File: /usr/src/linux/include/asm-s390x/bitops.h

1     /*
2      *  include/asm-s390/bitops.h
3      *
4      *  S390 version
5      *    Copyright (C) 1999 IBM Deutschland Entwicklung GmbH, IBM Corporation
6      *    Author(s): Martin Schwidefsky (schwidefsky@de.ibm.com)
7      *
8      *  Derived from "include/asm-i386/bitops.h"
9      *    Copyright (C) 1992, Linus Torvalds
10      *
11      */
12     
13     #ifndef _S390_BITOPS_H
14     #define _S390_BITOPS_H
15     
16     /*
17      * bit 0 is the LSB of *addr; bit 63 is the MSB of *addr;
18      * bit 64 is the LSB of *(addr+8). That combined with the
19      * big endian byte order on S390 give the following bit
20      * order in memory:
21      *    3f 3e 3d 3c 3b 3a 39 38 37 36 35 34 33 32 31 30
22      *    2f 2e 2d 2c 2b 2a 29 28 27 26 25 24 23 22 21 20
23      *    1f 1e 1d 1c 1b 1a 19 18 17 16 15 14 13 12 11 10
24      *    0f 0e 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00
25      * after that follows the next long with bit numbers
26      *    7f 7e 7d 7c 7b 7a 79 78 77 76 75 74 73 72 71 70
27      *    6f 6e 6d 6c 6b 6a 69 68 67 66 65 64 63 62 61 60
28      *    5f 5e 5d 5c 5b 5a 59 58 57 56 55 54 53 52 51 50
29      *    4f 4e 4d 4c 4b 4a 49 48 47 46 45 44 43 42 41 40
30      * The reason for this bit ordering is the fact that
31      * in the architecture independent code bits operations
32      * of the form "flags |= (1 << bitnr)" are used INTERMIXED
33      * with operation of the form "set_bit(bitnr, flags)".
34      */
35     #include <linux/config.h>
36     
37     /* set ALIGN_CS to 1 if the SMP safe bit operations should
38      * align the address to 4 byte boundary. It seems to work
39      * without the alignment. 
40      */
41     #ifdef __KERNEL__
42     #define ALIGN_CS 0
43     #else
44     #define ALIGN_CS 1
45     #ifndef CONFIG_SMP
46     #error "bitops won't work without CONFIG_SMP"
47     #endif
48     #endif
49     
50     /* bitmap tables from arch/S390/kernel/bitmap.S */
51     extern const char _oi_bitmap[];
52     extern const char _ni_bitmap[];
53     extern const char _zb_findmap[];
54     
55     #ifdef CONFIG_SMP
56     /*
57      * SMP save set_bit routine based on compare and swap (CS)
58      */
59     static __inline__ void set_bit_cs(unsigned long nr, volatile void * addr)
60     {
61             unsigned long bits, mask;
62             __asm__ __volatile__(
63     #if ALIGN_CS == 1
64                  "   lghi  %2,7\n"         /* CS must be aligned on 4 byte b. */
65                  "   ngr   %2,%1\n"        /* isolate last 2 bits of address */
66                  "   xgr   %1,%2\n"        /* make addr % 4 == 0 */
67                  "   sllg  %2,%2,3\n"
68                  "   agr   %0,%2\n"        /* add alignement to bitnr */
69     #endif
70                  "   lghi  %2,63\n"
71                  "   nr    %2,%0\n"        /* make shift value */
72                  "   xr    %0,%2\n"
73                  "   srlg  %0,%0,3\n"
74                  "   lghi  %3,1\n"
75                  "   la    %1,0(%0,%1)\n"  /* calc. address for CS */
76                  "   sllg  %3,%3,0(%2)\n"  /* make OR mask */
77                  "   lg    %0,0(%1)\n"
78                  "0: lgr   %2,%0\n"        /* CS loop starts here */
79                  "   ogr   %2,%3\n"        /* set bit */
80                  "   csg   %0,%2,0(%1)\n"
81                  "   jl    0b"
82                  : "+a" (nr), "+a" (addr), "=a" (bits), "=d" (mask) :
83                  : "cc", "memory" );
84     }
85     
86     /*
87      * SMP save clear_bit routine based on compare and swap (CS)
88      */
89     static __inline__ void clear_bit_cs(unsigned long nr, volatile void * addr)
90     {
91             unsigned long bits, mask;
92             __asm__ __volatile__(
93     #if ALIGN_CS == 1
94                  "   lghi  %2,7\n"         /* CS must be aligned on 4 byte b. */
95                  "   ngr   %2,%1\n"        /* isolate last 2 bits of address */
96                  "   xgr   %1,%2\n"        /* make addr % 4 == 0 */
97                  "   sllg  %2,%2,3\n"
98                  "   agr   %0,%2\n"        /* add alignement to bitnr */
99     #endif
100                  "   lghi  %2,63\n"
101                  "   nr    %2,%0\n"        /* make shift value */
102                  "   xr    %0,%2\n"
103                  "   srlg  %0,%0,3\n"
104                  "   lghi  %3,-2\n"
105                  "   la    %1,0(%0,%1)\n"  /* calc. address for CS */
106                  "   lghi  %3,-2\n"
107                  "   rllg  %3,%3,0(%2)\n"  /* make AND mask */
108                  "   lg    %0,0(%1)\n"
109                  "0: lgr   %2,%0\n"        /* CS loop starts here */
110                  "   ngr   %2,%3\n"        /* clear bit */
111                  "   csg   %0,%2,0(%1)\n"
