atomic.h revision 90515
1/*-
2 * Copyright (c) 1998 Doug Rabson
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 *
26 * $FreeBSD: head/sys/amd64/include/atomic.h 90515 2002-02-11 03:41:59Z bde $
27 */
28#ifndef _MACHINE_ATOMIC_H_
29#define _MACHINE_ATOMIC_H_
30
31/*
32 * Various simple arithmetic on memory which is atomic in the presence
33 * of interrupts and multiple processors.
34 *
35 * atomic_set_char(P, V)	(*(u_char*)(P) |= (V))
36 * atomic_clear_char(P, V)	(*(u_char*)(P) &= ~(V))
37 * atomic_add_char(P, V)	(*(u_char*)(P) += (V))
38 * atomic_subtract_char(P, V)	(*(u_char*)(P) -= (V))
39 *
40 * atomic_set_short(P, V)	(*(u_short*)(P) |= (V))
41 * atomic_clear_short(P, V)	(*(u_short*)(P) &= ~(V))
42 * atomic_add_short(P, V)	(*(u_short*)(P) += (V))
43 * atomic_subtract_short(P, V)	(*(u_short*)(P) -= (V))
44 *
45 * atomic_set_int(P, V)		(*(u_int*)(P) |= (V))
46 * atomic_clear_int(P, V)	(*(u_int*)(P) &= ~(V))
47 * atomic_add_int(P, V)		(*(u_int*)(P) += (V))
48 * atomic_subtract_int(P, V)	(*(u_int*)(P) -= (V))
49 * atomic_readandclear_int(P)	(return  *(u_int*)P; *(u_int*)P = 0;)
50 *
51 * atomic_set_long(P, V)	(*(u_long*)(P) |= (V))
52 * atomic_clear_long(P, V)	(*(u_long*)(P) &= ~(V))
53 * atomic_add_long(P, V)	(*(u_long*)(P) += (V))
54 * atomic_subtract_long(P, V)	(*(u_long*)(P) -= (V))
55 * atomic_readandclear_long(P)	(return  *(u_long*)P; *(u_long*)P = 0;)
56 */
57
58/*
59 * The above functions are expanded inline in the statically-linked
60 * kernel.  Lock prefixes are generated if an SMP kernel is being
61 * built.
62 *
63 * Kernel modules call real functions which are built into the kernel.
64 * This allows kernel modules to be portable between UP and SMP systems.
65 */
66#if defined(KLD_MODULE)
67#define ATOMIC_ASM(NAME, TYPE, OP, CONS, V)			\
68void atomic_##NAME##_##TYPE(volatile u_##TYPE *p, u_##TYPE v);
69
70int atomic_cmpset_int(volatile u_int *dst, u_int exp, u_int src);
71
72#define	ATOMIC_STORE_LOAD(TYPE, LOP, SOP)			\
73u_##TYPE	atomic_load_acq_##TYPE(volatile u_##TYPE *p);	\
74void		atomic_store_rel_##TYPE(volatile u_##TYPE *p, u_##TYPE v);
75
76#else /* !KLD_MODULE */
77
78/*
79 * For userland, assume the SMP case and use lock prefixes so that
80 * the binaries will run on both types of systems.
81 */
82#if defined(SMP) || !defined(_KERNEL)
83#define MPLOCKED	"lock ; "
84#else
85#define MPLOCKED
86#endif
87
88/*
89 * The assembly is volatilized to demark potential before-and-after side
90 * effects if an interrupt or SMP collision were to occur.
