Skip to contentSkip to conversation

arm64: link with -z norelro regardless of CONFIG_RELOCATABLE

PATCH Applied Nick Desaulniers 6 participants 14 messages Oct 16, 2020 – Jan 5, 2021

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 arm64: link with -z norelro regardless of CONFIG_RELOCATABLE

With CONFIG_EXPERT=y, CONFIG_KASAN=y, CONFIG_RANDOMIZE_BASE=n, CONFIG_RELOCATABLE=n, we observe the following failure when trying to link the kernel image with LD=ld.lld:

error: section: .exit.data is not contiguous with other relro sections

ld.lld defaults to -z relro while ld.bfd defaults to -z norelro. This was previously fixed, but only for CONFIG_RELOCATABLE=y.

Fixes commit 3bbd3db86470 ("arm64: relocatable: fix inconsistencies in linker script and options")

---
While upgrading our toolchains for Android, we started seeing the above failure for a particular config that enabled KASAN but disabled KASLR.
This was on a 5.4 stable branch. It looks like
commit dd4bc6076587 ("arm64: warn on incorrect placement of the kernel by the bootloader") made RELOCATABLE=y the default and depend on EXPERT=y. With those two enabled, we can then reproduce the same failure on mainline.

 arch/arm64/Makefile | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
arch/arm64/Makefile +2 -2
index f4717facf31e..674241df91ab 100644
@@ -10,13 +10,13 @@
 #
 # Copyright (C) 1995-2001 by Russell King
 
-LDFLAGS_vmlinux	:=--no-undefined -X
+LDFLAGS_vmlinux	:=--no-undefined -X -z norelro
 
 ifeq ($(CONFIG_RELOCATABLE), y)
 # Pass --no-apply-dynamic-relocs to restore pre-binutils-2.27 behaviour
 # for relative relocs, since this leads to better Image compression
 # with the relocation offsets always being zero.
-LDFLAGS_vmlinux		+= -shared -Bsymbolic -z notext -z norelro \
+LDFLAGS_vmlinux		+= -shared -Bsymbolic -z notext \
 			$(call ld-option, --no-apply-dynamic-relocs)
 endif
 
Signature · 1 line
2.29.0.rc1.297.gfa9743e501-goog
maintainer Applied
#1 lore raw
Quoted 8 lines · Nick Desaulniers

With CONFIG_EXPERT=y, CONFIG_KASAN=y, CONFIG_RANDOMIZE_BASE=n, CONFIG_RELOCATABLE=n, we observe the following failure when trying to link the kernel image with LD=ld.lld:

error: section: .exit.data is not contiguous with other relro sections

ld.lld defaults to -z relro while ld.bfd defaults to -z norelro. This was previously fixed, but only for CONFIG_RELOCATABLE=y.

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

[1/1] arm64: link with -z norelro regardless of CONFIG_RELOCATABLE

      https://git.kernel.org/arm64/c/3b92fa7485eb

Cheers,

Signature · 5 lines
Will

https://fixes.arm64.dev
https://next.arm64.dev
https://will.arm64.dev
#2 lore raw
Quoted 8 lines · Will Deacon
Quoted 6 lines · Nick Desaulniers

With CONFIG_EXPERT=y, CONFIG_KASAN=y, CONFIG_RANDOMIZE_BASE=n, CONFIG_RELOCATABLE=n, we observe the following failure when trying to link the kernel image with LD=ld.lld:

error: section: .exit.data is not contiguous with other relro sections

Quoted 7 lines
Quoted 2 lines

ld.lld defaults to -z relro while ld.bfd defaults to -z norelro. This was previously fixed, but only for CONFIG_RELOCATABLE=y.

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

[1/1] arm64: link with -z norelro regardless of CONFIG_RELOCATABLE

      https://git.kernel.org/arm64/c/3b92fa7485eb

IF we wanted to go further and remove -z norelro, or even enable -z relro for aarch64, then we would have to detangle some KASAN/GCOV generated section discard spaghetti. Fangrui did some more digging and found that .fini_array.* sections were relro (read only after relocations, IIUC), so adding them to EXIT_DATA (include/asm-generic/vmlinux.lds.h) was causing them to get included in .exit.data (arch/arm64/kernel/vmlinux.lds.S) making that relro. There's some history here with commits:

- e41f501d39126 ("vmlinux.lds: account for destructor sections")
- 8dcf86caa1e3da ("vmlinux.lds.h: Fix incomplete .text.exit discards")
- d812db78288d7 ("vmlinux.lds.h: Avoid KASAN and KCSAN's unwanted sections")

It seems the following works for quite a few different configs/toolchains I played with, but the big IF is whether enabling -z relro is worthwhile? If the kernel does respect that mapping, then I assume that's a yes, but I haven't checked yet whether relro is respected within the kernel (grep -rn RELRO turns up nothing interesting). I also haven't checked yet whether all supported versions of GNU ld.bfd support -z relro (guessing not, since a quick test warns: aarch64-linux-gnu-ld: warning: -z relro ignored for v2.34.90.20200706, may be holding it wrong).

