Index: head/share/man/man9/atomic.9 =================================================================== --- head/share/man/man9/atomic.9 (revision 315717) +++ head/share/man/man9/atomic.9 (revision 315718) @@ -1,501 +1,501 @@ .\" Copyright (c) 2000-2001 John H. Baldwin .\" 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 DEVELOPERS ``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 DEVELOPERS 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. .\" .\" $FreeBSD$ .\" -.Dd Jan 3, 2017 +.Dd March 22, 2017 .Dt ATOMIC 9 .Os .Sh NAME .Nm atomic_add , .Nm atomic_clear , .Nm atomic_cmpset , .Nm atomic_fcmpset , .Nm atomic_fetchadd , .Nm atomic_load , .Nm atomic_readandclear , .Nm atomic_set , .Nm atomic_subtract , .Nm atomic_store .Nd atomic operations .Sh SYNOPSIS .In sys/types.h .In machine/atomic.h .Ft void .Fn atomic_add_[acq_|rel_] "volatile *p" " v" .Ft void .Fn atomic_clear_[acq_|rel_] "volatile *p" " v" .Ft int .Fo atomic_cmpset_[acq_|rel_] .Fa "volatile *dst" .Fa " old" .Fa " new" .Fc .Ft int .Fo atomic_fcmpset_[acq_|rel_] .Fa "volatile *dst" .Fa " *old" .Fa " new" .Fc .Ft .Fn atomic_fetchadd_ "volatile *p" " v" .Ft .Fn atomic_load_acq_ "volatile *p" .Ft .Fn atomic_readandclear_ "volatile *p" .Ft void .Fn atomic_set_[acq_|rel_] "volatile *p" " v" .Ft void .Fn atomic_subtract_[acq_|rel_] "volatile *p" " v" .Ft void .Fn atomic_store_rel_ "volatile *p" " v" .Ft .Fn atomic_swap_ "volatile *p" " v" .Ft int .Fn atomic_testandclear_ "volatile *p" "u_int v" .Ft int .Fn atomic_testandset_ "volatile *p" "u_int v" .Sh DESCRIPTION Each of the atomic operations is guaranteed to be atomic across multiple threads and in the presence of interrupts. They can be used to implement reference counts or as building blocks for more advanced synchronization primitives such as mutexes. .Ss Types Each atomic operation operates on a specific .Fa type . The type to use is indicated in the function name. The available types that can be used are: .Pp .Bl -tag -offset indent -width short -compact .It Li int unsigned integer .It Li long unsigned long integer .It Li ptr unsigned integer the size of a pointer .It Li 32 unsigned 32-bit integer .It Li 64 unsigned 64-bit integer .El .Pp For example, the function to atomically add two integers is called .Fn atomic_add_int . .Pp Certain architectures also provide operations for types smaller than .Dq Li int . .Pp .Bl -tag -offset indent -width short -compact .It Li char unsigned character .It Li short unsigned short integer .It Li 8 unsigned 8-bit integer .It Li 16 unsigned 16-bit integer .El .Pp These must not be used in MI code because the instructions to implement them efficiently might not be available. .Ss Acquire and Release Operations By default, a thread's accesses to different memory locations might not be performed in .Em program order , that is, the order in which the accesses appear in the source code. To optimize the program's execution, both the compiler and processor might reorder the thread's accesses. However, both ensure that their reordering of the accesses is not visible to the thread. Otherwise, the traditional memory model that is expected by single-threaded programs would be violated. Nonetheless, other threads in a multithreaded program, such as the .Fx kernel, might observe the reordering. Moreover, in some cases, such as the implementation of synchronization between threads, arbitrary reordering might result in the incorrect execution of the program. To constrain the reordering that both the compiler and processor might perform on a thread's accesses, the thread should use atomic operations with .Em acquire and .Em release semantics. .Pp Most of the atomic operations on memory have three variants. The first variant performs the operation without imposing any ordering constraints on memory accesses to other locations. The second variant has acquire semantics, and the third variant has release semantics. In effect, operations with acquire and release semantics establish one-way barriers to reordering. .Pp When an atomic operation has acquire semantics, the effects of the operation must have completed before any subsequent load or store (by program order) is performed. Conversely, acquire semantics do not require that prior loads or stores have completed before the atomic operation is performed. To denote acquire semantics, the suffix .Dq Li _acq is inserted into the function name immediately prior to the .Dq Li _ Ns Aq Fa type suffix. For example, to subtract two integers ensuring that subsequent loads and stores happen after the subtraction is performed, use .Fn atomic_subtract_acq_int . .Pp When an atomic operation has release semantics, the effects of all prior loads or stores (by program order) must have completed before the operation is performed. Conversely, release semantics do not require that the effects of the atomic operation must have completed before any subsequent load or store is performed. To denote release semantics, the suffix .Dq Li _rel is inserted into the function name immediately prior to the .Dq Li _ Ns Aq Fa type suffix. For example, to add two long integers ensuring that all prior loads and stores happen before the addition, use .Fn atomic_add_rel_long . .Pp The one-way barriers provided by acquire and release operations allow the implementations of common synchronization primitives to express their ordering requirements without also imposing unnecessary ordering. For example, for a critical section guarded by a mutex, an acquire operation when the mutex is locked and a release operation when the mutex is unlocked will prevent any loads or stores from moving outside of the critical section. However, they will not prevent the compiler or processor from moving loads or stores into the critical section, which does not violate the semantics of a mutex. .Ss Multiple Processors In multiprocessor systems, the atomicity of the atomic operations on memory depends on support for cache coherence in the underlying architecture. In general, cache coherence on the default memory type, .Dv VM_MEMATTR_DEFAULT , is guaranteed by all architectures that are supported by .Fx . For example, cache coherence is guaranteed on write-back memory by the .Tn amd64 and .Tn i386 architectures. However, on some architectures, cache coherence might not be enabled on all memory types. To determine if cache coherence is enabled for a non-default memory type, consult the architecture's documentation. .Ss Semantics This section describes the semantics of each operation using a C like notation. .Bl -hang .It Fn atomic_add p v .Bd -literal -compact *p += v; .Ed .It Fn atomic_clear p v .Bd -literal -compact *p &= ~v; .Ed .It Fn atomic_cmpset dst old new .Bd -literal -compact if (*dst == old) { *dst = new; return (1); } else return (0); .Ed .El .Pp -The +Some architectures do not implement the .Fn atomic_cmpset -functions are not implemented for the types +functions for the types .Dq Li char , .Dq Li short , .Dq Li 8 , and .Dq Li 16 . .Bl -hang .It Fn atomic_fcmpset dst *old new .El .Pp On architectures implementing .Em Compare And Swap operation in hardware, the functionality can be described as .Bd -literal -offset indent -compact if (*dst == *old) { *dst = new; return (1); } else { *old = *dst; return (0); } .Ed On architectures which provide .Em Load Linked/Store Conditional primitive, the write to .Dv *dst might also fail for several reasons, most important of which is a parallel write to .Dv *dst cache line by other CPU. In this case .Fn atomic_fcmpset function also returns .Dv false , despite .Dl *old == *dst . .Pp -The +Some architectures do not implement the .Fn atomic_fcmpset -functions are not implemented for the types +functions for the types .Dq Li char , .Dq Li short , .Dq Li 8 , and .Dq Li 16 . .Bl -hang .It Fn atomic_fetchadd p v .Bd -literal -compact tmp = *p; *p += v; return (tmp); .Ed .El .Pp The .Fn atomic_fetchadd functions are only implemented for the types .Dq Li int , .Dq Li long and .Dq Li 32 and do not have any variants with memory barriers at this time. .Bl -hang .It Fn atomic_load p .Bd -literal -compact return (*p); .Ed .El .Pp The .Fn atomic_load functions are only provided with acquire memory barriers. .Bl -hang .It Fn atomic_readandclear p .Bd -literal -compact tmp = *p; *p = 0; return (tmp); .Ed .El .Pp The .Fn atomic_readandclear functions are not implemented for the types .Dq Li char , .Dq Li short , .Dq Li ptr , .Dq Li 8 , and .Dq Li 16 and do not have any variants with memory barriers at this time. .Bl -hang .It Fn atomic_set p v .Bd -literal -compact *p |= v; .Ed .It Fn atomic_subtract p v .Bd -literal -compact *p -= v; .Ed .It Fn atomic_store p v .Bd -literal -compact *p = v; .Ed .El .Pp The .Fn atomic_store functions are only provided with release memory barriers. .Bl -hang .It Fn atomic_swap p v .Bd -literal -compact tmp = *p; *p = v; return (tmp); .Ed .El .Pp The .Fn atomic_swap functions are not implemented for the types .Dq Li char , .Dq Li short , .Dq Li ptr , .Dq Li 8 , and .Dq Li 16 and do not have any variants with memory barriers at this time. .Bl -hang .It Fn atomic_testandclear p v .Bd -literal -compact bit = 1 << (v % (sizeof(*p) * NBBY)); tmp = (*p & bit) != 0; *p &= ~bit; return (tmp); .Ed .El .Bl -hang .It Fn atomic_testandset p v .Bd -literal -compact bit = 1 << (v % (sizeof(*p) * NBBY)); tmp = (*p & bit) != 0; *p |= bit; return (tmp); .Ed .El .Pp The .Fn atomic_testandset and .Fn atomic_testandclear functions are only implemented for the types .Dq Li int , .Dq Li long and .Dq Li 32 and do not have any variants with memory barriers at this time. .Pp The type .Dq Li 64 is currently not implemented for any of the atomic operations on the .Tn arm , .Tn i386 , and .Tn powerpc architectures. .Sh RETURN VALUES The .Fn atomic_cmpset function returns the result of the compare operation. The .Fn atomic_fcmpset function returns .Dv true if the operation succeeded. Otherwise it returns .Dv false and sets .Va *old to the found value. The .Fn atomic_fetchadd , .Fn atomic_load , .Fn atomic_readandclear , and .Fn atomic_swap functions return the value at the specified address. The .Fn atomic_testandset and .Fn atomic_testandclear function returns the result of the test operation. .Sh EXAMPLES This example uses the .Fn atomic_cmpset_acq_ptr and .Fn atomic_set_ptr functions to obtain a sleep mutex and handle recursion. Since the .Va mtx_lock member of a .Vt "struct mtx" is a pointer, the .Dq Li ptr type is used. .Bd -literal /* Try to obtain mtx_lock once. */ #define _obtain_lock(mp, tid) \\ atomic_cmpset_acq_ptr(&(mp)->mtx_lock, MTX_UNOWNED, (tid)) /* Get a sleep lock, deal with recursion inline. */ #define _get_sleep_lock(mp, tid, opts, file, line) do { \\ uintptr_t _tid = (uintptr_t)(tid); \\ \\ if (!_obtain_lock(mp, tid)) { \\ if (((mp)->mtx_lock & MTX_FLAGMASK) != _tid) \\ _mtx_lock_sleep((mp), _tid, (opts), (file), (line));\\ else { \\ atomic_set_ptr(&(mp)->mtx_lock, MTX_RECURSE); \\ (mp)->mtx_recurse++; \\ } \\ } \\ } while (0) .Ed .Sh HISTORY The .Fn atomic_add , .Fn atomic_clear , .Fn atomic_set , and .Fn atomic_subtract operations were first introduced in .Fx 3.0 . This first set only supported the types .Dq Li char , .Dq Li short , .Dq Li int , and .Dq Li long . The .Fn atomic_cmpset , .Fn atomic_load , .Fn atomic_readandclear , and .Fn atomic_store operations were added in .Fx 5.0 . The types .Dq Li 8 , .Dq Li 16 , .Dq Li 32 , .Dq Li 64 , and .Dq Li ptr and all of the acquire and release variants were added in .Fx 5.0 as well. The .Fn atomic_fetchadd operations were added in .Fx 6.0 . The .Fn atomic_swap and .Fn atomic_testandset operations were added in .Fx 10.0 . .Fn atomic_testandclear operation was added in .Fx 11.0 . Index: head/sys/amd64/include/atomic.h =================================================================== --- head/sys/amd64/include/atomic.h (revision 315717) +++ head/sys/amd64/include/atomic.h (revision 315718) @@ -1,676 +1,678 @@ /*- * Copyright (c) 1998 Doug Rabson * 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. * * $FreeBSD$ */ #ifndef _MACHINE_ATOMIC_H_ #define _MACHINE_ATOMIC_H_ #ifndef _SYS_CDEFS_H_ #error this file needs sys/cdefs.h as a prerequisite #endif /* * To express interprocessor (as opposed to processor and device) memory * ordering constraints, use the atomic_*() functions with acquire and release * semantics rather than the *mb() functions. An architecture's memory * ordering (or memory consistency) model governs the order in which a * program's accesses to different locations may be performed by an * implementation of that architecture. In general, for memory regions * defined as writeback cacheable, the memory ordering implemented by amd64 * processors preserves the program ordering of a load followed by a load, a * load followed by a store, and a store followed by a store. Only a store * followed by a load to a different memory location may be reordered. * Therefore, except for special cases, like non-temporal memory accesses or * memory regions defined as write combining, the memory ordering effects * provided by the sfence instruction in the wmb() function and the lfence * instruction in the rmb() function are redundant. In contrast, the * atomic_*() functions with acquire and release semantics do not perform * redundant instructions for ordinary cases of interprocessor memory * ordering on any architecture. */ #define mb() __asm __volatile("mfence;" : : : "memory") #define wmb() __asm __volatile("sfence;" : : : "memory") #define rmb() __asm __volatile("lfence;" : : : "memory") /* * Various simple operations on memory, each of which is atomic in the * presence of interrupts and multiple processors. * * atomic_set_char(P, V) (*(u_char *)(P) |= (V)) * atomic_clear_char(P, V) (*(u_char *)(P) &= ~(V)) * atomic_add_char(P, V) (*(u_char *)(P) += (V)) * atomic_subtract_char(P, V) (*(u_char *)(P) -= (V)) * * atomic_set_short(P, V) (*(u_short *)(P) |= (V)) * atomic_clear_short(P, V) (*(u_short *)(P) &= ~(V)) * atomic_add_short(P, V) (*(u_short *)(P) += (V)) * atomic_subtract_short(P, V) (*(u_short *)(P) -= (V)) * * atomic_set_int(P, V) (*(u_int *)(P) |= (V)) * atomic_clear_int(P, V) (*(u_int *)(P) &= ~(V)) * atomic_add_int(P, V) (*(u_int *)(P) += (V)) * atomic_subtract_int(P, V) (*(u_int *)(P) -= (V)) * atomic_swap_int(P, V) (return (*(u_int *)(P)); *(u_int *)(P) = (V);) * atomic_readandclear_int(P) (return (*(u_int *)(P)); *(u_int *)(P) = 0;) * * atomic_set_long(P, V) (*(u_long *)(P) |= (V)) * atomic_clear_long(P, V) (*(u_long *)(P) &= ~(V)) * atomic_add_long(P, V) (*(u_long *)(P) += (V)) * atomic_subtract_long(P, V) (*(u_long *)(P) -= (V)) * atomic_swap_long(P, V) (return (*(u_long *)(P)); *(u_long *)(P) = (V);) * atomic_readandclear_long(P) (return (*(u_long *)(P)); *(u_long *)(P) = 0;) */ /* * The above functions are expanded inline in the statically-linked * kernel. Lock prefixes are generated if an SMP kernel is being * built. * * Kernel modules call real functions which are built into the kernel. * This allows kernel modules to be portable between UP and SMP systems. */ #if defined(KLD_MODULE) || !defined(__GNUCLIKE_ASM) #define ATOMIC_ASM(NAME, TYPE, OP, CONS, V) \ void atomic_##NAME##_##TYPE(volatile u_##TYPE *p, u_##TYPE v); \ void atomic_##NAME##_barr_##TYPE(volatile u_##TYPE *p, u_##TYPE v) +int atomic_cmpset_char(volatile u_char *dst, u_char expect, u_char src); +int atomic_cmpset_short(volatile u_short *dst, u_short expect, u_short src); int atomic_cmpset_int(volatile u_int *dst, u_int expect, u_int src); int atomic_cmpset_long(volatile u_long *dst, u_long expect, u_long src); +int atomic_fcmpset_char(volatile u_char *dst, u_char *expect, u_char src); +int atomic_fcmpset_short(volatile u_short *dst, u_short *expect, + u_short src); int atomic_fcmpset_int(volatile u_int *dst, u_int *expect, u_int src); int atomic_fcmpset_long(volatile u_long *dst, u_long *expect, u_long src); u_int atomic_fetchadd_int(volatile u_int *p, u_int v); u_long atomic_fetchadd_long(volatile u_long *p, u_long v); int atomic_testandset_int(volatile u_int *p, u_int v); int atomic_testandset_long(volatile u_long *p, u_int v); int atomic_testandclear_int(volatile u_int *p, u_int v); int atomic_testandclear_long(volatile u_long *p, u_int v); void atomic_thread_fence_acq(void); void atomic_thread_fence_acq_rel(void); void atomic_thread_fence_rel(void); void atomic_thread_fence_seq_cst(void); #define ATOMIC_LOAD(TYPE) \ u_##TYPE atomic_load_acq_##TYPE(volatile u_##TYPE *p) #define ATOMIC_STORE(TYPE) \ void atomic_store_rel_##TYPE(volatile u_##TYPE *p, u_##TYPE v) #else /* !KLD_MODULE && __GNUCLIKE_ASM */ /* * For userland, always use lock prefixes so that the binaries will run * on both SMP and !SMP systems. */ #if defined(SMP) || !defined(_KERNEL) #define MPLOCKED "lock ; " #else #define MPLOCKED #endif /* * The assembly is volatilized to avoid code chunk removal by the compiler. * GCC aggressively reorders operations and memory clobbering is necessary * in order to avoid that for memory barriers. */ #define ATOMIC_ASM(NAME, TYPE, OP, CONS, V) \ static __inline void \ atomic_##NAME##_##TYPE(volatile u_##TYPE *p, u_##TYPE v)\ { \ __asm __volatile(MPLOCKED OP \ : "+m" (*p) \ : CONS (V) \ : "cc"); \ } \ \ static __inline void \ atomic_##NAME##_barr_##TYPE(volatile u_##TYPE *p, u_##TYPE v)\ { \ __asm __volatile(MPLOCKED OP \ : "+m" (*p) \ : CONS (V) \ : "memory", "cc"); \ } \ struct __hack /* - * Atomic compare and set, used by the mutex functions + * Atomic compare and set, used by the mutex functions. * - * if (*dst == expect) *dst = src (all 32 bit words) + * cmpset: + * if (*dst == expect) + * *dst = src * - * Returns 0 on failure, non-zero on success + * fcmpset: + * if (*dst == *expect) + * *dst = src + * else + * *expect = *dst + * + * Returns 0 on failure, non-zero on success. */ - -static __inline int -atomic_cmpset_int(volatile u_int *dst, u_int expect, u_int src) -{ - u_char res; - - __asm __volatile( - " " MPLOCKED " " - " cmpxchgl %3,%1 ; " - " sete %0 ; " - "# atomic_cmpset_int" - : "=q" (res), /* 0 */ - "+m" (*dst), /* 1 */ - "+a" (expect) /* 2 */ - : "r" (src) /* 3 */ - : "memory", "cc"); - return (res); +#define ATOMIC_CMPSET(TYPE) \ +static __inline int \ +atomic_cmpset_##TYPE(volatile u_##TYPE *dst, u_##TYPE expect, u_##TYPE src) \ +{ \ + u_char res; \ + \ + __asm __volatile( \ + " " MPLOCKED " " \ + " cmpxchg %3,%1 ; " \ + " sete %0 ; " \ + "# atomic_cmpset_" #TYPE " " \ + : "=q" (res), /* 0 */ \ + "+m" (*dst), /* 1 */ \ + "+a" (expect) /* 2 */ \ + : "r" (src) /* 3 */ \ + : "memory", "cc"); \ + return (res); \ +} \ + \ +static __inline int \ +atomic_fcmpset_##TYPE(volatile u_##TYPE *dst, u_##TYPE *expect, u_##TYPE src) \ +{ \ + u_char res; \ + \ + __asm __volatile( \ + " " MPLOCKED " " \ + " cmpxchg %3,%1 ; " \ + " sete %0 ; " \ + "# atomic_fcmpset_" #TYPE " " \ + : "=q" (res), /* 0 */ \ + "+m" (*dst), /* 1 */ \ + "+a" (*expect) /* 2 */ \ + : "r" (src) /* 3 */ \ + : "memory", "cc"); \ + return (res); \ } -static __inline int -atomic_cmpset_long(volatile u_long *dst, u_long expect, u_long src) -{ - u_char res; +ATOMIC_CMPSET(char); +ATOMIC_CMPSET(short); +ATOMIC_CMPSET(int); +ATOMIC_CMPSET(long); - __asm __volatile( - " " MPLOCKED " " - " cmpxchgq %3,%1 ; " - " sete %0 ; " - "# atomic_cmpset_long" - : "=q" (res), /* 0 */ - "+m" (*dst), /* 1 */ - "+a" (expect) /* 2 */ - : "r" (src) /* 3 */ - : "memory", "cc"); - return (res); -} - -static __inline int -atomic_fcmpset_int(volatile u_int *dst, u_int *expect, u_int src) -{ - u_char res; - - __asm __volatile( - " " MPLOCKED " " - " cmpxchgl %3,%1 ; " - " sete %0 ; " - "# atomic_fcmpset_int" - : "=r" (res), /* 0 */ - "+m" (*dst), /* 1 */ - "+a" (*expect) /* 2 */ - : "r" (src) /* 3 */ - : "memory", "cc"); - return (res); -} - -static __inline int -atomic_fcmpset_long(volatile u_long *dst, u_long *expect, u_long src) -{ - u_char res; - - __asm __volatile( - " " MPLOCKED " " - " cmpxchgq %3,%1 ; " - " sete %0 ; " - "# atomic_fcmpset_long" - : "=r" (res), /* 0 */ - "+m" (*dst), /* 1 */ - "+a" (*expect) /* 2 */ - : "r" (src) /* 3 */ - : "memory", "cc"); - return (res); -} - /* * Atomically add the value of v to the integer pointed to by p and return * the previous value of *p. */ static __inline u_int atomic_fetchadd_int(volatile u_int *p, u_int v) { __asm __volatile( " " MPLOCKED " " " xaddl %0,%1 ; " "# atomic_fetchadd_int" : "+r" (v), /* 0 */ "+m" (*p) /* 1 */ : : "cc"); return (v); } /* * Atomically add the value of v to the long integer pointed to by p and return * the previous value of *p. */ static __inline u_long atomic_fetchadd_long(volatile u_long *p, u_long v) { __asm __volatile( " " MPLOCKED " " " xaddq %0,%1 ; " "# atomic_fetchadd_long" : "+r" (v), /* 0 */ "+m" (*p) /* 1 */ : : "cc"); return (v); } static __inline int atomic_testandset_int(volatile u_int *p, u_int v) { u_char res; __asm __volatile( " " MPLOCKED " " " btsl %2,%1 ; " " setc %0 ; " "# atomic_testandset_int" : "=q" (res), /* 0 */ "+m" (*p) /* 1 */ : "Ir" (v & 0x1f) /* 2 */ : "cc"); return (res); } static __inline int atomic_testandset_long(volatile u_long *p, u_int v) { u_char res; __asm __volatile( " " MPLOCKED " " " btsq %2,%1 ; " " setc %0 ; " "# atomic_testandset_long" : "=q" (res), /* 0 */ "+m" (*p) /* 1 */ : "Jr" ((u_long)(v & 0x3f)) /* 2 */ : "cc"); return (res); } static __inline int atomic_testandclear_int(volatile u_int *p, u_int v) { u_char res; __asm __volatile( " " MPLOCKED " " " btrl %2,%1 ; " " setc %0 ; " "# atomic_testandclear_int" : "=q" (res), /* 0 */ "+m" (*p) /* 1 */ : "Ir" (v & 0x1f) /* 2 */ : "cc"); return (res); } static __inline int atomic_testandclear_long(volatile u_long *p, u_int v) { u_char res; __asm __volatile( " " MPLOCKED " " " btrq %2,%1 ; " " setc %0 ; " "# atomic_testandclear_long" : "=q" (res), /* 0 */ "+m" (*p) /* 1 */ : "Jr" ((u_long)(v & 0x3f)) /* 2 */ : "cc"); return (res); } /* * We assume that a = b will do atomic loads and stores. Due to the * IA32 memory model, a simple store guarantees release semantics. * * However, a load may pass a store if they are performed on distinct * addresses, so we need a Store/Load barrier for sequentially * consistent fences in SMP kernels. We use "lock addl $0,mem" for a * Store/Load barrier, as recommended by the AMD Software Optimization * Guide, and not mfence. To avoid false data dependencies, we use a * special address for "mem". In the kernel, we use a private per-cpu * cache line. In user space, we use a word in the stack's red zone * (-8(%rsp)). * * For UP kernels, however, the memory of the single processor is * always consistent, so we only need to stop the compiler from * reordering accesses in a way that violates the semantics of acquire * and release. */ #if defined(_KERNEL) /* * OFFSETOF_MONITORBUF == __pcpu_offset(pc_monitorbuf). * * The open-coded number is used instead of the symbolic expression to * avoid a dependency on sys/pcpu.h in machine/atomic.h consumers. * An assertion in amd64/vm_machdep.c ensures that the value is correct. */ #define OFFSETOF_MONITORBUF 0x180 #if defined(SMP) static __inline void __storeload_barrier(void) { __asm __volatile("lock; addl $0,%%gs:%0" : "+m" (*(u_int *)OFFSETOF_MONITORBUF) : : "memory", "cc"); } #else /* _KERNEL && UP */ static __inline void __storeload_barrier(void) { __compiler_membar(); } #endif /* SMP */ #else /* !_KERNEL */ static __inline void __storeload_barrier(void) { __asm __volatile("lock; addl $0,-8(%%rsp)" : : : "memory", "cc"); } #endif /* _KERNEL*/ #define ATOMIC_LOAD(TYPE) \ static __inline u_##TYPE \ atomic_load_acq_##TYPE(volatile u_##TYPE *p) \ { \ u_##TYPE res; \ \ res = *p; \ __compiler_membar(); \ return (res); \ } \ struct __hack #define ATOMIC_STORE(TYPE) \ static __inline void \ atomic_store_rel_##TYPE(volatile u_##TYPE *p, u_##TYPE v) \ { \ \ __compiler_membar(); \ *p = v; \ } \ struct __hack 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) { __storeload_barrier(); } #endif /* KLD_MODULE || !__GNUCLIKE_ASM */ ATOMIC_ASM(set, char, "orb %b1,%0", "iq", v); ATOMIC_ASM(clear, char, "andb %b1,%0", "iq", ~v); ATOMIC_ASM(add, char, "addb %b1,%0", "iq", v); ATOMIC_ASM(subtract, char, "subb %b1,%0", "iq", v); ATOMIC_ASM(set, short, "orw %w1,%0", "ir", v); ATOMIC_ASM(clear, short, "andw %w1,%0", "ir", ~v); ATOMIC_ASM(add, short, "addw %w1,%0", "ir", v); ATOMIC_ASM(subtract, short, "subw %w1,%0", "ir", v); ATOMIC_ASM(set, int, "orl %1,%0", "ir", v); ATOMIC_ASM(clear, int, "andl %1,%0", "ir", ~v); ATOMIC_ASM(add, int, "addl %1,%0", "ir", v); ATOMIC_ASM(subtract, int, "subl %1,%0", "ir", v); ATOMIC_ASM(set, long, "orq %1,%0", "ir", v); ATOMIC_ASM(clear, long, "andq %1,%0", "ir", ~v); ATOMIC_ASM(add, long, "addq %1,%0", "ir", v); ATOMIC_ASM(subtract, long, "subq %1,%0", "ir", v); #define ATOMIC_LOADSTORE(TYPE) \ ATOMIC_LOAD(TYPE); \ ATOMIC_STORE(TYPE) ATOMIC_LOADSTORE(char); ATOMIC_LOADSTORE(short); ATOMIC_LOADSTORE(int); ATOMIC_LOADSTORE(long); #undef ATOMIC_ASM #undef ATOMIC_LOAD #undef ATOMIC_STORE #undef ATOMIC_LOADSTORE #ifndef WANT_FUNCTIONS /* Read the current value and store a new value in the destination. */ #ifdef __GNUCLIKE_ASM static __inline u_int atomic_swap_int(volatile u_int *p, u_int v) { __asm __volatile( " xchgl %1,%0 ; " "# atomic_swap_int" : "+r" (v), /* 0 */ "+m" (*p)); /* 1 */ return (v); } static __inline u_long atomic_swap_long(volatile u_long *p, u_long v) { __asm __volatile( " xchgq %1,%0 ; " "# atomic_swap_long" : "+r" (v), /* 0 */ "+m" (*p)); /* 1 */ return (v); } #else /* !__GNUCLIKE_ASM */ u_int atomic_swap_int(volatile u_int *p, u_int v); u_long atomic_swap_long(volatile u_long *p, u_long v); #endif /* __GNUCLIKE_ASM */ #define atomic_set_acq_char atomic_set_barr_char #define atomic_set_rel_char atomic_set_barr_char #define atomic_clear_acq_char atomic_clear_barr_char #define atomic_clear_rel_char atomic_clear_barr_char #define atomic_add_acq_char atomic_add_barr_char #define atomic_add_rel_char atomic_add_barr_char #define atomic_subtract_acq_char atomic_subtract_barr_char #define atomic_subtract_rel_char atomic_subtract_barr_char +#define atomic_cmpset_acq_char atomic_cmpset_char +#define atomic_cmpset_rel_char atomic_cmpset_char +#define atomic_fcmpset_acq_char atomic_fcmpset_char +#define atomic_fcmpset_rel_char atomic_fcmpset_char #define atomic_set_acq_short atomic_set_barr_short #define atomic_set_rel_short atomic_set_barr_short #define atomic_clear_acq_short atomic_clear_barr_short #define atomic_clear_rel_short atomic_clear_barr_short #define atomic_add_acq_short atomic_add_barr_short #define atomic_add_rel_short atomic_add_barr_short #define atomic_subtract_acq_short atomic_subtract_barr_short #define atomic_subtract_rel_short atomic_subtract_barr_short +#define atomic_cmpset_acq_short atomic_cmpset_short +#define atomic_cmpset_rel_short atomic_cmpset_short +#define atomic_fcmpset_acq_short atomic_fcmpset_short +#define atomic_fcmpset_rel_short atomic_fcmpset_short #define atomic_set_acq_int atomic_set_barr_int #define atomic_set_rel_int atomic_set_barr_int #define atomic_clear_acq_int atomic_clear_barr_int #define atomic_clear_rel_int atomic_clear_barr_int #define atomic_add_acq_int atomic_add_barr_int #define atomic_add_rel_int atomic_add_barr_int #define atomic_subtract_acq_int atomic_subtract_barr_int #define atomic_subtract_rel_int atomic_subtract_barr_int #define atomic_cmpset_acq_int atomic_cmpset_int #define atomic_cmpset_rel_int atomic_cmpset_int -#define atomic_fcmpset_acq_int atomic_fcmpset_int -#define atomic_fcmpset_rel_int atomic_fcmpset_int +#define atomic_fcmpset_acq_int atomic_fcmpset_int +#define atomic_fcmpset_rel_int atomic_fcmpset_int #define atomic_set_acq_long atomic_set_barr_long #define atomic_set_rel_long atomic_set_barr_long #define atomic_clear_acq_long atomic_clear_barr_long #define atomic_clear_rel_long atomic_clear_barr_long #define atomic_add_acq_long atomic_add_barr_long #define atomic_add_rel_long atomic_add_barr_long #define atomic_subtract_acq_long atomic_subtract_barr_long #define atomic_subtract_rel_long atomic_subtract_barr_long #define atomic_cmpset_acq_long atomic_cmpset_long #define atomic_cmpset_rel_long atomic_cmpset_long -#define atomic_fcmpset_acq_long atomic_fcmpset_long -#define atomic_fcmpset_rel_long atomic_fcmpset_long +#define atomic_fcmpset_acq_long atomic_fcmpset_long +#define atomic_fcmpset_rel_long atomic_fcmpset_long #define atomic_readandclear_int(p) atomic_swap_int(p, 0) #define atomic_readandclear_long(p) atomic_swap_long(p, 0) /* Operations on 8-bit bytes. */ #define atomic_set_8 atomic_set_char #define atomic_set_acq_8 atomic_set_acq_char #define atomic_set_rel_8 atomic_set_rel_char #define atomic_clear_8 atomic_clear_char #define atomic_clear_acq_8 atomic_clear_acq_char #define atomic_clear_rel_8 atomic_clear_rel_char #define atomic_add_8 atomic_add_char #define atomic_add_acq_8 atomic_add_acq_char #define atomic_add_rel_8 atomic_add_rel_char #define atomic_subtract_8 atomic_subtract_char #define atomic_subtract_acq_8 