91 */
92#define ATOMIC_ASM(NAME, TYPE, OP, CONS, V)		\
93static __inline void					\
94atomic_##NAME##_##TYPE(volatile u_##TYPE *p, u_##TYPE v)\
95{							\
96	__asm __volatile(MPLOCKED OP			\
97			 : "+m" (*p)			\
98			 : CONS (V));			\
99}
100
101/*
102 * Atomic compare and set, used by the mutex functions
103 *
104 * if (*dst == exp) *dst = src (all 32 bit words)
105 *
106 * Returns 0 on failure, non-zero on success
107 */
108
109#if defined(I386_CPU)
110static __inline int
111atomic_cmpset_int(volatile u_int *dst, u_int exp, u_int src)
112{
113	int res = exp;
114
115	__asm __volatile(
116	"	pushfl ;		"
117	"	cli ;			"
118	"	cmpl	%0,%2 ;		"
119	"	jne	1f ;		"
120	"	movl	%1,%2 ;		"
121	"1:				"
122	"       sete	%%al;		"
123	"	movzbl	%%al,%0 ;	"
124	"	popfl ;			"
125	"# atomic_cmpset_int"
126	: "+a" (res)			/* 0 (result) */
127	: "r" (src),			/* 1 */
128	  "m" (*(dst))			/* 2 */
129	: "memory");
130
131	return (res);
132}
133#else /* defined(I386_CPU) */
134static __inline int
135atomic_cmpset_int(volatile u_int *dst, u_int exp, u_int src)
136{
137	int res = exp;
138
139	__asm __volatile (
140	"	" MPLOCKED "		"
141	"	cmpxchgl %1,%2 ;	"
142	"       setz	%%al ;		"
143	"	movzbl	%%al,%0 ;	"
144	"1:				"
145	"# atomic_cmpset_int"
146	: "+a" (res)			/* 0 (result) */
147	: "r" (src),			/* 1 */
148	  "m" (*(dst))			/* 2 */
149	: "memory");
150
151	return (res);
152}
153#endif /* defined(I386_CPU) */
154
155#if defined(I386_CPU)
156/*
157 * We assume that a = b will do atomic loads and stores.
158 *
159 * XXX: This is _NOT_ safe on a P6 or higher because it does not guarantee
160 * memory ordering.  These should only be used on a 386.
161 */
162#define ATOMIC_STORE_LOAD(TYPE, LOP, SOP)		\
163static __inline u_##TYPE				\
164atomic_load_acq_##TYPE(volatile u_##TYPE *p)		\
165{							\
166	return (*p);					\
167}							\
168							\
169static __inline void					\
170atomic_store_rel_##TYPE(volatile u_##TYPE *p, u_##TYPE v)\
171{							\
172	*p = v;						\
173	__asm __volatile("" : : : "memory");		\
174}
175#else
176
177#define ATOMIC_STORE_LOAD(TYPE, LOP, SOP)		\
178static __inline u_##TYPE				\
179atomic_load_acq_##TYPE(volatile u_##TYPE *p)		\
180{							\
181	u_##TYPE res;					\
182							\
183	__asm __volatile(MPLOCKED LOP			\
184	: "=a" (res),			/* 0 (result) */\
185	  "+m" (*p)			/* 1 */		\
186	: : "memory");				 	\
187							\
188	return (res);					\
189}							\
190							\
191/*							\
192 * The XCHG instruction asserts LOCK automagically.	\
193 */							\
194static __inline void					\
195atomic_store_rel_##TYPE(volatile u_##TYPE *p, u_##TYPE v)\
196{							\
197	__asm __volatile(SOP				\
198	: "+m" (*p),			/* 0 */		\
199	  "+r" (v)			/* 1 */		\
200	: : "memory");				 	\
201}
202#endif	/* defined(I386_CPU) */
203#endif /* KLD_MODULE */
204
205ATOMIC_ASM(set,	     char,  "orb %b1,%0",  "iq",  v)
206ATOMIC_ASM(clear,    char,  "andb %b1,%0", "iq", ~v)
207ATOMIC_ASM(add,	     char,  "addb %b1,%0", "iq",  v)
208ATOMIC_ASM(subtract, char,  "subb %b1,%0", "iq",  v)
209
210ATOMIC_ASM(set,	     short, "orw %w1,%0",  "ir",  v)
211ATOMIC_ASM(clear,    short, "andw %w1,%0", "ir", ~v)
212ATOMIC_ASM(add,	     short, "addw %w1,%0", "ir",  v)
213ATOMIC_ASM(subtract, short, "subw %w1,%0", "ir",  v)
214
215ATOMIC_ASM(set,	     int,   "orl %1,%0",   "ir",  v)
216ATOMIC_ASM(clear,    int,   "andl %1,%0",  "ir", ~v)
217ATOMIC_ASM(add,	     int,   "addl %1,%0",  "ir",  v)
