Richard Kuo | a86a7ce | 2011-10-31 18:27:45 -0500 | [diff] [blame] | 1 | /* |
| 2 | * Bit operations for the Hexagon architecture |
| 3 | * |
Richard Kuo | e1858b2 | 2012-09-19 16:22:02 -0500 | [diff] [blame] | 4 | * Copyright (c) 2010-2011, The Linux Foundation. All rights reserved. |
Richard Kuo | a86a7ce | 2011-10-31 18:27:45 -0500 | [diff] [blame] | 5 | * |
| 6 | * |
| 7 | * This program is free software; you can redistribute it and/or modify |
| 8 | * it under the terms of the GNU General Public License version 2 and |
| 9 | * only version 2 as published by the Free Software Foundation. |
| 10 | * |
| 11 | * This program is distributed in the hope that it will be useful, |
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 14 | * GNU General Public License for more details. |
| 15 | * |
| 16 | * You should have received a copy of the GNU General Public License |
| 17 | * along with this program; if not, write to the Free Software |
| 18 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA |
| 19 | * 02110-1301, USA. |
| 20 | */ |
| 21 | |
| 22 | #ifndef _ASM_BITOPS_H |
| 23 | #define _ASM_BITOPS_H |
| 24 | |
| 25 | #include <linux/compiler.h> |
| 26 | #include <asm/byteorder.h> |
Richard Kuo | a86a7ce | 2011-10-31 18:27:45 -0500 | [diff] [blame] | 27 | #include <asm/atomic.h> |
| 28 | |
| 29 | #ifdef __KERNEL__ |
| 30 | |
| 31 | #define smp_mb__before_clear_bit() barrier() |
| 32 | #define smp_mb__after_clear_bit() barrier() |
| 33 | |
| 34 | /* |
| 35 | * The offset calculations for these are based on BITS_PER_LONG == 32 |
| 36 | * (i.e. I get to shift by #5-2 (32 bits per long, 4 bytes per access), |
| 37 | * mask by 0x0000001F) |
| 38 | * |
| 39 | * Typically, R10 is clobbered for address, R11 bit nr, and R12 is temp |
| 40 | */ |
| 41 | |
| 42 | /** |
| 43 | * test_and_clear_bit - clear a bit and return its old value |
| 44 | * @nr: bit number to clear |
| 45 | * @addr: pointer to memory |
| 46 | */ |
| 47 | static inline int test_and_clear_bit(int nr, volatile void *addr) |
| 48 | { |
| 49 | int oldval; |
| 50 | |
| 51 | __asm__ __volatile__ ( |
| 52 | " {R10 = %1; R11 = asr(%2,#5); }\n" |
| 53 | " {R10 += asl(R11,#2); R11 = and(%2,#0x1f)}\n" |
| 54 | "1: R12 = memw_locked(R10);\n" |
| 55 | " { P0 = tstbit(R12,R11); R12 = clrbit(R12,R11); }\n" |
| 56 | " memw_locked(R10,P1) = R12;\n" |
| 57 | " {if !P1 jump 1b; %0 = mux(P0,#1,#0);}\n" |
| 58 | : "=&r" (oldval) |
| 59 | : "r" (addr), "r" (nr) |
| 60 | : "r10", "r11", "r12", "p0", "p1", "memory" |
| 61 | ); |
| 62 | |
| 63 | return oldval; |
| 64 | } |
| 65 | |
| 66 | /** |
| 67 | * test_and_set_bit - set a bit and return its old value |
| 68 | * @nr: bit number to set |
| 69 | * @addr: pointer to memory |
| 70 | */ |
| 71 | static inline int test_and_set_bit(int nr, volatile void *addr) |
| 72 | { |
| 73 | int oldval; |
| 74 | |
| 75 | __asm__ __volatile__ ( |
| 76 | " {R10 = %1; R11 = asr(%2,#5); }\n" |
| 77 | " {R10 += asl(R11,#2); R11 = and(%2,#0x1f)}\n" |
| 78 | "1: R12 = memw_locked(R10);\n" |
| 79 | " { P0 = tstbit(R12,R11); R12 = setbit(R12,R11); }\n" |
| 80 | " memw_locked(R10,P1) = R12;\n" |
| 81 | " {if !P1 jump 1b; %0 = mux(P0,#1,#0);}\n" |
| 82 | : "=&r" (oldval) |
| 83 | : "r" (addr), "r" (nr) |
