Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/mm/mincore.c | |
3 | * | |
2f77d107 | 4 | * Copyright (C) 1994-2006 Linus Torvalds |
1da177e4 LT |
5 | */ |
6 | ||
7 | /* | |
8 | * The mincore() system call. | |
9 | */ | |
1da177e4 | 10 | #include <linux/pagemap.h> |
5a0e3ad6 | 11 | #include <linux/gfp.h> |
1da177e4 LT |
12 | #include <linux/mm.h> |
13 | #include <linux/mman.h> | |
14 | #include <linux/syscalls.h> | |
42da9cbd NP |
15 | #include <linux/swap.h> |
16 | #include <linux/swapops.h> | |
4f16fc10 | 17 | #include <linux/hugetlb.h> |
1da177e4 LT |
18 | |
19 | #include <asm/uaccess.h> | |
20 | #include <asm/pgtable.h> | |
21 | ||
f4884010 | 22 | static void mincore_hugetlb_page_range(struct vm_area_struct *vma, |
25ef0e50 | 23 | unsigned long addr, unsigned long end, |
f4884010 JW |
24 | unsigned char *vec) |
25 | { | |
26 | #ifdef CONFIG_HUGETLB_PAGE | |
27 | struct hstate *h; | |
f4884010 | 28 | |
f4884010 JW |
29 | h = hstate_vma(vma); |
30 | while (1) { | |
31 | unsigned char present; | |
32 | pte_t *ptep; | |
33 | /* | |
34 | * Huge pages are always in RAM for now, but | |
35 | * theoretically it needs to be checked. | |
36 | */ | |
37 | ptep = huge_pte_offset(current->mm, | |
38 | addr & huge_page_mask(h)); | |
39 | present = ptep && !huge_pte_none(huge_ptep_get(ptep)); | |
40 | while (1) { | |
25ef0e50 JW |
41 | *vec = present; |
42 | vec++; | |
f4884010 | 43 | addr += PAGE_SIZE; |
25ef0e50 | 44 | if (addr == end) |
f4884010 JW |
45 | return; |
46 | /* check hugepage border */ | |
47 | if (!(addr & ~huge_page_mask(h))) | |
48 | break; | |
49 | } | |
50 | } | |
51 | #else | |
52 | BUG(); | |
53 | #endif | |
54 | } | |
55 | ||
1da177e4 LT |
56 | /* |
57 | * Later we can get more picky about what "in core" means precisely. | |
58 | * For now, simply check to see if the page is in the page cache, | |
59 | * and is up to date; i.e. that no page-in operation would be required | |
60 | * at this time if an application were to map and access this page. | |
61 | */ | |
42da9cbd | 62 | static unsigned char mincore_page(struct address_space *mapping, pgoff_t pgoff) |
1da177e4 LT |
63 | { |
64 | unsigned char present = 0; | |
42da9cbd | 65 | struct page *page; |
1da177e4 | 66 | |
42da9cbd NP |
67 | /* |
68 | * When tmpfs swaps out a page from a file, any process mapping that | |
69 | * file will not get a swp_entry_t in its pte, but rather it is like | |
70 | * any other file mapping (ie. marked !present and faulted in with | |
3c18ddd1 | 71 | * tmpfs's .fault). So swapped out tmpfs mappings are tested here. |
42da9cbd NP |
72 | */ |
73 | page = find_get_page(mapping, pgoff); | |
31475dd6 HD |
74 | #ifdef CONFIG_SWAP |
75 | if (radix_tree_exceptional_entry(page)) { | |
76 | swp_entry_t swap = radix_to_swp_entry(page); | |
77 | page = find_get_page(&swapper_space, swap.val); | |
78 | } | |
79 | #endif | |
1da177e4 LT |
80 | if (page) { |
81 | present = PageUptodate(page); | |
82 | page_cache_release(page); | |
83 | } | |
84 | ||
85 | return present; | |
86 | } | |
87 | ||
f4884010 | 88 | static void mincore_unmapped_range(struct vm_area_struct *vma, |
25ef0e50 | 89 | unsigned long addr, unsigned long end, |
f4884010 JW |
90 | unsigned char *vec) |
91 | { | |
25ef0e50 | 92 | unsigned long nr = (end - addr) >> PAGE_SHIFT; |
f4884010 JW |
93 | int i; |
94 | ||
95 | if (vma->vm_file) { | |
96 | pgoff_t pgoff; | |
97 | ||
98 | pgoff = linear_page_index(vma, addr); | |
99 | for (i = 0; i < nr; i++, pgoff++) | |
100 | vec[i] = mincore_page(vma->vm_file->f_mapping, pgoff); | |
101 | } else { | |
102 | for (i = 0; i < nr; i++) | |
103 | vec[i] = 0; | |
104 | } | |
105 | } | |
106 | ||
107 | static void mincore_pte_range(struct vm_area_struct *vma, pmd_t *pmd, | |
25ef0e50 | 108 | unsigned long addr, unsigned long end, |
f4884010 JW |
109 | unsigned char *vec) |
110 | { | |
