Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /** |
| 2 | * inode.c - NTFS kernel inode handling. Part of the Linux-NTFS project. |
| 3 | * |
Anton Altaparmakov | a0646a1 | 2006-03-23 15:53:03 +0000 | [diff] [blame] | 4 | * Copyright (c) 2001-2006 Anton Altaparmakov |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 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 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 22 | #include <linux/buffer_head.h> |
Anton Altaparmakov | a778f21 | 2006-03-23 16:18:23 +0000 | [diff] [blame] | 23 | #include <linux/fs.h> |
| 24 | #include <linux/mm.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 25 | #include <linux/mount.h> |
Anton Altaparmakov | a0646a1 | 2006-03-23 15:53:03 +0000 | [diff] [blame] | 26 | #include <linux/mutex.h> |
Anton Altaparmakov | a778f21 | 2006-03-23 16:18:23 +0000 | [diff] [blame] | 27 | #include <linux/pagemap.h> |
| 28 | #include <linux/quotaops.h> |
| 29 | #include <linux/slab.h> |
| 30 | #include <linux/smp_lock.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 31 | |
| 32 | #include "aops.h" |
Anton Altaparmakov | a0646a1 | 2006-03-23 15:53:03 +0000 | [diff] [blame] | 33 | #include "attrib.h" |
Anton Altaparmakov | a778f21 | 2006-03-23 16:18:23 +0000 | [diff] [blame] | 34 | #include "bitmap.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 35 | #include "dir.h" |
| 36 | #include "debug.h" |
| 37 | #include "inode.h" |
| 38 | #include "attrib.h" |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 39 | #include "lcnalloc.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 40 | #include "malloc.h" |
| 41 | #include "mft.h" |
| 42 | #include "time.h" |
| 43 | #include "ntfs.h" |
| 44 | |
| 45 | /** |
| 46 | * ntfs_test_inode - compare two (possibly fake) inodes for equality |
| 47 | * @vi: vfs inode which to test |
| 48 | * @na: ntfs attribute which is being tested with |
| 49 | * |
| 50 | * Compare the ntfs attribute embedded in the ntfs specific part of the vfs |
| 51 | * inode @vi for equality with the ntfs attribute @na. |
| 52 | * |
| 53 | * If searching for the normal file/directory inode, set @na->type to AT_UNUSED. |
| 54 | * @na->name and @na->name_len are then ignored. |
| 55 | * |
| 56 | * Return 1 if the attributes match and 0 if not. |
| 57 | * |
| 58 | * NOTE: This function runs with the inode_lock spin lock held so it is not |
| 59 | * allowed to sleep. |
| 60 | */ |
| 61 | int ntfs_test_inode(struct inode *vi, ntfs_attr *na) |
| 62 | { |
| 63 | ntfs_inode *ni; |
| 64 | |
| 65 | if (vi->i_ino != na->mft_no) |
| 66 | return 0; |
| 67 | ni = NTFS_I(vi); |
| 68 | /* If !NInoAttr(ni), @vi is a normal file or directory inode. */ |
| 69 | if (likely(!NInoAttr(ni))) { |
| 70 | /* If not looking for a normal inode this is a mismatch. */ |
| 71 | if (unlikely(na->type != AT_UNUSED)) |
| 72 | return 0; |
| 73 | } else { |
| 74 | /* A fake inode describing an attribute. */ |
| 75 | if (ni->type != na->type) |
| 76 | return 0; |
| 77 | if (ni->name_len != na->name_len) |
| 78 | return 0; |
| 79 | if (na->name_len && memcmp(ni->name, na->name, |
| 80 | na->name_len * sizeof(ntfschar))) |
| 81 | return 0; |
| 82 | } |
| 83 | /* Match! */ |
| 84 | return 1; |
| 85 | } |
| 86 | |
| 87 | /** |
| 88 | * ntfs_init_locked_inode - initialize an inode |
| 89 | * @vi: vfs inode to initialize |
| 90 | * @na: ntfs attribute which to initialize @vi to |
| 91 | * |
| 92 | * Initialize the vfs inode @vi with the values from the ntfs attribute @na in |
| 93 | * order to enable ntfs_test_inode() to do its work. |
| 94 | * |
| 95 | * If initializing the normal file/directory inode, set @na->type to AT_UNUSED. |
| 96 | * In that case, @na->name and @na->name_len should be set to NULL and 0, |
| 97 | * respectively. Although that is not strictly necessary as |
| 98 | * ntfs_read_inode_locked() will fill them in later. |
| 99 | * |
| 100 | * Return 0 on success and -errno on error. |
| 101 | * |
| 102 | * NOTE: This function runs with the inode_lock spin lock held so it is not |
| 103 | * allowed to sleep. (Hence the GFP_ATOMIC allocation.) |
| 104 | */ |
| 105 | static int ntfs_init_locked_inode(struct inode *vi, ntfs_attr *na) |
| 106 | { |
| 107 | ntfs_inode *ni = NTFS_I(vi); |
| 108 | |
| 109 | vi->i_ino = na->mft_no; |
| 110 | |
| 111 | ni->type = na->type; |
| 112 | if (na->type == AT_INDEX_ALLOCATION) |
| 113 | NInoSetMstProtected(ni); |
| 114 | |
| 115 | ni->name = na->name; |
| 116 | ni->name_len = na->name_len; |
| 117 | |
| 118 | /* If initializing a normal inode, we are done. */ |
| 119 | if (likely(na->type == AT_UNUSED)) { |
| 120 | BUG_ON(na->name); |
| 121 | BUG_ON(na->name_len); |
| 122 | return 0; |
| 123 | } |
| 124 | |
| 125 | /* It is a fake inode. */ |
| 126 | NInoSetAttr(ni); |
| 127 | |
| 128 | /* |
| 129 | * We have I30 global constant as an optimization as it is the name |
| 130 | * in >99.9% of named attributes! The other <0.1% incur a GFP_ATOMIC |
| 131 | * allocation but that is ok. And most attributes are unnamed anyway, |
| 132 | * thus the fraction of named attributes with name != I30 is actually |
| 133 | * absolutely tiny. |
| 134 | */ |
| 135 | if (na->name_len && na->name != I30) { |
| 136 | unsigned int i; |
| 137 | |
| 138 | BUG_ON(!na->name); |
| 139 | i = na->name_len * sizeof(ntfschar); |
| 140 | ni->name = (ntfschar*)kmalloc(i + sizeof(ntfschar), GFP_ATOMIC); |
| 141 | if (!ni->name) |
| 142 | return -ENOMEM; |
| 143 | memcpy(ni->name, na->name, i); |
| 144 | ni->name[i] = 0; |
| 145 | } |
| 146 | return 0; |
| 147 | } |
| 148 | |
| 149 | typedef int (*set_t)(struct inode *, void *); |
| 150 | static int ntfs_read_locked_inode(struct inode *vi); |
| 151 | static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi); |
| 152 | static int ntfs_read_locked_index_inode(struct inode *base_vi, |
| 153 | struct inode *vi); |
| 154 | |
| 155 | /** |
| 156 | * ntfs_iget - obtain a struct inode corresponding to a specific normal inode |
| 157 | * @sb: super block of mounted volume |
| 158 | * @mft_no: mft record number / inode number to obtain |
| 159 | * |
| 160 | * Obtain the struct inode corresponding to a specific normal inode (i.e. a |
| 161 | * file or directory). |
| 162 | * |
| 163 | * If the inode is in the cache, it is just returned with an increased |
| 164 | * reference count. Otherwise, a new struct inode is allocated and initialized, |
| 165 | * and finally ntfs_read_locked_inode() is called to read in the inode and |
| 166 | * fill in the remainder of the inode structure. |
| 167 | * |
| 168 | * Return the struct inode on success. Check the return value with IS_ERR() and |
| 169 | * if true, the function failed and the error code is obtained from PTR_ERR(). |
| 170 | */ |
| 171 | struct inode *ntfs_iget(struct super_block *sb, unsigned long mft_no) |
| 172 | { |
| 173 | struct inode *vi; |
| 174 | ntfs_attr na; |
| 175 | int err; |
| 176 | |
| 177 | na.mft_no = mft_no; |
| 178 | na.type = AT_UNUSED; |
| 179 | na.name = NULL; |
| 180 | na.name_len = 0; |
| 181 | |
| 182 | vi = iget5_locked(sb, mft_no, (test_t)ntfs_test_inode, |
| 183 | (set_t)ntfs_init_locked_inode, &na); |
Anton Altaparmakov | f50f3ac | 2004-11-19 22:16:00 +0000 | [diff] [blame] | 184 | if (unlikely(!vi)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 185 | return ERR_PTR(-ENOMEM); |
| 186 | |
| 187 | err = 0; |
| 188 | |
| 189 | /* If this is a freshly allocated inode, need to read it now. */ |
| 190 | if (vi->i_state & I_NEW) { |
| 191 | err = ntfs_read_locked_inode(vi); |
| 192 | unlock_new_inode(vi); |
| 193 | } |
| 194 | /* |
| 195 | * There is no point in keeping bad inodes around if the failure was |
| 196 | * due to ENOMEM. We want to be able to retry again later. |
| 197 | */ |
Anton Altaparmakov | f50f3ac | 2004-11-19 22:16:00 +0000 | [diff] [blame] | 198 | if (unlikely(err == -ENOMEM)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 199 | iput(vi); |
| 200 | vi = ERR_PTR(err); |
| 201 | } |
| 202 | return vi; |
| 203 | } |
| 204 | |
| 205 | /** |
| 206 | * ntfs_attr_iget - obtain a struct inode corresponding to an attribute |
| 207 | * @base_vi: vfs base inode containing the attribute |
| 208 | * @type: attribute type |
| 209 | * @name: Unicode name of the attribute (NULL if unnamed) |
| 210 | * @name_len: length of @name in Unicode characters (0 if unnamed) |
| 211 | * |
| 212 | * Obtain the (fake) struct inode corresponding to the attribute specified by |
| 213 | * @type, @name, and @name_len, which is present in the base mft record |
| 214 | * specified by the vfs inode @base_vi. |
| 215 | * |
| 216 | * If the attribute inode is in the cache, it is just returned with an |
| 217 | * increased reference count. Otherwise, a new struct inode is allocated and |
| 218 | * initialized, and finally ntfs_read_locked_attr_inode() is called to read the |
| 219 | * attribute and fill in the inode structure. |
| 220 | * |
| 221 | * Note, for index allocation attributes, you need to use ntfs_index_iget() |
| 222 | * instead of ntfs_attr_iget() as working with indices is a lot more complex. |
| 223 | * |
| 224 | * Return the struct inode of the attribute inode on success. Check the return |
| 225 | * value with IS_ERR() and if true, the function failed and the error code is |
| 226 | * obtained from PTR_ERR(). |
| 227 | */ |
| 228 | struct inode *ntfs_attr_iget(struct inode *base_vi, ATTR_TYPE type, |
| 229 | ntfschar *name, u32 name_len) |
| 230 | { |
| 231 | struct inode *vi; |
| 232 | ntfs_attr na; |
| 233 | int err; |
| 234 | |
| 235 | /* Make sure no one calls ntfs_attr_iget() for indices. */ |
| 236 | BUG_ON(type == AT_INDEX_ALLOCATION); |
| 237 | |
| 238 | na.mft_no = base_vi->i_ino; |
| 239 | na.type = type; |
| 240 | na.name = name; |
| 241 | na.name_len = name_len; |
| 242 | |
| 243 | vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode, |
| 244 | (set_t)ntfs_init_locked_inode, &na); |
Anton Altaparmakov | f50f3ac | 2004-11-19 22:16:00 +0000 | [diff] [blame] | 245 | if (unlikely(!vi)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 246 | return ERR_PTR(-ENOMEM); |
| 247 | |
| 248 | err = 0; |
| 249 | |
| 250 | /* If this is a freshly allocated inode, need to read it now. */ |
| 251 | if (vi->i_state & I_NEW) { |
| 252 | err = ntfs_read_locked_attr_inode(base_vi, vi); |
| 253 | unlock_new_inode(vi); |
| 254 | } |
| 255 | /* |
| 256 | * There is no point in keeping bad attribute inodes around. This also |
| 257 | * simplifies things in that we never need to check for bad attribute |
| 258 | * inodes elsewhere. |
| 259 | */ |
Anton Altaparmakov | f50f3ac | 2004-11-19 22:16:00 +0000 | [diff] [blame] | 260 | if (unlikely(err)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 261 | iput(vi); |
| 262 | vi = ERR_PTR(err); |
| 263 | } |
| 264 | return vi; |
| 265 | } |
| 266 | |
| 267 | /** |
| 268 | * ntfs_index_iget - obtain a struct inode corresponding to an index |
| 269 | * @base_vi: vfs base inode containing the index related attributes |
| 270 | * @name: Unicode name of the index |
| 271 | * @name_len: length of @name in Unicode characters |
| 272 | * |
| 273 | * Obtain the (fake) struct inode corresponding to the index specified by @name |
| 274 | * and @name_len, which is present in the base mft record specified by the vfs |
| 275 | * inode @base_vi. |
| 276 | * |
| 277 | * If the index inode is in the cache, it is just returned with an increased |
| 278 | * reference count. Otherwise, a new struct inode is allocated and |
| 279 | * initialized, and finally ntfs_read_locked_index_inode() is called to read |
| 280 | * the index related attributes and fill in the inode structure. |
| 281 | * |
| 282 | * Return the struct inode of the index inode on success. Check the return |
| 283 | * value with IS_ERR() and if true, the function failed and the error code is |
| 284 | * obtained from PTR_ERR(). |
| 285 | */ |
| 286 | struct inode *ntfs_index_iget(struct inode *base_vi, ntfschar *name, |
| 287 | u32 name_len) |
| 288 | { |
| 289 | struct inode *vi; |
| 290 | ntfs_attr na; |
| 291 | int err; |
| 292 | |
| 293 | na.mft_no = base_vi->i_ino; |
| 294 | na.type = AT_INDEX_ALLOCATION; |
| 295 | na.name = name; |
| 296 | na.name_len = name_len; |
| 297 | |
| 298 | vi = iget5_locked(base_vi->i_sb, na.mft_no, (test_t)ntfs_test_inode, |
| 299 | (set_t)ntfs_init_locked_inode, &na); |
Anton Altaparmakov | f50f3ac | 2004-11-19 22:16:00 +0000 | [diff] [blame] | 300 | if (unlikely(!vi)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 301 | return ERR_PTR(-ENOMEM); |
| 302 | |
| 303 | err = 0; |
| 304 | |
| 305 | /* If this is a freshly allocated inode, need to read it now. */ |
| 306 | if (vi->i_state & I_NEW) { |
| 307 | err = ntfs_read_locked_index_inode(base_vi, vi); |
| 308 | unlock_new_inode(vi); |
| 309 | } |
| 310 | /* |
| 311 | * There is no point in keeping bad index inodes around. This also |
| 312 | * simplifies things in that we never need to check for bad index |
| 313 | * inodes elsewhere. |
| 314 | */ |
Anton Altaparmakov | f50f3ac | 2004-11-19 22:16:00 +0000 | [diff] [blame] | 315 | if (unlikely(err)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 316 | iput(vi); |
| 317 | vi = ERR_PTR(err); |
| 318 | } |
| 319 | return vi; |
| 320 | } |
| 321 | |
| 322 | struct inode *ntfs_alloc_big_inode(struct super_block *sb) |
| 323 | { |
| 324 | ntfs_inode *ni; |
| 325 | |
| 326 | ntfs_debug("Entering."); |
Pekka Enberg | 2fb21db | 2005-05-25 21:15:34 +0300 | [diff] [blame] | 327 | ni = kmem_cache_alloc(ntfs_big_inode_cache, SLAB_NOFS); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 328 | if (likely(ni != NULL)) { |
| 329 | ni->state = 0; |
| 330 | return VFS_I(ni); |
| 331 | } |
| 332 | ntfs_error(sb, "Allocation of NTFS big inode structure failed."); |
| 333 | return NULL; |
| 334 | } |
| 335 | |
| 336 | void ntfs_destroy_big_inode(struct inode *inode) |
| 337 | { |
| 338 | ntfs_inode *ni = NTFS_I(inode); |
| 339 | |
| 340 | ntfs_debug("Entering."); |
| 341 | BUG_ON(ni->page); |
| 342 | if (!atomic_dec_and_test(&ni->count)) |
| 343 | BUG(); |
| 344 | kmem_cache_free(ntfs_big_inode_cache, NTFS_I(inode)); |
| 345 | } |
| 346 | |
| 347 | static inline ntfs_inode *ntfs_alloc_extent_inode(void) |
| 348 | { |
| 349 | ntfs_inode *ni; |
| 350 | |
| 351 | ntfs_debug("Entering."); |
Pekka Enberg | 2fb21db | 2005-05-25 21:15:34 +0300 | [diff] [blame] | 352 | ni = kmem_cache_alloc(ntfs_inode_cache, SLAB_NOFS); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 353 | if (likely(ni != NULL)) { |
| 354 | ni->state = 0; |
| 355 | return ni; |
| 356 | } |
| 357 | ntfs_error(NULL, "Allocation of NTFS inode structure failed."); |
| 358 | return NULL; |
| 359 | } |
| 360 | |
| 361 | static void ntfs_destroy_extent_inode(ntfs_inode *ni) |
| 362 | { |
| 363 | ntfs_debug("Entering."); |
| 364 | BUG_ON(ni->page); |
| 365 | if (!atomic_dec_and_test(&ni->count)) |
| 366 | BUG(); |
| 367 | kmem_cache_free(ntfs_inode_cache, ni); |
| 368 | } |
| 369 | |
Ingo Molnar | 5934537 | 2006-07-03 00:25:18 -0700 | [diff] [blame^] | 370 | /* |
| 371 | * The attribute runlist lock has separate locking rules from the |
| 372 | * normal runlist lock, so split the two lock-classes: |
| 373 | */ |
| 374 | static struct lock_class_key attr_list_rl_lock_class; |
| 375 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 376 | /** |
| 377 | * __ntfs_init_inode - initialize ntfs specific part of an inode |
| 378 | * @sb: super block of mounted volume |
| 379 | * @ni: freshly allocated ntfs inode which to initialize |
| 380 | * |
| 381 | * Initialize an ntfs inode to defaults. |
| 382 | * |
| 383 | * NOTE: ni->mft_no, ni->state, ni->type, ni->name, and ni->name_len are left |
| 384 | * untouched. Make sure to initialize them elsewhere. |
| 385 | * |
| 386 | * Return zero on success and -ENOMEM on error. |
| 387 | */ |
| 388 | void __ntfs_init_inode(struct super_block *sb, ntfs_inode *ni) |
| 389 | { |
| 390 | ntfs_debug("Entering."); |
Anton Altaparmakov | 3676367 | 2004-11-18 13:46:45 +0000 | [diff] [blame] | 391 | rwlock_init(&ni->size_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 392 | ni->initialized_size = ni->allocated_size = 0; |
| 393 | ni->seq_no = 0; |
| 394 | atomic_set(&ni->count, 1); |
| 395 | ni->vol = NTFS_SB(sb); |
| 396 | ntfs_init_runlist(&ni->runlist); |
Ingo Molnar | 4e5e529 | 2006-03-23 16:57:48 +0000 | [diff] [blame] | 397 | mutex_init(&ni->mrec_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 398 | ni->page = NULL; |
| 399 | ni->page_ofs = 0; |
| 400 | ni->attr_list_size = 0; |
| 401 | ni->attr_list = NULL; |
| 402 | ntfs_init_runlist(&ni->attr_list_rl); |
Ingo Molnar | 5934537 | 2006-07-03 00:25:18 -0700 | [diff] [blame^] | 403 | lockdep_set_class(&ni->attr_list_rl.lock, |
| 404 | &attr_list_rl_lock_class); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 405 | ni->itype.index.bmp_ino = NULL; |
| 406 | ni->itype.index.block_size = 0; |
| 407 | ni->itype.index.vcn_size = 0; |
| 408 | ni->itype.index.collation_rule = 0; |
| 409 | ni->itype.index.block_size_bits = 0; |
| 410 | ni->itype.index.vcn_size_bits = 0; |
Ingo Molnar | 4e5e529 | 2006-03-23 16:57:48 +0000 | [diff] [blame] | 411 | mutex_init(&ni->extent_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 412 | ni->nr_extents = 0; |
| 413 | ni->ext.base_ntfs_ino = NULL; |
| 414 | } |
| 415 | |
Ingo Molnar | 5934537 | 2006-07-03 00:25:18 -0700 | [diff] [blame^] | 416 | /* |
| 417 | * Extent inodes get MFT-mapped in a nested way, while the base inode |
| 418 | * is still mapped. Teach this nesting to the lock validator by creating |
| 419 | * a separate class for nested inode's mrec_lock's: |
| 420 | */ |
| 421 | static struct lock_class_key extent_inode_mrec_lock_key; |
| 422 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 423 | inline ntfs_inode *ntfs_new_extent_inode(struct super_block *sb, |
| 424 | unsigned long mft_no) |
| 425 | { |
| 426 | ntfs_inode *ni = ntfs_alloc_extent_inode(); |
| 427 | |
| 428 | ntfs_debug("Entering."); |
| 429 | if (likely(ni != NULL)) { |
| 430 | __ntfs_init_inode(sb, ni); |
Ingo Molnar | 5934537 | 2006-07-03 00:25:18 -0700 | [diff] [blame^] | 431 | lockdep_set_class(&ni->mrec_lock, &extent_inode_mrec_lock_key); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 432 | ni->mft_no = mft_no; |
| 433 | ni->type = AT_UNUSED; |
| 434 | ni->name = NULL; |
| 435 | ni->name_len = 0; |
| 436 | } |
| 437 | return ni; |
