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>
21 extern int kaiser_enabled;
23 #define kaiser_enabled 0
27 * ZERO_PAGE is a global shared page that is always zero: used
28 * for zero-mapped memory areas etc..
30 extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)];
31 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
33 extern spinlock_t pgd_lock;
34 extern struct list_head pgd_list;
36 extern struct mm_struct *pgd_page_get_mm(struct page *page);
38 #ifdef CONFIG_PARAVIRT
39 #include <asm/paravirt.h>
40 #else /* !CONFIG_PARAVIRT */
41 #define set_pte(ptep, pte) native_set_pte(ptep, pte)
42 #define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte)
43 #define set_pmd_at(mm, addr, pmdp, pmd) native_set_pmd_at(mm, addr, pmdp, pmd)
45 #define set_pte_atomic(ptep, pte) \
46 native_set_pte_atomic(ptep, pte)
48 #define set_pmd(pmdp, pmd) native_set_pmd(pmdp, pmd)
50 #ifndef __PAGETABLE_PUD_FOLDED
51 #define set_pgd(pgdp, pgd) native_set_pgd(pgdp, pgd)
52 #define pgd_clear(pgd) native_pgd_clear(pgd)
56 # define set_pud(pudp, pud) native_set_pud(pudp, pud)
59 #ifndef __PAGETABLE_PMD_FOLDED
60 #define pud_clear(pud) native_pud_clear(pud)
63 #define pte_clear(mm, addr, ptep) native_pte_clear(mm, addr, ptep)
64 #define pmd_clear(pmd) native_pmd_clear(pmd)
66 #define pte_update(mm, addr, ptep) do { } while (0)
67 #define pte_update_defer(mm, addr, ptep) do { } while (0)
68 #define pmd_update(mm, addr, ptep) do { } while (0)
69 #define pmd_update_defer(mm, addr, ptep) do { } while (0)
71 #define pgd_val(x) native_pgd_val(x)
72 #define __pgd(x) native_make_pgd(x)
74 #ifndef __PAGETABLE_PUD_FOLDED
75 #define pud_val(x) native_pud_val(x)
76 #define __pud(x) native_make_pud(x)
79 #ifndef __PAGETABLE_PMD_FOLDED
80 #define pmd_val(x) native_pmd_val(x)
81 #define __pmd(x) native_make_pmd(x)
84 #define pte_val(x) native_pte_val(x)
85 #define __pte(x) native_make_pte(x)
87 #define arch_end_context_switch(prev) do {} while(0)
89 #endif /* CONFIG_PARAVIRT */
92 * The following only work if pte_present() is true.
93 * Undefined behaviour if not..
95 static inline int pte_dirty(pte_t pte)
97 return pte_flags(pte) & _PAGE_DIRTY;
100 static inline int pte_young(pte_t pte)
102 return pte_flags(pte) & _PAGE_ACCESSED;
105 static inline int pmd_young(pmd_t pmd)
107 return pmd_flags(pmd) & _PAGE_ACCESSED;
110 static inline int pte_write(pte_t pte)
112 return pte_flags(pte) & _PAGE_RW;
115 static inline int pte_file(pte_t pte)
117 return pte_flags(pte) & _PAGE_FILE;
120 static inline int pte_huge(pte_t pte)
122 return pte_flags(pte) & _PAGE_PSE;
125 static inline int pte_global(pte_t pte)
127 return pte_flags(pte) & _PAGE_GLOBAL;
130 static inline int pte_exec(pte_t pte)
132 return !(pte_flags(pte) & _PAGE_NX);
135 static inline int pte_special(pte_t pte)
137 return pte_flags(pte) & _PAGE_SPECIAL;
140 static inline unsigned long pte_pfn(pte_t pte)
142 return (pte_val(pte) & PTE_PFN_MASK) >> PAGE_SHIFT;
145 static inline unsigned long pmd_pfn(pmd_t pmd)
147 return (pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT;