112                  "   jl    0b"
113                  : "+a" (nr), "+a" (addr), "=a" (bits), "=d" (mask) :
114                  : "cc", "memory" );
115     }
116     
117     /*
118      * SMP save change_bit routine based on compare and swap (CS)
119      */
120     static __inline__ void change_bit_cs(unsigned long nr, volatile void * addr)
121     {
122             unsigned long bits, mask;
123             __asm__ __volatile__(
124     #if ALIGN_CS == 1
125                  "   lghi  %2,7\n"         /* CS must be aligned on 4 byte b. */
126                  "   ngr   %2,%1\n"        /* isolate last 2 bits of address */
127                  "   xgr   %1,%2\n"        /* make addr % 4 == 0 */
128                  "   sllg  %2,%2,3\n"
129                  "   agr   %0,%2\n"        /* add alignement to bitnr */
130     #endif
131                  "   lghi  %2,63\n"
132                  "   nr    %2,%0\n"        /* make shift value */
133                  "   xr    %0,%2\n"
134                  "   srlg  %0,%0,3\n"
135                  "   lghi  %3,1\n"
136                  "   la    %1,0(%0,%1)\n"  /* calc. address for CS */
137                  "   sllg  %3,%3,0(%2)\n"  /* make XR mask */
138                  "   lg    %0,0(%1)\n"
139                  "0: lgr   %2,%0\n"        /* CS loop starts here */
140                  "   xgr   %2,%3\n"        /* change bit */
141                  "   csg   %0,%2,0(%1)\n"
142                  "   jl    0b"
143                  : "+a" (nr), "+a" (addr), "=a" (bits), "=d" (mask) : 
144                  : "cc", "memory" );
145     }
146     
147     /*
148      * SMP save test_and_set_bit routine based on compare and swap (CS)
149      */
150     static __inline__ int 
151     test_and_set_bit_cs(unsigned long nr, volatile void * addr)
152     {
153             unsigned long bits, mask;
154             __asm__ __volatile__(
155     #if ALIGN_CS == 1
156                  "   lghi  %2,7\n"         /* CS must be aligned on 4 byte b. */
157                  "   ngr   %2,%1\n"        /* isolate last 2 bits of address */
158                  "   xgr   %1,%2\n"        /* make addr % 4 == 0 */
159                  "   sllg  %2,%2,3\n"
160                  "   agr   %0,%2\n"        /* add alignement to bitnr */
161     #endif
162                  "   lghi  %2,63\n"
163                  "   nr    %2,%0\n"        /* make shift value */
164                  "   xr    %0,%2\n"
165                  "   srlg  %0,%0,3\n"
166                  "   lghi  %3,1\n"
167                  "   la    %1,0(%0,%1)\n"  /* calc. address for CS */
168                  "   sllg  %3,%3,0(%2)\n"  /* make OR mask */
169                  "   lg    %0,0(%1)\n"
170                  "0: lgr   %2,%0\n"        /* CS loop starts here */
171                  "   ogr   %2,%3\n"        /* set bit */
172                  "   csg   %0,%2,0(%1)\n"
173                  "   jl    0b\n"
174                  "   ngr   %0,%3\n"        /* isolate old bit */
175                  : "+a" (nr), "+a" (addr), "=a" (bits), "=d" (mask) :
176                  : "cc", "memory" );
177             return nr != 0;
178     }
179     
180     /*
181      * SMP save test_and_clear_bit routine based on compare and swap (CS)
182      */
183     static __inline__ int
184     test_and_clear_bit_cs(unsigned long nr, volatile void * addr)
185     {
186             unsigned long bits, mask;
187             __asm__ __volatile__(
188     #if ALIGN_CS == 1
189                  "   lghi  %2,7\n"         /* CS must be aligned on 4 byte b. */
190                  "   ngr   %2,%1\n"        /* isolate last 2 bits of address */
191                  "   xgr   %1,%2\n"        /* make addr % 4 == 0 */
192                  "   sllg  %2,%2,3\n"
193                  "   agr   %0,%2\n"        /* add alignement to bitnr */
194     #endif
195                  "   lghi  %2,63\n"
196                  "   nr    %2,%0\n"        /* make shift value */
197                  "   xr    %0,%2\n"
198                  "   srlg  %0,%0,3\n"
199                  "   lghi  %3,-2\n"
200                  "   la    %1,0(%0,%1)\n"  /* calc. address for CS */
201                  "   rllg  %3,%3,0(%2)\n"  /* make AND mask */
202                  "   lg    %0,0(%1)\n"
203                  "0: lgr   %2,%0\n"        /* CS loop starts here */
204                  "   ngr   %2,%3\n"        /* clear bit */
205                  "   csg   %0,%2,0(%1)\n"
206                  "   jl    0b\n"
207                  "   xgr   %0,%2\n"        /* isolate old bit */
208                  : "+a" (nr), "+a" (addr), "=a" (bits), "=d" (mask) :
209                  : "cc", "memory" );
210             return nr != 0;
211     }
212     
213     /*
214      * SMP save test_and_change_bit routine based on compare and swap (CS) 
215      */
