atomic.h revision 262004
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: stable/10/sys/ia64/include/atomic.h 262004 2014-02-16 23:08:21Z marcel $
27 */
28
29#ifndef _MACHINE_ATOMIC_H_
30#define	_MACHINE_ATOMIC_H_
31
32#define	mb()	__asm __volatile("mf")
33#define	wmb()	mb()
34#define	rmb()	mb()
35
36/*
37 * Various simple arithmetic on memory which is atomic in the presence
38 * of interrupts and SMP safe.
39 */
40
41/*
42 * Everything is built out of cmpxchg.
43 */
44#define	IA64_CMPXCHG(sz, sem, p, cmpval, newval, ret)			\
45	__asm __volatile (						\
46		"mov ar.ccv=%2;;\n\t"					\
47		"cmpxchg" #sz "." #sem " %0=%4,%3,ar.ccv\n\t"		\
48		: "=r" (ret), "=m" (*p)					\
49		: "r" ((uint64_t)cmpval), "r" (newval), "m" (*p)	\
50		: "memory")
51
52/*
53 * Some common forms of cmpxch.
54 */
55static __inline uint32_t
56ia64_cmpxchg_acq_32(volatile uint32_t* p, uint32_t cmpval, uint32_t newval)
57{
58	uint32_t ret;
59	IA64_CMPXCHG(4, acq, p, cmpval, newval, ret);
60	return (ret);
61}
62
63static __inline uint32_t
64ia64_cmpxchg_rel_32(volatile uint32_t* p, uint32_t cmpval, uint32_t newval)
65{
66	uint32_t ret;
67	IA64_CMPXCHG(4, rel, p, cmpval, newval, ret);
68	return (ret);
69}
70
71static __inline uint64_t
72ia64_cmpxchg_acq_64(volatile uint64_t* p, uint64_t cmpval, uint64_t newval)
73{
74	uint64_t ret;
75	IA64_CMPXCHG(8, acq, p, cmpval, newval, ret);
76	return (ret);
77}
78
79static __inline uint64_t
80ia64_cmpxchg_rel_64(volatile uint64_t* p, uint64_t cmpval, uint64_t newval)
81{
82	uint64_t ret;
83	IA64_CMPXCHG(8, rel, p, cmpval, newval, ret);
84	return (ret);
85}
86
87#define	ATOMIC_STORE_LOAD(type, width, size)				\
88	static __inline uint##width##_t					\
89	ia64_ld_acq_##width(volatile uint##width##_t* p)		\
90	{								\
91		uint##width##_t v;					\
92		__asm __volatile ("ld" size ".acq %0=%1" : "=r" (v)	\
93		    : "m" (*p) : "memory");				\
94		return (v);						\
95	}								\
96									\
97	static __inline uint##width##_t					\
98	atomic_load_acq_##width(volatile uint##width##_t* p)		\
99	{								\
100		uint##width##_t v;					\
101		__asm __volatile ("ld" size ".acq %0=%1" : "=r" (v)	\
102		    : "m" (*p) : "memory");				\
103		return (v);						\
104	}								\
105									\
106	static __inline uint##width##_t					\
107	atomic_load_acq_##type(volatile uint##width##_t* p)		\
108	{								\
109		uint##width##_t v;					\
110		__asm __volatile ("ld" size ".acq %0=%1" : "=r" (v)	\
111		    : "m" (*p) : "memory");				\
112		return (v);						\
113	}								\
114								       	\
115	static __inline void						\
116	ia64_st_rel_##width(volatile uint##width##_t* p, uint##width##_t v) \
117	{								\
118		__asm __volatile ("st" size ".rel %0=%1" : "=m" (*p)	\
119		    : "r" (v) : "memory");				\
120	}								\
121									\
122	static __inline void						\
123	atomic_store_rel_##width(volatile uint##width##_t* p,		\
124	    uint##width##_t v)						\
125	{								\
126		__asm __volatile ("st" size ".rel %0=%1" : "=m" (*p)	\
127		    : "r" (v) : "memory");				\
128	}								\
129									\
130	static __inline void						\
131	atomic_store_rel_##type(volatile uint##width##_t* p,		\
132	    uint##width##_t v)						\
133	{								\
134		__asm __volatile ("st" size ".rel %0=%1" : "=m" (*p)	\