218ATOMIC_ASM(subtract, int,   "subl %1,%0",  "ir",  v)
219
220ATOMIC_ASM(set,	     long,  "orl %1,%0",   "ir",  v)
221ATOMIC_ASM(clear,    long,  "andl %1,%0",  "ir", ~v)
222ATOMIC_ASM(add,	     long,  "addl %1,%0",  "ir",  v)
223ATOMIC_ASM(subtract, long,  "subl %1,%0",  "ir",  v)
224
225ATOMIC_STORE_LOAD(char,	"cmpxchgb %b0,%1", "xchgb %b1,%0")
226ATOMIC_STORE_LOAD(short,"cmpxchgw %w0,%1", "xchgw %w1,%0")
227ATOMIC_STORE_LOAD(int,	"cmpxchgl %0,%1",  "xchgl %1,%0")
228ATOMIC_STORE_LOAD(long,	"cmpxchgl %0,%1",  "xchgl %1,%0")
229
230#undef ATOMIC_ASM
231#undef ATOMIC_STORE_LOAD
232
233#define	atomic_set_acq_char		atomic_set_char
234#define	atomic_set_rel_char		atomic_set_char
235#define	atomic_clear_acq_char		atomic_clear_char
236#define	atomic_clear_rel_char		atomic_clear_char
237#define	atomic_add_acq_char		atomic_add_char
238#define	atomic_add_rel_char		atomic_add_char
239#define	atomic_subtract_acq_char	atomic_subtract_char
240#define	atomic_subtract_rel_char	atomic_subtract_char
241
242#define	atomic_set_acq_short		atomic_set_short
243#define	atomic_set_rel_short		atomic_set_short
244#define	atomic_clear_acq_short		atomic_clear_short
245#define	atomic_clear_rel_short		atomic_clear_short
246#define	atomic_add_acq_short		atomic_add_short
247#define	atomic_add_rel_short		atomic_add_short
248#define	atomic_subtract_acq_short	atomic_subtract_short
249#define	atomic_subtract_rel_short	atomic_subtract_short
250
251#define	atomic_set_acq_int		atomic_set_int
252#define	atomic_set_rel_int		atomic_set_int
253#define	atomic_clear_acq_int		atomic_clear_int
254#define	atomic_clear_rel_int		atomic_clear_int
255#define	atomic_add_acq_int		atomic_add_int
256#define	atomic_add_rel_int		atomic_add_int
257#define	atomic_subtract_acq_int		atomic_subtract_int
258#define	atomic_subtract_rel_int		atomic_subtract_int
259#define atomic_cmpset_acq_int		atomic_cmpset_int
260#define atomic_cmpset_rel_int		atomic_cmpset_int
261
262#define	atomic_set_acq_long		atomic_set_long
263#define	atomic_set_rel_long		atomic_set_long
264#define	atomic_clear_acq_long		atomic_clear_long
265#define	atomic_clear_rel_long		atomic_clear_long
266#define	atomic_add_acq_long		atomic_add_long
267#define	atomic_add_rel_long		atomic_add_long
268#define	atomic_subtract_acq_long	atomic_subtract_long
269#define	atomic_subtract_rel_long	atomic_subtract_long
270#define	atomic_cmpset_long		atomic_cmpset_int
271#define	atomic_cmpset_acq_long		atomic_cmpset_acq_int
272#define	atomic_cmpset_rel_long		atomic_cmpset_rel_int
273
274#define atomic_cmpset_acq_ptr		atomic_cmpset_ptr
275#define atomic_cmpset_rel_ptr		atomic_cmpset_ptr
276
277#define	atomic_set_8		atomic_set_char
278#define	atomic_set_acq_8	atomic_set_acq_char
279#define	atomic_set_rel_8	atomic_set_rel_char
280#define	atomic_clear_8		atomic_clear_char
281#define	atomic_clear_acq_8	atomic_clear_acq_char
282#define	atomic_clear_rel_8	atomic_clear_rel_char
283#define	atomic_add_8		atomic_add_char
284#define	atomic_add_acq_8	atomic_add_acq_char
285#define	atomic_add_rel_8	atomic_add_rel_char
286#define	atomic_subtract_8	atomic_subtract_char
287#define	atomic_subtract_acq_8	atomic_subtract_acq_char
288#define	atomic_subtract_rel_8	atomic_subtract_rel_char
289#define	atomic_load_acq_8	atomic_load_acq_char
290#define	atomic_store_rel_8	atomic_store_rel_char
291
292#define	atomic_set_16		atomic_set_short
293#define	atomic_set_acq_16	atomic_set_acq_short