atomic_subtract_acq_char #define atomic_subtract_rel_8 atomic_subtract_rel_char #define atomic_load_acq_8 atomic_load_acq_char #define atomic_store_rel_8 atomic_store_rel_char +#define atomic_cmpset_8 atomic_cmpset_char +#define atomic_cmpset_acq_8 atomic_cmpset_acq_char +#define atomic_cmpset_rel_8 atomic_cmpset_rel_char +#define atomic_fcmpset_8 atomic_fcmpset_char +#define atomic_fcmpset_acq_8 atomic_fcmpset_acq_char +#define atomic_fcmpset_rel_8 atomic_fcmpset_rel_char /* Operations on 16-bit words. */ #define atomic_set_16 atomic_set_short #define atomic_set_acq_16 atomic_set_acq_short #define atomic_set_rel_16 atomic_set_rel_short #define atomic_clear_16 atomic_clear_short #define atomic_clear_acq_16 atomic_clear_acq_short #define atomic_clear_rel_16 atomic_clear_rel_short #define atomic_add_16 atomic_add_short #define atomic_add_acq_16 atomic_add_acq_short #define atomic_add_rel_16 atomic_add_rel_short #define atomic_subtract_16 atomic_subtract_short #define atomic_subtract_acq_16 atomic_subtract_acq_short #define atomic_subtract_rel_16 atomic_subtract_rel_short #define atomic_load_acq_16 atomic_load_acq_short #define atomic_store_rel_16 atomic_store_rel_short +#define atomic_cmpset_16 atomic_cmpset_short +#define atomic_cmpset_acq_16 atomic_cmpset_acq_short +#define atomic_cmpset_rel_16 atomic_cmpset_rel_short +#define atomic_fcmpset_16 atomic_fcmpset_short +#define atomic_fcmpset_acq_16 atomic_fcmpset_acq_short +#define atomic_fcmpset_rel_16 atomic_fcmpset_rel_short /* Operations on 32-bit double words. */ #define atomic_set_32 atomic_set_int #define atomic_set_acq_32 atomic_set_acq_int #define atomic_set_rel_32 atomic_set_rel_int #define atomic_clear_32 atomic_clear_int #define atomic_clear_acq_32 atomic_clear_acq_int #define atomic_clear_rel_32 atomic_clear_rel_int #define atomic_add_32 atomic_add_int #define atomic_add_acq_32 atomic_add_acq_int #define atomic_add_rel_32 atomic_add_rel_int #define atomic_subtract_32 atomic_subtract_int #define atomic_subtract_acq_32 atomic_subtract_acq_int #define atomic_subtract_rel_32 atomic_subtract_rel_int #define atomic_load_acq_32 atomic_load_acq_int #define atomic_store_rel_32 atomic_store_rel_int #define atomic_cmpset_32 atomic_cmpset_int #define atomic_cmpset_acq_32 atomic_cmpset_acq_int #define atomic_cmpset_rel_32 atomic_cmpset_rel_int #define atomic_fcmpset_32 atomic_fcmpset_int #define atomic_fcmpset_acq_32 atomic_fcmpset_acq_int #define atomic_fcmpset_rel_32 atomic_fcmpset_rel_int #define atomic_swap_32 atomic_swap_int #define atomic_readandclear_32 atomic_readandclear_int #define atomic_fetchadd_32 atomic_fetchadd_int #define atomic_testandset_32 atomic_testandset_int #define atomic_testandclear_32 atomic_testandclear_int /* Operations on 64-bit quad words. */ #define atomic_set_64 atomic_set_long #define atomic_set_acq_64 atomic_set_acq_long #define atomic_set_rel_64 atomic_set_rel_long #define atomic_clear_64 atomic_clear_long #define atomic_clear_acq_64 atomic_clear_acq_long #define atomic_clear_rel_64 atomic_clear_rel_long #define atomic_add_64 atomic_add_long #define atomic_add_acq_64 atomic_add_acq_long #define atomic_add_rel_64 atomic_add_rel_long #define atomic_subtract_64 atomic_subtract_long #define atomic_subtract_acq_64 atomic_subtract_acq_long #define atomic_subtract_rel_64 atomic_subtract_rel_long #define atomic_load_acq_64 atomic_load_acq_long #define atomic_store_rel_64 atomic_store_rel_long #define atomic_cmpset_64 atomic_cmpset_long #define atomic_cmpset_acq_64 atomic_cmpset_acq_long #define atomic_cmpset_rel_64 atomic_cmpset_rel_long #define atomic_fcmpset_64 atomic_fcmpset_long #define atomic_fcmpset_acq_64 atomic_fcmpset_acq_long #define atomic_fcmpset_rel_64 atomic_fcmpset_rel_long #define atomic_swap_64 atomic_swap_long #define atomic_readandclear_64 atomic_readandclear_long #define atomic_fetchadd_64 atomic_fetchadd_long #define atomic_testandset_64 atomic_testandset_long #define atomic_testandclear_64 atomic_testandclear_long /* Operations on pointers. */ #define atomic_set_ptr atomic_set_long #define atomic_set_acq_ptr atomic_set_acq_long #define atomic_set_rel_ptr atomic_set_rel_long #define atomic_clear_ptr atomic_clear_long #define atomic_clear_acq_ptr atomic_clear_acq_long #define atomic_clear_rel_ptr atomic_clear_rel_long #define atomic_add_ptr atomic_add_long #define atomic_add_acq_ptr atomic_add_acq_long #define atomic_add_rel_ptr atomic_add_rel_long #define atomic_subtract_ptr atomic_subtract_long #define atomic_subtract_acq_ptr atomic_subtract_acq_long #define atomic_subtract_rel_ptr atomic_subtract_rel_long #define atomic_load_acq_ptr atomic_load_acq_long #define atomic_store_rel_ptr atomic_store_rel_long #define atomic_cmpset_ptr atomic_cmpset_long #define atomic_cmpset_acq_ptr atomic_cmpset_acq_long #define atomic_cmpset_rel_ptr atomic_cmpset_rel_long #define atomic_fcmpset_ptr atomic_fcmpset_long #define atomic_fcmpset_acq_ptr atomic_fcmpset_acq_long #define atomic_fcmpset_rel_ptr atomic_fcmpset_rel_long #define atomic_swap_ptr atomic_swap_long #define atomic_readandclear_ptr atomic_readandclear_long #endif /* !WANT_FUNCTIONS */ #endif /* !_MACHINE_ATOMIC_H_ */ Index: head/sys/i386/include/atomic.h =================================================================== --- head/sys/i386/include/atomic.h (revision 315717) +++ head/sys/i386/include/atomic.h (revision 315718) @@ -1,817 +1,854 @@ /*- * Copyright (c) 1998 Doug Rabson * 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. * * $FreeBSD$ */ #ifndef _MACHINE_ATOMIC_H_ #define _MACHINE_ATOMIC_H_ #ifndef _SYS_CDEFS_H_ #error this file needs sys/cdefs.h as a prerequisite #endif #ifdef _KERNEL #include #include #endif #ifndef __OFFSETOF_MONITORBUF /* * __OFFSETOF_MONITORBUF == __pcpu_offset(pc_monitorbuf). * * The open-coded number is used instead of the symbolic expression to * avoid a dependency on sys/pcpu.h in machine/atomic.h consumers. * An assertion in i386/vm_machdep.c ensures that the value is correct. */ #define __OFFSETOF_MONITORBUF 0x180 static __inline void __mbk(void) { __asm __volatile("lock; addl $0,%%fs:%0" : "+m" (*(u_int *)__OFFSETOF_MONITORBUF) : : "memory", "cc"); } static __inline void __mbu(void) { __asm __volatile("lock; addl $0,(%%esp)" : : : "memory", "cc"); } #endif /* * Various simple operations on memory, each of which is atomic in the * presence of interrupts and multiple processors. * * atomic_set_char(P, V) (*(u_char *)(P) |= (V)) * atomic_clear_char(P, V) (*(u_char *)(P) &= ~(V)) * atomic_add_char(P, V) (*(u_char *)(P) += (V)) * atomic_subtract_char(P, V) (*(u_char *)(P) -= (V)) * * atomic_set_short(P, V) (*(u_short *)(P) |= (V)) * atomic_clear_short(P, V) (*(u_short *)(P) &= ~(V)) * atomic_add_short(P, V) (*(u_short *)(P) += (V)) * atomic_subtract_short(P, V) (*(u_short *)(P) -= (V)) * * atomic_set_int(P, V) (*(u_int *)(P) |= (V)) * atomic_clear_int(P, V) (*(u_int *)(P) &= ~(V)) * atomic_add_int(P, V) (*(u_int *)(P) += (V)) * atomic_subtract_int(P, V) (*(u_int *)(P) -= (V)) * atomic_swap_int(P, V) (return (*(u_int *)(P)); *(u_int *)(P) = (V);) * atomic_readandclear_int(P) (return (*(u_int *)(P)); *(u_int *)(P) = 0;) * * atomic_set_long(P, V) (*(u_long *)(P) |= (V)) * atomic_clear_long(P, V) (*(u_long *)(P) &= ~(V)) * atomic_add_long(P, V) (*(u_long *)(P) += (V)) * atomic_subtract_long(P, V) (*(u_long *)(P) -= (V)) * atomic_swap_long(P, V) (return (*(u_long *)(P)); *(u_long *)(P) = (V);) * atomic_readandclear_long(P) (return (*(u_long *)(P)); *(u_long *)(P) = 0;) */ /* * The above functions are expanded inline in the statically-linked * kernel. Lock prefixes are generated if an SMP kernel is being * built. * * Kernel modules call real functions which are built into the kernel. * This allows kernel modules to be portable between UP and SMP systems. */ #if defined(KLD_MODULE) || !defined(__GNUCLIKE_ASM) #define ATOMIC_ASM(NAME, TYPE, OP, CONS, V) \ void atomic_##NAME##_##TYPE(volatile u_##TYPE *p, u_##TYPE v); \ void atomic_##NAME##_barr_##TYPE(volatile u_##TYPE *p, u_##TYPE v) +int atomic_cmpset_char(volatile u_char *dst, u_char expect, u_char src); +int atomic_cmpset_short(volatile u_short *dst, u_short expect, u_short src); int atomic_cmpset_int(volatile u_int *dst, u_int expect, u_int src); +int atomic_fcmpset_char(volatile u_char *dst, u_char *expect, u_char src); +int atomic_fcmpset_short(volatile u_short *dst, u_short *expect, + u_short src); int atomic_fcmpset_int(volatile u_int *dst, u_int *expect, u_int src); u_int atomic_fetchadd_int(volatile u_int *p, u_int v); int atomic_testandset_int(volatile u_int *p, u_int v); int atomic_testandclear_int(volatile u_int *p, u_int v); void atomic_thread_fence_acq(void); void atomic_thread_fence_acq_rel(void); void atomic_thread_fence_rel(void); void atomic_thread_fence_seq_cst(void); #define ATOMIC_LOAD(TYPE) \ u_##TYPE atomic_load_acq_##TYPE(volatile u_##TYPE *p) #define ATOMIC_STORE(TYPE) \ void atomic_store_rel_##TYPE(volatile u_##TYPE *p, u_##TYPE v) int atomic_cmpset_64(volatile uint64_t *, uint64_t, uint64_t); uint64_t atomic_load_acq_64(volatile uint64_t *); void atomic_store_rel_64(volatile uint64_t *, uint64_t); uint64_t atomic_swap_64(volatile uint64_t *, uint64_t); #else /* !KLD_MODULE && __GNUCLIKE_ASM */ /* * For userland, always use lock prefixes so that the binaries will run * on both SMP and !SMP systems. */ #if defined(SMP) || !defined(_KERNEL) #define MPLOCKED "lock ; " #else #define MPLOCKED #endif /* * The assembly is volatilized to avoid code chunk removal by the compiler. * GCC aggressively reorders operations and memory clobbering is necessary * in order to avoid that for memory barriers. */ #define ATOMIC_ASM(NAME, TYPE, OP, CONS, V) \ static __inline void \ atomic_##NAME##_##TYPE(volatile u_##TYPE *p, u_##TYPE v)\ { \ __asm __volatile(MPLOCKED OP \ : "+m" (*p) \ : CONS (V) \ : "cc"); \ } \ \ static __inline void \ atomic_##NAME##_barr_##TYPE(volatile u_##TYPE *p, u_##TYPE v)\ { \ __asm __volatile(MPLOCKED OP \ : "+m" (*p) \ : CONS (V) \ : "memory", "cc"); \ } \ struct __hack /* - * Atomic compare and set, used by the mutex functions + * Atomic compare and set, used by the mutex functions. * - * if (*dst == expect) *dst = src (all 32 bit words) + * cmpset: + * if (*dst == expect) + * *dst = src * - * Returns 0 on failure, non-zero on success + * fcmpset: + * if (*dst == *expect) + * *dst = src + * else + * *expect = *dst + * + * Returns 0 on failure, non-zero on success. */ - -static __inline int -atomic_cmpset_int(volatile u_int *dst, u_int expect, u_int src) -{ - u_char res; - - __asm __volatile( - " " MPLOCKED " " - " cmpxchgl %3,%1 ; " - " sete %0 ; " - "# atomic_cmpset_int" - : "=q" (res), /* 0 */ - "+m" (*dst), /* 1 */ - "+a" (expect) /* 2 */ - : "r" (src) /* 3 */ - : "memory", "cc"); - return (res); +#define ATOMIC_CMPSET(TYPE) \ +static __inline int \ +atomic_cmpset_##TYPE(volatile u_##TYPE *dst, u_##TYPE expect, u_##TYPE src) \ +{ \ + u_char res; \ + \ + __asm __volatile( \ + " " MPLOCKED " " \ + " cmpxchg %3,%1 ; " \ + " sete %0 ; " \ + "# atomic_cmpset_" #TYPE " " \ + : "=q" (res), /* 0 */ \ + "+m" (*dst), /* 1 */ \ + "+a" (expect) /* 2 */ \ + : "r" (src) /* 3 */ \ + : "memory", "cc"); \ + return (res); \ +} \ + \ +static __inline int \ +atomic_fcmpset_##TYPE(volatile u_##TYPE *dst, u_##TYPE *expect, u_##TYPE src) \ +{ \ + u_char res; \ + \ + __asm __volatile( \ + " " MPLOCKED " " \ + " cmpxchg %3,%1 ; " \ + " sete %0 ; " \ + "# atomic_fcmpset_" #TYPE " " \ + : "=q" (res), /* 0 */ \ + "+m" (*dst), /* 1 */ \ + "+a" (*expect) /* 2 */ \ + : "r" (src) /* 3 */ \ + : "memory", "cc"); \ + return (res); \ } -static __inline int -atomic_fcmpset_int(volatile u_int *dst, u_int *expect, u_int src) -{ - u_char res; +ATOMIC_CMPSET(char); +ATOMIC_CMPSET(short); +ATOMIC_CMPSET(int); - __asm __volatile( - " " MPLOCKED " " - " cmpxchgl %3,%1 ; " - " sete %0 ; " - "# atomic_cmpset_int" - : "=q" (res), /* 0 */ - "+m" (*dst), /* 1 */ - "+a" (*expect) /* 2 */ - : "r" (src) /* 3 */ - : "memory", "cc"); - return (res); -} - /* * Atomically add the value of v to the integer pointed to by p and return * the previous value of *p. */ static __inline u_int atomic_fetchadd_int(volatile u_int *p, u_int v) { __asm __volatile( " " MPLOCKED " " " xaddl %0,%1 ; " "# atomic_fetchadd_int" : "+r" (v), /* 0 */ "+m" (*p) /* 1 */ : : "cc"); return (v); } static __inline int atomic_testandset_int(volatile u_int *p, u_int v) { u_char res; __asm __volatile( " " MPLOCKED " " " btsl %2,%1 ; " " setc %0 ; " "# atomic_testandset_int" : "=q" (res), /* 0 */ "+m" (*p) /* 1 */ : "Ir" (v & 0x1f) /* 2 */ : "cc"); return (res); } static __inline int atomic_testandclear_int(volatile u_int *p, u_int v) { u_char res; __asm __volatile( " " MPLOCKED " " " btrl %2,%1 ; " " setc %0 ; " "# atomic_testandclear_int" : "=q" (res), /* 0 */ "+m" (*p) /* 1 */ : "Ir" (v & 0x1f) /* 2 */ : "cc"); return (res); } /* * We assume that a = b will do atomic loads and stores. Due to the * IA32 memory model, a simple store guarantees release semantics. * * However, a load may pass a store if they are performed on distinct * addresses, so we need Store/Load barrier for sequentially * consistent fences in SMP kernels. We use "lock addl $0,mem" for a * Store/Load barrier, as recommended by the AMD Software Optimization * Guide, and not mfence. In the kernel, we use a private per-cpu * cache line for "mem", to avoid introducing false data * dependencies. In user space, we use the word at the top of the * stack. * * For UP kernels, however, the memory of the single processor is * always consistent, so we only need to stop the compiler from * reordering accesses in a way that violates the semantics of acquire * and release. */ #if defined(_KERNEL) #if defined(SMP) #define __storeload_barrier() __mbk() #else /* _KERNEL && UP */ #define __storeload_barrier() __compiler_membar() #endif /* SMP */ #else /* !