Skip to contentSkip to conversation

ARM: trivial assembly fixes to enable LLVM as

PATCH In review Stefan Agner 2 participants 8 messages Jan 3, 2019 – Jan 7, 2019

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

v2 cover ARM: trivial assembly fixes to enable LLVM as

This patchset is a starting point to enable LLVM integrated assembler and contains some trivial changes. With this patchset the LLVM integrated assembler can be used to assemble almost all C files.

Changes since v1:
- Explicitly use unified syntax for inline assembly where necessary

Stefan Agner (3):
ARM: fix argument count to match macro definition
ARM: uaccess: use unified assembler language syntax
ARM: spinlock: use unified assembler language syntax

 arch/arm/include/asm/spinlock.h | 3 ++-
 arch/arm/include/asm/uaccess.h  | 3 ++-
 arch/arm/lib/copy_template.S    | 2 +-
 3 files changed, 5 insertions(+), 3 deletions(-)
Signature · 1 line
2.20.1
#1 lore raw

PATCH v2 1/3 ARM: fix argument count to match macro definition

The macro str8w takes 10 arguments, abort being the 10th. In this particular instantiation the abort argument is passed as 11th argument leading to an error when using LLVM's integrated assembler:
<instantiation>:46:47: error: too many positional arguments

    str8w r0, r3, r4, r5, r6, r7, r8, r9, ip, , abort=19f
                                                ^
  arch/arm/lib/copy_template.S:277:5: note: while in macro instantiation
  18: forward_copy_shift pull=24 push=8
      ^

The argument is not used in the macro hence this does not change code generation.

---
 arch/arm/lib/copy_template.S | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm/lib/copy_template.S +1 -1
index 652e4d98cd47..2d54491b0e22 100644
@@ -241,7 +241,7 @@
 		orr	r9, r9, ip, lspush #\push
 		mov	ip, ip, lspull #\pull
 		orr	ip, ip, lr, lspush #\push
-		str8w	r0, r3, r4, r5, r6, r7, r8, r9, ip, , abort=19f
+		str8w	r0, r3, r4, r5, r6, r7, r8, r9, ip, abort=19f
 		bge	12b
 	PLD(	cmn	r2, #96			)
 	PLD(	bge	13b			)
Signature · 1 line
2.20.1
Reviewed-by on v2 1/3
#2 lore raw

Re: [PATCH v2 1/3] ARM: fix argument count to match macro definition

Quoted 11 lines · Stefan Agner

The macro str8w takes 10 arguments, abort being the 10th. In this particular instantiation the abort argument is passed as 11th argument leading to an error when using LLVM's integrated assembler:
<instantiation>:46:47: error: too many positional arguments

    str8w r0, r3, r4, r5, r6, r7, r8, r9, ip, , abort=19f
                                                ^
  arch/arm/lib/copy_template.S:277:5: note: while in macro instantiation
  18: forward_copy_shift pull=24 push=8
      ^
Quoted 4 lines

The argument is not used in the macro hence this does not change code generation.

Quoted 21 lines
---
 arch/arm/lib/copy_template.S | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm/lib/copy_template.S +1 -1
index 652e4d98cd47..2d54491b0e22 100644
@@ -241,7 +241,7 @@
 		orr	r9, r9, ip, lspush #\push
 		mov	ip, ip, lspull #\pull
 		orr	ip, ip, lr, lspush #\push
-		str8w	r0, r3, r4, r5, r6, r7, r8, r9, ip, , abort=19f
+		str8w	r0, r3, r4, r5, r6, r7, r8, r9, ip, abort=19f
 		bge	12b
 	PLD(	cmn	r2, #96			)
 	PLD(	bge	13b			)
Signature · 1 line
2.20.1
#3 lore raw

PATCH v2 2/3 ARM: uaccess: use unified assembler language syntax

In reply to Stefan Agner

Convert the conditional infix to a postfix to make sure this inline assembly is unified syntax. Since gcc assumes non-unified syntax when emitting ARM instructions, make sure to define the syntax as unified.

This allows to use LLVM's integrated assembler.

---
Changes since v1:
- Explicitly use unified syntax for inline assembly

 arch/arm/include/asm/uaccess.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
arch/arm/include/asm/uaccess.h +2 -1
index 6390a40f16e7..a50f9b4e2574 100644
@@ -86,7 +86,8 @@ static inline void set_fs(mm_segment_t fs)
 #define __range_ok(addr, size) ({ \
 	unsigned long flag, roksum; \
 	__chk_user_ptr(addr);	\
-	__asm__("adds %1, %2, %3; sbcccs %1, %1, %0; movcc %0, #0" \
+	__asm__(".syntax unified\n" \
+		"adds %1, %2, %3; sbcscc %1, %1, %0; movcc %0, #0" \
 		: "=&r" (flag), "=&r" (roksum) \
 		: "r" (addr), "Ir" (size), "0" (current_thread_info()->addr_limit) \
 		: "cc"); \
Signature · 1 line
2.20.1
#4 lore raw

Re: [PATCH v2 2/3] ARM: uaccess: use unified assembler language syntax

Quoted 21 lines · Stefan Agner

Convert the conditional infix to a postfix to make sure this inline assembly is unified syntax. Since gcc assumes non-unified syntax when emitting ARM instructions, make sure to define the syntax as unified.

This allows to use LLVM's integrated assembler.

