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 #define set_pte(ptep, pte) xen_set_pte(ptep, pte)
32 #define set_pte_at(mm, addr, ptep, pte) xen_set_pte_at(mm, addr, ptep, pte)
34 #define set_pte_atomic(ptep, pte) \
35 xen_set_pte_atomic(ptep, pte)
37 #define set_pmd(pmdp, pmd) xen_set_pmd(pmdp, pmd)
39 #ifndef __PAGETABLE_PUD_FOLDED
40 #define set_pgd(pgdp, pgd) xen_set_pgd(pgdp, pgd)
41 #define pgd_clear(pgd) xen_pgd_clear(pgd)
45 # define set_pud(pudp, pud) xen_set_pud(pudp, pud)
48 #ifndef __PAGETABLE_PMD_FOLDED
49 #define pud_clear(pud) xen_pud_clear(pud)
52 #define pte_clear(mm, addr, ptep) xen_pte_clear(mm, addr, ptep)
53 #define pmd_clear(pmd) xen_pmd_clear(pmd)
55 #define pte_update(mm, addr, ptep) do { } while (0)
56 #define pte_update_defer(mm, addr, ptep) do { } while (0)
58 #define pgd_val(x) xen_pgd_val(x)
59 #define __pgd(x) xen_make_pgd(x)
61 #ifndef __PAGETABLE_PUD_FOLDED
62 #define pud_val(x) xen_pud_val(x)
63 #define __pud(x) xen_make_pud(x)
66 #ifndef __PAGETABLE_PMD_FOLDED
67 #define pmd_val(x) xen_pmd_val(x)
68 #define __pmd(x) xen_make_pmd(x)
71 #define pte_val(x) xen_pte_val(x)
72 #define __pte(x) xen_make_pte(x)
74 #define arch_end_context_switch(prev) do {} while(0)
77 * The following only work if pte_present() is true.
78 * Undefined behaviour if not..
80 static inline int pte_dirty(pte_t pte)
82 return pte_flags(pte) & _PAGE_DIRTY;
85 static inline int pte_young(pte_t pte)
87 return pte_flags(pte) & _PAGE_ACCESSED;
90 static inline int pte_write(pte_t pte)
92 return pte_flags(pte) & _PAGE_RW;
95 static inline int pte_file(pte_t pte)
97 return pte_flags(pte) & _PAGE_FILE;
100 static inline int pte_huge(pte_t pte)
102 return pte_flags(pte) & _PAGE_PSE;
105 static inline int pte_global(pte_t pte)
110 static inline int pte_exec(pte_t pte)
112 return !(pte_flags(pte) & _PAGE_NX);
115 static inline int pte_special(pte_t pte)
117 return pte_flags(pte) & _PAGE_SPECIAL;
120 #define pte_mfn(_pte) ((_pte).pte_low & _PAGE_PRESENT ? \
121 __pte_mfn(_pte) : pfn_to_mfn(__pte_mfn(_pte)))
122 #define pte_pfn(_pte) ((_pte).pte_low & _PAGE_IOMAP ? max_mapnr : \
123 (_pte).pte_low & _PAGE_PRESENT ? \
124 mfn_to_local_pfn(__pte_mfn(_pte)) : \
127 #define pte_page(pte) pfn_to_page(pte_pfn(pte))
129 static inline unsigned long pmd_pfn(pmd_t pmd)
131 return (pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT;
134 static inline int pmd_large(pmd_t pte)
136 return (pmd_flags(pte) & (_PAGE_PSE | _PAGE_PRESENT)) ==
137 (_PAGE_PSE | _PAGE_PRESENT);
140 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
142 pteval_t v = __pte_val(pte);
144 return __pte_ma(v | set);
147 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
149 pteval_t v = __pte_val(pte);
151 return __pte_ma(v & ~clear);
154 static inline pte_t pte_mkclean(pte_t pte)
156 return pte_clear_flags(pte, _PAGE_DIRTY);
159 static inline pte_t pte_mkold(pte_t pte)
161 return pte_clear_flags(pte, _PAGE_ACCESSED);
164 static inline pte_t pte_wrprotect(pte_t pte)
166 return pte_clear_flags(pte, _PAGE_RW);
169 static inline pte_t pte_mkexec(pte_t pte)
171 return pte_clear_flags(pte, _PAGE_NX);
174 static inline pte_t pte_mkdirty(pte_t pte)
176 return pte_set_flags(pte, _PAGE_DIRTY);
179 static inline pte_t pte_mkyoung(pte_t pte)
181 return pte_set_flags(pte, _PAGE_ACCESSED);
184 static inline pte_t pte_mkwrite(pte_t pte)
186 return pte_set_flags(pte, _PAGE_RW);
189 static inline pte_t pte_mkhuge(pte_t pte)
191 return pte_set_flags(pte, _PAGE_PSE);
194 static inline pte_t pte_clrhuge(pte_t pte)
196 return pte_clear_flags(pte, _PAGE_PSE);
199 static inline pte_t pte_mkglobal(pte_t pte)
204 static inline pte_t pte_clrglobal(pte_t pte)
209 static inline pte_t pte_mkspecial(pte_t pte)
211 return pte_set_flags(pte, _PAGE_SPECIAL);
215 * Mask out unsupported bits in a present pgprot. Non-present pgprots
216 * can use those bits for other purposes, so leave them be.
