3 * Copyright IBM Corp. 1999, 2000
8 * Derived from "include/asm-i386/pgtable.h"
11 #ifndef _ASM_S390_PGTABLE_H
12 #define _ASM_S390_PGTABLE_H
15 * The Linux memory management assumes a three-level page table setup. For
16 * s390 31 bit we "fold" the mid level into the top-level page table, so
17 * that we physically have the same two-level page table as the s390 mmu
18 * expects in 31 bit mode. For s390 64 bit we use three of the five levels
19 * the hardware provides (region first and region second tables are not
22 * The "pgd_xxx()" functions are trivial for a folded two-level
23 * setup: the pgd is never bad, and a pmd always exists (as it's folded
26 * This file contains the functions and defines necessary to modify and use
27 * the S390 page table tree.
30 #include <linux/sched.h>
31 #include <linux/mm_types.h>
32 #include <linux/page-flags.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, ptep) do { } while (0)
45 #define update_mmu_cache_pmd(vma, address, ptep) do { } while (0)
48 * ZERO_PAGE is a global shared page that is always zero; used
49 * for zero-mapped memory areas etc..
52 extern unsigned long empty_zero_page;
53 extern unsigned long zero_page_mask;
55 #define ZERO_PAGE(vaddr) \
56 (virt_to_page((void *)(empty_zero_page + \
57 (((unsigned long)(vaddr)) &zero_page_mask))))
58 #define __HAVE_COLOR_ZERO_PAGE
60 /* TODO: s390 cannot support io_remap_pfn_range... */
61 #endif /* !__ASSEMBLY__ */
64 * PMD_SHIFT determines the size of the area a second-level page
66 * PGDIR_SHIFT determines what a third-level page table entry can map
71 # define PGDIR_SHIFT 20
72 #else /* CONFIG_64BIT */
75 # define PGDIR_SHIFT 42
76 #endif /* CONFIG_64BIT */
78 #define PMD_SIZE (1UL << PMD_SHIFT)
79 #define PMD_MASK (~(PMD_SIZE-1))
80 #define PUD_SIZE (1UL << PUD_SHIFT)
81 #define PUD_MASK (~(PUD_SIZE-1))
82 #define PGDIR_SIZE (1UL << PGDIR_SHIFT)
83 #define PGDIR_MASK (~(PGDIR_SIZE-1))
86 * entries per page directory level: the S390 is two-level, so
87 * we don't really have any PMD directory physically.
88 * for S390 segment-table entries are combined to one PGD
89 * that leads to 1024 pte per pgd
91 #define PTRS_PER_PTE 256
93 #define PTRS_PER_PMD 1
94 #define PTRS_PER_PUD 1
95 #else /* CONFIG_64BIT */
96 #define PTRS_PER_PMD 2048
97 #define PTRS_PER_PUD 2048
98 #endif /* CONFIG_64BIT */
99 #define PTRS_PER_PGD 2048
101 #define FIRST_USER_ADDRESS 0
103 #define pte_ERROR(e) \
104 printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e))
105 #define pmd_ERROR(e) \
106 printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e))
107 #define pud_ERROR(e) \
108 printk("%s:%d: bad pud %p.\n", __FILE__, __LINE__, (void *) pud_val(e))
109 #define pgd_ERROR(e) \
110 printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
114 * The vmalloc and module area will always be on the topmost area of the kernel
115 * mapping. We reserve 96MB (31bit) / 128GB (64bit) for vmalloc and modules.
116 * On 64 bit kernels we have a 2GB area at the top of the vmalloc area where
117 * modules will reside. That makes sure that inter module branches always
118 * happen without trampolines and in addition the placement within a 2GB frame
119 * is branch prediction unit friendly.
121 extern unsigned long VMALLOC_START;
122 extern unsigned long VMALLOC_END;
123 extern struct page *vmemmap;
125 #define VMEM_MAX_PHYS ((unsigned long) vmemmap)
128 extern unsigned long MODULES_VADDR;
129 extern unsigned long MODULES_END;
130 #define MODULES_VADDR MODULES_VADDR
131 #define MODULES_END MODULES_END
132 #define MODULES_LEN (1UL << 31)
136 * A 31 bit pagetable entry of S390 has following format:
139 * 00000000001111111111222222222233
140 * 01234567890123456789012345678901
142 * I Page-Invalid Bit: Page is not available for address-translation
143 * P Page-Protection Bit: Store access not possible for page
145 * A 31 bit segmenttable entry of S390 has following format:
146 * | P-table origin | |PTL
148 * 00000000001111111111222222222233
149 * 01234567890123456789012345678901
151 * I Segment-Invalid Bit: Segment is not available for address-translation
152 * C Common-Segment Bit: Segment is not private (PoP 3-30)
153 * PTL Page-Table-Length: Page-table length (PTL+1*16 entries -> up to 256)
155 * The 31 bit segmenttable origin of S390 has following format:
157 * |S-table origin | | STL |
159 * 00000000001111111111222222222233
160 * 01234567890123456789012345678901
162 * X Space-Switch event:
163 * G Segment-Invalid Bit: *
164 * P Private-Space Bit: Segment is not private (PoP 3-30)
165 * S Storage-Alteration:
166 * STL Segment-Table-Length: Segment-table length (STL+1*16 entries -> up to 2048)
168 * A 64 bit pagetable entry of S390 has following format:
170 * 0000000000111111111122222222223333333333444444444455555555556666
171 * 0123456789012345678901234567890123456789012345678901234567890123
173 * I Page-Invalid Bit: Page is not available for address-translation
174 * P Page-Protection Bit: Store access not possible for page
175 * C Change-bit override: HW is not required to set change bit
177 * A 64 bit segmenttable entry of S390 has following format:
178 * | P-table origin | TT
179 * 0000000000111111111122222222223333333333444444444455555555556666
180 * 0123456789012345678901234567890123456789012345678901234567890123
182 * I Segment-Invalid Bit: Segment is not available for address-translation
183 * C Common-Segment Bit: Segment is not private (PoP 3-30)
184 * P Page-Protection Bit: Store access not possible for page
187 * A 64 bit region table entry of S390 has following format:
188 * | S-table origin | TF TTTL
189 * 0000000000111111111122222222223333333333444444444455555555556666
190 * 0123456789012345678901234567890123456789012345678901234567890123
192 * I Segment-Invalid Bit: Segment is not available for address-translation
197 * The 64 bit regiontable origin of S390 has following format:
198 * | region table origon | DTTL
199 * 0000000000111111111122222222223333333333444444444455555555556666
200 * 0123456789012345678901234567890123456789012345678901234567890123
202 * X Space-Switch event:
203 * G Segment-Invalid Bit:
204 * P Private-Space Bit:
205 * S Storage-Alteration:
209 * A storage key has the following format:
213 * F : fetch protection bit
218 /* Hardware bits in the page table entry */
219 #define _PAGE_CO 0x100 /* HW Change-bit override */
220 #define _PAGE_PROTECT 0x200 /* HW read-only bit */
221 #define _PAGE_INVALID 0x400 /* HW invalid bit */
222 #define _PAGE_LARGE 0x800 /* Bit to mark a large pte */
224 /* Software bits in the page table entry */
225 #define _PAGE_PRESENT 0x001 /* SW pte present bit */
226 #define _PAGE_TYPE 0x002 /* SW pte type bit */
227 #define _PAGE_YOUNG 0x004 /* SW pte young bit */
228 #define _PAGE_DIRTY 0x008 /* SW pte dirty bit */
229 #define _PAGE_READ 0x010 /* SW pte read bit */
230 #define _PAGE_WRITE 0x020 /* SW pte write bit */
231 #define _PAGE_SPECIAL 0x040 /* SW associated with special page */
232 #define _PAGE_UNUSED 0x080 /* SW bit for pgste usage state */
233 #define __HAVE_ARCH_PTE_SPECIAL
235 /* Set of bits not changed in pte_modify */
236 #define _PAGE_CHG_MASK (PAGE_MASK | _PAGE_SPECIAL | _PAGE_CO | \
237 _PAGE_DIRTY | _PAGE_YOUNG)
240 * handle_pte_fault uses pte_present, pte_none and pte_file to find out the
241 * pte type WITHOUT holding the page table lock. The _PAGE_PRESENT bit
242 * is used to distinguish present from not-present ptes. It is changed only
243 * with the page table lock held.
