2 * include/asm-s390/pgtable.h
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)
10 * Derived from "include/asm-i386/pgtable.h"
13 #ifndef _ASM_S390_PGTABLE_H
14 #define _ASM_S390_PGTABLE_H
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
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
28 * This file contains the functions and defines necessary to modify and use
29 * the S390 page table tree.
32 #include <linux/mm_types.h>
34 #include <asm/processor.h>
36 extern pgd_t swapper_pg_dir[] __attribute__ ((aligned (4096)));
37 extern void paging_init(void);
38 extern void vmem_map_init(void);
41 * The S390 doesn't have any external MMU info: the kernel page
42 * tables contain all the necessary information.
44 #define update_mmu_cache(vma, address, pte) do { } while (0)
47 * ZERO_PAGE is a global shared page that is always zero: used
48 * for zero-mapped memory areas etc..
50 extern char empty_zero_page[PAGE_SIZE];
51 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
52 #endif /* !__ASSEMBLY__ */
55 * PMD_SHIFT determines the size of the area a second-level page
57 * PGDIR_SHIFT determines what a third-level page table entry can map
62 # define PGDIR_SHIFT 22
66 # define PGDIR_SHIFT 31
67 #endif /* __s390x__ */
69 #define PMD_SIZE (1UL << PMD_SHIFT)
70 #define PMD_MASK (~(PMD_SIZE-1))
71 #define PUD_SIZE (1UL << PUD_SHIFT)
72 #define PUD_MASK (~(PUD_SIZE-1))
73 #define PGDIR_SIZE (1UL << PGDIR_SHIFT)
74 #define PGDIR_MASK (~(PGDIR_SIZE-1))
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
83 # define PTRS_PER_PTE 1024
84 # define PTRS_PER_PMD 1
85 # define PTRS_PER_PUD 1
86 # define PTRS_PER_PGD 512
88 # define PTRS_PER_PTE 512
89 # define PTRS_PER_PMD 1024
90 # define PTRS_PER_PUD 1
91 # define PTRS_PER_PGD 2048
92 #endif /* __s390x__ */
94 #define FIRST_USER_ADDRESS 0
96 #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))
100 #define pud_ERROR(e) \
101 printk("%s:%d: bad pud %p.\n", __FILE__, __LINE__, (void *) pud_val(e))
102 #define pgd_ERROR(e) \
103 printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
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. ;)
113 * vmalloc area starts at 4GB to prevent syscall table entry exchanging
116 extern unsigned long vmalloc_end;
119 #define VMALLOC_ADDR (max(0x100000000UL, (unsigned long) high_memory))
121 #define VMALLOC_ADDR ((unsigned long) high_memory)
123 #define VMALLOC_OFFSET (8*1024*1024)
124 #define VMALLOC_START ((VMALLOC_ADDR + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1))
125 #define VMALLOC_END vmalloc_end
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
132 * with 4TB we make sure we have at least 128GB.
