| 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 ia32
|
|---|
| 31 | * @{
|
|---|
| 32 | */
|
|---|
| 33 | /** @file
|
|---|
| 34 | */
|
|---|
| 35 |
|
|---|
| 36 | #ifndef KERN_ia32_ATOMIC_H_
|
|---|
| 37 | #define KERN_ia32_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 incl %[count]\n"
|
|---|
| 49 | : [count] "+m" (val->count)
|
|---|
| 50 | );
|
|---|
| 51 | #else
|
|---|
| 52 | asm volatile (
|
|---|
| 53 | "incl %[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 decl %[count]\n"
|
|---|
| 64 | : [count] "+m" (val->count)
|
|---|
| 65 | );
|
|---|
| 66 | #else
|
|---|
| 67 | asm volatile (
|
|---|
| 68 | "decl %[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 xaddl %[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 xaddl %[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 | "xchgl %[v], %[count]\n"
|
|---|
| 109 | : [v] "+r" (v),
|
|---|
| 110 | [count] "+m" (val->count)
|
|---|
| 111 | );
|
|---|
| 112 |
|
|---|
| 113 | return v;
|
|---|
| 114 | }
|
|---|
| 115 |
|
|---|
| 116 |
|
|---|
| 117 | /** ia32 specific fast spinlock */
|
|---|
| 118 | NO_TRACE static inline void atomic_lock_arch(atomic_t *val)
|
|---|
| 119 | {
|
|---|
| 120 | atomic_count_t tmp;
|
|---|
| 121 |
|
|---|
| 122 | preemption_disable();
|
|---|
| 123 | asm volatile (
|
|---|
| 124 | "0:\n"
|
|---|
| 125 | #ifndef PROCESSOR_i486
|
|---|
| 126 | "pause\n" /* Pentium 4's HT love this instruction */
|
|---|
| 127 | #endif
|
|---|
| 128 | "mov %[count], %[tmp]\n"
|
|---|
| 129 | "testl %[tmp], %[tmp]\n"
|
|---|
| 130 | "jnz 0b\n" /* lightweight looping on locked spinlock */
|
|---|
| 131 |
|
|---|
| 132 | "incl %[tmp]\n" /* now use the atomic operation */
|
|---|
| 133 | "xchgl %[count], %[tmp]\n"
|
|---|
| 134 | "testl %[tmp], %[tmp]\n"
|
|---|
| 135 | "jnz 0b\n"
|
|---|
| 136 | : [count] "+m" (val->count),
|
|---|
| 137 | [tmp] "=&r" (tmp)
|
|---|
| 138 | );
|
|---|
| 139 |
|
|---|
| 140 | /*
|
|---|
| 141 | * Prevent critical section code from bleeding out this way up.
|
|---|
| 142 | */
|
|---|
| 143 | CS_ENTER_BARRIER();
|
|---|
| 144 | }
|
|---|
| 145 |
|
|---|
| 146 |
|
|---|
| 147 | #define _atomic_cas_impl(pptr, exp_val, new_val, old_val, prefix) \
|
|---|
| 148 | ({ \
|
|---|
| 149 | switch (sizeof(typeof(*(pptr)))) { \
|
|---|
| 150 | case 1: \
|
|---|
| 151 | asm volatile ( \
|
|---|
| 152 | prefix " cmpxchgb %[newval], %[ptr]\n" \
|
|---|
| 153 | : /* Output operands. */ \
|
|---|
| 154 | /* Old/current value is returned in eax. */ \
|
|---|
| 155 | [oldval] "=a" (old_val), \
|
|---|
| 156 | /* (*ptr) will be read and written to, hence "+" */ \
|
|---|
| 157 | [ptr] "+m" (*pptr) \
|
|---|
| 158 | : /* Input operands. */ \
|
|---|
| 159 | /* Expected value must be in eax. */ \
|
|---|
| 160 | [expval] "a" (exp_val), \
|
|---|
| 161 | /* The new value may be in any register. */ \
|
|---|
| 162 | [newval] "r" (new_val) \
|
|---|
| 163 | : "memory" \
|
|---|
| 164 | ); \
|
|---|
| 165 | break; \
|
|---|
| 166 | case 2: \
|
|---|
| 167 | asm volatile ( \
|
|---|
| 168 | prefix " cmpxchgw %[newval], %[ptr]\n" \
|
|---|
| 169 | : /* Output operands. */ \
|
|---|
| 170 | /* Old/current value is returned in eax. */ \
|
|---|
| 171 | [oldval] "=a" (old_val), \