216     static __inline__ int
217     test_and_change_bit_cs(unsigned long nr, volatile void * addr)
218     {
219             unsigned long bits, mask;
220             __asm__ __volatile__(
221     #if ALIGN_CS == 1
222                  "   lghi  %2,7\n"         /* CS must be aligned on 4 byte b. */
223                  "   ngr   %2,%1\n"        /* isolate last 2 bits of address */
224                  "   xgr   %1,%2\n"        /* make addr % 4 == 0 */
225                  "   sllg  %2,%2,3\n"
226                  "   agr   %0,%2\n"        /* add alignement to bitnr */
227     #endif
228                  "   lghi  %2,63\n"
229                  "   nr    %2,%0\n"        /* make shift value */
230                  "   xr    %0,%2\n"
231                  "   srlg  %0,%0,3\n"
232                  "   lghi  %3,1\n"
233                  "   la    %1,0(%0,%1)\n"  /* calc. address for CS */
234                  "   sllg  %3,%3,0(%2)\n"  /* make OR mask */
235                  "   lg    %0,0(%1)\n"
236                  "0: lgr   %2,%0\n"        /* CS loop starts here */
237                  "   xgr   %2,%3\n"        /* change bit */
238                  "   csg   %0,%2,0(%1)\n"
239                  "   jl    0b\n"
240                  "   ngr   %0,%3\n"        /* isolate old bit */
241                  : "+a" (nr), "+a" (addr), "=a" (bits), "=d" (mask) :
242                  : "cc", "memory" );
243             return nr != 0;
244     }
245     #endif /* CONFIG_SMP */
246     
247     /*
248      * fast, non-SMP set_bit routine
249      */
250     static __inline__ void __set_bit(unsigned long nr, volatile void * addr)
251     {
252             __asm__ __volatile__(
253                  "   lghi  2,56\n"
254                  "   lghi  1,7\n"
255                  "   xgr   2,%0\n"
256                  "   nr    1,%0\n"
257                  "   srlg  2,2,3\n"
258                  "   la    2,0(2,%1)\n"
259                  "   la    1,0(1,%2)\n"
260                  "   oc    0(1,2),0(1)"
261                  :  : "a" (nr), "a" (addr), "a" (&_oi_bitmap)
262                  : "cc", "memory", "1", "2" );
263     }
264     
265     static __inline__ void 
266     __constant_set_bit(const unsigned long nr, volatile void * addr)
267     {
268       switch (nr&7) {
269       case 0:
270         __asm__ __volatile__ ("la 1,%0\n\t"
271                               "oi 0(1),0x01"
272                               : "=m" (*((volatile char *) addr + ((nr>>3)^7))) 
273                               : : "1", "cc", "memory");
274         break;
275       case 1:
276         __asm__ __volatile__ ("la 1,%0\n\t"
277                               "oi 0(1),0x02"
278                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
279                               : : "1", "cc", "memory" );
280         break;
281       case 2:
282         __asm__ __volatile__ ("la 1,%0\n\t"
283                               "oi 0(1),0x04"
284                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
285                               : : "1", "cc", "memory" );
286         break;
287       case 3:
288         __asm__ __volatile__ ("la 1,%0\n\t"
289                               "oi 0(1),0x08"
290                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
291                               : : "1", "cc", "memory" );
292         break;
293       case 4:
294         __asm__ __volatile__ ("la 1,%0\n\t"
295                               "oi 0(1),0x10"
296                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
297                               : : "1", "cc", "memory" );
298         break;
299       case 5:
300         __asm__ __volatile__ ("la 1,%0\n\t"
301                               "oi 0(1),0x20"
302                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
303                               : : "1", "cc", "memory" );
304         break;
305       case 6:
306         __asm__ __volatile__ ("la 1,%0\n\t"
307                               "oi 0(1),0x40"
308                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
309                               : : "1", "cc", "memory" );
310         break;
311       case 7:
312         __asm__ __volatile__ ("la 1,%0\n\t"
313                               "oi 0(1),0x80"
314                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
315                               : : "1", "cc", "memory" );
316         break;
317       }
318     }
319     
320     #define set_bit_simple(nr,addr) \
321     (__builtin_constant_p((nr)) ? \
322      __constant_set_bit((nr),(addr)) : \
323      __set_bit((nr),(addr)) )
324     
325     /*
326      * fast, non-SMP clear_bit routine
327      */
328     static __inline__ void 
329     __clear_bit(unsigned long nr, volatile void * addr)
330     {
331             __asm__ __volatile__(