| 438 | } |
| 439 | |
| 440 | /** |
| 441 | * ntfs_is_extended_system_file - check if a file is in the $Extend directory |
| 442 | * @ctx: initialized attribute search context |
| 443 | * |
| 444 | * Search all file name attributes in the inode described by the attribute |
| 445 | * search context @ctx and check if any of the names are in the $Extend system |
| 446 | * directory. |
| 447 | * |
| 448 | * Return values: |
| 449 | * 1: file is in $Extend directory |
| 450 | * 0: file is not in $Extend directory |
| 451 | * -errno: failed to determine if the file is in the $Extend directory |
| 452 | */ |
| 453 | static int ntfs_is_extended_system_file(ntfs_attr_search_ctx *ctx) |
| 454 | { |
| 455 | int nr_links, err; |
| 456 | |
| 457 | /* Restart search. */ |
| 458 | ntfs_attr_reinit_search_ctx(ctx); |
| 459 | |
| 460 | /* Get number of hard links. */ |
| 461 | nr_links = le16_to_cpu(ctx->mrec->link_count); |
| 462 | |
| 463 | /* Loop through all hard links. */ |
| 464 | while (!(err = ntfs_attr_lookup(AT_FILE_NAME, NULL, 0, 0, 0, NULL, 0, |
| 465 | ctx))) { |
| 466 | FILE_NAME_ATTR *file_name_attr; |
| 467 | ATTR_RECORD *attr = ctx->attr; |
| 468 | u8 *p, *p2; |
| 469 | |
| 470 | nr_links--; |
| 471 | /* |
| 472 | * Maximum sanity checking as we are called on an inode that |
| 473 | * we suspect might be corrupt. |
| 474 | */ |
| 475 | p = (u8*)attr + le32_to_cpu(attr->length); |
| 476 | if (p < (u8*)ctx->mrec || (u8*)p > (u8*)ctx->mrec + |
| 477 | le32_to_cpu(ctx->mrec->bytes_in_use)) { |
| 478 | err_corrupt_attr: |
| 479 | ntfs_error(ctx->ntfs_ino->vol->sb, "Corrupt file name " |
| 480 | "attribute. You should run chkdsk."); |
| 481 | return -EIO; |
| 482 | } |
| 483 | if (attr->non_resident) { |
| 484 | ntfs_error(ctx->ntfs_ino->vol->sb, "Non-resident file " |
| 485 | "name. You should run chkdsk."); |
| 486 | return -EIO; |
| 487 | } |
| 488 | if (attr->flags) { |
| 489 | ntfs_error(ctx->ntfs_ino->vol->sb, "File name with " |
| 490 | "invalid flags. You should run " |
| 491 | "chkdsk."); |
| 492 | return -EIO; |
| 493 | } |
| 494 | if (!(attr->data.resident.flags & RESIDENT_ATTR_IS_INDEXED)) { |
| 495 | ntfs_error(ctx->ntfs_ino->vol->sb, "Unindexed file " |
| 496 | "name. You should run chkdsk."); |
| 497 | return -EIO; |
| 498 | } |
| 499 | file_name_attr = (FILE_NAME_ATTR*)((u8*)attr + |
| 500 | le16_to_cpu(attr->data.resident.value_offset)); |
| 501 | p2 = (u8*)attr + le32_to_cpu(attr->data.resident.value_length); |
| 502 | if (p2 < (u8*)attr || p2 > p) |
| 503 | goto err_corrupt_attr; |
| 504 | /* This attribute is ok, but is it in the $Extend directory? */ |
| 505 | if (MREF_LE(file_name_attr->parent_directory) == FILE_Extend) |
| 506 | return 1; /* YES, it's an extended system file. */ |
| 507 | } |
| 508 | if (unlikely(err != -ENOENT)) |
| 509 | return err; |
| 510 | if (unlikely(nr_links)) { |
| 511 | ntfs_error(ctx->ntfs_ino->vol->sb, "Inode hard link count " |
| 512 | "doesn't match number of name attributes. You " |
| 513 | "should run chkdsk."); |
| 514 | return -EIO; |
| 515 | } |
| 516 | return 0; /* NO, it is not an extended system file. */ |
| 517 | } |
| 518 | |
| 519 | /** |
| 520 | * ntfs_read_locked_inode - read an inode from its device |
| 521 | * @vi: inode to read |
| 522 | * |
| 523 | * ntfs_read_locked_inode() is called from ntfs_iget() to read the inode |
| 524 | * described by @vi into memory from the device. |
| 525 | * |
| 526 | * The only fields in @vi that we need to/can look at when the function is |
| 527 | * called are i_sb, pointing to the mounted device's super block, and i_ino, |
| 528 | * the number of the inode to load. |
| 529 | * |
| 530 | * ntfs_read_locked_inode() maps, pins and locks the mft record number i_ino |
| 531 | * for reading and sets up the necessary @vi fields as well as initializing |
| 532 | * the ntfs inode. |
| 533 | * |
| 534 | * Q: What locks are held when the function is called? |
| 535 | * A: i_state has I_LOCK set, hence the inode is locked, also |
| 536 | * i_count is set to 1, so it is not going to go away |
| 537 | * i_flags is set to 0 and we have no business touching it. Only an ioctl() |
| 538 | * is allowed to write to them. We should of course be honouring them but |
| 539 | * we need to do that using the IS_* macros defined in include/linux/fs.h. |
| 540 | * In any case ntfs_read_locked_inode() has nothing to do with i_flags. |
| 541 | * |
| 542 | * Return 0 on success and -errno on error. In the error case, the inode will |
| 543 | * have had make_bad_inode() executed on it. |
| 544 | */ |
| 545 | static int ntfs_read_locked_inode(struct inode *vi) |
| 546 | { |
| 547 | ntfs_volume *vol = NTFS_SB(vi->i_sb); |
| 548 | ntfs_inode *ni; |
| 549 | MFT_RECORD *m; |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 550 | ATTR_RECORD *a; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 551 | STANDARD_INFORMATION *si; |
| 552 | ntfs_attr_search_ctx *ctx; |
| 553 | int err = 0; |
| 554 | |
| 555 | ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino); |
| 556 | |
| 557 | /* Setup the generic vfs inode parts now. */ |
| 558 | |
| 559 | /* This is the optimal IO size (for stat), not the fs block size. */ |
| 560 | vi->i_blksize = PAGE_CACHE_SIZE; |
| 561 | /* |
| 562 | * This is for checking whether an inode has changed w.r.t. a file so |
| 563 | * that the file can be updated if necessary (compare with f_version). |
| 564 | */ |
| 565 | vi->i_version = 1; |
| 566 | |
| 567 | vi->i_uid = vol->uid; |
| 568 | vi->i_gid = vol->gid; |
| 569 | vi->i_mode = 0; |
| 570 | |
| 571 | /* |
| 572 | * Initialize the ntfs specific part of @vi special casing |
| 573 | * FILE_MFT which we need to do at mount time. |
| 574 | */ |
| 575 | if (vi->i_ino != FILE_MFT) |
| 576 | ntfs_init_big_inode(vi); |
| 577 | ni = NTFS_I(vi); |
| 578 | |
| 579 | m = map_mft_record(ni); |
| 580 | if (IS_ERR(m)) { |
| 581 | err = PTR_ERR(m); |
| 582 | goto err_out; |
| 583 | } |
| 584 | ctx = ntfs_attr_get_search_ctx(ni, m); |
| 585 | if (!ctx) { |
| 586 | err = -ENOMEM; |
| 587 | goto unm_err_out; |
| 588 | } |
| 589 | |
| 590 | if (!(m->flags & MFT_RECORD_IN_USE)) { |
| 591 | ntfs_error(vi->i_sb, "Inode is not in use!"); |
| 592 | goto unm_err_out; |
| 593 | } |
| 594 | if (m->base_mft_record) { |
| 595 | ntfs_error(vi->i_sb, "Inode is an extent inode!"); |
| 596 | goto unm_err_out; |
| 597 | } |
| 598 | |
| 599 | /* Transfer information from mft record into vfs and ntfs inodes. */ |
| 600 | vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number); |
| 601 | |
| 602 | /* |
| 603 | * FIXME: Keep in mind that link_count is two for files which have both |
| 604 | * a long file name and a short file name as separate entries, so if |
| 605 | * we are hiding short file names this will be too high. Either we need |
| 606 | * to account for the short file names by subtracting them or we need |
| 607 | * to make sure we delete files even though i_nlink is not zero which |
| 608 | * might be tricky due to vfs interactions. Need to think about this |
| 609 | * some more when implementing the unlink command. |
| 610 | */ |
| 611 | vi->i_nlink = le16_to_cpu(m->link_count); |
| 612 | /* |
| 613 | * FIXME: Reparse points can have the directory bit set even though |
| 614 | * they would be S_IFLNK. Need to deal with this further below when we |
| 615 | * implement reparse points / symbolic links but it will do for now. |
| 616 | * Also if not a directory, it could be something else, rather than |
| 617 | * a regular file. But again, will do for now. |
| 618 | */ |
| 619 | /* Everyone gets all permissions. */ |
| 620 | vi->i_mode |= S_IRWXUGO; |
| 621 | /* If read-only, noone gets write permissions. */ |
| 622 | if (IS_RDONLY(vi)) |
| 623 | vi->i_mode &= ~S_IWUGO; |
| 624 | if (m->flags & MFT_RECORD_IS_DIRECTORY) { |
| 625 | vi->i_mode |= S_IFDIR; |
| 626 | /* |
| 627 | * Apply the directory permissions mask set in the mount |
| 628 | * options. |
| 629 | */ |
| 630 | vi->i_mode &= ~vol->dmask; |
| 631 | /* Things break without this kludge! */ |
| 632 | if (vi->i_nlink > 1) |
| 633 | vi->i_nlink = 1; |
| 634 | } else { |
| 635 | vi->i_mode |= S_IFREG; |
| 636 | /* Apply the file permissions mask set in the mount options. */ |
| 637 | vi->i_mode &= ~vol->fmask; |
| 638 | } |
| 639 | /* |
| 640 | * Find the standard information attribute in the mft record. At this |
| 641 | * stage we haven't setup the attribute list stuff yet, so this could |
| 642 | * in fact fail if the standard information is in an extent record, but |
| 643 | * I don't think this actually ever happens. |
| 644 | */ |
| 645 | err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, 0, 0, NULL, 0, |
| 646 | ctx); |
| 647 | if (unlikely(err)) { |
| 648 | if (err == -ENOENT) { |
| 649 | /* |
| 650 | * TODO: We should be performing a hot fix here (if the |
| 651 | * recover mount option is set) by creating a new |
| 652 | * attribute. |
| 653 | */ |
| 654 | ntfs_error(vi->i_sb, "$STANDARD_INFORMATION attribute " |
| 655 | "is missing."); |
| 656 | } |
| 657 | goto unm_err_out; |
| 658 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 659 | a = ctx->attr; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 660 | /* Get the standard information attribute value. */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 661 | si = (STANDARD_INFORMATION*)((u8*)a + |
| 662 | le16_to_cpu(a->data.resident.value_offset)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 663 | |
| 664 | /* Transfer information from the standard information into vi. */ |
| 665 | /* |
| 666 | * Note: The i_?times do not quite map perfectly onto the NTFS times, |
| 667 | * but they are close enough, and in the end it doesn't really matter |
| 668 | * that much... |
| 669 | */ |
| 670 | /* |
| 671 | * mtime is the last change of the data within the file. Not changed |
| 672 | * when only metadata is changed, e.g. a rename doesn't affect mtime. |
| 673 | */ |
| 674 | vi->i_mtime = ntfs2utc(si->last_data_change_time); |
| 675 | /* |
| 676 | * ctime is the last change of the metadata of the file. This obviously |
| 677 | * always changes, when mtime is changed. ctime can be changed on its |
| 678 | * own, mtime is then not changed, e.g. when a file is renamed. |
| 679 | */ |
| 680 | vi->i_ctime = ntfs2utc(si->last_mft_change_time); |
| 681 | /* |
| 682 | * Last access to the data within the file. Not changed during a rename |
| 683 | * for example but changed whenever the file is written to. |
| 684 | */ |
| 685 | vi->i_atime = ntfs2utc(si->last_access_time); |
| 686 | |
| 687 | /* Find the attribute list attribute if present. */ |
| 688 | ntfs_attr_reinit_search_ctx(ctx); |
| 689 | err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx); |
| 690 | if (err) { |
| 691 | if (unlikely(err != -ENOENT)) { |
| 692 | ntfs_error(vi->i_sb, "Failed to lookup attribute list " |
| 693 | "attribute."); |
| 694 | goto unm_err_out; |
| 695 | } |
| 696 | } else /* if (!err) */ { |
| 697 | if (vi->i_ino == FILE_MFT) |
| 698 | goto skip_attr_list_load; |
| 699 | ntfs_debug("Attribute list found in inode 0x%lx.", vi->i_ino); |
| 700 | NInoSetAttrList(ni); |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 701 | a = ctx->attr; |
Anton Altaparmakov | 3672b63 | 2006-02-24 09:55:07 +0000 | [diff] [blame] | 702 | if (a->flags & ATTR_COMPRESSION_MASK) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 703 | ntfs_error(vi->i_sb, "Attribute list attribute is " |
Anton Altaparmakov | 3672b63 | 2006-02-24 09:55:07 +0000 | [diff] [blame] | 704 | "compressed."); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 705 | goto unm_err_out; |
| 706 | } |
Anton Altaparmakov | 3672b63 | 2006-02-24 09:55:07 +0000 | [diff] [blame] | 707 | if (a->flags & ATTR_IS_ENCRYPTED || |
| 708 | a->flags & ATTR_IS_SPARSE) { |
| 709 | if (a->non_resident) { |
| 710 | ntfs_error(vi->i_sb, "Non-resident attribute " |
| 711 | "list attribute is encrypted/" |
| 712 | "sparse."); |
| 713 | goto unm_err_out; |
| 714 | } |
| 715 | ntfs_warning(vi->i_sb, "Resident attribute list " |
| 716 | "attribute in inode 0x%lx is marked " |
| 717 | "encrypted/sparse which is not true. " |
| 718 | "However, Windows allows this and " |
| 719 | "chkdsk does not detect or correct it " |
| 720 | "so we will just ignore the invalid " |
| 721 | "flags and pretend they are not set.", |
| 722 | vi->i_ino); |
| 723 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 724 | /* Now allocate memory for the attribute list. */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 725 | ni->attr_list_size = (u32)ntfs_attr_size(a); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 726 | ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size); |
| 727 | if (!ni->attr_list) { |
| 728 | ntfs_error(vi->i_sb, "Not enough memory to allocate " |
| 729 | "buffer for attribute list."); |
| 730 | err = -ENOMEM; |
| 731 | goto unm_err_out; |
| 732 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 733 | if (a->non_resident) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 734 | NInoSetAttrListNonResident(ni); |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 735 | if (a->data.non_resident.lowest_vcn) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 736 | ntfs_error(vi->i_sb, "Attribute list has non " |
| 737 | "zero lowest_vcn."); |
| 738 | goto unm_err_out; |
| 739 | } |
| 740 | /* |
| 741 | * Setup the runlist. No need for locking as we have |
| 742 | * exclusive access to the inode at this time. |
| 743 | */ |
| 744 | ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol, |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 745 | a, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 746 | if (IS_ERR(ni->attr_list_rl.rl)) { |
| 747 | err = PTR_ERR(ni->attr_list_rl.rl); |
| 748 | ni->attr_list_rl.rl = NULL; |
| 749 | ntfs_error(vi->i_sb, "Mapping pairs " |
| 750 | "decompression failed."); |
| 751 | goto unm_err_out; |
| 752 | } |
| 753 | /* Now load the attribute list. */ |
| 754 | if ((err = load_attribute_list(vol, &ni->attr_list_rl, |
| 755 | ni->attr_list, ni->attr_list_size, |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 756 | sle64_to_cpu(a->data.non_resident. |
| 757 | initialized_size)))) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 758 | ntfs_error(vi->i_sb, "Failed to load " |
| 759 | "attribute list attribute."); |
| 760 | goto unm_err_out; |
| 761 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 762 | } else /* if (!a->non_resident) */ { |
| 763 | if ((u8*)a + le16_to_cpu(a->data.resident.value_offset) |
| 764 | + le32_to_cpu( |
| 765 | a->data.resident.value_length) > |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 766 | (u8*)ctx->mrec + vol->mft_record_size) { |
| 767 | ntfs_error(vi->i_sb, "Corrupt attribute list " |
| 768 | "in inode."); |
| 769 | goto unm_err_out; |
| 770 | } |
| 771 | /* Now copy the attribute list. */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 772 | memcpy(ni->attr_list, (u8*)a + le16_to_cpu( |
| 773 | a->data.resident.value_offset), |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 774 | le32_to_cpu( |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 775 | a->data.resident.value_length)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 776 | } |
| 777 | } |
| 778 | skip_attr_list_load: |
| 779 | /* |
| 780 | * If an attribute list is present we now have the attribute list value |
| 781 | * in ntfs_ino->attr_list and it is ntfs_ino->attr_list_size bytes. |
| 782 | */ |
| 783 | if (S_ISDIR(vi->i_mode)) { |
Anton Altaparmakov | f50f3ac | 2004-11-19 22:16:00 +0000 | [diff] [blame] | 784 | loff_t bvi_size; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 785 | struct inode *bvi; |
| 786 | ntfs_inode *bni; |
| 787 | INDEX_ROOT *ir; |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 788 | u8 *ir_end, *index_end; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 789 | |
| 790 | /* It is a directory, find index root attribute. */ |
| 791 | ntfs_attr_reinit_search_ctx(ctx); |
| 792 | err = ntfs_attr_lookup(AT_INDEX_ROOT, I30, 4, CASE_SENSITIVE, |
| 793 | 0, NULL, 0, ctx); |
| 794 | if (unlikely(err)) { |
| 795 | if (err == -ENOENT) { |
| 796 | // FIXME: File is corrupt! Hot-fix with empty |
| 797 | // index root attribute if recovery option is |
| 798 | // set. |
| 799 | ntfs_error(vi->i_sb, "$INDEX_ROOT attribute " |
| 800 | "is missing."); |
| 801 | } |
| 802 | goto unm_err_out; |
| 803 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 804 | a = ctx->attr; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 805 | /* Set up the state. */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 806 | if (unlikely(a->non_resident)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 807 | ntfs_error(vol->sb, "$INDEX_ROOT attribute is not " |
| 808 | "resident."); |
| 809 | goto unm_err_out; |
| 810 | } |
| 811 | /* Ensure the attribute name is placed before the value. */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 812 | if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= |
| 813 | le16_to_cpu(a->data.resident.value_offset)))) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 814 | ntfs_error(vol->sb, "$INDEX_ROOT attribute name is " |
| 815 | "placed after the attribute value."); |
| 816 | goto unm_err_out; |
| 817 | } |
| 818 | /* |
| 819 | * Compressed/encrypted index root just means that the newly |
| 820 | * created files in that directory should be created compressed/ |
| 821 | * encrypted. However index root cannot be both compressed and |
| 822 | * encrypted. |
| 823 | */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 824 | if (a->flags & ATTR_COMPRESSION_MASK) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 825 | NInoSetCompressed(ni); |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 826 | if (a->flags & ATTR_IS_ENCRYPTED) { |
| 827 | if (a->flags & ATTR_COMPRESSION_MASK) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 828 | ntfs_error(vi->i_sb, "Found encrypted and " |
| 829 | "compressed attribute."); |
| 830 | goto unm_err_out; |
| 831 | } |
| 832 | NInoSetEncrypted(ni); |
| 833 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 834 | if (a->flags & ATTR_IS_SPARSE) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 835 | NInoSetSparse(ni); |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 836 | ir = (INDEX_ROOT*)((u8*)a + |
| 837 | le16_to_cpu(a->data.resident.value_offset)); |