245 * The following table gives the different possible bit combinations for
246 * the pte hardware and software bits in the last 12 bits of a pte:
255 * prot-none, clean, old .11...000001
256 * prot-none, clean, young .11...000101
257 * prot-none, dirty, old .10...001001
258 * prot-none, dirty, young .10...001101
259 * read-only, clean, old .11...010001
260 * read-only, clean, young .01...010101
261 * read-only, dirty, old .11...011001
262 * read-only, dirty, young .01...011101
263 * read-write, clean, old .11...110001
264 * read-write, clean, young .01...110101
265 * read-write, dirty, old .10...111001
266 * read-write, dirty, young .00...111101
268 * pte_present is true for the bit pattern .xx...xxxxx1, (pte & 0x001) == 0x001
269 * pte_none is true for the bit pattern .10...xxxx00, (pte & 0x603) == 0x400
270 * pte_file is true for the bit pattern .11...xxxxx0, (pte & 0x601) == 0x600
271 * pte_swap is true for the bit pattern .10...xxxx10, (pte & 0x603) == 0x402
276 /* Bits in the segment table address-space-control-element */
277 #define _ASCE_SPACE_SWITCH 0x80000000UL /* space switch event */
278 #define _ASCE_ORIGIN_MASK 0x7ffff000UL /* segment table origin */
279 #define _ASCE_PRIVATE_SPACE 0x100 /* private space control */
280 #define _ASCE_ALT_EVENT 0x80 /* storage alteration event control */
281 #define _ASCE_TABLE_LENGTH 0x7f /* 128 x 64 entries = 8k */
283 /* Bits in the segment table entry */
284 #define _SEGMENT_ENTRY_BITS 0x7fffffffUL /* Valid segment table bits */
285 #define _SEGMENT_ENTRY_ORIGIN 0x7fffffc0UL /* page table origin */
286 #define _SEGMENT_ENTRY_PROTECT 0x200 /* page protection bit */
287 #define _SEGMENT_ENTRY_INVALID 0x20 /* invalid segment table entry */
288 #define _SEGMENT_ENTRY_COMMON 0x10 /* common segment bit */
289 #define _SEGMENT_ENTRY_PTL 0x0f /* page table length */
290 #define _SEGMENT_ENTRY_NONE _SEGMENT_ENTRY_PROTECT
292 #define _SEGMENT_ENTRY (_SEGMENT_ENTRY_PTL)
293 #define _SEGMENT_ENTRY_EMPTY (_SEGMENT_ENTRY_INVALID)
296 * Segment table entry encoding (I = invalid, R = read-only bit):
298 * prot-none ..1...1.....
299 * read-only ..1...0.....
300 * read-write ..0...0.....
304 /* Page status table bits for virtualization */
305 #define PGSTE_ACC_BITS 0xf0000000UL
306 #define PGSTE_FP_BIT 0x08000000UL
307 #define PGSTE_PCL_BIT 0x00800000UL
308 #define PGSTE_HR_BIT 0x00400000UL
309 #define PGSTE_HC_BIT 0x00200000UL
310 #define PGSTE_GR_BIT 0x00040000UL
311 #define PGSTE_GC_BIT 0x00020000UL
312 #define PGSTE_IN_BIT 0x00008000UL /* IPTE notify bit */
314 #else /* CONFIG_64BIT */
316 /* Bits in the segment/region table address-space-control-element */
317 #define _ASCE_ORIGIN ~0xfffUL/* segment table origin */
318 #define _ASCE_PRIVATE_SPACE 0x100 /* private space control */
319 #define _ASCE_ALT_EVENT 0x80 /* storage alteration event control */
320 #define _ASCE_SPACE_SWITCH 0x40 /* space switch event */
321 #define _ASCE_REAL_SPACE 0x20 /* real space control */
322 #define _ASCE_TYPE_MASK 0x0c /* asce table type mask */
323 #define _ASCE_TYPE_REGION1 0x0c /* region first table type */
324 #define _ASCE_TYPE_REGION2 0x08 /* region second table type */
325 #define _ASCE_TYPE_REGION3 0x04 /* region third table type */
326 #define _ASCE_TYPE_SEGMENT 0x00 /* segment table type */
327 #define _ASCE_TABLE_LENGTH 0x03 /* region table length */
329 /* Bits in the region table entry */
330 #define _REGION_ENTRY_ORIGIN ~0xfffUL/* region/segment table origin */
331 #define _REGION_ENTRY_PROTECT 0x200 /* region protection bit */
332 #define _REGION_ENTRY_INVALID 0x20 /* invalid region table entry */
333 #define _REGION_ENTRY_TYPE_MASK 0x0c /* region/segment table type mask */
334 #define _REGION_ENTRY_TYPE_R1 0x0c /* region first table type */
335 #define _REGION_ENTRY_TYPE_R2 0x08 /* region second table type */
336 #define _REGION_ENTRY_TYPE_R3 0x04 /* region third table type */
337 #define _REGION_ENTRY_LENGTH 0x03 /* region third length */
339 #define _REGION1_ENTRY (_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
340 #define _REGION1_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INVALID)
341 #define _REGION2_ENTRY (_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
342 #define _REGION2_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INVALID)
343 #define _REGION3_ENTRY (_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
344 #define _REGION3_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INVALID)
346 #define _REGION3_ENTRY_LARGE 0x400 /* RTTE-format control, large page */
347 #define _REGION3_ENTRY_RO 0x200 /* page protection bit */
348 #define _REGION3_ENTRY_CO 0x100 /* change-recording override */
350 /* Bits in the segment table entry */
351 #define _SEGMENT_ENTRY_BITS 0xfffffffffffffe33UL
352 #define _SEGMENT_ENTRY_BITS_LARGE 0xfffffffffff1ff33UL
353 #define _SEGMENT_ENTRY_ORIGIN_LARGE ~0xfffffUL /* large page address */
354 #define _SEGMENT_ENTRY_ORIGIN ~0x7ffUL/* segment table origin */
355 #define _SEGMENT_ENTRY_PROTECT 0x200 /* page protection bit */
356 #define _SEGMENT_ENTRY_INVALID 0x20 /* invalid segment table entry */
358 #define _SEGMENT_ENTRY (0)
359 #define _SEGMENT_ENTRY_EMPTY (_SEGMENT_ENTRY_INVALID)
361 #define _SEGMENT_ENTRY_LARGE 0x400 /* STE-format control, large page */
362 #define _SEGMENT_ENTRY_CO 0x100 /* change-recording override */
363 #define _SEGMENT_ENTRY_SPLIT 0x001 /* THP splitting bit */
364 #define _SEGMENT_ENTRY_YOUNG 0x002 /* SW segment young bit */
365 #define _SEGMENT_ENTRY_NONE _SEGMENT_ENTRY_YOUNG
368 * Segment table entry encoding (R = read-only, I = invalid, y = young bit):
370 * prot-none, old ..0...1...1.