332                  "   lghi  2,56\n"
333                  "   lghi  1,7\n"
334                  "   xgr   2,%0\n"
335                  "   nr    1,%0\n"
336                  "   srlg  2,2,3\n"
337                  "   la    2,0(2,%1)\n"
338                  "   la    1,0(1,%2)\n"
339                  "   nc    0(1,2),0(1)"
340                  :  : "d" (nr), "a" (addr), "a" (&_ni_bitmap)
341                  : "cc", "memory", "1", "2" );
342     }
343     
344     static __inline__ void 
345     __constant_clear_bit(const unsigned long nr, volatile void * addr)
346     {
347       switch (nr&7) {
348       case 0:
349         __asm__ __volatile__ ("la 1,%0\n\t"
350                               "ni 0(1),0xFE"
351                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
352                               : : "1", "cc", "memory" );
353         break;
354       case 1:
355         __asm__ __volatile__ ("la 1,%0\n\t"
356                               "ni 0(1),0xFD"
357                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
358                               : : "1", "cc", "memory" );
359         break;
360       case 2:
361         __asm__ __volatile__ ("la 1,%0\n\t"
362                               "ni 0(1),0xFB"
363                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
364                               : : "1", "cc", "memory" );
365         break;
366       case 3:
367         __asm__ __volatile__ ("la 1,%0\n\t"
368                               "ni 0(1),0xF7"
369                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
370                               : : "1", "cc", "memory" );
371         break;
372       case 4:
373         __asm__ __volatile__ ("la 1,%0\n\t"
374                               "ni 0(1),0xEF"
375                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
376                               : : "cc", "memory" );
377         break;
378       case 5:
379         __asm__ __volatile__ ("la 1,%0\n\t"
380                               "ni 0(1),0xDF"
381                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
382                               : : "1", "cc", "memory" );
383         break;
384       case 6:
385         __asm__ __volatile__ ("la 1,%0\n\t"
386                               "ni 0(1),0xBF"
387                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
388                               : : "1", "cc", "memory" );
389         break;
390       case 7:
391         __asm__ __volatile__ ("la 1,%0\n\t"
392                               "ni 0(1),0x7F"
393                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
394                               : : "1", "cc", "memory" );
395         break;
396       }
397     }
398     
399     #define clear_bit_simple(nr,addr) \
400     (__builtin_constant_p((nr)) ? \
401      __constant_clear_bit((nr),(addr)) : \
402      __clear_bit((nr),(addr)) )
403     
404     /* 
405      * fast, non-SMP change_bit routine 
406      */
407     static __inline__ void __change_bit(unsigned long nr, volatile void * addr)
408     {
409             __asm__ __volatile__(
410                  "   lghi  2,56\n"
411                  "   lghi  1,7\n"
412                  "   xgr   2,%0\n"
413                  "   nr    1,%0\n"
414                  "   srlg  2,2,3\n"
415                  "   la    2,0(2,%1)\n"
416                  "   la    1,0(1,%2)\n"
417                  "   xc    0(1,2),0(1)"
418                  :  : "d" (nr), "a" (addr), "a" (&_oi_bitmap)
419                  : "cc", "memory", "1", "2" );
420     }
421     
422     static __inline__ void 
423     __constant_change_bit(const unsigned long nr, volatile void * addr) 
424     {
425       switch (nr&7) {
426       case 0:
427         __asm__ __volatile__ ("la 1,%0\n\t"
428                               "xi 0(1),0x01"
429                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
430                               : : "cc", "memory" );
431         break;
432       case 1:
433         __asm__ __volatile__ ("la 1,%0\n\t"
434                               "xi 0(1),0x02"
435                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
436                               : : "cc", "memory" );
437         break;
438       case 2:
439         __asm__ __volatile__ ("la 1,%0\n\t"
440                               "xi 0(1),0x04"
441                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
442                               : : "cc", "memory" );
443         break;
444       case 3:
445         __asm__ __volatile__ ("la 1,%0\n\t"
446                               "xi 0(1),0x08"
447                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
448                               : : "cc", "memory" );
449         break;
450       case 4:
451         __asm__ __volatile__ ("la 1,%0\n\t"
452                               "xi 0(1),0x10"
453                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
454                               : : "cc", "memory" );
455         break;
456       case 5:
457         __asm__ __volatile__ ("la 1,%0\n\t"
458                               "xi 0(1),0x20"
459                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
460                               : : "1", "cc", "memory" );
461         break;
462       case 6:
463         __asm__ __volatile__ ("la 1,%0\n\t"
464                               "xi 0(1),0x40"
465                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
466                               : : "1", "cc", "memory" );
467         break;
468       case 7:
469         __asm__ __volatile__ ("la 1,%0\n\t"
470                               "xi 0(1),0x80"
471                               : "=m" (*((volatile char *) addr + ((nr>>3)^7)))
472                               : : "1", "cc", "memory" );
473         break;
474       }
475     }
476     
477     #define change_bit_simple(nr,addr) \
478     (__builtin_constant_p((nr)) ? \
479      __constant_change_bit((nr),(addr)) : \
480      __change_bit((nr),(addr)) )
481     
482     /*
483      * fast, non-SMP test_and_set_bit routine
484      */
485     static __inline__ int
486     test_and_set_bit_simple(unsigned long nr, volatile void * addr)
487     {
488             int oldbit;
489             __asm__ __volatile__(
490                  "   lghi  1,56\n"
491                  "   lghi  2,7\n"
492                  "   xgr   1,%1\n"
493                  "   nr    2,%1\n"
494                  "   srlg  1,1,3\n"
495                  "   la    1,0(1,%2)\n"
496                  "   ic    %0,0(1)\n"
497                  "   srl   %0,0(2)\n"
498                  "   la    2,0(2,%3)\n"
499                  "   oc    0(1,1),0(2)"
500                  : "=&d" (oldbit) : "d" (nr), "a" (addr), "a" (&_oi_bitmap)
501                  : "cc", "memory", "1", "2" );
502             return oldbit & 1;
503     }
504     #define __test_and_set_bit(X,Y)		test_and_set_bit_simple(X,Y)
505     
506     /*
507      * fast, non-SMP test_and_clear_bit routine
508      */
509     static __inline__ int
510     test_and_clear_bit_simple(unsigned long nr, volatile void * addr)
511     {
512             int oldbit;
513     
514             __asm__ __volatile__(
515                  "   lghi  1,56\n"
516                  "   lghi  2,7\n"
517                  "   xgr   1,%1\n"
518                  "   nr    2,%1\n"
519                  "   srlg  1,1,3\n"
520                  "   la    1,0(1,%2)\n"
521                  "   ic    %0,0(1)\n"
522                  "   srl   %0,0(2)\n"
523                  "   la    2,0(2,%3)\n"
524                  "   nc    0(1,1),0(2)"
525                  : "=&d" (oldbit) : "d" (nr), "a" (addr), "a" (&_ni_bitmap)
526                  : "cc", "memory", "1", "2" );
527             return oldbit & 1;
528     }
529     #define __test_and_clear_bit(X,Y)	test_and_clear_bit_simple(X,Y)
530     
531     /*
532      * fast, non-SMP test_and_change_bit routine
533      */
534     static __inline__ int
535     test_and_change_bit_simple(unsigned long nr, volatile void * addr)
536     {
537             int oldbit;
538     
539             __asm__ __volatile__(
540                  "   lghi  1,56\n"
541                  "   lghi  2,7\n"
542                  "   xgr   1,%1\n"
543                  "   nr    2,%1\n"
544                  "   srlg  1,1,3\n"
545                  "   la    1,0(1,%2)\n"
546                  "   ic    %0,0(1)\n"
547                  "   srl   %0,0(2)\n"
548                  "   la    2,0(2,%3)\n"
549                  "   xc    0(1,1),0(2)"
550                  : "=&d" (oldbit) : "d" (nr), "a" (addr), "a" (&_oi_bitmap)
551                  : "cc", "memory", "1", "2" );
552             return oldbit & 1;
553     }
554     #define __test_and_change_bit(X,Y)	test_and_change_bit_simple(X,Y)
555     
556     #ifdef CONFIG_SMP
557     #define set_bit             set_bit_cs
558     #define clear_bit           clear_bit_cs
559     #define change_bit          change_bit_cs
560     #define test_and_set_bit    test_and_set_bit_cs
561     #define test_and_clear_bit  test_and_clear_bit_cs
562     #define test_and_change_bit test_and_change_bit_cs
563     #else
564     #define set_bit             set_bit_simple
565     #define clear_bit           clear_bit_simple
566     #define change_bit          change_bit_simple
567     #define test_and_set_bit    test_and_set_bit_simple
568     #define test_and_clear_bit  test_and_clear_bit_simple
569     #define test_and_change_bit test_and_change_bit_simple
570     #endif
571     
572     
573     /*
574      * This routine doesn't need to be atomic.
