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 * The vmalloc area will always be on the topmost area of the kernel
108 * mapping. We reserve 96MB (31bit) / 1GB (64bit) for vmalloc,
109 * which should be enough for any sane case.
110 * By putting vmalloc at the top, we maximise the gap between physical
111 * memory and vmalloc to catch misplaced memory accesses. As a side
112 * effect, this also makes sure that 64 bit module code cannot be used
113 * as system call address.
116 #define VMALLOC_START 0x78000000UL
117 #define VMALLOC_END 0x7e000000UL
118 #define VMEM_MAP_END 0x80000000UL
119 #else /* __s390x__ */
120 #define VMALLOC_START 0x3e000000000UL
121 #define VMALLOC_END 0x3e040000000UL
122 #define VMEM_MAP_END 0x40000000000UL
123 #endif /* __s390x__ */
126 * VMEM_MAX_PHYS is the highest physical address that can be added to the 1:1
127 * mapping. This needs to be calculated at compile time since the size of the
128 * VMEM_MAP is static but the size of struct page can change.
130 #define VMEM_MAX_PHYS min(VMALLOC_START, ((VMEM_MAP_END - VMALLOC_END) / \
131 sizeof(struct page) * PAGE_SIZE) & ~((16 << 20) - 1))
132 #define VMEM_MAP ((struct page *) VMALLOC_END)
135 * A 31 bit pagetable entry of S390 has following format:
138 * 00000000001111111111222222222233
139 * 01234567890123456789012345678901
141 * I Page-Invalid Bit: Page is not available for address-translation
142 * P Page-Protection Bit: Store access not possible for page
144 * A 31 bit segmenttable entry of S390 has following format:
145 * | P-table origin | |PTL
147 * 00000000001111111111222222222233
148 * 01234567890123456789012345678901
150 * I Segment-Invalid Bit: Segment is not available for address-translation
151 * C Common-Segment Bit: Segment is not private (PoP 3-30)
152 * PTL Page-Table-Length: Page-table length (PTL+1*16 entries -> up to 256)
154 * The 31 bit segmenttable origin of S390 has following format:
156 * |S-table origin | | STL |
158 * 00000000001111111111222222222233
159 * 01234567890123456789012345678901
161 * X Space-Switch event:
162 * G Segment-Invalid Bit: *
163 * P Private-Space Bit: Segment is not private (PoP 3-30)
164 * S Storage-Alteration:
165 * STL Segment-Table-Length: Segment-table length (STL+1*16 entries -> up to 2048)
167 * A 64 bit pagetable entry of S390 has following format:
169 * 0000000000111111111122222222223333333333444444444455555555556666
170 * 0123456789012345678901234567890123456789012345678901234567890123
172 * I Page-Invalid Bit: Page is not available for address-translation
173 * P Page-Protection Bit: Store access not possible for page
175 * A 64 bit segmenttable entry of S390 has following format:
176 * | P-table origin | TT
177 * 0000000000111111111122222222223333333333444444444455555555556666
178 * 0123456789012345678901234567890123456789012345678901234567890123
180 * I Segment-Invalid Bit: Segment is not available for address-translation
181 * C Common-Segment Bit: Segment is not private (PoP 3-30)
182 * P Page-Protection Bit: Store access not possible for page
185 * A 64 bit region table entry of S390 has following format:
186 * | S-table origin | TF TTTL
187 * 0000000000111111111122222222223333333333444444444455555555556666
188 * 0123456789012345678901234567890123456789012345678901234567890123
