freebsd-nq/sys/sparc64/include/atomic.h
Konstantin Belousov 8954a9a4e6 Add the atomic_thread_fence() family of functions with intent to
provide a semantic defined by the C11 fences with corresponding
memory_order.

atomic_thread_fence_acq() gives r | r, w, where r and w are read and
write accesses, and | denotes the fence itself.

atomic_thread_fence_rel() is r, w | w.

atomic_thread_fence_acq_rel() is the combination of the acquire and
release in single operation.  Note that reads after the acq+rel fence
could be made visible before writes preceeding the fence.

atomic_thread_fence_seq_cst() orders all accesses before/after the
fence, and the fence itself is globally ordered against other
sequentially consistent atomic operations.

Reviewed by:	alc
Discussed with:	bde
Sponsored by:	The FreeBSD Foundation
MFC after:	3 weeks
2015-07-08 18:12:24 +00:00

337 lines
9.5 KiB
C

/*-
* Copyright (c) 1998 Doug Rabson.
* Copyright (c) 2001 Jake Burkholder.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* from: FreeBSD: src/sys/i386/include/atomic.h,v 1.20 2001/02/11
* $FreeBSD$
*/
#ifndef _MACHINE_ATOMIC_H_
#define _MACHINE_ATOMIC_H_
#include <machine/cpufunc.h>
#define mb() __asm__ __volatile__ ("membar #MemIssue": : :"memory")
#define wmb() mb()
#define rmb() mb()
/* Userland needs different ASI's. */
#ifdef _KERNEL
#define __ASI_ATOMIC ASI_N
#else
#define __ASI_ATOMIC ASI_P
#endif
/*
* Various simple arithmetic on memory which is atomic in the presence
* of interrupts and multiple processors. See atomic(9) for details.
* Note that efficient hardware support exists only for the 32 and 64
* bit variants; the 8 and 16 bit versions are not provided and should
* not be used in MI code.
*
* This implementation takes advantage of the fact that the sparc64
* cas instruction is both a load and a store. The loop is often coded
* as follows:
*
* do {
* expect = *p;
* new = expect + 1;
* } while (cas(p, expect, new) != expect);
*
* which performs an unnnecessary load on each iteration that the cas
* operation fails. Modified as follows:
*
* expect = *p;
* for (;;) {
* new = expect + 1;
* result = cas(p, expect, new);
* if (result == expect)
* break;
* expect = result;
* }
*
* the return value of cas is used to avoid the extra reload.
*
* We only include a memory barrier in the rel variants as in total store
* order which we use for running the kernel and all of the userland atomic
* loads and stores behave as if the were followed by a membar with a mask
* of #LoadLoad | #LoadStore | #StoreStore. In order to be also sufficient
* for use of relaxed memory ordering, the atomic_cas() in the acq variants
* additionally would have to be followed by a membar #LoadLoad | #LoadStore.
* Due to the suggested assembly syntax of the membar operands containing a
* # character, they cannot be used in macros. The cmask and mmask bits thus
* are hard coded in machine/cpufunc.h and used here through macros.
* Hopefully the bit numbers won't change in the future.
*/
#define itype(sz) uint ## sz ## _t
#define atomic_cas_32(p, e, s) casa((p), (e), (s), __ASI_ATOMIC)
#define atomic_cas_64(p, e, s) casxa((p), (e), (s), __ASI_ATOMIC)
#define atomic_cas(p, e, s, sz) \
atomic_cas_ ## sz((p), (e), (s))
#define atomic_cas_acq(p, e, s, sz) ({ \
itype(sz) v; \
v = atomic_cas((p), (e), (s), sz); \
__compiler_membar(); \
v; \
})
#define atomic_cas_rel(p, e, s, sz) ({ \
itype(sz) v; \
membar(LoadStore | StoreStore); \
v = atomic_cas((p), (e), (s), sz); \
v; \
})
#define atomic_op(p, op, v, sz) ({ \
itype(sz) e, r, s; \
for (e = *(volatile itype(sz) *)(p);; e = r) { \
s = e op (v); \
r = atomic_cas_ ## sz((p), e, s); \
if (r == e) \
break; \
} \
e; \
})
#define atomic_op_acq(p, op, v, sz) ({ \
itype(sz) t; \
t = atomic_op((p), op, (v), sz); \
__compiler_membar(); \
t; \
})
#define atomic_op_rel(p, op, v, sz) ({ \
itype(sz) t; \
membar(LoadStore | StoreStore); \
t = atomic_op((p), op, (v), sz); \
t; \
})
#define atomic_ld_acq(p, sz) ({ \
itype(sz) v; \
v = atomic_cas((p), 0, 0, sz); \
__compiler_membar(); \
v; \
})
#define atomic_ld_clear(p, sz) ({ \
itype(sz) e, r; \
for (e = *(volatile itype(sz) *)(p);; e = r) { \
r = atomic_cas((p), e, 0, sz); \
