blob: 24312387fa244c453b6840441fe26eb6dcb974c6 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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__
34#include <asm/bug.h>
35#include <asm/processor.h>
36#include <linux/threads.h>
37
38struct vm_area_struct; /* forward declaration (include/linux/mm.h) */
Tim Schmielau8c65b4a2005-11-07 00:59:43 -080039struct mm_struct;
Linus Torvalds1da177e2005-04-16 15:20:36 -070040
41extern pgd_t swapper_pg_dir[] __attribute__ ((aligned (4096)));
42extern 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 */
54extern 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
92/*
93 * pgd entries used up by user/kernel:
94 */
95#ifndef __s390x__
96# define USER_PTRS_PER_PGD 512
97# define USER_PGD_PTRS 512
98# define KERNEL_PGD_PTRS 512
Linus Torvalds1da177e2005-04-16 15:20:36 -070099#else /* __s390x__ */
100# define USER_PTRS_PER_PGD 2048
101# define USER_PGD_PTRS 2048
102# define KERNEL_PGD_PTRS 2048
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103#endif /* __s390x__ */
104
Hugh Dickinsd455a362005-04-19 13:29:23 -0700105#define FIRST_USER_ADDRESS 0
106
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107#define pte_ERROR(e) \
108 printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e))
109#define pmd_ERROR(e) \
110 printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e))
111#define pgd_ERROR(e) \
112 printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
113
114#ifndef __ASSEMBLY__
115/*
116 * Just any arbitrary offset to the start of the vmalloc VM area: the
117 * current 8MB value just means that there will be a 8MB "hole" after the
118 * physical memory until the kernel virtual memory starts. That means that
119 * any out-of-bounds memory accesses will hopefully be caught.
120 * The vmalloc() routines leaves a hole of 4kB between each vmalloced
121 * area for the same reason. ;)
122 */
123#define VMALLOC_OFFSET (8*1024*1024)
124#define VMALLOC_START (((unsigned long) high_memory + VMALLOC_OFFSET) \
125 & ~(VMALLOC_OFFSET-1))
126#ifndef __s390x__
127# define VMALLOC_END (0x7fffffffL)
128#else /* __s390x__ */
129# define VMALLOC_END (0x40000000000L)
130#endif /* __s390x__ */
131
132
133/*
134 * A 31 bit pagetable entry of S390 has following format:
135 * | PFRA | | OS |
136 * 0 0IP0
137 * 00000000001111111111222222222233
138 * 01234567890123456789012345678901
139 *
140 * I Page-Invalid Bit: Page is not available for address-translation
141 * P Page-Protection Bit: Store access not possible for page
142 *
143 * A 31 bit segmenttable entry of S390 has following format:
144 * | P-table origin | |PTL
145 * 0 IC
146 * 00000000001111111111222222222233
147 * 01234567890123456789012345678901
148 *
149 * I Segment-Invalid Bit: Segment is not available for address-translation
150 * C Common-Segment Bit: Segment is not private (PoP 3-30)
151 * PTL Page-Table-Length: Page-table length (PTL+1*16 entries -> up to 256)
152 *
153 * The 31 bit segmenttable origin of S390 has following format:
154 *
155 * |S-table origin | | STL |
156 * X **GPS
157 * 00000000001111111111222222222233
158 * 01234567890123456789012345678901
159 *
160 * X Space-Switch event:
161 * G Segment-Invalid Bit: *
162 * P Private-Space Bit: Segment is not private (PoP 3-30)
163 * S Storage-Alteration:
164 * STL Segment-Table-Length: Segment-table length (STL+1*16 entries -> up to 2048)
165 *
166 * A 64 bit pagetable entry of S390 has following format:
167 * | PFRA |0IP0| OS |
168 * 0000000000111111111122222222223333333333444444444455555555556666
169 * 0123456789012345678901234567890123456789012345678901234567890123
170 *
171 * I Page-Invalid Bit: Page is not available for address-translation
172 * P Page-Protection Bit: Store access not possible for page
173 *
174 * A 64 bit segmenttable entry of S390 has following format:
175 * | P-table origin | TT
176 * 0000000000111111111122222222223333333333444444444455555555556666
177 * 0123456789012345678901234567890123456789012345678901234567890123
