1 |
1275 |
phoenix |
/*
|
2 |
|
|
* PowerPC64 atomic bit operations.
|
3 |
|
|
* Dave Engebretsen, Todd Inglett, Don Reed, Pat McCarthy, Peter Bergner,
|
4 |
|
|
* Anton Blanchard
|
5 |
|
|
*
|
6 |
|
|
* Originally taken from the 32b PPC code. Modified to use 64b values for
|
7 |
|
|
* the various counters & memory references.
|
8 |
|
|
*
|
9 |
|
|
* Bitops are odd when viewed on big-endian systems. They were designed
|
10 |
|
|
* on little endian so the size of the bitset doesn't matter (low order bytes
|
11 |
|
|
* come first) as long as the bit in question is valid.
|
12 |
|
|
*
|
13 |
|
|
* Bits are "tested" often using the C expression (val & (1<<nr)) so we do
|
14 |
|
|
* our best to stay compatible with that. The assumption is that val will
|
15 |
|
|
* be unsigned long for such tests. As such, we assume the bits are stored
|
16 |
|
|
* as an array of unsigned long (the usual case is a single unsigned long,
|
17 |
|
|
* of course). Here's an example bitset with bit numbering:
|
18 |
|
|
*
|
19 |
|
|
* |63..........0|127........64|195.......128|255.......196|
|
20 |
|
|
*
|
21 |
|
|
* This leads to a problem. If an int, short or char is passed as a bitset
|
22 |
|
|
* it will be a bad memory reference since we want to store in chunks
|
23 |
|
|
* of unsigned long (64 bits here) size.
|
24 |
|
|
*
|
25 |
|
|
* This program is free software; you can redistribute it and/or
|
26 |
|
|
* modify it under the terms of the GNU General Public License
|
27 |
|
|
* as published by the Free Software Foundation; either version
|
28 |
|
|
* 2 of the License, or (at your option) any later version.
|
29 |
|
|
*/
|
30 |
|
|
|
31 |
|
|
#ifndef _PPC64_BITOPS_H
|
32 |
|
|
#define _PPC64_BITOPS_H
|
33 |
|
|
|
34 |
|
|
#ifdef __KERNEL__
|
35 |
|
|
|
36 |
|
|
#include <asm/memory.h>
|
37 |
|
|
|
38 |
|
|
/*
|
39 |
|
|
* clear_bit doesn't imply a memory barrier
|
40 |
|
|
*/
|
41 |
|
|
#define smp_mb__before_clear_bit() smp_mb()
|
42 |
|
|
#define smp_mb__after_clear_bit() smp_mb()
|
43 |
|
|
|
44 |
|
|
static __inline__ int test_bit(unsigned long nr, __const__ volatile void *addr)
|
45 |
|
|
{
|
46 |
|
|
return (1UL & (((__const__ long *) addr)[nr >> 6] >> (nr & 63)));
|
47 |
|
|
}
|
48 |
|
|
|
49 |
|
|
static __inline__ void set_bit(unsigned long nr, volatile void *addr)
|
50 |
|
|
{
|
51 |
|
|
unsigned long old;
|
52 |
|
|
unsigned long mask = 1UL << (nr & 0x3f);
|
53 |
|
|
unsigned long *p = ((unsigned long *)addr) + (nr >> 6);
|
54 |
|
|
|
55 |
|
|
__asm__ __volatile__(
|
56 |
|
|
"1: ldarx %0,0,%3 # set_bit\n\
|
57 |
|
|
or %0,%0,%2\n\
|
58 |
|
|
stdcx. %0,0,%3\n\
|
59 |
|
|
bne- 1b"
|
60 |
|
|
: "=&r" (old), "=m" (*p)
|
61 |
|
|
: "r" (mask), "r" (p), "m" (*p)
|
62 |
|
|
: "cc");
|
63 |
|
|
}
|
64 |
|
|
|
65 |
|
|
static __inline__ void clear_bit(unsigned long nr, volatile void *addr)
