]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/mm/mincore.c | |
4 | * | |
2f77d107 | 5 | * Copyright (C) 1994-2006 Linus Torvalds |
1da177e4 LT |
6 | */ |
7 | ||
8 | /* | |
9 | * The mincore() system call. | |
10 | */ | |
1da177e4 | 11 | #include <linux/pagemap.h> |
5a0e3ad6 | 12 | #include <linux/gfp.h> |
a520110e | 13 | #include <linux/pagewalk.h> |
1da177e4 LT |
14 | #include <linux/mman.h> |
15 | #include <linux/syscalls.h> | |
42da9cbd NP |
16 | #include <linux/swap.h> |
17 | #include <linux/swapops.h> | |
3a4f8a0b | 18 | #include <linux/shmem_fs.h> |
4f16fc10 | 19 | #include <linux/hugetlb.h> |
65fddcfc | 20 | #include <linux/pgtable.h> |
1da177e4 | 21 | |
7c0f6ba6 | 22 | #include <linux/uaccess.h> |
014bb1de | 23 | #include "swap.h" |
1da177e4 | 24 | |
1e25a271 NH |
25 | static int mincore_hugetlb(pte_t *pte, unsigned long hmask, unsigned long addr, |
26 | unsigned long end, struct mm_walk *walk) | |
f4884010 JW |
27 | { |
28 | #ifdef CONFIG_HUGETLB_PAGE | |
1e25a271 NH |
29 | unsigned char present; |
30 | unsigned char *vec = walk->private; | |
f4884010 | 31 | |
1e25a271 NH |
32 | /* |
33 | * Hugepages under user process are always in RAM and never | |
34 | * swapped out, but theoretically it needs to be checked. | |
35 | */ | |
36 | present = pte && !huge_pte_none(huge_ptep_get(pte)); | |
37 | for (; addr != end; vec++, addr += PAGE_SIZE) | |
38 | *vec = present; | |
39 | walk->private = vec; | |
f4884010 JW |
40 | #else |
41 | BUG(); | |
42 | #endif | |
1e25a271 | 43 | return 0; |
f4884010 JW |
44 | } |
45 | ||
30bac164 LT |
46 | /* |
47 | * Later we can get more picky about what "in core" means precisely. | |
48 | * For now, simply check to see if the page is in the page cache, | |
49 | * and is up to date; i.e. that no page-in operation would be required | |
50 | * at this time if an application were to map and access this page. | |
51 | */ | |
61ef1865 | 52 | static unsigned char mincore_page(struct address_space *mapping, pgoff_t index) |
30bac164 LT |
53 | { |
54 | unsigned char present = 0; | |
524984ff | 55 | struct folio *folio; |
30bac164 LT |
56 | |
57 | /* | |
58 | * When tmpfs swaps out a page from a file, any process mapping that | |
59 | * file will not get a swp_entry_t in its pte, but rather it is like | |
60 | * any other file mapping (ie. marked !present and faulted in with | |
61 | * tmpfs's .fault). So swapped out tmpfs mappings are tested here. | |
62 | */ | |
524984ff MWO |
63 | folio = filemap_get_incore_folio(mapping, index); |
64 | if (folio) { | |
65 | present = folio_test_uptodate(folio); | |
66 | folio_put(folio); | |
30bac164 LT |
67 | } |
68 | ||
69 | return present; | |
70 | } | |
71 | ||
72 | static int __mincore_unmapped_range(unsigned long addr, unsigned long end, | |
73 | struct vm_area_struct *vma, unsigned char *vec) | |
1e25a271 | 74 | { |
574823bf | 75 | unsigned long nr = (end - addr) >> PAGE_SHIFT; |
30bac164 | 76 | int i; |
574823bf | 77 | |
30bac164 LT |
78 | if (vma->vm_file) { |
79 | pgoff_t pgoff; | |
80 | ||
81 | pgoff = linear_page_index(vma, addr); | |
82 | for (i = 0; i < nr; i++, pgoff++) | |
83 | vec[i] = mincore_page(vma->vm_file->f_mapping, pgoff); | |
84 | } else { | |
85 | for (i = 0; i < nr; i++) | |
86 | vec[i] = 0; | |
87 | } | |
88 | return nr; | |
89 | } | |
90 | ||
91 | static int mincore_unmapped_range(unsigned long addr, unsigned long end, | |
b7a16c7a | 92 | __always_unused int depth, |
30bac164 LT |
93 | struct mm_walk *walk) |
94 | { | |
95 | walk->private += __mincore_unmapped_range(addr, end, | |
96 | walk->vma, walk->private); | |
