blob: 3b6a883d0926dcd900da448872bf391893fdf3f8
1 | /* |
2 | * linux/mm/mincore.c |
3 | * |
4 | * Copyright (C) 1994-2006 Linus Torvalds |
5 | */ |
6 | |
7 | /* |
8 | * The mincore() system call. |
9 | */ |
10 | #include <linux/pagemap.h> |
11 | #include <linux/gfp.h> |
12 | #include <linux/mm.h> |
13 | #include <linux/mman.h> |
14 | #include <linux/syscalls.h> |
15 | #include <linux/swap.h> |
16 | #include <linux/swapops.h> |
17 | #include <linux/hugetlb.h> |
18 | |
19 | #include <asm/uaccess.h> |
20 | #include <asm/pgtable.h> |
21 | |
22 | static int mincore_hugetlb(pte_t *pte, unsigned long hmask, unsigned long addr, |
23 | unsigned long end, struct mm_walk *walk) |
24 | { |
25 | #ifdef CONFIG_HUGETLB_PAGE |
26 | unsigned char present; |
27 | unsigned char *vec = walk->private; |
28 | |
29 | /* |
30 | * Hugepages under user process are always in RAM and never |
31 | * swapped out, but theoretically it needs to be checked. |
32 | */ |
33 | present = pte && !huge_pte_none(huge_ptep_get(pte)); |
34 | for (; addr != end; vec++, addr += PAGE_SIZE) |
35 | *vec = present; |
36 | walk->private = vec; |
37 | #else |
38 | BUG(); |
39 | #endif |
40 | return 0; |
41 | } |
42 | |
43 | /* |
44 | * Later we can get more picky about what "in core" means precisely. |
45 | * For now, simply check to see if the page is in the page cache, |
46 | * and is up to date; i.e. that no page-in operation would be required |
47 | * at this time if an application were to map and access this page. |
48 | */ |
49 | static unsigned char mincore_page(struct address_space *mapping, pgoff_t pgoff) |
50 | { |
51 | unsigned char present = 0; |
52 | struct page *page; |
53 | |
54 | /* |
55 | * When tmpfs swaps out a page from a file, any process mapping that |
56 | * file will not get a swp_entry_t in its pte, but rather it is like |
57 | * any other file mapping (ie. marked !present and faulted in with |
58 | * tmpfs's .fault). So swapped out tmpfs mappings are tested here. |
59 | */ |
60 | #ifdef CONFIG_SWAP |
61 | if (shmem_mapping(mapping)) { |
62 | page = find_get_entry(mapping, pgoff); |
63 | /* |
64 | * shmem/tmpfs may return swap: account for swapcache |
65 | * page too. |
66 | */ |
67 | if (radix_tree_exceptional_entry(page)) { |
68 | swp_entry_t swp = radix_to_swp_entry(page); |
69 | page = find_get_page(swap_address_space(swp), |
70 | swp_offset(swp)); |
71 | } |
72 | } else |
73 | page = find_get_page(mapping, pgoff); |
74 | #else |
75 | page = find_get_page(mapping, pgoff); |
76 | #endif |
77 | if (page) { |
78 | present = PageUptodate(page); |
79 | put_page(page); |
80 | } |
81 | |
82 | return present; |
83 | } |
84 | |
85 | static int __mincore_unmapped_range(unsigned long addr, unsigned long end, |
86 | struct vm_area_struct *vma, unsigned char *vec) |
87 | { |
88 | unsigned long nr = (end - addr) >> PAGE_SHIFT; |
89 | int i; |
90 | |
91 | if (vma->vm_file) { |
92 | pgoff_t pgoff; |
93 | |
94 | pgoff = linear_page_index(vma, addr); |
95 | for (i = 0; i < nr; i++, pgoff++) |
96 | vec[i] = mincore_page(vma->vm_file->f_mapping, pgoff); |
97 | } else { |
98 | for (i = 0; i < nr; i++) |
99 | vec[i] = 0; |
100 | } |
101 | return nr; |
102 | } |
103 | |
104 | static int mincore_unmapped_range(unsigned long addr, unsigned long end, |
105 | struct mm_walk *walk) |
106 | { |
107 | walk->private += __mincore_unmapped_range(addr, end, |
108 | walk->vma, walk->private); |
109 | return 0; |
110 | } |
111 | |
112 | static int mincore_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end, |
113 | struct mm_walk *walk) |
114 | { |
115 | spinlock_t *ptl; |
116 | struct vm_area_struct *vma = walk->vma; |
117 | pte_t *ptep; |
118 | unsigned char *vec = walk->private; |
119 | int nr = (end - addr) >> PAGE_SHIFT; |
120 | |
121 | ptl = pmd_trans_huge_lock(pmd, vma); |
122 | if (ptl) { |
123 | memset(vec, 1, nr); |
124 | spin_unlock(ptl); |
125 | goto out; |
126 | } |
127 | |
128 | if (pmd_trans_unstable(pmd)) { |
129 | __mincore_unmapped_range(addr, end, vma, vec); |
130 | goto out; |
131 | } |
132 | |
133 | ptep = pte_offset_map_lock(walk->mm, pmd, addr, &ptl); |
134 | for (; addr != end; ptep++, addr += PAGE_SIZE) { |
135 | pte_t pte = *ptep; |
136 | |
137 | if (pte_none(pte)) |
138 | __mincore_unmapped_range(addr, addr + PAGE_SIZE, |
139 | vma, vec); |
140 | else if (pte_present(pte)) |
141 | *vec = 1; |
142 | else { /* pte is a swap entry */ |
143 | swp_entry_t entry = pte_to_swp_entry(pte); |
144 | |
145 | if (non_swap_entry(entry)) { |
146 | /* |
147 | * migration or hwpoison entries are always |
148 | * uptodate |
149 | */ |
150 | *vec = 1; |
151 | } else { |
152 | #ifdef CONFIG_SWAP |