|
66 |
|
|
{
|
67 |
|
|
unsigned long old;
|
68 |
|
|
unsigned long mask = 1UL << (nr & 0x3f);
|
69 |
|
|
unsigned long *p = ((unsigned long *)addr) + (nr >> 6);
|
70 |
|
|
|
71 |
|
|
__asm__ __volatile__(
|
72 |
|
|
"1: ldarx %0,0,%3 # clear_bit\n\
|
73 |
|
|
andc %0,%0,%2\n\
|
74 |
|
|
stdcx. %0,0,%3\n\
|
75 |
|
|
bne- 1b"
|
76 |
|
|
: "=&r" (old), "=m" (*p)
|
77 |
|
|
: "r" (mask), "r" (p), "m" (*p)
|
78 |
|
|
: "cc");
|
79 |
|
|
}
|
80 |
|
|
|
81 |
|
|
static __inline__ void change_bit(unsigned long nr, volatile void *addr)
|
82 |
|
|
{
|
83 |
|
|
unsigned long old;
|
84 |
|
|
unsigned long mask = 1UL << (nr & 0x3f);
|
85 |
|
|
unsigned long *p = ((unsigned long *)addr) + (nr >> 6);
|
86 |
|
|
|
87 |
|
|
__asm__ __volatile__(
|
88 |
|
|
"1: ldarx %0,0,%3 # change_bit\n\
|
89 |
|
|
xor %0,%0,%2\n\
|
90 |
|
|
stdcx. %0,0,%3\n\
|
91 |
|
|
bne- 1b"
|
92 |
|
|
: "=&r" (old), "=m" (*p)
|
93 |
|
|
: "r" (mask), "r" (p), "m" (*p)
|
94 |
|
|
: "cc");
|
95 |
|
|
}
|
96 |
|
|
|
97 |
|
|
static __inline__ int test_and_set_bit(unsigned long nr, volatile void *addr)
|
98 |
|
|
{
|
99 |
|
|
unsigned long old, t;
|
100 |
|
|
unsigned long mask = 1UL << (nr & 0x3f);
|
101 |
|
|
unsigned long *p = ((unsigned long *)addr) + (nr >> 6);
|
102 |
|
|
|
103 |
|
|
__asm__ __volatile__(
|
104 |
|
|
EIEIO_ON_SMP
|
105 |
|
|
"1: ldarx %0,0,%3 # test_and_set_bit\n\
|
106 |
|
|
or %1,%0,%2 \n\
|
107 |
|
|
stdcx. %1,0,%3 \n\
|
108 |
|
|
bne- 1b"
|
109 |
|
|
ISYNC_ON_SMP
|
110 |
|
|
: "=&r" (old), "=&r" (t)
|
111 |
|
|
: "r" (mask), "r" (p)
|
112 |
|
|
: "cc", "memory");
|
113 |
|
|
|
114 |
|
|
return (old & mask) != 0;
|
115 |
|
|
}
|
116 |
|
|
|
117 |
|
|
static __inline__ int test_and_clear_bit(unsigned long nr, volatile void *addr)
|
118 |
|
|
{
|
119 |
|
|
unsigned long old, t;
|
120 |
|
|
unsigned long mask = 1UL << (nr & 0x3f);
|
121 |
|
|
unsigned long *p = ((unsigned long *)addr) + (nr >> 6);
|
122 |
|
|
|
123 |
|
|
__asm__ __volatile__(
|
124 |
|
|
EIEIO_ON_SMP
|
125 |
|
|
"1: ldarx %0,0,%3 # test_and_clear_bit\n\
|
126 |
|
|
andc %1,%0,%2\n\
|
127 |
|
|
stdcx. %1,0,%3\n\
|
128 |
|
|
bne- 1b"
|
129 |
|
|
ISYNC_ON_SMP
|
130 |
|
|
: "=&r" (old), "=&r" (t)
|
131 |
|
|
: "r" (mask), "r" (p)
|
132 |
|
|
: "cc", "memory");
|
133 |
|
|
|
134 |
|
|
return (old & mask) != 0;
|
135 |
|
|
}
|
136 |
|
|
|
137 |
|
|
static __inline__ int test_and_change_bit(unsigned long nr, volatile void *addr)
|
138 |
|
|
{
|
139 |
|
|
unsigned long old, t;
|
140 |
|
|
unsigned long mask = 1UL << (nr & 0x3f);
|
141 |
|
|
unsigned long *p = ((unsigned long *)addr) + (nr >> 6);
|
142 |
|
|
|
143 |
|
|
__asm__ __volatile__(
|
144 |
|
|
EIEIO_ON_SMP
|
145 |
|
|
"1: ldarx %0,0,%3 # test_and_change_bit\n\
|
146 |
|
|
xor %1,%0,%2\n\
|
147 |
|
|
stdcx. %1,0,%3\n\
|
148 |
|
|