135		    : "r" (v) : "memory");				\
136	}
137
138ATOMIC_STORE_LOAD(char,	 8,  "1")
139ATOMIC_STORE_LOAD(short, 16, "2")
140ATOMIC_STORE_LOAD(int,	 32, "4")
141ATOMIC_STORE_LOAD(long,	 64, "8")
142
143#undef ATOMIC_STORE_LOAD
144
145#define	atomic_load_acq_ptr(p)		\
146    ((void *)atomic_load_acq_64((volatile uint64_t *)p))
147
148#define	atomic_store_rel_ptr(p, v)	\
149    atomic_store_rel_64((volatile uint64_t *)p, (uint64_t)v)
150
151#define	IA64_ATOMIC(sz, type, name, width, op)				\
152	static __inline type						\
153	atomic_##name##_acq_##width(volatile type *p, type v)		\
154	{								\
155		type old, ret;						\
156		do {							\
157			old = *p;					\
158			IA64_CMPXCHG(sz, acq, p, old, old op v, ret);	\
159		} while (ret != old);					\
160		return (old);						\
161	}								\
162									\
163	static __inline type						\
164	atomic_##name##_rel_##width(volatile type *p, type v)		\
165	{								\
166		type old, ret;						\
167		do {							\
168			old = *p;					\
169			IA64_CMPXCHG(sz, rel, p, old, old op v, ret);	\
170		} while (ret != old);					\
171		return (old);						\
172	}
173
174IA64_ATOMIC(1, uint8_t,	 set, 8,  |)
175IA64_ATOMIC(2, uint16_t, set, 16, |)
176IA64_ATOMIC(4, uint32_t, set, 32, |)
177IA64_ATOMIC(8, uint64_t, set, 64, |)
178
179IA64_ATOMIC(1, uint8_t,  clear,	8,  &~)
180IA64_ATOMIC(2, uint16_t, clear,	16, &~)
181IA64_ATOMIC(4, uint32_t, clear,	32, &~)
182IA64_ATOMIC(8, uint64_t, clear,	64, &~)
183
184IA64_ATOMIC(1, uint8_t,  add, 8,  +)
185IA64_ATOMIC(2, uint16_t, add, 16, +)
186IA64_ATOMIC(4, uint32_t, add, 32, +)
187IA64_ATOMIC(8, uint64_t, add, 64, +)
188
189IA64_ATOMIC(1, uint8_t,  subtract, 8,  -)
190IA64_ATOMIC(2, uint16_t, subtract, 16, -)
191IA64_ATOMIC(4, uint32_t, subtract, 32, -)
192IA64_ATOMIC(8, uint64_t, subtract, 64, -)
193
194#undef IA64_ATOMIC
195
196#define	atomic_set_8			atomic_set_acq_8
197#define	atomic_clear_8			atomic_clear_acq_8
198#define	atomic_add_8			atomic_add_acq_8
199#define	atomic_subtract_8		atomic_subtract_acq_8
200
201#define	atomic_set_16			atomic_set_acq_16
202#define	atomic_clear_16			atomic_clear_acq_16
203#define	atomic_add_16			atomic_add_acq_16
204#define	atomic_subtract_16		atomic_subtract_acq_16
205
206#define	atomic_set_32			atomic_set_acq_32
207#define	atomic_clear_32			atomic_clear_acq_32
208#define	atomic_add_32			atomic_add_acq_32
209#define	atomic_subtract_32		atomic_subtract_acq_32
210
211#define	atomic_set_64			atomic_set_acq_64
212#define	atomic_clear_64			atomic_clear_acq_64
213#define	atomic_add_64			atomic_add_acq_64
214#define	atomic_subtract_64		atomic_subtract_acq_64
215
216#define	atomic_set_char			atomic_set_8
217#define	atomic_clear_char		atomic_clear_8
218#define	atomic_add_char			atomic_add_8
219#define	atomic_subtract_char		atomic_subtract_8
220#define	atomic_set_acq_char		atomic_set_acq_8
221#define	atomic_clear_acq_char		atomic_clear_acq_8
222#define	atomic_add_acq_char		atomic_add_acq_8
223#define	atomic_subtract_acq_char	atomic_subtract_acq_8
224#define	atomic_set_rel_char		atomic_set_rel_8
225#define	atomic_clear_rel_char		atomic_clear_rel_8
226#define	atomic_add_rel_char		atomic_add_rel_8
227#define	atomic_subtract_rel_char	atomic_subtract_rel_8
228
229#define	atomic_set_short		atomic_set_16
