[PATCH] Makefile: fix GDB warning with CONFIG

Nick Desaulniers

unread,

Mar 18, 2021, 8:07:17 PM3/18/21

to Masahiro Yamada, clang-bu...@googlegroups.com, Fangrui Song, Elliot Berman, Sami Tolvanen, Nick Desaulniers, Peter Collingbourne, Michal Marek, linux-...@vger.kernel.org, linux-...@vger.kernel.org

GDB produces the following warning when debugging kernels built with
CONFIG_RELR:

BFD: /android0/linux-next/vmlinux: unknown type [0x13] section `.relr.dyn'

when loading a kernel built with CONFIG_RELR into GDB. It can also
prevent debugging symbols using such relocations.

Peter sugguests:
[That flag] means that lld will use dynamic tags and section type
numbers in the OS-specific range rather than the generic range. The
kernel itself doesn't care about these numbers; it determines the
location of the RELR section using symbols defined by a linker script.

Link: https://github.com/ClangBuiltLinux/linux/issues/1057
Suggested-by: Peter Collingbourne <p...@google.com>
Signed-off-by: Nick Desaulniers <ndesau...@google.com>
---
Makefile | 2 +-
scripts/tools-support-relr.sh | 3 ++-
2 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/Makefile b/Makefile
index 5160ff8903c1..47741cb60995 100644
--- a/Makefile
+++ b/Makefile
@@ -1088,7 +1088,7 @@ LDFLAGS_vmlinux += $(call ld-option, -X,)
endif

ifeq ($(CONFIG_RELR),y)
-LDFLAGS_vmlinux += --pack-dyn-relocs=relr
+LDFLAGS_vmlinux += --pack-dyn-relocs=relr --use-android-relr-tags
endif

# We never want expected sections to be placed heuristically by the
diff --git a/scripts/tools-support-relr.sh b/scripts/tools-support-relr.sh
index 45e8aa360b45..cb55878bd5b8 100755
--- a/scripts/tools-support-relr.sh
+++ b/scripts/tools-support-relr.sh
@@ -7,7 +7,8 @@ trap "rm -f $tmp_file.o $tmp_file $tmp_file.bin" EXIT
cat << "END" | $CC -c -x c - -o $tmp_file.o >/dev/null 2>&1
void *p = &p;
END
-$LD $tmp_file.o -shared -Bsymbolic --pack-dyn-relocs=relr -o $tmp_file
+$LD $tmp_file.o -shared -Bsymbolic --pack-dyn-relocs=relr \
+ --use-android-relr-tags -o $tmp_file

# Despite printing an error message, GNU nm still exits with exit code 0 if it
# sees a relr section. So we need to check that nothing is printed to stderr.
--
2.31.0.rc2.261.g7f71774620-goog

Nathan Chancellor

unread,

Mar 23, 2021, 3:05:39 PM3/23/21

to Nick Desaulniers, Masahiro Yamada, clang-bu...@googlegroups.com, Fangrui Song, Elliot Berman, Sami Tolvanen, Peter Collingbourne, Michal Marek, linux-...@vger.kernel.org, linux-...@vger.kernel.org

On Thu, Mar 18, 2021 at 05:07:06PM -0700, Nick Desaulniers wrote:
> GDB produces the following warning when debugging kernels built with
> CONFIG_RELR:
>
> BFD: /android0/linux-next/vmlinux: unknown type [0x13] section `.relr.dyn'
>
> when loading a kernel built with CONFIG_RELR into GDB. It can also
> prevent debugging symbols using such relocations.
>
> Peter sugguests:
> [That flag] means that lld will use dynamic tags and section type
> numbers in the OS-specific range rather than the generic range. The
> kernel itself doesn't care about these numbers; it determines the
> location of the RELR section using symbols defined by a linker script.
>
> Link: https://github.com/ClangBuiltLinux/linux/issues/1057
> Suggested-by: Peter Collingbourne <p...@google.com>
> Signed-off-by: Nick Desaulniers <ndesau...@google.com>