(Fangrui also filed https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97507
in regards to GCOV+GCC)

(quoted diff) +0 -0

b/include/asm-generic/vmlinux.lds.h
index cd14444bf600..64578c998e53 100644

include/asm-generic/vmlinux.lds.h +9 -1
@@ -744,7 +744,6 @@
 
 #define EXIT_DATA                                                      \
        *(.exit.data .exit.data.*)                                      \
-       *(.fini_array .fini_array.*)                                    \
        *(.dtors .dtors.*)                                              \
        MEM_DISCARD(exit.data*)                                         \
        MEM_DISCARD(exit.rodata*)
@@ -995,6 +994,7 @@
 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KCSAN)
 # ifdef CONFIG_CONSTRUCTORS
 #  define SANITIZER_DISCARDS                                           \
+       *(.fini_array .fini_array.*)                                    \
        *(.eh_frame)
 # else
 #  define SANITIZER_DISCARDS                                           \
@@ -1005,8 +1005,16 @@
 # define SANITIZER_DISCARDS
 #endif
 
+#if defined(CONFIG_GCOV_KERNEL) && defined(CONFIG_CC_IS_GCC)
+# define GCOV_DISCARDS                                                 \
+       *(.fini_array .fini_array.*)
+#else
+# define GCOV_DISCARDS
+#endif
+
 #define COMMON_DISCARDS
         \
        SANITIZER_DISCARDS                                              \
+       GCOV_DISCARDS                                                   \
        *(.discard)                                                     \
        *(.discard.*)                                                   \
        *(.modinfo)                                                     \
Signature · 2 lines
Thanks,
~Nick Desaulniers
#3 lore raw
Quoted 13 lines · Nick Desaulniers
Quoted 11 lines · Will Deacon
Quoted 8 lines · Nick Desaulniers

With CONFIG_EXPERT=y, CONFIG_KASAN=y, CONFIG_RANDOMIZE_BASE=n, CONFIG_RELOCATABLE=n, we observe the following failure when trying to link the kernel image with LD=ld.lld:

error: section: .exit.data is not contiguous with other relro sections

ld.lld defaults to -z relro while ld.bfd defaults to -z norelro. This was previously fixed, but only for CONFIG_RELOCATABLE=y.

Quoted 8 lines
Quoted 4 lines

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

[1/1] arm64: link with -z norelro regardless of CONFIG_RELOCATABLE

      https://git.kernel.org/arm64/c/3b92fa7485eb

IF we wanted to go further and remove -z norelro, or even enable -z relro for aarch64, then we would have to detangle some KASAN/GCOV generated section discard spaghetti.

Why on earth would we want that?

Quoted 11 lines

Fangrui did some more digging
and found that .fini_array.* sections were relro (read only after relocations, IIUC), so adding them to EXIT_DATA (include/asm-generic/vmlinux.lds.h) was causing them to get included in .exit.data (arch/arm64/kernel/vmlinux.lds.S) making that relro. There's some history here with commits:

- e41f501d39126 ("vmlinux.lds: account for destructor sections")
- 8dcf86caa1e3da ("vmlinux.lds.h: Fix incomplete .text.exit discards")
- d812db78288d7 ("vmlinux.lds.h: Avoid KASAN and KCSAN's unwanted sections")

Quoted 10 lines

It seems the following works for quite a few different configs/toolchains I played with, but the big IF is whether enabling -z relro is worthwhile? If the kernel does respect that mapping, then I assume that's a yes, but I haven't checked yet whether relro is respected within the kernel (grep -rn RELRO turns up nothing interesting). I also haven't checked yet whether all supported versions of GNU ld.bfd support -z relro (guessing not, since a quick test warns: aarch64-linux-gnu-ld: warning: -z relro ignored for v2.34.90.20200706, may be holding it wrong).

RELRO just moves statically initialized const pointers into a separate section so we can place it in a way that allows us to easily map it r/w during load, and switch it over to r/o once the relocations have been applied.

On AArch64, we don't even use -fpic to build the kernel, and load time relocations may appear everywhere in .text, .rodata etc etc, which is absolutely fine given that we apply the relocations way before we finalize the kernel mappings. This means that, in our case, statically initialized const pointers will be mapped r/o already, and we don't need RELRO.