294#define	atomic_set_rel_16	atomic_set_rel_short
295#define	atomic_clear_16		atomic_clear_short
296#define	atomic_clear_acq_16	atomic_clear_acq_short
297#define	atomic_clear_rel_16	atomic_clear_rel_short
298#define	atomic_add_16		atomic_add_short
299#define	atomic_add_acq_16	atomic_add_acq_short
300#define	atomic_add_rel_16	atomic_add_rel_short
301#define	atomic_subtract_16	atomic_subtract_short
302#define	atomic_subtract_acq_16	atomic_subtract_acq_short
303#define	atomic_subtract_rel_16	atomic_subtract_rel_short
304#define	atomic_load_acq_16	atomic_load_acq_short
305#define	atomic_store_rel_16	atomic_store_rel_short
306
307#define	atomic_set_32		atomic_set_int
308#define	atomic_set_acq_32	atomic_set_acq_int
309#define	atomic_set_rel_32	atomic_set_rel_int
310#define	atomic_clear_32		atomic_clear_int
311#define	atomic_clear_acq_32	atomic_clear_acq_int
312#define	atomic_clear_rel_32	atomic_clear_rel_int
313#define	atomic_add_32		atomic_add_int
314#define	atomic_add_acq_32	atomic_add_acq_int
315#define	atomic_add_rel_32	atomic_add_rel_int
316#define	atomic_subtract_32	atomic_subtract_int
317#define	atomic_subtract_acq_32	atomic_subtract_acq_int
318#define	atomic_subtract_rel_32	atomic_subtract_rel_int
319#define	atomic_load_acq_32	atomic_load_acq_int
320#define	atomic_store_rel_32	atomic_store_rel_int
321#define	atomic_cmpset_32	atomic_cmpset_int
322#define	atomic_cmpset_acq_32	atomic_cmpset_acq_int
323#define	atomic_cmpset_rel_32	atomic_cmpset_rel_int
324#define	atomic_readandclear_32	atomic_readandclear_int
325
326#if !defined(WANT_FUNCTIONS)
327static __inline int
328atomic_cmpset_ptr(volatile void *dst, void *exp, void *src)
329{
330
331	return (atomic_cmpset_int((volatile u_int *)dst, (u_int)exp,
332	    (u_int)src));
333}
334
335static __inline void *
336atomic_load_acq_ptr(volatile void *p)
337{
338	return (void *)atomic_load_acq_int((volatile u_int *)p);
339}
340
341static __inline void
342atomic_store_rel_ptr(volatile void *p, void *v)
343{
344	atomic_store_rel_int((volatile u_int *)p, (u_int)v);
345}
346
347#define ATOMIC_PTR(NAME)				\
348static __inline void					\
349atomic_##NAME##_ptr(volatile void *p, uintptr_t v)	\
350{							\
351	atomic_##NAME##_int((volatile u_int *)p, v);	\
352}							\
353							\
354static __inline void					\
355atomic_##NAME##_acq_ptr(volatile void *p, uintptr_t v)	\
356{							\
357	atomic_##NAME##_acq_int((volatile u_int *)p, v);\
358}							\
359							\
360static __inline void					\
361atomic_##NAME##_rel_ptr(volatile void *p, uintptr_t v)	\
362{							\
363	atomic_##NAME##_rel_int((volatile u_int *)p, v);\
364}
365
366ATOMIC_PTR(set)
367ATOMIC_PTR(clear)
368ATOMIC_PTR(add)
369ATOMIC_PTR(subtract)
370
371#undef ATOMIC_PTR
372
373static __inline u_int
374atomic_readandclear_int(volatile u_int *addr)
375{
376	u_int result;
377
378	__asm __volatile (
379	"	xorl	%0,%0 ;		"
380	"	xchgl	%1,%0 ;		"
381	"# atomic_readandclear_int"
382	: "=&r" (result)		/* 0 (result) */
383	: "m" (*addr));			/* 1 (addr) */
384
385	return (result);
386}
387
388static __inline u_long
389atomic_readandclear_long(volatile u_long *addr)
390{
391	u_long result;
392
393	__asm __volatile (
394	"	xorl	%0,%0 ;		"
395	"	xchgl	%1,%0 ;		"
396	"# atomic_readandclear_int"
397	: "=&r" (result)		/* 0 (result) */
398	: "m" (*addr));			/* 1 (addr) */
399
400	return (result);
401}
402#endif	/* !defined(WANT_FUNCTIONS) */
403#endif /* ! _MACHINE_ATOMIC_H_ */
404