bne- 1b"
|
149 |
|
|
ISYNC_ON_SMP
|
150 |
|
|
: "=&r" (old), "=&r" (t)
|
151 |
|
|
: "r" (mask), "r" (p)
|
152 |
|
|
: "cc", "memory");
|
153 |
|
|
|
154 |
|
|
return (old & mask) != 0;
|
155 |
|
|
}
|
156 |
|
|
|
157 |
|
|
/*
|
158 |
|
|
* non-atomic versions
|
159 |
|
|
*/
|
160 |
|
|
static __inline__ void __set_bit(unsigned long nr, volatile void *addr)
|
161 |
|
|
{
|
162 |
|
|
unsigned long mask = 1UL << (nr & 0x3f);
|
163 |
|
|
unsigned long *p = ((unsigned long *)addr) + (nr >> 6);
|
164 |
|
|
|
165 |
|
|
*p |= mask;
|
166 |
|
|
}
|
167 |
|
|
|
168 |
|
|
static __inline__ void __clear_bit(unsigned long nr, volatile void *addr)
|
169 |
|
|
{
|
170 |
|
|
unsigned long mask = 1UL << (nr & 0x3f);
|
171 |
|
|
unsigned long *p = ((unsigned long *)addr) + (nr >> 6);
|
172 |
|
|
|
173 |
|
|
*p &= ~mask;
|
174 |
|
|
}
|
175 |
|
|
|
176 |
|
|
static __inline__ void __change_bit(unsigned long nr, volatile void *addr)
|
177 |
|
|
{
|
178 |
|
|
unsigned long mask = 1UL << (nr & 0x3f);
|
179 |
|
|
unsigned long *p = ((unsigned long *)addr) + (nr >> 6);
|
180 |
|
|
|
181 |
|
|
*p ^= mask;
|
182 |
|
|
}
|
183 |
|
|
|
184 |
|
|
static __inline__ int __test_and_set_bit(unsigned long nr, volatile void *addr)
|
185 |
|
|
{
|
186 |
|
|
unsigned long mask = 1UL << (nr & 0x3f);
|
187 |
|
|
unsigned long *p = ((unsigned long *)addr) + (nr >> 6);
|
188 |
|
|
unsigned long old = *p;
|
189 |
|
|
|
190 |
|
|
*p = old | mask;
|
191 |
|
|
return (old & mask) != 0;
|
192 |
|
|
}
|
193 |
|
|
|
194 |
|
|
static __inline__ int __test_and_clear_bit(unsigned long nr, volatile void *addr)
|
195 |
|
|
{
|
196 |
|
|
unsigned long mask = 1UL << (nr & 0x3f);
|
197 |
|
|
unsigned long *p = ((unsigned long *)addr) + (nr >> 6);
|
198 |
|
|
unsigned long old = *p;
|
199 |
|
|
|
200 |
|
|
*p = old & ~mask;
|
201 |
|
|
return (old & mask) != 0;
|
202 |
|
|
}
|
203 |
|
|
|
204 |
|
|
static __inline__ int __test_and_change_bit(unsigned long nr, volatile void *addr)
|
205 |
|
|
{
|
206 |
|
|
unsigned long mask = 1UL << (nr & 0x3f);
|
207 |
|
|
unsigned long *p = ((unsigned long *)addr) + (nr >> 6);
|
208 |
|
|
unsigned long old = *p;
|
209 |
|
|
|
210 |
|
|
*p = old ^ mask;
|
211 |
|
|
return (old & mask) != 0;
|
212 |
|
|
}
|
213 |
|
|
|
214 |
|
|
/*
|
215 |
|
|
* Return the zero-based bit position (from RIGHT TO LEFT, 63 -> 0) of the
|
216 |
|
|
* most significant (left-most) 1-bit in a double word.
|
217 |
|
|
*/
|
218 |
|
|
static __inline__ int __ilog2(unsigned long x)
|
219 |
|
|
{
|
220 |
|
|
int lz;
|
221 |
|
|
|
222 |
|
|
asm ("cntlzd %0,%1" : "=r" (lz) : "r" (x));
|
223 |
|
|
return 63 - lz;
|
224 |
|
|
}
|
225 |
|
|
|
226 |
|
|
/* Return the zero-based bit position
|
227 |
|
|
* from RIGHT TO LEFT 63 --> 0
|
228 |
|
|
* of the most significant (left-most) 1-bit in an 8-byte area.