190 * I Segment-Invalid Bit: Segment is not available for address-translation
195 * The 64 bit regiontable origin of S390 has following format:
196 * | region table origon | DTTL
197 * 0000000000111111111122222222223333333333444444444455555555556666
198 * 0123456789012345678901234567890123456789012345678901234567890123
200 * X Space-Switch event:
201 * G Segment-Invalid Bit:
202 * P Private-Space Bit:
203 * S Storage-Alteration:
207 * A storage key has the following format:
211 * F : fetch protection bit
216 /* Hardware bits in the page table entry */
217 #define _PAGE_RO 0x200 /* HW read-only bit */
218 #define _PAGE_INVALID 0x400 /* HW invalid bit */
220 /* Software bits in the page table entry */
221 #define _PAGE_SWT 0x001 /* SW pte type bit t */
222 #define _PAGE_SWX 0x002 /* SW pte type bit x */
224 /* Six different types of pages. */
225 #define _PAGE_TYPE_EMPTY 0x400
226 #define _PAGE_TYPE_NONE 0x401
227 #define _PAGE_TYPE_SWAP 0x403
228 #define _PAGE_TYPE_FILE 0x601 /* bit 0x002 is used for offset !! */
229 #define _PAGE_TYPE_RO 0x200
230 #define _PAGE_TYPE_RW 0x000
231 #define _PAGE_TYPE_EX_RO 0x202
232 #define _PAGE_TYPE_EX_RW 0x002
235 * PTE type bits are rather complicated. handle_pte_fault uses pte_present,
236 * pte_none and pte_file to find out the pte type WITHOUT holding the page
237 * table lock. ptep_clear_flush on the other hand uses ptep_clear_flush to
238 * invalidate a given pte. ipte sets the hw invalid bit and clears all tlbs
239 * for the page. The page table entry is set to _PAGE_TYPE_EMPTY afterwards.
240 * This change is done while holding the lock, but the intermediate step
241 * of a previously valid pte with the hw invalid bit set can be observed by
242 * handle_pte_fault. That makes it necessary that all valid pte types with
243 * the hw invalid bit set must be distinguishable from the four pte types
244 * empty, none, swap and file.
247 * _PAGE_TYPE_EMPTY 1000 -> 1000
248 * _PAGE_TYPE_NONE 1001 -> 1001
249 * _PAGE_TYPE_SWAP 1011 -> 1011
250 * _PAGE_TYPE_FILE 11?1 -> 11?1
251 * _PAGE_TYPE_RO 0100 -> 1100
252 * _PAGE_TYPE_RW 0000 -> 1000
253 * _PAGE_TYPE_EX_RO 0110 -> 1110
254 * _PAGE_TYPE_EX_RW 0010 -> 1010
256 * pte_none is true for bits combinations 1000, 1010, 1100, 1110
257 * pte_present is true for bits combinations 0000, 0010, 0100, 0110, 1001
258 * pte_file is true for bits combinations 1101, 1111
259 * swap pte is 1011 and 0001, 0011, 0101, 0111 are invalid.
264 /* Bits in the segment table address-space-control-element */
265 #define _ASCE_SPACE_SWITCH 0x80000000UL /* space switch event */
266 #define _ASCE_ORIGIN_MASK 0x7ffff000UL /* segment table origin */
267 #define _ASCE_PRIVATE_SPACE 0x100 /* private space control */
268 #define _ASCE_ALT_EVENT 0x80 /* storage alteration event control */
269 #define _ASCE_TABLE_LENGTH 0x7f /* 128 x 64 entries = 8k */
271 /* Bits in the segment table entry */
272 #define _SEGMENT_ENTRY_ORIGIN 0x7fffffc0UL /* page table origin */
273 #define _SEGMENT_ENTRY_INV 0x20 /* invalid segment table entry */
274 #define _SEGMENT_ENTRY_COMMON 0x10 /* common segment bit */
275 #define _SEGMENT_ENTRY_PTL 0x0f /* page table length */
277 #define _SEGMENT_ENTRY (_SEGMENT_ENTRY_PTL)
278 #define _SEGMENT_ENTRY_EMPTY (_SEGMENT_ENTRY_INV)
280 #else /* __s390x__ */