In general, we should ensure that the 'relocatable bare metal' case doesn't get snowed under, as toolchain development is [understandably] very focused on hosted binaries that use shared libraries, where things like CoW footprint, ELF symbol preemption, text relocations and RELRO sections actually matter. For bare metal, it is quite the
opposite: text relocations are fine, there is no CoW so minimizing the footprint of the .so pages that are modified due to relocations is unnecessary, and symbols cannot be preempted either. So many of the shared library tricks actually make things worse for us, because we have to work around them while they have no benefit for us.

I have suggested this before, but perhaps we should have a -mcmodel=kernel (like x86 does) that takes these things into account? As a start, it could imply -cmodel=small (which we rely on today), but with guarantees that the generated code is position independent, but without GOT indirections, and that the resulting object code can be linked with -pie (so that we have access to the load time relocations in the bare metal binary itself). This is something we rely on today, and happens to work in practice, but this could easily break in the future.

Quoted 45 lines

(Fangrui also filed https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97507
in regards to GCOV+GCC)

include/asm-generic/vmlinux.lds.h +9 -1
index cd14444bf600..64578c998e53 100644
@@ -744,7 +744,6 @@
 
 #define EXIT_DATA                                                      \
        *(.exit.data .exit.data.*)                                      \
-       *(.fini_array .fini_array.*)                                    \
        *(.dtors .dtors.*)                                              \
        MEM_DISCARD(exit.data*)                                         \
        MEM_DISCARD(exit.rodata*)
@@ -995,6 +994,7 @@
 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KCSAN)
 # ifdef CONFIG_CONSTRUCTORS
 #  define SANITIZER_DISCARDS                                           \
+       *(.fini_array .fini_array.*)                                    \
        *(.eh_frame)
 # else
 #  define SANITIZER_DISCARDS                                           \
@@ -1005,8 +1005,16 @@
 # define SANITIZER_DISCARDS
 #endif
 
+#if defined(CONFIG_GCOV_KERNEL) && defined(CONFIG_CC_IS_GCC)
+# define GCOV_DISCARDS                                                 \
+       *(.fini_array .fini_array.*)
+#else
+# define GCOV_DISCARDS
+#endif
+
 #define COMMON_DISCARDS
         \
        SANITIZER_DISCARDS                                              \
+       GCOV_DISCARDS                                                   \
        *(.discard)                                                     \
        *(.discard.*)                                                   \
        *(.modinfo)                                                     \

--
Thanks,
~Nick Desaulniers

#4 lore raw
In reply to Will Deacon
Quoted 8 lines · Will Deacon
Quoted 6 lines · Nick Desaulniers

With CONFIG_EXPERT=y, CONFIG_KASAN=y, CONFIG_RANDOMIZE_BASE=n, CONFIG_RELOCATABLE=n, we observe the following failure when trying to link the kernel image with LD=ld.lld:

error: section: .exit.data is not contiguous with other relro sections

Quoted 7 lines
Quoted 2 lines

ld.lld defaults to -z relro while ld.bfd defaults to -z norelro. This was previously fixed, but only for CONFIG_RELOCATABLE=y.

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

[1/1] arm64: link with -z norelro regardless of CONFIG_RELOCATABLE

      https://git.kernel.org/arm64/c/3b92fa7485eb

It looks like this is now producing warnings when linking with BFD.

$ make ...
...
  LD      .tmp_vmlinux.kallsyms1
aarch64-linux-gnu-ld: warning: -z norelro ignored
  KSYMS   .tmp_vmlinux.kallsyms1.S
  AS      .tmp_vmlinux.kallsyms1.S
  LD      .tmp_vmlinux.kallsyms2
aarch64-linux-gnu-ld: warning: -z norelro ignored
  KSYMS   .tmp_vmlinux.kallsyms2.S
  AS      .tmp_vmlinux.kallsyms2.S
  LD      vmlinux
aarch64-linux-gnu-ld: warning: -z norelro ignored

Alan, looking at binutils-gdb commit 5fd104addfddb ("Emit a warning when -z relro is unsupported") mentions targets lacking relro support will produce this warning. I thought aarch64 supports relro though...?
Looks like we're invoking:
+ aarch64-linux-gnu-ld -EL -maarch64elf --no-undefined -X -z norelro -shared -Bsymbolic -z notext --no-apply-dynamic-relocs --fix-cortex-a53-843419 --build-id=sha1 --orphan-handling=warn --strip-debug -o .tmp_vmlinux.kallsyms1 -T ./arch/arm64/kernel/vmlinux.lds --whole-archive arch/arm64/kernel/head.o init/built-in.a usr/built-in.a arch/arm64/built-in.a kernel/built-in.a certs/built-in.a mm/built-in.a fs/built-in.a ipc/built-in.a security/built-in.a crypto/built-in.a block/built-in.a arch/arm64/lib/built-in.a lib/built-in.a arch/arm64/lib/lib.a lib/lib.a drivers/built-in.a sound/built-in.a net/built-in.a virt/built-in.a --no-whole-archive --start-group ./drivers/firmware/efi/libstub/lib.a --end-group
aarch64-linux-gnu-ld: warning: -z norelro ignored

