xref: /linux/arch/hexagon/include/asm/bitops.h (revision f2ee442115c9b6219083c019939a9cc0c9abb2f8)
1 /*
2  * Bit operations for the Hexagon architecture
3  *
4  * Copyright (c) 2010-2011, Code Aurora Forum. All rights reserved.
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>
27 #include <asm/system.h>
28 #include <asm/atomic.h>
29 
30 #ifdef __KERNEL__
31 
32 #define smp_mb__before_clear_bit()	barrier()
33 #define smp_mb__after_clear_bit()	barrier()
34 
35 /*
36  * The offset calculations for these are based on BITS_PER_LONG == 32
37  * (i.e. I get to shift by #5-2 (32 bits per long, 4 bytes per access),
38  * mask by 0x0000001F)
39  *
40  * Typically, R10 is clobbered for address, R11 bit nr, and R12 is temp
41  */
42 
43 /**
44  * test_and_clear_bit - clear a bit and return its old value
45  * @nr:  bit number to clear
46  * @addr:  pointer to memory
47  */
48 static inline int test_and_clear_bit(int nr, volatile void *addr)
49 {
50 	int oldval;
51 
52 	__asm__ __volatile__ (
53 	"	{R10 = %1; R11 = asr(%2,#5); }\n"
54 	"	{R10 += asl(R11,#2); R11 = and(%2,#0x1f)}\n"
55 	"1:	R12 = memw_locked(R10);\n"
56 	"	{ P0 = tstbit(R12,R11); R12 = clrbit(R12,R11); }\n"
57 	"	memw_locked(R10,P1) = R12;\n"
58 	"	{if !P1 jump 1b; %0 = mux(P0,#1,#0);}\n"
59 	: "=&r" (oldval)
60 	: "r" (addr), "r" (nr)
61 	: "r10", "r11", "r12", "p0", "p1", "memory"
62 	);
63 
64 	return oldval;
65 }
66 
67 /**
68  * test_and_set_bit - set a bit and return its old value
69  * @nr:  bit number to set
70  * @addr:  pointer to memory
71  */
72 static inline int test_and_set_bit(int nr, volatile void *addr)
73 {
74 	int oldval;
75 
76 	__asm__ __volatile__ (
77 	"	{R10 = %1; R11 = asr(%2,#5); }\n"
78 	"	{R10 += asl(R11,#2); R11 = and(%2,#0x1f)}\n"
79 	"1:	R12 = memw_locked(R10);\n"
80 	"	{ P0 = tstbit(R12,R11); R12 = setbit(R12,R11); }\n"
81 	"	memw_locked(R10,P1) = R12;\n"
82 	"	{if !P1 jump 1b; %0 = mux(P0,#1,#0);}\n"
83 	: "=&r" (oldval)
84 	: "r" (addr), "r" (nr)
85 	: "r10", "r11", "r12", "p0", "p1", "memory"
86 	);
87 
88 
89 	return oldval;
90 
91 }
92 
93 /**
94  * test_and_change_bit - toggle a bit and return its old value
95  * @nr:  bit number to set
96  * @addr:  pointer to memory
97  */
98 static inline int test_and_change_bit(int nr, volatile void *addr)
99 {
100 	int oldval;
101 
102 	__asm__ __volatile__ (
103 	"	{R10 = %1; R11 = asr(%2,#5); }\n"
104 	"	{R10 += asl(R11,#2); R11 = and(%2,#0x1f)}\n"
105 	"1:	R12 = memw_locked(R10);\n"
106 	"	{ P0 = tstbit(R12,R11); R12 = togglebit(R12,R11); }\n"
107 	"	memw_locked(R10,P1) = R12;\n"
108 	"	{if !P1 jump 1b; %0 = mux(P0,#1,#0);}\n"
109 	: "=&r" (oldval)
110 	: "r" (addr), "r" (nr)
111 	: "r10", "r11", "r12", "p0", "p1", "memory"
112 	);
113 
114 	return oldval;
115 
116 }
117 
118 /*
119  * Atomic, but doesn't care about the return value.
120  * Rewrite later to save a cycle or two.
121  */
122 
123 static inline void clear_bit(int nr, volatile void *addr)
124 {
125 	test_and_clear_bit(nr, addr);
126 }
127 
128 static inline void set_bit(int nr, volatile void *addr)
129 {
130 	test_and_set_bit(nr, addr);
131 }
132 
133 static inline void change_bit(int nr, volatile void *addr)
134 {
135 	test_and_change_bit(nr, addr);
136 }
137 
138 
139 /*
140  * These are allowed to be non-atomic.  In fact the generic flavors are
141  * in non-atomic.h.  Would it be better to use intrinsics for this?
142  *
143  * OK, writes in our architecture do not invalidate LL/SC, so this has to
144  * be atomic, particularly for things like slab_lock and slab_unlock.
