blob: f2cc25b74adf7edfc424ef84251051e558827053 [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
Linus Torvalds1da177e2005-04-16 15:20:36 -070016/*
17 * The Linux memory management assumes a three-level page table setup. For
18 * s390 31 bit we "fold" the mid level into the top-level page table, so
19 * that we physically have the same two-level page table as the s390 mmu
20 * expects in 31 bit mode. For s390 64 bit we use three of the five levels
21 * the hardware provides (region first and region second tables are not
22 * used).
23 *
24 * The "pgd_xxx()" functions are trivial for a folded two-level
25 * setup: the pgd is never bad, and a pmd always exists (as it's folded
26 * into the pgd entry)
27 *
28 * This file contains the functions and defines necessary to modify and use
29 * the S390 page table tree.
30 */
31#ifndef __ASSEMBLY__
Heiko Carstens2dcea572006-09-29 01:58:41 -070032#include <linux/mm_types.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#include <asm/bug.h>
34#include <asm/processor.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070035
Linus Torvalds1da177e2005-04-16 15:20:36 -070036extern pgd_t swapper_pg_dir[] __attribute__ ((aligned (4096)));
37extern void paging_init(void);
Heiko Carstens2b67fc42007-02-05 21:16:47 +010038extern void vmem_map_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -070039
40/*
41 * The S390 doesn't have any external MMU info: the kernel page
42 * tables contain all the necessary information.
43 */
44#define update_mmu_cache(vma, address, pte) do { } while (0)
45
46/*
47 * ZERO_PAGE is a global shared page that is always zero: used
48 * for zero-mapped memory areas etc..
49 */
50extern char empty_zero_page[PAGE_SIZE];
51#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
52#endif /* !__ASSEMBLY__ */
53
54/*
55 * PMD_SHIFT determines the size of the area a second-level page
56 * table can map
57 * PGDIR_SHIFT determines what a third-level page table entry can map
58 */
59#ifndef __s390x__
60# define PMD_SHIFT 22
Martin Schwidefsky190a1d72007-10-22 12:52:48 +020061# define PUD_SHIFT 22
Linus Torvalds1da177e2005-04-16 15:20:36 -070062# define PGDIR_SHIFT 22
63#else /* __s390x__ */
64# define PMD_SHIFT 21
Martin Schwidefsky190a1d72007-10-22 12:52:48 +020065# define PUD_SHIFT 31
Linus Torvalds1da177e2005-04-16 15:20:36 -070066# define PGDIR_SHIFT 31
67#endif /* __s390x__ */
68
69#define PMD_SIZE (1UL << PMD_SHIFT)
70#define PMD_MASK (~(PMD_SIZE-1))
Martin Schwidefsky190a1d72007-10-22 12:52:48 +020071#define PUD_SIZE (1UL << PUD_SHIFT)
72#define PUD_MASK (~(PUD_SIZE-1))
Linus Torvalds1da177e2005-04-16 15:20:36 -070073#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
Martin Schwidefsky190a1d72007-10-22 12:52:48 +020085# define PTRS_PER_PUD 1
Linus Torvalds1da177e2005-04-16 15:20:36 -070086# define PTRS_PER_PGD 512
87#else /* __s390x__ */
88# define PTRS_PER_PTE 512
89# define PTRS_PER_PMD 1024
Martin Schwidefsky190a1d72007-10-22 12:52:48 +020090# define PTRS_PER_PUD 1
Linus Torvalds1da177e2005-04-16 15:20:36 -070091# define PTRS_PER_PGD 2048
92#endif /* __s390x__ */
93
Hugh Dickinsd455a362005-04-19 13:29:23 -070094#define FIRST_USER_ADDRESS 0
95
Linus Torvalds1da177e2005-04-16 15:20:36 -070096#define pte_ERROR(e) \
97 printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e))
98#define pmd_ERROR(e) \
99 printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e))
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200100#define pud_ERROR(e) \
101 printk("%s:%d: bad pud %p.\n", __FILE__, __LINE__, (void *) pud_val(e))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102#define pgd_ERROR(e) \
103 printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
104
105#ifndef __ASSEMBLY__
106/*
107 * Just any arbitrary offset to the start of the vmalloc VM area: the
108 * current 8MB value just means that there will be a 8MB "hole" after the
109 * physical memory until the kernel virtual memory starts. That means that
110 * any out-of-bounds memory accesses will hopefully be caught.
111 * The vmalloc() routines leaves a hole of 4kB between each vmalloced
112 * area for the same reason. ;)
Heiko Carstense39394b2007-10-12 16:11:45 +0200113 * vmalloc area starts at 4GB to prevent syscall table entry exchanging
114 * from modules.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115 */
Heiko Carstensf4eb07c2006-12-08 15:56:07 +0100116extern unsigned long vmalloc_end;
Heiko Carstense39394b2007-10-12 16:11:45 +0200117
118#ifdef CONFIG_64BIT
119#define VMALLOC_ADDR (max(0x100000000UL, (unsigned long) high_memory))
120#else
121#define VMALLOC_ADDR ((unsigned long) high_memory)
122#endif
123#define VMALLOC_OFFSET (8*1024*1024)
124#define VMALLOC_START ((VMALLOC_ADDR + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1))
Heiko Carstensf4eb07c2006-12-08 15:56:07 +0100125#define VMALLOC_END vmalloc_end
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126
Heiko Carstens8b62bc92006-12-04 15:40:56 +0100127/*
128 * We need some free virtual space to be able to do vmalloc.
