blob: a4e6b9d75c76198be4f04d41e996f2067a97db27 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/mm/mincore.c
3 *
Linus Torvalds2f77d102006-12-16 09:44:32 -08004 * Copyright (C) 1994-2006 Linus Torvalds
Linus Torvalds1da177e2005-04-16 15:20:36 -07005 */
6
7/*
8 * The mincore() system call.
9 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070010#include <linux/pagemap.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090011#include <linux/gfp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/mm.h>
13#include <linux/mman.h>
14#include <linux/syscalls.h>
Nick Piggin42da9cb2007-02-12 00:51:39 -080015#include <linux/swap.h>
16#include <linux/swapops.h>
Naoya Horiguchi4f16fc12009-12-14 17:59:58 -080017#include <linux/hugetlb.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018
19#include <asm/uaccess.h>
20#include <asm/pgtable.h>
21
Johannes Weinerf4884012010-05-24 14:32:10 -070022static void mincore_hugetlb_page_range(struct vm_area_struct *vma,
Johannes Weiner25ef0e52010-05-24 14:32:11 -070023 unsigned long addr, unsigned long end,
Johannes Weinerf4884012010-05-24 14:32:10 -070024 unsigned char *vec)
25{
26#ifdef CONFIG_HUGETLB_PAGE
27 struct hstate *h;
Johannes Weinerf4884012010-05-24 14:32:10 -070028
Johannes Weinerf4884012010-05-24 14:32:10 -070029 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) {
Johannes Weiner25ef0e52010-05-24 14:32:11 -070041 *vec = present;
42 vec++;
Johannes Weinerf4884012010-05-24 14:32:10 -070043 addr += PAGE_SIZE;
Johannes Weiner25ef0e52010-05-24 14:32:11 -070044 if (addr == end)
Johannes Weinerf4884012010-05-24 14:32:10 -070045 return;
46 /* check hugepage border */
47 if (!(addr & ~huge_page_mask(h)))
48 break;
49 }
50 }
51#else
52 BUG();
53#endif
54}
55
Linus Torvalds1da177e2005-04-16 15:20:36 -070056/*
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 */
Nick Piggin42da9cb2007-02-12 00:51:39 -080062static unsigned char mincore_page(struct address_space *mapping, pgoff_t pgoff)
Linus Torvalds1da177e2005-04-16 15:20:36 -070063{
64 unsigned char present = 0;
Nick Piggin42da9cb2007-02-12 00:51:39 -080065 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -070066
Nick Piggin42da9cb2007-02-12 00:51:39 -080067 /*
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
Nick Piggin3c18ddd2008-04-28 02:12:10 -070071 * tmpfs's .fault). So swapped out tmpfs mappings are tested here.
Nick Piggin42da9cb2007-02-12 00:51:39 -080072 *
73 * However when tmpfs moves the page from pagecache and into swapcache,
74 * it is still in core, but the find_get_page below won't find it.
75 * No big deal, but make a note of it.
76 */
77 page = find_get_page(mapping, pgoff);
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 if (page) {
79 present = PageUptodate(page);
80 page_cache_release(page);
81 }
82
83 return present;
84}
85
Johannes Weinerf4884012010-05-24 14:32:10 -070086static void mincore_unmapped_range(struct vm_area_struct *vma,
Johannes Weiner25ef0e52010-05-24 14:32:11 -070087 unsigned long addr, unsigned long end,
Johannes Weinerf4884012010-05-24 14:32:10 -070088 unsigned char *vec)
Linus Torvalds1da177e2005-04-16 15:20:36 -070089{
Johannes Weiner25ef0e52010-05-24 14:32:11 -070090 unsigned long nr = (end - addr) >> PAGE_SHIFT;
Nick Piggin42da9cb2007-02-12 00:51:39 -080091 int i;
Linus Torvalds1da177e2005-04-16 15:20:36 -070092
Johannes Weinerf4884012010-05-24 14:32:10 -070093 if (vma->vm_file) {
94 pgoff_t pgoff;
Linus Torvalds2f77d102006-12-16 09:44:32 -080095
Johannes Weinerf4884012010-05-24 14:32:10 -070096 pgoff = linear_page_index(vma, addr);
97 for (i = 0; i < nr; i++, pgoff++)
98 vec[i] = mincore_page(vma->vm_file->f_mapping, pgoff);
99 } else {
100 for (i = 0; i < nr; i++)
101 vec[i] = 0;
Naoya Horiguchi4f16fc12009-12-14 17:59:58 -0800102 }
Johannes Weinerf4884012010-05-24 14:32:10 -0700103}
Naoya Horiguchi4f16fc12009-12-14 17:59:58 -0800104
Johannes Weinerf4884012010-05-24 14:32:10 -0700105static void mincore_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700106 unsigned long addr, unsigned long end,
Johannes Weinerf4884012010-05-24 14:32:10 -0700107 unsigned char *vec)
108{
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700109 unsigned long next;
Johannes Weinerf4884012010-05-24 14:32:10 -0700110 spinlock_t *ptl;
111 pte_t *ptep;
