1 #ifndef _ASM_X86_PGTABLE_H
2 #define _ASM_X86_PGTABLE_H
7 #include <asm/pgtable_types.h>
10 * Macro to mark a page protection value as UC-
12 #define pgprot_noncached(prot) \
13 ((boot_cpu_data.x86 > 3) \
14 ? (__pgprot(pgprot_val(prot) | _PAGE_CACHE_UC_MINUS)) \
19 #include <asm/x86_init.h>
22 * ZERO_PAGE is a global shared page that is always zero: used
23 * for zero-mapped memory areas etc..
25 extern unsigned long empty_zero_page
[PAGE_SIZE
/ sizeof(unsigned long)];
26 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
28 extern spinlock_t pgd_lock
;
29 extern struct list_head pgd_list
;
31 extern struct mm_struct
*pgd_page_get_mm(struct page
*page
);
33 #ifdef CONFIG_PARAVIRT
34 #include <asm/paravirt.h>
35 #else /* !CONFIG_PARAVIRT */
36 #define set_pte(ptep, pte) native_set_pte(ptep, pte)
37 #define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte)
38 #define set_pmd_at(mm, addr, pmdp, pmd) native_set_pmd_at(mm, addr, pmdp, pmd)
40 #define set_pte_atomic(ptep, pte) \
41 native_set_pte_atomic(ptep, pte)
43 #define set_pmd(pmdp, pmd) native_set_pmd(pmdp, pmd)
45 #ifndef __PAGETABLE_PUD_FOLDED
46 #define set_pgd(pgdp, pgd) native_set_pgd(pgdp, pgd)
47 #define pgd_clear(pgd) native_pgd_clear(pgd)
51 # define set_pud(pudp, pud) native_set_pud(pudp, pud)
54 #ifndef __PAGETABLE_PMD_FOLDED
55 #define pud_clear(pud) native_pud_clear(pud)
58 #define pte_clear(mm, addr, ptep) native_pte_clear(mm, addr, ptep)
59 #define pmd_clear(pmd) native_pmd_clear(pmd)
61 #define pte_update(mm, addr, ptep) do { } while (0)
62 #define pte_update_defer(mm, addr, ptep) do { } while (0)
63 #define pmd_update(mm, addr, ptep) do { } while (0)
64 #define pmd_update_defer(mm, addr, ptep) do { } while (0)
66 #define pgd_val(x) native_pgd_val(x)
67 #define __pgd(x) native_make_pgd(x)
69 #ifndef __PAGETABLE_PUD_FOLDED
70 #define pud_val(x) native_pud_val(x)
71 #define __pud(x) native_make_pud(x)
74 #ifndef __PAGETABLE_PMD_FOLDED
75 #define pmd_val(x) native_pmd_val(x)
76 #define __pmd(x) native_make_pmd(x)
79 #define pte_val(x) native_pte_val(x)
80 #define __pte(x) native_make_pte(x)
82 #define arch_end_context_switch(prev) do {} while(0)
84 #endif /* CONFIG_PARAVIRT */
87 * The following only work if pte_present() is true.
88 * Undefined behaviour if not..