129 * VMALLOC_MIN_SIZE defines the minimum size of the vmalloc
130 * area. On a machine with 2GB memory we make sure that we
131 * have at least 128MB free space for vmalloc. On a machine
Heiko Carstensf4eb07c2006-12-08 15:56:07 +0100132 * with 4TB we make sure we have at least 128GB.
Heiko Carstens8b62bc92006-12-04 15:40:56 +0100133 */
134#ifndef __s390x__
135#define VMALLOC_MIN_SIZE 0x8000000UL
Heiko Carstensf4eb07c2006-12-08 15:56:07 +0100136#define VMALLOC_END_INIT 0x80000000UL
Heiko Carstens8b62bc92006-12-04 15:40:56 +0100137#else /* __s390x__ */
Heiko Carstensf4eb07c2006-12-08 15:56:07 +0100138#define VMALLOC_MIN_SIZE 0x2000000000UL
139#define VMALLOC_END_INIT 0x40000000000UL
Heiko Carstens8b62bc92006-12-04 15:40:56 +0100140#endif /* __s390x__ */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141
142/*
143 * A 31 bit pagetable entry of S390 has following format:
144 * | PFRA | | OS |
145 * 0 0IP0
146 * 00000000001111111111222222222233
147 * 01234567890123456789012345678901
148 *
149 * I Page-Invalid Bit: Page is not available for address-translation
150 * P Page-Protection Bit: Store access not possible for page
151 *
152 * A 31 bit segmenttable entry of S390 has following format:
153 * | P-table origin | |PTL
154 * 0 IC
155 * 00000000001111111111222222222233
156 * 01234567890123456789012345678901
157 *
158 * I Segment-Invalid Bit: Segment is not available for address-translation
159 * C Common-Segment Bit: Segment is not private (PoP 3-30)
160 * PTL Page-Table-Length: Page-table length (PTL+1*16 entries -> up to 256)
161 *
162 * The 31 bit segmenttable origin of S390 has following format:
163 *
164 * |S-table origin | | STL |
165 * X **GPS
166 * 00000000001111111111222222222233
167 * 01234567890123456789012345678901
168 *
169 * X Space-Switch event:
170 * G Segment-Invalid Bit: *
171 * P Private-Space Bit: Segment is not private (PoP 3-30)
172 * S Storage-Alteration:
173 * STL Segment-Table-Length: Segment-table length (STL+1*16 entries -> up to 2048)
174 *
175 * A 64 bit pagetable entry of S390 has following format:
176 * | PFRA |0IP0| OS |
177 * 0000000000111111111122222222223333333333444444444455555555556666
178 * 0123456789012345678901234567890123456789012345678901234567890123
179 *
180 * I Page-Invalid Bit: Page is not available for address-translation
181 * P Page-Protection Bit: Store access not possible for page
182 *
183 * A 64 bit segmenttable entry of S390 has following format:
184 * | P-table origin | TT
185 * 0000000000111111111122222222223333333333444444444455555555556666
186 * 0123456789012345678901234567890123456789012345678901234567890123
187 *
188 * I Segment-Invalid Bit: Segment is not available for address-translation
189 * C Common-Segment Bit: Segment is not private (PoP 3-30)
190 * P Page-Protection Bit: Store access not possible for page
191 * TT Type 00
192 *
193 * A 64 bit region table entry of S390 has following format:
194 * | S-table origin | TF TTTL
195 * 0000000000111111111122222222223333333333444444444455555555556666
196 * 0123456789012345678901234567890123456789012345678901234567890123
197 *
198 * I Segment-Invalid Bit: Segment is not available for address-translation
199 * TT Type 01
200 * TF
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200201 * TL Table length
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202 *
203 * The 64 bit regiontable origin of S390 has following format:
204 * | region table origon | DTTL
205 * 0000000000111111111122222222223333333333444444444455555555556666
206 * 0123456789012345678901234567890123456789012345678901234567890123
207 *
208 * X Space-Switch event:
209 * G Segment-Invalid Bit:
210 * P Private-Space Bit:
211 * S Storage-Alteration:
212 * R Real space
213 * TL Table-Length:
214 *
215 * A storage key has the following format:
216 * | ACC |F|R|C|0|
217 * 0 3 4 5 6 7
218 * ACC: access key
219 * F : fetch protection bit
220 * R : referenced bit
221 * C : changed bit
222 */
223
224/* Hardware bits in the page table entry */
Martin Schwidefsky83377482006-10-18 18:30:51 +0200225#define _PAGE_RO 0x200 /* HW read-only bit */
226#define _PAGE_INVALID 0x400 /* HW invalid bit */
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200227
228/* Software bits in the page table entry */
Martin Schwidefsky83377482006-10-18 18:30:51 +0200229#define _PAGE_SWT 0x001 /* SW pte type bit t */
230#define _PAGE_SWX 0x002 /* SW pte type bit x */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231
Martin Schwidefsky83377482006-10-18 18:30:51 +0200232/* Six different types of pages. */
Gerald Schaefer9282ed92006-09-20 15:59:37 +0200233#define _PAGE_TYPE_EMPTY 0x400
234#define _PAGE_TYPE_NONE 0x401
Martin Schwidefsky83377482006-10-18 18:30:51 +0200235#define _PAGE_TYPE_SWAP 0x403
236#define _PAGE_TYPE_FILE 0x601 /* bit 0x002 is used for offset !! */
Gerald Schaefer9282ed92006-09-20 15:59:37 +0200237#define _PAGE_TYPE_RO 0x200
238#define _PAGE_TYPE_RW 0x000
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100239#define _PAGE_TYPE_EX_RO 0x202
240#define _PAGE_TYPE_EX_RW 0x002
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241
Martin Schwidefsky83377482006-10-18 18:30:51 +0200242/*
243 * PTE type bits are rather complicated. handle_pte_fault uses pte_present,
244 * pte_none and pte_file to find out the pte type WITHOUT holding the page
245 * table lock. ptep_clear_flush on the other hand uses ptep_clear_flush to
246 * invalidate a given pte. ipte sets the hw invalid bit and clears all tlbs
247 * for the page. The page table entry is set to _PAGE_TYPE_EMPTY afterwards.