371 * prot-none, young ..1...1...1.
372 * read-only, old ..1...1...0.
373 * read-only, young ..1...0...1.
374 * read-write, old ..0...1...0.
375 * read-write, young ..0...0...1.
376 * The segment table origin is used to distinguish empty (origin==0) from
377 * read-write, old segment table entries (origin!=0)
380 #define _SEGMENT_ENTRY_SPLIT_BIT 0 /* THP splitting bit number */
382 /* Set of bits not changed in pmd_modify */
383 #define _SEGMENT_CHG_MASK (_SEGMENT_ENTRY_ORIGIN | _SEGMENT_ENTRY_LARGE \
384 | _SEGMENT_ENTRY_SPLIT | _SEGMENT_ENTRY_CO)
386 /* Page status table bits for virtualization */
387 #define PGSTE_ACC_BITS 0xf000000000000000UL
388 #define PGSTE_FP_BIT 0x0800000000000000UL
389 #define PGSTE_PCL_BIT 0x0080000000000000UL
390 #define PGSTE_HR_BIT 0x0040000000000000UL
391 #define PGSTE_HC_BIT 0x0020000000000000UL
392 #define PGSTE_GR_BIT 0x0004000000000000UL
393 #define PGSTE_GC_BIT 0x0002000000000000UL
394 #define PGSTE_IN_BIT 0x0000800000000000UL /* IPTE notify bit */
396 #endif /* CONFIG_64BIT */
398 /* Guest Page State used for virtualization */
399 #define _PGSTE_GPS_ZERO 0x0000000080000000UL
400 #define _PGSTE_GPS_USAGE_MASK 0x0000000003000000UL
401 #define _PGSTE_GPS_USAGE_STABLE 0x0000000000000000UL
402 #define _PGSTE_GPS_USAGE_UNUSED 0x0000000001000000UL
405 * A user page table pointer has the space-switch-event bit, the
406 * private-space-control bit and the storage-alteration-event-control
407 * bit set. A kernel page table pointer doesn't need them.
409 #define _ASCE_USER_BITS (_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
413 * Page protection definitions.
415 #define PAGE_NONE __pgprot(_PAGE_PRESENT | _PAGE_INVALID)
416 #define PAGE_READ __pgprot(_PAGE_PRESENT | _PAGE_READ | \
417 _PAGE_INVALID | _PAGE_PROTECT)
418 #define PAGE_WRITE __pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
419 _PAGE_INVALID | _PAGE_PROTECT)
421 #define PAGE_SHARED __pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
422 _PAGE_YOUNG | _PAGE_DIRTY)
423 #define PAGE_KERNEL __pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
424 _PAGE_YOUNG | _PAGE_DIRTY)
425 #define PAGE_KERNEL_RO __pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_YOUNG | \
429 * On s390 the page table entry has an invalid bit and a read-only bit.
430 * Read permission implies execute permission and write permission
431 * implies read permission.
434 #define __P000 PAGE_NONE
435 #define __P001 PAGE_READ
436 #define __P010 PAGE_READ
437 #define __P011 PAGE_READ
438 #define __P100 PAGE_READ
439 #define __P101 PAGE_READ
440 #define __P110 PAGE_READ
441 #define __P111 PAGE_READ
443 #define __S000 PAGE_NONE
444 #define __S001 PAGE_READ
445 #define __S010 PAGE_WRITE
446 #define __S011 PAGE_WRITE
447 #define __S100 PAGE_READ
448 #define __S101 PAGE_READ
449 #define __S110 PAGE_WRITE
450 #define __S111 PAGE_WRITE
453 * Segment entry (large page) protection definitions.
455 #define SEGMENT_NONE __pgprot(_SEGMENT_ENTRY_INVALID | \
457 #define SEGMENT_READ __pgprot(_SEGMENT_ENTRY_INVALID | \
458 _SEGMENT_ENTRY_PROTECT)
459 #define SEGMENT_WRITE __pgprot(_SEGMENT_ENTRY_INVALID)
461 static inline int mm_has_pgste(struct mm_struct *mm)
464 if (unlikely(mm->context.has_pgste))
470 * pgd/pmd/pte query functions
474 static inline int pgd_present(pgd_t pgd) { return 1; }
475 static inline int pgd_none(pgd_t pgd) { return 0; }
476 static inline int pgd_bad(pgd_t pgd) { return 0; }
478 static inline int pud_present(pud_t pud) { return 1; }
479 static inline int pud_none(pud_t pud) { return 0; }
480 static inline int pud_large(pud_t pud) { return 0; }
481 static inline int pud_bad(pud_t pud) { return 0; }
483 #else /* CONFIG_64BIT */
485 static inline int pgd_present(pgd_t pgd)
487 if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
489 return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
492 static inline int pgd_none(pgd_t pgd)
494 if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
496 return (pgd_val(pgd) & _REGION_ENTRY_INVALID) != 0UL;
499 static inline int pgd_bad(pgd_t pgd)
502 * With dynamic page table levels the pgd can be a region table
503 * entry or a segment table entry. Check for the bit that are
504 * invalid for either table entry.
507 ~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INVALID &
508 ~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
509 return (pgd_val(pgd) & mask) != 0;
512 static inline int pud_present(pud_t pud)
514 if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
516 return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
519 static inline int pud_none(pud_t pud)
521 if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
523 return (pud_val(pud) & _REGION_ENTRY_INVALID) != 0UL;
526 static inline int pud_large(pud_t pud)
528 if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) != _REGION_ENTRY_TYPE_R3)
530 return !!(pud_val(pud) & _REGION3_ENTRY_LARGE);
533 static inline int pud_bad(pud_t pud)
536 * With dynamic page table levels the pud can be a region table
537 * entry or a segment table entry. Check for the bit that are
538 * invalid for either table entry.