So we set the emulation mode via -maarch64elf, and our preprocessed linker script has OUTPUT_ARCH(aarch64). From that commit, there's a linked mailing list discussion:
https://sourceware.org/legacy-ml/binutils/2017-01/msg00441.html

Is there something more we need to do to our linker script
(https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/kernel/vmlinux.lds.S)
for BFD not to warn when passing -z norelro? It looks like common page size might need to be specified? I tried:

arch/arm64/kernel/vmlinux.lds.S +1 -1
index 1bda604f4c70..ae8cce140fdf 100644
@@ -121,7 +121,7 @@ SECTIONS
                _text = .;
                HEAD_TEXT
        }
-       .text : {                       /* Real text segment            */
+       .text ALIGN (CONSTANT (COMMONPAGESIZE)): {      /* Real text

segment */

and passing -z common-page-size=4096 but neither seemed to do the
trick. (https://docs.adacore.com/live/wave/binutils-stable/html/ld/ld.html#index-COMMONPAGESIZE-553

Worst case, we add -z norelro just for LLD:

arch/arm64/Makefile +5 -1
index 6a87d592bd00..6a6235e1e8a9 100644
@@ -10,7 +10,7 @@
 #
 # Copyright (C) 1995-2001 by Russell King
 
-LDFLAGS_vmlinux        :=--no-undefined -X -z norelro
+LDFLAGS_vmlinux        :=--no-undefined -X
 
 ifeq ($(CONFIG_RELOCATABLE), y)
 # Pass --no-apply-dynamic-relocs to restore pre-binutils-2.27 behaviour
@@ -28,6 +28,10 @@ LDFLAGS_vmlinux      += --fix-cortex-a53-843419
   endif
 endif
 
+ifeq ($(CONFIG_LD_IS_LLD), y)
+LDFLAGS_vmlinux        += -z norelro
+endif
+
 ifeq ($(CONFIG_ARM64_USE_LSE_ATOMICS), y)
   ifneq ($(CONFIG_ARM64_LSE_ATOMICS), y)
 $(warning LSE atomics not supported by binutils)
Signature · 7 lines
Thanks,
~Nick Desaulniers

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
#5 lore raw
Quoted 3 lines · Nick Desaulniers

aarch64-linux-gnu-ld: warning: -z norelro ignored

So we set the emulation mode via -maarch64elf, and our preprocessed

The default linker emulation for an aarch64-linux ld.bfd is -maarch64linux, the default for an aarch64-elf linker is -maarch64elf. They are not equivalent. If you choose -maarch64elf you get an emulation that doesn't support -z relro.

Now I don't know why the kernel uses -maarch64elf so you shouldn't interpret my comment as a recommendation to use -maarch64linux instead. That may have other unwanted effects.

Signature · 2 lines
Alan Modra
Australia Development Lab, IBM
#6 lore raw
Quoted 11 lines · Alan Modra
Quoted 3 lines · Nick Desaulniers

aarch64-linux-gnu-ld: warning: -z norelro ignored

So we set the emulation mode via -maarch64elf, and our preprocessed

The default linker emulation for an aarch64-linux ld.bfd is

(quoted diff) +0 -2
-maarch64linux, the default for an aarch64-elf linker is
-maarch64elf.  They are not equivalent.  If you choose -maarch64elf

you get an emulation that doesn't support -z relro.

Quoted 3 lines

Now I don't know why the kernel uses -maarch64elf so you shouldn't interpret my comment as a recommendation to use -maarch64linux instead. That may have other unwanted effects.

Cool, thanks for the context. The kernel currently has:

arch/arm64/Makefile:
...
ifeq ($(CONFIG_CPU_BIG_ENDIAN), y)
...
# Prefer the baremetal ELF build target, but not all toolchains include

# it so fall back to the standard linux version if needed.
LDFLAGS                += -EB $(call ld-option, -maarch64elfb, -maarch64linuxb)
...
else
...
# Same as above, prefer ELF but fall back to linux target if needed.
LDFLAGS                += -EL $(call ld-option, -maarch64elf, -maarch64linux)
...

Then

$ git log -S maarch64elf arch/arm64/Makefile
provides more context:
1. https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=c931d34ea0853d41349e93f871bd3f17f1c03a6b
2. https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=96f95a17c1cfe65a002e525114d96616e91a8f2d
3. https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=38fc4248677552ce35efc09902fdcb06b61d7ef9

So it seems more that the kernel is relying on whichever emulation targets are supported in local distros, not necessarily the semantic differences between the two. Since the kernel might be linked as either, I'll send the patch described in my last post to add -z norelro just when linking with LLD, since it sounds like it's only possible to specify when -maarch64linux/-maarch64linuxb is used, which is unlikely.