248 * This change is done while holding the lock, but the intermediate step
249 * of a previously valid pte with the hw invalid bit set can be observed by
250 * handle_pte_fault. That makes it necessary that all valid pte types with
251 * the hw invalid bit set must be distinguishable from the four pte types
252 * empty, none, swap and file.
253 *
254 * irxt ipte irxt
255 * _PAGE_TYPE_EMPTY 1000 -> 1000
256 * _PAGE_TYPE_NONE 1001 -> 1001
257 * _PAGE_TYPE_SWAP 1011 -> 1011
258 * _PAGE_TYPE_FILE 11?1 -> 11?1
259 * _PAGE_TYPE_RO 0100 -> 1100
260 * _PAGE_TYPE_RW 0000 -> 1000
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100261 * _PAGE_TYPE_EX_RO 0110 -> 1110
262 * _PAGE_TYPE_EX_RW 0010 -> 1010
Martin Schwidefsky83377482006-10-18 18:30:51 +0200263 *
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100264 * pte_none is true for bits combinations 1000, 1010, 1100, 1110
Martin Schwidefsky83377482006-10-18 18:30:51 +0200265 * pte_present is true for bits combinations 0000, 0010, 0100, 0110, 1001
266 * pte_file is true for bits combinations 1101, 1111
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100267 * swap pte is 1011 and 0001, 0011, 0101, 0111 are invalid.
Martin Schwidefsky83377482006-10-18 18:30:51 +0200268 */
269
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270#ifndef __s390x__
271
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200272/* Bits in the segment table address-space-control-element */
273#define _ASCE_SPACE_SWITCH 0x80000000UL /* space switch event */
274#define _ASCE_ORIGIN_MASK 0x7ffff000UL /* segment table origin */
275#define _ASCE_PRIVATE_SPACE 0x100 /* private space control */
276#define _ASCE_ALT_EVENT 0x80 /* storage alteration event control */
277#define _ASCE_TABLE_LENGTH 0x7f /* 128 x 64 entries = 8k */
278
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279/* Bits in the segment table entry */
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200280#define _SEGMENT_ENTRY_ORIGIN 0x7fffffc0UL /* page table origin */
281#define _SEGMENT_ENTRY_INV 0x20 /* invalid segment table entry */
282#define _SEGMENT_ENTRY_COMMON 0x10 /* common segment bit */
283#define _SEGMENT_ENTRY_PTL 0x0f /* page table length */
284
285#define _SEGMENT_ENTRY (_SEGMENT_ENTRY_PTL)
286#define _SEGMENT_ENTRY_EMPTY (_SEGMENT_ENTRY_INV)
287
288#else /* __s390x__ */
289
290/* Bits in the segment/region table address-space-control-element */
291#define _ASCE_ORIGIN ~0xfffUL/* segment table origin */
292#define _ASCE_PRIVATE_SPACE 0x100 /* private space control */
293#define _ASCE_ALT_EVENT 0x80 /* storage alteration event control */
294#define _ASCE_SPACE_SWITCH 0x40 /* space switch event */
295#define _ASCE_REAL_SPACE 0x20 /* real space control */
296#define _ASCE_TYPE_MASK 0x0c /* asce table type mask */
297#define _ASCE_TYPE_REGION1 0x0c /* region first table type */
298#define _ASCE_TYPE_REGION2 0x08 /* region second table type */
299#define _ASCE_TYPE_REGION3 0x04 /* region third table type */
300#define _ASCE_TYPE_SEGMENT 0x00 /* segment table type */
301#define _ASCE_TABLE_LENGTH 0x03 /* region table length */
302
303/* Bits in the region table entry */
304#define _REGION_ENTRY_ORIGIN ~0xfffUL/* region/segment table origin */
305#define _REGION_ENTRY_INV 0x20 /* invalid region table entry */
306#define _REGION_ENTRY_TYPE_MASK 0x0c /* region/segment table type mask */
307#define _REGION_ENTRY_TYPE_R1 0x0c /* region first table type */
308#define _REGION_ENTRY_TYPE_R2 0x08 /* region second table type */
309#define _REGION_ENTRY_TYPE_R3 0x04 /* region third table type */
310#define _REGION_ENTRY_LENGTH 0x03 /* region third length */
311
312#define _REGION1_ENTRY (_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
313#define _REGION1_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INV)
314#define _REGION2_ENTRY (_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
315#define _REGION2_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INV)
316#define _REGION3_ENTRY (_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
317#define _REGION3_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INV)
318
319/* Bits in the segment table entry */
320#define _SEGMENT_ENTRY_ORIGIN ~0x7ffUL/* segment table origin */
321#define _SEGMENT_ENTRY_RO 0x200 /* page protection bit */
322#define _SEGMENT_ENTRY_INV 0x20 /* invalid segment table entry */
323
324#define _SEGMENT_ENTRY (0)
325#define _SEGMENT_ENTRY_EMPTY (_SEGMENT_ENTRY_INV)
326
327#endif /* __s390x__ */
328
329/*
330 * A user page table pointer has the space-switch-event bit, the
331 * private-space-control bit and the storage-alteration-event-control
332 * bit set. A kernel page table pointer doesn't need them.
