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