|
|---|
| 172 | /* (*ptr) will be read and written to, hence "+" */ \
|
|---|
| 173 | [ptr] "+m" (*pptr) \
|
|---|
| 174 | : /* Input operands. */ \
|
|---|
| 175 | /* Expected value must be in eax. */ \
|
|---|
| 176 | [expval] "a" (exp_val), \
|
|---|
| 177 | /* The new value may be in any register. */ \
|
|---|
| 178 | [newval] "r" (new_val) \
|
|---|
| 179 | : "memory" \
|
|---|
| 180 | ); \
|
|---|
| 181 | break; \
|
|---|
| 182 | case 4: \
|
|---|
| 183 | asm volatile ( \
|
|---|
| 184 | prefix " cmpxchgl %[newval], %[ptr]\n" \
|
|---|
| 185 | : /* Output operands. */ \
|
|---|
| 186 | /* Old/current value is returned in eax. */ \
|
|---|
| 187 | [oldval] "=a" (old_val), \
|
|---|
| 188 | /* (*ptr) will be read and written to, hence "+" */ \
|
|---|
| 189 | [ptr] "+m" (*pptr) \
|
|---|
| 190 | : /* Input operands. */ \
|
|---|
| 191 | /* Expected value must be in eax. */ \
|
|---|
| 192 | [expval] "a" (exp_val), \
|
|---|
| 193 | /* The new value may be in any register. */ \
|
|---|
| 194 | [newval] "r" (new_val) \
|
|---|
| 195 | : "memory" \
|
|---|
| 196 | ); \
|
|---|
| 197 | break; \
|
|---|
| 198 | } \
|
|---|
| 199 | })
|
|---|
| 200 |
|
|---|
| 201 |
|
|---|
| 202 | #ifndef local_atomic_cas
|
|---|
| 203 |
|
|---|
| 204 | #define local_atomic_cas(pptr, exp_val, new_val) \
|
|---|
| 205 | ({ \
|
|---|
| 206 | /* Use proper types and avoid name clashes */ \
|
|---|
| 207 | typeof(*(pptr)) _old_val_cas; \
|
|---|
| 208 | typeof(*(pptr)) _exp_val_cas = exp_val; \
|
|---|
| 209 | typeof(*(pptr)) _new_val_cas = new_val; \
|
|---|
| 210 | _atomic_cas_impl(pptr, _exp_val_cas, _new_val_cas, _old_val_cas, ""); \
|
|---|
| 211 | \
|
|---|
| 212 | _old_val_cas; \
|
|---|
| 213 | })
|
|---|
| 214 |
|
|---|
| 215 | #else
|
|---|
| 216 | /* Check if arch/atomic.h does not accidentally include /atomic.h .*/
|
|---|
| 217 | #error Architecture specific cpu local atomics already defined! Check your includes.
|
|---|
| 218 | #endif
|
|---|
| 219 |
|
|---|
| 220 |
|
|---|
| 221 | #ifndef local_atomic_exchange
|
|---|
| 222 | /*
|
|---|
| 223 | * Issuing a xchg instruction always implies lock prefix semantics.
|
|---|
| 224 | * Therefore, it is cheaper to use a cmpxchg without a lock prefix
|
|---|
| 225 | * in a loop.
|
|---|
| 226 | */
|
|---|
| 227 | #define local_atomic_exchange(pptr, new_val) \
|
|---|
| 228 | ({ \
|
|---|
| 229 | /* Use proper types and avoid name clashes */ \
|
|---|
| 230 | typeof(*(pptr)) _exp_val_x; \
|
|---|
| 231 | typeof(*(pptr)) _old_val_x; \
|
|---|
| 232 | typeof(*(pptr)) _new_val_x = new_val; \
|
|---|
| 233 | \
|
|---|
| 234 | do { \
|
|---|
| 235 | _exp_val_x = *pptr; \
|
|---|
| 236 | _old_val_x = local_atomic_cas(pptr, _exp_val_x, _new_val_x); \
|
|---|
| 237 | } while (_old_val_x != _exp_val_x); \
|
|---|
| 238 | \
|
|---|
| 239 | _old_val_x; \
|
|---|
| 240 | })
|
|---|
| 241 |
|
|---|
| 242 | #else
|
|---|
| 243 | /* Check if arch/atomic.h does not accidentally include /atomic.h .*/
|
|---|
| 244 | #error Architecture specific cpu local atomics already defined! Check your includes.
|
|---|
| 245 | #endif
|
|---|
| 246 |
|
|---|
| 247 |
|
|---|
| 248 | #endif
|
|---|
| 249 |
|
|---|
| 250 | /** @}
|
|---|
| 251 | */
|
|---|