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>
22 static int mincore_hugetlb(pte_t *pte, unsigned long hmask, unsigned long addr,
23 unsigned long end, struct mm_walk *walk)
25 #ifdef CONFIG_HUGETLB_PAGE
26 unsigned char present;
27 unsigned char *vec = walk->private;
30 * Hugepages under user process are always in RAM and never
31 * swapped out, but theoretically it needs to be checked.
33 present = pte && !huge_pte_none(huge_ptep_get(pte));
34 for (; addr != end; vec++, addr += PAGE_SIZE)
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.
49 static unsigned char mincore_page(struct address_space *mapping, pgoff_t pgoff)
51 unsigned char present = 0;
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.
61 if (shmem_mapping(mapping)) {
62 page = find_get_entry(mapping, pgoff);
64 * shmem/tmpfs may return swap: account for swapcache
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),
73 page = find_get_page(mapping, pgoff);
75 page = find_get_page(mapping, pgoff);
78 present = PageUptodate(page);
85 static int __mincore_unmapped_range(unsigned long addr, unsigned long end,
86 struct vm_area_struct *vma, unsigned char *vec)
88 unsigned long nr = (end - addr) >> PAGE_SHIFT;
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);
98 for (i = 0; i < nr; i++)
104 static int mincore_unmapped_range(unsigned long addr, unsigned long end,
105 struct mm_walk *walk)
107 walk->private += __mincore_unmapped_range(addr, end,
108 walk->vma, walk->private);
112 static int mincore_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
113 struct mm_walk *walk)
116 struct vm_area_struct *vma = walk->vma;
118 unsigned char *vec = walk->private;
119 int nr = (end - addr) >> PAGE_SHIFT;
121 ptl = pmd_trans_huge_lock(pmd, vma);
128 if (pmd_trans_unstable(pmd)) {
129 __mincore_unmapped_range(addr, end, vma, vec);
133 ptep = pte_offset_map_lock(walk->mm, pmd, addr, &ptl);
134 for (; addr != end; ptep++, addr += PAGE_SIZE) {
138 __mincore_unmapped_range(addr, addr + PAGE_SIZE,
140 else if (pte_present(pte))
142 else { /* pte is a swap entry */
143 swp_entry_t entry = pte_to_swp_entry(pte);
145 if (non_swap_entry(entry)) {
147 * migration or hwpoison entries are always
153 *vec = mincore_page(swap_address_space(entry),
163 pte_unmap_unlock(ptep - 1, ptl);
170 static inline bool can_do_mincore(struct vm_area_struct *vma)
172 if (vma_is_anonymous(vma))
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.
182 return inode_owner_or_capable(file_inode(vma->vm_file)) ||
183 inode_permission(file_inode(vma->vm_file), MAY_WRITE) == 0;
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.
191 static long do_mincore(unsigned long addr, unsigned long pages, unsigned char *vec)
193 struct vm_area_struct *vma;
196 struct mm_walk mincore_walk = {
197 .pmd_entry = mincore_pte_range,
198 .pte_hole = mincore_unmapped_range,
199 .hugetlb_entry = mincore_hugetlb,
203 vma = find_vma(current->mm, addr);
204 if (!vma || addr < vma->vm_start)
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);
212 mincore_walk.mm = vma->vm_mm;
213 err = walk_page_range(addr, end, &mincore_walk);
216 return (end - addr) >> PAGE_SHIFT;
220 * The mincore(2) system call.
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
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
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
241 * -EAGAIN - A kernel resource was temporarily unavailable.
243 SYSCALL_DEFINE3(mincore, unsigned long, start, size_t, len,
244 unsigned char __user *, vec)
250 /* Check the start address: needs to be page-aligned.. */
251 if (start & ~PAGE_MASK)
254 /* ..and we need to be passed a valid user-space range */
255 if (!access_ok(VERIFY_READ, (void __user *) start, len))
258 /* This also avoids any overflows on PAGE_ALIGN */
259 pages = len >> PAGE_SHIFT;
260 pages += (offset_in_page(len)) != 0;
262 if (!access_ok(VERIFY_WRITE, vec, pages))
265 tmp = (void *) __get_free_page(GFP_USER);
272 * Do at most PAGE_SIZE entries per iteration, due to
273 * the temporary buffer size.
275 down_read(¤t->mm->mmap_sem);
276 retval = do_mincore(start, min(pages, PAGE_SIZE), tmp);
277 up_read(¤t->mm->mmap_sem);
281 if (copy_to_user(vec, tmp, retval)) {
287 start += retval << PAGE_SHIFT;
290 free_page((unsigned long) tmp);