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 static inline unsigned long pud_pfn(pud_t pud)
147 return (pud_val(pud) & PTE_PFN_MASK) >> PAGE_SHIFT;
150 #define pte_page(pte) pfn_to_page(pte_pfn(pte))
152 static inline int pmd_large(pmd_t pte)
154 return pmd_flags(pte) & _PAGE_PSE;
157 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
158 static inline int pmd_trans_splitting(pmd_t pmd)
160 return pmd_val(pmd) & _PAGE_SPLITTING;
163 static inline int pmd_trans_huge(pmd_t pmd)
165 return pmd_val(pmd) & _PAGE_PSE;
168 static inline int has_transparent_hugepage(void)
172 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
174 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
176 pteval_t v = native_pte_val(pte);
178 return native_make_pte(v | set);
181 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
183 pteval_t v = native_pte_val(pte);
185 return native_make_pte(v & ~clear);
188 static inline pte_t pte_mkclean(pte_t pte)
190 return pte_clear_flags(pte, _PAGE_DIRTY);
193 static inline pte_t pte_mkold(pte_t pte)
195 return pte_clear_flags(pte, _PAGE_ACCESSED);
198 static inline pte_t pte_wrprotect(pte_t pte)
200 return pte_clear_flags(pte, _PAGE_RW);
203 static inline pte_t pte_mkexec(pte_t pte)
205 return pte_clear_flags(pte, _PAGE_NX);
208 static inline pte_t pte_mkdirty(pte_t pte)
210 return pte_set_flags(pte, _PAGE_DIRTY);
213 static inline pte_t pte_mkyoung(pte_t pte)
215 return pte_set_flags(pte, _PAGE_ACCESSED);
218 static inline pte_t pte_mkwrite(pte_t pte)
220 return pte_set_flags(pte, _PAGE_RW);
223 static inline pte_t pte_mkhuge(pte_t pte)
225 return pte_set_flags(pte, _PAGE_PSE);
228 static inline pte_t pte_clrhuge(pte_t pte)
230 return pte_clear_flags(pte, _PAGE_PSE);
233 static inline pte_t pte_mkglobal(pte_t pte)
235 return pte_set_flags(pte, _PAGE_GLOBAL);
238 static inline pte_t pte_clrglobal(pte_t pte)
240 return pte_clear_flags(pte, _PAGE_GLOBAL);
243 static inline pte_t pte_mkspecial(pte_t pte)
245 return pte_set_flags(pte, _PAGE_SPECIAL);
248 static inline pmd_t pmd_set_flags(pmd_t pmd, pmdval_t set)
250 pmdval_t v = native_pmd_val(pmd);
252 return __pmd(v | set);
255 static inline pmd_t pmd_clear_flags(pmd_t pmd, pmdval_t clear)
257 pmdval_t v = native_pmd_val(pmd);
259 return __pmd(v & ~clear);
262 static inline pmd_t pmd_mkold(pmd_t pmd)
264 return pmd_clear_flags(pmd, _PAGE_ACCESSED);
267 static inline pmd_t pmd_wrprotect(pmd_t pmd)
269 return pmd_clear_flags(pmd, _PAGE_RW);
272 static inline pmd_t pmd_mkdirty(pmd_t pmd)
274 return pmd_set_flags(pmd, _PAGE_DIRTY);
277 static inline pmd_t pmd_mkhuge(pmd_t pmd)
279 return pmd_set_flags(pmd, _PAGE_PSE);
282 static inline pmd_t pmd_mkyoung(pmd_t pmd)
284 return pmd_set_flags(pmd, _PAGE_ACCESSED);
287 static inline pmd_t pmd_mkwrite(pmd_t pmd)
289 return pmd_set_flags(pmd, _PAGE_RW);
292 static inline pmd_t pmd_mknotpresent(pmd_t pmd)
294 return pmd_clear_flags(pmd, _PAGE_PRESENT);
298 * Mask out unsupported bits in a present pgprot. Non-present pgprots
299 * can use those bits for other purposes, so leave them be.
301 static inline pgprotval_t massage_pgprot(pgprot_t pgprot)
303 pgprotval_t protval = pgprot_val(pgprot);
305 if (protval & _PAGE_PRESENT)
306 protval &= __supported_pte_mask;
311 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
313 return __pte(((phys_addr_t)page_nr << PAGE_SHIFT) |
314 massage_pgprot(pgprot));
317 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
319 return __pmd(((phys_addr_t)page_nr << PAGE_SHIFT) |
320 massage_pgprot(pgprot));
323 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
325 pteval_t val = pte_val(pte);
328 * Chop off the NX bit (if present), and add the NX portion of
329 * the newprot (if present):
331 val &= _PAGE_CHG_MASK;
332 val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
337 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
339 pmdval_t val = pmd_val(pmd);
341 val &= _HPAGE_CHG_MASK;
342 val |= massage_pgprot(newprot) & ~_HPAGE_CHG_MASK;
347 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
348 #define pgprot_modify pgprot_modify
349 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
351 pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
352 pgprotval_t addbits = pgprot_val(newprot);
353 return __pgprot(preservebits | addbits);
356 #define pte_pgprot(x) __pgprot(pte_flags(x) & PTE_FLAGS_MASK)
358 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
360 static inline int is_new_memtype_allowed(u64 paddr, unsigned long size,
362 unsigned long new_flags)
365 * PAT type is always WB for untracked ranges, so no need to check.