541 ~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INVALID &
542 ~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
543 return (pud_val(pud) & mask) != 0;
546 #endif /* CONFIG_64BIT */
548 static inline int pmd_present(pmd_t pmd)
550 return pmd_val(pmd) != _SEGMENT_ENTRY_INVALID;
553 static inline int pmd_none(pmd_t pmd)
555 return pmd_val(pmd) == _SEGMENT_ENTRY_INVALID;
558 static inline int pmd_large(pmd_t pmd)
561 return (pmd_val(pmd) & _SEGMENT_ENTRY_LARGE) != 0;
567 static inline int pmd_prot_none(pmd_t pmd)
569 return (pmd_val(pmd) & _SEGMENT_ENTRY_INVALID) &&
570 (pmd_val(pmd) & _SEGMENT_ENTRY_NONE);
573 static inline int pmd_bad(pmd_t pmd)
577 return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS_LARGE) != 0;
579 return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS) != 0;
582 #define __HAVE_ARCH_PMDP_SPLITTING_FLUSH
583 extern void pmdp_splitting_flush(struct vm_area_struct *vma,
584 unsigned long addr, pmd_t *pmdp);
586 #define __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
587 extern int pmdp_set_access_flags(struct vm_area_struct *vma,
588 unsigned long address, pmd_t *pmdp,
589 pmd_t entry, int dirty);
591 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
592 extern int pmdp_clear_flush_young(struct vm_area_struct *vma,
593 unsigned long address, pmd_t *pmdp);
595 #define __HAVE_ARCH_PMD_WRITE
596 static inline int pmd_write(pmd_t pmd)
598 if (pmd_prot_none(pmd))
600 return (pmd_val(pmd) & _SEGMENT_ENTRY_PROTECT) == 0;
603 static inline int pmd_young(pmd_t pmd)
607 if (pmd_prot_none(pmd))
608 young = (pmd_val(pmd) & _SEGMENT_ENTRY_PROTECT) != 0;
610 young = (pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG) != 0;
615 static inline int pte_present(pte_t pte)
617 /* Bit pattern: (pte & 0x001) == 0x001 */
618 return (pte_val(pte) & _PAGE_PRESENT) != 0;
621 static inline int pte_none(pte_t pte)
623 /* Bit pattern: pte == 0x400 */
624 return pte_val(pte) == _PAGE_INVALID;
627 static inline int pte_swap(pte_t pte)
629 /* Bit pattern: (pte & 0x603) == 0x402 */
630 return (pte_val(pte) & (_PAGE_INVALID | _PAGE_PROTECT |
631 _PAGE_TYPE | _PAGE_PRESENT))
632 == (_PAGE_INVALID | _PAGE_TYPE);
635 static inline int pte_file(pte_t pte)
637 /* Bit pattern: (pte & 0x601) == 0x600 */
638 return (pte_val(pte) & (_PAGE_INVALID | _PAGE_PROTECT | _PAGE_PRESENT))
639 == (_PAGE_INVALID | _PAGE_PROTECT);
642 static inline int pte_special(pte_t pte)
644 return (pte_val(pte) & _PAGE_SPECIAL);
647 #define __HAVE_ARCH_PTE_SAME
648 static inline int pte_same(pte_t a, pte_t b)
650 return pte_val(a) == pte_val(b);
653 static inline pgste_t pgste_get_lock(pte_t *ptep)
655 unsigned long new = 0;
663 " nihh %0,0xff7f\n" /* clear PCL bit in old */
664 " oihh %1,0x0080\n" /* set PCL bit in new */
667 : "=&d" (old), "=&d" (new), "=Q" (ptep[PTRS_PER_PTE])
668 : "Q" (ptep[PTRS_PER_PTE]) : "cc", "memory");
673 static inline void pgste_set_unlock(pte_t *ptep, pgste_t pgste)
677 " nihh %1,0xff7f\n" /* clear PCL bit */
679 : "=Q" (ptep[PTRS_PER_PTE])
680 : "d" (pgste_val(pgste)), "Q" (ptep[PTRS_PER_PTE])
686 static inline pgste_t pgste_get(pte_t *ptep)
688 unsigned long pgste = 0;
690 pgste = *(unsigned long *)(ptep + PTRS_PER_PTE);
692 return __pgste(pgste);
695 static inline void pgste_set(pte_t *ptep, pgste_t pgste)
698 *(pgste_t *)(ptep + PTRS_PER_PTE) = pgste;
702 static inline pgste_t pgste_update_all(pte_t *ptep, pgste_t pgste)
705 unsigned long address, bits, skey;
707 if (pte_val(*ptep) & _PAGE_INVALID)
709 address = pte_val(*ptep) & PAGE_MASK;
710 skey = (unsigned long) page_get_storage_key(address);
711 bits = skey & (_PAGE_CHANGED | _PAGE_REFERENCED);
712 if (!(pgste_val(pgste) & PGSTE_HC_BIT) && (bits & _PAGE_CHANGED)) {
713 /* Transfer dirty + referenced bit to host bits in pgste */
714 pgste_val(pgste) |= bits << 52;
715 page_set_storage_key(address, skey ^ bits, 0);
716 } else if (!(pgste_val(pgste) & PGSTE_HR_BIT) &&
717 (bits & _PAGE_REFERENCED)) {
718 /* Transfer referenced bit to host bit in pgste */
719 pgste_val(pgste) |= PGSTE_HR_BIT;
720 page_reset_referenced(address);
722 /* Transfer page changed & referenced bit to guest bits in pgste */
723 pgste_val(pgste) |= bits << 48; /* GR bit & GC bit */
724 /* Copy page access key and fetch protection bit to pgste */
725 pgste_val(pgste) &= ~(PGSTE_ACC_BITS | PGSTE_FP_BIT);
726 pgste_val(pgste) |= (skey & (_PAGE_ACC_BITS | _PAGE_FP_BIT)) << 56;
732 static inline pgste_t pgste_update_young(pte_t *ptep, pgste_t pgste)
735 if (pte_val(*ptep) & _PAGE_INVALID)
737 /* Get referenced bit from storage key */
738 if (page_reset_referenced(pte_val(*ptep) & PAGE_MASK))
739 pgste_val(pgste) |= PGSTE_HR_BIT | PGSTE_GR_BIT;
744 static inline void pgste_set_key(pte_t *ptep, pgste_t pgste, pte_t entry)
747 unsigned long address;
750 if (pte_val(entry) & _PAGE_INVALID)
752 VM_BUG_ON(!(pte_val(*ptep) & _PAGE_INVALID));
753 address = pte_val(entry) & PAGE_MASK;
755 * Set page access key and fetch protection bit from pgste.
756 * The guest C/R information is still in the PGSTE, set real
759 nkey = (pgste_val(pgste) & (PGSTE_ACC_BITS | PGSTE_FP_BIT)) >> 56;
760 page_set_storage_key(address, nkey, 0);
764 static inline void pgste_set_pte(pte_t *ptep, pte_t entry)
766 if (!MACHINE_HAS_ESOP &&
767 (pte_val(entry) & _PAGE_PRESENT) &&
768 (pte_val(entry) & _PAGE_WRITE)) {
770 * Without enhanced suppression-on-protection force
771 * the dirty bit on for all writable ptes.
