1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * linux/fs/ext4/super.c 4 * 5 * Copyright (C) 1992, 1993, 1994, 1995 6 * Remy Card (card@masi.ibp.fr) 7 * Laboratoire MASI - Institut Blaise Pascal 8 * Universite Pierre et Marie Curie (Paris VI) 9 * 10 * from 11 * 12 * linux/fs/minix/inode.c 13 * 14 * Copyright (C) 1991, 1992 Linus Torvalds 15 * 16 * Big-endian to little-endian byte-swapping/bitmaps by 17 * David S. Miller (davem@caip.rutgers.edu), 1995 18 */ 19 20 #include <linux/module.h> 21 #include <linux/string.h> 22 #include <linux/fs.h> 23 #include <linux/time.h> 24 #include <linux/vmalloc.h> 25 #include <linux/slab.h> 26 #include <linux/init.h> 27 #include <linux/blkdev.h> 28 #include <linux/backing-dev.h> 29 #include <linux/parser.h> 30 #include <linux/buffer_head.h> 31 #include <linux/exportfs.h> 32 #include <linux/vfs.h> 33 #include <linux/random.h> 34 #include <linux/mount.h> 35 #include <linux/namei.h> 36 #include <linux/quotaops.h> 37 #include <linux/seq_file.h> 38 #include <linux/ctype.h> 39 #include <linux/log2.h> 40 #include <linux/crc16.h> 41 #include <linux/dax.h> 42 #include <linux/cleancache.h> 43 #include <linux/uaccess.h> 44 #include <linux/iversion.h> 45 #include <linux/unicode.h> 46 #include <linux/part_stat.h> 47 #include <linux/kthread.h> 48 #include <linux/freezer.h> 49 #include <linux/fsnotify.h> 50 #include <linux/fs_context.h> 51 #include <linux/fs_parser.h> 52 53 #include "ext4.h" 54 #include "ext4_extents.h" /* Needed for trace points definition */ 55 #include "ext4_jbd2.h" 56 #include "xattr.h" 57 #include "acl.h" 58 #include "mballoc.h" 59 #include "fsmap.h" 60 61 #define CREATE_TRACE_POINTS 62 #include <trace/events/ext4.h> 63 64 static struct ext4_lazy_init *ext4_li_info; 65 static DEFINE_MUTEX(ext4_li_mtx); 66 static struct ratelimit_state ext4_mount_msg_ratelimit; 67 68 static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 69 unsigned long journal_devnum); 70 static int ext4_show_options(struct seq_file *seq, struct dentry *root); 71 static void ext4_update_super(struct super_block *sb); 72 static int ext4_commit_super(struct super_block *sb); 73 static int ext4_mark_recovery_complete(struct super_block *sb, 74 struct ext4_super_block *es); 75 static int ext4_clear_journal_err(struct super_block *sb, 76 struct ext4_super_block *es); 77 static int ext4_sync_fs(struct super_block *sb, int wait); 78 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 79 static int ext4_unfreeze(struct super_block *sb); 80 static int ext4_freeze(struct super_block *sb); 81 static inline int ext2_feature_set_ok(struct super_block *sb); 82 static inline int ext3_feature_set_ok(struct super_block *sb); 83 static void ext4_destroy_lazyinit_thread(void); 84 static void ext4_unregister_li_request(struct super_block *sb); 85 static void ext4_clear_request_list(void); 86 static struct inode *ext4_get_journal_inode(struct super_block *sb, 87 unsigned int journal_inum); 88 static int ext4_validate_options(struct fs_context *fc); 89 static int ext4_check_opt_consistency(struct fs_context *fc, 90 struct super_block *sb); 91 static int ext4_apply_options(struct fs_context *fc, struct super_block *sb); 92 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param); 93 static int ext4_get_tree(struct fs_context *fc); 94 static int ext4_reconfigure(struct fs_context *fc); 95 static void ext4_fc_free(struct fs_context *fc); 96 static int ext4_init_fs_context(struct fs_context *fc); 97 static const struct fs_parameter_spec ext4_param_specs[]; 98 99 /* 100 * Lock ordering 101 * 102 * page fault path: 103 * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start 104 * -> page lock -> i_data_sem (rw) 105 * 106 * buffered write path: 107 * sb_start_write -> i_mutex -> mmap_lock 108 * sb_start_write -> i_mutex -> transaction start -> page lock -> 109 * i_data_sem (rw) 110 * 111 * truncate: 112 * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) -> 113 * page lock 114 * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start -> 115 * i_data_sem (rw) 116 * 117 * direct IO: 118 * sb_start_write -> i_mutex -> mmap_lock 119 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw) 120 * 121 * writepages: 122 * transaction start -> page lock(s) -> i_data_sem (rw) 123 */ 124 125 static const struct fs_context_operations ext4_context_ops = { 126 .parse_param = ext4_parse_param, 127 .get_tree = ext4_get_tree, 128 .reconfigure = ext4_reconfigure, 129 .free = ext4_fc_free, 130 }; 131 132 133 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 134 static struct file_system_type ext2_fs_type = { 135 .owner = THIS_MODULE, 136 .name = "ext2", 137 .init_fs_context = ext4_init_fs_context, 138 .parameters = ext4_param_specs, 139 .kill_sb = kill_block_super, 140 .fs_flags = FS_REQUIRES_DEV, 141 }; 142 MODULE_ALIAS_FS("ext2"); 143 MODULE_ALIAS("ext2"); 144 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) 145 #else 146 #define IS_EXT2_SB(sb) (0) 147 #endif 148 149 150 static struct file_system_type ext3_fs_type = { 151 .owner = THIS_MODULE, 152 .name = "ext3", 153 .init_fs_context = ext4_init_fs_context, 154 .parameters = ext4_param_specs, 155 .kill_sb = kill_block_super, 156 .fs_flags = FS_REQUIRES_DEV, 157 }; 158 MODULE_ALIAS_FS("ext3"); 159 MODULE_ALIAS("ext3"); 160 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) 161 162 163 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags, 164 bh_end_io_t *end_io) 165 { 166 /* 167 * buffer's verified bit is no longer valid after reading from 168 * disk again due to write out error, clear it to make sure we 169 * recheck the buffer contents. 170 */ 171 clear_buffer_verified(bh); 172 173 bh->b_end_io = end_io ? end_io : end_buffer_read_sync; 174 get_bh(bh); 175 submit_bh(REQ_OP_READ, op_flags, bh); 176 } 177 178 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags, 179 bh_end_io_t *end_io) 180 { 181 BUG_ON(!buffer_locked(bh)); 182 183 if (ext4_buffer_uptodate(bh)) { 184 unlock_buffer(bh); 185 return; 186 } 187 __ext4_read_bh(bh, op_flags, end_io); 188 } 189 190 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io) 191 { 192 BUG_ON(!buffer_locked(bh)); 193 194 if (ext4_buffer_uptodate(bh)) { 195 unlock_buffer(bh); 196 return 0; 197 } 198 199 __ext4_read_bh(bh, op_flags, end_io); 200 201 wait_on_buffer(bh); 202 if (buffer_uptodate(bh)) 203 return 0; 204 return -EIO; 205 } 206 207 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait) 208 { 209 if (trylock_buffer(bh)) { 210 if (wait) 211 return ext4_read_bh(bh, op_flags, NULL); 212 ext4_read_bh_nowait(bh, op_flags, NULL); 213 return 0; 214 } 215 if (wait) { 216 wait_on_buffer(bh); 217 if (buffer_uptodate(bh)) 218 return 0; 219 return -EIO; 220 } 221 return 0; 222 } 223 224 /* 225 * This works like __bread_gfp() except it uses ERR_PTR for error 226 * returns. Currently with sb_bread it's impossible to distinguish 227 * between ENOMEM and EIO situations (since both result in a NULL 228 * return. 229 */ 230 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb, 231 sector_t block, int op_flags, 232 gfp_t gfp) 233 { 234 struct buffer_head *bh; 235 int ret; 236 237 bh = sb_getblk_gfp(sb, block, gfp); 238 if (bh == NULL) 239 return ERR_PTR(-ENOMEM); 240 if (ext4_buffer_uptodate(bh)) 241 return bh; 242 243 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true); 244 if (ret) { 245 put_bh(bh); 246 return ERR_PTR(ret); 247 } 248 return bh; 249 } 250 251 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block, 252 int op_flags) 253 { 254 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE); 255 } 256 257 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb, 258 sector_t block) 259 { 260 return __ext4_sb_bread_gfp(sb, block, 0, 0); 261 } 262 263 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block) 264 { 265 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0); 266 267 if (likely(bh)) { 268 ext4_read_bh_lock(bh, REQ_RAHEAD, false); 269 brelse(bh); 270 } 271 } 272 273 static int ext4_verify_csum_type(struct super_block *sb, 274 struct ext4_super_block *es) 275 { 276 if (!ext4_has_feature_metadata_csum(sb)) 277 return 1; 278 279 return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 280 } 281 282 __le32 ext4_superblock_csum(struct super_block *sb, 283 struct ext4_super_block *es) 284 { 285 struct ext4_sb_info *sbi = EXT4_SB(sb); 286 int offset = offsetof(struct ext4_super_block, s_checksum); 287 __u32 csum; 288 289 csum = ext4_chksum(sbi, ~0, (char *)es, offset); 290 291 return cpu_to_le32(csum); 292 } 293 294 static int ext4_superblock_csum_verify(struct super_block *sb, 295 struct ext4_super_block *es) 296 { 297 if (!ext4_has_metadata_csum(sb)) 298 return 1; 299 300 return es->s_checksum == ext4_superblock_csum(sb, es); 301 } 302 303 void ext4_superblock_csum_set(struct super_block *sb) 304 { 305 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 306 307 if (!ext4_has_metadata_csum(sb)) 308 return; 309 310 es->s_checksum = ext4_superblock_csum(sb, es); 311 } 312 313 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 314 struct ext4_group_desc *bg) 315 { 316 return le32_to_cpu(bg->bg_block_bitmap_lo) | 317 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 318 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 319 } 320 321 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 322 struct ext4_group_desc *bg) 323 { 324 return le32_to_cpu(bg->bg_inode_bitmap_lo) | 325 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 326 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 327 } 328 329 ext4_fsblk_t ext4_inode_table(struct super_block *sb, 330 struct ext4_group_desc *bg) 331 { 332 return le32_to_cpu(bg->bg_inode_table_lo) | 333 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 334 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 335 } 336 337 __u32 ext4_free_group_clusters(struct super_block *sb, 338 struct ext4_group_desc *bg) 339 { 340 return le16_to_cpu(bg->bg_free_blocks_count_lo) | 341 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 342 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 343 } 344 345 __u32 ext4_free_inodes_count(struct super_block *sb, 346 struct ext4_group_desc *bg) 347 { 348 return le16_to_cpu(bg->bg_free_inodes_count_lo) | 349 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 350 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); 351 } 352 353 __u32 ext4_used_dirs_count(struct super_block *sb, 354 struct ext4_group_desc *bg) 355 { 356 return le16_to_cpu(bg->bg_used_dirs_count_lo) | 357 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 358 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 359 } 360 361 __u32 ext4_itable_unused_count(struct super_block *sb, 362 struct ext4_group_desc *bg) 363 { 364 return le16_to_cpu(bg->bg_itable_unused_lo) | 365 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 366 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 367 } 368 369 void ext4_block_bitmap_set(struct super_block *sb, 370 struct ext4_group_desc *bg, ext4_fsblk_t blk) 371 { 372 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 373 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 374 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 375 } 376 377 void ext4_inode_bitmap_set(struct super_block *sb, 378 struct ext4_group_desc *bg, ext4_fsblk_t blk) 379 { 380 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 381 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 382 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 383 } 384 385 void ext4_inode_table_set(struct super_block *sb, 386 struct ext4_group_desc *bg, ext4_fsblk_t blk) 387 { 388 bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 389 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 390 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 391 } 392 393 void ext4_free_group_clusters_set(struct super_block *sb, 394 struct ext4_group_desc *bg, __u32 count) 395 { 396 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 397 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 398 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 399 } 400 401 void ext4_free_inodes_set(struct super_block *sb, 402 struct ext4_group_desc *bg, __u32 count) 403 { 404 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); 405 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 406 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); 407 } 408 409 void ext4_used_dirs_set(struct super_block *sb, 410 struct ext4_group_desc *bg, __u32 count) 411 { 412 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 413 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 414 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 415 } 416 417 void ext4_itable_unused_set(struct super_block *sb, 418 struct ext4_group_desc *bg, __u32 count) 419 { 420 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 421 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 422 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 423 } 424 425 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now) 426 { 427 now = clamp_val(now, 0, (1ull << 40) - 1); 428 429 *lo = cpu_to_le32(lower_32_bits(now)); 430 *hi = upper_32_bits(now); 431 } 432 433 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi) 434 { 435 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo); 436 } 437 #define ext4_update_tstamp(es, tstamp) \ 438 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \ 439 ktime_get_real_seconds()) 440 #define ext4_get_tstamp(es, tstamp) \ 441 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 442 443 /* 444 * The del_gendisk() function uninitializes the disk-specific data 445 * structures, including the bdi structure, without telling anyone 446 * else. Once this happens, any attempt to call mark_buffer_dirty() 447 * (for example, by ext4_commit_super), will cause a kernel OOPS. 448 * This is a kludge to prevent these oops until we can put in a proper 449 * hook in del_gendisk() to inform the VFS and file system layers. 450 */ 451 static int block_device_ejected(struct super_block *sb) 452 { 453 struct inode *bd_inode = sb->s_bdev->bd_inode; 454 struct backing_dev_info *bdi = inode_to_bdi(bd_inode); 455 456 return bdi->dev == NULL; 457 } 458 459 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 460 { 461 struct super_block *sb = journal->j_private; 462 struct ext4_sb_info *sbi = EXT4_SB(sb); 463 int error = is_journal_aborted(journal); 464 struct ext4_journal_cb_entry *jce; 465 466 BUG_ON(txn->t_state == T_FINISHED); 467 468 ext4_process_freed_data(sb, txn->t_tid); 469 470 spin_lock(&sbi->s_md_lock); 471 while (!list_empty(&txn->t_private_list)) { 472 jce = list_entry(txn->t_private_list.next, 473 struct ext4_journal_cb_entry, jce_list); 474 list_del_init(&jce->jce_list); 475 spin_unlock(&sbi->s_md_lock); 476 jce->jce_func(sb, jce, error); 477 spin_lock(&sbi->s_md_lock); 478 } 479 spin_unlock(&sbi->s_md_lock); 480 } 481 482 /* 483 * This writepage callback for write_cache_pages() 484 * takes care of a few cases after page cleaning. 485 * 486 * write_cache_pages() already checks for dirty pages 487 * and calls clear_page_dirty_for_io(), which we want, 488 * to write protect the pages. 489 * 490 * However, we may have to redirty a page (see below.) 491 */ 492 static int ext4_journalled_writepage_callback(struct page *page, 493 struct writeback_control *wbc, 494 void *data) 495 { 496 transaction_t *transaction = (transaction_t *) data; 497 struct buffer_head *bh, *head; 498 struct journal_head *jh; 499 500 bh = head = page_buffers(page); 501 do { 502 /* 503 * We have to redirty a page in these cases: 504 * 1) If buffer is dirty, it means the page was dirty because it 505 * contains a buffer that needs checkpointing. So the dirty bit 506 * needs to be preserved so that checkpointing writes the buffer 507 * properly. 508 * 2) If buffer is not part of the committing transaction 509 * (we may have just accidentally come across this buffer because 510 * inode range tracking is not exact) or if the currently running 511 * transaction already contains this buffer as well, dirty bit 512 * needs to be preserved so that the buffer gets writeprotected 513 * properly on running transaction's commit. 514 */ 515 jh = bh2jh(bh); 516 if (buffer_dirty(bh) || 517 (jh && (jh->b_transaction != transaction || 518 jh->b_next_transaction))) { 519 redirty_page_for_writepage(wbc, page); 520 goto out; 521 } 522 } while ((bh = bh->b_this_page) != head); 523 524 out: 525 return AOP_WRITEPAGE_ACTIVATE; 526 } 527 528 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode) 529 { 530 struct address_space *mapping = jinode->i_vfs_inode->i_mapping; 531 struct writeback_control wbc = { 532 .sync_mode = WB_SYNC_ALL, 533 .nr_to_write = LONG_MAX, 534 .range_start = jinode->i_dirty_start, 535 .range_end = jinode->i_dirty_end, 536 }; 537 538 return write_cache_pages(mapping, &wbc, 539 ext4_journalled_writepage_callback, 540 jinode->i_transaction); 541 } 542 543 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode) 544 { 545 int ret; 546 547 if (ext4_should_journal_data(jinode->i_vfs_inode)) 548 ret = ext4_journalled_submit_inode_data_buffers(jinode); 549 else 550 ret = jbd2_journal_submit_inode_data_buffers(jinode); 551 552 return ret; 553 } 554 555 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode) 556 { 557 int ret = 0; 558 559 if (!ext4_should_journal_data(jinode->i_vfs_inode)) 560 ret = jbd2_journal_finish_inode_data_buffers(jinode); 561 562 return ret; 563 } 564 565 static bool system_going_down(void) 566 { 567 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 568 || system_state == SYSTEM_RESTART; 569 } 570 571 struct ext4_err_translation { 572 int code; 573 int errno; 574 }; 575 576 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err } 577 578 static struct ext4_err_translation err_translation[] = { 579 EXT4_ERR_TRANSLATE(EIO), 580 EXT4_ERR_TRANSLATE(ENOMEM), 581 EXT4_ERR_TRANSLATE(EFSBADCRC), 582 EXT4_ERR_TRANSLATE(EFSCORRUPTED), 583 EXT4_ERR_TRANSLATE(ENOSPC), 584 EXT4_ERR_TRANSLATE(ENOKEY), 585 EXT4_ERR_TRANSLATE(EROFS), 586 EXT4_ERR_TRANSLATE(EFBIG), 587 EXT4_ERR_TRANSLATE(EEXIST), 588 EXT4_ERR_TRANSLATE(ERANGE), 589 EXT4_ERR_TRANSLATE(EOVERFLOW), 590 EXT4_ERR_TRANSLATE(EBUSY), 591 EXT4_ERR_TRANSLATE(ENOTDIR), 592 EXT4_ERR_TRANSLATE(ENOTEMPTY), 593 EXT4_ERR_TRANSLATE(ESHUTDOWN), 594 EXT4_ERR_TRANSLATE(EFAULT), 595 }; 596 597 static int ext4_errno_to_code(int errno) 598 { 599 int i; 600 601 for (i = 0; i < ARRAY_SIZE(err_translation); i++) 602 if (err_translation[i].errno == errno) 603 return err_translation[i].code; 604 return EXT4_ERR_UNKNOWN; 605 } 606 607 static void save_error_info(struct super_block *sb, int error, 608 __u32 ino, __u64 block, 609 const char *func, unsigned int line) 610 { 611 struct ext4_sb_info *sbi = EXT4_SB(sb); 612 613 /* We default to EFSCORRUPTED error... */ 614 if (error == 0) 615 error = EFSCORRUPTED; 616 617 spin_lock(&sbi->s_error_lock); 618 sbi->s_add_error_count++; 619 sbi->s_last_error_code = error; 620 sbi->s_last_error_line = line; 621 sbi->s_last_error_ino = ino; 622 sbi->s_last_error_block = block; 623 sbi->s_last_error_func = func; 624 sbi->s_last_error_time = ktime_get_real_seconds(); 625 if (!sbi->s_first_error_time) { 626 sbi->s_first_error_code = error; 627 sbi->s_first_error_line = line; 628 sbi->s_first_error_ino = ino; 629 sbi->s_first_error_block = block; 630 sbi->s_first_error_func = func; 631 sbi->s_first_error_time = sbi->s_last_error_time; 632 } 633 spin_unlock(&sbi->s_error_lock); 634 } 635 636 /* Deal with the reporting of failure conditions on a filesystem such as 637 * inconsistencies detected or read IO failures. 638 * 639 * On ext2, we can store the error state of the filesystem in the 640 * superblock. That is not possible on ext4, because we may have other 641 * write ordering constraints on the superblock which prevent us from 642 * writing it out straight away; and given that the journal is about to 643 * be aborted, we can't rely on the current, or future, transactions to 644 * write out the superblock safely. 645 * 646 * We'll just use the jbd2_journal_abort() error code to record an error in 647 * the journal instead. On recovery, the journal will complain about 648 * that error until we've noted it down and cleared it. 649 * 650 * If force_ro is set, we unconditionally force the filesystem into an 651 * ABORT|READONLY state, unless the error response on the fs has been set to 652 * panic in which case we take the easy way out and panic immediately. This is 653 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM 654 * at a critical moment in log management. 655 */ 656 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error, 657 __u32 ino, __u64 block, 658 const char *func, unsigned int line) 659 { 660 journal_t *journal = EXT4_SB(sb)->s_journal; 661 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT); 662 663 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 664 if (test_opt(sb, WARN_ON_ERROR)) 665 WARN_ON_ONCE(1); 666 667 if (!continue_fs && !sb_rdonly(sb)) { 668 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED); 669 if (journal) 670 jbd2_journal_abort(journal, -EIO); 671 } 672 673 if (!bdev_read_only(sb->s_bdev)) { 674 save_error_info(sb, error, ino, block, func, line); 675 /* 676 * In case the fs should keep running, we need to writeout 677 * superblock through the journal. Due to lock ordering 678 * constraints, it may not be safe to do it right here so we 679 * defer superblock flushing to a workqueue. 680 */ 681 if (continue_fs && journal) 682 schedule_work(&EXT4_SB(sb)->s_error_work); 683 else 684 ext4_commit_super(sb); 685 } 686 687 /* 688 * We force ERRORS_RO behavior when system is rebooting. Otherwise we 689 * could panic during 'reboot -f' as the underlying device got already 690 * disabled. 691 */ 692 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) { 693 panic("EXT4-fs (device %s): panic forced after error\n", 694 sb->s_id); 695 } 696 697 if (sb_rdonly(sb) || continue_fs) 698 return; 699 700 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 701 /* 702 * Make sure updated value of ->s_mount_flags will be visible before 703 * ->s_flags update 704 */ 705 smp_wmb(); 706 sb->s_flags |= SB_RDONLY; 707 } 708 709 static void flush_stashed_error_work(struct work_struct *work) 710 { 711 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info, 712 s_error_work); 713 journal_t *journal = sbi->s_journal; 714 handle_t *handle; 715 716 /* 717 * If the journal is still running, we have to write out superblock 718 * through the journal to avoid collisions of other journalled sb 719 * updates. 720 * 721 * We use directly jbd2 functions here to avoid recursing back into 722 * ext4 error handling code during handling of previous errors. 723 */ 724 if (!sb_rdonly(sbi->s_sb) && journal) { 725 struct buffer_head *sbh = sbi->s_sbh; 726 handle = jbd2_journal_start(journal, 1); 727 if (IS_ERR(handle)) 728 goto write_directly; 729 if (jbd2_journal_get_write_access(handle, sbh)) { 730 jbd2_journal_stop(handle); 731 goto write_directly; 732 } 733 ext4_update_super(sbi->s_sb); 734 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 735 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to " 736 "superblock detected"); 737 clear_buffer_write_io_error(sbh); 738 set_buffer_uptodate(sbh); 739 } 740 741 if (jbd2_journal_dirty_metadata(handle, sbh)) { 742 jbd2_journal_stop(handle); 743 goto write_directly; 744 } 745 jbd2_journal_stop(handle); 746 ext4_notify_error_sysfs(sbi); 747 return; 748 } 749 write_directly: 750 /* 751 * Write through journal failed. Write sb directly to get error info 752 * out and hope for the best. 753 */ 754 ext4_commit_super(sbi->s_sb); 755 ext4_notify_error_sysfs(sbi); 756 } 757 758 #define ext4_error_ratelimit(sb) \ 759 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ 760 "EXT4-fs error") 761 762 void __ext4_error(struct super_block *sb, const char *function, 763 unsigned int line, bool force_ro, int error, __u64 block, 764 const char *fmt, ...) 765 { 766 struct va_format vaf; 767 va_list args; 768 769 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 770 return; 771 772 trace_ext4_error(sb, function, line); 773 if (ext4_error_ratelimit(sb)) { 774 va_start(args, fmt); 775 vaf.fmt = fmt; 776 vaf.va = &args; 777 printk(KERN_CRIT 778 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 779 sb->s_id, function, line, current->comm, &vaf); 780 va_end(args); 781 } 782 fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED); 783 784 ext4_handle_error(sb, force_ro, error, 0, block, function, line); 785 } 786 787 void __ext4_error_inode(struct inode *inode, const char *function, 788 unsigned int line, ext4_fsblk_t block, int error, 789 const char *fmt, ...) 790 { 791 va_list args; 792 struct va_format vaf; 793 794 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 795 return; 796 797 trace_ext4_error(inode->i_sb, function, line); 798 if (ext4_error_ratelimit(inode->i_sb)) { 799 va_start(args, fmt); 800 vaf.fmt = fmt; 801 vaf.va = &args; 802 if (block) 803 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 804 "inode #%lu: block %llu: comm %s: %pV\n", 805 inode->i_sb->s_id, function, line, inode->i_ino, 806 block, current->comm, &vaf); 807 else 808 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 809 "inode #%lu: comm %s: %pV\n", 810 inode->i_sb->s_id, function, line, inode->i_ino, 811 current->comm, &vaf); 812 va_end(args); 813 } 814 fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED); 815 816 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block, 817 function, line); 818 } 819 820 void __ext4_error_file(struct file *file, const char *function, 821 unsigned int line, ext4_fsblk_t block, 822 const char *fmt, ...) 823 { 824 va_list args; 825 struct va_format vaf; 826 struct inode *inode = file_inode(file); 827 char pathname[80], *path; 828 829 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 830 return; 831 832 trace_ext4_error(inode->i_sb, function, line); 833 if (ext4_error_ratelimit(inode->i_sb)) { 834 path = file_path(file, pathname, sizeof(pathname)); 835 if (IS_ERR(path)) 836 path = "(unknown)"; 837 va_start(args, fmt); 838 vaf.fmt = fmt; 839 vaf.va = &args; 840 if (block) 841 printk(KERN_CRIT 842 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 843 "block %llu: comm %s: path %s: %pV\n", 844 inode->i_sb->s_id, function, line, inode->i_ino, 845 block, current->comm, path, &vaf); 846 else 847 printk(KERN_CRIT 848 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 849 "comm %s: path %s: %pV\n", 850 inode->i_sb->s_id, function, line, inode->i_ino, 851 current->comm, path, &vaf); 852 va_end(args); 853 } 854 fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED); 855 856 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block, 857 function, line); 858 } 859 860 const char *ext4_decode_error(struct super_block *sb, int errno, 861 char nbuf[16]) 862 { 863 char *errstr = NULL; 864 865 switch (errno) { 866 case -EFSCORRUPTED: 867 errstr = "Corrupt filesystem"; 868 break; 869 case -EFSBADCRC: 870 errstr = "Filesystem failed CRC"; 871 break; 872 case -EIO: 873 errstr = "IO failure"; 874 break; 875 case -ENOMEM: 876 errstr = "Out of memory"; 877 break; 878 case -EROFS: 879 if (!sb || (EXT4_SB(sb)->s_journal && 880 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 881 errstr = "Journal has aborted"; 882 else 883 errstr = "Readonly filesystem"; 884 break; 885 default: 886 /* If the caller passed in an extra buffer for unknown 887 * errors, textualise them now. Else we just return 888 * NULL. */ 889 if (nbuf) { 890 /* Check for truncated error codes... */ 891 if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 892 errstr = nbuf; 893 } 894 break; 895 } 896 897 return errstr; 898 } 899 900 /* __ext4_std_error decodes expected errors from journaling functions 901 * automatically and invokes the appropriate error response. */ 902 903 void __ext4_std_error(struct super_block *sb, const char *function, 904 unsigned int line, int errno) 905 { 906 char nbuf[16]; 907 const char *errstr; 908 909 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 910 return; 911 912 /* Special case: if the error is EROFS, and we're not already 913 * inside a transaction, then there's really no point in logging 914 * an error. */ 915 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb)) 916 return; 917 918 if (ext4_error_ratelimit(sb)) { 919 errstr = ext4_decode_error(sb, errno, nbuf); 920 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 921 sb->s_id, function, line, errstr); 922 } 923 fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED); 924 925 ext4_handle_error(sb, false, -errno, 0, 0, function, line); 926 } 927 928 void __ext4_msg(struct super_block *sb, 929 const char *prefix, const char *fmt, ...) 