367 if (x86_platform.is_untracked_pat_range(paddr, paddr + size))
371 * Certain new memtypes are not allowed with certain
373 * - request is uncached, return cannot be write-back
374 * - request is write-combine, return cannot be write-back
376 if ((flags == _PAGE_CACHE_UC_MINUS &&
377 new_flags == _PAGE_CACHE_WB) ||
378 (flags == _PAGE_CACHE_WC &&
379 new_flags == _PAGE_CACHE_WB)) {
386 pmd_t *populate_extra_pmd(unsigned long vaddr);
387 pte_t *populate_extra_pte(unsigned long vaddr);
388 #endif /* __ASSEMBLY__ */
391 # include "pgtable_32.h"
393 # include "pgtable_64.h"
397 #include <linux/mm_types.h>
399 static inline int pte_none(pte_t pte)
404 #define __HAVE_ARCH_PTE_SAME
405 static inline int pte_same(pte_t a, pte_t b)
407 return a.pte == b.pte;
410 static inline int pte_present(pte_t a)
412 return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
415 static inline int pte_hidden(pte_t pte)
417 return pte_flags(pte) & _PAGE_HIDDEN;
420 static inline int pmd_present(pmd_t pmd)
423 * Checking for _PAGE_PSE is needed too because
424 * split_huge_page will temporarily clear the present bit (but
425 * the _PAGE_PSE flag will remain set at all times while the
426 * _PAGE_PRESENT bit is clear).
428 return pmd_flags(pmd) & (_PAGE_PRESENT | _PAGE_PROTNONE | _PAGE_PSE);
431 static inline int pmd_none(pmd_t pmd)
433 /* Only check low word on 32-bit platforms, since it might be
434 out of sync with upper half. */
435 return (unsigned long)native_pmd_val(pmd) == 0;
438 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
440 return (unsigned long)__va(pmd_val(pmd) & PTE_PFN_MASK);
444 * Currently stuck as a macro due to indirect forward reference to
445 * linux/mmzone.h's __section_mem_map_addr() definition:
447 #define pmd_page(pmd) pfn_to_page((pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT)
450 * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
452 * this macro returns the index of the entry in the pmd page which would
453 * control the given virtual address
455 static inline unsigned long pmd_index(unsigned long address)
457 return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
461 * Conversion functions: convert a page and protection to a page entry,
462 * and a page entry and page directory to the page they refer to.
464 * (Currently stuck as a macro because of indirect forward reference
465 * to linux/mm.h:page_to_nid())
467 #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
470 * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
472 * this function returns the index of the entry in the pte page which would
473 * control the given virtual address
475 static inline unsigned long pte_index(unsigned long address)
477 return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
480 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
482 return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
485 static inline int pmd_bad(pmd_t pmd)
487 return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
490 static inline unsigned long pages_to_mb(unsigned long npg)
492 return npg >> (20 - PAGE_SHIFT);
495 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
496 remap_pfn_range(vma, vaddr, pfn, size, prot)
498 #if PAGETABLE_LEVELS > 2
499 static inline int pud_none(pud_t pud)
501 return native_pud_val(pud) == 0;
504 static inline int pud_present(pud_t pud)
506 return pud_flags(pud) & _PAGE_PRESENT;
509 static inline unsigned long pud_page_vaddr(pud_t pud)
511 return (unsigned long)__va((unsigned long)pud_val(pud) & PTE_PFN_MASK);
515 * Currently stuck as a macro due to indirect forward reference to
516 * linux/mmzone.h's __section_mem_map_addr() definition:
518 #define pud_page(pud) pfn_to_page(pud_val(pud) >> PAGE_SHIFT)
520 /* Find an entry in the second-level page table.. */
521 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
523 return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
526 static inline int pud_large(pud_t pud)
528 return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
529 (_PAGE_PSE | _PAGE_PRESENT);
532 static inline int pud_bad(pud_t pud)
534 return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
537 static inline int pud_large(pud_t pud)
541 #endif /* PAGETABLE_LEVELS > 2 */
543 #if PAGETABLE_LEVELS > 3
544 static inline int pgd_present(pgd_t pgd)
546 return pgd_flags(pgd) & _PAGE_PRESENT;
549 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
551 return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
555 * Currently stuck as a macro due to indirect forward reference to
556 * linux/mmzone.h's __section_mem_map_addr() definition:
558 #define pgd_page(pgd) pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
560 /* to find an entry in a page-table-directory. */
561 static inline unsigned long pud_index(unsigned long address)
563 return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
566 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
568 return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
571 static inline int pgd_bad(pgd_t pgd)
573 return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
576 static inline int pgd_none(pgd_t pgd)
578 return !native_pgd_val(pgd);
580 #endif /* PAGETABLE_LEVELS > 3 */
582 #endif /* __ASSEMBLY__ */
585 * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
587 * this macro returns the index of the entry in the pgd page which would
588 * control the given virtual address
590 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
593 * pgd_offset() returns a (pgd_t *)
594 * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
596 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
598 * a shortcut which implies the use of the kernel's pgd, instead
601 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
604 #define KERNEL_PGD_BOUNDARY pgd_index(PAGE_OFFSET)
605 #define KERNEL_PGD_PTRS (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
609 extern int direct_gbpages;
611 /* local pte updates need not use xchg for locking */
612 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
616 /* Pure native function needs no input for mm, addr */
617 native_pte_clear(NULL, 0, ptep);
621 static inline pmd_t native_local_pmdp_get_and_clear(pmd_t *pmdp)
625 native_pmd_clear(pmdp);
629 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
630 pte_t *ptep , pte_t pte)
632 native_set_pte(ptep, pte);
635 static inline void native_set_pmd_at(struct mm_struct *mm, unsigned long addr,
636 pmd_t *pmdp , pmd_t pmd)
638 native_set_pmd(pmdp, pmd);
641 #ifndef CONFIG_PARAVIRT
643 * Rules for using pte_update - it must be called after any PTE update which
644 * has not been done using the set_pte / clear_pte interfaces. It is used by
645 * shadow mode hypervisors to resynchronize the shadow page tables. Kernel PTE
646 * updates should either be sets, clears, or set_pte_atomic for P->P
647 * transitions, which means this hook should only be called for user PTEs.