773 pte_val(entry) |= _PAGE_DIRTY;
774 pte_val(entry) &= ~_PAGE_PROTECT;
780 * struct gmap_struct - guest address space
781 * @mm: pointer to the parent mm_struct
782 * @table: pointer to the page directory
783 * @asce: address space control element for gmap page table
784 * @crst_list: list of all crst tables used in the guest address space
785 * @pfault_enabled: defines if pfaults are applicable for the guest
788 struct list_head list;
789 struct mm_struct *mm;
790 unsigned long *table;
793 struct list_head crst_list;
798 * struct gmap_rmap - reverse mapping for segment table entries
799 * @gmap: pointer to the gmap_struct
800 * @entry: pointer to a segment table entry
801 * @vmaddr: virtual address in the guest address space
804 struct list_head list;
806 unsigned long *entry;
807 unsigned long vmaddr;
811 * struct gmap_pgtable - gmap information attached to a page table
812 * @vmaddr: address of the 1MB segment in the process virtual memory
813 * @mapper: list of segment table entries mapping a page table
815 struct gmap_pgtable {
816 unsigned long vmaddr;
817 struct list_head mapper;
821 * struct gmap_notifier - notify function block for page invalidation
822 * @notifier_call: address of callback function
824 struct gmap_notifier {
825 struct list_head list;
826 void (*notifier_call)(struct gmap *gmap, unsigned long address);
829 struct gmap *gmap_alloc(struct mm_struct *mm);
830 void gmap_free(struct gmap *gmap);
831 void gmap_enable(struct gmap *gmap);
832 void gmap_disable(struct gmap *gmap);
833 int gmap_map_segment(struct gmap *gmap, unsigned long from,
834 unsigned long to, unsigned long len);
835 int gmap_unmap_segment(struct gmap *gmap, unsigned long to, unsigned long len);
836 unsigned long __gmap_translate(unsigned long address, struct gmap *);
837 unsigned long gmap_translate(unsigned long address, struct gmap *);
838 unsigned long __gmap_fault(unsigned long address, struct gmap *);
839 unsigned long gmap_fault(unsigned long address, struct gmap *);
840 void gmap_discard(unsigned long from, unsigned long to, struct gmap *);
841 void __gmap_zap(unsigned long address, struct gmap *);
843 void gmap_register_ipte_notifier(struct gmap_notifier *);
844 void gmap_unregister_ipte_notifier(struct gmap_notifier *);
845 int gmap_ipte_notify(struct gmap *, unsigned long start, unsigned long len);
846 void gmap_do_ipte_notify(struct mm_struct *, pte_t *);
848 static inline pgste_t pgste_ipte_notify(struct mm_struct *mm,
849 pte_t *ptep, pgste_t pgste)
852 if (pgste_val(pgste) & PGSTE_IN_BIT) {
853 pgste_val(pgste) &= ~PGSTE_IN_BIT;
854 gmap_do_ipte_notify(mm, ptep);
861 * Certain architectures need to do special things when PTEs
862 * within a page table are directly modified. Thus, the following
863 * hook is made available.
865 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
866 pte_t *ptep, pte_t entry)
870 if (mm_has_pgste(mm)) {
871 pgste = pgste_get_lock(ptep);
872 pgste_val(pgste) &= ~_PGSTE_GPS_ZERO;
873 pgste_set_key(ptep, pgste, entry);
874 pgste_set_pte(ptep, entry);
875 pgste_set_unlock(ptep, pgste);
877 if (!(pte_val(entry) & _PAGE_INVALID) && MACHINE_HAS_EDAT1)
878 pte_val(entry) |= _PAGE_CO;
884 * query functions pte_write/pte_dirty/pte_young only work if
885 * pte_present() is true. Undefined behaviour if not..
887 static inline int pte_write(pte_t pte)
889 return (pte_val(pte) & _PAGE_WRITE) != 0;
892 static inline int pte_dirty(pte_t pte)
894 return (pte_val(pte) & _PAGE_DIRTY) != 0;
897 static inline int pte_young(pte_t pte)
899 return (pte_val(pte) & _PAGE_YOUNG) != 0;
902 #define __HAVE_ARCH_PTE_UNUSED
903 static inline int pte_unused(pte_t pte)
905 return pte_val(pte) & _PAGE_UNUSED;
909 * pgd/pmd/pte modification functions
912 static inline void pgd_clear(pgd_t *pgd)
915 if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
916 pgd_val(*pgd) = _REGION2_ENTRY_EMPTY;
920 static inline void pud_clear(pud_t *pud)
923 if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
924 pud_val(*pud) = _REGION3_ENTRY_EMPTY;
928 static inline void pmd_clear(pmd_t *pmdp)
930 pmd_val(*pmdp) = _SEGMENT_ENTRY_INVALID;
933 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
935 pte_val(*ptep) = _PAGE_INVALID;
939 * The following pte modification functions only work if
940 * pte_present() is true. Undefined behaviour if not..
942 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
944 pte_val(pte) &= _PAGE_CHG_MASK;
945 pte_val(pte) |= pgprot_val(newprot);
947 * newprot for PAGE_NONE, PAGE_READ and PAGE_WRITE has the
948 * invalid bit set, clear it again for readable, young pages
950 if ((pte_val(pte) & _PAGE_YOUNG) && (pte_val(pte) & _PAGE_READ))
951 pte_val(pte) &= ~_PAGE_INVALID;
953 * newprot for PAGE_READ and PAGE_WRITE has the page protection
954 * bit set, clear it again for writable, dirty pages
956 if ((pte_val(pte) & _PAGE_DIRTY) && (pte_val(pte) & _PAGE_WRITE))
957 pte_val(pte) &= ~_PAGE_PROTECT;
961 static inline pte_t pte_wrprotect(pte_t pte)
963 pte_val(pte) &= ~_PAGE_WRITE;
964 pte_val(pte) |= _PAGE_PROTECT;
968 static inline pte_t pte_mkwrite(pte_t pte)
970 pte_val(pte) |= _PAGE_WRITE;
971 if (pte_val(pte) & _PAGE_DIRTY)
972 pte_val(pte) &= ~_PAGE_PROTECT;
976 static inline pte_t pte_mkclean(pte_t pte)
978 pte_val(pte) &= ~_PAGE_DIRTY;
979 pte_val(pte) |= _PAGE_PROTECT;
983 static inline pte_t pte_mkdirty(pte_t pte)
985 pte_val(pte) |= _PAGE_DIRTY;
986 if (pte_val(pte) & _PAGE_WRITE)
987 pte_val(pte) &= ~_PAGE_PROTECT;
991 static inline pte_t pte_mkold(pte_t pte)
993 pte_val(pte) &= ~_PAGE_YOUNG;
994 pte_val(pte) |= _PAGE_INVALID;
998 static inline pte_t pte_mkyoung(pte_t pte)
1000 pte_val(pte) |= _PAGE_YOUNG;
1001 if (pte_val(pte) & _PAGE_READ)
1002 pte_val(pte) &= ~_PAGE_INVALID;
1006 static inline pte_t pte_mkspecial(pte_t pte)
1008 pte_val(pte) |= _PAGE_SPECIAL;
1012 #ifdef CONFIG_HUGETLB_PAGE
1013 static inline pte_t pte_mkhuge(pte_t pte)
1015 pte_val(pte) |= _PAGE_LARGE;
1021 * Get (and clear) the user dirty bit for a pte.
