Commit | Line | Data |
---|---|---|
e6473092 MM |
1 | #include <linux/mm.h> |
2 | #include <linux/highmem.h> | |
3 | #include <linux/sched.h> | |
d33b9f45 | 4 | #include <linux/hugetlb.h> |
e6473092 MM |
5 | |
6 | static int walk_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end, | |
2165009b | 7 | struct mm_walk *walk) |
e6473092 MM |
8 | { |
9 | pte_t *pte; | |
10 | int err = 0; | |
11 | ||
12 | pte = pte_offset_map(pmd, addr); | |
556637cd | 13 | for (;;) { |
2165009b | 14 | err = walk->pte_entry(pte, addr, addr + PAGE_SIZE, walk); |
e6473092 MM |
15 | if (err) |
16 | break; | |
556637cd JW |
17 | addr += PAGE_SIZE; |
18 | if (addr == end) | |
19 | break; | |
20 | pte++; | |
21 | } | |
e6473092 MM |
22 | |
23 | pte_unmap(pte); | |
24 | return err; | |
25 | } | |
26 | ||
27 | static int walk_pmd_range(pud_t *pud, unsigned long addr, unsigned long end, | |
2165009b | 28 | struct mm_walk *walk) |
e6473092 MM |
29 | { |
30 | pmd_t *pmd; | |
31 | unsigned long next; | |
32 | int err = 0; | |
33 | ||
34 | pmd = pmd_offset(pud, addr); | |
35 | do { | |
36 | next = pmd_addr_end(addr, end); | |
37 | if (pmd_none_or_clear_bad(pmd)) { | |
38 | if (walk->pte_hole) | |
2165009b | 39 | err = walk->pte_hole(addr, next, walk); |
e6473092 MM |
40 | if (err) |
41 | break; | |
42 | continue; | |
43 | } | |
44 | if (walk->pmd_entry) | |
2165009b | 45 | err = walk->pmd_entry(pmd, addr, next, walk); |
e6473092 | 46 | if (!err && walk->pte_entry) |
2165009b | 47 | err = walk_pte_range(pmd, addr, next, walk); |
e6473092 MM |
48 | if (err) |
49 | break; | |
50 | } while (pmd++, addr = next, addr != end); | |
51 | ||
52 | return err; | |
53 | } | |
54 | ||
55 | static int walk_pud_range(pgd_t *pgd, unsigned long addr, unsigned long end, | |
2165009b | 56 | struct mm_walk *walk) |
e6473092 MM |
57 | { |
58 | pud_t *pud; | |
59 | unsigned long next; | |
60 | int err = 0; | |
61 | ||
62 | pud = pud_offset(pgd, addr); | |
63 | do { | |
64 | next = pud_addr_end(addr, end); | |
65 | if (pud_none_or_clear_bad(pud)) { | |
66 | if (walk->pte_hole) | |
2165009b | 67 | err = walk->pte_hole(addr, next, walk); |
e6473092 MM |
68 | if (err) |
69 | break; | |
70 | continue; | |
71 | } | |
72 | if (walk->pud_entry) | |
2165009b | 73 | err = walk->pud_entry(pud, addr, next, walk); |
e6473092 | 74 | if (!err && (walk->pmd_entry || walk->pte_entry)) |
2165009b | 75 | err = walk_pmd_range(pud, addr, next, walk); |
e6473092 MM |
76 | if (err) |
77 | break; | |
78 | } while (pud++, addr = next, addr != end); | |
79 | ||
80 | return err; | |
81 | } | |
82 | ||
83 | /** | |
84 | * walk_page_range - walk a memory map's page tables with a callback | |
7682486b RD |
85 | * @mm: memory map to walk |
86 | * @addr: starting address | |
87 | * @end: ending address | |
88 | * @walk: set of callbacks to invoke for each level of the tree | |
e6473092 MM |
89 | * |
90 | * Recursively walk the page table for the memory area in a VMA, | |