333 */
334#define _ASCE_USER_BITS (_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
335 _ASCE_ALT_EVENT)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336
337/* Bits int the storage key */
338#define _PAGE_CHANGED 0x02 /* HW changed bit */
339#define _PAGE_REFERENCED 0x04 /* HW referenced bit */
340
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341/*
Gerald Schaefer9282ed92006-09-20 15:59:37 +0200342 * Page protection definitions.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343 */
Gerald Schaefer9282ed92006-09-20 15:59:37 +0200344#define PAGE_NONE __pgprot(_PAGE_TYPE_NONE)
345#define PAGE_RO __pgprot(_PAGE_TYPE_RO)
346#define PAGE_RW __pgprot(_PAGE_TYPE_RW)
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100347#define PAGE_EX_RO __pgprot(_PAGE_TYPE_EX_RO)
348#define PAGE_EX_RW __pgprot(_PAGE_TYPE_EX_RW)
Gerald Schaefer9282ed92006-09-20 15:59:37 +0200349
350#define PAGE_KERNEL PAGE_RW
351#define PAGE_COPY PAGE_RO
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352
353/*
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100354 * Dependent on the EXEC_PROTECT option s390 can do execute protection.
355 * Write permission always implies read permission. In theory with a
356 * primary/secondary page table execute only can be implemented but
357 * it would cost an additional bit in the pte to distinguish all the
358 * different pte types. To avoid that execute permission currently
359 * implies read permission as well.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 */
361 /*xwr*/
Gerald Schaefer9282ed92006-09-20 15:59:37 +0200362#define __P000 PAGE_NONE
363#define __P001 PAGE_RO
364#define __P010 PAGE_RO
365#define __P011 PAGE_RO
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100366#define __P100 PAGE_EX_RO
367#define __P101 PAGE_EX_RO
368#define __P110 PAGE_EX_RO
369#define __P111 PAGE_EX_RO
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370
Gerald Schaefer9282ed92006-09-20 15:59:37 +0200371#define __S000 PAGE_NONE
372#define __S001 PAGE_RO
373#define __S010 PAGE_RW
374#define __S011 PAGE_RW
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100375#define __S100 PAGE_EX_RO
376#define __S101 PAGE_EX_RO
377#define __S110 PAGE_EX_RW
378#define __S111 PAGE_EX_RW
379
380#ifndef __s390x__
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200381# define PxD_SHADOW_SHIFT 1
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100382#else /* __s390x__ */
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200383# define PxD_SHADOW_SHIFT 2
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100384#endif /* __s390x__ */
385
386static inline struct page *get_shadow_page(struct page *page)
387{
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200388 if (s390_noexec && page->index)
389 return virt_to_page((void *)(addr_t) page->index);
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100390 return NULL;
391}
392
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200393static inline void *get_shadow_pte(void *table)
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100394{
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200395 unsigned long addr, offset;
396 struct page *page;
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100397
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200398 addr = (unsigned long) table;
399 offset = addr & (PAGE_SIZE - 1);
400 page = virt_to_page((void *)(addr ^ offset));
401 return (void *)(addr_t)(page->index ? (page->index | offset) : 0UL);
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100402}
403
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200404static inline void *get_shadow_table(void *table)
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100405{
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200406 unsigned long addr, offset;
407 struct page *page;
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100408
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200409 addr = (unsigned long) table;
410 offset = addr & ((PAGE_SIZE << PxD_SHADOW_SHIFT) - 1);
411 page = virt_to_page((void *)(addr ^ offset));
412 return (void *)(addr_t)(page->index ? (page->index | offset) : 0UL);
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100413}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414
415/*
416 * Certain architectures need to do special things when PTEs
417 * within a page table are directly modified. Thus, the following
418 * hook is made available.
