ARM: add workaround for ambiguous C99 stdint.h types
RFC PATCH 1/1 ARM: add workaround for ambiguous C99 stdint.h types
The C99 types uintXX_t that are usually defined in 'stdint.h' are not as unambiguous on ARM as you would expect. For the types below, there is a difference on ARM between GCC built for bare metal ARM, GCC built for glibc and the kernel itself, which results in build errors if you try to build with -ffreestanding and include 'stdint.h' (such as when you include 'arm_neon.h' in order to use NEON intrinsics)
As the typedefs for these types in 'stdint.h' are based on builtin defines supplied by GCC, we can tweak these to align with the kernel's idea of those types, so 'linux/types.h' and 'stdint.h' can be safely included from the same source file (provided that -ffreestanding is used).
int32_t uint32_t uintptr_t bare metal GCC long unsigned long unsigned long glibc GCC int unsigned int unsigned int kernel int unsigned int unsigned long
---
Hi all,
This is another, less obtrusive way of handling the problem for which I previously sent an RFC patch 'types.h: use GCC supplied typedefs if appropriate' In this case, only source files that in fact include stdint.h and are being built for ARM should be affected.
Regards,
Signature · 6 lines
Ard. arch/arm/include/asm/types.h | 40 ++++++++++++++++++++++++++++++++++++++++ 1 file changed, 40 insertions(+) create mode 100644 arch/arm/include/asm/types.h
arch/arm/include/asm/types.h new file +40 -0
@@ -0,0 +1,40 @@ +#ifndef _ASM_TYPES_H +#define _ASM_TYPES_H + +#include <asm-generic/int-ll64.h> + +/* +* The C99 types uintXX_t that are usually defined in 'stdint.h' are not as +* unambiguous on ARM as you would expect. For the types below, there is a +* difference on ARM between GCC built for bare metal ARM, GCC built for glibc +* and the kernel itself, which results in build errors if you try to build with +* -ffreestanding and include 'stdint.h' (such as when you include 'arm_neon.h' +* in order to use NEON intrinsics) +* +* As the typedefs for these types in 'stdint.h' are based on builtin defines +* supplied by GCC, we can tweak these to align with the kernel's idea of those +* types, so 'linux/types.h' and 'stdint.h' can be safely included from the same +* source file (provided that -ffreestanding is used). +* +* int32_t uint32_t uintptr_t +* bare metal GCC long unsigned long unsigned long +* glibc GCC int unsigned int unsigned int +* kernel int unsigned int unsigned long +*/ + +#ifdef __INT32__TYPE__ +#undef __INT32__TYPE__ +#define __INT32__TYPE__ int +#endif + +#ifdef __UINT32__TYPE__ +#undef __UINT32__TYPE__ +#define __UINT32__TYPE__ unsigned int +#endif + +#ifdef __UINTPTR_TYPE__ +#undef __UINTPTR_TYPE__ +#define __UINTPTR_TYPE__ unsigned long +#endif + +#endif /* _ASM_TYPES_H */
Signature · 1 line
1.8.1.2
RFC PATCH 1/1 ARM: add workaround for ambiguous C99 stdint.h types
Quoted 70 lines · Ard Biesheuvel
The C99 types
uintXX_tthat are usually defined in 'stdint.h' are not as unambiguous on ARM as you would expect. For the types below, there is a difference on ARM between GCC built for bare metal ARM, GCC built for glibc and the kernel itself, which results in build errors if you try to build with -ffreestanding and include 'stdint.h' (such as when you include 'arm_neon.h' in order to use NEON intrinsics)As the typedefs for these types in '
stdint.h' are based on builtin defines supplied by GCC, we can tweak these to align with the kernel's idea of those types, so 'linux/types.h' and 'stdint.h' can be safely included from the same source file (provided that -ffreestanding is used).int32_t uint32_t uintptr_t bare metal GCC long unsigned long unsigned long glibc GCC int unsigned int unsigned int kernel int unsigned int unsigned longSigned-off-by Ard Biesheuvel <[email protected]>---
Hi all,
This is another, less obtrusive way of handling the problem for which I previously sent an RFC patch '
types.h: use GCC supplied typedefs if appropriate' In this case, only source files that in fact includestdint.hand are being built for ARM should be affected.Regards,
Signature · 6 lines
Ard. arch/arm/include/asm/types.h | 40 ++++++++++++++++++++++++++++++++++++++++ 1 file changed, 40 insertions(+) create mode 100644 arch/arm/include/asm/types.harch/arm/include/asm/types.h new file +28
-0@@ -0,0 +1,40 @@ +#ifndef _ASM_TYPES_H +#define _ASM_TYPES_H + +#include <asm-generic/int-ll64.h> + +/* +* The C99 types uintXX_t that are usually defined in 'stdint.h' are not as +* unambiguous on ARM as you would expect. For the types below, there is a +* difference on ARM between GCC built for bare metal ARM, GCC built for glibc +* and the kernel itself, which results in build errors if you try to build with +* -ffreestanding and include 'stdint.h' (such as when you include 'arm_neon.h' +* in order to use NEON intrinsics) +* +* As the typedefs for these types in 'stdint.h' are based on builtin defines +* supplied by GCC, we can tweak these to align with the kernel's idea of those +* types, so 'linux/types.h' and 'stdint.h' can be safely included from the same +* source file (provided that -ffreestanding is used). +* +* int32_t uint32_t uintptr_t +* bare metal GCC long unsigned long unsigned long +* glibc GCC int unsigned int unsigned int +* kernel int unsigned int unsigned long +*/ + +#ifdef __INT32__TYPE__ +#undef __INT32__TYPE__ +#define __INT32__TYPE__ int +#endif
arch/arm/include/asm/types.h line 38
Quoted 10 lines
arch/arm/include/asm/types.h new file +10
-0+ +#ifdef __UINT32__TYPE__ +#undef __UINT32__TYPE__ +#define __UINT32__TYPE__ unsigned int +#endif + +#ifdef __UINTPTR_TYPE__ +#undef __UINTPTR_TYPE__ +#define __UINTPTR_TYPE__ unsigned long +#endif
Somebody else might have the opposite problem to ARM, so I'm doubtful about whether it's safe to do this for all arches. The arch maintainers would have to comment on that. This looks ugly in an otherwise generic header.
