mm/mincore.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  *      linux/mm/mincore.c
   4  *
   5  * Copyright (C) 1994-2006  Linus Torvalds
   6  */
   7
   8 /*
   9  * The mincore() system call.
  10  */
  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>
  20
  21 #include <linux/uaccess.h>
  22 #include <asm/pgtable.h>
  23
  24 static int mincore_hugetlb(pte_t *pte, unsigned long hmask, unsigned long addr,
  25                         unsigned long end, struct mm_walk *walk)
  26 {
  27 #ifdef CONFIG_HUGETLB_PAGE
  28         unsigned char present;
  29         unsigned char *vec = walk->private;
  30
  31         /*
  32          * Hugepages under user process are always in RAM and never
  33          * swapped out, but theoretically it needs to be checked.
  34          */
  35         present = pte && !huge_pte_none(huge_ptep_get(pte));
  36         for (; addr != end; vec++, addr += PAGE_SIZE)
  37                 *vec = present;
  38         walk->private = vec;
  39 #else
  40         BUG();
  41 #endif
  42         return 0;
  43 }
  44
  45 /*
  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.
  50  */
  51 static unsigned char mincore_page(struct address_space *mapping, pgoff_t pgoff)
  52 {
  53         unsigned char present = 0;
  54         struct page *page;
  55
  56         /*
  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.
  61          */
  62 #ifdef CONFIG_SWAP
  63         if (shmem_mapping(mapping)) {
  64                 page = find_get_entry(mapping, pgoff);
  65                 /*
  66                  * shmem/tmpfs may return swap: account for swapcache
  67                  * page too.
  68                  */
  69                 if (xa_is_value(page)) {
  70                         swp_entry_t swp = radix_to_swp_entry(page);
  71                         struct swap_info_struct *si;
  72
  73                         /* Prevent swap device to being swapoff under us */
  74                         si = get_swap_device(swp);
  75                         if (si) {
  76                                 page = find_get_page(swap_address_space(swp),
  77                                                      swp_offset(swp));
  78                                 put_swap_device(si);
  79                         } else
  80                                 page = NULL;
  81                 }
  82         } else
  83                 page = find_get_page(mapping, pgoff);
  84 #else
  85         page = find_get_page(mapping, pgoff);
  86 #endif
  87         if (page) {
  88                 present = PageUptodate(page);
  89                 put_page(page);
  90         }
  91
  92         return present;
  93 }
  94
  95 static int __mincore_unmapped_range(unsigned long addr, unsigned long end,
  96                                 struct vm_area_struct *vma, unsigned char *vec)
  97 {
  98         unsigned long nr = (end - addr) >> PAGE_SHIFT;
  99         int i;
 100
 101         if (vma->vm_file) {
 102                 pgoff_t pgoff;
 103
 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);
 107         } else {
 108                 for (i = 0; i < nr; i++)
 109                         vec[i] = 0;
 110         }
 111         return nr;
 112 }
 113
 114 static int mincore_unmapped_range(unsigned long addr, unsigned long end,
 115                                    __always_unused int depth,
 116                                    struct mm_walk *walk)
 117 {
 118         walk->private += __mincore_unmapped_range(addr, end,
 119                                                   walk->vma, walk->private);
 120         return 0;
 121 }
 122
 123 static int mincore_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
 124                         struct mm_walk *walk)
 125 {
 126         spinlock_t *ptl;
 127         struct vm_area_struct *vma = walk->vma;
 128         pte_t *ptep;
 129         unsigned char *vec = walk->private;
 130         int nr = (end - addr) >> PAGE_SHIFT;
 131
 132         ptl = pmd_trans_huge_lock(pmd, vma);
 133         if (ptl) {
 134                 memset(vec, 1, nr);
 135                 spin_unlock(ptl);
 136                 goto out;
 137         }
 138
 139         if (pmd_trans_unstable(pmd)) {
 140                 __mincore_unmapped_range(addr, end, vma, vec);
 141                 goto out;
 142         }
 143
 144         ptep = pte_offset_map_lock(walk->mm, pmd, addr, &ptl);
 145         for (; addr != end; ptep++, addr += PAGE_SIZE) {
 146                 pte_t pte = *ptep;
 147
 148                 if (pte_none(pte))
 149                         __mincore_unmapped_range(addr, addr + PAGE_SIZE,
 150                                                  vma, vec);
 151                 else if (pte_present(pte))
 152                         *vec = 1;
 153                 else { /* pte is a swap entry */
 154                         swp_entry_t entry = pte_to_swp_entry(pte);
 155
 156                         if (non_swap_entry(entry)) {
 157                                 /*
 158                                  * migration or hwpoison entries are always
 159                                  * uptodate
 160                                  */
 161                                 *vec = 1;
 162                         } else {
 163 #ifdef CONFIG_SWAP
 164                                 *vec = mincore_page(swap_address_space(entry),
 165                                                     swp_offset(entry));
 166 #else
 167                                 WARN_ON(1);
 168                                 *vec = 1;
 169 #endif
 170                         }
 171                 }
 172                 vec++;
 173         }
 174         pte_unmap_unlock(ptep - 1, ptl);
 175 out:
 176         walk->private += nr;
 177         cond_resched();
 178         return 0;
 179 }
 180
 181 static inline bool can_do_mincore(struct vm_area_struct *vma)
 182 {
 183         if (vma_is_anonymous(vma))
 184                 return true;
 185         if (!vma->vm_file)
 186                 return false;
 187         /*
 188          * Reveal pagecache information only for non-anonymous mappings that
 189          * correspond to the files the calling process could (if tried) open
 190          * for writing; otherwise we'd be including shared non-exclusive
 191          * mappings, which opens a side channel.
