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 void mincore_hugetlb_page_range(struct vm_area_struct *vma,
23 unsigned long addr, unsigned long end,
26 #ifdef CONFIG_HUGETLB_PAGE
31 unsigned char present;
34 * Huge pages are always in RAM for now, but
35 * theoretically it needs to be checked.
37 ptep = huge_pte_offset(current->mm,
38 addr & huge_page_mask(h));
39 present = ptep && !huge_pte_none(huge_ptep_get(ptep));
46 /* check hugepage border */
47 if (!(addr & ~huge_page_mask(h)))
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.
62 static unsigned char mincore_page(struct address_space *mapping, pgoff_t pgoff)
64 unsigned char present = 0;
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
71 * tmpfs's .fault). So swapped out tmpfs mappings are tested here.
73 * However when tmpfs moves the page from pagecache and into swapcache,
74 * it is still in core, but the find_get_page below won't find it.
75 * No big deal, but make a note of it.
77 page = find_get_page(mapping, pgoff);
79 present = PageUptodate(page);
80 page_cache_release(page);
86 static void mincore_unmapped_range(struct vm_area_struct *vma,
87 unsigned long addr, unsigned long end,
90 unsigned long nr = (end - addr) >> PAGE_SHIFT;
96 pgoff = linear_page_index(vma, addr);
97 for (i = 0; i < nr; i++, pgoff++)
98 vec[i] = mincore_page(vma->vm_file->f_mapping, pgoff);
100 for (i = 0; i < nr; i++)
105 static void mincore_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
106 unsigned long addr, unsigned long end,
113 ptep = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
118 next = addr + PAGE_SIZE;
120 mincore_unmapped_range(vma, addr, next, vec);
121 else if (pte_present(pte))
123 else if (pte_file(pte)) {
124 pgoff = pte_to_pgoff(pte);
125 *vec = mincore_page(vma->vm_file->f_mapping, pgoff);
126 } else { /* pte is a swap entry */
127 swp_entry_t entry = pte_to_swp_entry(pte);
129 if (is_migration_entry(entry)) {
130 /* migration entries are always uptodate */
135 *vec = mincore_page(&swapper_space, pgoff);
143 } while (ptep++, addr = next, addr != end);
144 pte_unmap_unlock(ptep - 1, ptl);
148 * Do a chunk of "sys_mincore()". We've already checked
149 * all the arguments, we hold the mmap semaphore: we should
150 * just return the amount of info we're asked for.
152 static long do_mincore(unsigned long addr, unsigned long pages, unsigned char *vec)
157 struct vm_area_struct *vma;
160 vma = find_vma(current->mm, addr);
161 if (!vma || addr < vma->vm_start)
164 end = min(vma->vm_end, addr + (pages << PAGE_SHIFT));
166 if (is_vm_hugetlb_page(vma)) {
167 mincore_hugetlb_page_range(vma, addr, end, vec);
168 return (end - addr) >> PAGE_SHIFT;
171 end = pmd_addr_end(addr, end);
173 pgd = pgd_offset(vma->vm_mm, addr);
174 if (pgd_none_or_clear_bad(pgd))
176 pud = pud_offset(pgd, addr);
177 if (pud_none_or_clear_bad(pud))
179 pmd = pmd_offset(pud, addr);
180 if (pmd_none_or_clear_bad(pmd))
183 mincore_pte_range(vma, pmd, addr, end, vec);
184 return (end - addr) >> PAGE_SHIFT;
187 mincore_unmapped_range(vma, addr, end, vec);
188 return (end - addr) >> PAGE_SHIFT;
192 * The mincore(2) system call.
194 * mincore() returns the memory residency status of the pages in the
195 * current process's address space specified by [addr, addr + len).
196 * The status is returned in a vector of bytes. The least significant
197 * bit of each byte is 1 if the referenced page is in memory, otherwise
200 * Because the status of a page can change after mincore() checks it
201 * but before it returns to the application, the returned vector may
202 * contain stale information. Only locked pages are guaranteed to
207 * -EFAULT - vec points to an illegal address
208 * -EINVAL - addr is not a multiple of PAGE_CACHE_SIZE
209 * -ENOMEM - Addresses in the range [addr, addr + len] are
210 * invalid for the address space of this process, or
211 * specify one or more pages which are not currently
213 * -EAGAIN - A kernel resource was temporarily unavailable.
215 SYSCALL_DEFINE3(mincore, unsigned long, start, size_t, len,
216 unsigned char __user *, vec)
222 /* Check the start address: needs to be page-aligned.. */
223 if (start & ~PAGE_CACHE_MASK)
226 /* ..and we need to be passed a valid user-space range */
227 if (!access_ok(VERIFY_READ, (void __user *) start, len))
230 /* This also avoids any overflows on PAGE_CACHE_ALIGN */
231 pages = len >> PAGE_SHIFT;
232 pages += (len & ~PAGE_MASK) != 0;
234 if (!access_ok(VERIFY_WRITE, vec, pages))
237 tmp = (void *) __get_free_page(GFP_USER);
244 * Do at most PAGE_SIZE entries per iteration, due to
245 * the temporary buffer size.
247 down_read(¤t->mm->mmap_sem);
248 retval = do_mincore(start, min(pages, PAGE_SIZE), tmp);
249 up_read(¤t->mm->mmap_sem);
253 if (copy_to_user(vec, tmp, retval)) {
259 start += retval << PAGE_SHIFT;
262 free_page((unsigned long) tmp);