1 #ifndef _PPC64_PGTABLE_H
2 #define _PPC64_PGTABLE_H
5 * This file contains the functions and defines necessary to modify and use
6 * the ppc64 hashed page table.
10 #include <linux/config.h>
11 #include <linux/stddef.h>
12 #include <asm/processor.h> /* For TASK_SIZE */
15 #include <asm/tlbflush.h>
16 #endif /* __ASSEMBLY__ */
18 #include <asm-generic/pgtable-nopud.h>
21 * Entries per page directory level. The PTE level must use a 64b record
22 * for each page table entry. The PMD and PGD level use a 32b record for
23 * each entry by assuming that each entry is page aligned.
25 #define PTE_INDEX_SIZE 9
26 #define PMD_INDEX_SIZE 10
27 #define PGD_INDEX_SIZE 10
29 #define PTRS_PER_PTE (1 << PTE_INDEX_SIZE)
30 #define PTRS_PER_PMD (1 << PMD_INDEX_SIZE)
31 #define PTRS_PER_PGD (1 << PGD_INDEX_SIZE)
33 /* PMD_SHIFT determines what a second-level page table entry can map */
34 #define PMD_SHIFT (PAGE_SHIFT + PTE_INDEX_SIZE)
35 #define PMD_SIZE (1UL << PMD_SHIFT)
36 #define PMD_MASK (~(PMD_SIZE-1))
38 /* PGDIR_SHIFT determines what a third-level page table entry can map */
39 #define PGDIR_SHIFT (PMD_SHIFT + PMD_INDEX_SIZE)
40 #define PGDIR_SIZE (1UL << PGDIR_SHIFT)
41 #define PGDIR_MASK (~(PGDIR_SIZE-1))
43 #define FIRST_USER_ADDRESS 0
46 * Size of EA range mapped by our pagetables.
48 #define EADDR_SIZE (PTE_INDEX_SIZE + PMD_INDEX_SIZE + \
49 PGD_INDEX_SIZE + PAGE_SHIFT)
50 #define EADDR_MASK ((1UL << EADDR_SIZE) - 1)
53 * Define the address range of the vmalloc VM area.
55 #define VMALLOC_START (0xD000000000000000ul)
56 #define VMALLOC_SIZE (0x10000000000UL)
57 #define VMALLOC_END (VMALLOC_START + VMALLOC_SIZE)
60 * Bits in a linux-style PTE. These match the bits in the
61 * (hardware-defined) PowerPC PTE as closely as possible.
63 #define _PAGE_PRESENT 0x0001 /* software: pte contains a translation */
64 #define _PAGE_USER 0x0002 /* matches one of the PP bits */
65 #define _PAGE_FILE 0x0002 /* (!present only) software: pte holds file offset */
66 #define _PAGE_EXEC 0x0004 /* No execute on POWER4 and newer (we invert) */
67 #define _PAGE_GUARDED 0x0008
68 #define _PAGE_COHERENT 0x0010 /* M: enforce memory coherence (SMP systems) */
69 #define _PAGE_NO_CACHE 0x0020 /* I: cache inhibit */
70 #define _PAGE_WRITETHRU 0x0040 /* W: cache write-through */
71 #define _PAGE_DIRTY 0x0080 /* C: page changed */
72 #define _PAGE_ACCESSED 0x0100 /* R: page referenced */
73 #define _PAGE_RW 0x0200 /* software: user write access allowed */
74 #define _PAGE_HASHPTE 0x0400 /* software: pte has an associated HPTE */
75 #define _PAGE_BUSY 0x0800 /* software: PTE & hash are busy */
76 #define _PAGE_SECONDARY 0x8000 /* software: HPTE is in secondary group */
77 #define _PAGE_GROUP_IX 0x7000 /* software: HPTE index within group */
78 #define _PAGE_HUGE 0x10000 /* 16MB page */
79 /* Bits 0x7000 identify the index within an HPT Group */
80 #define _PAGE_HPTEFLAGS (_PAGE_BUSY | _PAGE_HASHPTE | _PAGE_SECONDARY | _PAGE_GROUP_IX)
81 /* PAGE_MASK gives the right answer below, but only by accident */