930 { 931 struct va_format vaf; 932 va_list args; 933 934 if (sb) { 935 atomic_inc(&EXT4_SB(sb)->s_msg_count); 936 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), 937 "EXT4-fs")) 938 return; 939 } 940 941 va_start(args, fmt); 942 vaf.fmt = fmt; 943 vaf.va = &args; 944 if (sb) 945 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 946 else 947 printk("%sEXT4-fs: %pV\n", prefix, &vaf); 948 va_end(args); 949 } 950 951 static int ext4_warning_ratelimit(struct super_block *sb) 952 { 953 atomic_inc(&EXT4_SB(sb)->s_warning_count); 954 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), 955 "EXT4-fs warning"); 956 } 957 958 void __ext4_warning(struct super_block *sb, const char *function, 959 unsigned int line, const char *fmt, ...) 960 { 961 struct va_format vaf; 962 va_list args; 963 964 if (!ext4_warning_ratelimit(sb)) 965 return; 966 967 va_start(args, fmt); 968 vaf.fmt = fmt; 969 vaf.va = &args; 970 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 971 sb->s_id, function, line, &vaf); 972 va_end(args); 973 } 974 975 void __ext4_warning_inode(const struct inode *inode, const char *function, 976 unsigned int line, const char *fmt, ...) 977 { 978 struct va_format vaf; 979 va_list args; 980 981 if (!ext4_warning_ratelimit(inode->i_sb)) 982 return; 983 984 va_start(args, fmt); 985 vaf.fmt = fmt; 986 vaf.va = &args; 987 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: " 988 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id, 989 function, line, inode->i_ino, current->comm, &vaf); 990 va_end(args); 991 } 992 993 void __ext4_grp_locked_error(const char *function, unsigned int line, 994 struct super_block *sb, ext4_group_t grp, 995 unsigned long ino, ext4_fsblk_t block, 996 const char *fmt, ...) 997 __releases(bitlock) 998 __acquires(bitlock) 999 { 1000 struct va_format vaf; 1001 va_list args; 1002 1003 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 1004 return; 1005 1006 trace_ext4_error(sb, function, line); 1007 if (ext4_error_ratelimit(sb)) { 1008 va_start(args, fmt); 1009 vaf.fmt = fmt; 1010 vaf.va = &args; 1011 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 1012 sb->s_id, function, line, grp); 1013 if (ino) 1014 printk(KERN_CONT "inode %lu: ", ino); 1015 if (block) 1016 printk(KERN_CONT "block %llu:", 1017 (unsigned long long) block); 1018 printk(KERN_CONT "%pV\n", &vaf); 1019 va_end(args); 1020 } 1021 1022 if (test_opt(sb, ERRORS_CONT)) { 1023 if (test_opt(sb, WARN_ON_ERROR)) 1024 WARN_ON_ONCE(1); 1025 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 1026 if (!bdev_read_only(sb->s_bdev)) { 1027 save_error_info(sb, EFSCORRUPTED, ino, block, function, 1028 line); 1029 schedule_work(&EXT4_SB(sb)->s_error_work); 1030 } 1031 return; 1032 } 1033 ext4_unlock_group(sb, grp); 1034 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line); 1035 /* 1036 * We only get here in the ERRORS_RO case; relocking the group 1037 * may be dangerous, but nothing bad will happen since the 1038 * filesystem will have already been marked read/only and the 1039 * journal has been aborted. We return 1 as a hint to callers 1040 * who might what to use the return value from 1041 * ext4_grp_locked_error() to distinguish between the 1042 * ERRORS_CONT and ERRORS_RO case, and perhaps return more 1043 * aggressively from the ext4 function in question, with a 1044 * more appropriate error code. 1045 */ 1046 ext4_lock_group(sb, grp); 1047 return; 1048 } 1049 1050 void ext4_mark_group_bitmap_corrupted(struct super_block *sb, 1051 ext4_group_t group, 1052 unsigned int flags) 1053 { 1054 struct ext4_sb_info *sbi = EXT4_SB(sb); 1055 struct ext4_group_info *grp = ext4_get_group_info(sb, group); 1056 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL); 1057 int ret; 1058 1059 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) { 1060 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, 1061 &grp->bb_state); 1062 if (!ret) 1063 percpu_counter_sub(&sbi->s_freeclusters_counter, 1064 grp->bb_free); 1065 } 1066 1067 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) { 1068 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, 1069 &grp->bb_state); 1070 if (!ret && gdp) { 1071 int count; 1072 1073 count = ext4_free_inodes_count(sb, gdp); 1074 percpu_counter_sub(&sbi->s_freeinodes_counter, 1075 count); 1076 } 1077 } 1078 } 1079 1080 void ext4_update_dynamic_rev(struct super_block *sb) 1081 { 1082 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 1083 1084 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 1085 return; 1086 1087 ext4_warning(sb, 1088 "updating to rev %d because of new feature flag, " 1089 "running e2fsck is recommended", 1090 EXT4_DYNAMIC_REV); 1091 1092 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 1093 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 1094 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 1095 /* leave es->s_feature_*compat flags alone */ 1096 /* es->s_uuid will be set by e2fsck if empty */ 1097 1098 /* 1099 * The rest of the superblock fields should be zero, and if not it 1100 * means they are likely already in use, so leave them alone. We 1101 * can leave it up to e2fsck to clean up any inconsistencies there. 1102 */ 1103 } 1104 1105 /* 1106 * Open the external journal device 1107 */ 1108 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 1109 { 1110 struct block_device *bdev; 1111 1112 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); 1113 if (IS_ERR(bdev)) 1114 goto fail; 1115 return bdev; 1116 1117 fail: 1118 ext4_msg(sb, KERN_ERR, 1119 "failed to open journal device unknown-block(%u,%u) %ld", 1120 MAJOR(dev), MINOR(dev), PTR_ERR(bdev)); 1121 return NULL; 1122 } 1123 1124 /* 1125 * Release the journal device 1126 */ 1127 static void ext4_blkdev_put(struct block_device *bdev) 1128 { 1129 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 1130 } 1131 1132 static void ext4_blkdev_remove(struct ext4_sb_info *sbi) 1133 { 1134 struct block_device *bdev; 1135 bdev = sbi->s_journal_bdev; 1136 if (bdev) { 1137 ext4_blkdev_put(bdev); 1138 sbi->s_journal_bdev = NULL; 1139 } 1140 } 1141 1142 static inline struct inode *orphan_list_entry(struct list_head *l) 1143 { 1144 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 1145 } 1146 1147 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 1148 { 1149 struct list_head *l; 1150 1151 ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 1152 le32_to_cpu(sbi->s_es->s_last_orphan)); 1153 1154 printk(KERN_ERR "sb_info orphan list:\n"); 1155 list_for_each(l, &sbi->s_orphan) { 1156 struct inode *inode = orphan_list_entry(l); 1157 printk(KERN_ERR " " 1158 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 1159 inode->i_sb->s_id, inode->i_ino, inode, 1160 inode->i_mode, inode->i_nlink, 1161 NEXT_ORPHAN(inode)); 1162 } 1163 } 1164 1165 #ifdef CONFIG_QUOTA 1166 static int ext4_quota_off(struct super_block *sb, int type); 1167 1168 static inline void ext4_quota_off_umount(struct super_block *sb) 1169 { 1170 int type; 1171 1172 /* Use our quota_off function to clear inode flags etc. */ 1173 for (type = 0; type < EXT4_MAXQUOTAS; type++) 1174 ext4_quota_off(sb, type); 1175 } 1176 1177 /* 1178 * This is a helper function which is used in the mount/remount 1179 * codepaths (which holds s_umount) to fetch the quota file name. 1180 */ 1181 static inline char *get_qf_name(struct super_block *sb, 1182 struct ext4_sb_info *sbi, 1183 int type) 1184 { 1185 return rcu_dereference_protected(sbi->s_qf_names[type], 1186 lockdep_is_held(&sb->s_umount)); 1187 } 1188 #else 1189 static inline void ext4_quota_off_umount(struct super_block *sb) 1190 { 1191 } 1192 #endif 1193 1194 static void ext4_put_super(struct super_block *sb) 1195 { 1196 struct ext4_sb_info *sbi = EXT4_SB(sb); 1197 struct ext4_super_block *es = sbi->s_es; 1198 struct buffer_head **group_desc; 1199 struct flex_groups **flex_groups; 1200 int aborted = 0; 1201 int i, err; 1202 1203 ext4_unregister_li_request(sb); 1204 ext4_quota_off_umount(sb); 1205 1206 flush_work(&sbi->s_error_work); 1207 destroy_workqueue(sbi->rsv_conversion_wq); 1208 ext4_release_orphan_info(sb); 1209 1210 /* 1211 * Unregister sysfs before destroying jbd2 journal. 1212 * Since we could still access attr_journal_task attribute via sysfs 1213 * path which could have sbi->s_journal->j_task as NULL 1214 */ 1215 ext4_unregister_sysfs(sb); 1216 1217 if (sbi->s_journal) { 1218 aborted = is_journal_aborted(sbi->s_journal); 1219 err = jbd2_journal_destroy(sbi->s_journal); 1220 sbi->s_journal = NULL; 1221 if ((err < 0) && !aborted) { 1222 ext4_abort(sb, -err, "Couldn't clean up the journal"); 1223 } 1224 } 1225 1226 ext4_es_unregister_shrinker(sbi); 1227 del_timer_sync(&sbi->s_err_report); 1228 ext4_release_system_zone(sb); 1229 ext4_mb_release(sb); 1230 ext4_ext_release(sb); 1231 1232 if (!sb_rdonly(sb) && !aborted) { 1233 ext4_clear_feature_journal_needs_recovery(sb); 1234 ext4_clear_feature_orphan_present(sb); 1235 es->s_state = cpu_to_le16(sbi->s_mount_state); 1236 } 1237 if (!sb_rdonly(sb)) 1238 ext4_commit_super(sb); 1239 1240 rcu_read_lock(); 1241 group_desc = rcu_dereference(sbi->s_group_desc); 1242 for (i = 0; i < sbi->s_gdb_count; i++) 1243 brelse(group_desc[i]); 1244 kvfree(group_desc); 1245 flex_groups = rcu_dereference(sbi->s_flex_groups); 1246 if (flex_groups) { 1247 for (i = 0; i < sbi->s_flex_groups_allocated; i++) 1248 kvfree(flex_groups[i]); 1249 kvfree(flex_groups); 1250 } 1251 rcu_read_unlock(); 1252 percpu_counter_destroy(&sbi->s_freeclusters_counter); 1253 percpu_counter_destroy(&sbi->s_freeinodes_counter); 1254 percpu_counter_destroy(&sbi->s_dirs_counter); 1255 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 1256 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit); 1257 percpu_free_rwsem(&sbi->s_writepages_rwsem); 1258 #ifdef CONFIG_QUOTA 1259 for (i = 0; i < EXT4_MAXQUOTAS; i++) 1260 kfree(get_qf_name(sb, sbi, i)); 1261 #endif 1262 1263 /* Debugging code just in case the in-memory inode orphan list 1264 * isn't empty. The on-disk one can be non-empty if we've 1265 * detected an error and taken the fs readonly, but the 1266 * in-memory list had better be clean by this point. */ 1267 if (!list_empty(&sbi->s_orphan)) 1268 dump_orphan_list(sb, sbi); 1269 ASSERT(list_empty(&sbi->s_orphan)); 1270 1271 sync_blockdev(sb->s_bdev); 1272 invalidate_bdev(sb->s_bdev); 1273 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) { 1274 /* 1275 * Invalidate the journal device's buffers. We don't want them 1276 * floating about in memory - the physical journal device may 1277 * hotswapped, and it breaks the `ro-after' testing code. 1278 */ 1279 sync_blockdev(sbi->s_journal_bdev); 1280 invalidate_bdev(sbi->s_journal_bdev); 1281 ext4_blkdev_remove(sbi); 1282 } 1283 1284 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 1285 sbi->s_ea_inode_cache = NULL; 1286 1287 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 1288 sbi->s_ea_block_cache = NULL; 1289 1290 ext4_stop_mmpd(sbi); 1291 1292 brelse(sbi->s_sbh); 1293 sb->s_fs_info = NULL; 1294 /* 1295 * Now that we are completely done shutting down the 1296 * superblock, we need to actually destroy the kobject. 1297 */ 1298 kobject_put(&sbi->s_kobj); 1299 wait_for_completion(&sbi->s_kobj_unregister); 1300 if (sbi->s_chksum_driver) 1301 crypto_free_shash(sbi->s_chksum_driver); 1302 kfree(sbi->s_blockgroup_lock); 1303 fs_put_dax(sbi->s_daxdev); 1304 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy); 1305 #ifdef CONFIG_UNICODE 1306 utf8_unload(sb->s_encoding); 1307 #endif 1308 kfree(sbi); 1309 } 1310 1311 static struct kmem_cache *ext4_inode_cachep; 1312 1313 /* 1314 * Called inside transaction, so use GFP_NOFS 1315 */ 1316 static struct inode *ext4_alloc_inode(struct super_block *sb) 1317 { 1318 struct ext4_inode_info *ei; 1319 1320 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); 1321 if (!ei) 1322 return NULL; 1323 1324 inode_set_iversion(&ei->vfs_inode, 1); 1325 spin_lock_init(&ei->i_raw_lock); 1326 INIT_LIST_HEAD(&ei->i_prealloc_list); 1327 atomic_set(&ei->i_prealloc_active, 0); 1328 spin_lock_init(&ei->i_prealloc_lock); 1329 ext4_es_init_tree(&ei->i_es_tree); 1330 rwlock_init(&ei->i_es_lock); 1331 INIT_LIST_HEAD(&ei->i_es_list); 1332 ei->i_es_all_nr = 0; 1333 ei->i_es_shk_nr = 0; 1334 ei->i_es_shrink_lblk = 0; 1335 ei->i_reserved_data_blocks = 0; 1336 spin_lock_init(&(ei->i_block_reservation_lock)); 1337 ext4_init_pending_tree(&ei->i_pending_tree); 1338 #ifdef CONFIG_QUOTA 1339 ei->i_reserved_quota = 0; 1340 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); 1341 #endif 1342 ei->jinode = NULL; 1343 INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 1344 spin_lock_init(&ei->i_completed_io_lock); 1345 ei->i_sync_tid = 0; 1346 ei->i_datasync_tid = 0; 1347 atomic_set(&ei->i_unwritten, 0); 1348 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 1349 ext4_fc_init_inode(&ei->vfs_inode); 1350 mutex_init(&ei->i_fc_lock); 1351 return &ei->vfs_inode; 1352 } 1353 1354 static int ext4_drop_inode(struct inode *inode) 1355 { 1356 int drop = generic_drop_inode(inode); 1357 1358 if (!drop) 1359 drop = fscrypt_drop_inode(inode); 1360 1361 trace_ext4_drop_inode(inode, drop); 1362 return drop; 1363 } 1364 1365 static void ext4_free_in_core_inode(struct inode *inode) 1366 { 1367 fscrypt_free_inode(inode); 1368 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) { 1369 pr_warn("%s: inode %ld still in fc list", 1370 __func__, inode->i_ino); 1371 } 1372 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 1373 } 1374 1375 static void ext4_destroy_inode(struct inode *inode) 1376 { 1377 if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 1378 ext4_msg(inode->i_sb, KERN_ERR, 1379 "Inode %lu (%p): orphan list check failed!", 1380 inode->i_ino, EXT4_I(inode)); 1381 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 1382 EXT4_I(inode), sizeof(struct ext4_inode_info), 1383 true); 1384 dump_stack(); 1385 } 1386 1387 if (EXT4_I(inode)->i_reserved_data_blocks) 1388 ext4_msg(inode->i_sb, KERN_ERR, 1389 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!", 1390 inode->i_ino, EXT4_I(inode), 1391 EXT4_I(inode)->i_reserved_data_blocks); 1392 } 1393 1394 static void init_once(void *foo) 1395 { 1396 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 1397 1398 INIT_LIST_HEAD(&ei->i_orphan); 1399 init_rwsem(&ei->xattr_sem); 1400 init_rwsem(&ei->i_data_sem); 1401 inode_init_once(&ei->vfs_inode); 1402 ext4_fc_init_inode(&ei->vfs_inode); 1403 } 1404 1405 static int __init init_inodecache(void) 1406 { 1407 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache", 1408 sizeof(struct ext4_inode_info), 0, 1409 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD| 1410 SLAB_ACCOUNT), 1411 offsetof(struct ext4_inode_info, i_data), 1412 sizeof_field(struct ext4_inode_info, i_data), 1413 init_once); 1414 if (ext4_inode_cachep == NULL) 1415 return -ENOMEM; 1416 return 0; 1417 } 1418 1419 static void destroy_inodecache(void) 1420 { 1421 /* 1422 * Make sure all delayed rcu free inodes are flushed before we 1423 * destroy cache. 1424 */ 1425 rcu_barrier(); 1426 kmem_cache_destroy(ext4_inode_cachep); 1427 } 1428 1429 void ext4_clear_inode(struct inode *inode) 1430 { 1431 ext4_fc_del(inode); 1432 invalidate_inode_buffers(inode); 1433 clear_inode(inode); 1434 ext4_discard_preallocations(inode, 0); 1435 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 1436 dquot_drop(inode); 1437 if (EXT4_I(inode)->jinode) { 1438 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 1439 EXT4_I(inode)->jinode); 1440 jbd2_free_inode(EXT4_I(inode)->jinode); 1441 EXT4_I(inode)->jinode = NULL; 1442 } 1443 fscrypt_put_encryption_info(inode); 1444 fsverity_cleanup_inode(inode); 1445 } 1446 1447 static struct inode *ext4_nfs_get_inode(struct super_block *sb, 1448 u64 ino, u32 generation) 1449 { 1450 struct inode *inode; 1451 1452 /* 1453 * Currently we don't know the generation for parent directory, so 1454 * a generation of 0 means "accept any" 1455 */ 1456 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE); 1457 if (IS_ERR(inode)) 1458 return ERR_CAST(inode); 1459 if (generation && inode->i_generation != generation) { 1460 iput(inode); 1461 return ERR_PTR(-ESTALE); 1462 } 1463 1464 return inode; 1465 } 1466 1467 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 1468 int fh_len, int fh_type) 1469 { 1470 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1471 ext4_nfs_get_inode); 1472 } 1473 1474 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 1475 int fh_len, int fh_type) 1476 { 1477 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1478 ext4_nfs_get_inode); 1479 } 1480 1481 static int ext4_nfs_commit_metadata(struct inode *inode) 1482 { 1483 struct writeback_control wbc = { 1484 .sync_mode = WB_SYNC_ALL 1485 }; 1486 1487 trace_ext4_nfs_commit_metadata(inode); 1488 return ext4_write_inode(inode, &wbc); 1489 } 1490 1491 #ifdef CONFIG_FS_ENCRYPTION 1492 static int ext4_get_context(struct inode *inode, void *ctx, size_t len) 1493 { 1494 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION, 1495 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len); 1496 } 1497 1498 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len, 1499 void *fs_data) 1500 { 1501 handle_t *handle = fs_data; 1502 int res, res2, credits, retries = 0; 1503 1504 /* 1505 * Encrypting the root directory is not allowed because e2fsck expects 1506 * lost+found to exist and be unencrypted, and encrypting the root 1507 * directory would imply encrypting the lost+found directory as well as 1508 * the filename "lost+found" itself. 1509 */ 1510 if (inode->i_ino == EXT4_ROOT_INO) 1511 return -EPERM; 1512 1513 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode))) 1514 return -EINVAL; 1515 1516 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX)) 1517 return -EOPNOTSUPP; 1518 1519 res = ext4_convert_inline_data(inode); 1520 if (res) 1521 return res; 1522 1523 /* 1524 * If a journal handle was specified, then the encryption context is 1525 * being set on a new inode via inheritance and is part of a larger 1526 * transaction to create the inode. Otherwise the encryption context is 1527 * being set on an existing inode in its own transaction. Only in the 1528 * latter case should the "retry on ENOSPC" logic be used. 1529 */ 1530 1531 if (handle) { 1532 res = ext4_xattr_set_handle(handle, inode, 1533 EXT4_XATTR_INDEX_ENCRYPTION, 1534 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 1535 ctx, len, 0); 1536 if (!res) { 1537 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1538 ext4_clear_inode_state(inode, 1539 EXT4_STATE_MAY_INLINE_DATA); 1540 /* 1541 * Update inode->i_flags - S_ENCRYPTED will be enabled, 1542 * S_DAX may be disabled 1543 */ 1544 ext4_set_inode_flags(inode, false); 1545 } 1546 return res; 1547 } 1548 1549 res = dquot_initialize(inode); 1550 if (res) 1551 return res; 1552 retry: 1553 res = ext4_xattr_set_credits(inode, len, false /* is_create */, 1554 &credits); 1555 if (res) 1556 return res; 1557 1558 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits); 1559 if (IS_ERR(handle)) 1560 return PTR_ERR(handle); 1561 1562 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION, 1563 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 1564 ctx, len, 0); 1565 if (!res) { 1566 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1567 /* 1568 * Update inode->i_flags - S_ENCRYPTED will be enabled, 1569 * S_DAX may be disabled 1570 */ 1571 ext4_set_inode_flags(inode, false); 1572 res = ext4_mark_inode_dirty(handle, inode); 1573 if (res) 1574 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty"); 1575 } 1576 res2 = ext4_journal_stop(handle); 1577 1578 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) 1579 goto retry; 1580 if (!res) 1581 res = res2; 1582 return res; 1583 } 1584 1585 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb) 1586 { 1587 return EXT4_SB(sb)->s_dummy_enc_policy.policy; 1588 } 1589 1590 static bool ext4_has_stable_inodes(struct super_block *sb) 1591 { 1592 return ext4_has_feature_stable_inodes(sb); 1593 } 1594 1595 static void ext4_get_ino_and_lblk_bits(struct super_block *sb, 1596 int *ino_bits_ret, int *lblk_bits_ret) 1597 { 1598 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count); 1599 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t); 1600 } 1601 1602 static const struct fscrypt_operations ext4_cryptops = { 1603 .key_prefix = "ext4:", 1604 .get_context = ext4_get_context, 1605 .set_context = ext4_set_context, 1606 .get_dummy_policy = ext4_get_dummy_policy, 1607 .empty_dir = ext4_empty_dir, 1608 .has_stable_inodes = ext4_has_stable_inodes, 1609 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits, 1610 }; 1611 #endif 1612 1613 #ifdef CONFIG_QUOTA 1614 static const char * const quotatypes[] = INITQFNAMES; 1615 #define QTYPE2NAME(t) (quotatypes[t]) 1616 1617 static int ext4_write_dquot(struct dquot *dquot); 1618 static int ext4_acquire_dquot(struct dquot *dquot); 1619 static int ext4_release_dquot(struct dquot *dquot); 1620 static int ext4_mark_dquot_dirty(struct dquot *dquot); 1621 static int ext4_write_info(struct super_block *sb, int type); 1622 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1623 const struct path *path); 1624 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1625 size_t len, loff_t off); 1626 static ssize_t ext4_quota_write(struct super_block *sb, int type, 1627 const char *data, size_t len, loff_t off); 1628 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 1629 unsigned int flags); 1630 1631 static struct dquot **ext4_get_dquots(struct inode *inode) 1632 { 1633 return EXT4_I(inode)->i_dquot; 1634 } 1635 1636 static const struct dquot_operations ext4_quota_operations = { 1637 .get_reserved_space = ext4_get_reserved_space, 1638 .write_dquot = ext4_write_dquot, 1639 .acquire_dquot = ext4_acquire_dquot, 1640 .release_dquot = ext4_release_dquot, 1641 .mark_dirty = ext4_mark_dquot_dirty, 1642 .write_info = ext4_write_info, 1643 .alloc_dquot = dquot_alloc, 1644 .destroy_dquot = dquot_destroy, 1645 .get_projid = ext4_get_projid, 1646 .get_inode_usage = ext4_get_inode_usage, 1647 .get_next_id = dquot_get_next_id, 1648 }; 1649 1650 static const struct quotactl_ops ext4_qctl_operations = { 1651 .quota_on = ext4_quota_on, 1652 .quota_off = ext4_quota_off, 1653 .quota_sync = dquot_quota_sync, 1654 .get_state = dquot_get_state, 1655 .set_info = dquot_set_dqinfo, 1656 .get_dqblk = dquot_get_dqblk, 1657 .set_dqblk = dquot_set_dqblk, 1658 .get_nextdqblk = dquot_get_next_dqblk, 1659 }; 1660 #endif 1661 1662 static const struct super_operations ext4_sops = { 1663 .alloc_inode = ext4_alloc_inode, 1664 .free_inode = ext4_free_in_core_inode, 1665 .destroy_inode = ext4_destroy_inode, 1666 .write_inode = ext4_write_inode, 1667 .dirty_inode = ext4_dirty_inode, 1668 .drop_inode = ext4_drop_inode, 1669 .evict_inode = ext4_evict_inode, 1670 .put_super = ext4_put_super, 1671 .sync_fs = ext4_sync_fs, 1672 .freeze_fs = ext4_freeze, 1673 .unfreeze_fs = ext4_unfreeze, 1674 .statfs = ext4_statfs, 1675 .show_options = ext4_show_options, 1676 #ifdef CONFIG_QUOTA 1677 .quota_read = ext4_quota_read, 1678 .quota_write = ext4_quota_write, 1679 .get_dquots = ext4_get_dquots, 1680 #endif 1681 }; 1682 1683 static const struct export_operations ext4_export_ops = { 1684 .fh_to_dentry = ext4_fh_to_dentry, 1685 .fh_to_parent = ext4_fh_to_parent, 1686 .get_parent = ext4_get_parent, 1687 .commit_metadata = ext4_nfs_commit_metadata, 1688 }; 1689 1690 enum { 1691 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1692 Opt_resgid, Opt_resuid, Opt_sb, 1693 Opt_nouid32, Opt_debug, Opt_removed, 1694 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 1695 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1696 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1697 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1698 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1699 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 1700 Opt_inlinecrypt, 1701 Opt_usrjquota, Opt_grpjquota, Opt_quota, 1702 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1703 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, 1704 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never, 1705 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error, 1706 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize, 1707 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 1708 Opt_inode_readahead_blks, Opt_journal_ioprio, 1709 Opt_dioread_nolock, Opt_dioread_lock, 1710 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1711 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, 1712 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan, 1713 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type, 1714 #ifdef CONFIG_EXT4_DEBUG 1715 Opt_fc_debug_max_replay, Opt_fc_debug_force 1716 #endif 1717 }; 1718 1719 static const struct constant_table ext4_param_errors[] = { 1720 {"continue", EXT4_MOUNT_ERRORS_CONT}, 1721 {"panic", EXT4_MOUNT_ERRORS_PANIC}, 1722 {"remount-ro", EXT4_MOUNT_ERRORS_RO}, 1723 {} 1724 }; 1725 1726 static const struct constant_table ext4_param_data[] = { 1727 {"journal", EXT4_MOUNT_JOURNAL_DATA}, 1728 {"ordered", EXT4_MOUNT_ORDERED_DATA}, 1729 {"writeback", EXT4_MOUNT_WRITEBACK_DATA}, 1730 {} 1731 }; 1732 1733 static const struct constant_table ext4_param_data_err[] = { 1734 {"abort", Opt_data_err_abort}, 1735 {"ignore", Opt_data_err_ignore}, 1736 {} 1737 }; 1738 1739 static const struct constant_table ext4_param_jqfmt[] = { 1740 {"vfsold", QFMT_VFS_OLD}, 1741 {"vfsv0", QFMT_VFS_V0}, 1742 {"vfsv1", QFMT_VFS_V1}, 1743 {} 1744 }; 1745 1746 static const struct constant_table ext4_param_dax[] = { 1747 {"always", Opt_dax_always}, 1748 {"inode", Opt_dax_inode}, 1749 {"never", Opt_dax_never}, 1750 {} 1751 }; 1752 1753 /* String parameter that allows empty argument */ 1754 #define fsparam_string_empty(NAME, OPT) \ 1755 __fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL) 1756 1757 /* 1758 * Mount option specification 1759 * We don't use fsparam_flag_no because of the way we set the 1760 * options and the way we show them in _ext4_show_options(). To 1761 * keep the changes to a minimum, let's keep the negative options 1762 * separate for now. 1763 */ 1764 static const struct fs_parameter_spec ext4_param_specs[] = { 1765 fsparam_flag ("bsddf", Opt_bsd_df), 1766 fsparam_flag ("minixdf", Opt_minix_df), 1767 fsparam_flag ("grpid", Opt_grpid), 1768 fsparam_flag ("bsdgroups", Opt_grpid), 1769 fsparam_flag ("nogrpid", Opt_nogrpid), 1770 fsparam_flag ("sysvgroups", Opt_nogrpid), 1771 fsparam_u32 ("resgid", Opt_resgid), 1772 fsparam_u32 ("resuid", Opt_resuid), 1773 fsparam_u32 ("sb", Opt_sb), 1774 fsparam_enum ("errors", Opt_errors, ext4_param_errors), 1775 fsparam_flag ("nouid32", Opt_nouid32), 1776 fsparam_flag ("debug", Opt_debug), 1777 fsparam_flag ("oldalloc", Opt_removed), 1778 fsparam_flag ("orlov", Opt_removed), 1779 fsparam_flag ("user_xattr", Opt_user_xattr), 1780 fsparam_flag ("nouser_xattr", Opt_nouser_xattr), 1781 fsparam_flag ("acl", Opt_acl), 1782 fsparam_flag ("noacl", Opt_noacl), 1783 fsparam_flag ("norecovery", Opt_noload), 1784 fsparam_flag ("noload", Opt_noload), 1785 fsparam_flag ("bh", Opt_removed), 1786 fsparam_flag ("nobh", Opt_removed), 1787 fsparam_u32 ("commit", Opt_commit), 1788 fsparam_u32 ("min_batch_time", Opt_min_batch_time), 1789 fsparam_u32 ("max_batch_time", Opt_max_batch_time), 1790 fsparam_u32 ("journal_dev", Opt_journal_dev), 1791 fsparam_bdev ("journal_path", Opt_journal_path), 1792 fsparam_flag ("journal_checksum", Opt_journal_checksum), 1793 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum), 1794 fsparam_flag ("journal_async_commit",Opt_journal_async_commit), 1795 