[PATCH v2] efistub: Only link libstub to final vmlinux

[Huacai Chen]

On Sun, Sep 28, 2025 at 10:40 PM Ard Biesheuvel <ardb at kernel.org> wrote:
>
> On Sun, 28 Sept 2025 at 15:52, Huacai Chen <chenhuacai at kernel.org> wrote:
> >
> > Hi, Ard,
> >
> > On Sun, Sep 28, 2025 at 9:42 PM Ard Biesheuvel <ardb at kernel.org> wrote:
> > >
> > > On Sun, 28 Sept 2025 at 10:55, Tiezhu Yang <yangtiezhu at loongson.cn> wrote:
> > > >
> > > > When compiling with LLVM and CONFIG_LTO_CLANG is set, there exists
> > > > the following objtool warning on LoongArch:
> > > >
> > > >   vmlinux.o: warning: objtool: __efistub_efi_boot_kernel()
> > > >   falls through to next function __efistub_exit_boot_func()
> > > >
> > > > This is because efi_boot_kernel() doesn't end with a return instruction
> > > > or an unconditional jump, then objtool has determined that the function
> > > > can fall through into the next function.
> > > >
> > > > At the beginning, try to do something to make efi_boot_kernel() ends with
> > > > an unconditional jump instruction, but this modification seems not proper.
> > > >
> > > > Since the efistub functions are useless for stack unwinder, they can be
> > > > ignored by objtool. After many discussions, no need to link libstub to
> > > > the vmlinux.o, only link libstub to the final vmlinux.
> > > >
> > >
> > > Please try keeping these changes confined to arch/loongarch. This
> > > problem does not exist on other architectures, and changing the way
> > > vmlinux is constructed might create other issues down the road.
> > ARM, RISC-V and LoongArch do things exactly in the same way. Now
> > LoongArch is the first of the three to enable objtool, so we meet the
> > problem first.
> >
> > But yes, I also don't want to change the way of constructing vmlinux.
> > So I prefer the earliest way to fix this problem.
> > https://lore.kernel.org/loongarch/CAAhV-H7fRHGFVKV8HitRgmuoDPt5ODt--iSuV0EmeeUb9d5FNw@mail.gmail.com/T/#meef7411abd14f4c28c85e686614aa9211fccdca0
> >
>
> Can we just drop the __noreturn annotation from kernel_entry_t, and
> return EFI_SUCCESS from efi_boot_kernel()?
Not good, because kernel_entry_t is really "noreturn", and at present
no architecture returns EFI_SUCCESS at the end of efi_boot_kernel().


Huacai