178 *
179 * I Segment-Invalid Bit: Segment is not available for address-translation
180 * C Common-Segment Bit: Segment is not private (PoP 3-30)
181 * P Page-Protection Bit: Store access not possible for page
182 * TT Type 00
183 *
184 * A 64 bit region table entry of S390 has following format:
185 * | S-table origin | TF TTTL
186 * 0000000000111111111122222222223333333333444444444455555555556666
187 * 0123456789012345678901234567890123456789012345678901234567890123
188 *
189 * I Segment-Invalid Bit: Segment is not available for address-translation
190 * TT Type 01
191 * TF
192 * TL Table lenght
193 *
194 * The 64 bit regiontable origin of S390 has following format:
195 * | region table origon | DTTL
196 * 0000000000111111111122222222223333333333444444444455555555556666
197 * 0123456789012345678901234567890123456789012345678901234567890123
198 *
199 * X Space-Switch event:
200 * G Segment-Invalid Bit:
201 * P Private-Space Bit:
202 * S Storage-Alteration:
203 * R Real space
204 * TL Table-Length:
205 *
206 * A storage key has the following format:
207 * | ACC |F|R|C|0|
208 * 0 3 4 5 6 7
209 * ACC: access key
210 * F : fetch protection bit
211 * R : referenced bit
212 * C : changed bit
213 */
214
215/* Hardware bits in the page table entry */
216#define _PAGE_RO 0x200 /* HW read-only */
217#define _PAGE_INVALID 0x400 /* HW invalid */
218
219/* Mask and four different kinds of invalid pages. */
220#define _PAGE_INVALID_MASK 0x601
221#define _PAGE_INVALID_EMPTY 0x400
222#define _PAGE_INVALID_NONE 0x401
223#define _PAGE_INVALID_SWAP 0x600
224#define _PAGE_INVALID_FILE 0x601
225
226#ifndef __s390x__
227
228/* Bits in the segment table entry */
229#define _PAGE_TABLE_LEN 0xf /* only full page-tables */
230#define _PAGE_TABLE_COM 0x10 /* common page-table */
231#define _PAGE_TABLE_INV 0x20 /* invalid page-table */
232#define _SEG_PRESENT 0x001 /* Software (overlap with PTL) */
233
234/* Bits int the storage key */
235#define _PAGE_CHANGED 0x02 /* HW changed bit */
236#define _PAGE_REFERENCED 0x04 /* HW referenced bit */
237
238#define _USER_SEG_TABLE_LEN 0x7f /* user-segment-table up to 2 GB */
239#define _KERNEL_SEG_TABLE_LEN 0x7f /* kernel-segment-table up to 2 GB */
240
241/*
242 * User and Kernel pagetables are identical
243 */
244#define _PAGE_TABLE _PAGE_TABLE_LEN
245#define _KERNPG_TABLE _PAGE_TABLE_LEN
246
247/*
248 * The Kernel segment-tables includes the User segment-table
249 */
250
251#define _SEGMENT_TABLE (_USER_SEG_TABLE_LEN|0x80000000|0x100)
252#define _KERNSEG_TABLE _KERNEL_SEG_TABLE_LEN
253
254#define USER_STD_MASK 0x00000080UL
255
256#else /* __s390x__ */
257
258/* Bits in the segment table entry */
259#define _PMD_ENTRY_INV 0x20 /* invalid segment table entry */
260#define _PMD_ENTRY 0x00
261
262/* Bits in the region third table entry */
263#define _PGD_ENTRY_INV 0x20 /* invalid region table entry */
264#define _PGD_ENTRY 0x07
265
266/*
267 * User and kernel page directory
268 */
269#define _REGION_THIRD 0x4
270#define _REGION_THIRD_LEN 0x3
271#define _REGION_TABLE (_REGION_THIRD|_REGION_THIRD_LEN|0x40|0x100)
272#define _KERN_REGION_TABLE (_REGION_THIRD|_REGION_THIRD_LEN)
273
274#define USER_STD_MASK 0x0000000000000080UL
275
276/* Bits in the storage key */
277#define _PAGE_CHANGED 0x02 /* HW changed bit */
278#define _PAGE_REFERENCED 0x04 /* HW referenced bit */
279
280#endif /* __s390x__ */
281
282/*
283 * No mapping available
284 */
285#define PAGE_NONE_SHARED __pgprot(_PAGE_INVALID_NONE)
286#define PAGE_NONE_PRIVATE __pgprot(_PAGE_INVALID_NONE)
287#define PAGE_RO_SHARED __pgprot(_PAGE_RO)
288#define PAGE_RO_PRIVATE __pgprot(_PAGE_RO)
289#define PAGE_COPY __pgprot(_PAGE_RO)
290#define PAGE_SHARED __pgprot(0)
291#define PAGE_KERNEL __pgprot(0)
292
293/*
294 * The S390 can't do page protection for execute, and considers that the
295 * same are read. Also, write permissions imply read permissions. This is
296 * the closest we can get..
