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)) \
20 * ZERO_PAGE is a global shared page that is always zero: used
21 * for zero-mapped memory areas etc..
23 extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)];
24 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
26 extern spinlock_t pgd_lock;
27 extern struct list_head pgd_list;
29 #define set_pte(ptep, pte) xen_set_pte(ptep, pte)
30 #define set_pte_at(mm, addr, ptep, pte) xen_set_pte_at(mm, addr, ptep, pte)
32 #define set_pte_atomic(ptep, pte) \
33 xen_set_pte_atomic(ptep, pte)
35 #define set_pmd(pmdp, pmd) xen_set_pmd(pmdp, pmd)
37 #ifndef __PAGETABLE_PUD_FOLDED
38 #define set_pgd(pgdp, pgd) xen_set_pgd(pgdp, pgd)
39 #define pgd_clear(pgd) xen_pgd_clear(pgd)
43 # define set_pud(pudp, pud) xen_set_pud(pudp, pud)
46 #ifndef __PAGETABLE_PMD_FOLDED
47 #define pud_clear(pud) xen_pud_clear(pud)
50 #define pte_clear(mm, addr, ptep) xen_pte_clear(mm, addr, ptep)
51 #define pmd_clear(pmd) xen_pmd_clear(pmd)
53 #define pte_update(mm, addr, ptep) do { } while (0)
54 #define pte_update_defer(mm, addr, ptep) do { } while (0)
56 #define pgd_val(x) xen_pgd_val(x)
57 #define __pgd(x) xen_make_pgd(x)
59 #ifndef __PAGETABLE_PUD_FOLDED
60 #define pud_val(x) xen_pud_val(x)
61 #define __pud(x) xen_make_pud(x)
64 #ifndef __PAGETABLE_PMD_FOLDED
65 #define pmd_val(x) xen_pmd_val(x)
66 #define __pmd(x) xen_make_pmd(x)
69 #define pte_val(x) xen_pte_val(x)
70 #define __pte(x) xen_make_pte(x)
72 #define arch_end_context_switch(prev) do {} while(0)
75 * The following only work if pte_present() is true.
76 * Undefined behaviour if not..
78 static inline int pte_dirty(pte_t pte)
80 return pte_flags(pte) & _PAGE_DIRTY;
83 static inline int pte_young(pte_t pte)
85 return pte_flags(pte) & _PAGE_ACCESSED;
88 static inline int pte_write(pte_t pte)
90 return pte_flags(pte) & _PAGE_RW;
93 static inline int pte_file(pte_t pte)
95 return pte_flags(pte) & _PAGE_FILE;
98 static inline int pte_huge(pte_t pte)
100 return pte_flags(pte) & _PAGE_PSE;
103 static inline int pte_global(pte_t pte)
108 static inline int pte_exec(pte_t pte)
110 return !(pte_flags(pte) & _PAGE_NX);
113 static inline int pte_special(pte_t pte)
115 return pte_flags(pte) & _PAGE_SPECIAL;
118 #define pte_mfn(_pte) ((_pte).pte_low & _PAGE_PRESENT ? \
119 __pte_mfn(_pte) : pfn_to_mfn(__pte_mfn(_pte)))
120 #define pte_pfn(_pte) ((_pte).pte_low & _PAGE_IOMAP ? max_mapnr : \
121 (_pte).pte_low & _PAGE_PRESENT ? \
122 mfn_to_local_pfn(__pte_mfn(_pte)) : \
125 #define pte_page(pte) pfn_to_page(pte_pfn(pte))
127 static inline unsigned long pmd_pfn(pmd_t pmd)
129 return (pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT;
132 static inline int pmd_large(pmd_t pte)
134 return (pmd_flags(pte) & (_PAGE_PSE | _PAGE_PRESENT)) ==
135 (_PAGE_PSE | _PAGE_PRESENT);
138 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
140 pteval_t v = __pte_val(pte);
142 return __pte_ma(v | set);
145 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
147 pteval_t v = __pte_val(pte);
149 return __pte_ma(v & ~clear);
152 static inline pte_t pte_mkclean(pte_t pte)
154 return pte_clear_flags(pte, _PAGE_DIRTY);
157 static inline pte_t pte_mkold(pte_t pte)
159 return pte_clear_flags(pte, _PAGE_ACCESSED);
162 static inline pte_t pte_wrprotect(pte_t pte)
164 return pte_clear_flags(pte, _PAGE_RW);
167 static inline pte_t pte_mkexec(pte_t pte)
169 return pte_clear_flags(pte, _PAGE_NX);
172 static inline pte_t pte_mkdirty(pte_t pte)
174 return pte_set_flags(pte, _PAGE_DIRTY);
177 static inline pte_t pte_mkyoung(pte_t pte)
179 return pte_set_flags(pte, _PAGE_ACCESSED);
182 static inline pte_t pte_mkwrite(pte_t pte)
184 return pte_set_flags(pte, _PAGE_RW);
187 static inline pte_t pte_mkhuge(pte_t pte)
189 return pte_set_flags(pte, _PAGE_PSE);
192 static inline pte_t pte_clrhuge(pte_t pte)
194 return pte_clear_flags(pte, _PAGE_PSE);
197 static inline pte_t pte_mkglobal(pte_t pte)
202 static inline pte_t pte_clrglobal(pte_t pte)
207 static inline pte_t pte_mkspecial(pte_t pte)
209 return pte_set_flags(pte, _PAGE_SPECIAL);
213 * Mask out unsupported bits in a present pgprot. Non-present pgprots
214 * can use those bits for other purposes, so leave them be.