419 */
Martin Schwidefskyba8a9222007-10-22 12:52:44 +0200420static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
421 pte_t *pteptr, pte_t pteval)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422{
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100423 pte_t *shadow_pte = get_shadow_pte(pteptr);
424
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 *pteptr = pteval;
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100426 if (shadow_pte) {
427 if (!(pte_val(pteval) & _PAGE_INVALID) &&
428 (pte_val(pteval) & _PAGE_SWX))
429 pte_val(*shadow_pte) = pte_val(pteval) | _PAGE_RO;
430 else
431 pte_val(*shadow_pte) = _PAGE_TYPE_EMPTY;
432 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434
435/*
436 * pgd/pmd/pte query functions
437 */
438#ifndef __s390x__
439
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800440static inline int pgd_present(pgd_t pgd) { return 1; }
441static inline int pgd_none(pgd_t pgd) { return 0; }
442static inline int pgd_bad(pgd_t pgd) { return 0; }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200444static inline int pud_present(pud_t pud) { return 1; }
445static inline int pud_none(pud_t pud) { return 0; }
446static inline int pud_bad(pud_t pud) { return 0; }
447
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448#else /* __s390x__ */
449
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200450static inline int pgd_present(pgd_t pgd) { return 1; }
451static inline int pgd_none(pgd_t pgd) { return 0; }
452static inline int pgd_bad(pgd_t pgd) { return 0; }
453
454static inline int pud_present(pud_t pud)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455{
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200456 return pud_val(pud) & _REGION_ENTRY_ORIGIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457}
458
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200459static inline int pud_none(pud_t pud)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460{
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200461 return pud_val(pud) & _REGION_ENTRY_INV;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462}
463
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200464static inline int pud_bad(pud_t pud)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465{
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200466 unsigned long mask = ~_REGION_ENTRY_ORIGIN & ~_REGION_ENTRY_INV;
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200467 return (pud_val(pud) & mask) != _REGION3_ENTRY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468}
469
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200470#endif /* __s390x__ */
471
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800472static inline int pmd_present(pmd_t pmd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473{
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200474 return pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475}
476
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800477static inline int pmd_none(pmd_t pmd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478{
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200479 return pmd_val(pmd) & _SEGMENT_ENTRY_INV;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480}
481
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800482static inline int pmd_bad(pmd_t pmd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483{
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200484 unsigned long mask = ~_SEGMENT_ENTRY_ORIGIN & ~_SEGMENT_ENTRY_INV;
485 return (pmd_val(pmd) & mask) != _SEGMENT_ENTRY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486}
487
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800488static inline int pte_none(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489{
Martin Schwidefsky83377482006-10-18 18:30:51 +0200490 return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491}
492
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800493static inline int pte_present(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494{
Martin Schwidefsky83377482006-10-18 18:30:51 +0200495 unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT | _PAGE_SWX;
496 return (pte_val(pte) & mask) == _PAGE_TYPE_NONE ||
497 (!(pte_val(pte) & _PAGE_INVALID) &&
498 !(pte_val(pte) & _PAGE_SWT));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499}
500
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800501static inline int pte_file(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502{
Martin Schwidefsky83377482006-10-18 18:30:51 +0200503 unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT;
504 return (pte_val(pte) & mask) == _PAGE_TYPE_FILE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505}
506
Martin Schwidefskyba8a9222007-10-22 12:52:44 +0200507#define __HAVE_ARCH_PTE_SAME
508#define pte_same(a,b) (pte_val(a) == pte_val(b))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509
510/*
511 * query functions pte_write/pte_dirty/pte_young only work if
512 * pte_present() is true. Undefined behaviour if not..
513 */
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800514static inline int pte_write(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515{
516 return (pte_val(pte) & _PAGE_RO) == 0;
517}
518
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800519static inline int pte_dirty(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520{
521 /* A pte is neither clean nor dirty on s/390. The dirty bit
522 * is in the storage key. See page_test_and_clear_dirty for
523 * details.
524 */
525 return 0;
526}
527
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800528static inline int pte_young(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529{
530 /* A pte is neither young nor old on s/390. The young bit
531 * is in the storage key. See page_test_and_clear_young for
532 * details.
533 */
534 return 0;
535}
536
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537/*
538 * pgd/pmd/pte modification functions
539 */
540
541#ifndef __s390x__
542
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200543#define pgd_clear(pgd) do { } while (0)
544#define pud_clear(pud) do { } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100546static inline void pmd_clear_kernel(pmd_t * pmdp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547{
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200548 pmd_val(pmdp[0]) = _SEGMENT_ENTRY_EMPTY;
549 pmd_val(pmdp[1]) = _SEGMENT_ENTRY_EMPTY;
550 pmd_val(pmdp[2]) = _SEGMENT_ENTRY_EMPTY;
551 pmd_val(pmdp[3]) = _SEGMENT_ENTRY_EMPTY;
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100552}
553
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554#else /* __s390x__ */
555
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200556#define pgd_clear(pgd) do { } while (0)
557
558static inline void pud_clear_kernel(pud_t *pud)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559{
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200560 pud_val(*pud) = _REGION3_ENTRY_EMPTY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561}
562
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200563static inline void pud_clear(pud_t * pud)
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100564{
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200565 pud_t *shadow = get_shadow_table(pud);
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100566
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200567 pud_clear_kernel(pud);
568 if (shadow)
569 pud_clear_kernel(shadow);
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100570}
571
572static inline void pmd_clear_kernel(pmd_t * pmdp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573{
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200574 pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY;
575 pmd_val1(*pmdp) = _SEGMENT_ENTRY_EMPTY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576}
577
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200578#endif /* __s390x__ */
579
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100580static inline void pmd_clear(pmd_t * pmdp)
581{
Martin Schwidefsky3610cce2007-10-22 12:52:47 +0200582 pmd_t *shadow_pmd = get_shadow_table(pmdp);
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100583
584 pmd_clear_kernel(pmdp);
585 if (shadow_pmd)
586 pmd_clear_kernel(shadow_pmd);
587}
588
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800589static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590{
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100591 pte_t *shadow_pte = get_shadow_pte(ptep);
592
Gerald Schaefer9282ed92006-09-20 15:59:37 +0200593 pte_val(*ptep) = _PAGE_TYPE_EMPTY;
Gerald Schaeferc1821c22007-02-05 21:18:17 +0100594 if (shadow_pte)
595 pte_val(*shadow_pte) = _PAGE_TYPE_EMPTY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596}
597
598/*
599 * The following pte modification functions only work if
600 * pte_present() is true. Undefined behaviour if not..