Reviewed-by: Nathan Chancellor <nat...@kernel.org>

Lee Jones

unread,

May 21, 2021, 5:36:52 AM5/21/21

to Nathan Chancellor, Nick Desaulniers, Masahiro Yamada, clang-built-linux, Fangrui Song, Elliot Berman, Sami Tolvanen, Peter Collingbourne, Michal Marek, linux-kbuild, open list

On Tue, 23 Mar 2021 at 19:06, Nathan Chancellor <nat...@kernel.org> wrote:

On Thu, Mar 18, 2021 at 05:07:06PM -0700, Nick Desaulniers wrote:
> GDB produces the following warning when debugging kernels built with
> CONFIG_RELR:
>
> BFD: /android0/linux-next/vmlinux: unknown type [0x13] section `.relr.dyn'
>
> when loading a kernel built with CONFIG_RELR into GDB. It can also
> prevent debugging symbols using such relocations.
>
> Peter sugguests:
> [That flag] means that lld will use dynamic tags and section type
> numbers in the OS-specific range rather than the generic range. The
> kernel itself doesn't care about these numbers; it determines the
> location of the RELR section using symbols defined by a linker script.
>
> Link: https://github.com/ClangBuiltLinux/linux/issues/1057
> Suggested-by: Peter Collingbourne <p...@google.com>
> Signed-off-by: Nick Desaulniers <ndesau...@google.com>

Reviewed-by: Nathan Chancellor <nat...@kernel.org>

Masahiro,

Would you mind sharing your plans for this reviewed patch please?

--

Lee Jones [李琼斯]

Linaro Services Senior Technical Lead

Linaro.org │ Open source software for ARM SoCs

Follow Linaro: Facebook | Twitter | Blog

Masahiro Yamada

unread,

May 21, 2021, 9:07:16 PM5/21/21

to Lee Jones, Nathan Chancellor, Nick Desaulniers, clang-built-linux, Fangrui Song, Elliot Berman, Sami Tolvanen, Peter Collingbourne, Michal Marek, linux-kbuild, open list

On Fri, May 21, 2021 at 6:36 PM Lee Jones <lee....@linaro.org> wrote:
>
> On Tue, 23 Mar 2021 at 19:06, Nathan Chancellor <nat...@kernel.org> wrote:
>>
>> On Thu, Mar 18, 2021 at 05:07:06PM -0700, Nick Desaulniers wrote:
>> > GDB produces the following warning when debugging kernels built with
>> > CONFIG_RELR:
>> >
>> > BFD: /android0/linux-next/vmlinux: unknown type [0x13] section `.relr.dyn'
>> >
>> > when loading a kernel built with CONFIG_RELR into GDB. It can also
>> > prevent debugging symbols using such relocations.
>> >
>> > Peter sugguests:
>> > [That flag] means that lld will use dynamic tags and section type
>> > numbers in the OS-specific range rather than the generic range. The
>> > kernel itself doesn't care about these numbers; it determines the
>> > location of the RELR section using symbols defined by a linker script.
>> >
>> > Link: https://github.com/ClangBuiltLinux/linux/issues/1057
>> > Suggested-by: Peter Collingbourne <p...@google.com>
>> > Signed-off-by: Nick Desaulniers <ndesau...@google.com>
>>
>> Reviewed-by: Nathan Chancellor <nat...@kernel.org>
>
>
> Masahiro,
>
> Would you mind sharing your plans for this reviewed patch please?
>

Do you want me to pick up this?

Or, do you think it should be done by the committer of
5cf896fb6be3effd9aea455b22213e27be8bdb1d ?

--
Best Regards
Masahiro Yamada

Nick Desaulniers

unread,

May 21, 2021, 9:26:30 PM5/21/21

to Will Deacon, Catalin Marinas, linux-ar...@lists.infradead.org, Lee Jones, Masahiro Yamada, clang-built-linux, Fangrui Song, Elliot Berman, Sami Tolvanen, Peter Collingbourne, Michal Marek, linux-...@vger.kernel.org, linux-...@vger.kernel.org, Nick Desaulniers, Nathan Chancellor

GDB produces the following warning when debugging kernels built with
CONFIG_RELR:

BFD: /android0/linux-next/vmlinux: unknown type [0x13] section `.relr.dyn'

when loading a kernel built with CONFIG_RELR into GDB. It can also
prevent debugging symbols using such relocations.