82 /* It should be preserving the high 48 bits and then specifically */
83 /* preserving _PAGE_SECONDARY | _PAGE_GROUP_IX */
84 #define _PAGE_CHG_MASK (PAGE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY | _PAGE_HPTEFLAGS)
86 #define _PAGE_BASE (_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_COHERENT)
88 #define _PAGE_WRENABLE (_PAGE_RW | _PAGE_DIRTY)
90 /* __pgprot defined in asm-ppc64/page.h */
91 #define PAGE_NONE __pgprot(_PAGE_PRESENT | _PAGE_ACCESSED)
93 #define PAGE_SHARED __pgprot(_PAGE_BASE | _PAGE_RW | _PAGE_USER)
94 #define PAGE_SHARED_X __pgprot(_PAGE_BASE | _PAGE_RW | _PAGE_USER | _PAGE_EXEC)
95 #define PAGE_COPY __pgprot(_PAGE_BASE | _PAGE_USER)
96 #define PAGE_COPY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
97 #define PAGE_READONLY __pgprot(_PAGE_BASE | _PAGE_USER)
98 #define PAGE_READONLY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
99 #define PAGE_KERNEL __pgprot(_PAGE_BASE | _PAGE_WRENABLE)
100 #define PAGE_KERNEL_CI __pgprot(_PAGE_PRESENT | _PAGE_ACCESSED | \
101 _PAGE_WRENABLE | _PAGE_NO_CACHE | _PAGE_GUARDED)
102 #define PAGE_KERNEL_EXEC __pgprot(_PAGE_BASE | _PAGE_WRENABLE | _PAGE_EXEC)
104 #define PAGE_AGP __pgprot(_PAGE_BASE | _PAGE_WRENABLE | _PAGE_NO_CACHE)
105 #define HAVE_PAGE_AGP
108 * This bit in a hardware PTE indicates that the page is *not* executable.
110 #define HW_NO_EXEC _PAGE_EXEC
113 * POWER4 and newer have per page execute protection, older chips can only
114 * do this on a segment (256MB) basis.
116 * Also, write permissions imply read permissions.
117 * This is the closest we can get..
119 * Note due to the way vm flags are laid out, the bits are XWR
121 #define __P000 PAGE_NONE
122 #define __P001 PAGE_READONLY
123 #define __P010 PAGE_COPY
124 #define __P011 PAGE_COPY
125 #define __P100 PAGE_READONLY_X
126 #define __P101 PAGE_READONLY_X
127 #define __P110 PAGE_COPY_X
128 #define __P111 PAGE_COPY_X
130 #define __S000 PAGE_NONE
131 #define __S001 PAGE_READONLY
132 #define __S010 PAGE_SHARED
133 #define __S011 PAGE_SHARED
134 #define __S100 PAGE_READONLY_X
135 #define __S101 PAGE_READONLY_X
136 #define __S110 PAGE_SHARED_X
137 #define __S111 PAGE_SHARED_X
142 * ZERO_PAGE is a global shared page that is always zero: used
143 * for zero-mapped memory areas etc..
145 extern unsigned long empty_zero_page
[PAGE_SIZE
/sizeof(unsigned long)];
146 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
147 #endif /* __ASSEMBLY__ */
149 /* shift to put page number into pte */
150 #define PTE_SHIFT (17)
152 #ifdef CONFIG_HUGETLB_PAGE
155 int hash_huge_page(struct mm_struct
*mm
, unsigned long access
,
156 unsigned long ea
, unsigned long vsid
, int local
);
158 void hugetlb_mm_free_pgd(struct mm_struct
*mm
);
159 #endif /* __ASSEMBLY__ */
161 #define HAVE_ARCH_UNMAPPED_AREA
162 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
165 #define hash_huge_page(mm,a,ea,vsid,local) -1
166 #define hugetlb_mm_free_pgd(mm) do {} while (0)
173 * Conversion functions: convert a page and protection to a page entry,
174 * and a page entry and page directory to the page they refer to.