fsparam_flag ("abort", Opt_abort), 1796 fsparam_enum ("data", Opt_data, ext4_param_data), 1797 fsparam_enum ("data_err", Opt_data_err, 1798 ext4_param_data_err), 1799 fsparam_string_empty 1800 ("usrjquota", Opt_usrjquota), 1801 fsparam_string_empty 1802 ("grpjquota", Opt_grpjquota), 1803 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt), 1804 fsparam_flag ("grpquota", Opt_grpquota), 1805 fsparam_flag ("quota", Opt_quota), 1806 fsparam_flag ("noquota", Opt_noquota), 1807 fsparam_flag ("usrquota", Opt_usrquota), 1808 fsparam_flag ("prjquota", Opt_prjquota), 1809 fsparam_flag ("barrier", Opt_barrier), 1810 fsparam_u32 ("barrier", Opt_barrier), 1811 fsparam_flag ("nobarrier", Opt_nobarrier), 1812 fsparam_flag ("i_version", Opt_i_version), 1813 fsparam_flag ("dax", Opt_dax), 1814 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax), 1815 fsparam_u32 ("stripe", Opt_stripe), 1816 fsparam_flag ("delalloc", Opt_delalloc), 1817 fsparam_flag ("nodelalloc", Opt_nodelalloc), 1818 fsparam_flag ("warn_on_error", Opt_warn_on_error), 1819 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error), 1820 fsparam_u32 ("debug_want_extra_isize", 1821 Opt_debug_want_extra_isize), 1822 fsparam_flag ("mblk_io_submit", Opt_removed), 1823 fsparam_flag ("nomblk_io_submit", Opt_removed), 1824 fsparam_flag ("block_validity", Opt_block_validity), 1825 fsparam_flag ("noblock_validity", Opt_noblock_validity), 1826 fsparam_u32 ("inode_readahead_blks", 1827 Opt_inode_readahead_blks), 1828 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio), 1829 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc), 1830 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc), 1831 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc), 1832 fsparam_flag ("dioread_nolock", Opt_dioread_nolock), 1833 fsparam_flag ("nodioread_nolock", Opt_dioread_lock), 1834 fsparam_flag ("dioread_lock", Opt_dioread_lock), 1835 fsparam_flag ("discard", Opt_discard), 1836 fsparam_flag ("nodiscard", Opt_nodiscard), 1837 fsparam_u32 ("init_itable", Opt_init_itable), 1838 fsparam_flag ("init_itable", Opt_init_itable), 1839 fsparam_flag ("noinit_itable", Opt_noinit_itable), 1840 #ifdef CONFIG_EXT4_DEBUG 1841 fsparam_flag ("fc_debug_force", Opt_fc_debug_force), 1842 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay), 1843 #endif 1844 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb), 1845 fsparam_flag ("test_dummy_encryption", 1846 Opt_test_dummy_encryption), 1847 fsparam_string ("test_dummy_encryption", 1848 Opt_test_dummy_encryption), 1849 fsparam_flag ("inlinecrypt", Opt_inlinecrypt), 1850 fsparam_flag ("nombcache", Opt_nombcache), 1851 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */ 1852 fsparam_flag ("prefetch_block_bitmaps", 1853 Opt_removed), 1854 fsparam_flag ("no_prefetch_block_bitmaps", 1855 Opt_no_prefetch_block_bitmaps), 1856 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan), 1857 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */ 1858 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */ 1859 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */ 1860 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */ 1861 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */ 1862 {} 1863 }; 1864 1865 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1866 #define DEFAULT_MB_OPTIMIZE_SCAN (-1) 1867 1868 static const char deprecated_msg[] = 1869 "Mount option \"%s\" will be removed by %s\n" 1870 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1871 1872 #define MOPT_SET 0x0001 1873 #define MOPT_CLEAR 0x0002 1874 #define MOPT_NOSUPPORT 0x0004 1875 #define MOPT_EXPLICIT 0x0008 1876 #ifdef CONFIG_QUOTA 1877 #define MOPT_Q 0 1878 #define MOPT_QFMT 0x0010 1879 #else 1880 #define MOPT_Q MOPT_NOSUPPORT 1881 #define MOPT_QFMT MOPT_NOSUPPORT 1882 #endif 1883 #define MOPT_NO_EXT2 0x0020 1884 #define MOPT_NO_EXT3 0x0040 1885 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1886 #define MOPT_SKIP 0x0080 1887 #define MOPT_2 0x0100 1888 1889 static const struct mount_opts { 1890 int token; 1891 int mount_opt; 1892 int flags; 1893 } ext4_mount_opts[] = { 1894 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 1895 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 1896 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 1897 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 1898 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 1899 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 1900 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 1901 MOPT_EXT4_ONLY | MOPT_SET}, 1902 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 1903 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1904 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 1905 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 1906 {Opt_delalloc, EXT4_MOUNT_DELALLOC, 1907 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1908 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 1909 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1910 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET}, 1911 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR}, 1912 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1913 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1914 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1915 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1916 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 1917 EXT4_MOUNT_JOURNAL_CHECKSUM), 1918 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1919 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 1920 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2}, 1921 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 1922 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 1923 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 1924 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 1925 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 1926 {Opt_dax_type, 0, MOPT_EXT4_ONLY}, 1927 {Opt_journal_dev, 0, MOPT_NO_EXT2}, 1928 {Opt_journal_path, 0, MOPT_NO_EXT2}, 1929 {Opt_journal_ioprio, 0, MOPT_NO_EXT2}, 1930 {Opt_data, 0, MOPT_NO_EXT2}, 1931 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 1932 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 1933 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1934 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 1935 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 1936 #else 1937 {Opt_acl, 0, MOPT_NOSUPPORT}, 1938 {Opt_noacl, 0, MOPT_NOSUPPORT}, 1939 #endif 1940 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 1941 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1942 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 1943 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 1944 MOPT_SET | MOPT_Q}, 1945 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 1946 MOPT_SET | MOPT_Q}, 1947 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, 1948 MOPT_SET | MOPT_Q}, 1949 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 1950 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), 1951 MOPT_CLEAR | MOPT_Q}, 1952 {Opt_usrjquota, 0, MOPT_Q}, 1953 {Opt_grpjquota, 0, MOPT_Q}, 1954 {Opt_jqfmt, 0, MOPT_QFMT}, 1955 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, 1956 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS, 1957 MOPT_SET}, 1958 #ifdef CONFIG_EXT4_DEBUG 1959 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT, 1960 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY}, 1961 #endif 1962 {Opt_err, 0, 0} 1963 }; 1964 1965 #ifdef CONFIG_UNICODE 1966 static const struct ext4_sb_encodings { 1967 __u16 magic; 1968 char *name; 1969 unsigned int version; 1970 } ext4_sb_encoding_map[] = { 1971 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)}, 1972 }; 1973 1974 static const struct ext4_sb_encodings * 1975 ext4_sb_read_encoding(const struct ext4_super_block *es) 1976 { 1977 __u16 magic = le16_to_cpu(es->s_encoding); 1978 int i; 1979 1980 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++) 1981 if (magic == ext4_sb_encoding_map[i].magic) 1982 return &ext4_sb_encoding_map[i]; 1983 1984 return NULL; 1985 } 1986 #endif 1987 1988 static int ext4_set_test_dummy_encryption(struct super_block *sb, char *arg) 1989 { 1990 #ifdef CONFIG_FS_ENCRYPTION 1991 struct ext4_sb_info *sbi = EXT4_SB(sb); 1992 int err; 1993 1994 err = fscrypt_set_test_dummy_encryption(sb, arg, 1995 &sbi->s_dummy_enc_policy); 1996 if (err) { 1997 ext4_msg(sb, KERN_WARNING, 1998 "Error while setting test dummy encryption [%d]", err); 1999 return err; 2000 } 2001 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled"); 2002 #endif 2003 return 0; 2004 } 2005 2006 #define EXT4_SPEC_JQUOTA (1 << 0) 2007 #define EXT4_SPEC_JQFMT (1 << 1) 2008 #define EXT4_SPEC_DATAJ (1 << 2) 2009 #define EXT4_SPEC_SB_BLOCK (1 << 3) 2010 #define EXT4_SPEC_JOURNAL_DEV (1 << 4) 2011 #define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5) 2012 #define EXT4_SPEC_DUMMY_ENCRYPTION (1 << 6) 2013 #define EXT4_SPEC_s_want_extra_isize (1 << 7) 2014 #define EXT4_SPEC_s_max_batch_time (1 << 8) 2015 #define EXT4_SPEC_s_min_batch_time (1 << 9) 2016 #define EXT4_SPEC_s_inode_readahead_blks (1 << 10) 2017 #define EXT4_SPEC_s_li_wait_mult (1 << 11) 2018 #define EXT4_SPEC_s_max_dir_size_kb (1 << 12) 2019 #define EXT4_SPEC_s_stripe (1 << 13) 2020 #define EXT4_SPEC_s_resuid (1 << 14) 2021 #define EXT4_SPEC_s_resgid (1 << 15) 2022 #define EXT4_SPEC_s_commit_interval (1 << 16) 2023 #define EXT4_SPEC_s_fc_debug_max_replay (1 << 17) 2024 #define EXT4_SPEC_s_sb_block (1 << 18) 2025 2026 struct ext4_fs_context { 2027 char *s_qf_names[EXT4_MAXQUOTAS]; 2028 char *test_dummy_enc_arg; 2029 int s_jquota_fmt; /* Format of quota to use */ 2030 int mb_optimize_scan; 2031 #ifdef CONFIG_EXT4_DEBUG 2032 int s_fc_debug_max_replay; 2033 #endif 2034 unsigned short qname_spec; 2035 unsigned long vals_s_flags; /* Bits to set in s_flags */ 2036 unsigned long mask_s_flags; /* Bits changed in s_flags */ 2037 unsigned long journal_devnum; 2038 unsigned long s_commit_interval; 2039 unsigned long s_stripe; 2040 unsigned int s_inode_readahead_blks; 2041 unsigned int s_want_extra_isize; 2042 unsigned int s_li_wait_mult; 2043 unsigned int s_max_dir_size_kb; 2044 unsigned int journal_ioprio; 2045 unsigned int vals_s_mount_opt; 2046 unsigned int mask_s_mount_opt; 2047 unsigned int vals_s_mount_opt2; 2048 unsigned int mask_s_mount_opt2; 2049 unsigned int vals_s_mount_flags; 2050 unsigned int mask_s_mount_flags; 2051 unsigned int opt_flags; /* MOPT flags */ 2052 unsigned int spec; 2053 u32 s_max_batch_time; 2054 u32 s_min_batch_time; 2055 kuid_t s_resuid; 2056 kgid_t s_resgid; 2057 ext4_fsblk_t s_sb_block; 2058 }; 2059 2060 static void ext4_fc_free(struct fs_context *fc) 2061 { 2062 struct ext4_fs_context *ctx = fc->fs_private; 2063 int i; 2064 2065 if (!ctx) 2066 return; 2067 2068 for (i = 0; i < EXT4_MAXQUOTAS; i++) 2069 kfree(ctx->s_qf_names[i]); 2070 2071 kfree(ctx->test_dummy_enc_arg); 2072 kfree(ctx); 2073 } 2074 2075 int ext4_init_fs_context(struct fs_context *fc) 2076 { 2077 struct ext4_fs_context *ctx; 2078 2079 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL); 2080 if (!ctx) 2081 return -ENOMEM; 2082 2083 fc->fs_private = ctx; 2084 fc->ops = &ext4_context_ops; 2085 2086 return 0; 2087 } 2088 2089 #ifdef CONFIG_QUOTA 2090 /* 2091 * Note the name of the specified quota file. 2092 */ 2093 static int note_qf_name(struct fs_context *fc, int qtype, 2094 struct fs_parameter *param) 2095 { 2096 struct ext4_fs_context *ctx = fc->fs_private; 2097 char *qname; 2098 2099 if (param->size < 1) { 2100 ext4_msg(NULL, KERN_ERR, "Missing quota name"); 2101 return -EINVAL; 2102 } 2103 if (strchr(param->string, '/')) { 2104 ext4_msg(NULL, KERN_ERR, 2105 "quotafile must be on filesystem root"); 2106 return -EINVAL; 2107 } 2108 if (ctx->s_qf_names[qtype]) { 2109 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) { 2110 ext4_msg(NULL, KERN_ERR, 2111 "%s quota file already specified", 2112 QTYPE2NAME(qtype)); 2113 return -EINVAL; 2114 } 2115 return 0; 2116 } 2117 2118 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL); 2119 if (!qname) { 2120 ext4_msg(NULL, KERN_ERR, 2121 "Not enough memory for storing quotafile name"); 2122 return -ENOMEM; 2123 } 2124 ctx->s_qf_names[qtype] = qname; 2125 ctx->qname_spec |= 1 << qtype; 2126 ctx->spec |= EXT4_SPEC_JQUOTA; 2127 return 0; 2128 } 2129 2130 /* 2131 * Clear the name of the specified quota file. 2132 */ 2133 static int unnote_qf_name(struct fs_context *fc, int qtype) 2134 { 2135 struct ext4_fs_context *ctx = fc->fs_private; 2136 2137 if (ctx->s_qf_names[qtype]) 2138 kfree(ctx->s_qf_names[qtype]); 2139 2140 ctx->s_qf_names[qtype] = NULL; 2141 ctx->qname_spec |= 1 << qtype; 2142 ctx->spec |= EXT4_SPEC_JQUOTA; 2143 return 0; 2144 } 2145 #endif 2146 2147 #define EXT4_SET_CTX(name) \ 2148 static inline void ctx_set_##name(struct ext4_fs_context *ctx, \ 2149 unsigned long flag) \ 2150 { \ 2151 ctx->mask_s_##name |= flag; \ 2152 ctx->vals_s_##name |= flag; \ 2153 } \ 2154 static inline void ctx_clear_##name(struct ext4_fs_context *ctx, \ 2155 unsigned long flag) \ 2156 { \ 2157 ctx->mask_s_##name |= flag; \ 2158 ctx->vals_s_##name &= ~flag; \ 2159 } \ 2160 static inline unsigned long \ 2161 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \ 2162 { \ 2163 return (ctx->vals_s_##name & flag); \ 2164 } \ 2165 2166 EXT4_SET_CTX(flags); 2167 EXT4_SET_CTX(mount_opt); 2168 EXT4_SET_CTX(mount_opt2); 2169 EXT4_SET_CTX(mount_flags); 2170 2171 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param) 2172 { 2173 struct ext4_fs_context *ctx = fc->fs_private; 2174 struct fs_parse_result result; 2175 const struct mount_opts *m; 2176 int is_remount; 2177 kuid_t uid; 2178 kgid_t gid; 2179 int token; 2180 2181 token = fs_parse(fc, ext4_param_specs, param, &result); 2182 if (token < 0) 2183 return token; 2184 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE; 2185 2186 for (m = ext4_mount_opts; m->token != Opt_err; m++) 2187 if (token == m->token) 2188 break; 2189 2190 ctx->opt_flags |= m->flags; 2191 2192 if (m->flags & MOPT_EXPLICIT) { 2193 if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 2194 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC); 2195 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 2196 ctx_set_mount_opt2(ctx, 2197 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM); 2198 } else 2199 return -EINVAL; 2200 } 2201 2202 if (m->flags & MOPT_NOSUPPORT) { 2203 ext4_msg(NULL, KERN_ERR, "%s option not supported", 2204 param->key); 2205 return 0; 2206 } 2207 2208 switch (token) { 2209 #ifdef CONFIG_QUOTA 2210 case Opt_usrjquota: 2211 if (!*param->string) 2212 return unnote_qf_name(fc, USRQUOTA); 2213 else 2214 return note_qf_name(fc, USRQUOTA, param); 2215 case Opt_grpjquota: 2216 if (!*param->string) 2217 return unnote_qf_name(fc, GRPQUOTA); 2218 else 2219 return note_qf_name(fc, GRPQUOTA, param); 2220 #endif 2221 case Opt_noacl: 2222 case Opt_nouser_xattr: 2223 ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "3.5"); 2224 break; 2225 case Opt_sb: 2226 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) { 2227 ext4_msg(NULL, KERN_WARNING, 2228 "Ignoring %s option on remount", param->key); 2229 } else { 2230 ctx->s_sb_block = result.uint_32; 2231 ctx->spec |= EXT4_SPEC_s_sb_block; 2232 } 2233 return 0; 2234 case Opt_removed: 2235 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option", 2236 param->key); 2237 return 0; 2238 case Opt_abort: 2239 ctx_set_mount_flags(ctx, EXT4_MF_FS_ABORTED); 2240 return 0; 2241 case Opt_i_version: 2242 ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "5.20"); 2243 ext4_msg(NULL, KERN_WARNING, "Use iversion instead\n"); 2244 ctx_set_flags(ctx, SB_I_VERSION); 2245 return 0; 2246 case Opt_inlinecrypt: 2247 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 2248 ctx_set_flags(ctx, SB_INLINECRYPT); 2249 #else 2250 ext4_msg(NULL, KERN_ERR, "inline encryption not supported"); 2251 #endif 2252 return 0; 2253 case Opt_errors: 2254 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK); 2255 ctx_set_mount_opt(ctx, result.uint_32); 2256 return 0; 2257 #ifdef CONFIG_QUOTA 2258 case Opt_jqfmt: 2259 ctx->s_jquota_fmt = result.uint_32; 2260 ctx->spec |= EXT4_SPEC_JQFMT; 2261 return 0; 2262 #endif 2263 case Opt_data: 2264 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS); 2265 ctx_set_mount_opt(ctx, result.uint_32); 2266 ctx->spec |= EXT4_SPEC_DATAJ; 2267 return 0; 2268 case Opt_commit: 2269 if (result.uint_32 == 0) 2270 ctx->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE; 2271 else if (result.uint_32 > INT_MAX / HZ) { 2272 ext4_msg(NULL, KERN_ERR, 2273 "Invalid commit interval %d, " 2274 "must be smaller than %d", 2275 result.uint_32, INT_MAX / HZ); 2276 return -EINVAL; 2277 } 2278 ctx->s_commit_interval = HZ * result.uint_32; 2279 ctx->spec |= EXT4_SPEC_s_commit_interval; 2280 return 0; 2281 case Opt_debug_want_extra_isize: 2282 if ((result.uint_32 & 1) || (result.uint_32 < 4)) { 2283 ext4_msg(NULL, KERN_ERR, 2284 "Invalid want_extra_isize %d", result.uint_32); 2285 return -EINVAL; 2286 } 2287 ctx->s_want_extra_isize = result.uint_32; 2288 ctx->spec |= EXT4_SPEC_s_want_extra_isize; 2289 return 0; 2290 case Opt_max_batch_time: 2291 ctx->s_max_batch_time = result.uint_32; 2292 ctx->spec |= EXT4_SPEC_s_max_batch_time; 2293 return 0; 2294 case Opt_min_batch_time: 2295 ctx->s_min_batch_time = result.uint_32; 2296 ctx->spec |= EXT4_SPEC_s_min_batch_time; 2297 return 0; 2298 case Opt_inode_readahead_blks: 2299 if (result.uint_32 && 2300 (result.uint_32 > (1 << 30) || 2301 !is_power_of_2(result.uint_32))) { 2302 ext4_msg(NULL, KERN_ERR, 2303 "EXT4-fs: inode_readahead_blks must be " 2304 "0 or a power of 2 smaller than 2^31"); 2305 return -EINVAL; 2306 } 2307 ctx->s_inode_readahead_blks = result.uint_32; 2308 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks; 2309 return 0; 2310 case Opt_init_itable: 2311 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE); 2312 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 2313 if (param->type == fs_value_is_string) 2314 ctx->s_li_wait_mult = result.uint_32; 2315 ctx->spec |= EXT4_SPEC_s_li_wait_mult; 2316 return 0; 2317 case Opt_max_dir_size_kb: 2318 ctx->s_max_dir_size_kb = result.uint_32; 2319 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb; 2320 return 0; 2321 #ifdef CONFIG_EXT4_DEBUG 2322 case Opt_fc_debug_max_replay: 2323 ctx->s_fc_debug_max_replay = result.uint_32; 2324 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay; 2325 return 0; 2326 #endif 2327 case Opt_stripe: 2328 ctx->s_stripe = result.uint_32; 2329 ctx->spec |= EXT4_SPEC_s_stripe; 2330 return 0; 2331 case Opt_resuid: 2332 uid = make_kuid(current_user_ns(), result.uint_32); 2333 if (!uid_valid(uid)) { 2334 ext4_msg(NULL, KERN_ERR, "Invalid uid value %d", 2335 result.uint_32); 2336 return -EINVAL; 2337 } 2338 ctx->s_resuid = uid; 2339 ctx->spec |= EXT4_SPEC_s_resuid; 2340 return 0; 2341 case Opt_resgid: 2342 gid = make_kgid(current_user_ns(), result.uint_32); 2343 if (!gid_valid(gid)) { 2344 ext4_msg(NULL, KERN_ERR, "Invalid gid value %d", 2345 result.uint_32); 2346 return -EINVAL; 2347 } 2348 ctx->s_resgid = gid; 2349 ctx->spec |= EXT4_SPEC_s_resgid; 2350 return 0; 2351 case Opt_journal_dev: 2352 if (is_remount) { 2353 ext4_msg(NULL, KERN_ERR, 2354 "Cannot specify journal on remount"); 2355 return -EINVAL; 2356 } 2357 ctx->journal_devnum = result.uint_32; 2358 ctx->spec |= EXT4_SPEC_JOURNAL_DEV; 2359 return 0; 2360 case Opt_journal_path: 2361 { 2362 struct inode *journal_inode; 2363 struct path path; 2364 int error; 2365 2366 if (is_remount) { 2367 ext4_msg(NULL, KERN_ERR, 2368 "Cannot specify journal on remount"); 2369 return -EINVAL; 2370 } 2371 2372 error = fs_lookup_param(fc, param, 1, &path); 2373 if (error) { 2374 ext4_msg(NULL, KERN_ERR, "error: could not find " 2375 "journal device path"); 2376 return -EINVAL; 2377 } 2378 2379 journal_inode = d_inode(path.dentry); 2380 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev); 2381 ctx->spec |= EXT4_SPEC_JOURNAL_DEV; 2382 path_put(&path); 2383 return 0; 2384 } 2385 case Opt_journal_ioprio: 2386 if (result.uint_32 > 7) { 2387 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority" 2388 " (must be 0-7)"); 2389 return -EINVAL; 2390 } 2391 ctx->journal_ioprio = 2392 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32); 2393 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO; 2394 return 0; 2395 case Opt_test_dummy_encryption: 2396 #ifdef CONFIG_FS_ENCRYPTION 2397 if (param->type == fs_value_is_flag) { 2398 ctx->spec |= EXT4_SPEC_DUMMY_ENCRYPTION; 2399 ctx->test_dummy_enc_arg = NULL; 2400 return 0; 2401 } 2402 if (*param->string && 2403 !(!strcmp(param->string, "v1") || 2404 !strcmp(param->string, "v2"))) { 2405 ext4_msg(NULL, KERN_WARNING, 2406 "Value of option \"%s\" is unrecognized", 2407 param->key); 2408 return -EINVAL; 2409 } 2410 ctx->spec |= EXT4_SPEC_DUMMY_ENCRYPTION; 2411 ctx->test_dummy_enc_arg = kmemdup_nul(param->string, param->size, 2412 GFP_KERNEL); 2413 #else 2414 ext4_msg(NULL, KERN_WARNING, 2415 "Test dummy encryption mount option ignored"); 2416 #endif 2417 return 0; 2418 case Opt_dax: 2419 case Opt_dax_type: 2420 #ifdef CONFIG_FS_DAX 2421 { 2422 int type = (token == Opt_dax) ? 2423 Opt_dax : result.uint_32; 2424 2425 switch (type) { 2426 case Opt_dax: 2427 case Opt_dax_always: 2428 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); 2429 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); 2430 break; 2431 case Opt_dax_never: 2432 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); 2433 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); 2434 break; 2435 case Opt_dax_inode: 2436 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); 2437 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); 2438 /* Strictly for printing options */ 2439 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE); 2440 break; 2441 } 2442 return 0; 2443 } 2444 #else 2445 ext4_msg(NULL, KERN_INFO, "dax option not supported"); 2446 return -EINVAL; 2447 #endif 2448 case Opt_data_err: 2449 if (result.uint_32 == Opt_data_err_abort) 2450 ctx_set_mount_opt(ctx, m->mount_opt); 2451 else if (result.uint_32 == Opt_data_err_ignore) 2452 ctx_clear_mount_opt(ctx, m->mount_opt); 2453 return 0; 2454 case Opt_mb_optimize_scan: 2455 if (result.int_32 != 0 && result.int_32 != 1) { 2456 ext4_msg(NULL, KERN_WARNING, 2457 "mb_optimize_scan should be set to 0 or 1."); 2458 return -EINVAL; 2459 } 2460 ctx->mb_optimize_scan = result.int_32; 2461 return 0; 2462 } 2463 2464 /* 2465 * At this point we should only be getting options requiring MOPT_SET, 2466 * or MOPT_CLEAR. Anything else is a bug 2467 */ 2468 if (m->token == Opt_err) { 2469 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s", 2470 param->key); 2471 WARN_ON(1); 2472 return -EINVAL; 2473 } 2474 2475 else { 2476 unsigned int set = 0; 2477 2478 if ((param->type == fs_value_is_flag) || 2479 result.uint_32 > 0) 2480 set = 1; 2481 2482 if (m->flags & MOPT_CLEAR) 2483 set = !set; 2484 else if (unlikely(!(m->flags & MOPT_SET))) { 2485 ext4_msg(NULL, KERN_WARNING, 2486 "buggy handling of option %s", 2487 param->key); 2488 WARN_ON(1); 2489 return -EINVAL; 2490 } 2491 if (m->flags & MOPT_2) { 2492 if (set != 0) 2493 ctx_set_mount_opt2(ctx, m->mount_opt); 2494 else 2495 ctx_clear_mount_opt2(ctx, m->mount_opt); 2496 } else { 2497 if (set != 0) 2498 ctx_set_mount_opt(ctx, m->mount_opt); 2499 else 2500 ctx_clear_mount_opt(ctx, m->mount_opt); 2501 } 2502 } 2503 2504 return 0; 2505 } 2506 2507 static int parse_options(struct fs_context *fc, char *options) 2508 { 2509 struct fs_parameter param; 2510 int ret; 2511 char *key; 2512 2513 if (!options) 2514 return 0; 2515 2516 while ((key = strsep(&options, ",")) != NULL) { 2517 if (*key) { 2518 size_t v_len = 0; 2519 char *value = strchr(key, '='); 2520 2521 param.type = fs_value_is_flag; 2522 param.string = NULL; 2523 2524 if (value) { 2525 if (value == key) 2526 continue; 2527 2528 *value++ = 0; 2529 v_len = strlen(value); 2530 param.string = kmemdup_nul(value, v_len, 2531 GFP_KERNEL); 2532 if (!param.string) 2533 return -ENOMEM; 2534 param.type = fs_value_is_string; 2535 } 2536 2537 param.key = key; 2538 param.size = v_len; 2539 2540 ret = ext4_parse_param(fc, ¶m); 2541 if (param.string) 2542 kfree(param.string); 2543 if (ret < 0) 2544 return ret; 2545 } 2546 } 2547 2548 ret = ext4_validate_options(fc); 2549 if (ret < 0) 2550 return ret; 2551 2552 return 0; 2553 } 2554 2555 static int parse_apply_sb_mount_options(struct super_block *sb, 2556 struct ext4_fs_context *m_ctx) 2557 { 2558 struct ext4_sb_info *sbi = EXT4_SB(sb); 2559 char *s_mount_opts = NULL; 2560 struct ext4_fs_context *s_ctx = NULL; 2561 struct fs_context *fc = NULL; 2562 int ret = -ENOMEM; 2563 2564 if (!sbi->s_es->s_mount_opts[0]) 2565 return 0; 2566 2567 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts, 2568 sizeof(sbi->s_es->s_mount_opts), 2569 GFP_KERNEL); 2570 if (!s_mount_opts) 2571 return ret; 2572 2573 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL); 2574 if (!fc) 2575 goto out_free; 2576 2577 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL); 2578 if (!s_ctx) 2579 goto out_free; 2580 2581 fc->fs_private = s_ctx; 2582 fc->s_fs_info = sbi; 2583 2584 ret = parse_options(fc, s_mount_opts); 2585 if (ret < 0) 2586 goto parse_failed; 2587 2588 ret = ext4_check_opt_consistency(fc, sb); 2589 if (ret < 0) { 2590 parse_failed: 2591 ext4_msg(sb, KERN_WARNING, 2592 "failed to parse options in superblock: %s", 2593 s_mount_opts); 2594 ret = 0; 2595 goto out_free; 2596 } 2597 2598 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV) 2599 m_ctx->journal_devnum = s_ctx->journal_devnum; 2600 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO) 2601 m_ctx->journal_ioprio = s_ctx->journal_ioprio; 2602 2603 ret = ext4_apply_options(fc, sb); 2604 2605 out_free: 2606 kfree(s_ctx); 2607 kfree(fc); 2608 kfree(s_mount_opts); 2609 return ret; 2610 } 2611 2612 static void ext4_apply_quota_options(struct fs_context *fc, 2613 struct super_block *sb) 2614 { 2615 #ifdef CONFIG_QUOTA 2616 bool quota_feature = ext4_has_feature_quota(sb); 2617 struct ext4_fs_context *ctx = fc->fs_private; 2618 struct ext4_sb_info *sbi = EXT4_SB(sb); 2619 char *qname; 2620 int i; 2621 2622 if (quota_feature) 2623 return; 2624 2625 if (ctx->spec & EXT4_SPEC_JQUOTA) { 2626 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2627 if (!(ctx->qname_spec & (1 << i))) 2628 continue; 2629 2630 qname = ctx->s_qf_names[i]; /* May be NULL */ 2631 if (qname) 2632 set_opt(sb, QUOTA); 2633 ctx->s_qf_names[i] = NULL; 2634 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname, 2635 lockdep_is_held(&sb->s_umount)); 2636 if (qname) 2637 kfree_rcu(qname); 2638 } 2639 } 2640 2641 if (ctx->spec & EXT4_SPEC_JQFMT) 2642 sbi->s_jquota_fmt = ctx->s_jquota_fmt; 2643 #endif 2644 } 2645 2646 /* 2647 * Check quota settings consistency. 2648 */ 2649 static int ext4_check_quota_consistency(struct fs_context *fc, 2650 struct super_block *sb) 2651 { 2652 #ifdef CONFIG_QUOTA 2653 struct ext4_fs_context *ctx = fc->fs_private; 2654 struct ext4_sb_info *sbi = EXT4_SB(sb); 2655 bool quota_feature = ext4_has_feature_quota(sb); 2656 bool quota_loaded = sb_any_quota_loaded(sb); 2657 bool usr_qf_name, grp_qf_name, usrquota, grpquota; 2658 int quota_flags, i; 2659 2660 /* 2661 * We do the test below only for project quotas. 'usrquota' and 2662 * 'grpquota' mount options are allowed even without quota feature 2663 * to support legacy quotas in quota files. 2664 */ 2665 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) && 2666 !ext4_has_feature_project(sb)) { 2667 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. " 2668 "Cannot enable project quota enforcement."); 2669 return -EINVAL; 2670 } 2671 2672 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 2673 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA; 2674 if (quota_loaded && 2675 ctx->mask_s_mount_opt & quota_flags && 2676 !ctx_test_mount_opt(ctx, quota_flags)) 2677 goto err_quota_change; 2678 2679 if (ctx->spec & EXT4_SPEC_JQUOTA) { 2680 2681 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2682 if (!