Peter sugguests:
[That flag] means that lld will use dynamic tags and section type
numbers in the OS-specific range rather than the generic range. The
kernel itself doesn't care about these numbers; it determines the
location of the RELR section using symbols defined by a linker script.

Link: https://github.com/ClangBuiltLinux/linux/issues/1057
Suggested-by: Peter Collingbourne <p...@google.com>

Reviewed-by: Nathan Chancellor <nat...@kernel.org>

Signed-off-by: Nick Desaulniers <ndesau...@google.com>
---

Changes V1 -> V2:
* rebase
* pick up Nathan's reviewed by tag.

Makefile | 2 +-
scripts/tools-support-relr.sh | 3 ++-
2 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/Makefile b/Makefile

index 0ed7e061c8e9..2dbb56f831d4 100644
--- a/Makefile
+++ b/Makefile
@@ -1031,7 +1031,7 @@ LDFLAGS_vmlinux += $(call ld-option, -X,)

endif

ifeq ($(CONFIG_RELR),y)
-LDFLAGS_vmlinux += --pack-dyn-relocs=relr
+LDFLAGS_vmlinux += --pack-dyn-relocs=relr --use-android-relr-tags
endif

# We never want expected sections to be placed heuristically by the
diff --git a/scripts/tools-support-relr.sh b/scripts/tools-support-relr.sh
index 45e8aa360b45..cb55878bd5b8 100755
--- a/scripts/tools-support-relr.sh
+++ b/scripts/tools-support-relr.sh
@@ -7,7 +7,8 @@ trap "rm -f $tmp_file.o $tmp_file $tmp_file.bin" EXIT
cat << "END" | $CC -c -x c - -o $tmp_file.o >/dev/null 2>&1
void *p = &p;
END
-$LD $tmp_file.o -shared -Bsymbolic --pack-dyn-relocs=relr -o $tmp_file
+$LD $tmp_file.o -shared -Bsymbolic --pack-dyn-relocs=relr \
+ --use-android-relr-tags -o $tmp_file

# Despite printing an error message, GNU nm still exits with exit code 0 if it
# sees a relr section. So we need to check that nothing is printed to stderr.

base-commit: 45af60e7ced07ae3def41368c3d260dbf496fbce
--
2.31.1.818.g46aad6cb9e-goog

Lee Jones

unread,

May 24, 2021, 4:03:13 AM5/24/21

to Masahiro Yamada, Nathan Chancellor, Nick Desaulniers, clang-built-linux, Fangrui Song, Elliot Berman, Sami Tolvanen, Peter Collingbourne, Michal Marek, linux-kbuild, open list

Not sure it matters a whole bunch TBH. If you have the bandwidth and
are willing to merge it, please feel free to do so.

--
Lee Jones [李琼斯]
Senior Technical Lead - Developer Services
Linaro.org │ Open source software for Arm SoCs

Will Deacon

unread,

May 26, 2021, 1:09:11 PM5/26/21

to Nick Desaulniers, Catalin Marinas, linux-ar...@lists.infradead.org, Lee Jones, Masahiro Yamada, clang-built-linux, Fangrui Song, Elliot Berman, Sami Tolvanen, Peter Collingbourne, Michal Marek, linux-...@vger.kernel.org, linux-...@vger.kernel.org, Nathan Chancellor