| 838 | ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length); |
| 839 | if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 840 | ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is " |
| 841 | "corrupt."); |
| 842 | goto unm_err_out; |
| 843 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 844 | index_end = (u8*)&ir->index + |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 845 | le32_to_cpu(ir->index.index_length); |
| 846 | if (index_end > ir_end) { |
| 847 | ntfs_error(vi->i_sb, "Directory index is corrupt."); |
| 848 | goto unm_err_out; |
| 849 | } |
| 850 | if (ir->type != AT_FILE_NAME) { |
| 851 | ntfs_error(vi->i_sb, "Indexed attribute is not " |
| 852 | "$FILE_NAME."); |
| 853 | goto unm_err_out; |
| 854 | } |
| 855 | if (ir->collation_rule != COLLATION_FILE_NAME) { |
| 856 | ntfs_error(vi->i_sb, "Index collation rule is not " |
| 857 | "COLLATION_FILE_NAME."); |
| 858 | goto unm_err_out; |
| 859 | } |
| 860 | ni->itype.index.collation_rule = ir->collation_rule; |
| 861 | ni->itype.index.block_size = le32_to_cpu(ir->index_block_size); |
| 862 | if (ni->itype.index.block_size & |
| 863 | (ni->itype.index.block_size - 1)) { |
| 864 | ntfs_error(vi->i_sb, "Index block size (%u) is not a " |
| 865 | "power of two.", |
| 866 | ni->itype.index.block_size); |
| 867 | goto unm_err_out; |
| 868 | } |
| 869 | if (ni->itype.index.block_size > PAGE_CACHE_SIZE) { |
| 870 | ntfs_error(vi->i_sb, "Index block size (%u) > " |
| 871 | "PAGE_CACHE_SIZE (%ld) is not " |
| 872 | "supported. Sorry.", |
| 873 | ni->itype.index.block_size, |
| 874 | PAGE_CACHE_SIZE); |
| 875 | err = -EOPNOTSUPP; |
| 876 | goto unm_err_out; |
| 877 | } |
| 878 | if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) { |
| 879 | ntfs_error(vi->i_sb, "Index block size (%u) < " |
| 880 | "NTFS_BLOCK_SIZE (%i) is not " |
| 881 | "supported. Sorry.", |
| 882 | ni->itype.index.block_size, |
| 883 | NTFS_BLOCK_SIZE); |
| 884 | err = -EOPNOTSUPP; |
| 885 | goto unm_err_out; |
| 886 | } |
| 887 | ni->itype.index.block_size_bits = |
| 888 | ffs(ni->itype.index.block_size) - 1; |
| 889 | /* Determine the size of a vcn in the directory index. */ |
| 890 | if (vol->cluster_size <= ni->itype.index.block_size) { |
| 891 | ni->itype.index.vcn_size = vol->cluster_size; |
| 892 | ni->itype.index.vcn_size_bits = vol->cluster_size_bits; |
| 893 | } else { |
| 894 | ni->itype.index.vcn_size = vol->sector_size; |
| 895 | ni->itype.index.vcn_size_bits = vol->sector_size_bits; |
| 896 | } |
| 897 | |
| 898 | /* Setup the index allocation attribute, even if not present. */ |
| 899 | NInoSetMstProtected(ni); |
| 900 | ni->type = AT_INDEX_ALLOCATION; |
| 901 | ni->name = I30; |
| 902 | ni->name_len = 4; |
| 903 | |
| 904 | if (!(ir->index.flags & LARGE_INDEX)) { |
| 905 | /* No index allocation. */ |
| 906 | vi->i_size = ni->initialized_size = |
| 907 | ni->allocated_size = 0; |
| 908 | /* We are done with the mft record, so we release it. */ |
| 909 | ntfs_attr_put_search_ctx(ctx); |
| 910 | unmap_mft_record(ni); |
| 911 | m = NULL; |
| 912 | ctx = NULL; |
| 913 | goto skip_large_dir_stuff; |
| 914 | } /* LARGE_INDEX: Index allocation present. Setup state. */ |
| 915 | NInoSetIndexAllocPresent(ni); |
| 916 | /* Find index allocation attribute. */ |
| 917 | ntfs_attr_reinit_search_ctx(ctx); |
| 918 | err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, I30, 4, |
| 919 | CASE_SENSITIVE, 0, NULL, 0, ctx); |
| 920 | if (unlikely(err)) { |
| 921 | if (err == -ENOENT) |
| 922 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION " |
| 923 | "attribute is not present but " |
| 924 | "$INDEX_ROOT indicated it is."); |
| 925 | else |
| 926 | ntfs_error(vi->i_sb, "Failed to lookup " |
| 927 | "$INDEX_ALLOCATION " |
| 928 | "attribute."); |
| 929 | goto unm_err_out; |
| 930 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 931 | a = ctx->attr; |
| 932 | if (!a->non_resident) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 933 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute " |
| 934 | "is resident."); |
| 935 | goto unm_err_out; |
| 936 | } |
| 937 | /* |
| 938 | * Ensure the attribute name is placed before the mapping pairs |
| 939 | * array. |
| 940 | */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 941 | if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= |
| 942 | le16_to_cpu( |
| 943 | a->data.non_resident.mapping_pairs_offset)))) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 944 | ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name " |
| 945 | "is placed after the mapping pairs " |
| 946 | "array."); |
| 947 | goto unm_err_out; |
| 948 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 949 | if (a->flags & ATTR_IS_ENCRYPTED) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 950 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute " |
| 951 | "is encrypted."); |
| 952 | goto unm_err_out; |
| 953 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 954 | if (a->flags & ATTR_IS_SPARSE) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 955 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute " |
| 956 | "is sparse."); |
| 957 | goto unm_err_out; |
| 958 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 959 | if (a->flags & ATTR_COMPRESSION_MASK) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 960 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute " |
| 961 | "is compressed."); |
| 962 | goto unm_err_out; |
| 963 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 964 | if (a->data.non_resident.lowest_vcn) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 965 | ntfs_error(vi->i_sb, "First extent of " |
| 966 | "$INDEX_ALLOCATION attribute has non " |
| 967 | "zero lowest_vcn."); |
| 968 | goto unm_err_out; |
| 969 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 970 | vi->i_size = sle64_to_cpu(a->data.non_resident.data_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 971 | ni->initialized_size = sle64_to_cpu( |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 972 | a->data.non_resident.initialized_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 973 | ni->allocated_size = sle64_to_cpu( |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 974 | a->data.non_resident.allocated_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 975 | /* |
| 976 | * We are done with the mft record, so we release it. Otherwise |
| 977 | * we would deadlock in ntfs_attr_iget(). |
| 978 | */ |
| 979 | ntfs_attr_put_search_ctx(ctx); |
| 980 | unmap_mft_record(ni); |
| 981 | m = NULL; |
| 982 | ctx = NULL; |
| 983 | /* Get the index bitmap attribute inode. */ |
| 984 | bvi = ntfs_attr_iget(vi, AT_BITMAP, I30, 4); |
| 985 | if (IS_ERR(bvi)) { |
| 986 | ntfs_error(vi->i_sb, "Failed to get bitmap attribute."); |
| 987 | err = PTR_ERR(bvi); |
| 988 | goto unm_err_out; |
| 989 | } |
| 990 | ni->itype.index.bmp_ino = bvi; |
| 991 | bni = NTFS_I(bvi); |
| 992 | if (NInoCompressed(bni) || NInoEncrypted(bni) || |
| 993 | NInoSparse(bni)) { |
| 994 | ntfs_error(vi->i_sb, "$BITMAP attribute is compressed " |
| 995 | "and/or encrypted and/or sparse."); |
| 996 | goto unm_err_out; |
| 997 | } |
| 998 | /* Consistency check bitmap size vs. index allocation size. */ |
Anton Altaparmakov | f50f3ac | 2004-11-19 22:16:00 +0000 | [diff] [blame] | 999 | bvi_size = i_size_read(bvi); |
| 1000 | if ((bvi_size << 3) < (vi->i_size >> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1001 | ni->itype.index.block_size_bits)) { |
| 1002 | ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) " |
| 1003 | "for index allocation (0x%llx).", |
Anton Altaparmakov | f50f3ac | 2004-11-19 22:16:00 +0000 | [diff] [blame] | 1004 | bvi_size << 3, vi->i_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1005 | goto unm_err_out; |
| 1006 | } |
| 1007 | skip_large_dir_stuff: |
| 1008 | /* Setup the operations for this inode. */ |
| 1009 | vi->i_op = &ntfs_dir_inode_ops; |
| 1010 | vi->i_fop = &ntfs_dir_ops; |
| 1011 | } else { |
| 1012 | /* It is a file. */ |
| 1013 | ntfs_attr_reinit_search_ctx(ctx); |
| 1014 | |
| 1015 | /* Setup the data attribute, even if not present. */ |
| 1016 | ni->type = AT_DATA; |
| 1017 | ni->name = NULL; |
| 1018 | ni->name_len = 0; |
| 1019 | |
| 1020 | /* Find first extent of the unnamed data attribute. */ |
| 1021 | err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx); |
| 1022 | if (unlikely(err)) { |
| 1023 | vi->i_size = ni->initialized_size = |
| 1024 | ni->allocated_size = 0; |
| 1025 | if (err != -ENOENT) { |
| 1026 | ntfs_error(vi->i_sb, "Failed to lookup $DATA " |
| 1027 | "attribute."); |
| 1028 | goto unm_err_out; |
| 1029 | } |
| 1030 | /* |
| 1031 | * FILE_Secure does not have an unnamed $DATA |
| 1032 | * attribute, so we special case it here. |
| 1033 | */ |
| 1034 | if (vi->i_ino == FILE_Secure) |
| 1035 | goto no_data_attr_special_case; |
| 1036 | /* |
| 1037 | * Most if not all the system files in the $Extend |
| 1038 | * system directory do not have unnamed data |
| 1039 | * attributes so we need to check if the parent |
| 1040 | * directory of the file is FILE_Extend and if it is |
| 1041 | * ignore this error. To do this we need to get the |
| 1042 | * name of this inode from the mft record as the name |
| 1043 | * contains the back reference to the parent directory. |
| 1044 | */ |
| 1045 | if (ntfs_is_extended_system_file(ctx) > 0) |
| 1046 | goto no_data_attr_special_case; |
| 1047 | // FIXME: File is corrupt! Hot-fix with empty data |
| 1048 | // attribute if recovery option is set. |
| 1049 | ntfs_error(vi->i_sb, "$DATA attribute is missing."); |
| 1050 | goto unm_err_out; |
| 1051 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1052 | a = ctx->attr; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1053 | /* Setup the state. */ |
Anton Altaparmakov | 67bb103 | 2005-09-08 21:19:45 +0100 | [diff] [blame] | 1054 | if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) { |
| 1055 | if (a->flags & ATTR_COMPRESSION_MASK) { |
| 1056 | NInoSetCompressed(ni); |
| 1057 | if (vol->cluster_size > 4096) { |
| 1058 | ntfs_error(vi->i_sb, "Found " |
Anton Altaparmakov | 9451f85 | 2005-03-03 14:43:43 +0000 | [diff] [blame] | 1059 | "compressed data but " |
| 1060 | "compression is " |
| 1061 | "disabled due to " |
| 1062 | "cluster size (%i) > " |
| 1063 | "4kiB.", |
| 1064 | vol->cluster_size); |
Anton Altaparmakov | 67bb103 | 2005-09-08 21:19:45 +0100 | [diff] [blame] | 1065 | goto unm_err_out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1066 | } |
Anton Altaparmakov | 67bb103 | 2005-09-08 21:19:45 +0100 | [diff] [blame] | 1067 | if ((a->flags & ATTR_COMPRESSION_MASK) |
| 1068 | != ATTR_IS_COMPRESSED) { |
| 1069 | ntfs_error(vi->i_sb, "Found unknown " |
| 1070 | "compression method " |
| 1071 | "or corrupt file."); |
| 1072 | goto unm_err_out; |
| 1073 | } |
| 1074 | } |
| 1075 | if (a->flags & ATTR_IS_SPARSE) |
| 1076 | NInoSetSparse(ni); |
| 1077 | } |
| 1078 | if (a->flags & ATTR_IS_ENCRYPTED) { |
| 1079 | if (NInoCompressed(ni)) { |
| 1080 | ntfs_error(vi->i_sb, "Found encrypted and " |
| 1081 | "compressed data."); |
| 1082 | goto unm_err_out; |
| 1083 | } |
| 1084 | NInoSetEncrypted(ni); |
| 1085 | } |
| 1086 | if (a->non_resident) { |
| 1087 | NInoSetNonResident(ni); |
| 1088 | if (NInoCompressed(ni) || NInoSparse(ni)) { |
Anton Altaparmakov | a0646a1 | 2006-03-23 15:53:03 +0000 | [diff] [blame] | 1089 | if (NInoCompressed(ni) && a->data.non_resident. |
| 1090 | compression_unit != 4) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1091 | ntfs_error(vi->i_sb, "Found " |
Anton Altaparmakov | a0646a1 | 2006-03-23 15:53:03 +0000 | [diff] [blame] | 1092 | "non-standard " |
Anton Altaparmakov | 67bb103 | 2005-09-08 21:19:45 +0100 | [diff] [blame] | 1093 | "compression unit (%u " |
| 1094 | "instead of 4). " |
| 1095 | "Cannot handle this.", |
| 1096 | a->data.non_resident. |
| 1097 | compression_unit); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1098 | err = -EOPNOTSUPP; |
| 1099 | goto unm_err_out; |
| 1100 | } |
Anton Altaparmakov | a0646a1 | 2006-03-23 15:53:03 +0000 | [diff] [blame] | 1101 | if (a->data.non_resident.compression_unit) { |
| 1102 | ni->itype.compressed.block_size = 1U << |
| 1103 | (a->data.non_resident. |
| 1104 | compression_unit + |
| 1105 | vol->cluster_size_bits); |
| 1106 | ni->itype.compressed.block_size_bits = |
| 1107 | ffs(ni->itype. |
| 1108 | compressed. |
| 1109 | block_size) - 1; |
| 1110 | ni->itype.compressed.block_clusters = |
| 1111 | 1U << a->data. |
| 1112 | non_resident. |
| 1113 | compression_unit; |
| 1114 | } else { |
| 1115 | ni->itype.compressed.block_size = 0; |
| 1116 | ni->itype.compressed.block_size_bits = |
| 1117 | 0; |
| 1118 | ni->itype.compressed.block_clusters = |
| 1119 | 0; |
| 1120 | } |
Anton Altaparmakov | 9451f85 | 2005-03-03 14:43:43 +0000 | [diff] [blame] | 1121 | ni->itype.compressed.size = sle64_to_cpu( |
| 1122 | a->data.non_resident. |
| 1123 | compressed_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1124 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1125 | if (a->data.non_resident.lowest_vcn) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1126 | ntfs_error(vi->i_sb, "First extent of $DATA " |
| 1127 | "attribute has non zero " |
| 1128 | "lowest_vcn."); |
| 1129 | goto unm_err_out; |
| 1130 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1131 | vi->i_size = sle64_to_cpu( |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1132 | a->data.non_resident.data_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1133 | ni->initialized_size = sle64_to_cpu( |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1134 | a->data.non_resident.initialized_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1135 | ni->allocated_size = sle64_to_cpu( |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1136 | a->data.non_resident.allocated_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1137 | } else { /* Resident attribute. */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1138 | vi->i_size = ni->initialized_size = le32_to_cpu( |
| 1139 | a->data.resident.value_length); |
| 1140 | ni->allocated_size = le32_to_cpu(a->length) - |
| 1141 | le16_to_cpu( |
| 1142 | a->data.resident.value_offset); |
| 1143 | if (vi->i_size > ni->allocated_size) { |
| 1144 | ntfs_error(vi->i_sb, "Resident data attribute " |
| 1145 | "is corrupt (size exceeds " |
| 1146 | "allocation)."); |
| 1147 | goto unm_err_out; |
| 1148 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1149 | } |
| 1150 | no_data_attr_special_case: |
| 1151 | /* We are done with the mft record, so we release it. */ |
| 1152 | ntfs_attr_put_search_ctx(ctx); |
| 1153 | unmap_mft_record(ni); |
| 1154 | m = NULL; |
| 1155 | ctx = NULL; |
| 1156 | /* Setup the operations for this inode. */ |
| 1157 | vi->i_op = &ntfs_file_inode_ops; |
| 1158 | vi->i_fop = &ntfs_file_ops; |
| 1159 | } |
| 1160 | if (NInoMstProtected(ni)) |
| 1161 | vi->i_mapping->a_ops = &ntfs_mst_aops; |
| 1162 | else |
| 1163 | vi->i_mapping->a_ops = &ntfs_aops; |
| 1164 | /* |
| 1165 | * The number of 512-byte blocks used on disk (for stat). This is in so |
| 1166 | * far inaccurate as it doesn't account for any named streams or other |
| 1167 | * special non-resident attributes, but that is how Windows works, too, |
| 1168 | * so we are at least consistent with Windows, if not entirely |
| 1169 | * consistent with the Linux Way. Doing it the Linux Way would cause a |
| 1170 | * significant slowdown as it would involve iterating over all |
| 1171 | * attributes in the mft record and adding the allocated/compressed |
| 1172 | * sizes of all non-resident attributes present to give us the Linux |
| 1173 | * correct size that should go into i_blocks (after division by 512). |
| 1174 | */ |
Anton Altaparmakov | 9451f85 | 2005-03-03 14:43:43 +0000 | [diff] [blame] | 1175 | if (S_ISREG(vi->i_mode) && (NInoCompressed(ni) || NInoSparse(ni))) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1176 | vi->i_blocks = ni->itype.compressed.size >> 9; |
Anton Altaparmakov | 9451f85 | 2005-03-03 14:43:43 +0000 | [diff] [blame] | 1177 | else |
| 1178 | vi->i_blocks = ni->allocated_size >> 9; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1179 | ntfs_debug("Done."); |
| 1180 | return 0; |
| 1181 | |
| 1182 | unm_err_out: |
| 1183 | if (!err) |
| 1184 | err = -EIO; |
| 1185 | if (ctx) |
| 1186 | ntfs_attr_put_search_ctx(ctx); |
| 1187 | if (m) |
| 1188 | unmap_mft_record(ni); |
| 1189 | err_out: |
| 1190 | ntfs_error(vol->sb, "Failed with error code %i. Marking corrupt " |
| 1191 | "inode 0x%lx as bad. Run chkdsk.", err, vi->i_ino); |
| 1192 | make_bad_inode(vi); |
| 1193 | if (err != -EOPNOTSUPP && err != -ENOMEM) |
| 1194 | NVolSetErrors(vol); |
| 1195 | return err; |
| 1196 | } |
| 1197 | |
| 1198 | /** |
| 1199 | * ntfs_read_locked_attr_inode - read an attribute inode from its base inode |
| 1200 | * @base_vi: base inode |
| 1201 | * @vi: attribute inode to read |
| 1202 | * |
| 1203 | * ntfs_read_locked_attr_inode() is called from ntfs_attr_iget() to read the |
| 1204 | * attribute inode described by @vi into memory from the base mft record |
| 1205 | * described by @base_ni. |
| 1206 | * |
| 1207 | * ntfs_read_locked_attr_inode() maps, pins and locks the base inode for |
| 1208 | * reading and looks up the attribute described by @vi before setting up the |
| 1209 | * necessary fields in @vi as well as initializing the ntfs inode. |
| 1210 | * |
| 1211 | * Q: What locks are held when the function is called? |
| 1212 | * A: i_state has I_LOCK set, hence the inode is locked, also |
| 1213 | * i_count is set to 1, so it is not going to go away |
| 1214 | * |
| 1215 | * Return 0 on success and -errno on error. In the error case, the inode will |
| 1216 | * have had make_bad_inode() executed on it. |
Anton Altaparmakov | f6098cf | 2005-09-19 09:41:39 +0100 | [diff] [blame] | 1217 | * |
| 1218 | * Note this cannot be called for AT_INDEX_ALLOCATION. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1219 | */ |
| 1220 | static int ntfs_read_locked_attr_inode(struct inode *base_vi, struct inode *vi) |
| 1221 | { |
| 1222 | ntfs_volume *vol = NTFS_SB(vi->i_sb); |
| 1223 | ntfs_inode *ni, *base_ni; |
| 1224 | MFT_RECORD *m; |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1225 | ATTR_RECORD *a; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1226 | ntfs_attr_search_ctx *ctx; |
