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/uaccess.h> 43 #include <linux/iversion.h> 44 #include <linux/unicode.h> 45 #include <linux/part_stat.h> 46 #include <linux/kthread.h> 47 #include <linux/freezer.h> 48 #include <linux/fsnotify.h> 49 #include <linux/fs_context.h> 50 #include <linux/fs_parser.h> 51 52 #include "ext4.h" 53 #include "ext4_extents.h" /* Needed for trace points definition */ 54 #include "ext4_jbd2.h" 55 #include "xattr.h" 56 #include "acl.h" 57 #include "mballoc.h" 58 #include "fsmap.h" 59 60 #define CREATE_TRACE_POINTS 61 #include <trace/events/ext4.h> 62 63 static struct ext4_lazy_init *ext4_li_info; 64 static DEFINE_MUTEX(ext4_li_mtx); 65 static struct ratelimit_state ext4_mount_msg_ratelimit; 66 67 static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 68 unsigned long journal_devnum); 69 static int ext4_show_options(struct seq_file *seq, struct dentry *root); 70 static void ext4_update_super(struct super_block *sb); 71 static int ext4_commit_super(struct super_block *sb); 72 static int ext4_mark_recovery_complete(struct super_block *sb, 73 struct ext4_super_block *es); 74 static int ext4_clear_journal_err(struct super_block *sb, 75 struct ext4_super_block *es); 76 static int ext4_sync_fs(struct super_block *sb, int wait); 77 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 78 static int ext4_unfreeze(struct super_block *sb); 79 static int ext4_freeze(struct super_block *sb); 80 static inline int ext2_feature_set_ok(struct super_block *sb); 81 static inline int ext3_feature_set_ok(struct super_block *sb); 82 static void ext4_destroy_lazyinit_thread(void); 83 static void ext4_unregister_li_request(struct super_block *sb); 84 static void ext4_clear_request_list(void); 85 static struct inode *ext4_get_journal_inode(struct super_block *sb, 86 unsigned int journal_inum); 87 static int ext4_validate_options(struct fs_context *fc); 88 static int ext4_check_opt_consistency(struct fs_context *fc, 89 struct super_block *sb); 90 static void ext4_apply_options(struct fs_context *fc, struct super_block *sb); 91 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param); 92 static int ext4_get_tree(struct fs_context *fc); 93 static int ext4_reconfigure(struct fs_context *fc); 94 static void ext4_fc_free(struct fs_context *fc); 95 static int ext4_init_fs_context(struct fs_context *fc); 96 static const struct fs_parameter_spec ext4_param_specs[]; 97 98 /* 99 * Lock ordering 100 * 101 * page fault path: 102 * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start 103 * -> page lock -> i_data_sem (rw) 104 * 105 * buffered write path: 106 * sb_start_write -> i_mutex -> mmap_lock 107 * sb_start_write -> i_mutex -> transaction start -> page lock -> 108 * i_data_sem (rw) 109 * 110 * truncate: 111 * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) -> 112 * page lock 113 * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start -> 114 * i_data_sem (rw) 115 * 116 * direct IO: 117 * sb_start_write -> i_mutex -> mmap_lock 118 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw) 119 * 120 * writepages: 121 * transaction start -> page lock(s) -> i_data_sem (rw) 122 */ 123 124 static const struct fs_context_operations ext4_context_ops = { 125 .parse_param = ext4_parse_param, 126 .get_tree = ext4_get_tree, 127 .reconfigure = ext4_reconfigure, 128 .free = ext4_fc_free, 129 }; 130 131 132 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 133 static struct file_system_type ext2_fs_type = { 134 .owner = THIS_MODULE, 135 .name = "ext2", 136 .init_fs_context = ext4_init_fs_context, 137 .parameters = ext4_param_specs, 138 .kill_sb = kill_block_super, 139 .fs_flags = FS_REQUIRES_DEV, 140 }; 141 MODULE_ALIAS_FS("ext2"); 142 MODULE_ALIAS("ext2"); 143 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) 144 #else 145 #define IS_EXT2_SB(sb) (0) 146 #endif 147 148 149 static struct file_system_type ext3_fs_type = { 150 .owner = THIS_MODULE, 151 .name = "ext3", 152 .init_fs_context = ext4_init_fs_context, 153 .parameters = ext4_param_specs, 154 .kill_sb = kill_block_super, 155 .fs_flags = FS_REQUIRES_DEV, 156 }; 157 MODULE_ALIAS_FS("ext3"); 158 MODULE_ALIAS("ext3"); 159 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) 160 161 162 static inline void __ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags, 163 bh_end_io_t *end_io) 164 { 165 /* 166 * buffer's verified bit is no longer valid after reading from 167 * disk again due to write out error, clear it to make sure we 168 * recheck the buffer contents. 169 */ 170 clear_buffer_verified(bh); 171 172 bh->b_end_io = end_io ? end_io : end_buffer_read_sync; 173 get_bh(bh); 174 submit_bh(REQ_OP_READ | op_flags, bh); 175 } 176 177 void ext4_read_bh_nowait(struct buffer_head *bh, blk_opf_t op_flags, 178 bh_end_io_t *end_io) 179 { 180 BUG_ON(!buffer_locked(bh)); 181 182 if (ext4_buffer_uptodate(bh)) { 183 unlock_buffer(bh); 184 return; 185 } 186 __ext4_read_bh(bh, op_flags, end_io); 187 } 188 189 int ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags, bh_end_io_t *end_io) 190 { 191 BUG_ON(!buffer_locked(bh)); 192 193 if (ext4_buffer_uptodate(bh)) { 194 unlock_buffer(bh); 195 return 0; 196 } 197 198 __ext4_read_bh(bh, op_flags, end_io); 199 200 wait_on_buffer(bh); 201 if (buffer_uptodate(bh)) 202 return 0; 203 return -EIO; 204 } 205 206 int ext4_read_bh_lock(struct buffer_head *bh, blk_opf_t op_flags, bool wait) 207 { 208 lock_buffer(bh); 209 if (!wait) { 210 ext4_read_bh_nowait(bh, op_flags, NULL); 211 return 0; 212 } 213 return ext4_read_bh(bh, op_flags, NULL); 214 } 215 216 /* 217 * This works like __bread_gfp() except it uses ERR_PTR for error 218 * returns. Currently with sb_bread it's impossible to distinguish 219 * between ENOMEM and EIO situations (since both result in a NULL 220 * return. 221 */ 222 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb, 223 sector_t block, 224 blk_opf_t op_flags, gfp_t gfp) 225 { 226 struct buffer_head *bh; 227 int ret; 228 229 bh = sb_getblk_gfp(sb, block, gfp); 230 if (bh == NULL) 231 return ERR_PTR(-ENOMEM); 232 if (ext4_buffer_uptodate(bh)) 233 return bh; 234 235 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true); 236 if (ret) { 237 put_bh(bh); 238 return ERR_PTR(ret); 239 } 240 return bh; 241 } 242 243 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block, 244 blk_opf_t op_flags) 245 { 246 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE); 247 } 248 249 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb, 250 sector_t block) 251 { 252 return __ext4_sb_bread_gfp(sb, block, 0, 0); 253 } 254 255 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block) 256 { 257 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0); 258 259 if (likely(bh)) { 260 if (trylock_buffer(bh)) 261 ext4_read_bh_nowait(bh, REQ_RAHEAD, NULL); 262 brelse(bh); 263 } 264 } 265 266 static int ext4_verify_csum_type(struct super_block *sb, 267 struct ext4_super_block *es) 268 { 269 if (!ext4_has_feature_metadata_csum(sb)) 270 return 1; 271 272 return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 273 } 274 275 __le32 ext4_superblock_csum(struct super_block *sb, 276 struct ext4_super_block *es) 277 { 278 struct ext4_sb_info *sbi = EXT4_SB(sb); 279 int offset = offsetof(struct ext4_super_block, s_checksum); 280 __u32 csum; 281 282 csum = ext4_chksum(sbi, ~0, (char *)es, offset); 283 284 return cpu_to_le32(csum); 285 } 286 287 static int ext4_superblock_csum_verify(struct super_block *sb, 288 struct ext4_super_block *es) 289 { 290 if (!ext4_has_metadata_csum(sb)) 291 return 1; 292 293 return es->s_checksum == ext4_superblock_csum(sb, es); 294 } 295 296 void ext4_superblock_csum_set(struct super_block *sb) 297 { 298 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 299 300 if (!ext4_has_metadata_csum(sb)) 301 return; 302 303 es->s_checksum = ext4_superblock_csum(sb, es); 304 } 305 306 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 307 struct ext4_group_desc *bg) 308 { 309 return le32_to_cpu(bg->bg_block_bitmap_lo) | 310 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 311 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 312 } 313 314 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 315 struct ext4_group_desc *bg) 316 { 317 return le32_to_cpu(bg->bg_inode_bitmap_lo) | 318 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 319 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 320 } 321 322 ext4_fsblk_t ext4_inode_table(struct super_block *sb, 323 struct ext4_group_desc *bg) 324 { 325 return le32_to_cpu(bg->bg_inode_table_lo) | 326 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 327 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 328 } 329 330 __u32 ext4_free_group_clusters(struct super_block *sb, 331 struct ext4_group_desc *bg) 332 { 333 return le16_to_cpu(bg->bg_free_blocks_count_lo) | 334 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 335 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 336 } 337 338 __u32 ext4_free_inodes_count(struct super_block *sb, 339 struct ext4_group_desc *bg) 340 { 341 return le16_to_cpu(bg->bg_free_inodes_count_lo) | 342 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 343 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); 344 } 345 346 __u32 ext4_used_dirs_count(struct super_block *sb, 347 struct ext4_group_desc *bg) 348 { 349 return le16_to_cpu(bg->bg_used_dirs_count_lo) | 350 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 351 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 352 } 353 354 __u32 ext4_itable_unused_count(struct super_block *sb, 355 struct ext4_group_desc *bg) 356 { 357 return le16_to_cpu(bg->bg_itable_unused_lo) | 358 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 359 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 360 } 361 362 void ext4_block_bitmap_set(struct super_block *sb, 363 struct ext4_group_desc *bg, ext4_fsblk_t blk) 364 { 365 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 366 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 367 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 368 } 369 370 void ext4_inode_bitmap_set(struct super_block *sb, 371 struct ext4_group_desc *bg, ext4_fsblk_t blk) 372 { 373 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 374 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 375 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 376 } 377 378 void ext4_inode_table_set(struct super_block *sb, 379 struct ext4_group_desc *bg, ext4_fsblk_t blk) 380 { 381 bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 382 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 383 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 384 } 385 386 void ext4_free_group_clusters_set(struct super_block *sb, 387 struct ext4_group_desc *bg, __u32 count) 388 { 389 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 390 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 391 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 392 } 393 394 void ext4_free_inodes_set(struct super_block *sb, 395 struct ext4_group_desc *bg, __u32 count) 396 { 397 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); 398 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 399 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); 400 } 401 402 void ext4_used_dirs_set(struct super_block *sb, 403 struct ext4_group_desc *bg, __u32 count) 404 { 405 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 406 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 407 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 408 } 409 410 void ext4_itable_unused_set(struct super_block *sb, 411 struct ext4_group_desc *bg, __u32 count) 412 { 413 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 414 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 415 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 416 } 417 418 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now) 419 { 420 now = clamp_val(now, 0, (1ull << 40) - 1); 421 422 *lo = cpu_to_le32(lower_32_bits(now)); 423 *hi = upper_32_bits(now); 424 } 425 426 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi) 427 { 428 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo); 429 } 430 #define ext4_update_tstamp(es, tstamp) \ 431 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \ 432 ktime_get_real_seconds()) 433 #define ext4_get_tstamp(es, tstamp) \ 434 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 435 436 /* 437 * The del_gendisk() function uninitializes the disk-specific data 438 * structures, including the bdi structure, without telling anyone 439 * else. Once this happens, any attempt to call mark_buffer_dirty() 440 * (for example, by ext4_commit_super), will cause a kernel OOPS. 441 * This is a kludge to prevent these oops until we can put in a proper 442 * hook in del_gendisk() to inform the VFS and file system layers. 443 */ 444 static int block_device_ejected(struct super_block *sb) 445 { 446 struct inode *bd_inode = sb->s_bdev->bd_inode; 447 struct backing_dev_info *bdi = inode_to_bdi(bd_inode); 448 449 return bdi->dev == NULL; 450 } 451 452 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 453 { 454 struct super_block *sb = journal->j_private; 455 struct ext4_sb_info *sbi = EXT4_SB(sb); 456 int error = is_journal_aborted(journal); 457 struct ext4_journal_cb_entry *jce; 458 459 BUG_ON(txn->t_state == T_FINISHED); 460 461 ext4_process_freed_data(sb, txn->t_tid); 462 463 spin_lock(&sbi->s_md_lock); 464 while (!list_empty(&txn->t_private_list)) { 465 jce = list_entry(txn->t_private_list.next, 466 struct ext4_journal_cb_entry, jce_list); 467 list_del_init(&jce->jce_list); 468 spin_unlock(&sbi->s_md_lock); 469 jce->jce_func(sb, jce, error); 470 spin_lock(&sbi->s_md_lock); 471 } 472 spin_unlock(&sbi->s_md_lock); 473 } 474 475 /* 476 * This writepage callback for write_cache_pages() 477 * takes care of a few cases after page cleaning. 478 * 479 * write_cache_pages() already checks for dirty pages 480 * and calls clear_page_dirty_for_io(), which we want, 481 * to write protect the pages. 482 * 483 * However, we may have to redirty a page (see below.) 484 */ 485 static int ext4_journalled_writepage_callback(struct folio *folio, 486 struct writeback_control *wbc, 487 void *data) 488 { 489 transaction_t *transaction = (transaction_t *) data; 490 struct buffer_head *bh, *head; 491 struct journal_head *jh; 492 493 bh = head = folio_buffers(folio); 494 do { 495 /* 496 * We have to redirty a page in these cases: 497 * 1) If buffer is dirty, it means the page was dirty because it 498 * contains a buffer that needs checkpointing. So the dirty bit 499 * needs to be preserved so that checkpointing writes the buffer 500 * properly. 501 * 2) If buffer is not part of the committing transaction 502 * (we may have just accidentally come across this buffer because 503 * inode range tracking is not exact) or if the currently running 504 * transaction already contains this buffer as well, dirty bit 505 * needs to be preserved so that the buffer gets writeprotected 506 * properly on running transaction's commit. 507 */ 508 jh = bh2jh(bh); 509 if (buffer_dirty(bh) || 510 (jh && (jh->b_transaction != transaction || 511 jh->b_next_transaction))) { 512 folio_redirty_for_writepage(wbc, folio); 513 goto out; 514 } 515 } while ((bh = bh->b_this_page) != head); 516 517 out: 518 return AOP_WRITEPAGE_ACTIVATE; 519 } 520 521 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode) 522 { 523 struct address_space *mapping = jinode->i_vfs_inode->i_mapping; 524 struct writeback_control wbc = { 525 .sync_mode = WB_SYNC_ALL, 526 .nr_to_write = LONG_MAX, 527 .range_start = jinode->i_dirty_start, 528 .range_end = jinode->i_dirty_end, 529 }; 530 531 return write_cache_pages(mapping, &wbc, 532 ext4_journalled_writepage_callback, 533 jinode->i_transaction); 534 } 535 536 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode) 537 { 538 int ret; 539 540 if (ext4_should_journal_data(jinode->i_vfs_inode)) 541 ret = ext4_journalled_submit_inode_data_buffers(jinode); 542 else 543 ret = ext4_normal_submit_inode_data_buffers(jinode); 544 return ret; 545 } 546 547 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode) 548 { 549 int ret = 0; 550 551 if (!ext4_should_journal_data(jinode->i_vfs_inode)) 552 ret = jbd2_journal_finish_inode_data_buffers(jinode); 553 554 return ret; 555 } 556 557 static bool system_going_down(void) 558 { 559 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 560 || system_state == SYSTEM_RESTART; 561 } 562 563 struct ext4_err_translation { 564 int code; 565 int errno; 566 }; 567 568 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err } 569 570 static struct ext4_err_translation err_translation[] = { 571 EXT4_ERR_TRANSLATE(EIO), 572 EXT4_ERR_TRANSLATE(ENOMEM), 573 EXT4_ERR_TRANSLATE(EFSBADCRC), 574 EXT4_ERR_TRANSLATE(EFSCORRUPTED), 575 EXT4_ERR_TRANSLATE(ENOSPC), 576 EXT4_ERR_TRANSLATE(ENOKEY), 577 EXT4_ERR_TRANSLATE(EROFS), 578 EXT4_ERR_TRANSLATE(EFBIG), 579 EXT4_ERR_TRANSLATE(EEXIST), 580 EXT4_ERR_TRANSLATE(ERANGE), 581 EXT4_ERR_TRANSLATE(EOVERFLOW), 582 EXT4_ERR_TRANSLATE(EBUSY), 583 EXT4_ERR_TRANSLATE(ENOTDIR), 584 EXT4_ERR_TRANSLATE(ENOTEMPTY), 585 EXT4_ERR_TRANSLATE(ESHUTDOWN), 586 EXT4_ERR_TRANSLATE(EFAULT), 587 }; 588 589 static int ext4_errno_to_code(int errno) 590 { 591 int i; 592 593 for (i = 0; i < ARRAY_SIZE(err_translation); i++) 594 if (err_translation[i].errno == errno) 595 return err_translation[i].code; 596 return EXT4_ERR_UNKNOWN; 597 } 598 599 static void save_error_info(struct super_block *sb, int error, 600 __u32 ino, __u64 block, 601 const char *func, unsigned int line) 602 { 603 struct ext4_sb_info *sbi = EXT4_SB(sb); 604 605 /* We default to EFSCORRUPTED error... */ 606 if (error == 0) 607 error = EFSCORRUPTED; 608 609 spin_lock(&sbi->s_error_lock); 610 sbi->s_add_error_count++; 611 sbi->s_last_error_code = error; 612 sbi->s_last_error_line = line; 613 sbi->s_last_error_ino = ino; 614 sbi->s_last_error_block = block; 615 sbi->s_last_error_func = func; 616 sbi->s_last_error_time = ktime_get_real_seconds(); 617 if (!sbi->s_first_error_time) { 618 sbi->s_first_error_code = error; 619 sbi->s_first_error_line = line; 620 sbi->s_first_error_ino = ino; 621 sbi->s_first_error_block = block; 622 sbi->s_first_error_func = func; 623 sbi->s_first_error_time = sbi->s_last_error_time; 624 } 625 spin_unlock(&sbi->s_error_lock); 626 } 627 628 /* Deal with the reporting of failure conditions on a filesystem such as 629 * inconsistencies detected or read IO failures. 