blob: 636a86876ff217a7f6d18f29c18524c2f2645a63 [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 page = find_get_page(mapping, pgoff);
Hugh Dickins31475dd2011-08-03 16:21:27 -070074#ifdef CONFIG_SWAP
Hugh Dickins8079b1c2011-08-03 16:21:28 -070075 /* shmem/tmpfs may return swap: account for swapcache page too. */
Hugh Dickins31475dd2011-08-03 16:21:27 -070076 if (radix_tree_exceptional_entry(page)) {
77 swp_entry_t swap = radix_to_swp_entry(page);
78 page = find_get_page(&swapper_space, swap.val);
79 }
80#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070081 if (page) {
82 present = PageUptodate(page);
83 page_cache_release(page);
84 }
85
86 return present;
87}
88
Johannes Weinerf4884012010-05-24 14:32:10 -070089static void mincore_unmapped_range(struct vm_area_struct *vma,
Johannes Weiner25ef0e52010-05-24 14:32:11 -070090 unsigned long addr, unsigned long end,
Johannes Weinerf4884012010-05-24 14:32:10 -070091 unsigned char *vec)
Linus Torvalds1da177e2005-04-16 15:20:36 -070092{
Johannes Weiner25ef0e52010-05-24 14:32:11 -070093 unsigned long nr = (end - addr) >> PAGE_SHIFT;
Nick Piggin42da9cb2007-02-12 00:51:39 -080094 int i;
Linus Torvalds1da177e2005-04-16 15:20:36 -070095
Johannes Weinerf4884012010-05-24 14:32:10 -070096 if (vma->vm_file) {
97 pgoff_t pgoff;
Linus Torvalds2f77d102006-12-16 09:44:32 -080098
Johannes Weinerf4884012010-05-24 14:32:10 -070099 pgoff = linear_page_index(vma, addr);
100 for (i = 0; i < nr; i++, pgoff++)
101 vec[i] = mincore_page(vma->vm_file->f_mapping, pgoff);
102 } else {
103 for (i = 0; i < nr; i++)
104 vec[i] = 0;
Naoya Horiguchi4f16fc12009-12-14 17:59:58 -0800105 }
Johannes Weinerf4884012010-05-24 14:32:10 -0700106}
Naoya Horiguchi4f16fc12009-12-14 17:59:58 -0800107
Johannes Weinerf4884012010-05-24 14:32:10 -0700108static void mincore_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700109 unsigned long addr, unsigned long end,
Johannes Weinerf4884012010-05-24 14:32:10 -0700110 unsigned char *vec)
111{
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700112 unsigned long next;
Johannes Weinerf4884012010-05-24 14:32:10 -0700113 spinlock_t *ptl;
114 pte_t *ptep;
Linus Torvalds2f77d102006-12-16 09:44:32 -0800115
Nick Piggin42da9cb2007-02-12 00:51:39 -0800116 ptep = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700117 do {
Nick Piggin42da9cb2007-02-12 00:51:39 -0800118 pte_t pte = *ptep;
Johannes Weinerf4884012010-05-24 14:32:10 -0700119 pgoff_t pgoff;
Nick Piggin42da9cb2007-02-12 00:51:39 -0800120
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700121 next = addr + PAGE_SIZE;
Johannes Weinerf4884012010-05-24 14:32:10 -0700122 if (pte_none(pte))
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700123 mincore_unmapped_range(vma, addr, next, vec);
Johannes Weinerf4884012010-05-24 14:32:10 -0700124 else if (pte_present(pte))
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700125 *vec = 1;
Johannes Weiner6a60f1b2010-05-24 14:32:09 -0700126 else if (pte_file(pte)) {
Nick Piggin42da9cb2007-02-12 00:51:39 -0800127 pgoff = pte_to_pgoff(pte);
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700128 *vec = mincore_page(vma->vm_file->f_mapping, pgoff);
Nick Piggin42da9cb2007-02-12 00:51:39 -0800129 } else { /* pte is a swap entry */