91 | * calling supplied callbacks. Callbacks are called in-order (first | |
92 | * PGD, first PUD, first PMD, first PTE, second PTE... second PMD, | |
93 | * etc.). If lower-level callbacks are omitted, walking depth is reduced. | |
94 | * | |
2165009b DH |
95 | * Each callback receives an entry pointer and the start and end of the |
96 | * associated range, and a copy of the original mm_walk for access to | |
97 | * the ->private or ->mm fields. | |
e6473092 MM |
98 | * |
99 | * No locks are taken, but the bottom level iterator will map PTE | |
100 | * directories from highmem if necessary. | |
101 | * | |
102 | * If any callback returns a non-zero value, the walk is aborted and | |
103 | * the return value is propagated back to the caller. Otherwise 0 is returned. | |
104 | */ | |
2165009b DH |
105 | int walk_page_range(unsigned long addr, unsigned long end, |
106 | struct mm_walk *walk) | |
e6473092 MM |
107 | { |
108 | pgd_t *pgd; | |
109 | unsigned long next; | |
110 | int err = 0; | |
d33b9f45 | 111 | struct vm_area_struct *vma; |
e6473092 MM |
112 | |
113 | if (addr >= end) | |
114 | return err; | |
115 | ||
2165009b DH |
116 | if (!walk->mm) |
117 | return -EINVAL; | |
118 | ||
119 | pgd = pgd_offset(walk->mm, addr); | |
e6473092 MM |
120 | do { |
121 | next = pgd_addr_end(addr, end); | |
d33b9f45 | 122 | |
5dc37642 NH |
123 | /* |
124 | * handle hugetlb vma individually because pagetable walk for | |
125 | * the hugetlb page is dependent on the architecture and | |
126 | * we can't handled it in the same manner as non-huge pages. | |
127 | */ | |
d33b9f45 | 128 | vma = find_vma(walk->mm, addr); |
5dc37642 | 129 | #ifdef CONFIG_HUGETLB_PAGE |
d33b9f45 | 130 | if (vma && is_vm_hugetlb_page(vma)) { |
5dc37642 NH |
131 | pte_t *pte; |
132 | struct hstate *hs; | |
133 | ||
d33b9f45 NH |
134 | if (vma->vm_end < next) |
135 | next = vma->vm_end; | |
5dc37642 NH |
136 | hs = hstate_vma(vma); |
137 | pte = huge_pte_offset(walk->mm, | |
138 | addr & huge_page_mask(hs)); | |
139 | if (pte && !huge_pte_none(huge_ptep_get(pte)) | |
140 | && walk->hugetlb_entry) | |
141 | err = walk->hugetlb_entry(pte, addr, | |
142 | next, walk); | |
143 | if (err) | |
144 | break; | |
d33b9f45 NH |
145 | continue; |
146 | } | |
5dc37642 | 147 | #endif |
e6473092 MM |
148 | if (pgd_none_or_clear_bad(pgd)) { |
149 | if (walk->pte_hole) | |
2165009b | 150 | err = walk->pte_hole(addr, next, walk); |
e6473092 MM |
151 | if (err) |
152 | break; | |
d33b9f45 | 153 | pgd++; |
e6473092 MM |
154 | continue; |
155 | } | |
156 | if (walk->pgd_entry) | |
2165009b | 157 | err = walk->pgd_entry(pgd, addr, next, walk); |
e6473092 MM |
158 | if (!err && |
159 | (walk->pud_entry || walk->pmd_entry || walk->pte_entry)) | |
2165009b | 160 | err = walk_pud_range(pgd, addr, next, walk); |
e6473092 MM |
161 | if (err) |
162 | break; | |
d33b9f45 NH |
163 | pgd++; |
164 | } while (addr = next, addr != end); | |
e6473092 MM |
165 | |
166 | return err; | |
167 | } |