|
229 |
|
|
*/
|
230 |
|
|
static __inline__ long cnt_trailing_zeros(unsigned long mask)
|
231 |
|
|
{
|
232 |
|
|
long cnt;
|
233 |
|
|
|
234 |
|
|
asm(
|
235 |
|
|
" addi %0,%1,-1 \n\
|
236 |
|
|
andc %0,%0,%1 \n\
|
237 |
|
|
cntlzd %0,%0 \n\
|
238 |
|
|
subfic %0,%0,64"
|
239 |
|
|
: "=r" (cnt)
|
240 |
|
|
: "r" (mask));
|
241 |
|
|
return cnt;
|
242 |
|
|
}
|
243 |
|
|
|
244 |
|
|
|
245 |
|
|
/*
|
246 |
|
|
* ffz = Find First Zero in word.
|
247 |
|
|
* Determines the bit position of the least significant (rightmost) 0 bit
|
248 |
|
|
* in the specified double word. The returned bit position will be zero-based,
|
249 |
|
|
* starting from the right side (63 - 0).
|
250 |
|
|
*/
|
251 |
|
|
static __inline__ unsigned long ffz(unsigned long x)
|
252 |
|
|
{
|
253 |
|
|
/* no zero exists anywhere in the 8 byte area. */
|
254 |
|
|
if ((x = ~x) == 0)
|
255 |
|
|
return 64;
|
256 |
|
|
|
257 |
|
|
/*
|
258 |
|
|
* Calculate the bit position of the least signficant '1' bit in x
|
259 |
|
|
* (since x has been changed this will actually be the least signficant
|
260 |
|
|
* '0' bit in * the original x). Note: (x & -x) gives us a mask that
|
261 |
|
|
* is the least significant * (RIGHT-most) 1-bit of the value in x.
|
262 |
|
|
*/
|
263 |
|
|
return __ilog2(x & -x);
|
264 |
|
|
}
|
265 |
|
|
|
266 |
|
|
static __inline__ int __ffs(unsigned long x)
|
267 |
|
|
{
|
268 |
|
|
return __ilog2(x & -x);
|
269 |
|
|
}
|
270 |
|
|
|
271 |
|
|
/*
|
272 |
|
|
* ffs: find first bit set. This is defined the same way as
|
273 |
|
|
* the libc and compiler builtin ffs routines, therefore
|
274 |
|
|
* differs in spirit from the above ffz (man ffs).
|
275 |
|
|
*/
|
276 |
|
|
static __inline__ int ffs(int x)
|
277 |
|
|
{
|
278 |
|
|
unsigned long i = (unsigned long)x;
|
279 |
|
|
return __ilog2(i & -i) + 1;
|
280 |
|
|
}
|
281 |
|
|
|
282 |
|
|
/*
|
283 |
|
|
* fls: find last (most-significant) bit set.
|
284 |
|
|
* Note fls(0) = 0, fls(1) = 1, fls(0x80000000) = 32.
|
285 |
|
|
*/
|
286 |
|
|
#define fls(x) generic_fls(x)
|
287 |
|
|
|
288 |
|
|
/*
|
289 |
|
|
* hweightN: returns the hamming weight (i.e. the number
|
290 |
|
|
* of bits set) of a N-bit word
|
291 |
|
|
*/
|
292 |
|
|
#define hweight32(x) generic_hweight32(x)
|
293 |
|
|
#define hweight16(x) generic_hweight16(x)
|
294 |
|
|
#define hweight8(x) generic_hweight8(x)
|
295 |
|
|
|
296 |
|
|
extern unsigned long find_next_zero_bit(unsigned long* addr, unsigned long size, unsigned long offset);
|
297 |
|
|
#define find_first_zero_bit(addr, size) \
|
298 |
|
|
find_next_zero_bit((addr), (size), 0)
|
299 |
|
|
|
300 |
|
|
extern unsigned long find_next_bit(unsigned long* addr, unsigned long size, unsigned long offset);
|
301 |
|
|
#define find_first_bit(addr, size) \
|
302 |
|
|
find_next_bit((addr), (size), 0)
|
303 |
|
|
|
304 |
|
|