_KERNEL */ #define __storeload_barrier() __mbu() #endif /* _KERNEL*/ #define ATOMIC_LOAD(TYPE) \ static __inline u_##TYPE \ atomic_load_acq_##TYPE(volatile u_##TYPE *p) \ { \ u_##TYPE res; \ \ res = *p; \ __compiler_membar(); \ return (res); \ } \ struct __hack #define ATOMIC_STORE(TYPE) \ static __inline void \ atomic_store_rel_##TYPE(volatile u_##TYPE *p, u_##TYPE v) \ { \ \ __compiler_membar(); \ *p = v; \ } \ struct __hack 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) { __storeload_barrier(); } #ifdef _KERNEL #ifdef WANT_FUNCTIONS int atomic_cmpset_64_i386(volatile uint64_t *, uint64_t, uint64_t); int atomic_cmpset_64_i586(volatile uint64_t *, uint64_t, uint64_t); uint64_t atomic_load_acq_64_i386(volatile uint64_t *); uint64_t atomic_load_acq_64_i586(volatile uint64_t *); void atomic_store_rel_64_i386(volatile uint64_t *, uint64_t); void atomic_store_rel_64_i586(volatile uint64_t *, uint64_t); uint64_t atomic_swap_64_i386(volatile uint64_t *, uint64_t); uint64_t atomic_swap_64_i586(volatile uint64_t *, uint64_t); #endif /* I486 does not support SMP or CMPXCHG8B. */ static __inline int atomic_cmpset_64_i386(volatile uint64_t *dst, uint64_t expect, uint64_t src) { volatile uint32_t *p; u_char res; p = (volatile uint32_t *)dst; __asm __volatile( " pushfl ; " " cli ; " " xorl %1,%%eax ; " " xorl %2,%%edx ; " " orl %%edx,%%eax ; " " jne 1f ; " " movl %4,%1 ; " " movl %5,%2 ; " "1: " " sete %3 ; " " popfl" : "+A" (expect), /* 0 */ "+m" (*p), /* 1 */ "+m" (*(p + 1)), /* 2 */ "=q" (res) /* 3 */ : "r" ((uint32_t)src), /* 4 */ "r" ((uint32_t)(src >> 32)) /* 5 */ : "memory", "cc"); return (res); } static __inline uint64_t atomic_load_acq_64_i386(volatile uint64_t *p) { volatile uint32_t *q; uint64_t res; q = (volatile uint32_t *)p; __asm __volatile( " pushfl ; " " cli ; " " movl %1,%%eax ; " " movl %2,%%edx ; " " popfl" : "=&A" (res) /* 0 */ : "m" (*q), /* 1 */ "m" (*(q + 1)) /* 2 */ : "memory"); return (res); } static __inline void atomic_store_rel_64_i386(volatile uint64_t *p, uint64_t v) { volatile uint32_t *q; q = (volatile uint32_t *)p; __asm __volatile( " pushfl ; " " cli ; " " movl %%eax,%0 ; " " movl %%edx,%1 ; " " popfl" : "=m" (*q), /* 0 */ "=m" (*(q + 1)) /* 1 */ : "A" (v) /* 2 */ : "memory"); } static __inline uint64_t atomic_swap_64_i386(volatile uint64_t *p, uint64_t v) { volatile uint32_t *q; uint64_t res; q = (volatile uint32_t *)p; __asm __volatile( " pushfl ; " " cli ; " " movl %1,%%eax ; " " movl %2,%%edx ; " " movl %4,%2 ; " " movl %3,%1 ; " " popfl" : "=&A" (res), /* 0 */ "+m" (*q), /* 1 */ "+m" (*(q + 1)) /* 2 */ : "r" ((uint32_t)v), /* 3 */ "r" ((uint32_t)(v >> 32))); /* 4 */ return (res); } static __inline int atomic_cmpset_64_i586(volatile uint64_t *dst, uint64_t expect, uint64_t src) { u_char res; __asm __volatile( " " MPLOCKED " " " cmpxchg8b %1 ; " " sete %0" : "=q" (res), /* 0 */ "+m" (*dst), /* 1 */ "+A" (expect) /* 2 */ : "b" ((uint32_t)src), /* 3 */ "c" ((uint32_t)(src >> 32)) /* 4 */ : "memory", "cc"); return (res); } static __inline uint64_t atomic_load_acq_64_i586(volatile uint64_t *p) { uint64_t res; __asm __volatile( " movl %%ebx,%%eax ; " " movl %%ecx,%%edx ; " " " MPLOCKED " " " cmpxchg8b %1" : "=&A" (res), /* 0 */ "+m" (*p) /* 1 */ : : "memory", "cc"); return (res); } static __inline void atomic_store_rel_64_i586(volatile uint64_t *p, uint64_t v) { __asm __volatile( " movl %%eax,%%ebx ; " " movl %%edx,%%ecx ; " "1: " " " MPLOCKED " " " cmpxchg8b %0 ; " " jne 1b" : "+m" (*p), /* 0 */ "+A" (v) /* 1 */ : : "ebx", "ecx", "memory", "cc"); } static __inline uint64_t atomic_swap_64_i586(volatile uint64_t *p, uint64_t v) { __asm __volatile( " movl %%eax,%%ebx ; " " movl %%edx,%%ecx ; " "1: " " " MPLOCKED " " " cmpxchg8b %0 ; " " jne 1b" : "+m" (*p), /* 0 */ "+A" (v) /* 1 */ : : "ebx", "ecx", "memory", "cc"); return (v); } static __inline int atomic_cmpset_64(volatile uint64_t *dst, uint64_t expect, uint64_t src) { if ((cpu_feature & CPUID_CX8) == 0) return (atomic_cmpset_64_i386(dst, expect, src)); else return (atomic_cmpset_64_i586(dst, expect, src)); } static __inline uint64_t atomic_load_acq_64(volatile uint64_t *p) { if ((cpu_feature & CPUID_CX8) == 0) return (atomic_load_acq_64_i386(p)); else return (atomic_load_acq_64_i586(p)); } static __inline void atomic_store_rel_64(volatile uint64_t *p, uint64_t v) { if ((cpu_feature & CPUID_CX8) == 0) atomic_store_rel_64_i386(p, v); else atomic_store_rel_64_i586(p, v); } static __inline uint64_t atomic_swap_64(volatile uint64_t *p, uint64_t v) { if ((cpu_feature & CPUID_CX8) == 0) return (atomic_swap_64_i386(p, v)); else return (atomic_swap_64_i586(p, v)); } #endif /* _KERNEL */ #endif /* KLD_MODULE || !__GNUCLIKE_ASM */ ATOMIC_ASM(set, char, "orb %b1,%0", "iq", v); ATOMIC_ASM(clear, char, "andb %b1,%0", "iq", ~v); ATOMIC_ASM(add, char, "addb %b1,%0", "iq", v); ATOMIC_ASM(subtract, char, "subb %b1,%0", "iq", v); ATOMIC_ASM(set, short, "orw %w1,%0", "ir", v); ATOMIC_ASM(clear, short, "andw %w1,%0", "ir", ~v); ATOMIC_ASM(add, short, "addw %w1,%0", "ir", v); ATOMIC_ASM(subtract, short, "subw %w1,%0", "ir", v); ATOMIC_ASM(set, int, "orl %1,%0", "ir", v); ATOMIC_ASM(clear, int, "andl %1,%0", "ir", ~v); ATOMIC_ASM(add, int, "addl %1,%0", "ir", v); ATOMIC_ASM(subtract, int, "subl %1,%0", "ir", v); ATOMIC_ASM(set, long, "orl %1,%0", "ir", v); ATOMIC_ASM(clear, long, "andl %1,%0", "ir", ~v); ATOMIC_ASM(add, long, "addl %1,%0", "ir", v); ATOMIC_ASM(subtract, long, "subl %1,%0", "ir", v); #define ATOMIC_LOADSTORE(TYPE) \ ATOMIC_LOAD(TYPE); \ ATOMIC_STORE(TYPE) ATOMIC_LOADSTORE(char); ATOMIC_LOADSTORE(short); ATOMIC_LOADSTORE(int); ATOMIC_LOADSTORE(long); #undef ATOMIC_ASM #undef ATOMIC_LOAD #undef ATOMIC_STORE #undef ATOMIC_LOADSTORE #ifndef WANT_FUNCTIONS static __inline int atomic_cmpset_long(volatile u_long *dst, u_long expect, u_long src) { return (atomic_cmpset_int((volatile u_int *)dst, (u_int)expect, (u_int)src)); } static __inline u_long atomic_fetchadd_long(volatile u_long *p, u_long v) { return (atomic_fetchadd_int((volatile u_int *)p, (u_int)v)); } static __inline int atomic_testandset_long(volatile u_long *p, u_int v) { return (atomic_testandset_int((volatile u_int *)p, v)); } static __inline int atomic_testandclear_long(volatile u_long *p, u_int v) { return (atomic_testandclear_int((volatile u_int *)p, v)); } /* Read the current value and store a new value in the destination. */ #ifdef __GNUCLIKE_ASM static __inline u_int atomic_swap_int(volatile u_int *p, u_int v) { __asm __volatile( " xchgl %1,%0 ; " "# atomic_swap_int" : "+r" (v), /* 0 */ "+m" (*p)); /* 1 */ return (v); } static __inline u_long atomic_swap_long(volatile u_long *p, u_long v) { return (atomic_swap_int((volatile u_int *)p, (u_int)v)); } #else /* !__GNUCLIKE_ASM */ u_int atomic_swap_int(volatile u_int *p, u_int v); u_long atomic_swap_long(volatile u_long *p, u_long v); #endif /* __GNUCLIKE_ASM */ #define atomic_set_acq_char atomic_set_barr_char #define atomic_set_rel_char atomic_set_barr_char #define atomic_clear_acq_char atomic_clear_barr_char #define atomic_clear_rel_char atomic_clear_barr_char #define atomic_add_acq_char atomic_add_barr_char #define atomic_add_rel_char atomic_add_barr_char #define atomic_subtract_acq_char atomic_subtract_barr_char #define atomic_subtract_rel_char atomic_subtract_barr_char +#define atomic_cmpset_acq_char atomic_cmpset_char +#define atomic_cmpset_rel_char atomic_cmpset_char +#define atomic_fcmpset_acq_char atomic_fcmpset_char +#define atomic_fcmpset_rel_char atomic_fcmpset_char #define atomic_set_acq_short atomic_set_barr_short #define atomic_set_rel_short atomic_set_barr_short #define atomic_clear_acq_short atomic_clear_barr_short #define atomic_clear_rel_short atomic_clear_barr_short #define atomic_add_acq_short atomic_add_barr_short #define atomic_add_rel_short atomic_add_barr_short #define atomic_subtract_acq_short atomic_subtract_barr_short #define