1023 static inline int ptep_test_and_clear_user_dirty(struct mm_struct *mm,
1029 if (mm_has_pgste(mm)) {
1030 pgste = pgste_get_lock(ptep);
1031 pgste = pgste_update_all(ptep, pgste);
1032 dirty = !!(pgste_val(pgste) & PGSTE_HC_BIT);
1033 pgste_val(pgste) &= ~PGSTE_HC_BIT;
1034 pgste_set_unlock(ptep, pgste);
1041 * Get (and clear) the user referenced bit for a pte.
1043 static inline int ptep_test_and_clear_user_young(struct mm_struct *mm,
1049 if (mm_has_pgste(mm)) {
1050 pgste = pgste_get_lock(ptep);
1051 pgste = pgste_update_young(ptep, pgste);
1052 young = !!(pgste_val(pgste) & PGSTE_HR_BIT);
1053 pgste_val(pgste) &= ~PGSTE_HR_BIT;
1054 pgste_set_unlock(ptep, pgste);
1059 static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
1061 unsigned long pto = (unsigned long) ptep;
1063 #ifndef CONFIG_64BIT
1064 /* pto in ESA mode must point to the start of the segment table */
1067 /* Invalidation + global TLB flush for the pte */
1070 : "=m" (*ptep) : "m" (*ptep), "a" (pto), "a" (address));
1073 static inline void __ptep_ipte_local(unsigned long address, pte_t *ptep)
1075 unsigned long pto = (unsigned long) ptep;
1077 #ifndef CONFIG_64BIT
1078 /* pto in ESA mode must point to the start of the segment table */
1081 /* Invalidation + local TLB flush for the pte */
1083 " .insn rrf,0xb2210000,%2,%3,0,1"
1084 : "=m" (*ptep) : "m" (*ptep), "a" (pto), "a" (address));
1087 static inline void ptep_flush_direct(struct mm_struct *mm,
1088 unsigned long address, pte_t *ptep)
1092 if (pte_val(*ptep) & _PAGE_INVALID)
1094 active = (mm == current->active_mm) ? 1 : 0;
1095 count = atomic_add_return(0x10000, &mm->context.attach_count);
1096 if (MACHINE_HAS_TLB_LC && (count & 0xffff) <= active &&
1097 cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id())))
1098 __ptep_ipte_local(address, ptep);
1100 __ptep_ipte(address, ptep);
1101 atomic_sub(0x10000, &mm->context.attach_count);
1104 static inline void ptep_flush_lazy(struct mm_struct *mm,
1105 unsigned long address, pte_t *ptep)
1109 if (pte_val(*ptep) & _PAGE_INVALID)
1111 active = (mm == current->active_mm) ? 1 : 0;
1112 count = atomic_add_return(0x10000, &mm->context.attach_count);
1113 if ((count & 0xffff) <= active) {
1114 pte_val(*ptep) |= _PAGE_INVALID;
1115 mm->context.flush_mm = 1;
1117 __ptep_ipte(address, ptep);
1118 atomic_sub(0x10000, &mm->context.attach_count);
1121 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
1122 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
1123 unsigned long addr, pte_t *ptep)
1129 if (mm_has_pgste(vma->vm_mm)) {
1130 pgste = pgste_get_lock(ptep);
1131 pgste = pgste_ipte_notify(vma->vm_mm, ptep, pgste);
1135 ptep_flush_direct(vma->vm_mm, addr, ptep);
1136 young = pte_young(pte);
1137 pte = pte_mkold(pte);
1139 if (mm_has_pgste(vma->vm_mm)) {
1140 pgste_set_pte(ptep, pte);
1141 pgste_set_unlock(ptep, pgste);
1148 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
1149 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
1150 unsigned long address, pte_t *ptep)
1152 return ptep_test_and_clear_young(vma, address, ptep);
1156 * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
1157 * both clear the TLB for the unmapped pte. The reason is that
1158 * ptep_get_and_clear is used in common code (e.g. change_pte_range)
1159 * to modify an active pte. The sequence is
1160 * 1) ptep_get_and_clear
1162 * 3) flush_tlb_range
1163 * On s390 the tlb needs to get flushed with the modification of the pte
1164 * if the pte is active. The only way how this can be implemented is to
1165 * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
1168 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
1169 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
1170 unsigned long address, pte_t *ptep)
1175 if (mm_has_pgste(mm)) {
1176 pgste = pgste_get_lock(ptep);
1177 pgste = pgste_ipte_notify(mm, ptep, pgste);
1181 ptep_flush_lazy(mm, address, ptep);
1182 pte_val(*ptep) = _PAGE_INVALID;
1184 if (mm_has_pgste(mm)) {
1185 pgste = pgste_update_all(&pte, pgste);
1186 pgste_set_unlock(ptep, pgste);
1191 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
1192 static inline pte_t ptep_modify_prot_start(struct mm_struct *mm,
1193 unsigned long address,
1199 if (mm_has_pgste(mm)) {
1200 pgste = pgste_get_lock(ptep);
1201 pgste_ipte_notify(mm, ptep, pgste);
1205 ptep_flush_lazy(mm, address, ptep);
1207 if (mm_has_pgste(mm)) {
1208 pgste = pgste_update_all(&pte, pgste);
1209 pgste_set(ptep, pgste);
1214 static inline void ptep_modify_prot_commit(struct mm_struct *mm,
1215 unsigned long address,
1216 pte_t *ptep, pte_t pte)
1220 if (mm_has_pgste(mm)) {
1221 pgste = pgste_get(ptep);
1222 pgste_set_key(ptep, pgste, pte);
1223 pgste_set_pte(ptep, pte);
1224 pgste_set_unlock(ptep, pgste);
1229 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
1230 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1231 unsigned long address, pte_t *ptep)
1236 if (mm_has_pgste(vma->vm_mm)) {
1237 pgste = pgste_get_lock(ptep);
1238 pgste = pgste_ipte_notify(vma->vm_mm, ptep, pgste);
1242 ptep_flush_direct(vma->vm_mm, address, ptep);
1243 pte_val(*ptep) = _PAGE_INVALID;
1245 if (mm_has_pgste(vma->vm_mm)) {
1246 if ((pgste_val(pgste) & _PGSTE_GPS_USAGE_MASK) ==
1247 _PGSTE_GPS_USAGE_UNUSED)
1248 pte_val(pte) |= _PAGE_UNUSED;
1249 pgste = pgste_update_all(&pte, pgste);
1250 pgste_set_unlock(ptep, pgste);
1256 * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1257 * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1258 * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1259 * cannot be accessed while the batched unmap is running. In this case
1260 * full==1 and a simple pte_clear is enough. See tlb.h.