extern unsigned long find_next_zero_le_bit(unsigned long *addr, unsigned long size, unsigned long offset);
|
305 |
|
|
#define find_first_zero_le_bit(addr, size) \
|
306 |
|
|
find_next_zero_le_bit((addr), (size), 0)
|
307 |
|
|
|
308 |
|
|
static __inline__ int test_le_bit(unsigned long nr, __const__ unsigned long * addr)
|
309 |
|
|
{
|
310 |
|
|
__const__ unsigned char *ADDR = (__const__ unsigned char *) addr;
|
311 |
|
|
return (ADDR[nr >> 3] >> (nr & 7)) & 1;
|
312 |
|
|
}
|
313 |
|
|
|
314 |
|
|
/*
|
315 |
|
|
* non-atomic versions
|
316 |
|
|
*/
|
317 |
|
|
static __inline__ void __set_le_bit(unsigned long nr, unsigned long *addr)
|
318 |
|
|
{
|
319 |
|
|
unsigned char *ADDR = (unsigned char *)addr;
|
320 |
|
|
|
321 |
|
|
ADDR += nr >> 3;
|
322 |
|
|
*ADDR |= 1 << (nr & 0x07);
|
323 |
|
|
}
|
324 |
|
|
|
325 |
|
|
static __inline__ void __clear_le_bit(unsigned long nr, unsigned long *addr)
|
326 |
|
|
{
|
327 |
|
|
unsigned char *ADDR = (unsigned char *)addr;
|
328 |
|
|
|
329 |
|
|
ADDR += nr >> 3;
|
330 |
|
|
*ADDR &= ~(1 << (nr & 0x07));
|
331 |
|
|
}
|
332 |
|
|
|
333 |
|
|
static __inline__ int __test_and_set_le_bit(unsigned long nr, unsigned long *addr)
|
334 |
|
|
{
|
335 |
|
|
int mask, retval;
|
336 |
|
|
unsigned char *ADDR = (unsigned char *)addr;
|
337 |
|
|
|
338 |
|
|
ADDR += nr >> 3;
|
339 |
|
|
mask = 1 << (nr & 0x07);
|
340 |
|
|
retval = (mask & *ADDR) != 0;
|
341 |
|
|
*ADDR |= mask;
|
342 |
|
|
return retval;
|
343 |
|
|
}
|
344 |
|
|
|
345 |
|
|
static __inline__ int __test_and_clear_le_bit(unsigned long nr, unsigned long *addr)
|
346 |
|
|
{
|
347 |
|
|
int mask, retval;
|
348 |
|
|
unsigned char *ADDR = (unsigned char *)addr;
|
349 |
|
|
|
350 |
|
|
ADDR += nr >> 3;
|
351 |
|
|
mask = 1 << (nr & 0x07);
|
352 |
|
|
retval = (mask & *ADDR) != 0;
|
353 |
|
|
*ADDR &= ~mask;
|
354 |
|
|
return retval;
|
355 |
|
|
}
|
356 |
|
|
|
357 |
|
|
#define ext2_set_bit(nr,addr) \
|
358 |
|
|
__test_and_set_le_bit((nr),(unsigned long*)addr)
|
359 |
|
|
#define ext2_clear_bit(nr, addr) \
|
360 |
|
|
__test_and_clear_le_bit((nr),(unsigned long*)addr)
|
361 |
|
|
#define ext2_test_bit(nr, addr) test_le_bit((nr),(unsigned long*)addr)
|
362 |
|
|
#define ext2_find_first_zero_bit(addr, size) \
|
363 |
|
|
find_first_zero_le_bit((unsigned long*)addr, size)
|
364 |
|
|
#define ext2_find_next_zero_bit(addr, size, off) \
|
365 |
|
|
find_next_zero_le_bit((unsigned long*)addr, size, off)
|
366 |
|
|
|
367 |
|
|
#define minix_test_and_set_bit(nr,addr) test_and_set_bit(nr,addr)
|
368 |
|
|
#define minix_set_bit(nr,addr) set_bit(nr,addr)
|
369 |
|
|
#define minix_test_and_clear_bit(nr,addr) test_and_clear_bit(nr,addr)
|
370 |
|
|
#define minix_test_bit(nr,addr) test_bit(nr,addr)
|
371 |
|
|
#define minix_find_first_zero_bit(addr,size) find_first_zero_bit(addr,size)
|
372 |
|
|
|
373 |
|
|
#endif /* __KERNEL__ */
|
374 |
|
|
#endif /* _PPC64_BITOPS_H */
|