Signature · 2 lines
Thanks,
~Nick Desaulniers
#7 lore raw

PATCH 1/1 arm64: link with -z norelro for LLD or aarch64-elf

With newer GNU binutils, linking with BFD produces warnings for vmlinux:
aarch64-linux-gnu-ld: warning: -z norelro ignored

BFD can produce this warning when the target emulation mode does not support RELRO relocation types, and -z relro or -z norelro is passed.

Alan Modra clarifies:
The default linker emulation for an aarch64-linux ld.bfd is -maarch64linux, the default for an aarch64-elf linker is -maarch64elf. They are not equivalent. If you choose -maarch64elf you get an emulation that doesn't support -z relro.

The ARCH=arm64 kernel prefers -maarch64elf, but may fall back to -maarch64linux based on the toolchain configuration.

LLD will always create RELRO relocation types regardless of target emulation.

To avoid the above warning when linking with BFD, pass -z norelro only when linking with LLD or with -maarch64linux.

Cc Alan Modra <[email protected]>
Cc Ard Biesheuvel <[email protected]>
Cc Fāng-ruì Sòng <[email protected]>
Fixes 3b92fa7485eb ("arm64: link with -z norelro regardless of CONFIG_RELOCATABLE")
---
 arch/arm64/Makefile | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
arch/arm64/Makefile +7 -3
index 6be9b3750250..90309208bb28 100644
@@ -10,7 +10,7 @@
 #
 # Copyright (C) 1995-2001 by Russell King
 
-LDFLAGS_vmlinux	:=--no-undefined -X -z norelro
+LDFLAGS_vmlinux	:=--no-undefined -X
 
 ifeq ($(CONFIG_RELOCATABLE), y)
 # Pass --no-apply-dynamic-relocs to restore pre-binutils-2.27 behaviour
@@ -115,16 +115,20 @@ KBUILD_CPPFLAGS	+= -mbig-endian
 CHECKFLAGS	+= -D__AARCH64EB__
 # Prefer the baremetal ELF build target, but not all toolchains include
 # it so fall back to the standard linux version if needed.
-KBUILD_LDFLAGS	+= -EB $(call ld-option, -maarch64elfb, -maarch64linuxb)
+KBUILD_LDFLAGS	+= -EB $(call ld-option, -maarch64elfb, -maarch64linuxb -z norelro)
 UTS_MACHINE	:= aarch64_be
 else
 KBUILD_CPPFLAGS	+= -mlittle-endian
 CHECKFLAGS	+= -D__AARCH64EL__
 # Same as above, prefer ELF but fall back to linux target if needed.
-KBUILD_LDFLAGS	+= -EL $(call ld-option, -maarch64elf, -maarch64linux)
+KBUILD_LDFLAGS	+= -EL $(call ld-option, -maarch64elf, -maarch64linux -z norelro)
 UTS_MACHINE	:= aarch64
 endif
 
+ifeq ($(CONFIG_LD_IS_LLD), y)
+KBUILD_LDFLAGS	+= -z norelro
+endif
+
 CHECKFLAGS	+= -D__aarch64__
 
 ifeq ($(CONFIG_DYNAMIC_FTRACE_WITH_REGS),y)
Signature · 1 line
2.29.2.684.gfbc64c5ab5-goog
maintainer
#8 lore raw

Re: [PATCH] arm64: link with -z norelro for LLD or aarch64-elf

Quoted 12 lines · Nick Desaulniers

With newer GNU binutils, linking with BFD produces warnings for vmlinux:
aarch64-linux-gnu-ld: warning: -z norelro ignored

BFD can produce this warning when the target emulation mode does not support RELRO relocation types, and -z relro or -z norelro is passed.

Alan Modra clarifies:
The default linker emulation for an aarch64-linux ld.bfd is -maarch64linux, the default for an aarch64-elf linker is -maarch64elf. They are not equivalent. If you choose -maarch64elf you get an emulation that doesn't support -z relro.

Quoted 8 lines

The ARCH=arm64 kernel prefers -maarch64elf, but may fall back to -maarch64linux based on the toolchain configuration.

LLD will always create RELRO relocation types regardless of target emulation.

To avoid the above warning when linking with BFD, pass -z norelro only when linking with LLD or with -maarch64linux.

Given that, prior to 3b92fa7485eb, we used to pass '-z norelro' if CONFIG_RELOCATABLE then was this already broken with the ELF toolchain?

Will

#9 lore raw

Re: [PATCH] arm64: link with -z norelro for LLD or aarch64-elf

Quoted 7 lines · Will Deacon
Quoted 3 lines · Nick Desaulniers

With newer GNU binutils, linking with BFD produces warnings for vmlinux:
aarch64-linux-gnu-ld: warning: -z norelro ignored

Given that, prior to 3b92fa7485eb, we used to pass '-z norelro' if CONFIG_RELOCATABLE then was this already broken with the ELF toolchain?

