1 #ifndef _ASM_X86_PGTABLE_H
2 #define _ASM_X86_PGTABLE_H
6 #include <asm/pgtable_types.h>
9 * Macro to mark a page protection value as UC-
11 #define pgprot_noncached(prot) \
12 ((boot_cpu_data.x86 > 3) \
13 ? (__pgprot(pgprot_val(prot) | _PAGE_CACHE_UC_MINUS)) \
19 * ZERO_PAGE is a global shared page that is always zero: used
20 * for zero-mapped memory areas etc..
22 extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)];
23 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
25 extern spinlock_t pgd_lock;
26 extern struct list_head pgd_list;
28 #ifdef CONFIG_PARAVIRT
29 #include <asm/paravirt.h>
30 #else /* !CONFIG_PARAVIRT */
31 #define set_pte(ptep, pte) native_set_pte(ptep, pte)
32 #define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte)
34 #define set_pte_atomic(ptep, pte) \
35 native_set_pte_atomic(ptep, pte)
37 #define set_pmd(pmdp, pmd) native_set_pmd(pmdp, pmd)
39 #ifndef __PAGETABLE_PUD_FOLDED
40 #define set_pgd(pgdp, pgd) native_set_pgd(pgdp, pgd)
41 #define pgd_clear(pgd) native_pgd_clear(pgd)
45 # define set_pud(pudp, pud) native_set_pud(pudp, pud)
48 #ifndef __PAGETABLE_PMD_FOLDED
49 #define pud_clear(pud) native_pud_clear(pud)
52 #define pte_clear(mm, addr, ptep) native_pte_clear(mm, addr, ptep)
53 #define pmd_clear(pmd) native_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 static inline void __init paravirt_pagetable_setup_start(pgd_t *base)
60 native_pagetable_setup_start(base);
63 static inline void __init paravirt_pagetable_setup_done(pgd_t *base)
65 native_pagetable_setup_done(base);
68 #define pgd_val(x) native_pgd_val(x)
69 #define __pgd(x) native_make_pgd(x)
71 #ifndef __PAGETABLE_PUD_FOLDED
72 #define pud_val(x) native_pud_val(x)
73 #define __pud(x) native_make_pud(x)
76 #ifndef __PAGETABLE_PMD_FOLDED
77 #define pmd_val(x) native_pmd_val(x)
78 #define __pmd(x) native_make_pmd(x)
81 #define pte_val(x) native_pte_val(x)
82 #define __pte(x) native_make_pte(x)
84 #define arch_end_context_switch(prev) do {} while(0)
86 #endif /* CONFIG_PARAVIRT */
89 * The following only work if pte_present() is true.
90 * Undefined behaviour if not..
92 static inline int pte_dirty(pte_t pte)
94 return pte_flags(pte) & _PAGE_DIRTY;
97 static inline int pte_young(pte_t pte)
99 return pte_flags(pte) & _PAGE_ACCESSED;
102 static inline int pte_write(pte_t pte)
104 return pte_flags(pte) & _PAGE_RW;
107 static inline int pte_file(pte_t pte)
109 return pte_flags(pte) & _PAGE_FILE;
112 static inline int pte_huge(pte_t pte)
114 return pte_flags(pte) & _PAGE_PSE;
117 static inline int pte_global(pte_t pte)
119 return pte_flags(pte) & _PAGE_GLOBAL;
122 static inline int pte_exec(pte_t pte)
124 return !(pte_flags(pte) & _PAGE_NX);
127 static inline int pte_special(pte_t pte)
129 return pte_flags(pte) & _PAGE_SPECIAL;
132 static inline unsigned long pte_pfn(pte_t pte)
134 return (pte_val(pte) & PTE_PFN_MASK) >> PAGE_SHIFT;
137 #define pte_page(pte) pfn_to_page(pte_pfn(pte))
139 static inline int pmd_large(pmd_t pte)
141 return (pmd_flags(pte) & (_PAGE_PSE | _PAGE_PRESENT)) ==
142 (_PAGE_PSE | _PAGE_PRESENT);
145 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
147 pteval_t v = native_pte_val(pte);
149 return native_make_pte(v | set);
152 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
154 pteval_t v = native_pte_val(pte);
156 return native_make_pte(v & ~clear);
159 static inline pte_t pte_mkclean(pte_t pte)
161 return pte_clear_flags(pte, _PAGE_DIRTY);
164 static inline pte_t pte_mkold(pte_t pte)
166 return pte_clear_flags(pte, _PAGE_ACCESSED);
169 static inline pte_t pte_wrprotect(pte_t pte)
171 return pte_clear_flags(pte, _PAGE_RW);
174 static inline pte_t pte_mkexec(pte_t pte)
176 return pte_clear_flags(pte, _PAGE_NX);
179 static inline pte_t pte_mkdirty(pte_t pte)
181 return pte_set_flags(pte, _PAGE_DIRTY);
184 static inline pte_t pte_mkyoung(pte_t pte)
186 return pte_set_flags(pte, _PAGE_ACCESSED);
189 static inline pte_t pte_mkwrite(pte_t pte)
191 return pte_set_flags(pte, _PAGE_RW);
194 static inline pte_t pte_mkhuge(pte_t pte)
196 return pte_set_flags(pte, _PAGE_PSE);
199 static inline pte_t pte_clrhuge(pte_t pte)
201 return pte_clear_flags(pte, _PAGE_PSE);
204 static inline pte_t pte_mkglobal(pte_t pte)
206 return pte_set_flags(pte, _PAGE_GLOBAL);
209 static inline pte_t pte_clrglobal(pte_t pte)
211 return pte_clear_flags(pte, _PAGE_GLOBAL);
214 static inline pte_t pte_mkspecial(pte_t pte)
216 return pte_set_flags(pte, _PAGE_SPECIAL);
220 * Mask out unsupported bits in a present pgprot. Non-present pgprots
221 * can use those bits for other purposes, so leave them be.