130 swp_entry_t entry = pte_to_swp_entry(pte);
Johannes Weiner6a60f1b2010-05-24 14:32:09 -0700131
Nick Piggin42da9cb2007-02-12 00:51:39 -0800132 if (is_migration_entry(entry)) {
133 /* migration entries are always uptodate */
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700134 *vec = 1;
Nick Piggin42da9cb2007-02-12 00:51:39 -0800135 } else {
Nick Piggin30fcffe2007-02-14 12:35:02 +0100136#ifdef CONFIG_SWAP
Nick Piggin42da9cb2007-02-12 00:51:39 -0800137 pgoff = entry.val;
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700138 *vec = mincore_page(&swapper_space, pgoff);
Nick Piggin30fcffe2007-02-14 12:35:02 +0100139#else
140 WARN_ON(1);
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700141 *vec = 1;
Nick Piggin30fcffe2007-02-14 12:35:02 +0100142#endif
Nick Piggin42da9cb2007-02-12 00:51:39 -0800143 }
144 }
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700145 vec++;
146 } while (ptep++, addr = next, addr != end);
Johannes Weiner6a60f1b2010-05-24 14:32:09 -0700147 pte_unmap_unlock(ptep - 1, ptl);
Johannes Weinerf4884012010-05-24 14:32:10 -0700148}
Nick Piggin42da9cb2007-02-12 00:51:39 -0800149
Johannes Weinere48293f2010-05-24 14:32:11 -0700150static void mincore_pmd_range(struct vm_area_struct *vma, pud_t *pud,
151 unsigned long addr, unsigned long end,
152 unsigned char *vec)
153{
154 unsigned long next;
155 pmd_t *pmd;
156
157 pmd = pmd_offset(pud, addr);
158 do {
159 next = pmd_addr_end(addr, end);
Johannes Weiner0ca16342011-01-13 15:47:02 -0800160 if (pmd_trans_huge(*pmd)) {
161 if (mincore_huge_pmd(vma, pmd, addr, next, vec)) {
162 vec += (next - addr) >> PAGE_SHIFT;
163 continue;
164 }
165 /* fall through */
166 }
Johannes Weinere48293f2010-05-24 14:32:11 -0700167 if (pmd_none_or_clear_bad(pmd))
168 mincore_unmapped_range(vma, addr, next, vec);
169 else
170 mincore_pte_range(vma, pmd, addr, next, vec);
171 vec += (next - addr) >> PAGE_SHIFT;
172 } while (pmd++, addr = next, addr != end);
173}
174
175static void mincore_pud_range(struct vm_area_struct *vma, pgd_t *pgd,
176 unsigned long addr, unsigned long end,
177 unsigned char *vec)
178{
179 unsigned long next;
180 pud_t *pud;
181
182 pud = pud_offset(pgd, addr);
183 do {
184 next = pud_addr_end(addr, end);
185 if (pud_none_or_clear_bad(pud))
186 mincore_unmapped_range(vma, addr, next, vec);
187 else
188 mincore_pmd_range(vma, pud, addr, next, vec);
189 vec += (next - addr) >> PAGE_SHIFT;
190 } while (pud++, addr = next, addr != end);
191}
192
193static void mincore_page_range(struct vm_area_struct *vma,
194 unsigned long addr, unsigned long end,
195 unsigned char *vec)
196{
197 unsigned long next;
198 pgd_t *pgd;
199
200 pgd = pgd_offset(vma->vm_mm, addr);
201 do {
202 next = pgd_addr_end(addr, end);
203 if (pgd_none_or_clear_bad(pgd))
204 mincore_unmapped_range(vma, addr, next, vec);
205 else
206 mincore_pud_range(vma, pgd, addr, next, vec);
207 vec += (next - addr) >> PAGE_SHIFT;
208 } while (pgd++, addr = next, addr != end);
209}
210
Johannes Weinerf4884012010-05-24 14:32:10 -0700211/*
212 * Do a chunk of "sys_mincore()". We've already checked
213 * all the arguments, we hold the mmap semaphore: we should
214 * just return the amount of info we're asked for.