(ctx->qname_spec & (1 << i))) 2683 continue; 2684 2685 if (quota_loaded && 2686 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i]) 2687 goto err_jquota_change; 2688 2689 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] && 2690 strcmp(get_qf_name(sb, sbi, i), 2691 ctx->s_qf_names[i]) != 0) 2692 goto err_jquota_specified; 2693 } 2694 2695 if (quota_feature) { 2696 ext4_msg(NULL, KERN_INFO, 2697 "Journaled quota options ignored when " 2698 "QUOTA feature is enabled"); 2699 return 0; 2700 } 2701 } 2702 2703 if (ctx->spec & EXT4_SPEC_JQFMT) { 2704 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded) 2705 goto err_jquota_change; 2706 if (quota_feature) { 2707 ext4_msg(NULL, KERN_INFO, "Quota format mount options " 2708 "ignored when QUOTA feature is enabled"); 2709 return 0; 2710 } 2711 } 2712 2713 /* Make sure we don't mix old and new quota format */ 2714 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) || 2715 ctx->s_qf_names[USRQUOTA]); 2716 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) || 2717 ctx->s_qf_names[GRPQUOTA]); 2718 2719 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) || 2720 test_opt(sb, USRQUOTA)); 2721 2722 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) || 2723 test_opt(sb, GRPQUOTA)); 2724 2725 if (usr_qf_name) { 2726 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA); 2727 usrquota = false; 2728 } 2729 if (grp_qf_name) { 2730 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA); 2731 grpquota = false; 2732 } 2733 2734 if (usr_qf_name || grp_qf_name) { 2735 if (usrquota || grpquota) { 2736 ext4_msg(NULL, KERN_ERR, "old and new quota " 2737 "format mixing"); 2738 return -EINVAL; 2739 } 2740 2741 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) { 2742 ext4_msg(NULL, KERN_ERR, "journaled quota format " 2743 "not specified"); 2744 return -EINVAL; 2745 } 2746 } 2747 2748 return 0; 2749 2750 err_quota_change: 2751 ext4_msg(NULL, KERN_ERR, 2752 "Cannot change quota options when quota turned on"); 2753 return -EINVAL; 2754 err_jquota_change: 2755 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota " 2756 "options when quota turned on"); 2757 return -EINVAL; 2758 err_jquota_specified: 2759 ext4_msg(NULL, KERN_ERR, "%s quota file already specified", 2760 QTYPE2NAME(i)); 2761 return -EINVAL; 2762 #else 2763 return 0; 2764 #endif 2765 } 2766 2767 static int ext4_check_opt_consistency(struct fs_context *fc, 2768 struct super_block *sb) 2769 { 2770 struct ext4_fs_context *ctx = fc->fs_private; 2771 struct ext4_sb_info *sbi = fc->s_fs_info; 2772 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE; 2773 2774 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 2775 ext4_msg(NULL, KERN_ERR, 2776 "Mount option(s) incompatible with ext2"); 2777 return -EINVAL; 2778 } 2779 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 2780 ext4_msg(NULL, KERN_ERR, 2781 "Mount option(s) incompatible with ext3"); 2782 return -EINVAL; 2783 } 2784 2785 if (ctx->s_want_extra_isize > 2786 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) { 2787 ext4_msg(NULL, KERN_ERR, 2788 "Invalid want_extra_isize %d", 2789 ctx->s_want_extra_isize); 2790 return -EINVAL; 2791 } 2792 2793 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DIOREAD_NOLOCK)) { 2794 int blocksize = 2795 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 2796 if (blocksize < PAGE_SIZE) 2797 ext4_msg(NULL, KERN_WARNING, "Warning: mounting with an " 2798 "experimental mount option 'dioread_nolock' " 2799 "for blocksize < PAGE_SIZE"); 2800 } 2801 2802 #ifdef CONFIG_FS_ENCRYPTION 2803 /* 2804 * This mount option is just for testing, and it's not worthwhile to 2805 * implement the extra complexity (e.g. RCU protection) that would be 2806 * needed to allow it to be set or changed during remount. We do allow 2807 * it to be specified during remount, but only if there is no change. 2808 */ 2809 if ((ctx->spec & EXT4_SPEC_DUMMY_ENCRYPTION) && 2810 is_remount && !sbi->s_dummy_enc_policy.policy) { 2811 ext4_msg(NULL, KERN_WARNING, 2812 "Can't set test_dummy_encryption on remount"); 2813 return -1; 2814 } 2815 #endif 2816 2817 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) { 2818 if (!sbi->s_journal) { 2819 ext4_msg(NULL, KERN_WARNING, 2820 "Remounting file system with no journal " 2821 "so ignoring journalled data option"); 2822 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS); 2823 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) != 2824 test_opt(sb, DATA_FLAGS)) { 2825 ext4_msg(NULL, KERN_ERR, "Cannot change data mode " 2826 "on remount"); 2827 return -EINVAL; 2828 } 2829 } 2830 2831 if (is_remount) { 2832 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) && 2833 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) { 2834 ext4_msg(NULL, KERN_ERR, "can't mount with " 2835 "both data=journal and dax"); 2836 return -EINVAL; 2837 } 2838 2839 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) && 2840 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2841 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) { 2842 fail_dax_change_remount: 2843 ext4_msg(NULL, KERN_ERR, "can't change " 2844 "dax mount option while remounting"); 2845 return -EINVAL; 2846 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) && 2847 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2848 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) { 2849 goto fail_dax_change_remount; 2850 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) && 2851 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2852 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2853 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) { 2854 goto fail_dax_change_remount; 2855 } 2856 } 2857 2858 return ext4_check_quota_consistency(fc, sb); 2859 } 2860 2861 static int ext4_apply_options(struct fs_context *fc, struct super_block *sb) 2862 { 2863 struct ext4_fs_context *ctx = fc->fs_private; 2864 struct ext4_sb_info *sbi = fc->s_fs_info; 2865 int ret = 0; 2866 2867 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt; 2868 sbi->s_mount_opt |= ctx->vals_s_mount_opt; 2869 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2; 2870 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2; 2871 sbi->s_mount_flags &= ~ctx->mask_s_mount_flags; 2872 sbi->s_mount_flags |= ctx->vals_s_mount_flags; 2873 sb->s_flags &= ~ctx->mask_s_flags; 2874 sb->s_flags |= ctx->vals_s_flags; 2875 2876 /* 2877 * i_version differs from common mount option iversion so we have 2878 * to let vfs know that it was set, otherwise it would get cleared 2879 * on remount 2880 */ 2881 if (ctx->mask_s_flags & SB_I_VERSION) 2882 fc->sb_flags |= SB_I_VERSION; 2883 2884 #define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; }) 2885 APPLY(s_commit_interval); 2886 APPLY(s_stripe); 2887 APPLY(s_max_batch_time); 2888 APPLY(s_min_batch_time); 2889 APPLY(s_want_extra_isize); 2890 APPLY(s_inode_readahead_blks); 2891 APPLY(s_max_dir_size_kb); 2892 APPLY(s_li_wait_mult); 2893 APPLY(s_resgid); 2894 APPLY(s_resuid); 2895 2896 #ifdef CONFIG_EXT4_DEBUG 2897 APPLY(s_fc_debug_max_replay); 2898 #endif 2899 2900 ext4_apply_quota_options(fc, sb); 2901 2902 if (ctx->spec & EXT4_SPEC_DUMMY_ENCRYPTION) 2903 ret = ext4_set_test_dummy_encryption(sb, ctx->test_dummy_enc_arg); 2904 2905 return ret; 2906 } 2907 2908 2909 static int ext4_validate_options(struct fs_context *fc) 2910 { 2911 #ifdef CONFIG_QUOTA 2912 struct ext4_fs_context *ctx = fc->fs_private; 2913 char *usr_qf_name, *grp_qf_name; 2914 2915 usr_qf_name = ctx->s_qf_names[USRQUOTA]; 2916 grp_qf_name = ctx->s_qf_names[GRPQUOTA]; 2917 2918 if (usr_qf_name || grp_qf_name) { 2919 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name) 2920 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA); 2921 2922 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name) 2923 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA); 2924 2925 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) || 2926 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) { 2927 ext4_msg(NULL, KERN_ERR, "old and new quota " 2928 "format mixing"); 2929 return -EINVAL; 2930 } 2931 } 2932 #endif 2933 return 1; 2934 } 2935 2936 static inline void ext4_show_quota_options(struct seq_file *seq, 2937 struct super_block *sb) 2938 { 2939 #if defined(CONFIG_QUOTA) 2940 struct ext4_sb_info *sbi = EXT4_SB(sb); 2941 char *usr_qf_name, *grp_qf_name; 2942 2943 if (sbi->s_jquota_fmt) { 2944 char *fmtname = ""; 2945 2946 switch (sbi->s_jquota_fmt) { 2947 case QFMT_VFS_OLD: 2948 fmtname = "vfsold"; 2949 break; 2950 case QFMT_VFS_V0: 2951 fmtname = "vfsv0"; 2952 break; 2953 case QFMT_VFS_V1: 2954 fmtname = "vfsv1"; 2955 break; 2956 } 2957 seq_printf(seq, ",jqfmt=%s", fmtname); 2958 } 2959 2960 rcu_read_lock(); 2961 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); 2962 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); 2963 if (usr_qf_name) 2964 seq_show_option(seq, "usrjquota", usr_qf_name); 2965 if (grp_qf_name) 2966 seq_show_option(seq, "grpjquota", grp_qf_name); 2967 rcu_read_unlock(); 2968 #endif 2969 } 2970 2971 static const char *token2str(int token) 2972 { 2973 const struct fs_parameter_spec *spec; 2974 2975 for (spec = ext4_param_specs; spec->name != NULL; spec++) 2976 if (spec->opt == token && !spec->type) 2977 break; 2978 return spec->name; 2979 } 2980 2981 /* 2982 * Show an option if 2983 * - it's set to a non-default value OR 2984 * - if the per-sb default is different from the global default 2985 */ 2986 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 2987 int nodefs) 2988 { 2989 struct ext4_sb_info *sbi = EXT4_SB(sb); 2990 struct ext4_super_block *es = sbi->s_es; 2991 int def_errors, def_mount_opt = sbi->s_def_mount_opt; 2992 const struct mount_opts *m; 2993 char sep = nodefs ? '\n' : ','; 2994 2995 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 2996 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 2997 2998 if (sbi->s_sb_block != 1) 2999 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 3000 3001 for (m = ext4_mount_opts; m->token != Opt_err; m++) { 3002 int want_set = m->flags & MOPT_SET; 3003 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 3004 m->flags & MOPT_SKIP) 3005 continue; 3006 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 3007 continue; /* skip if same as the default */ 3008 if ((want_set && 3009 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 3010 (!want_set && (sbi->s_mount_opt & m->mount_opt))) 3011 continue; /* select Opt_noFoo vs Opt_Foo */ 3012 SEQ_OPTS_PRINT("%s", token2str(m->token)); 3013 } 3014 3015 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 3016 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 3017 SEQ_OPTS_PRINT("resuid=%u", 3018 from_kuid_munged(&init_user_ns, sbi->s_resuid)); 3019 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 3020 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 3021 SEQ_OPTS_PRINT("resgid=%u", 3022 from_kgid_munged(&init_user_ns, sbi->s_resgid)); 3023 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 3024 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 3025 SEQ_OPTS_PUTS("errors=remount-ro"); 3026 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 3027 SEQ_OPTS_PUTS("errors=continue"); 3028 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 3029 SEQ_OPTS_PUTS("errors=panic"); 3030 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 3031 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 3032 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 3033 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 3034 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 3035 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 3036 if (sb->s_flags & SB_I_VERSION) 3037 SEQ_OPTS_PUTS("i_version"); 3038 if (nodefs || sbi->s_stripe) 3039 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 3040 if (nodefs || EXT4_MOUNT_DATA_FLAGS & 3041 (sbi->s_mount_opt ^ def_mount_opt)) { 3042 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 3043 SEQ_OPTS_PUTS("data=journal"); 3044 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 3045 SEQ_OPTS_PUTS("data=ordered"); 3046 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 3047 SEQ_OPTS_PUTS("data=writeback"); 3048 } 3049 if (nodefs || 3050 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 3051 SEQ_OPTS_PRINT("inode_readahead_blks=%u", 3052 sbi->s_inode_readahead_blks); 3053 3054 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs || 3055 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 3056 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 3057 if (nodefs || sbi->s_max_dir_size_kb) 3058 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 3059 if (test_opt(sb, DATA_ERR_ABORT)) 3060 SEQ_OPTS_PUTS("data_err=abort"); 3061 3062 fscrypt_show_test_dummy_encryption(seq, sep, sb); 3063 3064 if (sb->s_flags & SB_INLINECRYPT) 3065 SEQ_OPTS_PUTS("inlinecrypt"); 3066 3067 if (test_opt(sb, DAX_ALWAYS)) { 3068 if (IS_EXT2_SB(sb)) 3069 SEQ_OPTS_PUTS("dax"); 3070 else 3071 SEQ_OPTS_PUTS("dax=always"); 3072 } else if (test_opt2(sb, DAX_NEVER)) { 3073 SEQ_OPTS_PUTS("dax=never"); 3074 } else if (test_opt2(sb, DAX_INODE)) { 3075 SEQ_OPTS_PUTS("dax=inode"); 3076 } 3077 ext4_show_quota_options(seq, sb); 3078 return 0; 3079 } 3080 3081 static int ext4_show_options(struct seq_file *seq, struct dentry *root) 3082 { 3083 return _ext4_show_options(seq, root->d_sb, 0); 3084 } 3085 3086 int ext4_seq_options_show(struct seq_file *seq, void *offset) 3087 { 3088 struct super_block *sb = seq->private; 3089 int rc; 3090 3091 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); 3092 rc = _ext4_show_options(seq, sb, 1); 3093 seq_puts(seq, "\n"); 3094 return rc; 3095 } 3096 3097 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 3098 int read_only) 3099 { 3100 struct ext4_sb_info *sbi = EXT4_SB(sb); 3101 int err = 0; 3102 3103 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 3104 ext4_msg(sb, KERN_ERR, "revision level too high, " 3105 "forcing read-only mode"); 3106 err = -EROFS; 3107 goto done; 3108 } 3109 if (read_only) 3110 goto done; 3111 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 3112 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 3113 "running e2fsck is recommended"); 3114 else if (sbi->s_mount_state & EXT4_ERROR_FS) 3115 ext4_msg(sb, KERN_WARNING, 3116 "warning: mounting fs with errors, " 3117 "running e2fsck is recommended"); 3118 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 3119 le16_to_cpu(es->s_mnt_count) >= 3120 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 3121 ext4_msg(sb, KERN_WARNING, 3122 "warning: maximal mount count reached, " 3123 "running e2fsck is recommended"); 3124 else if (le32_to_cpu(es->s_checkinterval) && 3125 (ext4_get_tstamp(es, s_lastcheck) + 3126 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds())) 3127 ext4_msg(sb, KERN_WARNING, 3128 "warning: checktime reached, " 3129 "running e2fsck is recommended"); 3130 if (!sbi->s_journal) 3131 es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 3132 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 3133 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 3134 le16_add_cpu(&es->s_mnt_count, 1); 3135 ext4_update_tstamp(es, s_mtime); 3136 if (sbi->s_journal) { 3137 ext4_set_feature_journal_needs_recovery(sb); 3138 if (ext4_has_feature_orphan_file(sb)) 3139 ext4_set_feature_orphan_present(sb); 3140 } 3141 3142 err = ext4_commit_super(sb); 3143 done: 3144 if (test_opt(sb, DEBUG)) 3145 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 3146 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 3147 sb->s_blocksize, 3148 sbi->s_groups_count, 3149 EXT4_BLOCKS_PER_GROUP(sb), 3150 EXT4_INODES_PER_GROUP(sb), 3151 sbi->s_mount_opt, sbi->s_mount_opt2); 3152 3153 cleancache_init_fs(sb); 3154 return err; 3155 } 3156 3157 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 3158 { 3159 struct ext4_sb_info *sbi = EXT4_SB(sb); 3160 struct flex_groups **old_groups, **new_groups; 3161 int size, i, j; 3162 3163 if (!sbi->s_log_groups_per_flex) 3164 return 0; 3165 3166 size = ext4_flex_group(sbi, ngroup - 1) + 1; 3167 if (size <= sbi->s_flex_groups_allocated) 3168 return 0; 3169 3170 new_groups = kvzalloc(roundup_pow_of_two(size * 3171 sizeof(*sbi->s_flex_groups)), GFP_KERNEL); 3172 if (!new_groups) { 3173 ext4_msg(sb, KERN_ERR, 3174 "not enough memory for %d flex group pointers", size); 3175 return -ENOMEM; 3176 } 3177 for (i = sbi->s_flex_groups_allocated; i < size; i++) { 3178 new_groups[i] = kvzalloc(roundup_pow_of_two( 3179 sizeof(struct flex_groups)), 3180 GFP_KERNEL); 3181 if (!new_groups[i]) { 3182 for (j = sbi->s_flex_groups_allocated; j < i; j++) 3183 kvfree(new_groups[j]); 3184 kvfree(new_groups); 3185 ext4_msg(sb, KERN_ERR, 3186 "not enough memory for %d flex groups", size); 3187 return -ENOMEM; 3188 } 3189 } 3190 rcu_read_lock(); 3191 old_groups = rcu_dereference(sbi->s_flex_groups); 3192 if (old_groups) 3193 memcpy(new_groups, old_groups, 3194 (sbi->s_flex_groups_allocated * 3195 sizeof(struct flex_groups *))); 3196 rcu_read_unlock(); 3197 rcu_assign_pointer(sbi->s_flex_groups, new_groups); 3198 sbi->s_flex_groups_allocated = size; 3199 if (old_groups) 3200 ext4_kvfree_array_rcu(old_groups); 3201 return 0; 3202 } 3203 3204 static int ext4_fill_flex_info(struct super_block *sb) 3205 { 3206 struct ext4_sb_info *sbi = EXT4_SB(sb); 3207 struct ext4_group_desc *gdp = NULL; 3208 struct flex_groups *fg; 3209 ext4_group_t flex_group; 3210 int i, err; 3211 3212 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 3213 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 3214 sbi->s_log_groups_per_flex = 0; 3215 return 1; 3216 } 3217 3218 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 3219 if (err) 3220 goto failed; 3221 3222 for (i = 0; i < sbi->s_groups_count; i++) { 3223 gdp = ext4_get_group_desc(sb, i, NULL); 3224 3225 flex_group = ext4_flex_group(sbi, i); 3226 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group); 3227 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes); 3228 atomic64_add(ext4_free_group_clusters(sb, gdp), 3229 &fg->free_clusters); 3230 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs); 3231 } 3232 3233 return 1; 3234 failed: 3235 return 0; 3236 } 3237 3238 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 3239 struct ext4_group_desc *gdp) 3240 { 3241 int offset = offsetof(struct ext4_group_desc, bg_checksum); 3242 __u16 crc = 0; 3243 __le32 le_group = cpu_to_le32(block_group); 3244 struct ext4_sb_info *sbi = EXT4_SB(sb); 3245 3246 if (ext4_has_metadata_csum(sbi->s_sb)) { 3247 /* Use new metadata_csum algorithm */ 3248 __u32 csum32; 3249 __u16 dummy_csum = 0; 3250 3251 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 3252 sizeof(le_group)); 3253 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); 3254 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, 3255 sizeof(dummy_csum)); 3256 offset += sizeof(dummy_csum); 3257 if (offset < sbi->s_desc_size) 3258 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, 3259 sbi->s_desc_size - offset); 3260 3261 crc = csum32 & 0xFFFF; 3262 goto out; 3263 } 3264 3265 /* old crc16 code */ 3266 if (!ext4_has_feature_gdt_csum(sb)) 3267 return 0; 3268 3269 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 3270 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 3271 crc = crc16(crc, (__u8 *)gdp, offset); 3272 offset += sizeof(gdp->bg_checksum); /* skip checksum */ 3273 /* for checksum of struct ext4_group_desc do the rest...*/ 3274 if (ext4_has_feature_64bit(sb) && 3275 offset < le16_to_cpu(sbi->s_es->s_desc_size)) 3276 crc = crc16(crc, (__u8 *)gdp + offset, 3277 le16_to_cpu(sbi->s_es->s_desc_size) - 3278 offset); 3279 3280 out: 3281 return cpu_to_le16(crc); 3282 } 3283 3284 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 3285 struct ext4_group_desc *gdp) 3286 { 3287 if (ext4_has_group_desc_csum(sb) && 3288 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 3289 return 0; 3290 3291 return 1; 3292 } 3293 3294 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 3295 struct ext4_group_desc *gdp) 3296 { 3297 if (!ext4_has_group_desc_csum(sb)) 3298 return; 3299 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 3300 } 3301 3302 /* Called at mount-time, super-block is locked */ 3303 static int ext4_check_descriptors(struct super_block *sb, 3304 ext4_fsblk_t sb_block, 3305 ext4_group_t *first_not_zeroed) 3306 { 3307 struct ext4_sb_info *sbi = EXT4_SB(sb); 3308 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 3309 ext4_fsblk_t last_block; 3310 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0); 3311 ext4_fsblk_t block_bitmap; 3312 ext4_fsblk_t inode_bitmap; 3313 ext4_fsblk_t inode_table; 3314 int flexbg_flag = 0; 3315 ext4_group_t i, grp = sbi->s_groups_count; 3316 3317 if (ext4_has_feature_flex_bg(sb)) 3318 flexbg_flag = 1; 3319 3320 ext4_debug("Checking group descriptors"); 3321 3322 for (i = 0; i < sbi->s_groups_count; i++) { 3323 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 3324 3325 if (i == sbi->s_groups_count - 1 || flexbg_flag) 3326 last_block = ext4_blocks_count(sbi->s_es) - 1; 3327 else 3328 last_block = first_block + 3329 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 3330 3331 if ((grp == sbi->s_groups_count) && 3332 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3333 grp = i; 3334 3335 block_bitmap = ext4_block_bitmap(sb, gdp); 3336 if (block_bitmap == sb_block) { 3337 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3338 "Block bitmap for group %u overlaps " 3339 "superblock", i); 3340 if (!sb_rdonly(sb)) 3341 return 0; 3342 } 3343 if (block_bitmap >= sb_block + 1 && 3344 block_bitmap <= last_bg_block) { 3345 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3346 "Block bitmap for group %u overlaps " 3347 "block group descriptors", i); 3348 if (!sb_rdonly(sb)) 3349 return 0; 3350 } 3351 if (block_bitmap < first_block || block_bitmap > last_block) { 3352 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3353 "Block bitmap for group %u not in group " 3354 "(block %llu)!", i, block_bitmap); 3355 return 0; 3356 } 3357 inode_bitmap = ext4_inode_bitmap(sb, gdp); 3358 if (inode_bitmap == sb_block) { 3359 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3360 "Inode bitmap for group %u overlaps " 3361 "superblock", i); 3362 if (!sb_rdonly(sb)) 3363 return 0; 3364 } 3365 if (inode_bitmap >= sb_block + 1 && 3366 inode_bitmap <= last_bg_block) { 3367 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3368 "Inode bitmap for group %u overlaps " 3369 "block group descriptors", i); 3370 if (!sb_rdonly(sb)) 3371 return 0; 3372 } 3373 if (inode_bitmap < first_block || inode_bitmap > last_block) { 3374 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3375 "Inode bitmap for group %u not in group " 3376 "(block %llu)!", i, inode_bitmap); 3377 return 0; 3378 } 3379 inode_table = ext4_inode_table(sb, gdp); 3380 if (inode_table == sb_block) { 3381 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3382 "Inode table for group %u overlaps " 3383 "superblock", i); 3384 if (!sb_rdonly(sb)) 3385 return 0; 3386 } 3387 if (inode_table >= sb_block + 1 && 3388 inode_table <= last_bg_block) { 3389 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3390 "Inode table for group %u overlaps " 3391 "block group descriptors", i); 3392 if (!sb_rdonly(sb)) 3393 return 0; 3394 } 3395 if (inode_table < first_block || 3396 inode_table + sbi->s_itb_per_group - 1 > last_block) { 3397 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3398 "Inode table for group %u not in group " 3399 "(block %llu)!", i, inode_table); 3400 return 0; 3401 } 3402 ext4_lock_group(sb, i); 3403 if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 3404 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3405 "Checksum for group %u failed (%u!=%u)", 3406 i, le16_to_cpu(ext4_group_desc_csum(sb, i, 3407 gdp)), le16_to_cpu(gdp->bg_checksum)); 3408 if (!sb_rdonly(sb)) { 3409 ext4_unlock_group(sb, i); 3410 return 0; 3411 } 3412 } 3413 ext4_unlock_group(sb, i); 3414 if (!flexbg_flag) 3415 first_block += EXT4_BLOCKS_PER_GROUP(sb); 3416 } 3417 if (NULL != first_not_zeroed) 3418 *first_not_zeroed = grp; 3419 return 1; 3420 } 3421 3422 /* 3423 * Maximal extent format file size. 3424 * Resulting logical blkno at s_maxbytes must fit in our on-disk 3425 * extent format containers, within a sector_t, and within i_blocks 3426 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 3427 * so that won't be a limiting factor. 3428 * 3429 * However there is other limiting factor. We do store extents in the form 3430 * of starting block and length, hence the resulting length of the extent 3431 * covering maximum file size must fit into on-disk format containers as 3432 * well. Given that length is always by 1 unit bigger than max unit (because 3433 * we count 0 as well) we have to lower the s_maxbytes by one fs block. 3434 * 3435 * Note, this does *not* consider any metadata overhead for vfs i_blocks. 3436 */ 3437 static loff_t ext4_max_size(int blkbits, int has_huge_files) 3438 { 3439 loff_t res; 3440 loff_t upper_limit = MAX_LFS_FILESIZE; 3441 3442 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64)); 3443 3444 if (!has_huge_files) { 3445 upper_limit = (1LL << 32) - 1; 3446 3447 /* total blocks in file system block size */ 3448 upper_limit >>= (blkbits - 9); 3449 upper_limit <<= blkbits; 3450 } 3451 3452 /* 3453 * 32-bit extent-start container, ee_block. We lower the maxbytes 3454 * by one fs block, so ee_len can cover the extent of maximum file 3455 * size 3456 */ 3457 res = (1LL << 32) - 1; 3458 res <<= blkbits; 3459 3460 /* Sanity check against vm- & vfs- imposed limits */ 3461 if (res > upper_limit) 3462 res = upper_limit; 3463 3464 return res; 3465 } 3466 3467 /* 3468 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 3469 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 3470 * We need to be 1 filesystem block less than the 2^48 sector limit. 3471 */ 3472 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 3473 { 3474 unsigned long long upper_limit, res = EXT4_NDIR_BLOCKS; 3475 int meta_blocks; 3476 3477 /* 3478 * This is calculated to be the largest file size for a dense, block 3479 * mapped file such that the file's total number of 512-byte sectors, 3480 * including data and all indirect blocks, does not exceed (2^48 - 1). 3481 * 3482 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 3483 * number of 512-byte sectors of the file. 3484 */ 3485 if (!has_huge_files) { 3486 /* 3487 * !has_huge_files or implies that the inode i_block field 3488 * represents total file blocks in 2^32 512-byte sectors == 3489 * size of vfs inode i_blocks * 8 3490 */ 3491 upper_limit = (1LL << 32) - 1; 3492 3493 /* total blocks in file system block size */ 3494 upper_limit >>= (bits - 9); 3495 3496 } else { 3497 /* 3498 * We use 48 bit ext4_inode i_blocks 3499 * With EXT4_HUGE_FILE_FL set the i_blocks 3500 * represent total number of blocks in 3501 * file system block size 3502 */ 3503 upper_limit = (1LL << 48) - 1; 3504 3505 } 3506 3507 /* indirect blocks */ 3508 meta_blocks = 1; 3509 /* double indirect blocks */ 3510 meta_blocks += 1 + (1LL << (bits-2)); 3511 /* tripple indirect blocks */ 3512 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 3513 3514 upper_limit -= meta_blocks; 3515 upper_limit <<= bits; 3516 3517 res += 1LL << (bits-2); 3518 res += 1LL << (2*(bits-2)); 3519 res += 1LL << (3*(bits-2)); 3520 res <<= bits; 3521 if (res > upper_limit) 3522 res = upper_limit; 3523 3524 if (res > MAX_LFS_FILESIZE) 3525 res = MAX_LFS_FILESIZE; 3526 3527 return (loff_t)res; 3528 } 3529 3530 static ext4_fsblk_t descriptor_loc(struct super_block *sb, 3531 ext4_fsblk_t logical_sb_block, int nr) 3532 { 3533 struct ext4_sb_info *sbi = EXT4_SB(sb); 3534 ext4_group_t bg, first_meta_bg; 3535 int has_super = 0; 3536 3537 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 3538 3539 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 3540 return logical_sb_block + nr + 1; 3541 bg = sbi->s_desc_per_block * nr; 3542 if (ext4_bg_has_super(sb, bg)) 3543 has_super = 1; 3544 3545 /* 3546 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 3547 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 3548 * on modern mke2fs or blksize > 1k on older mke2fs) then we must 3549 * compensate. 