Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * include/asm-s390/pgtable.h | |
3 | * | |
4 | * S390 version | |
5 | * Copyright (C) 1999,2000 IBM Deutschland Entwicklung GmbH, IBM Corporation | |
6 | * Author(s): Hartmut Penner (hp@de.ibm.com) | |
7 | * Ulrich Weigand (weigand@de.ibm.com) | |
8 | * Martin Schwidefsky (schwidefsky@de.ibm.com) | |
9 | * | |
10 | * Derived from "include/asm-i386/pgtable.h" | |
11 | */ | |
12 | ||
13 | #ifndef _ASM_S390_PGTABLE_H | |
14 | #define _ASM_S390_PGTABLE_H | |
15 | ||
16 | #include <asm-generic/4level-fixup.h> | |
17 | ||
18 | /* | |
19 | * The Linux memory management assumes a three-level page table setup. For | |
20 | * s390 31 bit we "fold" the mid level into the top-level page table, so | |
21 | * that we physically have the same two-level page table as the s390 mmu | |
22 | * expects in 31 bit mode. For s390 64 bit we use three of the five levels | |
23 | * the hardware provides (region first and region second tables are not | |
24 | * used). | |
25 | * | |
26 | * The "pgd_xxx()" functions are trivial for a folded two-level | |
27 | * setup: the pgd is never bad, and a pmd always exists (as it's folded | |
28 | * into the pgd entry) | |
29 | * | |
30 | * This file contains the functions and defines necessary to modify and use | |
31 | * the S390 page table tree. | |
32 | */ | |
33 | #ifndef __ASSEMBLY__ | |
2dcea57a | 34 | #include <linux/mm_types.h> |
1da177e4 LT |
35 | #include <asm/bug.h> |
36 | #include <asm/processor.h> | |
1da177e4 LT |
37 | |
38 | struct vm_area_struct; /* forward declaration (include/linux/mm.h) */ | |
8c65b4a6 | 39 | struct mm_struct; |
1da177e4 LT |
40 | |
41 | extern pgd_t swapper_pg_dir[] __attribute__ ((aligned (4096))); | |
42 | extern void paging_init(void); | |
43 | ||
44 | /* | |
45 | * The S390 doesn't have any external MMU info: the kernel page | |
46 | * tables contain all the necessary information. | |
47 | */ | |
48 | #define update_mmu_cache(vma, address, pte) do { } while (0) | |
49 | ||
50 | /* | |
51 | * ZERO_PAGE is a global shared page that is always zero: used | |
52 | * for zero-mapped memory areas etc.. | |
53 | */ | |
54 | extern char empty_zero_page[PAGE_SIZE]; | |
55 | #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page)) | |
56 | #endif /* !__ASSEMBLY__ */ | |
57 | ||
58 | /* | |
59 | * PMD_SHIFT determines the size of the area a second-level page | |
60 | * table can map | |
61 | * PGDIR_SHIFT determines what a third-level page table entry can map | |
62 | */ | |
63 | #ifndef __s390x__ | |
64 | # define PMD_SHIFT 22 | |
65 | # define PGDIR_SHIFT 22 | |
66 | #else /* __s390x__ */ | |
67 | # define PMD_SHIFT 21 | |
68 | # define PGDIR_SHIFT 31 | |
69 | #endif /* __s390x__ */ | |
70 | ||
71 | #define PMD_SIZE (1UL << PMD_SHIFT) | |
72 | #define PMD_MASK (~(PMD_SIZE-1)) | |
73 | #define PGDIR_SIZE (1UL << PGDIR_SHIFT) | |
74 | #define PGDIR_MASK (~(PGDIR_SIZE-1)) | |
75 | ||
76 | /* | |
77 | * entries per page directory level: the S390 is two-level, so | |
78 | * we don't really have any PMD directory physically. | |
79 | * for S390 segment-table entries are combined to one PGD | |
80 | * that leads to 1024 pte per pgd | |
81 | */ | |
82 | #ifndef __s390x__ | |
83 | # define PTRS_PER_PTE 1024 | |
84 | # define PTRS_PER_PMD 1 | |
85 | # define PTRS_PER_PGD 512 | |
86 | #else /* __s390x__ */ | |
87 | # define PTRS_PER_PTE 512 | |
88 | # define PTRS_PER_PMD 1024 | |
89 | # define PTRS_PER_PGD 2048 | |
90 | #endif /* __s390x__ */ | |
91 | ||
d455a369 HD |
92 | #define FIRST_USER_ADDRESS 0 |
93 | ||
1da177e4 LT |
94 | #define pte_ERROR(e) \ |
95 | printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e)) | |
96 | #define pmd_ERROR(e) \ | |
97 | printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e)) | |
98 | #define pgd_ERROR(e) \ | |
99 | printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e)) | |
100 | ||
101 | #ifndef __ASSEMBLY__ | |
102 | /* | |
103 | * Just any arbitrary offset to the start of the vmalloc VM area: the | |
104 | * current 8MB value just means that there will be a 8MB "hole" after the | |