150 static inline unsigned long pud_pfn(pud_t pud)
152 return (pud_val(pud) & PTE_PFN_MASK) >> PAGE_SHIFT;
155 #define pte_page(pte) pfn_to_page(pte_pfn(pte))
157 static inline int pmd_large(pmd_t pte)
159 return pmd_flags(pte) & _PAGE_PSE;
162 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
163 static inline int pmd_trans_splitting(pmd_t pmd)
165 return pmd_val(pmd) & _PAGE_SPLITTING;
168 static inline int pmd_trans_huge(pmd_t pmd)
170 return pmd_val(pmd) & _PAGE_PSE;
173 static inline int has_transparent_hugepage(void)
177 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
179 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
181 pteval_t v = native_pte_val(pte);
183 return native_make_pte(v | set);
186 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
188 pteval_t v = native_pte_val(pte);
190 return native_make_pte(v & ~clear);
193 static inline pte_t pte_mkclean(pte_t pte)
195 return pte_clear_flags(pte, _PAGE_DIRTY);
198 static inline pte_t pte_mkold(pte_t pte)
200 return pte_clear_flags(pte, _PAGE_ACCESSED);
203 static inline pte_t pte_wrprotect(pte_t pte)
205 return pte_clear_flags(pte, _PAGE_RW);
208 static inline pte_t pte_mkexec(pte_t pte)
210 return pte_clear_flags(pte, _PAGE_NX);
213 static inline pte_t pte_mkdirty(pte_t pte)
215 return pte_set_flags(pte, _PAGE_DIRTY);
218 static inline pte_t pte_mkyoung(pte_t pte)
220 return pte_set_flags(pte, _PAGE_ACCESSED);
223 static inline pte_t pte_mkwrite(pte_t pte)
225 return pte_set_flags(pte, _PAGE_RW);
228 static inline pte_t pte_mkhuge(pte_t pte)
230 return pte_set_flags(pte, _PAGE_PSE);
233 static inline pte_t pte_clrhuge(pte_t pte)
235 return pte_clear_flags(pte, _PAGE_PSE);
238 static inline pte_t pte_mkglobal(pte_t pte)
240 return pte_set_flags(pte, _PAGE_GLOBAL);
243 static inline pte_t pte_clrglobal(pte_t pte)
245 return pte_clear_flags(pte, _PAGE_GLOBAL);
248 static inline pte_t pte_mkspecial(pte_t pte)
250 return pte_set_flags(pte, _PAGE_SPECIAL);
253 static inline pmd_t pmd_set_flags(pmd_t pmd, pmdval_t set)
255 pmdval_t v = native_pmd_val(pmd);
257 return __pmd(v | set);
260 static inline pmd_t pmd_clear_flags(pmd_t pmd, pmdval_t clear)
262 pmdval_t v = native_pmd_val(pmd);
264 return __pmd(v & ~clear);
267 static inline pmd_t pmd_mkold(pmd_t pmd)
269 return pmd_clear_flags(pmd, _PAGE_ACCESSED);
272 static inline pmd_t pmd_wrprotect(pmd_t pmd)
274 return pmd_clear_flags(pmd, _PAGE_RW);
277 static inline pmd_t pmd_mkdirty(pmd_t pmd)
279 return pmd_set_flags(pmd, _PAGE_DIRTY);
282 static inline pmd_t pmd_mkhuge(pmd_t pmd)
284 return pmd_set_flags(pmd, _PAGE_PSE);
287 static inline pmd_t pmd_mkyoung(pmd_t pmd)
289 return pmd_set_flags(pmd, _PAGE_ACCESSED);
292 static inline pmd_t pmd_mkwrite(pmd_t pmd)
294 return pmd_set_flags(pmd, _PAGE_RW);
297 static inline pmd_t pmd_mknotpresent(pmd_t pmd)
299 return pmd_clear_flags(pmd, _PAGE_PRESENT);
303 * Mask out unsupported bits in a present pgprot. Non-present pgprots
304 * can use those bits for other purposes, so leave them be.