Yes, though it would have been hard to foresee the change to BFD ~6 months later.

Specifically, binutils-gdb
commit 5fd104addfddb ("Emit a warning when -z relro is unsupported") was committed Fri Jun 19 09:50:20 2020 +0930. The first git tag that describes this commit was binutils-2_35 which was tagged Fri Jul 24 11:05:23 2020 +0100.

I noticed about a month ago that the version of binutils-aarch64-linux-gnu installed on my gLinux workstation had auto updated to version 2.35.1; I was authoring kernel patches for DWARF v5 support, which relied on 2.35 for DWARF v5 assembler support. I suspect Quentin's host was auto updated as well, at which point he noticed and mentioned to me since I had touched -z norelro last.

But if we look at
commit 3bbd3db86470 ("arm64: relocatable: fix inconsistencies in linker script and options")
which was committed at Tue Dec 4 12:48:25 2018 +0000, it was not possible to foresee that binutils-gdb would change to produce such a warning for such an emulation mode.

So I'm not sure whether my patch should either:
- have a fixes tag for just the latest commit that touched anything related to -z norelro, mine, 3b92fa7485eb.
- have an additional fixes tag for 3bbd3db86470 which first introduced -z norelro.
- have no fixes tag

I'll respin a v2 folding in Ard's suggestions. Meanwhile, I've filed:
- https://bugs.llvm.org/show_bug.cgi?id=48549 against LLD
- https://sourceware.org/bugzilla/show_bug.cgi?id=27093 against BFD

Signature · 2 lines
Thanks,
~Nick Desaulniers
#10 lore raw

PATCH v2 1/1 arm64: link with -z norelro for LLD or aarch64-elf

With GNU binutils 2.35+, linking with BFD produces warnings for vmlinux:
aarch64-linux-gnu-ld: warning: -z norelro ignored

BFD can produce this warning when the target emulation mode does not support RELRO program headers, and -z relro or -z norelro is passed.

Alan Modra clarifies:
The default linker emulation for an aarch64-linux ld.bfd is -maarch64linux, the default for an aarch64-elf linker is -maarch64elf. They are not equivalent. If you choose -maarch64elf you get an emulation that doesn't support -z relro.

The ARCH=arm64 kernel prefers -maarch64elf, but may fall back to -maarch64linux based on the toolchain configuration.

LLD will always create RELRO program header regardless of target emulation.

To avoid the above warning when linking with BFD, pass -z norelro only when linking with LLD or with -maarch64linux.

Cc Alan Modra <[email protected]>
Cc Fāng-ruì Sòng <[email protected]>
Fixes 3b92fa7485eb ("arm64: link with -z norelro regardless of CONFIG_RELOCATABLE")

---
Changes V1 -> V2:
* s/relocation types/program headers/
* s/newer GNU binutils/GNU binutils 2.35+/
* Pick up Ard's Ack.

Note: maintainers may want to pick up the following tag:

Fixes 3bbd3db86470 ("arm64: relocatable: fix inconsistencies in linker script and options")

or drop the existing fixes tag (this patch is more so in response to change to BFD to warn than fix a kernel regression, IMO, but I don't care). Either way, it would be good to fix this for the newly minted v5.10.y.

I'll probably be offline for the next two weeks for the holidays, so no promises on quick replies. Happy holidays+new year!

 arch/arm64/Makefile | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
arch/arm64/Makefile +7 -3
index 6be9b3750250..90309208bb28 100644
@@ -10,7 +10,7 @@
 #
 # Copyright (C) 1995-2001 by Russell King
 
-LDFLAGS_vmlinux	:=--no-undefined -X -z norelro
+LDFLAGS_vmlinux	:=--no-undefined -X
 
 ifeq ($(CONFIG_RELOCATABLE), y)
 # Pass --no-apply-dynamic-relocs to restore pre-binutils-2.27 behaviour
@@ -115,16 +115,20 @@ KBUILD_CPPFLAGS	+= -mbig-endian
 CHECKFLAGS	+= -D__AARCH64EB__
 # Prefer the baremetal ELF build target, but not all toolchains include
 # it so fall back to the standard linux version if needed.
-KBUILD_LDFLAGS	+= -EB $(call ld-option, -maarch64elfb, -maarch64linuxb)
+KBUILD_LDFLAGS	+= -EB $(call ld-option, -maarch64elfb, -maarch64linuxb -z norelro)
 UTS_MACHINE	:= aarch64_be
 else
 KBUILD_CPPFLAGS	+= -mlittle-endian
 CHECKFLAGS	+= -D__AARCH64EL__
 # Same as above, prefer ELF but fall back to linux target if needed.
-KBUILD_LDFLAGS	+= -EL $(call ld-option, -maarch64elf, -maarch64linux)
+KBUILD_LDFLAGS	+= -EL $(call ld-option, -maarch64elf, -maarch64linux -z norelro)
 UTS_MACHINE	:= aarch64
 endif
 