On Fri, May 21, 2021 at 06:26:24PM -0700, Nick Desaulniers wrote:
> GDB produces the following warning when debugging kernels built with
> CONFIG_RELR:
>
> BFD: /android0/linux-next/vmlinux: unknown type [0x13] section `.relr.dyn'
>
> when loading a kernel built with CONFIG_RELR into GDB. It can also
> prevent debugging symbols using such relocations.
>
> Peter sugguests:
> [That flag] means that lld will use dynamic tags and section type
> numbers in the OS-specific range rather than the generic range. The
> kernel itself doesn't care about these numbers; it determines the
> location of the RELR section using symbols defined by a linker script.
>
> Link: https://github.com/ClangBuiltLinux/linux/issues/1057
> Suggested-by: Peter Collingbourne <p...@google.com>
> Reviewed-by: Nathan Chancellor <nat...@kernel.org>
> Signed-off-by: Nick Desaulniers <ndesau...@google.com>
> ---
> Changes V1 -> V2:
> * rebase
> * pick up Nathan's reviewed by tag.
>
> Makefile | 2 +-
> scripts/tools-support-relr.sh | 3 ++-
> 2 files changed, 3 insertions(+), 2 deletions(-)

Does lld support RELR relocations for any architectures other than arm64? If
so, is the "--use-android-relr-tags" option supported on all of those as
well?

Will

Nick Desaulniers

unread,

Jun 4, 2021, 2:55:18 PM6/4/21

to Will Deacon, Catalin Marinas, Linux ARM, Lee Jones, Masahiro Yamada, clang-built-linux, Fangrui Song, Elliot Berman, Sami Tolvanen, Peter Collingbourne, Michal Marek, Linux Kbuild mailing list, LKML, Nathan Chancellor

Yes; from what I can tell it's not an architecture specific relocation
type. Combing through LLVM's sources, it seems Fuchsia sets it always
(at least when using lld) and I'm pretty sure they support x86. At
least I don't get any errors out of LLD when building with
--pack-dyn-relocs=relr on x86.

I can force on RELR for x86 kernel builds with:
```
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 0045e1b44190..513272c77827 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -2117,6 +2117,7 @@ config PHYSICAL_START

config RELOCATABLE
bool "Build a relocatable kernel"
+ select ARCH_HAS_RELR
default y
help
This builds a kernel image that retains relocation information
```
That builds (it won't boot because we don't have the machinery in the
kernel to self relocate that type, yet).

> so, is the "--use-android-relr-tags" option supported on all of those as
> well?

I believe so; no issues building with this patch and with the above
diff applied on x86. All that flag does is change the elf section type
from SHT_RELR to SHT_ANDROID_RELR.

pcc@ can correct me if I'm wrong.
--
Thanks,
~Nick Desaulniers

Will Deacon

unread,

Jun 8, 2021, 9:22:07 AM6/8/21

to Nick Desaulniers, catalin...@arm.com, kerne...@android.com, Will Deacon, Lee Jones, Peter Collingbourne, Michal Marek, Nathan Chancellor, linux-...@vger.kernel.org, clang-built-linux, linux-ar...@lists.infradead.org, Elliot Berman, Masahiro Yamada, Sami Tolvanen, Fangrui Song, linux-...@vger.kernel.org

On Fri, 21 May 2021 18:26:24 -0700, Nick Desaulniers wrote:
> GDB produces the following warning when debugging kernels built with
> CONFIG_RELR:
>
> BFD: /android0/linux-next/vmlinux: unknown type [0x13] section `.relr.dyn'
>
> when loading a kernel built with CONFIG_RELR into GDB. It can also
> prevent debugging symbols using such relocations.
>

> [...]

Applied to arm64 (for-next/build), thanks!

[1/1] Makefile: fix GDB warning with CONFIG_RELR
https://git.kernel.org/arm64/c/27f2a4db76e8

Cheers,
--
Will

https://fixes.arm64.dev
https://next.arm64.dev
https://will.arm64.dev

Reply all

Reply to author

Forward

[PATCH] Makefile: fix GDB warning with CONFIG_RELR

Nick Desaulniers

Nathan Chancellor

Lee Jones

Masahiro Yamada

Nick Desaulniers

Lee Jones

Will Deacon

Nick Desaulniers

Will Deacon