630 * 631 * On ext2, we can store the error state of the filesystem in the 632 * superblock. That is not possible on ext4, because we may have other 633 * write ordering constraints on the superblock which prevent us from 634 * writing it out straight away; and given that the journal is about to 635 * be aborted, we can't rely on the current, or future, transactions to 636 * write out the superblock safely. 637 * 638 * We'll just use the jbd2_journal_abort() error code to record an error in 639 * the journal instead. On recovery, the journal will complain about 640 * that error until we've noted it down and cleared it. 641 * 642 * If force_ro is set, we unconditionally force the filesystem into an 643 * ABORT|READONLY state, unless the error response on the fs has been set to 644 * panic in which case we take the easy way out and panic immediately. This is 645 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM 646 * at a critical moment in log management. 647 */ 648 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error, 649 __u32 ino, __u64 block, 650 const char *func, unsigned int line) 651 { 652 journal_t *journal = EXT4_SB(sb)->s_journal; 653 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT); 654 655 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 656 if (test_opt(sb, WARN_ON_ERROR)) 657 WARN_ON_ONCE(1); 658 659 if (!continue_fs && !sb_rdonly(sb)) { 660 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED); 661 if (journal) 662 jbd2_journal_abort(journal, -EIO); 663 } 664 665 if (!bdev_read_only(sb->s_bdev)) { 666 save_error_info(sb, error, ino, block, func, line); 667 /* 668 * In case the fs should keep running, we need to writeout 669 * superblock through the journal. Due to lock ordering 670 * constraints, it may not be safe to do it right here so we 671 * defer superblock flushing to a workqueue. 672 */ 673 if (continue_fs && journal) 674 schedule_work(&EXT4_SB(sb)->s_error_work); 675 else 676 ext4_commit_super(sb); 677 } 678 679 /* 680 * We force ERRORS_RO behavior when system is rebooting. Otherwise we 681 * could panic during 'reboot -f' as the underlying device got already 682 * disabled. 683 */ 684 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) { 685 panic("EXT4-fs (device %s): panic forced after error\n", 686 sb->s_id); 687 } 688 689 if (sb_rdonly(sb) || continue_fs) 690 return; 691 692 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 693 /* 694 * Make sure updated value of ->s_mount_flags will be visible before 695 * ->s_flags update 696 */ 697 smp_wmb(); 698 sb->s_flags |= SB_RDONLY; 699 } 700 701 static void flush_stashed_error_work(struct work_struct *work) 702 { 703 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info, 704 s_error_work); 705 journal_t *journal = sbi->s_journal; 706 handle_t *handle; 707 708 /* 709 * If the journal is still running, we have to write out superblock 710 * through the journal to avoid collisions of other journalled sb 711 * updates. 712 * 713 * We use directly jbd2 functions here to avoid recursing back into 714 * ext4 error handling code during handling of previous errors. 715 */ 716 if (!sb_rdonly(sbi->s_sb) && journal) { 717 struct buffer_head *sbh = sbi->s_sbh; 718 handle = jbd2_journal_start(journal, 1); 719 if (IS_ERR(handle)) 720 goto write_directly; 721 if (jbd2_journal_get_write_access(handle, sbh)) { 722 jbd2_journal_stop(handle); 723 goto write_directly; 724 } 725 ext4_update_super(sbi->s_sb); 726 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 727 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to " 728 "superblock detected"); 729 clear_buffer_write_io_error(sbh); 730 set_buffer_uptodate(sbh); 731 } 732 733 if (jbd2_journal_dirty_metadata(handle, sbh)) { 734 jbd2_journal_stop(handle); 735 goto write_directly; 736 } 737 jbd2_journal_stop(handle); 738 ext4_notify_error_sysfs(sbi); 739 return; 740 } 741 write_directly: 742 /* 743 * Write through journal failed. Write sb directly to get error info 744 * out and hope for the best. 745 */ 746 ext4_commit_super(sbi->s_sb); 747 ext4_notify_error_sysfs(sbi); 748 } 749 750 #define ext4_error_ratelimit(sb) \ 751 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ 752 "EXT4-fs error") 753 754 void __ext4_error(struct super_block *sb, const char *function, 755 unsigned int line, bool force_ro, int error, __u64 block, 756 const char *fmt, ...) 757 { 758 struct va_format vaf; 759 va_list args; 760 761 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 762 return; 763 764 trace_ext4_error(sb, function, line); 765 if (ext4_error_ratelimit(sb)) { 766 va_start(args, fmt); 767 vaf.fmt = fmt; 768 vaf.va = &args; 769 printk(KERN_CRIT 770 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 771 sb->s_id, function, line, current->comm, &vaf); 772 va_end(args); 773 } 774 fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED); 775 776 ext4_handle_error(sb, force_ro, error, 0, block, function, line); 777 } 778 779 void __ext4_error_inode(struct inode *inode, const char *function, 780 unsigned int line, ext4_fsblk_t block, int error, 781 const char *fmt, ...) 782 { 783 va_list args; 784 struct va_format vaf; 785 786 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 787 return; 788 789 trace_ext4_error(inode->i_sb, function, line); 790 if (ext4_error_ratelimit(inode->i_sb)) { 791 va_start(args, fmt); 792 vaf.fmt = fmt; 793 vaf.va = &args; 794 if (block) 795 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 796 "inode #%lu: block %llu: comm %s: %pV\n", 797 inode->i_sb->s_id, function, line, inode->i_ino, 798 block, current->comm, &vaf); 799 else 800 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 801 "inode #%lu: comm %s: %pV\n", 802 inode->i_sb->s_id, function, line, inode->i_ino, 803 current->comm, &vaf); 804 va_end(args); 805 } 806 fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED); 807 808 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block, 809 function, line); 810 } 811 812 void __ext4_error_file(struct file *file, const char *function, 813 unsigned int line, ext4_fsblk_t block, 814 const char *fmt, ...) 815 { 816 va_list args; 817 struct va_format vaf; 818 struct inode *inode = file_inode(file); 819 char pathname[80], *path; 820 821 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 822 return; 823 824 trace_ext4_error(inode->i_sb, function, line); 825 if (ext4_error_ratelimit(inode->i_sb)) { 826 path = file_path(file, pathname, sizeof(pathname)); 827 if (IS_ERR(path)) 828 path = "(unknown)"; 829 va_start(args, fmt); 830 vaf.fmt = fmt; 831 vaf.va = &args; 832 if (block) 833 printk(KERN_CRIT 834 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 835 "block %llu: comm %s: path %s: %pV\n", 836 inode->i_sb->s_id, function, line, inode->i_ino, 837 block, current->comm, path, &vaf); 838 else 839 printk(KERN_CRIT 840 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 841 "comm %s: path %s: %pV\n", 842 inode->i_sb->s_id, function, line, inode->i_ino, 843 current->comm, path, &vaf); 844 va_end(args); 845 } 846 fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED); 847 848 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block, 849 function, line); 850 } 851 852 const char *ext4_decode_error(struct super_block *sb, int errno, 853 char nbuf[16]) 854 { 855 char *errstr = NULL; 856 857 switch (errno) { 858 case -EFSCORRUPTED: 859 errstr = "Corrupt filesystem"; 860 break; 861 case -EFSBADCRC: 862 errstr = "Filesystem failed CRC"; 863 break; 864 case -EIO: 865 errstr = "IO failure"; 866 break; 867 case -ENOMEM: 868 errstr = "Out of memory"; 869 break; 870 case -EROFS: 871 if (!sb || (EXT4_SB(sb)->s_journal && 872 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 873 errstr = "Journal has aborted"; 874 else 875 errstr = "Readonly filesystem"; 876 break; 877 default: 878 /* If the caller passed in an extra buffer for unknown 879 * errors, textualise them now. Else we just return 880 * NULL. */ 881 if (nbuf) { 882 /* Check for truncated error codes... */ 883 if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 884 errstr = nbuf; 885 } 886 break; 887 } 888 889 return errstr; 890 } 891 892 /* __ext4_std_error decodes expected errors from journaling functions 893 * automatically and invokes the appropriate error response. */ 894 895 void __ext4_std_error(struct super_block *sb, const char *function, 896 unsigned int line, int errno) 897 { 898 char nbuf[16]; 899 const char *errstr; 900 901 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 902 return; 903 904 /* Special case: if the error is EROFS, and we're not already 905 * inside a transaction, then there's really no point in logging 906 * an error. */ 907 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb)) 908 return; 909 910 if (ext4_error_ratelimit(sb)) { 911 errstr = ext4_decode_error(sb, errno, nbuf); 912 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 913 sb->s_id, function, line, errstr); 914 } 915 fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED); 916 917 ext4_handle_error(sb, false, -errno, 0, 0, function, line); 918 } 919 920 void __ext4_msg(struct super_block *sb, 921 const char *prefix, const char *fmt, ...) 922 { 923 struct va_format vaf; 924 va_list args; 925 926 if (sb) { 927 atomic_inc(&EXT4_SB(sb)->s_msg_count); 928 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), 929 "EXT4-fs")) 930 return; 931 } 932 933 va_start(args, fmt); 934 vaf.fmt = fmt; 935 vaf.va = &args; 936 if (sb) 937 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 938 else 939 printk("%sEXT4-fs: %pV\n", prefix, &vaf); 940 va_end(args); 941 } 942 943 static int ext4_warning_ratelimit(struct super_block *sb) 944 { 945 atomic_inc(&EXT4_SB(sb)->s_warning_count); 946 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), 947 "EXT4-fs warning"); 948 } 949 950 void __ext4_warning(struct super_block *sb, const char *function, 951 unsigned int line, const char *fmt, ...) 952 { 953 struct va_format vaf; 954 va_list args; 955 956 if (!ext4_warning_ratelimit(sb)) 957 return; 958 959 va_start(args, fmt); 960 vaf.fmt = fmt; 961 vaf.va = &args; 962 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 963 sb->s_id, function, line, &vaf); 964 va_end(args); 965 } 966 967 void __ext4_warning_inode(const struct inode *inode, const char *function, 968 unsigned int line, const char *fmt, ...) 969 { 970 struct va_format vaf; 971 va_list args; 972 973 if (!ext4_warning_ratelimit(inode->i_sb)) 974 return; 975 976 va_start(args, fmt); 977 vaf.fmt = fmt; 978 vaf.va = &args; 979 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: " 980 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id, 981 function, line, inode->i_ino, current->comm, &vaf); 982 va_end(args); 983 } 984 985 void __ext4_grp_locked_error(const char *function, unsigned int line, 986 struct super_block *sb, ext4_group_t grp, 987 unsigned long ino, ext4_fsblk_t block, 988 const char *fmt, ...) 989 __releases(bitlock) 990 __acquires(bitlock) 991 { 992 struct va_format vaf; 993 va_list args; 994 995 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 996 return; 997 998 trace_ext4_error(sb, function, line); 999 if (ext4_error_ratelimit(sb)) { 1000 va_start(args, fmt); 1001 vaf.fmt = fmt; 1002 vaf.va = &args; 1003 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 1004 sb->s_id, function, line, grp); 1005 if (ino) 1006 printk(KERN_CONT "inode %lu: ", ino); 1007 if (block) 1008 printk(KERN_CONT "block %llu:", 1009 (unsigned long long) block); 1010 printk(KERN_CONT "%pV\n", &vaf); 1011 va_end(args); 1012 } 1013 1014 if (test_opt(sb, ERRORS_CONT)) { 1015 if (test_opt(sb, WARN_ON_ERROR)) 1016 WARN_ON_ONCE(1); 1017 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 1018 if (!bdev_read_only(sb->s_bdev)) { 1019 save_error_info(sb, EFSCORRUPTED, ino, block, function, 1020 line); 1021 schedule_work(&EXT4_SB(sb)->s_error_work); 1022 } 1023 return; 1024 } 1025 ext4_unlock_group(sb, grp); 1026 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line); 1027 /* 1028 * We only get here in the ERRORS_RO case; relocking the group 1029 * may be dangerous, but nothing bad will happen since the 1030 * filesystem will have already been marked read/only and the 1031 * journal has been aborted. We return 1 as a hint to callers 1032 * who might what to use the return value from 1033 * ext4_grp_locked_error() to distinguish between the 1034 * ERRORS_CONT and ERRORS_RO case, and perhaps return more 1035 * aggressively from the ext4 function in question, with a 1036 * more appropriate error code. 1037 */ 1038 ext4_lock_group(sb, grp); 1039 return; 1040 } 1041 1042 void ext4_mark_group_bitmap_corrupted(struct super_block *sb, 1043 ext4_group_t group, 1044 unsigned int flags) 1045 { 1046 struct ext4_sb_info *sbi = EXT4_SB(sb); 1047 struct ext4_group_info *grp = ext4_get_group_info(sb, group); 1048 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL); 1049 int ret; 1050 1051 if (!grp || !gdp) 1052 return; 1053 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) { 1054 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, 1055 &grp->bb_state); 1056 if (!ret) 1057 percpu_counter_sub(&sbi->s_freeclusters_counter, 1058 grp->bb_free); 1059 } 1060 1061 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) { 1062 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, 1063 &grp->bb_state); 1064 if (!ret && gdp) { 1065 int count; 1066 1067 count = ext4_free_inodes_count(sb, gdp); 1068 percpu_counter_sub(&sbi->s_freeinodes_counter, 1069 count); 1070 } 1071 } 1072 } 1073 1074 void ext4_update_dynamic_rev(struct super_block *sb) 1075 { 1076 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 1077 1078 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 1079 return; 1080 1081 ext4_warning(sb, 1082 "updating to rev %d because of new feature flag, " 1083 "running e2fsck is recommended", 1084 EXT4_DYNAMIC_REV); 1085 1086 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 1087 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 1088 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 1089 /* leave es->s_feature_*compat flags alone */ 1090 /* es->s_uuid will be set by e2fsck if empty */ 1091 1092 /* 1093 * The rest of the superblock fields should be zero, and if not it 1094 * means they are likely already in use, so leave them alone. We 1095 * can leave it up to e2fsck to clean up any inconsistencies there. 