1 /** 2 * inode.c - NTFS kernel inode handling. Part of the Linux-NTFS project. 3 * 4 * Copyright (c) 2001-2007 Anton Altaparmakov 5 * 6 * This program/include file is free software; you can redistribute it and/or 7 * modify it under the terms of the GNU General Public License as published 8 * by the Free Software Foundation; either version 2 of the License, or 9 * (at your option) any later version. 10 * 11 * This program/include file is distributed in the hope that it will be 12 * useful, but WITHOUT ANY WARRANTY; without even the implied warranty 13 * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 14 * GNU General Public License for more details. 15 * 16 * You should have received a copy of the GNU General Public License 17 * along with this program (in the main directory of the Linux-NTFS 18 * distribution in the file COPYING); if not, write to the Free Software 19 * Foundation,Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA 20 */ 21 22 #include <linux/buffer_head.h> 23 #include <linux/fs.h> 24 #include <linux/mm.h> 25 #include <linux/mount.h> 26 #include <linux/mutex.h> 27 #include <linux/pagemap.h> 28 #include <linux/quotaops.h> 29 #include <linux/slab.h> 30 #include <linux/log2.h> 31 32 #include "aops.h" 33 #include "attrib.h" 34 #include "bitmap.h" 35 #include "dir.h" 36 #include "debug.h" 37 #include "inode.h" 38 #include "lcnalloc.h" 39 #include "malloc.h" 40 #include "mft.h" 41 #include "time.h" 42 #include "ntfs.h" 43 44 /** 45 * ntfs_test_inode - compare two (possibly fake) inodes for equality 46 * @vi: vfs inode which to test 47 * @na: ntfs attribute which is being tested with 48 * 49 * Compare the ntfs attribute embedded in the ntfs specific part of the vfs 50 * inode @vi for equality with the ntfs attribute @na. 51 * 52 * If searching for the normal file/directory inode, set @na->type to AT_UNUSED. 53 * @na->name and @na->name_len are then ignored. 54 * 55 * Return 1 if the attributes match and 0 if not. 56 * 57 * NOTE: This function runs with the inode_lock spin lock held so it is not 58 * allowed to sleep. 59 */ 60 int ntfs_test_inode(struct inode *vi, ntfs_attr *na) 61 { 62 ntfs_inode *ni; 63 64 if (vi->i_ino != na->mft_no) 65 return 0; 66 ni = NTFS_I(vi); 67 /* If !NInoAttr(ni), @vi is a normal file or directory inode. */ 68 if (likely(!NInoAttr(ni))) { 69 /* If not looking for a normal inode this is a mismatch. */ 70 if (unlikely(na->type != AT_UNUSED)) 71 return 0; 72 } else { 73 /* A fake inode describing an attribute. */ 74 if (ni->type != na->type) 75 return 0; 76 if (ni->name_len != na->name_len) 77 return 0; 78 if (na->name_len && memcmp(ni->name, na->name, 79 na->name_len * sizeof(ntfschar))) 80 return 0; 81 } 82 /* Match! */ 83 return 1; 84 } 85 86 /** 87 * ntfs_init_locked_inode - initialize an inode 88 * @vi: vfs inode to initialize 89 * @na: ntfs attribute which to initialize @vi to 90 * 91 * Initialize the vfs inode @vi with the values from the ntfs attribute @na in 92 * order to enable ntfs_test_inode() to do its work. 93 * 94 * If initializing the normal file/directory inode, set @na->type to AT_UNUSED. 95 * In that case, @na->name and @na->name_len should be set to NULL and 0, 96 * respectively. Although that is not strictly necessary as 97 * ntfs_read_locked_inode() will fill them in later. 98 * 99 * Return 0 on success and -errno on error. 100 * 101 * NOTE: This function runs with the inode_lock spin lock held so it is not 102 * allowed to sleep. (Hence the GFP_ATOMIC allocation.) 103 */ 104 static int ntfs_init_locked_inode(struct inode *vi, ntfs_attr *na) 105 { 106 ntfs_inode *ni = NTFS_I(vi); 107 108 vi->i_ino = na->mft_no; 109 110 ni->type = na->type; 111 if (na->type == AT_INDEX_ALLOCATION) 112 NInoSetMstProtected(ni); 113 114 ni->name = na->name; 115 ni->name_len = na->name_len; 116 117 /* If initializing a normal inode, we are done. */ 118 if (likely(na->type == AT_UNUSED)) { 119 BUG_ON(na->name); 120 BUG_ON(na->name_len); 121 return 0; 122 } 123 124 /* It is a fake inode. */ 125 NInoSetAttr(ni); 126 127 /* 128 * We have I30 global constant as an optimization as it is the name 129 * in >99.9% of named attributes! The other <0.1% incur a GFP_ATOMIC 130 * allocation but that is ok. And most attributes are unnamed anyway, 131 * thus the fraction of named attributes with name != I30 is actually 132 * absolutely tiny. 133 */ 134 if (na->name_len && na->name != I30) { 135 unsigned int i; 136 137 BUG_ON(!na->name); 138 i = na->name_len * sizeof(ntfschar); 139 ni->name = kmalloc(i + sizeof(ntfschar), GFP_ATOMIC); 140 if (!ni->name) 141 return -ENOMEM; 142 memcpy(ni->name, na->name, i); 143 ni->name[na->name_len] = 0; 144 } 145 return 0; 146 } 147 148 typedef int (*set_t)(struct inode *, void *); 149 static int ntfs_read_locked_inode(struct inode *vi); 150 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi); 151 static int ntfs_read_locked_index_inode(struct inode *base_vi, 152 struct inode *vi); 153 154 /** 155 * ntfs_iget - obtain a struct inode corresponding to a specific normal inode 156 * @sb: super block of mounted volume 157 * @mft_no: mft record number / inode number to obtain 158 * 159 * Obtain the struct inode corresponding to a specific normal inode (i.e. a 160 * file or directory). 161 * 162 * If the inode is in the cache, it is just returned with an increased 163 * reference count. Otherwise, a new struct inode is allocated and initialized, 164 * and finally ntfs_read_locked_inode() is called to read in the inode and 165 * fill in the remainder of the inode structure. 166 * 167 * Return the struct inode on success. Check the return value with IS_ERR() and 168 * if true, the function failed and the error code is obtained from PTR_ERR(). 169 */ 170 struct inode *ntfs_iget(struct super_block *sb, unsigned long mft_no) 171 { 172 struct inode *vi; 173 int err; 174 ntfs_attr na; 175 176 na.mft_no = mft_no; 177 na.type = AT_UNUSED; 178 na.name = NULL; 179 na.name_len = 0; 180 181 vi = iget5_locked(sb, mft_no, (test_t)ntfs_test_inode, 182 (set_t)ntfs_init_locked_inode, &na); 183 if (unlikely(!vi)) 184 return ERR_PTR(-ENOMEM); 185 186 err = 0; 187 188 /* If this is a freshly allocated inode, need to read it now. */ 189 if (vi->i_state & I_NEW) { 190 err = ntfs_read_locked_inode(vi); 191 unlock_new_inode(vi); 192 } 193 /* 194 * There is no point in keeping bad inodes around if the failure was 195 * due to ENOMEM. We want to be able to retry again later. 196 */ 197 if (unlikely(err == -ENOMEM)) { 198 iput(vi); 199 vi = ERR_PTR(err); 200 } 201 return vi; 202 } 203 204 /** 205 * ntfs_attr_iget - obtain a struct inode corresponding to an attribute 206 * @base_vi: vfs base inode containing the attribute 207 * @type: attribute type 208 * @name: Unicode name of the attribute (NULL if unnamed) 209 * @name_len: length of @name in Unicode characters (0 if unnamed) 210 * 211 * Obtain the (fake) struct inode corresponding to the attribute specified by 212 * @type, @name, and @name_len, which is present in the base mft record 213 * specified by the vfs inode @base_vi. 214 * 215 * If the attribute inode is in the cache, it is just returned with an 216 * increased reference count. Otherwise, a new struct inode is allocated and 217 * initialized, and finally ntfs_read_locked_attr_inode() is called to read the 218 * attribute and fill in the inode structure. 219 * 220 * Note, for index allocation attributes, you need to use ntfs_index_iget() 221 * instead of ntfs_attr_iget() as working with indices is a lot more complex. 222 * 223 * Return the struct inode of the attribute inode on success. Check the return 224 * value with IS_ERR() and if true, the function failed and the error code is 225 * obtained from PTR_ERR(). 226 */ 227 struct inode *ntfs_attr_iget(struct inode *base_vi, ATTR_TYPE type, 228 ntfschar *name, u32 name_len) 229 { 230 struct inode *vi; 231 int err; 232 ntfs_attr na; 233 234 /* Make sure no one calls ntfs_attr_iget() for indices. */ 235 BUG_ON(type == AT_INDEX_ALLOCATION); 236 237 na.mft_no = base_vi->i_ino; 238 na.type = type; 239 na.name = name; 240 na.name_len = name_len; 241 242 vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode, 243 (set_t)ntfs_init_locked_inode, &na); 244 if (unlikely(!vi)) 245 return ERR_PTR(-ENOMEM); 246 247 err = 0; 248 249 /* If this is a freshly allocated inode, need to read it now. */ 250 if (vi->i_state & I_NEW) { 251 err = ntfs_read_locked_attr_inode(base_vi, vi); 252 unlock_new_inode(vi); 253 } 254 /* 255 * There is no point in keeping bad attribute inodes around. This also 256 * simplifies things in that we never need to check for bad attribute 257 * inodes elsewhere. 258 */ 259 if (unlikely(err)) { 260 iput(vi); 261 vi = ERR_PTR(err); 262 } 263 return vi; 264 } 265 266 /** 267 * ntfs_index_iget - obtain a struct inode corresponding to an index 268 * @base_vi: vfs base inode containing the index related attributes 269 * @name: Unicode name of the index 270 * @name_len: length of @name in Unicode characters 271 * 272 * Obtain the (fake) struct inode corresponding to the index specified by @name 273 * and @name_len, which is present in the base mft record specified by the vfs 274 * inode @base_vi. 275 * 276 * If the index inode is in the cache, it is just returned with an increased 277 * reference count. Otherwise, a new struct inode is allocated and 278 * initialized, and finally ntfs_read_locked_index_inode() is called to read 279 * the index related attributes and fill in the inode structure. 280 * 281 * Return the struct inode of the index inode on success. Check the return 282 * value with IS_ERR() and if true, the function failed and the error code is 283 * obtained from PTR_ERR(). 284 */ 285 struct inode *ntfs_index_iget(struct inode *base_vi, ntfschar *name, 286 u32 name_len) 287 { 288 struct inode *vi; 289 int err; 290 ntfs_attr na; 291 292 na.mft_no = base_vi->i_ino; 293 na.type = AT_INDEX_ALLOCATION; 294 na.name = name; 295 na.name_len = name_len; 296 297 vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode, 298 (set_t)ntfs_init_locked_inode, &na); 299 if (unlikely(!vi)) 300 return ERR_PTR(-ENOMEM); 301 302 err = 0; 303 304 /* If this is a freshly allocated inode, need to read it now. */ 305 if (vi->i_state & I_NEW) { 306 err = ntfs_read_locked_index_inode(base_vi, vi); 307 unlock_new_inode(vi); 308 } 309 /* 310 * There is no point in keeping bad index inodes around. This also 311 * simplifies things in that we never need to check for bad index 312 * inodes elsewhere. 313 */ 314 if (unlikely(err)) { 315 iput(vi); 316 vi = ERR_PTR(err); 317 } 318 return vi; 319 } 320 321 struct inode *ntfs_alloc_big_inode(struct super_block *sb) 322 { 323 ntfs_inode *ni; 324 325 ntfs_debug("Entering."); 326 ni = kmem_cache_alloc(ntfs_big_inode_cache, GFP_NOFS); 327 if (likely(ni != NULL)) { 328 ni->state = 0; 329 return VFS_I(ni); 330 } 331 ntfs_error(sb, "Allocation of NTFS big inode structure failed."); 332 return NULL; 333 } 334 335 void ntfs_destroy_big_inode(struct inode *inode) 336 { 337 ntfs_inode *ni = NTFS_I(inode); 338 339 ntfs_debug("Entering."); 340 BUG_ON(ni->page); 341 if (!atomic_dec_and_test(&ni->count)) 342 BUG(); 343 kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode)); 344 } 345 346 static inline ntfs_inode *ntfs_alloc_extent_inode(void) 347 { 348 ntfs_inode *ni; 349 350 ntfs_debug("Entering."); 351 ni = kmem_cache_alloc(ntfs_inode_cache, GFP_NOFS); 352 if (likely(ni != NULL)) { 353 ni->state = 0; 354 return ni; 355 } 356 ntfs_error(NULL, "Allocation of NTFS inode structure failed."); 357 return NULL; 358 } 359 360 static void ntfs_destroy_extent_inode(ntfs_inode *ni) 361 { 362 ntfs_debug("Entering."); 363 BUG_ON(ni->page); 364 if (!atomic_dec_and_test(&ni->count)) 365 BUG(); 366 kmem_cache_free(ntfs_inode_cache, ni); 367 } 368 369 /* 370 * The attribute runlist lock has separate locking rules from the 371 * normal runlist lock, so split the two lock-classes: 372 */ 373 static struct lock_class_key attr_list_rl_lock_class; 374 375 /** 376 * __ntfs_init_inode - initialize ntfs specific part of an inode 377 * @sb: super block of mounted volume 378 * @ni: freshly allocated ntfs inode which to initialize 379 * 380 * Initialize an ntfs inode to defaults. 381 * 382 * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left 383 * untouched. Make sure to initialize them elsewhere. 384 * 385 * Return zero on success and -ENOMEM on error. 