282 /* Bits in the segment/region table address-space-control-element */
283 #define _ASCE_ORIGIN ~0xfffUL/* segment table origin */
284 #define _ASCE_PRIVATE_SPACE 0x100 /* private space control */
285 #define _ASCE_ALT_EVENT 0x80 /* storage alteration event control */
286 #define _ASCE_SPACE_SWITCH 0x40 /* space switch event */
287 #define _ASCE_REAL_SPACE 0x20 /* real space control */
288 #define _ASCE_TYPE_MASK 0x0c /* asce table type mask */
289 #define _ASCE_TYPE_REGION1 0x0c /* region first table type */
290 #define _ASCE_TYPE_REGION2 0x08 /* region second table type */
291 #define _ASCE_TYPE_REGION3 0x04 /* region third table type */
292 #define _ASCE_TYPE_SEGMENT 0x00 /* segment table type */
293 #define _ASCE_TABLE_LENGTH 0x03 /* region table length */
295 /* Bits in the region table entry */
296 #define _REGION_ENTRY_ORIGIN ~0xfffUL/* region/segment table origin */
297 #define _REGION_ENTRY_INV 0x20 /* invalid region table entry */
298 #define _REGION_ENTRY_TYPE_MASK 0x0c /* region/segment table type mask */
299 #define _REGION_ENTRY_TYPE_R1 0x0c /* region first table type */
300 #define _REGION_ENTRY_TYPE_R2 0x08 /* region second table type */
301 #define _REGION_ENTRY_TYPE_R3 0x04 /* region third table type */
302 #define _REGION_ENTRY_LENGTH 0x03 /* region third length */
304 #define _REGION1_ENTRY (_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
305 #define _REGION1_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INV)
306 #define _REGION2_ENTRY (_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
307 #define _REGION2_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INV)
308 #define _REGION3_ENTRY (_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
309 #define _REGION3_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INV)
311 /* Bits in the segment table entry */
312 #define _SEGMENT_ENTRY_ORIGIN ~0x7ffUL/* segment table origin */
313 #define _SEGMENT_ENTRY_RO 0x200 /* page protection bit */
314 #define _SEGMENT_ENTRY_INV 0x20 /* invalid segment table entry */
316 #define _SEGMENT_ENTRY (0)
317 #define _SEGMENT_ENTRY_EMPTY (_SEGMENT_ENTRY_INV)
319 #endif /* __s390x__ */
322 * A user page table pointer has the space-switch-event bit, the
323 * private-space-control bit and the storage-alteration-event-control
324 * bit set. A kernel page table pointer doesn't need them.
326 #define _ASCE_USER_BITS (_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
329 /* Bits int the storage key */
330 #define _PAGE_CHANGED 0x02 /* HW changed bit */
331 #define _PAGE_REFERENCED 0x04 /* HW referenced bit */
334 * Page protection definitions.
336 #define PAGE_NONE __pgprot(_PAGE_TYPE_NONE)
337 #define PAGE_RO __pgprot(_PAGE_TYPE_RO)
338 #define PAGE_RW __pgprot(_PAGE_TYPE_RW)
339 #define PAGE_EX_RO __pgprot(_PAGE_TYPE_EX_RO)
340 #define PAGE_EX_RW __pgprot(_PAGE_TYPE_EX_RW)
342 #define PAGE_KERNEL PAGE_RW
343 #define PAGE_COPY PAGE_RO
346 * Dependent on the EXEC_PROTECT option s390 can do execute protection.
347 * Write permission always implies read permission. In theory with a
348 * primary/secondary page table execute only can be implemented but
349 * it would cost an additional bit in the pte to distinguish all the
350 * different pte types. To avoid that execute permission currently
351 * implies read permission as well.
