1 #ifndef _ASM_X86_PGTABLE_3LEVEL_H
2 #define _ASM_X86_PGTABLE_3LEVEL_H
5 * Intel Physical Address Extension (PAE) Mode - three-level page
6 * tables on PPro+ CPUs.
8 * Copyright (C) 1999 Ingo Molnar <mingo@redhat.com>
11 #define pte_ERROR(e) \
12 printk("%s:%d: bad pte %p(%08lx%08lx).\n", \
13 __FILE__, __LINE__, &(e), (e).pte_high, (e).pte_low)
14 #define pmd_ERROR(e) \
15 printk("%s:%d: bad pmd %p(%016Lx).\n", \
16 __FILE__, __LINE__, &(e), pmd_val(e))
17 #define pgd_ERROR(e) \
18 printk("%s:%d: bad pgd %p(%016Lx).\n", \
19 __FILE__, __LINE__, &(e), pgd_val(e))
21 static inline int pud_none(pud_t pud
)
23 return pud_val(pud
) == 0;
26 static inline int pud_bad(pud_t pud
)
28 return (pud_val(pud
) & ~(PTE_PFN_MASK
| _KERNPG_TABLE
| _PAGE_USER
)) != 0;
31 /* Rules for using set_pte: the pte being assigned *must* be
32 * either not present or in a state where the hardware will
33 * not attempt to update the pte. In places where this is
34 * not possible, use pte_get_and_clear to obtain the old pte
35 * value and then use set_pte to update it. -ben
37 static inline void native_set_pte(pte_t
*ptep
, pte_t pte
)
39 ptep
->pte_high
= pte
.pte_high
;
41 ptep
->pte_low
= pte
.pte_low
;
45 * Since this is only called on user PTEs, and the page fault handler
46 * must handle the already racy situation of simultaneous page faults,
47 * we are justified in merely clearing the PTE present bit, followed
48 * by a set. The ordering here is important.
50 static inline void native_set_pte_present(struct mm_struct
*mm
,
52 pte_t
*ptep
, pte_t pte
)
56 ptep
->pte_high
= pte
.pte_high
;
58 ptep
->pte_low
= pte
.pte_low
;
61 static inline void native_set_pte_atomic(pte_t
*ptep
, pte_t pte
)
63 set_64bit((unsigned long long *)(ptep
), native_pte_val(pte
));
66 static inline void native_set_pmd(pmd_t
*pmdp
, pmd_t pmd
)
68 set_64bit((unsigned long long *)(pmdp
), native_pmd_val(pmd
));
71 static inline void native_set_pud(pud_t
*pudp
, pud_t pud
)
73 set_64bit((unsigned long long *)(pudp
), native_pud_val(pud
));
77 * For PTEs and PDEs, we must clear the P-bit first when clearing a page table
78 * entry, so clear the bottom half first and enforce ordering with a compiler
81 static inline void native_pte_clear(struct mm_struct
*mm
, unsigned long addr
,
89 static inline void native_pmd_clear(pmd_t
*pmd
)
91 u32
*tmp
= (u32
*)pmd
;
97 static inline void pud_clear(pud_t
*pudp
)
101 set_pud(pudp
, __pud(0));
104 * According to Intel App note "TLBs, Paging-Structure Caches,
105 * and Their Invalidation", April 2007, document 317080-001,
106 * section 8.1: in PAE mode we explicitly have to flush the
107 * TLB via cr3 if the top-level pgd is changed...
109 * Make sure the pud entry we're updating is within the
110 * current pgd to avoid unnecessary TLB flushes.
113 if (__pa(pudp
) >= pgd
&& __pa(pudp
) <
114 (pgd
+ sizeof(pgd_t
)*PTRS_PER_PGD
))
118 /* Find an entry in the second-level page table.. */
119 #define pmd_offset(pud, address) ((pmd_t *)pud_page_vaddr(*(pud)) + \
123 static inline pte_t
native_ptep_get_and_clear(pte_t
*ptep
)
127 /* xchg acts as a barrier before the setting of the high bits */
128 res
.pte_low
= xchg(&ptep
->pte_low
, 0);
129 res
.pte_high
= ptep
->pte_high
;
135 #define native_ptep_get_and_clear(xp) native_local_ptep_get_and_clear(xp)
139 * Bits 0, 6 and 7 are taken in the low part of the pte,
140 * put the 32 bits of offset into the high part.
142 #define pte_to_pgoff(pte) ((pte).pte_high)
143 #define pgoff_to_pte(off) \
144 ((pte_t) { { .pte_low = _PAGE_FILE, .pte_high = (off) } })
145 #define PTE_FILE_MAX_BITS 32
147 /* Encode and de-code a swap entry */
148 #define MAX_SWAPFILES_CHECK() BUILD_BUG_ON(MAX_SWAPFILES_SHIFT > 5)
149 #define __swp_type(x) (((x).val) & 0x1f)
150 #define __swp_offset(x) ((x).val >> 5)
151 #define __swp_entry(type, offset) ((swp_entry_t){(type) | (offset) << 5})
152 #define __pte_to_swp_entry(pte) ((swp_entry_t){ (pte).pte_high })
153 #define __swp_entry_to_pte(x) ((pte_t){ { .pte_high = (x).val } })
155 #endif /* _ASM_X86_PGTABLE_3LEVEL_H */