218 static inline pgprotval_t massage_pgprot(pgprot_t pgprot)
220 pgprotval_t protval = pgprot_val(pgprot);
222 if (protval & _PAGE_PRESENT)
223 protval &= __supported_pte_mask;
228 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
230 return __pte(((phys_addr_t)page_nr << PAGE_SHIFT) |
231 massage_pgprot(pgprot));
234 static inline pte_t pfn_pte_ma(unsigned long page_nr, pgprot_t pgprot)
236 return __pte_ma(((phys_addr_t)page_nr << PAGE_SHIFT) |
237 massage_pgprot(pgprot));
240 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
242 return __pmd(((phys_addr_t)page_nr << PAGE_SHIFT) |
243 massage_pgprot(pgprot));
246 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
248 pteval_t val = pte_val(pte) & _PAGE_CHG_MASK;
250 val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
255 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
256 #define pgprot_modify pgprot_modify
257 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
259 pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
260 pgprotval_t addbits = pgprot_val(newprot);
261 return __pgprot(preservebits | addbits);
264 #define pte_pgprot(x) __pgprot(pte_flags(x) & PTE_FLAGS_MASK)
266 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
268 static inline int is_new_memtype_allowed(u64 paddr, unsigned long size,
270 unsigned long new_flags)
273 * PAT type is always WB for untracked ranges, so no need to check.
275 if (x86_platform.is_untracked_pat_range(paddr, paddr + size))
279 * Certain new memtypes are not allowed with certain
281 * - request is uncached, return cannot be write-back
282 * - request is write-combine, return cannot be write-back
284 if ((flags == _PAGE_CACHE_UC_MINUS &&
285 new_flags == _PAGE_CACHE_WB) ||
286 (flags == _PAGE_CACHE_WC &&
287 new_flags == _PAGE_CACHE_WB)) {
294 pmd_t *populate_extra_pmd(unsigned long vaddr);
295 pte_t *populate_extra_pte(unsigned long vaddr);
296 #endif /* __ASSEMBLY__ */
299 # include "pgtable_32.h"
301 # include "pgtable_64.h"
305 #include <linux/mm_types.h>
307 static inline int pte_none(pte_t pte)
312 #define __HAVE_ARCH_PTE_SAME
313 static inline int pte_same(pte_t a, pte_t b)
315 return a.pte == b.pte;
318 static inline int pte_present(pte_t a)
320 return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
323 static inline int pte_hidden(pte_t pte)
325 return pte_flags(pte) & _PAGE_HIDDEN;
328 static inline int pmd_present(pmd_t pmd)
330 #if CONFIG_XEN_COMPAT <= 0x030002
331 /* pmd_present doesn't just test the _PAGE_PRESENT bit since wr.p.t.
332 can temporarily clear it. */
333 return __pmd_val(pmd) != 0;
335 return pmd_flags(pmd) & _PAGE_PRESENT;
339 static inline int pmd_none(pmd_t pmd)
341 /* Only check low word on 32-bit platforms, since it might be
342 out of sync with upper half. */
343 return (unsigned long)__pmd_val(pmd) == 0;
346 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
348 return (unsigned long)__va(pmd_val(pmd) & PTE_PFN_MASK);
352 * Currently stuck as a macro due to indirect forward reference to
353 * linux/mmzone.h's __section_mem_map_addr() definition:
355 #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
358 * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
360 * this macro returns the index of the entry in the pmd page which would
361 * control the given virtual address
363 static inline unsigned long pmd_index(unsigned long address)
365 return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
369 * Conversion functions: convert a page and protection to a page entry,
370 * and a page entry and page directory to the page they refer to.