386 */ 387 void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni) 388 { 389 ntfs_debug("Entering."); 390 rwlock_init(&ni->size_lock); 391 ni->initialized_size = ni->allocated_size = 0; 392 ni->seq_no = 0; 393 atomic_set(&ni->count, 1); 394 ni->vol = NTFS_SB(sb); 395 ntfs_init_runlist(&ni->runlist); 396 mutex_init(&ni->mrec_lock); 397 ni->page = NULL; 398 ni->page_ofs = 0; 399 ni->attr_list_size = 0; 400 ni->attr_list = NULL; 401 ntfs_init_runlist(&ni->attr_list_rl); 402 lockdep_set_class(&ni->attr_list_rl.lock, 403 &attr_list_rl_lock_class); 404 ni->itype.index.block_size = 0; 405 ni->itype.index.vcn_size = 0; 406 ni->itype.index.collation_rule = 0; 407 ni->itype.index.block_size_bits = 0; 408 ni->itype.index.vcn_size_bits = 0; 409 mutex_init(&ni->extent_lock); 410 ni->nr_extents = 0; 411 ni->ext.base_ntfs_ino = NULL; 412 } 413 414 /* 415 * Extent inodes get MFT-mapped in a nested way, while the base inode 416 * is still mapped. Teach this nesting to the lock validator by creating 417 * a separate class for nested inode's mrec_lock's: 418 */ 419 static struct lock_class_key extent_inode_mrec_lock_key; 420 421 inline ntfs_inode *ntfs_new_extent_inode(struct super_block *sb, 422 unsigned long mft_no) 423 { 424 ntfs_inode *ni = ntfs_alloc_extent_inode(); 425 426 ntfs_debug("Entering."); 427 if (likely(ni != NULL)) { 428 __ntfs_init_inode(sb, ni); 429 lockdep_set_class(&ni->mrec_lock, &extent_inode_mrec_lock_key); 430 ni->mft_no = mft_no; 431 ni->type = AT_UNUSED; 432 ni->name = NULL; 433 ni->name_len = 0; 434 } 435 return ni; 436 } 437 438 /** 439 * ntfs_is_extended_system_file - check if a file is in the $Extend directory 440 * @ctx: initialized attribute search context 441 * 442 * Search all file name attributes in the inode described by the attribute 443 * search context @ctx and check if any of the names are in the $Extend system 444 * directory. 445 * 446 * Return values: 447 * 1: file is in $Extend directory 448 * 0: file is not in $Extend directory 449 * -errno: failed to determine if the file is in the $Extend directory 450 */ 451 static int ntfs_is_extended_system_file(ntfs_attr_search_ctx *ctx) 452 { 453 int nr_links, err; 454 455 /* Restart search. */ 456 ntfs_attr_reinit_search_ctx(ctx); 457 458 /* Get number of hard links. */ 459 nr_links = le16_to_cpu(ctx->mrec->link_count); 460 461 /* Loop through all hard links. */ 462 while (!(err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0, NULL, 0, 463 ctx))) { 464 FILE_NAME_ATTR *file_name_attr; 465 ATTR_RECORD *attr = ctx->attr; 466 u8 *p, *p2; 467 468 nr_links--; 469 /* 470 * Maximum sanity checking as we are called on an inode that 471 * we suspect might be corrupt. 472 */ 473 p = (u8*)attr + le32_to_cpu(attr->length); 474 if (p < (u8*)ctx->mrec || (u8*)p > (u8*)ctx->mrec + 475 le32_to_cpu(ctx->mrec->bytes_in_use)) { 476 err_corrupt_attr: 477 ntfs_error(ctx->ntfs_ino->vol->sb, "Corrupt file name " 478 "attribute. You should run chkdsk."); 479 return -EIO; 480 } 481 if (attr->non_resident) { 482 ntfs_error(ctx->ntfs_ino->vol->sb, "Non-resident file " 483 "name. You should run chkdsk."); 484 return -EIO; 485 } 486 if (attr->flags) { 487 ntfs_error(ctx->ntfs_ino->vol->sb, "File name with " 488 "invalid flags. You should run " 489 "chkdsk."); 490 return -EIO; 491 } 492 if (!(attr->data.resident.flags & RESIDENT_ATTR_IS_INDEXED)) { 493 ntfs_error(ctx->ntfs_ino->vol->sb, "Unindexed file " 494 "name. You should run chkdsk."); 495 return -EIO; 496 } 497 file_name_attr = (FILE_NAME_ATTR*)((u8*)attr + 498 le16_to_cpu(attr->data.resident.value_offset)); 499 p2 = (u8*)attr + le32_to_cpu(attr->data.resident.value_length); 500 if (p2 < (u8*)attr || p2 > p) 501 goto err_corrupt_attr; 502 /* This attribute is ok, but is it in the $Extend directory? */ 503 if (MREF_LE(file_name_attr->parent_directory) == FILE_Extend) 504 return 1; /* YES, it's an extended system file. */ 505 } 506 if (unlikely(err != -ENOENT)) 507 return err; 508 if (unlikely(nr_links)) { 509 ntfs_error(ctx->ntfs_ino->vol->sb, "Inode hard link count " 510 "doesn't match number of name attributes. You " 511 "should run chkdsk."); 512 return -EIO; 513 } 514 return 0; /* NO, it is not an extended system file. */ 515 } 516 517 /** 518 * ntfs_read_locked_inode - read an inode from its device 519 * @vi: inode to read 520 * 521 * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode 522 * described by @vi into memory from the device. 523 * 524 * The only fields in @vi that we need to/can look at when the function is 525 * called are i_sb, pointing to the mounted device's super block, and i_ino, 526 * the number of the inode to load. 527 * 528 * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino 529 * for reading and sets up the necessary @vi fields as well as initializing 530 * the ntfs inode. 531 * 532 * Q: What locks are held when the function is called? 533 * A: i_state has I_NEW set, hence the inode is locked, also 534 * i_count is set to 1, so it is not going to go away 535 * i_flags is set to 0 and we have no business touching it. Only an ioctl() 536 * is allowed to write to them. We should of course be honouring them but 537 * we need to do that using the IS_* macros defined in include/linux/fs.h. 538 * In any case ntfs_read_locked_inode() has nothing to do with i_flags. 539 * 540 * Return 0 on success and -errno on error. In the error case, the inode will 541 * have had make_bad_inode() executed on it. 542 */ 543 static int ntfs_read_locked_inode(struct inode *vi) 544 { 545 ntfs_volume *vol = NTFS_SB(vi->i_sb); 546 ntfs_inode *ni; 547 struct inode *bvi; 548 MFT_RECORD *m; 549 ATTR_RECORD *a; 550 STANDARD_INFORMATION *si; 551 ntfs_attr_search_ctx *ctx; 552 int err = 0; 553 554 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino); 555 556 /* Setup the generic vfs inode parts now. */ 557 558 /* 559 * This is for checking whether an inode has changed w.r.t. a file so 560 * that the file can be updated if necessary (compare with f_version). 561 */ 562 vi->i_version = 1; 563 564 vi->i_uid = vol->uid; 565 vi->i_gid = vol->gid; 566 vi->i_mode = 0; 567 568 /* 569 * Initialize the ntfs specific part of @vi special casing 570 * FILE_MFT which we need to do at mount time. 571 */ 572 if (vi->i_ino != FILE_MFT) 573 ntfs_init_big_inode(vi); 574 ni = NTFS_I(vi); 575 576 m = map_mft_record(ni); 577 if (IS_ERR(m)) { 578 err = PTR_ERR(m); 579 goto err_out; 580 } 581 ctx = ntfs_attr_get_search_ctx(ni, m); 582 if (!ctx) { 583 err = -ENOMEM; 584 goto unm_err_out; 585 } 586 587 if (!(m->flags & MFT_RECORD_IN_USE)) { 588 ntfs_error(vi->i_sb, "Inode is not in use!"); 589 goto unm_err_out; 590 } 591 if (m->base_mft_record) { 592 ntfs_error(vi->i_sb, "Inode is an extent inode!"); 593 goto unm_err_out; 594 } 595 596 /* Transfer information from mft record into vfs and ntfs inodes. */ 597 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number); 598 599 /* 600 * FIXME: Keep in mind that link_count is two for files which have both 601 * a long file name and a short file name as separate entries, so if 602 * we are hiding short file names this will be too high. Either we need 603 * to account for the short file names by subtracting them or we need 604 * to make sure we delete files even though i_nlink is not zero which 605 * might be tricky due to vfs interactions. Need to think about this 606 * some more when implementing the unlink command. 607 */ 608 vi->i_nlink = le16_to_cpu(m->link_count); 609 /* 610 * FIXME: Reparse points can have the directory bit set even though 611 * they would be S_IFLNK. Need to deal with this further below when we 612 * implement reparse points / symbolic links but it will do for now. 613 * Also if not a directory, it could be something else, rather than 614 * a regular file. But again, will do for now. 615 */ 616 /* Everyone gets all permissions. */ 617 vi->i_mode |= S_IRWXUGO; 618 /* If read-only, noone gets write permissions. */ 619 if (IS_RDONLY(vi)) 620 vi->i_mode &= ~S_IWUGO; 621 if (m->flags & MFT_RECORD_IS_DIRECTORY) { 622 vi->i_mode |= S_IFDIR; 623 /* 624 * Apply the directory permissions mask set in the mount 625 * options. 626 */ 627 vi->i_mode &= ~vol->dmask; 628 /* Things break without this kludge! */ 629 if (vi->i_nlink > 1) 630 vi->i_nlink = 1; 631 } else { 632 vi->i_mode |= S_IFREG; 633 /* Apply the file permissions mask set in the mount options. */ 634 vi->i_mode &= ~vol->fmask; 635 } 636 /* 637 * Find the standard information attribute in the mft record. At this 638 * stage we haven't setup the attribute list stuff yet, so this could 639 * in fact fail if the standard information is in an extent record, but 640 * I don't think this actually ever happens. 641 */ 642 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0, 643 ctx); 644 if (unlikely(err)) { 645 if (err == -ENOENT) { 646 /* 647 * TODO: We should be performing a hot fix here (if the 648 * recover mount option is set) by creating a new 649 * attribute. 650 */ 651 ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute " 652 "is missing."); 653 } 654 goto unm_err_out; 655 } 656 a = ctx->attr; 657 /* Get the standard information attribute value. */ 658 si = (STANDARD_INFORMATION*)((u8*)a + 659 le16_to_cpu(a->data.resident.value_offset)); 660 661 /* Transfer information from the standard information into vi. */ 662 /* 663 * Note: The i_?times do not quite map perfectly onto the NTFS times, 664 * but they are close enough, and in the end it doesn't really matter 665 * that much... 666 */ 667 /* 668 * mtime is the last change of the data within the file. Not changed 669 * when only metadata is changed, e.g. a rename doesn't affect mtime. 670 */ 671 vi->i_mtime = ntfs2utc(si->last_data_change_time); 672 /* 673 * ctime is the last change of the metadata of the file. This obviously 674 * always changes, when mtime is changed. ctime can be changed on its 675 * own, mtime is then not changed, e.g. when a file is renamed. 676 */ 677 vi->i_ctime = ntfs2utc(si->last_mft_change_time); 678 /* 679 * Last access to the data within the file. Not changed during a rename 680 * for example but changed whenever the file is written to. 681 */ 682 vi->i_atime = ntfs2utc(si->last_access_time); 683 684 /* Find the attribute list attribute if present. */ 685 ntfs_attr_reinit_search_ctx(ctx); 686 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx); 687 if (err) { 688 if (unlikely(err != -ENOENT)) { 689 ntfs_error(vi->i_sb, "Failed to lookup attribute list " 690 "attribute."); 691 goto unm_err_out; 692 } 693 } else /* if (!err) */ { 694 if (vi->i_ino == FILE_MFT) 695 goto skip_attr_list_load; 696 ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino); 697 NInoSetAttrList(ni); 698 a = ctx->attr; 699 if (a->flags & ATTR_COMPRESSION_MASK) { 700 ntfs_error(vi->i_sb, "Attribute list attribute is " 701 "compressed."); 702 goto unm_err_out; 703 } 704 if (a->flags & ATTR_IS_ENCRYPTED || 705 a->flags & ATTR_IS_SPARSE) { 706 if (a->non_resident) { 707 ntfs_error(vi->i_sb, "Non-resident attribute " 708 "list attribute is encrypted/" 709 "sparse."); 710 goto unm_err_out; 711 } 712 ntfs_warning(vi->i_sb, "Resident attribute list " 713 "attribute in inode 0x%lx is marked " 714 "encrypted/sparse which is not true. " 715 "However, Windows allows this and " 716 "chkdsk does not detect or correct it " 717 "so we will just ignore the invalid " 718 "flags and pretend they are not set.", 719 vi->i_ino); 720 } 721 /* Now allocate memory for the attribute list. */ 722 ni->attr_list_size = (u32)ntfs_attr_size(a); 723 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size); 724 if (!ni->attr_list) { 725 ntfs_error(vi->i_sb, "Not enough memory to allocate " 726 "buffer for attribute list."); 727 err = -ENOMEM; 728 goto unm_err_out; 729 } 730 if (a->non_resident) { 731 NInoSetAttrListNonResident(ni); 732 if (a->data.non_resident.lowest_vcn) { 733 ntfs_error(vi->i_sb, "Attribute list has non " 734 "zero lowest_vcn."); 735 goto unm_err_out; 736 } 737 /* 738 * Setup the runlist. No need for locking as we have 739 * exclusive access to the inode at this time. 740 */ 741 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol, 742 a, NULL); 743 if (IS_ERR(ni->attr_list_rl.rl)) { 744 err = PTR_ERR(ni->attr_list_rl.rl); 745 ni->attr_list_rl.rl = NULL; 746 ntfs_error(vi->i_sb, "Mapping pairs " 747 "decompression failed."); 748 goto unm_err_out; 749 } 750 /* Now load the attribute list. */ 751 if ((err = load_attribute_list(vol, &ni->attr_list_rl, 752 ni->attr_list, ni->attr_list_size, 753 sle64_to_cpu(a->data.non_resident. 754 initialized_size)))) { 755 ntfs_error(vi->i_sb, "Failed to load " 756 "attribute list attribute."); 757 goto unm_err_out; 758 } 759 } else /* if (!a->non_resident) */ { 760 if ((u8*)a + le16_to_cpu(a->data.resident.value_offset) 761 + le32_to_cpu( 762 a->data.resident.value_length) > 763 (u8*)ctx->mrec + vol->mft_record_size) { 764 ntfs_error(vi->i_sb, "Corrupt attribute list " 765 "in inode."); 766 goto unm_err_out; 767 } 768 /* Now copy the attribute list. */ 769 memcpy(ni->attr_list, (u8*)a + le16_to_cpu( 770 a->data.resident.value_offset), 771 le32_to_cpu( 772 a->data.resident.value_length)); 773 } 774 } 775 skip_attr_list_load: 776 /* 777 * If an attribute list is present we now have the attribute list value 778 * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes. 779 */ 780 if (S_ISDIR(vi->i_mode)) { 781 loff_t bvi_size; 782 ntfs_inode *bni; 783 INDEX_ROOT *ir; 784 u8 *ir_end, *index_end; 785 786 /* It is a directory, find index root attribute. */ 787 ntfs_attr_reinit_search_ctx(ctx); 788 err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE, 789 0, NULL, 0, ctx); 790 if (unlikely(err)) { 791 if (err == -ENOENT) { 792 // FIXME: File is corrupt! Hot-fix with empty 793 // index root attribute if recovery option is 794 // set. 795 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute " 796 "is missing."); 797 } 798 goto unm_err_out; 799 } 800 a = ctx->attr; 801 /* Set up the state. */ 802 if (unlikely(a->non_resident)) { 803 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not " 804 "resident."); 805 goto unm_err_out; 806 } 807 /* Ensure the attribute name is placed before the value. */ 808 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= 809 le16_to_cpu(a->data.resident.value_offset)))) { 810 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is " 811 "placed after the attribute value."); 812 goto unm_err_out; 813 } 814 /* 815 * Compressed/encrypted index root just means that the newly 816 * created files in that directory should be created compressed/ 817 * encrypted. However index root cannot be both compressed and 818 * encrypted. 