1096 */ 1097 } 1098 1099 static void ext4_bdev_mark_dead(struct block_device *bdev) 1100 { 1101 ext4_force_shutdown(bdev->bd_holder, EXT4_GOING_FLAGS_NOLOGFLUSH); 1102 } 1103 1104 static const struct blk_holder_ops ext4_holder_ops = { 1105 .mark_dead = ext4_bdev_mark_dead, 1106 }; 1107 1108 /* 1109 * Open the external journal device 1110 */ 1111 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 1112 { 1113 struct block_device *bdev; 1114 1115 bdev = blkdev_get_by_dev(dev, BLK_OPEN_READ | BLK_OPEN_WRITE, sb, 1116 &ext4_holder_ops); 1117 if (IS_ERR(bdev)) 1118 goto fail; 1119 return bdev; 1120 1121 fail: 1122 ext4_msg(sb, KERN_ERR, 1123 "failed to open journal device unknown-block(%u,%u) %ld", 1124 MAJOR(dev), MINOR(dev), PTR_ERR(bdev)); 1125 return NULL; 1126 } 1127 1128 /* 1129 * Release the journal device 1130 */ 1131 static void ext4_blkdev_remove(struct ext4_sb_info *sbi) 1132 { 1133 struct block_device *bdev; 1134 bdev = sbi->s_journal_bdev; 1135 if (bdev) { 1136 blkdev_put(bdev, sbi->s_sb); 1137 sbi->s_journal_bdev = NULL; 1138 } 1139 } 1140 1141 static inline struct inode *orphan_list_entry(struct list_head *l) 1142 { 1143 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 1144 } 1145 1146 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 1147 { 1148 struct list_head *l; 1149 1150 ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 1151 le32_to_cpu(sbi->s_es->s_last_orphan)); 1152 1153 printk(KERN_ERR "sb_info orphan list:\n"); 1154 list_for_each(l, &sbi->s_orphan) { 1155 struct inode *inode = orphan_list_entry(l); 1156 printk(KERN_ERR " " 1157 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 1158 inode->i_sb->s_id, inode->i_ino, inode, 1159 inode->i_mode, inode->i_nlink, 1160 NEXT_ORPHAN(inode)); 1161 } 1162 } 1163 1164 #ifdef CONFIG_QUOTA 1165 static int ext4_quota_off(struct super_block *sb, int type); 1166 1167 static inline void ext4_quota_off_umount(struct super_block *sb) 1168 { 1169 int type; 1170 1171 /* Use our quota_off function to clear inode flags etc. */ 1172 for (type = 0; type < EXT4_MAXQUOTAS; type++) 1173 ext4_quota_off(sb, type); 1174 } 1175 1176 /* 1177 * This is a helper function which is used in the mount/remount 1178 * codepaths (which holds s_umount) to fetch the quota file name. 1179 */ 1180 static inline char *get_qf_name(struct super_block *sb, 1181 struct ext4_sb_info *sbi, 1182 int type) 1183 { 1184 return rcu_dereference_protected(sbi->s_qf_names[type], 1185 lockdep_is_held(&sb->s_umount)); 1186 } 1187 #else 1188 static inline void ext4_quota_off_umount(struct super_block *sb) 1189 { 1190 } 1191 #endif 1192 1193 static int ext4_percpu_param_init(struct ext4_sb_info *sbi) 1194 { 1195 ext4_fsblk_t block; 1196 int err; 1197 1198 block = ext4_count_free_clusters(sbi->s_sb); 1199 ext4_free_blocks_count_set(sbi->s_es, EXT4_C2B(sbi, block)); 1200 err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 1201 GFP_KERNEL); 1202 if (!err) { 1203 unsigned long freei = ext4_count_free_inodes(sbi->s_sb); 1204 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 1205 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 1206 GFP_KERNEL); 1207 } 1208 if (!err) 1209 err = percpu_counter_init(&sbi->s_dirs_counter, 1210 ext4_count_dirs(sbi->s_sb), GFP_KERNEL); 1211 if (!err) 1212 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 1213 GFP_KERNEL); 1214 if (!err) 1215 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0, 1216 GFP_KERNEL); 1217 if (!err) 1218 err = percpu_init_rwsem(&sbi->s_writepages_rwsem); 1219 1220 if (err) 1221 ext4_msg(sbi->s_sb, KERN_ERR, "insufficient memory"); 1222 1223 return err; 1224 } 1225 1226 static void ext4_percpu_param_destroy(struct ext4_sb_info *sbi) 1227 { 1228 percpu_counter_destroy(&sbi->s_freeclusters_counter); 1229 percpu_counter_destroy(&sbi->s_freeinodes_counter); 1230 percpu_counter_destroy(&sbi->s_dirs_counter); 1231 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 1232 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit); 1233 percpu_free_rwsem(&sbi->s_writepages_rwsem); 1234 } 1235 1236 static void ext4_group_desc_free(struct ext4_sb_info *sbi) 1237 { 1238 struct buffer_head **group_desc; 1239 int i; 1240 1241 rcu_read_lock(); 1242 group_desc = rcu_dereference(sbi->s_group_desc); 1243 for (i = 0; i < sbi->s_gdb_count; i++) 1244 brelse(group_desc[i]); 1245 kvfree(group_desc); 1246 rcu_read_unlock(); 1247 } 1248 1249 static void ext4_flex_groups_free(struct ext4_sb_info *sbi) 1250 { 1251 struct flex_groups **flex_groups; 1252 int i; 1253 1254 rcu_read_lock(); 1255 flex_groups = rcu_dereference(sbi->s_flex_groups); 1256 if (flex_groups) { 1257 for (i = 0; i < sbi->s_flex_groups_allocated; i++) 1258 kvfree(flex_groups[i]); 1259 kvfree(flex_groups); 1260 } 1261 rcu_read_unlock(); 1262 } 1263 1264 static void ext4_put_super(struct super_block *sb) 1265 { 1266 struct ext4_sb_info *sbi = EXT4_SB(sb); 1267 struct ext4_super_block *es = sbi->s_es; 1268 int aborted = 0; 1269 int err; 1270 1271 /* 1272 * Unregister sysfs before destroying jbd2 journal. 1273 * Since we could still access attr_journal_task attribute via sysfs 1274 * path which could have sbi->s_journal->j_task as NULL 1275 * Unregister sysfs before flush sbi->s_error_work. 1276 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If 1277 * read metadata verify failed then will queue error work. 1278 * flush_stashed_error_work will call start_this_handle may trigger 1279 * BUG_ON. 1280 */ 1281 ext4_unregister_sysfs(sb); 1282 1283 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount")) 1284 ext4_msg(sb, KERN_INFO, "unmounting filesystem %pU.", 1285 &sb->s_uuid); 1286 1287 ext4_unregister_li_request(sb); 1288 ext4_quota_off_umount(sb); 1289 1290 flush_work(&sbi->s_error_work); 1291 destroy_workqueue(sbi->rsv_conversion_wq); 1292 ext4_release_orphan_info(sb); 1293 1294 if (sbi->s_journal) { 1295 aborted = is_journal_aborted(sbi->s_journal); 1296 err = jbd2_journal_destroy(sbi->s_journal); 1297 sbi->s_journal = NULL; 1298 if ((err < 0) && !aborted) { 1299 ext4_abort(sb, -err, "Couldn't clean up the journal"); 1300 } 1301 } 1302 1303 ext4_es_unregister_shrinker(sbi); 1304 timer_shutdown_sync(&sbi->s_err_report); 1305 ext4_release_system_zone(sb); 1306 ext4_mb_release(sb); 1307 ext4_ext_release(sb); 1308 1309 if (!sb_rdonly(sb) && !aborted) { 1310 ext4_clear_feature_journal_needs_recovery(sb); 1311 ext4_clear_feature_orphan_present(sb); 1312 es->s_state = cpu_to_le16(sbi->s_mount_state); 1313 } 1314 if (!sb_rdonly(sb)) 1315 ext4_commit_super(sb); 1316 1317 ext4_group_desc_free(sbi); 1318 ext4_flex_groups_free(sbi); 1319 ext4_percpu_param_destroy(sbi); 1320 #ifdef CONFIG_QUOTA 1321 for (int i = 0; i < EXT4_MAXQUOTAS; i++) 1322 kfree(get_qf_name(sb, sbi, i)); 1323 #endif 1324 1325 /* Debugging code just in case the in-memory inode orphan list 1326 * isn't empty. The on-disk one can be non-empty if we've 1327 * detected an error and taken the fs readonly, but the 1328 * in-memory list had better be clean by this point. */ 1329 if (!list_empty(&sbi->s_orphan)) 1330 dump_orphan_list(sb, sbi); 1331 ASSERT(list_empty(&sbi->s_orphan)); 1332 1333 sync_blockdev(sb->s_bdev); 1334 invalidate_bdev(sb->s_bdev); 1335 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) { 1336 /* 1337 * Invalidate the journal device's buffers. We don't want them 1338 * floating about in memory - the physical journal device may 1339 * hotswapped, and it breaks the `ro-after' testing code. 1340 */ 1341 sync_blockdev(sbi->s_journal_bdev); 1342 invalidate_bdev(sbi->s_journal_bdev); 1343 ext4_blkdev_remove(sbi); 1344 } 1345 1346 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 1347 sbi->s_ea_inode_cache = NULL; 1348 1349 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 1350 sbi->s_ea_block_cache = NULL; 1351 1352 ext4_stop_mmpd(sbi); 1353 1354 brelse(sbi->s_sbh); 1355 sb->s_fs_info = NULL; 1356 /* 1357 * Now that we are completely done shutting down the 1358 * superblock, we need to actually destroy the kobject. 1359 */ 1360 kobject_put(&sbi->s_kobj); 1361 wait_for_completion(&sbi->s_kobj_unregister); 1362 if (sbi->s_chksum_driver) 1363 crypto_free_shash(sbi->s_chksum_driver); 1364 kfree(sbi->s_blockgroup_lock); 1365 fs_put_dax(sbi->s_daxdev, NULL); 1366 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy); 1367 #if IS_ENABLED(CONFIG_UNICODE) 1368 utf8_unload(sb->s_encoding); 1369 #endif 1370 kfree(sbi); 1371 } 1372 1373 static struct kmem_cache *ext4_inode_cachep; 1374 1375 /* 1376 * Called inside transaction, so use GFP_NOFS 1377 */ 1378 static struct inode *ext4_alloc_inode(struct super_block *sb) 1379 { 1380 struct ext4_inode_info *ei; 1381 1382 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS); 1383 if (!ei) 1384 return NULL; 1385 1386 inode_set_iversion(&ei->vfs_inode, 1); 1387 ei->i_flags = 0; 1388 spin_lock_init(&ei->i_raw_lock); 1389 ei->i_prealloc_node = RB_ROOT; 1390 atomic_set(&ei->i_prealloc_active, 0); 1391 rwlock_init(&ei->i_prealloc_lock); 1392 ext4_es_init_tree(&ei->i_es_tree); 1393 rwlock_init(&ei->i_es_lock); 1394 INIT_LIST_HEAD(&ei->i_es_list); 1395 ei->i_es_all_nr = 0; 1396 ei->i_es_shk_nr = 0; 1397 ei->i_es_shrink_lblk = 0; 1398 ei->i_reserved_data_blocks = 0; 1399 spin_lock_init(&(ei->i_block_reservation_lock)); 1400 ext4_init_pending_tree(&ei->i_pending_tree); 1401 #ifdef CONFIG_QUOTA 1402 ei->i_reserved_quota = 0; 1403 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); 1404 #endif 1405 ei->jinode = NULL; 1406 INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 1407 spin_lock_init(&ei->i_completed_io_lock); 1408 ei->i_sync_tid = 0; 1409 ei->i_datasync_tid = 0; 1410 atomic_set(&ei->i_unwritten, 0); 1411 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 1412 ext4_fc_init_inode(&ei->vfs_inode); 1413 mutex_init(&ei->i_fc_lock); 1414 return &ei->vfs_inode; 1415 } 1416 1417 static int ext4_drop_inode(struct inode *inode) 1418 { 1419 int drop = generic_drop_inode(inode); 1420 1421 if (!drop) 1422 drop = fscrypt_drop_inode(inode); 1423 1424 trace_ext4_drop_inode(inode, drop); 1425 return drop; 1426 } 1427 1428 static void ext4_free_in_core_inode(struct inode *inode) 1429 { 1430 fscrypt_free_inode(inode); 1431 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) { 1432 pr_warn("%s: inode %ld still in fc list", 1433 __func__, inode->i_ino); 1434 } 1435 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 1436 } 1437 1438 static void ext4_destroy_inode(struct inode *inode) 1439 { 1440 if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 1441 ext4_msg(inode->i_sb, KERN_ERR, 1442 "Inode %lu (%p): orphan list check failed!", 1443 inode->i_ino, EXT4_I(inode)); 1444 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 1445 EXT4_I(inode), sizeof(struct ext4_inode_info), 1446 true); 1447 dump_stack(); 1448 } 1449 1450 if (EXT4_I(inode)->i_reserved_data_blocks) 1451 ext4_msg(inode->i_sb, KERN_ERR, 1452 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!", 1453 inode->i_ino, EXT4_I(inode), 1454 EXT4_I(inode)->i_reserved_data_blocks); 1455 } 1456 1457 static void ext4_shutdown(struct super_block *sb) 1458 { 1459 ext4_force_shutdown(sb, EXT4_GOING_FLAGS_NOLOGFLUSH); 1460 } 1461 1462 static void init_once(void *foo) 1463 { 1464 struct ext4_inode_info *ei = foo; 1465 1466 INIT_LIST_HEAD(&ei->i_orphan); 1467 init_rwsem(&ei->xattr_sem); 1468 init_rwsem(&ei->i_data_sem); 1469 inode_init_once(&ei->vfs_inode); 1470 ext4_fc_init_inode(&ei->vfs_inode); 1471 } 1472 1473 static int __init init_inodecache(void) 1474 { 1475 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache", 1476 sizeof(struct ext4_inode_info), 0, 1477 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD| 1478 SLAB_ACCOUNT), 1479 offsetof(struct ext4_inode_info, i_data), 1480 sizeof_field(struct ext4_inode_info, i_data), 1481 init_once); 1482 if (ext4_inode_cachep == NULL) 1483 return -ENOMEM; 1484 return 0; 1485 } 1486 1487 static void destroy_inodecache(void) 1488 { 1489 /* 1490 * Make sure all delayed rcu free inodes are flushed before we 1491 * destroy cache. 1492 */ 1493 rcu_barrier(); 1494 kmem_cache_destroy(ext4_inode_cachep); 1495 } 1496 1497 void ext4_clear_inode(struct inode *inode) 1498 { 1499 ext4_fc_del(inode); 1500 invalidate_inode_buffers(inode); 1501 clear_inode(inode); 1502 ext4_discard_preallocations(inode, 0); 1503 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 1504 dquot_drop(inode); 1505 if (EXT4_I(inode)->jinode) { 1506 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 1507 EXT4_I(inode)->jinode); 1508 jbd2_free_inode(EXT4_I(inode)->jinode); 1509 EXT4_I(inode)->jinode = NULL; 1510 } 1511 fscrypt_put_encryption_info(inode); 1512 fsverity_cleanup_inode(inode); 1513 } 1514 1515 static struct inode *ext4_nfs_get_inode(struct super_block *sb, 1516 u64 ino, u32 generation) 1517 { 1518 struct inode *inode; 1519 1520 /* 1521 * Currently we don't know the generation for parent directory, so 1522 * a generation of 0 means "accept any" 1523 */ 1524 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE); 1525 if (IS_ERR(inode)) 1526 return ERR_CAST(inode); 1527 if (generation && inode->i_generation != generation) { 1528 iput(inode); 1529 return ERR_PTR(-ESTALE); 1530 } 1531 1532 return inode; 1533 } 1534 1535 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 1536 int fh_len, int fh_type) 1537 { 1538 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1539 ext4_nfs_get_inode); 1540 } 1541 1542 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 1543 int fh_len, int fh_type) 1544 { 1545 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1546 ext4_nfs_get_inode); 1547 } 1548 1549 static int ext4_nfs_commit_metadata(struct inode *inode) 1550 { 1551 struct writeback_control wbc = { 1552 .sync_mode = WB_SYNC_ALL 1553 }; 1554 1555 trace_ext4_nfs_commit_metadata(inode); 1556 return ext4_write_inode(inode, &wbc); 1557 } 1558 1559 #ifdef CONFIG_QUOTA 1560 static const char * const quotatypes[] = INITQFNAMES; 1561 #define QTYPE2NAME(t) (quotatypes[t]) 1562 1563 static int ext4_write_dquot(struct dquot *dquot); 1564 static int ext4_acquire_dquot(struct dquot *dquot); 1565 static int ext4_release_dquot(struct dquot *dquot); 1566 static int ext4_mark_dquot_dirty(struct dquot *dquot); 1567 static int ext4_write_info(struct super_block *sb, int type); 1568 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1569 const struct path *path); 1570 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1571 size_t len, loff_t off); 1572 static ssize_t ext4_quota_write(struct super_block *sb, int type, 1573 const char *data, size_t len, loff_t off); 1574 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 1575 unsigned int flags); 1576 1577 static struct dquot **ext4_get_dquots(struct inode *inode) 1578 { 1579 return EXT4_I(inode)->i_dquot; 1580 } 1581 1582 static const struct dquot_operations ext4_quota_operations = { 1583 .get_reserved_space = ext4_get_reserved_space, 1584 .write_dquot = ext4_write_dquot, 1585 .acquire_dquot = ext4_acquire_dquot, 1586 .release_dquot = ext4_release_dquot, 1587 .mark_dirty = ext4_mark_dquot_dirty, 1588 .write_info = ext4_write_info, 1589 .alloc_dquot = dquot_alloc, 1590 .destroy_dquot = dquot_destroy, 1591 .get_projid = ext4_get_projid, 1592 .get_inode_usage = ext4_get_inode_usage, 1593 .get_next_id = dquot_get_next_id, 1594 }; 1595 1596 static const struct quotactl_ops ext4_qctl_operations = { 1597 .quota_on = ext4_quota_on, 1598 .quota_off = ext4_quota_off, 1599 .quota_sync = dquot_quota_sync, 1600 .get_state = dquot_get_state, 1601 .set_info = dquot_set_dqinfo, 1602 .get_dqblk = dquot_get_dqblk, 1603 .set_dqblk = dquot_set_dqblk, 1604 .get_nextdqblk = dquot_get_next_dqblk, 1605 }; 1606 #endif 1607 1608 static const struct super_operations ext4_sops = { 1609 .alloc_inode = ext4_alloc_inode, 1610 .free_inode = ext4_free_in_core_inode, 1611 .destroy_inode = ext4_destroy_inode, 1612 .write_inode = ext4_write_inode, 1613 .dirty_inode = ext4_dirty_inode, 1614 .drop_inode = ext4_drop_inode, 1615 .evict_inode = ext4_evict_inode, 1616 .put_super = ext4_put_super, 1617 .sync_fs = ext4_sync_fs, 1618 .freeze_fs = ext4_freeze, 1619 .unfreeze_fs = ext4_unfreeze, 1620 .statfs = ext4_statfs, 1621 .show_options = ext4_show_options, 1622 .shutdown = ext4_shutdown, 1623 #ifdef CONFIG_QUOTA 1624 .quota_read = ext4_quota_read, 1625 .quota_write = ext4_quota_write, 1626 .get_dquots = ext4_get_dquots, 1627 #endif 1628 }; 1629 1630 static const struct export_operations ext4_export_ops = { 1631 .fh_to_dentry = ext4_fh_to_dentry, 1632 .fh_to_parent = ext4_fh_to_parent, 1633 .get_parent = ext4_get_parent, 1634 .commit_metadata = ext4_nfs_commit_metadata, 1635 }; 1636 1637 enum { 1638 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1639 Opt_resgid, Opt_resuid, Opt_sb, 1640 Opt_nouid32, Opt_debug, Opt_removed, 1641 Opt_user_xattr, Opt_acl, 1642 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1643 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1644 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1645 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1646 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 1647 Opt_inlinecrypt, 1648 Opt_usrjquota, Opt_grpjquota, Opt_quota, 1649 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1650 Opt_usrquota, Opt_grpquota, Opt_prjquota, 1651 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never, 1652 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error, 1653 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize, 1654 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 1655 Opt_inode_readahead_blks, Opt_journal_ioprio, 1656 Opt_dioread_nolock, Opt_dioread_lock, 1657 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1658 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, 1659 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan, 1660 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type, 1661 #ifdef CONFIG_EXT4_DEBUG 1662 Opt_fc_debug_max_replay, Opt_fc_debug_force 1663 #endif 1664 }; 1665 1666 static const struct constant_table ext4_param_errors[] = { 1667 {"continue", EXT4_MOUNT_ERRORS_CONT}, 1668 {"panic", EXT4_MOUNT_ERRORS_PANIC}, 1669 {"remount-ro", EXT4_MOUNT_ERRORS_RO}, 1670 {} 1671 }; 1672 1673 static const struct constant_table ext4_param_data[] = { 1674 {"journal", EXT4_MOUNT_JOURNAL_DATA}, 1675 {"ordered", EXT4_MOUNT_ORDERED_DATA}, 1676 {"writeback", EXT4_MOUNT_WRITEBACK_DATA}, 1677 {} 1678 }; 1679 1680 static const struct constant_table ext4_param_data_err[] = { 1681 {"abort", Opt_data_err_abort}, 1682 {"ignore", Opt_data_err_ignore}, 1683 {} 1684 }; 1685 1686 static const struct constant_table ext4_param_jqfmt[] = { 1687 {"vfsold", QFMT_VFS_OLD}, 1688 {"vfsv0", QFMT_VFS_V0}, 1689 {"vfsv1", QFMT_VFS_V1}, 1690 {} 1691 }; 1692 1693 static const struct constant_table ext4_param_dax[] = { 1694 {"always", Opt_dax_always}, 1695 {"inode", Opt_dax_inode}, 1696 {"never", Opt_dax_never}, 1697 {} 1698 }; 1699 1700 /* String parameter that allows empty argument */ 1701 #define fsparam_string_empty(NAME, OPT) \ 1702 __fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL) 1703 1704 /* 1705 * Mount option specification 1706 * We don't use fsparam_flag_no because of the way we set the 1707 * options and the way we show them in _ext4_show_options(). To 1708 * keep the changes to a minimum, let's keep the negative options 1709 * separate for now. 1710 */ 1711 static const struct fs_parameter_spec ext4_param_specs[] = { 1712 fsparam_flag ("bsddf", Opt_bsd_df), 1713 fsparam_flag ("minixdf", Opt_minix_df), 1714 fsparam_flag ("grpid", Opt_grpid), 1715 fsparam_flag ("bsdgroups", Opt_grpid), 1716 fsparam_flag ("nogrpid", Opt_nogrpid), 1717 fsparam_flag ("sysvgroups", Opt_nogrpid), 1718 fsparam_u32 ("resgid", Opt_resgid), 1719 fsparam_u32 ("resuid", Opt_resuid), 1720 fsparam_u32 ("sb", Opt_sb), 1721 fsparam_enum ("errors", Opt_errors, ext4_param_errors), 1722 fsparam_flag ("nouid32", Opt_nouid32), 1723 fsparam_flag ("debug", Opt_debug), 1724 fsparam_flag ("oldalloc", Opt_removed), 1725 fsparam_flag ("orlov", Opt_removed), 1726 fsparam_flag ("user_xattr", Opt_user_xattr), 1727 fsparam_flag ("acl", Opt_acl), 1728 fsparam_flag ("norecovery", Opt_noload), 1729 fsparam_flag ("noload", Opt_noload), 1730 fsparam_flag ("bh", Opt_removed), 1731 fsparam_flag ("nobh", Opt_removed), 1732 fsparam_u32 ("commit", Opt_commit), 1733 fsparam_u32 ("min_batch_time", Opt_min_batch_time), 1734 fsparam_u32 ("max_batch_time", Opt_max_batch_time), 1735 fsparam_u32 ("journal_dev", Opt_journal_dev), 1736 fsparam_bdev ("journal_path", Opt_journal_path), 1737 fsparam_flag ("journal_checksum", Opt_journal_checksum), 1738 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum), 1739 fsparam_flag ("journal_async_commit",Opt_journal_async_commit), 1740 fsparam_flag ("abort", Opt_abort), 1741 fsparam_enum ("data", Opt_data, ext4_param_data), 1742 fsparam_enum ("data_err", Opt_data_err, 1743 ext4_param_data_err), 1744 fsparam_string_empty 1745 ("usrjquota", Opt_usrjquota), 1746 fsparam_string_empty 1747 ("grpjquota", Opt_grpjquota), 1748 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt), 1749 fsparam_flag ("grpquota", Opt_grpquota), 1750 fsparam_flag ("quota", Opt_quota), 1751 fsparam_flag ("noquota", Opt_noquota), 1752 fsparam_flag ("usrquota", Opt_usrquota), 1753 fsparam_flag ("prjquota", Opt_prjquota), 1754 fsparam_flag ("barrier", Opt_barrier), 1755 fsparam_u32 ("barrier", Opt_barrier), 1756 fsparam_flag ("nobarrier", Opt_nobarrier), 1757 fsparam_flag ("i_version", Opt_removed), 1758 fsparam_flag ("dax", Opt_dax), 1759 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax), 1760 fsparam_u32 ("stripe", Opt_stripe), 1761 fsparam_flag ("delalloc", Opt_delalloc), 1762 fsparam_flag ("nodelalloc", Opt_nodelalloc), 1763 fsparam_flag ("warn_on_error", Opt_warn_on_error), 1764 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error), 1765 fsparam_u32 ("debug_want_extra_isize", 1766 Opt_debug_want_extra_isize), 1767 fsparam_flag ("mblk_io_submit", Opt_removed), 1768 fsparam_flag ("nomblk_io_submit", Opt_removed), 1769 fsparam_flag ("block_validity", Opt_block_validity), 1770 fsparam_flag ("noblock_validity", Opt_noblock_validity), 1771 fsparam_u32 ("inode_readahead_blks", 1772 Opt_inode_readahead_blks), 1773 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio), 1774 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc), 1775 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc), 1776 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc), 1777 fsparam_flag ("dioread_nolock", Opt_dioread_nolock), 1778 fsparam_flag ("nodioread_nolock", Opt_dioread_lock), 1779 fsparam_flag ("dioread_lock", Opt_dioread_lock), 1780 fsparam_flag ("discard", Opt_discard), 1781 fsparam_flag ("nodiscard", Opt_nodiscard), 1782 fsparam_u32 ("init_itable", Opt_init_itable), 1783 fsparam_flag ("init_itable", Opt_init_itable), 1784 fsparam_flag ("noinit_itable", Opt_noinit_itable), 1785 #ifdef CONFIG_EXT4_DEBUG 1786 fsparam_flag ("fc_debug_force", Opt_fc_debug_force), 1787 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay), 1788 #endif 1789 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb), 1790 fsparam_flag ("test_dummy_encryption", 1791 Opt_test_dummy_encryption), 1792 fsparam_string ("test_dummy_encryption", 1793 Opt_test_dummy_encryption), 1794 fsparam_flag ("inlinecrypt", Opt_inlinecrypt), 1795 fsparam_flag ("nombcache", Opt_nombcache), 1796 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */ 1797 fsparam_flag ("prefetch_block_bitmaps", 1798 Opt_removed), 1799 fsparam_flag ("no_prefetch_block_bitmaps", 1800 Opt_no_prefetch_block_bitmaps), 1801 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan), 1802 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */ 1803 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */ 1804 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */ 1805 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */ 1806 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */ 1807 {} 1808 }; 1809 1810 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1811 1812 #define MOPT_SET 0x0001 1813 #define MOPT_CLEAR 0x0002 1814 #define MOPT_NOSUPPORT 0x0004 1815 #define MOPT_EXPLICIT 0x0008 1816 #ifdef CONFIG_QUOTA 1817 #define MOPT_Q 0 1818 #define MOPT_QFMT 0x0010 1819 #else 1820 #define MOPT_Q MOPT_NOSUPPORT 1821 #define MOPT_QFMT MOPT_NOSUPPORT 1822 #endif 1823 #define MOPT_NO_EXT2 0x0020 1824 #define MOPT_NO_EXT3 0x0040 1825 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1826 #define MOPT_SKIP 0x0080 1827 #define MOPT_2 0x0100 1828 1829 static const struct mount_opts { 1830 int token; 1831 int mount_opt; 1832 int flags; 1833 } ext4_mount_opts[] = { 1834 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 1835 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 1836 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 1837 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 1838 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 1839 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 1840 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 1841 MOPT_EXT4_ONLY | MOPT_SET}, 1842 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 1843 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1844 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 1845 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 1846 {Opt_delalloc, EXT4_MOUNT_DELALLOC, 1847 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1848 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 1849 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1850 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET}, 1851 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR}, 1852 {Opt_commit, 0, MOPT_NO_EXT2}, 1853 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1854 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1855 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1856 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1857 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 1858 EXT4_MOUNT_JOURNAL_CHECKSUM), 1859 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1860 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 1861 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2}, 1862 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 1863 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 1864 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 1865 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 1866 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 1867 {Opt_dax_type, 0, MOPT_EXT4_ONLY}, 1868 {Opt_journal_dev, 0, MOPT_NO_EXT2}, 1869 {Opt_journal_path, 0, MOPT_NO_EXT2}, 1870 {Opt_journal_ioprio, 0, MOPT_NO_EXT2}, 1871 {Opt_data, 0, MOPT_NO_EXT2}, 1872 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 1873 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1874 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 1875 #else 1876 {Opt_acl, 0, MOPT_NOSUPPORT}, 1877 #endif 1878 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 1879 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1880 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 1881 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 1882 MOPT_SET | MOPT_Q}, 1883 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 1884 MOPT_SET | MOPT_Q}, 1885 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, 1886 MOPT_SET | MOPT_Q}, 1887 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 1888 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), 1889 MOPT_CLEAR | MOPT_Q}, 1890 {Opt_usrjquota, 0, MOPT_Q}, 1891 {Opt_grpjquota, 0, MOPT_Q}, 1892 {Opt_jqfmt, 0, MOPT_QFMT}, 1893 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, 1894 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS, 1895 MOPT_SET}, 1896 #ifdef CONFIG_EXT4_DEBUG 1897 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT, 1898 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY}, 1899 #endif 1900 {Opt_err, 0, 0} 1901 }; 1902 1903 #if IS_ENABLED(CONFIG_UNICODE) 1904 static const struct ext4_sb_encodings { 1905 __u16 magic; 1906 char *name; 1907 unsigned int version; 1908 } ext4_sb_encoding_map[] = { 1909 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)}, 1910 }; 1911 1912 static const struct ext4_sb_encodings * 1913 ext4_sb_read_encoding(const struct ext4_super_block *es) 1914 { 1915 __u16 magic = le16_to_cpu(es->s_encoding); 1916 int i; 1917 1918 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++) 1919 if (magic == ext4_sb_encoding_map[i].magic) 1920 return &ext4_sb_encoding_map[i]; 1921 1922 return NULL; 1923 } 1924 #endif 1925 1926 #define EXT4_SPEC_JQUOTA (1 << 0) 1927 #define EXT4_SPEC_JQFMT (1 << 1) 1928 #define EXT4_SPEC_DATAJ (1 << 2) 1929 #define EXT4_SPEC_SB_BLOCK (1 << 3) 1930 #define EXT4_SPEC_JOURNAL_DEV (1 << 4) 1931 #define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5) 1932 #define EXT4_SPEC_s_want_extra_isize (1 << 7) 1933 #define EXT4_SPEC_s_max_batch_time (1 << 8) 1934 #define EXT4_SPEC_s_min_batch_time (1 << 9) 1935 #define EXT4_SPEC_s_inode_readahead_blks (1 << 10) 1936 #define EXT4_SPEC_s_li_wait_mult (1 << 11) 1937 #define EXT4_SPEC_s_max_dir_size_kb (1 << 12) 1938 #define EXT4_SPEC_s_stripe (1 << 13) 1939 #define EXT4_SPEC_s_resuid (1 << 14) 1940 #define EXT4_SPEC_s_resgid (1 << 15) 1941 #define EXT4_SPEC_s_commit_interval (1 << 16) 1942 #define EXT4_SPEC_s_fc_debug_max_replay (1 << 17) 1943 #define EXT4_SPEC_s_sb_block (1 << 18) 1944 #define EXT4_SPEC_mb_optimize_scan (1 << 19) 1945 1946 struct ext4_fs_context { 1947 char *s_qf_names[EXT4_MAXQUOTAS]; 1948 struct fscrypt_dummy_policy dummy_enc_policy; 1949 int s_jquota_fmt; /* Format of quota to use */ 1950 #ifdef CONFIG_EXT4_DEBUG 1951 int s_fc_debug_max_replay; 1952 #endif 1953 unsigned short qname_spec; 1954 unsigned long vals_s_flags; /* Bits to set in s_flags */ 1955 unsigned long mask_s_flags; /* Bits changed in s_flags */ 1956 unsigned long journal_devnum; 1957 unsigned long s_commit_interval; 1958 unsigned long s_stripe; 1959 unsigned int s_inode_readahead_blks; 1960 unsigned int s_want_extra_isize; 1961 unsigned int s_li_wait_mult; 1962 unsigned int s_max_dir_size_kb; 1963 unsigned int journal_ioprio; 1964 unsigned int vals_s_mount_opt; 1965 unsigned int mask_s_mount_opt; 1966 unsigned int vals_s_mount_opt2; 1967 unsigned int mask_s_mount_opt2; 1968 unsigned long vals_s_mount_flags; 1969 unsigned long mask_s_mount_flags; 1970 unsigned int opt_flags; /* MOPT flags */ 1971 unsigned int spec; 1972 u32 s_max_batch_time; 1973 u32 s_min_batch_time; 1974 kuid_t s_resuid; 1975 kgid_t s_resgid; 1976 ext4_fsblk_t s_sb_block; 1977 }; 1978 1979 static void ext4_fc_free(struct fs_context *fc) 1980 { 1981 struct ext4_fs_context *ctx = fc->fs_private; 1982 int i; 1983 1984 if (!ctx) 1985 return; 1986 1987 for (i = 0; i < EXT4_MAXQUOTAS; i++) 1988 kfree(ctx->s_qf_names[i]); 1989 1990 fscrypt_free_dummy_policy(&ctx->dummy_enc_policy); 1991 kfree(ctx); 1992 } 1993 1994 int ext4_init_fs_context(struct fs_context *fc) 1995 { 1996 struct ext4_fs_context *ctx; 1997 1998 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL); 1999 if (!ctx) 2000 return -ENOMEM; 2001 2002 fc->fs_private = ctx; 2003 fc->ops = &ext4_context_ops; 2004 2005 return 0; 2006 } 2007 2008 #ifdef CONFIG_QUOTA 2009 /* 2010 * Note the name of the specified quota file. 2011 */ 2012 static int note_qf_name(struct fs_context *fc, int qtype, 2013 struct fs_parameter *param) 2014 { 2015 struct ext4_fs_context *ctx = fc->fs_private; 2016 char *qname; 2017 2018 if (param->size < 1) { 2019 ext4_msg(NULL, KERN_ERR, "Missing quota name"); 2020 return -EINVAL; 2021 } 2022 if (strchr(param->string, '/')) { 2023 ext4_msg(NULL, KERN_ERR, 2024 "quotafile must be on filesystem root"); 2025 return -EINVAL; 2026 } 2027 if (ctx->s_qf_names[qtype]) { 2028 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) { 2029 ext4_msg(NULL, KERN_ERR, 2030 "%s quota file already specified", 2031 QTYPE2NAME(qtype)); 2032 return -EINVAL; 2033 } 2034 return 0; 2035 } 2036 2037 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL); 2038 if (!qname) { 2039 ext4_msg(NULL, KERN_ERR, 2040 "Not enough memory for storing quotafile name"); 2041 return -ENOMEM; 2042 } 2043 ctx->s_qf_names[qtype] = qname; 2044 ctx->qname_spec |= 1 << qtype; 2045 ctx->spec |= EXT4_SPEC_JQUOTA; 2046 return 0; 2047 } 2048 2049 /* 2050 * Clear the name of the specified quota file. 2051 */ 2052 static int unnote_qf_name(struct fs_context *fc, int qtype) 2053 { 2054 struct ext4_fs_context *ctx = fc->fs_private; 2055 2056 if (ctx->s_qf_names[qtype]) 2057 kfree(ctx->s_qf_names[qtype]); 2058 2059 ctx->s_qf_names[qtype] = NULL; 2060 ctx->qname_spec |= 1 << qtype; 2061 ctx->spec |= EXT4_SPEC_JQUOTA; 2062 return 0; 2063 } 2064 #endif 2065 2066 static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param, 2067 struct ext4_fs_context *ctx) 2068 { 2069 int err; 2070 2071 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) { 2072 ext4_msg(NULL, KERN_WARNING, 2073 "test_dummy_encryption option not supported"); 2074 return -EINVAL; 2075 } 2076 err = fscrypt_parse_test_dummy_encryption(param, 2077 &ctx->dummy_enc_policy); 2078 if (err == -EINVAL) { 2079 ext4_msg(NULL, KERN_WARNING, 2080 "Value of option \"%s\" is unrecognized", param->key); 2081 } else if (err == -EEXIST) { 2082 ext4_msg(NULL, KERN_WARNING, 2083 "Conflicting test_dummy_encryption options"); 2084 return -EINVAL; 2085 } 2086 return err; 2087 } 2088 2089 #define EXT4_SET_CTX(name) \ 2090 static inline void ctx_set_##name(struct ext4_fs_context *ctx, \ 2091 unsigned long flag) \ 2092 { \ 2093 ctx->mask_s_##name |= flag; \ 2094 ctx->vals_s_##name |= flag; \ 2095 } 2096 2097 #define EXT4_CLEAR_CTX(name) \ 2098 static inline void ctx_clear_##name(struct ext4_fs_context *ctx, \ 2099 unsigned long flag) \ 2100 { \ 2101 ctx->mask_s_##name |= flag; \ 2102 ctx->vals_s_##name &= ~flag; \ 2103 } 2104 2105 #define EXT4_TEST_CTX(name) \ 2106 static inline unsigned long \ 2107 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \ 2108 { \ 2109 return (ctx->vals_s_##name & flag); \ 2110 } 2111 2112 EXT4_SET_CTX(flags); /* set only */ 2113 EXT4_SET_CTX(mount_opt); 2114 EXT4_CLEAR_CTX(mount_opt); 2115 EXT4_TEST_CTX(mount_opt); 2116 EXT4_SET_CTX(mount_opt2); 2117 EXT4_CLEAR_CTX(mount_opt2); 2118 EXT4_TEST_CTX(mount_opt2); 2119 2120 static inline void ctx_set_mount_flag(struct ext4_fs_context *ctx, int bit) 2121 { 2122 set_bit(bit, &ctx->mask_s_mount_flags); 2123 set_bit(bit, &ctx->vals_s_mount_flags); 2124 } 2125 2126 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param) 2127 { 2128 struct ext4_fs_context *ctx = fc->fs_private; 2129 struct fs_parse_result result; 2130 const struct mount_opts *m; 2131 int is_remount; 2132 kuid_t uid; 2133 kgid_t gid; 2134 int token; 2135 2136 token = fs_parse(fc, ext4_param_specs, param, &result); 2137 if (token < 0) 2138 return token; 2139 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE; 2140 2141 for (m = ext4_mount_opts; m->token != Opt_err; m++) 2142 if (token == m->token) 2143 break; 2144 2145 ctx->opt_flags |= m->flags; 2146 2147 if (m->flags & MOPT_EXPLICIT) { 2148 if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 2149 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC); 2150 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 2151 ctx_set_mount_opt2(ctx, 2152 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM); 2153 } else 2154 return -EINVAL; 2155 } 2156 2157 if (m->flags & MOPT_NOSUPPORT) { 2158 ext4_msg(NULL, KERN_ERR, "%s option not supported", 2159 param->key); 2160 return 0; 2161 } 2162 2163 switch (token) { 2164 #ifdef CONFIG_QUOTA 2165 case Opt_usrjquota: 2166 if (!