Skip to contentSkip to conversation

x86: fix clang integrated assembler build

PATCH In review Arnd Bergmann 2 participants 3 messages May 27, 2020 – Jun 1, 2020 1 message not in the archive

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 x86: fix clang integrated assembler build

clang and gas seem to interpret the symbols in memmove_64.S and memset_64.S differently, such that clang does not make them 'weak' as expected, which leads to a linker error, with both ld.bfd and ld.lld:

ld.lld: error: duplicate symbol: memmove

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

defined at common.c
kasan/common.o:(memmove) in archive mm/built-in.a
defined at memmove.o:(__memmove) in archive arch/arm64/lib/lib.a

ld.lld: error: duplicate symbol: memset

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

defined at common.c
kasan/common.o:(memset) in archive mm/built-in.a
defined at memset.o:(__memset) in archive arch/arm64/lib/lib.a

Copy the exact way these are written in memcpy_64.S, which does not have the same problem.

I don't know why this makes a difference, and it would be good to have someone with a better understanding of assembler internals review it.

It might be either a bug in the kernel or a bug in the assembler, no idea which one. My patch makes it work with all versions of clang and gcc, which is probably helpful even if it's a workaround for a clang bug.

---
 arch/x86/lib/memmove_64.S | 4 ++--
 arch/x86/lib/memset_64.S  | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)
arch/x86/lib/memmove_64.S +2 -2
index 7ff00ea64e4f..dcca01434be8 100644
@@ -26,8 +26,8 @@
  */
 .weak memmove
 
-SYM_FUNC_START_ALIAS(memmove)
-SYM_FUNC_START(__memmove)
+SYM_FUNC_START_ALIAS(__memmove)
+SYM_FUNC_START_LOCAL(memmove)
 
 	mov %rdi, %rax
 
arch/x86/lib/memset_64.S +2 -2
index 9ff15ee404a4..a97f2ea4e0b2 100644
@@ -19,8 +19,8 @@
  *
  * rax   original destination
  */
-SYM_FUNC_START_ALIAS(memset)
-SYM_FUNC_START(__memset)
+SYM_FUNC_START_ALIAS(__memset)
+SYM_FUNC_START_LOCAL(memset)
 	/*
 	 * Some CPUs support enhanced REP MOVSB/STOSB feature. It is recommended
 	 * to use it when possible. If not available, use fast string instructions.
Signature · 1 line
2.26.2
#1 lore raw
Quoted 23 lines · Arnd Bergmann

clang and gas seem to interpret the symbols in memmove_64.S and memset_64.S differently, such that clang does not make them 'weak' as expected, which leads to a linker error, with both ld.bfd and ld.lld:

ld.lld: error: duplicate symbol: memmove

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

defined at common.c
kasan/common.o:(memmove) in archive mm/built-in.a
defined at memmove.o:(__memmove) in archive arch/arm64/lib/lib.a

ld.lld: error: duplicate symbol: memset

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

defined at common.c
kasan/common.o:(memset) in archive mm/built-in.a
defined at memset.o:(__memset) in archive arch/arm64/lib/lib.a

Copy the exact way these are written in memcpy_64.S, which does not have the same problem.

I don't know why this makes a difference, and it would be good to have someone with a better understanding of assembler internals review it.

Quoted 7 lines

It might be either a bug in the kernel or a bug in the assembler, no idea which one. My patch makes it work with all versions of clang and gcc, which is probably helpful even if it's a workaround for a clang bug.

+ Bill, Fangrui, Jian
I think we saw this bug or a very similar bug internally around the ordering of .weak to .global.