216 static inline pgprotval_t massage_pgprot(pgprot_t pgprot)
218 pgprotval_t protval = pgprot_val(pgprot);
220 if (protval & _PAGE_PRESENT)
221 protval &= __supported_pte_mask;
226 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
228 return __pte(((phys_addr_t)page_nr << PAGE_SHIFT) |
229 massage_pgprot(pgprot));
232 static inline pte_t pfn_pte_ma(unsigned long page_nr, pgprot_t pgprot)
234 return __pte_ma(((phys_addr_t)page_nr << PAGE_SHIFT) |
235 massage_pgprot(pgprot));
238 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
240 return __pmd(((phys_addr_t)page_nr << PAGE_SHIFT) |
241 massage_pgprot(pgprot));
244 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
246 pteval_t val = pte_val(pte) & _PAGE_CHG_MASK;
248 val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
253 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
254 #define pgprot_modify pgprot_modify
255 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
257 pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
258 pgprotval_t addbits = pgprot_val(newprot);
259 return __pgprot(preservebits | addbits);
262 #define pte_pgprot(x) __pgprot(pte_flags(x) & PTE_FLAGS_MASK)
264 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
266 static inline int is_new_memtype_allowed(u64 paddr, unsigned long size,
268 unsigned long new_flags)
271 * PAT type is always WB for ISA. So no need to check.
273 if (is_ISA_range(paddr, paddr + size - 1))
277 * Certain new memtypes are not allowed with certain
279 * - request is uncached, return cannot be write-back
280 * - request is write-combine, return cannot be write-back
282 if ((flags == _PAGE_CACHE_UC_MINUS &&
283 new_flags == _PAGE_CACHE_WB) ||
284 (flags == _PAGE_CACHE_WC &&
285 new_flags == _PAGE_CACHE_WB)) {
292 pmd_t *populate_extra_pmd(unsigned long vaddr);
293 pte_t *populate_extra_pte(unsigned long vaddr);
294 #endif /* __ASSEMBLY__ */
297 # include "pgtable_32.h"
299 # include "pgtable_64.h"
303 #include <linux/mm_types.h>
305 static inline int pte_none(pte_t pte)
310 #define __HAVE_ARCH_PTE_SAME
311 static inline int pte_same(pte_t a, pte_t b)
313 return a.pte == b.pte;
316 static inline int pte_present(pte_t a)
318 return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
321 static inline int pte_hidden(pte_t pte)
323 return pte_flags(pte) & _PAGE_HIDDEN;
326 static inline int pmd_present(pmd_t pmd)
328 #if CONFIG_XEN_COMPAT <= 0x030002
329 /* pmd_present doesn't just test the _PAGE_PRESENT bit since wr.p.t.
330 can temporarily clear it. */
331 return __pmd_val(pmd) != 0;
333 return pmd_flags(pmd) & _PAGE_PRESENT;
337 static inline int pmd_none(pmd_t pmd)
339 /* Only check low word on 32-bit platforms, since it might be
340 out of sync with upper half. */
341 return (unsigned long)__pmd_val(pmd) == 0;
344 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
346 return (unsigned long)__va(pmd_val(pmd) & PTE_PFN_MASK);
350 * Currently stuck as a macro due to indirect forward reference to
351 * linux/mmzone.h's __section_mem_map_addr() definition:
353 #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
356 * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
358 * this macro returns the index of the entry in the pmd page which would
359 * control the given virtual address
361 static inline unsigned long pmd_index(unsigned long address)
363 return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
367 * Conversion functions: convert a page and protection to a page entry,
368 * and a page entry and page directory to the page they refer to.
