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 #include <linux/pgtable.h> 21 22 #include <linux/uaccess.h> 23 #include "swap.h" 24 #include "internal.h" 25 26 static int mincore_hugetlb(pte_t *pte, unsigned long hmask, unsigned long addr, 27 unsigned long end, struct mm_walk *walk) 28 { 29 #ifdef CONFIG_HUGETLB_PAGE 30 unsigned char present; 31 unsigned char *vec = walk->private; 32 spinlock_t *ptl; 33 34 ptl = huge_pte_lock(hstate_vma(walk->vma), walk->mm, pte); 35 /* 36 * Hugepages under user process are always in RAM and never 37 * swapped out, but theoretically it needs to be checked. 38 */ 39 present = pte && !huge_pte_none_mostly(huge_ptep_get(walk->mm, addr, pte)); 40 for (; addr != end; vec++, addr += PAGE_SIZE) 41 *vec = present; 42 walk->private = vec; 43 spin_unlock(ptl); 44 #else 45 BUG(); 46 #endif 47 return 0; 48 } 49 50 static unsigned char mincore_swap(swp_entry_t entry, bool shmem) 51 { 52 struct swap_info_struct *si; 53 struct folio *folio = NULL; 54 unsigned char present = 0; 55 56 if (!IS_ENABLED(CONFIG_SWAP)) { 57 WARN_ON(1); 58 return 0; 59 } 60 61 /* 62 * Shmem mapping may contain swapin error entries, which are 63 * absent. Page table may contain migration or hwpoison 64 * entries which are always uptodate. 65 */ 66 if (non_swap_entry(entry)) 67 return !shmem; 68 69 /* 70 * Shmem mapping lookup is lockless, so we need to grab the swap 71 * device. mincore page table walk locks the PTL, and the swap 72 * device is stable, avoid touching the si for better performance. 73 */ 74 if (shmem) { 75 si = get_swap_device(entry); 76 if (!si) 77 return 0; 78 } 79 folio = swap_cache_get_folio(entry); 80 if (shmem) 81 put_swap_device(si); 82 /* The swap cache space contains either folio, shadow or NULL */ 83 if (folio && !xa_is_value(folio)) { 84 present = folio_test_uptodate(folio); 85 folio_put(folio); 86 } 87 88 return present; 89 } 90 91 /* 92 * Later we can get more picky about what "in core" means precisely. 93 * For now, simply check to see if the page is in the page cache, 94 * and is up to date; i.e. that no page-in operation would be required 95 * at this time if an application were to map and access this page. 96 */ 97 static unsigned char mincore_page(struct address_space *mapping, pgoff_t index) 98 { 99 unsigned char present = 0; 100 struct folio *folio; 101 102 /* 103 * When tmpfs swaps out a page from a file, any process mapping that 104 * file will not get a swp_entry_t in its pte, but rather it is like 105 * any other file mapping (ie. marked !present and faulted in with 106 * tmpfs's .fault). So swapped out tmpfs mappings are tested here. 107 */ 108 folio = filemap_get_entry(mapping, index); 109 if (folio) { 110 if (xa_is_value(folio)) { 111 if (shmem_mapping(mapping)) 112 return mincore_swap(radix_to_swp_entry(folio), 113 true); 114 else 115 return 0; 116 } 117 present = folio_test_uptodate(folio); 118 folio_put(folio); 119 } 120 121 return present; 122 } 123 124 static int __mincore_unmapped_range(unsigned long addr, unsigned long end, 125 struct vm_area_struct *vma, unsigned char *vec) 126 { 127 unsigned long nr = (end - addr) >> PAGE_SHIFT; 128 int i; 129 130 if (vma->vm_file) { 131 pgoff_t pgoff; 132 133 pgoff = linear_page_index(vma, addr); 134 for (i = 0; i < nr; i++, pgoff++) 135 vec[i] = mincore_page(vma->vm_file->f_mapping, pgoff); 136 } else { 137 for (i = 0; i < nr; i++) 138 vec[i] = 0; 139 } 140 return nr; 141 } 142 143 static int mincore_unmapped_range(unsigned long addr, unsigned long end, 144 __always_unused int depth, 145 struct mm_walk *walk) 146 { 147 walk->private += __mincore_unmapped_range(addr, end, 148 walk->vma, walk->private); 149 return 0; 150 } 151 152 static int mincore_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end, 153 struct mm_walk *walk) 154 { 155 spinlock_t *ptl; 156 struct vm_area_struct *vma = walk->vma; 157 pte_t *ptep; 158 unsigned char *vec = walk->private; 159 int nr = (end - addr) >> PAGE_SHIFT; 160 int step, i; 161 162 ptl = pmd_trans_huge_lock(pmd, vma); 163 if (ptl) { 164 memset(vec, 1, nr); 165 spin_unlock(ptl); 166 goto out; 167 } 168 169 ptep = pte_offset_map_lock(walk->mm, pmd, addr, &ptl); 170 if (!ptep) { 171 walk->action = ACTION_AGAIN; 172 return 0; 173 } 174 for (; addr != end; ptep += step, addr += step * PAGE_SIZE) { 175 pte_t pte = ptep_get(ptep); 176 177 step = 1; 178 /* We need to do cache lookup too for pte markers */ 179 if (pte_none_mostly(pte)) 180 __mincore_unmapped_range(addr, addr + PAGE_SIZE, 181 vma, vec); 182 else if (pte_present(pte)) { 183 unsigned int batch = pte_batch_hint(ptep, pte); 184 185 if (batch > 1) { 186 unsigned int max_nr = (end - addr) >> PAGE_SHIFT; 187 188 step = min_t(unsigned int, batch, max_nr); 189 } 190 191 for (i = 0; i < step; i++) 192 vec[i] = 1; 193 } else { /* pte is a swap entry */ 194 *vec = mincore_swap(pte_to_swp_entry(pte), false); 195 } 196 vec += step; 197 } 198 pte_unmap_unlock(ptep - 1, ptl); 199 out: 200 walk->private += nr; 201 cond_resched(); 202 return 0; 203 } 204 205 static inline bool can_do_mincore(struct vm_area_struct *vma) 206 { 207 if (vma_is_anonymous(vma)) 208 return true; 209 if (!vma->vm_file) 210 return false; 211 /* 212 * Reveal pagecache information only for non-anonymous mappings that 213 * correspond to the files the calling process could (if tried) open 214 * for writing; otherwise we'd be including shared non-exclusive 215 * mappings, which opens a side channel. 216 */ 217 return inode_owner_or_capable(&nop_mnt_idmap, 218 file_inode(vma->vm_file)) || 219 file_permission(vma->vm_file, MAY_WRITE) == 0; 220 } 221 222 static const struct mm_walk_ops mincore_walk_ops = { 223 .pmd_entry = mincore_pte_range, 224 .pte_hole = mincore_unmapped_range, 225 .hugetlb_entry = mincore_hugetlb, 226 .walk_lock = PGWALK_RDLOCK, 227 }; 228 229 /* 230 * Do a chunk of "sys_mincore()". We've already checked 231 * all the arguments, we hold the mmap semaphore: we should 232 * just return the amount of info we're asked for. 233 */ 234 static long do_mincore(unsigned long addr, unsigned long pages, unsigned char *vec) 235 { 236 struct vm_area_struct *vma; 237 unsigned long end; 238 int err; 239 240 vma = vma_lookup(current->mm, addr); 241 if (!vma) 242 return -ENOMEM; 243 end = min(vma->vm_end, addr + (pages << PAGE_SHIFT)); 244 if (!can_do_mincore(vma)) { 245 unsigned long pages = DIV_ROUND_UP(end - addr, PAGE_SIZE); 246 memset(vec, 1, pages); 247 return pages; 248 } 249 err = walk_page_range(vma->vm_mm, addr, end, &mincore_walk_ops, vec); 250 if (err < 0) 251 return err; 252 return (end - addr) >> PAGE_SHIFT; 253 } 254 255 /* 256 * The mincore(2) system call. 257 * 258 * mincore() returns the memory residency status of the pages in the 259 * current process's address space specified by [addr, addr + len). 260 * The status is returned in a vector of bytes. The least significant 261 * bit of each byte is 1 if the referenced page is in memory, otherwise 262 * it is zero. 263 * 264 * Because the status of a page can change after mincore() checks it 265 * but before it returns to the application, the returned vector may 266 * contain stale information. Only locked pages are guaranteed to 267 * remain in memory. 268 * 269 * return values: 270 * zero - success 271 * -EFAULT - vec points to an illegal address 272 * -EINVAL - addr is not a multiple of PAGE_SIZE 273 * -ENOMEM - Addresses in the range [addr, addr + len] are 274 * invalid for the address space of this process, or 275 * specify one or more pages which are not currently 276 * mapped 277 * -EAGAIN - A kernel resource was temporarily unavailable. 278 */ 279 SYSCALL_DEFINE3(mincore, unsigned long, start, size_t, len, 280 unsigned char __user *, vec) 281 { 282 long retval; 283 unsigned long pages; 284 unsigned char *tmp; 285 286 start = untagged_addr(start); 287 288 /* Check the start address: needs to be page-aligned.. */ 289 if (unlikely(start & ~PAGE_MASK)) 290 return -EINVAL; 291 292 /* ..and we need to be passed a valid user-space range */ 293 if (!access_ok((void __user *) start, len)) 294 return -ENOMEM; 295 296 /* This also avoids any overflows on PAGE_ALIGN */ 297 pages = len >> PAGE_SHIFT; 298 pages += (offset_in_page(len)) != 0; 299 300 if (!access_ok(vec, pages)) 301 return -EFAULT; 302 303 tmp = (void *) __get_free_page(GFP_USER); 304 if (!tmp) 305 return -EAGAIN; 306 307 retval = 0; 308 while (pages) { 309 /* 310 * Do at most PAGE_SIZE entries per iteration, due to 311 * the temporary buffer size. 312 */ 313 mmap_read_lock(current->mm); 314 retval = do_mincore(start, min(pages, PAGE_SIZE), tmp); 315 mmap_read_unlock(current->mm); 316 317 if (retval <= 0) 318 break; 319 if (copy_to_user(vec, tmp, retval)) { 320 retval = -EFAULT; 321 break; 322 } 323 pages -= retval; 324 vec += retval; 325 start += retval << PAGE_SHIFT; 326 retval = 0; 327 } 328 free_page((unsigned long) tmp); 329 return retval; 330 } 331