4 * Copyright (C) 1994-2006 Linus Torvalds
8 * The mincore() system call.
10 #include <linux/pagemap.h>
11 #include <linux/gfp.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>
19 #include <asm/uaccess.h>
20 #include <asm/pgtable.h>
23 * Later we can get more picky about what "in core" means precisely.
24 * For now, simply check to see if the page is in the page cache,
25 * and is up to date; i.e. that no page-in operation would be required
26 * at this time if an application were to map and access this page.
28 static unsigned char mincore_page(struct address_space
*mapping
, pgoff_t pgoff
)
30 unsigned char present
= 0;
34 * When tmpfs swaps out a page from a file, any process mapping that
35 * file will not get a swp_entry_t in its pte, but rather it is like
36 * any other file mapping (ie. marked !present and faulted in with
37 * tmpfs's .fault). So swapped out tmpfs mappings are tested here.
39 * However when tmpfs moves the page from pagecache and into swapcache,
40 * it is still in core, but the find_get_page below won't find it.
41 * No big deal, but make a note of it.
43 page
= find_get_page(mapping
, pgoff
);
45 present
= PageUptodate(page
);
46 page_cache_release(page
);
53 * Do a chunk of "sys_mincore()". We've already checked
54 * all the arguments, we hold the mmap semaphore: we should
55 * just return the amount of info we're asked for.
57 static long do_mincore(unsigned long addr
, unsigned char *vec
, unsigned long pages
)
67 struct vm_area_struct
*vma
= find_vma(current
->mm
, addr
);
70 * find_vma() didn't find anything above us, or we're
71 * in an unmapped hole in the address space: ENOMEM.
73 if (!vma
|| addr
< vma
->vm_start
)
76 #ifdef CONFIG_HUGETLB_PAGE
77 if (is_vm_hugetlb_page(vma
)) {
79 unsigned long nr_huge
;
80 unsigned char present
;
83 nr
= min(pages
, (vma
->vm_end
- addr
) >> PAGE_SHIFT
);
85 nr_huge
= ((addr
+ pages
* PAGE_SIZE
- 1) >> huge_page_shift(h
))
86 - (addr
>> huge_page_shift(h
)) + 1;
87 nr_huge
= min(nr_huge
,
88 (vma
->vm_end
- addr
) >> huge_page_shift(h
));
90 /* hugepage always in RAM for now,
91 * but generally it needs to be check */
92 ptep
= huge_pte_offset(current
->mm
,
93 addr
& huge_page_mask(h
));
95 !huge_pte_none(huge_ptep_get(ptep
)));
99 /* reach buffer limit */
102 /* check hugepage border */
103 if (!((addr
& ~huge_page_mask(h
))
113 * Calculate how many pages there are left in the last level of the
114 * PTE array for our address.
116 nr
= PTRS_PER_PTE
- ((addr
>> PAGE_SHIFT
) & (PTRS_PER_PTE
-1));
119 * Don't overrun this vma
121 nr
= min(nr
, (vma
->vm_end
- addr
) >> PAGE_SHIFT
);
124 * Don't return more than the caller asked for
128 pgd
= pgd_offset(vma
->vm_mm
, addr
);
129 if (pgd_none_or_clear_bad(pgd
))
131 pud
= pud_offset(pgd
, addr
);
132 if (pud_none_or_clear_bad(pud
))
134 pmd
= pmd_offset(pud
, addr
);
135 if (pmd_none_or_clear_bad(pmd
))
138 ptep
= pte_offset_map_lock(vma
->vm_mm
, pmd
, addr
, &ptl
);
139 for (i
= 0; i
< nr
; i
++, ptep
++, addr
+= PAGE_SIZE
) {
140 unsigned char present
;
143 if (pte_present(pte
)) {
146 } else if (pte_none(pte
)) {
148 pgoff
= linear_page_index(vma
, addr
);
149 present
= mincore_page(vma
->vm_file
->f_mapping
,
154 } else if (pte_file(pte
)) {
155 pgoff
= pte_to_pgoff(pte
);
156 present
= mincore_page(vma
->vm_file
->f_mapping
, pgoff
);
158 } else { /* pte is a swap entry */
159 swp_entry_t entry
= pte_to_swp_entry(pte
);
160 if (is_migration_entry(entry
)) {
161 /* migration entries are always uptodate */
166 present
= mincore_page(&swapper_space
, pgoff
);
176 pte_unmap_unlock(ptep
-1, ptl
);
182 pgoff
= linear_page_index(vma
, addr
);
183 for (i
= 0; i
< nr
; i
++, pgoff
++)
184 vec
[i
] = mincore_page(vma
->vm_file
->f_mapping
, pgoff
);
186 for (i
= 0; i
< nr
; i
++)
194 * The mincore(2) system call.
196 * mincore() returns the memory residency status of the pages in the
197 * current process's address space specified by [addr, addr + len).
198 * The status is returned in a vector of bytes. The least significant
199 * bit of each byte is 1 if the referenced page is in memory, otherwise
202 * Because the status of a page can change after mincore() checks it
203 * but before it returns to the application, the returned vector may
204 * contain stale information. Only locked pages are guaranteed to
209 * -EFAULT - vec points to an illegal address
210 * -EINVAL - addr is not a multiple of PAGE_CACHE_SIZE
211 * -ENOMEM - Addresses in the range [addr, addr + len] are
212 * invalid for the address space of this process, or
213 * specify one or more pages which are not currently
215 * -EAGAIN - A kernel resource was temporarily unavailable.
217 SYSCALL_DEFINE3(mincore
, unsigned long, start
, size_t, len
,
218 unsigned char __user
*, vec
)
224 /* Check the start address: needs to be page-aligned.. */
225 if (start
& ~PAGE_CACHE_MASK
)
228 /* ..and we need to be passed a valid user-space range */
229 if (!access_ok(VERIFY_READ
, (void __user
*) start
, len
))
232 /* This also avoids any overflows on PAGE_CACHE_ALIGN */
233 pages
= len
>> PAGE_SHIFT
;
234 pages
+= (len
& ~PAGE_MASK
) != 0;
236 if (!access_ok(VERIFY_WRITE
, vec
, pages
))
239 tmp
= (void *) __get_free_page(GFP_USER
);
246 * Do at most PAGE_SIZE entries per iteration, due to
247 * the temporary buffer size.
249 down_read(¤t
->mm
->mmap_sem
);
250 retval
= do_mincore(start
, tmp
, min(pages
, PAGE_SIZE
));
251 up_read(¤t
->mm
->mmap_sem
);
255 if (copy_to_user(vec
, tmp
, retval
)) {
261 start
+= retval
<< PAGE_SHIFT
;
264 free_page((unsigned long) tmp
);