176 * mk_pte takes a (struct page *) as input
178 #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
180 static inline pte_t
pfn_pte(unsigned long pfn
, pgprot_t pgprot
)
185 pte_val(pte
) = (pfn
<< PTE_SHIFT
) | pgprot_val(pgprot
);
189 #define pte_modify(_pte, newprot) \
190 (__pte((pte_val(_pte) & _PAGE_CHG_MASK) | pgprot_val(newprot)))
192 #define pte_none(pte) ((pte_val(pte) & ~_PAGE_HPTEFLAGS) == 0)
193 #define pte_present(pte) (pte_val(pte) & _PAGE_PRESENT)
195 /* pte_clear moved to later in this file */
197 #define pte_pfn(x) ((unsigned long)((pte_val(x) >> PTE_SHIFT)))
198 #define pte_page(x) pfn_to_page(pte_pfn(x))
200 #define pmd_set(pmdp, ptep) \
201 (pmd_val(*(pmdp)) = __ba_to_bpn(ptep))
202 #define pmd_none(pmd) (!pmd_val(pmd))
203 #define pmd_bad(pmd) (pmd_val(pmd) == 0)
204 #define pmd_present(pmd) (pmd_val(pmd) != 0)
205 #define pmd_clear(pmdp) (pmd_val(*(pmdp)) = 0)
206 #define pmd_page_kernel(pmd) (__bpn_to_ba(pmd_val(pmd)))
207 #define pmd_page(pmd) virt_to_page(pmd_page_kernel(pmd))
209 #define pud_set(pudp, pmdp) (pud_val(*(pudp)) = (__ba_to_bpn(pmdp)))
210 #define pud_none(pud) (!pud_val(pud))
211 #define pud_bad(pud) ((pud_val(pud)) == 0UL)
212 #define pud_present(pud) (pud_val(pud) != 0UL)
213 #define pud_clear(pudp) (pud_val(*(pudp)) = 0UL)
214 #define pud_page(pud) (__bpn_to_ba(pud_val(pud)))
217 * Find an entry in a page-table-directory. We combine the address region
218 * (the high order N bits) and the pgd portion of the address.
220 /* to avoid overflow in free_pgtables we don't use PTRS_PER_PGD here */
221 #define pgd_index(address) (((address) >> (PGDIR_SHIFT)) & 0x7ff)
223 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
225 /* Find an entry in the second-level page table.. */
226 #define pmd_offset(pudp,addr) \
227 ((pmd_t *) pud_page(*(pudp)) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1)))
229 /* Find an entry in the third-level page table.. */
230 #define pte_offset_kernel(dir,addr) \
231 ((pte_t *) pmd_page_kernel(*(dir)) \
232 + (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)))
234 #define pte_offset_map(dir,addr) pte_offset_kernel((dir), (addr))
235 #define pte_offset_map_nested(dir,addr) pte_offset_kernel((dir), (addr))
236 #define pte_unmap(pte) do { } while(0)
237 #define pte_unmap_nested(pte) do { } while(0)
239 /* to find an entry in a kernel page-table-directory */
240 /* This now only contains the vmalloc pages */
241 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
244 * The following only work if pte_present() is true.
245 * Undefined behaviour if not..
