From: "Anthony G. Basile" blueness@gentoo.org
On systems where uClibc doesn't provide an arch specific byteswap.h, we fall back on bits/byteswap-common.h. However, there is a bug in this header in the __bswap_constant_64(x) macro. If, for example, a double is passed, we get "invalid operands to binary &" in which we mismatch a 'double' and 'long long unsigned int'. The newer glibc headers fix this and so we import them.
This is needed, for example, for f2fs-tools 1.6.0 on 32-bit big endian PowerPC.
Signed-off-by: Anthony G. Basile blueness@gentoo.org --- libc/sysdeps/linux/common/bits/byteswap-16.h | 34 +++++++ libc/sysdeps/linux/common/bits/byteswap-common.h | 109 ++++++++++++----------- 2 files changed, 91 insertions(+), 52 deletions(-) create mode 100644 libc/sysdeps/linux/common/bits/byteswap-16.h
diff --git a/libc/sysdeps/linux/common/bits/byteswap-16.h b/libc/sysdeps/linux/common/bits/byteswap-16.h new file mode 100644 index 0000000..8063aa8 --- /dev/null +++ b/libc/sysdeps/linux/common/bits/byteswap-16.h @@ -0,0 +1,34 @@ +/* Macros to swap the order of bytes in 16-bit integer values. + Copyright (C) 2012-2016 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + http://www.gnu.org/licenses/. */ + +#ifndef _BITS_BYTESWAP_H +# error "Never use <bits/byteswap-16.h> directly; include <byteswap.h> instead." +#endif + +#ifdef __GNUC__ +# define __bswap_16(x) \ + (__extension__ \ + ({ unsigned short int __bsx = (unsigned short int) (x); \ + __bswap_constant_16 (__bsx); })) +#else +static __inline unsigned short int +__bswap_16 (unsigned short int __bsx) +{ + return __bswap_constant_16 (__bsx); +} +#endif diff --git a/libc/sysdeps/linux/common/bits/byteswap-common.h b/libc/sysdeps/linux/common/bits/byteswap-common.h index 4941d47..0effea6 100644 --- a/libc/sysdeps/linux/common/bits/byteswap-common.h +++ b/libc/sysdeps/linux/common/bits/byteswap-common.h @@ -1,5 +1,5 @@ /* Macros to swap the order of bytes in integer values. - Copyright (C) 1997,1998,2000,2001,2002,2005 Free Software Foundation, Inc. + Copyright (C) 1997-2016 Free Software Foundation, Inc. This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or @@ -16,54 +16,40 @@ License along with the GNU C Library; if not, see http://www.gnu.org/licenses/. */
-#if !defined _BYTESWAP_H && !defined _NETINET_IN_H +#if !defined _BYTESWAP_H && !defined _NETINET_IN_H && !defined _ENDIAN_H # error "Never use <bits/byteswap.h> directly; include <byteswap.h> instead." #endif
#ifndef _BITS_BYTESWAP_H #define _BITS_BYTESWAP_H 1
+#include <features.h> +#include <bits/types.h> + /* Swap bytes in 16 bit value. */ #define __bswap_constant_16(x) \ - ((((x) >> 8) & 0xffu) | (((x) & 0xffu) << 8)) + ((unsigned short int)((((x) >> 8) & 0xffu) | (((x) & 0xffu) << 8)))
-#ifndef __bswap_non_constant_16 -# define __bswap_non_constant_16(x) __bswap_constant_16(x) -#endif -#ifdef __GNUC__ -# define __bswap_16(x) \ - (__extension__ \ - ({ unsigned short int __bsv, __bsx = (x); \ - if (__builtin_constant_p (__bsx)) \ - __bsv = __bswap_constant_16 (__bsx); \ - else \ - __bsv = __bswap_non_constant_16 (__bsx); \ - __bsv; })) -#else -static __inline unsigned short int -__bswap_16 (unsigned short int __bsx) -{ - return __bswap_constant_16 (__bsx); -} -#endif +/* Get __bswap_16. */ +#include <bits/byteswap-16.h>
