mirror of
https://github.com/AuxXxilium/linux_dsm_epyc7002.git
synced 2024-12-20 15:47:03 +07:00
42344113ba
Recent probing at the Linux Kernel Memory Model uncovered a
'surprise'. Strongly ordered architectures where the atomic RmW
primitive implies full memory ordering and
smp_mb__{before,after}_atomic() are a simple barrier() (such as MIPS
without WEAK_REORDERING_BEYOND_LLSC) fail for:
*x = 1;
atomic_inc(u);
smp_mb__after_atomic();
r0 = *y;
Because, while the atomic_inc() implies memory order, it
(surprisingly) does not provide a compiler barrier. This then allows
the compiler to re-order like so:
atomic_inc(u);
*x = 1;
smp_mb__after_atomic();
r0 = *y;
Which the CPU is then allowed to re-order (under TSO rules) like:
atomic_inc(u);
r0 = *y;
*x = 1;
And this very much was not intended. Therefore strengthen the atomic
RmW ops to include a compiler barrier.
Reported-by: Andrea Parri <andrea.parri@amarulasolutions.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Paul Burton <paul.burton@mips.com>
317 lines
8.1 KiB
C
317 lines
8.1 KiB
C
/*
|
|
* This file is subject to the terms and conditions of the GNU General Public
|
|
* License. See the file "COPYING" in the main directory of this archive
|
|
* for more details.
|
|
*
|
|
* Copyright (C) 2003, 06, 07 by Ralf Baechle (ralf@linux-mips.org)
|
|
*/
|
|
#ifndef __ASM_CMPXCHG_H
|
|
#define __ASM_CMPXCHG_H
|
|
|
|
#include <linux/bug.h>
|
|
#include <linux/irqflags.h>
|
|
#include <asm/compiler.h>
|
|
#include <asm/war.h>
|
|
|
|
/*
|
|
* Using a branch-likely instruction to check the result of an sc instruction
|
|
* works around a bug present in R10000 CPUs prior to revision 3.0 that could
|
|
* cause ll-sc sequences to execute non-atomically.
|
|
*/
|
|
#if R10000_LLSC_WAR
|
|
# define __scbeqz "beqzl"
|
|
#else
|
|
# define __scbeqz "beqz"
|
|
#endif
|
|
|
|
/*
|
|
* These functions doesn't exist, so if they are called you'll either:
|
|
*
|
|
* - Get an error at compile-time due to __compiletime_error, if supported by
|
|
* your compiler.
|
|
*
|
|
* or:
|
|
*
|
|
* - Get an error at link-time due to the call to the missing function.
|
|
*/
|
|
extern unsigned long __cmpxchg_called_with_bad_pointer(void)
|
|
__compiletime_error("Bad argument size for cmpxchg");
|
|
extern unsigned long __cmpxchg64_unsupported(void)
|
|
__compiletime_error("cmpxchg64 not available; cpu_has_64bits may be false");
|
|
extern unsigned long __xchg_called_with_bad_pointer(void)
|
|
__compiletime_error("Bad argument size for xchg");
|
|
|
|
#define __xchg_asm(ld, st, m, val) \
|
|
({ \
|
|
__typeof(*(m)) __ret; \
|
|
\
|
|
if (kernel_uses_llsc) { \
|
|
loongson_llsc_mb(); \
|
|
__asm__ __volatile__( \
|
|
" .set push \n" \
|
|
" .set noat \n" \
|
|
" .set push \n" \
|
|
" .set " MIPS_ISA_ARCH_LEVEL " \n" \
|
|
"1: " ld " %0, %2 # __xchg_asm \n" \
|
|
" .set pop \n" \
|
|
" move $1, %z3 \n" \
|
|
" .set " MIPS_ISA_ARCH_LEVEL " \n" \
|
|
" " st " $1, %1 \n" \
|
|
"\t" __scbeqz " $1, 1b \n" \
|
|
" .set pop \n" \
|
|
: "=&r" (__ret), "=" GCC_OFF_SMALL_ASM() (*m) \
|
|
: GCC_OFF_SMALL_ASM() (*m), "Jr" (val) \
|
|
: __LLSC_CLOBBER); \
|
|
} else { \
|
|
unsigned long __flags; \
|
|
\
|
|
raw_local_irq_save(__flags); \
|
|
__ret = *m; \
|
|
*m = val; \
|
|
raw_local_irq_restore(__flags); \
|
|
} \
|
|
\
|
|
__ret; \
|
|
})
|
|
|
|
extern unsigned long __xchg_small(volatile void *ptr, unsigned long val,
|
|
unsigned int size);
|
|
|
|
static inline unsigned long __xchg(volatile void *ptr, unsigned long x,
|
|
int size)
|
|
{
|
|
switch (size) {
|
|
case 1:
|
|
case 2:
|
|
return __xchg_small(ptr, x, size);
|
|
|
|
case 4:
|
|
return __xchg_asm("ll", "sc", (volatile u32 *)ptr, x);