247 static inline int pte_read(pte_t pte
) { return pte_val(pte
) & _PAGE_USER
;}
248 static inline int pte_write(pte_t pte
) { return pte_val(pte
) & _PAGE_RW
;}
249 static inline int pte_exec(pte_t pte
) { return pte_val(pte
) & _PAGE_EXEC
;}
250 static inline int pte_dirty(pte_t pte
) { return pte_val(pte
) & _PAGE_DIRTY
;}
251 static inline int pte_young(pte_t pte
) { return pte_val(pte
) & _PAGE_ACCESSED
;}
252 static inline int pte_file(pte_t pte
) { return pte_val(pte
) & _PAGE_FILE
;}
253 static inline int pte_huge(pte_t pte
) { return pte_val(pte
) & _PAGE_HUGE
;}
255 static inline void pte_uncache(pte_t pte
) { pte_val(pte
) |= _PAGE_NO_CACHE
; }
256 static inline void pte_cache(pte_t pte
) { pte_val(pte
) &= ~_PAGE_NO_CACHE
; }
258 static inline pte_t
pte_rdprotect(pte_t pte
) {
259 pte_val(pte
) &= ~_PAGE_USER
; return pte
; }
260 static inline pte_t
pte_exprotect(pte_t pte
) {
261 pte_val(pte
) &= ~_PAGE_EXEC
; return pte
; }
262 static inline pte_t
pte_wrprotect(pte_t pte
) {
263 pte_val(pte
) &= ~(_PAGE_RW
); return pte
; }
264 static inline pte_t
pte_mkclean(pte_t pte
) {
265 pte_val(pte
) &= ~(_PAGE_DIRTY
); return pte
; }
266 static inline pte_t
pte_mkold(pte_t pte
) {
267 pte_val(pte
) &= ~_PAGE_ACCESSED
; return pte
; }
269 static inline pte_t
pte_mkread(pte_t pte
) {
270 pte_val(pte
) |= _PAGE_USER
; return pte
; }
271 static inline pte_t
pte_mkexec(pte_t pte
) {
272 pte_val(pte
) |= _PAGE_USER
| _PAGE_EXEC
; return pte
; }
273 static inline pte_t
pte_mkwrite(pte_t pte
) {
274 pte_val(pte
) |= _PAGE_RW
; return pte
; }
275 static inline pte_t
pte_mkdirty(pte_t pte
) {
276 pte_val(pte
) |= _PAGE_DIRTY
; return pte
; }
277 static inline pte_t
pte_mkyoung(pte_t pte
) {
278 pte_val(pte
) |= _PAGE_ACCESSED
; return pte
; }
279 static inline pte_t
pte_mkhuge(pte_t pte
) {
280 pte_val(pte
) |= _PAGE_HUGE
; return pte
; }
282 /* Atomic PTE updates */
283 static inline unsigned long pte_update(pte_t
*p
, unsigned long clr
)
285 unsigned long old
, tmp
;
287 __asm__
__volatile__(
288 "1: ldarx %0,0,%3 # pte_update\n\
294 : "=&r" (old
), "=&r" (tmp
), "=m" (*p
)
295 : "r" (p
), "r" (clr
), "m" (*p
), "i" (_PAGE_BUSY
)
300 /* PTE updating functions, this function puts the PTE in the
301 * batch, doesn't actually triggers the hash flush immediately,
302 * you need to call flush_tlb_pending() to do that.
304 extern void hpte_update(struct mm_struct
*mm
, unsigned long addr
, unsigned long pte
,
307 static inline int __ptep_test_and_clear_young(struct mm_struct
*mm
, unsigned long addr
, pte_t
*ptep
)
311 if ((pte_val(*ptep
) & (_PAGE_ACCESSED
| _PAGE_HASHPTE
)) == 0)
313 old
= pte_update(ptep
, _PAGE_ACCESSED
);
314 if (old
& _PAGE_HASHPTE
) {
315 hpte_update(mm
, addr
, old
, 0);
318 return (old
& _PAGE_ACCESSED
) != 0;
320 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
321 #define ptep_test_and_clear_young(__vma, __addr, __ptep) \
324 __r = __ptep_test_and_clear_young((__vma)->vm_mm, __addr, __ptep); \
329 * On RW/DIRTY bit transitions we can avoid flushing the hpte. For the
330 * moment we always flush but we need to fix hpte_update and test if the
331 * optimisation is worth it.