306 static inline pgprotval_t massage_pgprot(pgprot_t pgprot)
308 pgprotval_t protval = pgprot_val(pgprot);
310 if (protval & _PAGE_PRESENT)
311 protval &= __supported_pte_mask;
316 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
318 return __pte(((phys_addr_t)page_nr << PAGE_SHIFT) |
319 massage_pgprot(pgprot));
322 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
324 return __pmd(((phys_addr_t)page_nr << PAGE_SHIFT) |
325 massage_pgprot(pgprot));
328 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
330 pteval_t val = pte_val(pte);
333 * Chop off the NX bit (if present), and add the NX portion of
334 * the newprot (if present):
336 val &= _PAGE_CHG_MASK;
337 val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
342 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
344 pmdval_t val = pmd_val(pmd);
346 val &= _HPAGE_CHG_MASK;
347 val |= massage_pgprot(newprot) & ~_HPAGE_CHG_MASK;
352 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
353 #define pgprot_modify pgprot_modify
354 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
356 pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
357 pgprotval_t addbits = pgprot_val(newprot);
358 return __pgprot(preservebits | addbits);
361 #define pte_pgprot(x) __pgprot(pte_flags(x) & PTE_FLAGS_MASK)
363 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
365 static inline int is_new_memtype_allowed(u64 paddr, unsigned long size,
367 unsigned long new_flags)
370 * PAT type is always WB for untracked ranges, so no need to check.
372 if (x86_platform.is_untracked_pat_range(paddr, paddr + size))
376 * Certain new memtypes are not allowed with certain
378 * - request is uncached, return cannot be write-back
379 * - request is write-combine, return cannot be write-back
381 if ((flags == _PAGE_CACHE_UC_MINUS &&
382 new_flags == _PAGE_CACHE_WB) ||
383 (flags == _PAGE_CACHE_WC &&
384 new_flags == _PAGE_CACHE_WB)) {
391 pmd_t *populate_extra_pmd(unsigned long vaddr);
392 pte_t *populate_extra_pte(unsigned long vaddr);
393 #endif /* __ASSEMBLY__ */
396 # include "pgtable_32.h"
398 # include "pgtable_64.h"
402 #include <linux/mm_types.h>
404 static inline int pte_none(pte_t pte)
409 #define __HAVE_ARCH_PTE_SAME
410 static inline int pte_same(pte_t a, pte_t b)
412 return a.pte == b.pte;
415 static inline int pte_present(pte_t a)
417 return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
420 static inline int pte_hidden(pte_t pte)
422 return pte_flags(pte) & _PAGE_HIDDEN;
425 static inline int pmd_present(pmd_t pmd)
428 * Checking for _PAGE_PSE is needed too because
429 * split_huge_page will temporarily clear the present bit (but
430 * the _PAGE_PSE flag will remain set at all times while the
431 * _PAGE_PRESENT bit is clear).
433 return pmd_flags(pmd) & (_PAGE_PRESENT | _PAGE_PROTNONE | _PAGE_PSE);
436 static inline int pmd_none(pmd_t pmd)
438 /* Only check low word on 32-bit platforms, since it might be
439 out of sync with upper half. */
440 return (unsigned long)native_pmd_val(pmd) == 0;
443 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
445 return (unsigned long)__va(pmd_val(pmd) & PTE_PFN_MASK);
449 * Currently stuck as a macro due to indirect forward reference to
450 * linux/mmzone.h's __section_mem_map_addr() definition:
452 #define pmd_page(pmd) pfn_to_page((pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT)
455 * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
457 * this macro returns the index of the entry in the pmd page which would
458 * control the given virtual address
460 static inline unsigned long pmd_index(unsigned long address)
462 return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
466 * Conversion functions: convert a page and protection to a page entry,
467 * and a page entry and page directory to the page they refer to.