*param->string) 2167 return unnote_qf_name(fc, USRQUOTA); 2168 else 2169 return note_qf_name(fc, USRQUOTA, param); 2170 case Opt_grpjquota: 2171 if (!*param->string) 2172 return unnote_qf_name(fc, GRPQUOTA); 2173 else 2174 return note_qf_name(fc, GRPQUOTA, param); 2175 #endif 2176 case Opt_sb: 2177 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) { 2178 ext4_msg(NULL, KERN_WARNING, 2179 "Ignoring %s option on remount", param->key); 2180 } else { 2181 ctx->s_sb_block = result.uint_32; 2182 ctx->spec |= EXT4_SPEC_s_sb_block; 2183 } 2184 return 0; 2185 case Opt_removed: 2186 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option", 2187 param->key); 2188 return 0; 2189 case Opt_abort: 2190 ctx_set_mount_flag(ctx, EXT4_MF_FS_ABORTED); 2191 return 0; 2192 case Opt_inlinecrypt: 2193 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 2194 ctx_set_flags(ctx, SB_INLINECRYPT); 2195 #else 2196 ext4_msg(NULL, KERN_ERR, "inline encryption not supported"); 2197 #endif 2198 return 0; 2199 case Opt_errors: 2200 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK); 2201 ctx_set_mount_opt(ctx, result.uint_32); 2202 return 0; 2203 #ifdef CONFIG_QUOTA 2204 case Opt_jqfmt: 2205 ctx->s_jquota_fmt = result.uint_32; 2206 ctx->spec |= EXT4_SPEC_JQFMT; 2207 return 0; 2208 #endif 2209 case Opt_data: 2210 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS); 2211 ctx_set_mount_opt(ctx, result.uint_32); 2212 ctx->spec |= EXT4_SPEC_DATAJ; 2213 return 0; 2214 case Opt_commit: 2215 if (result.uint_32 == 0) 2216 result.uint_32 = JBD2_DEFAULT_MAX_COMMIT_AGE; 2217 else if (result.uint_32 > INT_MAX / HZ) { 2218 ext4_msg(NULL, KERN_ERR, 2219 "Invalid commit interval %d, " 2220 "must be smaller than %d", 2221 result.uint_32, INT_MAX / HZ); 2222 return -EINVAL; 2223 } 2224 ctx->s_commit_interval = HZ * result.uint_32; 2225 ctx->spec |= EXT4_SPEC_s_commit_interval; 2226 return 0; 2227 case Opt_debug_want_extra_isize: 2228 if ((result.uint_32 & 1) || (result.uint_32 < 4)) { 2229 ext4_msg(NULL, KERN_ERR, 2230 "Invalid want_extra_isize %d", result.uint_32); 2231 return -EINVAL; 2232 } 2233 ctx->s_want_extra_isize = result.uint_32; 2234 ctx->spec |= EXT4_SPEC_s_want_extra_isize; 2235 return 0; 2236 case Opt_max_batch_time: 2237 ctx->s_max_batch_time = result.uint_32; 2238 ctx->spec |= EXT4_SPEC_s_max_batch_time; 2239 return 0; 2240 case Opt_min_batch_time: 2241 ctx->s_min_batch_time = result.uint_32; 2242 ctx->spec |= EXT4_SPEC_s_min_batch_time; 2243 return 0; 2244 case Opt_inode_readahead_blks: 2245 if (result.uint_32 && 2246 (result.uint_32 > (1 << 30) || 2247 !is_power_of_2(result.uint_32))) { 2248 ext4_msg(NULL, KERN_ERR, 2249 "EXT4-fs: inode_readahead_blks must be " 2250 "0 or a power of 2 smaller than 2^31"); 2251 return -EINVAL; 2252 } 2253 ctx->s_inode_readahead_blks = result.uint_32; 2254 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks; 2255 return 0; 2256 case Opt_init_itable: 2257 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE); 2258 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 2259 if (param->type == fs_value_is_string) 2260 ctx->s_li_wait_mult = result.uint_32; 2261 ctx->spec |= EXT4_SPEC_s_li_wait_mult; 2262 return 0; 2263 case Opt_max_dir_size_kb: 2264 ctx->s_max_dir_size_kb = result.uint_32; 2265 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb; 2266 return 0; 2267 #ifdef CONFIG_EXT4_DEBUG 2268 case Opt_fc_debug_max_replay: 2269 ctx->s_fc_debug_max_replay = result.uint_32; 2270 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay; 2271 return 0; 2272 #endif 2273 case Opt_stripe: 2274 ctx->s_stripe = result.uint_32; 2275 ctx->spec |= EXT4_SPEC_s_stripe; 2276 return 0; 2277 case Opt_resuid: 2278 uid = make_kuid(current_user_ns(), result.uint_32); 2279 if (!uid_valid(uid)) { 2280 ext4_msg(NULL, KERN_ERR, "Invalid uid value %d", 2281 result.uint_32); 2282 return -EINVAL; 2283 } 2284 ctx->s_resuid = uid; 2285 ctx->spec |= EXT4_SPEC_s_resuid; 2286 return 0; 2287 case Opt_resgid: 2288 gid = make_kgid(current_user_ns(), result.uint_32); 2289 if (!gid_valid(gid)) { 2290 ext4_msg(NULL, KERN_ERR, "Invalid gid value %d", 2291 result.uint_32); 2292 return -EINVAL; 2293 } 2294 ctx->s_resgid = gid; 2295 ctx->spec |= EXT4_SPEC_s_resgid; 2296 return 0; 2297 case Opt_journal_dev: 2298 if (is_remount) { 2299 ext4_msg(NULL, KERN_ERR, 2300 "Cannot specify journal on remount"); 2301 return -EINVAL; 2302 } 2303 ctx->journal_devnum = result.uint_32; 2304 ctx->spec |= EXT4_SPEC_JOURNAL_DEV; 2305 return 0; 2306 case Opt_journal_path: 2307 { 2308 struct inode *journal_inode; 2309 struct path path; 2310 int error; 2311 2312 if (is_remount) { 2313 ext4_msg(NULL, KERN_ERR, 2314 "Cannot specify journal on remount"); 2315 return -EINVAL; 2316 } 2317 2318 error = fs_lookup_param(fc, param, 1, LOOKUP_FOLLOW, &path); 2319 if (error) { 2320 ext4_msg(NULL, KERN_ERR, "error: could not find " 2321 "journal device path"); 2322 return -EINVAL; 2323 } 2324 2325 journal_inode = d_inode(path.dentry); 2326 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev); 2327 ctx->spec |= EXT4_SPEC_JOURNAL_DEV; 2328 path_put(&path); 2329 return 0; 2330 } 2331 case Opt_journal_ioprio: 2332 if (result.uint_32 > 7) { 2333 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority" 2334 " (must be 0-7)"); 2335 return -EINVAL; 2336 } 2337 ctx->journal_ioprio = 2338 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32); 2339 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO; 2340 return 0; 2341 case Opt_test_dummy_encryption: 2342 return ext4_parse_test_dummy_encryption(param, ctx); 2343 case Opt_dax: 2344 case Opt_dax_type: 2345 #ifdef CONFIG_FS_DAX 2346 { 2347 int type = (token == Opt_dax) ? 2348 Opt_dax : result.uint_32; 2349 2350 switch (type) { 2351 case Opt_dax: 2352 case Opt_dax_always: 2353 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); 2354 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); 2355 break; 2356 case Opt_dax_never: 2357 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); 2358 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); 2359 break; 2360 case Opt_dax_inode: 2361 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); 2362 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); 2363 /* Strictly for printing options */ 2364 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE); 2365 break; 2366 } 2367 return 0; 2368 } 2369 #else 2370 ext4_msg(NULL, KERN_INFO, "dax option not supported"); 2371 return -EINVAL; 2372 #endif 2373 case Opt_data_err: 2374 if (result.uint_32 == Opt_data_err_abort) 2375 ctx_set_mount_opt(ctx, m->mount_opt); 2376 else if (result.uint_32 == Opt_data_err_ignore) 2377 ctx_clear_mount_opt(ctx, m->mount_opt); 2378 return 0; 2379 case Opt_mb_optimize_scan: 2380 if (result.int_32 == 1) { 2381 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN); 2382 ctx->spec |= EXT4_SPEC_mb_optimize_scan; 2383 } else if (result.int_32 == 0) { 2384 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN); 2385 ctx->spec |= EXT4_SPEC_mb_optimize_scan; 2386 } else { 2387 ext4_msg(NULL, KERN_WARNING, 2388 "mb_optimize_scan should be set to 0 or 1."); 2389 return -EINVAL; 2390 } 2391 return 0; 2392 } 2393 2394 /* 2395 * At this point we should only be getting options requiring MOPT_SET, 2396 * or MOPT_CLEAR. Anything else is a bug 2397 */ 2398 if (m->token == Opt_err) { 2399 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s", 2400 param->key); 2401 WARN_ON(1); 2402 return -EINVAL; 2403 } 2404 2405 else { 2406 unsigned int set = 0; 2407 2408 if ((param->type == fs_value_is_flag) || 2409 result.uint_32 > 0) 2410 set = 1; 2411 2412 if (m->flags & MOPT_CLEAR) 2413 set = !set; 2414 else if (unlikely(!(m->flags & MOPT_SET))) { 2415 ext4_msg(NULL, KERN_WARNING, 2416 "buggy handling of option %s", 2417 param->key); 2418 WARN_ON(1); 2419 return -EINVAL; 2420 } 2421 if (m->flags & MOPT_2) { 2422 if (set != 0) 2423 ctx_set_mount_opt2(ctx, m->mount_opt); 2424 else 2425 ctx_clear_mount_opt2(ctx, m->mount_opt); 2426 } else { 2427 if (set != 0) 2428 ctx_set_mount_opt(ctx, m->mount_opt); 2429 else 2430 ctx_clear_mount_opt(ctx, m->mount_opt); 2431 } 2432 } 2433 2434 return 0; 2435 } 2436 2437 static int parse_options(struct fs_context *fc, char *options) 2438 { 2439 struct fs_parameter param; 2440 int ret; 2441 char *key; 2442 2443 if (!options) 2444 return 0; 2445 2446 while ((key = strsep(&options, ",")) != NULL) { 2447 if (*key) { 2448 size_t v_len = 0; 2449 char *value = strchr(key, '='); 2450 2451 param.type = fs_value_is_flag; 2452 param.string = NULL; 2453 2454 if (value) { 2455 if (value == key) 2456 continue; 2457 2458 *value++ = 0; 2459 v_len = strlen(value); 2460 param.string = kmemdup_nul(value, v_len, 2461 GFP_KERNEL); 2462 if (!param.string) 2463 return -ENOMEM; 2464 param.type = fs_value_is_string; 2465 } 2466 2467 param.key = key; 2468 param.size = v_len; 2469 2470 ret = ext4_parse_param(fc, ¶m); 2471 if (param.string) 2472 kfree(param.string); 2473 if (ret < 0) 2474 return ret; 2475 } 2476 } 2477 2478 ret = ext4_validate_options(fc); 2479 if (ret < 0) 2480 return ret; 2481 2482 return 0; 2483 } 2484 2485 static int parse_apply_sb_mount_options(struct super_block *sb, 2486 struct ext4_fs_context *m_ctx) 2487 { 2488 struct ext4_sb_info *sbi = EXT4_SB(sb); 2489 char *s_mount_opts = NULL; 2490 struct ext4_fs_context *s_ctx = NULL; 2491 struct fs_context *fc = NULL; 2492 int ret = -ENOMEM; 2493 2494 if (!sbi->s_es->s_mount_opts[0]) 2495 return 0; 2496 2497 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts, 2498 sizeof(sbi->s_es->s_mount_opts), 2499 GFP_KERNEL); 2500 if (!s_mount_opts) 2501 return ret; 2502 2503 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL); 2504 if (!fc) 2505 goto out_free; 2506 2507 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL); 2508 if (!s_ctx) 2509 goto out_free; 2510 2511 fc->fs_private = s_ctx; 2512 fc->s_fs_info = sbi; 2513 2514 ret = parse_options(fc, s_mount_opts); 2515 if (ret < 0) 2516 goto parse_failed; 2517 2518 ret = ext4_check_opt_consistency(fc, sb); 2519 if (ret < 0) { 2520 parse_failed: 2521 ext4_msg(sb, KERN_WARNING, 2522 "failed to parse options in superblock: %s", 2523 s_mount_opts); 2524 ret = 0; 2525 goto out_free; 2526 } 2527 2528 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV) 2529 m_ctx->journal_devnum = s_ctx->journal_devnum; 2530 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO) 2531 m_ctx->journal_ioprio = s_ctx->journal_ioprio; 2532 2533 ext4_apply_options(fc, sb); 2534 ret = 0; 2535 2536 out_free: 2537 if (fc) { 2538 ext4_fc_free(fc); 2539 kfree(fc); 2540 } 2541 kfree(s_mount_opts); 2542 return ret; 2543 } 2544 2545 static void ext4_apply_quota_options(struct fs_context *fc, 2546 struct super_block *sb) 2547 { 2548 #ifdef CONFIG_QUOTA 2549 bool quota_feature = ext4_has_feature_quota(sb); 2550 struct ext4_fs_context *ctx = fc->fs_private; 2551 struct ext4_sb_info *sbi = EXT4_SB(sb); 2552 char *qname; 2553 int i; 2554 2555 if (quota_feature) 2556 return; 2557 2558 if (ctx->spec & EXT4_SPEC_JQUOTA) { 2559 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2560 if (!(ctx->qname_spec & (1 << i))) 2561 continue; 2562 2563 qname = ctx->s_qf_names[i]; /* May be NULL */ 2564 if (qname) 2565 set_opt(sb, QUOTA); 2566 ctx->s_qf_names[i] = NULL; 2567 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname, 2568 lockdep_is_held(&sb->s_umount)); 2569 if (qname) 2570 kfree_rcu_mightsleep(qname); 2571 } 2572 } 2573 2574 if (ctx->spec & EXT4_SPEC_JQFMT) 2575 sbi->s_jquota_fmt = ctx->s_jquota_fmt; 2576 #endif 2577 } 2578 2579 /* 2580 * Check quota settings consistency. 2581 */ 2582 static int ext4_check_quota_consistency(struct fs_context *fc, 2583 struct super_block *sb) 2584 { 2585 #ifdef CONFIG_QUOTA 2586 struct ext4_fs_context *ctx = fc->fs_private; 2587 struct ext4_sb_info *sbi = EXT4_SB(sb); 2588 bool quota_feature = ext4_has_feature_quota(sb); 2589 bool quota_loaded = sb_any_quota_loaded(sb); 2590 bool usr_qf_name, grp_qf_name, usrquota, grpquota; 2591 int quota_flags, i; 2592 2593 /* 2594 * We do the test below only for project quotas. 'usrquota' and 2595 * 'grpquota' mount options are allowed even without quota feature 2596 * to support legacy quotas in quota files. 2597 */ 2598 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) && 2599 !ext4_has_feature_project(sb)) { 2600 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. " 2601 "Cannot enable project quota enforcement."); 2602 return -EINVAL; 2603 } 2604 2605 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 2606 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA; 2607 if (quota_loaded && 2608 ctx->mask_s_mount_opt & quota_flags && 2609 !ctx_test_mount_opt(ctx, quota_flags)) 2610 goto err_quota_change; 2611 2612 if (ctx->spec & EXT4_SPEC_JQUOTA) { 2613 2614 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2615 if (!(ctx->qname_spec & (1 << i))) 2616 continue; 2617 2618 if (quota_loaded && 2619 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i]) 2620 goto err_jquota_change; 2621 2622 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] && 2623 strcmp(get_qf_name(sb, sbi, i), 2624 ctx->s_qf_names[i]) != 0) 2625 goto err_jquota_specified; 2626 } 2627 2628 if (quota_feature) { 2629 ext4_msg(NULL, KERN_INFO, 2630 "Journaled quota options ignored when " 2631 "QUOTA feature is enabled"); 2632 return 0; 2633 } 2634 } 2635 2636 if (ctx->spec & EXT4_SPEC_JQFMT) { 2637 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded) 2638 goto err_jquota_change; 2639 if (quota_feature) { 2640 ext4_msg(NULL, KERN_INFO, "Quota format mount options " 2641 "ignored when QUOTA feature is enabled"); 2642 return 0; 2643 } 2644 } 2645 2646 /* Make sure we don't mix old and new quota format */ 2647 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) || 2648 ctx->s_qf_names[USRQUOTA]); 2649 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) || 2650 ctx->s_qf_names[GRPQUOTA]); 2651 2652 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) || 2653 test_opt(sb, USRQUOTA)); 2654 2655 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) || 2656 test_opt(sb, GRPQUOTA)); 2657 2658 if (usr_qf_name) { 2659 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA); 2660 usrquota = false; 2661 } 2662 if (grp_qf_name) { 2663 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA); 2664 grpquota = false; 2665 } 2666 2667 if (usr_qf_name || grp_qf_name) { 2668 if (usrquota || grpquota) { 2669 ext4_msg(NULL, KERN_ERR, "old and new quota " 2670 "format mixing"); 2671 return -EINVAL; 2672 } 2673 2674 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) { 2675 ext4_msg(NULL, KERN_ERR, "journaled quota format " 2676 "not specified"); 2677 return -EINVAL; 2678 } 2679 } 2680 2681 return 0; 2682 2683 err_quota_change: 2684 ext4_msg(NULL, KERN_ERR, 2685 "Cannot change quota options when quota turned on"); 2686 return -EINVAL; 2687 err_jquota_change: 2688 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota " 2689 "options when quota turned on"); 2690 return -EINVAL; 2691 err_jquota_specified: 2692 ext4_msg(NULL, KERN_ERR, "%s quota file already specified", 2693 QTYPE2NAME(i)); 2694 return -EINVAL; 2695 #else 2696 return 0; 2697 #endif 2698 } 2699 2700 static int ext4_check_test_dummy_encryption(const struct fs_context *fc, 2701 struct super_block *sb) 2702 { 2703 const struct ext4_fs_context *ctx = fc->fs_private; 2704 const struct ext4_sb_info *sbi = EXT4_SB(sb); 2705 2706 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy)) 2707 return 0; 2708 2709 if (!ext4_has_feature_encrypt(sb)) { 2710 ext4_msg(NULL, KERN_WARNING, 2711 "test_dummy_encryption requires encrypt feature"); 2712 return -EINVAL; 2713 } 2714 /* 2715 * This mount option is just for testing, and it's not worthwhile to 2716 * implement the extra complexity (e.g. RCU protection) that would be 2717 * needed to allow it to be set or changed during remount. We do allow 2718 * it to be specified during remount, but only if there is no change. 