Linus Torvalds2f77d102006-12-16 09:44:32 -0800112
Nick Piggin42da9cb2007-02-12 00:51:39 -0800113 ptep = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700114 do {
Nick Piggin42da9cb2007-02-12 00:51:39 -0800115 pte_t pte = *ptep;
Johannes Weinerf4884012010-05-24 14:32:10 -0700116 pgoff_t pgoff;
Nick Piggin42da9cb2007-02-12 00:51:39 -0800117
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700118 next = addr + PAGE_SIZE;
Johannes Weinerf4884012010-05-24 14:32:10 -0700119 if (pte_none(pte))
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700120 mincore_unmapped_range(vma, addr, next, vec);
Johannes Weinerf4884012010-05-24 14:32:10 -0700121 else if (pte_present(pte))
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700122 *vec = 1;
Johannes Weiner6a60f1b2010-05-24 14:32:09 -0700123 else if (pte_file(pte)) {
Nick Piggin42da9cb2007-02-12 00:51:39 -0800124 pgoff = pte_to_pgoff(pte);
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700125 *vec = mincore_page(vma->vm_file->f_mapping, pgoff);
Nick Piggin42da9cb2007-02-12 00:51:39 -0800126 } else { /* pte is a swap entry */
127 swp_entry_t entry = pte_to_swp_entry(pte);
Johannes Weiner6a60f1b2010-05-24 14:32:09 -0700128
Nick Piggin42da9cb2007-02-12 00:51:39 -0800129 if (is_migration_entry(entry)) {
130 /* migration entries are always uptodate */
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700131 *vec = 1;
Nick Piggin42da9cb2007-02-12 00:51:39 -0800132 } else {
Nick Piggin30fcffe2007-02-14 12:35:02 +0100133#ifdef CONFIG_SWAP
Nick Piggin42da9cb2007-02-12 00:51:39 -0800134 pgoff = entry.val;
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700135 *vec = mincore_page(&swapper_space, pgoff);
Nick Piggin30fcffe2007-02-14 12:35:02 +0100136#else
137 WARN_ON(1);
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700138 *vec = 1;
Nick Piggin30fcffe2007-02-14 12:35:02 +0100139#endif
Nick Piggin42da9cb2007-02-12 00:51:39 -0800140 }
141 }
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700142 vec++;
143 } while (ptep++, addr = next, addr != end);
Johannes Weiner6a60f1b2010-05-24 14:32:09 -0700144 pte_unmap_unlock(ptep - 1, ptl);
Johannes Weinerf4884012010-05-24 14:32:10 -0700145}
Nick Piggin42da9cb2007-02-12 00:51:39 -0800146
Johannes Weinere48293f2010-05-24 14:32:11 -0700147static void mincore_pmd_range(struct vm_area_struct *vma, pud_t *pud,
148 unsigned long addr, unsigned long end,
149 unsigned char *vec)
150{
151 unsigned long next;
152 pmd_t *pmd;
153
154 pmd = pmd_offset(pud, addr);
155 do {
156 next = pmd_addr_end(addr, end);
Johannes Weiner0ca16342011-01-13 15:47:02 -0800157 if (pmd_trans_huge(*pmd)) {
158 if (mincore_huge_pmd(vma, pmd, addr, next, vec)) {
159 vec += (next - addr) >> PAGE_SHIFT;
160 continue;
161 }
162 /* fall through */
163 }
Johannes Weinere48293f2010-05-24 14:32:11 -0700164 if (pmd_none_or_clear_bad(pmd))
165 mincore_unmapped_range(vma, addr, next, vec);
166 else
167 mincore_pte_range(vma, pmd, addr, next, vec);
168 vec += (next - addr) >> PAGE_SHIFT;
169 } while (pmd++, addr = next, addr != end);
170}
171
172static void mincore_pud_range(struct vm_area_struct *vma, pgd_t *pgd,
173 unsigned long addr, unsigned long end,
174 unsigned char *vec)
175{
176 unsigned long next;
177 pud_t *pud;
178
179 pud = pud_offset(pgd, addr);
180 do {
181 next = pud_addr_end(addr, end);
182 if (pud_none_or_clear_bad(pud))
183 mincore_unmapped_range(vma, addr, next, vec);
184 else
185 mincore_pmd_range(vma, pud, addr, next, vec);
186 vec += (next - addr) >> PAGE_SHIFT;
187 } while (pud++, addr = next, addr != end);
188}
189
190static void mincore_page_range(struct vm_area_struct *vma,
191 unsigned long addr, unsigned long end,
192 unsigned char *vec)
193{
194 unsigned long next;
195 pgd_t *pgd;
196
197 pgd = pgd_offset(vma->vm_mm, addr);
198 do {
199 next = pgd_addr_end(addr, end);
200 if (pgd_none_or_clear_bad(pgd))
201 mincore_unmapped_range(vma, addr, next, vec);
202 else
203 mincore_pud_range(vma, pgd, addr, next, vec);
204 vec += (next - addr) >> PAGE_SHIFT;
205 } while (pgd++, addr = next, addr != end);
206}
207
Johannes Weinerf4884012010-05-24 14:32:10 -0700208/*
209 * Do a chunk of "sys_mincore()". We've already checked
210 * all the arguments, we hold the mmap semaphore: we should
211 * just return the amount of info we're asked for.