575      */
576     
577     static __inline__ int __test_bit(unsigned long nr, volatile void * addr)
578     {
579             int oldbit;
580     
581             __asm__ __volatile__(
582                  "   lghi  2,56\n"
583                  "   lghi  1,7\n"
584                  "   xgr   2,%1\n"
585                  "   nr    1,%1\n"
586                  "   srlg  2,2,3\n"
587                  "   ic    %0,0(2,%2)\n"
588                  "   srl   %0,0(1)\n"
589                  : "=&d" (oldbit) : "d" (nr), "a" (addr)
590                  : "cc", "1", "2" );
591             return oldbit & 1;
592     }
593     
594     static __inline__ int 
595     __constant_test_bit(unsigned long nr, volatile void * addr) {
596         return (((volatile char *) addr)[(nr>>3)^7] & (1<<(nr&7))) != 0;
597     }
598     
599     #define test_bit(nr,addr) \
600     (__builtin_constant_p((nr)) ? \
601      __constant_test_bit((nr),(addr)) : \
602      __test_bit((nr),(addr)) )
603     
604     /*
605      * Find-bit routines..
606      */
607     static __inline__ unsigned long
608     find_first_zero_bit(void * addr, unsigned long size)
609     {
610             unsigned long res;
611     
612             if (!size)
613                     return 0;
614             __asm__("   lghi  0,-1\n"
615                     "   lgr   1,%1\n"
616                     "   slgr  %0,%0\n"
617                     "   aghi  1,63\n"
618                     "   srlg  1,1,6\n"
619                     "0: cg    0,0(%0,%2)\n"
620                     "   jne   1f\n"
621                     "   aghi  %0,8\n"
622                     "   brct  1,0b\n"
623                     "   lgr   %0,%1\n"
624                     "   j     5f\n"
625                     "1: lg    1,0(%0,%2)\n"
626                     "   sllg  %0,%0,3\n"
627                     "   clr   1,0\n"
628     		"   jne   2f\n"
629     		"   aghi  %0,32\n"
630                     "   srlg  1,1,32\n"
631     		"2: lghi  0,0xff\n"
632                     "   tmll  1,0xffff\n"
633                     "   jno   3f\n"
634                     "   aghi  %0,16\n"
635                     "   srl   1,16\n"
636                     "3: tmll  1,0x00ff\n"
637                     "   jno   4f\n"
638                     "   aghi  %0,8\n"
639                     "   srl   1,8\n"
640                     "4: ngr   1,0\n"
641                     "   ic    1,0(1,%3)\n"
642                     "   algr  %0,1\n"
643                     "5:"
644                     : "=&a" (res) : "a" (size), "a" (addr), "a" (&_zb_findmap)
645                     : "cc", "0", "1" );
646             return (res < size) ? res : size;
647     }
648     
649     static __inline__ unsigned long
650     find_next_zero_bit (void * addr, unsigned long size, unsigned long offset)
651     {
652             unsigned long * p = ((unsigned long *) addr) + (offset >> 6);
653             unsigned long bitvec;
654             unsigned long set, bit = offset & 63, res;
655     
656             if (bit) {
657                     /*
658                      * Look for zero in first word
659                      */
660                     bitvec = (*p) >> bit;
661                     __asm__("   lhi  0,-1\n"
662                             "   lgr  1,%1\n"
663                             "   slgr %0,%0\n"
664                             "   clr  1,0\n"
665                             "   jne  0f\n"
666                             "   aghi %0,32\n"
667                             "   srlg 1,1,32\n"
668     			"0: lghi 0,0xff\n"
669                             "   tmll 1,0xffff\n"
670                             "   jno  1f\n"
671                             "   aghi %0,16\n"
672                             "   srlg 1,1,16\n"
673                             "1: tmll 1,0x00ff\n"
674                             "   jno  2f\n"
675                             "   aghi %0,8\n"
676                             "   srlg 1,1,8\n"
677                             "2: ngr  1,0\n"
678                             "   ic   1,0(1,%2)\n"
679                             "   algr %0,1"
680                             : "=&d" (set) 
681                             : "d" (bitvec), "a" (&_zb_findmap)
682                             : "cc", "0", "1" );
683                     if (set < (64 - bit))
684                             return set + offset;
685                     offset += 64 - bit;
686                     p++;
687             }
688             /*
689              * No zero yet, search remaining full words for a zero
690              */
691             res = find_first_zero_bit (p, size - 64 * (p - (unsigned long *) addr));
692             return (offset + res);
693     }
694     
695     /*
696      * ffz = Find First Zero in word. Undefined if no zero exists,
697      * so code should check against ~0UL first..