819 */ 820 if (a->flags & ATTR_COMPRESSION_MASK) 821 NInoSetCompressed(ni); 822 if (a->flags & ATTR_IS_ENCRYPTED) { 823 if (a->flags & ATTR_COMPRESSION_MASK) { 824 ntfs_error(vi->i_sb, "Found encrypted and " 825 "compressed attribute."); 826 goto unm_err_out; 827 } 828 NInoSetEncrypted(ni); 829 } 830 if (a->flags & ATTR_IS_SPARSE) 831 NInoSetSparse(ni); 832 ir = (INDEX_ROOT*)((u8*)a + 833 le16_to_cpu(a->data.resident.value_offset)); 834 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length); 835 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) { 836 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is " 837 "corrupt."); 838 goto unm_err_out; 839 } 840 index_end = (u8*)&ir->index + 841 le32_to_cpu(ir->index.index_length); 842 if (index_end > ir_end) { 843 ntfs_error(vi->i_sb, "Directory index is corrupt."); 844 goto unm_err_out; 845 } 846 if (ir->type != AT_FILE_NAME) { 847 ntfs_error(vi->i_sb, "Indexed attribute is not " 848 "$FILE_NAME."); 849 goto unm_err_out; 850 } 851 if (ir->collation_rule != COLLATION_FILE_NAME) { 852 ntfs_error(vi->i_sb, "Index collation rule is not " 853 "COLLATION_FILE_NAME."); 854 goto unm_err_out; 855 } 856 ni->itype.index.collation_rule = ir->collation_rule; 857 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size); 858 if (ni->itype.index.block_size & 859 (ni->itype.index.block_size - 1)) { 860 ntfs_error(vi->i_sb, "Index block size (%u) is not a " 861 "power of two.", 862 ni->itype.index.block_size); 863 goto unm_err_out; 864 } 865 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) { 866 ntfs_error(vi->i_sb, "Index block size (%u) > " 867 "PAGE_CACHE_SIZE (%ld) is not " 868 "supported. Sorry.", 869 ni->itype.index.block_size, 870 PAGE_CACHE_SIZE); 871 err = -EOPNOTSUPP; 872 goto unm_err_out; 873 } 874 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) { 875 ntfs_error(vi->i_sb, "Index block size (%u) < " 876 "NTFS_BLOCK_SIZE (%i) is not " 877 "supported. Sorry.", 878 ni->itype.index.block_size, 879 NTFS_BLOCK_SIZE); 880 err = -EOPNOTSUPP; 881 goto unm_err_out; 882 } 883 ni->itype.index.block_size_bits = 884 ffs(ni->itype.index.block_size) - 1; 885 /* Determine the size of a vcn in the directory index. */ 886 if (vol->cluster_size <= ni->itype.index.block_size) { 887 ni->itype.index.vcn_size = vol->cluster_size; 888 ni->itype.index.vcn_size_bits = vol->cluster_size_bits; 889 } else { 890 ni->itype.index.vcn_size = vol->sector_size; 891 ni->itype.index.vcn_size_bits = vol->sector_size_bits; 892 } 893 894 /* Setup the index allocation attribute, even if not present. */ 895 NInoSetMstProtected(ni); 896 ni->type = AT_INDEX_ALLOCATION; 897 ni->name = I30; 898 ni->name_len = 4; 899 900 if (!(ir->index.flags & LARGE_INDEX)) { 901 /* No index allocation. */ 902 vi->i_size = ni->initialized_size = 903 ni->allocated_size = 0; 904 /* We are done with the mft record, so we release it. */ 905 ntfs_attr_put_search_ctx(ctx); 906 unmap_mft_record(ni); 907 m = NULL; 908 ctx = NULL; 909 goto skip_large_dir_stuff; 910 } /* LARGE_INDEX: Index allocation present. Setup state. */ 911 NInoSetIndexAllocPresent(ni); 912 /* Find index allocation attribute. */ 913 ntfs_attr_reinit_search_ctx(ctx); 914 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4, 915 CASE_SENSITIVE, 0, NULL, 0, ctx); 916 if (unlikely(err)) { 917 if (err == -ENOENT) 918 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION " 919 "attribute is not present but " 920 "$INDEX_ROOT indicated it is."); 921 else 922 ntfs_error(vi->i_sb, "Failed to lookup " 923 "$INDEX_ALLOCATION " 924 "attribute."); 925 goto unm_err_out; 926 } 927 a = ctx->attr; 928 if (!a->non_resident) { 929 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute " 930 "is resident."); 931 goto unm_err_out; 932 } 933 /* 934 * Ensure the attribute name is placed before the mapping pairs 935 * array. 936 */ 937 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= 938 le16_to_cpu( 939 a->data.non_resident.mapping_pairs_offset)))) { 940 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name " 941 "is placed after the mapping pairs " 942 "array."); 943 goto unm_err_out; 944 } 945 if (a->flags & ATTR_IS_ENCRYPTED) { 946 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute " 947 "is encrypted."); 948 goto unm_err_out; 949 } 950 if (a->flags & ATTR_IS_SPARSE) { 951 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute " 952 "is sparse."); 953 goto unm_err_out; 954 } 955 if (a->flags & ATTR_COMPRESSION_MASK) { 956 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute " 957 "is compressed."); 958 goto unm_err_out; 959 } 960 if (a->data.non_resident.lowest_vcn) { 961 ntfs_error(vi->i_sb, "First extent of " 962 "$INDEX_ALLOCATION attribute has non " 963 "zero lowest_vcn."); 964 goto unm_err_out; 965 } 966 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size); 967 ni->initialized_size = sle64_to_cpu( 968 a->data.non_resident.initialized_size); 969 ni->allocated_size = sle64_to_cpu( 970 a->data.non_resident.allocated_size); 971 /* 972 * We are done with the mft record, so we release it. Otherwise 973 * we would deadlock in ntfs_attr_iget(). 974 */ 975 ntfs_attr_put_search_ctx(ctx); 976 unmap_mft_record(ni); 977 m = NULL; 978 ctx = NULL; 979 /* Get the index bitmap attribute inode. */ 980 bvi = ntfs_attr_iget(vi, AT_BITMAP, I30, 4); 981 if (IS_ERR(bvi)) { 982 ntfs_error(vi->i_sb, "Failed to get bitmap attribute."); 983 err = PTR_ERR(bvi); 984 goto unm_err_out; 985 } 986 bni = NTFS_I(bvi); 987 if (NInoCompressed(bni) || NInoEncrypted(bni) || 988 NInoSparse(bni)) { 989 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed " 990 "and/or encrypted and/or sparse."); 991 goto iput_unm_err_out; 992 } 993 /* Consistency check bitmap size vs. index allocation size. */ 994 bvi_size = i_size_read(bvi); 995 if ((bvi_size << 3) < (vi->i_size >> 996 ni->itype.index.block_size_bits)) { 997 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) " 998 "for index allocation (0x%llx).", 999 bvi_size << 3, vi->i_size); 1000 goto iput_unm_err_out; 1001 } 1002 /* No longer need the bitmap attribute inode. */ 1003 iput(bvi); 1004 skip_large_dir_stuff: 1005 /* Setup the operations for this inode. */ 1006 vi->i_op = &ntfs_dir_inode_ops; 1007 vi->i_fop = &ntfs_dir_ops; 1008 } else { 1009 /* It is a file. */ 1010 ntfs_attr_reinit_search_ctx(ctx); 1011 1012 /* Setup the data attribute, even if not present. */ 1013 ni->type = AT_DATA; 1014 ni->name = NULL; 1015 ni->name_len = 0; 1016 1017 /* Find first extent of the unnamed data attribute. */ 1018 err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx); 1019 if (unlikely(err)) { 1020 vi->i_size = ni->initialized_size = 1021 ni->allocated_size = 0; 1022 if (err != -ENOENT) { 1023 ntfs_error(vi->i_sb, "Failed to lookup $DATA " 1024 "attribute."); 1025 goto unm_err_out; 1026 } 1027 /* 1028 * FILE_Secure does not have an unnamed $DATA 1029 * attribute, so we special case it here. 1030 */ 1031 if (vi->i_ino == FILE_Secure) 1032 goto no_data_attr_special_case; 1033 /* 1034 * Most if not all the system files in the $Extend 1035 * system directory do not have unnamed data 1036 * attributes so we need to check if the parent 1037 * directory of the file is FILE_Extend and if it is 1038 * ignore this error. To do this we need to get the 1039 * name of this inode from the mft record as the name 1040 * contains the back reference to the parent directory. 1041 */ 1042 if (ntfs_is_extended_system_file(ctx) > 0) 1043 goto no_data_attr_special_case; 1044 // FIXME: File is corrupt! Hot-fix with empty data 1045 // attribute if recovery option is set. 1046 ntfs_error(vi->i_sb, "$DATA attribute is missing."); 1047 goto unm_err_out; 1048 } 1049 a = ctx->attr; 1050 /* Setup the state. */ 1051 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) { 1052 if (a->flags & ATTR_COMPRESSION_MASK) { 1053 NInoSetCompressed(ni); 1054 if (vol->cluster_size > 4096) { 1055 ntfs_error(vi->i_sb, "Found " 1056 "compressed data but " 1057 "compression is " 1058 "disabled due to " 1059 "cluster size (%i) > " 1060 "4kiB.", 1061 vol->cluster_size); 1062 goto unm_err_out; 1063 } 1064 if ((a->flags & ATTR_COMPRESSION_MASK) 1065 != ATTR_IS_COMPRESSED) { 1066 ntfs_error(vi->i_sb, "Found unknown " 1067 "compression method " 1068 "or corrupt file."); 1069 goto unm_err_out; 1070 } 1071 } 1072 if (a->flags & ATTR_IS_SPARSE) 1073 NInoSetSparse(ni); 1074 } 1075 if (a->flags & ATTR_IS_ENCRYPTED) { 1076 if (NInoCompressed(ni)) { 1077 ntfs_error(vi->i_sb, "Found encrypted and " 1078 "compressed data."); 1079 goto unm_err_out; 1080 } 1081 NInoSetEncrypted(ni); 1082 } 1083 if (a->non_resident) { 1084 NInoSetNonResident(ni); 1085 if (NInoCompressed(ni) || NInoSparse(ni)) { 1086 if (NInoCompressed(ni) && a->data.non_resident. 1087 compression_unit != 4) { 1088 ntfs_error(vi->i_sb, "Found " 1089 "non-standard " 1090 "compression unit (%u " 1091 "instead of 4). " 1092 "Cannot handle this.", 1093 a->data.non_resident. 1094 compression_unit); 1095 err = -EOPNOTSUPP; 1096 goto unm_err_out; 1097 } 1098 if (a->data.non_resident.compression_unit) { 1099 ni->itype.compressed.block_size = 1U << 1100 (a->data.non_resident. 1101 compression_unit + 1102 vol->cluster_size_bits); 1103 ni->itype.compressed.block_size_bits = 1104 ffs(ni->itype. 1105 compressed. 1106 block_size) - 1; 1107 ni->itype.compressed.block_clusters = 1108 1U << a->data. 1109 non_resident. 1110 compression_unit; 1111 } else { 1112 ni->itype.compressed.block_size = 0; 1113 ni->itype.compressed.block_size_bits = 1114 0; 1115 ni->itype.compressed.block_clusters = 1116 0; 1117 } 1118 ni->itype.compressed.size = sle64_to_cpu( 1119 a->data.non_resident. 1120 compressed_size); 1121 } 1122 if (a->data.non_resident.lowest_vcn) { 1123 ntfs_error(vi->i_sb, "First extent of $DATA " 1124 "attribute has non zero " 1125 "lowest_vcn."); 1126 goto unm_err_out; 1127 } 1128 vi->i_size = sle64_to_cpu( 1129 a->data.non_resident.data_size); 1130 ni->initialized_size = sle64_to_cpu( 1131 a->data.non_resident.initialized_size); 1132 ni->allocated_size = sle64_to_cpu( 1133 a->data.non_resident.allocated_size); 1134 } else { /* Resident attribute. */ 1135 vi->i_size = ni->initialized_size = le32_to_cpu( 1136 a->data.resident.value_length); 1137 ni->allocated_size = le32_to_cpu(a->length) - 1138 le16_to_cpu( 1139 a->data.resident.value_offset); 1140 if (vi->i_size > ni->allocated_size) { 1141 ntfs_error(vi->i_sb, "Resident data attribute " 1142 "is corrupt (size exceeds " 1143 "allocation)."); 1144 goto unm_err_out; 1145 } 1146 } 1147 no_data_attr_special_case: 1148 /* We are done with the mft record, so we release it. */ 1149 ntfs_attr_put_search_ctx(ctx); 1150 unmap_mft_record(ni); 1151 m = NULL; 1152 ctx = NULL; 1153 /* Setup the operations for this inode. */ 1154 vi->i_op = &ntfs_file_inode_ops; 1155 vi->i_fop = &ntfs_file_ops; 1156 } 1157 if (NInoMstProtected(ni)) 1158 vi->i_mapping->a_ops = &ntfs_mst_aops; 1159 else 1160 vi->i_mapping->a_ops = &ntfs_aops; 1161 /* 1162 * The number of 512-byte blocks used on disk (for stat). This is in so 1163 * far inaccurate as it doesn't account for any named streams or other 1164 * special non-resident attributes, but that is how Windows works, too, 1165 * so we are at least consistent with Windows, if not entirely 1166 * consistent with the Linux Way. Doing it the Linux Way would cause a 1167 * significant slowdown as it would involve iterating over all 1168 * attributes in the mft record and adding the allocated/compressed 1169 * sizes of all non-resident attributes present to give us the Linux 1170 * correct size that should go into i_blocks (after division by 512). 1171 */ 1172 if (S_ISREG(vi->i_mode) && (NInoCompressed(ni) || NInoSparse(ni))) 1173 vi->i_blocks = ni->itype.compressed.size >> 9; 1174 else 1175 vi->i_blocks = ni->allocated_size >> 9; 1176 ntfs_debug("Done."); 1177 return 0; 1178 iput_unm_err_out: 1179 iput(bvi); 1180 unm_err_out: 1181 if (!err) 1182 err = -EIO; 1183 if (ctx) 1184 ntfs_attr_put_search_ctx(ctx); 1185 if (m) 1186 unmap_mft_record(ni); 1187 err_out: 1188 ntfs_error(vol->sb, "Failed with error code %i. Marking corrupt " 1189 "inode 0x%lx as bad. Run chkdsk.", err, vi->i_ino); 1190 make_bad_inode(vi); 1191 if (err != -EOPNOTSUPP && err != -ENOMEM) 1192 NVolSetErrors(vol); 1193 return err; 1194 } 1195 1196 /** 1197 * ntfs_read_locked_attr_inode - read an attribute inode from its base inode 1198 * @base_vi: base inode 1199 * @vi: attribute inode to read 1200 * 1201 * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the 1202 * attribute inode described by @vi into memory from the base mft record 1203 * described by @base_ni. 1204 * 1205 * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for 1206 * reading and looks up the attribute described by @vi before setting up the 1207 * necessary fields in @vi as well as initializing the ntfs inode. 1208 * 1209 * Q: What locks are held when the function is called? 