105 | * physical memory until the kernel virtual memory starts. That means that | |
106 | * any out-of-bounds memory accesses will hopefully be caught. | |
107 | * The vmalloc() routines leaves a hole of 4kB between each vmalloced | |
108 | * area for the same reason. ;) | |
109 | */ | |
110 | #define VMALLOC_OFFSET (8*1024*1024) | |
111 | #define VMALLOC_START (((unsigned long) high_memory + VMALLOC_OFFSET) \ | |
112 | & ~(VMALLOC_OFFSET-1)) | |
113 | #ifndef __s390x__ | |
114 | # define VMALLOC_END (0x7fffffffL) | |
115 | #else /* __s390x__ */ | |
116 | # define VMALLOC_END (0x40000000000L) | |
117 | #endif /* __s390x__ */ | |
118 | ||
119 | ||
120 | /* | |
121 | * A 31 bit pagetable entry of S390 has following format: | |
122 | * | PFRA | | OS | | |
123 | * 0 0IP0 | |
124 | * 00000000001111111111222222222233 | |
125 | * 01234567890123456789012345678901 | |
126 | * | |
127 | * I Page-Invalid Bit: Page is not available for address-translation | |
128 | * P Page-Protection Bit: Store access not possible for page | |
129 | * | |
130 | * A 31 bit segmenttable entry of S390 has following format: | |
131 | * | P-table origin | |PTL | |
132 | * 0 IC | |
133 | * 00000000001111111111222222222233 | |
134 | * 01234567890123456789012345678901 | |
135 | * | |
136 | * I Segment-Invalid Bit: Segment is not available for address-translation | |
137 | * C Common-Segment Bit: Segment is not private (PoP 3-30) | |
138 | * PTL Page-Table-Length: Page-table length (PTL+1*16 entries -> up to 256) | |
139 | * | |
140 | * The 31 bit segmenttable origin of S390 has following format: | |
141 | * | |
142 | * |S-table origin | | STL | | |
143 | * X **GPS | |
144 | * 00000000001111111111222222222233 | |
145 | * 01234567890123456789012345678901 | |
146 | * | |
147 | * X Space-Switch event: | |
148 | * G Segment-Invalid Bit: * | |
149 | * P Private-Space Bit: Segment is not private (PoP 3-30) | |
150 | * S Storage-Alteration: | |
151 | * STL Segment-Table-Length: Segment-table length (STL+1*16 entries -> up to 2048) | |
152 | * | |
153 | * A 64 bit pagetable entry of S390 has following format: | |
154 | * | PFRA |0IP0| OS | | |
155 | * 0000000000111111111122222222223333333333444444444455555555556666 | |
156 | * 0123456789012345678901234567890123456789012345678901234567890123 | |
157 | * | |
158 | * I Page-Invalid Bit: Page is not available for address-translation | |
159 | * P Page-Protection Bit: Store access not possible for page | |
160 | * | |
161 | * A 64 bit segmenttable entry of S390 has following format: | |
162 | * | P-table origin | TT | |
163 | * 0000000000111111111122222222223333333333444444444455555555556666 | |
164 | * 0123456789012345678901234567890123456789012345678901234567890123 | |
165 | * | |
166 | * I Segment-Invalid Bit: Segment is not available for address-translation | |
167 | * C Common-Segment Bit: Segment is not private (PoP 3-30) | |
168 | * P Page-Protection Bit: Store access not possible for page | |
169 | * TT Type 00 | |
170 | * | |
171 | * A 64 bit region table entry of S390 has following format: | |
172 | * | S-table origin | TF TTTL | |
173 | * 0000000000111111111122222222223333333333444444444455555555556666 | |
174 | * 0123456789012345678901234567890123456789012345678901234567890123 | |
175 | * | |
176 | * I Segment-Invalid Bit: Segment is not available for address-translation | |
177 | * TT Type 01 | |
178 | * TF | |
179 | * TL Table lenght | |
180 | * | |
181 | * The 64 bit regiontable origin of S390 has following format: | |
182 | * | region table origon | DTTL | |
183 | * 0000000000111111111122222222223333333333444444444455555555556666 | |
184 | * 0123456789012345678901234567890123456789012345678901234567890123 | |
185 | * | |
186 | * X Space-Switch event: | |
187 | * G Segment-Invalid Bit: | |
188 | * P Private-Space Bit: | |
189 | * S Storage-Alteration: | |
190 | * R Real space | |
191 | * TL Table-Length: | |
192 | * | |
193 | * A storage key has the following format: | |
194 | * | ACC |F|R|C|0| | |
195 | * 0 3 4 5 6 7 | |
196 | * ACC: access key | |
197 | * F : fetch protection bit | |
198 | * R : referenced bit | |