145  *
146  */
147 static inline void __clear_bit(int nr, volatile unsigned long *addr)
148 {
149 	test_and_clear_bit(nr, addr);
150 }
151 
152 static inline void __set_bit(int nr, volatile unsigned long *addr)
153 {
154 	test_and_set_bit(nr, addr);
155 }
156 
157 static inline void __change_bit(int nr, volatile unsigned long *addr)
158 {
159 	test_and_change_bit(nr, addr);
160 }
161 
162 /*  Apparently, at least some of these are allowed to be non-atomic  */
163 static inline int __test_and_clear_bit(int nr, volatile unsigned long *addr)
164 {
165 	return test_and_clear_bit(nr, addr);
166 }
167 
168 static inline int __test_and_set_bit(int nr, volatile unsigned long *addr)
169 {
170 	return test_and_set_bit(nr, addr);
171 }
172 
173 static inline int __test_and_change_bit(int nr, volatile unsigned long *addr)
174 {
175 	return test_and_change_bit(nr, addr);
176 }
177 
178 static inline int __test_bit(int nr, const volatile unsigned long *addr)
179 {
180 	int retval;
181 
182 	asm volatile(
183 	"{P0 = tstbit(%1,%2); if (P0.new) %0 = #1; if (!P0.new) %0 = #0;}\n"
184 	: "=&r" (retval)
185 	: "r" (addr[BIT_WORD(nr)]), "r" (nr % BITS_PER_LONG)
186 	: "p0"
187 	);
188 
189 	return retval;
190 }
191 
192 #define test_bit(nr, addr) __test_bit(nr, addr)
193 
194 /*
195  * ffz - find first zero in word.
196  * @word: The word to search
197  *
198  * Undefined if no zero exists, so code should check against ~0UL first.
199  */
200 static inline long ffz(int x)
201 {
202 	int r;
203 
204 	asm("%0 = ct1(%1);\n"
205 		: "=&r" (r)
206 		: "r" (x));
207 	return r;
208 }
209 
210 /*
211  * fls - find last (most-significant) bit set
212  * @x: the word to search
213  *
214  * This is defined the same way as ffs.
215  * Note fls(0) = 0, fls(1) = 1, fls(0x80000000) = 32.
216  */
217 static inline long fls(int x)
218 {
219 	int r;
220 
221 	asm("{ %0 = cl0(%1);}\n"
222 		"%0 = sub(#32,%0);\n"
223 		: "=&r" (r)
224 		: "r" (x)
225 		: "p0");
226 
227 	return r;
228 }
229 
230 /*
231  * ffs - find first bit set
232  * @x: the word to search
233  *
234  * This is defined the same way as
235  * the libc and compiler builtin ffs routines, therefore
236  * differs in spirit from the above ffz (man ffs).
237  */
238 static inline long ffs(int x)
239 {
240 	int r;
241 
242 	asm("{ P0 = cmp.eq(%1,#0); %0 = ct0(%1);}\n"
243 		"{ if P0 %0 = #0; if !P0 %0 = add(%0,#1);}\n"
244 		: "=&r" (r)
245 		: "r" (x)
246 		: "p0");
247 
248 	return r;
249 }
250 
251 /*
252  * __ffs - find first bit in word.
253  * @word: The word to search
254  *
255  * Undefined if no bit exists, so code should check against 0 first.
256  *
257  * bits_per_long assumed to be 32
258  * numbering starts at 0 I think (instead of 1 like ffs)
259  */
260 static inline unsigned long __ffs(unsigned long word)
261 {
262 	int num;
263 
264 	asm("%0 = ct0(%1);\n"
265 		: "=&r" (num)
266 		: "r" (word));
267 
268 	return num;
269 }
270 
271 /*
272  * __fls - find last (most-significant) set bit in a long word
273  * @word: the word to search
274  *
275  * Undefined if no set bit exists, so code should check against 0 first.
276  * bits_per_long assumed to be 32
277  */
278 static inline unsigned long __fls(unsigned long word)
279 {
280 	int num;
281 
282 	asm("%0 = cl0(%1);\n"
283 		"%0 = sub(#31,%0);\n"
284 		: "=&r" (num)
285 		: "r" (word));
286 
287 	return num;
288 }
289 
290 #include <asm-generic/bitops/lock.h>
291 #include <asm-generic/bitops/find.h>
292 
293 #include <asm-generic/bitops/fls64.h>
294 #include <asm-generic/bitops/sched.h>
295 #include <asm-generic/bitops/hweight.h>
296 
297 #include <asm-generic/bitops/le.h>
298 #include <asm-generic/bitops/ext2-atomic.h>
299 
300 #endif /* __KERNEL__ */
301 #endif
302