/* Swap bytes in 32 bit value. */ #define __bswap_constant_32(x) \ ((((x) & 0xff000000u) >> 24) | (((x) & 0x00ff0000u) >> 8) | \ (((x) & 0x0000ff00u) << 8) | (((x) & 0x000000ffu) << 24))
-#ifndef __bswap_non_constant_32 -# define __bswap_non_constant_32(x) __bswap_constant_32(x) -#endif #ifdef __GNUC__ -# define __bswap_32(x) \ - (__extension__ \ - ({ unsigned int __bsv, __bsx = (x); \ - if (__builtin_constant_p (__bsx)) \ - __bsv = __bswap_constant_32 (__bsx); \ - else \ - __bsv = __bswap_non_constant_32 (__bsx); \ - __bsv; })) +# if __GNUC_PREREQ (4, 3) +static __inline unsigned int +__bswap_32 (unsigned int __bsx) +{ + return __builtin_bswap32 (__bsx); +} +# else +# define __bswap_32(x) \ + (__extension__ \ + ({ unsigned int __bsx = (x); __bswap_constant_32 (__bsx); })) +# endif #else static __inline unsigned int __bswap_32 (unsigned int __bsx) @@ -72,8 +58,40 @@ __bswap_32 (unsigned int __bsx) } #endif
-#if defined __GNUC__ && __GNUC__ >= 2 /* Swap bytes in 64 bit value. */ +#if __GNUC_PREREQ (2, 0) +# define __bswap_constant_64(x) \ + (__extension__ ((((x) & 0xff00000000000000ull) >> 56) \ + | (((x) & 0x00ff000000000000ull) >> 40) \ + | (((x) & 0x0000ff0000000000ull) >> 24) \ + | (((x) & 0x000000ff00000000ull) >> 8) \ + | (((x) & 0x00000000ff000000ull) << 8) \ + | (((x) & 0x0000000000ff0000ull) << 24) \ + | (((x) & 0x000000000000ff00ull) << 40) \ + | (((x) & 0x00000000000000ffull) << 56))) + +# if __GNUC_PREREQ (4, 3) +static __inline __uint64_t +__bswap_64 (__uint64_t __bsx) +{ + return __builtin_bswap64 (__bsx); +} +# else +# define __bswap_64(x) \ + (__extension__ \ + ({ union { __extension__ __uint64_t __ll; \ + unsigned int __l[2]; } __w, __r; \ + if (__builtin_constant_p (x)) \ + __r.__ll = __bswap_constant_64 (x); \ + else \ + { \ + __w.__ll = (x); \ + __r.__l[0] = __bswap_32 (__w.__l[1]); \ + __r.__l[1] = __bswap_32 (__w.__l[0]); \ + } \ + __r.__ll; })) +# endif +#else # define __bswap_constant_64(x) \ ((((x) & 0xff00000000000000ull) >> 56) \ | (((x) & 0x00ff000000000000ull) >> 40) \ @@ -84,24 +102,11 @@ __bswap_32 (unsigned int __bsx) | (((x) & 0x000000000000ff00ull) << 40) \ | (((x) & 0x00000000000000ffull) << 56))
-# ifndef __bswap_non_constant_64 -# define __bswap_non_constant_64(x) \ - (__extension__ \ - ({ union { __extension__ unsigned long long int __ll; \ - unsigned int __l[2]; } __w, __r; \ - __w.__ll = (x); \ - __r.__l[0] = __bswap_non_constant_32 (__w.__l[1]); \ - __r.__l[1] = __bswap_non_constant_32 (__w.__l[0]); \ - __r.__ll; })) -# endif -# define __bswap_64(x) \ - (__extension__ \ - ({ __extension__ unsigned long long int __ll; \ - if (__builtin_constant_p (x)) \ - __ll = __bswap_constant_64 (x); \ - else \ - __ll = __bswap_non_constant_64 (x); \ - __ll; })) +static __inline __uint64_t +__bswap_64 (__uint64_t __bsx) +{ + return __bswap_constant_64 (__bsx); +} #endif
#endif /* _BITS_BYTESWAP_H */
Hi Anthony, Anthony G. Basile wrote,
From: "Anthony G. Basile" blueness@gentoo.org
On systems where uClibc doesn't provide an arch specific byteswap.h, we fall back on bits/byteswap-common.h. However, there is a bug in this header in the __bswap_constant_64(x) macro. If, for example, a double is passed, we get "invalid operands to binary &" in which we mismatch a 'double' and 'long long unsigned int'. The newer glibc headers fix this and so we import them.