135 #define VMALLOC_MIN_SIZE 0x8000000UL
136 #define VMALLOC_END_INIT 0x80000000UL
137 #else /* __s390x__ */
138 #define VMALLOC_MIN_SIZE 0x2000000000UL
139 #define VMALLOC_END_INIT 0x40000000000UL
140 #endif /* __s390x__ */
143 * A 31 bit pagetable entry of S390 has following format:
146 * 00000000001111111111222222222233
147 * 01234567890123456789012345678901
149 * I Page-Invalid Bit: Page is not available for address-translation
150 * P Page-Protection Bit: Store access not possible for page
152 * A 31 bit segmenttable entry of S390 has following format:
153 * | P-table origin | |PTL
155 * 00000000001111111111222222222233
156 * 01234567890123456789012345678901
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)
162 * The 31 bit segmenttable origin of S390 has following format:
164 * |S-table origin | | STL |
166 * 00000000001111111111222222222233
167 * 01234567890123456789012345678901
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)
175 * A 64 bit pagetable entry of S390 has following format:
177 * 0000000000111111111122222222223333333333444444444455555555556666
178 * 0123456789012345678901234567890123456789012345678901234567890123
180 * I Page-Invalid Bit: Page is not available for address-translation
181 * P Page-Protection Bit: Store access not possible for page
183 * A 64 bit segmenttable entry of S390 has following format:
184 * | P-table origin | TT
185 * 0000000000111111111122222222223333333333444444444455555555556666
186 * 0123456789012345678901234567890123456789012345678901234567890123
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
193 * A 64 bit region table entry of S390 has following format:
194 * | S-table origin | TF TTTL
195 * 0000000000111111111122222222223333333333444444444455555555556666
196 * 0123456789012345678901234567890123456789012345678901234567890123
198 * I Segment-Invalid Bit: Segment is not available for address-translation
203 * The 64 bit regiontable origin of S390 has following format:
204 * | region table origon | DTTL
205 * 0000000000111111111122222222223333333333444444444455555555556666
206 * 0123456789012345678901234567890123456789012345678901234567890123
208 * X Space-Switch event:
209 * G Segment-Invalid Bit:
210 * P Private-Space Bit:
211 * S Storage-Alteration:
215 * A storage key has the following format:
219 * F : fetch protection bit
224 /* Hardware bits in the page table entry */
225 #define _PAGE_RO 0x200 /* HW read-only bit */
226 #define _PAGE_INVALID 0x400 /* HW invalid bit */
228 /* Software bits in the page table entry */
229 #define _PAGE_SWT 0x001 /* SW pte type bit t */
230 #define _PAGE_SWX 0x002 /* SW pte type bit x */
232 /* Six different types of pages. */
233 #define _PAGE_TYPE_EMPTY 0x400
234 #define _PAGE_TYPE_NONE 0x401
235 #define _PAGE_TYPE_SWAP 0x403
236 #define _PAGE_TYPE_FILE 0x601 /* bit 0x002 is used for offset !! */
237 #define _PAGE_TYPE_RO 0x200
238 #define _PAGE_TYPE_RW 0x000
239 #define _PAGE_TYPE_EX_RO 0x202
240 #define _PAGE_TYPE_EX_RW 0x002
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.
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
261 * _PAGE_TYPE_EX_RO 0110 -> 1110
262 * _PAGE_TYPE_EX_RW 0010 -> 1010
264 * pte_none is true for bits combinations 1000, 1010, 1100, 1110
265 * pte_present is true for bits combinations 0000, 0010, 0100, 0110, 1001
266 * pte_file is true for bits combinations 1101, 1111
267 * swap pte is 1011 and 0001, 0011, 0101, 0111 are invalid.
272 /* 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 */
279 /* Bits in the segment table entry */
280 #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 */
285 #define _SEGMENT_ENTRY (_SEGMENT_ENTRY_PTL)
286 #define _SEGMENT_ENTRY_EMPTY (_SEGMENT_ENTRY_INV)
288 #else /* __s390x__ */
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 */
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 */
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)
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 */
324 #define _SEGMENT_ENTRY (0)
325 #define _SEGMENT_ENTRY_EMPTY (_SEGMENT_ENTRY_INV)
327 #endif /* __s390x__ */
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.
334 #define _ASCE_USER_BITS (_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
337 /* Bits int the storage key */
338 #define _PAGE_CHANGED 0x02 /* HW changed bit */
339 #define _PAGE_REFERENCED 0x04 /* HW referenced bit */
342 * Page protection definitions.
344 #define PAGE_NONE __pgprot(_PAGE_TYPE_NONE)
345 #define PAGE_RO __pgprot(_PAGE_TYPE_RO)
346 #define PAGE_RW __pgprot(_PAGE_TYPE_RW)
347 #define PAGE_EX_RO __pgprot(_PAGE_TYPE_EX_RO)
348 #define PAGE_EX_RW __pgprot(_PAGE_TYPE_EX_RW)
350 #define PAGE_KERNEL PAGE_RW
351 #define PAGE_COPY PAGE_RO
354 * 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.