333 static inline int __ptep_test_and_clear_dirty(struct mm_struct
*mm
, unsigned long addr
, pte_t
*ptep
)
337 if ((pte_val(*ptep
) & _PAGE_DIRTY
) == 0)
339 old
= pte_update(ptep
, _PAGE_DIRTY
);
340 if (old
& _PAGE_HASHPTE
)
341 hpte_update(mm
, addr
, old
, 0);
342 return (old
& _PAGE_DIRTY
) != 0;
344 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
345 #define ptep_test_and_clear_dirty(__vma, __addr, __ptep) \
348 __r = __ptep_test_and_clear_dirty((__vma)->vm_mm, __addr, __ptep); \
352 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
353 static inline void ptep_set_wrprotect(struct mm_struct
*mm
, unsigned long addr
, pte_t
*ptep
)
357 if ((pte_val(*ptep
) & _PAGE_RW
) == 0)
359 old
= pte_update(ptep
, _PAGE_RW
);
360 if (old
& _PAGE_HASHPTE
)
361 hpte_update(mm
, addr
, old
, 0);
365 * We currently remove entries from the hashtable regardless of whether
366 * the entry was young or dirty. The generic routines only flush if the
367 * entry was young or dirty which is not good enough.
369 * We should be more intelligent about this but for the moment we override
370 * these functions and force a tlb flush unconditionally
372 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
373 #define ptep_clear_flush_young(__vma, __address, __ptep) \
375 int __young = __ptep_test_and_clear_young((__vma)->vm_mm, __address, \
380 #define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH
381 #define ptep_clear_flush_dirty(__vma, __address, __ptep) \
383 int __dirty = __ptep_test_and_clear_dirty((__vma)->vm_mm, __address, \
385 flush_tlb_page(__vma, __address); \
389 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
390 static inline pte_t
ptep_get_and_clear(struct mm_struct
*mm
, unsigned long addr
, pte_t
*ptep
)
392 unsigned long old
= pte_update(ptep
, ~0UL);
394 if (old
& _PAGE_HASHPTE
)
395 hpte_update(mm
, addr
, old
, 0);
399 static inline void pte_clear(struct mm_struct
*mm
, unsigned long addr
, pte_t
* ptep
)
401 unsigned long old
= pte_update(ptep
, ~0UL);
403 if (old
& _PAGE_HASHPTE
)
404 hpte_update(mm
, addr
, old
, 0);
408 * set_pte stores a linux PTE into the linux page table.
410 static inline void set_pte_at(struct mm_struct
*mm
, unsigned long addr
,
411 pte_t
*ptep
, pte_t pte
)
413 if (pte_present(*ptep
)) {
414 pte_clear(mm
, addr
, ptep
);
417 *ptep
= __pte(pte_val(pte
) & ~_PAGE_HPTEFLAGS
);
420 /* Set the dirty and/or accessed bits atomically in a linux PTE, this
421 * function doesn't need to flush the hash entry
423 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
424 static inline void __ptep_set_access_flags(pte_t
*ptep
, pte_t entry
, int dirty
)
426 unsigned long bits
= pte_val(entry
) &
427 (_PAGE_DIRTY
| _PAGE_ACCESSED
| _PAGE_RW
| _PAGE_EXEC
);
428 unsigned long old
, tmp
;
430 __asm__
__volatile__(
437 :"=&r" (old
), "=&r" (tmp
), "=m" (*ptep
)
438 :"r" (bits
), "r" (ptep
), "m" (*ptep
), "i" (_PAGE_BUSY
)
441 #define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
443 __ptep_set_access_flags(__ptep, __entry, __dirty); \
444 flush_tlb_page_nohash(__vma, __address); \
448 * Macro to mark a page protection value as "uncacheable".