3550 */ 3551 if (sb->s_blocksize == 1024 && nr == 0 && 3552 le32_to_cpu(sbi->s_es->s_first_data_block) == 0) 3553 has_super++; 3554 3555 return (has_super + ext4_group_first_block_no(sb, bg)); 3556 } 3557 3558 /** 3559 * ext4_get_stripe_size: Get the stripe size. 3560 * @sbi: In memory super block info 3561 * 3562 * If we have specified it via mount option, then 3563 * use the mount option value. If the value specified at mount time is 3564 * greater than the blocks per group use the super block value. 3565 * If the super block value is greater than blocks per group return 0. 3566 * Allocator needs it be less than blocks per group. 3567 * 3568 */ 3569 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 3570 { 3571 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 3572 unsigned long stripe_width = 3573 le32_to_cpu(sbi->s_es->s_raid_stripe_width); 3574 int ret; 3575 3576 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 3577 ret = sbi->s_stripe; 3578 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) 3579 ret = stripe_width; 3580 else if (stride && stride <= sbi->s_blocks_per_group) 3581 ret = stride; 3582 else 3583 ret = 0; 3584 3585 /* 3586 * If the stripe width is 1, this makes no sense and 3587 * we set it to 0 to turn off stripe handling code. 3588 */ 3589 if (ret <= 1) 3590 ret = 0; 3591 3592 return ret; 3593 } 3594 3595 /* 3596 * Check whether this filesystem can be mounted based on 3597 * the features present and the RDONLY/RDWR mount requested. 3598 * Returns 1 if this filesystem can be mounted as requested, 3599 * 0 if it cannot be. 3600 */ 3601 int ext4_feature_set_ok(struct super_block *sb, int readonly) 3602 { 3603 if (ext4_has_unknown_ext4_incompat_features(sb)) { 3604 ext4_msg(sb, KERN_ERR, 3605 "Couldn't mount because of " 3606 "unsupported optional features (%x)", 3607 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 3608 ~EXT4_FEATURE_INCOMPAT_SUPP)); 3609 return 0; 3610 } 3611 3612 #ifndef CONFIG_UNICODE 3613 if (ext4_has_feature_casefold(sb)) { 3614 ext4_msg(sb, KERN_ERR, 3615 "Filesystem with casefold feature cannot be " 3616 "mounted without CONFIG_UNICODE"); 3617 return 0; 3618 } 3619 #endif 3620 3621 if (readonly) 3622 return 1; 3623 3624 if (ext4_has_feature_readonly(sb)) { 3625 ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 3626 sb->s_flags |= SB_RDONLY; 3627 return 1; 3628 } 3629 3630 /* Check that feature set is OK for a read-write mount */ 3631 if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 3632 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 3633 "unsupported optional features (%x)", 3634 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 3635 ~EXT4_FEATURE_RO_COMPAT_SUPP)); 3636 return 0; 3637 } 3638 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 3639 ext4_msg(sb, KERN_ERR, 3640 "Can't support bigalloc feature without " 3641 "extents feature\n"); 3642 return 0; 3643 } 3644 3645 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2) 3646 if (!readonly && (ext4_has_feature_quota(sb) || 3647 ext4_has_feature_project(sb))) { 3648 ext4_msg(sb, KERN_ERR, 3649 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2"); 3650 return 0; 3651 } 3652 #endif /* CONFIG_QUOTA */ 3653 return 1; 3654 } 3655 3656 /* 3657 * This function is called once a day if we have errors logged 3658 * on the file system 3659 */ 3660 static void print_daily_error_info(struct timer_list *t) 3661 { 3662 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report); 3663 struct super_block *sb = sbi->s_sb; 3664 struct ext4_super_block *es = sbi->s_es; 3665 3666 if (es->s_error_count) 3667 /* fsck newer than v1.41.13 is needed to clean this condition. */ 3668 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 3669 le32_to_cpu(es->s_error_count)); 3670 if (es->s_first_error_time) { 3671 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d", 3672 sb->s_id, 3673 ext4_get_tstamp(es, s_first_error_time), 3674 (int) sizeof(es->s_first_error_func), 3675 es->s_first_error_func, 3676 le32_to_cpu(es->s_first_error_line)); 3677 if (es->s_first_error_ino) 3678 printk(KERN_CONT ": inode %u", 3679 le32_to_cpu(es->s_first_error_ino)); 3680 if (es->s_first_error_block) 3681 printk(KERN_CONT ": block %llu", (unsigned long long) 3682 le64_to_cpu(es->s_first_error_block)); 3683 printk(KERN_CONT "\n"); 3684 } 3685 if (es->s_last_error_time) { 3686 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d", 3687 sb->s_id, 3688 ext4_get_tstamp(es, s_last_error_time), 3689 (int) sizeof(es->s_last_error_func), 3690 es->s_last_error_func, 3691 le32_to_cpu(es->s_last_error_line)); 3692 if (es->s_last_error_ino) 3693 printk(KERN_CONT ": inode %u", 3694 le32_to_cpu(es->s_last_error_ino)); 3695 if (es->s_last_error_block) 3696 printk(KERN_CONT ": block %llu", (unsigned long long) 3697 le64_to_cpu(es->s_last_error_block)); 3698 printk(KERN_CONT "\n"); 3699 } 3700 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 3701 } 3702 3703 /* Find next suitable group and run ext4_init_inode_table */ 3704 static int ext4_run_li_request(struct ext4_li_request *elr) 3705 { 3706 struct ext4_group_desc *gdp = NULL; 3707 struct super_block *sb = elr->lr_super; 3708 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 3709 ext4_group_t group = elr->lr_next_group; 3710 unsigned int prefetch_ios = 0; 3711 int ret = 0; 3712 u64 start_time; 3713 3714 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) { 3715 elr->lr_next_group = ext4_mb_prefetch(sb, group, 3716 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios); 3717 if (prefetch_ios) 3718 ext4_mb_prefetch_fini(sb, elr->lr_next_group, 3719 prefetch_ios); 3720 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, 3721 prefetch_ios); 3722 if (group >= elr->lr_next_group) { 3723 ret = 1; 3724 if (elr->lr_first_not_zeroed != ngroups && 3725 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) { 3726 elr->lr_next_group = elr->lr_first_not_zeroed; 3727 elr->lr_mode = EXT4_LI_MODE_ITABLE; 3728 ret = 0; 3729 } 3730 } 3731 return ret; 3732 } 3733 3734 for (; group < ngroups; group++) { 3735 gdp = ext4_get_group_desc(sb, group, NULL); 3736 if (!gdp) { 3737 ret = 1; 3738 break; 3739 } 3740 3741 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3742 break; 3743 } 3744 3745 if (group >= ngroups) 3746 ret = 1; 3747 3748 if (!ret) { 3749 start_time = ktime_get_real_ns(); 3750 ret = ext4_init_inode_table(sb, group, 3751 elr->lr_timeout ? 0 : 1); 3752 trace_ext4_lazy_itable_init(sb, group); 3753 if (elr->lr_timeout == 0) { 3754 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) * 3755 EXT4_SB(elr->lr_super)->s_li_wait_mult); 3756 } 3757 elr->lr_next_sched = jiffies + elr->lr_timeout; 3758 elr->lr_next_group = group + 1; 3759 } 3760 return ret; 3761 } 3762 3763 /* 3764 * Remove lr_request from the list_request and free the 3765 * request structure. Should be called with li_list_mtx held 3766 */ 3767 static void ext4_remove_li_request(struct ext4_li_request *elr) 3768 { 3769 if (!elr) 3770 return; 3771 3772 list_del(&elr->lr_request); 3773 EXT4_SB(elr->lr_super)->s_li_request = NULL; 3774 kfree(elr); 3775 } 3776 3777 static void ext4_unregister_li_request(struct super_block *sb) 3778 { 3779 mutex_lock(&ext4_li_mtx); 3780 if (!ext4_li_info) { 3781 mutex_unlock(&ext4_li_mtx); 3782 return; 3783 } 3784 3785 mutex_lock(&ext4_li_info->li_list_mtx); 3786 ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 3787 mutex_unlock(&ext4_li_info->li_list_mtx); 3788 mutex_unlock(&ext4_li_mtx); 3789 } 3790 3791 static struct task_struct *ext4_lazyinit_task; 3792 3793 /* 3794 * This is the function where ext4lazyinit thread lives. It walks 3795 * through the request list searching for next scheduled filesystem. 3796 * When such a fs is found, run the lazy initialization request 3797 * (ext4_rn_li_request) and keep track of the time spend in this 3798 * function. Based on that time we compute next schedule time of 3799 * the request. When walking through the list is complete, compute 3800 * next waking time and put itself into sleep. 3801 */ 3802 static int ext4_lazyinit_thread(void *arg) 3803 { 3804 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 3805 struct list_head *pos, *n; 3806 struct ext4_li_request *elr; 3807 unsigned long next_wakeup, cur; 3808 3809 BUG_ON(NULL == eli); 3810 3811 cont_thread: 3812 while (true) { 3813 next_wakeup = MAX_JIFFY_OFFSET; 3814 3815 mutex_lock(&eli->li_list_mtx); 3816 if (list_empty(&eli->li_request_list)) { 3817 mutex_unlock(&eli->li_list_mtx); 3818 goto exit_thread; 3819 } 3820 list_for_each_safe(pos, n, &eli->li_request_list) { 3821 int err = 0; 3822 int progress = 0; 3823 elr = list_entry(pos, struct ext4_li_request, 3824 lr_request); 3825 3826 if (time_before(jiffies, elr->lr_next_sched)) { 3827 if (time_before(elr->lr_next_sched, next_wakeup)) 3828 next_wakeup = elr->lr_next_sched; 3829 continue; 3830 } 3831 if (down_read_trylock(&elr->lr_super->s_umount)) { 3832 if (sb_start_write_trylock(elr->lr_super)) { 3833 progress = 1; 3834 /* 3835 * We hold sb->s_umount, sb can not 3836 * be removed from the list, it is 3837 * now safe to drop li_list_mtx 3838 */ 3839 mutex_unlock(&eli->li_list_mtx); 3840 err = ext4_run_li_request(elr); 3841 sb_end_write(elr->lr_super); 3842 mutex_lock(&eli->li_list_mtx); 3843 n = pos->next; 3844 } 3845 up_read((&elr->lr_super->s_umount)); 3846 } 3847 /* error, remove the lazy_init job */ 3848 if (err) { 3849 ext4_remove_li_request(elr); 3850 continue; 3851 } 3852 if (!progress) { 3853 elr->lr_next_sched = jiffies + 3854 (prandom_u32() 3855 % (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3856 } 3857 if (time_before(elr->lr_next_sched, next_wakeup)) 3858 next_wakeup = elr->lr_next_sched; 3859 } 3860 mutex_unlock(&eli->li_list_mtx); 3861 3862 try_to_freeze(); 3863 3864 cur = jiffies; 3865 if ((time_after_eq(cur, next_wakeup)) || 3866 (MAX_JIFFY_OFFSET == next_wakeup)) { 3867 cond_resched(); 3868 continue; 3869 } 3870 3871 schedule_timeout_interruptible(next_wakeup - cur); 3872 3873 if (kthread_should_stop()) { 3874 ext4_clear_request_list(); 3875 goto exit_thread; 3876 } 3877 } 3878 3879 exit_thread: 3880 /* 3881 * It looks like the request list is empty, but we need 3882 * to check it under the li_list_mtx lock, to prevent any 3883 * additions into it, and of course we should lock ext4_li_mtx 3884 * to atomically free the list and ext4_li_info, because at 3885 * this point another ext4 filesystem could be registering 3886 * new one. 3887 */ 3888 mutex_lock(&ext4_li_mtx); 3889 mutex_lock(&eli->li_list_mtx); 3890 if (!list_empty(&eli->li_request_list)) { 3891 mutex_unlock(&eli->li_list_mtx); 3892 mutex_unlock(&ext4_li_mtx); 3893 goto cont_thread; 3894 } 3895 mutex_unlock(&eli->li_list_mtx); 3896 kfree(ext4_li_info); 3897 ext4_li_info = NULL; 3898 mutex_unlock(&ext4_li_mtx); 3899 3900 return 0; 3901 } 3902 3903 static void ext4_clear_request_list(void) 3904 { 3905 struct list_head *pos, *n; 3906 struct ext4_li_request *elr; 3907 3908 mutex_lock(&ext4_li_info->li_list_mtx); 3909 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3910 elr = list_entry(pos, struct ext4_li_request, 3911 lr_request); 3912 ext4_remove_li_request(elr); 3913 } 3914 mutex_unlock(&ext4_li_info->li_list_mtx); 3915 } 3916 3917 static int ext4_run_lazyinit_thread(void) 3918 { 3919 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 3920 ext4_li_info, "ext4lazyinit"); 3921 if (IS_ERR(ext4_lazyinit_task)) { 3922 int err = PTR_ERR(ext4_lazyinit_task); 3923 ext4_clear_request_list(); 3924 kfree(ext4_li_info); 3925 ext4_li_info = NULL; 3926 printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3927 "initialization thread\n", 3928 err); 3929 return err; 3930 } 3931 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3932 return 0; 3933 } 3934 3935 /* 3936 * Check whether it make sense to run itable init. thread or not. 3937 * If there is at least one uninitialized inode table, return 3938 * corresponding group number, else the loop goes through all 3939 * groups and return total number of groups. 3940 */ 3941 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3942 { 3943 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3944 struct ext4_group_desc *gdp = NULL; 3945 3946 if (!ext4_has_group_desc_csum(sb)) 3947 return ngroups; 3948 3949 for (group = 0; group < ngroups; group++) { 3950 gdp = ext4_get_group_desc(sb, group, NULL); 3951 if (!gdp) 3952 continue; 3953 3954 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3955 break; 3956 } 3957 3958 return group; 3959 } 3960 3961 static int ext4_li_info_new(void) 3962 { 3963 struct ext4_lazy_init *eli = NULL; 3964 3965 eli = kzalloc(sizeof(*eli), GFP_KERNEL); 3966 if (!eli) 3967 return -ENOMEM; 3968 3969 INIT_LIST_HEAD(&eli->li_request_list); 3970 mutex_init(&eli->li_list_mtx); 3971 3972 eli->li_state |= EXT4_LAZYINIT_QUIT; 3973 3974 ext4_li_info = eli; 3975 3976 return 0; 3977 } 3978 3979 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3980 ext4_group_t start) 3981 { 3982 struct ext4_li_request *elr; 3983 3984 elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3985 if (!elr) 3986 return NULL; 3987 3988 elr->lr_super = sb; 3989 elr->lr_first_not_zeroed = start; 3990 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) { 3991 elr->lr_mode = EXT4_LI_MODE_ITABLE; 3992 elr->lr_next_group = start; 3993 } else { 3994 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP; 3995 } 3996 3997 /* 3998 * Randomize first schedule time of the request to 3999 * spread the inode table initialization requests 4000 * better. 4001 */ 4002 elr->lr_next_sched = jiffies + (prandom_u32() % 4003 (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 4004 return elr; 4005 } 4006 4007 int ext4_register_li_request(struct super_block *sb, 4008 ext4_group_t first_not_zeroed) 4009 { 4010 struct ext4_sb_info *sbi = EXT4_SB(sb); 4011 struct ext4_li_request *elr = NULL; 4012 ext4_group_t ngroups = sbi->s_groups_count; 4013 int ret = 0; 4014 4015 mutex_lock(&ext4_li_mtx); 4016 if (sbi->s_li_request != NULL) { 4017 /* 4018 * Reset timeout so it can be computed again, because 4019 * s_li_wait_mult might have changed. 4020 */ 4021 sbi->s_li_request->lr_timeout = 0; 4022 goto out; 4023 } 4024 4025 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) && 4026 (first_not_zeroed == ngroups || sb_rdonly(sb) || 4027 !test_opt(sb, INIT_INODE_TABLE))) 4028 goto out; 4029 4030 elr = ext4_li_request_new(sb, first_not_zeroed); 4031 if (!elr) { 4032 ret = -ENOMEM; 4033 goto out; 4034 } 4035 4036 if (NULL == ext4_li_info) { 4037 ret = ext4_li_info_new(); 4038 if (ret) 4039 goto out; 4040 } 4041 4042 mutex_lock(&ext4_li_info->li_list_mtx); 4043 list_add(&elr->lr_request, &ext4_li_info->li_request_list); 4044 mutex_unlock(&ext4_li_info->li_list_mtx); 4045 4046 sbi->s_li_request = elr; 4047 /* 4048 * set elr to NULL here since it has been inserted to 4049 * the request_list and the removal and free of it is 4050 * handled by ext4_clear_request_list from now on. 4051 */ 4052 elr = NULL; 4053 4054 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 4055 ret = ext4_run_lazyinit_thread(); 4056 if (ret) 4057 goto out; 4058 } 4059 out: 4060 mutex_unlock(&ext4_li_mtx); 4061 if (ret) 4062 kfree(elr); 4063 return ret; 4064 } 4065 4066 /* 4067 * We do not need to lock anything since this is called on 4068 * module unload. 4069 */ 4070 static void ext4_destroy_lazyinit_thread(void) 4071 { 4072 /* 4073 * If thread exited earlier 4074 * there's nothing to be done. 4075 */ 4076 if (!ext4_li_info || !ext4_lazyinit_task) 4077 return; 4078 4079 kthread_stop(ext4_lazyinit_task); 4080 } 4081 4082 static int set_journal_csum_feature_set(struct super_block *sb) 4083 { 4084 int ret = 1; 4085 int compat, incompat; 4086 struct ext4_sb_info *sbi = EXT4_SB(sb); 4087 4088 if (ext4_has_metadata_csum(sb)) { 4089 /* journal checksum v3 */ 4090 compat = 0; 4091 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 4092 } else { 4093 /* journal checksum v1 */ 4094 compat = JBD2_FEATURE_COMPAT_CHECKSUM; 4095 incompat = 0; 4096 } 4097 4098 jbd2_journal_clear_features(sbi->s_journal, 4099 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 4100 JBD2_FEATURE_INCOMPAT_CSUM_V3 | 4101 JBD2_FEATURE_INCOMPAT_CSUM_V2); 4102 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4103 ret = jbd2_journal_set_features(sbi->s_journal, 4104 compat, 0, 4105 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 4106 incompat); 4107 } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 4108 ret = jbd2_journal_set_features(sbi->s_journal, 4109 compat, 0, 4110 incompat); 4111 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 4112 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 4113 } else { 4114 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 4115 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 4116 } 4117 4118 return ret; 4119 } 4120 4121 /* 4122 * Note: calculating the overhead so we can be compatible with 4123 * historical BSD practice is quite difficult in the face of 4124 * clusters/bigalloc. This is because multiple metadata blocks from 4125 * different block group can end up in the same allocation cluster. 4126 * Calculating the exact overhead in the face of clustered allocation 4127 * requires either O(all block bitmaps) in memory or O(number of block 4128 * groups**2) in time. We will still calculate the superblock for 4129 * older file systems --- and if we come across with a bigalloc file 4130 * system with zero in s_overhead_clusters the estimate will be close to 4131 * correct especially for very large cluster sizes --- but for newer 4132 * file systems, it's better to calculate this figure once at mkfs 4133 * time, and store it in the superblock. If the superblock value is 4134 * present (even for non-bigalloc file systems), we will use it. 4135 */ 4136 static int count_overhead(struct super_block *sb, ext4_group_t grp, 4137 char *buf) 4138 { 4139 struct ext4_sb_info *sbi = EXT4_SB(sb); 4140 struct ext4_group_desc *gdp; 4141 ext4_fsblk_t first_block, last_block, b; 4142 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 4143 int s, j, count = 0; 4144 4145 if (!ext4_has_feature_bigalloc(sb)) 4146 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + 4147 sbi->s_itb_per_group + 2); 4148 4149 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 4150 (grp * EXT4_BLOCKS_PER_GROUP(sb)); 4151 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 4152 for (i = 0; i < ngroups; i++) { 4153 gdp = ext4_get_group_desc(sb, i, NULL); 4154 b = ext4_block_bitmap(sb, gdp); 4155 if (b >= first_block && b <= last_block) { 4156 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 4157 count++; 4158 } 4159 b = ext4_inode_bitmap(sb, gdp); 4160 if (b >= first_block && b <= last_block) { 4161 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 4162 count++; 4163 } 4164 b = ext4_inode_table(sb, gdp); 4165 if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 4166 for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 4167 int c = EXT4_B2C(sbi, b - first_block); 4168 ext4_set_bit(c, buf); 4169 count++; 4170 } 4171 if (i != grp) 4172 continue; 4173 s = 0; 4174 if (ext4_bg_has_super(sb, grp)) { 4175 ext4_set_bit(s++, buf); 4176 count++; 4177 } 4178 j = ext4_bg_num_gdb(sb, grp); 4179 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { 4180 ext4_error(sb, "Invalid number of block group " 4181 "descriptor blocks: %d", j); 4182 j = EXT4_BLOCKS_PER_GROUP(sb) - s; 4183 } 4184 count += j; 4185 for (; j > 0; j--) 4186 ext4_set_bit(EXT4_B2C(sbi, s++), buf); 4187 } 4188 if (!count) 4189 return 0; 4190 return EXT4_CLUSTERS_PER_GROUP(sb) - 4191 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 4192 } 4193 4194 /* 4195 * Compute the overhead and stash it in sbi->s_overhead 4196 */ 4197 int ext4_calculate_overhead(struct super_block *sb) 4198 { 4199 struct ext4_sb_info *sbi = EXT4_SB(sb); 4200 struct ext4_super_block *es = sbi->s_es; 4201 struct inode *j_inode; 4202 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); 4203 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 4204 ext4_fsblk_t overhead = 0; 4205 char *buf = (char *) get_zeroed_page(GFP_NOFS); 4206 4207 if (!buf) 4208 return -ENOMEM; 4209 4210 /* 4211 * Compute the overhead (FS structures). This is constant 4212 * for a given filesystem unless the number of block groups 4213 * changes so we cache the previous value until it does. 4214 */ 4215 4216 /* 4217 * All of the blocks before first_data_block are overhead 4218 */ 4219 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 4220 4221 /* 4222 * Add the overhead found in each block group 4223 */ 4224 for (i = 0; i < ngroups; i++) { 4225 int blks; 4226 4227 blks = count_overhead(sb, i, buf); 4228 overhead += blks; 4229 if (blks) 4230 memset(buf, 0, PAGE_SIZE); 4231 cond_resched(); 4232 } 4233 4234 /* 4235 * Add the internal journal blocks whether the journal has been 4236 * loaded or not 4237 */ 4238 if (sbi->s_journal && !sbi->s_journal_bdev) 4239 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len); 4240 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) { 4241 /* j_inum for internal journal is non-zero */ 4242 j_inode = ext4_get_journal_inode(sb, j_inum); 4243 if (j_inode) { 4244 j_blocks = j_inode->i_size >> sb->s_blocksize_bits; 4245 overhead += EXT4_NUM_B2C(sbi, j_blocks); 4246 iput(j_inode); 4247 } else { 4248 ext4_msg(sb, KERN_ERR, "can't get journal size"); 4249 } 4250 } 4251 sbi->s_overhead = overhead; 4252 smp_wmb(); 4253 free_page((unsigned long) buf); 4254 return 0; 4255 } 4256 4257 static void ext4_set_resv_clusters(struct super_block *sb) 4258 { 4259 ext4_fsblk_t resv_clusters; 4260 struct ext4_sb_info *sbi = EXT4_SB(sb); 4261 4262 /* 4263 * There's no need to reserve anything when we aren't using extents. 4264 * The space estimates are exact, there are no unwritten extents, 4265 * hole punching doesn't need new metadata... This is needed especially 4266 * to keep ext2/3 backward compatibility. 4267 */ 4268 if (!ext4_has_feature_extents(sb)) 4269 return; 4270 /* 4271 * By default we reserve 2% or 4096 clusters, whichever is smaller. 4272 * This should cover the situations where we can not afford to run 4273 * out of space like for example punch hole, or converting 4274 * unwritten extents in delalloc path. In most cases such 4275 * allocation would require 1, or 2 blocks, higher numbers are 4276 * very rare. 4277 */ 4278 resv_clusters = (ext4_blocks_count(sbi->s_es) >> 4279 sbi->s_cluster_bits); 4280 4281 do_div(resv_clusters, 50); 4282 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 4283 4284 atomic64_set(&sbi->s_resv_clusters, resv_clusters); 4285 } 4286 4287 static const char *ext4_quota_mode(struct super_block *sb) 4288 { 4289 #ifdef CONFIG_QUOTA 4290 if (!ext4_quota_capable(sb)) 4291 return "none"; 4292 4293 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb)) 4294 return "journalled"; 4295 else 4296 return "writeback"; 4297 #else 4298 return "disabled"; 4299 #endif 4300 } 4301 4302 static void ext4_setup_csum_trigger(struct super_block *sb, 4303 enum ext4_journal_trigger_type type, 4304 void (*trigger)( 4305 struct jbd2_buffer_trigger_type *type, 4306 struct buffer_head *bh, 4307 void *mapped_data, 4308 size_t size)) 4309 { 4310 struct ext4_sb_info *sbi = EXT4_SB(sb); 4311 4312 sbi->s_journal_triggers[type].sb = sb; 4313 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger; 4314 } 4315 4316 static void ext4_free_sbi(struct ext4_sb_info *sbi) 4317 { 4318 if (!sbi) 4319 return; 4320 4321 kfree(sbi->s_blockgroup_lock); 4322 fs_put_dax(sbi->s_daxdev); 4323 kfree(sbi); 4324 } 4325 4326 static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb) 4327 { 4328 struct ext4_sb_info *sbi; 4329 4330 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 4331 if (!sbi) 4332 return NULL; 4333 4334 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off); 4335 4336 sbi->s_blockgroup_lock = 4337 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 4338 4339 if (!sbi->s_blockgroup_lock) 4340 goto err_out; 4341 4342 sb->s_fs_info = sbi; 4343 sbi->s_sb = sb; 4344 return sbi; 4345 err_out: 4346 fs_put_dax(sbi->s_daxdev); 4347 kfree(sbi); 4348 return NULL; 4349 } 4350 4351 static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb) 4352 { 4353 struct buffer_head *bh, **group_desc; 4354 struct ext4_super_block *es = NULL; 4355 struct ext4_sb_info *sbi = EXT4_SB(sb); 4356 struct flex_groups **flex_groups; 4357 ext4_fsblk_t block; 4358 ext4_fsblk_t logical_sb_block; 4359 unsigned long offset = 0; 4360 unsigned long def_mount_opts; 4361 struct inode *root; 4362 int ret = -ENOMEM; 4363 int blocksize, clustersize; 4364 unsigned int db_count; 4365 unsigned int i; 4366 int needs_recovery, has_huge_files; 4367 __u64 blocks_count; 4368 int err = 0; 4369 ext4_group_t first_not_zeroed; 4370 struct ext4_fs_context *ctx = fc->fs_private; 4371 int silent = fc->sb_flags & SB_SILENT; 4372 4373 /* Set defaults for the variables that will be set during parsing */ 4374 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 4375 ctx->mb_optimize_scan = DEFAULT_MB_OPTIMIZE_SCAN; 4376 4377 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 4378 sbi->s_sectors_written_start = 4379 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]); 4380 4381 /* -EINVAL is default */ 4382 ret = -EINVAL; 4383 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 4384 if (!blocksize) { 4385 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 4386 goto out_fail; 4387 } 4388 4389 /* 4390 * The ext4 superblock will not be buffer aligned for other than 1kB 4391 * block sizes. We need to calculate the offset from buffer start. 4392 */ 4393 if (blocksize != EXT4_MIN_BLOCK_SIZE) { 4394 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE; 4395 offset = do_div(logical_sb_block, blocksize); 4396 } else { 4397 logical_sb_block = sbi->s_sb_block; 4398 } 4399 4400 bh = ext4_sb_bread_unmovable(sb, logical_sb_block); 4401 if (IS_ERR(bh)) { 4402 ext4_msg(sb, KERN_ERR, "unable to read superblock"); 4403 ret = PTR_ERR(bh); 4404 goto out_fail; 4405 } 4406 /* 4407 * Note: s_es must be initialized as soon as possible because 4408 * some ext4 macro-instructions depend on its value 4409 */ 4410 es = (struct ext4_super_block *) (bh->b_data + offset); 4411 sbi->s_es = es; 4412 sb->s_magic = le16_to_cpu(es->s_magic); 4413 if (sb->s_magic != EXT4_SUPER_MAGIC) 4414 goto cantfind_ext4; 4415 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 4416 4417 /* Warn if metadata_csum and gdt_csum are both set. */ 4418 if (ext4_has_feature_metadata_csum(sb) && 4419 ext4_has_feature_gdt_csum(sb)) 4420 ext4_warning(sb, "metadata_csum and uninit_bg are " 4421 "redundant flags; please run fsck."); 4422 4423 /* Check for a known checksum algorithm */ 4424 if (!ext4_verify_csum_type(sb, es)) { 4425 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 4426 "unknown checksum algorithm."); 4427 silent = 1; 4428 goto cantfind_ext4; 4429 } 4430 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE, 4431 ext4_orphan_file_block_trigger); 4432 4433 /* Load the checksum driver */ 4434 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 4435 if (IS_ERR(sbi->s_chksum_driver)) { 4436 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 4437 ret = PTR_ERR(sbi->s_chksum_driver); 4438 sbi->s_chksum_driver = NULL; 4439 goto failed_mount; 4440 } 4441 4442 /* Check superblock checksum */ 4443 if (!ext4_superblock_csum_verify(sb, es)) { 4444 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 4445 "invalid superblock checksum. Run e2fsck?"); 4446 silent = 1; 4447 ret = -EFSBADCRC; 4448 goto cantfind_ext4; 4449 } 4450 4451 /* Precompute checksum seed for all metadata */ 4452 if (ext4_has_feature_csum_seed(sb)) 4453 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 4454 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb)) 4455 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 4456 sizeof(es->s_uuid)); 4457 4458 /* Set defaults before we parse the mount options */ 4459 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 4460 set_opt(sb, INIT_INODE_TABLE); 4461 if (def_mount_opts & EXT4_DEFM_DEBUG) 4462 set_opt(sb, DEBUG); 4463 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 4464 set_opt(sb, GRPID); 4465 if (def_mount_opts & EXT4_DEFM_UID16) 4466 set_opt(sb, NO_UID32); 4467 /* xattr user namespace & acls are now defaulted on */ 4468 set_opt(sb, XATTR_USER); 4469 #ifdef CONFIG_EXT4_FS_POSIX_ACL 4470 set_opt(sb, POSIX_ACL); 4471 #endif 4472 if (ext4_has_feature_fast_commit(sb)) 4473 set_opt2(sb, JOURNAL_FAST_COMMIT); 4474 /* don't forget to enable journal_csum when metadata_csum is enabled. */ 4475 if (ext4_has_metadata_csum(sb)) 4476 set_opt(sb, JOURNAL_CHECKSUM); 4477 4478 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 4479 set_opt(sb, JOURNAL_DATA); 4480 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 4481 set_opt(sb, ORDERED_DATA); 4482 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 4483 set_opt(sb, WRITEBACK_DATA); 4484 4485 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 4486 set_opt(sb, ERRORS_PANIC); 4487 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 4488 set_opt(sb, ERRORS_CONT); 4489 else 4490 set_opt(sb, ERRORS_RO); 4491 /* block_validity enabled by default; disable with noblock_validity */ 4492 set_opt(sb, BLOCK_VALIDITY); 4493 if (def_mount_opts & EXT4_DEFM_DISCARD) 4494 set_opt(sb, DISCARD); 4495 4496 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 4497 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 4498 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 4499 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 4500 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 4501 4502 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 4503 set_opt(sb, BARRIER); 4504 4505 /* 4506 * enable delayed allocation by default 4507 * Use -o nodelalloc to turn it off 4508 */ 4509 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 4510 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 4511 set_opt(sb, DELALLOC); 4512 4513 /* 4514 * set default s_li_wait_mult for lazyinit, for the case there is 4515 * no mount option specified. 