372 * (Currently stuck as a macro because of indirect forward reference
373 * to linux/mm.h:page_to_nid())
375 #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
378 * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
380 * this function returns the index of the entry in the pte page which would
381 * control the given virtual address
383 static inline unsigned long pte_index(unsigned long address)
385 return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
388 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
390 return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
393 static inline int pmd_bad(pmd_t pmd)
395 #if CONFIG_XEN_COMPAT <= 0x030002
396 return (pmd_flags(pmd) & ~_PAGE_USER & ~_PAGE_PRESENT)
397 != (_KERNPG_TABLE & ~_PAGE_PRESENT);
399 return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
403 static inline unsigned long pages_to_mb(unsigned long npg)
405 return npg >> (20 - PAGE_SHIFT);
408 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
409 direct_remap_pfn_range(vma, vaddr, pfn, size, prot, DOMID_IO)
411 #if PAGETABLE_LEVELS > 2
412 static inline int pud_none(pud_t pud)
414 return __pud_val(pud) == 0;
417 static inline int pud_present(pud_t pud)
419 return pud_flags(pud) & _PAGE_PRESENT;
422 static inline unsigned long pud_page_vaddr(pud_t pud)
424 return (unsigned long)__va((unsigned long)pud_val(pud) & PTE_PFN_MASK);
428 * Currently stuck as a macro due to indirect forward reference to
429 * linux/mmzone.h's __section_mem_map_addr() definition:
431 #define pud_page(pud) pfn_to_page(pud_val(pud) >> PAGE_SHIFT)
433 /* Find an entry in the second-level page table.. */
434 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
436 return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
439 static inline int pud_large(pud_t pud)
441 return (__pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
442 (_PAGE_PSE | _PAGE_PRESENT);
445 static inline int pud_bad(pud_t pud)
447 return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
450 static inline int pud_large(pud_t pud)
454 #endif /* PAGETABLE_LEVELS > 2 */
456 #if PAGETABLE_LEVELS > 3
457 static inline int pgd_present(pgd_t pgd)
459 return pgd_flags(pgd) & _PAGE_PRESENT;
462 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
464 return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
468 * Currently stuck as a macro due to indirect forward reference to
469 * linux/mmzone.h's __section_mem_map_addr() definition:
471 #define pgd_page(pgd) pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
473 /* to find an entry in a page-table-directory. */
474 static inline unsigned long pud_index(unsigned long address)
476 return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
479 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
481 return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
484 static inline int pgd_bad(pgd_t pgd)
486 return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
489 static inline int pgd_none(pgd_t pgd)
491 return !__pgd_val(pgd);
493 #endif /* PAGETABLE_LEVELS > 3 */
495 #endif /* __ASSEMBLY__ */
498 * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
500 * this macro returns the index of the entry in the pgd page which would
501 * control the given virtual address
503 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
506 * pgd_offset() returns a (pgd_t *)
507 * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
509 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
511 * a shortcut which implies the use of the kernel's pgd, instead
514 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
517 #define KERNEL_PGD_BOUNDARY pgd_index(PAGE_OFFSET)
518 #define KERNEL_PGD_PTRS (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
522 #define direct_gbpages 0
524 /* local pte updates need not use xchg for locking */
525 static inline pte_t xen_local_ptep_get_and_clear(pte_t *ptep, pte_t res)
527 xen_set_pte(ptep, __pte(0));
531 static inline void xen_set_pte_at(struct mm_struct *mm, unsigned long addr,
532 pte_t *ptep , pte_t pte)
534 if ((mm != current->mm && mm != &init_mm) ||
535 HYPERVISOR_update_va_mapping(addr, pte, 0))
536 xen_set_pte(ptep, pte);
539 static inline void xen_pte_clear(struct mm_struct *mm, unsigned long addr,
542 if ((mm != current->mm && mm != &init_mm)
543 || HYPERVISOR_update_va_mapping(addr, __pte(0), 0))
544 __xen_pte_clear(ptep);
547 #ifndef CONFIG_PARAVIRT
549 * Rules for using pte_update - it must be called after any PTE update which
550 * has not been done using the set_pte / clear_pte interfaces. It is used by
551 * shadow mode hypervisors to resynchronize the shadow page tables. Kernel PTE
552 * updates should either be sets, clears, or set_pte_atomic for P->P
553 * transitions, which means this hook should only be called for user PTEs.
554 * This hook implies a P->P protection or access change has taken place, which
555 * requires a subsequent TLB flush. The notification can optionally be delayed
556 * until the TLB flush event by using the pte_update_defer form of the
557 * interface, but care must be taken to assure that the flush happens while
558 * still holding the same page table lock so that the shadow and primary pages
559 * do not become out of sync on SMP.
561 #define pte_update(mm, addr, ptep) do { } while (0)
562 #define pte_update_defer(mm, addr, ptep) do { } while (0)
566 * We only update the dirty/accessed state if we set
567 * the dirty bit by hand in the kernel, since the hardware
568 * will do the accessed bit for us, and we don't want to
569 * race with other CPU's that might be updating the dirty
570 * bit at the same time.
