1 | /*
|
---|
2 | * Copyright (c) 2001-2004 Jakub Jermar
|
---|
3 | * Copyright (c) 2012 Adam Hraska
|
---|
4 | * All rights reserved.
|
---|
5 | *
|
---|
6 | * Redistribution and use in source and binary forms, with or without
|
---|
7 | * modification, are permitted provided that the following conditions
|
---|
8 | * are met:
|
---|
9 | *
|
---|
10 | * - Redistributions of source code must retain the above copyright
|
---|
11 | * notice, this list of conditions and the following disclaimer.
|
---|
12 | * - 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 | * - The name of the author may not be used to endorse or promote products
|
---|
16 | * derived from this software without specific prior written permission.
|
---|
17 | *
|
---|
18 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
|
---|
19 | * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
|
---|
20 | * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
|
---|
21 | * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
|
---|
22 | * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
|
---|
23 | * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
|
---|
24 | * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
|
---|
25 | * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
---|
26 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
|
---|
27 | * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
---|
28 | */
|
---|
29 |
|
---|
30 | /** @addtogroup amd64
|
---|
31 | * @{
|
---|
32 | */
|
---|
33 | /** @file
|
---|
34 | */
|
---|
35 |
|
---|
36 | #ifndef KERN_amd64_ATOMIC_H_
|
---|
37 | #define KERN_amd64_ATOMIC_H_
|
---|
38 |
|
---|
39 | #include <typedefs.h>
|
---|
40 | #include <arch/barrier.h>
|
---|
41 | #include <preemption.h>
|
---|
42 | #include <trace.h>
|
---|
43 |
|
---|
44 | NO_TRACE static inline void atomic_inc(atomic_t *val)
|
---|
45 | {
|
---|
46 | #ifdef CONFIG_SMP
|
---|
47 | asm volatile (
|
---|
48 | "lock incq %[count]\n"
|
---|
49 | : [count] "+m" (val->count)
|
---|
50 | );
|
---|
51 | #else
|
---|
52 | asm volatile (
|
---|
53 | "incq %[count]\n"
|
---|
54 | : [count] "+m" (val->count)
|
---|
55 | );
|
---|
56 | #endif /* CONFIG_SMP */
|
---|
57 | }
|
---|
58 |
|
---|
59 | NO_TRACE static inline void atomic_dec(atomic_t *val)
|
---|
60 | {
|
---|
61 | #ifdef CONFIG_SMP
|
---|
62 | asm volatile (
|
---|
63 | "lock decq %[count]\n"
|
---|
64 | : [count] "+m" (val->count)
|
---|
65 | );
|
---|
66 | #else
|
---|
67 | asm volatile (
|
---|
68 | "decq %[count]\n"
|
---|
69 | : [count] "+m" (val->count)
|
---|
70 | );
|
---|
71 | #endif /* CONFIG_SMP */
|
---|
72 | }
|
---|
73 |
|
---|
74 | NO_TRACE static inline atomic_count_t atomic_postinc(atomic_t *val)
|
---|
75 | {
|
---|
76 | atomic_count_t r = 1;
|
---|
77 |
|
---|
78 | asm volatile (
|
---|
79 | "lock xaddq %[r], %[count]\n"
|
---|
80 | : [count] "+m" (val->count),
|
---|
81 | [r] "+r" (r)
|
---|
82 | );
|
---|
83 |
|
---|
84 | return r;
|
---|
85 | }
|
---|
86 |
|
---|
87 | NO_TRACE static inline atomic_count_t atomic_postdec(atomic_t *val)
|
---|
88 | {
|
---|
89 | atomic_count_t r = -1;
|
---|
90 |
|
---|