| 1227 | int err = 0; |
| 1228 | |
| 1229 | ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino); |
| 1230 | |
| 1231 | ntfs_init_big_inode(vi); |
| 1232 | |
| 1233 | ni = NTFS_I(vi); |
| 1234 | base_ni = NTFS_I(base_vi); |
| 1235 | |
| 1236 | /* Just mirror the values from the base inode. */ |
| 1237 | vi->i_blksize = base_vi->i_blksize; |
| 1238 | vi->i_version = base_vi->i_version; |
| 1239 | vi->i_uid = base_vi->i_uid; |
| 1240 | vi->i_gid = base_vi->i_gid; |
| 1241 | vi->i_nlink = base_vi->i_nlink; |
| 1242 | vi->i_mtime = base_vi->i_mtime; |
| 1243 | vi->i_ctime = base_vi->i_ctime; |
| 1244 | vi->i_atime = base_vi->i_atime; |
| 1245 | vi->i_generation = ni->seq_no = base_ni->seq_no; |
| 1246 | |
| 1247 | /* Set inode type to zero but preserve permissions. */ |
| 1248 | vi->i_mode = base_vi->i_mode & ~S_IFMT; |
| 1249 | |
| 1250 | m = map_mft_record(base_ni); |
| 1251 | if (IS_ERR(m)) { |
| 1252 | err = PTR_ERR(m); |
| 1253 | goto err_out; |
| 1254 | } |
| 1255 | ctx = ntfs_attr_get_search_ctx(base_ni, m); |
| 1256 | if (!ctx) { |
| 1257 | err = -ENOMEM; |
| 1258 | goto unm_err_out; |
| 1259 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1260 | /* Find the attribute. */ |
| 1261 | err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len, |
| 1262 | CASE_SENSITIVE, 0, NULL, 0, ctx); |
| 1263 | if (unlikely(err)) |
| 1264 | goto unm_err_out; |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1265 | a = ctx->attr; |
Anton Altaparmakov | 67bb103 | 2005-09-08 21:19:45 +0100 | [diff] [blame] | 1266 | if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_SPARSE)) { |
| 1267 | if (a->flags & ATTR_COMPRESSION_MASK) { |
| 1268 | NInoSetCompressed(ni); |
| 1269 | if ((ni->type != AT_DATA) || (ni->type == AT_DATA && |
| 1270 | ni->name_len)) { |
| 1271 | ntfs_error(vi->i_sb, "Found compressed " |
| 1272 | "non-data or named data " |
| 1273 | "attribute. Please report " |
| 1274 | "you saw this message to " |
| 1275 | "linux-ntfs-dev@lists." |
| 1276 | "sourceforge.net"); |
| 1277 | goto unm_err_out; |
| 1278 | } |
| 1279 | if (vol->cluster_size > 4096) { |
| 1280 | ntfs_error(vi->i_sb, "Found compressed " |
| 1281 | "attribute but compression is " |
| 1282 | "disabled due to cluster size " |
| 1283 | "(%i) > 4kiB.", |
| 1284 | vol->cluster_size); |
| 1285 | goto unm_err_out; |
| 1286 | } |
| 1287 | if ((a->flags & ATTR_COMPRESSION_MASK) != |
| 1288 | ATTR_IS_COMPRESSED) { |
| 1289 | ntfs_error(vi->i_sb, "Found unknown " |
| 1290 | "compression method."); |
| 1291 | goto unm_err_out; |
| 1292 | } |
| 1293 | } |
| 1294 | /* |
Anton Altaparmakov | f6098cf | 2005-09-19 09:41:39 +0100 | [diff] [blame] | 1295 | * The compressed/sparse flag set in an index root just means |
| 1296 | * to compress all files. |
Anton Altaparmakov | 67bb103 | 2005-09-08 21:19:45 +0100 | [diff] [blame] | 1297 | */ |
| 1298 | if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) { |
| 1299 | ntfs_error(vi->i_sb, "Found mst protected attribute " |
| 1300 | "but the attribute is %s. Please " |
| 1301 | "report you saw this message to " |
| 1302 | "linux-ntfs-dev@lists.sourceforge.net", |
| 1303 | NInoCompressed(ni) ? "compressed" : |
| 1304 | "sparse"); |
| 1305 | goto unm_err_out; |
| 1306 | } |
| 1307 | if (a->flags & ATTR_IS_SPARSE) |
| 1308 | NInoSetSparse(ni); |
| 1309 | } |
| 1310 | if (a->flags & ATTR_IS_ENCRYPTED) { |
| 1311 | if (NInoCompressed(ni)) { |
| 1312 | ntfs_error(vi->i_sb, "Found encrypted and compressed " |
| 1313 | "data."); |
| 1314 | goto unm_err_out; |
| 1315 | } |
| 1316 | /* |
| 1317 | * The encryption flag set in an index root just means to |
| 1318 | * encrypt all files. |
| 1319 | */ |
| 1320 | if (NInoMstProtected(ni) && ni->type != AT_INDEX_ROOT) { |
| 1321 | ntfs_error(vi->i_sb, "Found mst protected attribute " |
| 1322 | "but the attribute is encrypted. " |
| 1323 | "Please report you saw this message " |
| 1324 | "to linux-ntfs-dev@lists.sourceforge." |
| 1325 | "net"); |
| 1326 | goto unm_err_out; |
| 1327 | } |
| 1328 | if (ni->type != AT_DATA) { |
| 1329 | ntfs_error(vi->i_sb, "Found encrypted non-data " |
| 1330 | "attribute."); |
| 1331 | goto unm_err_out; |
| 1332 | } |
| 1333 | NInoSetEncrypted(ni); |
| 1334 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1335 | if (!a->non_resident) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1336 | /* Ensure the attribute name is placed before the value. */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1337 | if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= |
| 1338 | le16_to_cpu(a->data.resident.value_offset)))) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1339 | ntfs_error(vol->sb, "Attribute name is placed after " |
| 1340 | "the attribute value."); |
| 1341 | goto unm_err_out; |
| 1342 | } |
Anton Altaparmakov | 67bb103 | 2005-09-08 21:19:45 +0100 | [diff] [blame] | 1343 | if (NInoMstProtected(ni)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1344 | ntfs_error(vi->i_sb, "Found mst protected attribute " |
Anton Altaparmakov | 67bb103 | 2005-09-08 21:19:45 +0100 | [diff] [blame] | 1345 | "but the attribute is resident. " |
| 1346 | "Please report you saw this message to " |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1347 | "linux-ntfs-dev@lists.sourceforge.net"); |
| 1348 | goto unm_err_out; |
| 1349 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1350 | vi->i_size = ni->initialized_size = le32_to_cpu( |
| 1351 | a->data.resident.value_length); |
| 1352 | ni->allocated_size = le32_to_cpu(a->length) - |
| 1353 | le16_to_cpu(a->data.resident.value_offset); |
| 1354 | if (vi->i_size > ni->allocated_size) { |
Anton Altaparmakov | 9451f85 | 2005-03-03 14:43:43 +0000 | [diff] [blame] | 1355 | ntfs_error(vi->i_sb, "Resident attribute is corrupt " |
| 1356 | "(size exceeds allocation)."); |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1357 | goto unm_err_out; |
| 1358 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1359 | } else { |
| 1360 | NInoSetNonResident(ni); |
| 1361 | /* |
| 1362 | * Ensure the attribute name is placed before the mapping pairs |
| 1363 | * array. |
| 1364 | */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1365 | if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= |
| 1366 | le16_to_cpu( |
| 1367 | a->data.non_resident.mapping_pairs_offset)))) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1368 | ntfs_error(vol->sb, "Attribute name is placed after " |
| 1369 | "the mapping pairs array."); |
| 1370 | goto unm_err_out; |
| 1371 | } |
Anton Altaparmakov | f6098cf | 2005-09-19 09:41:39 +0100 | [diff] [blame] | 1372 | if (NInoCompressed(ni) || NInoSparse(ni)) { |
Anton Altaparmakov | a0646a1 | 2006-03-23 15:53:03 +0000 | [diff] [blame] | 1373 | if (NInoCompressed(ni) && a->data.non_resident. |
| 1374 | compression_unit != 4) { |
| 1375 | ntfs_error(vi->i_sb, "Found non-standard " |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1376 | "compression unit (%u instead " |
| 1377 | "of 4). Cannot handle this.", |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1378 | a->data.non_resident. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1379 | compression_unit); |
| 1380 | err = -EOPNOTSUPP; |
| 1381 | goto unm_err_out; |
| 1382 | } |
Anton Altaparmakov | a0646a1 | 2006-03-23 15:53:03 +0000 | [diff] [blame] | 1383 | if (a->data.non_resident.compression_unit) { |
| 1384 | ni->itype.compressed.block_size = 1U << |
| 1385 | (a->data.non_resident. |
| 1386 | compression_unit + |
| 1387 | vol->cluster_size_bits); |
| 1388 | ni->itype.compressed.block_size_bits = |
| 1389 | ffs(ni->itype.compressed. |
| 1390 | block_size) - 1; |
| 1391 | ni->itype.compressed.block_clusters = 1U << |
| 1392 | a->data.non_resident. |
| 1393 | compression_unit; |
| 1394 | } else { |
| 1395 | ni->itype.compressed.block_size = 0; |
| 1396 | ni->itype.compressed.block_size_bits = 0; |
| 1397 | ni->itype.compressed.block_clusters = 0; |
| 1398 | } |
Anton Altaparmakov | 9451f85 | 2005-03-03 14:43:43 +0000 | [diff] [blame] | 1399 | ni->itype.compressed.size = sle64_to_cpu( |
| 1400 | a->data.non_resident.compressed_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1401 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1402 | if (a->data.non_resident.lowest_vcn) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1403 | ntfs_error(vi->i_sb, "First extent of attribute has " |
| 1404 | "non-zero lowest_vcn."); |
| 1405 | goto unm_err_out; |
| 1406 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1407 | vi->i_size = sle64_to_cpu(a->data.non_resident.data_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1408 | ni->initialized_size = sle64_to_cpu( |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1409 | a->data.non_resident.initialized_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1410 | ni->allocated_size = sle64_to_cpu( |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1411 | a->data.non_resident.allocated_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1412 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1413 | /* Setup the operations for this attribute inode. */ |
| 1414 | vi->i_op = NULL; |
| 1415 | vi->i_fop = NULL; |
| 1416 | if (NInoMstProtected(ni)) |
| 1417 | vi->i_mapping->a_ops = &ntfs_mst_aops; |
| 1418 | else |
| 1419 | vi->i_mapping->a_ops = &ntfs_aops; |
Anton Altaparmakov | 67bb103 | 2005-09-08 21:19:45 +0100 | [diff] [blame] | 1420 | if ((NInoCompressed(ni) || NInoSparse(ni)) && ni->type != AT_INDEX_ROOT) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1421 | vi->i_blocks = ni->itype.compressed.size >> 9; |
Anton Altaparmakov | 9451f85 | 2005-03-03 14:43:43 +0000 | [diff] [blame] | 1422 | else |
| 1423 | vi->i_blocks = ni->allocated_size >> 9; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1424 | /* |
Anton Altaparmakov | 67bb103 | 2005-09-08 21:19:45 +0100 | [diff] [blame] | 1425 | * Make sure the base inode does not go away and attach it to the |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1426 | * attribute inode. |
| 1427 | */ |
| 1428 | igrab(base_vi); |
| 1429 | ni->ext.base_ntfs_ino = base_ni; |
| 1430 | ni->nr_extents = -1; |
| 1431 | |
| 1432 | ntfs_attr_put_search_ctx(ctx); |
| 1433 | unmap_mft_record(base_ni); |
| 1434 | |
| 1435 | ntfs_debug("Done."); |
| 1436 | return 0; |
| 1437 | |
| 1438 | unm_err_out: |
| 1439 | if (!err) |
| 1440 | err = -EIO; |
| 1441 | if (ctx) |
| 1442 | ntfs_attr_put_search_ctx(ctx); |
| 1443 | unmap_mft_record(base_ni); |
| 1444 | err_out: |
| 1445 | ntfs_error(vol->sb, "Failed with error code %i while reading attribute " |
| 1446 | "inode (mft_no 0x%lx, type 0x%x, name_len %i). " |
| 1447 | "Marking corrupt inode and base inode 0x%lx as bad. " |
| 1448 | "Run chkdsk.", err, vi->i_ino, ni->type, ni->name_len, |
| 1449 | base_vi->i_ino); |
| 1450 | make_bad_inode(vi); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1451 | if (err != -ENOMEM) |
| 1452 | NVolSetErrors(vol); |
| 1453 | return err; |
| 1454 | } |
| 1455 | |
| 1456 | /** |
| 1457 | * ntfs_read_locked_index_inode - read an index inode from its base inode |
| 1458 | * @base_vi: base inode |
| 1459 | * @vi: index inode to read |
| 1460 | * |
| 1461 | * ntfs_read_locked_index_inode() is called from ntfs_index_iget() to read the |
| 1462 | * index inode described by @vi into memory from the base mft record described |
| 1463 | * by @base_ni. |
| 1464 | * |
| 1465 | * ntfs_read_locked_index_inode() maps, pins and locks the base inode for |
| 1466 | * reading and looks up the attributes relating to the index described by @vi |
| 1467 | * before setting up the necessary fields in @vi as well as initializing the |
| 1468 | * ntfs inode. |
| 1469 | * |
| 1470 | * Note, index inodes are essentially attribute inodes (NInoAttr() is true) |
| 1471 | * with the attribute type set to AT_INDEX_ALLOCATION. Apart from that, they |
| 1472 | * are setup like directory inodes since directories are a special case of |
| 1473 | * indices ao they need to be treated in much the same way. Most importantly, |
| 1474 | * for small indices the index allocation attribute might not actually exist. |
| 1475 | * However, the index root attribute always exists but this does not need to |
| 1476 | * have an inode associated with it and this is why we define a new inode type |
| 1477 | * index. Also, like for directories, we need to have an attribute inode for |
| 1478 | * the bitmap attribute corresponding to the index allocation attribute and we |
| 1479 | * can store this in the appropriate field of the inode, just like we do for |
| 1480 | * normal directory inodes. |
| 1481 | * |
| 1482 | * Q: What locks are held when the function is called? |
| 1483 | * A: i_state has I_LOCK set, hence the inode is locked, also |
| 1484 | * i_count is set to 1, so it is not going to go away |
| 1485 | * |
| 1486 | * Return 0 on success and -errno on error. In the error case, the inode will |
| 1487 | * have had make_bad_inode() executed on it. |
| 1488 | */ |
| 1489 | static int ntfs_read_locked_index_inode(struct inode *base_vi, struct inode *vi) |
| 1490 | { |
Anton Altaparmakov | f50f3ac | 2004-11-19 22:16:00 +0000 | [diff] [blame] | 1491 | loff_t bvi_size; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1492 | ntfs_volume *vol = NTFS_SB(vi->i_sb); |
| 1493 | ntfs_inode *ni, *base_ni, *bni; |
| 1494 | struct inode *bvi; |
| 1495 | MFT_RECORD *m; |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1496 | ATTR_RECORD *a; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1497 | ntfs_attr_search_ctx *ctx; |
| 1498 | INDEX_ROOT *ir; |
| 1499 | u8 *ir_end, *index_end; |
| 1500 | int err = 0; |
| 1501 | |
| 1502 | ntfs_debug("Entering for i_ino 0x%lx.", vi->i_ino); |
| 1503 | ntfs_init_big_inode(vi); |
| 1504 | ni = NTFS_I(vi); |
| 1505 | base_ni = NTFS_I(base_vi); |
| 1506 | /* Just mirror the values from the base inode. */ |
| 1507 | vi->i_blksize = base_vi->i_blksize; |
| 1508 | vi->i_version = base_vi->i_version; |
| 1509 | vi->i_uid = base_vi->i_uid; |
| 1510 | vi->i_gid = base_vi->i_gid; |
| 1511 | vi->i_nlink = base_vi->i_nlink; |
| 1512 | vi->i_mtime = base_vi->i_mtime; |
| 1513 | vi->i_ctime = base_vi->i_ctime; |
| 1514 | vi->i_atime = base_vi->i_atime; |
| 1515 | vi->i_generation = ni->seq_no = base_ni->seq_no; |
| 1516 | /* Set inode type to zero but preserve permissions. */ |
| 1517 | vi->i_mode = base_vi->i_mode & ~S_IFMT; |
| 1518 | /* Map the mft record for the base inode. */ |
| 1519 | m = map_mft_record(base_ni); |
| 1520 | if (IS_ERR(m)) { |
| 1521 | err = PTR_ERR(m); |
| 1522 | goto err_out; |
| 1523 | } |
| 1524 | ctx = ntfs_attr_get_search_ctx(base_ni, m); |
| 1525 | if (!ctx) { |
| 1526 | err = -ENOMEM; |
| 1527 | goto unm_err_out; |
| 1528 | } |
| 1529 | /* Find the index root attribute. */ |
| 1530 | err = ntfs_attr_lookup(AT_INDEX_ROOT, ni->name, ni->name_len, |
| 1531 | CASE_SENSITIVE, 0, NULL, 0, ctx); |
| 1532 | if (unlikely(err)) { |
| 1533 | if (err == -ENOENT) |
| 1534 | ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is " |
| 1535 | "missing."); |
| 1536 | goto unm_err_out; |
| 1537 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1538 | a = ctx->attr; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1539 | /* Set up the state. */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1540 | if (unlikely(a->non_resident)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1541 | ntfs_error(vol->sb, "$INDEX_ROOT attribute is not resident."); |
| 1542 | goto unm_err_out; |
| 1543 | } |
| 1544 | /* Ensure the attribute name is placed before the value. */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1545 | if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= |
| 1546 | le16_to_cpu(a->data.resident.value_offset)))) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1547 | ntfs_error(vol->sb, "$INDEX_ROOT attribute name is placed " |
| 1548 | "after the attribute value."); |
| 1549 | goto unm_err_out; |
| 1550 | } |
Anton Altaparmakov | 67bb103 | 2005-09-08 21:19:45 +0100 | [diff] [blame] | 1551 | /* |
| 1552 | * Compressed/encrypted/sparse index root is not allowed, except for |
| 1553 | * directories of course but those are not dealt with here. |
| 1554 | */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1555 | if (a->flags & (ATTR_COMPRESSION_MASK | ATTR_IS_ENCRYPTED | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1556 | ATTR_IS_SPARSE)) { |
| 1557 | ntfs_error(vi->i_sb, "Found compressed/encrypted/sparse index " |
| 1558 | "root attribute."); |
| 1559 | goto unm_err_out; |
| 1560 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1561 | ir = (INDEX_ROOT*)((u8*)a + le16_to_cpu(a->data.resident.value_offset)); |
| 1562 | ir_end = (u8*)ir + le32_to_cpu(a->data.resident.value_length); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1563 | if (ir_end > (u8*)ctx->mrec + vol->mft_record_size) { |
| 1564 | ntfs_error(vi->i_sb, "$INDEX_ROOT attribute is corrupt."); |
| 1565 | goto unm_err_out; |
| 1566 | } |
| 1567 | index_end = (u8*)&ir->index + le32_to_cpu(ir->index.index_length); |
| 1568 | if (index_end > ir_end) { |
| 1569 | ntfs_error(vi->i_sb, "Index is corrupt."); |
| 1570 | goto unm_err_out; |
| 1571 | } |
| 1572 | if (ir->type) { |
| 1573 | ntfs_error(vi->i_sb, "Index type is not 0 (type is 0x%x).", |
| 1574 | le32_to_cpu(ir->type)); |
| 1575 | goto unm_err_out; |
| 1576 | } |
| 1577 | ni->itype.index.collation_rule = ir->collation_rule; |
| 1578 | ntfs_debug("Index collation rule is 0x%x.", |
| 1579 | le32_to_cpu(ir->collation_rule)); |
| 1580 | ni->itype.index.block_size = le32_to_cpu(ir->index_block_size); |
| 1581 | if (ni->itype.index.block_size & (ni->itype.index.block_size - 1)) { |
| 1582 | ntfs_error(vi->i_sb, "Index block size (%u) is not a power of " |
| 1583 | "two.", ni->itype.index.block_size); |
| 1584 | goto unm_err_out; |
| 1585 | } |
| 1586 | if (ni->itype.index.block_size > PAGE_CACHE_SIZE) { |
| 1587 | ntfs_error(vi->i_sb, "Index block size (%u) > PAGE_CACHE_SIZE " |
| 1588 | "(%ld) is not supported. Sorry.", |
| 1589 | ni->itype.index.block_size, PAGE_CACHE_SIZE); |
| 1590 | err = -EOPNOTSUPP; |
| 1591 | goto unm_err_out; |
| 1592 | } |
| 1593 | if (ni->itype.index.block_size < NTFS_BLOCK_SIZE) { |
| 1594 | ntfs_error(vi->i_sb, "Index block size (%u) < NTFS_BLOCK_SIZE " |
| 1595 | "(%i) is not supported. Sorry.", |