572 struct vm_area_struct;
574 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
575 extern int ptep_set_access_flags(struct vm_area_struct *vma,
576 unsigned long address, pte_t *ptep,
577 pte_t entry, int dirty);
579 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
580 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
581 unsigned long addr, pte_t *ptep);
583 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
584 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
585 unsigned long address, pte_t *ptep);
587 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
588 #define ptep_clear_flush(vma, addr, ptep) \
590 pte_t *__ptep = (ptep); \
591 pte_t __res = *__ptep; \
592 if (!pte_none(__res) && \
593 ((vma)->vm_mm != current->mm || \
594 HYPERVISOR_update_va_mapping(addr, __pte(0), \
595 uvm_multi(mm_cpumask((vma)->vm_mm)) | \
597 __xen_pte_clear(__ptep); \
598 flush_tlb_page(vma, addr); \
603 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
604 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
610 || HYPERVISOR_update_va_mapping(addr, __pte(0), 0))) {
611 pte = xen_ptep_get_and_clear(ptep, pte);
612 pte_update(mm, addr, ptep);
617 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
618 #define ptep_get_and_clear_full(mm, addr, ptep, full) \
620 pte_t *__ptep = (ptep); \
621 pte_t __res = *__ptep; \
622 if (!PagePinned(virt_to_page((mm)->pgd))) \
623 __xen_pte_clear(__ptep); \
624 else if (!pte_none(__res)) \
625 xen_l1_entry_update(__ptep, __pte(0)); \
628 ptep_get_and_clear(mm, addr, ptep))
630 pte_t xen_ptep_get_and_clear_full(struct vm_area_struct *, unsigned long, pte_t *, int);
632 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
633 static inline void ptep_set_wrprotect(struct mm_struct *mm,
634 unsigned long addr, pte_t *ptep)
638 set_pte_at(mm, addr, ptep, pte_wrprotect(pte));
642 * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
644 * dst - pointer to pgd range anwhere on a pgd page
646 * count - the number of pgds to copy.
648 * dst and src can be on the same page, but the range must not overlap,
649 * and must not cross a page boundary.
651 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
653 memcpy(dst, src, count * sizeof(pgd_t));
656 #define arbitrary_virt_to_mfn(va) \
658 unsigned int __lvl; \
659 pte_t *__ptep = lookup_address((unsigned long)(va), &__lvl); \
660 BUG_ON(!__ptep || __lvl != PG_LEVEL_4K || !pte_present(*__ptep));\
664 #define arbitrary_virt_to_machine(va) \
665 (((maddr_t)arbitrary_virt_to_mfn(va) << PAGE_SHIFT) \
666 | ((unsigned long)(va) & (PAGE_SIZE - 1)))
668 #ifdef CONFIG_HIGHPTE
670 struct page *kmap_atomic_to_page(void *);
671 #define ptep_to_machine(ptep) \
673 pte_t *__ptep = (ptep); \
674 page_to_phys(kmap_atomic_to_page(__ptep)) \
675 | ((unsigned long)__ptep & (PAGE_SIZE - 1)); \
678 #define ptep_to_machine(ptep) virt_to_machine(ptep)
681 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
682 static inline pte_t ptep_modify_prot_start(struct mm_struct *mm, unsigned long addr,
685 #if CONFIG_XEN_COMPAT < 0x030300
686 if (unlikely(!xen_feature(XENFEAT_mmu_pt_update_preserve_ad)))
687 return ptep_get_and_clear(mm, addr, ptep);
692 static inline void ptep_modify_prot_commit(struct mm_struct *mm, unsigned long addr,
693 pte_t *ptep, pte_t pte)
697 #if CONFIG_XEN_COMPAT < 0x030300
698 if (unlikely(!xen_feature(XENFEAT_mmu_pt_update_preserve_ad))) {
699 set_pte_at(mm, addr, ptep, pte);
703 u.ptr = ptep_to_machine(ptep) | MMU_PT_UPDATE_PRESERVE_AD;
704 u.val = __pte_val(pte);
705 if (HYPERVISOR_mmu_update(&u, 1, NULL, DOMID_SELF))
709 #include <asm-generic/pgtable.h>
711 #include <xen/features.h>
712 void make_page_readonly(void *va, unsigned int feature);
713 void make_page_writable(void *va, unsigned int feature);
714 void make_pages_readonly(void *va, unsigned int nr, unsigned int feature);
715 void make_pages_writable(void *va, unsigned int nr, unsigned int feature);
717 struct vm_area_struct;
719 int direct_remap_pfn_range(struct vm_area_struct *vma,
720 unsigned long address,
725 int direct_kernel_remap_pfn_range(unsigned long address,
730 int create_lookup_pte_addr(struct mm_struct *mm,
731 unsigned long address,
733 int touch_pte_range(struct mm_struct *mm,
734 unsigned long address,
737 #endif /* __ASSEMBLY__ */
739 #endif /* _ASM_X86_PGTABLE_H */