354 #define __P000 PAGE_NONE
355 #define __P001 PAGE_RO
356 #define __P010 PAGE_RO
357 #define __P011 PAGE_RO
358 #define __P100 PAGE_EX_RO
359 #define __P101 PAGE_EX_RO
360 #define __P110 PAGE_EX_RO
361 #define __P111 PAGE_EX_RO
363 #define __S000 PAGE_NONE
364 #define __S001 PAGE_RO
365 #define __S010 PAGE_RW
366 #define __S011 PAGE_RW
367 #define __S100 PAGE_EX_RO
368 #define __S101 PAGE_EX_RO
369 #define __S110 PAGE_EX_RW
370 #define __S111 PAGE_EX_RW
373 # define PxD_SHADOW_SHIFT 1
374 #else /* __s390x__ */
375 # define PxD_SHADOW_SHIFT 2
376 #endif /* __s390x__ */
378 static inline struct page *get_shadow_page(struct page *page)
380 if (s390_noexec && page->index)
381 return virt_to_page((void *)(addr_t) page->index);
385 static inline void *get_shadow_pte(void *table)
387 unsigned long addr, offset;
390 addr = (unsigned long) table;
391 offset = addr & (PAGE_SIZE - 1);
392 page = virt_to_page((void *)(addr ^ offset));
393 return (void *)(addr_t)(page->index ? (page->index | offset) : 0UL);
396 static inline void *get_shadow_table(void *table)
398 unsigned long addr, offset;
401 addr = (unsigned long) table;
402 offset = addr & ((PAGE_SIZE << PxD_SHADOW_SHIFT) - 1);
403 page = virt_to_page((void *)(addr ^ offset));
404 return (void *)(addr_t)(page->index ? (page->index | offset) : 0UL);
408 * Certain architectures need to do special things when PTEs
409 * within a page table are directly modified. Thus, the following
410 * hook is made available.
412 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
413 pte_t *pteptr, pte_t pteval)
415 pte_t *shadow_pte = get_shadow_pte(pteptr);
419 if (!(pte_val(pteval) & _PAGE_INVALID) &&
420 (pte_val(pteval) & _PAGE_SWX))
421 pte_val(*shadow_pte) = pte_val(pteval) | _PAGE_RO;
423 pte_val(*shadow_pte) = _PAGE_TYPE_EMPTY;
428 * pgd/pmd/pte query functions
432 static inline int pgd_present(pgd_t pgd) { return 1; }
433 static inline int pgd_none(pgd_t pgd) { return 0; }
434 static inline int pgd_bad(pgd_t pgd) { return 0; }
436 static inline int pud_present(pud_t pud) { return 1; }
437 static inline int pud_none(pud_t pud) { return 0; }
438 static inline int pud_bad(pud_t pud) { return 0; }
440 #else /* __s390x__ */
442 static inline int pgd_present(pgd_t pgd) { return 1; }
443 static inline int pgd_none(pgd_t pgd) { return 0; }
444 static inline int pgd_bad(pgd_t pgd) { return 0; }
446 static inline int pud_present(pud_t pud)
448 return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
451 static inline int pud_none(pud_t pud)
453 return (pud_val(pud) & _REGION_ENTRY_INV) != 0UL;
456 static inline int pud_bad(pud_t pud)
458 unsigned long mask = ~_REGION_ENTRY_ORIGIN & ~_REGION_ENTRY_INV;
459 return (pud_val(pud) & mask) != _REGION3_ENTRY;
462 #endif /* __s390x__ */
464 static inline int pmd_present(pmd_t pmd)
466 return (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN) != 0UL;
469 static inline int pmd_none(pmd_t pmd)
471 return (pmd_val(pmd) & _SEGMENT_ENTRY_INV) != 0UL;
474 static inline int pmd_bad(pmd_t pmd)
476 unsigned long mask = ~_SEGMENT_ENTRY_ORIGIN & ~_SEGMENT_ENTRY_INV;
477 return (pmd_val(pmd) & mask) != _SEGMENT_ENTRY;
480 static inline int pte_none(pte_t pte)
482 return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT);
485 static inline int pte_present(pte_t pte)
487 unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT | _PAGE_SWX;
488 return (pte_val(pte) & mask) == _PAGE_TYPE_NONE ||
489 (!(pte_val(pte) & _PAGE_INVALID) &&
490 !(pte_val(pte) & _PAGE_SWT));
493 static inline int pte_file(pte_t pte)
495 unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT;
496 return (pte_val(pte) & mask) == _PAGE_TYPE_FILE;
499 #define __HAVE_ARCH_PTE_SAME
500 #define pte_same(a,b) (pte_val(a) == pte_val(b))
503 * query functions pte_write/pte_dirty/pte_young only work if
504 * pte_present() is true. Undefined behaviour if not..