| 1596 | ni->itype.index.block_size, NTFS_BLOCK_SIZE); |
| 1597 | err = -EOPNOTSUPP; |
| 1598 | goto unm_err_out; |
| 1599 | } |
| 1600 | ni->itype.index.block_size_bits = ffs(ni->itype.index.block_size) - 1; |
| 1601 | /* Determine the size of a vcn in the index. */ |
| 1602 | if (vol->cluster_size <= ni->itype.index.block_size) { |
| 1603 | ni->itype.index.vcn_size = vol->cluster_size; |
| 1604 | ni->itype.index.vcn_size_bits = vol->cluster_size_bits; |
| 1605 | } else { |
| 1606 | ni->itype.index.vcn_size = vol->sector_size; |
| 1607 | ni->itype.index.vcn_size_bits = vol->sector_size_bits; |
| 1608 | } |
| 1609 | /* Check for presence of index allocation attribute. */ |
| 1610 | if (!(ir->index.flags & LARGE_INDEX)) { |
| 1611 | /* No index allocation. */ |
| 1612 | vi->i_size = ni->initialized_size = ni->allocated_size = 0; |
| 1613 | /* We are done with the mft record, so we release it. */ |
| 1614 | ntfs_attr_put_search_ctx(ctx); |
| 1615 | unmap_mft_record(base_ni); |
| 1616 | m = NULL; |
| 1617 | ctx = NULL; |
| 1618 | goto skip_large_index_stuff; |
| 1619 | } /* LARGE_INDEX: Index allocation present. Setup state. */ |
| 1620 | NInoSetIndexAllocPresent(ni); |
| 1621 | /* Find index allocation attribute. */ |
| 1622 | ntfs_attr_reinit_search_ctx(ctx); |
| 1623 | err = ntfs_attr_lookup(AT_INDEX_ALLOCATION, ni->name, ni->name_len, |
| 1624 | CASE_SENSITIVE, 0, NULL, 0, ctx); |
| 1625 | if (unlikely(err)) { |
| 1626 | if (err == -ENOENT) |
| 1627 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is " |
| 1628 | "not present but $INDEX_ROOT " |
| 1629 | "indicated it is."); |
| 1630 | else |
| 1631 | ntfs_error(vi->i_sb, "Failed to lookup " |
| 1632 | "$INDEX_ALLOCATION attribute."); |
| 1633 | goto unm_err_out; |
| 1634 | } |
Anton Altaparmakov | a778f21 | 2006-03-23 16:18:23 +0000 | [diff] [blame] | 1635 | a = ctx->attr; |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1636 | if (!a->non_resident) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1637 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is " |
| 1638 | "resident."); |
| 1639 | goto unm_err_out; |
| 1640 | } |
| 1641 | /* |
| 1642 | * Ensure the attribute name is placed before the mapping pairs array. |
| 1643 | */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1644 | if (unlikely(a->name_length && (le16_to_cpu(a->name_offset) >= |
| 1645 | le16_to_cpu( |
| 1646 | a->data.non_resident.mapping_pairs_offset)))) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1647 | ntfs_error(vol->sb, "$INDEX_ALLOCATION attribute name is " |
| 1648 | "placed after the mapping pairs array."); |
| 1649 | goto unm_err_out; |
| 1650 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1651 | if (a->flags & ATTR_IS_ENCRYPTED) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1652 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is " |
| 1653 | "encrypted."); |
| 1654 | goto unm_err_out; |
| 1655 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1656 | if (a->flags & ATTR_IS_SPARSE) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1657 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is sparse."); |
| 1658 | goto unm_err_out; |
| 1659 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1660 | if (a->flags & ATTR_COMPRESSION_MASK) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1661 | ntfs_error(vi->i_sb, "$INDEX_ALLOCATION attribute is " |
| 1662 | "compressed."); |
| 1663 | goto unm_err_out; |
| 1664 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1665 | if (a->data.non_resident.lowest_vcn) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1666 | ntfs_error(vi->i_sb, "First extent of $INDEX_ALLOCATION " |
| 1667 | "attribute has non zero lowest_vcn."); |
| 1668 | goto unm_err_out; |
| 1669 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1670 | vi->i_size = sle64_to_cpu(a->data.non_resident.data_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1671 | ni->initialized_size = sle64_to_cpu( |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1672 | a->data.non_resident.initialized_size); |
| 1673 | ni->allocated_size = sle64_to_cpu(a->data.non_resident.allocated_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1674 | /* |
| 1675 | * We are done with the mft record, so we release it. Otherwise |
| 1676 | * we would deadlock in ntfs_attr_iget(). |
| 1677 | */ |
| 1678 | ntfs_attr_put_search_ctx(ctx); |
| 1679 | unmap_mft_record(base_ni); |
| 1680 | m = NULL; |
| 1681 | ctx = NULL; |
| 1682 | /* Get the index bitmap attribute inode. */ |
| 1683 | bvi = ntfs_attr_iget(base_vi, AT_BITMAP, ni->name, ni->name_len); |
| 1684 | if (IS_ERR(bvi)) { |
| 1685 | ntfs_error(vi->i_sb, "Failed to get bitmap attribute."); |
| 1686 | err = PTR_ERR(bvi); |
| 1687 | goto unm_err_out; |
| 1688 | } |
| 1689 | bni = NTFS_I(bvi); |
| 1690 | if (NInoCompressed(bni) || NInoEncrypted(bni) || |
| 1691 | NInoSparse(bni)) { |
| 1692 | ntfs_error(vi->i_sb, "$BITMAP attribute is compressed and/or " |
| 1693 | "encrypted and/or sparse."); |
| 1694 | goto iput_unm_err_out; |
| 1695 | } |
| 1696 | /* Consistency check bitmap size vs. index allocation size. */ |
Anton Altaparmakov | f50f3ac | 2004-11-19 22:16:00 +0000 | [diff] [blame] | 1697 | bvi_size = i_size_read(bvi); |
| 1698 | if ((bvi_size << 3) < (vi->i_size >> ni->itype.index.block_size_bits)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1699 | ntfs_error(vi->i_sb, "Index bitmap too small (0x%llx) for " |
Anton Altaparmakov | f50f3ac | 2004-11-19 22:16:00 +0000 | [diff] [blame] | 1700 | "index allocation (0x%llx).", bvi_size << 3, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1701 | vi->i_size); |
| 1702 | goto iput_unm_err_out; |
| 1703 | } |
| 1704 | ni->itype.index.bmp_ino = bvi; |
| 1705 | skip_large_index_stuff: |
| 1706 | /* Setup the operations for this index inode. */ |
| 1707 | vi->i_op = NULL; |
| 1708 | vi->i_fop = NULL; |
| 1709 | vi->i_mapping->a_ops = &ntfs_mst_aops; |
| 1710 | vi->i_blocks = ni->allocated_size >> 9; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1711 | /* |
| 1712 | * Make sure the base inode doesn't go away and attach it to the |
| 1713 | * index inode. |
| 1714 | */ |
| 1715 | igrab(base_vi); |
| 1716 | ni->ext.base_ntfs_ino = base_ni; |
| 1717 | ni->nr_extents = -1; |
| 1718 | |
| 1719 | ntfs_debug("Done."); |
| 1720 | return 0; |
| 1721 | |
| 1722 | iput_unm_err_out: |
| 1723 | iput(bvi); |
| 1724 | unm_err_out: |
| 1725 | if (!err) |
| 1726 | err = -EIO; |
| 1727 | if (ctx) |
| 1728 | ntfs_attr_put_search_ctx(ctx); |
| 1729 | if (m) |
| 1730 | unmap_mft_record(base_ni); |
| 1731 | err_out: |
| 1732 | ntfs_error(vi->i_sb, "Failed with error code %i while reading index " |
| 1733 | "inode (mft_no 0x%lx, name_len %i.", err, vi->i_ino, |
| 1734 | ni->name_len); |
| 1735 | make_bad_inode(vi); |
| 1736 | if (err != -EOPNOTSUPP && err != -ENOMEM) |
| 1737 | NVolSetErrors(vol); |
| 1738 | return err; |
| 1739 | } |
| 1740 | |
Ingo Molnar | 5934537 | 2006-07-03 00:25:18 -0700 | [diff] [blame^] | 1741 | /* |
| 1742 | * The MFT inode has special locking, so teach the lock validator |
| 1743 | * about this by splitting off the locking rules of the MFT from |
| 1744 | * the locking rules of other inodes. The MFT inode can never be |
| 1745 | * accessed from the VFS side (or even internally), only by the |
| 1746 | * map_mft functions. |
| 1747 | */ |
| 1748 | static struct lock_class_key mft_ni_runlist_lock_key, mft_ni_mrec_lock_key; |
| 1749 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1750 | /** |
| 1751 | * ntfs_read_inode_mount - special read_inode for mount time use only |
| 1752 | * @vi: inode to read |
| 1753 | * |
| 1754 | * Read inode FILE_MFT at mount time, only called with super_block lock |
| 1755 | * held from within the read_super() code path. |
| 1756 | * |
| 1757 | * This function exists because when it is called the page cache for $MFT/$DATA |
| 1758 | * is not initialized and hence we cannot get at the contents of mft records |
| 1759 | * by calling map_mft_record*(). |
| 1760 | * |
| 1761 | * Further it needs to cope with the circular references problem, i.e. cannot |
| 1762 | * load any attributes other than $ATTRIBUTE_LIST until $DATA is loaded, because |
| 1763 | * we do not know where the other extent mft records are yet and again, because |
| 1764 | * we cannot call map_mft_record*() yet. Obviously this applies only when an |
| 1765 | * attribute list is actually present in $MFT inode. |
| 1766 | * |
| 1767 | * We solve these problems by starting with the $DATA attribute before anything |
| 1768 | * else and iterating using ntfs_attr_lookup($DATA) over all extents. As each |
| 1769 | * extent is found, we ntfs_mapping_pairs_decompress() including the implied |
| 1770 | * ntfs_runlists_merge(). Each step of the iteration necessarily provides |
| 1771 | * sufficient information for the next step to complete. |
| 1772 | * |
| 1773 | * This should work but there are two possible pit falls (see inline comments |
| 1774 | * below), but only time will tell if they are real pits or just smoke... |
| 1775 | */ |
| 1776 | int ntfs_read_inode_mount(struct inode *vi) |
| 1777 | { |
| 1778 | VCN next_vcn, last_vcn, highest_vcn; |
| 1779 | s64 block; |
| 1780 | struct super_block *sb = vi->i_sb; |
| 1781 | ntfs_volume *vol = NTFS_SB(sb); |
| 1782 | struct buffer_head *bh; |
| 1783 | ntfs_inode *ni; |
| 1784 | MFT_RECORD *m = NULL; |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1785 | ATTR_RECORD *a; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1786 | ntfs_attr_search_ctx *ctx; |
| 1787 | unsigned int i, nr_blocks; |
| 1788 | int err; |
| 1789 | |
| 1790 | ntfs_debug("Entering."); |
| 1791 | |
| 1792 | /* Initialize the ntfs specific part of @vi. */ |
| 1793 | ntfs_init_big_inode(vi); |
| 1794 | |
| 1795 | ni = NTFS_I(vi); |
| 1796 | |
| 1797 | /* Setup the data attribute. It is special as it is mst protected. */ |
| 1798 | NInoSetNonResident(ni); |
| 1799 | NInoSetMstProtected(ni); |
Anton Altaparmakov | c002f42 | 2005-02-03 12:02:56 +0000 | [diff] [blame] | 1800 | NInoSetSparseDisabled(ni); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1801 | ni->type = AT_DATA; |
| 1802 | ni->name = NULL; |
| 1803 | ni->name_len = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1804 | /* |
| 1805 | * This sets up our little cheat allowing us to reuse the async read io |
| 1806 | * completion handler for directories. |
| 1807 | */ |
| 1808 | ni->itype.index.block_size = vol->mft_record_size; |
| 1809 | ni->itype.index.block_size_bits = vol->mft_record_size_bits; |
| 1810 | |
| 1811 | /* Very important! Needed to be able to call map_mft_record*(). */ |
| 1812 | vol->mft_ino = vi; |
| 1813 | |
| 1814 | /* Allocate enough memory to read the first mft record. */ |
| 1815 | if (vol->mft_record_size > 64 * 1024) { |
| 1816 | ntfs_error(sb, "Unsupported mft record size %i (max 64kiB).", |
| 1817 | vol->mft_record_size); |
| 1818 | goto err_out; |
| 1819 | } |
| 1820 | i = vol->mft_record_size; |
| 1821 | if (i < sb->s_blocksize) |
| 1822 | i = sb->s_blocksize; |
| 1823 | m = (MFT_RECORD*)ntfs_malloc_nofs(i); |
| 1824 | if (!m) { |
| 1825 | ntfs_error(sb, "Failed to allocate buffer for $MFT record 0."); |
| 1826 | goto err_out; |
| 1827 | } |
| 1828 | |
| 1829 | /* Determine the first block of the $MFT/$DATA attribute. */ |
| 1830 | block = vol->mft_lcn << vol->cluster_size_bits >> |
| 1831 | sb->s_blocksize_bits; |
| 1832 | nr_blocks = vol->mft_record_size >> sb->s_blocksize_bits; |
| 1833 | if (!nr_blocks) |
| 1834 | nr_blocks = 1; |
| 1835 | |
| 1836 | /* Load $MFT/$DATA's first mft record. */ |
| 1837 | for (i = 0; i < nr_blocks; i++) { |
| 1838 | bh = sb_bread(sb, block++); |
| 1839 | if (!bh) { |
| 1840 | ntfs_error(sb, "Device read failed."); |
| 1841 | goto err_out; |
| 1842 | } |
| 1843 | memcpy((char*)m + (i << sb->s_blocksize_bits), bh->b_data, |
| 1844 | sb->s_blocksize); |
| 1845 | brelse(bh); |
| 1846 | } |
| 1847 | |
| 1848 | /* Apply the mst fixups. */ |
| 1849 | if (post_read_mst_fixup((NTFS_RECORD*)m, vol->mft_record_size)) { |
| 1850 | /* FIXME: Try to use the $MFTMirr now. */ |
| 1851 | ntfs_error(sb, "MST fixup failed. $MFT is corrupt."); |
| 1852 | goto err_out; |
| 1853 | } |
| 1854 | |
| 1855 | /* Need this to sanity check attribute list references to $MFT. */ |
| 1856 | vi->i_generation = ni->seq_no = le16_to_cpu(m->sequence_number); |
| 1857 | |
| 1858 | /* Provides readpage() and sync_page() for map_mft_record(). */ |
| 1859 | vi->i_mapping->a_ops = &ntfs_mst_aops; |
| 1860 | |
| 1861 | ctx = ntfs_attr_get_search_ctx(ni, m); |
| 1862 | if (!ctx) { |
| 1863 | err = -ENOMEM; |
| 1864 | goto err_out; |
| 1865 | } |
| 1866 | |
| 1867 | /* Find the attribute list attribute if present. */ |
| 1868 | err = ntfs_attr_lookup(AT_ATTRIBUTE_LIST, NULL, 0, 0, 0, NULL, 0, ctx); |
| 1869 | if (err) { |
| 1870 | if (unlikely(err != -ENOENT)) { |
| 1871 | ntfs_error(sb, "Failed to lookup attribute list " |
| 1872 | "attribute. You should run chkdsk."); |
| 1873 | goto put_err_out; |
| 1874 | } |
| 1875 | } else /* if (!err) */ { |
| 1876 | ATTR_LIST_ENTRY *al_entry, *next_al_entry; |
| 1877 | u8 *al_end; |
Anton Altaparmakov | 3672b63 | 2006-02-24 09:55:07 +0000 | [diff] [blame] | 1878 | static const char *es = " Not allowed. $MFT is corrupt. " |
| 1879 | "You should run chkdsk."; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1880 | |
| 1881 | ntfs_debug("Attribute list attribute found in $MFT."); |
| 1882 | NInoSetAttrList(ni); |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1883 | a = ctx->attr; |
Anton Altaparmakov | 3672b63 | 2006-02-24 09:55:07 +0000 | [diff] [blame] | 1884 | if (a->flags & ATTR_COMPRESSION_MASK) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1885 | ntfs_error(sb, "Attribute list attribute is " |
Anton Altaparmakov | 3672b63 | 2006-02-24 09:55:07 +0000 | [diff] [blame] | 1886 | "compressed.%s", es); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1887 | goto put_err_out; |
| 1888 | } |
Anton Altaparmakov | 3672b63 | 2006-02-24 09:55:07 +0000 | [diff] [blame] | 1889 | if (a->flags & ATTR_IS_ENCRYPTED || |
| 1890 | a->flags & ATTR_IS_SPARSE) { |
| 1891 | if (a->non_resident) { |
| 1892 | ntfs_error(sb, "Non-resident attribute list " |
| 1893 | "attribute is encrypted/" |
| 1894 | "sparse.%s", es); |
| 1895 | goto put_err_out; |
| 1896 | } |
| 1897 | ntfs_warning(sb, "Resident attribute list attribute " |
| 1898 | "in $MFT system file is marked " |
| 1899 | "encrypted/sparse which is not true. " |
| 1900 | "However, Windows allows this and " |
| 1901 | "chkdsk does not detect or correct it " |
| 1902 | "so we will just ignore the invalid " |
| 1903 | "flags and pretend they are not set."); |
| 1904 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1905 | /* Now allocate memory for the attribute list. */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1906 | ni->attr_list_size = (u32)ntfs_attr_size(a); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1907 | ni->attr_list = ntfs_malloc_nofs(ni->attr_list_size); |
| 1908 | if (!ni->attr_list) { |
| 1909 | ntfs_error(sb, "Not enough memory to allocate buffer " |
| 1910 | "for attribute list."); |
| 1911 | goto put_err_out; |
| 1912 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1913 | if (a->non_resident) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1914 | NInoSetAttrListNonResident(ni); |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1915 | if (a->data.non_resident.lowest_vcn) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1916 | ntfs_error(sb, "Attribute list has non zero " |
| 1917 | "lowest_vcn. $MFT is corrupt. " |
| 1918 | "You should run chkdsk."); |
| 1919 | goto put_err_out; |
| 1920 | } |
| 1921 | /* Setup the runlist. */ |
| 1922 | ni->attr_list_rl.rl = ntfs_mapping_pairs_decompress(vol, |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1923 | a, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1924 | if (IS_ERR(ni->attr_list_rl.rl)) { |
| 1925 | err = PTR_ERR(ni->attr_list_rl.rl); |
| 1926 | ni->attr_list_rl.rl = NULL; |
| 1927 | ntfs_error(sb, "Mapping pairs decompression " |
| 1928 | "failed with error code %i.", |
| 1929 | -err); |
| 1930 | goto put_err_out; |
| 1931 | } |
| 1932 | /* Now load the attribute list. */ |
| 1933 | if ((err = load_attribute_list(vol, &ni->attr_list_rl, |
| 1934 | ni->attr_list, ni->attr_list_size, |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1935 | sle64_to_cpu(a->data. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1936 | non_resident.initialized_size)))) { |
| 1937 | ntfs_error(sb, "Failed to load attribute list " |
| 1938 | "attribute with error code %i.", |
| 1939 | -err); |
| 1940 | goto put_err_out; |
| 1941 | } |
| 1942 | } else /* if (!ctx.attr->non_resident) */ { |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1943 | if ((u8*)a + le16_to_cpu( |
| 1944 | a->data.resident.value_offset) + |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1945 | le32_to_cpu( |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1946 | a->data.resident.value_length) > |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1947 | (u8*)ctx->mrec + vol->mft_record_size) { |
| 1948 | ntfs_error(sb, "Corrupt attribute list " |
| 1949 | "attribute."); |
| 1950 | goto put_err_out; |
| 1951 | } |
| 1952 | /* Now copy the attribute list. */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1953 | memcpy(ni->attr_list, (u8*)a + le16_to_cpu( |
| 1954 | a->data.resident.value_offset), |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1955 | le32_to_cpu( |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 1956 | a->data.resident.value_length)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1957 | } |
| 1958 | /* The attribute list is now setup in memory. */ |
| 1959 | /* |
| 1960 | * FIXME: I don't know if this case is actually possible. |
| 1961 | * According to logic it is not possible but I have seen too |
| 1962 | * many weird things in MS software to rely on logic... Thus we |
| 1963 | * perform a manual search and make sure the first $MFT/$DATA |
| 1964 | * extent is in the base inode. If it is not we abort with an |
| 1965 | * error and if we ever see a report of this error we will need |
| 1966 | * to do some magic in order to have the necessary mft record |
| 1967 | * loaded and in the right place in the page cache. But |
| 1968 | * hopefully logic will prevail and this never happens... |