199 | * C : changed bit | |
200 | */ | |
201 | ||
202 | /* Hardware bits in the page table entry */ | |
83377484 MS |
203 | #define _PAGE_RO 0x200 /* HW read-only bit */ |
204 | #define _PAGE_INVALID 0x400 /* HW invalid bit */ | |
205 | #define _PAGE_SWT 0x001 /* SW pte type bit t */ | |
206 | #define _PAGE_SWX 0x002 /* SW pte type bit x */ | |
1da177e4 | 207 | |
83377484 | 208 | /* Six different types of pages. */ |
9282ed92 GS |
209 | #define _PAGE_TYPE_EMPTY 0x400 |
210 | #define _PAGE_TYPE_NONE 0x401 | |
83377484 MS |
211 | #define _PAGE_TYPE_SWAP 0x403 |
212 | #define _PAGE_TYPE_FILE 0x601 /* bit 0x002 is used for offset !! */ | |
9282ed92 GS |
213 | #define _PAGE_TYPE_RO 0x200 |
214 | #define _PAGE_TYPE_RW 0x000 | |
1da177e4 | 215 | |
83377484 MS |
216 | /* |
217 | * PTE type bits are rather complicated. handle_pte_fault uses pte_present, | |
218 | * pte_none and pte_file to find out the pte type WITHOUT holding the page | |
219 | * table lock. ptep_clear_flush on the other hand uses ptep_clear_flush to | |
220 | * invalidate a given pte. ipte sets the hw invalid bit and clears all tlbs | |
221 | * for the page. The page table entry is set to _PAGE_TYPE_EMPTY afterwards. | |
222 | * This change is done while holding the lock, but the intermediate step | |
223 | * of a previously valid pte with the hw invalid bit set can be observed by | |
224 | * handle_pte_fault. That makes it necessary that all valid pte types with | |
225 | * the hw invalid bit set must be distinguishable from the four pte types | |
226 | * empty, none, swap and file. | |
227 | * | |
228 | * irxt ipte irxt | |
229 | * _PAGE_TYPE_EMPTY 1000 -> 1000 | |
230 | * _PAGE_TYPE_NONE 1001 -> 1001 | |
231 | * _PAGE_TYPE_SWAP 1011 -> 1011 | |
232 | * _PAGE_TYPE_FILE 11?1 -> 11?1 | |
233 | * _PAGE_TYPE_RO 0100 -> 1100 | |
234 | * _PAGE_TYPE_RW 0000 -> 1000 | |
235 | * | |
236 | * pte_none is true for bits combinations 1000, 1100 | |
237 | * pte_present is true for bits combinations 0000, 0010, 0100, 0110, 1001 | |
238 | * pte_file is true for bits combinations 1101, 1111 | |
239 | * swap pte is 1011 and 0001, 0011, 0101, 0111, 1010 and 1110 are invalid. | |
240 | */ | |
241 | ||
1da177e4 LT |
242 | #ifndef __s390x__ |
243 | ||
244 | /* Bits in the segment table entry */ | |
245 | #define _PAGE_TABLE_LEN 0xf /* only full page-tables */ | |
246 | #define _PAGE_TABLE_COM 0x10 /* common page-table */ | |
247 | #define _PAGE_TABLE_INV 0x20 /* invalid page-table */ | |
248 | #define _SEG_PRESENT 0x001 /* Software (overlap with PTL) */ | |
249 | ||
250 | /* Bits int the storage key */ | |
251 | #define _PAGE_CHANGED 0x02 /* HW changed bit */ | |
252 | #define _PAGE_REFERENCED 0x04 /* HW referenced bit */ | |
253 | ||
254 | #define _USER_SEG_TABLE_LEN 0x7f /* user-segment-table up to 2 GB */ | |
255 | #define _KERNEL_SEG_TABLE_LEN 0x7f /* kernel-segment-table up to 2 GB */ | |
256 | ||
257 | /* | |
258 | * User and Kernel pagetables are identical | |
259 | */ | |
260 | #define _PAGE_TABLE _PAGE_TABLE_LEN | |
261 | #define _KERNPG_TABLE _PAGE_TABLE_LEN | |
262 | ||
263 | /* | |
264 | * The Kernel segment-tables includes the User segment-table | |
265 | */ | |
266 | ||
267 | #define _SEGMENT_TABLE (_USER_SEG_TABLE_LEN|0x80000000|0x100) | |
268 | #define _KERNSEG_TABLE _KERNEL_SEG_TABLE_LEN | |
269 | ||
270 | #define USER_STD_MASK 0x00000080UL | |
271 | ||
272 | #else /* __s390x__ */ | |
273 | ||
274 | /* Bits in the segment table entry */ | |
275 | #define _PMD_ENTRY_INV 0x20 /* invalid segment table entry */ | |
276 | #define _PMD_ENTRY 0x00 | |
277 | ||
278 | /* Bits in the region third table entry */ | |
279 | #define _PGD_ENTRY_INV 0x20 /* invalid region table entry */ | |
280 | #define _PGD_ENTRY 0x07 | |
281 | ||
282 | /* | |
283 | * User and kernel page directory | |
284 | */ | |
285 | #define _REGION_THIRD 0x4 | |
286 | #define _REGION_THIRD_LEN 0x3 | |
287 | #define _REGION_TABLE (_REGION_THIRD|_REGION_THIRD_LEN|0x40|0x100) | |
288 | #define _KERN_REGION_TABLE (_REGION_THIRD|_REGION_THIRD_LEN) | |
289 | ||
290 | #define USER_STD_MASK 0x0000000000000080UL | |