648 * This hook implies a P->P protection or access change has taken place, which
649 * requires a subsequent TLB flush. The notification can optionally be delayed
650 * until the TLB flush event by using the pte_update_defer form of the
651 * interface, but care must be taken to assure that the flush happens while
652 * still holding the same page table lock so that the shadow and primary pages
653 * do not become out of sync on SMP.
655 #define pte_update(mm, addr, ptep) do { } while (0)
656 #define pte_update_defer(mm, addr, ptep) do { } while (0)
660 * We only update the dirty/accessed state if we set
661 * the dirty bit by hand in the kernel, since the hardware
662 * will do the accessed bit for us, and we don't want to
663 * race with other CPU's that might be updating the dirty
664 * bit at the same time.
666 struct vm_area_struct;
668 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
669 extern int ptep_set_access_flags(struct vm_area_struct *vma,
670 unsigned long address, pte_t *ptep,
671 pte_t entry, int dirty);
673 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
674 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
675 unsigned long addr, pte_t *ptep);
677 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
678 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
679 unsigned long address, pte_t *ptep);
681 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
682 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
685 pte_t pte = native_ptep_get_and_clear(ptep);
686 pte_update(mm, addr, ptep);
690 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
691 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
692 unsigned long addr, pte_t *ptep,
698 * Full address destruction in progress; paravirt does not
699 * care about updates and native needs no locking
701 pte = native_local_ptep_get_and_clear(ptep);
703 pte = ptep_get_and_clear(mm, addr, ptep);
708 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
709 static inline void ptep_set_wrprotect(struct mm_struct *mm,
710 unsigned long addr, pte_t *ptep)
712 clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
713 pte_update(mm, addr, ptep);
716 #define flush_tlb_fix_spurious_fault(vma, address)
718 #define mk_pmd(page, pgprot) pfn_pmd(page_to_pfn(page), (pgprot))
720 #define __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
721 extern int pmdp_set_access_flags(struct vm_area_struct *vma,
722 unsigned long address, pmd_t *pmdp,
723 pmd_t entry, int dirty);
725 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
726 extern int pmdp_test_and_clear_young(struct vm_area_struct *vma,
727 unsigned long addr, pmd_t *pmdp);
729 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
730 extern int pmdp_clear_flush_young(struct vm_area_struct *vma,
731 unsigned long address, pmd_t *pmdp);
734 #define __HAVE_ARCH_PMDP_SPLITTING_FLUSH
735 extern void pmdp_splitting_flush(struct vm_area_struct *vma,
736 unsigned long addr, pmd_t *pmdp);
738 #define __HAVE_ARCH_PMD_WRITE
739 static inline int pmd_write(pmd_t pmd)
741 return pmd_flags(pmd) & _PAGE_RW;
744 #define __HAVE_ARCH_PMDP_GET_AND_CLEAR
745 static inline pmd_t pmdp_get_and_clear(struct mm_struct *mm, unsigned long addr,
748 pmd_t pmd = native_pmdp_get_and_clear(pmdp);
749 pmd_update(mm, addr, pmdp);
753 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
754 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
755 unsigned long addr, pmd_t *pmdp)
757 clear_bit(_PAGE_BIT_RW, (unsigned long *)pmdp);
758 pmd_update(mm, addr, pmdp);
762 * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
764 * dst - pointer to pgd range anwhere on a pgd page
766 * count - the number of pgds to copy.
768 * dst and src can be on the same page, but the range must not overlap,
769 * and must not cross a page boundary.
771 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
773 memcpy(dst, src, count * sizeof(pgd_t));
777 #include <asm-generic/pgtable.h>
778 #endif /* __ASSEMBLY__ */
780 #endif /* _ASM_X86_PGTABLE_H */