(As a cosmetic thing, you can lose the #ifdefs. #undef doesn't trigger an error of the specified macro doesn't already exist.)
Cheers
---Dave
RFC PATCH 1/1 ARM: add workaround for ambiguous C99 stdint.h types
Quoted 1 line · Dave Martin
On Fri, Aug 09, 2013 at 09:36:42AM +0200, Ard Biesheuvel wrote:
[...]
Quoted 5 lines
Somebody else might have the opposite problem to ARM, so I'm doubtful about whether it's safe to do this for all arches. The arch maintainers would have to comment on that. This looks ugly in an otherwise generic header.
This is actually under arch/arm so that should not be a problem. In fact, it's a copy of asm-generic/types.h with just the #defines added.
Quoted 3 lines
(As a cosmetic thing, you can lose the #ifdefs. #undef doesn't trigger an error of the specified macro doesn't already exist.)
I am aware of that, but I think it is cleaner not to pollute the namespace if the defines weren't there to begin with.
Cheers,
Ard.
RFC PATCH 1/1 ARM: add workaround for ambiguous C99 stdint.h types
Quoted 5 lines · Ard Biesheuvel
Quoted 1 line · Dave Martin
On Fri, Aug 09, 2013 at 09:36:42AM +0200, Ard Biesheuvel wrote:
[...]
Quoted 8 lines
Quoted 5 lines
Somebody else might have the opposite problem to ARM, so I'm doubtful about whether it's safe to do this for all arches. The arch maintainers would have to comment on that. This looks ugly in an otherwise generic header.
This is actually under
arch/armso that should not be a problem. In fact, it's a copy ofasm-generic/types.hwith just the #defines added.
Duh. Misread your patch, sorry about that.
Quoted 6 lines
Quoted 3 lines
(As a cosmetic thing, you can lose the #ifdefs. #undef doesn't trigger an error of the specified macro doesn't already exist.)
I am aware of that, but I think it is cleaner not to pollute the namespace if the defines weren't there to begin with.
Ah, I see what you mean. Yes, that makes sense. I'd read the #ifdefs as just being there to avoid redefinition errors.
I'll leave the thread for other people to comment, but this looks like a reasonable thing to do for now.
Because the kernel is not a hosted C environment, we shouldn't be including any foreign headers which care about the distinction, except for GCC's own headers like arm_neon.h.
Cheers
---Dave
RFC PATCH 1/1 ARM: add workaround for ambiguous C99 stdint.h types
In reply to Ard BiesheuvelKinda-acked-by Dave Martin, anyone else care to comment?