291 | ||
292 | /* Bits in the storage key */ | |
293 | #define _PAGE_CHANGED 0x02 /* HW changed bit */ | |
294 | #define _PAGE_REFERENCED 0x04 /* HW referenced bit */ | |
295 | ||
296 | #endif /* __s390x__ */ | |
297 | ||
298 | /* | |
9282ed92 | 299 | * Page protection definitions. |
1da177e4 | 300 | */ |
9282ed92 GS |
301 | #define PAGE_NONE __pgprot(_PAGE_TYPE_NONE) |
302 | #define PAGE_RO __pgprot(_PAGE_TYPE_RO) | |
303 | #define PAGE_RW __pgprot(_PAGE_TYPE_RW) | |
304 | ||
305 | #define PAGE_KERNEL PAGE_RW | |
306 | #define PAGE_COPY PAGE_RO | |
1da177e4 LT |
307 | |
308 | /* | |
309 | * The S390 can't do page protection for execute, and considers that the | |
310 | * same are read. Also, write permissions imply read permissions. This is | |
311 | * the closest we can get.. | |
312 | */ | |
313 | /*xwr*/ | |
9282ed92 GS |
314 | #define __P000 PAGE_NONE |
315 | #define __P001 PAGE_RO | |
316 | #define __P010 PAGE_RO | |
317 | #define __P011 PAGE_RO | |
318 | #define __P100 PAGE_RO | |
319 | #define __P101 PAGE_RO | |
320 | #define __P110 PAGE_RO | |
321 | #define __P111 PAGE_RO | |
322 | ||
323 | #define __S000 PAGE_NONE | |
324 | #define __S001 PAGE_RO | |
325 | #define __S010 PAGE_RW | |
326 | #define __S011 PAGE_RW | |
327 | #define __S100 PAGE_RO | |
328 | #define __S101 PAGE_RO | |
329 | #define __S110 PAGE_RW | |
330 | #define __S111 PAGE_RW | |
1da177e4 LT |
331 | |
332 | /* | |
333 | * Certain architectures need to do special things when PTEs | |
334 | * within a page table are directly modified. Thus, the following | |
335 | * hook is made available. | |
336 | */ | |
4448aaf0 | 337 | static inline void set_pte(pte_t *pteptr, pte_t pteval) |
1da177e4 LT |
338 | { |
339 | *pteptr = pteval; | |
340 | } | |
341 | #define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval) | |
342 | ||
343 | /* | |
344 | * pgd/pmd/pte query functions | |
345 | */ | |
346 | #ifndef __s390x__ | |
347 | ||
4448aaf0 AB |
348 | static inline int pgd_present(pgd_t pgd) { return 1; } |
349 | static inline int pgd_none(pgd_t pgd) { return 0; } | |
350 | static inline int pgd_bad(pgd_t pgd) { return 0; } | |
1da177e4 | 351 | |
4448aaf0 AB |
352 | static inline int pmd_present(pmd_t pmd) { return pmd_val(pmd) & _SEG_PRESENT; } |
353 | static inline int pmd_none(pmd_t pmd) { return pmd_val(pmd) & _PAGE_TABLE_INV; } | |
354 | static inline int pmd_bad(pmd_t pmd) | |
1da177e4 LT |
355 | { |
356 | return (pmd_val(pmd) & (~PAGE_MASK & ~_PAGE_TABLE_INV)) != _PAGE_TABLE; | |
357 | } | |
358 | ||
359 | #else /* __s390x__ */ | |
360 | ||
4448aaf0 | 361 | static inline int pgd_present(pgd_t pgd) |
1da177e4 LT |
362 | { |
363 | return (pgd_val(pgd) & ~PAGE_MASK) == _PGD_ENTRY; | |
364 | } | |
365 | ||
4448aaf0 | 366 | static inline int pgd_none(pgd_t pgd) |
1da177e4 LT |
367 | { |
368 | return pgd_val(pgd) & _PGD_ENTRY_INV; | |
369 | } | |
370 | ||
4448aaf0 | 371 | static inline int pgd_bad(pgd_t pgd) |
1da177e4 LT |
372 | { |
373 | return (pgd_val(pgd) & (~PAGE_MASK & ~_PGD_ENTRY_INV)) != _PGD_ENTRY; | |
374 | } | |
375 | ||
4448aaf0 | 376 | static inline int pmd_present(pmd_t pmd) |
1da177e4 LT |
377 | { |
378 | return (pmd_val(pmd) & ~PAGE_MASK) == _PMD_ENTRY; | |
379 | } | |
380 | ||
4448aaf0 | 381 | static inline int pmd_none(pmd_t pmd) |
1da177e4 LT |
382 | { |
383 | return pmd_val(pmd) & _PMD_ENTRY_INV; | |
384 | } | |
385 | ||
4448aaf0 | 386 | static inline int pmd_bad(pmd_t pmd) |
1da177e4 LT |
387 | { |
388 | return (pmd_val(pmd) & (~PAGE_MASK & ~_PMD_ENTRY_INV)) != _PMD_ENTRY; | |
389 | } | |
390 | ||
391 | #endif /* __s390x__ */ | |
392 | ||
4448aaf0 | 393 | static inline int pte_none(pte_t pte) |
1da177e4 | 394 | { |
83377484 | 395 | return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT); |
1da177e4 LT |
396 | } |
397 | ||
4448aaf0 | 398 | static inline int pte_present(pte_t pte) |
1da177e4 | 399 | { |
83377484 MS |
400 | unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT | _PAGE_SWX; |
401 | return (pte_val(pte) & mask) == _PAGE_TYPE_NONE || | |
402 | (!(pte_val(pte) & _PAGE_INVALID) && | |
403 | !(pte_val(pte) & _PAGE_SWT)); | |