|
|
|
|
case 8:
|
|
if (!IS_ENABLED(CONFIG_64BIT))
|
|
return __xchg_called_with_bad_pointer();
|
|
|
|
return __xchg_asm("lld", "scd", (volatile u64 *)ptr, x);
|
|
|
|
default:
|
|
return __xchg_called_with_bad_pointer();
|
|
}
|
|
}
|
|
|
|
#define xchg(ptr, x) \
|
|
({ \
|
|
__typeof__(*(ptr)) __res; \
|
|
\
|
|
smp_mb__before_llsc(); \
|
|
\
|
|
__res = (__typeof__(*(ptr))) \
|
|
__xchg((ptr), (unsigned long)(x), sizeof(*(ptr))); \
|
|
\
|
|
smp_llsc_mb(); \
|
|
\
|
|
__res; \
|
|
})
|
|
|
|
#define __cmpxchg_asm(ld, st, m, old, new) \
|
|
({ \
|
|
__typeof(*(m)) __ret; \
|
|
\
|
|
if (kernel_uses_llsc) { \
|
|
loongson_llsc_mb(); \
|
|
__asm__ __volatile__( \
|
|
" .set push \n" \
|
|
" .set noat \n" \
|
|
" .set push \n" \
|
|
" .set "MIPS_ISA_ARCH_LEVEL" \n" \
|
|
"1: " ld " %0, %2 # __cmpxchg_asm \n" \
|
|
" bne %0, %z3, 2f \n" \
|
|
" .set pop \n" \
|
|
" move $1, %z4 \n" \
|
|
" .set "MIPS_ISA_ARCH_LEVEL" \n" \
|
|
" " st " $1, %1 \n" \
|
|
"\t" __scbeqz " $1, 1b \n" \
|
|
" .set pop \n" \
|
|
"2: \n" \
|
|
: "=&r" (__ret), "=" GCC_OFF_SMALL_ASM() (*m) \
|
|
: GCC_OFF_SMALL_ASM() (*m), "Jr" (old), "Jr" (new) \
|
|
: __LLSC_CLOBBER); \
|
|
loongson_llsc_mb(); \
|
|
} else { \
|
|
unsigned long __flags; \
|
|
\
|
|
raw_local_irq_save(__flags); \
|
|
__ret = *m; \
|
|
if (__ret == old) \
|
|
*m = new; \
|
|
raw_local_irq_restore(__flags); \
|
|
} \
|
|
\
|
|
__ret; \
|
|
})
|
|
|
|
extern unsigned long __cmpxchg_small(volatile void *ptr, unsigned long old,
|
|
unsigned long new, unsigned int size);
|
|
|
|
static inline unsigned long __cmpxchg(volatile void *ptr, unsigned long old,
|
|
unsigned long new, unsigned int size)
|
|
{
|
|
switch (size) {
|
|
case 1:
|
|
case 2:
|
|
return __cmpxchg_small(ptr, old, new, size);
|
|
|
|
case 4:
|
|
return __cmpxchg_asm("ll", "sc", (volatile u32 *)ptr,
|
|
(u32)old, new);
|
|
|
|
case 8:
|
|
/* lld/scd are only available for MIPS64 */
|
|
if (!IS_ENABLED(CONFIG_64BIT))
|
|
return __cmpxchg_called_with_bad_pointer();
|
|
|
|
return __cmpxchg_asm("lld", "scd", (volatile u64 *)ptr,
|
|
(u64)old, new);
|
|
|
|
default:
|
|
return __cmpxchg_called_with_bad_pointer();
|
|
}
|
|
}
|
|
|
|
#define cmpxchg_local(ptr, old, new) \
|
|
((__typeof__(*(ptr))) \
|
|
__cmpxchg((ptr), \
|
|
(unsigned long)(__typeof__(*(ptr)))(old), \
|
|
(unsigned long)(__typeof__(*(ptr)))(new), \
|
|
sizeof(*(ptr))))
|
|
|
|
#define cmpxchg(ptr, old, new) \
|
|
({ \
|
|
__typeof__(*(ptr)) __res; \
|
|
\
|
|
smp_mb__before_llsc(); \
|
|
__res = cmpxchg_local((ptr), (old), (new)); \
|
|
smp_llsc_mb(); \
|
|
\
|
|
__res; \
|
|
})
|
|
|
|
#ifdef CONFIG_64BIT
|
|
#define cmpxchg64_local(ptr, o, n) \
|
|
({ \
|
|
BUILD_BUG_ON(sizeof(*(ptr)) != 8); \
|
|
cmpxchg_local((ptr), (o), (n)); \
|
|
})
|
|
|
|
#define cmpxchg64(ptr, o, n) \
|
|
({ \
|
|
BUILD_BUG_ON(sizeof(*(ptr)) != 8); \
|
|
cmpxchg((ptr), (o), (n)); \
|
|
})
|
|
#else
|
|
|
|
# include <asm-generic/cmpxchg-local.h>
|
|
# define cmpxchg64_local(ptr, o, n) __cmpxchg64_local_generic((ptr), (o), (n))
|
|
|
|
# ifdef CONFIG_SMP
|
|
|
|
static inline unsigned long __cmpxchg64(volatile void *ptr,
|
|
unsigned long long old,
|
|
unsigned long long new)
|
|
{
|
|
unsigned long long tmp, ret;
|
|
unsigned long flags;
|
|
|
|
/*
|
|
* The assembly below has to combine 32 bit values into a 64 bit
|
|
* register, and split 64 bit values from one register into two. If we
|
|
* were to take an interrupt in the middle of this we'd only save the
|
|
* least significant 32 bits of each register & probably clobber the
|
|
* most significant 32 bits of the 64 bit values we're using. In order
|
|
* to avoid this we must disable interrupts.
