1 /* 2 * This file is part of UBIFS. 3 * 4 * Copyright (C) 2006-2008 Nokia Corporation. 5 * 6 * This program is free software; you can redistribute it and/or modify it 7 * under the terms of the GNU General Public License version 2 as published by 8 * the Free Software Foundation. 9 * 10 * This program is distributed in the hope that it will be useful, but WITHOUT 11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for 13 * more details. 14 * 15 * You should have received a copy of the GNU General Public License along with 16 * this program; if not, write to the Free Software Foundation, Inc., 51 17 * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA 18 * 19 * Authors: Artem Bityutskiy (Битюцкий Артём) 20 * Adrian Hunter 21 */ 22 23 /* 24 * This file implements UBIFS initialization and VFS superblock operations. Some 25 * initialization stuff which is rather large and complex is placed at 26 * corresponding subsystems, but most of it is here. 27 */ 28 29 #include <linux/init.h> 30 #include <linux/slab.h> 31 #include <linux/module.h> 32 #include <linux/ctype.h> 33 #include <linux/kthread.h> 34 #include <linux/parser.h> 35 #include <linux/seq_file.h> 36 #include <linux/mount.h> 37 #include <linux/math64.h> 38 #include <linux/writeback.h> 39 #include "ubifs.h" 40 41 /* 42 * Maximum amount of memory we may 'kmalloc()' without worrying that we are 43 * allocating too much. 44 */ 45 #define UBIFS_KMALLOC_OK (128*1024) 46 47 /* Slab cache for UBIFS inodes */ 48 static struct kmem_cache *ubifs_inode_slab; 49 50 /* UBIFS TNC shrinker description */ 51 static struct shrinker ubifs_shrinker_info = { 52 .scan_objects = ubifs_shrink_scan, 53 .count_objects = ubifs_shrink_count, 54 .seeks = DEFAULT_SEEKS, 55 }; 56 57 /** 58 * validate_inode - validate inode. 59 * @c: UBIFS file-system description object 60 * @inode: the inode to validate 61 * 62 * This is a helper function for 'ubifs_iget()' which validates various fields 63 * of a newly built inode to make sure they contain sane values and prevent 64 * possible vulnerabilities. Returns zero if the inode is all right and 65 * a non-zero error code if not. 66 */ 67 static int validate_inode(struct ubifs_info *c, const struct inode *inode) 68 { 69 int err; 70 const struct ubifs_inode *ui = ubifs_inode(inode); 71 72 if (inode->i_size > c->max_inode_sz) { 73 ubifs_err(c, "inode is too large (%lld)", 74 (long long)inode->i_size); 75 return 1; 76 } 77 78 if (ui->compr_type >= UBIFS_COMPR_TYPES_CNT) { 79 ubifs_err(c, "unknown compression type %d", ui->compr_type); 80 return 2; 81 } 82 83 if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX) 84 return 3; 85 86 if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA) 87 return 4; 88 89 if (ui->xattr && !S_ISREG(inode->i_mode)) 90 return 5; 91 92 if (!ubifs_compr_present(ui->compr_type)) { 93 ubifs_warn(c, "inode %lu uses '%s' compression, but it was not compiled in", 94 inode->i_ino, ubifs_compr_name(ui->compr_type)); 95 } 96 97 err = dbg_check_dir(c, inode); 98 return err; 99 } 100 101 struct inode *ubifs_iget(struct super_block *sb, unsigned long inum) 102 { 103 int err; 104 union ubifs_key key; 105 struct ubifs_ino_node *ino; 106 struct ubifs_info *c = sb->s_fs_info; 107 struct inode *inode; 108 struct ubifs_inode *ui; 109 110 dbg_gen("inode %lu", inum); 111 112 inode = iget_locked(sb, inum); 113 if (!inode) 114 return ERR_PTR(-ENOMEM); 115 if (!(inode->i_state & I_NEW)) 116 return inode; 117 ui = ubifs_inode(inode); 118 119 ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS); 120 if (!ino) { 121 err = -ENOMEM; 122 goto out; 123 } 124 125 ino_key_init(c, &key, inode->i_ino); 126 127 err = ubifs_tnc_lookup(c, &key, ino); 128 if (err) 129 goto out_ino; 130 131 inode->i_flags |= S_NOCMTIME; 132 #ifndef CONFIG_UBIFS_ATIME_SUPPORT 133 inode->i_flags |= S_NOATIME; 134 #endif 135 set_nlink(inode, le32_to_cpu(ino->nlink)); 136 i_uid_write(inode, le32_to_cpu(ino->uid)); 137 i_gid_write(inode, le32_to_cpu(ino->gid)); 138 inode->i_atime.tv_sec = (int64_t)le64_to_cpu(ino->atime_sec); 139 inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec); 140 inode->i_mtime.tv_sec = (int64_t)le64_to_cpu(ino->mtime_sec); 141 inode->i_mtime.tv_nsec = le32_to_cpu(ino->mtime_nsec); 142 inode->i_ctime.tv_sec = (int64_t)le64_to_cpu(ino->ctime_sec); 143 inode->i_ctime.tv_nsec = le32_to_cpu(ino->ctime_nsec); 144 inode->i_mode = le32_to_cpu(ino->mode); 145 inode->i_size = le64_to_cpu(ino->size); 146 147 ui->data_len = le32_to_cpu(ino->data_len); 148 ui->flags = le32_to_cpu(ino->flags); 149 ui->compr_type = le16_to_cpu(ino->compr_type); 150 ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum); 151 ui->xattr_cnt = le32_to_cpu(ino->xattr_cnt); 152 ui->xattr_size = le32_to_cpu(ino->xattr_size); 153 ui->xattr_names = le32_to_cpu(ino->xattr_names); 154 ui->synced_i_size = ui->ui_size = inode->i_size; 155 156 ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0; 157 158 err = validate_inode(c, inode); 159 if (err) 160 goto out_invalid; 161 162 switch (inode->i_mode & S_IFMT) { 163 case S_IFREG: 164 inode->i_mapping->a_ops = &ubifs_file_address_operations; 165 inode->i_op = &ubifs_file_inode_operations; 166 inode->i_fop = &ubifs_file_operations; 167 if (ui->xattr) { 168 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS); 169 if (!ui->data) { 170 err = -ENOMEM; 171 goto out_ino; 172 } 173 memcpy(ui->data, ino->data, ui->data_len); 174 ((char *)ui->data)[ui->data_len] = '\0'; 175 } else if (ui->data_len != 0) { 176 err = 10; 177 goto out_invalid; 178 } 179 break; 180 case S_IFDIR: 181 inode->i_op = &ubifs_dir_inode_operations; 182 inode->i_fop = &ubifs_dir_operations; 183 if (ui->data_len != 0) { 184 err = 11; 185 goto out_invalid; 186 } 187 break; 188 case S_IFLNK: 189 inode->i_op = &ubifs_symlink_inode_operations; 190 if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) { 191 err = 12; 192 goto out_invalid; 193 } 194 ui->data = kmalloc(ui->data_len + 1, GFP_NOFS); 195 if (!ui->data) { 196 err = -ENOMEM; 197 goto out_ino; 198 } 199 memcpy(ui->data, ino->data, ui->data_len); 200 ((char *)ui->data)[ui->data_len] = '\0'; 201 break; 202 case S_IFBLK: 203 case S_IFCHR: 204 { 205 dev_t rdev; 206 union ubifs_dev_desc *dev; 207 208 ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS); 209 if (!ui->data) { 210 err = -ENOMEM; 211 goto out_ino; 212 } 213 214 dev = (union ubifs_dev_desc *)ino->data; 215 if (ui->data_len == sizeof(dev->new)) 216 rdev = new_decode_dev(le32_to_cpu(dev->new)); 217 else if (ui->data_len == sizeof(dev->huge)) 218 rdev = huge_decode_dev(le64_to_cpu(dev->huge)); 219 else { 220 err = 13; 221 goto out_invalid; 222 } 223 memcpy(ui->data, ino->data, ui->data_len); 224 inode->i_op = &ubifs_file_inode_operations; 225 init_special_inode(inode, inode->i_mode, rdev); 226 break; 227 } 228 case S_IFSOCK: 229 case S_IFIFO: 230 inode->i_op = &ubifs_file_inode_operations; 231 init_special_inode(inode, inode->i_mode, 0); 232 if (ui->data_len != 0) { 233 err = 14; 234 goto out_invalid; 235 } 236 break; 237 default: 238 err = 15; 239 goto out_invalid; 240 } 241 242 kfree(ino); 243 ubifs_set_inode_flags(inode); 244 unlock_new_inode(inode); 245 return inode; 246 247 out_invalid: 248 ubifs_err(c, "inode %lu validation failed, error %d", inode->i_ino, err); 249 ubifs_dump_node(c, ino); 250 ubifs_dump_inode(c, inode); 251 err = -EINVAL; 252 out_ino: 253 kfree(ino); 254 out: 255 ubifs_err(c, "failed to read inode %lu, error %d", inode->i_ino, err); 256 iget_failed(inode); 257 return ERR_PTR(err); 258 } 259 260 static struct inode *ubifs_alloc_inode(struct super_block *sb) 261 { 262 struct ubifs_inode *ui; 263 264 ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS); 265 if (!ui) 266 return NULL; 267 268 memset((void *)ui + sizeof(struct inode), 0, 269 sizeof(struct ubifs_inode) - sizeof(struct inode)); 270 mutex_init(&ui->ui_mutex); 271 spin_lock_init(&ui->ui_lock); 272 return &ui->vfs_inode; 273 }; 274 275 static void ubifs_i_callback(struct rcu_head *head) 276 { 277 struct inode *inode = container_of(head, struct inode, i_rcu); 278 struct ubifs_inode *ui = ubifs_inode(inode); 279 kmem_cache_free(ubifs_inode_slab, ui); 280 } 281 282 static void ubifs_destroy_inode(struct inode *inode) 283 { 284 struct ubifs_inode *ui = ubifs_inode(inode); 285 286 kfree(ui->data); 287 call_rcu(&inode->i_rcu, ubifs_i_callback); 288 } 289 290 /* 291 * Note, Linux write-back code calls this without 'i_mutex'. 