1da177e4 LT |
404 | } |
405 | ||
4448aaf0 | 406 | static inline int pte_file(pte_t pte) |
1da177e4 | 407 | { |
83377484 MS |
408 | unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT; |
409 | return (pte_val(pte) & mask) == _PAGE_TYPE_FILE; | |
1da177e4 LT |
410 | } |
411 | ||
412 | #define pte_same(a,b) (pte_val(a) == pte_val(b)) | |
413 | ||
414 | /* | |
415 | * query functions pte_write/pte_dirty/pte_young only work if | |
416 | * pte_present() is true. Undefined behaviour if not.. | |
417 | */ | |
4448aaf0 | 418 | static inline int pte_write(pte_t pte) |
1da177e4 LT |
419 | { |
420 | return (pte_val(pte) & _PAGE_RO) == 0; | |
421 | } | |
422 | ||
4448aaf0 | 423 | static inline int pte_dirty(pte_t pte) |
1da177e4 LT |
424 | { |
425 | /* A pte is neither clean nor dirty on s/390. The dirty bit | |
426 | * is in the storage key. See page_test_and_clear_dirty for | |
427 | * details. | |
428 | */ | |
429 | return 0; | |
430 | } | |
431 | ||
4448aaf0 | 432 | static inline int pte_young(pte_t pte) |
1da177e4 LT |
433 | { |
434 | /* A pte is neither young nor old on s/390. The young bit | |
435 | * is in the storage key. See page_test_and_clear_young for | |
436 | * details. | |
437 | */ | |
438 | return 0; | |
439 | } | |
440 | ||
4448aaf0 | 441 | static inline int pte_read(pte_t pte) |
1da177e4 LT |
442 | { |
443 | /* All pages are readable since we don't use the fetch | |
444 | * protection bit in the storage key. | |
445 | */ | |
446 | return 1; | |
447 | } | |
448 | ||
449 | /* | |
450 | * pgd/pmd/pte modification functions | |
451 | */ | |
452 | ||
453 | #ifndef __s390x__ | |
454 | ||
4448aaf0 | 455 | static inline void pgd_clear(pgd_t * pgdp) { } |
1da177e4 | 456 | |
4448aaf0 | 457 | static inline void pmd_clear(pmd_t * pmdp) |
1da177e4 LT |
458 | { |
459 | pmd_val(pmdp[0]) = _PAGE_TABLE_INV; | |
460 | pmd_val(pmdp[1]) = _PAGE_TABLE_INV; | |
461 | pmd_val(pmdp[2]) = _PAGE_TABLE_INV; | |
462 | pmd_val(pmdp[3]) = _PAGE_TABLE_INV; | |
463 | } | |
464 | ||
465 | #else /* __s390x__ */ | |
466 | ||
4448aaf0 | 467 | static inline void pgd_clear(pgd_t * pgdp) |
1da177e4 LT |
468 | { |
469 | pgd_val(*pgdp) = _PGD_ENTRY_INV | _PGD_ENTRY; | |
470 | } | |
471 | ||
4448aaf0 | 472 | static inline void pmd_clear(pmd_t * pmdp) |
1da177e4 LT |
473 | { |
474 | pmd_val(*pmdp) = _PMD_ENTRY_INV | _PMD_ENTRY; | |
475 | pmd_val1(*pmdp) = _PMD_ENTRY_INV | _PMD_ENTRY; | |
476 | } | |
477 | ||
478 | #endif /* __s390x__ */ | |
479 | ||
4448aaf0 | 480 | static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep) |
1da177e4 | 481 | { |
9282ed92 | 482 | pte_val(*ptep) = _PAGE_TYPE_EMPTY; |
1da177e4 LT |
483 | } |
484 | ||
485 | /* | |
486 | * The following pte modification functions only work if | |
487 | * pte_present() is true. Undefined behaviour if not.. | |
488 | */ | |
4448aaf0 | 489 | static inline pte_t pte_modify(pte_t pte, pgprot_t newprot) |
1da177e4 LT |
490 | { |
491 | pte_val(pte) &= PAGE_MASK; | |
492 | pte_val(pte) |= pgprot_val(newprot); | |
493 | return pte; | |
494 | } | |
495 | ||
4448aaf0 | 496 | static inline pte_t pte_wrprotect(pte_t pte) |
1da177e4 | 497 | { |
9282ed92 | 498 | /* Do not clobber _PAGE_TYPE_NONE pages! */ |
1da177e4 LT |
499 | if (!(pte_val(pte) & _PAGE_INVALID)) |
500 | pte_val(pte) |= _PAGE_RO; | |
501 | return pte; | |
502 | } | |
503 | ||
4448aaf0 | 504 | static inline pte_t pte_mkwrite(pte_t pte) |
1da177e4 LT |
505 | { |
506 | pte_val(pte) &= ~_PAGE_RO; | |
507 | return pte; | |
508 | } | |
509 | ||
4448aaf0 | 510 | static inline pte_t pte_mkclean(pte_t pte) |
1da177e4 LT |
511 | { |
512 | /* The only user of pte_mkclean is the fork() code. | |
513 | We must *not* clear the *physical* page dirty bit | |
514 | just because fork() wants to clear the dirty bit in | |
515 | *one* of the page's mappings. So we just do nothing. */ | |
516 | return pte; | |
517 | } | |
518 | ||
4448aaf0 | 519 | static inline pte_t pte_mkdirty(pte_t pte) |
1da177e4 LT |
520 | { |
521 | /* We do not explicitly set the dirty bit because the | |
522 | * sske instruction is slow. It is faster to let the | |