4516 */ 4517 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 4518 4519 if (le32_to_cpu(es->s_log_block_size) > 4520 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 4521 ext4_msg(sb, KERN_ERR, 4522 "Invalid log block size: %u", 4523 le32_to_cpu(es->s_log_block_size)); 4524 goto failed_mount; 4525 } 4526 if (le32_to_cpu(es->s_log_cluster_size) > 4527 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 4528 ext4_msg(sb, KERN_ERR, 4529 "Invalid log cluster size: %u", 4530 le32_to_cpu(es->s_log_cluster_size)); 4531 goto failed_mount; 4532 } 4533 4534 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 4535 4536 if (blocksize == PAGE_SIZE) 4537 set_opt(sb, DIOREAD_NOLOCK); 4538 4539 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 4540 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 4541 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 4542 } else { 4543 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 4544 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 4545 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) { 4546 ext4_msg(sb, KERN_ERR, "invalid first ino: %u", 4547 sbi->s_first_ino); 4548 goto failed_mount; 4549 } 4550 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 4551 (!is_power_of_2(sbi->s_inode_size)) || 4552 (sbi->s_inode_size > blocksize)) { 4553 ext4_msg(sb, KERN_ERR, 4554 "unsupported inode size: %d", 4555 sbi->s_inode_size); 4556 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize); 4557 goto failed_mount; 4558 } 4559 /* 4560 * i_atime_extra is the last extra field available for 4561 * [acm]times in struct ext4_inode. Checking for that 4562 * field should suffice to ensure we have extra space 4563 * for all three. 4564 */ 4565 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) + 4566 sizeof(((struct ext4_inode *)0)->i_atime_extra)) { 4567 sb->s_time_gran = 1; 4568 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX; 4569 } else { 4570 sb->s_time_gran = NSEC_PER_SEC; 4571 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX; 4572 } 4573 sb->s_time_min = EXT4_TIMESTAMP_MIN; 4574 } 4575 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 4576 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 4577 EXT4_GOOD_OLD_INODE_SIZE; 4578 if (ext4_has_feature_extra_isize(sb)) { 4579 unsigned v, max = (sbi->s_inode_size - 4580 EXT4_GOOD_OLD_INODE_SIZE); 4581 4582 v = le16_to_cpu(es->s_want_extra_isize); 4583 if (v > max) { 4584 ext4_msg(sb, KERN_ERR, 4585 "bad s_want_extra_isize: %d", v); 4586 goto failed_mount; 4587 } 4588 if (sbi->s_want_extra_isize < v) 4589 sbi->s_want_extra_isize = v; 4590 4591 v = le16_to_cpu(es->s_min_extra_isize); 4592 if (v > max) { 4593 ext4_msg(sb, KERN_ERR, 4594 "bad s_min_extra_isize: %d", v); 4595 goto failed_mount; 4596 } 4597 if (sbi->s_want_extra_isize < v) 4598 sbi->s_want_extra_isize = v; 4599 } 4600 } 4601 4602 err = parse_apply_sb_mount_options(sb, ctx); 4603 if (err < 0) 4604 goto failed_mount; 4605 4606 sbi->s_def_mount_opt = sbi->s_mount_opt; 4607 4608 err = ext4_check_opt_consistency(fc, sb); 4609 if (err < 0) 4610 goto failed_mount; 4611 4612 err = ext4_apply_options(fc, sb); 4613 if (err < 0) 4614 goto failed_mount; 4615 4616 #ifdef CONFIG_UNICODE 4617 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) { 4618 const struct ext4_sb_encodings *encoding_info; 4619 struct unicode_map *encoding; 4620 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags); 4621 4622 encoding_info = ext4_sb_read_encoding(es); 4623 if (!encoding_info) { 4624 ext4_msg(sb, KERN_ERR, 4625 "Encoding requested by superblock is unknown"); 4626 goto failed_mount; 4627 } 4628 4629 encoding = utf8_load(encoding_info->version); 4630 if (IS_ERR(encoding)) { 4631 ext4_msg(sb, KERN_ERR, 4632 "can't mount with superblock charset: %s-%u.%u.%u " 4633 "not supported by the kernel. flags: 0x%x.", 4634 encoding_info->name, 4635 unicode_major(encoding_info->version), 4636 unicode_minor(encoding_info->version), 4637 unicode_rev(encoding_info->version), 4638 encoding_flags); 4639 goto failed_mount; 4640 } 4641 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: " 4642 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name, 4643 unicode_major(encoding_info->version), 4644 unicode_minor(encoding_info->version), 4645 unicode_rev(encoding_info->version), 4646 encoding_flags); 4647 4648 sb->s_encoding = encoding; 4649 sb->s_encoding_flags = encoding_flags; 4650 } 4651 #endif 4652 4653 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 4654 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n"); 4655 /* can't mount with both data=journal and dioread_nolock. */ 4656 clear_opt(sb, DIOREAD_NOLOCK); 4657 clear_opt2(sb, JOURNAL_FAST_COMMIT); 4658 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 4659 ext4_msg(sb, KERN_ERR, "can't mount with " 4660 "both data=journal and delalloc"); 4661 goto failed_mount; 4662 } 4663 if (test_opt(sb, DAX_ALWAYS)) { 4664 ext4_msg(sb, KERN_ERR, "can't mount with " 4665 "both data=journal and dax"); 4666 goto failed_mount; 4667 } 4668 if (ext4_has_feature_encrypt(sb)) { 4669 ext4_msg(sb, KERN_WARNING, 4670 "encrypted files will use data=ordered " 4671 "instead of data journaling mode"); 4672 } 4673 if (test_opt(sb, DELALLOC)) 4674 clear_opt(sb, DELALLOC); 4675 } else { 4676 sb->s_iflags |= SB_I_CGROUPWB; 4677 } 4678 4679 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 4680 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 4681 4682 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 4683 (ext4_has_compat_features(sb) || 4684 ext4_has_ro_compat_features(sb) || 4685 ext4_has_incompat_features(sb))) 4686 ext4_msg(sb, KERN_WARNING, 4687 "feature flags set on rev 0 fs, " 4688 "running e2fsck is recommended"); 4689 4690 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 4691 set_opt2(sb, HURD_COMPAT); 4692 if (ext4_has_feature_64bit(sb)) { 4693 ext4_msg(sb, KERN_ERR, 4694 "The Hurd can't support 64-bit file systems"); 4695 goto failed_mount; 4696 } 4697 4698 /* 4699 * ea_inode feature uses l_i_version field which is not 4700 * available in HURD_COMPAT mode. 4701 */ 4702 if (ext4_has_feature_ea_inode(sb)) { 4703 ext4_msg(sb, KERN_ERR, 4704 "ea_inode feature is not supported for Hurd"); 4705 goto failed_mount; 4706 } 4707 } 4708 4709 if (IS_EXT2_SB(sb)) { 4710 if (ext2_feature_set_ok(sb)) 4711 ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 4712 "using the ext4 subsystem"); 4713 else { 4714 /* 4715 * If we're probing be silent, if this looks like 4716 * it's actually an ext[34] filesystem. 4717 */ 4718 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 4719 goto failed_mount; 4720 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 4721 "to feature incompatibilities"); 4722 goto failed_mount; 4723 } 4724 } 4725 4726 if (IS_EXT3_SB(sb)) { 4727 if (ext3_feature_set_ok(sb)) 4728 ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 4729 "using the ext4 subsystem"); 4730 else { 4731 /* 4732 * If we're probing be silent, if this looks like 4733 * it's actually an ext4 filesystem. 4734 */ 4735 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 4736 goto failed_mount; 4737 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 4738 "to feature incompatibilities"); 4739 goto failed_mount; 4740 } 4741 } 4742 4743 /* 4744 * Check feature flags regardless of the revision level, since we 4745 * previously didn't change the revision level when setting the flags, 4746 * so there is a chance incompat flags are set on a rev 0 filesystem. 4747 */ 4748 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) 4749 goto failed_mount; 4750 4751 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) { 4752 ext4_msg(sb, KERN_ERR, 4753 "Number of reserved GDT blocks insanely large: %d", 4754 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); 4755 goto failed_mount; 4756 } 4757 4758 if (sbi->s_daxdev) { 4759 if (blocksize == PAGE_SIZE) 4760 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags); 4761 else 4762 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n"); 4763 } 4764 4765 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) { 4766 if (ext4_has_feature_inline_data(sb)) { 4767 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem" 4768 " that may contain inline data"); 4769 goto failed_mount; 4770 } 4771 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) { 4772 ext4_msg(sb, KERN_ERR, 4773 "DAX unsupported by block device."); 4774 goto failed_mount; 4775 } 4776 } 4777 4778 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 4779 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 4780 es->s_encryption_level); 4781 goto failed_mount; 4782 } 4783 4784 if (sb->s_blocksize != blocksize) { 4785 /* 4786 * bh must be released before kill_bdev(), otherwise 4787 * it won't be freed and its page also. kill_bdev() 4788 * is called by sb_set_blocksize(). 4789 */ 4790 brelse(bh); 4791 /* Validate the filesystem blocksize */ 4792 if (!sb_set_blocksize(sb, blocksize)) { 4793 ext4_msg(sb, KERN_ERR, "bad block size %d", 4794 blocksize); 4795 bh = NULL; 4796 goto failed_mount; 4797 } 4798 4799 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE; 4800 offset = do_div(logical_sb_block, blocksize); 4801 bh = ext4_sb_bread_unmovable(sb, logical_sb_block); 4802 if (IS_ERR(bh)) { 4803 ext4_msg(sb, KERN_ERR, 4804 "Can't read superblock on 2nd try"); 4805 ret = PTR_ERR(bh); 4806 bh = NULL; 4807 goto failed_mount; 4808 } 4809 es = (struct ext4_super_block *)(bh->b_data + offset); 4810 sbi->s_es = es; 4811 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 4812 ext4_msg(sb, KERN_ERR, 4813 "Magic mismatch, very weird!"); 4814 goto failed_mount; 4815 } 4816 } 4817 4818 has_huge_files = ext4_has_feature_huge_file(sb); 4819 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 4820 has_huge_files); 4821 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 4822 4823 sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 4824 if (ext4_has_feature_64bit(sb)) { 4825 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 4826 sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 4827 !is_power_of_2(sbi->s_desc_size)) { 4828 ext4_msg(sb, KERN_ERR, 4829 "unsupported descriptor size %lu", 4830 sbi->s_desc_size); 4831 goto failed_mount; 4832 } 4833 } else 4834 sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 4835 4836 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 4837 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 4838 4839 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 4840 if (sbi->s_inodes_per_block == 0) 4841 goto cantfind_ext4; 4842 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || 4843 sbi->s_inodes_per_group > blocksize * 8) { 4844 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", 4845 sbi->s_inodes_per_group); 4846 goto failed_mount; 4847 } 4848 sbi->s_itb_per_group = sbi->s_inodes_per_group / 4849 sbi->s_inodes_per_block; 4850 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 4851 sbi->s_sbh = bh; 4852 sbi->s_mount_state = le16_to_cpu(es->s_state); 4853 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 4854 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 4855 4856 for (i = 0; i < 4; i++) 4857 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 4858 sbi->s_def_hash_version = es->s_def_hash_version; 4859 if (ext4_has_feature_dir_index(sb)) { 4860 i = le32_to_cpu(es->s_flags); 4861 if (i & EXT2_FLAGS_UNSIGNED_HASH) 4862 sbi->s_hash_unsigned = 3; 4863 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 4864 #ifdef __CHAR_UNSIGNED__ 4865 if (!sb_rdonly(sb)) 4866 es->s_flags |= 4867 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 4868 sbi->s_hash_unsigned = 3; 4869 #else 4870 if (!sb_rdonly(sb)) 4871 es->s_flags |= 4872 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 4873 #endif 4874 } 4875 } 4876 4877 /* Handle clustersize */ 4878 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 4879 if (ext4_has_feature_bigalloc(sb)) { 4880 if (clustersize < blocksize) { 4881 ext4_msg(sb, KERN_ERR, 4882 "cluster size (%d) smaller than " 4883 "block size (%d)", clustersize, blocksize); 4884 goto failed_mount; 4885 } 4886 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 4887 le32_to_cpu(es->s_log_block_size); 4888 sbi->s_clusters_per_group = 4889 le32_to_cpu(es->s_clusters_per_group); 4890 if (sbi->s_clusters_per_group > blocksize * 8) { 4891 ext4_msg(sb, KERN_ERR, 4892 "#clusters per group too big: %lu", 4893 sbi->s_clusters_per_group); 4894 goto failed_mount; 4895 } 4896 if (sbi->s_blocks_per_group != 4897 (sbi->s_clusters_per_group * (clustersize / blocksize))) { 4898 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 4899 "clusters per group (%lu) inconsistent", 4900 sbi->s_blocks_per_group, 4901 sbi->s_clusters_per_group); 4902 goto failed_mount; 4903 } 4904 } else { 4905 if (clustersize != blocksize) { 4906 ext4_msg(sb, KERN_ERR, 4907 "fragment/cluster size (%d) != " 4908 "block size (%d)", clustersize, blocksize); 4909 goto failed_mount; 4910 } 4911 if (sbi->s_blocks_per_group > blocksize * 8) { 4912 ext4_msg(sb, KERN_ERR, 4913 "#blocks per group too big: %lu", 4914 sbi->s_blocks_per_group); 4915 goto failed_mount; 4916 } 4917 sbi->s_clusters_per_group = sbi->s_blocks_per_group; 4918 sbi->s_cluster_bits = 0; 4919 } 4920 sbi->s_cluster_ratio = clustersize / blocksize; 4921 4922 /* Do we have standard group size of clustersize * 8 blocks ? */ 4923 if (sbi->s_blocks_per_group == clustersize << 3) 4924 set_opt2(sb, STD_GROUP_SIZE); 4925 4926 /* 4927 * Test whether we have more sectors than will fit in sector_t, 4928 * and whether the max offset is addressable by the page cache. 4929 */ 4930 err = generic_check_addressable(sb->s_blocksize_bits, 4931 ext4_blocks_count(es)); 4932 if (err) { 4933 ext4_msg(sb, KERN_ERR, "filesystem" 4934 " too large to mount safely on this system"); 4935 goto failed_mount; 4936 } 4937 4938 if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 4939 goto cantfind_ext4; 4940 4941 /* check blocks count against device size */ 4942 blocks_count = sb_bdev_nr_blocks(sb); 4943 if (blocks_count && ext4_blocks_count(es) > blocks_count) { 4944 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 4945 "exceeds size of device (%llu blocks)", 4946 ext4_blocks_count(es), blocks_count); 4947 goto failed_mount; 4948 } 4949 4950 /* 4951 * It makes no sense for the first data block to be beyond the end 4952 * of the filesystem. 4953 */ 4954 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 4955 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4956 "block %u is beyond end of filesystem (%llu)", 4957 le32_to_cpu(es->s_first_data_block), 4958 ext4_blocks_count(es)); 4959 goto failed_mount; 4960 } 4961 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) && 4962 (sbi->s_cluster_ratio == 1)) { 4963 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4964 "block is 0 with a 1k block and cluster size"); 4965 goto failed_mount; 4966 } 4967 4968 blocks_count = (ext4_blocks_count(es) - 4969 le32_to_cpu(es->s_first_data_block) + 4970 EXT4_BLOCKS_PER_GROUP(sb) - 1); 4971 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 4972 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 4973 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu " 4974 "(block count %llu, first data block %u, " 4975 "blocks per group %lu)", blocks_count, 4976 ext4_blocks_count(es), 4977 le32_to_cpu(es->s_first_data_block), 4978 EXT4_BLOCKS_PER_GROUP(sb)); 4979 goto failed_mount; 4980 } 4981 sbi->s_groups_count = blocks_count; 4982 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 4983 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 4984 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) != 4985 le32_to_cpu(es->s_inodes_count)) { 4986 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu", 4987 le32_to_cpu(es->s_inodes_count), 4988 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group)); 4989 ret = -EINVAL; 4990 goto failed_mount; 4991 } 4992 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 4993 EXT4_DESC_PER_BLOCK(sb); 4994 if (ext4_has_feature_meta_bg(sb)) { 4995 if (le32_to_cpu(es->s_first_meta_bg) > db_count) { 4996 ext4_msg(sb, KERN_WARNING, 4997 "first meta block group too large: %u " 4998 "(group descriptor block count %u)", 4999 le32_to_cpu(es->s_first_meta_bg), db_count); 5000 goto failed_mount; 5001 } 5002 } 5003 rcu_assign_pointer(sbi->s_group_desc, 5004 kvmalloc_array(db_count, 5005 sizeof(struct buffer_head *), 5006 GFP_KERNEL)); 5007 if (sbi->s_group_desc == NULL) { 5008 ext4_msg(sb, KERN_ERR, "not enough memory"); 5009 ret = -ENOMEM; 5010 goto failed_mount; 5011 } 5012 5013 bgl_lock_init(sbi->s_blockgroup_lock); 5014 5015 /* Pre-read the descriptors into the buffer cache */ 5016 for (i = 0; i < db_count; i++) { 5017 block = descriptor_loc(sb, logical_sb_block, i); 5018 ext4_sb_breadahead_unmovable(sb, block); 5019 } 5020 5021 for (i = 0; i < db_count; i++) { 5022 struct buffer_head *bh; 5023 5024 block = descriptor_loc(sb, logical_sb_block, i); 5025 bh = ext4_sb_bread_unmovable(sb, block); 5026 if (IS_ERR(bh)) { 5027 ext4_msg(sb, KERN_ERR, 5028 "can't read group descriptor %d", i); 5029 db_count = i; 5030 ret = PTR_ERR(bh); 5031 goto failed_mount2; 5032 } 5033 rcu_read_lock(); 5034 rcu_dereference(sbi->s_group_desc)[i] = bh; 5035 rcu_read_unlock(); 5036 } 5037 sbi->s_gdb_count = db_count; 5038 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) { 5039 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 5040 ret = -EFSCORRUPTED; 5041 goto failed_mount2; 5042 } 5043 5044 timer_setup(&sbi->s_err_report, print_daily_error_info, 0); 5045 spin_lock_init(&sbi->s_error_lock); 5046 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work); 5047 5048 /* Register extent status tree shrinker */ 5049 if (ext4_es_register_shrinker(sbi)) 5050 goto failed_mount3; 5051 5052 sbi->s_stripe = ext4_get_stripe_size(sbi); 5053 sbi->s_extent_max_zeroout_kb = 32; 5054 5055 /* 5056 * set up enough so that it can read an inode 5057 */ 5058 sb->s_op = &ext4_sops; 5059 sb->s_export_op = &ext4_export_ops; 5060 sb->s_xattr = ext4_xattr_handlers; 5061 #ifdef CONFIG_FS_ENCRYPTION 5062 sb->s_cop = &ext4_cryptops; 5063 #endif 5064 #ifdef CONFIG_FS_VERITY 5065 sb->s_vop = &ext4_verityops; 5066 #endif 5067 #ifdef CONFIG_QUOTA 5068 sb->dq_op = &ext4_quota_operations; 5069 if (ext4_has_feature_quota(sb)) 5070 sb->s_qcop = &dquot_quotactl_sysfile_ops; 5071 else 5072 sb->s_qcop = &ext4_qctl_operations; 5073 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 5074 #endif 5075 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 5076 5077 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 5078 mutex_init(&sbi->s_orphan_lock); 5079 5080 /* Initialize fast commit stuff */ 5081 atomic_set(&sbi->s_fc_subtid, 0); 5082 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]); 5083 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]); 5084 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]); 5085 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]); 5086 sbi->s_fc_bytes = 0; 5087 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE); 5088 ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING); 5089 spin_lock_init(&sbi->s_fc_lock); 5090 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats)); 5091 sbi->s_fc_replay_state.fc_regions = NULL; 5092 sbi->s_fc_replay_state.fc_regions_size = 0; 5093 sbi->s_fc_replay_state.fc_regions_used = 0; 5094 sbi->s_fc_replay_state.fc_regions_valid = 0; 5095 sbi->s_fc_replay_state.fc_modified_inodes = NULL; 5096 sbi->s_fc_replay_state.fc_modified_inodes_size = 0; 5097 sbi->s_fc_replay_state.fc_modified_inodes_used = 0; 5098 5099 sb->s_root = NULL; 5100 5101 needs_recovery = (es->s_last_orphan != 0 || 5102 ext4_has_feature_orphan_present(sb) || 5103 ext4_has_feature_journal_needs_recovery(sb)); 5104 5105 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) 5106 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 5107 goto failed_mount3a; 5108 5109 /* 5110 * The first inode we look at is the journal inode. Don't try 5111 * root first: it may be modified in the journal! 5112 */ 5113 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 5114 err = ext4_load_journal(sb, es, ctx->journal_devnum); 5115 if (err) 5116 goto failed_mount3a; 5117 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && 5118 ext4_has_feature_journal_needs_recovery(sb)) { 5119 ext4_msg(sb, KERN_ERR, "required journal recovery " 5120 "suppressed and not mounted read-only"); 5121 goto failed_mount_wq; 5122 } else { 5123 /* Nojournal mode, all journal mount options are illegal */ 5124 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { 5125 ext4_msg(sb, KERN_ERR, "can't mount with " 5126 "journal_checksum, fs mounted w/o journal"); 5127 goto failed_mount_wq; 5128 } 5129 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 5130 ext4_msg(sb, KERN_ERR, "can't mount with " 5131 "journal_async_commit, fs mounted w/o journal"); 5132 goto failed_mount_wq; 5133 } 5134 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { 5135 ext4_msg(sb, KERN_ERR, "can't mount with " 5136 "commit=%lu, fs mounted w/o journal", 5137 sbi->s_commit_interval / HZ); 5138 goto failed_mount_wq; 5139 } 5140 if (EXT4_MOUNT_DATA_FLAGS & 5141 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 5142 ext4_msg(sb, KERN_ERR, "can't mount with " 5143 "data=, fs mounted w/o journal"); 5144 goto failed_mount_wq; 5145 } 5146 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM; 5147 clear_opt(sb, JOURNAL_CHECKSUM); 5148 clear_opt(sb, DATA_FLAGS); 5149 clear_opt2(sb, JOURNAL_FAST_COMMIT); 5150 sbi->s_journal = NULL; 5151 needs_recovery = 0; 5152 goto no_journal; 5153 } 5154 5155 if (ext4_has_feature_64bit(sb) && 5156 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 5157 JBD2_FEATURE_INCOMPAT_64BIT)) { 5158 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 5159 goto failed_mount_wq; 5160 } 5161 5162 if (!set_journal_csum_feature_set(sb)) { 5163 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 5164 "feature set"); 5165 goto failed_mount_wq; 5166 } 5167 5168 if (test_opt2(sb, JOURNAL_FAST_COMMIT) && 5169 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 5170 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) { 5171 ext4_msg(sb, KERN_ERR, 5172 "Failed to set fast commit journal feature"); 5173 goto failed_mount_wq; 5174 } 5175 5176 /* We have now updated the journal if required, so we can 5177 * validate the data journaling mode. */ 5178 switch (test_opt(sb, DATA_FLAGS)) { 5179 case 0: 5180 /* No mode set, assume a default based on the journal 5181 * capabilities: ORDERED_DATA if the journal can 5182 * cope, else JOURNAL_DATA 5183 */ 5184 if (jbd2_journal_check_available_features 5185 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 5186 set_opt(sb, ORDERED_DATA); 5187 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA; 5188 } else { 5189 set_opt(sb, JOURNAL_DATA); 5190 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; 5191 } 5192 break; 5193 5194 case EXT4_MOUNT_ORDERED_DATA: 5195 case EXT4_MOUNT_WRITEBACK_DATA: 5196 if (!jbd2_journal_check_available_features 5197 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 5198 ext4_msg(sb, KERN_ERR, "Journal does not support " 5199 "requested data journaling mode"); 5200 goto failed_mount_wq; 5201 } 5202 break; 5203 default: 5204 break; 5205 } 5206 5207 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 5208 test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 5209 ext4_msg(sb, KERN_ERR, "can't mount with " 5210 "journal_async_commit in data=ordered mode"); 5211 goto failed_mount_wq; 5212 } 5213 5214 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio); 5215 5216 sbi->s_journal->j_submit_inode_data_buffers = 5217 ext4_journal_submit_inode_data_buffers; 5218 sbi->s_journal->j_finish_inode_data_buffers = 5219 ext4_journal_finish_inode_data_buffers; 5220 5221 no_journal: 5222 if (!test_opt(sb, NO_MBCACHE)) { 5223 sbi->s_ea_block_cache = ext4_xattr_create_cache(); 5224 if (!sbi->s_ea_block_cache) { 5225 ext4_msg(sb, KERN_ERR, 5226 "Failed to create ea_block_cache"); 5227 goto failed_mount_wq; 5228 } 5229 5230 if (ext4_has_feature_ea_inode(sb)) { 5231 sbi->s_ea_inode_cache = ext4_xattr_create_cache(); 5232 if (!sbi->s_ea_inode_cache) { 5233 ext4_msg(sb, KERN_ERR, 5234 "Failed to create ea_inode_cache"); 5235 goto failed_mount_wq; 5236 } 5237 } 5238 } 5239 5240 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) { 5241 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity"); 5242 goto failed_mount_wq; 5243 } 5244 5245 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) && 5246 !ext4_has_feature_encrypt(sb)) { 5247 ext4_set_feature_encrypt(sb); 5248 ext4_commit_super(sb); 5249 } 5250 5251 /* 5252 * Get the # of file system overhead blocks from the 5253 * superblock if present. 5254 */ 5255 if (es->s_overhead_clusters) 5256 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 5257 else { 5258 err = ext4_calculate_overhead(sb); 5259 if (err) 5260 goto failed_mount_wq; 5261 } 5262 5263 /* 5264 * The maximum number of concurrent works can be high and 5265 * concurrency isn't really necessary. Limit it to 1. 5266 */ 5267 EXT4_SB(sb)->rsv_conversion_wq = 5268 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 5269 if (!EXT4_SB(sb)->rsv_conversion_wq) { 5270 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 5271 ret = -ENOMEM; 5272 goto failed_mount4; 5273 } 5274 5275 /* 5276 * The jbd2_journal_load will have done any necessary log recovery, 5277 * so we can safely mount the rest of the filesystem now. 5278 */ 5279 5280 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL); 5281 if (IS_ERR(root)) { 5282 ext4_msg(sb, KERN_ERR, "get root inode failed"); 5283 ret = PTR_ERR(root); 5284 root = NULL; 5285 goto failed_mount4; 5286 } 5287 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 5288 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 5289 iput(root); 5290 goto failed_mount4; 5291 } 5292 5293 sb->s_root = d_make_root(root); 5294 if (!sb->s_root) { 5295 ext4_msg(sb, KERN_ERR, "get root dentry failed"); 5296 ret = -ENOMEM; 5297 goto failed_mount4; 5298 } 5299 5300 ret = ext4_setup_super(sb, es, sb_rdonly(sb)); 5301 if (ret == -EROFS) { 5302 sb->s_flags |= SB_RDONLY; 5303 ret = 0; 5304 } else if (ret) 5305 goto failed_mount4a; 5306 5307 ext4_set_resv_clusters(sb); 5308 5309 if (test_opt(sb, BLOCK_VALIDITY)) { 5310 err = ext4_setup_system_zone(sb); 5311 if (err) { 5312 ext4_msg(sb, KERN_ERR, "failed to initialize system " 5313 "zone (%d)", err); 5314 goto failed_mount4a; 5315 } 5316 } 5317 ext4_fc_replay_cleanup(sb); 5318 5319 ext4_ext_init(sb); 5320 5321 /* 5322 * Enable optimize_scan if number of groups is > threshold. This can be 5323 * turned off by passing "mb_optimize_scan=0". This can also be 5324 * turned on forcefully by passing "mb_optimize_scan=1". 