370 * (Currently stuck as a macro because of indirect forward reference
371 * to linux/mm.h:page_to_nid())
373 #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
376 * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
378 * this function returns the index of the entry in the pte page which would
379 * control the given virtual address
381 static inline unsigned long pte_index(unsigned long address)
383 return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
386 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
388 return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
391 static inline int pmd_bad(pmd_t pmd)
393 #if CONFIG_XEN_COMPAT <= 0x030002
394 return (pmd_flags(pmd) & ~_PAGE_USER & ~_PAGE_PRESENT)
395 != (_KERNPG_TABLE & ~_PAGE_PRESENT);
397 return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
401 static inline unsigned long pages_to_mb(unsigned long npg)
403 return npg >> (20 - PAGE_SHIFT);
406 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
407 direct_remap_pfn_range(vma, vaddr, pfn, size, prot, DOMID_IO)
409 #if PAGETABLE_LEVELS > 2
410 static inline int pud_none(pud_t pud)
412 return __pud_val(pud) == 0;
415 static inline int pud_present(pud_t pud)
417 return pud_flags(pud) & _PAGE_PRESENT;
420 static inline unsigned long pud_page_vaddr(pud_t pud)
422 return (unsigned long)__va((unsigned long)pud_val(pud) & PTE_PFN_MASK);
426 * Currently stuck as a macro due to indirect forward reference to
427 * linux/mmzone.h's __section_mem_map_addr() definition:
429 #define pud_page(pud) pfn_to_page(pud_val(pud) >> PAGE_SHIFT)
431 /* Find an entry in the second-level page table.. */
432 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
434 return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
437 static inline int pud_large(pud_t pud)
439 return (__pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
440 (_PAGE_PSE | _PAGE_PRESENT);
443 static inline int pud_bad(pud_t pud)
445 return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
448 static inline int pud_large(pud_t pud)
452 #endif /* PAGETABLE_LEVELS > 2 */
454 #if PAGETABLE_LEVELS > 3
455 static inline int pgd_present(pgd_t pgd)
457 return pgd_flags(pgd) & _PAGE_PRESENT;
460 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
462 return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
466 * Currently stuck as a macro due to indirect forward reference to
467 * linux/mmzone.h's __section_mem_map_addr() definition:
469 #define pgd_page(pgd) pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
471 /* to find an entry in a page-table-directory. */
472 static inline unsigned long pud_index(unsigned long address)
474 return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
477 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
479 return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
482 static inline int pgd_bad(pgd_t pgd)
484 return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
487 static inline int pgd_none(pgd_t pgd)
489 return !__pgd_val(pgd);
491 #endif /* PAGETABLE_LEVELS > 3 */
493 #endif /* __ASSEMBLY__ */
496 * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
498 * this macro returns the index of the entry in the pgd page which would
499 * control the given virtual address
501 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
504 * pgd_offset() returns a (pgd_t *)
505 * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
507 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
509 * a shortcut which implies the use of the kernel's pgd, instead
512 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
515 #define KERNEL_PGD_BOUNDARY pgd_index(PAGE_OFFSET)
516 #define KERNEL_PGD_PTRS (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
520 #define direct_gbpages 0
522 /* local pte updates need not use xchg for locking */
523 static inline pte_t xen_local_ptep_get_and_clear(pte_t *ptep, pte_t res)
525 xen_set_pte(ptep, __pte(0));
529 static inline void xen_set_pte_at(struct mm_struct *mm, unsigned long addr,
530 pte_t *ptep , pte_t pte)
532 if ((mm != current->mm && mm != &init_mm) ||
533 HYPERVISOR_update_va_mapping(addr, pte, 0))
534 xen_set_pte(ptep, pte);
537 static inline void xen_pte_clear(struct mm_struct *mm, unsigned long addr,
540 if ((mm != current->mm && mm != &init_mm)
541 || HYPERVISOR_update_va_mapping(addr, __pte(0), 0))
542 __xen_pte_clear(ptep);
545 #ifndef CONFIG_PARAVIRT
547 * Rules for using pte_update - it must be called after any PTE update which
548 * has not been done using the set_pte / clear_pte interfaces. It is used by
549 * shadow mode hypervisors to resynchronize the shadow page tables. Kernel PTE
550 * updates should either be sets, clears, or set_pte_atomic for P->P
551 * transitions, which means this hook should only be called for user PTEs.
552 * This hook implies a P->P protection or access change has taken place, which
553 * requires a subsequent TLB flush. The notification can optionally be delayed
554 * until the TLB flush event by using the pte_update_defer form of the
555 * interface, but care must be taken to assure that the flush happens while
556 * still holding the same page table lock so that the shadow and primary pages
557 * do not become out of sync on SMP.
559 #define pte_update(mm, addr, ptep) do { } while (0)
560 #define pte_update_defer(mm, addr, ptep) do { } while (0)
564 * We only update the dirty/accessed state if we set
565 * the dirty bit by hand in the kernel, since the hardware
566 * will do the accessed bit for us, and we don't want to
567 * race with other CPU's that might be updating the dirty
568 * bit at the same time.