523 | * next instruction set the dirty bit. | |
524 | */ | |
525 | return pte; | |
526 | } | |
527 | ||
4448aaf0 | 528 | static inline pte_t pte_mkold(pte_t pte) |
1da177e4 LT |
529 | { |
530 | /* S/390 doesn't keep its dirty/referenced bit in the pte. | |
531 | * There is no point in clearing the real referenced bit. | |
532 | */ | |
533 | return pte; | |
534 | } | |
535 | ||
4448aaf0 | 536 | static inline pte_t pte_mkyoung(pte_t pte) |
1da177e4 LT |
537 | { |
538 | /* S/390 doesn't keep its dirty/referenced bit in the pte. | |
539 | * There is no point in setting the real referenced bit. | |
540 | */ | |
541 | return pte; | |
542 | } | |
543 | ||
544 | static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep) | |
545 | { | |
546 | return 0; | |
547 | } | |
548 | ||
549 | static inline int | |
550 | ptep_clear_flush_young(struct vm_area_struct *vma, | |
551 | unsigned long address, pte_t *ptep) | |
552 | { | |
553 | /* No need to flush TLB; bits are in storage key */ | |
554 | return ptep_test_and_clear_young(vma, address, ptep); | |
555 | } | |
556 | ||
557 | static inline int ptep_test_and_clear_dirty(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep) | |
558 | { | |
559 | return 0; | |
560 | } | |
561 | ||
562 | static inline int | |
563 | ptep_clear_flush_dirty(struct vm_area_struct *vma, | |
564 | unsigned long address, pte_t *ptep) | |
565 | { | |
566 | /* No need to flush TLB; bits are in storage key */ | |
567 | return ptep_test_and_clear_dirty(vma, address, ptep); | |
568 | } | |
569 | ||
570 | static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep) | |
571 | { | |
572 | pte_t pte = *ptep; | |
573 | pte_clear(mm, addr, ptep); | |
574 | return pte; | |
575 | } | |
576 | ||
9282ed92 | 577 | static inline void __ptep_ipte(unsigned long address, pte_t *ptep) |
1da177e4 | 578 | { |
9282ed92 | 579 | if (!(pte_val(*ptep) & _PAGE_INVALID)) { |
1da177e4 | 580 | #ifndef __s390x__ |
1da177e4 LT |
581 | /* S390 has 1mb segments, we are emulating 4MB segments */ |
582 | pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00); | |
9282ed92 GS |
583 | #else |
584 | /* ipte in zarch mode can do the math */ | |
585 | pte_t *pto = ptep; | |
586 | #endif | |
94c12cc7 MS |
587 | asm volatile( |
588 | " ipte %2,%3" | |
589 | : "=m" (*ptep) : "m" (*ptep), | |
590 | "a" (pto), "a" (address)); | |
1da177e4 | 591 | } |
9282ed92 GS |
592 | pte_val(*ptep) = _PAGE_TYPE_EMPTY; |
593 | } | |
594 | ||
595 | static inline pte_t | |
596 | ptep_clear_flush(struct vm_area_struct *vma, | |
597 | unsigned long address, pte_t *ptep) | |
598 | { | |
599 | pte_t pte = *ptep; | |
600 | ||
601 | __ptep_ipte(address, ptep); | |
1da177e4 LT |
602 | return pte; |
603 | } | |
604 | ||
605 | static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep) | |
606 | { | |
607 | pte_t old_pte = *ptep; | |
608 | set_pte_at(mm, addr, ptep, pte_wrprotect(old_pte)); | |
609 | } | |
610 | ||
611 | static inline void | |
612 | ptep_establish(struct vm_area_struct *vma, | |
613 | unsigned long address, pte_t *ptep, | |
614 | pte_t entry) | |
615 | { | |
616 | ptep_clear_flush(vma, address, ptep); | |
617 | set_pte(ptep, entry); | |
618 | } | |
619 | ||
620 | #define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \ | |
621 | ptep_establish(__vma, __address, __ptep, __entry) | |
622 | ||
623 | /* | |
624 | * Test and clear dirty bit in storage key. | |
625 | * We can't clear the changed bit atomically. This is a potential | |
626 | * race against modification of the referenced bit. This function | |
627 | * should therefore only be called if it is not mapped in any | |
628 | * address space. | |
629 | */ | |
2dcea57a HC |
630 | static inline int page_test_and_clear_dirty(struct page *page) |
631 | { | |
0b2b6e1d | 632 | unsigned long physpage = page_to_phys(page); |
2dcea57a HC |
633 | int skey = page_get_storage_key(physpage); |
634 | ||
635 | if (skey & _PAGE_CHANGED) | |
636 | page_set_storage_key(physpage, skey & ~_PAGE_CHANGED); | |
637 | return skey & _PAGE_CHANGED; | |
638 | } | |
1da177e4 LT |
639 | |
640 | /* | |
641 | * Test and clear referenced bit in storage key. | |