292 */ 293 static int ubifs_write_inode(struct inode *inode, struct writeback_control *wbc) 294 { 295 int err = 0; 296 struct ubifs_info *c = inode->i_sb->s_fs_info; 297 struct ubifs_inode *ui = ubifs_inode(inode); 298 299 ubifs_assert(!ui->xattr); 300 if (is_bad_inode(inode)) 301 return 0; 302 303 mutex_lock(&ui->ui_mutex); 304 /* 305 * Due to races between write-back forced by budgeting 306 * (see 'sync_some_inodes()') and background write-back, the inode may 307 * have already been synchronized, do not do this again. This might 308 * also happen if it was synchronized in an VFS operation, e.g. 309 * 'ubifs_link()'. 310 */ 311 if (!ui->dirty) { 312 mutex_unlock(&ui->ui_mutex); 313 return 0; 314 } 315 316 /* 317 * As an optimization, do not write orphan inodes to the media just 318 * because this is not needed. 319 */ 320 dbg_gen("inode %lu, mode %#x, nlink %u", 321 inode->i_ino, (int)inode->i_mode, inode->i_nlink); 322 if (inode->i_nlink) { 323 err = ubifs_jnl_write_inode(c, inode); 324 if (err) 325 ubifs_err(c, "can't write inode %lu, error %d", 326 inode->i_ino, err); 327 else 328 err = dbg_check_inode_size(c, inode, ui->ui_size); 329 } 330 331 ui->dirty = 0; 332 mutex_unlock(&ui->ui_mutex); 333 ubifs_release_dirty_inode_budget(c, ui); 334 return err; 335 } 336 337 static void ubifs_evict_inode(struct inode *inode) 338 { 339 int err; 340 struct ubifs_info *c = inode->i_sb->s_fs_info; 341 struct ubifs_inode *ui = ubifs_inode(inode); 342 343 if (ui->xattr) 344 /* 345 * Extended attribute inode deletions are fully handled in 346 * 'ubifs_removexattr()'. These inodes are special and have 347 * limited usage, so there is nothing to do here. 348 */ 349 goto out; 350 351 dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode); 352 ubifs_assert(!atomic_read(&inode->i_count)); 353 354 truncate_inode_pages_final(&inode->i_data); 355 356 if (inode->i_nlink) 357 goto done; 358 359 if (is_bad_inode(inode)) 360 goto out; 361 362 ui->ui_size = inode->i_size = 0; 363 err = ubifs_jnl_delete_inode(c, inode); 364 if (err) 365 /* 366 * Worst case we have a lost orphan inode wasting space, so a 367 * simple error message is OK here. 368 */ 369 ubifs_err(c, "can't delete inode %lu, error %d", 370 inode->i_ino, err); 371 372 out: 373 if (ui->dirty) 374 ubifs_release_dirty_inode_budget(c, ui); 375 else { 376 /* We've deleted something - clean the "no space" flags */ 377 c->bi.nospace = c->bi.nospace_rp = 0; 378 smp_wmb(); 379 } 380 done: 381 clear_inode(inode); 382 #ifdef CONFIG_UBIFS_FS_ENCRYPTION 383 fscrypt_put_encryption_info(inode, NULL); 384 #endif 385 } 386 387 static void ubifs_dirty_inode(struct inode *inode, int flags) 388 { 389 struct ubifs_inode *ui = ubifs_inode(inode); 390 391 ubifs_assert(mutex_is_locked(&ui->ui_mutex)); 392 if (!ui->dirty) { 393 ui->dirty = 1; 394 dbg_gen("inode %lu", inode->i_ino); 395 } 396 } 397 398 static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf) 399 { 400 struct ubifs_info *c = dentry->d_sb->s_fs_info; 401 unsigned long long free; 402 __le32 *uuid = (__le32 *)c->uuid; 403 404 free = ubifs_get_free_space(c); 405 dbg_gen("free space %lld bytes (%lld blocks)", 406 free, free >> UBIFS_BLOCK_SHIFT); 407 408 buf->f_type = UBIFS_SUPER_MAGIC; 409 buf->f_bsize = UBIFS_BLOCK_SIZE; 410 buf->f_blocks = c->block_cnt; 411 buf->f_bfree = free >> UBIFS_BLOCK_SHIFT; 412 if (free > c->report_rp_size) 413 buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT; 414 else 415 buf->f_bavail = 0; 416 buf->f_files = 0; 417 buf->f_ffree = 0; 418 buf->f_namelen = UBIFS_MAX_NLEN; 419 buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]); 420 buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]); 421 ubifs_assert(buf->f_bfree <= c->block_cnt); 422 return 0; 423 } 424 425 static int ubifs_show_options(struct seq_file *s, struct dentry *root) 426 { 427 struct ubifs_info *c = root->d_sb->s_fs_info; 428 429 if (c->mount_opts.unmount_mode == 2) 430 seq_puts(s, ",fast_unmount"); 431 else if (c->mount_opts.unmount_mode == 1) 432 seq_puts(s, ",norm_unmount"); 433 434 if (c->mount_opts.bulk_read == 2) 435 seq_puts(s, ",bulk_read"); 436 else if (c->mount_opts.bulk_read == 1) 437 seq_puts(s, ",no_bulk_read"); 438 439 if (c->mount_opts.chk_data_crc == 2) 440 seq_puts(s, ",chk_data_crc"); 441 else if (c->mount_opts.chk_data_crc == 1) 442 seq_puts(s, ",no_chk_data_crc"); 443 444 if (c->mount_opts.override_compr) { 445 seq_printf(s, ",compr=%s", 446 ubifs_compr_name(c->mount_opts.compr_type)); 447 } 448 449 seq_printf(s, ",ubi=%d,vol=%d", c->vi.ubi_num, c->vi.vol_id); 450 451 return 0; 452 } 453 454 static int ubifs_sync_fs(struct super_block *sb, int wait) 455 { 456 int i, err; 457 struct ubifs_info *c = sb->s_fs_info; 458 459 /* 460 * Zero @wait is just an advisory thing to help the file system shove 461 * lots of data into the queues, and there will be the second 462 * '->sync_fs()' call, with non-zero @wait. 463 */ 464 if (!wait) 465 return 0; 466 467 /* 468 * Synchronize write buffers, because 'ubifs_run_commit()' does not 469 * do this if it waits for an already running commit. 470 */ 471 for (i = 0; i < c->jhead_cnt; i++) { 472 err = ubifs_wbuf_sync(&c->jheads[i].wbuf); 473 if (err) 474 return err; 475 } 476 477 /* 478 * Strictly speaking, it is not necessary to commit the journal here, 479 * synchronizing write-buffers would be enough. But committing makes 480 * UBIFS free space predictions much more accurate, so we want to let 481 * the user be able to get more accurate results of 'statfs()' after 482 * they synchronize the file system. 483 */ 484 err = ubifs_run_commit(c); 485 if (err) 486 return err; 487 488 return ubi_sync(c->vi.ubi_num); 489 } 490 491 /** 492 * init_constants_early - initialize UBIFS constants. 493 * @c: UBIFS file-system description object 494 * 495 * This function initialize UBIFS constants which do not need the superblock to 496 * be read. It also checks that the UBI volume satisfies basic UBIFS 497 * requirements. Returns zero in case of success and a negative error code in 498 * case of failure. 499 */ 500 static int init_constants_early(struct ubifs_info *c) 501 { 502 if (c->vi.corrupted) { 503 ubifs_warn(c, "UBI volume is corrupted - read-only mode"); 504 c->ro_media = 1; 505 } 506 507 if (c->di.ro_mode) { 508 ubifs_msg(c, "read-only UBI device"); 509 c->ro_media = 1; 510 } 511 512 if (c->vi.vol_type == UBI_STATIC_VOLUME) { 513 ubifs_msg(c, "static UBI volume - read-only mode"); 514 c->ro_media = 1; 515 } 516 517 c->leb_cnt = c->vi.size; 518 c->leb_size = c->vi.usable_leb_size; 519 c->leb_start = c->di.leb_start; 520 c->half_leb_size = c->leb_size / 2; 521 c->min_io_size = c->di.min_io_size; 522 c->min_io_shift = fls(c->min_io_size) - 1; 523 c->max_write_size = c->di.max_write_size; 524 c->max_write_shift = fls(c->max_write_size) - 1; 525 526 if (c->leb_size < UBIFS_MIN_LEB_SZ) { 527 ubifs_errc(c, "too small LEBs (%d bytes), min. is %d bytes", 528 c->leb_size, UBIFS_MIN_LEB_SZ); 529 return -EINVAL; 530 } 531 532 if (c->leb_cnt < UBIFS_MIN_LEB_CNT) { 533 ubifs_errc(c, "too few LEBs (%d), min. is %d", 534 c->leb_cnt, UBIFS_MIN_LEB_CNT); 535 return -EINVAL; 536 } 537 538 if (!is_power_of_2(c->min_io_size)) { 539 ubifs_errc(c, "bad min. I/O size %d", c->min_io_size); 540 return -EINVAL; 541 } 542 543 /* 544 * Maximum write size has to be greater or equivalent to min. I/O 545 * size, and be multiple of min. I/O size. 546 */ 547 if (c->max_write_size < c->min_io_size || 548 c->max_write_size % c->min_io_size || 549 !is_power_of_2(c->max_write_size)) { 550 ubifs_errc(c, "bad write buffer size %d for %d min. I/O unit", 551 c->max_write_size, c->min_io_size); 552 return -EINVAL; 553 } 554 555 /* 556 * UBIFS aligns all node to 8-byte boundary, so to make function in 557 * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is 558 * less than 8. 559 */ 560 if (c->min_io_size < 8) { 561 c->min_io_size = 8; 562 c->min_io_shift = 3; 563 if (c->max_write_size < c->min_io_size) { 564 c->max_write_size = c->min_io_size; 565 c->max_write_shift = c->min_io_shift; 566 } 567 } 568 569 c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size); 570 c->mst_node_alsz = ALIGN(UBIFS_MST_NODE_SZ, c->min_io_size); 571 572 /* 573 * Initialize node length ranges which are mostly needed for node 574 * length validation. 575 */ 576 c->ranges[UBIFS_PAD_NODE].len = UBIFS_PAD_NODE_SZ; 577 c->ranges[UBIFS_SB_NODE].len = UBIFS_SB_NODE_SZ; 578 c->ranges[UBIFS_MST_NODE].len = UBIFS_MST_NODE_SZ; 579 c->ranges[UBIFS_REF_NODE].len = UBIFS_REF_NODE_SZ; 580 c->ranges[UBIFS_TRUN_NODE].len = UBIFS_TRUN_NODE_SZ; 581 c->ranges[UBIFS_CS_NODE].len = UBIFS_CS_NODE_SZ; 582 583 c->ranges[UBIFS_INO_NODE].min_len = UBIFS_INO_NODE_SZ; 584 c->ranges[UBIFS_INO_NODE].max_len = UBIFS_MAX_INO_NODE_SZ; 585 c->ranges[UBIFS_ORPH_NODE].min_len = 586 UBIFS_ORPH_NODE_SZ + sizeof(__le64); 587 c->ranges[UBIFS_ORPH_NODE].max_len = c->leb_size; 588 c->ranges[UBIFS_DENT_NODE].min_len = UBIFS_DENT_NODE_SZ; 589 c->ranges[UBIFS_DENT_NODE].max_len = UBIFS_MAX_DENT_NODE_SZ; 590 c->ranges[UBIFS_XENT_NODE].min_len = UBIFS_XENT_NODE_SZ; 591 c->ranges[UBIFS_XENT_NODE].max_len = UBIFS_MAX_XENT_NODE_SZ; 592 c->ranges[UBIFS_DATA_NODE].min_len = UBIFS_DATA_NODE_SZ; 593 c->ranges[UBIFS_DATA_NODE].max_len = UBIFS_MAX_DATA_NODE_SZ; 594 /* 595 * Minimum indexing node size is amended later when superblock is 596 * read and the key length is known. 