1210 * A: i_state has I_NEW set, hence the inode is locked, also 1211 * i_count is set to 1, so it is not going to go away 1212 * 1213 * Return 0 on success and -errno on error. In the error case, the inode will 1214 * have had make_bad_inode() executed on it. 1215 * 1216 * Note this cannot be called for AT_INDEX_ALLOCATION. 1217 */ 1218 static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi) 1219 { 1220 ntfs_volume *vol = NTFS_SB(vi->i_sb); 1221 ntfs_inode *ni, *base_ni; 1222 MFT_RECORD *m; 1223 ATTR_RECORD *a; 1224 ntfs_attr_search_ctx *ctx; 1225 int err = 0; 1226 1227 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino); 1228 1229 ntfs_init_big_inode(vi); 1230 1231 ni = NTFS_I(vi); 1232 base_ni = NTFS_I(base_vi); 1233 1234 /* Just mirror the values from the base inode. */ 1235 vi->i_version = base_vi->i_version; 1236 vi->i_uid = base_vi->i_uid; 1237 vi->i_gid = base_vi->i_gid; 1238 vi->i_nlink = base_vi->i_nlink; 1239 vi->i_mtime = base_vi->i_mtime; 1240 vi->i_ctime = base_vi->i_ctime; 1241 vi->i_atime = base_vi->i_atime; 1242 vi->i_generation = ni->seq_no = base_ni->seq_no; 1243 1244 /* Set inode type to zero but preserve permissions. */ 1245 vi->i_mode = base_vi->i_mode & ~S_IFMT; 1246 1247 m = map_mft_record(base_ni); 1248 if (IS_ERR(m)) { 1249 err = PTR_ERR(m); 1250 goto err_out; 1251 } 1252 ctx = ntfs_attr_get_search_ctx(base_ni, m); 1253 if (!ctx) { 1254 err = -ENOMEM; 1255 goto unm_err_out; 1256 } 1257 /* Find the attribute. */ 1258 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len, 1259 CASE_SENSITIVE, 0, NULL, 0, ctx); 1260 if (unlikely(err)) 1261 goto unm_err_out; 1262 a = ctx->attr; 1263 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) { 1264 if (a->flags & ATTR_COMPRESSION_MASK) { 1265 NInoSetCompressed(ni); 1266 if ((ni->type != AT_DATA) || (ni->type == AT_DATA && 1267 ni->name_len)) { 1268 ntfs_error(vi->i_sb, "Found compressed " 1269 "non-data or named data " 1270 "attribute. Please report " 1271 "you saw this message to " 1272 "linux-ntfs-dev@lists." 1273 "sourceforge.net"); 1274 goto unm_err_out; 1275 } 1276 if (vol->cluster_size > 4096) { 1277 ntfs_error(vi->i_sb, "Found compressed " 1278 "attribute but compression is " 1279 "disabled due to cluster size " 1280 "(%i) > 4kiB.", 1281 vol->cluster_size); 1282 goto unm_err_out; 1283 } 1284 if ((a->flags & ATTR_COMPRESSION_MASK) != 1285 ATTR_IS_COMPRESSED) { 1286 ntfs_error(vi->i_sb, "Found unknown " 1287 "compression method."); 1288 goto unm_err_out; 1289 } 1290 } 1291 /* 1292 * The compressed/sparse flag set in an index root just means 1293 * to compress all files. 1294 */ 1295 if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) { 1296 ntfs_error(vi->i_sb, "Found mst protected attribute " 1297 "but the attribute is %s. Please " 1298 "report you saw this message to " 1299 "linux-ntfs-dev@lists.sourceforge.net", 1300 NInoCompressed(ni) ? "compressed" : 1301 "sparse"); 1302 goto unm_err_out; 1303 } 1304 if (a->flags & ATTR_IS_SPARSE) 1305 NInoSetSparse(ni); 1306 } 1307 if (a->flags & ATTR_IS_ENCRYPTED) { 1308 if (NInoCompressed(ni)) { 1309 ntfs_error(vi->i_sb, "Found encrypted and compressed " 1310 "data."); 1311 goto unm_err_out; 1312 } 1313 /* 1314 * The encryption flag set in an index root just means to 1315 * encrypt all files. 1316 */ 1317 if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) { 1318 ntfs_error(vi->i_sb, "Found mst protected attribute " 1319 "but the attribute is encrypted. " 1320 "Please report you saw this message " 1321 "to linux-ntfs-dev@lists.sourceforge." 1322 "net"); 1323 goto unm_err_out; 1324 } 1325 if (ni->type != AT_DATA) { 1326 ntfs_error(vi->i_sb, "Found encrypted non-data " 1327 "attribute."); 1328 goto unm_err_out; 1329 } 1330 NInoSetEncrypted(ni); 1331 } 1332 if (!a->non_resident) { 1333 /* Ensure the attribute name is placed before the value. */ 1334 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= 1335 le16_to_cpu(a->data.resident.value_offset)))) { 1336 ntfs_error(vol->sb, "Attribute name is placed after " 1337 "the attribute value."); 1338 goto unm_err_out; 1339 } 1340 if (NInoMstProtected(ni)) { 1341 ntfs_error(vi->i_sb, "Found mst protected attribute " 1342 "but the attribute is resident. " 1343 "Please report you saw this message to " 1344 "linux-ntfs-dev@lists.sourceforge.net"); 1345 goto unm_err_out; 1346 } 1347 vi->i_size = ni->initialized_size = le32_to_cpu( 1348 a->data.resident.value_length); 1349 ni->allocated_size = le32_to_cpu(a->length) - 1350 le16_to_cpu(a->data.resident.value_offset); 1351 if (vi->i_size > ni->allocated_size) { 1352 ntfs_error(vi->i_sb, "Resident attribute is corrupt " 1353 "(size exceeds allocation)."); 1354 goto unm_err_out; 1355 } 1356 } else { 1357 NInoSetNonResident(ni); 1358 /* 1359 * Ensure the attribute name is placed before the mapping pairs 1360 * array. 1361 */ 1362 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= 1363 le16_to_cpu( 1364 a->data.non_resident.mapping_pairs_offset)))) { 1365 ntfs_error(vol->sb, "Attribute name is placed after " 1366 "the mapping pairs array."); 1367 goto unm_err_out; 1368 } 1369 if (NInoCompressed(ni) || NInoSparse(ni)) { 1370 if (NInoCompressed(ni) && a->data.non_resident. 1371 compression_unit != 4) { 1372 ntfs_error(vi->i_sb, "Found non-standard " 1373 "compression unit (%u instead " 1374 "of 4). Cannot handle this.", 1375 a->data.non_resident. 1376 compression_unit); 1377 err = -EOPNOTSUPP; 1378 goto unm_err_out; 1379 } 1380 if (a->data.non_resident.compression_unit) { 1381 ni->itype.compressed.block_size = 1U << 1382 (a->data.non_resident. 1383 compression_unit + 1384 vol->cluster_size_bits); 1385 ni->itype.compressed.block_size_bits = 1386 ffs(ni->itype.compressed. 1387 block_size) - 1; 1388 ni->itype.compressed.block_clusters = 1U << 1389 a->data.non_resident. 1390 compression_unit; 1391 } else { 1392 ni->itype.compressed.block_size = 0; 1393 ni->itype.compressed.block_size_bits = 0; 1394 ni->itype.compressed.block_clusters = 0; 1395 } 1396 ni->itype.compressed.size = sle64_to_cpu( 1397 a->data.non_resident.compressed_size); 1398 } 1399 if (a->data.non_resident.lowest_vcn) { 1400 ntfs_error(vi->i_sb, "First extent of attribute has " 1401 "non-zero lowest_vcn."); 1402 goto unm_err_out; 1403 } 1404 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size); 1405 ni->initialized_size = sle64_to_cpu( 1406 a->data.non_resident.initialized_size); 1407 ni->allocated_size = sle64_to_cpu( 1408 a->data.non_resident.allocated_size); 1409 } 1410 if (NInoMstProtected(ni)) 1411 vi->i_mapping->a_ops = &ntfs_mst_aops; 1412 else 1413 vi->i_mapping->a_ops = &ntfs_aops; 1414 if ((NInoCompressed(ni) || NInoSparse(ni)) && ni->type != AT_INDEX_ROOT) 1415 vi->i_blocks = ni->itype.compressed.size >> 9; 1416 else 1417 vi->i_blocks = ni->allocated_size >> 9; 1418 /* 1419 * Make sure the base inode does not go away and attach it to the 1420 * attribute inode. 1421 */ 1422 igrab(base_vi); 1423 ni->ext.base_ntfs_ino = base_ni; 1424 ni->nr_extents = -1; 1425 1426 ntfs_attr_put_search_ctx(ctx); 1427 unmap_mft_record(base_ni); 1428 1429 ntfs_debug("Done."); 1430 return 0; 1431 1432 unm_err_out: 1433 if (!err) 1434 err = -EIO; 1435 if (ctx) 1436 ntfs_attr_put_search_ctx(ctx); 1437 unmap_mft_record(base_ni); 1438 err_out: 1439 ntfs_error(vol->sb, "Failed with error code %i while reading attribute " 1440 "inode (mft_no 0x%lx, type 0x%x, name_len %i). " 1441 "Marking corrupt inode and base inode 0x%lx as bad. " 1442 "Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len, 1443 base_vi->i_ino); 1444 make_bad_inode(vi); 1445 if (err != -ENOMEM) 1446 NVolSetErrors(vol); 1447 return err; 1448 } 1449 1450 /** 1451 * ntfs_read_locked_index_inode - read an index inode from its base inode 1452 * @base_vi: base inode 1453 * @vi: index inode to read 1454 * 1455 * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the 1456 * index inode described by @vi into memory from the base mft record described 1457 * by @base_ni. 1458 * 1459 * ntfs_read_locked_index_inode() maps, pins and locks the base inode for 1460 * reading and looks up the attributes relating to the index described by @vi 1461 * before setting up the necessary fields in @vi as well as initializing the 1462 * ntfs inode. 1463 * 1464 * Note, index inodes are essentially attribute inodes (NInoAttr() is true) 1465 * with the attribute type set to AT_INDEX_ALLOCATION. Apart from that, they 1466 * are setup like directory inodes since directories are a special case of 1467 * indices ao they need to be treated in much the same way. Most importantly, 1468 * for small indices the index allocation attribute might not actually exist. 1469 * However, the index root attribute always exists but this does not need to 1470 * have an inode associated with it and this is why we define a new inode type 1471 * index. Also, like for directories, we need to have an attribute inode for 1472 * the bitmap attribute corresponding to the index allocation attribute and we 1473 * can store this in the appropriate field of the inode, just like we do for 1474 * normal directory inodes. 1475 * 1476 * Q: What locks are held when the function is called? 1477 * A: i_state has I_NEW set, hence the inode is locked, also 1478 * i_count is set to 1, so it is not going to go away 1479 * 1480 * Return 0 on success and -errno on error. In the error case, the inode will 1481 * have had make_bad_inode() executed on it. 1482 */ 1483 static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi) 1484 { 1485 loff_t bvi_size; 1486 ntfs_volume *vol = NTFS_SB(vi->i_sb); 1487 ntfs_inode *ni, *base_ni, *bni; 1488 struct inode *bvi; 1489 MFT_RECORD *m; 1490 ATTR_RECORD *a; 1491 ntfs_attr_search_ctx *ctx; 1492 INDEX_ROOT *ir; 1493 u8 *ir_end, *index_end; 1494 int err = 0; 1495 1496 ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino); 1497 ntfs_init_big_inode(vi); 1498 ni = NTFS_I(vi); 1499 base_ni = NTFS_I(base_vi); 1500 /* Just mirror the values from the base inode. */ 1501 vi->i_version = base_vi->i_version; 1502 vi->i_uid = base_vi->i_uid; 1503 vi->i_gid = base_vi->i_gid; 1504 vi->i_nlink = base_vi->i_nlink; 1505 vi->i_mtime = base_vi->i_mtime; 1506 vi->i_ctime = base_vi->i_ctime; 1507 vi->i_atime = base_vi->i_atime; 1508 vi->i_generation = ni->seq_no = base_ni->seq_no; 1509 /* Set inode type to zero but preserve permissions. */ 1510 vi->i_mode = base_vi->i_mode & ~S_IFMT; 1511 /* Map the mft record for the base inode. */ 1512 m = map_mft_record(base_ni); 1513 if (IS_ERR(m)) { 1514 err = PTR_ERR(m); 1515 goto err_out; 1516 } 1517 ctx = ntfs_attr_get_search_ctx(base_ni, m); 1518 if (!ctx) { 1519 err = -ENOMEM; 1520 goto unm_err_out; 1521 } 1522 /* Find the index root attribute. */ 1523 err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len, 1524 CASE_SENSITIVE, 0, NULL, 0, ctx); 1525 if (unlikely(err)) { 1526 if (err == -ENOENT) 1527 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is " 1528 "missing."); 1529 goto unm_err_out; 1530 } 1531 a = ctx->attr; 1532 /* Set up the state. */ 1533 if (unlikely(a->non_resident)) { 1534 ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident."); 1535 goto unm_err_out; 1536 } 1537 /* Ensure the attribute name is placed before the value. */ 1538 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= 1539 le16_to_cpu(a->data.resident.value_offset)))) { 1540 ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed " 1541 "after the attribute value."); 1542 goto unm_err_out; 1543 } 1544 /* 1545 * Compressed/encrypted/sparse index root is not allowed, except for 1546 * directories of course but those are not dealt with here. 1547 */ 1548 if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED | 1549 ATTR_IS_SPARSE)) { 1550 ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index " 1551 "root attribute."); 1552 goto unm_err_out; 1553 } 1554 ir = (INDEX_ROOT*)((u8*)a + le16_to_cpu(a->data.resident.value_offset)); 1555 ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length); 1556 if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) { 1557 ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt."); 1558 goto unm_err_out; 1559 } 1560 index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length); 1561 if (index_end > ir_end) { 1562 ntfs_error(vi->i_sb, "Index is corrupt."); 1563 goto unm_err_out; 1564 } 1565 if (ir->type) { 1566 ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).", 1567 le32_to_cpu(ir->type)); 1568 goto unm_err_out; 1569 } 1570 ni->itype.index.collation_rule = ir->collation_rule; 1571 ntfs_debug("Index collation rule is 0x%x.", 1572 le32_to_cpu(ir->collation_rule)); 1573 ni->itype.index.block_size = le32_to_cpu(ir->index_block_size); 1574 if (!is_power_of_2(ni->itype.index.block_size)) { 1575 ntfs_error(vi->i_sb, "Index block size (%u) is not a power of " 1576 "two.", ni->itype.index.block_size); 1577 goto unm_err_out; 1578 } 1579 if (ni->itype.index.block_size > PAGE_CACHE_SIZE) { 1580 ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_CACHE_SIZE " 1581 "(%ld) is not supported. Sorry.", 1582 ni->itype.index.block_size, PAGE_CACHE_SIZE); 1583 err = -EOPNOTSUPP; 1584 goto unm_err_out; 1585 } 1586 if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) { 1587 ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE " 1588 "(%i) is not supported. Sorry.", 1589 ni->itype.index.block_size, NTFS_BLOCK_SIZE); 1590 err = -EOPNOTSUPP; 1591 goto unm_err_out; 1592 } 1593 ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1; 1594 /* Determine the size of a vcn in the index. */ 1595 if (vol->cluster_size <= ni->itype.index.block_size) { 1596 ni->itype.index.vcn_size = vol->cluster_size; 1597 ni->itype.index.vcn_size_bits = vol->cluster_size_bits; 1598 } else { 1599 ni->itype.index.vcn_size = vol->sector_size; 1600 ni->itype.index.vcn_size_bits = vol->sector_size_bits; 1601 } 1602 /* Check for presence of index allocation attribute. */ 1603 if (!