| 1969 | */ |
| 1970 | al_entry = (ATTR_LIST_ENTRY*)ni->attr_list; |
| 1971 | al_end = (u8*)al_entry + ni->attr_list_size; |
| 1972 | for (;; al_entry = next_al_entry) { |
| 1973 | /* Out of bounds check. */ |
| 1974 | if ((u8*)al_entry < ni->attr_list || |
| 1975 | (u8*)al_entry > al_end) |
| 1976 | goto em_put_err_out; |
| 1977 | /* Catch the end of the attribute list. */ |
| 1978 | if ((u8*)al_entry == al_end) |
| 1979 | goto em_put_err_out; |
| 1980 | if (!al_entry->length) |
| 1981 | goto em_put_err_out; |
| 1982 | if ((u8*)al_entry + 6 > al_end || (u8*)al_entry + |
| 1983 | le16_to_cpu(al_entry->length) > al_end) |
| 1984 | goto em_put_err_out; |
| 1985 | next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry + |
| 1986 | le16_to_cpu(al_entry->length)); |
| 1987 | if (le32_to_cpu(al_entry->type) > |
| 1988 | const_le32_to_cpu(AT_DATA)) |
| 1989 | goto em_put_err_out; |
| 1990 | if (AT_DATA != al_entry->type) |
| 1991 | continue; |
| 1992 | /* We want an unnamed attribute. */ |
| 1993 | if (al_entry->name_length) |
| 1994 | goto em_put_err_out; |
| 1995 | /* Want the first entry, i.e. lowest_vcn == 0. */ |
| 1996 | if (al_entry->lowest_vcn) |
| 1997 | goto em_put_err_out; |
| 1998 | /* First entry has to be in the base mft record. */ |
| 1999 | if (MREF_LE(al_entry->mft_reference) != vi->i_ino) { |
| 2000 | /* MFT references do not match, logic fails. */ |
| 2001 | ntfs_error(sb, "BUG: The first $DATA extent " |
| 2002 | "of $MFT is not in the base " |
| 2003 | "mft record. Please report " |
| 2004 | "you saw this message to " |
| 2005 | "linux-ntfs-dev@lists." |
| 2006 | "sourceforge.net"); |
| 2007 | goto put_err_out; |
| 2008 | } else { |
| 2009 | /* Sequence numbers must match. */ |
| 2010 | if (MSEQNO_LE(al_entry->mft_reference) != |
| 2011 | ni->seq_no) |
| 2012 | goto em_put_err_out; |
| 2013 | /* Got it. All is ok. We can stop now. */ |
| 2014 | break; |
| 2015 | } |
| 2016 | } |
| 2017 | } |
| 2018 | |
| 2019 | ntfs_attr_reinit_search_ctx(ctx); |
| 2020 | |
| 2021 | /* Now load all attribute extents. */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 2022 | a = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2023 | next_vcn = last_vcn = highest_vcn = 0; |
| 2024 | while (!(err = ntfs_attr_lookup(AT_DATA, NULL, 0, 0, next_vcn, NULL, 0, |
| 2025 | ctx))) { |
| 2026 | runlist_element *nrl; |
| 2027 | |
| 2028 | /* Cache the current attribute. */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 2029 | a = ctx->attr; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2030 | /* $MFT must be non-resident. */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 2031 | if (!a->non_resident) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2032 | ntfs_error(sb, "$MFT must be non-resident but a " |
| 2033 | "resident extent was found. $MFT is " |
| 2034 | "corrupt. Run chkdsk."); |
| 2035 | goto put_err_out; |
| 2036 | } |
| 2037 | /* $MFT must be uncompressed and unencrypted. */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 2038 | if (a->flags & ATTR_COMPRESSION_MASK || |
| 2039 | a->flags & ATTR_IS_ENCRYPTED || |
| 2040 | a->flags & ATTR_IS_SPARSE) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2041 | ntfs_error(sb, "$MFT must be uncompressed, " |
| 2042 | "non-sparse, and unencrypted but a " |
| 2043 | "compressed/sparse/encrypted extent " |
| 2044 | "was found. $MFT is corrupt. Run " |
| 2045 | "chkdsk."); |
| 2046 | goto put_err_out; |
| 2047 | } |
| 2048 | /* |
| 2049 | * Decompress the mapping pairs array of this extent and merge |
| 2050 | * the result into the existing runlist. No need for locking |
| 2051 | * as we have exclusive access to the inode at this time and we |
| 2052 | * are a mount in progress task, too. |
| 2053 | */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 2054 | nrl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2055 | if (IS_ERR(nrl)) { |
| 2056 | ntfs_error(sb, "ntfs_mapping_pairs_decompress() " |
| 2057 | "failed with error code %ld. $MFT is " |
| 2058 | "corrupt.", PTR_ERR(nrl)); |
| 2059 | goto put_err_out; |
| 2060 | } |
| 2061 | ni->runlist.rl = nrl; |
| 2062 | |
| 2063 | /* Are we in the first extent? */ |
| 2064 | if (!next_vcn) { |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 2065 | if (a->data.non_resident.lowest_vcn) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2066 | ntfs_error(sb, "First extent of $DATA " |
| 2067 | "attribute has non zero " |
| 2068 | "lowest_vcn. $MFT is corrupt. " |
| 2069 | "You should run chkdsk."); |
| 2070 | goto put_err_out; |
| 2071 | } |
| 2072 | /* Get the last vcn in the $DATA attribute. */ |
| 2073 | last_vcn = sle64_to_cpu( |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 2074 | a->data.non_resident.allocated_size) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2075 | >> vol->cluster_size_bits; |
| 2076 | /* Fill in the inode size. */ |
| 2077 | vi->i_size = sle64_to_cpu( |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 2078 | a->data.non_resident.data_size); |
| 2079 | ni->initialized_size = sle64_to_cpu( |
| 2080 | a->data.non_resident.initialized_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2081 | ni->allocated_size = sle64_to_cpu( |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 2082 | a->data.non_resident.allocated_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2083 | /* |
| 2084 | * Verify the number of mft records does not exceed |
| 2085 | * 2^32 - 1. |
| 2086 | */ |
| 2087 | if ((vi->i_size >> vol->mft_record_size_bits) >= |
| 2088 | (1ULL << 32)) { |
| 2089 | ntfs_error(sb, "$MFT is too big! Aborting."); |
| 2090 | goto put_err_out; |
| 2091 | } |
| 2092 | /* |
| 2093 | * We have got the first extent of the runlist for |
| 2094 | * $MFT which means it is now relatively safe to call |
| 2095 | * the normal ntfs_read_inode() function. |
| 2096 | * Complete reading the inode, this will actually |
| 2097 | * re-read the mft record for $MFT, this time entering |
| 2098 | * it into the page cache with which we complete the |
| 2099 | * kick start of the volume. It should be safe to do |
| 2100 | * this now as the first extent of $MFT/$DATA is |
| 2101 | * already known and we would hope that we don't need |
| 2102 | * further extents in order to find the other |
| 2103 | * attributes belonging to $MFT. Only time will tell if |
| 2104 | * this is really the case. If not we will have to play |
| 2105 | * magic at this point, possibly duplicating a lot of |
| 2106 | * ntfs_read_inode() at this point. We will need to |
| 2107 | * ensure we do enough of its work to be able to call |
| 2108 | * ntfs_read_inode() on extents of $MFT/$DATA. But lets |
| 2109 | * hope this never happens... |
| 2110 | */ |
| 2111 | ntfs_read_locked_inode(vi); |
| 2112 | if (is_bad_inode(vi)) { |
| 2113 | ntfs_error(sb, "ntfs_read_inode() of $MFT " |
| 2114 | "failed. BUG or corrupt $MFT. " |
| 2115 | "Run chkdsk and if no errors " |
| 2116 | "are found, please report you " |
| 2117 | "saw this message to " |
| 2118 | "linux-ntfs-dev@lists." |
| 2119 | "sourceforge.net"); |
| 2120 | ntfs_attr_put_search_ctx(ctx); |
| 2121 | /* Revert to the safe super operations. */ |
| 2122 | ntfs_free(m); |
| 2123 | return -1; |
| 2124 | } |
| 2125 | /* |
| 2126 | * Re-initialize some specifics about $MFT's inode as |
| 2127 | * ntfs_read_inode() will have set up the default ones. |
| 2128 | */ |
| 2129 | /* Set uid and gid to root. */ |
| 2130 | vi->i_uid = vi->i_gid = 0; |
| 2131 | /* Regular file. No access for anyone. */ |
| 2132 | vi->i_mode = S_IFREG; |
| 2133 | /* No VFS initiated operations allowed for $MFT. */ |
| 2134 | vi->i_op = &ntfs_empty_inode_ops; |
| 2135 | vi->i_fop = &ntfs_empty_file_ops; |
| 2136 | } |
| 2137 | |
| 2138 | /* Get the lowest vcn for the next extent. */ |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 2139 | highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2140 | next_vcn = highest_vcn + 1; |
| 2141 | |
| 2142 | /* Only one extent or error, which we catch below. */ |
| 2143 | if (next_vcn <= 0) |
| 2144 | break; |
| 2145 | |
| 2146 | /* Avoid endless loops due to corruption. */ |
| 2147 | if (next_vcn < sle64_to_cpu( |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 2148 | a->data.non_resident.lowest_vcn)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2149 | ntfs_error(sb, "$MFT has corrupt attribute list " |
| 2150 | "attribute. Run chkdsk."); |
| 2151 | goto put_err_out; |
| 2152 | } |
| 2153 | } |
| 2154 | if (err != -ENOENT) { |
| 2155 | ntfs_error(sb, "Failed to lookup $MFT/$DATA attribute extent. " |
| 2156 | "$MFT is corrupt. Run chkdsk."); |
| 2157 | goto put_err_out; |
| 2158 | } |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 2159 | if (!a) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2160 | ntfs_error(sb, "$MFT/$DATA attribute not found. $MFT is " |
| 2161 | "corrupt. Run chkdsk."); |
| 2162 | goto put_err_out; |
| 2163 | } |
| 2164 | if (highest_vcn && highest_vcn != last_vcn - 1) { |
| 2165 | ntfs_error(sb, "Failed to load the complete runlist for " |
| 2166 | "$MFT/$DATA. Driver bug or corrupt $MFT. " |
| 2167 | "Run chkdsk."); |
| 2168 | ntfs_debug("highest_vcn = 0x%llx, last_vcn - 1 = 0x%llx", |
| 2169 | (unsigned long long)highest_vcn, |
| 2170 | (unsigned long long)last_vcn - 1); |
| 2171 | goto put_err_out; |
| 2172 | } |
| 2173 | ntfs_attr_put_search_ctx(ctx); |
| 2174 | ntfs_debug("Done."); |
| 2175 | ntfs_free(m); |
Ingo Molnar | 5934537 | 2006-07-03 00:25:18 -0700 | [diff] [blame^] | 2176 | |
| 2177 | /* |
| 2178 | * Split the locking rules of the MFT inode from the |
| 2179 | * locking rules of other inodes: |
| 2180 | */ |
| 2181 | lockdep_set_class(&ni->runlist.lock, &mft_ni_runlist_lock_key); |
| 2182 | lockdep_set_class(&ni->mrec_lock, &mft_ni_mrec_lock_key); |
| 2183 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2184 | return 0; |
| 2185 | |
| 2186 | em_put_err_out: |
| 2187 | ntfs_error(sb, "Couldn't find first extent of $DATA attribute in " |
| 2188 | "attribute list. $MFT is corrupt. Run chkdsk."); |
| 2189 | put_err_out: |
| 2190 | ntfs_attr_put_search_ctx(ctx); |
| 2191 | err_out: |
| 2192 | ntfs_error(sb, "Failed. Marking inode as bad."); |
| 2193 | make_bad_inode(vi); |
| 2194 | ntfs_free(m); |
| 2195 | return -1; |
| 2196 | } |
| 2197 | |
| 2198 | /** |
| 2199 | * ntfs_put_inode - handler for when the inode reference count is decremented |
| 2200 | * @vi: vfs inode |
| 2201 | * |
| 2202 | * The VFS calls ntfs_put_inode() every time the inode reference count (i_count) |
| 2203 | * is about to be decremented (but before the decrement itself. |
| 2204 | * |
| 2205 | * If the inode @vi is a directory with two references, one of which is being |
| 2206 | * dropped, we need to put the attribute inode for the directory index bitmap, |
| 2207 | * if it is present, otherwise the directory inode would remain pinned for |
| 2208 | * ever. |
| 2209 | */ |
| 2210 | void ntfs_put_inode(struct inode *vi) |
| 2211 | { |
| 2212 | if (S_ISDIR(vi->i_mode) && atomic_read(&vi->i_count) == 2) { |
| 2213 | ntfs_inode *ni = NTFS_I(vi); |
| 2214 | if (NInoIndexAllocPresent(ni)) { |
| 2215 | struct inode *bvi = NULL; |
Jes Sorensen | 1b1dcc1 | 2006-01-09 15:59:24 -0800 | [diff] [blame] | 2216 | mutex_lock(&vi->i_mutex); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2217 | if (atomic_read(&vi->i_count) == 2) { |
| 2218 | bvi = ni->itype.index.bmp_ino; |
| 2219 | if (bvi) |
| 2220 | ni->itype.index.bmp_ino = NULL; |
| 2221 | } |
Jes Sorensen | 1b1dcc1 | 2006-01-09 15:59:24 -0800 | [diff] [blame] | 2222 | mutex_unlock(&vi->i_mutex); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2223 | if (bvi) |
| 2224 | iput(bvi); |
| 2225 | } |
| 2226 | } |
| 2227 | } |
| 2228 | |
| 2229 | static void __ntfs_clear_inode(ntfs_inode *ni) |
| 2230 | { |
| 2231 | /* Free all alocated memory. */ |
| 2232 | down_write(&ni->runlist.lock); |
| 2233 | if (ni->runlist.rl) { |
| 2234 | ntfs_free(ni->runlist.rl); |
| 2235 | ni->runlist.rl = NULL; |
| 2236 | } |
| 2237 | up_write(&ni->runlist.lock); |
| 2238 | |
| 2239 | if (ni->attr_list) { |
| 2240 | ntfs_free(ni->attr_list); |
| 2241 | ni->attr_list = NULL; |
| 2242 | } |
| 2243 | |
| 2244 | down_write(&ni->attr_list_rl.lock); |
| 2245 | if (ni->attr_list_rl.rl) { |
| 2246 | ntfs_free(ni->attr_list_rl.rl); |
| 2247 | ni->attr_list_rl.rl = NULL; |
| 2248 | } |
| 2249 | up_write(&ni->attr_list_rl.lock); |
| 2250 | |
| 2251 | if (ni->name_len && ni->name != I30) { |
| 2252 | /* Catch bugs... */ |
| 2253 | BUG_ON(!ni->name); |
| 2254 | kfree(ni->name); |
| 2255 | } |
| 2256 | } |
| 2257 | |
| 2258 | void ntfs_clear_extent_inode(ntfs_inode *ni) |
| 2259 | { |
| 2260 | ntfs_debug("Entering for inode 0x%lx.", ni->mft_no); |
| 2261 | |
| 2262 | BUG_ON(NInoAttr(ni)); |
| 2263 | BUG_ON(ni->nr_extents != -1); |
| 2264 | |
| 2265 | #ifdef NTFS_RW |
| 2266 | if (NInoDirty(ni)) { |
| 2267 | if (!is_bad_inode(VFS_I(ni->ext.base_ntfs_ino))) |
| 2268 | ntfs_error(ni->vol->sb, "Clearing dirty extent inode! " |
| 2269 | "Losing data! This is a BUG!!!"); |
| 2270 | // FIXME: Do something!!! |
| 2271 | } |
| 2272 | #endif /* NTFS_RW */ |
| 2273 | |
| 2274 | __ntfs_clear_inode(ni); |
| 2275 | |
| 2276 | /* Bye, bye... */ |
| 2277 | ntfs_destroy_extent_inode(ni); |
| 2278 | } |
| 2279 | |
| 2280 | /** |
| 2281 | * ntfs_clear_big_inode - clean up the ntfs specific part of an inode |
| 2282 | * @vi: vfs inode pending annihilation |
| 2283 | * |
| 2284 | * When the VFS is going to remove an inode from memory, ntfs_clear_big_inode() |
| 2285 | * is called, which deallocates all memory belonging to the NTFS specific part |
| 2286 | * of the inode and returns. |
| 2287 | * |
| 2288 | * If the MFT record is dirty, we commit it before doing anything else. |
| 2289 | */ |
| 2290 | void ntfs_clear_big_inode(struct inode *vi) |
| 2291 | { |
| 2292 | ntfs_inode *ni = NTFS_I(vi); |
| 2293 | |
| 2294 | /* |
| 2295 | * If the inode @vi is an index inode we need to put the attribute |
| 2296 | * inode for the index bitmap, if it is present, otherwise the index |
| 2297 | * inode would disappear and the attribute inode for the index bitmap |
| 2298 | * would no longer be referenced from anywhere and thus it would remain |
| 2299 | * pinned for ever. |
| 2300 | */ |
| 2301 | if (NInoAttr(ni) && (ni->type == AT_INDEX_ALLOCATION) && |
| 2302 | NInoIndexAllocPresent(ni) && ni->itype.index.bmp_ino) { |
| 2303 | iput(ni->itype.index.bmp_ino); |
| 2304 | ni->itype.index.bmp_ino = NULL; |
| 2305 | } |
| 2306 | #ifdef NTFS_RW |
| 2307 | if (NInoDirty(ni)) { |
| 2308 | BOOL was_bad = (is_bad_inode(vi)); |
| 2309 | |
| 2310 | /* Committing the inode also commits all extent inodes. */ |
| 2311 | ntfs_commit_inode(vi); |
| 2312 | |
| 2313 | if (!was_bad && (is_bad_inode(vi) || NInoDirty(ni))) { |
| 2314 | ntfs_error(vi->i_sb, "Failed to commit dirty inode " |
| 2315 | "0x%lx. Losing data!", vi->i_ino); |
| 2316 | // FIXME: Do something!!! |
| 2317 | } |
| 2318 | } |
| 2319 | #endif /* NTFS_RW */ |
| 2320 | |
| 2321 | /* No need to lock at this stage as no one else has a reference. */ |
| 2322 | if (ni->nr_extents > 0) { |
| 2323 | int i; |
| 2324 | |
| 2325 | for (i = 0; i < ni->nr_extents; i++) |
| 2326 | ntfs_clear_extent_inode(ni->ext.extent_ntfs_inos[i]); |
| 2327 | kfree(ni->ext.extent_ntfs_inos); |
| 2328 | } |
| 2329 | |
| 2330 | __ntfs_clear_inode(ni); |
| 2331 | |
| 2332 | if (NInoAttr(ni)) { |
| 2333 | /* Release the base inode if we are holding it. */ |
| 2334 | if (ni->nr_extents == -1) { |
| 2335 | iput(VFS_I(ni->ext.base_ntfs_ino)); |
| 2336 | ni->nr_extents = 0; |
| 2337 | ni->ext.base_ntfs_ino = NULL; |
| 2338 | } |
| 2339 | } |
| 2340 | return; |
| 2341 | } |
| 2342 | |
| 2343 | /** |
| 2344 | * ntfs_show_options - show mount options in /proc/mounts |
| 2345 | * @sf: seq_file in which to write our mount options |
| 2346 | * @mnt: vfs mount whose mount options to display |
| 2347 | * |
| 2348 | * Called by the VFS once for each mounted ntfs volume when someone reads |
| 2349 | * /proc/mounts in order to display the NTFS specific mount options of each |
| 2350 | * mount. The mount options of the vfs mount @mnt are written to the seq file |
| 2351 | * @sf and success is returned. |
| 2352 | */ |
| 2353 | int ntfs_show_options(struct seq_file *sf, struct vfsmount *mnt) |
| 2354 | { |
| 2355 | ntfs_volume *vol = NTFS_SB(mnt->mnt_sb); |
| 2356 | int i; |
| 2357 | |
| 2358 | seq_printf(sf, ",uid=%i", vol->uid); |
| 2359 | seq_printf(sf, ",gid=%i", vol->gid); |
| 2360 | if (vol->fmask == vol->dmask) |
| 2361 | seq_printf(sf, ",umask=0%o", vol->fmask); |
| 2362 | else { |
| 2363 | seq_printf(sf, ",fmask=0%o", vol->fmask); |
| 2364 | seq_printf(sf, ",dmask=0%o", vol->dmask); |
| 2365 | } |
| 2366 | seq_printf(sf, ",nls=%s", vol->nls_map->charset); |
| 2367 | if (NVolCaseSensitive(vol)) |
| 2368 | seq_printf(sf, ",case_sensitive"); |
| 2369 | if (NVolShowSystemFiles(vol)) |
| 2370 | seq_printf(sf, ",show_sys_files"); |
Anton Altaparmakov | c002f42 | 2005-02-03 12:02:56 +0000 | [diff] [blame] | 2371 | if (!NVolSparseEnabled(vol)) |
| 2372 | seq_printf(sf, ",disable_sparse"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2373 | for (i = 0; on_errors_arr[i].val; i++) { |
| 2374 | if (on_errors_arr[i].val & vol->on_errors) |
| 2375 | seq_printf(sf, ",errors=%s", on_errors_arr[i].str); |
| 2376 | } |
| 2377 | seq_printf(sf, ",mft_zone_multiplier=%i", vol->mft_zone_multiplier); |
| 2378 | return 0; |
| 2379 | } |
| 2380 | |
| 2381 | #ifdef NTFS_RW |
| 2382 | |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2383 | static const char *es = " Leaving inconsistent metadata. Unmount and run " |
| 2384 | "chkdsk."; |
| 2385 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2386 | /** |
| 2387 | * ntfs_truncate - called when the i_size of an ntfs inode is changed |
| 2388 | * @vi: inode for which the i_size was changed |
| 2389 | * |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2390 | * We only support i_size changes for normal files at present, i.e. not |
| 2391 | * compressed and not encrypted. This is enforced in ntfs_setattr(), see |
| 2392 | * below. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2393 | * |
| 2394 | * The kernel guarantees that @vi is a regular file (S_ISREG() is true) and |
| 2395 | * that the change is allowed. |
| 2396 | * |