+ifeq ($(CONFIG_LD_IS_LLD), y)
+KBUILD_LDFLAGS	+= -z norelro
+endif
+
 CHECKFLAGS	+= -D__aarch64__
 
 ifeq ($(CONFIG_DYNAMIC_FTRACE_WITH_REGS),y)
Signature · 1 line
2.29.2.684.gfbc64c5ab5-goog
Reviewed-by on v2 1/1
#11 lore raw

Re: [PATCH v2] arm64: link with -z norelro for LLD or aarch64-elf

Quoted 21 lines · Nick Desaulniers' via Clang Built Linux

With GNU binutils 2.35+, linking with BFD produces warnings for vmlinux:
aarch64-linux-gnu-ld: warning: -z norelro ignored

BFD can produce this warning when the target emulation mode does not support RELRO program headers, and -z relro or -z norelro is passed.

Alan Modra clarifies:
The default linker emulation for an aarch64-linux ld.bfd is -maarch64linux, the default for an aarch64-elf linker is -maarch64elf. They are not equivalent. If you choose -maarch64elf you get an emulation that doesn't support -z relro.

The ARCH=arm64 kernel prefers -maarch64elf, but may fall back to -maarch64linux based on the toolchain configuration.

LLD will always create RELRO program header regardless of target emulation.

To avoid the above warning when linking with BFD, pass -z norelro only when linking with LLD or with -maarch64linux.

Quoted 7 lines
Cc Alan Modra <[email protected]>
Cc Fāng-ruì Sòng <[email protected]>
Fixes 3b92fa7485eb ("arm64: link with -z norelro regardless of CONFIG_RELOCATABLE")
Quoted 6 lines

---
Changes V1 -> V2:
* s/relocation types/program headers/
* s/newer GNU binutils/GNU binutils 2.35+/
* Pick up Ard's Ack.

Quoted 8 lines

Note: maintainers may want to pick up the following tag:

Fixes 3bbd3db86470 ("arm64: relocatable: fix inconsistencies in linker script and options")

or drop the existing fixes tag (this patch is more so in response to change to BFD to warn than fix a kernel regression, IMO, but I don't care). Either way, it would be good to fix this for the newly minted v5.10.y.

Should probably have

then.

Quoted 46 lines

I'll probably be offline for the next two weeks for the holidays, so no promises on quick replies. Happy holidays+new year!

 arch/arm64/Makefile | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
arch/arm64/Makefile +7 -3
index 6be9b3750250..90309208bb28 100644
@@ -10,7 +10,7 @@
 #
 # Copyright (C) 1995-2001 by Russell King
 
-LDFLAGS_vmlinux	:=--no-undefined -X -z norelro
+LDFLAGS_vmlinux	:=--no-undefined -X
 
 ifeq ($(CONFIG_RELOCATABLE), y)
 # Pass --no-apply-dynamic-relocs to restore pre-binutils-2.27 behaviour
@@ -115,16 +115,20 @@ KBUILD_CPPFLAGS	+= -mbig-endian
 CHECKFLAGS	+= -D__AARCH64EB__
 # Prefer the baremetal ELF build target, but not all toolchains include
 # it so fall back to the standard linux version if needed.
-KBUILD_LDFLAGS	+= -EB $(call ld-option, -maarch64elfb, -maarch64linuxb)
+KBUILD_LDFLAGS	+= -EB $(call ld-option, -maarch64elfb, -maarch64linuxb -z norelro)
 UTS_MACHINE	:= aarch64_be
 else
 KBUILD_CPPFLAGS	+= -mlittle-endian
 CHECKFLAGS	+= -D__AARCH64EL__
 # Same as above, prefer ELF but fall back to linux target if needed.
-KBUILD_LDFLAGS	+= -EL $(call ld-option, -maarch64elf, -maarch64linux)
+KBUILD_LDFLAGS	+= -EL $(call ld-option, -maarch64elf, -maarch64linux -z norelro)
 UTS_MACHINE	:= aarch64
 endif
 
+ifeq ($(CONFIG_LD_IS_LLD), y)
+KBUILD_LDFLAGS	+= -z norelro
+endif
+
 CHECKFLAGS	+= -D__aarch64__
 
 ifeq ($(CONFIG_DYNAMIC_FTRACE_WITH_REGS),y)
Signature · 1 line
2.29.2.684.gfbc64c5ab5-goog
maintainer Applied
#12 lore raw

Re: [PATCH v2] arm64: link with -z norelro for LLD or aarch64-elf

In reply to Nick Desaulniers
Quoted 6 lines · Nick Desaulniers

With GNU binutils 2.35+, linking with BFD produces warnings for vmlinux:
aarch64-linux-gnu-ld: warning: -z norelro ignored

BFD can produce this warning when the target emulation mode does not support RELRO program headers, and -z relro or -z norelro is passed.