91 | asm volatile (
|
---|
92 | "lock xaddq %[r], %[count]\n"
|
---|
93 | : [count] "+m" (val->count),
|
---|
94 | [r] "+r" (r)
|
---|
95 | );
|
---|
96 |
|
---|
97 | return r;
|
---|
98 | }
|
---|
99 |
|
---|
100 | #define atomic_preinc(val) (atomic_postinc(val) + 1)
|
---|
101 | #define atomic_predec(val) (atomic_postdec(val) - 1)
|
---|
102 |
|
---|
103 | NO_TRACE static inline atomic_count_t test_and_set(atomic_t *val)
|
---|
104 | {
|
---|
105 | atomic_count_t v = 1;
|
---|
106 |
|
---|
107 | asm volatile (
|
---|
108 | "xchgq %[v], %[count]\n"
|
---|
109 | : [v] "+r" (v),
|
---|
110 | [count] "+m" (val->count)
|
---|
111 | );
|
---|
112 |
|
---|
113 | return v;
|
---|
114 | }
|
---|
115 |
|
---|
116 | /** amd64 specific fast spinlock */
|
---|
117 | NO_TRACE static inline void atomic_lock_arch(atomic_t *val)
|
---|
118 | {
|
---|
119 | atomic_count_t tmp;
|
---|
120 |
|
---|
121 | preemption_disable();
|
---|
122 | asm volatile (
|
---|
123 | "0:\n"
|
---|
124 | " pause\n"
|
---|
125 | " mov %[count], %[tmp]\n"
|
---|
126 | " testq %[tmp], %[tmp]\n"
|
---|
127 | " jnz 0b\n" /* lightweight looping on locked spinlock */
|
---|
128 |
|
---|
129 | " incq %[tmp]\n" /* now use the atomic operation */
|
---|
130 | " xchgq %[count], %[tmp]\n"
|
---|
131 | " testq %[tmp], %[tmp]\n"
|
---|
132 | " jnz 0b\n"
|
---|
133 | : [count] "+m" (val->count),
|
---|
134 | [tmp] "=&r" (tmp)
|
---|
135 | );
|
---|
136 |
|
---|
137 | /*
|
---|
138 | * Prevent critical section code from bleeding out this way up.
|
---|
139 | */
|
---|
140 | CS_ENTER_BARRIER();
|
---|
141 | }
|
---|
142 |
|
---|
143 |
|
---|
144 | #define _atomic_cas_impl(pptr, exp_val, new_val, old_val, prefix) \
|
---|
145 | ({ \
|
---|
146 | switch (sizeof(typeof(*(pptr)))) { \
|
---|
147 | case 1: \
|
---|
148 | asm volatile ( \
|
---|
149 | prefix " cmpxchgb %[newval], %[ptr]\n" \
|
---|
150 | : /* Output operands. */ \
|
---|
151 | /* Old/current value is returned in eax. */ \
|
---|
152 | [oldval] "=a" (old_val), \
|
---|
153 | /* (*ptr) will be read and written to, hence "+" */ \
|
---|
154 | [ptr] "+m" (*pptr) \
|
---|
155 | : /* Input operands. */ \
|
---|
156 | /* Expected value must be in eax. */ \
|
---|
157 | [expval] "a" (exp_val), \
|
---|
158 | /* The new value may be in any register. */ \
|
---|
159 | [newval] "r" (new_val) \
|
---|
160 | : "memory" \
|
---|
161 | ); \
|
---|
162 | break; \
|
---|
163 | case 2: \
|
---|
164 | asm volatile ( \
|
---|
165 | prefix " cmpxchgw %[newval], %[ptr]\n" \
|
---|
166 | : /* Output operands. */ \
|
---|
167 | /* Old/current value is returned in eax. */ \
|
---|
168 | [oldval] "=a" (old_val), \
|
---|
169 | /* (*ptr) will be read and written to, hence "+" */ \
|
---|
170 | [ptr] "+m" (*pptr) \
|
---|
171 | : /* Input operands. */ \
|
---|
172 | /* Expected value must be in eax. */ \
|
---|
173 | [expval] "a" (exp_val), \
|
---|
174 | /* The new value may be in any register. */ \
|
---|
175 | [newval] "r" (new_val) \
|
---|
176 | : "memory" \
|
---|
177 | ); \
|
---|
178 | break; \
|
---|
179 | case 4: \
|
---|
180 | asm volatile ( \
|
---|
181 | prefix " cmpxchgl %[newval], %[ptr]\n" \