| 2397 | * This implies for us that @vi is a file inode rather than a directory, index, |
| 2398 | * or attribute inode as well as that @vi is a base inode. |
| 2399 | * |
| 2400 | * Returns 0 on success or -errno on error. |
| 2401 | * |
Jes Sorensen | 1b1dcc1 | 2006-01-09 15:59:24 -0800 | [diff] [blame] | 2402 | * Called with ->i_mutex held. In all but one case ->i_alloc_sem is held for |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2403 | * writing. The only case in the kernel where ->i_alloc_sem is not held is |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2404 | * mm/filemap.c::generic_file_buffered_write() where vmtruncate() is called |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2405 | * with the current i_size as the offset. The analogous place in NTFS is in |
| 2406 | * fs/ntfs/file.c::ntfs_file_buffered_write() where we call vmtruncate() again |
| 2407 | * without holding ->i_alloc_sem. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2408 | */ |
| 2409 | int ntfs_truncate(struct inode *vi) |
| 2410 | { |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2411 | s64 new_size, old_size, nr_freed, new_alloc_size, old_alloc_size; |
| 2412 | VCN highest_vcn; |
| 2413 | unsigned long flags; |
| 2414 | ntfs_inode *base_ni, *ni = NTFS_I(vi); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2415 | ntfs_volume *vol = ni->vol; |
| 2416 | ntfs_attr_search_ctx *ctx; |
| 2417 | MFT_RECORD *m; |
Anton Altaparmakov | 5ae9fcf | 2005-03-02 17:03:24 +0000 | [diff] [blame] | 2418 | ATTR_RECORD *a; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2419 | const char *te = " Leaving file length out of sync with i_size."; |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2420 | int err, mp_size, size_change, alloc_change; |
| 2421 | u32 attr_len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2422 | |
| 2423 | ntfs_debug("Entering for inode 0x%lx.", vi->i_ino); |
| 2424 | BUG_ON(NInoAttr(ni)); |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2425 | BUG_ON(S_ISDIR(vi->i_mode)); |
| 2426 | BUG_ON(NInoMstProtected(ni)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2427 | BUG_ON(ni->nr_extents < 0); |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2428 | retry_truncate: |
| 2429 | /* |
| 2430 | * Lock the runlist for writing and map the mft record to ensure it is |
| 2431 | * safe to mess with the attribute runlist and sizes. |
| 2432 | */ |
| 2433 | down_write(&ni->runlist.lock); |
| 2434 | if (!NInoAttr(ni)) |
| 2435 | base_ni = ni; |
| 2436 | else |
| 2437 | base_ni = ni->ext.base_ntfs_ino; |
| 2438 | m = map_mft_record(base_ni); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2439 | if (IS_ERR(m)) { |
| 2440 | err = PTR_ERR(m); |
| 2441 | ntfs_error(vi->i_sb, "Failed to map mft record for inode 0x%lx " |
| 2442 | "(error code %d).%s", vi->i_ino, err, te); |
| 2443 | ctx = NULL; |
| 2444 | m = NULL; |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2445 | goto old_bad_out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2446 | } |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2447 | ctx = ntfs_attr_get_search_ctx(base_ni, m); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2448 | if (unlikely(!ctx)) { |
| 2449 | ntfs_error(vi->i_sb, "Failed to allocate a search context for " |
| 2450 | "inode 0x%lx (not enough memory).%s", |
| 2451 | vi->i_ino, te); |
| 2452 | err = -ENOMEM; |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2453 | goto old_bad_out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2454 | } |
| 2455 | err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len, |
| 2456 | CASE_SENSITIVE, 0, NULL, 0, ctx); |
| 2457 | if (unlikely(err)) { |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2458 | if (err == -ENOENT) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2459 | ntfs_error(vi->i_sb, "Open attribute is missing from " |
| 2460 | "mft record. Inode 0x%lx is corrupt. " |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2461 | "Run chkdsk.%s", vi->i_ino, te); |
| 2462 | err = -EIO; |
| 2463 | } else |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2464 | ntfs_error(vi->i_sb, "Failed to lookup attribute in " |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2465 | "inode 0x%lx (error code %d).%s", |
| 2466 | vi->i_ino, err, te); |
| 2467 | goto old_bad_out; |
| 2468 | } |
| 2469 | m = ctx->mrec; |
| 2470 | a = ctx->attr; |
| 2471 | /* |
| 2472 | * The i_size of the vfs inode is the new size for the attribute value. |
| 2473 | */ |
| 2474 | new_size = i_size_read(vi); |
| 2475 | /* The current size of the attribute value is the old size. */ |
| 2476 | old_size = ntfs_attr_size(a); |
| 2477 | /* Calculate the new allocated size. */ |
| 2478 | if (NInoNonResident(ni)) |
| 2479 | new_alloc_size = (new_size + vol->cluster_size - 1) & |
| 2480 | ~(s64)vol->cluster_size_mask; |
| 2481 | else |
| 2482 | new_alloc_size = (new_size + 7) & ~7; |
| 2483 | /* The current allocated size is the old allocated size. */ |
| 2484 | read_lock_irqsave(&ni->size_lock, flags); |
| 2485 | old_alloc_size = ni->allocated_size; |
| 2486 | read_unlock_irqrestore(&ni->size_lock, flags); |
| 2487 | /* |
| 2488 | * The change in the file size. This will be 0 if no change, >0 if the |
| 2489 | * size is growing, and <0 if the size is shrinking. |
| 2490 | */ |
| 2491 | size_change = -1; |
| 2492 | if (new_size - old_size >= 0) { |
| 2493 | size_change = 1; |
| 2494 | if (new_size == old_size) |
| 2495 | size_change = 0; |
| 2496 | } |
| 2497 | /* As above for the allocated size. */ |
| 2498 | alloc_change = -1; |
| 2499 | if (new_alloc_size - old_alloc_size >= 0) { |
| 2500 | alloc_change = 1; |
| 2501 | if (new_alloc_size == old_alloc_size) |
| 2502 | alloc_change = 0; |
| 2503 | } |
| 2504 | /* |
| 2505 | * If neither the size nor the allocation are being changed there is |
| 2506 | * nothing to do. |
| 2507 | */ |
| 2508 | if (!size_change && !alloc_change) |
| 2509 | goto unm_done; |
| 2510 | /* If the size is changing, check if new size is allowed in $AttrDef. */ |
| 2511 | if (size_change) { |
| 2512 | err = ntfs_attr_size_bounds_check(vol, ni->type, new_size); |
| 2513 | if (unlikely(err)) { |
| 2514 | if (err == -ERANGE) { |
| 2515 | ntfs_error(vol->sb, "Truncate would cause the " |
| 2516 | "inode 0x%lx to %simum size " |
| 2517 | "for its attribute type " |
| 2518 | "(0x%x). Aborting truncate.", |
| 2519 | vi->i_ino, |
| 2520 | new_size > old_size ? "exceed " |
| 2521 | "the max" : "go under the min", |
| 2522 | le32_to_cpu(ni->type)); |
| 2523 | err = -EFBIG; |
| 2524 | } else { |
| 2525 | ntfs_error(vol->sb, "Inode 0x%lx has unknown " |
| 2526 | "attribute type 0x%x. " |
| 2527 | "Aborting truncate.", |
| 2528 | vi->i_ino, |
| 2529 | le32_to_cpu(ni->type)); |
| 2530 | err = -EIO; |
| 2531 | } |
| 2532 | /* Reset the vfs inode size to the old size. */ |
| 2533 | i_size_write(vi, old_size); |
| 2534 | goto err_out; |
| 2535 | } |
| 2536 | } |
| 2537 | if (NInoCompressed(ni) || NInoEncrypted(ni)) { |
| 2538 | ntfs_warning(vi->i_sb, "Changes in inode size are not " |
| 2539 | "supported yet for %s files, ignoring.", |
| 2540 | NInoCompressed(ni) ? "compressed" : |
| 2541 | "encrypted"); |
| 2542 | err = -EOPNOTSUPP; |
| 2543 | goto bad_out; |
| 2544 | } |
| 2545 | if (a->non_resident) |
| 2546 | goto do_non_resident_truncate; |
| 2547 | BUG_ON(NInoNonResident(ni)); |
| 2548 | /* Resize the attribute record to best fit the new attribute size. */ |
| 2549 | if (new_size < vol->mft_record_size && |
| 2550 | !ntfs_resident_attr_value_resize(m, a, new_size)) { |
| 2551 | unsigned long flags; |
| 2552 | |
| 2553 | /* The resize succeeded! */ |
| 2554 | flush_dcache_mft_record_page(ctx->ntfs_ino); |
| 2555 | mark_mft_record_dirty(ctx->ntfs_ino); |
| 2556 | write_lock_irqsave(&ni->size_lock, flags); |
| 2557 | /* Update the sizes in the ntfs inode and all is done. */ |
| 2558 | ni->allocated_size = le32_to_cpu(a->length) - |
| 2559 | le16_to_cpu(a->data.resident.value_offset); |
| 2560 | /* |
| 2561 | * Note ntfs_resident_attr_value_resize() has already done any |
| 2562 | * necessary data clearing in the attribute record. When the |
| 2563 | * file is being shrunk vmtruncate() will already have cleared |
| 2564 | * the top part of the last partial page, i.e. since this is |
| 2565 | * the resident case this is the page with index 0. However, |
| 2566 | * when the file is being expanded, the page cache page data |
| 2567 | * between the old data_size, i.e. old_size, and the new_size |
| 2568 | * has not been zeroed. Fortunately, we do not need to zero it |
| 2569 | * either since on one hand it will either already be zero due |
| 2570 | * to both readpage and writepage clearing partial page data |
| 2571 | * beyond i_size in which case there is nothing to do or in the |
| 2572 | * case of the file being mmap()ped at the same time, POSIX |
| 2573 | * specifies that the behaviour is unspecified thus we do not |
| 2574 | * have to do anything. This means that in our implementation |
| 2575 | * in the rare case that the file is mmap()ped and a write |
| 2576 | * occured into the mmap()ped region just beyond the file size |
| 2577 | * and writepage has not yet been called to write out the page |
| 2578 | * (which would clear the area beyond the file size) and we now |
| 2579 | * extend the file size to incorporate this dirty region |
| 2580 | * outside the file size, a write of the page would result in |
| 2581 | * this data being written to disk instead of being cleared. |
| 2582 | * Given both POSIX and the Linux mmap(2) man page specify that |
| 2583 | * this corner case is undefined, we choose to leave it like |
| 2584 | * that as this is much simpler for us as we cannot lock the |
| 2585 | * relevant page now since we are holding too many ntfs locks |
| 2586 | * which would result in a lock reversal deadlock. |
| 2587 | */ |
| 2588 | ni->initialized_size = new_size; |
| 2589 | write_unlock_irqrestore(&ni->size_lock, flags); |
| 2590 | goto unm_done; |
| 2591 | } |
| 2592 | /* If the above resize failed, this must be an attribute extension. */ |
| 2593 | BUG_ON(size_change < 0); |
| 2594 | /* |
| 2595 | * We have to drop all the locks so we can call |
| 2596 | * ntfs_attr_make_non_resident(). This could be optimised by try- |
| 2597 | * locking the first page cache page and only if that fails dropping |
| 2598 | * the locks, locking the page, and redoing all the locking and |
| 2599 | * lookups. While this would be a huge optimisation, it is not worth |
| 2600 | * it as this is definitely a slow code path as it only ever can happen |
| 2601 | * once for any given file. |
| 2602 | */ |
| 2603 | ntfs_attr_put_search_ctx(ctx); |
| 2604 | unmap_mft_record(base_ni); |
| 2605 | up_write(&ni->runlist.lock); |
| 2606 | /* |
| 2607 | * Not enough space in the mft record, try to make the attribute |
| 2608 | * non-resident and if successful restart the truncation process. |
| 2609 | */ |
| 2610 | err = ntfs_attr_make_non_resident(ni, old_size); |
| 2611 | if (likely(!err)) |
| 2612 | goto retry_truncate; |
| 2613 | /* |
| 2614 | * Could not make non-resident. If this is due to this not being |
| 2615 | * permitted for this attribute type or there not being enough space, |
| 2616 | * try to make other attributes non-resident. Otherwise fail. |
| 2617 | */ |
| 2618 | if (unlikely(err != -EPERM && err != -ENOSPC)) { |
| 2619 | ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, attribute " |
| 2620 | "type 0x%x, because the conversion from " |
| 2621 | "resident to non-resident attribute failed " |
| 2622 | "with error code %i.", vi->i_ino, |
| 2623 | (unsigned)le32_to_cpu(ni->type), err); |
| 2624 | if (err != -ENOMEM) |
| 2625 | err = -EIO; |
| 2626 | goto conv_err_out; |
| 2627 | } |
| 2628 | /* TODO: Not implemented from here, abort. */ |
| 2629 | if (err == -ENOSPC) |
| 2630 | ntfs_error(vol->sb, "Not enough space in the mft record/on " |
| 2631 | "disk for the non-resident attribute value. " |
| 2632 | "This case is not implemented yet."); |
| 2633 | else /* if (err == -EPERM) */ |
| 2634 | ntfs_error(vol->sb, "This attribute type may not be " |
| 2635 | "non-resident. This case is not implemented " |
| 2636 | "yet."); |
| 2637 | err = -EOPNOTSUPP; |
| 2638 | goto conv_err_out; |
| 2639 | #if 0 |
| 2640 | // TODO: Attempt to make other attributes non-resident. |
| 2641 | if (!err) |
| 2642 | goto do_resident_extend; |
| 2643 | /* |
| 2644 | * Both the attribute list attribute and the standard information |
| 2645 | * attribute must remain in the base inode. Thus, if this is one of |
| 2646 | * these attributes, we have to try to move other attributes out into |
| 2647 | * extent mft records instead. |
| 2648 | */ |
| 2649 | if (ni->type == AT_ATTRIBUTE_LIST || |
| 2650 | ni->type == AT_STANDARD_INFORMATION) { |
| 2651 | // TODO: Attempt to move other attributes into extent mft |
| 2652 | // records. |
| 2653 | err = -EOPNOTSUPP; |
| 2654 | if (!err) |
| 2655 | goto do_resident_extend; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2656 | goto err_out; |
| 2657 | } |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2658 | // TODO: Attempt to move this attribute to an extent mft record, but |
| 2659 | // only if it is not already the only attribute in an mft record in |
| 2660 | // which case there would be nothing to gain. |
| 2661 | err = -EOPNOTSUPP; |
| 2662 | if (!err) |
| 2663 | goto do_resident_extend; |
| 2664 | /* There is nothing we can do to make enough space. )-: */ |
| 2665 | goto err_out; |
| 2666 | #endif |
| 2667 | do_non_resident_truncate: |
| 2668 | BUG_ON(!NInoNonResident(ni)); |
| 2669 | if (alloc_change < 0) { |
| 2670 | highest_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn); |
| 2671 | if (highest_vcn > 0 && |
| 2672 | old_alloc_size >> vol->cluster_size_bits > |
| 2673 | highest_vcn + 1) { |
| 2674 | /* |
| 2675 | * This attribute has multiple extents. Not yet |
| 2676 | * supported. |
| 2677 | */ |
| 2678 | ntfs_error(vol->sb, "Cannot truncate inode 0x%lx, " |
| 2679 | "attribute type 0x%x, because the " |
| 2680 | "attribute is highly fragmented (it " |
| 2681 | "consists of multiple extents) and " |
| 2682 | "this case is not implemented yet.", |
| 2683 | vi->i_ino, |
| 2684 | (unsigned)le32_to_cpu(ni->type)); |
| 2685 | err = -EOPNOTSUPP; |
| 2686 | goto bad_out; |
| 2687 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2688 | } |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2689 | /* |
| 2690 | * If the size is shrinking, need to reduce the initialized_size and |
| 2691 | * the data_size before reducing the allocation. |
| 2692 | */ |
| 2693 | if (size_change < 0) { |
| 2694 | /* |
| 2695 | * Make the valid size smaller (i_size is already up-to-date). |
| 2696 | */ |
| 2697 | write_lock_irqsave(&ni->size_lock, flags); |
| 2698 | if (new_size < ni->initialized_size) { |
| 2699 | ni->initialized_size = new_size; |
| 2700 | a->data.non_resident.initialized_size = |
| 2701 | cpu_to_sle64(new_size); |
| 2702 | } |
| 2703 | a->data.non_resident.data_size = cpu_to_sle64(new_size); |
| 2704 | write_unlock_irqrestore(&ni->size_lock, flags); |
| 2705 | flush_dcache_mft_record_page(ctx->ntfs_ino); |
| 2706 | mark_mft_record_dirty(ctx->ntfs_ino); |
| 2707 | /* If the allocated size is not changing, we are done. */ |
| 2708 | if (!alloc_change) |
| 2709 | goto unm_done; |
| 2710 | /* |
| 2711 | * If the size is shrinking it makes no sense for the |
| 2712 | * allocation to be growing. |
| 2713 | */ |
| 2714 | BUG_ON(alloc_change > 0); |
| 2715 | } else /* if (size_change >= 0) */ { |
| 2716 | /* |
| 2717 | * The file size is growing or staying the same but the |
| 2718 | * allocation can be shrinking, growing or staying the same. |
| 2719 | */ |
| 2720 | if (alloc_change > 0) { |
| 2721 | /* |
| 2722 | * We need to extend the allocation and possibly update |
| 2723 | * the data size. If we are updating the data size, |
| 2724 | * since we are not touching the initialized_size we do |
| 2725 | * not need to worry about the actual data on disk. |
| 2726 | * And as far as the page cache is concerned, there |
| 2727 | * will be no pages beyond the old data size and any |
| 2728 | * partial region in the last page between the old and |
| 2729 | * new data size (or the end of the page if the new |
| 2730 | * data size is outside the page) does not need to be |
| 2731 | * modified as explained above for the resident |
| 2732 | * attribute truncate case. To do this, we simply drop |
| 2733 | * the locks we hold and leave all the work to our |
| 2734 | * friendly helper ntfs_attr_extend_allocation(). |
| 2735 | */ |
| 2736 | ntfs_attr_put_search_ctx(ctx); |
| 2737 | unmap_mft_record(base_ni); |
| 2738 | up_write(&ni->runlist.lock); |
| 2739 | err = ntfs_attr_extend_allocation(ni, new_size, |
| 2740 | size_change > 0 ? new_size : -1, -1); |
| 2741 | /* |
| 2742 | * ntfs_attr_extend_allocation() will have done error |
| 2743 | * output already. |
| 2744 | */ |
| 2745 | goto done; |
| 2746 | } |
| 2747 | if (!alloc_change) |
| 2748 | goto alloc_done; |
| 2749 | } |
| 2750 | /* alloc_change < 0 */ |
| 2751 | /* Free the clusters. */ |
| 2752 | nr_freed = ntfs_cluster_free(ni, new_alloc_size >> |
| 2753 | vol->cluster_size_bits, -1, ctx); |
| 2754 | m = ctx->mrec; |
| 2755 | a = ctx->attr; |
| 2756 | if (unlikely(nr_freed < 0)) { |
| 2757 | ntfs_error(vol->sb, "Failed to release cluster(s) (error code " |
| 2758 | "%lli). Unmount and run chkdsk to recover " |
| 2759 | "the lost cluster(s).", (long long)nr_freed); |
| 2760 | NVolSetErrors(vol); |
| 2761 | nr_freed = 0; |
| 2762 | } |
| 2763 | /* Truncate the runlist. */ |
| 2764 | err = ntfs_rl_truncate_nolock(vol, &ni->runlist, |
| 2765 | new_alloc_size >> vol->cluster_size_bits); |
| 2766 | /* |
| 2767 | * If the runlist truncation failed and/or the search context is no |
| 2768 | * longer valid, we cannot resize the attribute record or build the |
| 2769 | * mapping pairs array thus we mark the inode bad so that no access to |
| 2770 | * the freed clusters can happen. |
| 2771 | */ |
| 2772 | if (unlikely(err || IS_ERR(m))) { |
| 2773 | ntfs_error(vol->sb, "Failed to %s (error code %li).%s", |
| 2774 | IS_ERR(m) ? |
| 2775 | "restore attribute search context" : |
| 2776 | "truncate attribute runlist", |
| 2777 | IS_ERR(m) ? PTR_ERR(m) : err, es); |
| 2778 | err = -EIO; |
| 2779 | goto bad_out; |
| 2780 | } |