642 | */ | |
2dcea57a HC |
643 | static inline int page_test_and_clear_young(struct page *page) |
644 | { | |
0b2b6e1d | 645 | unsigned long physpage = page_to_phys(page); |
2dcea57a HC |
646 | int ccode; |
647 | ||
0b2b6e1d HC |
648 | asm volatile( |
649 | " rrbe 0,%1\n" | |
650 | " ipm %0\n" | |
651 | " srl %0,28\n" | |
2dcea57a HC |
652 | : "=d" (ccode) : "a" (physpage) : "cc" ); |
653 | return ccode & 2; | |
654 | } | |
1da177e4 LT |
655 | |
656 | /* | |
657 | * Conversion functions: convert a page and protection to a page entry, | |
658 | * and a page entry and page directory to the page they refer to. | |
659 | */ | |
660 | static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot) | |
661 | { | |
662 | pte_t __pte; | |
663 | pte_val(__pte) = physpage + pgprot_val(pgprot); | |
664 | return __pte; | |
665 | } | |
666 | ||
2dcea57a HC |
667 | static inline pte_t mk_pte(struct page *page, pgprot_t pgprot) |
668 | { | |
0b2b6e1d | 669 | unsigned long physpage = page_to_phys(page); |
1da177e4 | 670 | |
2dcea57a HC |
671 | return mk_pte_phys(physpage, pgprot); |
672 | } | |
673 | ||
674 | static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot) | |
675 | { | |
676 | unsigned long physpage = __pa((pfn) << PAGE_SHIFT); | |
677 | ||
678 | return mk_pte_phys(physpage, pgprot); | |
679 | } | |
1da177e4 | 680 | |
1da177e4 LT |
681 | #ifdef __s390x__ |
682 | ||
2dcea57a HC |
683 | static inline pmd_t pfn_pmd(unsigned long pfn, pgprot_t pgprot) |
684 | { | |
685 | unsigned long physpage = __pa((pfn) << PAGE_SHIFT); | |
686 | ||
687 | return __pmd(physpage + pgprot_val(pgprot)); | |
688 | } | |
1da177e4 LT |
689 | |
690 | #endif /* __s390x__ */ | |
691 | ||
692 | #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT) | |
693 | #define pte_page(x) pfn_to_page(pte_pfn(x)) | |
694 | ||
46a82b2d | 695 | #define pmd_page_vaddr(pmd) (pmd_val(pmd) & PAGE_MASK) |
1da177e4 | 696 | |
0b2b6e1d | 697 | #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT) |
1da177e4 | 698 | |
46a82b2d DM |
699 | #define pgd_page_vaddr(pgd) (pgd_val(pgd) & PAGE_MASK) |
700 | ||
0b2b6e1d | 701 | #define pgd_page(pgd) pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT) |
1da177e4 LT |
702 | |
703 | /* to find an entry in a page-table-directory */ | |
704 | #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1)) | |
705 | #define pgd_offset(mm, address) ((mm)->pgd+pgd_index(address)) | |
706 | ||
707 | /* to find an entry in a kernel page-table-directory */ | |
708 | #define pgd_offset_k(address) pgd_offset(&init_mm, address) | |
709 | ||
710 | #ifndef __s390x__ | |
711 | ||
712 | /* Find an entry in the second-level page table.. */ | |
4448aaf0 | 713 | static inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address) |
1da177e4 LT |
714 | { |
715 | return (pmd_t *) dir; | |
716 | } | |
717 | ||
718 | #else /* __s390x__ */ | |
719 | ||
720 | /* Find an entry in the second-level page table.. */ | |
721 | #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1)) | |
722 | #define pmd_offset(dir,addr) \ | |
46a82b2d | 723 | ((pmd_t *) pgd_page_vaddr(*(dir)) + pmd_index(addr)) |
1da177e4 LT |
724 | |
725 | #endif /* __s390x__ */ | |
726 | ||
727 | /* Find an entry in the third-level page table.. */ | |
728 | #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1)) | |
729 | #define pte_offset_kernel(pmd, address) \ | |
46a82b2d | 730 | ((pte_t *) pmd_page_vaddr(*(pmd)) + pte_index(address)) |
1da177e4 LT |
731 | #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address) |
732 | #define pte_offset_map_nested(pmd, address) pte_offset_kernel(pmd, address) | |
733 | #define pte_unmap(pte) do { } while (0) | |
734 | #define pte_unmap_nested(pte) do { } while (0) | |
735 | ||
736 | /* | |
737 | * 31 bit swap entry format: | |
738 | * A page-table entry has some bits we have to treat in a special way. | |
739 | * Bits 0, 20 and bit 23 have to be zero, otherwise an specification | |
740 | * exception will occur instead of a page translation exception. The | |
741 | * specifiation exception has the bad habit not to store necessary | |