601 */
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800602static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603{
604 pte_val(pte) &= PAGE_MASK;
605 pte_val(pte) |= pgprot_val(newprot);
606 return pte;
607}
608
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800609static inline pte_t pte_wrprotect(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610{
Gerald Schaefer9282ed92006-09-20 15:59:37 +0200611 /* Do not clobber _PAGE_TYPE_NONE pages! */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612 if (!(pte_val(pte) & _PAGE_INVALID))
613 pte_val(pte) |= _PAGE_RO;
614 return pte;
615}
616
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800617static inline pte_t pte_mkwrite(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618{
619 pte_val(pte) &= ~_PAGE_RO;
620 return pte;
621}
622
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800623static inline pte_t pte_mkclean(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624{
625 /* The only user of pte_mkclean is the fork() code.
626 We must *not* clear the *physical* page dirty bit
627 just because fork() wants to clear the dirty bit in
628 *one* of the page's mappings. So we just do nothing. */
629 return pte;
630}
631
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800632static inline pte_t pte_mkdirty(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633{
634 /* We do not explicitly set the dirty bit because the
635 * sske instruction is slow. It is faster to let the
636 * next instruction set the dirty bit.
637 */
638 return pte;
639}
640
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800641static inline pte_t pte_mkold(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642{
643 /* S/390 doesn't keep its dirty/referenced bit in the pte.
644 * There is no point in clearing the real referenced bit.
645 */
646 return pte;
647}
648
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800649static inline pte_t pte_mkyoung(pte_t pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650{
651 /* S/390 doesn't keep its dirty/referenced bit in the pte.
652 * There is no point in setting the real referenced bit.
653 */
654 return pte;
655}
656
Martin Schwidefskyba8a9222007-10-22 12:52:44 +0200657#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
658static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
659 unsigned long addr, pte_t *ptep)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660{
661 return 0;
662}
663
Martin Schwidefskyba8a9222007-10-22 12:52:44 +0200664#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
665static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
666 unsigned long address, pte_t *ptep)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667{
668 /* No need to flush TLB; bits are in storage key */
Martin Schwidefskyba8a9222007-10-22 12:52:44 +0200669 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670}
671
Gerald Schaefer9282ed92006-09-20 15:59:37 +0200672static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
673{
674 if (!(pte_val(*ptep) & _PAGE_INVALID)) {
675#ifndef __s390x__
676 /* S390 has 1mb segments, we are emulating 4MB segments */
677 pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00);
678#else
679 /* ipte in zarch mode can do the math */
680 pte_t *pto = ptep;
681#endif
Martin Schwidefsky94c12cc2006-09-28 16:56:43 +0200682 asm volatile(
683 " ipte %2,%3"
684 : "=m" (*ptep) : "m" (*ptep),
685 "a" (pto), "a" (address));
Gerald Schaefer9282ed92006-09-20 15:59:37 +0200686 }
687 pte_val(*ptep) = _PAGE_TYPE_EMPTY;
688}
689
Martin Schwidefskyf0e47c22007-07-17 04:03:03 -0700690static inline void ptep_invalidate(unsigned long address, pte_t *ptep)
691{
692 __ptep_ipte(address, ptep);
693 ptep = get_shadow_pte(ptep);
694 if (ptep)
695 __ptep_ipte(address, ptep);
696}
697
Martin Schwidefskyba8a9222007-10-22 12:52:44 +0200698/*
699 * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
700 * both clear the TLB for the unmapped pte. The reason is that
701 * ptep_get_and_clear is used in common code (e.g. change_pte_range)
702 * to modify an active pte. The sequence is
703 * 1) ptep_get_and_clear
704 * 2) set_pte_at
705 * 3) flush_tlb_range
706 * On s390 the tlb needs to get flushed with the modification of the pte
707 * if the pte is active. The only way how this can be implemented is to
708 * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
709 * is a nop.