506 static inline int pte_write(pte_t pte)
508 return (pte_val(pte) & _PAGE_RO) == 0;
511 static inline int pte_dirty(pte_t pte)
513 /* A pte is neither clean nor dirty on s/390. The dirty bit
514 * is in the storage key. See page_test_and_clear_dirty for
520 static inline int pte_young(pte_t pte)
522 /* A pte is neither young nor old on s/390. The young bit
523 * is in the storage key. See page_test_and_clear_young for
530 * pgd/pmd/pte modification functions
535 #define pgd_clear(pgd) do { } while (0)
536 #define pud_clear(pud) do { } while (0)
538 static inline void pmd_clear_kernel(pmd_t * pmdp)
540 pmd_val(pmdp[0]) = _SEGMENT_ENTRY_EMPTY;
541 pmd_val(pmdp[1]) = _SEGMENT_ENTRY_EMPTY;
542 pmd_val(pmdp[2]) = _SEGMENT_ENTRY_EMPTY;
543 pmd_val(pmdp[3]) = _SEGMENT_ENTRY_EMPTY;
546 #else /* __s390x__ */
548 #define pgd_clear(pgd) do { } while (0)
550 static inline void pud_clear_kernel(pud_t *pud)
552 pud_val(*pud) = _REGION3_ENTRY_EMPTY;
555 static inline void pud_clear(pud_t * pud)
557 pud_t *shadow = get_shadow_table(pud);
559 pud_clear_kernel(pud);
561 pud_clear_kernel(shadow);
564 static inline void pmd_clear_kernel(pmd_t * pmdp)
566 pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY;
567 pmd_val1(*pmdp) = _SEGMENT_ENTRY_EMPTY;
570 #endif /* __s390x__ */
572 static inline void pmd_clear(pmd_t * pmdp)
574 pmd_t *shadow_pmd = get_shadow_table(pmdp);
576 pmd_clear_kernel(pmdp);
578 pmd_clear_kernel(shadow_pmd);
581 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
583 pte_t *shadow_pte = get_shadow_pte(ptep);
585 pte_val(*ptep) = _PAGE_TYPE_EMPTY;
587 pte_val(*shadow_pte) = _PAGE_TYPE_EMPTY;
591 * The following pte modification functions only work if
592 * pte_present() is true. Undefined behaviour if not..
594 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
596 pte_val(pte) &= PAGE_MASK;
597 pte_val(pte) |= pgprot_val(newprot);
601 static inline pte_t pte_wrprotect(pte_t pte)
603 /* Do not clobber _PAGE_TYPE_NONE pages! */
604 if (!(pte_val(pte) & _PAGE_INVALID))
605 pte_val(pte) |= _PAGE_RO;
609 static inline pte_t pte_mkwrite(pte_t pte)
611 pte_val(pte) &= ~_PAGE_RO;
615 static inline pte_t pte_mkclean(pte_t pte)
617 /* The only user of pte_mkclean is the fork() code.
618 We must *not* clear the *physical* page dirty bit
619 just because fork() wants to clear the dirty bit in
620 *one* of the page's mappings. So we just do nothing. */
624 static inline pte_t pte_mkdirty(pte_t pte)
626 /* We do not explicitly set the dirty bit because the
627 * sske instruction is slow. It is faster to let the
628 * next instruction set the dirty bit.
633 static inline pte_t pte_mkold(pte_t pte)
635 /* S/390 doesn't keep its dirty/referenced bit in the pte.
636 * There is no point in clearing the real referenced bit.
641 static inline pte_t pte_mkyoung(pte_t pte)
643 /* S/390 doesn't keep its dirty/referenced bit in the pte.
644 * There is no point in setting the real referenced bit.