469 * (Currently stuck as a macro because of indirect forward reference
470 * to linux/mm.h:page_to_nid())
472 #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
475 * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
477 * this function returns the index of the entry in the pte page which would
478 * control the given virtual address
480 static inline unsigned long pte_index(unsigned long address)
482 return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
485 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
487 return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
490 static inline int pmd_bad(pmd_t pmd)
492 return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
495 static inline unsigned long pages_to_mb(unsigned long npg)
497 return npg >> (20 - PAGE_SHIFT);
500 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
501 remap_pfn_range(vma, vaddr, pfn, size, prot)
503 #if PAGETABLE_LEVELS > 2
504 static inline int pud_none(pud_t pud)
506 return native_pud_val(pud) == 0;
509 static inline int pud_present(pud_t pud)
511 return pud_flags(pud) & _PAGE_PRESENT;
514 static inline unsigned long pud_page_vaddr(pud_t pud)
516 return (unsigned long)__va((unsigned long)pud_val(pud) & PTE_PFN_MASK);
520 * Currently stuck as a macro due to indirect forward reference to
521 * linux/mmzone.h's __section_mem_map_addr() definition:
523 #define pud_page(pud) pfn_to_page(pud_val(pud) >> PAGE_SHIFT)
525 /* Find an entry in the second-level page table.. */
526 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
528 return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
531 static inline int pud_large(pud_t pud)
533 return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
534 (_PAGE_PSE | _PAGE_PRESENT);
537 static inline int pud_bad(pud_t pud)
539 return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
542 static inline int pud_large(pud_t pud)
546 #endif /* PAGETABLE_LEVELS > 2 */
548 #if PAGETABLE_LEVELS > 3
549 static inline int pgd_present(pgd_t pgd)
551 return pgd_flags(pgd) & _PAGE_PRESENT;
554 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
556 return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
560 * Currently stuck as a macro due to indirect forward reference to
561 * linux/mmzone.h's __section_mem_map_addr() definition:
563 #define pgd_page(pgd) pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
565 /* to find an entry in a page-table-directory. */
566 static inline unsigned long pud_index(unsigned long address)
568 return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
571 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
573 return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
576 static inline int pgd_bad(pgd_t pgd)
578 pgdval_t ignore_flags = _PAGE_USER;
580 * We set NX on KAISER pgds that map userspace memory so
581 * that userspace can not meaningfully use the kernel
582 * page table by accident; it will fault on the first
583 * instruction it tries to run. See native_set_pgd().
586 ignore_flags |= _PAGE_NX;
588 return (pgd_flags(pgd) & ~ignore_flags) != _KERNPG_TABLE;
591 static inline int pgd_none(pgd_t pgd)
593 return !native_pgd_val(pgd);
595 #endif /* PAGETABLE_LEVELS > 3 */
597 #endif /* __ASSEMBLY__ */
600 * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
602 * this macro returns the index of the entry in the pgd page which would
603 * control the given virtual address
605 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
608 * pgd_offset() returns a (pgd_t *)
609 * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
611 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
613 * a shortcut which implies the use of the kernel's pgd, instead
616 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
619 #define KERNEL_PGD_BOUNDARY pgd_index(PAGE_OFFSET)
620 #define KERNEL_PGD_PTRS (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
624 extern int direct_gbpages;
626 /* local pte updates need not use xchg for locking */
627 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
631 /* Pure native function needs no input for mm, addr */
632 native_pte_clear(NULL, 0, ptep);
636 static inline pmd_t native_local_pmdp_get_and_clear(pmd_t *pmdp)
640 native_pmd_clear(pmdp);
644 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
645 pte_t *ptep , pte_t pte)
647 native_set_pte(ptep, pte);
650 static inline void native_set_pmd_at(struct mm_struct *mm, unsigned long addr,
651 pmd_t *pmdp , pmd_t pmd)
653 native_set_pmd(pmdp, pmd);
656 #ifndef CONFIG_PARAVIRT
658 * Rules for using pte_update - it must be called after any PTE update which
659 * has not been done using the set_pte / clear_pte interfaces. It is used by
660 * shadow mode hypervisors to resynchronize the shadow page tables. Kernel PTE
661 * updates should either be sets, clears, or set_pte_atomic for P->P
662 * transitions, which means this hook should only be called for user PTEs.