450 #define pgprot_noncached(prot) (__pgprot(pgprot_val(prot) | _PAGE_NO_CACHE | _PAGE_GUARDED))
453 extern pgprot_t
phys_mem_access_prot(struct file
*file
, unsigned long addr
,
454 unsigned long size
, pgprot_t vma_prot
);
455 #define __HAVE_PHYS_MEM_ACCESS_PROT
457 #define __HAVE_ARCH_PTE_SAME
458 #define pte_same(A,B) (((pte_val(A) ^ pte_val(B)) & ~_PAGE_HPTEFLAGS) == 0)
460 #define pmd_ERROR(e) \
461 printk("%s:%d: bad pmd %08x.\n", __FILE__, __LINE__, pmd_val(e))
462 #define pgd_ERROR(e) \
463 printk("%s:%d: bad pgd %08x.\n", __FILE__, __LINE__, pgd_val(e))
465 extern pgd_t swapper_pg_dir
[];
467 extern void paging_init(void);
470 * Because the huge pgtables are only 2 level, they can take
471 * at most around 4M, much less than one hugepage which the
472 * process is presumably entitled to use. So we don't bother
473 * freeing up the pagetables on unmap, and wait until
474 * destroy_context() to clean up the lot.
476 #define hugetlb_free_pgd_range(tlb, addr, end, floor, ceiling) \
480 * This gets called at the end of handling a page fault, when
481 * the kernel has put a new PTE into the page table for the process.
482 * We use it to put a corresponding HPTE into the hash table
483 * ahead of time, instead of waiting for the inevitable extra
484 * hash-table miss exception.
486 struct vm_area_struct
;
487 extern void update_mmu_cache(struct vm_area_struct
*, unsigned long, pte_t
);
489 /* Encode and de-code a swap entry */
490 #define __swp_type(entry) (((entry).val >> 1) & 0x3f)
491 #define __swp_offset(entry) ((entry).val >> 8)
492 #define __swp_entry(type, offset) ((swp_entry_t) { ((type) << 1) | ((offset) << 8) })
493 #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) >> PTE_SHIFT })
494 #define __swp_entry_to_pte(x) ((pte_t) { (x).val << PTE_SHIFT })
495 #define pte_to_pgoff(pte) (pte_val(pte) >> PTE_SHIFT)
496 #define pgoff_to_pte(off) ((pte_t) {((off) << PTE_SHIFT)|_PAGE_FILE})
497 #define PTE_FILE_MAX_BITS (BITS_PER_LONG - PTE_SHIFT)
500 * kern_addr_valid is intended to indicate whether an address is a valid
501 * kernel address. Most 32-bit archs define it as always true (like this)
502 * but most 64-bit archs actually perform a test. What should we do here?
503 * The only use is in fs/ncpfs/dir.c
505 #define kern_addr_valid(addr) (1)
507 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
508 remap_pfn_range(vma, vaddr, pfn, size, prot)
510 void pgtable_cache_init(void);
513 * find_linux_pte returns the address of a linux pte for a given
514 * effective address and directory. If not found, it returns zero.
516 static inline pte_t
*find_linux_pte(pgd_t
*pgdir
, unsigned long ea
)
524 pg
= pgdir
+ pgd_index(ea
);
525 if (!pgd_none(*pg
)) {
526 pu
= pud_offset(pg
, ea
);
527 if (!pud_none(*pu
)) {
528 pm
= pmd_offset(pu
, ea
);
529 if (pmd_present(*pm
)) {
530 pt
= pte_offset_kernel(pm
, ea
);
532 if (!pte_present(pte
))
541 #include <asm-generic/pgtable.h>
543 #endif /* __ASSEMBLY__ */
545 #endif /* _PPC64_PGTABLE_H */