649 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
650 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
651 unsigned long addr, pte_t *ptep)
656 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
657 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
658 unsigned long address, pte_t *ptep)
660 /* No need to flush TLB; bits are in storage key */
664 static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
666 if (!(pte_val(*ptep) & _PAGE_INVALID)) {
668 /* S390 has 1mb segments, we are emulating 4MB segments */
669 pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00);
671 /* ipte in zarch mode can do the math */
676 : "=m" (*ptep) : "m" (*ptep),
677 "a" (pto), "a" (address));
679 pte_val(*ptep) = _PAGE_TYPE_EMPTY;
682 static inline void ptep_invalidate(unsigned long address, pte_t *ptep)
684 __ptep_ipte(address, ptep);
685 ptep = get_shadow_pte(ptep);
687 __ptep_ipte(address, ptep);
691 * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
692 * both clear the TLB for the unmapped pte. The reason is that
693 * ptep_get_and_clear is used in common code (e.g. change_pte_range)
694 * to modify an active pte. The sequence is
695 * 1) ptep_get_and_clear
698 * On s390 the tlb needs to get flushed with the modification of the pte
699 * if the pte is active. The only way how this can be implemented is to
700 * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
703 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
704 #define ptep_get_and_clear(__mm, __address, __ptep) \
706 pte_t __pte = *(__ptep); \
707 if (atomic_read(&(__mm)->mm_users) > 1 || \
708 (__mm) != current->active_mm) \
709 ptep_invalidate(__address, __ptep); \
711 pte_clear((__mm), (__address), (__ptep)); \
715 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
716 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
717 unsigned long address, pte_t *ptep)
720 ptep_invalidate(address, ptep);
725 * The batched pte unmap code uses ptep_get_and_clear_full to clear the
726 * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
727 * tlbs of an mm if it can guarantee that the ptes of the mm_struct
728 * cannot be accessed while the batched unmap is running. In this case
729 * full==1 and a simple pte_clear is enough. See tlb.h.
731 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
732 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
734 pte_t *ptep, int full)
739 pte_clear(mm, addr, ptep);
741 ptep_invalidate(addr, ptep);
745 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
746 #define ptep_set_wrprotect(__mm, __addr, __ptep) \
748 pte_t __pte = *(__ptep); \
749 if (pte_write(__pte)) { \
750 if (atomic_read(&(__mm)->mm_users) > 1 || \
751 (__mm) != current->active_mm) \
752 ptep_invalidate(__addr, __ptep); \
753 set_pte_at(__mm, __addr, __ptep, pte_wrprotect(__pte)); \
757 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
758 #define ptep_set_access_flags(__vma, __addr, __ptep, __entry, __dirty) \
760 int __changed = !pte_same(*(__ptep), __entry); \
762 ptep_invalidate(__addr, __ptep); \
763 set_pte_at((__vma)->vm_mm, __addr, __ptep, __entry); \
769 * Test and clear dirty bit in storage key.
770 * We can't clear the changed bit atomically. This is a potential
771 * race against modification of the referenced bit. This function
772 * should therefore only be called if it is not mapped in any
775 #define __HAVE_ARCH_PAGE_TEST_DIRTY
776 static inline int page_test_dirty(struct page *page)
778 return (page_get_storage_key(page_to_phys(page)) & _PAGE_CHANGED) != 0;
781 #define __HAVE_ARCH_PAGE_CLEAR_DIRTY
782 static inline void page_clear_dirty(struct page *page)
784 page_set_storage_key(page_to_phys(page), PAGE_DEFAULT_KEY);
788 * Test and clear referenced bit in storage key.
790 #define __HAVE_ARCH_PAGE_TEST_AND_CLEAR_YOUNG
791 static inline int page_test_and_clear_young(struct page *page)
793 unsigned long physpage = page_to_phys(page);
800 : "=d" (ccode) : "a" (physpage) : "cc" );
805 * Conversion functions: convert a page and protection to a page entry,
806 * and a page entry and page directory to the page they refer to.