2719 */ 2720 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) { 2721 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy, 2722 &ctx->dummy_enc_policy)) 2723 return 0; 2724 ext4_msg(NULL, KERN_WARNING, 2725 "Can't set or change test_dummy_encryption on remount"); 2726 return -EINVAL; 2727 } 2728 /* Also make sure s_mount_opts didn't contain a conflicting value. */ 2729 if (fscrypt_is_dummy_policy_set(&sbi->s_dummy_enc_policy)) { 2730 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy, 2731 &ctx->dummy_enc_policy)) 2732 return 0; 2733 ext4_msg(NULL, KERN_WARNING, 2734 "Conflicting test_dummy_encryption options"); 2735 return -EINVAL; 2736 } 2737 return 0; 2738 } 2739 2740 static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx, 2741 struct super_block *sb) 2742 { 2743 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy) || 2744 /* if already set, it was already verified to be the same */ 2745 fscrypt_is_dummy_policy_set(&EXT4_SB(sb)->s_dummy_enc_policy)) 2746 return; 2747 EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy; 2748 memset(&ctx->dummy_enc_policy, 0, sizeof(ctx->dummy_enc_policy)); 2749 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled"); 2750 } 2751 2752 static int ext4_check_opt_consistency(struct fs_context *fc, 2753 struct super_block *sb) 2754 { 2755 struct ext4_fs_context *ctx = fc->fs_private; 2756 struct ext4_sb_info *sbi = fc->s_fs_info; 2757 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE; 2758 int err; 2759 2760 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 2761 ext4_msg(NULL, KERN_ERR, 2762 "Mount option(s) incompatible with ext2"); 2763 return -EINVAL; 2764 } 2765 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 2766 ext4_msg(NULL, KERN_ERR, 2767 "Mount option(s) incompatible with ext3"); 2768 return -EINVAL; 2769 } 2770 2771 if (ctx->s_want_extra_isize > 2772 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) { 2773 ext4_msg(NULL, KERN_ERR, 2774 "Invalid want_extra_isize %d", 2775 ctx->s_want_extra_isize); 2776 return -EINVAL; 2777 } 2778 2779 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DIOREAD_NOLOCK)) { 2780 int blocksize = 2781 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 2782 if (blocksize < PAGE_SIZE) 2783 ext4_msg(NULL, KERN_WARNING, "Warning: mounting with an " 2784 "experimental mount option 'dioread_nolock' " 2785 "for blocksize < PAGE_SIZE"); 2786 } 2787 2788 err = ext4_check_test_dummy_encryption(fc, sb); 2789 if (err) 2790 return err; 2791 2792 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) { 2793 if (!sbi->s_journal) { 2794 ext4_msg(NULL, KERN_WARNING, 2795 "Remounting file system with no journal " 2796 "so ignoring journalled data option"); 2797 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS); 2798 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) != 2799 test_opt(sb, DATA_FLAGS)) { 2800 ext4_msg(NULL, KERN_ERR, "Cannot change data mode " 2801 "on remount"); 2802 return -EINVAL; 2803 } 2804 } 2805 2806 if (is_remount) { 2807 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) && 2808 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) { 2809 ext4_msg(NULL, KERN_ERR, "can't mount with " 2810 "both data=journal and dax"); 2811 return -EINVAL; 2812 } 2813 2814 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) && 2815 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2816 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) { 2817 fail_dax_change_remount: 2818 ext4_msg(NULL, KERN_ERR, "can't change " 2819 "dax mount option while remounting"); 2820 return -EINVAL; 2821 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) && 2822 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2823 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) { 2824 goto fail_dax_change_remount; 2825 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) && 2826 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2827 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2828 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) { 2829 goto fail_dax_change_remount; 2830 } 2831 } 2832 2833 return ext4_check_quota_consistency(fc, sb); 2834 } 2835 2836 static void ext4_apply_options(struct fs_context *fc, struct super_block *sb) 2837 { 2838 struct ext4_fs_context *ctx = fc->fs_private; 2839 struct ext4_sb_info *sbi = fc->s_fs_info; 2840 2841 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt; 2842 sbi->s_mount_opt |= ctx->vals_s_mount_opt; 2843 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2; 2844 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2; 2845 sbi->s_mount_flags &= ~ctx->mask_s_mount_flags; 2846 sbi->s_mount_flags |= ctx->vals_s_mount_flags; 2847 sb->s_flags &= ~ctx->mask_s_flags; 2848 sb->s_flags |= ctx->vals_s_flags; 2849 2850 #define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; }) 2851 APPLY(s_commit_interval); 2852 APPLY(s_stripe); 2853 APPLY(s_max_batch_time); 2854 APPLY(s_min_batch_time); 2855 APPLY(s_want_extra_isize); 2856 APPLY(s_inode_readahead_blks); 2857 APPLY(s_max_dir_size_kb); 2858 APPLY(s_li_wait_mult); 2859 APPLY(s_resgid); 2860 APPLY(s_resuid); 2861 2862 #ifdef CONFIG_EXT4_DEBUG 2863 APPLY(s_fc_debug_max_replay); 2864 #endif 2865 2866 ext4_apply_quota_options(fc, sb); 2867 ext4_apply_test_dummy_encryption(ctx, sb); 2868 } 2869 2870 2871 static int ext4_validate_options(struct fs_context *fc) 2872 { 2873 #ifdef CONFIG_QUOTA 2874 struct ext4_fs_context *ctx = fc->fs_private; 2875 char *usr_qf_name, *grp_qf_name; 2876 2877 usr_qf_name = ctx->s_qf_names[USRQUOTA]; 2878 grp_qf_name = ctx->s_qf_names[GRPQUOTA]; 2879 2880 if (usr_qf_name || grp_qf_name) { 2881 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name) 2882 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA); 2883 2884 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name) 2885 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA); 2886 2887 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) || 2888 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) { 2889 ext4_msg(NULL, KERN_ERR, "old and new quota " 2890 "format mixing"); 2891 return -EINVAL; 2892 } 2893 } 2894 #endif 2895 return 1; 2896 } 2897 2898 static inline void ext4_show_quota_options(struct seq_file *seq, 2899 struct super_block *sb) 2900 { 2901 #if defined(CONFIG_QUOTA) 2902 struct ext4_sb_info *sbi = EXT4_SB(sb); 2903 char *usr_qf_name, *grp_qf_name; 2904 2905 if (sbi->s_jquota_fmt) { 2906 char *fmtname = ""; 2907 2908 switch (sbi->s_jquota_fmt) { 2909 case QFMT_VFS_OLD: 2910 fmtname = "vfsold"; 2911 break; 2912 case QFMT_VFS_V0: 2913 fmtname = "vfsv0"; 2914 break; 2915 case QFMT_VFS_V1: 2916 fmtname = "vfsv1"; 2917 break; 2918 } 2919 seq_printf(seq, ",jqfmt=%s", fmtname); 2920 } 2921 2922 rcu_read_lock(); 2923 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); 2924 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); 2925 if (usr_qf_name) 2926 seq_show_option(seq, "usrjquota", usr_qf_name); 2927 if (grp_qf_name) 2928 seq_show_option(seq, "grpjquota", grp_qf_name); 2929 rcu_read_unlock(); 2930 #endif 2931 } 2932 2933 static const char *token2str(int token) 2934 { 2935 const struct fs_parameter_spec *spec; 2936 2937 for (spec = ext4_param_specs; spec->name != NULL; spec++) 2938 if (spec->opt == token && !spec->type) 2939 break; 2940 return spec->name; 2941 } 2942 2943 /* 2944 * Show an option if 2945 * - it's set to a non-default value OR 2946 * - if the per-sb default is different from the global default 2947 */ 2948 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 2949 int nodefs) 2950 { 2951 struct ext4_sb_info *sbi = EXT4_SB(sb); 2952 struct ext4_super_block *es = sbi->s_es; 2953 int def_errors; 2954 const struct mount_opts *m; 2955 char sep = nodefs ? '\n' : ','; 2956 2957 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 2958 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 2959 2960 if (sbi->s_sb_block != 1) 2961 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 2962 2963 for (m = ext4_mount_opts; m->token != Opt_err; m++) { 2964 int want_set = m->flags & MOPT_SET; 2965 int opt_2 = m->flags & MOPT_2; 2966 unsigned int mount_opt, def_mount_opt; 2967 2968 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 2969 m->flags & MOPT_SKIP) 2970 continue; 2971 2972 if (opt_2) { 2973 mount_opt = sbi->s_mount_opt2; 2974 def_mount_opt = sbi->s_def_mount_opt2; 2975 } else { 2976 mount_opt = sbi->s_mount_opt; 2977 def_mount_opt = sbi->s_def_mount_opt; 2978 } 2979 /* skip if same as the default */ 2980 if (!nodefs && !(m->mount_opt & (mount_opt ^ def_mount_opt))) 2981 continue; 2982 /* select Opt_noFoo vs Opt_Foo */ 2983 if ((want_set && 2984 (mount_opt & m->mount_opt) != m->mount_opt) || 2985 (!want_set && (mount_opt & m->mount_opt))) 2986 continue; 2987 SEQ_OPTS_PRINT("%s", token2str(m->token)); 2988 } 2989 2990 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 2991 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 2992 SEQ_OPTS_PRINT("resuid=%u", 2993 from_kuid_munged(&init_user_ns, sbi->s_resuid)); 2994 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 2995 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 2996 SEQ_OPTS_PRINT("resgid=%u", 2997 from_kgid_munged(&init_user_ns, sbi->s_resgid)); 2998 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 2999 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 3000 SEQ_OPTS_PUTS("errors=remount-ro"); 3001 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 3002 SEQ_OPTS_PUTS("errors=continue"); 3003 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 3004 SEQ_OPTS_PUTS("errors=panic"); 3005 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 3006 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 3007 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 3008 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 3009 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 3010 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 3011 if (nodefs || sbi->s_stripe) 3012 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 3013 if (nodefs || EXT4_MOUNT_DATA_FLAGS & 3014 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 3015 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 3016 SEQ_OPTS_PUTS("data=journal"); 3017 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 3018 SEQ_OPTS_PUTS("data=ordered"); 3019 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 3020 SEQ_OPTS_PUTS("data=writeback"); 3021 } 3022 if (nodefs || 3023 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 3024 SEQ_OPTS_PRINT("inode_readahead_blks=%u", 3025 sbi->s_inode_readahead_blks); 3026 3027 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs || 3028 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 3029 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 3030 if (nodefs || sbi->s_max_dir_size_kb) 3031 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 3032 if (test_opt(sb, DATA_ERR_ABORT)) 3033 SEQ_OPTS_PUTS("data_err=abort"); 3034 3035 fscrypt_show_test_dummy_encryption(seq, sep, sb); 3036 3037 if (sb->s_flags & SB_INLINECRYPT) 3038 SEQ_OPTS_PUTS("inlinecrypt"); 3039 3040 if (test_opt(sb, DAX_ALWAYS)) { 3041 if (IS_EXT2_SB(sb)) 3042 SEQ_OPTS_PUTS("dax"); 3043 else 3044 SEQ_OPTS_PUTS("dax=always"); 3045 } else if (test_opt2(sb, DAX_NEVER)) { 3046 SEQ_OPTS_PUTS("dax=never"); 3047 } else if (test_opt2(sb, DAX_INODE)) { 3048 SEQ_OPTS_PUTS("dax=inode"); 3049 } 3050 3051 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD && 3052 !test_opt2(sb, MB_OPTIMIZE_SCAN)) { 3053 SEQ_OPTS_PUTS("mb_optimize_scan=0"); 3054 } else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD && 3055 test_opt2(sb, MB_OPTIMIZE_SCAN)) { 3056 SEQ_OPTS_PUTS("mb_optimize_scan=1"); 3057 } 3058 3059 ext4_show_quota_options(seq, sb); 3060 return 0; 3061 } 3062 3063 static int ext4_show_options(struct seq_file *seq, struct dentry *root) 3064 { 3065 return _ext4_show_options(seq, root->d_sb, 0); 3066 } 3067 3068 int ext4_seq_options_show(struct seq_file *seq, void *offset) 3069 { 3070 struct super_block *sb = seq->private; 3071 int rc; 3072 3073 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); 3074 rc = _ext4_show_options(seq, sb, 1); 3075 seq_puts(seq, "\n"); 3076 return rc; 3077 } 3078 3079 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 3080 int read_only) 3081 { 3082 struct ext4_sb_info *sbi = EXT4_SB(sb); 3083 int err = 0; 3084 3085 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 3086 ext4_msg(sb, KERN_ERR, "revision level too high, " 3087 "forcing read-only mode"); 3088 err = -EROFS; 3089 goto done; 3090 } 3091 if (read_only) 3092 goto done; 3093 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 3094 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 3095 "running e2fsck is recommended"); 3096 else if (sbi->s_mount_state & EXT4_ERROR_FS) 3097 ext4_msg(sb, KERN_WARNING, 3098 "warning: mounting fs with errors, " 3099 "running e2fsck is recommended"); 3100 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 3101 le16_to_cpu(es->s_mnt_count) >= 3102 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 3103 ext4_msg(sb, KERN_WARNING, 3104 "warning: maximal mount count reached, " 3105 "running e2fsck is recommended"); 3106 else if (le32_to_cpu(es->s_checkinterval) && 3107 (ext4_get_tstamp(es, s_lastcheck) + 3108 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds())) 3109 ext4_msg(sb, KERN_WARNING, 3110 "warning: checktime reached, " 3111 "running e2fsck is recommended"); 3112 if (!sbi->s_journal) 3113 es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 3114 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 3115 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 3116 le16_add_cpu(&es->s_mnt_count, 1); 3117 ext4_update_tstamp(es, s_mtime); 3118 if (sbi->s_journal) { 3119 ext4_set_feature_journal_needs_recovery(sb); 3120 if (ext4_has_feature_orphan_file(sb)) 3121 ext4_set_feature_orphan_present(sb); 3122 } 3123 3124 err = ext4_commit_super(sb); 3125 done: 3126 if (test_opt(sb, DEBUG)) 3127 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 3128 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 3129 sb->s_blocksize, 3130 sbi->s_groups_count, 3131 EXT4_BLOCKS_PER_GROUP(sb), 3132 EXT4_INODES_PER_GROUP(sb), 3133 sbi->s_mount_opt, sbi->s_mount_opt2); 3134 return err; 3135 } 3136 3137 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 3138 { 3139 struct ext4_sb_info *sbi = EXT4_SB(sb); 3140 struct flex_groups **old_groups, **new_groups; 3141 int size, i, j; 3142 3143 if (!sbi->s_log_groups_per_flex) 3144 return 0; 3145 3146 size = ext4_flex_group(sbi, ngroup - 1) + 1; 3147 if (size <= sbi->s_flex_groups_allocated) 3148 return 0; 3149 3150 new_groups = kvzalloc(roundup_pow_of_two(size * 3151 sizeof(*sbi->s_flex_groups)), GFP_KERNEL); 3152 if (!new_groups) { 3153 ext4_msg(sb, KERN_ERR, 3154 "not enough memory for %d flex group pointers", size); 3155 return -ENOMEM; 3156 } 3157 for (i = sbi->s_flex_groups_allocated; i < size; i++) { 3158 new_groups[i] = kvzalloc(roundup_pow_of_two( 3159 sizeof(struct flex_groups)), 3160 GFP_KERNEL); 3161 if (!new_groups[i]) { 3162 for (j = sbi->s_flex_groups_allocated; j < i; j++) 3163 kvfree(new_groups[j]); 3164 kvfree(new_groups); 3165 ext4_msg(sb, KERN_ERR, 3166 "not enough memory for %d flex groups", size); 3167 return -ENOMEM; 3168 } 3169 } 3170 rcu_read_lock(); 3171 old_groups = rcu_dereference(sbi->s_flex_groups); 3172 if (old_groups) 3173 memcpy(new_groups, old_groups, 3174 (sbi->s_flex_groups_allocated * 3175 sizeof(struct flex_groups *))); 3176 rcu_read_unlock(); 3177 rcu_assign_pointer(sbi->s_flex_groups, new_groups); 3178 sbi->s_flex_groups_allocated = size; 3179 if (old_groups) 3180 ext4_kvfree_array_rcu(old_groups); 3181 return 0; 3182 } 3183 3184 static int ext4_fill_flex_info(struct super_block *sb) 3185 { 3186 struct ext4_sb_info *sbi = EXT4_SB(sb); 3187 struct ext4_group_desc *gdp = NULL; 3188 struct flex_groups *fg; 3189 ext4_group_t flex_group; 3190 int i, err; 3191 3192 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 3193 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 3194 sbi->s_log_groups_per_flex = 0; 3195 return 1; 3196 } 3197 3198 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 3199 if (err) 3200 goto failed; 3201 3202 for (i = 0; i < sbi->s_groups_count; i++) { 3203 gdp = ext4_get_group_desc(sb, i, NULL); 3204 3205 flex_group = ext4_flex_group(sbi, i); 3206 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group); 3207 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes); 3208 atomic64_add(ext4_free_group_clusters(sb, gdp), 3209 &fg->free_clusters); 3210 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs); 3211 } 3212 3213 return 1; 3214 failed: 3215 return 0; 3216 } 3217 3218 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 3219 struct ext4_group_desc *gdp) 3220 { 3221 int offset = offsetof(struct ext4_group_desc, bg_checksum); 3222 __u16 crc = 0; 3223 __le32 le_group = cpu_to_le32(block_group); 3224 struct ext4_sb_info *sbi = EXT4_SB(sb); 3225 3226 if (ext4_has_metadata_csum(sbi->s_sb)) { 3227 /* Use new metadata_csum algorithm */ 3228 __u32 csum32; 3229 __u16 dummy_csum = 0; 3230 3231 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 3232 sizeof(le_group)); 3233 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); 3234 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, 3235 sizeof(dummy_csum)); 3236 offset += sizeof(dummy_csum); 3237 if (offset < sbi->s_desc_size) 3238 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, 3239 sbi->s_desc_size - offset); 3240 3241 crc = csum32 & 0xFFFF; 3242 goto out; 3243 } 3244 3245 /* old crc16 code */ 3246 if (!ext4_has_feature_gdt_csum(sb)) 3247 return 0; 3248 3249 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 3250 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 3251 crc = crc16(crc, (__u8 *)gdp, offset); 3252 offset += sizeof(gdp->bg_checksum); /* skip checksum */ 3253 /* for checksum of struct ext4_group_desc do the rest...