Re: [PATCH v4 05/17] ctype: Work around Clang -mbranch-protection=none bug

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Mon, Jun 29, 2020 at 8:06 AM Kees Cook <keescook@xxxxxxxxxxxx> wrote:
>
> On Mon, Jun 29, 2020 at 10:15:47AM +0200, Ard Biesheuvel wrote:
> > On Mon, 29 Jun 2020 at 08:18, Kees Cook <keescook@xxxxxxxxxxxx> wrote:
> > >
> > > In preparation for building efi/libstub with -mbranch-protection=none
> > > (EFI does not support branch protection features[1]), add no-op code
> > > to work around a Clang bug that emits an unwanted .note.gnu.property
> > > section for object files without code[2].
> > >
> > > [1] https://lore.kernel.org/lkml/CAMj1kXHck12juGi=E=P4hWP_8vQhQ+-x3vBMc3TGeRWdQ-XkxQ@xxxxxxxxxxxxxx
> > > [2] https://bugs.llvm.org/show_bug.cgi?id=46480
> > >
> > > Cc: Ard Biesheuvel <ardb@xxxxxxxxxx>
> > > Cc: Will Deacon <will@xxxxxxxxxx>
> > > Cc: Dave Martin <Dave.Martin@xxxxxxx>
> > > Cc: clang-built-linux@xxxxxxxxxxxxxxxx
> > > Signed-off-by: Kees Cook <keescook@xxxxxxxxxxxx>
> > > ---
> > >  lib/ctype.c | 10 ++++++++++
> > >  1 file changed, 10 insertions(+)
> > >
> > > diff --git a/lib/ctype.c b/lib/ctype.c
> > > index c819fe269eb2..21245ed57d90 100644
> > > --- a/lib/ctype.c
> > > +++ b/lib/ctype.c
> > > @@ -36,3 +36,13 @@ _L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,_L,     /* 224-239 */
> > >  _L,_L,_L,_L,_L,_L,_L,_P,_L,_L,_L,_L,_L,_L,_L,_L};      /* 240-255 */
> > >
> > >  EXPORT_SYMBOL(_ctype);
> > > +
> > > +/*
> > > + * Clang will generate .note.gnu.property sections for object files
> > > + * without code, even in the presence of -mbranch-protection=none.
> > > + * To work around this, define an unused static function.
> > > + * https://bugs.llvm.org/show_bug.cgi?id=46480
> > > + */
> > > +#ifdef CONFIG_CC_IS_CLANG
> > > +void __maybe_unused __clang_needs_code_here(void) { }
> > > +#endif
> > > --
> > > 2.25.1
> > >
> >
> > I take it we don't need this horrible hack if we build the EFI stub
> > with branch protections and filter out the .note.gnu.property section
> > explicitly?
> >
> > Sorry to backpedal, but that is probably a better approach after all,
> > given that the instructions don't hurt, and we will hopefully be able
> > to arm them once UEFI (as well as PE/COFF) gets around to describing
> > this in a way that both the firmware and the OS can consume.
>
> How does this look?
>
>
> commit 051ef0b75a386c3fe2f216d16246468147a48c5b
> Author: Kees Cook <keescook@xxxxxxxxxxxx>
> Date:   Tue Jun 23 18:02:56 2020 -0700
>
>     efi/libstub: Disable -mbranch-protection
>
>     In preparation for adding --orphan-handling=warn to more architectures,
>     disable -mbranch-protection, as EFI does not yet support it[1].  This was
>     noticed due to it producing unwanted .note.gnu.property sections (prefixed
>     with .init due to the objcopy build step).
>
>     However, we must also work around a bug in Clang where the section is
>     still emitted for code-less object files[2], so also remove the section
>     during the objcopy.
>
>     [1] https://lore.kernel.org/lkml/CAMj1kXHck12juGi=E=P4hWP_8vQhQ+-x3vBMc3TGeRWdQ-XkxQ@xxxxxxxxxxxxxx
>     [2] https://bugs.llvm.org/show_bug.cgi?id=46480
>
>     Cc: Ard Biesheuvel <ardb@xxxxxxxxxx>
>     Cc: Arvind Sankar <nivedita@xxxxxxxxxxxx>
>     Cc: Atish Patra <atish.patra@xxxxxxx>
>     Cc: linux-efi@xxxxxxxxxxxxxxx
>     Signed-off-by: Kees Cook <keescook@xxxxxxxxxxxx>
>
> diff --git a/drivers/firmware/efi/libstub/Makefile b/drivers/firmware/efi/libstub/Makefile
> index 75daaf20374e..f9f1922f8f28 100644
> --- a/drivers/firmware/efi/libstub/Makefile
> +++ b/drivers/firmware/efi/libstub/Makefile
> @@ -18,7 +18,8 @@ cflags-$(CONFIG_X86)          += -m$(BITS) -D__KERNEL__ \
>  # arm64 uses the full KBUILD_CFLAGS so it's necessary to explicitly
>  # disable the stackleak plugin
>  cflags-$(CONFIG_ARM64)         := $(subst $(CC_FLAGS_FTRACE),,$(KBUILD_CFLAGS)) \
> -                                  -fpie $(DISABLE_STACKLEAK_PLUGIN)
> +                                  -fpie $(DISABLE_STACKLEAK_PLUGIN) \
> +                                  $(call cc-option,-mbranch-protection=none)
>  cflags-$(CONFIG_ARM)           := $(subst $(CC_FLAGS_FTRACE),,$(KBUILD_CFLAGS)) \
>                                    -fno-builtin -fpic \
>                                    $(call cc-option,-mno-single-pic-base)
> @@ -66,6 +67,12 @@ lib-$(CONFIG_X86)            += x86-stub.o
>  CFLAGS_arm32-stub.o            := -DTEXT_OFFSET=$(TEXT_OFFSET)
>  CFLAGS_arm64-stub.o            := -DTEXT_OFFSET=$(TEXT_OFFSET)
>
> +# Even when -mbranch-protection=none is set, Clang will generate a
> +# .note.gnu.property for code-less object files (like lib/ctype.c),
> +# so work around this by explicitly removing the unwanted section.
> +# https://bugs.llvm.org/show_bug.cgi?id=46480
> +STUBCOPY_FLAGS-y               += --remove-section=.note.gnu.property

This is definitely better than the empty function.  Though a patch is
posted for fixing this in LLVM. Assuming that lands before this, we
might not actually need this workaround?

arch/arm64/Kconfig
1625 config ARM64_BTI_KERNEL
...
1633   # https://reviews.llvm.org/rGb8ae3fdfa579dbf366b1bb1cbfdbf8c51db7fa55
1634   depends on !CC_IS_CLANG || CLANG_VERSION >= 100001

So if Daniel's patch lands AND is backported into the clang 10.0.1
release, then we might not need to carry this workaround?  Either way,

Reviewed-by: Nick Desaulniers <ndesaulniers@xxxxxxxxxx>

> +
>  #
>  # For x86, bootloaders like systemd-boot or grub-efi do not zero-initialize the
>  # .bss section, so the .bss section of the EFI stub needs to be included in the
>
> --
> Kees Cook

-- 
Thanks,
~Nick Desaulniers



[Index of Archives]     [Linux Kernel]     [Kernel Newbies]     [x86 Platform Driver]     [Netdev]     [Linux Wireless]     [Netfilter]     [Bugtraq]     [Linux Filesystems]     [Yosemite Discussion]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Samba]     [Device Mapper]

  Powered by Linux