597 */ 598 c->ranges[UBIFS_IDX_NODE].min_len = UBIFS_IDX_NODE_SZ + UBIFS_BRANCH_SZ; 599 /* 600 * Maximum indexing node size is amended later when superblock is 601 * read and the fanout is known. 602 */ 603 c->ranges[UBIFS_IDX_NODE].max_len = INT_MAX; 604 605 /* 606 * Initialize dead and dark LEB space watermarks. See gc.c for comments 607 * about these values. 608 */ 609 c->dead_wm = ALIGN(MIN_WRITE_SZ, c->min_io_size); 610 c->dark_wm = ALIGN(UBIFS_MAX_NODE_SZ, c->min_io_size); 611 612 /* 613 * Calculate how many bytes would be wasted at the end of LEB if it was 614 * fully filled with data nodes of maximum size. This is used in 615 * calculations when reporting free space. 616 */ 617 c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ; 618 619 /* Buffer size for bulk-reads */ 620 c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ; 621 if (c->max_bu_buf_len > c->leb_size) 622 c->max_bu_buf_len = c->leb_size; 623 return 0; 624 } 625 626 /** 627 * bud_wbuf_callback - bud LEB write-buffer synchronization call-back. 628 * @c: UBIFS file-system description object 629 * @lnum: LEB the write-buffer was synchronized to 630 * @free: how many free bytes left in this LEB 631 * @pad: how many bytes were padded 632 * 633 * This is a callback function which is called by the I/O unit when the 634 * write-buffer is synchronized. We need this to correctly maintain space 635 * accounting in bud logical eraseblocks. This function returns zero in case of 636 * success and a negative error code in case of failure. 637 * 638 * This function actually belongs to the journal, but we keep it here because 639 * we want to keep it static. 640 */ 641 static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad) 642 { 643 return ubifs_update_one_lp(c, lnum, free, pad, 0, 0); 644 } 645 646 /* 647 * init_constants_sb - initialize UBIFS constants. 648 * @c: UBIFS file-system description object 649 * 650 * This is a helper function which initializes various UBIFS constants after 651 * the superblock has been read. It also checks various UBIFS parameters and 652 * makes sure they are all right. Returns zero in case of success and a 653 * negative error code in case of failure. 654 */ 655 static int init_constants_sb(struct ubifs_info *c) 656 { 657 int tmp, err; 658 long long tmp64; 659 660 c->main_bytes = (long long)c->main_lebs * c->leb_size; 661 c->max_znode_sz = sizeof(struct ubifs_znode) + 662 c->fanout * sizeof(struct ubifs_zbranch); 663 664 tmp = ubifs_idx_node_sz(c, 1); 665 c->ranges[UBIFS_IDX_NODE].min_len = tmp; 666 c->min_idx_node_sz = ALIGN(tmp, 8); 667 668 tmp = ubifs_idx_node_sz(c, c->fanout); 669 c->ranges[UBIFS_IDX_NODE].max_len = tmp; 670 c->max_idx_node_sz = ALIGN(tmp, 8); 671 672 /* Make sure LEB size is large enough to fit full commit */ 673 tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt; 674 tmp = ALIGN(tmp, c->min_io_size); 675 if (tmp > c->leb_size) { 676 ubifs_err(c, "too small LEB size %d, at least %d needed", 677 c->leb_size, tmp); 678 return -EINVAL; 679 } 680 681 /* 682 * Make sure that the log is large enough to fit reference nodes for 683 * all buds plus one reserved LEB. 684 */ 685 tmp64 = c->max_bud_bytes + c->leb_size - 1; 686 c->max_bud_cnt = div_u64(tmp64, c->leb_size); 687 tmp = (c->ref_node_alsz * c->max_bud_cnt + c->leb_size - 1); 688 tmp /= c->leb_size; 689 tmp += 1; 690 if (c->log_lebs < tmp) { 691 ubifs_err(c, "too small log %d LEBs, required min. %d LEBs", 692 c->log_lebs, tmp); 693 return -EINVAL; 694 } 695 696 /* 697 * When budgeting we assume worst-case scenarios when the pages are not 698 * be compressed and direntries are of the maximum size. 699 * 700 * Note, data, which may be stored in inodes is budgeted separately, so 701 * it is not included into 'c->bi.inode_budget'. 702 */ 703 c->bi.page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE; 704 c->bi.inode_budget = UBIFS_INO_NODE_SZ; 705 c->bi.dent_budget = UBIFS_MAX_DENT_NODE_SZ; 706 707 /* 708 * When the amount of flash space used by buds becomes 709 * 'c->max_bud_bytes', UBIFS just blocks all writers and starts commit. 710 * The writers are unblocked when the commit is finished. To avoid 711 * writers to be blocked UBIFS initiates background commit in advance, 712 * when number of bud bytes becomes above the limit defined below. 713 */ 714 c->bg_bud_bytes = (c->max_bud_bytes * 13) >> 4; 715 716 /* 717 * Ensure minimum journal size. All the bytes in the journal heads are 718 * considered to be used, when calculating the current journal usage. 719 * Consequently, if the journal is too small, UBIFS will treat it as 720 * always full. 721 */ 722 tmp64 = (long long)(c->jhead_cnt + 1) * c->leb_size + 1; 723 if (c->bg_bud_bytes < tmp64) 724 c->bg_bud_bytes = tmp64; 725 if (c->max_bud_bytes < tmp64 + c->leb_size) 726 c->max_bud_bytes = tmp64 + c->leb_size; 727 728 err = ubifs_calc_lpt_geom(c); 729 if (err) 730 return err; 731 732 /* Initialize effective LEB size used in budgeting calculations */ 733 c->idx_leb_size = c->leb_size - c->max_idx_node_sz; 734 return 0; 735 } 736 737 /* 738 * init_constants_master - initialize UBIFS constants. 739 * @c: UBIFS file-system description object 740 * 741 * This is a helper function which initializes various UBIFS constants after 742 * the master node has been read. It also checks various UBIFS parameters and 743 * makes sure they are all right. 744 */ 745 static void init_constants_master(struct ubifs_info *c) 746 { 747 long long tmp64; 748 749 c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c); 750 c->report_rp_size = ubifs_reported_space(c, c->rp_size); 751 752 /* 753 * Calculate total amount of FS blocks. This number is not used 754 * internally because it does not make much sense for UBIFS, but it is 755 * necessary to report something for the 'statfs()' call. 756 * 757 * Subtract the LEB reserved for GC, the LEB which is reserved for 758 * deletions, minimum LEBs for the index, and assume only one journal 759 * head is available. 760 */ 761 tmp64 = c->main_lebs - 1 - 1 - MIN_INDEX_LEBS - c->jhead_cnt + 1; 762 tmp64 *= (long long)c->leb_size - c->leb_overhead; 763 tmp64 = ubifs_reported_space(c, tmp64); 764 c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT; 765 } 766 767 /** 768 * take_gc_lnum - reserve GC LEB. 769 * @c: UBIFS file-system description object 770 * 771 * This function ensures that the LEB reserved for garbage collection is marked 772 * as "taken" in lprops. We also have to set free space to LEB size and dirty 773 * space to zero, because lprops may contain out-of-date information if the 774 * file-system was un-mounted before it has been committed. This function 775 * returns zero in case of success and a negative error code in case of 776 * failure. 777 */ 778 static int take_gc_lnum(struct ubifs_info *c) 779 { 780 int err; 781 782 if (c->gc_lnum == -1) { 783 ubifs_err(c, "no LEB for GC"); 784 return -EINVAL; 785 } 786 787 /* And we have to tell lprops that this LEB is taken */ 788 err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0, 789 LPROPS_TAKEN, 0, 0); 790 return err; 791 } 792 793 /** 794 * alloc_wbufs - allocate write-buffers. 795 * @c: UBIFS file-system description object 796 * 797 * This helper function allocates and initializes UBIFS write-buffers. Returns 798 * zero in case of success and %-ENOMEM in case of failure. 799 */ 800 static int alloc_wbufs(struct ubifs_info *c) 801 { 802 int i, err; 803 804 c->jheads = kcalloc(c->jhead_cnt, sizeof(struct ubifs_jhead), 805 GFP_KERNEL); 806 if (!c->jheads) 807 return -ENOMEM; 808 809 /* Initialize journal heads */ 810 for (i = 0; i < c->jhead_cnt; i++) { 811 INIT_LIST_HEAD(&c->jheads[i].buds_list); 812 err = ubifs_wbuf_init(c, &c->jheads[i].wbuf); 813 if (err) 814 return err; 815 816 c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback; 817 c->jheads[i].wbuf.jhead = i; 818 c->jheads[i].grouped = 1; 819 } 820 821 /* 822 * Garbage Collector head does not need to be synchronized by timer. 823 * Also GC head nodes are not grouped. 824 */ 825 c->jheads[GCHD].wbuf.no_timer = 1; 826 c->jheads[GCHD].grouped = 0; 827 828 return 0; 829 } 830 831 /** 832 * free_wbufs - free write-buffers. 