5325 */ 5326 if (ctx->mb_optimize_scan == 1) 5327 set_opt2(sb, MB_OPTIMIZE_SCAN); 5328 else if (ctx->mb_optimize_scan == 0) 5329 clear_opt2(sb, MB_OPTIMIZE_SCAN); 5330 else if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD) 5331 set_opt2(sb, MB_OPTIMIZE_SCAN); 5332 5333 err = ext4_mb_init(sb); 5334 if (err) { 5335 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 5336 err); 5337 goto failed_mount5; 5338 } 5339 5340 /* 5341 * We can only set up the journal commit callback once 5342 * mballoc is initialized 5343 */ 5344 if (sbi->s_journal) 5345 sbi->s_journal->j_commit_callback = 5346 ext4_journal_commit_callback; 5347 5348 block = ext4_count_free_clusters(sb); 5349 ext4_free_blocks_count_set(sbi->s_es, 5350 EXT4_C2B(sbi, block)); 5351 err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 5352 GFP_KERNEL); 5353 if (!err) { 5354 unsigned long freei = ext4_count_free_inodes(sb); 5355 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 5356 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 5357 GFP_KERNEL); 5358 } 5359 /* 5360 * Update the checksum after updating free space/inode 5361 * counters. Otherwise the superblock can have an incorrect 5362 * checksum in the buffer cache until it is written out and 5363 * e2fsprogs programs trying to open a file system immediately 5364 * after it is mounted can fail. 5365 */ 5366 ext4_superblock_csum_set(sb); 5367 if (!err) 5368 err = percpu_counter_init(&sbi->s_dirs_counter, 5369 ext4_count_dirs(sb), GFP_KERNEL); 5370 if (!err) 5371 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 5372 GFP_KERNEL); 5373 if (!err) 5374 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0, 5375 GFP_KERNEL); 5376 if (!err) 5377 err = percpu_init_rwsem(&sbi->s_writepages_rwsem); 5378 5379 if (err) { 5380 ext4_msg(sb, KERN_ERR, "insufficient memory"); 5381 goto failed_mount6; 5382 } 5383 5384 if (ext4_has_feature_flex_bg(sb)) 5385 if (!ext4_fill_flex_info(sb)) { 5386 ext4_msg(sb, KERN_ERR, 5387 "unable to initialize " 5388 "flex_bg meta info!"); 5389 ret = -ENOMEM; 5390 goto failed_mount6; 5391 } 5392 5393 err = ext4_register_li_request(sb, first_not_zeroed); 5394 if (err) 5395 goto failed_mount6; 5396 5397 err = ext4_register_sysfs(sb); 5398 if (err) 5399 goto failed_mount7; 5400 5401 err = ext4_init_orphan_info(sb); 5402 if (err) 5403 goto failed_mount8; 5404 #ifdef CONFIG_QUOTA 5405 /* Enable quota usage during mount. */ 5406 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { 5407 err = ext4_enable_quotas(sb); 5408 if (err) 5409 goto failed_mount9; 5410 } 5411 #endif /* CONFIG_QUOTA */ 5412 5413 /* 5414 * Save the original bdev mapping's wb_err value which could be 5415 * used to detect the metadata async write error. 5416 */ 5417 spin_lock_init(&sbi->s_bdev_wb_lock); 5418 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err, 5419 &sbi->s_bdev_wb_err); 5420 sb->s_bdev->bd_super = sb; 5421 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 5422 ext4_orphan_cleanup(sb, es); 5423 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 5424 if (needs_recovery) { 5425 ext4_msg(sb, KERN_INFO, "recovery complete"); 5426 err = ext4_mark_recovery_complete(sb, es); 5427 if (err) 5428 goto failed_mount9; 5429 } 5430 5431 if (test_opt(sb, DISCARD)) { 5432 struct request_queue *q = bdev_get_queue(sb->s_bdev); 5433 if (!blk_queue_discard(q)) 5434 ext4_msg(sb, KERN_WARNING, 5435 "mounting with \"discard\" option, but " 5436 "the device does not support discard"); 5437 } 5438 5439 if (es->s_error_count) 5440 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 5441 5442 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 5443 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 5444 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 5445 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 5446 atomic_set(&sbi->s_warning_count, 0); 5447 atomic_set(&sbi->s_msg_count, 0); 5448 5449 return 0; 5450 5451 cantfind_ext4: 5452 if (!silent) 5453 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 5454 goto failed_mount; 5455 5456 failed_mount9: 5457 ext4_release_orphan_info(sb); 5458 failed_mount8: 5459 ext4_unregister_sysfs(sb); 5460 kobject_put(&sbi->s_kobj); 5461 failed_mount7: 5462 ext4_unregister_li_request(sb); 5463 failed_mount6: 5464 ext4_mb_release(sb); 5465 rcu_read_lock(); 5466 flex_groups = rcu_dereference(sbi->s_flex_groups); 5467 if (flex_groups) { 5468 for (i = 0; i < sbi->s_flex_groups_allocated; i++) 5469 kvfree(flex_groups[i]); 5470 kvfree(flex_groups); 5471 } 5472 rcu_read_unlock(); 5473 percpu_counter_destroy(&sbi->s_freeclusters_counter); 5474 percpu_counter_destroy(&sbi->s_freeinodes_counter); 5475 percpu_counter_destroy(&sbi->s_dirs_counter); 5476 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 5477 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit); 5478 percpu_free_rwsem(&sbi->s_writepages_rwsem); 5479 failed_mount5: 5480 ext4_ext_release(sb); 5481 ext4_release_system_zone(sb); 5482 failed_mount4a: 5483 dput(sb->s_root); 5484 sb->s_root = NULL; 5485 failed_mount4: 5486 ext4_msg(sb, KERN_ERR, "mount failed"); 5487 if (EXT4_SB(sb)->rsv_conversion_wq) 5488 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 5489 failed_mount_wq: 5490 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 5491 sbi->s_ea_inode_cache = NULL; 5492 5493 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 5494 sbi->s_ea_block_cache = NULL; 5495 5496 if (sbi->s_journal) { 5497 /* flush s_error_work before journal destroy. */ 5498 flush_work(&sbi->s_error_work); 5499 jbd2_journal_destroy(sbi->s_journal); 5500 sbi->s_journal = NULL; 5501 } 5502 failed_mount3a: 5503 ext4_es_unregister_shrinker(sbi); 5504 failed_mount3: 5505 /* flush s_error_work before sbi destroy */ 5506 flush_work(&sbi->s_error_work); 5507 del_timer_sync(&sbi->s_err_report); 5508 ext4_stop_mmpd(sbi); 5509 failed_mount2: 5510 rcu_read_lock(); 5511 group_desc = rcu_dereference(sbi->s_group_desc); 5512 for (i = 0; i < db_count; i++) 5513 brelse(group_desc[i]); 5514 kvfree(group_desc); 5515 rcu_read_unlock(); 5516 failed_mount: 5517 if (sbi->s_chksum_driver) 5518 crypto_free_shash(sbi->s_chksum_driver); 5519 5520 #ifdef CONFIG_UNICODE 5521 utf8_unload(sb->s_encoding); 5522 #endif 5523 5524 #ifdef CONFIG_QUOTA 5525 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5526 kfree(get_qf_name(sb, sbi, i)); 5527 #endif 5528 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy); 5529 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */ 5530 brelse(bh); 5531 ext4_blkdev_remove(sbi); 5532 out_fail: 5533 sb->s_fs_info = NULL; 5534 return err ? err : ret; 5535 } 5536 5537 static int ext4_fill_super(struct super_block *sb, struct fs_context *fc) 5538 { 5539 struct ext4_fs_context *ctx = fc->fs_private; 5540 struct ext4_sb_info *sbi; 5541 const char *descr; 5542 int ret; 5543 5544 sbi = ext4_alloc_sbi(sb); 5545 if (!sbi) 5546 ret = -ENOMEM; 5547 5548 fc->s_fs_info = sbi; 5549 5550 /* Cleanup superblock name */ 5551 strreplace(sb->s_id, '/', '!'); 5552 5553 sbi->s_sb_block = 1; /* Default super block location */ 5554 if (ctx->spec & EXT4_SPEC_s_sb_block) 5555 sbi->s_sb_block = ctx->s_sb_block; 5556 5557 ret = __ext4_fill_super(fc, sb); 5558 if (ret < 0) 5559 goto free_sbi; 5560 5561 if (sbi->s_journal) { 5562 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 5563 descr = " journalled data mode"; 5564 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 5565 descr = " ordered data mode"; 5566 else 5567 descr = " writeback data mode"; 5568 } else 5569 descr = "out journal"; 5570 5571 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 5572 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 5573 "Quota mode: %s.", descr, ext4_quota_mode(sb)); 5574 5575 return 0; 5576 5577 free_sbi: 5578 ext4_free_sbi(sbi); 5579 fc->s_fs_info = NULL; 5580 return ret; 5581 } 5582 5583 static int ext4_get_tree(struct fs_context *fc) 5584 { 5585 return get_tree_bdev(fc, ext4_fill_super); 5586 } 5587 5588 /* 5589 * Setup any per-fs journal parameters now. We'll do this both on 5590 * initial mount, once the journal has been initialised but before we've 5591 * done any recovery; and again on any subsequent remount. 5592 */ 5593 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 5594 { 5595 struct ext4_sb_info *sbi = EXT4_SB(sb); 5596 5597 journal->j_commit_interval = sbi->s_commit_interval; 5598 journal->j_min_batch_time = sbi->s_min_batch_time; 5599 journal->j_max_batch_time = sbi->s_max_batch_time; 5600 ext4_fc_init(sb, journal); 5601 5602 write_lock(&journal->j_state_lock); 5603 if (test_opt(sb, BARRIER)) 5604 journal->j_flags |= JBD2_BARRIER; 5605 else 5606 journal->j_flags &= ~JBD2_BARRIER; 5607 if (test_opt(sb, DATA_ERR_ABORT)) 5608 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 5609 else 5610 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 5611 write_unlock(&journal->j_state_lock); 5612 } 5613 5614 static struct inode *ext4_get_journal_inode(struct super_block *sb, 5615 unsigned int journal_inum) 5616 { 5617 struct inode *journal_inode; 5618 5619 /* 5620 * Test for the existence of a valid inode on disk. Bad things 5621 * happen if we iget() an unused inode, as the subsequent iput() 5622 * will try to delete it. 5623 */ 5624 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL); 5625 if (IS_ERR(journal_inode)) { 5626 ext4_msg(sb, KERN_ERR, "no journal found"); 5627 return NULL; 5628 } 5629 if (!journal_inode->i_nlink) { 5630 make_bad_inode(journal_inode); 5631 iput(journal_inode); 5632 ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 5633 return NULL; 5634 } 5635 5636 jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 5637 journal_inode, journal_inode->i_size); 5638 if (!S_ISREG(journal_inode->i_mode)) { 5639 ext4_msg(sb, KERN_ERR, "invalid journal inode"); 5640 iput(journal_inode); 5641 return NULL; 5642 } 5643 return journal_inode; 5644 } 5645 5646 static journal_t *ext4_get_journal(struct super_block *sb, 5647 unsigned int journal_inum) 5648 { 5649 struct inode *journal_inode; 5650 journal_t *journal; 5651 5652 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 5653 return NULL; 5654 5655 journal_inode = ext4_get_journal_inode(sb, journal_inum); 5656 if (!journal_inode) 5657 return NULL; 5658 5659 journal = jbd2_journal_init_inode(journal_inode); 5660 if (!journal) { 5661 ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 5662 iput(journal_inode); 5663 return NULL; 5664 } 5665 journal->j_private = sb; 5666 ext4_init_journal_params(sb, journal); 5667 return journal; 5668 } 5669 5670 static journal_t *ext4_get_dev_journal(struct super_block *sb, 5671 dev_t j_dev) 5672 { 5673 struct buffer_head *bh; 5674 journal_t *journal; 5675 ext4_fsblk_t start; 5676 ext4_fsblk_t len; 5677 int hblock, blocksize; 5678 ext4_fsblk_t sb_block; 5679 unsigned long offset; 5680 struct ext4_super_block *es; 5681 struct block_device *bdev; 5682 5683 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 5684 return NULL; 5685 5686 bdev = ext4_blkdev_get(j_dev, sb); 5687 if (bdev == NULL) 5688 return NULL; 5689 5690 blocksize = sb->s_blocksize; 5691 hblock = bdev_logical_block_size(bdev); 5692 if (blocksize < hblock) { 5693 ext4_msg(sb, KERN_ERR, 5694 "blocksize too small for journal device"); 5695 goto out_bdev; 5696 } 5697 5698 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 5699 offset = EXT4_MIN_BLOCK_SIZE % blocksize; 5700 set_blocksize(bdev, blocksize); 5701 if (!(bh = __bread(bdev, sb_block, blocksize))) { 5702 ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 5703 "external journal"); 5704 goto out_bdev; 5705 } 5706 5707 es = (struct ext4_super_block *) (bh->b_data + offset); 5708 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 5709 !(le32_to_cpu(es->s_feature_incompat) & 5710 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 5711 ext4_msg(sb, KERN_ERR, "external journal has " 5712 "bad superblock"); 5713 brelse(bh); 5714 goto out_bdev; 5715 } 5716 5717 if ((le32_to_cpu(es->s_feature_ro_compat) & 5718 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 5719 es->s_checksum != ext4_superblock_csum(sb, es)) { 5720 ext4_msg(sb, KERN_ERR, "external journal has " 5721 "corrupt superblock"); 5722 brelse(bh); 5723 goto out_bdev; 5724 } 5725 5726 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 5727 ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 5728 brelse(bh); 5729 goto out_bdev; 5730 } 5731 5732 len = ext4_blocks_count(es); 5733 start = sb_block + 1; 5734 brelse(bh); /* we're done with the superblock */ 5735 5736 journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 5737 start, len, blocksize); 5738 if (!journal) { 5739 ext4_msg(sb, KERN_ERR, "failed to create device journal"); 5740 goto out_bdev; 5741 } 5742 journal->j_private = sb; 5743 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) { 5744 ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 5745 goto out_journal; 5746 } 5747 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 5748 ext4_msg(sb, KERN_ERR, "External journal has more than one " 5749 "user (unsupported) - %d", 5750 be32_to_cpu(journal->j_superblock->s_nr_users)); 5751 goto out_journal; 5752 } 5753 EXT4_SB(sb)->s_journal_bdev = bdev; 5754 ext4_init_journal_params(sb, journal); 5755 return journal; 5756 5757 out_journal: 5758 jbd2_journal_destroy(journal); 5759 out_bdev: 5760 ext4_blkdev_put(bdev); 5761 return NULL; 5762 } 5763 5764 static int ext4_load_journal(struct super_block *sb, 5765 struct ext4_super_block *es, 5766 unsigned long journal_devnum) 5767 { 5768 journal_t *journal; 5769 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 5770 dev_t journal_dev; 5771 int err = 0; 5772 int really_read_only; 5773 int journal_dev_ro; 5774 5775 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 5776 return -EFSCORRUPTED; 5777 5778 if (journal_devnum && 5779 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 5780 ext4_msg(sb, KERN_INFO, "external journal device major/minor " 5781 "numbers have changed"); 5782 journal_dev = new_decode_dev(journal_devnum); 5783 } else 5784 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 5785 5786 if (journal_inum && journal_dev) { 5787 ext4_msg(sb, KERN_ERR, 5788 "filesystem has both journal inode and journal device!"); 5789 return -EINVAL; 5790 } 5791 5792 if (journal_inum) { 5793 journal = ext4_get_journal(sb, journal_inum); 5794 if (!journal) 5795 return -EINVAL; 5796 } else { 5797 journal = ext4_get_dev_journal(sb, journal_dev); 5798 if (!journal) 5799 return -EINVAL; 5800 } 5801 5802 journal_dev_ro = bdev_read_only(journal->j_dev); 5803 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro; 5804 5805 if (journal_dev_ro && !sb_rdonly(sb)) { 5806 ext4_msg(sb, KERN_ERR, 5807 "journal device read-only, try mounting with '-o ro'"); 5808 err = -EROFS; 5809 goto err_out; 5810 } 5811 5812 /* 5813 * Are we loading a blank journal or performing recovery after a 5814 * crash? For recovery, we need to check in advance whether we 5815 * can get read-write access to the device. 5816 */ 5817 if (ext4_has_feature_journal_needs_recovery(sb)) { 5818 if (sb_rdonly(sb)) { 5819 ext4_msg(sb, KERN_INFO, "INFO: recovery " 5820 "required on readonly filesystem"); 5821 if (really_read_only) { 5822 ext4_msg(sb, KERN_ERR, "write access " 5823 "unavailable, cannot proceed " 5824 "(try mounting with noload)"); 5825 err = -EROFS; 5826 goto err_out; 5827 } 5828 ext4_msg(sb, KERN_INFO, "write access will " 5829 "be enabled during recovery"); 5830 } 5831 } 5832 5833 if (!(journal->j_flags & JBD2_BARRIER)) 5834 ext4_msg(sb, KERN_INFO, "barriers disabled"); 5835 5836 if (!ext4_has_feature_journal_needs_recovery(sb)) 5837 err = jbd2_journal_wipe(journal, !really_read_only); 5838 if (!err) { 5839 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 5840 if (save) 5841 memcpy(save, ((char *) es) + 5842 EXT4_S_ERR_START, EXT4_S_ERR_LEN); 5843 err = jbd2_journal_load(journal); 5844 if (save) 5845 memcpy(((char *) es) + EXT4_S_ERR_START, 5846 save, EXT4_S_ERR_LEN); 5847 kfree(save); 5848 } 5849 5850 if (err) { 5851 ext4_msg(sb, KERN_ERR, "error loading journal"); 5852 goto err_out; 5853 } 5854 5855 EXT4_SB(sb)->s_journal = journal; 5856 err = ext4_clear_journal_err(sb, es); 5857 if (err) { 5858 EXT4_SB(sb)->s_journal = NULL; 5859 jbd2_journal_destroy(journal); 5860 return err; 5861 } 5862 5863 if (!really_read_only && journal_devnum && 5864 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 5865 es->s_journal_dev = cpu_to_le32(journal_devnum); 5866 5867 /* Make sure we flush the recovery flag to disk. */ 5868 ext4_commit_super(sb); 5869 } 5870 5871 return 0; 5872 5873 err_out: 5874 jbd2_journal_destroy(journal); 5875 return err; 5876 } 5877 5878 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */ 5879 static void ext4_update_super(struct super_block *sb) 5880 { 5881 struct ext4_sb_info *sbi = EXT4_SB(sb); 5882 struct ext4_super_block *es = sbi->s_es; 5883 struct buffer_head *sbh = sbi->s_sbh; 5884 5885 lock_buffer(sbh); 5886 /* 5887 * If the file system is mounted read-only, don't update the 5888 * superblock write time. This avoids updating the superblock 5889 * write time when we are mounting the root file system 5890 * read/only but we need to replay the journal; at that point, 5891 * for people who are east of GMT and who make their clock 5892 * tick in localtime for Windows bug-for-bug compatibility, 5893 * the clock is set in the future, and this will cause e2fsck 5894 * to complain and force a full file system check. 5895 */ 5896 if (!(sb->s_flags & SB_RDONLY)) 5897 ext4_update_tstamp(es, s_wtime); 5898 es->s_kbytes_written = 5899 cpu_to_le64(sbi->s_kbytes_written + 5900 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) - 5901 sbi->s_sectors_written_start) >> 1)); 5902 if (percpu_counter_initialized(&sbi->s_freeclusters_counter)) 5903 ext4_free_blocks_count_set(es, 5904 EXT4_C2B(sbi, percpu_counter_sum_positive( 5905 &sbi->s_freeclusters_counter))); 5906 if (percpu_counter_initialized(&sbi->s_freeinodes_counter)) 5907 es->s_free_inodes_count = 5908 cpu_to_le32(percpu_counter_sum_positive( 5909 &sbi->s_freeinodes_counter)); 5910 /* Copy error information to the on-disk superblock */ 5911 spin_lock(&sbi->s_error_lock); 5912 if (sbi->s_add_error_count > 0) { 5913 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 5914 if (!es->s_first_error_time && !es->s_first_error_time_hi) { 5915 __ext4_update_tstamp(&es->s_first_error_time, 5916 &es->s_first_error_time_hi, 5917 sbi->s_first_error_time); 5918 strncpy(es->s_first_error_func, sbi->s_first_error_func, 5919 sizeof(es->s_first_error_func)); 5920 es->s_first_error_line = 5921 cpu_to_le32(sbi->s_first_error_line); 5922 es->s_first_error_ino = 5923 cpu_to_le32(sbi->s_first_error_ino); 5924 es->s_first_error_block = 5925 cpu_to_le64(sbi->s_first_error_block); 5926 es->s_first_error_errcode = 5927 ext4_errno_to_code(sbi->s_first_error_code); 5928 } 5929 __ext4_update_tstamp(&es->s_last_error_time, 5930 &es->s_last_error_time_hi, 5931 sbi->s_last_error_time); 5932 strncpy(es->s_last_error_func, sbi->s_last_error_func, 5933 sizeof(es->s_last_error_func)); 5934 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line); 5935 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino); 5936 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block); 5937 es->s_last_error_errcode = 5938 ext4_errno_to_code(sbi->s_last_error_code); 5939 /* 5940 * Start the daily error reporting function if it hasn't been 5941 * started already 5942 */ 5943 if (!es->s_error_count) 5944 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); 5945 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count); 5946 sbi->s_add_error_count = 0; 5947 } 5948 spin_unlock(&sbi->s_error_lock); 5949 5950 ext4_superblock_csum_set(sb); 5951 unlock_buffer(sbh); 5952 } 5953 5954 static int ext4_commit_super(struct super_block *sb) 5955 { 5956 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 5957 int error = 0; 5958 5959 if (!sbh) 5960 return -EINVAL; 5961 if (block_device_ejected(sb)) 5962 return -ENODEV; 5963 5964 ext4_update_super(sb); 5965 5966 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 5967 /* 5968 * Oh, dear. A previous attempt to write the 5969 * superblock failed. This could happen because the 5970 * USB device was yanked out. Or it could happen to 5971 * be a transient write error and maybe the block will 5972 * be remapped. Nothing we can do but to retry the 5973 * write and hope for the best. 5974 */ 5975 ext4_msg(sb, KERN_ERR, "previous I/O error to " 5976 "superblock detected"); 5977 clear_buffer_write_io_error(sbh); 5978 set_buffer_uptodate(sbh); 5979 } 5980 BUFFER_TRACE(sbh, "marking dirty"); 5981 mark_buffer_dirty(sbh); 5982 error = __sync_dirty_buffer(sbh, 5983 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0)); 5984 if (buffer_write_io_error(sbh)) { 5985 ext4_msg(sb, KERN_ERR, "I/O error while writing " 5986 "superblock"); 5987 clear_buffer_write_io_error(sbh); 5988 set_buffer_uptodate(sbh); 5989 } 5990 return error; 5991 } 5992 5993 /* 5994 * Have we just finished recovery? If so, and if we are mounting (or 5995 * remounting) the filesystem readonly, then we will end up with a 5996 * consistent fs on disk. Record that fact. 5997 */ 5998 static int ext4_mark_recovery_complete(struct super_block *sb, 5999 struct ext4_super_block *es) 6000 { 6001 int err; 6002 journal_t *journal = EXT4_SB(sb)->s_journal; 6003 6004 if (!ext4_has_feature_journal(sb)) { 6005 if (journal != NULL) { 6006 ext4_error(sb, "Journal got removed while the fs was " 6007 "mounted!"); 6008 return -EFSCORRUPTED; 6009 } 6010 return 0; 6011 } 6012 jbd2_journal_lock_updates(journal); 6013 err = jbd2_journal_flush(journal, 0); 6014 if (err < 0) 6015 goto out; 6016 6017 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) || 6018 ext4_has_feature_orphan_present(sb))) { 6019 if (!ext4_orphan_file_empty(sb)) { 6020 ext4_error(sb, "Orphan file not empty on read-only fs."); 6021 err = -EFSCORRUPTED; 6022 goto out; 6023 } 6024 ext4_clear_feature_journal_needs_recovery(sb); 6025 ext4_clear_feature_orphan_present(sb); 6026 ext4_commit_super(sb); 6027 } 6028 out: 6029 jbd2_journal_unlock_updates(journal); 6030 return err; 6031 } 6032 6033 /* 6034 * If we are mounting (or read-write remounting) a filesystem whose journal 6035 * has recorded an error from a previous lifetime, move that error to the 6036 * main filesystem now. 6037 */ 6038 static int ext4_clear_journal_err(struct super_block *sb, 6039 struct ext4_super_block *es) 6040 { 6041 journal_t *journal; 6042 int j_errno; 6043 const char *errstr; 6044 6045 if (!ext4_has_feature_journal(sb)) { 6046 ext4_error(sb, "Journal got removed while the fs was mounted!"); 6047 return -EFSCORRUPTED; 6048 } 6049 6050 journal = EXT4_SB(sb)->s_journal; 6051 6052 /* 6053 * Now check for any error status which may have been recorded in the 6054 * journal by a prior ext4_error() or ext4_abort() 6055 */ 6056 6057 j_errno = jbd2_journal_errno(journal); 6058 if (j_errno) { 6059 char nbuf[16]; 6060 6061 errstr = ext4_decode_error(sb, j_errno, nbuf); 6062 ext4_warning(sb, "Filesystem error recorded " 6063 "from previous mount: %s", errstr); 6064 ext4_warning(sb, "Marking fs in need of filesystem check."); 6065 6066 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 6067 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 6068 ext4_commit_super(sb); 6069 6070 jbd2_journal_clear_err(journal); 6071 jbd2_journal_update_sb_errno(journal); 6072 } 6073 return 0; 6074 } 6075 6076 /* 6077 * Force the running and committing transactions to commit, 6078 * and wait on the commit. 6079 */ 6080 int ext4_force_commit(struct super_block *sb) 6081 { 6082 journal_t *journal; 6083 6084 if (sb_rdonly(sb)) 6085 return 0; 6086 6087 journal = EXT4_SB(sb)->s_journal; 6088 return ext4_journal_force_commit(journal); 6089 } 6090 6091 static int ext4_sync_fs(struct super_block *sb, int wait) 6092 { 6093 int ret = 0; 6094 tid_t target; 6095 bool needs_barrier = false; 6096 struct ext4_sb_info *sbi = EXT4_SB(sb); 6097 6098 if (unlikely(ext4_forced_shutdown(sbi))) 6099 return 0; 6100 6101 trace_ext4_sync_fs(sb, wait); 6102 flush_workqueue(sbi->rsv_conversion_wq); 6103 /* 6104 * Writeback quota in non-journalled quota case - journalled quota has 6105 * no dirty dquots 6106 */ 6107 dquot_writeback_dquots(sb, -1); 6108 /* 6109 * Data writeback is possible w/o journal transaction, so barrier must 6110 * being sent at the end of the function. But we can skip it if 6111 * transaction_commit will do it for us. 6112 */ 6113 if (sbi->s_journal) { 6114 target = jbd2_get_latest_transaction(sbi->s_journal); 6115 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 6116 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 6117 needs_barrier = true; 6118 6119 if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 6120 if (wait) 6121 ret = jbd2_log_wait_commit(sbi->s_journal, 6122 target); 6123 } 6124 } else if (wait && test_opt(sb, BARRIER)) 6125 needs_barrier = true; 6126 if (needs_barrier) { 6127 int err; 6128 err = blkdev_issue_flush(sb->s_bdev); 6129 if (!ret) 6130 ret = err; 6131 } 6132 6133 return ret; 6134 } 6135 6136 /* 6137 * LVM calls this function before a (read-only) snapshot is created. This 6138 * gives us a chance to flush the journal completely and mark the fs clean. 6139 * 6140 * Note that only this function cannot bring a filesystem to be in a clean 6141 * state independently. It relies on upper layer to stop all data & metadata 6142 * modifications. 6143 */ 6144 static int ext4_freeze(struct super_block *sb) 6145 { 6146 int error = 0; 6147 journal_t *journal; 6148 6149 if (sb_rdonly(sb)) 6150 return 0; 6151 6152 journal = EXT4_SB(sb)->s_journal; 6153 6154 if (journal) { 6155 /* Now we set up the journal barrier. */ 6156 jbd2_journal_lock_updates(journal); 6157 6158 /* 6159 * Don't clear the needs_recovery flag if we failed to 6160 * flush the journal. 