if (r == e) \
break; \
} \
e; \
})
#define atomic_st(p, v, sz) do { \
itype(sz) e, r; \
for (e = *(volatile itype(sz) *)(p);; e = r) { \
r = atomic_cas((p), e, (v), sz); \
if (r == e) \
break; \
} \
} while (0)
#define atomic_st_acq(p, v, sz) do { \
atomic_st((p), (v), sz); \
__compiler_membar(); \
} while (0)
#define atomic_st_rel(p, v, sz) do { \
membar(LoadStore | StoreStore); \
atomic_st((p), (v), sz); \
} while (0)
#define ATOMIC_GEN(name, ptype, vtype, atype, sz) \
\
static __inline vtype \
atomic_add_ ## name(volatile ptype p, atype v) \
{ \
return ((vtype)atomic_op((p), +, (v), sz)); \
} \
static __inline vtype \
atomic_add_acq_ ## name(volatile ptype p, atype v) \
{ \
return ((vtype)atomic_op_acq((p), +, (v), sz)); \
} \
static __inline vtype \
atomic_add_rel_ ## name(volatile ptype p, atype v) \
{ \
return ((vtype)atomic_op_rel((p), +, (v), sz)); \
} \
\
static __inline vtype \
atomic_clear_ ## name(volatile ptype p, atype v) \
{ \
return ((vtype)atomic_op((p), &, ~(v), sz)); \
} \
static __inline vtype \
atomic_clear_acq_ ## name(volatile ptype p, atype v) \
{ \
return ((vtype)atomic_op_acq((p), &, ~(v), sz)); \
} \
static __inline vtype \
atomic_clear_rel_ ## name(volatile ptype p, atype v) \
{ \
return ((vtype)atomic_op_rel((p), &, ~(v), sz)); \
} \
\
static __inline int \
atomic_cmpset_ ## name(volatile ptype p, vtype e, vtype s) \
{ \
return (((vtype)atomic_cas((p), (e), (s), sz)) == (e)); \
} \
static __inline int \
atomic_cmpset_acq_ ## name(volatile ptype p, vtype e, vtype s) \
{ \
return (((vtype)atomic_cas_acq((p), (e), (s), sz)) == (e)); \
} \
static __inline int \
atomic_cmpset_rel_ ## name(volatile ptype p, vtype e, vtype s) \
{ \
return (((vtype)atomic_cas_rel((p), (e), (s), sz)) == (e)); \
} \
\
static __inline vtype \
atomic_load_ ## name(volatile ptype p) \
{ \
return ((vtype)atomic_cas((p), 0, 0, sz)); \
} \
static __inline vtype \
atomic_load_acq_ ## name(volatile ptype p) \
{ \
return ((vtype)atomic_cas_acq((p), 0, 0, sz)); \
} \
\
static __inline vtype \
atomic_readandclear_ ## name(volatile ptype p) \
{ \
return ((vtype)atomic_ld_clear((p), sz)); \
} \
\
static __inline vtype \
atomic_set_ ## name(volatile ptype p, atype v) \
{ \
return ((vtype)atomic_op((p), |, (v), sz)); \
} \
static __inline vtype \
atomic_set_acq_ ## name(volatile ptype p, atype v) \
{ \
return ((vtype)atomic_op_acq((p), |, (v), sz)); \
} \
static __inline vtype \
atomic_set_rel_ ## name(volatile ptype p, atype v) \
{ \
return ((vtype)atomic_op_rel((p), |, (v), sz)); \
} \
\
static __inline vtype \
atomic_subtract_ ## name(volatile ptype p, atype v) \
{ \
return ((vtype)atomic_op((p), -, (v), sz)); \
} \
static __inline vtype \
atomic_subtract_acq_ ## name(volatile ptype p, atype v) \
{ \
return ((vtype)atomic_op_acq((p), -, (v), sz)); \
} \
static __inline vtype \
atomic_subtract_rel_ ## name(volatile ptype p, atype v) \
{ \
return ((vtype)atomic_op_rel((p), -, (v), sz)); \
} \
\
static __inline void \
atomic_store_acq_ ## name(volatile ptype p, vtype v) \
{ \
atomic_st_acq((p), (v), sz); \
} \
static __inline void \
atomic_store_rel_ ## name(volatile ptype p, vtype v) \
{ \
atomic_st_rel((p), (v), sz); \
}
static __inline void
atomic_thread_fence_acq(void)
{
__compiler_membar();
}
static __inline void
atomic_thread_fence_rel(void)
{
__compiler_membar();
}
static __inline void
atomic_thread_fence_acq_rel(void)
{
__compiler_membar();
}
static __inline void
atomic_thread_fence_seq_cst(void)
{
membar(LoadLoad | LoadStore | StoreStore | StoreLoad);
}
ATOMIC_GEN(int, u_int *, u_int, u_int, 32);
ATOMIC_GEN(32, uint32_t *, uint32_t, uint32_t, 32);
ATOMIC_GEN(long, u_long *, u_long, u_long, 64);
ATOMIC_GEN(64, uint64_t *, uint64_t, uint64_t, 64);
ATOMIC_GEN(ptr, uintptr_t *, uintptr_t, uintptr_t, 64);
#define atomic_fetchadd_int atomic_add_int
#define atomic_fetchadd_32 atomic_add_32
#define atomic_fetchadd_long atomic_add_long
#undef ATOMIC_GEN
#undef atomic_cas
#undef atomic_cas_acq
#undef atomic_cas_rel
#undef atomic_op
#undef atomic_op_acq
#undef atomic_op_rel
#undef atomic_ld_acq
#undef atomic_ld_clear
#undef atomic_st
#undef atomic_st_acq
#undef atomic_st_rel
#endif /* !_MACHINE_ATOMIC_H_ */