25ef0e50 | 111 | unsigned long next; |
f4884010 JW |
112 | spinlock_t *ptl; |
113 | pte_t *ptep; | |
f4884010 JW |
114 | |
115 | ptep = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl); | |
25ef0e50 | 116 | do { |
f4884010 JW |
117 | pte_t pte = *ptep; |
118 | pgoff_t pgoff; | |
119 | ||
25ef0e50 | 120 | next = addr + PAGE_SIZE; |
f4884010 | 121 | if (pte_none(pte)) |
25ef0e50 | 122 | mincore_unmapped_range(vma, addr, next, vec); |
f4884010 | 123 | else if (pte_present(pte)) |
25ef0e50 | 124 | *vec = 1; |
f4884010 JW |
125 | else if (pte_file(pte)) { |
126 | pgoff = pte_to_pgoff(pte); | |
25ef0e50 | 127 | *vec = mincore_page(vma->vm_file->f_mapping, pgoff); |
f4884010 JW |
128 | } else { /* pte is a swap entry */ |
129 | swp_entry_t entry = pte_to_swp_entry(pte); | |
130 | ||
131 | if (is_migration_entry(entry)) { | |
132 | /* migration entries are always uptodate */ | |
25ef0e50 | 133 | *vec = 1; |
f4884010 JW |
134 | } else { |
135 | #ifdef CONFIG_SWAP | |
136 | pgoff = entry.val; | |
25ef0e50 | 137 | *vec = mincore_page(&swapper_space, pgoff); |
f4884010 JW |
138 | #else |
139 | WARN_ON(1); | |
25ef0e50 | 140 | *vec = 1; |
f4884010 JW |
141 | #endif |
142 | } | |
143 | } | |
25ef0e50 JW |
144 | vec++; |
145 | } while (ptep++, addr = next, addr != end); | |
f4884010 JW |
146 | pte_unmap_unlock(ptep - 1, ptl); |
147 | } | |
148 | ||
e48293fd JW |
149 | static void mincore_pmd_range(struct vm_area_struct *vma, pud_t *pud, |
150 | unsigned long addr, unsigned long end, | |
151 | unsigned char *vec) | |
152 | { | |
153 | unsigned long next; | |
154 | pmd_t *pmd; | |
155 | ||
156 | pmd = pmd_offset(pud, addr); | |
157 | do { | |
158 | next = pmd_addr_end(addr, end); | |
0ca1634d JW |
159 | if (pmd_trans_huge(*pmd)) { |
160 | if (mincore_huge_pmd(vma, pmd, addr, next, vec)) { | |
161 | vec += (next - addr) >> PAGE_SHIFT; | |
162 | continue; | |
163 | } | |
164 | /* fall through */ | |
165 | } | |
e48293fd JW |
166 | if (pmd_none_or_clear_bad(pmd)) |
167 | mincore_unmapped_range(vma, addr, next, vec); | |
168 | else | |
169 | mincore_pte_range(vma, pmd, addr, next, vec); | |
170 | vec += (next - addr) >> PAGE_SHIFT; | |
171 | } while (pmd++, addr = next, addr != end); | |
172 | } | |
173 | ||
174 | static void mincore_pud_range(struct vm_area_struct *vma, pgd_t *pgd, | |
175 | unsigned long addr, unsigned long end, | |
176 | unsigned char *vec) | |
177 | { | |
178 | unsigned long next; | |
179 | pud_t *pud; | |
180 | ||
181 | pud = pud_offset(pgd, addr); | |
182 | do { | |
183 | next = pud_addr_end(addr, end); | |
184 | if (pud_none_or_clear_bad(pud)) | |
185 | mincore_unmapped_range(vma, addr, next, vec); | |
186 | else | |
187 | mincore_pmd_range(vma, pud, addr, next, vec); | |
188 | vec += (next - addr) >> PAGE_SHIFT; | |
189 | } while (pud++, addr = next, addr != end); | |
190 | } | |
191 | ||
192 | static void mincore_page_range(struct vm_area_struct *vma, | |
193 | unsigned long addr, unsigned long end, | |
194 | unsigned char *vec) | |
195 | { | |
196 | unsigned long next; | |
197 | pgd_t *pgd; | |
198 | ||
199 | pgd = pgd_offset(vma->vm_mm, addr); | |
200 | do { | |
201 | next = pgd_addr_end(addr, end); | |
202 | if (pgd_none_or_clear_bad(pgd)) | |
203 | mincore_unmapped_range(vma, addr, next, vec); | |
204 | else | |
205 | mincore_pud_range(vma, pgd, addr, next, vec); | |
206 | vec += (next - addr) >> PAGE_SHIFT; | |
207 | } while (pgd++, addr = next, addr != end); | |
208 | } | |
209 | ||
2f77d107 LT |
210 | /* |
211 | * Do a chunk of "sys_mincore()". We've already checked | |
212 | * all the arguments, we hold the mmap semaphore: we should | |
213 | * just return the amount of info we're asked for. | |
214 | */ | |
6a60f1b3 | 215 | static long do_mincore(unsigned long addr, unsigned long pages, unsigned char *vec) |
1da177e4 | 216 | { |