808 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
811 pte_val(__pte) = physpage + pgprot_val(pgprot);
815 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
817 unsigned long physpage = page_to_phys(page);
819 return mk_pte_phys(physpage, pgprot);
822 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
823 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
824 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
825 #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
827 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
828 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
832 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
833 #define pud_deref(pmd) ({ BUG(); 0UL; })
834 #define pgd_deref(pmd) ({ BUG(); 0UL; })
836 #define pud_offset(pgd, address) ((pud_t *) pgd)
837 #define pmd_offset(pud, address) ((pmd_t *) pud + pmd_index(address))
839 #else /* __s390x__ */
841 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
842 #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)
843 #define pgd_deref(pgd) ({ BUG(); 0UL; })
845 #define pud_offset(pgd, address) ((pud_t *) pgd)
847 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
849 pmd_t *pmd = (pmd_t *) pud_deref(*pud);
850 return pmd + pmd_index(address);
853 #endif /* __s390x__ */
855 #define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
856 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
857 #define pte_page(x) pfn_to_page(pte_pfn(x))
859 #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
861 /* Find an entry in the lowest level page table.. */
862 #define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr))
863 #define pte_offset_kernel(pmd, address) pte_offset(pmd,address)
864 #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
865 #define pte_offset_map_nested(pmd, address) pte_offset_kernel(pmd, address)
866 #define pte_unmap(pte) do { } while (0)
867 #define pte_unmap_nested(pte) do { } while (0)
870 * 31 bit swap entry format:
871 * A page-table entry has some bits we have to treat in a special way.
872 * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
873 * exception will occur instead of a page translation exception. The
874 * specifiation exception has the bad habit not to store necessary
875 * information in the lowcore.
876 * Bit 21 and bit 22 are the page invalid bit and the page protection
877 * bit. We set both to indicate a swapped page.
878 * Bit 30 and 31 are used to distinguish the different page types. For
879 * a swapped page these bits need to be zero.
880 * This leaves the bits 1-19 and bits 24-29 to store type and offset.
881 * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
882 * plus 24 for the offset.
883 * 0| offset |0110|o|type |00|
884 * 0 0000000001111111111 2222 2 22222 33
885 * 0 1234567890123456789 0123 4 56789 01
887 * 64 bit swap entry format:
888 * A page-table entry has some bits we have to treat in a special way.
889 * Bits 52 and bit 55 have to be zero, otherwise an specification
890 * exception will occur instead of a page translation exception. The
891 * specifiation exception has the bad habit not to store necessary
892 * information in the lowcore.
893 * Bit 53 and bit 54 are the page invalid bit and the page protection
894 * bit. We set both to indicate a swapped page.
895 * Bit 62 and 63 are used to distinguish the different page types. For
896 * a swapped page these bits need to be zero.
897 * This leaves the bits 0-51 and bits 56-61 to store type and offset.
898 * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
899 * plus 56 for the offset.
900 * | offset |0110|o|type |00|
901 * 0000000000111111111122222222223333333333444444444455 5555 5 55566 66
902 * 0123456789012345678901234567890123456789012345678901 2345 6 78901 23
905 #define __SWP_OFFSET_MASK (~0UL >> 12)
907 #define __SWP_OFFSET_MASK (~0UL >> 11)
909 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
912 offset &= __SWP_OFFSET_MASK;
913 pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) |
914 ((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
918 #define __swp_type(entry) (((entry).val >> 2) & 0x1f)
919 #define __swp_offset(entry) (((entry).val >> 11) | (((entry).val >> 7) & 1))
920 #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
922 #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
923 #define __swp_entry_to_pte(x) ((pte_t) { (x).val })
926 # define PTE_FILE_MAX_BITS 26
927 #else /* __s390x__ */
928 # define PTE_FILE_MAX_BITS 59
929 #endif /* __s390x__ */
931 #define pte_to_pgoff(__pte) \
932 ((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
934 #define pgoff_to_pte(__off) \
935 ((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
938 #endif /* !__ASSEMBLY__ */
940 #define kern_addr_valid(addr) (1)
942 extern int add_shared_memory(unsigned long start, unsigned long size);
943 extern int remove_shared_memory(unsigned long start, unsigned long size);
946 * No page table caches to initialise
948 #define pgtable_cache_init() do { } while (0)
950 #define __HAVE_ARCH_MEMMAP_INIT
951 extern void memmap_init(unsigned long, int, unsigned long, unsigned long);
953 #include <asm-generic/pgtable.h>
955 #endif /* _S390_PAGE_H */