| 2781 | /* Get the size for the shrunk mapping pairs array for the runlist. */ |
| 2782 | mp_size = ntfs_get_size_for_mapping_pairs(vol, ni->runlist.rl, 0, -1); |
| 2783 | if (unlikely(mp_size <= 0)) { |
| 2784 | ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, " |
| 2785 | "attribute type 0x%x, because determining the " |
| 2786 | "size for the mapping pairs failed with error " |
| 2787 | "code %i.%s", vi->i_ino, |
| 2788 | (unsigned)le32_to_cpu(ni->type), mp_size, es); |
| 2789 | err = -EIO; |
| 2790 | goto bad_out; |
| 2791 | } |
| 2792 | /* |
| 2793 | * Shrink the attribute record for the new mapping pairs array. Note, |
| 2794 | * this cannot fail since we are making the attribute smaller thus by |
| 2795 | * definition there is enough space to do so. |
| 2796 | */ |
| 2797 | attr_len = le32_to_cpu(a->length); |
| 2798 | err = ntfs_attr_record_resize(m, a, mp_size + |
| 2799 | le16_to_cpu(a->data.non_resident.mapping_pairs_offset)); |
| 2800 | BUG_ON(err); |
| 2801 | /* |
| 2802 | * Generate the mapping pairs array directly into the attribute record. |
| 2803 | */ |
| 2804 | err = ntfs_mapping_pairs_build(vol, (u8*)a + |
| 2805 | le16_to_cpu(a->data.non_resident.mapping_pairs_offset), |
| 2806 | mp_size, ni->runlist.rl, 0, -1, NULL); |
| 2807 | if (unlikely(err)) { |
| 2808 | ntfs_error(vol->sb, "Cannot shrink allocation of inode 0x%lx, " |
| 2809 | "attribute type 0x%x, because building the " |
| 2810 | "mapping pairs failed with error code %i.%s", |
| 2811 | vi->i_ino, (unsigned)le32_to_cpu(ni->type), |
| 2812 | err, es); |
| 2813 | err = -EIO; |
| 2814 | goto bad_out; |
| 2815 | } |
| 2816 | /* Update the allocated/compressed size as well as the highest vcn. */ |
| 2817 | a->data.non_resident.highest_vcn = cpu_to_sle64((new_alloc_size >> |
| 2818 | vol->cluster_size_bits) - 1); |
| 2819 | write_lock_irqsave(&ni->size_lock, flags); |
| 2820 | ni->allocated_size = new_alloc_size; |
| 2821 | a->data.non_resident.allocated_size = cpu_to_sle64(new_alloc_size); |
| 2822 | if (NInoSparse(ni) || NInoCompressed(ni)) { |
| 2823 | if (nr_freed) { |
| 2824 | ni->itype.compressed.size -= nr_freed << |
| 2825 | vol->cluster_size_bits; |
| 2826 | BUG_ON(ni->itype.compressed.size < 0); |
| 2827 | a->data.non_resident.compressed_size = cpu_to_sle64( |
| 2828 | ni->itype.compressed.size); |
| 2829 | vi->i_blocks = ni->itype.compressed.size >> 9; |
| 2830 | } |
| 2831 | } else |
| 2832 | vi->i_blocks = new_alloc_size >> 9; |
| 2833 | write_unlock_irqrestore(&ni->size_lock, flags); |
| 2834 | /* |
| 2835 | * We have shrunk the allocation. If this is a shrinking truncate we |
| 2836 | * have already dealt with the initialized_size and the data_size above |
| 2837 | * and we are done. If the truncate is only changing the allocation |
| 2838 | * and not the data_size, we are also done. If this is an extending |
| 2839 | * truncate, need to extend the data_size now which is ensured by the |
| 2840 | * fact that @size_change is positive. |
| 2841 | */ |
| 2842 | alloc_done: |
| 2843 | /* |
| 2844 | * If the size is growing, need to update it now. If it is shrinking, |
| 2845 | * we have already updated it above (before the allocation change). |
| 2846 | */ |
| 2847 | if (size_change > 0) |
| 2848 | a->data.non_resident.data_size = cpu_to_sle64(new_size); |
| 2849 | /* Ensure the modified mft record is written out. */ |
| 2850 | flush_dcache_mft_record_page(ctx->ntfs_ino); |
| 2851 | mark_mft_record_dirty(ctx->ntfs_ino); |
| 2852 | unm_done: |
| 2853 | ntfs_attr_put_search_ctx(ctx); |
| 2854 | unmap_mft_record(base_ni); |
| 2855 | up_write(&ni->runlist.lock); |
| 2856 | done: |
| 2857 | /* Update the mtime and ctime on the base inode. */ |
Christoph Hellwig | 870f481 | 2006-01-09 20:52:01 -0800 | [diff] [blame] | 2858 | /* normally ->truncate shouldn't update ctime or mtime, |
| 2859 | * but ntfs did before so it got a copy & paste version |
| 2860 | * of file_update_time. one day someone should fix this |
| 2861 | * for real. |
| 2862 | */ |
| 2863 | if (!IS_NOCMTIME(VFS_I(base_ni)) && !IS_RDONLY(VFS_I(base_ni))) { |
| 2864 | struct timespec now = current_fs_time(VFS_I(base_ni)->i_sb); |
| 2865 | int sync_it = 0; |
| 2866 | |
| 2867 | if (!timespec_equal(&VFS_I(base_ni)->i_mtime, &now) || |
| 2868 | !timespec_equal(&VFS_I(base_ni)->i_ctime, &now)) |
| 2869 | sync_it = 1; |
| 2870 | VFS_I(base_ni)->i_mtime = now; |
| 2871 | VFS_I(base_ni)->i_ctime = now; |
| 2872 | |
| 2873 | if (sync_it) |
| 2874 | mark_inode_dirty_sync(VFS_I(base_ni)); |
| 2875 | } |
| 2876 | |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2877 | if (likely(!err)) { |
| 2878 | NInoClearTruncateFailed(ni); |
| 2879 | ntfs_debug("Done."); |
| 2880 | } |
| 2881 | return err; |
| 2882 | old_bad_out: |
| 2883 | old_size = -1; |
| 2884 | bad_out: |
Anton Altaparmakov | a778f21 | 2006-03-23 16:18:23 +0000 | [diff] [blame] | 2885 | if (err != -ENOMEM && err != -EOPNOTSUPP) |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2886 | NVolSetErrors(vol); |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2887 | if (err != -EOPNOTSUPP) |
| 2888 | NInoSetTruncateFailed(ni); |
| 2889 | else if (old_size >= 0) |
| 2890 | i_size_write(vi, old_size); |
| 2891 | err_out: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2892 | if (ctx) |
| 2893 | ntfs_attr_put_search_ctx(ctx); |
| 2894 | if (m) |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2895 | unmap_mft_record(base_ni); |
| 2896 | up_write(&ni->runlist.lock); |
| 2897 | out: |
| 2898 | ntfs_debug("Failed. Returning error code %i.", err); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2899 | return err; |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2900 | conv_err_out: |
Anton Altaparmakov | a778f21 | 2006-03-23 16:18:23 +0000 | [diff] [blame] | 2901 | if (err != -ENOMEM && err != -EOPNOTSUPP) |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2902 | NVolSetErrors(vol); |
Anton Altaparmakov | dd07233 | 2005-10-04 15:39:02 +0100 | [diff] [blame] | 2903 | if (err != -EOPNOTSUPP) |
| 2904 | NInoSetTruncateFailed(ni); |
| 2905 | else |
| 2906 | i_size_write(vi, old_size); |
| 2907 | goto out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2908 | } |
| 2909 | |
| 2910 | /** |
| 2911 | * ntfs_truncate_vfs - wrapper for ntfs_truncate() that has no return value |
| 2912 | * @vi: inode for which the i_size was changed |
| 2913 | * |
| 2914 | * Wrapper for ntfs_truncate() that has no return value. |
| 2915 | * |
| 2916 | * See ntfs_truncate() description above for details. |
| 2917 | */ |
| 2918 | void ntfs_truncate_vfs(struct inode *vi) { |
| 2919 | ntfs_truncate(vi); |
| 2920 | } |
| 2921 | |
| 2922 | /** |
| 2923 | * ntfs_setattr - called from notify_change() when an attribute is being changed |
| 2924 | * @dentry: dentry whose attributes to change |
| 2925 | * @attr: structure describing the attributes and the changes |
| 2926 | * |
| 2927 | * We have to trap VFS attempts to truncate the file described by @dentry as |
| 2928 | * soon as possible, because we do not implement changes in i_size yet. So we |
| 2929 | * abort all i_size changes here. |
| 2930 | * |
| 2931 | * We also abort all changes of user, group, and mode as we do not implement |
| 2932 | * the NTFS ACLs yet. |
| 2933 | * |
Jes Sorensen | 1b1dcc1 | 2006-01-09 15:59:24 -0800 | [diff] [blame] | 2934 | * Called with ->i_mutex held. For the ATTR_SIZE (i.e. ->truncate) case, also |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2935 | * called with ->i_alloc_sem held for writing. |
| 2936 | * |
| 2937 | * Basically this is a copy of generic notify_change() and inode_setattr() |
| 2938 | * functionality, except we intercept and abort changes in i_size. |
| 2939 | */ |
| 2940 | int ntfs_setattr(struct dentry *dentry, struct iattr *attr) |
| 2941 | { |
| 2942 | struct inode *vi = dentry->d_inode; |
| 2943 | int err; |
| 2944 | unsigned int ia_valid = attr->ia_valid; |
| 2945 | |
| 2946 | err = inode_change_ok(vi, attr); |
| 2947 | if (err) |
Anton Altaparmakov | e943825 | 2005-10-04 16:01:06 +0100 | [diff] [blame] | 2948 | goto out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2949 | /* We do not support NTFS ACLs yet. */ |
| 2950 | if (ia_valid & (ATTR_UID | ATTR_GID | ATTR_MODE)) { |
| 2951 | ntfs_warning(vi->i_sb, "Changes in user/group/mode are not " |
| 2952 | "supported yet, ignoring."); |
| 2953 | err = -EOPNOTSUPP; |
| 2954 | goto out; |
| 2955 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2956 | if (ia_valid & ATTR_SIZE) { |
| 2957 | if (attr->ia_size != i_size_read(vi)) { |
Anton Altaparmakov | e943825 | 2005-10-04 16:01:06 +0100 | [diff] [blame] | 2958 | ntfs_inode *ni = NTFS_I(vi); |
| 2959 | /* |
| 2960 | * FIXME: For now we do not support resizing of |
| 2961 | * compressed or encrypted files yet. |
| 2962 | */ |
| 2963 | if (NInoCompressed(ni) || NInoEncrypted(ni)) { |
| 2964 | ntfs_warning(vi->i_sb, "Changes in inode size " |
| 2965 | "are not supported yet for " |
| 2966 | "%s files, ignoring.", |
| 2967 | NInoCompressed(ni) ? |
| 2968 | "compressed" : "encrypted"); |
| 2969 | err = -EOPNOTSUPP; |
| 2970 | } else |
| 2971 | err = vmtruncate(vi, attr->ia_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2972 | if (err || ia_valid == ATTR_SIZE) |
| 2973 | goto out; |
| 2974 | } else { |
| 2975 | /* |
| 2976 | * We skipped the truncate but must still update |
| 2977 | * timestamps. |
| 2978 | */ |
Anton Altaparmakov | 1c7d469 | 2005-09-08 21:15:09 +0100 | [diff] [blame] | 2979 | ia_valid |= ATTR_MTIME | ATTR_CTIME; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2980 | } |
| 2981 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2982 | if (ia_valid & ATTR_ATIME) |
Anton Altaparmakov | 1c7d469 | 2005-09-08 21:15:09 +0100 | [diff] [blame] | 2983 | vi->i_atime = timespec_trunc(attr->ia_atime, |
| 2984 | vi->i_sb->s_time_gran); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2985 | if (ia_valid & ATTR_MTIME) |
Anton Altaparmakov | 1c7d469 | 2005-09-08 21:15:09 +0100 | [diff] [blame] | 2986 | vi->i_mtime = timespec_trunc(attr->ia_mtime, |
| 2987 | vi->i_sb->s_time_gran); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2988 | if (ia_valid & ATTR_CTIME) |
Anton Altaparmakov | 1c7d469 | 2005-09-08 21:15:09 +0100 | [diff] [blame] | 2989 | vi->i_ctime = timespec_trunc(attr->ia_ctime, |
| 2990 | vi->i_sb->s_time_gran); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2991 | mark_inode_dirty(vi); |
| 2992 | out: |
| 2993 | return err; |
| 2994 | } |
| 2995 | |
| 2996 | /** |
| 2997 | * ntfs_write_inode - write out a dirty inode |
| 2998 | * @vi: inode to write out |
| 2999 | * @sync: if true, write out synchronously |
| 3000 | * |
| 3001 | * Write out a dirty inode to disk including any extent inodes if present. |
| 3002 | * |
| 3003 | * If @sync is true, commit the inode to disk and wait for io completion. This |
| 3004 | * is done using write_mft_record(). |
| 3005 | * |
| 3006 | * If @sync is false, just schedule the write to happen but do not wait for i/o |
| 3007 | * completion. In 2.6 kernels, scheduling usually happens just by virtue of |
| 3008 | * marking the page (and in this case mft record) dirty but we do not implement |
| 3009 | * this yet as write_mft_record() largely ignores the @sync parameter and |
| 3010 | * always performs synchronous writes. |
| 3011 | * |
| 3012 | * Return 0 on success and -errno on error. |
| 3013 | */ |
| 3014 | int ntfs_write_inode(struct inode *vi, int sync) |
| 3015 | { |
| 3016 | sle64 nt; |
| 3017 | ntfs_inode *ni = NTFS_I(vi); |
| 3018 | ntfs_attr_search_ctx *ctx; |
| 3019 | MFT_RECORD *m; |
| 3020 | STANDARD_INFORMATION *si; |
| 3021 | int err = 0; |
| 3022 | BOOL modified = FALSE; |
| 3023 | |
| 3024 | ntfs_debug("Entering for %sinode 0x%lx.", NInoAttr(ni) ? "attr " : "", |
| 3025 | vi->i_ino); |
| 3026 | /* |
| 3027 | * Dirty attribute inodes are written via their real inodes so just |
| 3028 | * clean them here. Access time updates are taken care off when the |
| 3029 | * real inode is written. |
| 3030 | */ |
| 3031 | if (NInoAttr(ni)) { |
| 3032 | NInoClearDirty(ni); |
| 3033 | ntfs_debug("Done."); |
| 3034 | return 0; |
| 3035 | } |
| 3036 | /* Map, pin, and lock the mft record belonging to the inode. */ |
| 3037 | m = map_mft_record(ni); |
| 3038 | if (IS_ERR(m)) { |
| 3039 | err = PTR_ERR(m); |
| 3040 | goto err_out; |
| 3041 | } |
| 3042 | /* Update the access times in the standard information attribute. */ |
| 3043 | ctx = ntfs_attr_get_search_ctx(ni, m); |
| 3044 | if (unlikely(!ctx)) { |
| 3045 | err = -ENOMEM; |
| 3046 | goto unm_err_out; |
| 3047 | } |
| 3048 | err = ntfs_attr_lookup(AT_STANDARD_INFORMATION, NULL, 0, |
| 3049 | CASE_SENSITIVE, 0, NULL, 0, ctx); |
| 3050 | if (unlikely(err)) { |
| 3051 | ntfs_attr_put_search_ctx(ctx); |
| 3052 | goto unm_err_out; |
| 3053 | } |
| 3054 | si = (STANDARD_INFORMATION*)((u8*)ctx->attr + |
| 3055 | le16_to_cpu(ctx->attr->data.resident.value_offset)); |
| 3056 | /* Update the access times if they have changed. */ |
| 3057 | nt = utc2ntfs(vi->i_mtime); |
| 3058 | if (si->last_data_change_time != nt) { |
| 3059 | ntfs_debug("Updating mtime for inode 0x%lx: old = 0x%llx, " |
Randy Dunlap | 8907547 | 2005-03-03 11:19:53 +0000 | [diff] [blame] | 3060 | "new = 0x%llx", vi->i_ino, (long long) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3061 | sle64_to_cpu(si->last_data_change_time), |
Randy Dunlap | 8907547 | 2005-03-03 11:19:53 +0000 | [diff] [blame] | 3062 | (long long)sle64_to_cpu(nt)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3063 | si->last_data_change_time = nt; |
| 3064 | modified = TRUE; |
| 3065 | } |
| 3066 | nt = utc2ntfs(vi->i_ctime); |
| 3067 | if (si->last_mft_change_time != nt) { |
| 3068 | ntfs_debug("Updating ctime for inode 0x%lx: old = 0x%llx, " |
Randy Dunlap | 8907547 | 2005-03-03 11:19:53 +0000 | [diff] [blame] | 3069 | "new = 0x%llx", vi->i_ino, (long long) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3070 | sle64_to_cpu(si->last_mft_change_time), |
Randy Dunlap | 8907547 | 2005-03-03 11:19:53 +0000 | [diff] [blame] | 3071 | (long long)sle64_to_cpu(nt)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3072 | si->last_mft_change_time = nt; |
| 3073 | modified = TRUE; |
| 3074 | } |
| 3075 | nt = utc2ntfs(vi->i_atime); |
| 3076 | if (si->last_access_time != nt) { |
| 3077 | ntfs_debug("Updating atime for inode 0x%lx: old = 0x%llx, " |
| 3078 | "new = 0x%llx", vi->i_ino, |
Randy Dunlap | 8907547 | 2005-03-03 11:19:53 +0000 | [diff] [blame] | 3079 | (long long)sle64_to_cpu(si->last_access_time), |
| 3080 | (long long)sle64_to_cpu(nt)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3081 | si->last_access_time = nt; |
| 3082 | modified = TRUE; |
| 3083 | } |
| 3084 | /* |
| 3085 | * If we just modified the standard information attribute we need to |
| 3086 | * mark the mft record it is in dirty. We do this manually so that |
| 3087 | * mark_inode_dirty() is not called which would redirty the inode and |
| 3088 | * hence result in an infinite loop of trying to write the inode. |
| 3089 | * There is no need to mark the base inode nor the base mft record |
| 3090 | * dirty, since we are going to write this mft record below in any case |
| 3091 | * and the base mft record may actually not have been modified so it |
| 3092 | * might not need to be written out. |
| 3093 | * NOTE: It is not a problem when the inode for $MFT itself is being |
| 3094 | * written out as mark_ntfs_record_dirty() will only set I_DIRTY_PAGES |
| 3095 | * on the $MFT inode and hence ntfs_write_inode() will not be |
| 3096 | * re-invoked because of it which in turn is ok since the dirtied mft |
| 3097 | * record will be cleaned and written out to disk below, i.e. before |
| 3098 | * this function returns. |
| 3099 | */ |
Anton Altaparmakov | 20fdcf1 | 2006-03-23 16:21:02 +0000 | [diff] [blame] | 3100 | if (modified) { |
| 3101 | flush_dcache_mft_record_page(ctx->ntfs_ino); |
Ingo Molnar | 4e5e529 | 2006-03-23 16:57:48 +0000 | [diff] [blame] | 3102 | if (!NInoTestSetDirty(ctx->ntfs_ino)) |
Anton Altaparmakov | 20fdcf1 | 2006-03-23 16:21:02 +0000 | [diff] [blame] | 3103 | mark_ntfs_record_dirty(ctx->ntfs_ino->page, |
| 3104 | ctx->ntfs_ino->page_ofs); |
| 3105 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3106 | ntfs_attr_put_search_ctx(ctx); |
| 3107 | /* Now the access times are updated, write the base mft record. */ |
| 3108 | if (NInoDirty(ni)) |
| 3109 | err = write_mft_record(ni, m, sync); |
| 3110 | /* Write all attached extent mft records. */ |
Ingo Molnar | 4e5e529 | 2006-03-23 16:57:48 +0000 | [diff] [blame] | 3111 | mutex_lock(&ni->extent_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3112 | if (ni->nr_extents > 0) { |
| 3113 | ntfs_inode **extent_nis = ni->ext.extent_ntfs_inos; |
| 3114 | int i; |
| 3115 | |
| 3116 | ntfs_debug("Writing %i extent inodes.", ni->nr_extents); |
| 3117 | for (i = 0; i < ni->nr_extents; i++) { |
| 3118 | ntfs_inode *tni = extent_nis[i]; |
| 3119 | |
| 3120 | if (NInoDirty(tni)) { |
| 3121 | MFT_RECORD *tm = map_mft_record(tni); |
| 3122 | int ret; |
| 3123 | |
| 3124 | if (IS_ERR(tm)) { |
| 3125 | if (!err || err == -ENOMEM) |
| 3126 | err = PTR_ERR(tm); |
| 3127 | continue; |
| 3128 | } |
| 3129 | ret = write_mft_record(tni, tm, sync); |
| 3130 | unmap_mft_record(tni); |
| 3131 | if (unlikely(ret)) { |
| 3132 | if (!err || err == -ENOMEM) |
| 3133 | err = ret; |
| 3134 | } |
| 3135 | } |
| 3136 | } |
| 3137 | } |
Ingo Molnar | 4e5e529 | 2006-03-23 16:57:48 +0000 | [diff] [blame] | 3138 | mutex_unlock(&ni->extent_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3139 | unmap_mft_record(ni); |
| 3140 | if (unlikely(err)) |
| 3141 | goto err_out; |
| 3142 | ntfs_debug("Done."); |
| 3143 | return 0; |
| 3144 | unm_err_out: |
| 3145 | unmap_mft_record(ni); |
| 3146 | err_out: |
| 3147 | if (err == -ENOMEM) { |
| 3148 | ntfs_warning(vi->i_sb, "Not enough memory to write inode. " |
| 3149 | "Marking the inode dirty again, so the VFS " |
| 3150 | "retries later."); |
| 3151 | mark_inode_dirty(vi); |
| 3152 | } else { |
Anton Altaparmakov | a778f21 | 2006-03-23 16:18:23 +0000 | [diff] [blame] | 3153 | ntfs_error(vi->i_sb, "Failed (error %i): Run chkdsk.", -err); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3154 | NVolSetErrors(ni->vol); |
| 3155 | } |
| 3156 | return err; |
| 3157 | } |
| 3158 | |
| 3159 | #endif /* NTFS_RW */ |