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/shmem_fs.h>
18 #include <linux/hugetlb.h>
20 #include <linux/uaccess.h>
21 #include <asm/pgtable.h>
23 static int mincore_hugetlb(pte_t *pte, unsigned long hmask, unsigned long addr,
24 unsigned long end, struct mm_walk *walk)
26 #ifdef CONFIG_HUGETLB_PAGE
27 unsigned char present;
28 unsigned char *vec = walk->private;
31 * Hugepages under user process are always in RAM and never
32 * swapped out, but theoretically it needs to be checked.
34 present = pte && !huge_pte_none(huge_ptep_get(pte));
35 for (; addr != end; vec++, addr += PAGE_SIZE)
45 * Later we can get more picky about what "in core" means precisely.
46 * For now, simply check to see if the page is in the page cache,
47 * and is up to date; i.e. that no page-in operation would be required
48 * at this time if an application were to map and access this page.
50 static unsigned char mincore_page(struct address_space *mapping, pgoff_t pgoff)
52 unsigned char present = 0;
56 * When tmpfs swaps out a page from a file, any process mapping that
57 * file will not get a swp_entry_t in its pte, but rather it is like
58 * any other file mapping (ie. marked !present and faulted in with
59 * tmpfs's .fault). So swapped out tmpfs mappings are tested here.
62 if (shmem_mapping(mapping)) {
63 page = find_get_entry(mapping, pgoff);
65 * shmem/tmpfs may return swap: account for swapcache
68 if (radix_tree_exceptional_entry(page)) {
69 swp_entry_t swp = radix_to_swp_entry(page);
70 page = find_get_page(swap_address_space(swp),
74 page = find_get_page(mapping, pgoff);
76 page = find_get_page(mapping, pgoff);
79 present = PageUptodate(page);
86 static int __mincore_unmapped_range(unsigned long addr, unsigned long end,
87 struct vm_area_struct *vma, unsigned char *vec)
89 unsigned long nr = (end - addr) >> PAGE_SHIFT;
95 pgoff = linear_page_index(vma, addr);
96 for (i = 0; i < nr; i++, pgoff++)
97 vec[i] = mincore_page(vma->vm_file->f_mapping, pgoff);
99 for (i = 0; i < nr; i++)
105 static int mincore_unmapped_range(unsigned long addr, unsigned long end,
106 struct mm_walk *walk)
108 walk->private += __mincore_unmapped_range(addr, end,
109 walk->vma, walk->private);
113 static int mincore_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
114 struct mm_walk *walk)
117 struct vm_area_struct *vma = walk->vma;
119 unsigned char *vec = walk->private;
120 int nr = (end - addr) >> PAGE_SHIFT;
122 ptl = pmd_trans_huge_lock(pmd, vma);
129 if (pmd_trans_unstable(pmd)) {
130 __mincore_unmapped_range(addr, end, vma, vec);
134 ptep = pte_offset_map_lock(walk->mm, pmd, addr, &ptl);
135 for (; addr != end; ptep++, addr += PAGE_SIZE) {
139 __mincore_unmapped_range(addr, addr + PAGE_SIZE,
141 else if (pte_present(pte))
143 else { /* pte is a swap entry */
144 swp_entry_t entry = pte_to_swp_entry(pte);
146 if (non_swap_entry(entry)) {
148 * migration or hwpoison entries are always
154 *vec = mincore_page(swap_address_space(entry),
164 pte_unmap_unlock(ptep - 1, ptl);
172 * Do a chunk of "sys_mincore()". We've already checked
173 * all the arguments, we hold the mmap semaphore: we should
174 * just return the amount of info we're asked for.
176 static long do_mincore(unsigned long addr, unsigned long pages, unsigned char *vec)
178 struct vm_area_struct *vma;
181 struct mm_walk mincore_walk = {
182 .pmd_entry = mincore_pte_range,
183 .pte_hole = mincore_unmapped_range,
184 .hugetlb_entry = mincore_hugetlb,
188 vma = find_vma(current->mm, addr);
189 if (!vma || addr < vma->vm_start)
191 mincore_walk.mm = vma->vm_mm;
192 end = min(vma->vm_end, addr + (pages << PAGE_SHIFT));
193 err = walk_page_range(addr, end, &mincore_walk);
196 return (end - addr) >> PAGE_SHIFT;
200 * The mincore(2) system call.
202 * mincore() returns the memory residency status of the pages in the
203 * current process's address space specified by [addr, addr + len).
204 * The status is returned in a vector of bytes. The least significant
205 * bit of each byte is 1 if the referenced page is in memory, otherwise
208 * Because the status of a page can change after mincore() checks it
209 * but before it returns to the application, the returned vector may
210 * contain stale information. Only locked pages are guaranteed to
215 * -EFAULT - vec points to an illegal address
216 * -EINVAL - addr is not a multiple of PAGE_SIZE
217 * -ENOMEM - Addresses in the range [addr, addr + len] are
218 * invalid for the address space of this process, or
219 * specify one or more pages which are not currently
221 * -EAGAIN - A kernel resource was temporarily unavailable.
223 SYSCALL_DEFINE3(mincore, unsigned long, start, size_t, len,
224 unsigned char __user *, vec)
230 /* Check the start address: needs to be page-aligned.. */
231 if (start & ~PAGE_MASK)
234 /* ..and we need to be passed a valid user-space range */
235 if (!access_ok(VERIFY_READ, (void __user *) start, len))
238 /* This also avoids any overflows on PAGE_ALIGN */
239 pages = len >> PAGE_SHIFT;
240 pages += (offset_in_page(len)) != 0;
242 if (!access_ok(VERIFY_WRITE, vec, pages))
245 tmp = (void *) __get_free_page(GFP_USER);
252 * Do at most PAGE_SIZE entries per iteration, due to
253 * the temporary buffer size.
255 down_read(¤t->mm->mmap_sem);
256 retval = do_mincore(start, min(pages, PAGE_SIZE), tmp);
257 up_read(¤t->mm->mmap_sem);
261 if (copy_to_user(vec, tmp, retval)) {
267 start += retval << PAGE_SHIFT;
270 free_page((unsigned long) tmp);