[PATCH 00/12] ARM: use adr_l/ldr_l macros for PC-relative references

[Ard Biesheuvel]

On Wed, 16 Sep 2020 at 22:53, Nick Desaulniers <ndesaulniers at google.com> wrote:
>
> On Tue, Sep 15, 2020 at 10:55 PM Ard Biesheuvel <ardb at kernel.org> wrote:
> >
> > On Wed, 16 Sep 2020 at 02:31, Nick Desaulniers <ndesaulniers at google.com> wrote:
> > >
> > > On Tue, Sep 15, 2020 at 2:30 PM Ard Biesheuvel <ardb at kernel.org> wrote:
> > > >
> > > > Excellent, thanks for testing. Do you have any coverage for Thumb2
> > > > builds as well? (CONFIG_THUMB2_KERNEL=y)
> > >
> > > Ah, right, manually testing ARCH=arm defconfig with
> > > CONFIG_THUMB2_KERNEL enabled via menuconfig:
> > >
> > > Builds and boots for clang.
> > >
> > > (Also needed https://lore.kernel.org/lkml/20200915225751.274531-1-ndesaulniers@google.com/T/#u
> > > for an unrelated issue).
> > >
> > > For GCC, I observe:
> > >
> > > arch/arm/vfp/vfphw.o: in function `vfp_support_entry':
> > > (.text+0xa): relocation truncated to fit: R_ARM_THM_JUMP19 against
> > > symbol `vfp_kmode_exception' defined in .text.unlikely section in
> > > arch/arm/vfp/vfpmodule.o
> > >
> >
> > Interesting. And this is using ld.bfd ?
>
> $ arm-linux-gnueabihf-ld --version
> GNU ld (GNU Binutils for Debian) 2.34
>
> .text.unlikely reminds me of the sections created when profiling data
> is present.  That's obviously not the case here, so maybe there's
> other ways this section can be created by GCC?  I may have added a
> patch recently for placing this section explicitly, which was a
> prerequisite for Kees' series explicitly placing all sections.
> https://lore.kernel.org/lkml/159896087937.20229.4955362311782724603.tip-bot2@tip-bot2/
> Maybe that can be improved?
>
> IIRC, LLD is able to emit range extension thunks for these cases, but
> BFD does not.

ld.bfd will usually emit veneers for branches that turn out to be out
of range in the final link stage.

Does the following help?

diff --git a/arch/arm/vfp/vfphw.S b/arch/arm/vfp/vfphw.S
index 4fcff9f59947..f1468702fbc9 100644
--- a/arch/arm/vfp/vfphw.S
+++ b/arch/arm/vfp/vfphw.S
@@ -82,6 +82,7 @@ ENTRY(vfp_support_entry)
        ldr     r3, [sp, #S_PSR]        @ Neither lazy restore nor FP exceptions
        and     r3, r3, #MODE_MASK      @ are supported in kernel mode
        teq     r3, #USR_MODE
+THUMB( it      ne                      )
        bne     vfp_kmode_exception     @ Returns through lr

        VFPFMRX r1, FPEXC               @ Is the VFP enabled?