297 */
298 /*xwr*/
299#define __P000 PAGE_NONE_PRIVATE
300#define __P001 PAGE_RO_PRIVATE
301#define __P010 PAGE_COPY
302#define __P011 PAGE_COPY
303#define __P100 PAGE_RO_PRIVATE
304#define __P101 PAGE_RO_PRIVATE
305#define __P110 PAGE_COPY
306#define __P111 PAGE_COPY
307
308#define __S000 PAGE_NONE_SHARED
309#define __S001 PAGE_RO_SHARED
310#define __S010 PAGE_SHARED
311#define __S011 PAGE_SHARED
312#define __S100 PAGE_RO_SHARED
313#define __S101 PAGE_RO_SHARED
314#define __S110 PAGE_SHARED
315#define __S111 PAGE_SHARED
316
317/*
318 * Certain architectures need to do special things when PTEs
319 * within a page table are directly modified. Thus, the following
320 * hook is made available.
321 */
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800322static inline void set_pte(pte_t *pteptr, pte_t pteval)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323{
324 *pteptr = pteval;
325}
326#define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval)
327
328/*
329 * pgd/pmd/pte query functions
330 */
331#ifndef __s390x__
332
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800333static inline int pgd_present(pgd_t pgd) { return 1; }
334static inline int pgd_none(pgd_t pgd) { return 0; }
335static inline int pgd_bad(pgd_t pgd) { return 0; }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800337static inline int pmd_present(pmd_t pmd) { return pmd_val(pmd) & _SEG_PRESENT; }
338static inline int pmd_none(pmd_t pmd) { return pmd_val(pmd) & _PAGE_TABLE_INV; }
339static inline int pmd_bad(pmd_t pmd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340{
341 return (pmd_val(pmd) & (~PAGE_MASK & ~_PAGE_TABLE_INV)) != _PAGE_TABLE;
342}
343
344#else /* __s390x__ */
345
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800346static inline int pgd_present(pgd_t pgd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347{
348 return (pgd_val(pgd) & ~PAGE_MASK) == _PGD_ENTRY;
349}
350
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800351static inline int pgd_none(pgd_t pgd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352{
353 return pgd_val(pgd) & _PGD_ENTRY_INV;
354}
355
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800356static inline int pgd_bad(pgd_t pgd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357{
358 return (pgd_val(pgd) & (~PAGE_MASK & ~_PGD_ENTRY_INV)) != _PGD_ENTRY;
359}
360
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800361static inline int pmd_present(pmd_t pmd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362{
363 return (pmd_val(pmd) & ~PAGE_MASK) == _PMD_ENTRY;
364}
365
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800366static inline int pmd_none(pmd_t pmd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367{
368 return pmd_val(pmd) & _PMD_ENTRY_INV;
369}
370
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800371static inline int pmd_bad(pmd_t pmd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372{
373 return (pmd_val(pmd) & (~PAGE_MASK & ~_PMD_ENTRY_INV)) != _PMD_ENTRY;
374}
375
376#endif /* __s390x__ */
377
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800378static inline int pte_none(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379{
380 return (pte_val(pte) & _PAGE_INVALID_MASK) == _PAGE_INVALID_EMPTY;
381}
382
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800383static inline int pte_present(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384{
385 return !(pte_val(pte) & _PAGE_INVALID) ||
386 (pte_val(pte) & _PAGE_INVALID_MASK) == _PAGE_INVALID_NONE;
387}
388
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800389static inline int pte_file(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390{
391 return (pte_val(pte) & _PAGE_INVALID_MASK) == _PAGE_INVALID_FILE;
392}
393
394#define pte_same(a,b) (pte_val(a) == pte_val(b))
395
396/*
397 * query functions pte_write/pte_dirty/pte_young only work if
398 * pte_present() is true. Undefined behaviour if not..
399 */
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800400static inline int pte_write(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401{
402 return (pte_val(pte) & _PAGE_RO) == 0;
403}
404
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800405static inline int pte_dirty(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406{
407 /* A pte is neither clean nor dirty on s/390. The dirty bit
408 * is in the storage key. See page_test_and_clear_dirty for
409 * details.
