1/*	$NetBSD$	*/
2
3/*-
4 * Copyright (c) 1998 Doug Rabson
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
27 *
28 * $FreeBSD: src/sys/ia64/include/atomic.h,v 1.10 2005/09/27 17:39:10 jhb Exp $
29 */
30
31#ifndef _MACHINE_ATOMIC_H_
32#define	_MACHINE_ATOMIC_H_
33
34/*
35 * Various simple arithmetic on memory which is atomic in the presence
36 * of interrupts and SMP safe.
37 */
38
39/*
40 * Everything is built out of cmpxchg.
41 */
42#define	IA64_CMPXCHG(sz, sem, p, cmpval, newval, ret)			\
43	__asm __volatile (						\
44		"mov ar.ccv=%2;;\n\t"					\
45		"cmpxchg" #sz "." #sem " %0=%4,%3,ar.ccv\n\t"		\
46		: "=r" (ret), "=m" (*p)					\
47		: "r" (cmpval), "r" (newval), "m" (*p)			\
48		: "memory")
49
50/*
51 * Some common forms of cmpxch.
52 */
53static __inline uint32_t
54ia64_cmpxchg_acq_32(volatile uint32_t* p, uint32_t cmpval, uint32_t newval)
55{
56	uint32_t ret;
57	IA64_CMPXCHG(4, acq, p, cmpval, newval, ret);
58	return (ret);
59}
60
61static __inline uint32_t
62ia64_cmpxchg_rel_32(volatile uint32_t* p, uint32_t cmpval, uint32_t newval)
63{
64	uint32_t ret;
65	IA64_CMPXCHG(4, rel, p, cmpval, newval, ret);
66	return (ret);
67}
68
69static __inline uint64_t
70ia64_cmpxchg_acq_64(volatile uint64_t* p, uint64_t cmpval, uint64_t newval)
71{
72	uint64_t ret;
73	IA64_CMPXCHG(8, acq, p, cmpval, newval, ret);
74	return (ret);
75}
76
77static __inline uint64_t
78ia64_cmpxchg_rel_64(volatile uint64_t* p, uint64_t cmpval, uint64_t newval)
79{
80	uint64_t ret;
81	IA64_CMPXCHG(8, rel, p, cmpval, newval, ret);
82	return (ret);
83}
84
85#define	ATOMIC_STORE_LOAD(type, width, size)				\
86	static __inline uint##width##_t					\
87	ia64_ld_acq_##width(volatile uint##width##_t* p)		\
88	{								\
89		uint##width##_t v;					\
90		__asm __volatile ("ld" size ".acq %0=%1" : "=r" (v)	\
91		    : "m" (*p) : "memory");				\
92		return (v);						\
93	}								\
94									\
95	static __inline uint##width##_t					\
96	atomic_load_acq_##width(volatile uint##width##_t* p)		\
97	{								\
98		uint##width##_t v;					\
99		__asm __volatile ("ld" size ".acq %0=%1" : "=r" (v)	\
100		    : "m" (*p) : "memory");				\
101		return (v);						\
102	}								\
103									\
104	static __inline uint##width##_t					\
105	atomic_load_acq_##type(volatile uint##width##_t* p)		\
106	{								\
107		uint##width##_t v;					\
108		__asm __volatile ("ld" size ".acq %0=%1" : "=r" (v)	\
109		    : "m" (*p) : "memory");				\
110		return (v);						\
111	}								\
112								       	\
113	static __inline void						\
114	ia64_st_rel_##width(volatile uint##width##_t* p, uint##width##_t v) \
115	{								\
116		__asm __volatile ("st" size ".rel %0=%1" : "=m" (*p)	\
117		    : "r" (v) : "memory");				\
118	}								\
119									\
120	static __inline void						\
121	atomic_store_rel_##width(volatile uint##width##_t* p,		\
122	    uint##width##_t v)						\
123	{								\
124		__asm __volatile ("st" size ".rel %0=%1" : "=m" (*p)	\
125		    : "r" (v) : "memory");				\
126	}								\
127									\
128	static __inline void						\
129	atomic_store_rel_##type(volatile uint##width##_t* p,		\
130	    uint##width##_t v)						\
131	{								\
132		__asm __volatile ("st" size ".rel %0=%1" : "=m" (*p)	\
133		    : "r" (v) : "memory");				\
134	}
135
136ATOMIC_STORE_LOAD(char,	 8,  "1")
137ATOMIC_STORE_LOAD(short, 16, "2")
138ATOMIC_STORE_LOAD(int,	 32, "4")
139ATOMIC_STORE_LOAD(long,	 64, "8")
140
141#undef ATOMIC_STORE_LOAD
142
143#define	atomic_load_acq_ptr	atomic_load_acq_64
144#define	atomic_store_rel_ptr	atomic_store_rel_64
145
146#define	IA64_ATOMIC(sz, type, name, width, op)				\
147	static __inline type						\
148	atomic_##name##_acq_##width(volatile type *p, type v)		\
149	{								\
150		type old, ret;						\
151		do {							\
152			old = *p;					\
153			IA64_CMPXCHG(sz, acq, p, old, old op v, ret);	\