Quoted 7 lines

Alan Modra clarifies:
The default linker emulation for an aarch64-linux ld.bfd is -maarch64linux, the default for an aarch64-elf linker is -maarch64elf. They are not equivalent. If you choose -maarch64elf you get an emulation that doesn't support -z relro.

[...]

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

[1/1] arm64: link with -z norelro for LLD or aarch64-elf

      https://git.kernel.org/arm64/c/311bea3cb9ee

Also added the second Fixes tag with a cc stable.

Signature · 1 line
Catalin
Acked-by on v1 1/1
#13 lore raw

Re: [PATCH] arm64: link with -z norelro for LLD or aarch64-elf

In reply to Nick Desaulniers
Quoted 7 lines · Nick Desaulniers

With newer GNU binutils, linking with BFD produces warnings for vmlinux:
aarch64-linux-gnu-ld: warning: -z norelro ignored

BFD can produce this warning when the target emulation mode does not support RELRO relocation types, and -z relro or -z norelro is passed.

RELRO is not a relocation type, it is a type of program header which we might simply ignore, if it weren't for the fact that it can only be emitted if the layout of the sections adheres to certain rules (and ours doesn't), and we get an error otherwise.

It amounts to implicit __ro_after_init annotations for statically initialized const pointers, but given that we don't compile with -fpie, those const pointers reside in .rodata already, so RELRO adds no value for us.

Quoted 6 lines

Alan Modra clarifies:
The default linker emulation for an aarch64-linux ld.bfd is -maarch64linux, the default for an aarch64-elf linker is -maarch64elf. They are not equivalent. If you choose -maarch64elf you get an emulation that doesn't support -z relro.

Quoted 6 lines

The ARCH=arm64 kernel prefers -maarch64elf, but may fall back to -maarch64linux based on the toolchain configuration.

LLD will always create RELRO relocation types regardless of target emulation.

RELRO program header

Quoted 10 lines

To avoid the above warning when linking with BFD, pass -z norelro only when linking with LLD or with -maarch64linux.

Cc Alan Modra <[email protected]>
Cc Ard Biesheuvel <[email protected]>
Cc Fāng-ruì Sòng <[email protected]>
Fixes 3b92fa7485eb ("arm64: link with -z norelro regardless of CONFIG_RELOCATABLE")

With mentions of 'RELRO relocation types' fixed:

Quoted 43 lines
---
 arch/arm64/Makefile | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
arch/arm64/Makefile +7 -4
index 6be9b3750250..90309208bb28 100644
@@ -10,7 +10,7 @@
 #
 # Copyright (C) 1995-2001 by Russell King
 
-LDFLAGS_vmlinux        :=--no-undefined -X -z norelro
+LDFLAGS_vmlinux        :=--no-undefined -X
 
 ifeq ($(CONFIG_RELOCATABLE), y)
 # Pass --no-apply-dynamic-relocs to restore pre-binutils-2.27 behaviour
@@ -115,16 +115,20 @@ KBUILD_CPPFLAGS   += -mbig-endian
 CHECKFLAGS     += -D__AARCH64EB__
 # Prefer the baremetal ELF build target, but not all toolchains include
 # it so fall back to the standard linux version if needed.
-KBUILD_LDFLAGS += -EB $(call ld-option, -maarch64elfb, -maarch64linuxb)
+KBUILD_LDFLAGS += -EB $(call ld-option, -maarch64elfb, -maarch64linuxb -z norelro)
 UTS_MACHINE    := aarch64_be
 else
 KBUILD_CPPFLAGS        += -mlittle-endian
 CHECKFLAGS     += -D__AARCH64EL__
 # Same as above, prefer ELF but fall back to linux target if needed.
-KBUILD_LDFLAGS += -EL $(call ld-option, -maarch64elf, -maarch64linux)
+KBUILD_LDFLAGS += -EL $(call ld-option, -maarch64elf, -maarch64linux -z norelro)
 UTS_MACHINE    := aarch64
 endif
 
+ifeq ($(CONFIG_LD_IS_LLD), y)
+KBUILD_LDFLAGS += -z norelro
+endif
+
 CHECKFLAGS     += -D__aarch64__
 
 ifeq ($(CONFIG_DYNAMIC_FTRACE_WITH_REGS),y)
--

2.29.2.684.gfbc64c5ab5-goog