| 84 | : "r10", "r11", "r12", "p0", "p1", "memory" |
| 85 | ); |
| 86 | |
| 87 | |
| 88 | return oldval; |
| 89 | |
| 90 | } |
| 91 | |
| 92 | /** |
| 93 | * test_and_change_bit - toggle a bit and return its old value |
| 94 | * @nr: bit number to set |
| 95 | * @addr: pointer to memory |
| 96 | */ |
| 97 | static inline int test_and_change_bit(int nr, volatile void *addr) |
| 98 | { |
| 99 | int oldval; |
| 100 | |
| 101 | __asm__ __volatile__ ( |
| 102 | " {R10 = %1; R11 = asr(%2,#5); }\n" |
| 103 | " {R10 += asl(R11,#2); R11 = and(%2,#0x1f)}\n" |
| 104 | "1: R12 = memw_locked(R10);\n" |
| 105 | " { P0 = tstbit(R12,R11); R12 = togglebit(R12,R11); }\n" |
| 106 | " memw_locked(R10,P1) = R12;\n" |
| 107 | " {if !P1 jump 1b; %0 = mux(P0,#1,#0);}\n" |
| 108 | : "=&r" (oldval) |
| 109 | : "r" (addr), "r" (nr) |
| 110 | : "r10", "r11", "r12", "p0", "p1", "memory" |
| 111 | ); |
| 112 | |
| 113 | return oldval; |
| 114 | |
| 115 | } |
| 116 | |
| 117 | /* |
| 118 | * Atomic, but doesn't care about the return value. |
| 119 | * Rewrite later to save a cycle or two. |
| 120 | */ |
| 121 | |
| 122 | static inline void clear_bit(int nr, volatile void *addr) |
| 123 | { |
| 124 | test_and_clear_bit(nr, addr); |
| 125 | } |
| 126 | |
| 127 | static inline void set_bit(int nr, volatile void *addr) |
| 128 | { |
| 129 | test_and_set_bit(nr, addr); |
| 130 | } |
| 131 | |
| 132 | static inline void change_bit(int nr, volatile void *addr) |
| 133 | { |
| 134 | test_and_change_bit(nr, addr); |
| 135 | } |
| 136 | |
| 137 | |
| 138 | /* |
| 139 | * These are allowed to be non-atomic. In fact the generic flavors are |
| 140 | * in non-atomic.h. Would it be better to use intrinsics for this? |
| 141 | * |
| 142 | * OK, writes in our architecture do not invalidate LL/SC, so this has to |
| 143 | * be atomic, particularly for things like slab_lock and slab_unlock. |
| 144 | * |
| 145 | */ |
| 146 | static inline void __clear_bit(int nr, volatile unsigned long *addr) |
| 147 | { |
| 148 | test_and_clear_bit(nr, addr); |
| 149 | } |
| 150 | |
| 151 | static inline void __set_bit(int nr, volatile unsigned long *addr) |
| 152 | { |
| 153 | test_and_set_bit(nr, addr); |
| 154 | } |
| 155 | |
| 156 | static inline void __change_bit(int nr, volatile unsigned long *addr) |
| 157 | { |
| 158 | test_and_change_bit(nr, addr); |
| 159 | } |
| 160 | |
| 161 | /* Apparently, at least some of these are allowed to be non-atomic */ |
| 162 | static inline int __test_and_clear_bit(int nr, volatile unsigned long *addr) |
| 163 | { |
| 164 | return test_and_clear_bit(nr, addr); |
| 165 | } |
| 166 | |
| 167 | static inline int __test_and_set_bit(int nr, volatile unsigned long *addr) |
| 168 | { |
| 169 | return test_and_set_bit(nr, addr); |
| 170 | } |
| 171 | |
| 172 | static inline int __test_and_change_bit(int nr, volatile unsigned long *addr) |
| 173 | { |
| 174 | return test_and_change_bit(nr, addr); |
| 175 | } |
| 176 | |
| 177 | static inline int __test_bit(int nr, const volatile unsigned long *addr) |
| 178 | { |
| 179 | int retval; |
| 180 | |
| 181 | asm volatile( |
| 182 | "{P0 = tstbit(%1,%2); if (P0.new) %0 = #1; if (!P0.new) %0 = #0;}\n" |
| 183 | : "=&r" (retval) |