410 */
411 return 0;
412}
413
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800414static inline int pte_young(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415{
416 /* A pte is neither young nor old on s/390. The young bit
417 * is in the storage key. See page_test_and_clear_young for
418 * details.
419 */
420 return 0;
421}
422
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800423static inline int pte_read(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424{
425 /* All pages are readable since we don't use the fetch
426 * protection bit in the storage key.
427 */
428 return 1;
429}
430
431/*
432 * pgd/pmd/pte modification functions
433 */
434
435#ifndef __s390x__
436
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800437static inline void pgd_clear(pgd_t * pgdp) { }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800439static inline void pmd_clear(pmd_t * pmdp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440{
441 pmd_val(pmdp[0]) = _PAGE_TABLE_INV;
442 pmd_val(pmdp[1]) = _PAGE_TABLE_INV;
443 pmd_val(pmdp[2]) = _PAGE_TABLE_INV;
444 pmd_val(pmdp[3]) = _PAGE_TABLE_INV;
445}
446
447#else /* __s390x__ */
448
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800449static inline void pgd_clear(pgd_t * pgdp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450{
451 pgd_val(*pgdp) = _PGD_ENTRY_INV | _PGD_ENTRY;
452}
453
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800454static inline void pmd_clear(pmd_t * pmdp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455{
456 pmd_val(*pmdp) = _PMD_ENTRY_INV | _PMD_ENTRY;
457 pmd_val1(*pmdp) = _PMD_ENTRY_INV | _PMD_ENTRY;
458}
459
460#endif /* __s390x__ */
461
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800462static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463{
464 pte_val(*ptep) = _PAGE_INVALID_EMPTY;
465}
466
467/*
468 * The following pte modification functions only work if
469 * pte_present() is true. Undefined behaviour if not..
470 */
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800471static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472{
473 pte_val(pte) &= PAGE_MASK;
474 pte_val(pte) |= pgprot_val(newprot);
475 return pte;
476}
477
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800478static inline pte_t pte_wrprotect(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479{
480 /* Do not clobber _PAGE_INVALID_NONE pages! */
481 if (!(pte_val(pte) & _PAGE_INVALID))
482 pte_val(pte) |= _PAGE_RO;
483 return pte;
484}
485
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800486static inline pte_t pte_mkwrite(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487{
488 pte_val(pte) &= ~_PAGE_RO;
489 return pte;
490}
491
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800492static inline pte_t pte_mkclean(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493{
494 /* The only user of pte_mkclean is the fork() code.
495 We must *not* clear the *physical* page dirty bit
496 just because fork() wants to clear the dirty bit in
497 *one* of the page's mappings. So we just do nothing. */
498 return pte;
499}
500
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800501static inline pte_t pte_mkdirty(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502{
503 /* We do not explicitly set the dirty bit because the
504 * sske instruction is slow. It is faster to let the
505 * next instruction set the dirty bit.
506 */
507 return pte;
508}
509
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800510static inline pte_t pte_mkold(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511{
512 /* S/390 doesn't keep its dirty/referenced bit in the pte.
513 * There is no point in clearing the real referenced bit.
514 */
515 return pte;
516}
517
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800518static inline pte_t pte_mkyoung(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519{
520 /* S/390 doesn't keep its dirty/referenced bit in the pte.
521 * There is no point in setting the real referenced bit.