742 | * information in the lowcore. | |
743 | * Bit 21 and bit 22 are the page invalid bit and the page protection | |
744 | * bit. We set both to indicate a swapped page. | |
745 | * Bit 30 and 31 are used to distinguish the different page types. For | |
746 | * a swapped page these bits need to be zero. | |
747 | * This leaves the bits 1-19 and bits 24-29 to store type and offset. | |
748 | * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19 | |
749 | * plus 24 for the offset. | |
750 | * 0| offset |0110|o|type |00| | |
751 | * 0 0000000001111111111 2222 2 22222 33 | |
752 | * 0 1234567890123456789 0123 4 56789 01 | |
753 | * | |
754 | * 64 bit swap entry format: | |
755 | * A page-table entry has some bits we have to treat in a special way. | |
756 | * Bits 52 and bit 55 have to be zero, otherwise an specification | |
757 | * exception will occur instead of a page translation exception. The | |
758 | * specifiation exception has the bad habit not to store necessary | |
759 | * information in the lowcore. | |
760 | * Bit 53 and bit 54 are the page invalid bit and the page protection | |
761 | * bit. We set both to indicate a swapped page. | |
762 | * Bit 62 and 63 are used to distinguish the different page types. For | |
763 | * a swapped page these bits need to be zero. | |
764 | * This leaves the bits 0-51 and bits 56-61 to store type and offset. | |
765 | * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51 | |
766 | * plus 56 for the offset. | |
767 | * | offset |0110|o|type |00| | |
768 | * 0000000000111111111122222222223333333333444444444455 5555 5 55566 66 | |
769 | * 0123456789012345678901234567890123456789012345678901 2345 6 78901 23 | |
770 | */ | |
771 | #ifndef __s390x__ | |
772 | #define __SWP_OFFSET_MASK (~0UL >> 12) | |
773 | #else | |
774 | #define __SWP_OFFSET_MASK (~0UL >> 11) | |
775 | #endif | |
4448aaf0 | 776 | static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset) |
1da177e4 LT |
777 | { |
778 | pte_t pte; | |
779 | offset &= __SWP_OFFSET_MASK; | |
9282ed92 | 780 | pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) | |
1da177e4 LT |
781 | ((offset & 1UL) << 7) | ((offset & ~1UL) << 11); |
782 | return pte; | |
783 | } | |
784 | ||
785 | #define __swp_type(entry) (((entry).val >> 2) & 0x1f) | |
786 | #define __swp_offset(entry) (((entry).val >> 11) | (((entry).val >> 7) & 1)) | |
787 | #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) }) | |
788 | ||
789 | #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) }) | |
790 | #define __swp_entry_to_pte(x) ((pte_t) { (x).val }) | |
791 | ||
792 | #ifndef __s390x__ | |
793 | # define PTE_FILE_MAX_BITS 26 | |
794 | #else /* __s390x__ */ | |
795 | # define PTE_FILE_MAX_BITS 59 | |
796 | #endif /* __s390x__ */ | |
797 | ||
798 | #define pte_to_pgoff(__pte) \ | |
799 | ((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f)) | |
800 | ||
801 | #define pgoff_to_pte(__off) \ | |
802 | ((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \ | |
9282ed92 | 803 | | _PAGE_TYPE_FILE }) |
1da177e4 LT |
804 | |
805 | #endif /* !__ASSEMBLY__ */ | |
806 | ||
807 | #define kern_addr_valid(addr) (1) | |
808 | ||
809 | /* | |
810 | * No page table caches to initialise | |
811 | */ | |
812 | #define pgtable_cache_init() do { } while (0) | |
813 | ||
814 | #define __HAVE_ARCH_PTEP_ESTABLISH | |
815 | #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS | |
816 | #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG | |
817 | #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH | |
818 | #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY | |
819 | #define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH | |
820 | #define __HAVE_ARCH_PTEP_GET_AND_CLEAR | |
821 | #define __HAVE_ARCH_PTEP_CLEAR_FLUSH | |
822 | #define __HAVE_ARCH_PTEP_SET_WRPROTECT | |
823 | #define __HAVE_ARCH_PTE_SAME | |
824 | #define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_DIRTY | |
825 | #define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_YOUNG | |
826 | #include <asm-generic/pgtable.h> | |
827 | ||
828 | #endif /* _S390_PAGE_H */ | |
829 |