94a3f78566
Eric Sandeen tracked an XFS on ARM corruption bug down to a function under fs/xfs/ involving some get_unaligned() calls on u64 pointers. As it turns out, calling ARM's get_unaligned() on a u64 pointer pointing to the following byte sequence: 80 81 82 83 84 85 86 87 would return ffffffff83828180 (LE mode.) This turns out to be because of implicit u8 -> int promotion in ARM's implementation of various helpers for get_unaligned(), causing them to accidentally return signed instead of unsigned values, which in turn caused the subsequent casts to unsigned long long in __get_unaligned_8_[bl]e() to sign-extend the lower words. Fix by casting the return values of __get_unaligned_[24]_[bl]e() to unsigned int. Cc: Eric Sandeen <sandeen@sandeen.net> Cc: Rabeeh Khoury <rabeeh@marvell.com> Cc: Nicolas Pitre <nico@marvell.com> Signed-off-by: Lennert Buytenhek <buytenh@marvell.com> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
182 lines
5.1 KiB
C
182 lines
5.1 KiB
C
#ifndef __ASM_ARM_UNALIGNED_H
|
|
#define __ASM_ARM_UNALIGNED_H
|
|
|
|
#include <asm/types.h>
|
|
|
|
extern int __bug_unaligned_x(const void *ptr);
|
|
|
|
/*
|
|
* What is the most efficient way of loading/storing an unaligned value?
|
|
*
|
|
* That is the subject of this file. Efficiency here is defined as
|
|
* minimum code size with minimum register usage for the common cases.
|
|
* It is currently not believed that long longs are common, so we
|
|
* trade efficiency for the chars, shorts and longs against the long
|
|
* longs.
|
|
*
|
|
* Current stats with gcc 2.7.2.2 for these functions:
|
|
*
|
|
* ptrsize get: code regs put: code regs
|
|
* 1 1 1 1 2
|
|
* 2 3 2 3 2
|
|
* 4 7 3 7 3
|
|
* 8 20 6 16 6
|
|
*
|
|
* gcc 2.95.1 seems to code differently:
|
|
*
|
|
* ptrsize get: code regs put: code regs
|
|
* 1 1 1 1 2
|
|
* 2 3 2 3 2
|
|
* 4 7 4 7 4
|
|
* 8 19 8 15 6
|
|
*
|
|
* which may or may not be more efficient (depending upon whether
|
|
* you can afford the extra registers). Hopefully the gcc 2.95
|
|
* is inteligent enough to decide if it is better to use the
|
|
* extra register, but evidence so far seems to suggest otherwise.
|
|
*
|
|
* Unfortunately, gcc is not able to optimise the high word
|
|
* out of long long >> 32, or the low word from long long << 32
|
|
*/
|
|
|
|
#define __get_unaligned_2_le(__p) \
|
|
(unsigned int)(__p[0] | __p[1] << 8)
|
|
|
|
#define __get_unaligned_2_be(__p) \
|
|
(unsigned int)(__p[0] << 8 | __p[1])
|
|
|
|
#define __get_unaligned_4_le(__p) \
|
|
(unsigned int)(__p[0] | __p[1] << 8 | __p[2] << 16 | __p[3] << 24)
|
|
|
|
#define __get_unaligned_4_be(__p) \
|
|
(unsigned int)(__p[0] << 24 | __p[1] << 16 | __p[2] << 8 | __p[3])
|
|
|
|
#define __get_unaligned_8_le(__p) \
|
|
((unsigned long long)__get_unaligned_4_le((__p+4)) << 32 | \
|
|
__get_unaligned_4_le(__p))
|
|
|
|
#define __get_unaligned_8_be(__p) \
|
|
((unsigned long long)__get_unaligned_4_be(__p) << 32 | \
|
|
__get_unaligned_4_be((__p+4)))
|
|
|
|
#define __get_unaligned_le(ptr) \
|
|
((__force typeof(*(ptr)))({ \
|
|
const __u8 *__p = (const __u8 *)(ptr); \
|
|
__builtin_choose_expr(sizeof(*(ptr)) == 1, *__p, \
|
|
__builtin_choose_expr(sizeof(*(ptr)) == 2, __get_unaligned_2_le(__p), \
|
|
__builtin_choose_expr(sizeof(*(ptr)) == 4, __get_unaligned_4_le(__p), \
|