(ir->index.flags & LARGE_INDEX)) { 1604 /* No index allocation. */ 1605 vi->i_size = ni->initialized_size = ni->allocated_size = 0; 1606 /* We are done with the mft record, so we release it. */ 1607 ntfs_attr_put_search_ctx(ctx); 1608 unmap_mft_record(base_ni); 1609 m = NULL; 1610 ctx = NULL; 1611 goto skip_large_index_stuff; 1612 } /* LARGE_INDEX: Index allocation present. Setup state. */ 1613 NInoSetIndexAllocPresent(ni); 1614 /* Find index allocation attribute. */ 1615 ntfs_attr_reinit_search_ctx(ctx); 1616 err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len, 1617 CASE_SENSITIVE, 0, NULL, 0, ctx); 1618 if (unlikely(err)) { 1619 if (err == -ENOENT) 1620 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is " 1621 "not present but $INDEX_ROOT " 1622 "indicated it is."); 1623 else 1624 ntfs_error(vi->i_sb, "Failed to lookup " 1625 "$INDEX_ALLOCATION attribute."); 1626 goto unm_err_out; 1627 } 1628 a = ctx->attr; 1629 if (!a->non_resident) { 1630 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is " 1631 "resident."); 1632 goto unm_err_out; 1633 } 1634 /* 1635 * Ensure the attribute name is placed before the mapping pairs array. 1636 */ 1637 if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= 1638 le16_to_cpu( 1639 a->data.non_resident.mapping_pairs_offset)))) { 1640 ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is " 1641 "placed after the mapping pairs array."); 1642 goto unm_err_out; 1643 } 1644 if (a->flags & ATTR_IS_ENCRYPTED) { 1645 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is " 1646 "encrypted."); 1647 goto unm_err_out; 1648 } 1649 if (a->flags & ATTR_IS_SPARSE) { 1650 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse."); 1651 goto unm_err_out; 1652 } 1653 if (a->flags & ATTR_COMPRESSION_MASK) { 1654 ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is " 1655 "compressed."); 1656 goto unm_err_out; 1657 } 1658 if (a->data.non_resident.lowest_vcn) { 1659 ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION " 1660 "attribute has non zero lowest_vcn."); 1661 goto unm_err_out; 1662 } 1663 vi->i_size = sle64_to_cpu(a->data.non_resident.data_size); 1664 ni->initialized_size = sle64_to_cpu( 1665 a->data.non_resident.initialized_size); 1666 ni->allocated_size = sle64_to_cpu(a->data.non_resident.allocated_size); 1667 /* 1668 * We are done with the mft record, so we release it. Otherwise 1669 * we would deadlock in ntfs_attr_iget(). 1670 */ 1671 ntfs_attr_put_search_ctx(ctx); 1672 unmap_mft_record(base_ni); 1673 m = NULL; 1674 ctx = NULL; 1675 /* Get the index bitmap attribute inode. */ 1676 bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len); 1677 if (IS_ERR(bvi)) { 1678 ntfs_error(vi->i_sb, "Failed to get bitmap attribute."); 1679 err = PTR_ERR(bvi); 1680 goto unm_err_out; 1681 } 1682 bni = NTFS_I(bvi); 1683 if (NInoCompressed(bni) || NInoEncrypted(bni) || 1684 NInoSparse(bni)) { 1685 ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or " 1686 "encrypted and/or sparse."); 1687 goto iput_unm_err_out; 1688 } 1689 /* Consistency check bitmap size vs. index allocation size. */ 1690 bvi_size = i_size_read(bvi); 1691 if ((bvi_size << 3) < (vi->i_size >> ni->itype.index.block_size_bits)) { 1692 ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for " 1693 "index allocation (0x%llx).", bvi_size << 3, 1694 vi->i_size); 1695 goto iput_unm_err_out; 1696 } 1697 iput(bvi); 1698 skip_large_index_stuff: 1699 /* Setup the operations for this index inode. */ 1700 vi->i_op = NULL; 1701 vi->i_fop = NULL; 1702 vi->i_mapping->a_ops = &ntfs_mst_aops; 1703 vi->i_blocks = ni->allocated_size >> 9; 1704 /* 1705 * Make sure the base inode doesn't go away and attach it to the 1706 * index inode. 1707 */ 1708 igrab(base_vi); 1709 ni->ext.base_ntfs_ino = base_ni; 1710 ni->nr_extents = -1; 1711 1712 ntfs_debug("Done."); 1713 return 0; 1714 iput_unm_err_out: 1715 iput(bvi); 1716 unm_err_out: 1717 if (!err) 1718 err = -EIO; 1719 if (ctx) 1720 ntfs_attr_put_search_ctx(ctx); 1721 if (m) 1722 unmap_mft_record(base_ni); 1723 err_out: 1724 ntfs_error(vi->i_sb, "Failed with error code %i while reading index " 1725 "inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino, 1726 ni->name_len); 1727 make_bad_inode(vi); 1728 if (err != -EOPNOTSUPP && err != -ENOMEM) 1729 NVolSetErrors(vol); 1730 return err; 1731 } 1732 1733 /* 1734 * The MFT inode has special locking, so teach the lock validator 1735 * about this by splitting off the locking rules of the MFT from 1736 * the locking rules of other inodes. The MFT inode can never be 1737 * accessed from the VFS side (or even internally), only by the 1738 * map_mft functions. 1739 */ 1740 static struct lock_class_key mft_ni_runlist_lock_key, mft_ni_mrec_lock_key; 1741 1742 /** 1743 * ntfs_read_inode_mount - special read_inode for mount time use only 1744 * @vi: inode to read 1745 * 1746 * Read inode FILE_MFT at mount time, only called with super_block lock 1747 * held from within the read_super() code path. 1748 * 1749 * This function exists because when it is called the page cache for $MFT/$DATA 1750 * is not initialized and hence we cannot get at the contents of mft records 1751 * by calling map_mft_record*(). 1752 * 1753 * Further it needs to cope with the circular references problem, i.e. cannot 1754 * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because 1755 * we do not know where the other extent mft records are yet and again, because 1756 * we cannot call map_mft_record*() yet. Obviously this applies only when an 1757 * attribute list is actually present in $MFT inode. 1758 * 1759 * We solve these problems by starting with the $DATA attribute before anything 1760 * else and iterating using ntfs_attr_lookup($DATA) over all extents. As each 1761 * extent is found, we ntfs_mapping_pairs_decompress() including the implied 1762 * ntfs_runlists_merge(). Each step of the iteration necessarily provides 1763 * sufficient information for the next step to complete. 1764 * 1765 * This should work but there are two possible pit falls (see inline comments 1766 * below), but only time will tell if they are real pits or just smoke... 1767 */ 1768 int ntfs_read_inode_mount(struct inode *vi) 1769 { 1770 VCN next_vcn, last_vcn, highest_vcn; 1771 s64 block; 1772 struct super_block *sb = vi->i_sb; 1773 ntfs_volume *vol = NTFS_SB(sb); 1774 struct buffer_head *bh; 1775 ntfs_inode *ni; 1776 MFT_RECORD *m = NULL; 1777 ATTR_RECORD *a; 1778 ntfs_attr_search_ctx *ctx; 1779 unsigned int i, nr_blocks; 1780 int err; 1781 1782 ntfs_debug("Entering."); 1783 1784 /* Initialize the ntfs specific part of @vi. */ 1785 ntfs_init_big_inode(vi); 1786 1787 ni = NTFS_I(vi); 1788 1789 /* Setup the data attribute. It is special as it is mst protected. */ 1790 NInoSetNonResident(ni); 1791 NInoSetMstProtected(ni); 1792 NInoSetSparseDisabled(ni); 1793 ni->type = AT_DATA; 1794 ni->name = NULL; 1795 ni->name_len = 0; 1796 /* 1797 * This sets up our little cheat allowing us to reuse the async read io 1798 * completion handler for directories. 1799 */ 1800 ni->itype.index.block_size = vol->mft_record_size; 1801 ni->itype.index.block_size_bits = vol->mft_record_size_bits; 1802 1803 /* Very important! Needed to be able to call map_mft_record*(). */ 1804 vol->mft_ino = vi; 1805 1806 /* Allocate enough memory to read the first mft record. */ 1807 if (vol->mft_record_size > 64 * 1024) { 1808 ntfs_error(sb, "Unsupported mft record size %i (max 64kiB).", 1809 vol->mft_record_size); 1810 goto err_out; 1811 } 1812 i = vol->mft_record_size; 1813 if (i < sb->s_blocksize) 1814 i = sb->s_blocksize; 1815 m = (MFT_RECORD*)ntfs_malloc_nofs(i); 1816 if (!m) { 1817 ntfs_error(sb, "Failed to allocate buffer for $MFT record 0."); 1818 goto err_out; 1819 } 1820 1821 /* Determine the first block of the $MFT/$DATA attribute. */ 1822 block = vol->mft_lcn << vol->cluster_size_bits >> 1823 sb->s_blocksize_bits; 1824 nr_blocks = vol->mft_record_size >> sb->s_blocksize_bits; 1825 if (!nr_blocks) 1826 nr_blocks = 1; 1827 1828 /* Load $MFT/$DATA's first mft record. */ 1829 for (i = 0; i < nr_blocks; i++) { 1830 bh = sb_bread(sb, block++); 1831 if (!bh) { 1832 ntfs_error(sb, "Device read failed."); 1833 goto err_out; 1834 } 1835 memcpy((char*)m + (i << sb->s_blocksize_bits), bh->b_data, 1836 sb->s_blocksize); 1837 brelse(bh); 1838 } 1839 1840 /* Apply the mst fixups. */ 1841 if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) { 1842 /* FIXME: Try to use the $MFTMirr now. */ 1843 ntfs_error(sb, "MST fixup failed. $MFT is corrupt."); 1844 goto err_out; 1845 } 1846 1847 /* Need this to sanity check attribute list references to $MFT. */ 1848 vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number); 1849 1850 /* Provides readpage() and sync_page() for map_mft_record(). */ 1851 vi->i_mapping->a_ops = &ntfs_mst_aops; 1852 1853 ctx = ntfs_attr_get_search_ctx(ni, m); 1854 if (!ctx) { 1855 err = -ENOMEM; 1856 goto err_out; 1857 } 1858 1859 /* Find the attribute list attribute if present. */ 1860 err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx); 1861 if (err) { 1862 if (unlikely(err != -ENOENT)) { 1863 ntfs_error(sb, "Failed to lookup attribute list " 1864 "attribute. You should run chkdsk."); 1865 goto put_err_out; 1866 } 1867 } else /* if (!err) */ { 1868 ATTR_LIST_ENTRY *al_entry, *next_al_entry; 1869 u8 *al_end; 1870 static const char *es = " Not allowed. $MFT is corrupt. " 1871 "You should run chkdsk."; 1872 1873 ntfs_debug("Attribute list attribute found in $MFT."); 1874 NInoSetAttrList(ni); 1875 a = ctx->attr; 1876 if (a->flags & ATTR_COMPRESSION_MASK) { 1877 ntfs_error(sb, "Attribute list attribute is " 1878 "compressed.%s", es); 1879 goto put_err_out; 1880 } 1881 if (a->flags & ATTR_IS_ENCRYPTED || 1882 a->flags & ATTR_IS_SPARSE) { 1883 if (a->non_resident) { 1884 ntfs_error(sb, "Non-resident attribute list " 1885 "attribute is encrypted/" 1886 "sparse.%s", es); 1887 goto put_err_out; 1888 } 1889 ntfs_warning(sb, "Resident attribute list attribute " 1890 "in $MFT system file is marked " 1891 "encrypted/sparse which is not true. " 1892 "However, Windows allows this and " 1893 "chkdsk does not detect or correct it " 1894 "so we will just ignore the invalid " 1895 "flags and pretend they are not set."); 1896 } 1897 /* Now allocate memory for the attribute list. */ 1898 ni->attr_list_size = (u32)ntfs_attr_size(a); 1899 ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size); 1900 if (!ni->attr_list) { 1901 ntfs_error(sb, "Not enough memory to allocate buffer " 1902 "for attribute list."); 1903 goto put_err_out; 1904 } 1905 if (a->non_resident) { 1906 NInoSetAttrListNonResident(ni); 1907 if (a->data.non_resident.lowest_vcn) { 1908 ntfs_error(sb, "Attribute list has non zero " 1909 "lowest_vcn. $MFT is corrupt. " 1910 "You should run chkdsk."); 1911 goto put_err_out; 1912 } 1913 /* Setup the runlist. */ 1914 ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol, 1915 a, NULL); 1916 if (IS_ERR(ni->attr_list_rl.rl)) { 1917 err = PTR_ERR(ni->attr_list_rl.rl); 1918 ni->attr_list_rl.rl = NULL; 1919 ntfs_error(sb, "Mapping pairs decompression " 1920 "failed with error code %i.", 1921 -err); 1922 goto put_err_out; 1923 } 1924 /* Now load the attribute list. */ 1925 if ((err = load_attribute_list(vol, &ni->attr_list_rl, 1926 ni->attr_list, ni->attr_list_size, 1927 sle64_to_cpu(a->data. 1928 non_resident.initialized_size)))) { 1929 ntfs_error(sb, "Failed to load attribute list " 1930 "attribute with error code %i.", 1931 -err); 1932 goto put_err_out; 1933 } 1934 } else /* if (!ctx.attr->non_resident) */ { 1935 if ((u8*)a + le16_to_cpu( 1936 a->data.resident.value_offset) + 1937 le32_to_cpu( 1938 a->data.resident.value_length) > 1939 (u8*)ctx->mrec + vol->mft_record_size) { 1940 ntfs_error(sb, "Corrupt attribute list " 1941 "attribute."); 1942 goto put_err_out; 1943 } 1944 /* Now copy the attribute list. */ 1945 memcpy(ni->attr_list, (u8*)a + le16_to_cpu( 1946 a->data.resident.value_offset), 1947 le32_to_cpu( 1948 a->data.resident.value_length)); 1949 } 1950 /* The attribute list is now setup in memory. */ 1951 /* 1952 * FIXME: I don't know if this case is actually possible. 1953 * According to logic it is not possible but I have seen too 1954 * many weird things in MS software to rely on logic... Thus we 1955 * perform a manual search and make sure the first $MFT/$DATA 1956 * extent is in the base inode. If it is not we abort with an 1957 * error and if we ever see a report of this error we will need 1958 * to do some magic in order to have the necessary mft record 1959 * loaded and in the right place in the page cache. But 1960 * hopefully logic will prevail and this never happens... 