Quoted 42 lines
---
 arch/x86/lib/memmove_64.S | 4 ++--
 arch/x86/lib/memset_64.S  | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)
arch/x86/lib/memmove_64.S +2 -2
index 7ff00ea64e4f..dcca01434be8 100644
@@ -26,8 +26,8 @@
  */
 .weak memmove
 
-SYM_FUNC_START_ALIAS(memmove)
-SYM_FUNC_START(__memmove)
+SYM_FUNC_START_ALIAS(__memmove)
+SYM_FUNC_START_LOCAL(memmove)
 
        mov %rdi, %rax
 
arch/x86/lib/memset_64.S +2 -3
index 9ff15ee404a4..a97f2ea4e0b2 100644
@@ -19,8 +19,8 @@
  *
  * rax   original destination
  */
-SYM_FUNC_START_ALIAS(memset)
-SYM_FUNC_START(__memset)
+SYM_FUNC_START_ALIAS(__memset)
+SYM_FUNC_START_LOCAL(memset)
        /*
         * Some CPUs support enhanced REP MOVSB/STOSB feature. It is recommended
         * to use it when possible. If not available, use fast string instructions.
--

2.26.2

--
You received this message because you are subscribed to the Google Groups "Clang Built Linux" group.
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/clang-built-linux/20200527141553.1768675-1-arnd%40arndb.de.

Signature · 2 lines
Thanks,
~Nick Desaulniers

A message that is not in the archive (sent off-list or not yet received): CAGG=3QWwafjBUPaGLbJxyU7K65H0SnHBHZ-mnxtAC2A1xfjWYA@mail.gmail.com

#3 lore raw
Quoted 37 lines · Bill Wendling
Quoted 33 lines · Nick Desaulniers
Quoted 30 lines · Arnd Bergmann

clang and gas seem to interpret the symbols in memmove_64.S and memset_64.S differently, such that clang does not make them 'weak' as expected, which leads to a linker error, with both ld.bfd and ld.lld:

ld.lld: error: duplicate symbol: memmove

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

defined at common.c
kasan/common.o:(memmove) in archive mm/built-in.a
defined at memmove.o:(__memmove) in archive arch/arm64/lib/lib.a

ld.lld: error: duplicate symbol: memset

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

defined at common.c
kasan/common.o:(memset) in archive mm/built-in.a
defined at memset.o:(__memset) in archive arch/arm64/lib/lib.a

Copy the exact way these are written in memcpy_64.S, which does not have the same problem.

I don't know why this makes a difference, and it would be good to have someone with a better understanding of assembler internals review it.

It might be either a bug in the kernel or a bug in the assembler, no idea which one. My patch makes it work with all versions of clang and gcc, which is probably helpful even if it's a workaround for a clang bug.

Quoted 7 lines
Quoted 4 lines

+ Bill, Fangrui, Jian
I think we saw this bug or a very similar bug internally around the ordering of .weak to .global.

I think that would be this patch by Fangrui:

https://prodkernel-review.git.corp.google.com/c/kernel/release/9xx/+/292011

Thanks. That patch references this discussion:
https://lore.kernel.org/linuxppc-dev/[email protected]/

Quoted 52 lines

-bw

Quoted 49 lines
Quoted 39 lines
---
 arch/x86/lib/memmove_64.S | 4 ++--
 arch/x86/lib/memset_64.S  | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)
arch/x86/lib/memmove_64.S +2 -2
index 7ff00ea64e4f..dcca01434be8 100644
@@ -26,8 +26,8 @@
  */
 .weak memmove
 
-SYM_FUNC_START_ALIAS(memmove)
-SYM_FUNC_START(__memmove)
+SYM_FUNC_START_ALIAS(__memmove)
+SYM_FUNC_START_LOCAL(memmove)
 
        mov %rdi, %rax
 
arch/x86/lib/memset_64.S +2 -3
index 9ff15ee404a4..a97f2ea4e0b2 100644
@@ -19,8 +19,8 @@
  *
  * rax   original destination
  */
-SYM_FUNC_START_ALIAS(memset)
-SYM_FUNC_START(__memset)
+SYM_FUNC_START_ALIAS(__memset)
+SYM_FUNC_START_LOCAL(memset)
        /*
         * Some CPUs support enhanced REP MOVSB/STOSB feature. It is recommended
         * to use it when possible. If not available, use fast string instructions.
--

2.26.2

Quoted 4 lines

--
You received this message because you are subscribed to the Google Groups "Clang Built Linux" group.
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/clang-built-linux/20200527141553.1768675-1-arnd%40arndb.de.

--
Thanks,
~Nick Desaulniers

Signature · 2 lines
Thanks,
~Nick Desaulniers