154		} while (ret != old);					\
155		return (old);						\
156	}								\
157									\
158	static __inline type						\
159	atomic_##name##_rel_##width(volatile type *p, type v)		\
160	{								\
161		type old, ret;						\
162		do {							\
163			old = *p;					\
164			IA64_CMPXCHG(sz, rel, p, old, old op v, ret);	\
165		} while (ret != old);					\
166		return (old);						\
167	}
168
169IA64_ATOMIC(1, uint8_t,	 set, 8,  |)
170IA64_ATOMIC(2, uint16_t, set, 16, |)
171IA64_ATOMIC(4, uint32_t, set, 32, |)
172IA64_ATOMIC(8, uint64_t, set, 64, |)
173
174IA64_ATOMIC(1, uint8_t,  clear,	8,  &~)
175IA64_ATOMIC(2, uint16_t, clear,	16, &~)
176IA64_ATOMIC(4, uint32_t, clear,	32, &~)
177IA64_ATOMIC(8, uint64_t, clear,	64, &~)
178
179IA64_ATOMIC(1, uint8_t,  add, 8,  +)
180IA64_ATOMIC(2, uint16_t, add, 16, +)
181IA64_ATOMIC(4, uint32_t, add, 32, +)
182IA64_ATOMIC(8, uint64_t, add, 64, +)
183
184IA64_ATOMIC(1, uint8_t,  subtract, 8,  -)
185IA64_ATOMIC(2, uint16_t, subtract, 16, -)
186IA64_ATOMIC(4, uint32_t, subtract, 32, -)
187IA64_ATOMIC(8, uint64_t, subtract, 64, -)
188
189#undef IA64_ATOMIC
190
191#define	atomic_set_8			atomic_set_acq_8
192#define	atomic_clear_8			atomic_clear_acq_8
193#define	atomic_add_8			atomic_add_acq_8
194#define	atomic_subtract_8		atomic_subtract_acq_8
195
196#define	atomic_set_16			atomic_set_acq_16
197#define	atomic_clear_16			atomic_clear_acq_16
198#define	atomic_add_16			atomic_add_acq_16
199#define	atomic_subtract_16		atomic_subtract_acq_16
200
201#define	atomic_set_32			atomic_set_acq_32
202#define	atomic_clear_32			atomic_clear_acq_32
203#define	atomic_add_32			atomic_add_acq_32
204#define	atomic_subtract_32		atomic_subtract_acq_32
205
206#define	atomic_set_64			atomic_set_acq_64
207#define	atomic_clear_64			atomic_clear_acq_64
208#define	atomic_add_64			atomic_add_acq_64
209#define	atomic_subtract_64		atomic_subtract_acq_64
210
211#define	atomic_set_char			atomic_set_8
212#define	atomic_clear_char		atomic_clear_8
213#define	atomic_add_char			atomic_add_8
214#define	atomic_subtract_char		atomic_subtract_8
215#define	atomic_set_acq_char		atomic_set_acq_8
216#define	atomic_clear_acq_char		atomic_clear_acq_8
217#define	atomic_add_acq_char		atomic_add_acq_8
218#define	atomic_subtract_acq_char	atomic_subtract_acq_8
219#define	atomic_set_rel_char		atomic_set_rel_8
220#define	atomic_clear_rel_char		atomic_clear_rel_8
221#define	atomic_add_rel_char		atomic_add_rel_8
222#define	atomic_subtract_rel_char	atomic_subtract_rel_8
223
224#define	atomic_set_short		atomic_set_16
225#define	atomic_clear_short		atomic_clear_16
226#define	atomic_add_short		atomic_add_16
227#define	atomic_subtract_short		atomic_subtract_16
228#define	atomic_set_acq_short		atomic_set_acq_16
229#define	atomic_clear_acq_short		atomic_clear_acq_16
230#define	atomic_add_acq_short		atomic_add_acq_16
231#define	atomic_subtract_acq_short	atomic_subtract_acq_16
232#define	atomic_set_rel_short		atomic_set_rel_16
233#define	atomic_clear_rel_short		atomic_clear_rel_16
234#define	atomic_add_rel_short		atomic_add_rel_16
235#define	atomic_subtract_rel_short	atomic_subtract_rel_16
236
237#define	atomic_set_int			atomic_set_32
238#define	atomic_clear_int		atomic_clear_32
239#define	atomic_add_int			atomic_add_32
240#define	atomic_subtract_int		atomic_subtract_32
241#define	atomic_set_acq_int		atomic_set_acq_32
242#define	atomic_clear_acq_int		atomic_clear_acq_32
243#define	atomic_add_acq_int		atomic_add_acq_32
244#define	atomic_subtract_acq_int		atomic_subtract_acq_32
245#define	atomic_set_rel_int		atomic_set_rel_32
246#define	atomic_clear_rel_int		atomic_clear_rel_32
247#define	atomic_add_rel_int		atomic_add_rel_32
248#define	atomic_subtract_rel_int		atomic_subtract_rel_32
249
250#define	atomic_set_long			atomic_set_64
251#define	atomic_clear_long		atomic_clear_64
252#define	atomic_add_long			atomic_add_64