*/ 3254 if (ext4_has_feature_64bit(sb) && offset < sbi->s_desc_size) 3255 crc = crc16(crc, (__u8 *)gdp + offset, 3256 sbi->s_desc_size - offset); 3257 3258 out: 3259 return cpu_to_le16(crc); 3260 } 3261 3262 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 3263 struct ext4_group_desc *gdp) 3264 { 3265 if (ext4_has_group_desc_csum(sb) && 3266 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 3267 return 0; 3268 3269 return 1; 3270 } 3271 3272 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 3273 struct ext4_group_desc *gdp) 3274 { 3275 if (!ext4_has_group_desc_csum(sb)) 3276 return; 3277 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 3278 } 3279 3280 /* Called at mount-time, super-block is locked */ 3281 static int ext4_check_descriptors(struct super_block *sb, 3282 ext4_fsblk_t sb_block, 3283 ext4_group_t *first_not_zeroed) 3284 { 3285 struct ext4_sb_info *sbi = EXT4_SB(sb); 3286 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 3287 ext4_fsblk_t last_block; 3288 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0); 3289 ext4_fsblk_t block_bitmap; 3290 ext4_fsblk_t inode_bitmap; 3291 ext4_fsblk_t inode_table; 3292 int flexbg_flag = 0; 3293 ext4_group_t i, grp = sbi->s_groups_count; 3294 3295 if (ext4_has_feature_flex_bg(sb)) 3296 flexbg_flag = 1; 3297 3298 ext4_debug("Checking group descriptors"); 3299 3300 for (i = 0; i < sbi->s_groups_count; i++) { 3301 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 3302 3303 if (i == sbi->s_groups_count - 1 || flexbg_flag) 3304 last_block = ext4_blocks_count(sbi->s_es) - 1; 3305 else 3306 last_block = first_block + 3307 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 3308 3309 if ((grp == sbi->s_groups_count) && 3310 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3311 grp = i; 3312 3313 block_bitmap = ext4_block_bitmap(sb, gdp); 3314 if (block_bitmap == sb_block) { 3315 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3316 "Block bitmap for group %u overlaps " 3317 "superblock", i); 3318 if (!sb_rdonly(sb)) 3319 return 0; 3320 } 3321 if (block_bitmap >= sb_block + 1 && 3322 block_bitmap <= last_bg_block) { 3323 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3324 "Block bitmap for group %u overlaps " 3325 "block group descriptors", i); 3326 if (!sb_rdonly(sb)) 3327 return 0; 3328 } 3329 if (block_bitmap < first_block || block_bitmap > last_block) { 3330 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3331 "Block bitmap for group %u not in group " 3332 "(block %llu)!", i, block_bitmap); 3333 return 0; 3334 } 3335 inode_bitmap = ext4_inode_bitmap(sb, gdp); 3336 if (inode_bitmap == sb_block) { 3337 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3338 "Inode bitmap for group %u overlaps " 3339 "superblock", i); 3340 if (!sb_rdonly(sb)) 3341 return 0; 3342 } 3343 if (inode_bitmap >= sb_block + 1 && 3344 inode_bitmap <= last_bg_block) { 3345 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3346 "Inode bitmap for group %u overlaps " 3347 "block group descriptors", i); 3348 if (!sb_rdonly(sb)) 3349 return 0; 3350 } 3351 if (inode_bitmap < first_block || inode_bitmap > last_block) { 3352 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3353 "Inode bitmap for group %u not in group " 3354 "(block %llu)!", i, inode_bitmap); 3355 return 0; 3356 } 3357 inode_table = ext4_inode_table(sb, gdp); 3358 if (inode_table == sb_block) { 3359 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3360 "Inode table for group %u overlaps " 3361 "superblock", i); 3362 if (!sb_rdonly(sb)) 3363 return 0; 3364 } 3365 if (inode_table >= sb_block + 1 && 3366 inode_table <= last_bg_block) { 3367 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3368 "Inode table for group %u overlaps " 3369 "block group descriptors", i); 3370 if (!sb_rdonly(sb)) 3371 return 0; 3372 } 3373 if (inode_table < first_block || 3374 inode_table + sbi->s_itb_per_group - 1 > last_block) { 3375 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3376 "Inode table for group %u not in group " 3377 "(block %llu)!", i, inode_table); 3378 return 0; 3379 } 3380 ext4_lock_group(sb, i); 3381 if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 3382 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3383 "Checksum for group %u failed (%u!=%u)", 3384 i, le16_to_cpu(ext4_group_desc_csum(sb, i, 3385 gdp)), le16_to_cpu(gdp->bg_checksum)); 3386 if (!sb_rdonly(sb)) { 3387 ext4_unlock_group(sb, i); 3388 return 0; 3389 } 3390 } 3391 ext4_unlock_group(sb, i); 3392 if (!flexbg_flag) 3393 first_block += EXT4_BLOCKS_PER_GROUP(sb); 3394 } 3395 if (NULL != first_not_zeroed) 3396 *first_not_zeroed = grp; 3397 return 1; 3398 } 3399 3400 /* 3401 * Maximal extent format file size. 3402 * Resulting logical blkno at s_maxbytes must fit in our on-disk 3403 * extent format containers, within a sector_t, and within i_blocks 3404 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 3405 * so that won't be a limiting factor. 3406 * 3407 * However there is other limiting factor. We do store extents in the form 3408 * of starting block and length, hence the resulting length of the extent 3409 * covering maximum file size must fit into on-disk format containers as 3410 * well. Given that length is always by 1 unit bigger than max unit (because 3411 * we count 0 as well) we have to lower the s_maxbytes by one fs block. 3412 * 3413 * Note, this does *not* consider any metadata overhead for vfs i_blocks. 3414 */ 3415 static loff_t ext4_max_size(int blkbits, int has_huge_files) 3416 { 3417 loff_t res; 3418 loff_t upper_limit = MAX_LFS_FILESIZE; 3419 3420 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64)); 3421 3422 if (!has_huge_files) { 3423 upper_limit = (1LL << 32) - 1; 3424 3425 /* total blocks in file system block size */ 3426 upper_limit >>= (blkbits - 9); 3427 upper_limit <<= blkbits; 3428 } 3429 3430 /* 3431 * 32-bit extent-start container, ee_block. We lower the maxbytes 3432 * by one fs block, so ee_len can cover the extent of maximum file 3433 * size 3434 */ 3435 res = (1LL << 32) - 1; 3436 res <<= blkbits; 3437 3438 /* Sanity check against vm- & vfs- imposed limits */ 3439 if (res > upper_limit) 3440 res = upper_limit; 3441 3442 return res; 3443 } 3444 3445 /* 3446 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 3447 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 3448 * We need to be 1 filesystem block less than the 2^48 sector limit. 3449 */ 3450 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 3451 { 3452 loff_t upper_limit, res = EXT4_NDIR_BLOCKS; 3453 int meta_blocks; 3454 unsigned int ppb = 1 << (bits - 2); 3455 3456 /* 3457 * This is calculated to be the largest file size for a dense, block 3458 * mapped file such that the file's total number of 512-byte sectors, 3459 * including data and all indirect blocks, does not exceed (2^48 - 1). 3460 * 3461 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 3462 * number of 512-byte sectors of the file. 3463 */ 3464 if (!has_huge_files) { 3465 /* 3466 * !has_huge_files or implies that the inode i_block field 3467 * represents total file blocks in 2^32 512-byte sectors == 3468 * size of vfs inode i_blocks * 8 3469 */ 3470 upper_limit = (1LL << 32) - 1; 3471 3472 /* total blocks in file system block size */ 3473 upper_limit >>= (bits - 9); 3474 3475 } else { 3476 /* 3477 * We use 48 bit ext4_inode i_blocks 3478 * With EXT4_HUGE_FILE_FL set the i_blocks 3479 * represent total number of blocks in 3480 * file system block size 3481 */ 3482 upper_limit = (1LL << 48) - 1; 3483 3484 } 3485 3486 /* Compute how many blocks we can address by block tree */ 3487 res += ppb; 3488 res += ppb * ppb; 3489 res += ((loff_t)ppb) * ppb * ppb; 3490 /* Compute how many metadata blocks are needed */ 3491 meta_blocks = 1; 3492 meta_blocks += 1 + ppb; 3493 meta_blocks += 1 + ppb + ppb * ppb; 3494 /* Does block tree limit file size? */ 3495 if (res + meta_blocks <= upper_limit) 3496 goto check_lfs; 3497 3498 res = upper_limit; 3499 /* How many metadata blocks are needed for addressing upper_limit? */ 3500 upper_limit -= EXT4_NDIR_BLOCKS; 3501 /* indirect blocks */ 3502 meta_blocks = 1; 3503 upper_limit -= ppb; 3504 /* double indirect blocks */ 3505 if (upper_limit < ppb * ppb) { 3506 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb); 3507 res -= meta_blocks; 3508 goto check_lfs; 3509 } 3510 meta_blocks += 1 + ppb; 3511 upper_limit -= ppb * ppb; 3512 /* tripple indirect blocks for the rest */ 3513 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) + 3514 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb); 3515 res -= meta_blocks; 3516 check_lfs: 3517 res <<= bits; 3518 if (res > MAX_LFS_FILESIZE) 3519 res = MAX_LFS_FILESIZE; 3520 3521 return res; 3522 } 3523 3524 static ext4_fsblk_t descriptor_loc(struct super_block *sb, 3525 ext4_fsblk_t logical_sb_block, int nr) 3526 { 3527 struct ext4_sb_info *sbi = EXT4_SB(sb); 3528 ext4_group_t bg, first_meta_bg; 3529 int has_super = 0; 3530 3531 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 3532 3533 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 3534 return logical_sb_block + nr + 1; 3535 bg = sbi->s_desc_per_block * nr; 3536 if (ext4_bg_has_super(sb, bg)) 3537 has_super = 1; 3538 3539 /* 3540 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 3541 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 3542 * on modern mke2fs or blksize > 1k on older mke2fs) then we must 3543 * compensate. 3544 */ 3545 if (sb->s_blocksize == 1024 && nr == 0 && 3546 le32_to_cpu(sbi->s_es->s_first_data_block) == 0) 3547 has_super++; 3548 3549 return (has_super + ext4_group_first_block_no(sb, bg)); 3550 } 3551 3552 /** 3553 * ext4_get_stripe_size: Get the stripe size. 3554 * @sbi: In memory super block info 3555 * 3556 * If we have specified it via mount option, then 3557 * use the mount option value. If the value specified at mount time is 3558 * greater than the blocks per group use the super block value. 3559 * If the super block value is greater than blocks per group return 0. 3560 * Allocator needs it be less than blocks per group. 3561 * 3562 */ 3563 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 3564 { 3565 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 3566 unsigned long stripe_width = 3567 le32_to_cpu(sbi->s_es->s_raid_stripe_width); 3568 int ret; 3569 3570 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 3571 ret = sbi->s_stripe; 3572 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) 3573 ret = stripe_width; 3574 else if (stride && stride <= sbi->s_blocks_per_group) 3575 ret = stride; 3576 else 3577 ret = 0; 3578 3579 /* 3580 * If the stripe width is 1, this makes no sense and 3581 * we set it to 0 to turn off stripe handling code. 3582 */ 3583 if (ret <= 1) 3584 ret = 0; 3585 3586 return ret; 3587 } 3588 3589 /* 3590 * Check whether this filesystem can be mounted based on 3591 * the features present and the RDONLY/RDWR mount requested. 3592 * Returns 1 if this filesystem can be mounted as requested, 3593 * 0 if it cannot be. 3594 */ 3595 int ext4_feature_set_ok(struct super_block *sb, int readonly) 3596 { 3597 if (ext4_has_unknown_ext4_incompat_features(sb)) { 3598 ext4_msg(sb, KERN_ERR, 3599 "Couldn't mount because of " 3600 "unsupported optional features (%x)", 3601 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 3602 ~EXT4_FEATURE_INCOMPAT_SUPP)); 3603 return 0; 3604 } 3605 3606 #if !IS_ENABLED(CONFIG_UNICODE) 3607 if (ext4_has_feature_casefold(sb)) { 3608 ext4_msg(sb, KERN_ERR, 3609 "Filesystem with casefold feature cannot be " 3610 "mounted without CONFIG_UNICODE"); 3611 return 0; 3612 } 3613 #endif 3614 3615 if (readonly) 3616 return 1; 3617 3618 if (ext4_has_feature_readonly(sb)) { 3619 ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 3620 sb->s_flags |= SB_RDONLY; 3621 return 1; 3622 } 3623 3624 /* Check that feature set is OK for a read-write mount */ 3625 if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 3626 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 3627 "unsupported optional features (%x)", 3628 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 3629 ~EXT4_FEATURE_RO_COMPAT_SUPP)); 3630 return 0; 3631 } 3632 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 3633 ext4_msg(sb, KERN_ERR, 3634 "Can't support bigalloc feature without " 3635 "extents feature\n"); 3636 return 0; 3637 } 3638 3639 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2) 3640 if (!readonly && (ext4_has_feature_quota(sb) || 3641 ext4_has_feature_project(sb))) { 3642 ext4_msg(sb, KERN_ERR, 3643 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2"); 3644 return 0; 3645 } 3646 #endif /* CONFIG_QUOTA */ 3647 return 1; 3648 } 3649 3650 /* 3651 * This function is called once a day if we have errors logged 3652 * on the file system 3653 */ 3654 static void print_daily_error_info(struct timer_list *t) 3655 { 3656 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report); 3657 struct super_block *sb = sbi->s_sb; 3658 struct ext4_super_block *es = sbi->s_es; 3659 3660 if (es->s_error_count) 3661 /* fsck newer than v1.41.13 is needed to clean this condition. */ 3662 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 3663 le32_to_cpu(es->s_error_count)); 3664 if (es->s_first_error_time) { 3665 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d", 3666 sb->s_id, 3667 ext4_get_tstamp(es, s_first_error_time), 3668 (int) sizeof(es->s_first_error_func), 3669 es->s_first_error_func, 3670 le32_to_cpu(es->s_first_error_line)); 3671 if (es->s_first_error_ino) 3672 printk(KERN_CONT ": inode %u", 3673 le32_to_cpu(es->s_first_error_ino)); 3674 if (es->s_first_error_block) 3675 printk(KERN_CONT ": block %llu", (unsigned long long) 3676 le64_to_cpu(es->s_first_error_block)); 3677 printk(KERN_CONT "\n"); 3678 } 3679 if (es->s_last_error_time) { 3680 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d", 3681 sb->s_id, 3682 ext4_get_tstamp(es, s_last_error_time), 3683 (int) sizeof(es->s_last_error_func), 3684 es->s_last_error_func, 3685 le32_to_cpu(es->s_last_error_line)); 3686 if (es->s_last_error_ino) 3687 printk(KERN_CONT ": inode %u", 3688 le32_to_cpu(es->s_last_error_ino)); 3689 if (es->s_last_error_block) 3690 printk(KERN_CONT ": block %llu", (unsigned long long) 3691 le64_to_cpu(es->s_last_error_block)); 3692 printk(KERN_CONT "\n"); 3693 } 3694 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 3695 } 3696 3697 /* Find next suitable group and run ext4_init_inode_table */ 3698 static int ext4_run_li_request(struct ext4_li_request *elr) 3699 { 3700 struct ext4_group_desc *gdp = NULL; 3701 struct super_block *sb = elr->lr_super; 3702 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 3703 ext4_group_t group = elr->lr_next_group; 3704 unsigned int prefetch_ios = 0; 3705 int ret = 0; 3706 u64 start_time; 3707 3708 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) { 3709 elr->lr_next_group = ext4_mb_prefetch(sb, group, 3710 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios); 3711 if (prefetch_ios) 3712 ext4_mb_prefetch_fini(sb, elr->lr_next_group, 3713 prefetch_ios); 3714 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, 3715 prefetch_ios); 3716 if (group >= elr->lr_next_group) { 3717 ret = 1; 3718 if (elr->lr_first_not_zeroed != ngroups && 3719 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) { 3720 elr->lr_next_group = elr->lr_first_not_zeroed; 3721 elr->lr_mode = EXT4_LI_MODE_ITABLE; 3722 ret = 0; 3723 } 3724 } 3725 return ret; 3726 } 3727 3728 for (; group < ngroups; group++) { 3729 gdp = ext4_get_group_desc(sb, group, NULL); 3730 if (!gdp) { 3731 ret = 1; 3732 break; 3733 } 3734 3735 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3736 break; 3737 } 3738 3739 if (group >= ngroups) 3740 ret = 1; 3741 3742 if (!ret) { 3743 start_time = ktime_get_real_ns(); 3744 ret = ext4_init_inode_table(sb, group, 3745 elr->lr_timeout ? 0 : 1); 3746 trace_ext4_lazy_itable_init(sb, group); 3747 if (elr->lr_timeout == 0) { 3748 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) * 3749 EXT4_SB(elr->lr_super)->s_li_wait_mult); 3750 } 3751 elr->lr_next_sched = jiffies + elr->lr_timeout; 3752 elr->lr_next_group = group + 1; 3753 } 3754 return ret; 3755 } 3756 3757 /* 3758 * Remove lr_request from the list_request and free the 3759 * request structure. Should be called with li_list_mtx held 3760 */ 3761 static void ext4_remove_li_request(struct ext4_li_request *elr) 3762 { 3763 if (!elr) 3764 return; 3765 3766 list_del(&elr->lr_request); 3767 EXT4_SB(elr->lr_super)->s_li_request = NULL; 3768 kfree(elr); 3769 } 3770 3771 static void ext4_unregister_li_request(struct super_block *sb) 3772 { 3773 mutex_lock(&ext4_li_mtx); 3774 if (!ext4_li_info) { 3775 mutex_unlock(&ext4_li_mtx); 3776 return; 3777 } 3778 3779 mutex_lock(&ext4_li_info->li_list_mtx); 3780 ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 3781 mutex_unlock(&ext4_li_info->li_list_mtx); 3782 mutex_unlock(&ext4_li_mtx); 3783 } 3784 3785 static struct task_struct *ext4_lazyinit_task; 3786 3787 /* 3788 * This is the function where ext4lazyinit thread lives. It walks 3789 * through the request list searching for next scheduled filesystem. 