362 #define __P000 PAGE_NONE
363 #define __P001 PAGE_RO
364 #define __P010 PAGE_RO
365 #define __P011 PAGE_RO
366 #define __P100 PAGE_EX_RO
367 #define __P101 PAGE_EX_RO
368 #define __P110 PAGE_EX_RO
369 #define __P111 PAGE_EX_RO
371 #define __S000 PAGE_NONE
372 #define __S001 PAGE_RO
373 #define __S010 PAGE_RW
374 #define __S011 PAGE_RW
375 #define __S100 PAGE_EX_RO
376 #define __S101 PAGE_EX_RO
377 #define __S110 PAGE_EX_RW
378 #define __S111 PAGE_EX_RW
381 # define PxD_SHADOW_SHIFT 1
382 #else /* __s390x__ */
383 # define PxD_SHADOW_SHIFT 2
384 #endif /* __s390x__ */
386 static inline struct page *get_shadow_page(struct page *page)
388 if (s390_noexec && page->index)
389 return virt_to_page((void *)(addr_t) page->index);
393 static inline void *get_shadow_pte(void *table)
395 unsigned long addr, offset;
398 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);
404 static inline void *get_shadow_table(void *table)
406 unsigned long addr, offset;
409 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);
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.
420 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
421 pte_t *pteptr, pte_t pteval)
423 pte_t *shadow_pte = get_shadow_pte(pteptr);
427 if (!(pte_val(pteval) & _PAGE_INVALID) &&
428 (pte_val(pteval) & _PAGE_SWX))
429 pte_val(*shadow_pte) = pte_val(pteval) | _PAGE_RO;
431 pte_val(*shadow_pte) = _PAGE_TYPE_EMPTY;
436 * pgd/pmd/pte query functions
440 static inline int pgd_present(pgd_t pgd) { return 1; }
441 static inline int pgd_none(pgd_t pgd) { return 0; }
442 static inline int pgd_bad(pgd_t pgd) { return 0; }
444 static inline int pud_present(pud_t pud) { return 1; }
445 static inline int pud_none(pud_t pud) { return 0; }
446 static inline int pud_bad(pud_t pud) { return 0; }
448 #else /* __s390x__ */
450 static inline int pgd_present(pgd_t pgd) { return 1; }
451 static inline int pgd_none(pgd_t pgd) { return 0; }
452 static inline int pgd_bad(pgd_t pgd) { return 0; }
454 static inline int pud_present(pud_t pud)
456 return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
459 static inline int pud_none(pud_t pud)
461 return (pud_val(pud) & _REGION_ENTRY_INV) != 0UL;
464 static inline int pud_bad(pud_t pud)
466 unsigned long mask = ~_REGION_ENTRY_ORIGIN & ~_REGION_ENTRY_INV;
467 return (pud_val(pud) & mask) != _REGION3_ENTRY;
470 #endif /* __s390x__ */
472 static inline int pmd_present(pmd_t pmd)
474 return (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN) != 0UL;
477 static inline int pmd_none(pmd_t pmd)
479 return (pmd_val(pmd) & _SEGMENT_ENTRY_INV) != 0UL;
482 static inline int pmd_bad(pmd_t pmd)
484 unsigned long mask = ~_SEGMENT_ENTRY_ORIGIN & ~_SEGMENT_ENTRY_INV;
485 return (pmd_val(pmd) & mask) != _SEGMENT_ENTRY;
488 static inline int pte_none(pte_t pte)
490 return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT);
493 static inline int pte_present(pte_t pte)
495 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));
501 static inline int pte_file(pte_t pte)
503 unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT;
504 return (pte_val(pte) & mask) == _PAGE_TYPE_FILE;
507 #define __HAVE_ARCH_PTE_SAME
508 #define pte_same(a,b) (pte_val(a) == pte_val(b))
511 * query functions pte_write/pte_dirty/pte_young only work if
512 * pte_present() is true. Undefined behaviour if not..