1e25a271 | 97 | return 0; |
f4884010 JW |
98 | } |
99 | ||
1e25a271 NH |
100 | static int mincore_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end, |
101 | struct mm_walk *walk) | |
f4884010 JW |
102 | { |
103 | spinlock_t *ptl; | |
1e25a271 | 104 | struct vm_area_struct *vma = walk->vma; |
f4884010 | 105 | pte_t *ptep; |
1e25a271 NH |
106 | unsigned char *vec = walk->private; |
107 | int nr = (end - addr) >> PAGE_SHIFT; | |
108 | ||
b6ec57f4 KS |
109 | ptl = pmd_trans_huge_lock(pmd, vma); |
110 | if (ptl) { | |
1e25a271 NH |
111 | memset(vec, 1, nr); |
112 | spin_unlock(ptl); | |
113 | goto out; | |
114 | } | |
f4884010 | 115 | |
1e25a271 | 116 | if (pmd_trans_unstable(pmd)) { |
30bac164 | 117 | __mincore_unmapped_range(addr, end, vma, vec); |
1e25a271 NH |
118 | goto out; |
119 | } | |
120 | ||
121 | ptep = pte_offset_map_lock(walk->mm, pmd, addr, &ptl); | |
122 | for (; addr != end; ptep++, addr += PAGE_SIZE) { | |
f4884010 | 123 | pte_t pte = *ptep; |
f4884010 | 124 | |
5c041f5d PX |
125 | /* We need to do cache lookup too for pte markers */ |
126 | if (pte_none_mostly(pte)) | |
30bac164 LT |
127 | __mincore_unmapped_range(addr, addr + PAGE_SIZE, |
128 | vma, vec); | |
f4884010 | 129 | else if (pte_present(pte)) |
25ef0e50 | 130 | *vec = 1; |
0661a336 | 131 | else { /* pte is a swap entry */ |
f4884010 JW |
132 | swp_entry_t entry = pte_to_swp_entry(pte); |
133 | ||
30bac164 LT |
134 | if (non_swap_entry(entry)) { |
135 | /* | |
136 | * migration or hwpoison entries are always | |
137 | * uptodate | |
138 | */ | |
139 | *vec = 1; | |
140 | } else { | |
141 | #ifdef CONFIG_SWAP | |
142 | *vec = mincore_page(swap_address_space(entry), | |
143 | swp_offset(entry)); | |
144 | #else | |
145 | WARN_ON(1); | |
146 | *vec = 1; | |
147 | #endif | |
148 | } | |
f4884010 | 149 | } |
25ef0e50 | 150 | vec++; |
1e25a271 | 151 | } |
f4884010 | 152 | pte_unmap_unlock(ptep - 1, ptl); |
1e25a271 NH |
153 | out: |
154 | walk->private += nr; | |
155 | cond_resched(); | |
156 | return 0; | |
e48293fd JW |
157 | } |
158 | ||
134fca90 JK |
159 | static inline bool can_do_mincore(struct vm_area_struct *vma) |
160 | { | |
161 | if (vma_is_anonymous(vma)) | |
162 | return true; | |
163 | if (!vma->vm_file) | |
164 | return false; | |
165 | /* | |
166 | * Reveal pagecache information only for non-anonymous mappings that | |
167 | * correspond to the files the calling process could (if tried) open | |
168 | * for writing; otherwise we'd be including shared non-exclusive | |
169 | * mappings, which opens a side channel. | |
170 | */ | |
01beba79 | 171 | return inode_owner_or_capable(&nop_mnt_idmap, |
21cb47be | 172 | file_inode(vma->vm_file)) || |
02f92b38 | 173 | file_permission(vma->vm_file, MAY_WRITE) == 0; |
134fca90 JK |
174 | } |
175 | ||
7b86ac33 CH |
176 | static const struct mm_walk_ops mincore_walk_ops = { |
177 | .pmd_entry = mincore_pte_range, | |
178 | .pte_hole = mincore_unmapped_range, | |
179 | .hugetlb_entry = mincore_hugetlb, | |
180 | }; | |
181 | ||
2f77d107 LT |
182 | /* |
183 | * Do a chunk of "sys_mincore()". We've already checked | |
184 | * all the arguments, we hold the mmap semaphore: we should | |
185 | * just return the amount of info we're asked for. | |
186 | */ | |
6a60f1b3 | 187 | static long do_mincore(unsigned long addr, unsigned long pages, unsigned char *vec) |
1da177e4 | 188 | { |
6a60f1b3 | 189 | struct vm_area_struct *vma; |
25ef0e50 | 190 | unsigned long end; |
1e25a271 | 191 | int err; |
1da177e4 | 192 | |
97955f69 DW |
193 | vma = vma_lookup(current->mm, addr); |
194 | if (!vma) | |