215 */
216static long do_mincore(unsigned long addr, unsigned long pages, unsigned char *vec)
217{
Johannes Weinerf4884012010-05-24 14:32:10 -0700218 struct vm_area_struct *vma;
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700219 unsigned long end;
Johannes Weinerf4884012010-05-24 14:32:10 -0700220
221 vma = find_vma(current->mm, addr);
222 if (!vma || addr < vma->vm_start)
223 return -ENOMEM;
224
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700225 end = min(vma->vm_end, addr + (pages << PAGE_SHIFT));
Johannes Weinerf4884012010-05-24 14:32:10 -0700226
227 if (is_vm_hugetlb_page(vma)) {
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700228 mincore_hugetlb_page_range(vma, addr, end, vec);
229 return (end - addr) >> PAGE_SHIFT;
Johannes Weinerf4884012010-05-24 14:32:10 -0700230 }
231
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700232 end = pmd_addr_end(addr, end);
Johannes Weinerf4884012010-05-24 14:32:10 -0700233
Johannes Weinere48293f2010-05-24 14:32:11 -0700234 if (is_vm_hugetlb_page(vma))
235 mincore_hugetlb_page_range(vma, addr, end, vec);
236 else
237 mincore_page_range(vma, addr, end, vec);
Johannes Weinerf4884012010-05-24 14:32:10 -0700238
Johannes Weiner25ef0e52010-05-24 14:32:11 -0700239 return (end - addr) >> PAGE_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240}
241
242/*
243 * The mincore(2) system call.
244 *
245 * mincore() returns the memory residency status of the pages in the
246 * current process's address space specified by [addr, addr + len).
247 * The status is returned in a vector of bytes. The least significant
248 * bit of each byte is 1 if the referenced page is in memory, otherwise
249 * it is zero.
250 *
251 * Because the status of a page can change after mincore() checks it
252 * but before it returns to the application, the returned vector may
253 * contain stale information. Only locked pages are guaranteed to
254 * remain in memory.
255 *
256 * return values:
257 * zero - success
258 * -EFAULT - vec points to an illegal address
259 * -EINVAL - addr is not a multiple of PAGE_CACHE_SIZE
260 * -ENOMEM - Addresses in the range [addr, addr + len] are
261 * invalid for the address space of this process, or
262 * specify one or more pages which are not currently
263 * mapped
264 * -EAGAIN - A kernel resource was temporarily unavailable.
265 */
Heiko Carstens3480b252009-01-14 14:14:16 +0100266SYSCALL_DEFINE3(mincore, unsigned long, start, size_t, len,
267 unsigned char __user *, vec)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268{
Linus Torvalds2f77d102006-12-16 09:44:32 -0800269 long retval;
270 unsigned long pages;
271 unsigned char *tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272
Linus Torvalds2f77d102006-12-16 09:44:32 -0800273 /* Check the start address: needs to be page-aligned.. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 if (start & ~PAGE_CACHE_MASK)
Linus Torvalds2f77d102006-12-16 09:44:32 -0800275 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276
Linus Torvalds2f77d102006-12-16 09:44:32 -0800277 /* ..and we need to be passed a valid user-space range */
278 if (!access_ok(VERIFY_READ, (void __user *) start, len))
279 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280
Linus Torvalds2f77d102006-12-16 09:44:32 -0800281 /* This also avoids any overflows on PAGE_CACHE_ALIGN */
282 pages = len >> PAGE_SHIFT;
283 pages += (len & ~PAGE_MASK) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284
Linus Torvalds2f77d102006-12-16 09:44:32 -0800285 if (!access_ok(VERIFY_WRITE, vec, pages))
286 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287
Linus Torvalds2f77d102006-12-16 09:44:32 -0800288 tmp = (void *) __get_free_page(GFP_USER);
289 if (!tmp)
Linus Torvalds4fb23e42006-12-16 16:01:50 -0800290 return -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291
Linus Torvalds2f77d102006-12-16 09:44:32 -0800292 retval = 0;
293 while (pages) {
294 /*
295 * Do at most PAGE_SIZE entries per iteration, due to
296 * the temporary buffer size.
297 */
298 down_read(&current->mm->mmap_sem);
Johannes Weiner6a60f1b2010-05-24 14:32:09 -0700299 retval = do_mincore(start, min(pages, PAGE_SIZE), tmp);
Linus Torvalds2f77d102006-12-16 09:44:32 -0800300 up_read(&current->mm->mmap_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301
Linus Torvalds2f77d102006-12-16 09:44:32 -0800302 if (retval <= 0)
303 break;
304 if (copy_to_user(vec, tmp, retval)) {
305 retval = -EFAULT;
306 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 }
Linus Torvalds2f77d102006-12-16 09:44:32 -0800308 pages -= retval;
309 vec += retval;
310 start += retval << PAGE_SHIFT;
311 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 }
Linus Torvalds2f77d102006-12-16 09:44:32 -0800313 free_page((unsigned long) tmp);
314 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315}