514 static inline int pte_write(pte_t pte)
516 return (pte_val(pte) & _PAGE_RO) == 0;
519 static inline int pte_dirty(pte_t pte)
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
528 static inline int pte_young(pte_t pte)
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
538 * pgd/pmd/pte modification functions
543 #define pgd_clear(pgd) do { } while (0)
544 #define pud_clear(pud) do { } while (0)
546 static inline void pmd_clear_kernel(pmd_t * pmdp)
548 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;
554 #else /* __s390x__ */
556 #define pgd_clear(pgd) do { } while (0)
558 static inline void pud_clear_kernel(pud_t *pud)
560 pud_val(*pud) = _REGION3_ENTRY_EMPTY;
563 static inline void pud_clear(pud_t * pud)
565 pud_t *shadow = get_shadow_table(pud);
567 pud_clear_kernel(pud);
569 pud_clear_kernel(shadow);
572 static inline void pmd_clear_kernel(pmd_t * pmdp)
574 pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY;
575 pmd_val1(*pmdp) = _SEGMENT_ENTRY_EMPTY;
578 #endif /* __s390x__ */
580 static inline void pmd_clear(pmd_t * pmdp)
582 pmd_t *shadow_pmd = get_shadow_table(pmdp);
584 pmd_clear_kernel(pmdp);
586 pmd_clear_kernel(shadow_pmd);
589 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
591 pte_t *shadow_pte = get_shadow_pte(ptep);
593 pte_val(*ptep) = _PAGE_TYPE_EMPTY;
595 pte_val(*shadow_pte) = _PAGE_TYPE_EMPTY;
599 * The following pte modification functions only work if
600 * pte_present() is true. Undefined behaviour if not..
602 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
604 pte_val(pte) &= PAGE_MASK;
605 pte_val(pte) |= pgprot_val(newprot);
609 static inline pte_t pte_wrprotect(pte_t pte)
611 /* Do not clobber _PAGE_TYPE_NONE pages! */
612 if (!(pte_val(pte) & _PAGE_INVALID))
613 pte_val(pte) |= _PAGE_RO;
617 static inline pte_t pte_mkwrite(pte_t pte)
619 pte_val(pte) &= ~_PAGE_RO;
623 static inline pte_t pte_mkclean(pte_t pte)
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. */
632 static inline pte_t pte_mkdirty(pte_t pte)
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.
641 static inline pte_t pte_mkold(pte_t pte)
643 /* S/390 doesn't keep its dirty/referenced bit in the pte.
644 * There is no point in clearing the real referenced bit.
649 static inline pte_t pte_mkyoung(pte_t pte)
651 /* S/390 doesn't keep its dirty/referenced bit in the pte.
652 * There is no point in setting the real referenced bit.
657 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
658 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
659 unsigned long addr, pte_t *ptep)
664 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
665 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
666 unsigned long address, pte_t *ptep)
668 /* No need to flush TLB; bits are in storage key */
672 static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
674 if (!(pte_val(*ptep) & _PAGE_INVALID)) {
676 /* S390 has 1mb segments, we are emulating 4MB segments */
677 pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00);
679 /* ipte in zarch mode can do the math */
684 : "=m" (*ptep) : "m" (*ptep),
685 "a" (pto), "a" (address));
687 pte_val(*ptep) = _PAGE_TYPE_EMPTY;
690 static inline void ptep_invalidate(unsigned long address, pte_t *ptep)
692 __ptep_ipte(address, ptep);
693 ptep = get_shadow_pte(ptep);
695 __ptep_ipte(address, ptep);
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
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
711 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
712 #define ptep_get_and_clear(__mm, __address, __ptep) \
714 pte_t __pte = *(__ptep); \
715 if (atomic_read(&(__mm)->mm_users) > 1 || \
716 (__mm) != current->active_mm) \
717 ptep_invalidate(__address, __ptep); \
719 pte_clear((__mm), (__address), (__ptep)); \
723 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
724 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
725 unsigned long address, pte_t *ptep)
728 ptep_invalidate(address, ptep);
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.