833 * @c: UBIFS file-system description object 834 */ 835 static void free_wbufs(struct ubifs_info *c) 836 { 837 int i; 838 839 if (c->jheads) { 840 for (i = 0; i < c->jhead_cnt; i++) { 841 kfree(c->jheads[i].wbuf.buf); 842 kfree(c->jheads[i].wbuf.inodes); 843 } 844 kfree(c->jheads); 845 c->jheads = NULL; 846 } 847 } 848 849 /** 850 * free_orphans - free orphans. 851 * @c: UBIFS file-system description object 852 */ 853 static void free_orphans(struct ubifs_info *c) 854 { 855 struct ubifs_orphan *orph; 856 857 while (c->orph_dnext) { 858 orph = c->orph_dnext; 859 c->orph_dnext = orph->dnext; 860 list_del(&orph->list); 861 kfree(orph); 862 } 863 864 while (!list_empty(&c->orph_list)) { 865 orph = list_entry(c->orph_list.next, struct ubifs_orphan, list); 866 list_del(&orph->list); 867 kfree(orph); 868 ubifs_err(c, "orphan list not empty at unmount"); 869 } 870 871 vfree(c->orph_buf); 872 c->orph_buf = NULL; 873 } 874 875 /** 876 * free_buds - free per-bud objects. 877 * @c: UBIFS file-system description object 878 */ 879 static void free_buds(struct ubifs_info *c) 880 { 881 struct ubifs_bud *bud, *n; 882 883 rbtree_postorder_for_each_entry_safe(bud, n, &c->buds, rb) 884 kfree(bud); 885 } 886 887 /** 888 * check_volume_empty - check if the UBI volume is empty. 889 * @c: UBIFS file-system description object 890 * 891 * This function checks if the UBIFS volume is empty by looking if its LEBs are 892 * mapped or not. The result of checking is stored in the @c->empty variable. 893 * Returns zero in case of success and a negative error code in case of 894 * failure. 895 */ 896 static int check_volume_empty(struct ubifs_info *c) 897 { 898 int lnum, err; 899 900 c->empty = 1; 901 for (lnum = 0; lnum < c->leb_cnt; lnum++) { 902 err = ubifs_is_mapped(c, lnum); 903 if (unlikely(err < 0)) 904 return err; 905 if (err == 1) { 906 c->empty = 0; 907 break; 908 } 909 910 cond_resched(); 911 } 912 913 return 0; 914 } 915 916 /* 917 * UBIFS mount options. 918 * 919 * Opt_fast_unmount: do not run a journal commit before un-mounting 920 * Opt_norm_unmount: run a journal commit before un-mounting 921 * Opt_bulk_read: enable bulk-reads 922 * Opt_no_bulk_read: disable bulk-reads 923 * Opt_chk_data_crc: check CRCs when reading data nodes 924 * Opt_no_chk_data_crc: do not check CRCs when reading data nodes 925 * Opt_override_compr: override default compressor 926 * Opt_err: just end of array marker 927 */ 928 enum { 929 Opt_fast_unmount, 930 Opt_norm_unmount, 931 Opt_bulk_read, 932 Opt_no_bulk_read, 933 Opt_chk_data_crc, 934 Opt_no_chk_data_crc, 935 Opt_override_compr, 936 Opt_ignore, 937 Opt_err, 938 }; 939 940 static const match_table_t tokens = { 941 {Opt_fast_unmount, "fast_unmount"}, 942 {Opt_norm_unmount, "norm_unmount"}, 943 {Opt_bulk_read, "bulk_read"}, 944 {Opt_no_bulk_read, "no_bulk_read"}, 945 {Opt_chk_data_crc, "chk_data_crc"}, 946 {Opt_no_chk_data_crc, "no_chk_data_crc"}, 947 {Opt_override_compr, "compr=%s"}, 948 {Opt_ignore, "ubi=%s"}, 949 {Opt_ignore, "vol=%s"}, 950 {Opt_err, NULL}, 951 }; 952 953 /** 954 * parse_standard_option - parse a standard mount option. 955 * @option: the option to parse 956 * 957 * Normally, standard mount options like "sync" are passed to file-systems as 958 * flags. However, when a "rootflags=" kernel boot parameter is used, they may 959 * be present in the options string. This function tries to deal with this 960 * situation and parse standard options. Returns 0 if the option was not 961 * recognized, and the corresponding integer flag if it was. 962 * 963 * UBIFS is only interested in the "sync" option, so do not check for anything 964 * else. 965 */ 966 static int parse_standard_option(const char *option) 967 { 968 969 pr_notice("UBIFS: parse %s\n", option); 970 if (!strcmp(option, "sync")) 971 return MS_SYNCHRONOUS; 972 return 0; 973 } 974 975 /** 976 * ubifs_parse_options - parse mount parameters. 977 * @c: UBIFS file-system description object 978 * @options: parameters to parse 979 * @is_remount: non-zero if this is FS re-mount 980 * 981 * This function parses UBIFS mount options and returns zero in case success 982 * and a negative error code in case of failure. 983 */ 984 static int ubifs_parse_options(struct ubifs_info *c, char *options, 985 int is_remount) 986 { 987 char *p; 988 substring_t args[MAX_OPT_ARGS]; 989 990 if (!options) 991 return 0; 992 993 while ((p = strsep(&options, ","))) { 994 int token; 995 996 if (!*p) 997 continue; 998 999 token = match_token(p, tokens, args); 1000 switch (token) { 1001 /* 1002 * %Opt_fast_unmount and %Opt_norm_unmount options are ignored. 1003 * We accept them in order to be backward-compatible. But this 1004 * should be removed at some point. 1005 */ 1006 case Opt_fast_unmount: 1007 c->mount_opts.unmount_mode = 2; 1008 break; 1009 case Opt_norm_unmount: 1010 c->mount_opts.unmount_mode = 1; 1011 break; 1012 case Opt_bulk_read: 1013 c->mount_opts.bulk_read = 2; 1014 c->bulk_read = 1; 1015 break; 1016 case Opt_no_bulk_read: 1017 c->mount_opts.bulk_read = 1; 1018 c->bulk_read = 0; 1019 break; 1020 case Opt_chk_data_crc: 1021 c->mount_opts.chk_data_crc = 2; 1022 c->no_chk_data_crc = 0; 1023 break; 1024 case Opt_no_chk_data_crc: 1025 c->mount_opts.chk_data_crc = 1; 1026 c->no_chk_data_crc = 1; 1027 break; 1028 case Opt_override_compr: 1029 { 1030 char *name = match_strdup(&args[0]); 1031 1032 if (!name) 1033 return -ENOMEM; 1034 if (!strcmp(name, "none")) 1035 c->mount_opts.compr_type = UBIFS_COMPR_NONE; 1036 else if (!strcmp(name, "lzo")) 1037 c->mount_opts.compr_type = UBIFS_COMPR_LZO; 1038 else if (!strcmp(name, "zlib")) 1039 c->mount_opts.compr_type = UBIFS_COMPR_ZLIB; 1040 else { 1041 ubifs_err(c, "unknown compressor \"%s\"", name); //FIXME: is c ready? 1042 kfree(name); 1043 return -EINVAL; 1044 } 1045 kfree(name); 1046 c->mount_opts.override_compr = 1; 1047 c->default_compr = c->mount_opts.compr_type; 1048 break; 1049 } 1050 case Opt_ignore: 1051 break; 1052 default: 1053 { 1054 unsigned long flag; 1055 struct super_block *sb = c->vfs_sb; 1056 1057 flag = parse_standard_option(p); 1058 if (!flag) { 1059 ubifs_err(c, "unrecognized mount option \"%s\" or missing value", 1060 p); 1061 return -EINVAL; 1062 } 1063 sb->s_flags |= flag; 1064 break; 1065 } 1066 } 1067 } 1068 1069 return 0; 1070 } 1071 1072 /** 1073 * destroy_journal - destroy journal data structures. 1074 * @c: UBIFS file-system description object 1075 * 1076 * This function destroys journal data structures including those that may have 1077 * been created by recovery functions. 1078 */ 1079 static void destroy_journal(struct ubifs_info *c) 1080 { 1081 while (!list_empty(&c->unclean_leb_list)) { 1082 struct ubifs_unclean_leb *ucleb; 1083 1084 ucleb = list_entry(c->unclean_leb_list.next, 1085 struct ubifs_unclean_leb, list); 1086 list_del(&ucleb->list); 1087 kfree(ucleb); 1088 } 1089 while (!list_empty(&c->old_buds)) { 1090 struct ubifs_bud *bud; 1091 1092 bud = list_entry(c->old_buds.next, struct ubifs_bud, list); 1093 list_del(&bud->list); 1094 kfree(bud); 1095 } 1096 ubifs_destroy_idx_gc(c); 1097 ubifs_destroy_size_tree(c); 1098 ubifs_tnc_close(c); 1099 free_buds(c); 1100 } 1101 1102 /** 1103 * bu_init - initialize bulk-read information. 1104 * @c: UBIFS file-system description object 1105 */ 1106 static void bu_init(struct ubifs_info *c) 1107 { 1108 ubifs_assert(c->bulk_read == 1); 1109 1110 if (c->bu.buf) 1111 return; /* Already initialized */ 1112 1113 again: 1114 c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN); 1115 if (!c->bu.buf) { 1116 if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) { 1117 c->max_bu_buf_len = UBIFS_KMALLOC_OK; 1118 goto again; 1119 } 1120 1121 /* Just disable bulk-read */ 1122 ubifs_warn(c, "cannot allocate %d bytes of memory for bulk-read, disabling it", 1123 c->max_bu_buf_len); 1124 c->mount_opts.bulk_read = 1; 1125 c->bulk_read = 0; 1126 return; 1127 } 1128 } 1129 1130 /** 1131 * check_free_space - check if there is enough free space to mount. 1132 * @c: UBIFS file-system description object 1133 * 1134 * This function makes sure UBIFS has enough free space to be mounted in 1135 * read/write mode. UBIFS must always have some free space to allow deletions. 1136 */ 1137 static int check_free_space(struct ubifs_info *c) 1138 { 1139 ubifs_assert(c->dark_wm > 0); 1140 if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) { 1141 ubifs_err(c, "insufficient free space to mount in R/W mode"); 1142 ubifs_dump_budg(c, &c->bi); 1143 ubifs_dump_lprops(c); 1144 return -ENOSPC; 1145 } 1146 return 0; 1147 } 1148 1149 /** 1150 * mount_ubifs - mount UBIFS file-system. 1151 * @c: UBIFS file-system description object 1152 * 1153 * This function mounts UBIFS file system. Returns zero in case of success and 1154 * a negative error code in case of failure. 