230#define	atomic_clear_short		atomic_clear_16
231#define	atomic_add_short		atomic_add_16
232#define	atomic_subtract_short		atomic_subtract_16
233#define	atomic_set_acq_short		atomic_set_acq_16
234#define	atomic_clear_acq_short		atomic_clear_acq_16
235#define	atomic_add_acq_short		atomic_add_acq_16
236#define	atomic_subtract_acq_short	atomic_subtract_acq_16
237#define	atomic_set_rel_short		atomic_set_rel_16
238#define	atomic_clear_rel_short		atomic_clear_rel_16
239#define	atomic_add_rel_short		atomic_add_rel_16
240#define	atomic_subtract_rel_short	atomic_subtract_rel_16
241
242#define	atomic_set_int			atomic_set_32
243#define	atomic_clear_int		atomic_clear_32
244#define	atomic_add_int			atomic_add_32
245#define	atomic_subtract_int		atomic_subtract_32
246#define	atomic_set_acq_int		atomic_set_acq_32
247#define	atomic_clear_acq_int		atomic_clear_acq_32
248#define	atomic_add_acq_int		atomic_add_acq_32
249#define	atomic_subtract_acq_int		atomic_subtract_acq_32
250#define	atomic_set_rel_int		atomic_set_rel_32
251#define	atomic_clear_rel_int		atomic_clear_rel_32
252#define	atomic_add_rel_int		atomic_add_rel_32
253#define	atomic_subtract_rel_int		atomic_subtract_rel_32
254
255#define	atomic_set_long			atomic_set_64
256#define	atomic_clear_long		atomic_clear_64
257#define	atomic_add_long			atomic_add_64
258#define	atomic_subtract_long		atomic_subtract_64
259#define	atomic_set_acq_long		atomic_set_acq_64
260#define	atomic_clear_acq_long		atomic_clear_acq_64
261#define	atomic_add_acq_long		atomic_add_acq_64
262#define	atomic_subtract_acq_long	atomic_subtract_acq_64
263#define	atomic_set_rel_long		atomic_set_rel_64
264#define	atomic_clear_rel_long		atomic_clear_rel_64
265#define	atomic_add_rel_long		atomic_add_rel_64
266#define	atomic_subtract_rel_long	atomic_subtract_rel_64
267
268/* XXX Needs casting. */
269#define	atomic_set_ptr			atomic_set_64
270#define	atomic_clear_ptr		atomic_clear_64
271#define	atomic_add_ptr			atomic_add_64
272#define	atomic_subtract_ptr		atomic_subtract_64
273#define	atomic_set_acq_ptr		atomic_set_acq_64
274#define	atomic_clear_acq_ptr		atomic_clear_acq_64
275#define	atomic_add_acq_ptr		atomic_add_acq_64
276#define	atomic_subtract_acq_ptr		atomic_subtract_acq_64
277#define	atomic_set_rel_ptr		atomic_set_rel_64
278#define	atomic_clear_rel_ptr		atomic_clear_rel_64
279#define	atomic_add_rel_ptr		atomic_add_rel_64
280#define	atomic_subtract_rel_ptr		atomic_subtract_rel_64
281
282#undef IA64_CMPXCHG
283
284/*
285 * Atomically compare the value stored at *p with cmpval and if the
286 * two values are equal, update the value of *p with newval. Returns
287 * zero if the compare failed, nonzero otherwise.
288 */
289static __inline int
290atomic_cmpset_acq_32(volatile uint32_t* p, uint32_t cmpval, uint32_t newval)
291{
292	return (ia64_cmpxchg_acq_32(p, cmpval, newval) == cmpval);
293}
294
295static __inline int
296atomic_cmpset_rel_32(volatile uint32_t* p, uint32_t cmpval, uint32_t newval)
297{
298	return (ia64_cmpxchg_rel_32(p, cmpval, newval) == cmpval);
299}
300
301/*
302 * Atomically compare the value stored at *p with cmpval and if the
303 * two values are equal, update the value of *p with newval. Returns
304 * zero if the compare failed, nonzero otherwise.