698      */
699     static __inline__ unsigned long ffz(unsigned long word)
700     {
701             int result;
702     
703             __asm__("   lhi  0,-1\n"
704                     "   lgr  1,%1\n"
705                     "   slgr %0,%0\n"
706                     "   clr  1,0\n"
707                     "   jne  0f\n"
708                     "   aghi %0,32\n"
709                     "   srlg 1,1,32\n"
710                     "0: lghi 0,0xff\n"
711                     "   tmll 1,0xffff\n"
712                     "   jno  1f\n"
713                     "   aghi %0,16\n"
714                     "   srlg 1,1,16\n"
715                     "1: tmll 1,0x00ff\n"
716                     "   jno  2f\n"
717                     "   aghi %0,8\n"
718                     "   srlg 1,1,8\n"
719                     "2: ngr  1,0\n"
720                     "   ic   1,0(1,%2)\n"
721                     "   algr %0,1"
722                     : "=&d" (result)
723                     : "d" (word), "a" (&_zb_findmap)
724                     : "cc", "0", "1" );
725     
726             return result;
727     }
728     
729     /*
730      * ffs: find first bit set. This is defined the same way as
731      * the libc and compiler builtin ffs routines, therefore
732      * differs in spirit from the above ffz (man ffs).
733      */
734     
735     extern int __inline__ ffs (int x)
736     {
737             int r;
738     
739             if (x == 0)
740               return 0;
741             __asm__("    lr   1,%1\n"
742                     "    slr  %0,%0\n"
743                     "    tml  1,0xFFFF\n"
744                     "    jnz  0f\n"
745                     "    ahi  %0,16\n"
746                     "    srl  1,16\n"
747                     "0:  tml  1,0x00FF\n"
748                     "    jnz  1f\n"
749                     "    ahi  %0,8\n"
750                     "    srl  1,8\n"
751                     "1:  tml  1,0x000F\n"
752                     "    jnz  2f\n"
753                     "    ahi  %0,4\n"
754                     "    srl  1,4\n"
755                     "2:  tml  1,0x0003\n"
756                     "    jnz  3f\n"
757                     "    ahi  %0,2\n"
758                     "    srl  1,2\n"
759                     "3:  tml  1,0x0001\n"
760                     "    jnz  4f\n"
761                     "    ahi  %0,1\n"
762                     "4:"
763                     : "=&d" (r) : "d" (x) : "cc", "1" );
764             return r+1;
765     }
766     
767     /*
768      * hweightN: returns the hamming weight (i.e. the number
769      * of bits set) of a N-bit word
770      */
771     
772     #define hweight32(x) generic_hweight32(x)
773     #define hweight16(x) generic_hweight16(x)
774     #define hweight8(x) generic_hweight8(x)
775     
776     
777     #ifdef __KERNEL__
778     
779     /*
780      * ATTENTION: intel byte ordering convention for ext2 and minix !!
781      * bit 0 is the LSB of addr; bit 31 is the MSB of addr;
782      * bit 32 is the LSB of (addr+4).