570 struct vm_area_struct;
572 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
573 extern int ptep_set_access_flags(struct vm_area_struct *vma,
574 unsigned long address, pte_t *ptep,
575 pte_t entry, int dirty);
577 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
578 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
579 unsigned long addr, pte_t *ptep);
581 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
582 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
583 unsigned long address, pte_t *ptep);
585 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
586 #define ptep_clear_flush(vma, addr, ptep) \
588 pte_t *__ptep = (ptep); \
589 pte_t __res = *__ptep; \
590 if (!pte_none(__res) && \
591 ((vma)->vm_mm != current->mm || \
592 HYPERVISOR_update_va_mapping(addr, __pte(0), \
593 uvm_multi(mm_cpumask((vma)->vm_mm)) | \
595 __xen_pte_clear(__ptep); \
596 flush_tlb_page(vma, addr); \
601 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
602 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
608 || HYPERVISOR_update_va_mapping(addr, __pte(0), 0))) {
609 pte = xen_ptep_get_and_clear(ptep, pte);
610 pte_update(mm, addr, ptep);
615 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
616 #define ptep_get_and_clear_full(mm, addr, ptep, full) \
618 pte_t *__ptep = (ptep); \
619 pte_t __res = *__ptep; \
620 if (!PagePinned(virt_to_page((mm)->pgd))) \
621 __xen_pte_clear(__ptep); \
622 else if (!pte_none(__res)) \
623 xen_l1_entry_update(__ptep, __pte(0)); \
626 ptep_get_and_clear(mm, addr, ptep))
628 pte_t xen_ptep_get_and_clear_full(struct vm_area_struct *, unsigned long, pte_t *, int);
630 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
631 static inline void ptep_set_wrprotect(struct mm_struct *mm,
632 unsigned long addr, pte_t *ptep)
636 set_pte_at(mm, addr, ptep, pte_wrprotect(pte));
640 * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
642 * dst - pointer to pgd range anwhere on a pgd page
644 * count - the number of pgds to copy.
646 * dst and src can be on the same page, but the range must not overlap,
647 * and must not cross a page boundary.
649 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
651 memcpy(dst, src, count * sizeof(pgd_t));
654 #define arbitrary_virt_to_mfn(va) \
656 unsigned int __lvl; \
657 pte_t *__ptep = lookup_address((unsigned long)(va), &__lvl); \
658 BUG_ON(!__ptep || __lvl != PG_LEVEL_4K || !pte_present(*__ptep));\
662 #define arbitrary_virt_to_machine(va) \
663 (((maddr_t)arbitrary_virt_to_mfn(va) << PAGE_SHIFT) \
664 | ((unsigned long)(va) & (PAGE_SIZE - 1)))
666 #ifdef CONFIG_HIGHPTE
668 struct page *kmap_atomic_to_page(void *);
669 #define ptep_to_machine(ptep) \
671 pte_t *__ptep = (ptep); \
672 page_to_phys(kmap_atomic_to_page(__ptep)) \
673 | ((unsigned long)__ptep & (PAGE_SIZE - 1)); \
676 #define ptep_to_machine(ptep) virt_to_machine(ptep)
679 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
680 static inline pte_t ptep_modify_prot_start(struct mm_struct *mm, unsigned long addr,
683 #if CONFIG_XEN_COMPAT < 0x030300
684 if (unlikely(!xen_feature(XENFEAT_mmu_pt_update_preserve_ad)))
685 return ptep_get_and_clear(mm, addr, ptep);
690 static inline void ptep_modify_prot_commit(struct mm_struct *mm, unsigned long addr,
691 pte_t *ptep, pte_t pte)
695 #if CONFIG_XEN_COMPAT < 0x030300
696 if (unlikely(!xen_feature(XENFEAT_mmu_pt_update_preserve_ad))) {
697 set_pte_at(mm, addr, ptep, pte);
701 u.ptr = ptep_to_machine(ptep) | MMU_PT_UPDATE_PRESERVE_AD;
702 u.val = __pte_val(pte);
703 if (HYPERVISOR_mmu_update(&u, 1, NULL, DOMID_SELF))
707 #include <asm-generic/pgtable.h>
709 #include <xen/features.h>
710 void make_page_readonly(void *va, unsigned int feature);
711 void make_page_writable(void *va, unsigned int feature);
712 void make_pages_readonly(void *va, unsigned int nr, unsigned int feature);
713 void make_pages_writable(void *va, unsigned int nr, unsigned int feature);
715 struct vm_area_struct;
717 int direct_remap_pfn_range(struct vm_area_struct *vma,
718 unsigned long address,
723 int direct_kernel_remap_pfn_range(unsigned long address,
728 int create_lookup_pte_addr(struct mm_struct *mm,
729 unsigned long address,
731 int touch_pte_range(struct mm_struct *mm,
732 unsigned long address,
735 #endif /* __ASSEMBLY__ */
737 #endif /* _ASM_X86_PGTABLE_H */