305 */
306static __inline int
307atomic_cmpset_acq_64(volatile uint64_t* p, uint64_t cmpval, uint64_t newval)
308{
309	return (ia64_cmpxchg_acq_64(p, cmpval, newval) == cmpval);
310}
311
312static __inline int
313atomic_cmpset_rel_64(volatile uint64_t* p, uint64_t cmpval, uint64_t newval)
314{
315	return (ia64_cmpxchg_rel_64(p, cmpval, newval) == cmpval);
316}
317
318#define	atomic_cmpset_32		atomic_cmpset_acq_32
319#define	atomic_cmpset_64		atomic_cmpset_acq_64
320#define	atomic_cmpset_int		atomic_cmpset_32
321#define	atomic_cmpset_long		atomic_cmpset_64
322#define	atomic_cmpset_acq_int		atomic_cmpset_acq_32
323#define	atomic_cmpset_rel_int		atomic_cmpset_rel_32
324#define	atomic_cmpset_acq_long		atomic_cmpset_acq_64
325#define	atomic_cmpset_rel_long		atomic_cmpset_rel_64
326
327#define	atomic_cmpset_acq_ptr(p, o, n)	\
328    (atomic_cmpset_acq_64((volatile uint64_t *)p, (uint64_t)o, (uint64_t)n))
329
330#define	atomic_cmpset_ptr		atomic_cmpset_acq_ptr
331
332#define	atomic_cmpset_rel_ptr(p, o, n)	\
333    (atomic_cmpset_rel_64((volatile uint64_t *)p, (uint64_t)o, (uint64_t)n))
334
335static __inline uint32_t
336atomic_readandclear_32(volatile uint32_t* p)
337{
338	uint32_t val;
339	do {
340		val = *p;
341	} while (!atomic_cmpset_32(p, val, 0));
342	return (val);
343}
344
345static __inline uint64_t
346atomic_readandclear_64(volatile uint64_t* p)
347{
348	uint64_t val;
349	do {
350		val = *p;
351	} while (!atomic_cmpset_64(p, val, 0));
352	return (val);
353}
354
355#define	atomic_readandclear_int		atomic_readandclear_32
356#define	atomic_readandclear_long	atomic_readandclear_64
357#define	atomic_readandclear_ptr		atomic_readandclear_64
358
359/*
360 * Atomically add the value of v to the integer pointed to by p and return
361 * the previous value of *p.
362 *
363 * XXX: Should we use the fetchadd instruction here?
364 */
365static __inline uint32_t
366atomic_fetchadd_32(volatile uint32_t *p, uint32_t v)
367{
368	uint32_t value;
369
370	do {
371		value = *p;
372	} while (!atomic_cmpset_32(p, value, value + v));
373	return (value);
374}
375
376#define	atomic_fetchadd_int		atomic_fetchadd_32
377
378static __inline u_long
379atomic_fetchadd_long(volatile u_long *p, u_long v)
380{
381	u_long value;
382
383	do {
384		value = *p;
385	} while (!atomic_cmpset_64(p, value, value + v));
386	return (value);
387}
388
389/*
390 * <type> atomic_swap_<type>(volatile <type> *p, <type> v);
391 */
392
393static __inline uint32_t
394atomic_swap_32(volatile uint32_t *p, uint32_t v)
395{
396	uint32_t r;
397
398	__asm __volatile ("xchg4 %0 = %3, %2;;" : "=r"(r), "=m"(*p) :
399	    "r"(v), "m"(*p) : "memory");
400	return (r);
401}
402
403static __inline uint64_t
404atomic_swap_64(volatile uint64_t *p, uint64_t v)
405{
406	uint64_t r;
407
408	__asm __volatile ("xchg8 %0 = %3, %2;;" : "=r"(r), "=m"(*p) :
409	    "r"(v), "m"(*p) : "memory");
410	return (r);
411}
412
413#define	atomic_swap_int		atomic_swap_32
414#define	atomic_swap_long	atomic_swap_64
415#define	atomic_swap_ptr		atomic_swap_64
416
417#endif /* ! _MACHINE_ATOMIC_H_ */
418