCH v1] kbuild: Add KBUILD_VMLINUX_LIBS_PRELINK

[Nathan Chancellor <nathan@kernel.org>]

On Sat, Nov 22, 2025 at 07:26:52PM +0800, Tiezhu Yang wrote:
> On 11/22/25 09:35, Nathan Chancellor wrote:
> > On Wed, Nov 19, 2025 at 12:27:08PM +0800, Tiezhu Yang wrote:
> > ...
> > > index 433849ff7529..e72d3254b93f 100755
> > > --- a/scripts/link-vmlinux.sh
> > > +++ b/scripts/link-vmlinux.sh
> > > @@ -61,12 +61,11 @@ vmlinux_link()
> > >   	shift
> > >   	if is_enabled CONFIG_LTO_CLANG || is_enabled CONFIG_X86_KERNEL_IBT; then
> > > -		# Use vmlinux.o instead of performing the slow LTO link again.
> > 
> > Why is this comment getting removed?
> 
> When using KBUILD_VMLINUX_LIBS_PRELINK or KBUILD_VMLINUX_LIBS_FINAL,
> drivers/firmware/efi/libstub/lib.a needs to be linked to the final
> vmlinux with vmlinux.o, so I thought the above comment is not proper,
> it I misunderstood it, I will keep the comment as is.

I think this comment is referring to the use of vmlinux.o instead of
vmlinux.a, nothing to do with libs=, so I think the comment should
remain.

> > >   		objs=vmlinux.o
> > > -		libs=
> > > +		libs="${KBUILD_VMLINUX_LIBS_PRELINK}"
> > >   	else
> > >   		objs=vmlinux.a
> > > -		libs="${KBUILD_VMLINUX_LIBS}"
> > > +		libs="${KBUILD_VMLINUX_LIBS} ${KBUILD_VMLINUX_LIBS_PRELINK}"
> > >   	fi
> > >   	if is_enabled CONFIG_GENERIC_BUILTIN_DTB; then
> 
> FYI, the previous patch is still under discussion yesterday [1],
> I do not know whether this patch is necessary and what is the next
> step [2].
> 
> [1] https://lore.kernel.org/loongarch/CAAhV-H4AasfFet_Gi_mVyte3RPMH3qBS73dBfF-=Gd7HJ6ZPEw@mail.gmail.com/
> [2] https://lore.kernel.org/loongarch/39617a3e-c476-abac-8425-bbcece769cdb@loongson.cn/

Thanks, we will wait to see the conclusion of those threads before
applying this.

Cheers,
Nathan