 192          */
 193         return inode_owner_or_capable(file_inode(vma->vm_file)) ||
 194                 inode_permission(file_inode(vma->vm_file), MAY_WRITE) == 0;
 195 }
 196
 197 static const struct mm_walk_ops mincore_walk_ops = {
 198         .pmd_entry              = mincore_pte_range,
 199         .pte_hole               = mincore_unmapped_range,
 200         .hugetlb_entry          = mincore_hugetlb,
 201 };
 202
 203 /*
 204  * Do a chunk of "sys_mincore()". We've already checked
 205  * all the arguments, we hold the mmap semaphore: we should
 206  * just return the amount of info we're asked for.
 207  */
 208 static long do_mincore(unsigned long addr, unsigned long pages, unsigned char *vec)
 209 {
 210         struct vm_area_struct *vma;
 211         unsigned long end;
 212         int err;
 213
 214         vma = find_vma(current->mm, addr);
 215         if (!vma || addr < vma->vm_start)
 216                 return -ENOMEM;
 217         end = min(vma->vm_end, addr + (pages << PAGE_SHIFT));
 218         if (!can_do_mincore(vma)) {
 219                 unsigned long pages = DIV_ROUND_UP(end - addr, PAGE_SIZE);
 220                 memset(vec, 1, pages);
 221                 return pages;
 222         }
 223         err = walk_page_range(vma->vm_mm, addr, end, &mincore_walk_ops, vec);
 224         if (err < 0)
 225                 return err;
 226         return (end - addr) >> PAGE_SHIFT;
 227 }
 228
 229 /*
 230  * The mincore(2) system call.
 231  *
 232  * mincore() returns the memory residency status of the pages in the
 233  * current process's address space specified by [addr, addr + len).
 234  * The status is returned in a vector of bytes.  The least significant
 235  * bit of each byte is 1 if the referenced page is in memory, otherwise
 236  * it is zero.
 237  *
 238  * Because the status of a page can change after mincore() checks it
 239  * but before it returns to the application, the returned vector may
 240  * contain stale information.  Only locked pages are guaranteed to
 241  * remain in memory.
 242  *
 243  * return values:
 244  *  zero    - success
 245  *  -EFAULT - vec points to an illegal address
 246  *  -EINVAL - addr is not a multiple of PAGE_SIZE
 247  *  -ENOMEM - Addresses in the range [addr, addr + len] are
 248  *              invalid for the address space of this process, or
 249  *              specify one or more pages which are not currently
 250  *              mapped
 251  *  -EAGAIN - A kernel resource was temporarily unavailable.
 252  */
 253 SYSCALL_DEFINE3(mincore, unsigned long, start, size_t, len,
 254                 unsigned char __user *, vec)
 255 {
 256         long retval;
 257         unsigned long pages;
 258         unsigned char *tmp;
 259
 260         start = untagged_addr(start);
 261
 262         /* Check the start address: needs to be page-aligned.. */
 263         if (start & ~PAGE_MASK)
 264                 return -EINVAL;
 265
 266         /* ..and we need to be passed a valid user-space range */
 267         if (!access_ok((void __user *) start, len))
 268                 return -ENOMEM;
 269
 270         /* This also avoids any overflows on PAGE_ALIGN */
 271         pages = len >> PAGE_SHIFT;
 272         pages += (offset_in_page(len)) != 0;
 273
 274         if (!access_ok(vec, pages))
 275                 return -EFAULT;
 276
 277         tmp = (void *) __get_free_page(GFP_USER);
 278         if (!tmp)
 279                 return -EAGAIN;
 280
 281         retval = 0;
 282         while (pages) {
 283                 /*
 284                  * Do at most PAGE_SIZE entries per iteration, due to
 285                  * the temporary buffer size.
 286                  */
 287                 down_read(&current->mm->mmap_sem);
 288                 retval = do_mincore(start, min(pages, PAGE_SIZE), tmp);
 289                 up_read(&current->mm->mmap_sem);
 290
 291                 if (retval <= 0)
 292                         break;
 293                 if (copy_to_user(vec, tmp, retval)) {
 294                         retval = -EFAULT;
 295                         break;
 296                 }
 297                 pages -= retval;
 298                 vec += retval;
 299                 start += retval << PAGE_SHIFT;
 300                 retval = 0;
 301         }
 302         free_page((unsigned long) tmp);
 303         return retval;
 304 }