90 static inline int pte_dirty(pte_t pte
)
92 return pte_flags(pte
) & _PAGE_DIRTY
;
95 static inline int pte_young(pte_t pte
)
97 return pte_flags(pte
) & _PAGE_ACCESSED
;
100 static inline int pmd_young(pmd_t pmd
)
102 return pmd_flags(pmd
) & _PAGE_ACCESSED
;
105 static inline int pte_write(pte_t pte
)
107 return pte_flags(pte
) & _PAGE_RW
;
110 static inline int pte_file(pte_t pte
)
112 return pte_flags(pte
) & _PAGE_FILE
;
115 static inline int pte_huge(pte_t pte
)
117 return pte_flags(pte
) & _PAGE_PSE
;
120 static inline int pte_global(pte_t pte
)
122 return pte_flags(pte
) & _PAGE_GLOBAL
;
125 static inline int pte_exec(pte_t pte
)
127 return !(pte_flags(pte
) & _PAGE_NX
);
130 static inline int pte_special(pte_t pte
)
132 return pte_flags(pte
) & _PAGE_SPECIAL
;
135 static inline unsigned long pte_pfn(pte_t pte
)
137 return (pte_val(pte
) & PTE_PFN_MASK
) >> PAGE_SHIFT
;
140 static inline unsigned long pmd_pfn(pmd_t pmd
)
142 return (pmd_val(pmd
) & PTE_PFN_MASK
) >> PAGE_SHIFT
;
145 #define pte_page(pte) pfn_to_page(pte_pfn(pte))
147 static inline int pmd_large(pmd_t pte
)
149 return (pmd_flags(pte
) & (_PAGE_PSE
| _PAGE_PRESENT
)) ==
150 (_PAGE_PSE
| _PAGE_PRESENT
);
153 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
154 static inline int pmd_trans_splitting(pmd_t pmd
)
156 return pmd_val(pmd
) & _PAGE_SPLITTING
;
159 static inline int pmd_trans_huge(pmd_t pmd
)
161 return pmd_val(pmd
) & _PAGE_PSE
;
164 static inline int has_transparent_hugepage(void)
168 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
170 static inline pte_t
pte_set_flags(pte_t pte
, pteval_t set
)
172 pteval_t v
= native_pte_val(pte
);
174 return native_make_pte(v
| set
);
177 static inline pte_t
pte_clear_flags(pte_t pte
, pteval_t clear
)
179 pteval_t v
= native_pte_val(pte
);
181 return native_make_pte(v
& ~clear
);
184 static inline pte_t
pte_mkclean(pte_t pte
)
186 return pte_clear_flags(pte
, _PAGE_DIRTY
);
189 static inline pte_t
pte_mkold(pte_t pte
)
191 return pte_clear_flags(pte
, _PAGE_ACCESSED
);
194 static inline pte_t
pte_wrprotect(pte_t pte
)
196 return pte_clear_flags(pte
, _PAGE_RW
);
199 static inline pte_t
pte_mkexec(pte_t pte
)
201 return pte_clear_flags(pte
, _PAGE_NX
);
204 static inline pte_t
pte_mkdirty(pte_t pte
)
206 return pte_set_flags(pte
, _PAGE_DIRTY
);
209 static inline pte_t
pte_mkyoung(pte_t pte
)
211 return pte_set_flags(pte
, _PAGE_ACCESSED
);
214 static inline pte_t
pte_mkwrite(pte_t pte
)
216 return pte_set_flags(pte
, _PAGE_RW
);
219 static inline pte_t
pte_mkhuge(pte_t pte
)
221 return pte_set_flags(pte
, _PAGE_PSE
);
224 static inline pte_t
pte_clrhuge(pte_t pte
)
226 return pte_clear_flags(pte
, _PAGE_PSE
);
229 static inline pte_t
pte_mkglobal(pte_t pte
)
231 return pte_set_flags(pte
, _PAGE_GLOBAL
);
234 static inline pte_t
pte_clrglobal(pte_t pte
)
236 return pte_clear_flags(pte
, _PAGE_GLOBAL
);
239 static inline pte_t
pte_mkspecial(pte_t pte
)
241 return pte_set_flags(pte
, _PAGE_SPECIAL
);
244 static inline pmd_t
pmd_set_flags(pmd_t pmd
, pmdval_t set
)
246 pmdval_t v
= native_pmd_val(pmd
);
248 return __pmd(v
| set
);
251 static inline pmd_t
pmd_clear_flags(pmd_t pmd
, pmdval_t clear
)
253 pmdval_t v
= native_pmd_val(pmd
);
255 return __pmd(v
& ~clear
);
258 static inline pmd_t
pmd_mkold(pmd_t pmd
)
260 return pmd_clear_flags(pmd
, _PAGE_ACCESSED
);
263 static inline pmd_t
pmd_wrprotect(pmd_t pmd
)
265 return pmd_clear_flags(pmd
, _PAGE_RW
);
268 static inline pmd_t
pmd_mkdirty(pmd_t pmd
)
270 return pmd_set_flags(pmd
, _PAGE_DIRTY
);
273 static inline pmd_t
pmd_mkhuge(pmd_t pmd
)
275 return pmd_set_flags(pmd
, _PAGE_PSE
);
278 static inline pmd_t
pmd_mkyoung(pmd_t pmd
)
280 return pmd_set_flags(pmd
, _PAGE_ACCESSED
);
283 static inline pmd_t
pmd_mkwrite(pmd_t pmd
)
285 return pmd_set_flags(pmd
, _PAGE_RW
);
288 static inline pmd_t
pmd_mknotpresent(pmd_t pmd
)
290 return pmd_clear_flags(pmd
, _PAGE_PRESENT
);
294 * Mask out unsupported bits in a present pgprot. Non-present pgprots
295 * can use those bits for other purposes, so leave them be.