253#define	atomic_subtract_long		atomic_subtract_64
254#define	atomic_set_acq_long		atomic_set_acq_64
255#define	atomic_clear_acq_long		atomic_clear_acq_64
256#define	atomic_add_acq_long		atomic_add_acq_64
257#define	atomic_subtract_acq_long	atomic_subtract_acq_64
258#define	atomic_set_rel_long		atomic_set_rel_64
259#define	atomic_clear_rel_long		atomic_clear_rel_64
260#define	atomic_add_rel_long		atomic_add_rel_64
261#define	atomic_subtract_rel_long	atomic_subtract_rel_64
262
263#define	atomic_set_ptr			atomic_set_64
264#define	atomic_clear_ptr		atomic_clear_64
265#define	atomic_add_ptr			atomic_add_64
266#define	atomic_subtract_ptr		atomic_subtract_64
267#define	atomic_set_acq_ptr		atomic_set_acq_64
268#define	atomic_clear_acq_ptr		atomic_clear_acq_64
269#define	atomic_add_acq_ptr		atomic_add_acq_64
270#define	atomic_subtract_acq_ptr		atomic_subtract_acq_64
271#define	atomic_set_rel_ptr		atomic_set_rel_64
272#define	atomic_clear_rel_ptr		atomic_clear_rel_64
273#define	atomic_add_rel_ptr		atomic_add_rel_64
274#define	atomic_subtract_rel_ptr		atomic_subtract_rel_64
275
276#undef IA64_CMPXCHG
277
278/*
279 * Atomically compare the value stored at *p with cmpval and if the
280 * two values are equal, update the value of *p with newval. Returns
281 * zero if the compare failed, nonzero otherwise.
282 */
283static __inline int
284atomic_cmpset_acq_32(volatile uint32_t* p, uint32_t cmpval, uint32_t newval)
285{
286	return (ia64_cmpxchg_acq_32(p, cmpval, newval) == cmpval);
287}
288
289static __inline int
290atomic_cmpset_rel_32(volatile uint32_t* p, uint32_t cmpval, uint32_t newval)
291{
292	return (ia64_cmpxchg_rel_32(p, cmpval, newval) == cmpval);
293}
294
295/*
296 * Atomically compare the value stored at *p with cmpval and if the
297 * two values are equal, update the value of *p with newval. Returns
298 * zero if the compare failed, nonzero otherwise.
299 */
300static __inline int
301atomic_cmpset_acq_64(volatile uint64_t* p, uint64_t cmpval, uint64_t newval)
302{
303	return (ia64_cmpxchg_acq_64(p, cmpval, newval) == cmpval);
304}
305
306static __inline int
307atomic_cmpset_rel_64(volatile uint64_t* p, uint64_t cmpval, uint64_t newval)
308{
309	return (ia64_cmpxchg_rel_64(p, cmpval, newval) == cmpval);
310}
311
312#define	atomic_cmpset_32		atomic_cmpset_acq_32
313#define	atomic_cmpset_64		atomic_cmpset_acq_64
314#define	atomic_cmpset_int		atomic_cmpset_32
315#define	atomic_cmpset_long		atomic_cmpset_64
316#define	atomic_cmpset_ptr		atomic_cmpset_64
317#define	atomic_cmpset_acq_int		atomic_cmpset_acq_32
318#define	atomic_cmpset_rel_int		atomic_cmpset_rel_32
319#define	atomic_cmpset_acq_long		atomic_cmpset_acq_64
320#define	atomic_cmpset_rel_long		atomic_cmpset_rel_64
321#define	atomic_cmpset_acq_ptr		atomic_cmpset_acq_64
322#define	atomic_cmpset_rel_ptr		atomic_cmpset_rel_64
323
324static __inline uint32_t
325atomic_readandclear_32(volatile uint32_t* p)
326{
327	uint32_t val;
328	do {
329		val = *p;
330	} while (!atomic_cmpset_32(p, val, 0));
331	return (val);
332}
333
334static __inline uint64_t
335atomic_readandclear_64(volatile uint64_t* p)
336{
337	uint64_t val;
338	do {
339		val = *p;
340	} while (!atomic_cmpset_64(p, val, 0));
341	return (val);
342}
343
344#define	atomic_readandclear_int		atomic_readandclear_32
345#define	atomic_readandclear_long	atomic_readandclear_64
346
347/*
348 * Atomically add the value of v to the integer pointed to by p and return
349 * the previous value of *p.
350 *
351 * XXX: Should we use the fetchadd instruction here?
352 */
353static __inline uint32_t
354atomic_fetchadd_32(volatile uint32_t *p, uint32_t v)
355{
356	uint32_t value;
357
358	do {
359		value = *p;
360	} while (!atomic_cmpset_32(p, value, value + v));
361	return (value);
362}
363
364#define	atomic_fetchadd_int		atomic_fetchadd_32
365
366#endif /* ! _MACHINE_ATOMIC_H_ */
367