212 */
213static long do_mincore(unsigned long addr, unsigned long pages, unsigned char *vec)
214{
Johannes Weinerf4884012010-05-24 14:32:10 -0700215 struct vm_area_struct *vma;
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700216 unsigned long end;
Johannes Weinerf4884012010-05-24 14:32:10 -0700217
218 vma = find_vma(current->mm, addr);
219 if (!vma || addr < vma->vm_start)
220 return -ENOMEM;
221
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700222 end = min(vma->vm_end, addr + (pages << PAGE_SHIFT));
Johannes Weinerf4884012010-05-24 14:32:10 -0700223
224 if (is_vm_hugetlb_page(vma)) {
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700225 mincore_hugetlb_page_range(vma, addr, end, vec);
226 return (end - addr) >> PAGE_SHIFT;
Johannes Weinerf4884012010-05-24 14:32:10 -0700227 }
228
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700229 end = pmd_addr_end(addr, end);
Johannes Weinerf4884012010-05-24 14:32:10 -0700230
Johannes Weinere48293f2010-05-24 14:32:11 -0700231 if (is_vm_hugetlb_page(vma))
232 mincore_hugetlb_page_range(vma, addr, end, vec);
233 else
234 mincore_page_range(vma, addr, end, vec);
Johannes Weinerf4884012010-05-24 14:32:10 -0700235
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700236 return (end - addr) >> PAGE_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237}
238
239/*
240 * The mincore(2) system call.
241 *
242 * mincore() returns the memory residency status of the pages in the
243 * current process's address space specified by [addr, addr + len).
244 * The status is returned in a vector of bytes. The least significant
245 * bit of each byte is 1 if the referenced page is in memory, otherwise
246 * it is zero.
247 *
248 * Because the status of a page can change after mincore() checks it
249 * but before it returns to the application, the returned vector may
250 * contain stale information. Only locked pages are guaranteed to
251 * remain in memory.
252 *
253 * return values:
254 * zero - success
255 * -EFAULT - vec points to an illegal address
256 * -EINVAL - addr is not a multiple of PAGE_CACHE_SIZE
257 * -ENOMEM - Addresses in the range [addr, addr + len] are
258 * invalid for the address space of this process, or
259 * specify one or more pages which are not currently
260 * mapped
261 * -EAGAIN - A kernel resource was temporarily unavailable.
262 */
Heiko Carstens3480b252009-01-14 14:14:16 +0100263SYSCALL_DEFINE3(mincore, unsigned long, start, size_t, len,
264 unsigned char __user *, vec)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265{
Linus Torvalds2f77d102006-12-16 09:44:32 -0800266 long retval;
267 unsigned long pages;
268 unsigned char *tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269
Linus Torvalds2f77d102006-12-16 09:44:32 -0800270 /* Check the start address: needs to be page-aligned.. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271 if (start & ~PAGE_CACHE_MASK)
Linus Torvalds2f77d102006-12-16 09:44:32 -0800272 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273
Linus Torvalds2f77d102006-12-16 09:44:32 -0800274 /* ..and we need to be passed a valid user-space range */
275 if (!access_ok(VERIFY_READ, (void __user *) start, len))
276 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277
Linus Torvalds2f77d102006-12-16 09:44:32 -0800278 /* This also avoids any overflows on PAGE_CACHE_ALIGN */
279 pages = len >> PAGE_SHIFT;
280 pages += (len & ~PAGE_MASK) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281
Linus Torvalds2f77d102006-12-16 09:44:32 -0800282 if (!access_ok(VERIFY_WRITE, vec, pages))
283 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284
Linus Torvalds2f77d102006-12-16 09:44:32 -0800285 tmp = (void *) __get_free_page(GFP_USER);
286 if (!tmp)
Linus Torvalds4fb23e42006-12-16 16:01:50 -0800287 return -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288
Linus Torvalds2f77d102006-12-16 09:44:32 -0800289 retval = 0;
290 while (pages) {
291 /*
292 * Do at most PAGE_SIZE entries per iteration, due to
293 * the temporary buffer size.
294 */
295 down_read(&current->mm->mmap_sem);
Johannes Weiner6a60f1b2010-05-24 14:32:09 -0700296 retval = do_mincore(start, min(pages, PAGE_SIZE), tmp);
Linus Torvalds2f77d102006-12-16 09:44:32 -0800297 up_read(&current->mm->mmap_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298
Linus Torvalds2f77d102006-12-16 09:44:32 -0800299 if (retval <= 0)
300 break;
301 if (copy_to_user(vec, tmp, retval)) {
302 retval = -EFAULT;
303 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 }
Linus Torvalds2f77d102006-12-16 09:44:32 -0800305 pages -= retval;
306 vec += retval;
307 start += retval << PAGE_SHIFT;
308 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 }
Linus Torvalds2f77d102006-12-16 09:44:32 -0800310 free_page((unsigned long) tmp);
311 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312}