783      * That combined with the little endian byte order of Intel gives the
784      * following bit order in memory:
785      *    07 06 05 04 03 02 01 00 15 14 13 12 11 10 09 08 \
786      *    23 22 21 20 19 18 17 16 31 30 29 28 27 26 25 24
787      */
788     
789     #define ext2_set_bit(nr, addr)       test_and_set_bit((nr)^56, addr)
790     #define ext2_clear_bit(nr, addr)     test_and_clear_bit((nr)^56, addr)
791     #define ext2_test_bit(nr, addr)      test_bit((nr)^56, addr)
792     static __inline__ unsigned long
793     ext2_find_first_zero_bit(void *vaddr, unsigned long size)
794     {
795             unsigned long res;
796     
797             if (!size)
798                     return 0;
799             __asm__("   lghi  0,-1\n"
800                     "   lgr   1,%1\n"
801                     "   aghi  1,63\n"
802                     "   srlg  1,1,6\n"
803                     "   slgr  %0,%0\n"
804                     "0: clg   0,0(%0,%2)\n"
805                     "   jne   1f\n"
806                     "   aghi  %0,8\n"
807                     "   brct  1,0b\n"
808                     "   lgr   %0,%1\n"
809                     "   j     5f\n"
810                     "1: cl    0,0(%0,%2)\n"
811     		"   jne   2f\n"
812     		"   aghi  %0,4\n"
813     		"2: l     1,0(%0,%2)\n"
814                     "   sllg  %0,%0,3\n"
815                     "   aghi  %0,24\n"
816                     "   lghi  0,0xff\n"
817                     "   tmlh  1,0xffff\n"
818                     "   jo    3f\n"
819                     "   aghi  %0,-16\n"
820                     "   srl   1,16\n"
821                     "3: tmll  1,0xff00\n"
822                     "   jo    4f\n"
823                     "   aghi  %0,-8\n"
824                     "   srl   1,8\n"
825                     "4: ngr   1,0\n"
826                     "   ic    1,0(1,%3)\n"
827                     "   algr  %0,1\n"
828                     "5:"
829                     : "=&a" (res) : "a" (size), "a" (vaddr), "a" (&_zb_findmap)
830                     : "cc", "0", "1" );
831             return (res < size) ? res : size;
832     }
833     
834     static __inline__ unsigned long
835     ext2_find_next_zero_bit(void *vaddr, unsigned long size, unsigned long offset)
836     {
837             unsigned long *addr = vaddr;
838             unsigned long *p = addr + (offset >> 6);
839             unsigned long word;
840             unsigned long bit = offset & 63UL, res;
841     
842             if (offset >= size)
843                     return size;
844     
845             if (bit) {
846                     __asm__("   lrvg %0,%1" /* load reversed, neat instruction */
847                             : "=a" (word) : "m" (*p) );
848                     word >>= bit;
849                     res = bit;
850                     /* Look for zero in first 8 byte word */
851                     __asm__("   lgr  1,%1\n"
852                             "   lghi 0,0xff\n"
853     			"   tmll 1,0xffff\n"
854     			"   jno  2f\n"
855     			"   ahi  %0,16\n"
856     			"   srlg 1,1,16\n"
857                     	"0: tmll 1,0xffff\n"
858                             "   jno  2f\n"
859                             "   ahi  %0,16\n"
860                             "   srlg 1,1,16\n"
861                             "1: tmll 1,0xffff\n"
862                             "   jno  2f\n"
863                             "   ahi  %0,16\n"
864                             "   srl  1,16\n"
865                             "2: tmll 1,0x00ff\n"
866                     	"   jno  3f\n"
867                     	"   ahi  %0,8\n"
868                     	"   srl  1,8\n"
869                     	"3: ngr  1,0\n"
870                     	"   ic   1,0(1,%2)\n"
871                     	"   alr  %0,1"
872                     	: "+&d" (res) 
873                       	: "d" (word), "a" (&_zb_findmap)
874                     	: "cc", "0", "1" );
875                     if (res < 64)
876     			return (p - addr)*64 + res;
877                     p++;
878             }
879             /* No zero yet, search remaining full bytes for a zero */
880             res = ext2_find_first_zero_bit (p, size - 64 * (p - addr));
881             return (p - addr) * 64 + res;
882     }
883     
884     /* Bitmap functions for the minix filesystem.  */
885     /* FIXME !!! */
886     #define minix_test_and_set_bit(nr,addr) test_and_set_bit(nr,addr)
887     #define minix_set_bit(nr,addr) set_bit(nr,addr)
888     #define minix_test_and_clear_bit(nr,addr) test_and_clear_bit(nr,addr)
889     #define minix_test_bit(nr,addr) test_bit(nr,addr)
890     #define minix_find_first_zero_bit(addr,size) find_first_zero_bit(addr,size)
891     
892     #endif /* __KERNEL__ */
893     
894     #endif /* _S390_BITOPS_H */
895