297 static inline pgprotval_t
massage_pgprot(pgprot_t pgprot
)
299 pgprotval_t protval
= pgprot_val(pgprot
);
301 if (protval
& _PAGE_PRESENT
)
302 protval
&= __supported_pte_mask
;
307 static inline pte_t
pfn_pte(unsigned long page_nr
, pgprot_t pgprot
)
309 return __pte(((phys_addr_t
)page_nr
<< PAGE_SHIFT
) |
310 massage_pgprot(pgprot
));
313 static inline pmd_t
pfn_pmd(unsigned long page_nr
, pgprot_t pgprot
)
315 return __pmd(((phys_addr_t
)page_nr
<< PAGE_SHIFT
) |
316 massage_pgprot(pgprot
));
319 static inline pte_t
pte_modify(pte_t pte
, pgprot_t newprot
)
321 pteval_t val
= pte_val(pte
);
324 * Chop off the NX bit (if present), and add the NX portion of
325 * the newprot (if present):
327 val
&= _PAGE_CHG_MASK
;
328 val
|= massage_pgprot(newprot
) & ~_PAGE_CHG_MASK
;
333 static inline pmd_t
pmd_modify(pmd_t pmd
, pgprot_t newprot
)
335 pmdval_t val
= pmd_val(pmd
);
337 val
&= _HPAGE_CHG_MASK
;
338 val
|= massage_pgprot(newprot
) & ~_HPAGE_CHG_MASK
;
343 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
344 #define pgprot_modify pgprot_modify
345 static inline pgprot_t
pgprot_modify(pgprot_t oldprot
, pgprot_t newprot
)
347 pgprotval_t preservebits
= pgprot_val(oldprot
) & _PAGE_CHG_MASK
;
348 pgprotval_t addbits
= pgprot_val(newprot
);
349 return __pgprot(preservebits
| addbits
);
352 #define pte_pgprot(x) __pgprot(pte_flags(x) & PTE_FLAGS_MASK)
354 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
356 static inline int is_new_memtype_allowed(u64 paddr
, unsigned long size
,
358 unsigned long new_flags
)
361 * PAT type is always WB for untracked ranges, so no need to check.