Quoted 85 lines · Ard Biesheuvel
The C99 types
uintXX_tthat are usually defined in 'stdint.h' are not as unambiguous on ARM as you would expect. For the types below, there is a difference on ARM between GCC built for bare metal ARM, GCC built for glibc and the kernel itself, which results in build errors if you try to build with -ffreestanding and include 'stdint.h' (such as when you include 'arm_neon.h' in order to use NEON intrinsics)As the typedefs for these types in '
stdint.h' are based on builtin defines supplied by GCC, we can tweak these to align with the kernel's idea of those types, so 'linux/types.h' and 'stdint.h' can be safely included from the same source file (provided that -ffreestanding is used).int32_t uint32_t uintptr_t bare metal GCC long unsigned long unsigned long glibc GCC int unsigned int unsigned int kernel int unsigned int unsigned longSigned-off-by Ard Biesheuvel <[email protected]>---
Hi all,
This is another, less obtrusive way of handling the problem for which I previously sent an RFC patch '
types.h: use GCC supplied typedefs if appropriate' In this case, only source files that in fact includestdint.hand are being built for ARM should be affected.Regards,
--
Ard.arch/arm/include/asm/types.h | 40 ++++++++++++++++++++++++++++++++++++++++ 1 file changed, 40 insertions(+) create mode 100644 arch/arm/include/asm/types.harch/arm/include/asm/types.h new file +40
-1@@ -0,0 +1,40 @@ +#ifndef _ASM_TYPES_H +#define _ASM_TYPES_H + +#include <asm-generic/int-ll64.h> + +/* +* The C99 types uintXX_t that are usually defined in 'stdint.h' are not as +* unambiguous on ARM as you would expect. For the types below, there is a +* difference on ARM between GCC built for bare metal ARM, GCC built for glibc +* and the kernel itself, which results in build errors if you try to build with +* -ffreestanding and include 'stdint.h' (such as when you include 'arm_neon.h' +* in order to use NEON intrinsics) +* +* As the typedefs for these types in 'stdint.h' are based on builtin defines +* supplied by GCC, we can tweak these to align with the kernel's idea of those +* types, so 'linux/types.h' and 'stdint.h' can be safely included from the same +* source file (provided that -ffreestanding is used). +* +* int32_t uint32_t uintptr_t +* bare metal GCC long unsigned long unsigned long +* glibc GCC int unsigned int unsigned int +* kernel int unsigned int unsigned long +*/ + +#ifdef __INT32__TYPE__ +#undef __INT32__TYPE__ +#define __INT32__TYPE__ int +#endif + +#ifdef __UINT32__TYPE__ +#undef __UINT32__TYPE__ +#define __UINT32__TYPE__ unsigned int +#endif + +#ifdef __UINTPTR_TYPE__ +#undef __UINTPTR_TYPE__ +#define __UINTPTR_TYPE__ unsigned long +#endif + +#endif /* _ASM_TYPES_H */ --1.8.1.2
RFC PATCH 1/1 ARM: add workaround for ambiguous C99 stdint.h types
Quoted 2 lines · Ard Biesheuvel
Kinda-acked-by Dave Martin, anyone else care to comment?
Quoted 87 lines · Ard Biesheuvel
Quoted 85 lines
The C99 types
uintXX_tthat are usually defined in 'stdint.h' are not as unambiguous on ARM as you would expect. For the types below, there is a difference on ARM between GCC built for bare metal ARM, GCC built for glibc and the kernel itself, which results in build errors if you try to build with -ffreestanding and include 'stdint.h' (such as when you include 'arm_neon.h' in order to use NEON intrinsics)As the typedefs for these types in '
stdint.h' are based on builtin defines supplied by GCC, we can tweak these to align with the kernel's idea of those types, so 'linux/types.h' and 'stdint.h' can be safely included from the same source file (provided that -ffreestanding is used).int32_t uint32_t uintptr_t bare metal GCC long unsigned long unsigned long glibc GCC int unsigned int unsigned int kernel int unsigned int unsigned longSigned-off-by Ard Biesheuvel <[email protected]>---
Hi all,
This is another, less obtrusive way of handling the problem for which I previously sent an RFC patch '
types.h: use GCC supplied typedefs if appropriate' In this case, only source files that in fact includestdint.hand are being built for ARM should be affected.Regards,
--
Ard.arch/arm/include/asm/types.h | 40 ++++++++++++++++++++++++++++++++++++++++ 1 file changed, 40 insertions(+) create mode 100644 arch/arm/include/asm/types.harch/arm/include/asm/types.h new file +40
-1@@ -0,0 +1,40 @@ +#ifndef _ASM_TYPES_H +#define _ASM_TYPES_H + +#include <asm-generic/int-ll64.h> + +/* +* The C99 types uintXX_t that are usually defined in 'stdint.h' are not as +* unambiguous on ARM as you would expect. For the types below, there is a +* difference on ARM between GCC built for bare metal ARM, GCC built for glibc +* and the kernel itself, which results in build errors if you try to build with +* -ffreestanding and include 'stdint.h' (such as when you include 'arm_neon.h' +* in order to use NEON intrinsics) +* +* As the typedefs for these types in 'stdint.h' are based on builtin defines +* supplied by GCC, we can tweak these to align with the kernel's idea of those +* types, so 'linux/types.h' and 'stdint.h' can be safely included from the same +* source file (provided that -ffreestanding is used). +* +* int32_t uint32_t uintptr_t +* bare metal GCC long unsigned long unsigned long +* glibc GCC int unsigned int unsigned int +* kernel int unsigned int unsigned long +*/ + +#ifdef __INT32__TYPE__ +#undef __INT32__TYPE__ +#define __INT32__TYPE__ int +#endif + +#ifdef __UINT32__TYPE__ +#undef __UINT32__TYPE__ +#define __UINT32__TYPE__ unsigned int +#endif + +#ifdef __UINTPTR_TYPE__ +#undef __UINTPTR_TYPE__ +#define __UINTPTR_TYPE__ unsigned long +#endif + +#endif /* _ASM_TYPES_H */ --1.8.1.2