1155 */ 1156 static int mount_ubifs(struct ubifs_info *c) 1157 { 1158 int err; 1159 long long x, y; 1160 size_t sz; 1161 1162 c->ro_mount = !!sb_rdonly(c->vfs_sb); 1163 /* Suppress error messages while probing if MS_SILENT is set */ 1164 c->probing = !!(c->vfs_sb->s_flags & MS_SILENT); 1165 1166 err = init_constants_early(c); 1167 if (err) 1168 return err; 1169 1170 err = ubifs_debugging_init(c); 1171 if (err) 1172 return err; 1173 1174 err = check_volume_empty(c); 1175 if (err) 1176 goto out_free; 1177 1178 if (c->empty && (c->ro_mount || c->ro_media)) { 1179 /* 1180 * This UBI volume is empty, and read-only, or the file system 1181 * is mounted read-only - we cannot format it. 1182 */ 1183 ubifs_err(c, "can't format empty UBI volume: read-only %s", 1184 c->ro_media ? "UBI volume" : "mount"); 1185 err = -EROFS; 1186 goto out_free; 1187 } 1188 1189 if (c->ro_media && !c->ro_mount) { 1190 ubifs_err(c, "cannot mount read-write - read-only media"); 1191 err = -EROFS; 1192 goto out_free; 1193 } 1194 1195 /* 1196 * The requirement for the buffer is that it should fit indexing B-tree 1197 * height amount of integers. We assume the height if the TNC tree will 1198 * never exceed 64. 1199 */ 1200 err = -ENOMEM; 1201 c->bottom_up_buf = kmalloc(BOTTOM_UP_HEIGHT * sizeof(int), GFP_KERNEL); 1202 if (!c->bottom_up_buf) 1203 goto out_free; 1204 1205 c->sbuf = vmalloc(c->leb_size); 1206 if (!c->sbuf) 1207 goto out_free; 1208 1209 if (!c->ro_mount) { 1210 c->ileb_buf = vmalloc(c->leb_size); 1211 if (!c->ileb_buf) 1212 goto out_free; 1213 } 1214 1215 if (c->bulk_read == 1) 1216 bu_init(c); 1217 1218 if (!c->ro_mount) { 1219 c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ + \ 1220 UBIFS_CIPHER_BLOCK_SIZE, 1221 GFP_KERNEL); 1222 if (!c->write_reserve_buf) 1223 goto out_free; 1224 } 1225 1226 c->mounting = 1; 1227 1228 err = ubifs_read_superblock(c); 1229 if (err) 1230 goto out_free; 1231 1232 c->probing = 0; 1233 1234 /* 1235 * Make sure the compressor which is set as default in the superblock 1236 * or overridden by mount options is actually compiled in. 1237 */ 1238 if (!ubifs_compr_present(c->default_compr)) { 1239 ubifs_err(c, "'compressor \"%s\" is not compiled in", 1240 ubifs_compr_name(c->default_compr)); 1241 err = -ENOTSUPP; 1242 goto out_free; 1243 } 1244 1245 err = init_constants_sb(c); 1246 if (err) 1247 goto out_free; 1248 1249 sz = ALIGN(c->max_idx_node_sz, c->min_io_size); 1250 sz = ALIGN(sz + c->max_idx_node_sz, c->min_io_size); 1251 c->cbuf = kmalloc(sz, GFP_NOFS); 1252 if (!c->cbuf) { 1253 err = -ENOMEM; 1254 goto out_free; 1255 } 1256 1257 err = alloc_wbufs(c); 1258 if (err) 1259 goto out_cbuf; 1260 1261 sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id); 1262 if (!c->ro_mount) { 1263 /* Create background thread */ 1264 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name); 1265 if (IS_ERR(c->bgt)) { 1266 err = PTR_ERR(c->bgt); 1267 c->bgt = NULL; 1268 ubifs_err(c, "cannot spawn \"%s\", error %d", 1269 c->bgt_name, err); 1270 goto out_wbufs; 1271 } 1272 wake_up_process(c->bgt); 1273 } 1274 1275 err = ubifs_read_master(c); 1276 if (err) 1277 goto out_master; 1278 1279 init_constants_master(c); 1280 1281 if ((c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY)) != 0) { 1282 ubifs_msg(c, "recovery needed"); 1283 c->need_recovery = 1; 1284 } 1285 1286 if (c->need_recovery && !c->ro_mount) { 1287 err = ubifs_recover_inl_heads(c, c->sbuf); 1288 if (err) 1289 goto out_master; 1290 } 1291 1292 err = ubifs_lpt_init(c, 1, !c->ro_mount); 1293 if (err) 1294 goto out_master; 1295 1296 if (!c->ro_mount && c->space_fixup) { 1297 err = ubifs_fixup_free_space(c); 1298 if (err) 1299 goto out_lpt; 1300 } 1301 1302 if (!c->ro_mount && !c->need_recovery) { 1303 /* 1304 * Set the "dirty" flag so that if we reboot uncleanly we 1305 * will notice this immediately on the next mount. 1306 */ 1307 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY); 1308 err = ubifs_write_master(c); 1309 if (err) 1310 goto out_lpt; 1311 } 1312 1313 err = dbg_check_idx_size(c, c->bi.old_idx_sz); 1314 if (err) 1315 goto out_lpt; 1316 1317 err = ubifs_replay_journal(c); 1318 if (err) 1319 goto out_journal; 1320 1321 /* Calculate 'min_idx_lebs' after journal replay */ 1322 c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c); 1323 1324 err = ubifs_mount_orphans(c, c->need_recovery, c->ro_mount); 1325 if (err) 1326 goto out_orphans; 1327 1328 if (!c->ro_mount) { 1329 int lnum; 1330 1331 err = check_free_space(c); 1332 if (err) 1333 goto out_orphans; 1334 1335 /* Check for enough log space */ 1336 lnum = c->lhead_lnum + 1; 1337 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs) 1338 lnum = UBIFS_LOG_LNUM; 1339 if (lnum == c->ltail_lnum) { 1340 err = ubifs_consolidate_log(c); 1341 if (err) 1342 goto out_orphans; 1343 } 1344 1345 if (c->need_recovery) { 1346 err = ubifs_recover_size(c); 1347 if (err) 1348 goto out_orphans; 1349 err = ubifs_rcvry_gc_commit(c); 1350 if (err) 1351 goto out_orphans; 1352 } else { 1353 err = take_gc_lnum(c); 1354 if (err) 1355 goto out_orphans; 1356 1357 /* 1358 * GC LEB may contain garbage if there was an unclean 1359 * reboot, and it should be un-mapped. 1360 */ 1361 err = ubifs_leb_unmap(c, c->gc_lnum); 1362 if (err) 1363 goto out_orphans; 1364 } 1365 1366 err = dbg_check_lprops(c); 1367 if (err) 1368 goto out_orphans; 1369 } else if (c->need_recovery) { 1370 err = ubifs_recover_size(c); 1371 if (err) 1372 goto out_orphans; 1373 } else { 1374 /* 1375 * Even if we mount read-only, we have to set space in GC LEB 1376 * to proper value because this affects UBIFS free space 1377 * reporting. We do not want to have a situation when 1378 * re-mounting from R/O to R/W changes amount of free space. 1379 */ 1380 err = take_gc_lnum(c); 1381 if (err) 1382 goto out_orphans; 1383 } 1384 1385 spin_lock(&ubifs_infos_lock); 1386 list_add_tail(&c->infos_list, &ubifs_infos); 1387 spin_unlock(&ubifs_infos_lock); 1388 1389 if (c->need_recovery) { 1390 if (c->ro_mount) 1391 ubifs_msg(c, "recovery deferred"); 1392 else { 1393 c->need_recovery = 0; 1394 ubifs_msg(c, "recovery completed"); 1395 /* 1396 * GC LEB has to be empty and taken at this point. But 1397 * the journal head LEBs may also be accounted as 1398 * "empty taken" if they are empty. 1399 */ 1400 ubifs_assert(c->lst.taken_empty_lebs > 0); 1401 } 1402 } else 1403 ubifs_assert(c->lst.taken_empty_lebs > 0); 1404 1405 err = dbg_check_filesystem(c); 1406 if (err) 1407 goto out_infos; 1408 1409 err = dbg_debugfs_init_fs(c); 1410 if (err) 1411 goto out_infos; 1412 1413 c->mounting = 0; 1414 1415 ubifs_msg(c, "UBIFS: mounted UBI device %d, volume %d, name \"%s\"%s", 1416 c->vi.ubi_num, c->vi.vol_id, c->vi.name, 1417 c->ro_mount ? ", R/O mode" : ""); 1418 x = (long long)c->main_lebs * c->leb_size; 1419 y = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes; 1420 ubifs_msg(c, "LEB size: %d bytes (%d KiB), min./max. I/O unit sizes: %d bytes/%d bytes", 1421 c->leb_size, c->leb_size >> 10, c->min_io_size, 1422 c->max_write_size); 1423 ubifs_msg(c, "FS size: %lld bytes (%lld MiB, %d LEBs), journal size %lld bytes (%lld MiB, %d LEBs)", 1424 x, x >> 20, c->main_lebs, 1425 y, y >> 20, c->log_lebs + c->max_bud_cnt); 1426 ubifs_msg(c, "reserved for root: %llu bytes (%llu KiB)", 1427 c->report_rp_size, c->report_rp_size >> 10); 1428 ubifs_msg(c, "media format: w%d/r%d (latest is w%d/r%d), UUID %pUB%s", 1429 c->fmt_version, c->ro_compat_version, 1430 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION, c->uuid, 1431 c->big_lpt ? ", big LPT model" : ", small LPT model"); 1432 1433 dbg_gen("default compressor: %s", ubifs_compr_name(c->default_compr)); 1434 dbg_gen("data journal heads: %d", 1435 c->jhead_cnt - NONDATA_JHEADS_CNT); 1436 dbg_gen("log LEBs: %d (%d - %d)", 1437 c->log_lebs, UBIFS_LOG_LNUM, c->log_last); 1438 dbg_gen("LPT area LEBs: %d (%d - %d)", 1439 c->lpt_lebs, c->lpt_first, c->lpt_last); 1440 dbg_gen("orphan area LEBs: %d (%d - %d)", 1441 c->orph_lebs, c->orph_first, c->orph_last); 1442 dbg_gen("main area LEBs: %d (%d - %d)", 1443 c->main_lebs, c->main_first, c->leb_cnt - 1); 1444 dbg_gen("index LEBs: %d", c->lst.idx_lebs); 1445 dbg_gen("total index bytes: %lld (%lld KiB, %lld MiB)", 1446 c->bi.old_idx_sz, c->bi.old_idx_sz >> 10, 1447 c->bi.old_idx_sz >> 20); 1448 dbg_gen("key hash type: %d", c->key_hash_type); 1449 dbg_gen("tree fanout: %d", c->fanout); 1450 dbg_gen("reserved GC LEB: %d", c->gc_lnum); 1451 dbg_gen("max. znode size %d", c->max_znode_sz); 1452 dbg_gen("max. index node size %d", c->max_idx_node_sz); 1453 dbg_gen("node sizes: data %zu, inode %zu, dentry %zu", 1454 UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ); 1455 dbg_gen("node sizes: trun %zu, sb %zu, master %zu", 1456 UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ); 1457 dbg_gen("node sizes: ref %zu, cmt. start %zu, orph %zu", 1458 UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ); 1459 dbg_gen("max. node sizes: data %zu, inode %zu dentry %zu, idx %d", 1460 UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ, 1461 UBIFS_MAX_DENT_NODE_SZ, ubifs_idx_node_sz(c, c->fanout)); 1462 dbg_gen("dead watermark: %d", c->dead_wm); 1463 dbg_gen("dark watermark: %d", c->dark_wm); 1464 dbg_gen("LEB overhead: %d", c->leb_overhead); 1465 x = (long long)c->main_lebs * c->dark_wm; 1466 dbg_gen("max. dark space: %lld (%lld KiB, %lld MiB)", 1467 x, x >> 10, x >> 20); 1468 dbg_gen("maximum bud bytes: %lld (%lld KiB, %lld MiB)", 1469 c->max_bud_bytes, c->max_bud_bytes >> 10, 1470 c->max_bud_bytes >> 20); 1471 dbg_gen("BG commit bud bytes: %lld (%lld KiB, %lld MiB)", 1472 c->bg_bud_bytes, c->bg_bud_bytes >> 10, 1473 c->bg_bud_bytes >> 20); 1474 dbg_gen("current bud bytes %lld (%lld KiB, %lld MiB)", 1475 c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20); 1476 dbg_gen("max. seq. number: %llu", c->max_sqnum); 1477 dbg_gen("commit number: %llu", c->cmt_no); 1478 1479 return 0; 1480 1481 out_infos: 1482 spin_lock(&ubifs_infos_lock); 1483 list_del(&c->infos_list); 1484 spin_unlock(&ubifs_infos_lock); 1485 out_orphans: 1486 free_orphans(c); 1487 out_journal: 1488 destroy_journal(c); 1489 out_lpt: 1490 ubifs_lpt_free(c, 0); 1491 out_master: 1492 kfree(c->mst_node); 1493 kfree(c->rcvrd_mst_node); 1494 if (c->bgt) 1495 kthread_stop(c->bgt); 1496 out_wbufs: 1497 free_wbufs(c); 1498 out_cbuf: 1499 kfree(c->cbuf); 1500 out_free: 1501 kfree(c->write_reserve_buf); 1502 kfree(c->bu.buf); 1503 vfree(c->ileb_buf); 1504 vfree(c->sbuf); 1505 kfree(c->bottom_up_buf); 1506 ubifs_debugging_exit(c); 1507 return err; 1508 } 1509 1510 /** 1511 * ubifs_umount - un-mount UBIFS file-system. 1512 * @c: UBIFS file-system description object 1513 * 1514 * Note, this function is called to free allocated resourced when un-mounting, 1515 * as well as free resources when an error occurred while we were half way 1516 * through mounting (error path cleanup function). So it has to make sure the 1517 * resource was actually allocated before freeing it. 1518 */ 1519 static void ubifs_umount(struct ubifs_info *c) 1520 { 1521 dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num, 1522 c->vi.vol_id); 1523 1524 dbg_debugfs_exit_fs(c); 1525 spin_lock(&ubifs_infos_lock); 1526 list_del(&c->infos_list); 1527 spin_unlock(&ubifs_infos_lock); 1528 1529 if (c->bgt) 1530 kthread_stop(c->bgt); 1531 1532 destroy_journal(c); 1533 free_wbufs(c); 1534 free_orphans(c); 1535 ubifs_lpt_free(c, 0); 1536 1537 kfree(c->cbuf); 1538 kfree(c->rcvrd_mst_node); 1539 kfree(c->mst_node); 1540 kfree(c->write_reserve_buf); 1541 kfree(c->bu.buf); 1542 vfree(c->ileb_buf); 1543 vfree(c->sbuf); 1544 kfree(c->bottom_up_buf); 1545 ubifs_debugging_exit(c); 1546 } 1547 1548 /** 1549 * ubifs_remount_rw - re-mount in read-write mode. 1550 * @c: UBIFS file-system description object 1551 * 1552 * UBIFS avoids allocating many unnecessary resources when mounted in read-only 1553 * mode. This function allocates the needed resources and re-mounts UBIFS in 1554 * read-write mode. 1555 */ 1556 static int ubifs_remount_rw(struct ubifs_info *c) 1557 { 1558 int err, lnum; 1559 1560 if (c->rw_incompat) { 1561 ubifs_err(c, "the file-system is not R/W-compatible"); 1562 ubifs_msg(c, "on-flash format version is w%d/r%d, but software only supports up to version w%d/r%d", 1563 c->fmt_version, c->ro_compat_version, 1564 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION); 1565 return -EROFS; 1566 } 1567 1568 mutex_lock(&c->umount_mutex); 1569 dbg_save_space_info(c); 1570 c->remounting_rw = 1; 1571 c->ro_mount = 0; 1572 1573 if (c->space_fixup) { 1574 err = ubifs_fixup_free_space(c); 1575 if (err) 1576 goto out; 1577 } 1578 1579 err = check_free_space(c); 1580 if (err) 1581 goto out; 1582 1583 if (c->old_leb_cnt != c->leb_cnt) { 1584 struct ubifs_sb_node *sup; 1585 1586 sup = ubifs_read_sb_node(c); 1587 if (IS_ERR(sup)) { 1588 err = PTR_ERR(sup); 1589 goto out; 1590 } 1591 sup->leb_cnt = cpu_to_le32(c->leb_cnt); 1592 err = ubifs_write_sb_node(c, sup); 1593 kfree(sup); 1594 if (err) 1595 goto out; 1596 } 1597 1598 if (c->need_recovery) { 1599 ubifs_msg(c, "completing deferred recovery"); 1600 err = ubifs_write_rcvrd_mst_node(c); 1601 if (err) 1602 goto out; 1603 err = ubifs_recover_size(c); 1604 if (err) 1605 goto out; 1606 err = ubifs_clean_lebs(c, c->sbuf); 1607 if (err) 1608 goto out; 1609 err = ubifs_recover_inl_heads(c, c->sbuf); 1610 if (err) 1611 goto out; 1612 } else { 1613 /* A readonly mount is not allowed to have orphans */ 1614 ubifs_assert(c->tot_orphans == 0); 1615 err = ubifs_clear_orphans(c); 1616 if (err) 1617 goto out; 1618 } 1619 1620 if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) { 1621 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY); 1622 err = ubifs_write_master(c); 1623 if (err) 1624 goto out; 1625 } 1626 1627 c->ileb_buf = vmalloc(c->leb_size); 1628 if (!c->ileb_buf) { 1629 err = -ENOMEM; 1630 goto out; 1631 } 1632 1633 c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ + \ 1634 UBIFS_CIPHER_BLOCK_SIZE, GFP_KERNEL); 1635 if (!c->write_reserve_buf) { 1636 err = -ENOMEM; 1637 goto out; 1638 } 1639 1640 err = ubifs_lpt_init(c, 0, 1); 1641 if (err) 1642 goto out; 1643 1644 /* Create background thread */ 1645 c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name); 1646 if (IS_ERR(c->bgt)) { 1647 err = PTR_ERR(c->bgt); 1648 c->bgt = NULL; 1649 ubifs_err(c, "cannot spawn \"%s\", error %d", 1650 c->bgt_name, err); 1651 goto out; 1652 } 1653 wake_up_process(c->bgt); 1654 1655 c->orph_buf = vmalloc(c->leb_size); 1656 if (!c->orph_buf) { 1657 err = -ENOMEM; 1658 goto out; 1659 } 1660 1661 /* Check for enough log space */ 1662 lnum = c->lhead_lnum + 1; 1663 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs) 1664 lnum = UBIFS_LOG_LNUM; 1665 if (lnum == c->ltail_lnum) { 1666 err = ubifs_consolidate_log(c); 1667 if (err) 1668 goto out; 1669 } 1670 1671 if (c->need_recovery) 1672 err = ubifs_rcvry_gc_commit(c); 1673 else 1674 err = ubifs_leb_unmap(c, c->gc_lnum); 1675 if (err) 1676 goto out; 1677 1678 dbg_gen("re-mounted read-write"); 1679 c->remounting_rw = 0; 1680 1681 if (c->need_recovery) { 1682 c->need_recovery = 0; 1683 ubifs_msg(c, "deferred recovery completed"); 1684 } else { 1685 /* 1686 * Do not run the debugging space check if the were doing 1687 * recovery, because when we saved the information we had the 1688 * file-system in a state where the TNC and lprops has been 1689 * modified in memory, but all the I/O operations (including a 1690 * commit) were deferred. So the file-system was in 1691 * "non-committed" state. Now the file-system is in committed 1692 * state, and of course the amount of free space will change 1693 * because, for example, the old index size was imprecise. 1694 */ 1695 err = dbg_check_space_info(c); 1696 } 1697 1698 mutex_unlock(&c->umount_mutex); 1699 return err; 1700 1701 out: 1702 c->ro_mount = 1; 1703 vfree(c->orph_buf); 1704 c->orph_buf = NULL; 1705 if (c->bgt) { 1706 kthread_stop(c->bgt); 1707 c->bgt = NULL; 1708 } 1709 free_wbufs(c); 1710 kfree(c->write_reserve_buf); 1711 c->write_reserve_buf = NULL; 1712 vfree(c->ileb_buf); 1713 c->ileb_buf = NULL; 1714 ubifs_lpt_free(c, 1); 1715 c->remounting_rw = 0; 1716 mutex_unlock(&c->umount_mutex); 1717 return err; 1718 } 1719 1720 /** 1721 * ubifs_remount_ro - re-mount in read-only mode. 1722 * @c: UBIFS file-system description object 1723 * 1724 * We assume VFS has stopped writing. Possibly the background thread could be 1725 * running a commit, however kthread_stop will wait in that case. 1726 */ 1727 static void ubifs_remount_ro(struct ubifs_info *c) 1728 { 1729 int i, err; 1730 1731 ubifs_assert(!c->need_recovery); 1732 ubifs_assert(!c->ro_mount); 1733 1734 mutex_lock(&c->umount_mutex); 1735 if (c->bgt) { 1736 kthread_stop(c->bgt); 1737 c->bgt = NULL; 1738 } 1739 1740 dbg_save_space_info(c); 1741 1742 for (i = 0; i < c->jhead_cnt; i++) 1743 ubifs_wbuf_sync(&c->jheads[i].wbuf); 1744 1745 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY); 1746 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS); 1747 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum); 1748 err = ubifs_write_master(c); 1749 if (err) 1750 ubifs_ro_mode(c, err); 1751 1752 vfree(c->orph_buf); 1753 c->orph_buf = NULL; 1754 kfree(c->write_reserve_buf); 1755 c->write_reserve_buf = NULL; 1756 vfree(c->ileb_buf); 1757 c->ileb_buf = NULL; 1758 ubifs_lpt_free(c, 1); 1759 c->ro_mount = 1; 1760 err = dbg_check_space_info(c); 1761 if (err) 1762 ubifs_ro_mode(c, err); 1763 mutex_unlock(&c->umount_mutex); 1764 } 1765 1766 static void ubifs_put_super(struct super_block *sb) 1767 { 1768 int i; 1769 struct ubifs_info *c = sb->s_fs_info; 1770 1771 ubifs_msg(c, "un-mount UBI device %d", c->vi.ubi_num); 1772 1773 /* 1774 * The following asserts are only valid if there has not been a failure 1775 * of the media. For example, there will be dirty inodes if we failed 1776 * to write them back because of I/O errors. 