710 */
711#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
712#define ptep_get_and_clear(__mm, __address, __ptep) \
713({ \
714 pte_t __pte = *(__ptep); \
715 if (atomic_read(&(__mm)->mm_users) > 1 || \
716 (__mm) != current->active_mm) \
717 ptep_invalidate(__address, __ptep); \
718 else \
719 pte_clear((__mm), (__address), (__ptep)); \
720 __pte; \
721})
722
723#define __HAVE_ARCH_PTEP_CLEAR_FLUSH
Martin Schwidefskyf0e47c22007-07-17 04:03:03 -0700724static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
725 unsigned long address, pte_t *ptep)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726{
727 pte_t pte = *ptep;
Martin Schwidefskyf0e47c22007-07-17 04:03:03 -0700728 ptep_invalidate(address, ptep);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 return pte;
730}
731
Martin Schwidefskyba8a9222007-10-22 12:52:44 +0200732/*
733 * The batched pte unmap code uses ptep_get_and_clear_full to clear the
734 * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
735 * tlbs of an mm if it can guarantee that the ptes of the mm_struct
736 * cannot be accessed while the batched unmap is running. In this case
737 * full==1 and a simple pte_clear is enough. See tlb.h.
738 */
739#define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
740static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
741 unsigned long addr,
742 pte_t *ptep, int full)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743{
Martin Schwidefskyba8a9222007-10-22 12:52:44 +0200744 pte_t pte = *ptep;
745
746 if (full)
747 pte_clear(mm, addr, ptep);
748 else
749 ptep_invalidate(addr, ptep);
750 return pte;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751}
752
Martin Schwidefskyba8a9222007-10-22 12:52:44 +0200753#define __HAVE_ARCH_PTEP_SET_WRPROTECT
754#define ptep_set_wrprotect(__mm, __addr, __ptep) \
755({ \
756 pte_t __pte = *(__ptep); \
757 if (pte_write(__pte)) { \
758 if (atomic_read(&(__mm)->mm_users) > 1 || \
759 (__mm) != current->active_mm) \
760 ptep_invalidate(__addr, __ptep); \
761 set_pte_at(__mm, __addr, __ptep, pte_wrprotect(__pte)); \
762 } \
763})
764
765#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
Martin Schwidefskyf0e47c22007-07-17 04:03:03 -0700766#define ptep_set_access_flags(__vma, __addr, __ptep, __entry, __dirty) \
767({ \
768 int __changed = !pte_same(*(__ptep), __entry); \
769 if (__changed) { \
770 ptep_invalidate(__addr, __ptep); \
771 set_pte_at((__vma)->vm_mm, __addr, __ptep, __entry); \
772 } \
773 __changed; \
Benjamin Herrenschmidt8dab5242007-06-16 10:16:12 -0700774})
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775
776/*
777 * Test and clear dirty bit in storage key.
778 * We can't clear the changed bit atomically. This is a potential
779 * race against modification of the referenced bit. This function
780 * should therefore only be called if it is not mapped in any
781 * address space.
782 */
Martin Schwidefskyba8a9222007-10-22 12:52:44 +0200783#define __HAVE_ARCH_PAGE_TEST_DIRTY
Martin Schwidefsky6c210482007-04-27 16:01:57 +0200784static inline int page_test_dirty(struct page *page)
Heiko Carstens2dcea572006-09-29 01:58:41 -0700785{
Martin Schwidefsky6c210482007-04-27 16:01:57 +0200786 return (page_get_storage_key(page_to_phys(page)) & _PAGE_CHANGED) != 0;
787}
Heiko Carstens2dcea572006-09-29 01:58:41 -0700788
Martin Schwidefskyba8a9222007-10-22 12:52:44 +0200789#define __HAVE_ARCH_PAGE_CLEAR_DIRTY
Martin Schwidefsky6c210482007-04-27 16:01:57 +0200790static inline void page_clear_dirty(struct page *page)
791{
792 page_set_storage_key(page_to_phys(page), PAGE_DEFAULT_KEY);
Heiko Carstens2dcea572006-09-29 01:58:41 -0700793}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794
795/*
796 * Test and clear referenced bit in storage key.
797 */
Martin Schwidefskyba8a9222007-10-22 12:52:44 +0200798#define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_YOUNG
Heiko Carstens2dcea572006-09-29 01:58:41 -0700799static inline int page_test_and_clear_young(struct page *page)
800{
Heiko Carstens0b2b6e1d2006-10-04 20:02:23 +0200801 unsigned long physpage = page_to_phys(page);
Heiko Carstens2dcea572006-09-29 01:58:41 -0700802 int ccode;
803
Heiko Carstens0b2b6e1d2006-10-04 20:02:23 +0200804 asm volatile(
805 " rrbe 0,%1\n"
806 " ipm %0\n"
807 " srl %0,28\n"
Heiko Carstens2dcea572006-09-29 01:58:41 -0700808 : "=d" (ccode) : "a" (physpage) : "cc" );
809 return ccode & 2;
810}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811
812/*
813 * Conversion functions: convert a page and protection to a page entry,
814 * and a page entry and page directory to the page they refer to.