|
__builtin_choose_expr(sizeof(*(ptr)) == 8, __get_unaligned_8_le(__p), \
|
|
(void)__bug_unaligned_x(__p))))); \
|
|
}))
|
|
|
|
#define __get_unaligned_be(ptr) \
|
|
((__force typeof(*(ptr)))({ \
|
|
const __u8 *__p = (const __u8 *)(ptr); \
|
|
__builtin_choose_expr(sizeof(*(ptr)) == 1, *__p, \
|
|
__builtin_choose_expr(sizeof(*(ptr)) == 2, __get_unaligned_2_be(__p), \
|
|
__builtin_choose_expr(sizeof(*(ptr)) == 4, __get_unaligned_4_be(__p), \
|
|
__builtin_choose_expr(sizeof(*(ptr)) == 8, __get_unaligned_8_be(__p), \
|
|
(void)__bug_unaligned_x(__p))))); \
|
|
}))
|
|
|
|
|
|
static inline void __put_unaligned_2_le(__u32 __v, register __u8 *__p)
|
|
{
|
|
*__p++ = __v;
|
|
*__p++ = __v >> 8;
|
|
}
|
|
|
|
static inline void __put_unaligned_2_be(__u32 __v, register __u8 *__p)
|
|
{
|
|
*__p++ = __v >> 8;
|
|
*__p++ = __v;
|
|
}
|
|
|
|
static inline void __put_unaligned_4_le(__u32 __v, register __u8 *__p)
|
|
{
|
|
__put_unaligned_2_le(__v >> 16, __p + 2);
|
|
__put_unaligned_2_le(__v, __p);
|
|
}
|
|
|
|
static inline void __put_unaligned_4_be(__u32 __v, register __u8 *__p)
|
|
{
|
|
__put_unaligned_2_be(__v >> 16, __p);
|
|
__put_unaligned_2_be(__v, __p + 2);
|
|
}
|
|
|
|
static inline void __put_unaligned_8_le(const unsigned long long __v, register __u8 *__p)
|
|
{
|
|
/*
|
|
* tradeoff: 8 bytes of stack for all unaligned puts (2
|
|
* instructions), or an extra register in the long long
|
|
* case - go for the extra register.
|
|
*/
|
|
__put_unaligned_4_le(__v >> 32, __p+4);
|
|
__put_unaligned_4_le(__v, __p);
|
|
}
|
|
|
|
static inline void __put_unaligned_8_be(const unsigned long long __v, register __u8 *__p)
|
|
{
|
|
/*
|
|
* tradeoff: 8 bytes of stack for all unaligned puts (2
|
|
* instructions), or an extra register in the long long
|
|
* case - go for the extra register.
|
|
*/
|
|
__put_unaligned_4_be(__v >> 32, __p);
|
|
__put_unaligned_4_be(__v, __p+4);
|
|
}
|
|
|
|
/*
|
|
* Try to store an unaligned value as efficiently as possible.
|
|
*/
|
|
#define __put_unaligned_le(val,ptr) \
|
|
({ \
|
|
(void)sizeof(*(ptr) = (val)); \
|
|
switch (sizeof(*(ptr))) { \
|
|
case 1: \
|
|
*(ptr) = (val); \
|
|
break; \
|
|
case 2: __put_unaligned_2_le((__force u16)(val),(__u8 *)(ptr)); \
|
|
break; \
|
|
case 4: __put_unaligned_4_le((__force u32)(val),(__u8 *)(ptr)); \
|
|
break; \
|
|
case 8: __put_unaligned_8_le((__force u64)(val),(__u8 *)(ptr)); \
|
|
break; \
|
|
default: __bug_unaligned_x(ptr); \
|
|
break; \
|
|
} \
|
|
(void) 0; \
|
|
})
|
|
|
|
#define __put_unaligned_be(val,ptr) \
|
|
({ \
|
|
(void)sizeof(*(ptr) = (val)); \
|
|
switch (sizeof(*(ptr))) { \
|
|
case 1: \
|
|
*(ptr) = (val); \
|
|
break; \
|
|
case 2: __put_unaligned_2_be((__force u16)(val),(__u8 *)(ptr)); \
|
|
break; \
|
|
case 4: __put_unaligned_4_be((__force u32)(val),(__u8 *)(ptr)); \
|
|
break; \
|
|
case 8: __put_unaligned_8_be((__force u64)(val),(__u8 *)(ptr)); \
|
|
break; \
|
|
default: __bug_unaligned_x(ptr); \
|
|
break; \
|
|
} \
|
|
(void) 0; \
|
|
})
|
|
|
|
/*
|
|
* Select endianness
|
|
*/
|
|
#ifndef __ARMEB__
|
|
#define get_unaligned __get_unaligned_le
|
|
#define put_unaligned __put_unaligned_le
|
|
#else
|
|
#define get_unaligned __get_unaligned_be
|
|
#define put_unaligned __put_unaligned_be
|
|
#endif
|
|
|
|
#endif
|