6a60f1b3 | 217 | struct vm_area_struct *vma; |
25ef0e50 | 218 | unsigned long end; |
1da177e4 | 219 | |
6a60f1b3 | 220 | vma = find_vma(current->mm, addr); |
4fb23e43 LT |
221 | if (!vma || addr < vma->vm_start) |
222 | return -ENOMEM; | |
1da177e4 | 223 | |
25ef0e50 | 224 | end = min(vma->vm_end, addr + (pages << PAGE_SHIFT)); |
6a60f1b3 | 225 | |
4f16fc10 | 226 | if (is_vm_hugetlb_page(vma)) { |
25ef0e50 JW |
227 | mincore_hugetlb_page_range(vma, addr, end, vec); |
228 | return (end - addr) >> PAGE_SHIFT; | |
4f16fc10 | 229 | } |
4f16fc10 | 230 | |
25ef0e50 | 231 | end = pmd_addr_end(addr, end); |
1da177e4 | 232 | |
e48293fd JW |
233 | if (is_vm_hugetlb_page(vma)) |
234 | mincore_hugetlb_page_range(vma, addr, end, vec); | |
235 | else | |
236 | mincore_page_range(vma, addr, end, vec); | |
42da9cbd | 237 | |
25ef0e50 | 238 | return (end - addr) >> PAGE_SHIFT; |
1da177e4 LT |
239 | } |
240 | ||
241 | /* | |
242 | * The mincore(2) system call. | |
243 | * | |
244 | * mincore() returns the memory residency status of the pages in the | |
245 | * current process's address space specified by [addr, addr + len). | |
246 | * The status is returned in a vector of bytes. The least significant | |
247 | * bit of each byte is 1 if the referenced page is in memory, otherwise | |
248 | * it is zero. | |
249 | * | |
250 | * Because the status of a page can change after mincore() checks it | |
251 | * but before it returns to the application, the returned vector may | |
252 | * contain stale information. Only locked pages are guaranteed to | |
253 | * remain in memory. | |
254 | * | |
255 | * return values: | |
256 | * zero - success | |
257 | * -EFAULT - vec points to an illegal address | |
258 | * -EINVAL - addr is not a multiple of PAGE_CACHE_SIZE | |
259 | * -ENOMEM - Addresses in the range [addr, addr + len] are | |
260 | * invalid for the address space of this process, or | |
261 | * specify one or more pages which are not currently | |
262 | * mapped | |
263 | * -EAGAIN - A kernel resource was temporarily unavailable. | |
264 | */ | |
3480b257 HC |
265 | SYSCALL_DEFINE3(mincore, unsigned long, start, size_t, len, |
266 | unsigned char __user *, vec) | |
1da177e4 | 267 | { |
2f77d107 LT |
268 | long retval; |
269 | unsigned long pages; | |
270 | unsigned char *tmp; | |
1da177e4 | 271 | |
2f77d107 LT |
272 | /* Check the start address: needs to be page-aligned.. */ |
273 | if (start & ~PAGE_CACHE_MASK) | |
274 | return -EINVAL; | |
1da177e4 | 275 | |
2f77d107 LT |
276 | /* ..and we need to be passed a valid user-space range */ |
277 | if (!access_ok(VERIFY_READ, (void __user *) start, len)) | |
278 | return -ENOMEM; | |
1da177e4 | 279 | |
2f77d107 LT |
280 | /* This also avoids any overflows on PAGE_CACHE_ALIGN */ |
281 | pages = len >> PAGE_SHIFT; | |
282 | pages += (len & ~PAGE_MASK) != 0; | |
1da177e4 | 283 | |
2f77d107 LT |
284 | if (!access_ok(VERIFY_WRITE, vec, pages)) |
285 | return -EFAULT; | |
1da177e4 | 286 | |
2f77d107 LT |
287 | tmp = (void *) __get_free_page(GFP_USER); |
288 | if (!tmp) | |
4fb23e43 | 289 | return -EAGAIN; |
2f77d107 LT |
290 | |
291 | retval = 0; | |
292 | while (pages) { | |
293 | /* | |
294 | * Do at most PAGE_SIZE entries per iteration, due to | |
295 | * the temporary buffer size. | |
296 | */ | |
297 | down_read(¤t->mm->mmap_sem); | |
6a60f1b3 | 298 | retval = do_mincore(start, min(pages, PAGE_SIZE), tmp); |
2f77d107 LT |
299 | up_read(¤t->mm->mmap_sem); |
300 | ||
301 | if (retval <= 0) | |
302 | break; | |
303 | if (copy_to_user(vec, tmp, retval)) { | |
304 | retval = -EFAULT; | |
305 | break; | |
1da177e4 | 306 | } |
2f77d107 LT |
307 | pages -= retval; |
308 | vec += retval; | |
309 | start += retval << PAGE_SHIFT; | |
310 | retval = 0; | |
1da177e4 | 311 | } |
2f77d107 LT |
312 | free_page((unsigned long) tmp); |
313 | return retval; | |
1da177e4 | 314 | } |