223 static inline pgprotval_t massage_pgprot(pgprot_t pgprot)
225 pgprotval_t protval = pgprot_val(pgprot);
227 if (protval & _PAGE_PRESENT)
228 protval &= __supported_pte_mask;
233 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
235 return __pte(((phys_addr_t)page_nr << PAGE_SHIFT) |
236 massage_pgprot(pgprot));
239 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
241 return __pmd(((phys_addr_t)page_nr << PAGE_SHIFT) |
242 massage_pgprot(pgprot));
245 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
247 pteval_t val = pte_val(pte);
250 * Chop off the NX bit (if present), and add the NX portion of
251 * the newprot (if present):
253 val &= _PAGE_CHG_MASK;
254 val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
259 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
260 #define pgprot_modify pgprot_modify
261 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
263 pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
264 pgprotval_t addbits = pgprot_val(newprot);
265 return __pgprot(preservebits | addbits);
268 #define pte_pgprot(x) __pgprot(pte_flags(x) & PTE_FLAGS_MASK)
270 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
272 static inline int is_new_memtype_allowed(unsigned long flags,
273 unsigned long new_flags)
276 * Certain new memtypes are not allowed with certain
278 * - request is uncached, return cannot be write-back
279 * - request is write-combine, return cannot be write-back
281 if ((flags == _PAGE_CACHE_UC_MINUS &&
282 new_flags == _PAGE_CACHE_WB) ||
283 (flags == _PAGE_CACHE_WC &&
284 new_flags == _PAGE_CACHE_WB)) {
291 pmd_t *populate_extra_pmd(unsigned long vaddr);
292 pte_t *populate_extra_pte(unsigned long vaddr);
293 #endif /* __ASSEMBLY__ */
296 # include "pgtable_32.h"
298 # include "pgtable_64.h"
302 #include <linux/mm_types.h>
304 static inline int pte_none(pte_t pte)
309 #define __HAVE_ARCH_PTE_SAME
310 static inline int pte_same(pte_t a, pte_t b)
312 return a.pte == b.pte;
315 static inline int pte_present(pte_t a)
317 return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
320 static inline int pte_hidden(pte_t x)
322 #ifdef CONFIG_KMEMCHECK
323 return pte_flags(x) & _PAGE_HIDDEN;
329 static inline int pmd_present(pmd_t pmd)
331 return pmd_flags(pmd) & _PAGE_PRESENT;
334 static inline int pmd_none(pmd_t pmd)
336 /* Only check low word on 32-bit platforms, since it might be
337 out of sync with upper half. */
338 return (unsigned long)native_pmd_val(pmd) == 0;
341 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
343 return (unsigned long)__va(pmd_val(pmd) & PTE_PFN_MASK);
347 * Currently stuck as a macro due to indirect forward reference to
348 * linux/mmzone.h's __section_mem_map_addr() definition:
350 #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
353 * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
355 * this macro returns the index of the entry in the pmd page which would
356 * control the given virtual address
358 static inline unsigned pmd_index(unsigned long address)
360 return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
364 * Conversion functions: convert a page and protection to a page entry,
365 * and a page entry and page directory to the page they refer to.
