pgtable-3level.h 4.3 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157
  1. #ifndef _ASM_X86_PGTABLE_3LEVEL_H
  2. #define _ASM_X86_PGTABLE_3LEVEL_H
  3. /*
  4. * Intel Physical Address Extension (PAE) Mode - three-level page
  5. * tables on PPro+ CPUs.
  6. *
  7. * Copyright (C) 1999 Ingo Molnar <mingo@redhat.com>
  8. */
  9. #define pte_ERROR(e) \
  10. printk("%s:%d: bad pte %p(%08lx%08lx).\n", \
  11. __FILE__, __LINE__, &(e), (e).pte_high, (e).pte_low)
  12. #define pmd_ERROR(e) \
  13. printk("%s:%d: bad pmd %p(%016Lx).\n", \
  14. __FILE__, __LINE__, &(e), pmd_val(e))
  15. #define pgd_ERROR(e) \
  16. printk("%s:%d: bad pgd %p(%016Lx).\n", \
  17. __FILE__, __LINE__, &(e), pgd_val(e))
  18. static inline int pud_none(pud_t pud)
  19. {
  20. return pud_val(pud) == 0;
  21. }
  22. static inline int pud_bad(pud_t pud)
  23. {
  24. return (pud_val(pud) & ~(PTE_PFN_MASK | _KERNPG_TABLE | _PAGE_USER)) != 0;
  25. }
  26. /* Rules for using set_pte: the pte being assigned *must* be
  27. * either not present or in a state where the hardware will
  28. * not attempt to update the pte. In places where this is
  29. * not possible, use pte_get_and_clear to obtain the old pte
  30. * value and then use set_pte to update it. -ben
  31. */
  32. static inline void native_set_pte(pte_t *ptep, pte_t pte)
  33. {
  34. ptep->pte_high = pte.pte_high;
  35. smp_wmb();
  36. ptep->pte_low = pte.pte_low;
  37. }
  38. /*
  39. * Since this is only called on user PTEs, and the page fault handler
  40. * must handle the already racy situation of simultaneous page faults,
  41. * we are justified in merely clearing the PTE present bit, followed
  42. * by a set. The ordering here is important.
  43. */
  44. static inline void native_set_pte_present(struct mm_struct *mm,
  45. unsigned long addr,
  46. pte_t *ptep, pte_t pte)
  47. {
  48. ptep->pte_low = 0;
  49. smp_wmb();
  50. ptep->pte_high = pte.pte_high;
  51. smp_wmb();
  52. ptep->pte_low = pte.pte_low;
  53. }
  54. static inline void native_set_pte_atomic(pte_t *ptep, pte_t pte)
  55. {
  56. set_64bit((unsigned long long *)(ptep), native_pte_val(pte));
  57. }
  58. static inline void native_set_pmd(pmd_t *pmdp, pmd_t pmd)
  59. {
  60. set_64bit((unsigned long long *)(pmdp), native_pmd_val(pmd));
  61. }
  62. static inline void native_set_pud(pud_t *pudp, pud_t pud)
  63. {
  64. set_64bit((unsigned long long *)(pudp), native_pud_val(pud));
  65. }
  66. /*
  67. * For PTEs and PDEs, we must clear the P-bit first when clearing a page table
  68. * entry, so clear the bottom half first and enforce ordering with a compiler
  69. * barrier.
  70. */
  71. static inline void native_pte_clear(struct mm_struct *mm, unsigned long addr,
  72. pte_t *ptep)
  73. {
  74. ptep->pte_low = 0;
  75. smp_wmb();
  76. ptep->pte_high = 0;
  77. }
  78. static inline void native_pmd_clear(pmd_t *pmd)
  79. {
  80. u32 *tmp = (u32 *)pmd;
  81. *tmp = 0;
  82. smp_wmb();
  83. *(tmp + 1) = 0;
  84. }
  85. static inline void pud_clear(pud_t *pudp)
  86. {
  87. unsigned long pgd;
  88. set_pud(pudp, __pud(0));
  89. /*
  90. * According to Intel App note "TLBs, Paging-Structure Caches,
  91. * and Their Invalidation", April 2007, document 317080-001,
  92. * section 8.1: in PAE mode we explicitly have to flush the
  93. * TLB via cr3 if the top-level pgd is changed...
  94. *
  95. * Make sure the pud entry we're updating is within the
  96. * current pgd to avoid unnecessary TLB flushes.
  97. */
  98. pgd = read_cr3();
  99. if (__pa(pudp) >= pgd && __pa(pudp) <
  100. (pgd + sizeof(pgd_t)*PTRS_PER_PGD))
  101. write_cr3(pgd);
  102. }
  103. #define pud_page(pud) pfn_to_page(pud_val(pud) >> PAGE_SHIFT)
  104. /* Find an entry in the second-level page table.. */
  105. #define pmd_offset(pud, address) ((pmd_t *)pud_page_vaddr(*(pud)) + \
  106. pmd_index(address))
  107. #ifdef CONFIG_SMP
  108. static inline pte_t native_ptep_get_and_clear(pte_t *ptep)
  109. {
  110. pte_t res;
  111. /* xchg acts as a barrier before the setting of the high bits */
  112. res.pte_low = xchg(&ptep->pte_low, 0);
  113. res.pte_high = ptep->pte_high;
  114. ptep->pte_high = 0;
  115. return res;
  116. }
  117. #else
  118. #define native_ptep_get_and_clear(xp) native_local_ptep_get_and_clear(xp)
  119. #endif
  120. /*
  121. * Bits 0, 6 and 7 are taken in the low part of the pte,
  122. * put the 32 bits of offset into the high part.
  123. */
  124. #define pte_to_pgoff(pte) ((pte).pte_high)
  125. #define pgoff_to_pte(off) \
  126. ((pte_t) { { .pte_low = _PAGE_FILE, .pte_high = (off) } })
  127. #define PTE_FILE_MAX_BITS 32
  128. /* Encode and de-code a swap entry */
  129. #define MAX_SWAPFILES_CHECK() BUILD_BUG_ON(MAX_SWAPFILES_SHIFT > 5)
  130. #define __swp_type(x) (((x).val) & 0x1f)
  131. #define __swp_offset(x) ((x).val >> 5)
  132. #define __swp_entry(type, offset) ((swp_entry_t){(type) | (offset) << 5})
  133. #define __pte_to_swp_entry(pte) ((swp_entry_t){ (pte).pte_high })
  134. #define __swp_entry_to_pte(x) ((pte_t){ { .pte_high = (x).val } })
  135. #endif /* _ASM_X86_PGTABLE_3LEVEL_H */