815 */
816static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
817{
818 pte_t __pte;
819 pte_val(__pte) = physpage + pgprot_val(pgprot);
820 return __pte;
821}
822
Heiko Carstens2dcea572006-09-29 01:58:41 -0700823static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
824{
Heiko Carstens0b2b6e1d2006-10-04 20:02:23 +0200825 unsigned long physpage = page_to_phys(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826
Heiko Carstens2dcea572006-09-29 01:58:41 -0700827 return mk_pte_phys(physpage, pgprot);
828}
829
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830#define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200831#define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
832#define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
833#define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200835#define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836#define pgd_offset_k(address) pgd_offset(&init_mm, address)
837
838#ifndef __s390x__
839
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200840#define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
841#define pud_deref(pmd) ({ BUG(); 0UL; })
842#define pgd_deref(pmd) ({ BUG(); 0UL; })
843
844#define pud_offset(pgd, address) ((pud_t *) pgd)
845#define pmd_offset(pud, address) ((pmd_t *) pud + pmd_index(address))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846
847#else /* __s390x__ */
848
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200849#define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
850#define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)
851#define pgd_deref(pgd) ({ BUG(); 0UL; })
852
853#define pud_offset(pgd, address) ((pud_t *) pgd)
854
855static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
856{
857 pmd_t *pmd = (pmd_t *) pud_deref(*pud);
858 return pmd + pmd_index(address);
859}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860
861#endif /* __s390x__ */
862
Martin Schwidefsky190a1d72007-10-22 12:52:48 +0200863#define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
864#define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
865#define pte_page(x) pfn_to_page(pte_pfn(x))
866
867#define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
868
869/* Find an entry in the lowest level page table.. */
870#define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr))
871#define pte_offset_kernel(pmd, address) pte_offset(pmd,address)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872#define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
873#define pte_offset_map_nested(pmd, address) pte_offset_kernel(pmd, address)
874#define pte_unmap(pte) do { } while (0)
875#define pte_unmap_nested(pte) do { } while (0)
876
877/*
878 * 31 bit swap entry format:
879 * A page-table entry has some bits we have to treat in a special way.
880 * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
881 * exception will occur instead of a page translation exception. The
882 * specifiation exception has the bad habit not to store necessary
883 * information in the lowcore.
884 * Bit 21 and bit 22 are the page invalid bit and the page protection
885 * bit. We set both to indicate a swapped page.
886 * Bit 30 and 31 are used to distinguish the different page types. For
887 * a swapped page these bits need to be zero.
888 * This leaves the bits 1-19 and bits 24-29 to store type and offset.
889 * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
890 * plus 24 for the offset.
891 * 0| offset |0110|o|type |00|
892 * 0 0000000001111111111 2222 2 22222 33
893 * 0 1234567890123456789 0123 4 56789 01
894 *
895 * 64 bit swap entry format:
896 * A page-table entry has some bits we have to treat in a special way.
897 * Bits 52 and bit 55 have to be zero, otherwise an specification
898 * exception will occur instead of a page translation exception. The
899 * specifiation exception has the bad habit not to store necessary
900 * information in the lowcore.
901 * Bit 53 and bit 54 are the page invalid bit and the page protection
902 * bit. We set both to indicate a swapped page.
903 * Bit 62 and 63 are used to distinguish the different page types. For
904 * a swapped page these bits need to be zero.
905 * This leaves the bits 0-51 and bits 56-61 to store type and offset.
906 * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
907 * plus 56 for the offset.
908 * | offset |0110|o|type |00|
909 * 0000000000111111111122222222223333333333444444444455 5555 5 55566 66
910 * 0123456789012345678901234567890123456789012345678901 2345 6 78901 23
911 */
912#ifndef __s390x__
913#define __SWP_OFFSET_MASK (~0UL >> 12)
914#else
915#define __SWP_OFFSET_MASK (~0UL >> 11)
916#endif
Adrian Bunk4448aaf2005-11-08 21:34:42 -0800917static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918{
919 pte_t pte;
920 offset &= __SWP_OFFSET_MASK;
Gerald Schaefer9282ed92006-09-20 15:59:37 +0200921 pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) |
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 ((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
923 return pte;
924}
925
926#define __swp_type(entry) (((entry).val >> 2) & 0x1f)
927#define __swp_offset(entry) (((entry).val >> 11) | (((entry).val >> 7) & 1))
928#define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
929
930#define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
931#define __swp_entry_to_pte(x) ((pte_t) { (x).val })
932
933#ifndef __s390x__
934# define PTE_FILE_MAX_BITS 26
935#else /* __s390x__ */
936# define PTE_FILE_MAX_BITS 59
937#endif /* __s390x__ */
938
939#define pte_to_pgoff(__pte) \
940 ((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
941
942#define pgoff_to_pte(__off) \
943 ((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
Gerald Schaefer9282ed92006-09-20 15:59:37 +0200944 | _PAGE_TYPE_FILE })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945
946#endif /* !__ASSEMBLY__ */
947
948#define kern_addr_valid(addr) (1)
949
Heiko Carstensf4eb07c2006-12-08 15:56:07 +0100950extern int add_shared_memory(unsigned long start, unsigned long size);
951extern int remove_shared_memory(unsigned long start, unsigned long size);
952
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953/*
954 * No page table caches to initialise
955 */
956#define pgtable_cache_init() do { } while (0)
957
Heiko Carstensf4eb07c2006-12-08 15:56:07 +0100958#define __HAVE_ARCH_MEMMAP_INIT
959extern void memmap_init(unsigned long, int, unsigned long, unsigned long);
960
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961#include <asm-generic/pgtable.h>
962
963#endif /* _S390_PAGE_H */