---
Changes since v1:
- Explicitly use unified syntax for inline assembly

 arch/arm/include/asm/uaccess.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
arch/arm/include/asm/uaccess.h +0 -0
index 6390a40f16e7..a50f9b4e2574 100644

arch/arm/include/asm/uaccess.h static inline void set_fs(mm_segment_t fs) · line 90

Quoted 6 lines
arch/arm/include/asm/uaccess.h +2 -1
@@ -86,7 +86,8 @@ static inline void set_fs(mm_segment_t fs)
 #define __range_ok(addr, size) ({ \
 	unsigned long flag, roksum; \
 	__chk_user_ptr(addr);	\
-	__asm__("adds %1, %2, %3; sbcccs %1, %1, %0; movcc %0, #0" \
+	__asm__(".syntax unified\n" \
+		"adds %1, %2, %3; sbcscc %1, %1, %0; movcc %0, #0" \

Instead of sprinkling ".syntax unified" around, you could consider including <asm/unified.h> when needed.

Nicolas

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

#5 lore raw

Re: [PATCH v2 2/3] ARM: uaccess: use unified assembler language syntax

Quoted 18 lines · Nicolas Pitre
Quoted 16 lines · Stefan Agner

Convert the conditional infix to a postfix to make sure this inline assembly is unified syntax. Since gcc assumes non-unified syntax when emitting ARM instructions, make sure to define the syntax as unified.

This allows to use LLVM's integrated assembler.

---
Changes since v1:
- Explicitly use unified syntax for inline assembly

 arch/arm/include/asm/uaccess.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
Quoted 14 lines

arch/arm/include/asm/uaccess.h static inline void set_fs(mm_segment_t fs) · line 90

Quoted 11 lines
arch/arm/include/asm/uaccess.h +2 -1
index 6390a40f16e7..a50f9b4e2574 100644
@@ -86,7 +86,8 @@ static inline void set_fs(mm_segment_t fs)
 #define __range_ok(addr, size) ({ \
 	unsigned long flag, roksum; \
 	__chk_user_ptr(addr);	\
-	__asm__("adds %1, %2, %3; sbcccs %1, %1, %0; movcc %0, #0" \
+	__asm__(".syntax unified\n" \
+		"adds %1, %2, %3; sbcscc %1, %1, %0; movcc %0, #0" \

Instead of sprinkling ".syntax unified" around, you could consider including <asm/unified.h> when needed.

This seems not to help in this case. When I include <asm/unified.h> at the beginning of uaccess.h GCC still seems to emit ".syntax unified".

I guess this is since we are dealing with inline assembly. As far as I understand its really gcc which emits a ".syntax divided" by default when compiling ARM mode. There is an option "-masm-syntax-unified" which according to the gcc man page should do what we need, but it seems broken, see also: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88648

I agree this patch is not particularly pretty, but it is the only solution I found to make this inline assembly work with gcc and Clang.

--
Stefan

#6 lore raw

PATCH v2 3/3 ARM: spinlock: use unified assembler language syntax

In reply to Stefan Agner

Convert the conditional infix to a postfix to make sure this inline assembly is unified syntax. Since gcc assumes non-unified syntax when emitting ARM instructions, make sure to define the syntax as unified.

This allows to use LLVM's integrated assembler.

---
Changes since v1:
- Explicitly use unified syntax for inline assembly

 arch/arm/include/asm/spinlock.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
arch/arm/include/asm/spinlock.h +2 -1
index 099c78fcf62d..8f009e788ad4 100644
@@ -210,11 +210,12 @@ static inline void arch_read_lock(arch_rwlock_t *rw)
 
 	prefetchw(&rw->lock);
 	__asm__ __volatile__(
+"	.syntax unified\n"
 "1:	ldrex	%0, [%2]\n"
 "	adds	%0, %0, #1\n"
 "	strexpl	%1, %0, [%2]\n"
 	WFE("mi")
-"	rsbpls	%0, %1, #0\n"
+"	rsbspl	%0, %1, #0\n"
 "	bmi	1b"
 	: "=&r" (tmp), "=&r" (tmp2)
 	: "r" (&rw->lock)
Signature · 1 line
2.20.1
#7 lore raw

Re: [PATCH v2 3/3] ARM: spinlock: use unified assembler language syntax

Quoted 6 lines · Stefan Agner

Convert the conditional infix to a postfix to make sure this inline assembly is unified syntax. Since gcc assumes non-unified syntax when emitting ARM instructions, make sure to define the syntax as unified.

This allows to use LLVM's integrated assembler.

Same comment as for patch 2/3.

Quoted 32 lines

---
Changes since v1:
- Explicitly use unified syntax for inline assembly

 arch/arm/include/asm/spinlock.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
arch/arm/include/asm/spinlock.h +2 -1
index 099c78fcf62d..8f009e788ad4 100644
@@ -210,11 +210,12 @@ static inline void arch_read_lock(arch_rwlock_t *rw)
 
 	prefetchw(&rw->lock);
 	__asm__ __volatile__(
+"	.syntax unified\n"
 "1:	ldrex	%0, [%2]\n"
 "	adds	%0, %0, #1\n"
 "	strexpl	%1, %0, [%2]\n"
 	WFE("mi")
-"	rsbpls	%0, %1, #0\n"
+"	rsbspl	%0, %1, #0\n"
 "	bmi	1b"
 	: "=&r" (tmp), "=&r" (tmp2)
 	: "r" (&rw->lock)
Signature · 1 line
2.20.1