1961 */ 1962 al_entry = (ATTR_LIST_ENTRY*)ni->attr_list; 1963 al_end = (u8*)al_entry + ni->attr_list_size; 1964 for (;; al_entry = next_al_entry) { 1965 /* Out of bounds check. */ 1966 if ((u8*)al_entry < ni->attr_list || 1967 (u8*)al_entry > al_end) 1968 goto em_put_err_out; 1969 /* Catch the end of the attribute list. */ 1970 if ((u8*)al_entry == al_end) 1971 goto em_put_err_out; 1972 if (!al_entry->length) 1973 goto em_put_err_out; 1974 if ((u8*)al_entry + 6 > al_end || (u8*)al_entry + 1975 le16_to_cpu(al_entry->length) > al_end) 1976 goto em_put_err_out; 1977 next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry + 1978 le16_to_cpu(al_entry->length)); 1979 if (le32_to_cpu(al_entry->type) > le32_to_cpu(AT_DATA)) 1980 goto em_put_err_out; 1981 if (AT_DATA != al_entry->type) 1982 continue; 1983 /* We want an unnamed attribute. */ 1984 if (al_entry->name_length) 1985 goto em_put_err_out; 1986 /* Want the first entry, i.e. lowest_vcn == 0. */ 1987 if (al_entry->lowest_vcn) 1988 goto em_put_err_out; 1989 /* First entry has to be in the base mft record. */ 1990 if (MREF_LE(al_entry->mft_reference) != vi->i_ino) { 1991 /* MFT references do not match, logic fails. */ 1992 ntfs_error(sb, "BUG: The first $DATA extent " 1993 "of $MFT is not in the base " 1994 "mft record. Please report " 1995 "you saw this message to " 1996 "linux-ntfs-dev@lists." 1997 "sourceforge.net"); 1998 goto put_err_out; 1999 } else { 2000 /* Sequence numbers must match. */ 2001 if (MSEQNO_LE(al_entry->mft_reference) != 2002 ni->seq_no) 2003 goto em_put_err_out; 2004 /* Got it. All is ok. We can stop now. */ 2005 break; 2006 } 2007 } 2008 } 2009 2010 ntfs_attr_reinit_search_ctx(ctx); 2011 2012 /* Now load all attribute extents. */ 2013 a = NULL; 2014 next_vcn = last_vcn = highest_vcn = 0; 2015 while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0, 2016 ctx))) { 2017 runlist_element *nrl; 2018 2019 /* Cache the current attribute. */ 2020 a = ctx->attr; 2021 /* $MFT must be non-resident. */ 2022 if (!a->non_resident) { 2023 ntfs_error(sb, "$MFT must be non-resident but a " 2024 "resident extent was found. $MFT is " 2025 "corrupt. Run chkdsk."); 2026 goto put_err_out; 2027 } 2028 /* $MFT must be uncompressed and unencrypted. */ 2029 if (a->flags & ATTR_COMPRESSION_MASK || 2030 a->flags & ATTR_IS_ENCRYPTED || 2031 a->flags & ATTR_IS_SPARSE) { 2032 ntfs_error(sb, "$MFT must be uncompressed, " 2033 "non-sparse, and unencrypted but a " 2034 "compressed/sparse/encrypted extent " 2035 "was found. $MFT is corrupt. Run " 2036 "chkdsk."); 2037 goto put_err_out; 2038 } 2039 /* 2040 * Decompress the mapping pairs array of this extent and merge 2041 * the result into the existing runlist. No need for locking 2042 * as we have exclusive access to the inode at this time and we 2043 * are a mount in progress task, too. 2044 */ 2045 nrl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl); 2046 if (IS_ERR(nrl)) { 2047 ntfs_error(sb, "ntfs_mapping_pairs_decompress() " 2048 "failed with error code %ld. $MFT is " 2049 "corrupt.", PTR_ERR(nrl)); 2050 goto put_err_out; 2051 } 2052 ni->runlist.rl = nrl; 2053 2054 /* Are we in the first extent? */ 2055 if (!next_vcn) { 2056 if (a->data.non_resident.lowest_vcn) { 2057 ntfs_error(sb, "First extent of $DATA " 2058 "attribute has non zero " 2059 "lowest_vcn. $MFT is corrupt. " 2060 "You should run chkdsk."); 2061 goto put_err_out; 2062 } 2063 /* Get the last vcn in the $DATA attribute. */ 2064 last_vcn = sle64_to_cpu( 2065 a->data.non_resident.allocated_size) 2066 >> vol->cluster_size_bits; 2067 /* Fill in the inode size. */ 2068 vi->i_size = sle64_to_cpu( 2069 a->data.non_resident.data_size); 2070 ni->initialized_size = sle64_to_cpu( 2071 a->data.non_resident.initialized_size); 2072 ni->allocated_size = sle64_to_cpu( 2073 a->data.non_resident.allocated_size); 2074 /* 2075 * Verify the number of mft records does not exceed 2076 * 2^32 - 1. 2077 */ 2078 if ((vi->i_size >> vol->mft_record_size_bits) >= 2079 (1ULL << 32)) { 2080 ntfs_error(sb, "$MFT is too big! Aborting."); 2081 goto put_err_out; 2082 } 2083 /* 2084 * We have got the first extent of the runlist for 2085 * $MFT which means it is now relatively safe to call 2086 * the normal ntfs_read_inode() function. 2087 * Complete reading the inode, this will actually 2088 * re-read the mft record for $MFT, this time entering 2089 * it into the page cache with which we complete the 2090 * kick start of the volume. It should be safe to do 2091 * this now as the first extent of $MFT/$DATA is 2092 * already known and we would hope that we don't need 2093 * further extents in order to find the other 2094 * attributes belonging to $MFT. Only time will tell if 2095 * this is really the case. If not we will have to play 2096 * magic at this point, possibly duplicating a lot of 2097 * ntfs_read_inode() at this point. We will need to 2098 * ensure we do enough of its work to be able to call 2099 * ntfs_read_inode() on extents of $MFT/$DATA. But lets 2100 * hope this never happens... 2101 */ 2102 ntfs_read_locked_inode(vi); 2103 if (is_bad_inode(vi)) { 2104 ntfs_error(sb, "ntfs_read_inode() of $MFT " 2105 "failed. BUG or corrupt $MFT. " 2106 "Run chkdsk and if no errors " 2107 "are found, please report you " 2108 "saw this message to " 2109 "linux-ntfs-dev@lists." 2110 "sourceforge.net"); 2111 ntfs_attr_put_search_ctx(ctx); 2112 /* Revert to the safe super operations. */ 2113 ntfs_free(m); 2114 return -1; 2115 } 2116 /* 2117 * Re-initialize some specifics about $MFT's inode as 2118 * ntfs_read_inode() will have set up the default ones. 2119 */ 2120 /* Set uid and gid to root. */ 2121 vi->i_uid = vi->i_gid = 0; 2122 /* Regular file. No access for anyone. */ 2123 vi->i_mode = S_IFREG; 2124 /* No VFS initiated operations allowed for $MFT. */ 2125 vi->i_op = &ntfs_empty_inode_ops; 2126 vi->i_fop = &ntfs_empty_file_ops; 2127 } 2128 2129 /* Get the lowest vcn for the next extent. */ 2130 highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn); 2131 next_vcn = highest_vcn + 1; 2132 2133 /* Only one extent or error, which we catch below. */ 2134 if (next_vcn <= 0) 2135 break; 2136 2137 /* Avoid endless loops due to corruption. */ 2138 if (next_vcn < sle64_to_cpu( 2139 a->data.non_resident.lowest_vcn)) { 2140 ntfs_error(sb, "$MFT has corrupt attribute list " 2141 "attribute. Run chkdsk."); 2142 goto put_err_out; 2143 } 2144 } 2145 if (err != -ENOENT) { 2146 ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. " 2147 "$MFT is corrupt. Run chkdsk."); 2148 goto put_err_out; 2149 } 2150 if (!a) { 2151 ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is " 2152 "corrupt. Run chkdsk."); 2153 goto put_err_out; 2154 } 2155 if (highest_vcn && highest_vcn != last_vcn - 1) { 2156 ntfs_error(sb, "Failed to load the complete runlist for " 2157 "$MFT/$DATA. Driver bug or corrupt $MFT. " 2158 "Run chkdsk."); 2159 ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx", 2160 (unsigned long long)highest_vcn, 2161 (unsigned long long)last_vcn - 1); 2162 goto put_err_out; 2163 } 2164 ntfs_attr_put_search_ctx(ctx); 2165 ntfs_debug("Done."); 2166 ntfs_free(m); 2167 2168 /* 2169 * Split the locking rules of the MFT inode from the 2170 * locking rules of other inodes: 2171 */ 2172 lockdep_set_class(&ni->runlist.lock, &mft_ni_runlist_lock_key); 2173 lockdep_set_class(&ni->mrec_lock, &mft_ni_mrec_lock_key); 2174 2175 return 0; 2176 2177 em_put_err_out: 2178 ntfs_error(sb, "Couldn't find first extent of $DATA attribute in " 2179 "attribute list. $MFT is corrupt. Run chkdsk."); 2180 put_err_out: 2181 ntfs_attr_put_search_ctx(ctx); 2182 err_out: 2183 ntfs_error(sb, "Failed. Marking inode as bad."); 2184 make_bad_inode(vi); 2185 ntfs_free(m); 2186 return -1; 2187 } 2188 2189 static void __ntfs_clear_inode(ntfs_inode *ni) 2190 { 2191 /* Free all alocated memory. */ 2192 down_write(&ni->runlist.lock); 2193 if (ni->runlist.rl) { 2194 ntfs_free(ni->runlist.rl); 2195 ni->runlist.rl = NULL; 2196 } 2197 up_write(&ni->runlist.lock); 2198 2199 if (ni->attr_list) { 2200 ntfs_free(ni->attr_list); 2201 ni->attr_list = NULL; 2202 } 2203 2204 down_write(&ni->attr_list_rl.lock); 2205 if (ni->attr_list_rl.rl) { 2206 ntfs_free(ni->attr_list_rl.rl); 2207 ni->attr_list_rl.rl = NULL; 2208 } 2209 up_write(&ni->attr_list_rl.lock); 2210 2211 if (ni->name_len && ni->name != I30) { 2212 /* Catch bugs... */ 2213 BUG_ON(!ni->name); 2214 kfree(ni->name); 2215 } 2216 } 2217 2218 void ntfs_clear_extent_inode(ntfs_inode *ni) 2219 { 2220 ntfs_debug("Entering for inode 0x%lx.", ni->mft_no); 2221 2222 BUG_ON(NInoAttr(ni)); 2223 BUG_ON(ni->nr_extents != -1); 2224 2225 #ifdef NTFS_RW 2226 if (NInoDirty(ni)) { 2227 if (!is_bad_inode(VFS_I(ni->ext.base_ntfs_ino))) 2228 ntfs_error(ni->vol->sb, "Clearing dirty extent inode! " 2229 "Losing data! This is a BUG!!!"); 2230 // FIXME: Do something!!! 2231 } 2232 #endif /* NTFS_RW */ 2233 2234 __ntfs_clear_inode(ni); 2235 2236 /* Bye, bye... */ 2237 ntfs_destroy_extent_inode(ni); 2238 } 2239 2240 /** 2241 * ntfs_clear_big_inode - clean up the ntfs specific part of an inode 2242 * @vi: vfs inode pending annihilation 2243 * 2244 * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode() 2245 * is called, which deallocates all memory belonging to the NTFS specific part 2246 * of the inode and returns. 2247 * 2248 * If the MFT record is dirty, we commit it before doing anything else. 2249 */ 2250 void ntfs_clear_big_inode(struct inode *vi) 2251 { 2252 ntfs_inode *ni = NTFS_I(vi); 2253 2254 #ifdef NTFS_RW 2255 if (NInoDirty(ni)) { 2256 bool was_bad = (is_bad_inode(vi)); 2257 2258 /* Committing the inode also commits all extent inodes. */ 2259 ntfs_commit_inode(vi); 2260 2261 if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) { 2262 ntfs_error(vi->i_sb, "Failed to commit dirty inode " 2263 "0x%lx. Losing data!", vi->i_ino); 2264 // FIXME: Do something!!! 2265 } 2266 } 2267 #endif /* NTFS_RW */ 2268 2269 /* No need to lock at this stage as no one else has a reference. */ 2270 if (ni->nr_extents > 0) { 2271 int i; 2272 2273 for (i = 0; i < ni->nr_extents; i++) 2274 ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]); 2275 kfree(ni->ext.extent_ntfs_inos); 2276 } 2277 2278 __ntfs_clear_inode(ni); 2279 2280 if (NInoAttr(ni)) { 2281 /* Release the base inode if we are holding it. */ 2282 if (ni->nr_extents == -1) { 2283 iput(VFS_I(ni->ext.base_ntfs_ino)); 2284 ni->nr_extents = 0; 2285 ni->ext.base_ntfs_ino = NULL; 2286 } 2287 } 2288 return; 2289 } 2290 2291 /** 2292 * ntfs_show_options - show mount options in /proc/mounts 2293 * @sf: seq_file in which to write our mount options 2294 * @mnt: vfs mount whose mount options to display 2295 * 2296 * Called by the VFS once for each mounted ntfs volume when someone reads 2297 * /proc/mounts in order to display the NTFS specific mount options of each 2298 * mount. The mount options of the vfs mount @mnt are written to the seq file 2299 * @sf and success is returned. 2300 */ 2301 int ntfs_show_options(struct seq_file *sf, struct vfsmount *mnt) 2302 { 2303 ntfs_volume *vol = NTFS_SB(mnt->mnt_sb); 2304 int i; 2305 2306 seq_printf(sf, ",uid=%i", vol->uid); 2307 seq_printf(sf, ",gid=%i", vol->gid); 2308 if (vol->fmask == vol->dmask) 2309 seq_printf(sf, ",umask=0%o", vol->fmask); 2310 else { 2311 seq_printf(sf, ",fmask=0%o", vol->fmask); 2312 seq_printf(sf, ",dmask=0%o", vol->dmask); 2313 } 2314 seq_printf(sf, ",nls=%s", vol->nls_map->charset); 2315 if (NVolCaseSensitive(vol)) 2316 seq_printf(sf, ",case_sensitive"); 2317 if (NVolShowSystemFiles(vol)) 2318 seq_printf(sf, ",show_sys_files"); 2319 if (!NVolSparseEnabled(vol)) 2320 seq_printf(sf, ",disable_sparse"); 2321 for (i = 0; on_errors_arr[i].val; i++) { 2322 if (on_errors_arr[i].val & vol->on_errors) 2323 seq_printf(sf, ",errors=%s", on_errors_arr[i].str); 2324 } 2325 seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier); 2326 return 0; 2327 } 2328 2329 #ifdef NTFS_RW 2330 2331 static const char *es = " Leaving inconsistent metadata. Unmount and run " 2332 "chkdsk."; 2333 2334 /** 2335 * ntfs_truncate - called when the i_size of an ntfs inode is changed 2336 * @vi: inode for which the i_size was changed 2337 * 2338 * We only support i_size changes for normal files at present, i.e. not 2339 * compressed and not encrypted. This is enforced in ntfs_setattr(), see 2340 * below. 2341 * 2342 * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and 2343 * that the change is allowed. 2344 * 2345 * This implies for us that @vi is a file inode rather than a directory, index, 2346 * or attribute inode as well as that @vi is a base inode. 2347 * 2348 * Returns 0 on success or -errno on error. 2349 * 2350 * Called with ->i_mutex held. In all but one case ->i_alloc_sem is held for 2351 * writing. The only case in the kernel where ->i_alloc_sem is not held is 2352 * mm/filemap.c::generic_file_buffered_write() where vmtruncate() is called 2353 * with the current i_size as the offset. The analogous place in NTFS is in 2354 * fs/ntfs/file.c::ntfs_file_buffered_write() where we call vmtruncate() again 2355 * without holding ->i_alloc_sem. 2356 */ 2357 int ntfs_truncate(struct inode *vi) 2358 { 2359 s64 new_size, old_size, nr_freed, new_alloc_size, old_alloc_size; 2360 VCN highest_vcn; 2361 unsigned long flags; 2362 ntfs_inode *base_ni, *ni = NTFS_I(vi); 2363 ntfs_volume *vol = ni->vol; 2364 ntfs_attr_search_ctx *ctx; 2365 MFT_RECORD *m; 2366 ATTR_RECORD *a; 2367 const char *te = " Leaving file length out of sync with i_size."; 2368 int err, mp_size, size_change, alloc_change; 2369 u32 attr_len; 2370 2371 ntfs_debug("Entering for inode 0x%lx.", vi->i_ino); 2372 BUG_ON(NInoAttr(ni)); 2373 BUG_ON(S_ISDIR(vi->i_mode)); 2374 BUG_ON(NInoMstProtected(ni)); 2375 BUG_ON(ni->nr_extents < 0); 2376 retry_truncate: 2377 /* 2378 * Lock the runlist for writing and map the mft record to ensure it is 2379 * safe to mess with the attribute runlist and sizes. 