1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (C) 1994-2006 Linus Torvalds
9 * The mincore() system call.
11 #include <linux/pagemap.h>
12 #include <linux/gfp.h>
13 #include <linux/pagewalk.h>
14 #include <linux/mman.h>
15 #include <linux/syscalls.h>
16 #include <linux/swap.h>
17 #include <linux/swapops.h>
18 #include <linux/shmem_fs.h>
19 #include <linux/hugetlb.h>
21 #include <linux/uaccess.h>
22 #include <asm/pgtable.h>
24 static int mincore_hugetlb(pte_t *pte, unsigned long hmask, unsigned long addr,
25 unsigned long end, struct mm_walk *walk)
27 #ifdef CONFIG_HUGETLB_PAGE
28 unsigned char present;
29 unsigned char *vec = walk->private;
32 * Hugepages under user process are always in RAM and never
33 * swapped out, but theoretically it needs to be checked.
35 present = pte && !huge_pte_none(huge_ptep_get(pte));
36 for (; addr != end; vec++, addr += PAGE_SIZE)
46 * Later we can get more picky about what "in core" means precisely.
47 * For now, simply check to see if the page is in the page cache,
48 * and is up to date; i.e. that no page-in operation would be required
49 * at this time if an application were to map and access this page.
51 static unsigned char mincore_page(struct address_space *mapping, pgoff_t pgoff)
53 unsigned char present = 0;
57 * When tmpfs swaps out a page from a file, any process mapping that
58 * file will not get a swp_entry_t in its pte, but rather it is like
59 * any other file mapping (ie. marked !present and faulted in with
60 * tmpfs's .fault). So swapped out tmpfs mappings are tested here.
63 if (shmem_mapping(mapping)) {
64 page = find_get_entry(mapping, pgoff);
66 * shmem/tmpfs may return swap: account for swapcache
69 if (xa_is_value(page)) {
70 swp_entry_t swp = radix_to_swp_entry(page);
71 struct swap_info_struct *si;
73 /* Prevent swap device to being swapoff under us */
74 si = get_swap_device(swp);
76 page = find_get_page(swap_address_space(swp),
83 page = find_get_page(mapping, pgoff);
85 page = find_get_page(mapping, pgoff);
88 present = PageUptodate(page);
95 static int __mincore_unmapped_range(unsigned long addr, unsigned long end,
96 struct vm_area_struct *vma, unsigned char *vec)
98 unsigned long nr = (end - addr) >> PAGE_SHIFT;
104 pgoff = linear_page_index(vma, addr);
105 for (i = 0; i < nr; i++, pgoff++)
106 vec[i] = mincore_page(vma->vm_file->f_mapping, pgoff);
108 for (i = 0; i < nr; i++)
114 static int mincore_unmapped_range(unsigned long addr, unsigned long end,
115 struct mm_walk *walk)
117 walk->private += __mincore_unmapped_range(addr, end,
118 walk->vma, walk->private);
122 static int mincore_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
123 struct mm_walk *walk)
126 struct vm_area_struct *vma = walk->vma;
128 unsigned char *vec = walk->private;
129 int nr = (end - addr) >> PAGE_SHIFT;
131 ptl = pmd_trans_huge_lock(pmd, vma);
138 if (pmd_trans_unstable(pmd)) {
139 __mincore_unmapped_range(addr, end, vma, vec);
143 ptep = pte_offset_map_lock(walk->mm, pmd, addr, &ptl);
144 for (; addr != end; ptep++, addr += PAGE_SIZE) {
148 __mincore_unmapped_range(addr, addr + PAGE_SIZE,
150 else if (pte_present(pte))
152 else { /* pte is a swap entry */
153 swp_entry_t entry = pte_to_swp_entry(pte);
155 if (non_swap_entry(entry)) {
157 * migration or hwpoison entries are always
163 *vec = mincore_page(swap_address_space(entry),
173 pte_unmap_unlock(ptep - 1, ptl);
180 static inline bool can_do_mincore(struct vm_area_struct *vma)
182 if (vma_is_anonymous(vma))
187 * Reveal pagecache information only for non-anonymous mappings that
188 * correspond to the files the calling process could (if tried) open
189 * for writing; otherwise we'd be including shared non-exclusive
190 * mappings, which opens a side channel.
192 return inode_owner_or_capable(file_inode(vma->vm_file)) ||
193 inode_permission(file_inode(vma->vm_file), MAY_WRITE) == 0;
196 static const struct mm_walk_ops mincore_walk_ops = {
197 .pmd_entry = mincore_pte_range,
198 .pte_hole = mincore_unmapped_range,
199 .hugetlb_entry = mincore_hugetlb,
203 * Do a chunk of "sys_mincore()". We've already checked
204 * all the arguments, we hold the mmap semaphore: we should
205 * just return the amount of info we're asked for.
207 static long do_mincore(unsigned long addr, unsigned long pages, unsigned char *vec)
209 struct vm_area_struct *vma;
213 vma = find_vma(current->mm, addr);
214 if (!vma || addr < vma->vm_start)
216 end = min(vma->vm_end, addr + (pages << PAGE_SHIFT));
217 if (!can_do_mincore(vma)) {
218 unsigned long pages = DIV_ROUND_UP(end - addr, PAGE_SIZE);
219 memset(vec, 1, pages);
222 err = walk_page_range(vma->vm_mm, addr, end, &mincore_walk_ops, vec);
225 return (end - addr) >> PAGE_SHIFT;
229 * The mincore(2) system call.
231 * mincore() returns the memory residency status of the pages in the
232 * current process's address space specified by [addr, addr + len).
233 * The status is returned in a vector of bytes. The least significant
234 * bit of each byte is 1 if the referenced page is in memory, otherwise
237 * Because the status of a page can change after mincore() checks it
238 * but before it returns to the application, the returned vector may
239 * contain stale information. Only locked pages are guaranteed to
244 * -EFAULT - vec points to an illegal address
245 * -EINVAL - addr is not a multiple of PAGE_SIZE
246 * -ENOMEM - Addresses in the range [addr, addr + len] are
247 * invalid for the address space of this process, or
248 * specify one or more pages which are not currently
250 * -EAGAIN - A kernel resource was temporarily unavailable.
252 SYSCALL_DEFINE3(mincore, unsigned long, start, size_t, len,
253 unsigned char __user *, vec)
259 start = untagged_addr(start);
261 /* Check the start address: needs to be page-aligned.. */
262 if (start & ~PAGE_MASK)
265 /* ..and we need to be passed a valid user-space range */
266 if (!access_ok((void __user *) start, len))
269 /* This also avoids any overflows on PAGE_ALIGN */
270 pages = len >> PAGE_SHIFT;
271 pages += (offset_in_page(len)) != 0;
273 if (!access_ok(vec, pages))
276 tmp = (void *) __get_free_page(GFP_USER);
283 * Do at most PAGE_SIZE entries per iteration, due to
284 * the temporary buffer size.
286 down_read(¤t->mm->mmap_sem);
287 retval = do_mincore(start, min(pages, PAGE_SIZE), tmp);
288 up_read(¤t->mm->mmap_sem);
292 if (copy_to_user(vec, tmp, retval)) {
298 start += retval << PAGE_SHIFT;
301 free_page((unsigned long) tmp);