atomic_subtract_rel_short atomic_subtract_barr_short +#define atomic_cmpset_acq_short atomic_cmpset_short +#define atomic_cmpset_rel_short atomic_cmpset_short +#define atomic_fcmpset_acq_short atomic_fcmpset_short +#define atomic_fcmpset_rel_short atomic_fcmpset_short #define atomic_set_acq_int atomic_set_barr_int #define atomic_set_rel_int atomic_set_barr_int #define atomic_clear_acq_int atomic_clear_barr_int #define atomic_clear_rel_int atomic_clear_barr_int #define atomic_add_acq_int atomic_add_barr_int #define atomic_add_rel_int atomic_add_barr_int #define atomic_subtract_acq_int atomic_subtract_barr_int #define atomic_subtract_rel_int atomic_subtract_barr_int #define atomic_cmpset_acq_int atomic_cmpset_int #define atomic_cmpset_rel_int atomic_cmpset_int #define atomic_fcmpset_acq_int atomic_fcmpset_int #define atomic_fcmpset_rel_int atomic_fcmpset_int #define atomic_set_acq_long atomic_set_barr_long #define atomic_set_rel_long atomic_set_barr_long #define atomic_clear_acq_long atomic_clear_barr_long #define atomic_clear_rel_long atomic_clear_barr_long #define atomic_add_acq_long atomic_add_barr_long #define atomic_add_rel_long atomic_add_barr_long #define atomic_subtract_acq_long atomic_subtract_barr_long #define atomic_subtract_rel_long atomic_subtract_barr_long #define atomic_cmpset_acq_long atomic_cmpset_long #define atomic_cmpset_rel_long atomic_cmpset_long #define atomic_fcmpset_acq_long atomic_fcmpset_long #define atomic_fcmpset_rel_long atomic_fcmpset_long #define atomic_readandclear_int(p) atomic_swap_int(p, 0) #define atomic_readandclear_long(p) atomic_swap_long(p, 0) /* Operations on 8-bit bytes. */ #define atomic_set_8 atomic_set_char #define atomic_set_acq_8 atomic_set_acq_char #define atomic_set_rel_8 atomic_set_rel_char #define atomic_clear_8 atomic_clear_char #define atomic_clear_acq_8 atomic_clear_acq_char #define atomic_clear_rel_8 atomic_clear_rel_char #define atomic_add_8 atomic_add_char #define atomic_add_acq_8 atomic_add_acq_char #define atomic_add_rel_8 atomic_add_rel_char #define atomic_subtract_8 atomic_subtract_char #define atomic_subtract_acq_8 atomic_subtract_acq_char #define atomic_subtract_rel_8 atomic_subtract_rel_char #define atomic_load_acq_8 atomic_load_acq_char #define atomic_store_rel_8 atomic_store_rel_char +#define atomic_cmpset_8 atomic_cmpset_char +#define atomic_cmpset_acq_8 atomic_cmpset_acq_char +#define atomic_cmpset_rel_8 atomic_cmpset_rel_char +#define atomic_fcmpset_8 atomic_fcmpset_char +#define atomic_fcmpset_acq_8 atomic_fcmpset_acq_char +#define atomic_fcmpset_rel_8 atomic_fcmpset_rel_char /* Operations on 16-bit words. */ #define atomic_set_16 atomic_set_short #define atomic_set_acq_16 atomic_set_acq_short #define atomic_set_rel_16 atomic_set_rel_short #define atomic_clear_16 atomic_clear_short #define atomic_clear_acq_16 atomic_clear_acq_short #define atomic_clear_rel_16 atomic_clear_rel_short #define atomic_add_16 atomic_add_short #define atomic_add_acq_16 atomic_add_acq_short #define atomic_add_rel_16 atomic_add_rel_short #define atomic_subtract_16 atomic_subtract_short #define atomic_subtract_acq_16 atomic_subtract_acq_short #define atomic_subtract_rel_16 atomic_subtract_rel_short #define atomic_load_acq_16 atomic_load_acq_short #define atomic_store_rel_16 atomic_store_rel_short +#define atomic_cmpset_16 atomic_cmpset_short +#define atomic_cmpset_acq_16 atomic_cmpset_acq_short +#define atomic_cmpset_rel_16 atomic_cmpset_rel_short +#define atomic_fcmpset_16 atomic_fcmpset_short +#define atomic_fcmpset_acq_16 atomic_fcmpset_acq_short +#define atomic_fcmpset_rel_16 atomic_fcmpset_rel_short /* Operations on 32-bit double words. */ #define atomic_set_32 atomic_set_int #define atomic_set_acq_32 atomic_set_acq_int #define atomic_set_rel_32 atomic_set_rel_int #define atomic_clear_32 atomic_clear_int #define atomic_clear_acq_32 atomic_clear_acq_int #define atomic_clear_rel_32 atomic_clear_rel_int #define atomic_add_32 atomic_add_int #define atomic_add_acq_32 atomic_add_acq_int #define atomic_add_rel_32 atomic_add_rel_int #define atomic_subtract_32 atomic_subtract_int #define atomic_subtract_acq_32 atomic_subtract_acq_int #define atomic_subtract_rel_32 atomic_subtract_rel_int #define atomic_load_acq_32 atomic_load_acq_int #define atomic_store_rel_32 atomic_store_rel_int #define atomic_cmpset_32 atomic_cmpset_int #define atomic_cmpset_acq_32 atomic_cmpset_acq_int #define atomic_cmpset_rel_32 atomic_cmpset_rel_int #define atomic_fcmpset_32 atomic_fcmpset_int #define atomic_fcmpset_acq_32 atomic_fcmpset_acq_int #define atomic_fcmpset_rel_32 atomic_fcmpset_rel_int #define atomic_swap_32 atomic_swap_int #define atomic_readandclear_32 atomic_readandclear_int #define atomic_fetchadd_32 atomic_fetchadd_int #define atomic_testandset_32 atomic_testandset_int #define atomic_testandclear_32 atomic_testandclear_int /* Operations on pointers. */ #define atomic_set_ptr(p, v) \ atomic_set_int((volatile u_int *)(p), (u_int)(v)) #define atomic_set_acq_ptr(p, v) \ atomic_set_acq_int((volatile u_int *)(p), (u_int)(v)) #define atomic_set_rel_ptr(p, v) \ atomic_set_rel_int((volatile u_int *)(p), (u_int)(v)) #define atomic_clear_ptr(p, v) \ atomic_clear_int((volatile u_int *)(p), (u_int)(v)) #define atomic_clear_acq_ptr(p, v) \ atomic_clear_acq_int((volatile u_int *)(p), (u_int)(v)) #define atomic_clear_rel_ptr(p, v) \ atomic_clear_rel_int((volatile u_int *)(p), (u_int)(v)) #define atomic_add_ptr(p, v) \ atomic_add_int((volatile u_int *)(p), (u_int)(v)) #define atomic_add_acq_ptr(p, v) \ atomic_add_acq_int((volatile u_int *)(p), (u_int)(v)) #define atomic_add_rel_ptr(p, v) \ atomic_add_rel_int((volatile u_int *)(p), (u_int)(v)) #define atomic_subtract_ptr(p, v) \ atomic_subtract_int((volatile u_int *)(p), (u_int)(v)) #define atomic_subtract_acq_ptr(p, v) \ atomic_subtract_acq_int((volatile u_int *)(p), (u_int)(v)) #define atomic_subtract_rel_ptr(p, v) \ atomic_subtract_rel_int((volatile u_int *)(p), (u_int)(v)) #define atomic_load_acq_ptr(p) \ atomic_load_acq_int((volatile u_int *)(p)) #define atomic_store_rel_ptr(p, v) \ atomic_store_rel_int((volatile u_int *)(p), (v)) #define atomic_cmpset_ptr(dst, old, new) \ atomic_cmpset_int((volatile u_int *)(dst), (u_int)(old), (u_int)(new)) #define atomic_cmpset_acq_ptr(dst, old, new) \ atomic_cmpset_acq_int((volatile u_int *)(dst), (u_int)(old), \ (u_int)(new)) #define atomic_cmpset_rel_ptr(dst, old, new) \ atomic_cmpset_rel_int((volatile u_int *)(dst), (u_int)(old), \ (u_int)(new)) #define atomic_fcmpset_ptr(dst, old, new) \ atomic_fcmpset_int((volatile u_int *)(dst), (u_int *)(old), (u_int)(new)) #define atomic_fcmpset_acq_ptr(dst, old, new) \ atomic_fcmpset_acq_int((volatile u_int *)(dst), (u_int *)(old), \ (u_int)(new)) #define atomic_fcmpset_rel_ptr(dst, old, new) \ atomic_fcmpset_rel_int((volatile u_int *)(dst), (u_int *)(old), \ (u_int)(new)) #define atomic_swap_ptr(p, v) \ atomic_swap_int((volatile u_int *)(p), (u_int)(v)) #define atomic_readandclear_ptr(p) \ atomic_readandclear_int((volatile u_int *)(p)) #endif /* !WANT_FUNCTIONS */ #if defined(_KERNEL) #define mb() __mbk() #define wmb() __mbk() #define rmb() __mbk() #else #define mb() __mbu() #define wmb() __mbu() #define rmb() __mbu() #endif #endif /* !_MACHINE_ATOMIC_H_ */