739 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
740 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
742 pte_t *ptep, int full)
747 pte_clear(mm, addr, ptep);
749 ptep_invalidate(addr, ptep);
753 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
754 #define ptep_set_wrprotect(__mm, __addr, __ptep) \
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)); \
765 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
766 #define ptep_set_access_flags(__vma, __addr, __ptep, __entry, __dirty) \
768 int __changed = !pte_same(*(__ptep), __entry); \
770 ptep_invalidate(__addr, __ptep); \
771 set_pte_at((__vma)->vm_mm, __addr, __ptep, __entry); \
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
783 #define __HAVE_ARCH_PAGE_TEST_DIRTY
784 static inline int page_test_dirty(struct page *page)
786 return (page_get_storage_key(page_to_phys(page)) & _PAGE_CHANGED) != 0;
789 #define __HAVE_ARCH_PAGE_CLEAR_DIRTY
790 static inline void page_clear_dirty(struct page *page)
792 page_set_storage_key(page_to_phys(page), PAGE_DEFAULT_KEY);
796 * Test and clear referenced bit in storage key.
798 #define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_YOUNG
799 static inline int page_test_and_clear_young(struct page *page)
801 unsigned long physpage = page_to_phys(page);
808 : "=d" (ccode) : "a" (physpage) : "cc" );
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.
816 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
819 pte_val(__pte) = physpage + pgprot_val(pgprot);
823 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
825 unsigned long physpage = page_to_phys(page);
827 return mk_pte_phys(physpage, pgprot);
830 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
831 #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))
835 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
836 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
840 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
841 #define pud_deref(pmd) ({ BUG(); 0UL; })
842 #define pgd_deref(pmd) ({ BUG(); 0UL; })
844 #define pud_offset(pgd, address) ((pud_t *) pgd)
845 #define pmd_offset(pud, address) ((pmd_t *) pud + pmd_index(address))
847 #else /* __s390x__ */
849 #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; })
853 #define pud_offset(pgd, address) ((pud_t *) pgd)
855 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
857 pmd_t *pmd = (pmd_t *) pud_deref(*pud);
858 return pmd + pmd_index(address);
861 #endif /* __s390x__ */
863 #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))
867 #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
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)
872 #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)
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
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
913 #define __SWP_OFFSET_MASK (~0UL >> 12)
915 #define __SWP_OFFSET_MASK (~0UL >> 11)
917 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
920 offset &= __SWP_OFFSET_MASK;
921 pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) |
922 ((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
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))) })
930 #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
931 #define __swp_entry_to_pte(x) ((pte_t) { (x).val })
934 # define PTE_FILE_MAX_BITS 26
935 #else /* __s390x__ */
936 # define PTE_FILE_MAX_BITS 59
937 #endif /* __s390x__ */
939 #define pte_to_pgoff(__pte) \
940 ((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
942 #define pgoff_to_pte(__off) \
943 ((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
946 #endif /* !__ASSEMBLY__ */
948 #define kern_addr_valid(addr) (1)
950 extern int add_shared_memory(unsigned long start, unsigned long size);
951 extern int remove_shared_memory(unsigned long start, unsigned long size);
954 * No page table caches to initialise
956 #define pgtable_cache_init() do { } while (0)
958 #define __HAVE_ARCH_MEMMAP_INIT
959 extern void memmap_init(unsigned long, int, unsigned long, unsigned long);
961 #include <asm-generic/pgtable.h>
963 #endif /* _S390_PAGE_H */