I have an issue with the patch on ARM Cortex-M4 (no MMU).
Any idea? Compile works fine without your patch.
/home/wbx/embedded-test/openadk/toolchain_generic-arm_uclibc-ng_cortex_m4_soft_eabi/usr/bin/arm-openadk-uclinux-uclibceabi-gcc -c libc/string/arm/memcpy.S -o libc/string/arm/memcpy.o - Wall -Wstrict-prototypes -Wstrict-aliasing -funsigned-char -fno-builtin -fno-asm -fmerge-all-constants -msoft-float -mlittle-endian -mthumb -fno-stack-protector -nostdinc -I./include - I./include -include libc-symbols.h -I./libc/sysdeps/linux/arm -I./libc/sysdeps/linux -I./ldso/ldso/arm -I./ldso/include -I. -DSTATIC -Os -fstrict-aliasing -fwrapv -fno-ident -fhonour-c opts -mcpu=cortex-m4 -mfloat-abi=soft -Wl,-elf2flt -Os -pipe -fomit-frame-pointer -fno-unwind-tables -fno-asynchronous-unwind-tables -mthumb -Wa,-mimplicit-it=thumb -mno-thumb-interwork -Wa,-mfloat-abi=soft -I./libc/sysdeps/linux/common -isystem /home/wbx/embedded-test/openadk/toolchain_generic-arm_uclibc-ng_cortex_m4_soft_eabi/usr/lib/gcc/arm-openadk-uclinux-uclibc eabi/5.3.0/include-fixed -isystem /home/wbx/embedded-test/openadk/toolchain_generic-arm_uclibc-ng_cortex_m4_soft_eabi/usr/lib/gcc/arm-openadk-uclinux-uclibceabi/5.3.0/include -I/home/w bx/embedded-test/openadk/target_generic-arm_uclibc-ng_cortex_m4_soft_eabi/usr/include/ -DNDEBUG -DIN_LIB=libc -MT libc/string/arm/memcpy.o -MD -MP -MF libc/string/arm/.memcpy.o.dep -D_ _ASSEMBLER__ -Wa,--noexecstack ./include/bits/kernel_types.h: Assembler messages: ./include/bits/kernel_types.h:10: Error: bad instruction `typedef unsigned short __kernel_dev_t' ./include/bits/kernel_types.h:11: Error: bad instruction `typedef unsigned long __kernel_ino_t' ./include/bits/kernel_types.h:12: Error: bad instruction `typedef unsigned short __kernel_mode_t' ./include/bits/kernel_types.h:13: Error: bad instruction `typedef unsigned short __kernel_nlink_t' ./include/bits/kernel_types.h:14: Error: bad instruction `typedef long __kernel_off_t' ./include/bits/kernel_types.h:15: Error: bad instruction `typedef int __kernel_pid_t' ./include/bits/kernel_types.h:16: Error: bad instruction `typedef unsigned short __kernel_ipc_pid_t' ./include/bits/kernel_types.h:17: Error: bad instruction `typedef unsigned short __kernel_uid_t' ./include/bits/kernel_types.h:18: Error: bad instruction `typedef unsigned short __kernel_gid_t' ./include/bits/kernel_types.h:19: Error: bad instruction `typedef unsigned int __kernel_size_t'
On 3/25/16 6:35 AM, Waldemar Brodkorb wrote:
Hi Anthony, Anthony G. Basile wrote,
From: "Anthony G. Basile" blueness@gentoo.org
On systems where uClibc doesn't provide an arch specific byteswap.h, we fall back on bits/byteswap-common.h. However, there is a bug in this header in the __bswap_constant_64(x) macro. If, for example, a double is passed, we get "invalid operands to binary &" in which we mismatch a 'double' and 'long long unsigned int'. The newer glibc headers fix this and so we import them.
I have an issue with the patch on ARM Cortex-M4 (no MMU).
Any idea? Compile works fine without your patch.