2380 */ 2381 down_write(&ni->runlist.lock); 2382 if (!NInoAttr(ni)) 2383 base_ni = ni; 2384 else 2385 base_ni = ni->ext.base_ntfs_ino; 2386 m = map_mft_record(base_ni); 2387 if (IS_ERR(m)) { 2388 err = PTR_ERR(m); 2389 ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx " 2390 "(error code %d).%s", vi->i_ino, err, te); 2391 ctx = NULL; 2392 m = NULL; 2393 goto old_bad_out; 2394 } 2395 ctx = ntfs_attr_get_search_ctx(base_ni, m); 2396 if (unlikely(!ctx)) { 2397 ntfs_error(vi->i_sb, "Failed to allocate a search context for " 2398 "inode 0x%lx (not enough memory).%s", 2399 vi->i_ino, te); 2400 err = -ENOMEM; 2401 goto old_bad_out; 2402 } 2403 err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len, 2404 CASE_SENSITIVE, 0, NULL, 0, ctx); 2405 if (unlikely(err)) { 2406 if (err == -ENOENT) { 2407 ntfs_error(vi->i_sb, "Open attribute is missing from " 2408 "mft record. Inode 0x%lx is corrupt. " 2409 "Run chkdsk.%s", vi->i_ino, te); 2410 err = -EIO; 2411 } else 2412 ntfs_error(vi->i_sb, "Failed to lookup attribute in " 2413 "inode 0x%lx (error code %d).%s", 2414 vi->i_ino, err, te); 2415 goto old_bad_out; 2416 } 2417 m = ctx->mrec; 2418 a = ctx->attr; 2419 /* 2420 * The i_size of the vfs inode is the new size for the attribute value. 2421 */ 2422 new_size = i_size_read(vi); 2423 /* The current size of the attribute value is the old size. */ 2424 old_size = ntfs_attr_size(a); 2425 /* Calculate the new allocated size. */ 2426 if (NInoNonResident(ni)) 2427 new_alloc_size = (new_size + vol->cluster_size - 1) & 2428 ~(s64)vol->cluster_size_mask; 2429 else 2430 new_alloc_size = (new_size + 7) & ~7; 2431 /* The current allocated size is the old allocated size. */ 2432 read_lock_irqsave(&ni->size_lock, flags); 2433 old_alloc_size = ni->allocated_size; 2434 read_unlock_irqrestore(&ni->size_lock, flags); 2435 /* 2436 * The change in the file size. This will be 0 if no change, >0 if the 2437 * size is growing, and <0 if the size is shrinking. 2438 */ 2439 size_change = -1; 2440 if (new_size - old_size >= 0) { 2441 size_change = 1; 2442 if (new_size == old_size) 2443 size_change = 0; 2444 } 2445 /* As above for the allocated size. */ 2446 alloc_change = -1; 2447 if (new_alloc_size - old_alloc_size >= 0) { 2448 alloc_change = 1; 2449 if (new_alloc_size == old_alloc_size) 2450 alloc_change = 0; 2451 } 2452 /* 2453 * If neither the size nor the allocation are being changed there is 2454 * nothing to do. 2455 */ 2456 if (!size_change && !alloc_change) 2457 goto unm_done; 2458 /* If the size is changing, check if new size is allowed in $AttrDef. */ 2459 if (size_change) { 2460 err = ntfs_attr_size_bounds_check(vol, ni->type, new_size); 2461 if (unlikely(err)) { 2462 if (err == -ERANGE) { 2463 ntfs_error(vol->sb, "Truncate would cause the " 2464 "inode 0x%lx to %simum size " 2465 "for its attribute type " 2466 "(0x%x). Aborting truncate.", 2467 vi->i_ino, 2468 new_size > old_size ? "exceed " 2469 "the max" : "go under the min", 2470 le32_to_cpu(ni->type)); 2471 err = -EFBIG; 2472 } else { 2473 ntfs_error(vol->sb, "Inode 0x%lx has unknown " 2474 "attribute type 0x%x. " 2475 "Aborting truncate.", 2476 vi->i_ino, 2477 le32_to_cpu(ni->type)); 2478 err = -EIO; 2479 } 2480 /* Reset the vfs inode size to the old size. */ 2481 i_size_write(vi, old_size); 2482 goto err_out; 2483 } 2484 } 2485 if (NInoCompressed(ni) || NInoEncrypted(ni)) { 2486 ntfs_warning(vi->i_sb, "Changes in inode size are not " 2487 "supported yet for %s files, ignoring.", 2488 NInoCompressed(ni) ? "compressed" : 2489 "encrypted"); 2490 err = -EOPNOTSUPP; 2491 goto bad_out; 2492 } 2493 if (a->non_resident) 2494 goto do_non_resident_truncate; 2495 BUG_ON(NInoNonResident(ni)); 2496 /* Resize the attribute record to best fit the new attribute size. */ 2497 if (new_size < vol->mft_record_size && 2498 !ntfs_resident_attr_value_resize(m, a, new_size)) { 2499 /* The resize succeeded! */ 2500 flush_dcache_mft_record_page(ctx->ntfs_ino); 2501 mark_mft_record_dirty(ctx->ntfs_ino); 2502 write_lock_irqsave(&ni->size_lock, flags); 2503 /* Update the sizes in the ntfs inode and all is done. */ 2504 ni->allocated_size = le32_to_cpu(a->length) - 2505 le16_to_cpu(a->data.resident.value_offset); 2506 /* 2507 * Note ntfs_resident_attr_value_resize() has already done any 2508 * necessary data clearing in the attribute record. When the 2509 * file is being shrunk vmtruncate() will already have cleared 2510 * the top part of the last partial page, i.e. since this is 2511 * the resident case this is the page with index 0. However, 2512 * when the file is being expanded, the page cache page data 2513 * between the old data_size, i.e. old_size, and the new_size 2514 * has not been zeroed. Fortunately, we do not need to zero it 2515 * either since on one hand it will either already be zero due 2516 * to both readpage and writepage clearing partial page data 2517 * beyond i_size in which case there is nothing to do or in the 2518 * case of the file being mmap()ped at the same time, POSIX 2519 * specifies that the behaviour is unspecified thus we do not 2520 * have to do anything. This means that in our implementation 2521 * in the rare case that the file is mmap()ped and a write 2522 * occured into the mmap()ped region just beyond the file size 2523 * and writepage has not yet been called to write out the page 2524 * (which would clear the area beyond the file size) and we now 2525 * extend the file size to incorporate this dirty region 2526 * outside the file size, a write of the page would result in 2527 * this data being written to disk instead of being cleared. 2528 * Given both POSIX and the Linux mmap(2) man page specify that 2529 * this corner case is undefined, we choose to leave it like 2530 * that as this is much simpler for us as we cannot lock the 2531 * relevant page now since we are holding too many ntfs locks 2532 * which would result in a lock reversal deadlock. 2533 */ 2534 ni->initialized_size = new_size; 2535 write_unlock_irqrestore(&ni->size_lock, flags); 2536 goto unm_done; 2537 } 2538 /* If the above resize failed, this must be an attribute extension. */ 2539 BUG_ON(size_change < 0); 2540 /* 2541 * We have to drop all the locks so we can call 2542 * ntfs_attr_make_non_resident(). This could be optimised by try- 2543 * locking the first page cache page and only if that fails dropping 2544 * the locks, locking the page, and redoing all the locking and 2545 * lookups. While this would be a huge optimisation, it is not worth 2546 * it as this is definitely a slow code path as it only ever can happen 2547 * once for any given file. 2548 */ 2549 ntfs_attr_put_search_ctx(ctx); 2550 unmap_mft_record(base_ni); 2551 up_write(&ni->runlist.lock); 2552 /* 2553 * Not enough space in the mft record, try to make the attribute 2554 * non-resident and if successful restart the truncation process. 2555 */ 2556 err = ntfs_attr_make_non_resident(ni, old_size); 2557 if (likely(!err)) 2558 goto retry_truncate; 2559 /* 2560 * Could not make non-resident. If this is due to this not being 2561 * permitted for this attribute type or there not being enough space, 2562 * try to make other attributes non-resident. Otherwise fail. 2563 */ 2564 if (unlikely(err != -EPERM && err != -ENOSPC)) { 2565 ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, attribute " 2566 "type 0x%x, because the conversion from " 2567 "resident to non-resident attribute failed " 2568 "with error code %i.", vi->i_ino, 2569 (unsigned)le32_to_cpu(ni->type), err); 2570 if (err != -ENOMEM) 2571 err = -EIO; 2572 goto conv_err_out; 2573 } 2574 /* TODO: Not implemented from here, abort. */ 2575 if (err == -ENOSPC) 2576 ntfs_error(vol->sb, "Not enough space in the mft record/on " 2577 "disk for the non-resident attribute value. " 2578 "This case is not implemented yet."); 2579 else /* if (err == -EPERM) */ 2580 ntfs_error(vol->sb, "This attribute type may not be " 2581 "non-resident. This case is not implemented " 2582 "yet."); 2583 err = -EOPNOTSUPP; 2584 goto conv_err_out; 2585 #if 0 2586 // TODO: Attempt to make other attributes non-resident. 2587 if (!err) 2588 goto do_resident_extend; 2589 /* 2590 * Both the attribute list attribute and the standard information 2591 * attribute must remain in the base inode. Thus, if this is one of 2592 * these attributes, we have to try to move other attributes out into 2593 * extent mft records instead. 2594 */ 2595 if (ni->type == AT_ATTRIBUTE_LIST || 2596 ni->type == AT_STANDARD_INFORMATION) { 2597 // TODO: Attempt to move other attributes into extent mft 2598 // records. 2599 err = -EOPNOTSUPP; 2600 if (!err) 2601 goto do_resident_extend; 2602 goto err_out; 2603 } 2604 // TODO: Attempt to move this attribute to an extent mft record, but 2605 // only if it is not already the only attribute in an mft record in 2606 // which case there would be nothing to gain. 2607 err = -EOPNOTSUPP; 2608 if (!err) 2609 goto do_resident_extend; 2610 /* There is nothing we can do to make enough space. )-: */ 2611 goto err_out; 2612 #endif 2613 do_non_resident_truncate: 2614 BUG_ON(!NInoNonResident(ni)); 2615 if (alloc_change < 0) { 2616 highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn); 2617 if (highest_vcn > 0 && 2618 old_alloc_size >> vol->cluster_size_bits > 2619 highest_vcn + 1) { 2620 /* 2621 * This attribute has multiple extents. Not yet 2622 * supported. 2623 */ 2624 ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, " 2625 "attribute type 0x%x, because the " 2626 "attribute is highly fragmented (it " 2627 "consists of multiple extents) and " 2628 "this case is not implemented yet.", 2629 vi->i_ino, 2630 (unsigned)le32_to_cpu(ni->type)); 2631 err = -EOPNOTSUPP; 2632 goto bad_out; 2633 } 2634 } 2635 /* 2636 * If the size is shrinking, need to reduce the initialized_size and 2637 * the data_size before reducing the allocation. 2638 */ 2639 if (size_change < 0) { 2640 /* 2641 * Make the valid size smaller (i_size is already up-to-date). 2642 */ 2643 write_lock_irqsave(&ni->size_lock, flags); 2644 if (new_size < ni->initialized_size) { 2645 ni->initialized_size = new_size; 2646 a->data.non_resident.initialized_size = 2647 cpu_to_sle64(new_size); 2648 } 2649 a->data.non_resident.data_size = cpu_to_sle64(new_size); 2650 write_unlock_irqrestore(&ni->size_lock, flags); 2651 flush_dcache_mft_record_page(ctx->ntfs_ino); 2652 mark_mft_record_dirty(ctx->ntfs_ino); 2653 /* If the allocated size is not changing, we are done. */ 2654 if (!alloc_change) 2655 goto unm_done; 2656 /* 2657 * If the size is shrinking it makes no sense for the 2658 * allocation to be growing. 2659 */ 2660 BUG_ON(alloc_change > 0); 2661 } else /* if (size_change >= 0) */ { 2662 /* 2663 * The file size is growing or staying the same but the 2664 * allocation can be shrinking, growing or staying the same. 2665 */ 2666 if (alloc_change > 0) { 2667 /* 2668 * We need to extend the allocation and possibly update 2669 * the data size. If we are updating the data size, 2670 * since we are not touching the initialized_size we do 2671 * not need to worry about the actual data on disk. 2672 * And as far as the page cache is concerned, there 2673 * will be no pages beyond the old data size and any 2674 * partial region in the last page between the old and 2675 * new data size (or the end of the page if the new 2676 * data size is outside the page) does not need to be 2677 * modified as explained above for the resident 2678 * attribute truncate case. To do this, we simply drop 2679 * the locks we hold and leave all the work to our 2680 * friendly helper ntfs_attr_extend_allocation(). 2681 */ 2682 ntfs_attr_put_search_ctx(ctx); 2683 unmap_mft_record(base_ni); 2684 up_write(&ni->runlist.lock); 2685 err = ntfs_attr_extend_allocation(ni, new_size, 2686 size_change > 0 ? new_size : -1, -1); 2687 /* 2688 * ntfs_attr_extend_allocation() will have done error 2689 * output already. 2690 */ 2691 goto done; 2692 } 2693 if (!alloc_change) 2694 goto alloc_done; 2695 } 2696 /* alloc_change < 0 */ 2697 /* Free the clusters. */ 2698 nr_freed = ntfs_cluster_free(ni, new_alloc_size >> 2699 vol->cluster_size_bits, -1, ctx); 2700 m = ctx->mrec; 2701 a = ctx->attr; 2702 if (unlikely(nr_freed < 0)) { 2703 ntfs_error(vol->sb, "Failed to release cluster(s) (error code " 2704 "%lli). Unmount and run chkdsk to recover " 2705 "the lost cluster(s).", (long long)nr_freed); 2706 NVolSetErrors(vol); 2707 nr_freed = 0; 2708 } 2709 /* Truncate the runlist. */ 2710 err = ntfs_rl_truncate_nolock(vol, &ni->runlist, 2711 new_alloc_size >> vol->cluster_size_bits); 2712 /* 2713 * If the runlist truncation failed and/or the search context is no 2714 * longer valid, we cannot resize the attribute record or build the 2715 * mapping pairs array thus we mark the inode bad so that no access to 2716 * the freed clusters can happen. 2717 */ 2718 if (unlikely(err || IS_ERR(m))) { 2719 ntfs_error(vol->sb, "Failed to %s (error code %li).%s", 2720 IS_ERR(m) ? 