522 */
523 return pte;
524}
525
526static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
527{
528 return 0;
529}
530
531static inline int
532ptep_clear_flush_young(struct vm_area_struct *vma,
533 unsigned long address, pte_t *ptep)
534{
535 /* No need to flush TLB; bits are in storage key */
536 return ptep_test_and_clear_young(vma, address, ptep);
537}
538
539static inline int ptep_test_and_clear_dirty(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
540{
541 return 0;
542}
543
544static inline int
545ptep_clear_flush_dirty(struct vm_area_struct *vma,
546 unsigned long address, pte_t *ptep)
547{
548 /* No need to flush TLB; bits are in storage key */
549 return ptep_test_and_clear_dirty(vma, address, ptep);
550}
551
552static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
553{
554 pte_t pte = *ptep;
555 pte_clear(mm, addr, ptep);
556 return pte;
557}
558
559static inline pte_t
560ptep_clear_flush(struct vm_area_struct *vma,
561 unsigned long address, pte_t *ptep)
562{
563 pte_t pte = *ptep;
564#ifndef __s390x__
565 if (!(pte_val(pte) & _PAGE_INVALID)) {
566 /* S390 has 1mb segments, we are emulating 4MB segments */
567 pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00);
568 __asm__ __volatile__ ("ipte %2,%3"
569 : "=m" (*ptep) : "m" (*ptep),
570 "a" (pto), "a" (address) );
571 }
572#else /* __s390x__ */
573 if (!(pte_val(pte) & _PAGE_INVALID))
574 __asm__ __volatile__ ("ipte %2,%3"
575 : "=m" (*ptep) : "m" (*ptep),
576 "a" (ptep), "a" (address) );
577#endif /* __s390x__ */
578 pte_val(*ptep) = _PAGE_INVALID_EMPTY;
579 return pte;
580}
581
582static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
583{
584 pte_t old_pte = *ptep;
585 set_pte_at(mm, addr, ptep, pte_wrprotect(old_pte));
586}
587
588static inline void
589ptep_establish(struct vm_area_struct *vma,
590 unsigned long address, pte_t *ptep,
591 pte_t entry)
592{
593 ptep_clear_flush(vma, address, ptep);
594 set_pte(ptep, entry);
595}
596
597#define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
598 ptep_establish(__vma, __address, __ptep, __entry)
599
600/*
601 * Test and clear dirty bit in storage key.
602 * We can't clear the changed bit atomically. This is a potential
603 * race against modification of the referenced bit. This function
604 * should therefore only be called if it is not mapped in any
605 * address space.
606 */
607#define page_test_and_clear_dirty(_page) \
608({ \
609 struct page *__page = (_page); \
610 unsigned long __physpage = __pa((__page-mem_map) << PAGE_SHIFT); \
611 int __skey = page_get_storage_key(__physpage); \
612 if (__skey & _PAGE_CHANGED) \
613 page_set_storage_key(__physpage, __skey & ~_PAGE_CHANGED);\
614 (__skey & _PAGE_CHANGED); \
615})
616
617/*
618 * Test and clear referenced bit in storage key.
619 */
620#define page_test_and_clear_young(page) \
621({ \
622 struct page *__page = (page); \
623 unsigned long __physpage = __pa((__page-mem_map) << PAGE_SHIFT); \
624 int __ccode; \
625 asm volatile ("rrbe 0,%1\n\t" \
626 "ipm %0\n\t" \
627 "srl %0,28\n\t" \
628 : "=d" (__ccode) : "a" (__physpage) : "cc" ); \
629 (__ccode & 2); \
630})
631
632/*
633 * Conversion functions: convert a page and protection to a page entry,
634 * and a page entry and page directory to the page they refer to.
635 */
636static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
637{
638 pte_t __pte;
639 pte_val(__pte) = physpage + pgprot_val(pgprot);
640 return __pte;
641}
642
643#define mk_pte(pg, pgprot) \
644({ \
645 struct page *__page = (pg); \
646 pgprot_t __pgprot = (pgprot); \
647 unsigned long __physpage = __pa((__page-mem_map) << PAGE_SHIFT); \
648 pte_t __pte = mk_pte_phys(__physpage, __pgprot); \
649 __pte; \
650})
651
652#define pfn_pte(pfn, pgprot) \
653({ \
654 pgprot_t __pgprot = (pgprot); \
655 unsigned long __physpage = __pa((pfn) << PAGE_SHIFT); \
656 pte_t __pte = mk_pte_phys(__physpage, __pgprot); \
657 __pte; \
658})
659
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660#ifdef __s390x__
661
662#define pfn_pmd(pfn, pgprot) \
663({ \
664 pgprot_t __pgprot = (pgprot); \
665 unsigned long __physpage = __pa((pfn) << PAGE_SHIFT); \
666 pmd_t __pmd = __pmd(__physpage + pgprot_val(__pgprot)); \
667 __pmd; \
668})
669
670#endif /* __s390x__ */
671
672#define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
673#define pte_page(x) pfn_to_page(pte_pfn(x))
674
675#define pmd_page_kernel(pmd) (pmd_val(pmd) & PAGE_MASK)
676
677#define pmd_page(pmd) (mem_map+(pmd_val(pmd) >> PAGE_SHIFT))
678
679#define pgd_page_kernel(pgd) (pgd_val(pgd) & PAGE_MASK)
680
681/* to find an entry in a page-table-directory */
682#define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
683#define pgd_offset(mm, address) ((mm)->pgd+pgd_index(address))
684
685/* to find an entry in a kernel page-table-directory */
686#define pgd_offset_k(address) pgd_offset(&init_mm, address)
687
688#ifndef __s390x__
689
690/* Find an entry in the second-level page table.. */
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800691static inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692{
693 return (pmd_t *) dir;
694}
695
696#else /* __s390x__ */
697
698/* Find an entry in the second-level page table.. */
699#define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
700#define pmd_offset(dir,addr) \
701 ((pmd_t *) pgd_page_kernel(*(dir)) + pmd_index(addr))
702
703#endif /* __s390x__ */
704
705/* Find an entry in the third-level page table.. */
706#define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
707#define pte_offset_kernel(pmd, address) \
708 ((pte_t *) pmd_page_kernel(*(pmd)) + pte_index(address))
709#define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
710#define pte_offset_map_nested(pmd, address) pte_offset_kernel(pmd, address)
711#define pte_unmap(pte) do { } while (0)
712#define pte_unmap_nested(pte) do { } while (0)
713
714/*
715 * 31 bit swap entry format:
716 * A page-table entry has some bits we have to treat in a special way.