6161 */ 6162 error = jbd2_journal_flush(journal, 0); 6163 if (error < 0) 6164 goto out; 6165 6166 /* Journal blocked and flushed, clear needs_recovery flag. */ 6167 ext4_clear_feature_journal_needs_recovery(sb); 6168 if (ext4_orphan_file_empty(sb)) 6169 ext4_clear_feature_orphan_present(sb); 6170 } 6171 6172 error = ext4_commit_super(sb); 6173 out: 6174 if (journal) 6175 /* we rely on upper layer to stop further updates */ 6176 jbd2_journal_unlock_updates(journal); 6177 return error; 6178 } 6179 6180 /* 6181 * Called by LVM after the snapshot is done. We need to reset the RECOVER 6182 * flag here, even though the filesystem is not technically dirty yet. 6183 */ 6184 static int ext4_unfreeze(struct super_block *sb) 6185 { 6186 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb))) 6187 return 0; 6188 6189 if (EXT4_SB(sb)->s_journal) { 6190 /* Reset the needs_recovery flag before the fs is unlocked. */ 6191 ext4_set_feature_journal_needs_recovery(sb); 6192 if (ext4_has_feature_orphan_file(sb)) 6193 ext4_set_feature_orphan_present(sb); 6194 } 6195 6196 ext4_commit_super(sb); 6197 return 0; 6198 } 6199 6200 /* 6201 * Structure to save mount options for ext4_remount's benefit 6202 */ 6203 struct ext4_mount_options { 6204 unsigned long s_mount_opt; 6205 unsigned long s_mount_opt2; 6206 kuid_t s_resuid; 6207 kgid_t s_resgid; 6208 unsigned long s_commit_interval; 6209 u32 s_min_batch_time, s_max_batch_time; 6210 #ifdef CONFIG_QUOTA 6211 int s_jquota_fmt; 6212 char *s_qf_names[EXT4_MAXQUOTAS]; 6213 #endif 6214 }; 6215 6216 static int __ext4_remount(struct fs_context *fc, struct super_block *sb) 6217 { 6218 struct ext4_fs_context *ctx = fc->fs_private; 6219 struct ext4_super_block *es; 6220 struct ext4_sb_info *sbi = EXT4_SB(sb); 6221 unsigned long old_sb_flags; 6222 struct ext4_mount_options old_opts; 6223 ext4_group_t g; 6224 int err = 0; 6225 #ifdef CONFIG_QUOTA 6226 int enable_quota = 0; 6227 int i, j; 6228 char *to_free[EXT4_MAXQUOTAS]; 6229 #endif 6230 6231 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 6232 6233 /* Store the original options */ 6234 old_sb_flags = sb->s_flags; 6235 old_opts.s_mount_opt = sbi->s_mount_opt; 6236 old_opts.s_mount_opt2 = sbi->s_mount_opt2; 6237 old_opts.s_resuid = sbi->s_resuid; 6238 old_opts.s_resgid = sbi->s_resgid; 6239 old_opts.s_commit_interval = sbi->s_commit_interval; 6240 old_opts.s_min_batch_time = sbi->s_min_batch_time; 6241 old_opts.s_max_batch_time = sbi->s_max_batch_time; 6242 #ifdef CONFIG_QUOTA 6243 old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 6244 for (i = 0; i < EXT4_MAXQUOTAS; i++) 6245 if (sbi->s_qf_names[i]) { 6246 char *qf_name = get_qf_name(sb, sbi, i); 6247 6248 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); 6249 if (!old_opts.s_qf_names[i]) { 6250 for (j = 0; j < i; j++) 6251 kfree(old_opts.s_qf_names[j]); 6252 return -ENOMEM; 6253 } 6254 } else 6255 old_opts.s_qf_names[i] = NULL; 6256 #endif 6257 if (sbi->s_journal && sbi->s_journal->j_task->io_context) 6258 ctx->journal_ioprio = 6259 sbi->s_journal->j_task->io_context->ioprio; 6260 6261 ext4_apply_options(fc, sb); 6262 6263 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 6264 test_opt(sb, JOURNAL_CHECKSUM)) { 6265 ext4_msg(sb, KERN_ERR, "changing journal_checksum " 6266 "during remount not supported; ignoring"); 6267 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 6268 } 6269 6270 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 6271 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 6272 ext4_msg(sb, KERN_ERR, "can't mount with " 6273 "both data=journal and delalloc"); 6274 err = -EINVAL; 6275 goto restore_opts; 6276 } 6277 if (test_opt(sb, DIOREAD_NOLOCK)) { 6278 ext4_msg(sb, KERN_ERR, "can't mount with " 6279 "both data=journal and dioread_nolock"); 6280 err = -EINVAL; 6281 goto restore_opts; 6282 } 6283 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { 6284 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 6285 ext4_msg(sb, KERN_ERR, "can't mount with " 6286 "journal_async_commit in data=ordered mode"); 6287 err = -EINVAL; 6288 goto restore_opts; 6289 } 6290 } 6291 6292 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { 6293 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); 6294 err = -EINVAL; 6295 goto restore_opts; 6296 } 6297 6298 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) 6299 ext4_abort(sb, ESHUTDOWN, "Abort forced by user"); 6300 6301 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 6302 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 6303 6304 es = sbi->s_es; 6305 6306 if (sbi->s_journal) { 6307 ext4_init_journal_params(sb, sbi->s_journal); 6308 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio); 6309 } 6310 6311 /* Flush outstanding errors before changing fs state */ 6312 flush_work(&sbi->s_error_work); 6313 6314 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) { 6315 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) { 6316 err = -EROFS; 6317 goto restore_opts; 6318 } 6319 6320 if (fc->sb_flags & SB_RDONLY) { 6321 err = sync_filesystem(sb); 6322 if (err < 0) 6323 goto restore_opts; 6324 err = dquot_suspend(sb, -1); 6325 if (err < 0) 6326 goto restore_opts; 6327 6328 /* 6329 * First of all, the unconditional stuff we have to do 6330 * to disable replay of the journal when we next remount 6331 */ 6332 sb->s_flags |= SB_RDONLY; 6333 6334 /* 6335 * OK, test if we are remounting a valid rw partition 6336 * readonly, and if so set the rdonly flag and then 6337 * mark the partition as valid again. 6338 */ 6339 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 6340 (sbi->s_mount_state & EXT4_VALID_FS)) 6341 es->s_state = cpu_to_le16(sbi->s_mount_state); 6342 6343 if (sbi->s_journal) { 6344 /* 6345 * We let remount-ro finish even if marking fs 6346 * as clean failed... 6347 */ 6348 ext4_mark_recovery_complete(sb, es); 6349 } 6350 } else { 6351 /* Make sure we can mount this feature set readwrite */ 6352 if (ext4_has_feature_readonly(sb) || 6353 !ext4_feature_set_ok(sb, 0)) { 6354 err = -EROFS; 6355 goto restore_opts; 6356 } 6357 /* 6358 * Make sure the group descriptor checksums 6359 * are sane. If they aren't, refuse to remount r/w. 6360 */ 6361 for (g = 0; g < sbi->s_groups_count; g++) { 6362 struct ext4_group_desc *gdp = 6363 ext4_get_group_desc(sb, g, NULL); 6364 6365 if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 6366 ext4_msg(sb, KERN_ERR, 6367 "ext4_remount: Checksum for group %u failed (%u!=%u)", 6368 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 6369 le16_to_cpu(gdp->bg_checksum)); 6370 err = -EFSBADCRC; 6371 goto restore_opts; 6372 } 6373 } 6374 6375 /* 6376 * If we have an unprocessed orphan list hanging 6377 * around from a previously readonly bdev mount, 6378 * require a full umount/remount for now. 6379 */ 6380 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) { 6381 ext4_msg(sb, KERN_WARNING, "Couldn't " 6382 "remount RDWR because of unprocessed " 6383 "orphan inode list. Please " 6384 "umount/remount instead"); 6385 err = -EINVAL; 6386 goto restore_opts; 6387 } 6388 6389 /* 6390 * Mounting a RDONLY partition read-write, so reread 6391 * and store the current valid flag. (It may have 6392 * been changed by e2fsck since we originally mounted 6393 * the partition.) 6394 */ 6395 if (sbi->s_journal) { 6396 err = ext4_clear_journal_err(sb, es); 6397 if (err) 6398 goto restore_opts; 6399 } 6400 sbi->s_mount_state = le16_to_cpu(es->s_state); 6401 6402 err = ext4_setup_super(sb, es, 0); 6403 if (err) 6404 goto restore_opts; 6405 6406 sb->s_flags &= ~SB_RDONLY; 6407 if (ext4_has_feature_mmp(sb)) 6408 if (ext4_multi_mount_protect(sb, 6409 le64_to_cpu(es->s_mmp_block))) { 6410 err = -EROFS; 6411 goto restore_opts; 6412 } 6413 #ifdef CONFIG_QUOTA 6414 enable_quota = 1; 6415 #endif 6416 } 6417 } 6418 6419 /* 6420 * Reinitialize lazy itable initialization thread based on 6421 * current settings 6422 */ 6423 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) 6424 ext4_unregister_li_request(sb); 6425 else { 6426 ext4_group_t first_not_zeroed; 6427 first_not_zeroed = ext4_has_uninit_itable(sb); 6428 ext4_register_li_request(sb, first_not_zeroed); 6429 } 6430 6431 /* 6432 * Handle creation of system zone data early because it can fail. 6433 * Releasing of existing data is done when we are sure remount will 6434 * succeed. 6435 */ 6436 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) { 6437 err = ext4_setup_system_zone(sb); 6438 if (err) 6439 goto restore_opts; 6440 } 6441 6442 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) { 6443 err = ext4_commit_super(sb); 6444 if (err) 6445 goto restore_opts; 6446 } 6447 6448 #ifdef CONFIG_QUOTA 6449 /* Release old quota file names */ 6450 for (i = 0; i < EXT4_MAXQUOTAS; i++) 6451 kfree(old_opts.s_qf_names[i]); 6452 if (enable_quota) { 6453 if (sb_any_quota_suspended(sb)) 6454 dquot_resume(sb, -1); 6455 else if (ext4_has_feature_quota(sb)) { 6456 err = ext4_enable_quotas(sb); 6457 if (err) 6458 goto restore_opts; 6459 } 6460 } 6461 #endif 6462 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 6463 ext4_release_system_zone(sb); 6464 6465 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb)) 6466 ext4_stop_mmpd(sbi); 6467 6468 return 0; 6469 6470 restore_opts: 6471 sb->s_flags = old_sb_flags; 6472 sbi->s_mount_opt = old_opts.s_mount_opt; 6473 sbi->s_mount_opt2 = old_opts.s_mount_opt2; 6474 sbi->s_resuid = old_opts.s_resuid; 6475 sbi->s_resgid = old_opts.s_resgid; 6476 sbi->s_commit_interval = old_opts.s_commit_interval; 6477 sbi->s_min_batch_time = old_opts.s_min_batch_time; 6478 sbi->s_max_batch_time = old_opts.s_max_batch_time; 6479 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 6480 ext4_release_system_zone(sb); 6481 #ifdef CONFIG_QUOTA 6482 sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 6483 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 6484 to_free[i] = get_qf_name(sb, sbi, i); 6485 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); 6486 } 6487 synchronize_rcu(); 6488 for (i = 0; i < EXT4_MAXQUOTAS; i++) 6489 kfree(to_free[i]); 6490 #endif 6491 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb)) 6492 ext4_stop_mmpd(sbi); 6493 return err; 6494 } 6495 6496 static int ext4_reconfigure(struct fs_context *fc) 6497 { 6498 struct super_block *sb = fc->root->d_sb; 6499 int ret; 6500 6501 fc->s_fs_info = EXT4_SB(sb); 6502 6503 ret = ext4_check_opt_consistency(fc, sb); 6504 if (ret < 0) 6505 return ret; 6506 6507 ret = __ext4_remount(fc, sb); 6508 if (ret < 0) 6509 return ret; 6510 6511 ext4_msg(sb, KERN_INFO, "re-mounted. Quota mode: %s.", 6512 ext4_quota_mode(sb)); 6513 6514 return 0; 6515 } 6516 6517 #ifdef CONFIG_QUOTA 6518 static int ext4_statfs_project(struct super_block *sb, 6519 kprojid_t projid, struct kstatfs *buf) 6520 { 6521 struct kqid qid; 6522 struct dquot *dquot; 6523 u64 limit; 6524 u64 curblock; 6525 6526 qid = make_kqid_projid(projid); 6527 dquot = dqget(sb, qid); 6528 if (IS_ERR(dquot)) 6529 return PTR_ERR(dquot); 6530 spin_lock(&dquot->dq_dqb_lock); 6531 6532 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 6533 dquot->dq_dqb.dqb_bhardlimit); 6534 limit >>= sb->s_blocksize_bits; 6535 6536 if (limit && buf->f_blocks > limit) { 6537 curblock = (dquot->dq_dqb.dqb_curspace + 6538 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 6539 buf->f_blocks = limit; 6540 buf->f_bfree = buf->f_bavail = 6541 (buf->f_blocks > curblock) ? 6542 (buf->f_blocks - curblock) : 0; 6543 } 6544 6545 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 6546 dquot->dq_dqb.dqb_ihardlimit); 6547 if (limit && buf->f_files > limit) { 6548 buf->f_files = limit; 6549 buf->f_ffree = 6550 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 6551 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 6552 } 6553 6554 spin_unlock(&dquot->dq_dqb_lock); 6555 dqput(dquot); 6556 return 0; 6557 } 6558 #endif 6559 6560 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 6561 { 6562 struct super_block *sb = dentry->d_sb; 6563 struct ext4_sb_info *sbi = EXT4_SB(sb); 6564 struct ext4_super_block *es = sbi->s_es; 6565 ext4_fsblk_t overhead = 0, resv_blocks; 6566 s64 bfree; 6567 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 6568 6569 if (!test_opt(sb, MINIX_DF)) 6570 overhead = sbi->s_overhead; 6571 6572 buf->f_type = EXT4_SUPER_MAGIC; 6573 buf->f_bsize = sb->s_blocksize; 6574 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 6575 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 6576 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 6577 /* prevent underflow in case that few free space is available */ 6578 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 6579 buf->f_bavail = buf->f_bfree - 6580 (ext4_r_blocks_count(es) + resv_blocks); 6581 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 6582 buf->f_bavail = 0; 6583 buf->f_files = le32_to_cpu(es->s_inodes_count); 6584 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 6585 buf->f_namelen = EXT4_NAME_LEN; 6586 buf->f_fsid = uuid_to_fsid(es->s_uuid); 6587 6588 #ifdef CONFIG_QUOTA 6589 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 6590 sb_has_quota_limits_enabled(sb, PRJQUOTA)) 6591 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 6592 #endif 6593 return 0; 6594 } 6595 6596 6597 #ifdef CONFIG_QUOTA 6598 6599 /* 6600 * Helper functions so that transaction is started before we acquire dqio_sem 6601 * to keep correct lock ordering of transaction > dqio_sem 6602 */ 6603 static inline struct inode *dquot_to_inode(struct dquot *dquot) 6604 { 6605 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 6606 } 6607 6608 static int ext4_write_dquot(struct dquot *dquot) 6609 { 6610 int ret, err; 6611 handle_t *handle; 6612 struct inode *inode; 6613 6614 inode = dquot_to_inode(dquot); 6615 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 6616 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 6617 if (IS_ERR(handle)) 6618 return PTR_ERR(handle); 6619 ret = dquot_commit(dquot); 6620 err = ext4_journal_stop(handle); 6621 if (!ret) 6622 ret = err; 6623 return ret; 6624 } 6625 6626 static int ext4_acquire_dquot(struct dquot *dquot) 6627 { 6628 int ret, err; 6629 handle_t *handle; 6630 6631 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 6632 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 6633 if (IS_ERR(handle)) 6634 return PTR_ERR(handle); 6635 ret = dquot_acquire(dquot); 6636 err = ext4_journal_stop(handle); 6637 if (!ret) 6638 ret = err; 6639 return ret; 6640 } 6641 6642 static int ext4_release_dquot(struct dquot *dquot) 6643 { 6644 int ret, err; 6645 handle_t *handle; 6646 6647 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 6648 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 6649 if (IS_ERR(handle)) { 6650 /* Release dquot anyway to avoid endless cycle in dqput() */ 6651 dquot_release(dquot); 6652 return PTR_ERR(handle); 6653 } 6654 ret = dquot_release(dquot); 6655 err = ext4_journal_stop(handle); 6656 if (!ret) 6657 ret = err; 6658 return ret; 6659 } 6660 6661 static int ext4_mark_dquot_dirty(struct dquot *dquot) 6662 { 6663 struct super_block *sb = dquot->dq_sb; 6664 6665 if (ext4_is_quota_journalled(sb)) { 6666 dquot_mark_dquot_dirty(dquot); 6667 return ext4_write_dquot(dquot); 6668 } else { 6669 return dquot_mark_dquot_dirty(dquot); 6670 } 6671 } 6672 6673 static int ext4_write_info(struct super_block *sb, int type) 6674 { 6675 int ret, err; 6676 handle_t *handle; 6677 6678 /* Data block + inode block */ 6679 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2); 6680 if (IS_ERR(handle)) 6681 return PTR_ERR(handle); 6682 ret = dquot_commit_info(sb, type); 6683 err = ext4_journal_stop(handle); 6684 if (!ret) 6685 ret = err; 6686 return ret; 6687 } 6688 6689 static void lockdep_set_quota_inode(struct inode *inode, int subclass) 6690 { 6691 struct ext4_inode_info *ei = EXT4_I(inode); 6692 6693 /* The first argument of lockdep_set_subclass has to be 6694 * *exactly* the same as the argument to init_rwsem() --- in 6695 * this case, in init_once() --- or lockdep gets unhappy 6696 * because the name of the lock is set using the 6697 * stringification of the argument to init_rwsem(). 6698 */ 6699 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 6700 lockdep_set_subclass(&ei->i_data_sem, subclass); 6701 } 6702 6703 /* 6704 * Standard function to be called on quota_on 6705 */ 6706 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 6707 const struct path *path) 6708 { 6709 int err; 6710 6711 if (!test_opt(sb, QUOTA)) 6712 return -EINVAL; 6713 6714 /* Quotafile not on the same filesystem? */ 6715 if (path->dentry->d_sb != sb) 6716 return -EXDEV; 6717 6718 /* Quota already enabled for this file? */ 6719 if (IS_NOQUOTA(d_inode(path->dentry))) 6720 return -EBUSY; 6721 6722 /* Journaling quota? */ 6723 if (EXT4_SB(sb)->s_qf_names[type]) { 6724 /* Quotafile not in fs root? */ 6725 if (path->dentry->d_parent != sb->s_root) 6726 ext4_msg(sb, KERN_WARNING, 6727 "Quota file not on filesystem root. " 6728 "Journaled quota will not work"); 6729 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY; 6730 } else { 6731 /* 6732 * Clear the flag just in case mount options changed since 6733 * last time. 6734 */ 6735 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY; 6736 } 6737 6738 /* 6739 * When we journal data on quota file, we have to flush journal to see 6740 * all updates to the file when we bypass pagecache... 6741 */ 6742 if (EXT4_SB(sb)->s_journal && 6743 ext4_should_journal_data(d_inode(path->dentry))) { 6744 /* 6745 * We don't need to lock updates but journal_flush() could 6746 * otherwise be livelocked... 6747 */ 6748 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 6749 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0); 6750 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 6751 if (err) 6752 return err; 6753 } 6754 6755 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 6756 err = dquot_quota_on(sb, type, format_id, path); 6757 if (!err) { 6758 struct inode *inode = d_inode(path->dentry); 6759 handle_t *handle; 6760 6761 /* 6762 * Set inode flags to prevent userspace from messing with quota 6763 * files. If this fails, we return success anyway since quotas 6764 * are already enabled and this is not a hard failure. 6765 */ 6766 inode_lock(inode); 6767 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 6768 if (IS_ERR(handle)) 6769 goto unlock_inode; 6770 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; 6771 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 6772 S_NOATIME | S_IMMUTABLE); 6773 err = ext4_mark_inode_dirty(handle, inode); 6774 ext4_journal_stop(handle); 6775 unlock_inode: 6776 inode_unlock(inode); 6777 if (err) 6778 dquot_quota_off(sb, type); 6779 } 6780 if (err) 6781 lockdep_set_quota_inode(path->dentry->d_inode, 6782 I_DATA_SEM_NORMAL); 6783 return err; 6784 } 6785 6786 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 6787 unsigned int flags) 6788 { 6789 int err; 6790 struct inode *qf_inode; 6791 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 6792 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 6793 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 6794 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 6795 }; 6796 6797 BUG_ON(!ext4_has_feature_quota(sb)); 6798 6799 if (!qf_inums[type]) 6800 return -EPERM; 6801 6802 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL); 6803 if (IS_ERR(qf_inode)) { 6804 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 6805 return PTR_ERR(qf_inode); 6806 } 6807 6808 /* Don't account quota for quota files to avoid recursion */ 6809 qf_inode->i_flags |= S_NOQUOTA; 6810 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 6811 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 6812 if (err) 6813 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 6814 iput(qf_inode); 6815 6816 return err; 6817 } 6818 6819 /* Enable usage tracking for all quota types. */ 6820 int ext4_enable_quotas(struct super_block *sb) 6821 { 6822 int type, err = 0; 6823 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 6824 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 6825 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 6826 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 6827 }; 6828 bool quota_mopt[EXT4_MAXQUOTAS] = { 6829 test_opt(sb, USRQUOTA), 6830 test_opt(sb, GRPQUOTA), 6831 test_opt(sb, PRJQUOTA), 6832 }; 6833 6834 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 6835 for (type = 0; type < EXT4_MAXQUOTAS; type++) { 6836 if (qf_inums[type]) { 6837 err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 6838 DQUOT_USAGE_ENABLED | 6839 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 6840 if (err) { 6841 ext4_warning(sb, 6842 "Failed to enable quota tracking " 6843 "(type=%d, err=%d). Please run " 6844 "e2fsck to fix.", type, err); 6845 for (type--; type >= 0; type--) { 6846 struct inode *inode; 6847 6848 inode = sb_dqopt(sb)->files[type]; 6849 if (inode) 6850 inode = igrab(inode); 6851 dquot_quota_off(sb, type); 6852 if (inode) { 6853 lockdep_set_quota_inode(inode, 6854 I_DATA_SEM_NORMAL); 6855 iput(inode); 6856 } 6857 } 6858 6859 return err; 6860 } 6861 } 6862 } 6863 return 0; 6864 } 6865 6866 static int ext4_quota_off(struct super_block *sb, int type) 6867 { 6868 struct inode *inode = sb_dqopt(sb)->files[type]; 6869 handle_t *handle; 6870 int err; 6871 6872 /* Force all delayed allocation blocks to be allocated. 6873 * Caller already holds s_umount sem */ 6874 if (test_opt(sb, DELALLOC)) 6875 sync_filesystem(sb); 6876 6877 if (!inode || !igrab(inode)) 6878 goto out; 6879 6880 err = dquot_quota_off(sb, type); 6881 if (err || ext4_has_feature_quota(sb)) 6882 goto out_put; 6883 6884 inode_lock(inode); 6885 /* 6886 * Update modification times of quota files when userspace can 6887 * start looking at them. If we fail, we return success anyway since 6888 * this is not a hard failure and quotas are already disabled. 6889 */ 6890 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 6891 if (IS_ERR(handle)) { 6892 err = PTR_ERR(handle); 6893 goto out_unlock; 6894 } 6895 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); 6896 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 6897 inode->i_mtime = inode->i_ctime = current_time(inode); 6898 err = ext4_mark_inode_dirty(handle, inode); 6899 ext4_journal_stop(handle); 6900 out_unlock: 6901 inode_unlock(inode); 6902 out_put: 6903 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); 6904 iput(inode); 6905 return err; 6906 out: 6907 return dquot_quota_off(sb, type); 6908 } 6909 6910 /* Read data from quotafile - avoid pagecache and such because we cannot afford 6911 * acquiring the locks... As quota files are never truncated and quota code 6912 * itself serializes the operations (and no one else should touch the files) 6913 * we don't have to be afraid of races */ 6914 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 6915 size_t len, loff_t off) 6916 { 6917 struct inode *inode = sb_dqopt(sb)->files[type]; 6918 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 6919 int offset = off & (sb->s_blocksize - 1); 6920 int tocopy; 6921 size_t toread; 6922 struct buffer_head *bh; 6923 loff_t i_size = i_size_read(inode); 6924 6925 if (off > i_size) 6926 return 0; 6927 if (off+len > i_size) 6928 len = i_size-off; 6929 toread = len; 6930 while (toread > 0) { 6931 tocopy = sb->s_blocksize - offset < toread ? 6932 sb->s_blocksize - offset : toread; 6933 bh = ext4_bread(NULL, inode, blk, 0); 6934 if (IS_ERR(bh)) 6935 return PTR_ERR(bh); 6936 if (!bh) /* A hole? */ 6937 memset(data, 0, tocopy); 6938 else 6939 memcpy(data, bh->b_data+offset, tocopy); 6940 brelse(bh); 6941 offset = 0; 6942 toread -= tocopy; 6943 data += tocopy; 6944 blk++; 6945 } 6946 return len; 6947 } 6948 6949 /* Write to quotafile (we know the transaction is already started and has 6950 * enough credits) */ 6951 static ssize_t ext4_quota_write(struct super_block *sb, int type, 6952 const char *data, size_t len, loff_t off) 6953 { 6954 struct inode *inode = sb_dqopt(sb)->files[type]; 6955 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 6956 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1); 6957 int retries = 0; 6958 struct buffer_head *bh; 6959 handle_t *handle = journal_current_handle(); 6960 6961 if (!handle) { 6962 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 6963 " cancelled because transaction is not started", 6964 (unsigned long long)off, (unsigned long long)len); 6965 return -EIO; 6966 } 6967 /* 6968 * Since we account only one data block in transaction credits, 6969 * then it is impossible to cross a block boundary. 6970 */ 6971 if (sb->s_blocksize - offset < len) { 6972 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 6973 " cancelled because not block aligned", 6974 (unsigned long long)off, (unsigned long long)len); 6975 return -EIO; 6976 } 6977 6978 do { 6979 bh = ext4_bread(handle, inode, blk, 6980 EXT4_GET_BLOCKS_CREATE | 6981 EXT4_GET_BLOCKS_METADATA_NOFAIL); 6982 } while (PTR_ERR(bh) == -ENOSPC && 6983 ext4_should_retry_alloc(inode->i_sb, &retries)); 6984 if (IS_ERR(bh)) 6985 return PTR_ERR(bh); 6986 if (!bh) 6987 goto out; 6988 BUFFER_TRACE(bh, "get write access"); 6989 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE); 6990 if (err) { 6991 brelse(bh); 6992 return err; 6993 } 6994 lock_buffer(bh); 6995 memcpy(bh->b_data+offset, data, len); 6996 flush_dcache_page(bh->b_page); 6997 unlock_buffer(bh); 6998 err = ext4_handle_dirty_metadata(handle, NULL, bh); 6999 brelse(bh); 7000 out: 7001 if (inode->i_size < off + len) { 7002 i_size_write(inode, off + len); 7003 EXT4_I(inode)->i_disksize = inode->i_size; 7004 err2 = ext4_mark_inode_dirty(handle, inode); 7005 if (unlikely(err2 && !err)) 7006 err = err2; 7007 } 7008 return err ? err : len; 7009 } 7010 #endif 7011 7012 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 7013 static inline void register_as_ext2(void) 7014 { 7015 int err = register_filesystem(&ext2_fs_type); 7016 if (err) 7017 printk(KERN_WARNING 7018 "EXT4-fs: Unable to register as ext2 (%d)\n", err); 7019 } 7020 7021 static inline void unregister_as_ext2(void) 7022 { 7023 unregister_filesystem(&ext2_fs_type); 7024 } 7025 7026 static inline int ext2_feature_set_ok(struct super_block *sb) 7027 { 7028 if (ext4_has_unknown_ext2_incompat_features(sb)) 7029 return 0; 7030 if (sb_rdonly(sb)) 7031 return 1; 7032 if (ext4_has_unknown_ext2_ro_compat_features(sb)) 7033 return 0; 7034 return 1; 7035 } 7036 #else 7037 static inline void register_as_ext2(void) { } 7038 static inline void unregister_as_ext2(void) { } 7039 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 7040 #endif 7041 7042 static inline void register_as_ext3(void) 7043 { 7044 int err = register_filesystem(&ext3_fs_type); 7045 if (err) 7046 printk(KERN_WARNING 7047 "EXT4-fs: Unable to register as ext3 (%d)\n", err); 7048 } 7049 7050 static inline void unregister_as_ext3(void) 7051 { 7052 unregister_filesystem(&ext3_fs_type); 7053 } 7054 7055 static inline int ext3_feature_set_ok(struct super_block *sb) 7056 { 7057 if (ext4_has_unknown_ext3_incompat_features(sb)) 7058 return 0; 7059 if (!ext4_has_feature_journal(sb)) 7060 return 0; 7061 if (sb_rdonly(sb)) 7062 return 1; 7063 if (ext4_has_unknown_ext3_ro_compat_features(sb)) 7064 return 0; 7065 return 1; 7066 } 7067 7068 static struct file_system_type ext4_fs_type = { 7069 .owner = THIS_MODULE, 7070 .name = "ext4", 7071 .init_fs_context = ext4_init_fs_context, 7072 .parameters = ext4_param_specs, 7073 .kill_sb = kill_block_super, 7074 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP, 7075 }; 7076 MODULE_ALIAS_FS("ext4"); 7077 7078 /* Shared across all ext4 file systems */ 7079 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 7080 7081 static int __init ext4_init_fs(void) 7082 { 7083 int i, err; 7084 7085 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 7086 ext4_li_info = NULL; 7087 7088 /* Build-time check for flags consistency */ 7089 ext4_check_flag_values(); 7090 7091 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) 7092 init_waitqueue_head(&ext4__ioend_wq[i]); 7093 7094 err = ext4_init_es(); 7095 if (err) 7096 return err; 7097 7098 err = ext4_init_pending(); 7099 if (err) 7100 goto out7; 7101 7102 err = ext4_init_post_read_processing(); 7103 if (err) 7104 goto out6; 7105 7106 err = ext4_init_pageio(); 7107 if (err) 7108 goto out5; 7109 7110 err = ext4_init_system_zone(); 7111 if (err) 7112 goto out4; 7113 7114 err = ext4_init_sysfs(); 7115 if (err) 7116 goto out3; 7117 7118 err = ext4_init_mballoc(); 7119 if (err) 7120 goto out2; 7121 err = init_inodecache(); 7122 if (err) 7123 goto out1; 7124 7125 err = ext4_fc_init_dentry_cache(); 7126 if (err) 7127 goto out05; 7128 7129 register_as_ext3(); 7130 register_as_ext2(); 7131 err = register_filesystem(&ext4_fs_type); 7132 if (err) 7133 goto out; 7134 7135 return 0; 7136 out: 7137 unregister_as_ext2(); 7138 unregister_as_ext3(); 7139 ext4_fc_destroy_dentry_cache(); 7140 out05: 7141 destroy_inodecache(); 7142 out1: 7143 ext4_exit_mballoc(); 7144 out2: 7145 ext4_exit_sysfs(); 7146 out3: 7147 ext4_exit_system_zone(); 7148 out4: 7149 ext4_exit_pageio(); 7150 out5: 7151 ext4_exit_post_read_processing(); 7152 out6: 7153 ext4_exit_pending(); 7154 out7: 7155 ext4_exit_es(); 7156 7157 return err; 7158 } 7159 7160 static void __exit ext4_exit_fs(void) 7161 { 7162 ext4_destroy_lazyinit_thread(); 7163 unregister_as_ext2(); 7164 unregister_as_ext3(); 7165 unregister_filesystem(&ext4_fs_type); 7166 ext4_fc_destroy_dentry_cache(); 7167 destroy_inodecache(); 7168 ext4_exit_mballoc(); 7169 ext4_exit_sysfs(); 7170 ext4_exit_system_zone(); 7171 ext4_exit_pageio(); 7172 ext4_exit_post_read_processing(); 7173 ext4_exit_es(); 7174 ext4_exit_pending(); 7175 } 7176 7177 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 7178 MODULE_DESCRIPTION("Fourth Extended Filesystem"); 7179 MODULE_LICENSE("GPL"); 7180 MODULE_SOFTDEP("pre: crc32c"); 7181 module_init(ext4_init_fs) 7182 module_exit(ext4_exit_fs) 7183