4fb23e43 | 195 | return -ENOMEM; |
25ef0e50 | 196 | end = min(vma->vm_end, addr + (pages << PAGE_SHIFT)); |
134fca90 JK |
197 | if (!can_do_mincore(vma)) { |
198 | unsigned long pages = DIV_ROUND_UP(end - addr, PAGE_SIZE); | |
199 | memset(vec, 1, pages); | |
200 | return pages; | |
201 | } | |
7b86ac33 | 202 | err = walk_page_range(vma->vm_mm, addr, end, &mincore_walk_ops, vec); |
1e25a271 NH |
203 | if (err < 0) |
204 | return err; | |
25ef0e50 | 205 | return (end - addr) >> PAGE_SHIFT; |
1da177e4 LT |
206 | } |
207 | ||
208 | /* | |
209 | * The mincore(2) system call. | |
210 | * | |
211 | * mincore() returns the memory residency status of the pages in the | |
212 | * current process's address space specified by [addr, addr + len). | |
213 | * The status is returned in a vector of bytes. The least significant | |
214 | * bit of each byte is 1 if the referenced page is in memory, otherwise | |
215 | * it is zero. | |
216 | * | |
217 | * Because the status of a page can change after mincore() checks it | |
218 | * but before it returns to the application, the returned vector may | |
219 | * contain stale information. Only locked pages are guaranteed to | |
220 | * remain in memory. | |
221 | * | |
222 | * return values: | |
223 | * zero - success | |
224 | * -EFAULT - vec points to an illegal address | |
ea1754a0 | 225 | * -EINVAL - addr is not a multiple of PAGE_SIZE |
1da177e4 LT |
226 | * -ENOMEM - Addresses in the range [addr, addr + len] are |
227 | * invalid for the address space of this process, or | |
228 | * specify one or more pages which are not currently | |
229 | * mapped | |
230 | * -EAGAIN - A kernel resource was temporarily unavailable. | |
231 | */ | |
3480b257 HC |
232 | SYSCALL_DEFINE3(mincore, unsigned long, start, size_t, len, |
233 | unsigned char __user *, vec) | |
1da177e4 | 234 | { |
2f77d107 LT |
235 | long retval; |
236 | unsigned long pages; | |
237 | unsigned char *tmp; | |
1da177e4 | 238 | |
057d3389 AK |
239 | start = untagged_addr(start); |
240 | ||
2f77d107 | 241 | /* Check the start address: needs to be page-aligned.. */ |
09cbfeaf | 242 | if (start & ~PAGE_MASK) |
2f77d107 | 243 | return -EINVAL; |
1da177e4 | 244 | |
2f77d107 | 245 | /* ..and we need to be passed a valid user-space range */ |
96d4f267 | 246 | if (!access_ok((void __user *) start, len)) |
2f77d107 | 247 | return -ENOMEM; |
1da177e4 | 248 | |
ea1754a0 | 249 | /* This also avoids any overflows on PAGE_ALIGN */ |
2f77d107 | 250 | pages = len >> PAGE_SHIFT; |
e7bbdd07 | 251 | pages += (offset_in_page(len)) != 0; |
1da177e4 | 252 | |
96d4f267 | 253 | if (!access_ok(vec, pages)) |
2f77d107 | 254 | return -EFAULT; |
1da177e4 | 255 | |
2f77d107 LT |
256 | tmp = (void *) __get_free_page(GFP_USER); |
257 | if (!tmp) | |
4fb23e43 | 258 | return -EAGAIN; |
2f77d107 LT |
259 | |
260 | retval = 0; | |
261 | while (pages) { | |
262 | /* | |
263 | * Do at most PAGE_SIZE entries per iteration, due to | |
264 | * the temporary buffer size. | |
265 | */ | |
d8ed45c5 | 266 | mmap_read_lock(current->mm); |
6a60f1b3 | 267 | retval = do_mincore(start, min(pages, PAGE_SIZE), tmp); |
d8ed45c5 | 268 | mmap_read_unlock(current->mm); |
2f77d107 LT |
269 | |
270 | if (retval <= 0) | |
271 | break; | |
272 | if (copy_to_user(vec, tmp, retval)) { | |
273 | retval = -EFAULT; | |
274 | break; | |
1da177e4 | 275 | } |
2f77d107 LT |
276 | pages -= retval; |
277 | vec += retval; | |
278 | start += retval << PAGE_SHIFT; | |
279 | retval = 0; | |
1da177e4 | 280 | } |
2f77d107 LT |
281 | free_page((unsigned long) tmp); |
282 | return retval; | |
1da177e4 | 283 | } |