1777 */ 1778 if (!c->ro_error) { 1779 ubifs_assert(c->bi.idx_growth == 0); 1780 ubifs_assert(c->bi.dd_growth == 0); 1781 ubifs_assert(c->bi.data_growth == 0); 1782 } 1783 1784 /* 1785 * The 'c->umount_lock' prevents races between UBIFS memory shrinker 1786 * and file system un-mount. Namely, it prevents the shrinker from 1787 * picking this superblock for shrinking - it will be just skipped if 1788 * the mutex is locked. 1789 */ 1790 mutex_lock(&c->umount_mutex); 1791 if (!c->ro_mount) { 1792 /* 1793 * First of all kill the background thread to make sure it does 1794 * not interfere with un-mounting and freeing resources. 1795 */ 1796 if (c->bgt) { 1797 kthread_stop(c->bgt); 1798 c->bgt = NULL; 1799 } 1800 1801 /* 1802 * On fatal errors c->ro_error is set to 1, in which case we do 1803 * not write the master node. 1804 */ 1805 if (!c->ro_error) { 1806 int err; 1807 1808 /* Synchronize write-buffers */ 1809 for (i = 0; i < c->jhead_cnt; i++) 1810 ubifs_wbuf_sync(&c->jheads[i].wbuf); 1811 1812 /* 1813 * We are being cleanly unmounted which means the 1814 * orphans were killed - indicate this in the master 1815 * node. Also save the reserved GC LEB number. 1816 */ 1817 c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY); 1818 c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS); 1819 c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum); 1820 err = ubifs_write_master(c); 1821 if (err) 1822 /* 1823 * Recovery will attempt to fix the master area 1824 * next mount, so we just print a message and 1825 * continue to unmount normally. 1826 */ 1827 ubifs_err(c, "failed to write master node, error %d", 1828 err); 1829 } else { 1830 for (i = 0; i < c->jhead_cnt; i++) 1831 /* Make sure write-buffer timers are canceled */ 1832 hrtimer_cancel(&c->jheads[i].wbuf.timer); 1833 } 1834 } 1835 1836 ubifs_umount(c); 1837 ubi_close_volume(c->ubi); 1838 mutex_unlock(&c->umount_mutex); 1839 } 1840 1841 static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data) 1842 { 1843 int err; 1844 struct ubifs_info *c = sb->s_fs_info; 1845 1846 sync_filesystem(sb); 1847 dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags); 1848 1849 err = ubifs_parse_options(c, data, 1); 1850 if (err) { 1851 ubifs_err(c, "invalid or unknown remount parameter"); 1852 return err; 1853 } 1854 1855 if (c->ro_mount && !(*flags & MS_RDONLY)) { 1856 if (c->ro_error) { 1857 ubifs_msg(c, "cannot re-mount R/W due to prior errors"); 1858 return -EROFS; 1859 } 1860 if (c->ro_media) { 1861 ubifs_msg(c, "cannot re-mount R/W - UBI volume is R/O"); 1862 return -EROFS; 1863 } 1864 err = ubifs_remount_rw(c); 1865 if (err) 1866 return err; 1867 } else if (!c->ro_mount && (*flags & MS_RDONLY)) { 1868 if (c->ro_error) { 1869 ubifs_msg(c, "cannot re-mount R/O due to prior errors"); 1870 return -EROFS; 1871 } 1872 ubifs_remount_ro(c); 1873 } 1874 1875 if (c->bulk_read == 1) 1876 bu_init(c); 1877 else { 1878 dbg_gen("disable bulk-read"); 1879 mutex_lock(&c->bu_mutex); 1880 kfree(c->bu.buf); 1881 c->bu.buf = NULL; 1882 mutex_unlock(&c->bu_mutex); 1883 } 1884 1885 ubifs_assert(c->lst.taken_empty_lebs > 0); 1886 return 0; 1887 } 1888 1889 const struct super_operations ubifs_super_operations = { 1890 .alloc_inode = ubifs_alloc_inode, 1891 .destroy_inode = ubifs_destroy_inode, 1892 .put_super = ubifs_put_super, 1893 .write_inode = ubifs_write_inode, 1894 .evict_inode = ubifs_evict_inode, 1895 .statfs = ubifs_statfs, 1896 .dirty_inode = ubifs_dirty_inode, 1897 .remount_fs = ubifs_remount_fs, 1898 .show_options = ubifs_show_options, 1899 .sync_fs = ubifs_sync_fs, 1900 }; 1901 1902 /** 1903 * open_ubi - parse UBI device name string and open the UBI device. 1904 * @name: UBI volume name 1905 * @mode: UBI volume open mode 1906 * 1907 * The primary method of mounting UBIFS is by specifying the UBI volume 1908 * character device node path. However, UBIFS may also be mounted withoug any 1909 * character device node using one of the following methods: 1910 * 1911 * o ubiX_Y - mount UBI device number X, volume Y; 1912 * o ubiY - mount UBI device number 0, volume Y; 1913 * o ubiX:NAME - mount UBI device X, volume with name NAME; 1914 * o ubi:NAME - mount UBI device 0, volume with name NAME. 1915 * 1916 * Alternative '!' separator may be used instead of ':' (because some shells 1917 * like busybox may interpret ':' as an NFS host name separator). This function 1918 * returns UBI volume description object in case of success and a negative 1919 * error code in case of failure. 1920 */ 1921 static struct ubi_volume_desc *open_ubi(const char *name, int mode) 1922 { 1923 struct ubi_volume_desc *ubi; 1924 int dev, vol; 1925 char *endptr; 1926 1927 /* First, try to open using the device node path method */ 1928 ubi = ubi_open_volume_path(name, mode); 1929 if (!IS_ERR(ubi)) 1930 return ubi; 1931 1932 /* Try the "nodev" method */ 1933 if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i') 1934 return ERR_PTR(-EINVAL); 1935 1936 /* ubi:NAME method */ 1937 if ((name[3] == ':' || name[3] == '!') && name[4] != '\0') 1938 return ubi_open_volume_nm(0, name + 4, mode); 1939 1940 if (!isdigit(name[3])) 1941 return ERR_PTR(-EINVAL); 1942 1943 dev = simple_strtoul(name + 3, &endptr, 0); 1944 1945 /* ubiY method */ 1946 if (*endptr == '\0') 1947 return ubi_open_volume(0, dev, mode); 1948 1949 /* ubiX_Y method */ 1950 if (*endptr == '_' && isdigit(endptr[1])) { 1951 vol = simple_strtoul(endptr + 1, &endptr, 0); 1952 if (*endptr != '\0') 1953 return ERR_PTR(-EINVAL); 1954 return ubi_open_volume(dev, vol, mode); 1955 } 1956 1957 /* ubiX:NAME method */ 1958 if ((*endptr == ':' || *endptr == '!') && endptr[1] != '\0') 1959 return ubi_open_volume_nm(dev, ++endptr, mode); 1960 1961 return ERR_PTR(-EINVAL); 1962 } 1963 1964 static struct ubifs_info *alloc_ubifs_info(struct ubi_volume_desc *ubi) 1965 { 1966 struct ubifs_info *c; 1967 1968 c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL); 1969 if (c) { 1970 spin_lock_init(&c->cnt_lock); 1971 spin_lock_init(&c->cs_lock); 1972 spin_lock_init(&c->buds_lock); 1973 spin_lock_init(&c->space_lock); 1974 spin_lock_init(&c->orphan_lock); 1975 init_rwsem(&c->commit_sem); 1976 mutex_init(&c->lp_mutex); 1977 mutex_init(&c->tnc_mutex); 1978 mutex_init(&c->log_mutex); 1979 mutex_init(&c->umount_mutex); 1980 mutex_init(&c->bu_mutex); 1981 mutex_init(&c->write_reserve_mutex); 1982 init_waitqueue_head(&c->cmt_wq); 1983 c->buds = RB_ROOT; 1984 c->old_idx = RB_ROOT; 1985 c->size_tree = RB_ROOT; 1986 c->orph_tree = RB_ROOT; 1987 INIT_LIST_HEAD(&c->infos_list); 1988 INIT_LIST_HEAD(&c->idx_gc); 1989 INIT_LIST_HEAD(&c->replay_list); 1990 INIT_LIST_HEAD(&c->replay_buds); 1991 INIT_LIST_HEAD(&c->uncat_list); 1992 INIT_LIST_HEAD(&c->empty_list); 1993 INIT_LIST_HEAD(&c->freeable_list); 1994 INIT_LIST_HEAD(&c->frdi_idx_list); 1995 INIT_LIST_HEAD(&c->unclean_leb_list); 1996 INIT_LIST_HEAD(&c->old_buds); 1997 INIT_LIST_HEAD(&c->orph_list); 1998 INIT_LIST_HEAD(&c->orph_new); 1999 c->no_chk_data_crc = 1; 2000 2001 c->highest_inum = UBIFS_FIRST_INO; 2002 c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM; 2003 2004 ubi_get_volume_info(ubi, &c->vi); 2005 ubi_get_device_info(c->vi.ubi_num, &c->di); 2006 } 2007 return c; 2008 } 2009 2010 #ifndef CONFIG_UBIFS_FS_ENCRYPTION 2011 const struct fscrypt_operations ubifs_crypt_operations = { 2012 .is_encrypted = __ubifs_crypt_is_encrypted, 2013 }; 2014 #endif 2015 2016 static int ubifs_fill_super(struct super_block *sb, void *data, int silent) 2017 { 2018 struct ubifs_info *c = sb->s_fs_info; 2019 struct inode *root; 2020 int err; 2021 2022 c->vfs_sb = sb; 2023 /* Re-open the UBI device in read-write mode */ 2024 c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE); 2025 if (IS_ERR(c->ubi)) { 2026 err = PTR_ERR(c->ubi); 2027 goto out; 2028 } 2029 2030 err = ubifs_parse_options(c, data, 0); 2031 if (err) 2032 goto out_close; 2033 2034 /* 2035 * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For 2036 * UBIFS, I/O is not deferred, it is done immediately in readpage, 2037 * which means the user would have to wait not just for their own I/O 2038 * but the read-ahead I/O as well i.e. completely pointless. 