363 if (x86_platform
.is_untracked_pat_range(paddr
, paddr
+ size
))
367 * Certain new memtypes are not allowed with certain
369 * - request is uncached, return cannot be write-back
370 * - request is write-combine, return cannot be write-back
372 if ((flags
== _PAGE_CACHE_UC_MINUS
&&
373 new_flags
== _PAGE_CACHE_WB
) ||
374 (flags
== _PAGE_CACHE_WC
&&
375 new_flags
== _PAGE_CACHE_WB
)) {
382 pmd_t
*populate_extra_pmd(unsigned long vaddr
);
383 pte_t
*populate_extra_pte(unsigned long vaddr
);
384 #endif /* __ASSEMBLY__ */
387 # include "pgtable_32.h"
389 # include "pgtable_64.h"
393 #include <linux/mm_types.h>
395 static inline int pte_none(pte_t pte
)
400 #define __HAVE_ARCH_PTE_SAME
401 static inline int pte_same(pte_t a
, pte_t b
)
403 return a
.pte
== b
.pte
;
406 static inline int pte_present(pte_t a
)
408 return pte_flags(a
) & (_PAGE_PRESENT
| _PAGE_PROTNONE
);
411 static inline int pte_hidden(pte_t pte
)
413 return pte_flags(pte
) & _PAGE_HIDDEN
;
416 static inline int pmd_present(pmd_t pmd
)
418 return pmd_flags(pmd
) & _PAGE_PRESENT
;
421 static inline int pmd_none(pmd_t pmd
)
423 /* Only check low word on 32-bit platforms, since it might be
424 out of sync with upper half. */
425 return (unsigned long)native_pmd_val(pmd
) == 0;
428 static inline unsigned long pmd_page_vaddr(pmd_t pmd
)
430 return (unsigned long)__va(pmd_val(pmd
) & PTE_PFN_MASK
);
434 * Currently stuck as a macro due to indirect forward reference to
435 * linux/mmzone.h's __section_mem_map_addr() definition:
437 #define pmd_page(pmd) pfn_to_page((pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT)
440 * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
442 * this macro returns the index of the entry in the pmd page which would
443 * control the given virtual address
445 static inline unsigned long pmd_index(unsigned long address
)
447 return (address
>> PMD_SHIFT
) & (PTRS_PER_PMD
- 1);
451 * Conversion functions: convert a page and protection to a page entry,
452 * and a page entry and page directory to the page they refer to.
454 * (Currently stuck as a macro because of indirect forward reference
455 * to linux/mm.h:page_to_nid())
457 #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
460 * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
462 * this function returns the index of the entry in the pte page which would
463 * control the given virtual address
465 static inline unsigned long pte_index(unsigned long address
)
467 return (address
>> PAGE_SHIFT
) & (PTRS_PER_PTE
- 1);
470 static inline pte_t
*pte_offset_kernel(pmd_t
*pmd
, unsigned long address
)
472 return (pte_t
*)pmd_page_vaddr(*pmd
) + pte_index(address
);
475 static inline int pmd_bad(pmd_t pmd
)
477 return (pmd_flags(pmd
) & ~_PAGE_USER
) != _KERNPG_TABLE
;
480 static inline unsigned long pages_to_mb(unsigned long npg
)
482 return npg
>> (20 - PAGE_SHIFT
);
485 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
486 remap_pfn_range(vma, vaddr, pfn, size, prot)
488 #if PAGETABLE_LEVELS > 2
489 static inline int pud_none(pud_t pud
)
491 return native_pud_val(pud
) == 0;
494 static inline int pud_present(pud_t pud
)
496 return pud_flags(pud
) & _PAGE_PRESENT
;
499 static inline unsigned long pud_page_vaddr(pud_t pud
)
501 return (unsigned long)__va((unsigned long)pud_val(pud
) & PTE_PFN_MASK
);
505 * Currently stuck as a macro due to indirect forward reference to
506 * linux/mmzone.h's __section_mem_map_addr() definition:
508 #define pud_page(pud) pfn_to_page(pud_val(pud) >> PAGE_SHIFT)
510 /* Find an entry in the second-level page table.. */
511 static inline pmd_t
*pmd_offset(pud_t
*pud
, unsigned long address
)
513 return (pmd_t
*)pud_page_vaddr(*pud
) + pmd_index(address
);
516 static inline int pud_large(pud_t pud
)
518 return (pud_val(pud
) & (_PAGE_PSE
| _PAGE_PRESENT
)) ==