|
---|
182 | : /* Output operands. */ \
|
---|
183 | /* Old/current value is returned in eax. */ \
|
---|
184 | [oldval] "=a" (old_val), \
|
---|
185 | /* (*ptr) will be read and written to, hence "+" */ \
|
---|
186 | [ptr] "+m" (*pptr) \
|
---|
187 | : /* Input operands. */ \
|
---|
188 | /* Expected value must be in eax. */ \
|
---|
189 | [expval] "a" (exp_val), \
|
---|
190 | /* The new value may be in any register. */ \
|
---|
191 | [newval] "r" (new_val) \
|
---|
192 | : "memory" \
|
---|
193 | ); \
|
---|
194 | break; \
|
---|
195 | case 8: \
|
---|
196 | asm volatile ( \
|
---|
197 | prefix " cmpxchgq %[newval], %[ptr]\n" \
|
---|
198 | : /* Output operands. */ \
|
---|
199 | /* Old/current value is returned in eax. */ \
|
---|
200 | [oldval] "=a" (old_val), \
|
---|
201 | /* (*ptr) will be read and written to, hence "+" */ \
|
---|
202 | [ptr] "+m" (*pptr) \
|
---|
203 | : /* Input operands. */ \
|
---|
204 | /* Expected value must be in eax. */ \
|
---|
205 | [expval] "a" (exp_val), \
|
---|
206 | /* The new value may be in any register. */ \
|
---|
207 | [newval] "r" (new_val) \
|
---|
208 | : "memory" \
|
---|
209 | ); \
|
---|
210 | break; \
|
---|
211 | } \
|
---|
212 | })
|
---|
213 |
|
---|
214 |
|
---|
215 | #ifndef local_atomic_cas
|
---|
216 |
|
---|
217 | #define local_atomic_cas(pptr, exp_val, new_val) \
|
---|
218 | ({ \
|
---|
219 | /* Use proper types and avoid name clashes */ \
|
---|
220 | typeof(*(pptr)) _old_val_cas; \
|
---|
221 | typeof(*(pptr)) _exp_val_cas = exp_val; \
|
---|
222 | typeof(*(pptr)) _new_val_cas = new_val; \
|
---|
223 | _atomic_cas_impl(pptr, _exp_val_cas, _new_val_cas, _old_val_cas, ""); \
|
---|
224 | \
|
---|
225 | _old_val_cas; \
|
---|
226 | })
|
---|
227 |
|
---|
228 | #else
|
---|
229 | /* Check if arch/atomic.h does not accidentally include /atomic.h .*/
|
---|
230 | #error Architecture specific cpu local atomics already defined! Check your includes.
|
---|
231 | #endif
|
---|
232 |
|
---|
233 |
|
---|
234 | #ifndef local_atomic_exchange
|
---|
235 | /*
|
---|
236 | * Issuing a xchg instruction always implies lock prefix semantics.
|
---|
237 | * Therefore, it is cheaper to use a cmpxchg without a lock prefix
|
---|
238 | * in a loop.
|
---|
239 | */
|
---|
240 | #define local_atomic_exchange(pptr, new_val) \
|
---|
241 | ({ \
|
---|
242 | /* Use proper types and avoid name clashes */ \
|
---|
243 | typeof(*(pptr)) _exp_val_x; \
|
---|
244 | typeof(*(pptr)) _old_val_x; \
|
---|
245 | typeof(*(pptr)) _new_val_x = new_val; \
|
---|
246 | \
|
---|
247 | do { \
|
---|
248 | _exp_val_x = *pptr; \
|
---|
249 | _old_val_x = local_atomic_cas(pptr, _exp_val_x, _new_val_x); \
|
---|
250 | } while (_old_val_x != _exp_val_x); \
|
---|
251 | \
|
---|
252 | _old_val_x; \
|
---|
253 | })
|
---|
254 |
|
---|
255 | #else
|
---|
256 | /* Check if arch/atomic.h does not accidentally include /atomic.h .*/
|
---|
257 | #error Architecture specific cpu local atomics already defined! Check your includes.
|
---|
258 | #endif
|
---|
259 |
|
---|
260 |
|
---|
261 | #endif
|
---|
262 |
|
---|
263 | /** @}
|
---|
264 | */
|
---|