1262 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1263 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1264 unsigned long address,
1265 pte_t *ptep, int full)
1270 if (!full && mm_has_pgste(mm)) {
1271 pgste = pgste_get_lock(ptep);
1272 pgste = pgste_ipte_notify(mm, ptep, pgste);
1277 ptep_flush_lazy(mm, address, ptep);
1278 pte_val(*ptep) = _PAGE_INVALID;
1280 if (!full && mm_has_pgste(mm)) {
1281 pgste = pgste_update_all(&pte, pgste);
1282 pgste_set_unlock(ptep, pgste);
1287 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1288 static inline pte_t ptep_set_wrprotect(struct mm_struct *mm,
1289 unsigned long address, pte_t *ptep)
1294 if (pte_write(pte)) {
1295 if (mm_has_pgste(mm)) {
1296 pgste = pgste_get_lock(ptep);
1297 pgste = pgste_ipte_notify(mm, ptep, pgste);
1300 ptep_flush_lazy(mm, address, ptep);
1301 pte = pte_wrprotect(pte);
1303 if (mm_has_pgste(mm)) {
1304 pgste_set_pte(ptep, pte);
1305 pgste_set_unlock(ptep, pgste);
1312 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1313 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1314 unsigned long address, pte_t *ptep,
1315 pte_t entry, int dirty)
1319 if (pte_same(*ptep, entry))
1321 if (mm_has_pgste(vma->vm_mm)) {
1322 pgste = pgste_get_lock(ptep);
1323 pgste = pgste_ipte_notify(vma->vm_mm, ptep, pgste);
1326 ptep_flush_direct(vma->vm_mm, address, ptep);
1328 if (mm_has_pgste(vma->vm_mm)) {
1329 pgste_set_pte(ptep, entry);
1330 pgste_set_unlock(ptep, pgste);
1337 * Conversion functions: convert a page and protection to a page entry,
1338 * and a page entry and page directory to the page they refer to.
1340 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1343 pte_val(__pte) = physpage + pgprot_val(pgprot);
1344 return pte_mkyoung(__pte);
1347 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1349 unsigned long physpage = page_to_phys(page);
1350 pte_t __pte = mk_pte_phys(physpage, pgprot);
1352 if (pte_write(__pte) && PageDirty(page))
1353 __pte = pte_mkdirty(__pte);
1357 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1358 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1359 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1360 #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
1362 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
1363 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
1365 #ifndef CONFIG_64BIT
1367 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1368 #define pud_deref(pmd) ({ BUG(); 0UL; })
1369 #define pgd_deref(pmd) ({ BUG(); 0UL; })
1371 #define pud_offset(pgd, address) ((pud_t *) pgd)
1372 #define pmd_offset(pud, address) ((pmd_t *) pud + pmd_index(address))
1374 #else /* CONFIG_64BIT */
1376 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1377 #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)
1378 #define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN)
1380 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
1382 pud_t *pud = (pud_t *) pgd;
1383 if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
1384 pud = (pud_t *) pgd_deref(*pgd);
1385 return pud + pud_index(address);
1388 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
1390 pmd_t *pmd = (pmd_t *) pud;
1391 if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
1392 pmd = (pmd_t *) pud_deref(*pud);
1393 return pmd + pmd_index(address);
1396 #endif /* CONFIG_64BIT */
1398 #define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
1399 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1400 #define pte_page(x) pfn_to_page(pte_pfn(x))
1402 #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
1404 /* Find an entry in the lowest level page table.. */
1405 #define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr))
1406 #define pte_offset_kernel(pmd, address) pte_offset(pmd,address)
1407 #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
1408 #define pte_unmap(pte) do { } while (0)
1410 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLB_PAGE)
1411 static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
1414 * pgprot is PAGE_NONE, PAGE_READ, or PAGE_WRITE (see __Pxxx / __Sxxx)
1415 * Convert to segment table entry format.
1417 if (pgprot_val(pgprot) == pgprot_val(PAGE_NONE))
1418 return pgprot_val(SEGMENT_NONE);
1419 if (pgprot_val(pgprot) == pgprot_val(PAGE_READ))
1420 return pgprot_val(SEGMENT_READ);
1421 return pgprot_val(SEGMENT_WRITE);
1424 static inline pmd_t pmd_mkyoung(pmd_t pmd)
1427 if (pmd_prot_none(pmd)) {
1428 pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1430 pmd_val(pmd) |= _SEGMENT_ENTRY_YOUNG;
1431 pmd_val(pmd) &= ~_SEGMENT_ENTRY_INVALID;
1437 static inline pmd_t pmd_mkold(pmd_t pmd)
1440 if (pmd_prot_none(pmd)) {
1441 pmd_val(pmd) &= ~_SEGMENT_ENTRY_PROTECT;
1443 pmd_val(pmd) &= ~_SEGMENT_ENTRY_YOUNG;
1444 pmd_val(pmd) |= _SEGMENT_ENTRY_INVALID;
1450 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
1454 young = pmd_young(pmd);
1455 pmd_val(pmd) &= _SEGMENT_CHG_MASK;
1456 pmd_val(pmd) |= massage_pgprot_pmd(newprot);
1458 pmd = pmd_mkyoung(pmd);
1462 static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
1465 pmd_val(__pmd) = physpage + massage_pgprot_pmd(pgprot);
1466 return pmd_mkyoung(__pmd);
1469 static inline pmd_t pmd_mkwrite(pmd_t pmd)
1471 /* Do not clobber PROT_NONE segments! */
1472 if (!pmd_prot_none(pmd))
1473 pmd_val(pmd) &= ~_SEGMENT_ENTRY_PROTECT;
1476 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLB_PAGE */
1478 static inline void __pmdp_csp(pmd_t *pmdp)
1480 register unsigned long reg2 asm("2") = pmd_val(*pmdp);
1481 register unsigned long reg3 asm("3") = pmd_val(*pmdp) |
1482 _SEGMENT_ENTRY_INVALID;
1483 register unsigned long reg4 asm("4") = ((unsigned long) pmdp) + 5;
1488 : "d" (reg2), "d" (reg3), "d" (reg4), "m" (*pmdp) : "cc");
1491 static inline void __pmdp_idte(unsigned long address, pmd_t *pmdp)
1495 sto = (unsigned long) pmdp - pmd_index(address) * sizeof(pmd_t);
1497 " .insn rrf,0xb98e0000,%2,%3,0,0"
1499 : "m" (*pmdp), "a" (sto), "a" ((address & HPAGE_MASK))
1503 static inline void __pmdp_idte_local(unsigned long address, pmd_t *pmdp)
1507 sto = (unsigned long) pmdp - pmd_index(address) * sizeof(pmd_t);
1509 " .insn rrf,0xb98e0000,%2,%3,0,1"
1511 : "m" (*pmdp), "a" (sto), "a" ((address & HPAGE_MASK))
1515 static inline void pmdp_flush_direct(struct mm_struct *mm,
1516 unsigned long address, pmd_t *pmdp)
1520 if (pmd_val(*pmdp) & _SEGMENT_ENTRY_INVALID)
1522 if (!MACHINE_HAS_IDTE) {
1526 active = (mm == current->active_mm) ? 1 : 0;
1527 count = atomic_add_return(0x10000, &mm->context.attach_count);