| 184 | : "r" (addr[BIT_WORD(nr)]), "r" (nr % BITS_PER_LONG) |
| 185 | : "p0" |
| 186 | ); |
| 187 | |
| 188 | return retval; |
| 189 | } |
| 190 | |
| 191 | #define test_bit(nr, addr) __test_bit(nr, addr) |
| 192 | |
| 193 | /* |
| 194 | * ffz - find first zero in word. |
| 195 | * @word: The word to search |
| 196 | * |
| 197 | * Undefined if no zero exists, so code should check against ~0UL first. |
| 198 | */ |
| 199 | static inline long ffz(int x) |
| 200 | { |
| 201 | int r; |
| 202 | |
| 203 | asm("%0 = ct1(%1);\n" |
| 204 | : "=&r" (r) |
| 205 | : "r" (x)); |
| 206 | return r; |
| 207 | } |
| 208 | |
| 209 | /* |
| 210 | * fls - find last (most-significant) bit set |
| 211 | * @x: the word to search |
| 212 | * |
| 213 | * This is defined the same way as ffs. |
| 214 | * Note fls(0) = 0, fls(1) = 1, fls(0x80000000) = 32. |
| 215 | */ |
| 216 | static inline long fls(int x) |
| 217 | { |
| 218 | int r; |
| 219 | |
| 220 | asm("{ %0 = cl0(%1);}\n" |
| 221 | "%0 = sub(#32,%0);\n" |
| 222 | : "=&r" (r) |
| 223 | : "r" (x) |
| 224 | : "p0"); |
| 225 | |
| 226 | return r; |
| 227 | } |
| 228 | |
| 229 | /* |
| 230 | * ffs - find first bit set |
| 231 | * @x: the word to search |
| 232 | * |
| 233 | * This is defined the same way as |
| 234 | * the libc and compiler builtin ffs routines, therefore |
| 235 | * differs in spirit from the above ffz (man ffs). |
| 236 | */ |
| 237 | static inline long ffs(int x) |
| 238 | { |
| 239 | int r; |
| 240 | |
| 241 | asm("{ P0 = cmp.eq(%1,#0); %0 = ct0(%1);}\n" |
| 242 | "{ if P0 %0 = #0; if !P0 %0 = add(%0,#1);}\n" |
| 243 | : "=&r" (r) |
| 244 | : "r" (x) |
| 245 | : "p0"); |
| 246 | |
| 247 | return r; |
| 248 | } |
| 249 | |
| 250 | /* |
| 251 | * __ffs - find first bit in word. |
| 252 | * @word: The word to search |
| 253 | * |
| 254 | * Undefined if no bit exists, so code should check against 0 first. |
| 255 | * |
| 256 | * bits_per_long assumed to be 32 |
| 257 | * numbering starts at 0 I think (instead of 1 like ffs) |
| 258 | */ |
| 259 | static inline unsigned long __ffs(unsigned long word) |
| 260 | { |
| 261 | int num; |
| 262 | |
| 263 | asm("%0 = ct0(%1);\n" |
| 264 | : "=&r" (num) |
| 265 | : "r" (word)); |
| 266 | |
| 267 | return num; |
| 268 | } |
| 269 | |
| 270 | /* |
| 271 | * __fls - find last (most-significant) set bit in a long word |
| 272 | * @word: the word to search |
| 273 | * |
| 274 | * Undefined if no set bit exists, so code should check against 0 first. |
| 275 | * bits_per_long assumed to be 32 |
| 276 | */ |
| 277 | static inline unsigned long __fls(unsigned long word) |
| 278 | { |
| 279 | int num; |
| 280 | |
| 281 | asm("%0 = cl0(%1);\n" |
| 282 | "%0 = sub(#31,%0);\n" |
| 283 | : "=&r" (num) |
| 284 | : "r" (word)); |
| 285 | |
| 286 | return num; |
| 287 | } |
| 288 | |
| 289 | #include <asm-generic/bitops/lock.h> |
| 290 | #include <asm-generic/bitops/find.h> |
| 291 | |
| 292 | #include <asm-generic/bitops/fls64.h> |
| 293 | #include <asm-generic/bitops/sched.h> |
| 294 | #include <asm-generic/bitops/hweight.h> |
| 295 | |
| 296 | #include <asm-generic/bitops/le.h> |
| 297 | #include <asm-generic/bitops/ext2-atomic.h> |
| 298 | |
| 299 | #endif /* __KERNEL__ */ |
| 300 | #endif |