367 * (Currently stuck as a macro because of indirect forward reference
368 * to linux/mm.h:page_to_nid())
370 #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
373 * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
375 * this function returns the index of the entry in the pte page which would
376 * control the given virtual address
378 static inline unsigned pte_index(unsigned long address)
380 return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
383 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
385 return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
388 static inline int pmd_bad(pmd_t pmd)
390 return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
393 static inline unsigned long pages_to_mb(unsigned long npg)
395 return npg >> (20 - PAGE_SHIFT);
398 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
399 remap_pfn_range(vma, vaddr, pfn, size, prot)
401 #if PAGETABLE_LEVELS > 2
402 static inline int pud_none(pud_t pud)
404 return native_pud_val(pud) == 0;
407 static inline int pud_present(pud_t pud)
409 return pud_flags(pud) & _PAGE_PRESENT;
412 static inline unsigned long pud_page_vaddr(pud_t pud)
414 return (unsigned long)__va((unsigned long)pud_val(pud) & PTE_PFN_MASK);
418 * Currently stuck as a macro due to indirect forward reference to
419 * linux/mmzone.h's __section_mem_map_addr() definition:
421 #define pud_page(pud) pfn_to_page(pud_val(pud) >> PAGE_SHIFT)
423 /* Find an entry in the second-level page table.. */
424 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
426 return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
429 static inline unsigned long pmd_pfn(pmd_t pmd)
431 return (pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT;
434 static inline int pud_large(pud_t pud)
436 return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
437 (_PAGE_PSE | _PAGE_PRESENT);
440 static inline int pud_bad(pud_t pud)
442 return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
445 static inline int pud_large(pud_t pud)
449 #endif /* PAGETABLE_LEVELS > 2 */
451 #if PAGETABLE_LEVELS > 3
452 static inline int pgd_present(pgd_t pgd)
454 return pgd_flags(pgd) & _PAGE_PRESENT;
457 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
459 return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
463 * Currently stuck as a macro due to indirect forward reference to
464 * linux/mmzone.h's __section_mem_map_addr() definition:
466 #define pgd_page(pgd) pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
468 /* to find an entry in a page-table-directory. */
469 static inline unsigned pud_index(unsigned long address)
471 return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
474 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
476 return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
479 static inline int pgd_bad(pgd_t pgd)
481 return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
484 static inline int pgd_none(pgd_t pgd)
486 return !native_pgd_val(pgd);
488 #endif /* PAGETABLE_LEVELS > 3 */
490 #endif /* __ASSEMBLY__ */
493 * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
495 * this macro returns the index of the entry in the pgd page which would
496 * control the given virtual address
498 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
501 * pgd_offset() returns a (pgd_t *)
502 * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
504 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
506 * a shortcut which implies the use of the kernel's pgd, instead
509 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
512 #define KERNEL_PGD_BOUNDARY pgd_index(PAGE_OFFSET)
513 #define KERNEL_PGD_PTRS (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
517 extern int direct_gbpages;
519 /* local pte updates need not use xchg for locking */
520 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
524 /* Pure native function needs no input for mm, addr */
525 native_pte_clear(NULL, 0, ptep);
529 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
530 pte_t *ptep , pte_t pte)
532 native_set_pte(ptep, pte);
535 #ifndef CONFIG_PARAVIRT
537 * Rules for using pte_update - it must be called after any PTE update which
538 * has not been done using the set_pte / clear_pte interfaces. It is used by
539 * shadow mode hypervisors to resynchronize the shadow page tables. Kernel PTE
540 * updates should either be sets, clears, or set_pte_atomic for P->P
541 * transitions, which means this hook should only be called for user PTEs.
542 * This hook implies a P->P protection or access change has taken place, which
543 * requires a subsequent TLB flush. The notification can optionally be delayed
544 * until the TLB flush event by using the pte_update_defer form of the
545 * interface, but care must be taken to assure that the flush happens while
546 * still holding the same page table lock so that the shadow and primary pages
547 * do not become out of sync on SMP.
549 #define pte_update(mm, addr, ptep) do { } while (0)
550 #define pte_update_defer(mm, addr, ptep) do { } while (0)
554 * We only update the dirty/accessed state if we set
555 * the dirty bit by hand in the kernel, since the hardware
556 * will do the accessed bit for us, and we don't want to
557 * race with other CPU's that might be updating the dirty
558 * bit at the same time.
560 struct vm_area_struct;
562 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
563 extern int ptep_set_access_flags(struct vm_area_struct *vma,
564 unsigned long address, pte_t *ptep,
565 pte_t entry, int dirty);
567 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
568 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
569 unsigned long addr, pte_t *ptep);
571 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
572 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
573 unsigned long address, pte_t *ptep);
575 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
576 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
579 pte_t pte = native_ptep_get_and_clear(ptep);
580 pte_update(mm, addr, ptep);
584 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
585 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
586 unsigned long addr, pte_t *ptep,
592 * Full address destruction in progress; paravirt does not
593 * care about updates and native needs no locking
595 pte = native_local_ptep_get_and_clear(ptep);
597 pte = ptep_get_and_clear(mm, addr, ptep);
602 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
603 static inline void ptep_set_wrprotect(struct mm_struct *mm,
604 unsigned long addr, pte_t *ptep)
606 clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
607 pte_update(mm, addr, ptep);
611 * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
613 * dst - pointer to pgd range anwhere on a pgd page
615 * count - the number of pgds to copy.
617 * dst and src can be on the same page, but the range must not overlap,
618 * and must not cross a page boundary.
620 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
622 memcpy(dst, src, count * sizeof(pgd_t));
626 #include <asm-generic/pgtable.h>
627 #endif /* __ASSEMBLY__ */
629 #endif /* _ASM_X86_PGTABLE_H */