2039 * 2040 * Read-ahead will be disabled because @sb->s_bdi->ra_pages is 0. Also 2041 * @sb->s_bdi->capabilities are initialized to 0 so there won't be any 2042 * writeback happening. 2043 */ 2044 err = super_setup_bdi_name(sb, "ubifs_%d_%d", c->vi.ubi_num, 2045 c->vi.vol_id); 2046 if (err) 2047 goto out_close; 2048 2049 sb->s_fs_info = c; 2050 sb->s_magic = UBIFS_SUPER_MAGIC; 2051 sb->s_blocksize = UBIFS_BLOCK_SIZE; 2052 sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT; 2053 sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c); 2054 if (c->max_inode_sz > MAX_LFS_FILESIZE) 2055 sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE; 2056 sb->s_op = &ubifs_super_operations; 2057 sb->s_xattr = ubifs_xattr_handlers; 2058 sb->s_cop = &ubifs_crypt_operations; 2059 2060 mutex_lock(&c->umount_mutex); 2061 err = mount_ubifs(c); 2062 if (err) { 2063 ubifs_assert(err < 0); 2064 goto out_unlock; 2065 } 2066 2067 /* Read the root inode */ 2068 root = ubifs_iget(sb, UBIFS_ROOT_INO); 2069 if (IS_ERR(root)) { 2070 err = PTR_ERR(root); 2071 goto out_umount; 2072 } 2073 2074 sb->s_root = d_make_root(root); 2075 if (!sb->s_root) { 2076 err = -ENOMEM; 2077 goto out_umount; 2078 } 2079 2080 mutex_unlock(&c->umount_mutex); 2081 return 0; 2082 2083 out_umount: 2084 ubifs_umount(c); 2085 out_unlock: 2086 mutex_unlock(&c->umount_mutex); 2087 out_close: 2088 ubi_close_volume(c->ubi); 2089 out: 2090 return err; 2091 } 2092 2093 static int sb_test(struct super_block *sb, void *data) 2094 { 2095 struct ubifs_info *c1 = data; 2096 struct ubifs_info *c = sb->s_fs_info; 2097 2098 return c->vi.cdev == c1->vi.cdev; 2099 } 2100 2101 static int sb_set(struct super_block *sb, void *data) 2102 { 2103 sb->s_fs_info = data; 2104 return set_anon_super(sb, NULL); 2105 } 2106 2107 static struct dentry *ubifs_mount(struct file_system_type *fs_type, int flags, 2108 const char *name, void *data) 2109 { 2110 struct ubi_volume_desc *ubi; 2111 struct ubifs_info *c; 2112 struct super_block *sb; 2113 int err; 2114 2115 dbg_gen("name %s, flags %#x", name, flags); 2116 2117 /* 2118 * Get UBI device number and volume ID. Mount it read-only so far 2119 * because this might be a new mount point, and UBI allows only one 2120 * read-write user at a time. 2121 */ 2122 ubi = open_ubi(name, UBI_READONLY); 2123 if (IS_ERR(ubi)) { 2124 if (!(flags & MS_SILENT)) 2125 pr_err("UBIFS error (pid: %d): cannot open \"%s\", error %d", 2126 current->pid, name, (int)PTR_ERR(ubi)); 2127 return ERR_CAST(ubi); 2128 } 2129 2130 c = alloc_ubifs_info(ubi); 2131 if (!c) { 2132 err = -ENOMEM; 2133 goto out_close; 2134 } 2135 2136 dbg_gen("opened ubi%d_%d", c->vi.ubi_num, c->vi.vol_id); 2137 2138 sb = sget(fs_type, sb_test, sb_set, flags, c); 2139 if (IS_ERR(sb)) { 2140 err = PTR_ERR(sb); 2141 kfree(c); 2142 goto out_close; 2143 } 2144 2145 if (sb->s_root) { 2146 struct ubifs_info *c1 = sb->s_fs_info; 2147 kfree(c); 2148 /* A new mount point for already mounted UBIFS */ 2149 dbg_gen("this ubi volume is already mounted"); 2150 if (!!(flags & MS_RDONLY) != c1->ro_mount) { 2151 err = -EBUSY; 2152 goto out_deact; 2153 } 2154 } else { 2155 err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0); 2156 if (err) 2157 goto out_deact; 2158 /* We do not support atime */ 2159 sb->s_flags |= MS_ACTIVE; 2160 #ifndef CONFIG_UBIFS_ATIME_SUPPORT 2161 sb->s_flags |= MS_NOATIME; 2162 #else 2163 ubifs_msg(c, "full atime support is enabled."); 2164 #endif 2165 } 2166 2167 /* 'fill_super()' opens ubi again so we must close it here */ 2168 ubi_close_volume(ubi); 2169 2170 return dget(sb->s_root); 2171 2172 out_deact: 2173 deactivate_locked_super(sb); 2174 out_close: 2175 ubi_close_volume(ubi); 2176 return ERR_PTR(err); 2177 } 2178 2179 static void kill_ubifs_super(struct super_block *s) 2180 { 2181 struct ubifs_info *c = s->s_fs_info; 2182 kill_anon_super(s); 2183 kfree(c); 2184 } 2185 2186 static struct file_system_type ubifs_fs_type = { 2187 .name = "ubifs", 2188 .owner = THIS_MODULE, 2189 .mount = ubifs_mount, 2190 .kill_sb = kill_ubifs_super, 2191 }; 2192 MODULE_ALIAS_FS("ubifs"); 2193 2194 /* 2195 * Inode slab cache constructor. 2196 */ 2197 static void inode_slab_ctor(void *obj) 2198 { 2199 struct ubifs_inode *ui = obj; 2200 inode_init_once(&ui->vfs_inode); 2201 } 2202 2203 static int __init ubifs_init(void) 2204 { 2205 int err; 2206 2207 BUILD_BUG_ON(sizeof(struct ubifs_ch) != 24); 2208 2209 /* Make sure node sizes are 8-byte aligned */ 2210 BUILD_BUG_ON(UBIFS_CH_SZ & 7); 2211 BUILD_BUG_ON(UBIFS_INO_NODE_SZ & 7); 2212 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ & 7); 2213 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ & 7); 2214 BUILD_BUG_ON(UBIFS_DATA_NODE_SZ & 7); 2215 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ & 7); 2216 BUILD_BUG_ON(UBIFS_SB_NODE_SZ & 7); 2217 BUILD_BUG_ON(UBIFS_MST_NODE_SZ & 7); 2218 BUILD_BUG_ON(UBIFS_REF_NODE_SZ & 7); 2219 BUILD_BUG_ON(UBIFS_CS_NODE_SZ & 7); 2220 BUILD_BUG_ON(UBIFS_ORPH_NODE_SZ & 7); 2221 2222 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ & 7); 2223 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ & 7); 2224 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ & 7); 2225 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ & 7); 2226 BUILD_BUG_ON(UBIFS_MAX_NODE_SZ & 7); 2227 BUILD_BUG_ON(MIN_WRITE_SZ & 7); 2228 2229 /* Check min. node size */ 2230 BUILD_BUG_ON(UBIFS_INO_NODE_SZ < MIN_WRITE_SZ); 2231 BUILD_BUG_ON(UBIFS_DENT_NODE_SZ < MIN_WRITE_SZ); 2232 BUILD_BUG_ON(UBIFS_XENT_NODE_SZ < MIN_WRITE_SZ); 2233 BUILD_BUG_ON(UBIFS_TRUN_NODE_SZ < MIN_WRITE_SZ); 2234 2235 BUILD_BUG_ON(UBIFS_MAX_DENT_NODE_SZ > UBIFS_MAX_NODE_SZ); 2236 BUILD_BUG_ON(UBIFS_MAX_XENT_NODE_SZ > UBIFS_MAX_NODE_SZ); 2237 BUILD_BUG_ON(UBIFS_MAX_DATA_NODE_SZ > UBIFS_MAX_NODE_SZ); 2238 BUILD_BUG_ON(UBIFS_MAX_INO_NODE_SZ > UBIFS_MAX_NODE_SZ); 2239 2240 /* Defined node sizes */ 2241 BUILD_BUG_ON(UBIFS_SB_NODE_SZ != 4096); 2242 BUILD_BUG_ON(UBIFS_MST_NODE_SZ != 512); 2243 BUILD_BUG_ON(UBIFS_INO_NODE_SZ != 160); 2244 BUILD_BUG_ON(UBIFS_REF_NODE_SZ != 64); 2245 2246 /* 2247 * We use 2 bit wide bit-fields to store compression type, which should 2248 * be amended if more compressors are added. The bit-fields are: 2249 * @compr_type in 'struct ubifs_inode', @default_compr in 2250 * 'struct ubifs_info' and @compr_type in 'struct ubifs_mount_opts'. 2251 */ 2252 BUILD_BUG_ON(UBIFS_COMPR_TYPES_CNT > 4); 2253 2254 /* 2255 * We require that PAGE_SIZE is greater-than-or-equal-to 2256 * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2. 2257 */ 2258 if (PAGE_SIZE < UBIFS_BLOCK_SIZE) { 2259 pr_err("UBIFS error (pid %d): VFS page cache size is %u bytes, but UBIFS requires at least 4096 bytes", 2260 current->pid, (unsigned int)PAGE_SIZE); 2261 return -EINVAL; 2262 } 2263 2264 ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab", 2265 sizeof(struct ubifs_inode), 0, 2266 SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT | 2267 SLAB_ACCOUNT, &inode_slab_ctor); 2268 if (!ubifs_inode_slab) 2269 return -ENOMEM; 2270 2271 err = register_shrinker(&ubifs_shrinker_info); 2272 if (err) 2273 goto out_slab; 2274 2275 err = ubifs_compressors_init(); 2276 if (err) 2277 goto out_shrinker; 2278 2279 err = dbg_debugfs_init(); 2280 if (err) 2281 goto out_compr; 2282 2283 err = register_filesystem(&ubifs_fs_type); 2284 if (err) { 2285 pr_err("UBIFS error (pid %d): cannot register file system, error %d", 2286 current->pid, err); 2287 goto out_dbg; 2288 } 2289 return 0; 2290 2291 out_dbg: 2292 dbg_debugfs_exit(); 2293 out_compr: 2294 ubifs_compressors_exit(); 2295 out_shrinker: 2296 unregister_shrinker(&ubifs_shrinker_info); 2297 out_slab: 2298 kmem_cache_destroy(ubifs_inode_slab); 2299 return err; 2300 } 2301 /* late_initcall to let compressors initialize first */ 2302 late_initcall(ubifs_init); 2303 2304 static void __exit ubifs_exit(void) 2305 { 2306 ubifs_assert(list_empty(&ubifs_infos)); 2307 ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0); 2308 2309 dbg_debugfs_exit(); 2310 ubifs_compressors_exit(); 2311 unregister_shrinker(&ubifs_shrinker_info); 2312 2313 /* 2314 * Make sure all delayed rcu free inodes are flushed before we 2315 * destroy cache. 2316 */ 2317 rcu_barrier(); 2318 kmem_cache_destroy(ubifs_inode_slab); 2319 unregister_filesystem(&ubifs_fs_type); 2320 } 2321 module_exit(ubifs_exit); 2322 2323 MODULE_LICENSE("GPL"); 2324 MODULE_VERSION(__stringify(UBIFS_VERSION)); 2325 MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter"); 2326 MODULE_DESCRIPTION("UBIFS - UBI File System"); 2327