|
|
*/
|
|
local_irq_save(flags);
|
|
|
|
loongson_llsc_mb();
|
|
asm volatile(
|
|
" .set push \n"
|
|
" .set " MIPS_ISA_ARCH_LEVEL " \n"
|
|
/* Load 64 bits from ptr */
|
|
"1: lld %L0, %3 # __cmpxchg64 \n"
|
|
/*
|
|
* Split the 64 bit value we loaded into the 2 registers that hold the
|
|
* ret variable.
|
|
*/
|
|
" dsra %M0, %L0, 32 \n"
|
|
" sll %L0, %L0, 0 \n"
|
|
/*
|
|
* Compare ret against old, breaking out of the loop if they don't
|
|
* match.
|
|
*/
|
|
" bne %M0, %M4, 2f \n"
|
|
" bne %L0, %L4, 2f \n"
|
|
/*
|
|
* Combine the 32 bit halves from the 2 registers that hold the new
|
|
* variable into a single 64 bit register.
|
|
*/
|
|
# if MIPS_ISA_REV >= 2
|
|
" move %L1, %L5 \n"
|
|
" dins %L1, %M5, 32, 32 \n"
|
|
# else
|
|
" dsll %L1, %L5, 32 \n"
|
|
" dsrl %L1, %L1, 32 \n"
|
|
" .set noat \n"
|
|
" dsll $at, %M5, 32 \n"
|
|
" or %L1, %L1, $at \n"
|
|
" .set at \n"
|
|
# endif
|
|
/* Attempt to store new at ptr */
|
|
" scd %L1, %2 \n"
|
|
/* If we failed, loop! */
|
|
"\t" __scbeqz " %L1, 1b \n"
|
|
" .set pop \n"
|
|
"2: \n"
|
|
: "=&r"(ret),
|
|
"=&r"(tmp),
|
|
"=" GCC_OFF_SMALL_ASM() (*(unsigned long long *)ptr)
|
|
: GCC_OFF_SMALL_ASM() (*(unsigned long long *)ptr),
|
|
"r" (old),
|
|
"r" (new)
|
|
: "memory");
|
|
loongson_llsc_mb();
|
|
|
|
local_irq_restore(flags);
|
|
return ret;
|
|
}
|
|
|
|
# define cmpxchg64(ptr, o, n) ({ \
|
|
unsigned long long __old = (__typeof__(*(ptr)))(o); \
|
|
unsigned long long __new = (__typeof__(*(ptr)))(n); \
|
|
__typeof__(*(ptr)) __res; \
|
|
\
|
|
/* \
|
|
* We can only use cmpxchg64 if we know that the CPU supports \
|
|
* 64-bits, ie. lld & scd. Our call to __cmpxchg64_unsupported \
|
|
* will cause a build error unless cpu_has_64bits is a \
|
|
* compile-time constant 1. \
|
|
*/ \
|
|
if (cpu_has_64bits && kernel_uses_llsc) { \
|
|
smp_mb__before_llsc(); \
|
|
__res = __cmpxchg64((ptr), __old, __new); \
|
|
smp_llsc_mb(); \
|
|
} else { \
|
|
__res = __cmpxchg64_unsupported(); \
|
|
} \
|
|
\
|
|
__res; \
|
|
})
|
|
|
|
# else /* !CONFIG_SMP */
|
|
# define cmpxchg64(ptr, o, n) cmpxchg64_local((ptr), (o), (n))
|
|
# endif /* !CONFIG_SMP */
|
|
#endif /* !CONFIG_64BIT */
|
|
|
|
#undef __scbeqz
|
|
|
|
#endif /* __ASM_CMPXCHG_H */
|