519 (_PAGE_PSE
| _PAGE_PRESENT
);
522 static inline int pud_bad(pud_t pud
)
524 return (pud_flags(pud
) & ~(_KERNPG_TABLE
| _PAGE_USER
)) != 0;
527 static inline int pud_large(pud_t pud
)
531 #endif /* PAGETABLE_LEVELS > 2 */
533 #if PAGETABLE_LEVELS > 3
534 static inline int pgd_present(pgd_t pgd
)
536 return pgd_flags(pgd
) & _PAGE_PRESENT
;
539 static inline unsigned long pgd_page_vaddr(pgd_t pgd
)
541 return (unsigned long)__va((unsigned long)pgd_val(pgd
) & PTE_PFN_MASK
);
545 * Currently stuck as a macro due to indirect forward reference to
546 * linux/mmzone.h's __section_mem_map_addr() definition:
548 #define pgd_page(pgd) pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
550 /* to find an entry in a page-table-directory. */
551 static inline unsigned long pud_index(unsigned long address
)
553 return (address
>> PUD_SHIFT
) & (PTRS_PER_PUD
- 1);
556 static inline pud_t
*pud_offset(pgd_t
*pgd
, unsigned long address
)
558 return (pud_t
*)pgd_page_vaddr(*pgd
) + pud_index(address
);
561 static inline int pgd_bad(pgd_t pgd
)
563 return (pgd_flags(pgd
) & ~_PAGE_USER
) != _KERNPG_TABLE
;
566 static inline int pgd_none(pgd_t pgd
)
568 return !native_pgd_val(pgd
);
570 #endif /* PAGETABLE_LEVELS > 3 */
572 #endif /* __ASSEMBLY__ */
575 * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
577 * this macro returns the index of the entry in the pgd page which would
578 * control the given virtual address
580 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
583 * pgd_offset() returns a (pgd_t *)
584 * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
586 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
588 * a shortcut which implies the use of the kernel's pgd, instead
591 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
594 #define KERNEL_PGD_BOUNDARY pgd_index(PAGE_OFFSET)
595 #define KERNEL_PGD_PTRS (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
599 extern int direct_gbpages
;
601 /* local pte updates need not use xchg for locking */
602 static inline pte_t
native_local_ptep_get_and_clear(pte_t
*ptep
)
606 /* Pure native function needs no input for mm, addr */
607 native_pte_clear(NULL
, 0, ptep
);
611 static inline pmd_t
native_local_pmdp_get_and_clear(pmd_t
*pmdp
)
615 native_pmd_clear(pmdp
);
619 static inline void native_set_pte_at(struct mm_struct
*mm
, unsigned long addr
,
620 pte_t
*ptep
, pte_t pte
)
622 native_set_pte(ptep
, pte
);
625 static inline void native_set_pmd_at(struct mm_struct
*mm
, unsigned long addr
,
626 pmd_t
*pmdp
, pmd_t pmd
)
628 native_set_pmd(pmdp
, pmd
);
631 #ifndef CONFIG_PARAVIRT
633 * Rules for using pte_update - it must be called after any PTE update which
634 * has not been done using the set_pte / clear_pte interfaces. It is used by
635 * shadow mode hypervisors to resynchronize the shadow page tables. Kernel PTE
636 * updates should either be sets, clears, or set_pte_atomic for P->P
637 * transitions, which means this hook should only be called for user PTEs.
638 * This hook implies a P->P protection or access change has taken place, which
639 * requires a subsequent TLB flush. The notification can optionally be delayed
640 * until the TLB flush event by using the pte_update_defer form of the
641 * interface, but care must be taken to assure that the flush happens while
642 * still holding the same page table lock so that the shadow and primary pages
643 * do not become out of sync on SMP.
645 #define pte_update(mm, addr, ptep) do { } while (0)
646 #define pte_update_defer(mm, addr, ptep) do { } while (0)
650 * We only update the dirty/accessed state if we set
651 * the dirty bit by hand in the kernel, since the hardware
652 * will do the accessed bit for us, and we don't want to
653 * race with other CPU's that might be updating the dirty
654 * bit at the same time.