153 | *vec = mincore_page(swap_address_space(entry), |
154 | swp_offset(entry)); |
155 | #else |
156 | WARN_ON(1); |
157 | *vec = 1; |
158 | #endif |
159 | } |
160 | } |
161 | vec++; |
162 | } |
163 | pte_unmap_unlock(ptep - 1, ptl); |
164 | out: |
165 | walk->private += nr; |
166 | cond_resched(); |
167 | return 0; |
168 | } |
169 | |
170 | static inline bool can_do_mincore(struct vm_area_struct *vma) |
171 | { |
172 | if (vma_is_anonymous(vma)) |
173 | return true; |
174 | if (!vma->vm_file) |
175 | return false; |
176 | /* |
177 | * Reveal pagecache information only for non-anonymous mappings that |
178 | * correspond to the files the calling process could (if tried) open |
179 | * for writing; otherwise we'd be including shared non-exclusive |
180 | * mappings, which opens a side channel. |
181 | */ |
182 | return inode_owner_or_capable(file_inode(vma->vm_file)) || |
183 | inode_permission(file_inode(vma->vm_file), MAY_WRITE) == 0; |
184 | } |
185 | |
186 | /* |
187 | * Do a chunk of "sys_mincore()". We've already checked |
188 | * all the arguments, we hold the mmap semaphore: we should |
189 | * just return the amount of info we're asked for. |
190 | */ |
191 | static long do_mincore(unsigned long addr, unsigned long pages, unsigned char *vec) |
192 | { |
193 | struct vm_area_struct *vma; |
194 | unsigned long end; |
195 | int err; |
196 | struct mm_walk mincore_walk = { |
197 | .pmd_entry = mincore_pte_range, |
198 | .pte_hole = mincore_unmapped_range, |
199 | .hugetlb_entry = mincore_hugetlb, |
200 | .private = vec, |
201 | }; |
202 | |
203 | vma = find_vma(current->mm, addr); |
204 | if (!vma || addr < vma->vm_start) |
205 | return -ENOMEM; |
206 | end = min(vma->vm_end, addr + (pages << PAGE_SHIFT)); |
207 | if (!can_do_mincore(vma)) { |
208 | unsigned long pages = DIV_ROUND_UP(end - addr, PAGE_SIZE); |
209 | memset(vec, 1, pages); |
210 | return pages; |
211 | } |
212 | mincore_walk.mm = vma->vm_mm; |
213 | err = walk_page_range(addr, end, &mincore_walk); |
214 | if (err < 0) |
215 | return err; |
216 | return (end - addr) >> PAGE_SHIFT; |
217 | } |
218 | |
219 | /* |
220 | * The mincore(2) system call. |
221 | * |
222 | * mincore() returns the memory residency status of the pages in the |
223 | * current process's address space specified by [addr, addr + len). |
224 | * The status is returned in a vector of bytes. The least significant |
225 | * bit of each byte is 1 if the referenced page is in memory, otherwise |
226 | * it is zero. |
227 | * |
228 | * Because the status of a page can change after mincore() checks it |
229 | * but before it returns to the application, the returned vector may |
230 | * contain stale information. Only locked pages are guaranteed to |
231 | * remain in memory. |
232 | * |
233 | * return values: |
234 | * zero - success |
235 | * -EFAULT - vec points to an illegal address |
236 | * -EINVAL - addr is not a multiple of PAGE_SIZE |
237 | * -ENOMEM - Addresses in the range [addr, addr + len] are |
238 | * invalid for the address space of this process, or |
239 | * specify one or more pages which are not currently |
240 | * mapped |
241 | * -EAGAIN - A kernel resource was temporarily unavailable. |
242 | */ |
243 | SYSCALL_DEFINE3(mincore, unsigned long, start, size_t, len, |
244 | unsigned char __user *, vec) |
245 | { |
246 | long retval; |
247 | unsigned long pages; |
248 | unsigned char *tmp; |
249 | |
250 | /* Check the start address: needs to be page-aligned.. */ |
251 | if (start & ~PAGE_MASK) |
252 | return -EINVAL; |
253 | |
254 | /* ..and we need to be passed a valid user-space range */ |
255 | if (!access_ok(VERIFY_READ, (void __user *) start, len)) |
256 | return -ENOMEM; |
257 | |
258 | /* This also avoids any overflows on PAGE_ALIGN */ |
259 | pages = len >> PAGE_SHIFT; |
260 | pages += (offset_in_page(len)) != 0; |
261 | |
262 | if (!access_ok(VERIFY_WRITE, vec, pages)) |
263 | return -EFAULT; |
264 | |
265 | tmp = (void *) __get_free_page(GFP_USER); |
266 | if (!tmp) |
267 | return -EAGAIN; |
268 | |
269 | retval = 0; |
270 | while (pages) { |
271 | /* |
272 | * Do at most PAGE_SIZE entries per iteration, due to |
273 | * the temporary buffer size. |
274 | */ |
275 | down_read(¤t->mm->mmap_sem); |
276 | retval = do_mincore(start, min(pages, PAGE_SIZE), tmp); |
277 | up_read(¤t->mm->mmap_sem); |
278 | |
279 | if (retval <= 0) |
280 | break; |
281 | if (copy_to_user(vec, tmp, retval)) { |
282 | retval = -EFAULT; |
283 | break; |
284 | } |
285 | pages -= retval; |
286 | vec += retval; |
287 | start += retval << PAGE_SHIFT; |
288 | retval = 0; |
289 | } |
290 | free_page((unsigned long) tmp); |
291 | return retval; |
292 | } |
293 |