663 * This hook implies a P->P protection or access change has taken place, which
664 * requires a subsequent TLB flush. The notification can optionally be delayed
665 * until the TLB flush event by using the pte_update_defer form of the
666 * interface, but care must be taken to assure that the flush happens while
667 * still holding the same page table lock so that the shadow and primary pages
668 * do not become out of sync on SMP.
670 #define pte_update(mm, addr, ptep) do { } while (0)
671 #define pte_update_defer(mm, addr, ptep) do { } while (0)
675 * We only update the dirty/accessed state if we set
676 * the dirty bit by hand in the kernel, since the hardware
677 * will do the accessed bit for us, and we don't want to
678 * race with other CPU's that might be updating the dirty
679 * bit at the same time.
681 struct vm_area_struct;
683 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
684 extern int ptep_set_access_flags(struct vm_area_struct *vma,
685 unsigned long address, pte_t *ptep,
686 pte_t entry, int dirty);
688 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
689 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
690 unsigned long addr, pte_t *ptep);
692 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
693 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
694 unsigned long address, pte_t *ptep);
696 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
697 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
700 pte_t pte = native_ptep_get_and_clear(ptep);
701 pte_update(mm, addr, ptep);
705 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
706 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
707 unsigned long addr, pte_t *ptep,
713 * Full address destruction in progress; paravirt does not
714 * care about updates and native needs no locking
716 pte = native_local_ptep_get_and_clear(ptep);
718 pte = ptep_get_and_clear(mm, addr, ptep);
723 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
724 static inline void ptep_set_wrprotect(struct mm_struct *mm,
725 unsigned long addr, pte_t *ptep)
727 clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
728 pte_update(mm, addr, ptep);
731 #define flush_tlb_fix_spurious_fault(vma, address)
733 #define mk_pmd(page, pgprot) pfn_pmd(page_to_pfn(page), (pgprot))
735 #define __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
736 extern int pmdp_set_access_flags(struct vm_area_struct *vma,
737 unsigned long address, pmd_t *pmdp,
738 pmd_t entry, int dirty);
740 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
741 extern int pmdp_test_and_clear_young(struct vm_area_struct *vma,
742 unsigned long addr, pmd_t *pmdp);
744 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
745 extern int pmdp_clear_flush_young(struct vm_area_struct *vma,
746 unsigned long address, pmd_t *pmdp);
749 #define __HAVE_ARCH_PMDP_SPLITTING_FLUSH
750 extern void pmdp_splitting_flush(struct vm_area_struct *vma,
751 unsigned long addr, pmd_t *pmdp);
753 #define __HAVE_ARCH_PMD_WRITE
754 static inline int pmd_write(pmd_t pmd)
756 return pmd_flags(pmd) & _PAGE_RW;
759 #define __HAVE_ARCH_PMDP_GET_AND_CLEAR
760 static inline pmd_t pmdp_get_and_clear(struct mm_struct *mm, unsigned long addr,
763 pmd_t pmd = native_pmdp_get_and_clear(pmdp);
764 pmd_update(mm, addr, pmdp);
768 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
769 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
770 unsigned long addr, pmd_t *pmdp)
772 clear_bit(_PAGE_BIT_RW, (unsigned long *)pmdp);
773 pmd_update(mm, addr, pmdp);
777 * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
779 * dst - pointer to pgd range anwhere on a pgd page
781 * count - the number of pgds to copy.
783 * dst and src can be on the same page, but the range must not overlap,
784 * and must not cross a page boundary.
786 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
788 memcpy(dst, src, count * sizeof(pgd_t));
790 if (kaiser_enabled) {
791 /* Clone the shadow pgd part as well */
792 memcpy(native_get_shadow_pgd(dst),
793 native_get_shadow_pgd(src),
794 count * sizeof(pgd_t));
800 #include <asm-generic/pgtable.h>
801 #endif /* __ASSEMBLY__ */
803 #endif /* _ASM_X86_PGTABLE_H */