717 * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
718 * exception will occur instead of a page translation exception. The
719 * specifiation exception has the bad habit not to store necessary
720 * information in the lowcore.
721 * Bit 21 and bit 22 are the page invalid bit and the page protection
722 * bit. We set both to indicate a swapped page.
723 * Bit 30 and 31 are used to distinguish the different page types. For
724 * a swapped page these bits need to be zero.
725 * This leaves the bits 1-19 and bits 24-29 to store type and offset.
726 * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
727 * plus 24 for the offset.
728 * 0| offset |0110|o|type |00|
729 * 0 0000000001111111111 2222 2 22222 33
730 * 0 1234567890123456789 0123 4 56789 01
731 *
732 * 64 bit swap entry format:
733 * A page-table entry has some bits we have to treat in a special way.
734 * Bits 52 and bit 55 have to be zero, otherwise an specification
735 * exception will occur instead of a page translation exception. The
736 * specifiation exception has the bad habit not to store necessary
737 * information in the lowcore.
738 * Bit 53 and bit 54 are the page invalid bit and the page protection
739 * bit. We set both to indicate a swapped page.
740 * Bit 62 and 63 are used to distinguish the different page types. For
741 * a swapped page these bits need to be zero.
742 * This leaves the bits 0-51 and bits 56-61 to store type and offset.
743 * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
744 * plus 56 for the offset.
745 * | offset |0110|o|type |00|
746 * 0000000000111111111122222222223333333333444444444455 5555 5 55566 66
747 * 0123456789012345678901234567890123456789012345678901 2345 6 78901 23
748 */
749#ifndef __s390x__
750#define __SWP_OFFSET_MASK (~0UL >> 12)
751#else
752#define __SWP_OFFSET_MASK (~0UL >> 11)
753#endif
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800754static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755{
756 pte_t pte;
757 offset &= __SWP_OFFSET_MASK;
758 pte_val(pte) = _PAGE_INVALID_SWAP | ((type & 0x1f) << 2) |
759 ((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
760 return pte;
761}
762
763#define __swp_type(entry) (((entry).val >> 2) & 0x1f)
764#define __swp_offset(entry) (((entry).val >> 11) | (((entry).val >> 7) & 1))
765#define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
766
767#define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
768#define __swp_entry_to_pte(x) ((pte_t) { (x).val })
769
770#ifndef __s390x__
771# define PTE_FILE_MAX_BITS 26
772#else /* __s390x__ */
773# define PTE_FILE_MAX_BITS 59
774#endif /* __s390x__ */
775
776#define pte_to_pgoff(__pte) \
777 ((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
778
779#define pgoff_to_pte(__off) \
780 ((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
781 | _PAGE_INVALID_FILE })
782
783#endif /* !__ASSEMBLY__ */
784
785#define kern_addr_valid(addr) (1)
786
787/*
788 * No page table caches to initialise
789 */
790#define pgtable_cache_init() do { } while (0)
791
792#define __HAVE_ARCH_PTEP_ESTABLISH
793#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
794#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
795#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
796#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
797#define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH
798#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
799#define __HAVE_ARCH_PTEP_CLEAR_FLUSH
800#define __HAVE_ARCH_PTEP_SET_WRPROTECT
801#define __HAVE_ARCH_PTE_SAME
802#define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_DIRTY
803#define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_YOUNG
804#include <asm-generic/pgtable.h>
805
806#endif /* _S390_PAGE_H */
807