1528 if (MACHINE_HAS_TLB_LC && (count & 0xffff) <= active &&
1529 cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id())))
1530 __pmdp_idte_local(address, pmdp);
1532 __pmdp_idte(address, pmdp);
1533 atomic_sub(0x10000, &mm->context.attach_count);
1536 static inline void pmdp_flush_lazy(struct mm_struct *mm,
1537 unsigned long address, pmd_t *pmdp)
1541 if (pmd_val(*pmdp) & _SEGMENT_ENTRY_INVALID)
1543 active = (mm == current->active_mm) ? 1 : 0;
1544 count = atomic_add_return(0x10000, &mm->context.attach_count);
1545 if ((count & 0xffff) <= active) {
1546 pmd_val(*pmdp) |= _SEGMENT_ENTRY_INVALID;
1547 mm->context.flush_mm = 1;
1548 } else if (MACHINE_HAS_IDTE)
1549 __pmdp_idte(address, pmdp);
1552 atomic_sub(0x10000, &mm->context.attach_count);
1555 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1557 #define __HAVE_ARCH_PGTABLE_DEPOSIT
1558 extern void pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
1561 #define __HAVE_ARCH_PGTABLE_WITHDRAW
1562 extern pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp);
1564 static inline int pmd_trans_splitting(pmd_t pmd)
1566 return pmd_val(pmd) & _SEGMENT_ENTRY_SPLIT;
1569 static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
1570 pmd_t *pmdp, pmd_t entry)
1572 if (!(pmd_val(entry) & _SEGMENT_ENTRY_INVALID) && MACHINE_HAS_EDAT1)
1573 pmd_val(entry) |= _SEGMENT_ENTRY_CO;
1577 static inline pmd_t pmd_mkhuge(pmd_t pmd)
1579 pmd_val(pmd) |= _SEGMENT_ENTRY_LARGE;
1583 static inline pmd_t pmd_wrprotect(pmd_t pmd)
1585 /* Do not clobber PROT_NONE segments! */
1586 if (!pmd_prot_none(pmd))
1587 pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1591 static inline pmd_t pmd_mkdirty(pmd_t pmd)
1593 /* No dirty bit in the segment table entry. */
1597 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
1598 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
1599 unsigned long address, pmd_t *pmdp)
1604 pmdp_flush_direct(vma->vm_mm, address, pmdp);
1605 *pmdp = pmd_mkold(pmd);
1606 return pmd_young(pmd);
1609 #define __HAVE_ARCH_PMDP_GET_AND_CLEAR
1610 static inline pmd_t pmdp_get_and_clear(struct mm_struct *mm,
1611 unsigned long address, pmd_t *pmdp)
1615 pmdp_flush_direct(mm, address, pmdp);
1620 #define __HAVE_ARCH_PMDP_CLEAR_FLUSH
1621 static inline pmd_t pmdp_clear_flush(struct vm_area_struct *vma,
1622 unsigned long address, pmd_t *pmdp)
1624 return pmdp_get_and_clear(vma->vm_mm, address, pmdp);
1627 #define __HAVE_ARCH_PMDP_INVALIDATE
1628 static inline void pmdp_invalidate(struct vm_area_struct *vma,
1629 unsigned long address, pmd_t *pmdp)
1631 pmdp_flush_direct(vma->vm_mm, address, pmdp);
1634 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
1635 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
1636 unsigned long address, pmd_t *pmdp)
1640 if (pmd_write(pmd)) {
1641 pmdp_flush_direct(mm, address, pmdp);
1642 set_pmd_at(mm, address, pmdp, pmd_wrprotect(pmd));
1646 #define pfn_pmd(pfn, pgprot) mk_pmd_phys(__pa((pfn) << PAGE_SHIFT), (pgprot))
1647 #define mk_pmd(page, pgprot) pfn_pmd(page_to_pfn(page), (pgprot))
1649 static inline int pmd_trans_huge(pmd_t pmd)
1651 return pmd_val(pmd) & _SEGMENT_ENTRY_LARGE;
1654 static inline int has_transparent_hugepage(void)
1656 return MACHINE_HAS_HPAGE ? 1 : 0;
1659 static inline unsigned long pmd_pfn(pmd_t pmd)
1661 return pmd_val(pmd) >> PAGE_SHIFT;
1663 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
1666 * 31 bit swap entry format:
1667 * A page-table entry has some bits we have to treat in a special way.
1668 * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
1669 * exception will occur instead of a page translation exception. The
1670 * specifiation exception has the bad habit not to store necessary
1671 * information in the lowcore.
1672 * Bits 21, 22, 30 and 31 are used to indicate the page type.
1673 * A swap pte is indicated by bit pattern (pte & 0x603) == 0x402
1674 * This leaves the bits 1-19 and bits 24-29 to store type and offset.
1675 * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
1676 * plus 24 for the offset.
1677 * 0| offset |0110|o|type |00|
1678 * 0 0000000001111111111 2222 2 22222 33
1679 * 0 1234567890123456789 0123 4 56789 01
1681 * 64 bit swap entry format:
1682 * A page-table entry has some bits we have to treat in a special way.
1683 * Bits 52 and bit 55 have to be zero, otherwise an specification
1684 * exception will occur instead of a page translation exception. The
1685 * specifiation exception has the bad habit not to store necessary
1686 * information in the lowcore.
1687 * Bits 53, 54, 62 and 63 are used to indicate the page type.
1688 * A swap pte is indicated by bit pattern (pte & 0x603) == 0x402
1689 * This leaves the bits 0-51 and bits 56-61 to store type and offset.
1690 * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
1691 * plus 56 for the offset.
1692 * | offset |0110|o|type |00|
1693 * 0000000000111111111122222222223333333333444444444455 5555 5 55566 66
1694 * 0123456789012345678901234567890123456789012345678901 2345 6 78901 23
1696 #ifndef CONFIG_64BIT
1697 #define __SWP_OFFSET_MASK (~0UL >> 12)
1699 #define __SWP_OFFSET_MASK (~0UL >> 11)
1701 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1704 offset &= __SWP_OFFSET_MASK;
1705 pte_val(pte) = _PAGE_INVALID | _PAGE_TYPE | ((type & 0x1f) << 2) |
1706 ((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
1710 #define __swp_type(entry) (((entry).val >> 2) & 0x1f)
1711 #define __swp_offset(entry) (((entry).val >> 11) | (((entry).val >> 7) & 1))
1712 #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
1714 #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
1715 #define __swp_entry_to_pte(x) ((pte_t) { (x).val })
1717 #ifndef CONFIG_64BIT
1718 # define PTE_FILE_MAX_BITS 26
1719 #else /* CONFIG_64BIT */
1720 # define PTE_FILE_MAX_BITS 59
1721 #endif /* CONFIG_64BIT */
1723 #define pte_to_pgoff(__pte) \
1724 ((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
1726 #define pgoff_to_pte(__off) \
1727 ((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
1728 | _PAGE_INVALID | _PAGE_PROTECT })
1730 #endif /* !__ASSEMBLY__ */
1732 #define kern_addr_valid(addr) (1)
1734 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1735 extern int vmem_remove_mapping(unsigned long start, unsigned long size);
1736 extern int s390_enable_sie(void);
1739 * No page table caches to initialise
1741 static inline void pgtable_cache_init(void) { }
1742 static inline void check_pgt_cache(void) { }
1744 #include <asm-generic/pgtable.h>
1746 #endif /* _S390_PAGE_H */