656 struct vm_area_struct
;
658 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
659 extern int ptep_set_access_flags(struct vm_area_struct
*vma
,
660 unsigned long address
, pte_t
*ptep
,
661 pte_t entry
, int dirty
);
663 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
664 extern int ptep_test_and_clear_young(struct vm_area_struct
*vma
,
665 unsigned long addr
, pte_t
*ptep
);
667 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
668 extern int ptep_clear_flush_young(struct vm_area_struct
*vma
,
669 unsigned long address
, pte_t
*ptep
);
671 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
672 static inline pte_t
ptep_get_and_clear(struct mm_struct
*mm
, unsigned long addr
,
675 pte_t pte
= native_ptep_get_and_clear(ptep
);
676 pte_update(mm
, addr
, ptep
);
680 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
681 static inline pte_t
ptep_get_and_clear_full(struct mm_struct
*mm
,
682 unsigned long addr
, pte_t
*ptep
,
688 * Full address destruction in progress; paravirt does not
689 * care about updates and native needs no locking
691 pte
= native_local_ptep_get_and_clear(ptep
);
693 pte
= ptep_get_and_clear(mm
, addr
, ptep
);
698 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
699 static inline void ptep_set_wrprotect(struct mm_struct
*mm
,
700 unsigned long addr
, pte_t
*ptep
)
702 clear_bit(_PAGE_BIT_RW
, (unsigned long *)&ptep
->pte
);
703 pte_update(mm
, addr
, ptep
);
706 #define flush_tlb_fix_spurious_fault(vma, address)
708 #define mk_pmd(page, pgprot) pfn_pmd(page_to_pfn(page), (pgprot))
710 #define __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
711 extern int pmdp_set_access_flags(struct vm_area_struct
*vma
,
712 unsigned long address
, pmd_t
*pmdp
,
713 pmd_t entry
, int dirty
);
715 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
716 extern int pmdp_test_and_clear_young(struct vm_area_struct
*vma
,
717 unsigned long addr
, pmd_t
*pmdp
);
719 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
720 extern int pmdp_clear_flush_young(struct vm_area_struct
*vma
,
721 unsigned long address
, pmd_t
*pmdp
);
724 #define __HAVE_ARCH_PMDP_SPLITTING_FLUSH
725 extern void pmdp_splitting_flush(struct vm_area_struct
*vma
,
726 unsigned long addr
, pmd_t
*pmdp
);
728 #define __HAVE_ARCH_PMD_WRITE
729 static inline int pmd_write(pmd_t pmd
)
731 return pmd_flags(pmd
) & _PAGE_RW
;
734 #define __HAVE_ARCH_PMDP_GET_AND_CLEAR
735 static inline pmd_t
pmdp_get_and_clear(struct mm_struct
*mm
, unsigned long addr
,
738 pmd_t pmd
= native_pmdp_get_and_clear(pmdp
);
739 pmd_update(mm
, addr
, pmdp
);
743 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
744 static inline void pmdp_set_wrprotect(struct mm_struct
*mm
,
745 unsigned long addr
, pmd_t
*pmdp
)
747 clear_bit(_PAGE_BIT_RW
, (unsigned long *)pmdp
);
748 pmd_update(mm
, addr
, pmdp
);
752 * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
754 * dst - pointer to pgd range anwhere on a pgd page
756 * count - the number of pgds to copy.
758 * dst and src can be on the same page, but the range must not overlap,
759 * and must not cross a page boundary.
761 static inline void clone_pgd_range(pgd_t
*dst
, pgd_t
*src
, int count
)
763 memcpy(dst
, src
, count
* sizeof(pgd_t
));
767 #include <asm-generic/pgtable.h>
768 #endif /* __ASSEMBLY__ */
770 #endif /* _ASM_X86_PGTABLE_H */