/home/wbx/embedded-test/openadk/toolchain_generic-arm_uclibc-ng_cortex_m4_soft_eabi/usr/bin/arm-openadk-uclinux-uclibceabi-gcc -c libc/string/arm/memcpy.S -o libc/string/arm/memcpy.o - Wall -Wstrict-prototypes -Wstrict-aliasing -funsigned-char -fno-builtin -fno-asm -fmerge-all-constants -msoft-float -mlittle-endian -mthumb -fno-stack-protector -nostdinc -I./include - I./include -include libc-symbols.h -I./libc/sysdeps/linux/arm -I./libc/sysdeps/linux -I./ldso/ldso/arm -I./ldso/include -I. -DSTATIC -Os -fstrict-aliasing -fwrapv -fno-ident -fhonour-c opts -mcpu=cortex-m4 -mfloat-abi=soft -Wl,-elf2flt -Os -pipe -fomit-frame-pointer -fno-unwind-tables -fno-asynchronous-unwind-tables -mthumb -Wa,-mimplicit-it=thumb -mno-thumb-interwork -Wa,-mfloat-abi=soft -I./libc/sysdeps/linux/common -isystem /home/wbx/embedded-test/openadk/toolchain_generic-arm_uclibc-ng_cortex_m4_soft_eabi/usr/lib/gcc/arm-openadk-uclinux-uclibc eabi/5.3.0/include-fixed -isystem /home/wbx/embedded-test/openadk/toolchain_generic-arm_uclibc-ng_cortex_m4_soft_eabi/usr/lib/gcc/arm-openadk-uclinux-uclibceabi/5.3.0/include -I/home/w bx/embedded-test/openadk/target_generic-arm_uclibc-ng_cortex_m4_soft_eabi/usr/include/ -DNDEBUG -DIN_LIB=libc -MT libc/string/arm/memcpy.o -MD -MP -MF libc/string/arm/.memcpy.o.dep -D_ _ASSEMBLER__ -Wa,--noexecstack ./include/bits/kernel_types.h: Assembler messages: ./include/bits/kernel_types.h:10: Error: bad instruction `typedef unsigned short __kernel_dev_t' ./include/bits/kernel_types.h:11: Error: bad instruction `typedef unsigned long __kernel_ino_t' ./include/bits/kernel_types.h:12: Error: bad instruction `typedef unsigned short __kernel_mode_t' ./include/bits/kernel_types.h:13: Error: bad instruction `typedef unsigned short __kernel_nlink_t' ./include/bits/kernel_types.h:14: Error: bad instruction `typedef long __kernel_off_t' ./include/bits/kernel_types.h:15: Error: bad instruction `typedef int __kernel_pid_t' ./include/bits/kernel_types.h:16: Error: bad instruction `typedef unsigned short __kernel_ipc_pid_t' ./include/bits/kernel_types.h:17: Error: bad instruction `typedef unsigned short __kernel_uid_t' ./include/bits/kernel_types.h:18: Error: bad instruction `typedef unsigned short __kernel_gid_t' ./include/bits/kernel_types.h:19: Error: bad instruction `typedef unsigned int __kernel_size_t'
Thanks for catching this. I get what's going on but I'm not sure of an easy fix. byteswap.h from glibc introduces #include <bits/types.h> which in turn includes <bits/kernel_types.h> (ifdef _LIBC), both of which have typedefs. This is not a problem when these headers get included in .c files but they are a problem when they are included in .S files. This happens in this case through #include <endian.h> which is in libc/string/arm/_memcpy.S and and strlen.S. You have a similar case with arch=sh.
If you're wondering what glibc does, its header stacking is different avoiding the issue altogether.
For now I have to withdraw this patch until I figure out a way around it. While it is the correct way to address the original issue, it introduces a regression. So now I'm thinking of how maybe to edit the .S files.
On 3/25/16 1:17 PM, Anthony G. Basile wrote:
For now I have to withdraw this patch until I figure out a way around it. While it is the correct way to address the original issue, it introduces a regression. So now I'm thinking of how maybe to edit the .S files.
The latest patch fixes this issue. I finally figured out that I could just guard against including byteswap.h through endian.h by masking part of it out. glibc does this too but for different reasons as far as I can tell because its headers stack differently.
-
Anthony G. Basile -
Waldemar Brodkorb