2721 "restore attribute search context" : 2722 "truncate attribute runlist", 2723 IS_ERR(m) ? PTR_ERR(m) : err, es); 2724 err = -EIO; 2725 goto bad_out; 2726 } 2727 /* Get the size for the shrunk mapping pairs array for the runlist. */ 2728 mp_size = ntfs_get_size_for_mapping_pairs(vol, ni->runlist.rl, 0, -1); 2729 if (unlikely(mp_size <= 0)) { 2730 ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, " 2731 "attribute type 0x%x, because determining the " 2732 "size for the mapping pairs failed with error " 2733 "code %i.%s", vi->i_ino, 2734 (unsigned)le32_to_cpu(ni->type), mp_size, es); 2735 err = -EIO; 2736 goto bad_out; 2737 } 2738 /* 2739 * Shrink the attribute record for the new mapping pairs array. Note, 2740 * this cannot fail since we are making the attribute smaller thus by 2741 * definition there is enough space to do so. 2742 */ 2743 attr_len = le32_to_cpu(a->length); 2744 err = ntfs_attr_record_resize(m, a, mp_size + 2745 le16_to_cpu(a->data.non_resident.mapping_pairs_offset)); 2746 BUG_ON(err); 2747 /* 2748 * Generate the mapping pairs array directly into the attribute record. 2749 */ 2750 err = ntfs_mapping_pairs_build(vol, (u8*)a + 2751 le16_to_cpu(a->data.non_resident.mapping_pairs_offset), 2752 mp_size, ni->runlist.rl, 0, -1, NULL); 2753 if (unlikely(err)) { 2754 ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, " 2755 "attribute type 0x%x, because building the " 2756 "mapping pairs failed with error code %i.%s", 2757 vi->i_ino, (unsigned)le32_to_cpu(ni->type), 2758 err, es); 2759 err = -EIO; 2760 goto bad_out; 2761 } 2762 /* Update the allocated/compressed size as well as the highest vcn. */ 2763 a->data.non_resident.highest_vcn = cpu_to_sle64((new_alloc_size >> 2764 vol->cluster_size_bits) - 1); 2765 write_lock_irqsave(&ni->size_lock, flags); 2766 ni->allocated_size = new_alloc_size; 2767 a->data.non_resident.allocated_size = cpu_to_sle64(new_alloc_size); 2768 if (NInoSparse(ni) || NInoCompressed(ni)) { 2769 if (nr_freed) { 2770 ni->itype.compressed.size -= nr_freed << 2771 vol->cluster_size_bits; 2772 BUG_ON(ni->itype.compressed.size < 0); 2773 a->data.non_resident.compressed_size = cpu_to_sle64( 2774 ni->itype.compressed.size); 2775 vi->i_blocks = ni->itype.compressed.size >> 9; 2776 } 2777 } else 2778 vi->i_blocks = new_alloc_size >> 9; 2779 write_unlock_irqrestore(&ni->size_lock, flags); 2780 /* 2781 * We have shrunk the allocation. If this is a shrinking truncate we 2782 * have already dealt with the initialized_size and the data_size above 2783 * and we are done. If the truncate is only changing the allocation 2784 * and not the data_size, we are also done. If this is an extending 2785 * truncate, need to extend the data_size now which is ensured by the 2786 * fact that @size_change is positive. 2787 */ 2788 alloc_done: 2789 /* 2790 * If the size is growing, need to update it now. If it is shrinking, 2791 * we have already updated it above (before the allocation change). 2792 */ 2793 if (size_change > 0) 2794 a->data.non_resident.data_size = cpu_to_sle64(new_size); 2795 /* Ensure the modified mft record is written out. */ 2796 flush_dcache_mft_record_page(ctx->ntfs_ino); 2797 mark_mft_record_dirty(ctx->ntfs_ino); 2798 unm_done: 2799 ntfs_attr_put_search_ctx(ctx); 2800 unmap_mft_record(base_ni); 2801 up_write(&ni->runlist.lock); 2802 done: 2803 /* Update the mtime and ctime on the base inode. */ 2804 /* normally ->truncate shouldn't update ctime or mtime, 2805 * but ntfs did before so it got a copy & paste version 2806 * of file_update_time. one day someone should fix this 2807 * for real. 2808 */ 2809 if (!IS_NOCMTIME(VFS_I(base_ni)) && !IS_RDONLY(VFS_I(base_ni))) { 2810 struct timespec now = current_fs_time(VFS_I(base_ni)->i_sb); 2811 int sync_it = 0; 2812 2813 if (!timespec_equal(&VFS_I(base_ni)->i_mtime, &now) || 2814 !timespec_equal(&VFS_I(base_ni)->i_ctime, &now)) 2815 sync_it = 1; 2816 VFS_I(base_ni)->i_mtime = now; 2817 VFS_I(base_ni)->i_ctime = now; 2818 2819 if (sync_it) 2820 mark_inode_dirty_sync(VFS_I(base_ni)); 2821 } 2822 2823 if (likely(!err)) { 2824 NInoClearTruncateFailed(ni); 2825 ntfs_debug("Done."); 2826 } 2827 return err; 2828 old_bad_out: 2829 old_size = -1; 2830 bad_out: 2831 if (err != -ENOMEM && err != -EOPNOTSUPP) 2832 NVolSetErrors(vol); 2833 if (err != -EOPNOTSUPP) 2834 NInoSetTruncateFailed(ni); 2835 else if (old_size >= 0) 2836 i_size_write(vi, old_size); 2837 err_out: 2838 if (ctx) 2839 ntfs_attr_put_search_ctx(ctx); 2840 if (m) 2841 unmap_mft_record(base_ni); 2842 up_write(&ni->runlist.lock); 2843 out: 2844 ntfs_debug("Failed. Returning error code %i.", err); 2845 return err; 2846 conv_err_out: 2847 if (err != -ENOMEM && err != -EOPNOTSUPP) 2848 NVolSetErrors(vol); 2849 if (err != -EOPNOTSUPP) 2850 NInoSetTruncateFailed(ni); 2851 else 2852 i_size_write(vi, old_size); 2853 goto out; 2854 } 2855 2856 /** 2857 * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value 2858 * @vi: inode for which the i_size was changed 2859 * 2860 * Wrapper for ntfs_truncate() that has no return value. 2861 * 2862 * See ntfs_truncate() description above for details. 2863 */ 2864 void ntfs_truncate_vfs(struct inode *vi) { 2865 ntfs_truncate(vi); 2866 } 2867 2868 /** 2869 * ntfs_setattr - called from notify_change() when an attribute is being changed 2870 * @dentry: dentry whose attributes to change 2871 * @attr: structure describing the attributes and the changes 2872 * 2873 * We have to trap VFS attempts to truncate the file described by @dentry as 2874 * soon as possible, because we do not implement changes in i_size yet. So we 2875 * abort all i_size changes here. 2876 * 2877 * We also abort all changes of user, group, and mode as we do not implement 2878 * the NTFS ACLs yet. 2879 * 2880 * Called with ->i_mutex held. For the ATTR_SIZE (i.e. ->truncate) case, also 2881 * called with ->i_alloc_sem held for writing. 2882 * 2883 * Basically this is a copy of generic notify_change() and inode_setattr() 2884 * functionality, except we intercept and abort changes in i_size. 2885 */ 2886 int ntfs_setattr(struct dentry *dentry, struct iattr *attr) 2887 { 2888 struct inode *vi = dentry->d_inode; 2889 int err; 2890 unsigned int ia_valid = attr->ia_valid; 2891 2892 err = inode_change_ok(vi, attr); 2893 if (err) 2894 goto out; 2895 /* We do not support NTFS ACLs yet. */ 2896 if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) { 2897 ntfs_warning(vi->i_sb, "Changes in user/group/mode are not " 2898 "supported yet, ignoring."); 2899 err = -EOPNOTSUPP; 2900 goto out; 2901 } 2902 if (ia_valid & ATTR_SIZE) { 2903 if (attr->ia_size != i_size_read(vi)) { 2904 ntfs_inode *ni = NTFS_I(vi); 2905 /* 2906 * FIXME: For now we do not support resizing of 2907 * compressed or encrypted files yet. 2908 */ 2909 if (NInoCompressed(ni) || NInoEncrypted(ni)) { 2910 ntfs_warning(vi->i_sb, "Changes in inode size " 2911 "are not supported yet for " 2912 "%s files, ignoring.", 2913 NInoCompressed(ni) ? 2914 "compressed" : "encrypted"); 2915 err = -EOPNOTSUPP; 2916 } else 2917 err = vmtruncate(vi, attr->ia_size); 2918 if (err || ia_valid == ATTR_SIZE) 2919 goto out; 2920 } else { 2921 /* 2922 * We skipped the truncate but must still update 2923 * timestamps. 2924 */ 2925 ia_valid |= ATTR_MTIME | ATTR_CTIME; 2926 } 2927 } 2928 if (ia_valid & ATTR_ATIME) 2929 vi->i_atime = timespec_trunc(attr->ia_atime, 2930 vi->i_sb->s_time_gran); 2931 if (ia_valid & ATTR_MTIME) 2932 vi->i_mtime = timespec_trunc(attr->ia_mtime, 2933 vi->i_sb->s_time_gran); 2934 if (ia_valid & ATTR_CTIME) 2935 vi->i_ctime = timespec_trunc(attr->ia_ctime, 2936 vi->i_sb->s_time_gran); 2937 mark_inode_dirty(vi); 2938 out: 2939 return err; 2940 } 2941 2942 /** 2943 * ntfs_write_inode - write out a dirty inode 2944 * @vi: inode to write out 2945 * @sync: if true, write out synchronously 2946 * 2947 * Write out a dirty inode to disk including any extent inodes if present. 2948 * 2949 * If @sync is true, commit the inode to disk and wait for io completion. This 2950 * is done using write_mft_record(). 2951 * 2952 * If @sync is false, just schedule the write to happen but do not wait for i/o 2953 * completion. In 2.6 kernels, scheduling usually happens just by virtue of 2954 * marking the page (and in this case mft record) dirty but we do not implement 2955 * this yet as write_mft_record() largely ignores the @sync parameter and 2956 * always performs synchronous writes. 2957 * 2958 * Return 0 on success and -errno on error. 2959 */ 2960 int ntfs_write_inode(struct inode *vi, int sync) 2961 { 2962 sle64 nt; 2963 ntfs_inode *ni = NTFS_I(vi); 2964 ntfs_attr_search_ctx *ctx; 2965 MFT_RECORD *m; 2966 STANDARD_INFORMATION *si; 2967 int err = 0; 2968 bool modified = false; 2969 2970 ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "", 2971 vi->i_ino); 2972 /* 2973 * Dirty attribute inodes are written via their real inodes so just 2974 * clean them here. Access time updates are taken care off when the 2975 * real inode is written. 2976 */ 2977 if (NInoAttr(ni)) { 2978 NInoClearDirty(ni); 2979 ntfs_debug("Done."); 2980 return 0; 2981 } 2982 /* Map, pin, and lock the mft record belonging to the inode. */ 2983 m = map_mft_record(ni); 2984 if (IS_ERR(m)) { 2985 err = PTR_ERR(m); 2986 goto err_out; 2987 } 2988 /* Update the access times in the standard information attribute. */ 2989 ctx = ntfs_attr_get_search_ctx(ni, m); 2990 if (unlikely(!ctx)) { 2991 err = -ENOMEM; 2992 goto unm_err_out; 2993 } 2994 err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 2995 CASE_SENSITIVE, 0, NULL, 0, ctx); 2996 if (unlikely(err)) { 2997 ntfs_attr_put_search_ctx(ctx); 2998 goto unm_err_out; 2999 } 3000 si = (STANDARD_INFORMATION*)((u8*)ctx->attr + 3001 le16_to_cpu(ctx->attr->data.resident.value_offset)); 3002 /* Update the access times if they have changed. */ 3003 nt = utc2ntfs(vi->i_mtime); 3004 if (si->last_data_change_time != nt) { 3005 ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, " 3006 "new = 0x%llx", vi->i_ino, (long long) 3007 sle64_to_cpu(si->last_data_change_time), 3008 (long long)sle64_to_cpu(nt)); 3009 si->last_data_change_time = nt; 3010 modified = true; 3011 } 3012 nt = utc2ntfs(vi->i_ctime); 3013 if (si->last_mft_change_time != nt) { 3014 ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, " 3015 "new = 0x%llx", vi->i_ino, (long long) 3016 sle64_to_cpu(si->last_mft_change_time), 3017 (long long)sle64_to_cpu(nt)); 3018 si->last_mft_change_time = nt; 3019 modified = true; 3020 } 3021 nt = utc2ntfs(vi->i_atime); 3022 if (si->last_access_time != nt) { 3023 ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, " 3024 "new = 0x%llx", vi->i_ino, 3025 (long long)sle64_to_cpu(si->last_access_time), 3026 (long long)sle64_to_cpu(nt)); 3027 si->last_access_time = nt; 3028 modified = true; 3029 } 3030 /* 3031 * If we just modified the standard information attribute we need to 3032 * mark the mft record it is in dirty. We do this manually so that 3033 * mark_inode_dirty() is not called which would redirty the inode and 3034 * hence result in an infinite loop of trying to write the inode. 3035 * There is no need to mark the base inode nor the base mft record 3036 * dirty, since we are going to write this mft record below in any case 3037 * and the base mft record may actually not have been modified so it 3038 * might not need to be written out. 3039 * NOTE: It is not a problem when the inode for $MFT itself is being 3040 * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES 3041 * on the $MFT inode and hence ntfs_write_inode() will not be 3042 * re-invoked because of it which in turn is ok since the dirtied mft 3043 * record will be cleaned and written out to disk below, i.e. before 3044 * this function returns. 3045 */ 3046 if (modified) { 3047 flush_dcache_mft_record_page(ctx->ntfs_ino); 3048 if (!NInoTestSetDirty(ctx->ntfs_ino)) 3049 mark_ntfs_record_dirty(ctx->ntfs_ino->page, 3050 ctx->ntfs_ino->page_ofs); 3051 } 3052 ntfs_attr_put_search_ctx(ctx); 3053 /* Now the access times are updated, write the base mft record. */ 3054 if (NInoDirty(ni)) 3055 err = write_mft_record(ni, m, sync); 3056 /* Write all attached extent mft records. */ 3057 mutex_lock(&ni->extent_lock); 3058 if (ni->nr_extents > 0) { 3059 ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos; 3060 int i; 3061 3062 ntfs_debug("Writing %i extent inodes.", ni->nr_extents); 3063 for (i = 0; i < ni->nr_extents; i++) { 3064 ntfs_inode *tni = extent_nis[i]; 3065 3066 if (NInoDirty(tni)) { 3067 MFT_RECORD *tm = map_mft_record(tni); 3068 int ret; 3069 3070 if (IS_ERR(tm)) { 3071 if (!err || err == -ENOMEM) 3072 err = PTR_ERR(tm); 3073 continue; 3074 } 3075 ret = write_mft_record(tni, tm, sync); 3076 unmap_mft_record(tni); 3077 if (unlikely(ret)) { 3078 if (!err || err == -ENOMEM) 3079 err = ret; 3080 } 3081 } 3082 } 3083 } 3084 mutex_unlock(&ni->extent_lock); 3085 unmap_mft_record(ni); 3086 if (unlikely(err)) 3087 goto err_out; 3088 ntfs_debug("Done."); 3089 return 0; 3090 unm_err_out: 3091 unmap_mft_record(ni); 3092 err_out: 3093 if (err == -ENOMEM) { 3094 ntfs_warning(vi->i_sb, "Not enough memory to write inode. " 3095 "Marking the inode dirty again, so the VFS " 3096 "retries later."); 3097 mark_inode_dirty(vi); 3098 } else { 3099 ntfs_error(vi->i_sb, "Failed (error %i): Run chkdsk.", -err); 3100 NVolSetErrors(ni->vol); 3101 } 3102 return err; 3103 } 3104 3105 #endif /* NTFS_RW */ 3106