3790 * When such a fs is found, run the lazy initialization request 3791 * (ext4_rn_li_request) and keep track of the time spend in this 3792 * function. Based on that time we compute next schedule time of 3793 * the request. When walking through the list is complete, compute 3794 * next waking time and put itself into sleep. 3795 */ 3796 static int ext4_lazyinit_thread(void *arg) 3797 { 3798 struct ext4_lazy_init *eli = arg; 3799 struct list_head *pos, *n; 3800 struct ext4_li_request *elr; 3801 unsigned long next_wakeup, cur; 3802 3803 BUG_ON(NULL == eli); 3804 set_freezable(); 3805 3806 cont_thread: 3807 while (true) { 3808 next_wakeup = MAX_JIFFY_OFFSET; 3809 3810 mutex_lock(&eli->li_list_mtx); 3811 if (list_empty(&eli->li_request_list)) { 3812 mutex_unlock(&eli->li_list_mtx); 3813 goto exit_thread; 3814 } 3815 list_for_each_safe(pos, n, &eli->li_request_list) { 3816 int err = 0; 3817 int progress = 0; 3818 elr = list_entry(pos, struct ext4_li_request, 3819 lr_request); 3820 3821 if (time_before(jiffies, elr->lr_next_sched)) { 3822 if (time_before(elr->lr_next_sched, next_wakeup)) 3823 next_wakeup = elr->lr_next_sched; 3824 continue; 3825 } 3826 if (down_read_trylock(&elr->lr_super->s_umount)) { 3827 if (sb_start_write_trylock(elr->lr_super)) { 3828 progress = 1; 3829 /* 3830 * We hold sb->s_umount, sb can not 3831 * be removed from the list, it is 3832 * now safe to drop li_list_mtx 3833 */ 3834 mutex_unlock(&eli->li_list_mtx); 3835 err = ext4_run_li_request(elr); 3836 sb_end_write(elr->lr_super); 3837 mutex_lock(&eli->li_list_mtx); 3838 n = pos->next; 3839 } 3840 up_read((&elr->lr_super->s_umount)); 3841 } 3842 /* error, remove the lazy_init job */ 3843 if (err) { 3844 ext4_remove_li_request(elr); 3845 continue; 3846 } 3847 if (!progress) { 3848 elr->lr_next_sched = jiffies + 3849 get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ); 3850 } 3851 if (time_before(elr->lr_next_sched, next_wakeup)) 3852 next_wakeup = elr->lr_next_sched; 3853 } 3854 mutex_unlock(&eli->li_list_mtx); 3855 3856 try_to_freeze(); 3857 3858 cur = jiffies; 3859 if ((time_after_eq(cur, next_wakeup)) || 3860 (MAX_JIFFY_OFFSET == next_wakeup)) { 3861 cond_resched(); 3862 continue; 3863 } 3864 3865 schedule_timeout_interruptible(next_wakeup - cur); 3866 3867 if (kthread_should_stop()) { 3868 ext4_clear_request_list(); 3869 goto exit_thread; 3870 } 3871 } 3872 3873 exit_thread: 3874 /* 3875 * It looks like the request list is empty, but we need 3876 * to check it under the li_list_mtx lock, to prevent any 3877 * additions into it, and of course we should lock ext4_li_mtx 3878 * to atomically free the list and ext4_li_info, because at 3879 * this point another ext4 filesystem could be registering 3880 * new one. 3881 */ 3882 mutex_lock(&ext4_li_mtx); 3883 mutex_lock(&eli->li_list_mtx); 3884 if (!list_empty(&eli->li_request_list)) { 3885 mutex_unlock(&eli->li_list_mtx); 3886 mutex_unlock(&ext4_li_mtx); 3887 goto cont_thread; 3888 } 3889 mutex_unlock(&eli->li_list_mtx); 3890 kfree(ext4_li_info); 3891 ext4_li_info = NULL; 3892 mutex_unlock(&ext4_li_mtx); 3893 3894 return 0; 3895 } 3896 3897 static void ext4_clear_request_list(void) 3898 { 3899 struct list_head *pos, *n; 3900 struct ext4_li_request *elr; 3901 3902 mutex_lock(&ext4_li_info->li_list_mtx); 3903 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3904 elr = list_entry(pos, struct ext4_li_request, 3905 lr_request); 3906 ext4_remove_li_request(elr); 3907 } 3908 mutex_unlock(&ext4_li_info->li_list_mtx); 3909 } 3910 3911 static int ext4_run_lazyinit_thread(void) 3912 { 3913 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 3914 ext4_li_info, "ext4lazyinit"); 3915 if (IS_ERR(ext4_lazyinit_task)) { 3916 int err = PTR_ERR(ext4_lazyinit_task); 3917 ext4_clear_request_list(); 3918 kfree(ext4_li_info); 3919 ext4_li_info = NULL; 3920 printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3921 "initialization thread\n", 3922 err); 3923 return err; 3924 } 3925 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3926 return 0; 3927 } 3928 3929 /* 3930 * Check whether it make sense to run itable init. thread or not. 3931 * If there is at least one uninitialized inode table, return 3932 * corresponding group number, else the loop goes through all 3933 * groups and return total number of groups. 3934 */ 3935 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3936 { 3937 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3938 struct ext4_group_desc *gdp = NULL; 3939 3940 if (!ext4_has_group_desc_csum(sb)) 3941 return ngroups; 3942 3943 for (group = 0; group < ngroups; group++) { 3944 gdp = ext4_get_group_desc(sb, group, NULL); 3945 if (!gdp) 3946 continue; 3947 3948 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3949 break; 3950 } 3951 3952 return group; 3953 } 3954 3955 static int ext4_li_info_new(void) 3956 { 3957 struct ext4_lazy_init *eli = NULL; 3958 3959 eli = kzalloc(sizeof(*eli), GFP_KERNEL); 3960 if (!eli) 3961 return -ENOMEM; 3962 3963 INIT_LIST_HEAD(&eli->li_request_list); 3964 mutex_init(&eli->li_list_mtx); 3965 3966 eli->li_state |= EXT4_LAZYINIT_QUIT; 3967 3968 ext4_li_info = eli; 3969 3970 return 0; 3971 } 3972 3973 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3974 ext4_group_t start) 3975 { 3976 struct ext4_li_request *elr; 3977 3978 elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3979 if (!elr) 3980 return NULL; 3981 3982 elr->lr_super = sb; 3983 elr->lr_first_not_zeroed = start; 3984 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) { 3985 elr->lr_mode = EXT4_LI_MODE_ITABLE; 3986 elr->lr_next_group = start; 3987 } else { 3988 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP; 3989 } 3990 3991 /* 3992 * Randomize first schedule time of the request to 3993 * spread the inode table initialization requests 3994 * better. 3995 */ 3996 elr->lr_next_sched = jiffies + get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ); 3997 return elr; 3998 } 3999 4000 int ext4_register_li_request(struct super_block *sb, 4001 ext4_group_t first_not_zeroed) 4002 { 4003 struct ext4_sb_info *sbi = EXT4_SB(sb); 4004 struct ext4_li_request *elr = NULL; 4005 ext4_group_t ngroups = sbi->s_groups_count; 4006 int ret = 0; 4007 4008 mutex_lock(&ext4_li_mtx); 4009 if (sbi->s_li_request != NULL) { 4010 /* 4011 * Reset timeout so it can be computed again, because 4012 * s_li_wait_mult might have changed. 4013 */ 4014 sbi->s_li_request->lr_timeout = 0; 4015 goto out; 4016 } 4017 4018 if (sb_rdonly(sb) || 4019 (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) && 4020 (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE)))) 4021 goto out; 4022 4023 elr = ext4_li_request_new(sb, first_not_zeroed); 4024 if (!elr) { 4025 ret = -ENOMEM; 4026 goto out; 4027 } 4028 4029 if (NULL == ext4_li_info) { 4030 ret = ext4_li_info_new(); 4031 if (ret) 4032 goto out; 4033 } 4034 4035 mutex_lock(&ext4_li_info->li_list_mtx); 4036 list_add(&elr->lr_request, &ext4_li_info->li_request_list); 4037 mutex_unlock(&ext4_li_info->li_list_mtx); 4038 4039 sbi->s_li_request = elr; 4040 /* 4041 * set elr to NULL here since it has been inserted to 4042 * the request_list and the removal and free of it is 4043 * handled by ext4_clear_request_list from now on. 4044 */ 4045 elr = NULL; 4046 4047 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 4048 ret = ext4_run_lazyinit_thread(); 4049 if (ret) 4050 goto out; 4051 } 4052 out: 4053 mutex_unlock(&ext4_li_mtx); 4054 if (ret) 4055 kfree(elr); 4056 return ret; 4057 } 4058 4059 /* 4060 * We do not need to lock anything since this is called on 4061 * module unload. 4062 */ 4063 static void ext4_destroy_lazyinit_thread(void) 4064 { 4065 /* 4066 * If thread exited earlier 4067 * there's nothing to be done. 4068 */ 4069 if (!ext4_li_info || !ext4_lazyinit_task) 4070 return; 4071 4072 kthread_stop(ext4_lazyinit_task); 4073 } 4074 4075 static int set_journal_csum_feature_set(struct super_block *sb) 4076 { 4077 int ret = 1; 4078 int compat, incompat; 4079 struct ext4_sb_info *sbi = EXT4_SB(sb); 4080 4081 if (ext4_has_metadata_csum(sb)) { 4082 /* journal checksum v3 */ 4083 compat = 0; 4084 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 4085 } else { 4086 /* journal checksum v1 */ 4087 compat = JBD2_FEATURE_COMPAT_CHECKSUM; 4088 incompat = 0; 4089 } 4090 4091 jbd2_journal_clear_features(sbi->s_journal, 4092 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 4093 JBD2_FEATURE_INCOMPAT_CSUM_V3 | 4094 JBD2_FEATURE_INCOMPAT_CSUM_V2); 4095 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4096 ret = jbd2_journal_set_features(sbi->s_journal, 4097 compat, 0, 4098 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 4099 incompat); 4100 } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 4101 ret = jbd2_journal_set_features(sbi->s_journal, 4102 compat, 0, 4103 incompat); 4104 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 4105 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 4106 } else { 4107 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 4108 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 4109 } 4110 4111 return ret; 4112 } 4113 4114 /* 4115 * Note: calculating the overhead so we can be compatible with 4116 * historical BSD practice is quite difficult in the face of 4117 * clusters/bigalloc. This is because multiple metadata blocks from 4118 * different block group can end up in the same allocation cluster. 4119 * Calculating the exact overhead in the face of clustered allocation 4120 * requires either O(all block bitmaps) in memory or O(number of block 4121 * groups**2) in time. We will still calculate the superblock for 4122 * older file systems --- and if we come across with a bigalloc file 4123 * system with zero in s_overhead_clusters the estimate will be close to 4124 * correct especially for very large cluster sizes --- but for newer 4125 * file systems, it's better to calculate this figure once at mkfs 4126 * time, and store it in the superblock. If the superblock value is 4127 * present (even for non-bigalloc file systems), we will use it. 4128 */ 4129 static int count_overhead(struct super_block *sb, ext4_group_t grp, 4130 char *buf) 4131 { 4132 struct ext4_sb_info *sbi = EXT4_SB(sb); 4133 struct ext4_group_desc *gdp; 4134 ext4_fsblk_t first_block, last_block, b; 4135 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 4136 int s, j, count = 0; 4137 int has_super = ext4_bg_has_super(sb, grp); 4138 4139 if (!ext4_has_feature_bigalloc(sb)) 4140 return (has_super + ext4_bg_num_gdb(sb, grp) + 4141 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) + 4142 sbi->s_itb_per_group + 2); 4143 4144 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 4145 (grp * EXT4_BLOCKS_PER_GROUP(sb)); 4146 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 4147 for (i = 0; i < ngroups; i++) { 4148 gdp = ext4_get_group_desc(sb, i, NULL); 4149 b = ext4_block_bitmap(sb, gdp); 4150 if (b >= first_block && b <= last_block) { 4151 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 4152 count++; 4153 } 4154 b = ext4_inode_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_table(sb, gdp); 4160 if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 4161 for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 4162 int c = EXT4_B2C(sbi, b - first_block); 4163 ext4_set_bit(c, buf); 4164 count++; 4165 } 4166 if (i != grp) 4167 continue; 4168 s = 0; 4169 if (ext4_bg_has_super(sb, grp)) { 4170 ext4_set_bit(s++, buf); 4171 count++; 4172 } 4173 j = ext4_bg_num_gdb(sb, grp); 4174 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { 4175 ext4_error(sb, "Invalid number of block group " 4176 "descriptor blocks: %d", j); 4177 j = EXT4_BLOCKS_PER_GROUP(sb) - s; 4178 } 4179 count += j; 4180 for (; j > 0; j--) 4181 ext4_set_bit(EXT4_B2C(sbi, s++), buf); 4182 } 4183 if (!count) 4184 return 0; 4185 return EXT4_CLUSTERS_PER_GROUP(sb) - 4186 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 4187 } 4188 4189 /* 4190 * Compute the overhead and stash it in sbi->s_overhead 4191 */ 4192 int ext4_calculate_overhead(struct super_block *sb) 4193 { 4194 struct ext4_sb_info *sbi = EXT4_SB(sb); 4195 struct ext4_super_block *es = sbi->s_es; 4196 struct inode *j_inode; 4197 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); 4198 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 4199 ext4_fsblk_t overhead = 0; 4200 char *buf = (char *) get_zeroed_page(GFP_NOFS); 4201 4202 if (!buf) 4203 return -ENOMEM; 4204 4205 /* 4206 * Compute the overhead (FS structures). This is constant 4207 * for a given filesystem unless the number of block groups 4208 * changes so we cache the previous value until it does. 4209 */ 4210 4211 /* 4212 * All of the blocks before first_data_block are overhead 4213 */ 4214 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 4215 4216 /* 4217 * Add the overhead found in each block group 4218 */ 4219 for (i = 0; i < ngroups; i++) { 4220 int blks; 4221 4222 blks = count_overhead(sb, i, buf); 4223 overhead += blks; 4224 if (blks) 4225 memset(buf, 0, PAGE_SIZE); 4226 cond_resched(); 4227 } 4228 4229 /* 4230 * Add the internal journal blocks whether the journal has been 4231 * loaded or not 4232 */ 4233 if (sbi->s_journal && !sbi->s_journal_bdev) 4234 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len); 4235 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) { 4236 /* j_inum for internal journal is non-zero */ 4237 j_inode = ext4_get_journal_inode(sb, j_inum); 4238 if (j_inode) { 4239 j_blocks = j_inode->i_size >> sb->s_blocksize_bits; 4240 overhead += EXT4_NUM_B2C(sbi, j_blocks); 4241 iput(j_inode); 4242 } else { 4243 ext4_msg(sb, KERN_ERR, "can't get journal size"); 4244 } 4245 } 4246 sbi->s_overhead = overhead; 4247 smp_wmb(); 4248 free_page((unsigned long) buf); 4249 return 0; 4250 } 4251 4252 static void ext4_set_resv_clusters(struct super_block *sb) 4253 { 4254 ext4_fsblk_t resv_clusters; 4255 struct ext4_sb_info *sbi = EXT4_SB(sb); 4256 4257 /* 4258 * There's no need to reserve anything when we aren't using extents. 4259 * The space estimates are exact, there are no unwritten extents, 4260 * hole punching doesn't need new metadata... This is needed especially 4261 * to keep ext2/3 backward compatibility. 4262 */ 4263 if (!ext4_has_feature_extents(sb)) 4264 return; 4265 /* 4266 * By default we reserve 2% or 4096 clusters, whichever is smaller. 4267 * This should cover the situations where we can not afford to run 4268 * out of space like for example punch hole, or converting 4269 * unwritten extents in delalloc path. In most cases such 4270 * allocation would require 1, or 2 blocks, higher numbers are 4271 * very rare. 4272 */ 4273 resv_clusters = (ext4_blocks_count(sbi->s_es) >> 4274 sbi->s_cluster_bits); 4275 4276 do_div(resv_clusters, 50); 4277 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 4278 4279 atomic64_set(&sbi->s_resv_clusters, resv_clusters); 4280 } 4281 4282 static const char *ext4_quota_mode(struct super_block *sb) 4283 { 4284 #ifdef CONFIG_QUOTA 4285 if (!ext4_quota_capable(sb)) 4286 return "none"; 4287 4288 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb)) 4289 return "journalled"; 4290 else 4291 return "writeback"; 4292 #else 4293 return "disabled"; 4294 #endif 4295 } 4296 4297 static void ext4_setup_csum_trigger(struct super_block *sb, 4298 enum ext4_journal_trigger_type type, 4299 void (*trigger)( 4300 struct jbd2_buffer_trigger_type *type, 4301 struct buffer_head *bh, 4302 void *mapped_data, 4303 size_t size)) 4304 { 4305 struct ext4_sb_info *sbi = EXT4_SB(sb); 4306 4307 sbi->s_journal_triggers[type].sb = sb; 4308 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger; 4309 } 4310 4311 static void ext4_free_sbi(struct ext4_sb_info *sbi) 4312 { 4313 if (!sbi) 4314 return; 4315 4316 kfree(sbi->s_blockgroup_lock); 4317 fs_put_dax(sbi->s_daxdev, NULL); 4318 kfree(sbi); 4319 } 4320 4321 static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb) 4322 { 4323 struct ext4_sb_info *sbi; 4324 4325 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 4326 if (!sbi) 4327 return NULL; 4328 4329 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off, 4330 NULL, NULL); 4331 4332 sbi->s_blockgroup_lock = 4333 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 4334 4335 if (!sbi->s_blockgroup_lock) 4336 goto err_out; 4337 4338 sb->s_fs_info = sbi; 4339 sbi->s_sb = sb; 4340 return sbi; 4341 err_out: 4342 fs_put_dax(sbi->s_daxdev, NULL); 4343 kfree(sbi); 4344 return NULL; 4345 } 4346 4347 static void ext4_set_def_opts(struct super_block *sb, 4348 struct ext4_super_block *es) 4349 { 4350 unsigned long def_mount_opts; 4351 4352 /* Set defaults before we parse the mount options */ 4353 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 4354 set_opt(sb, INIT_INODE_TABLE); 4355 if (def_mount_opts & EXT4_DEFM_DEBUG) 4356 set_opt(sb, DEBUG); 4357 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 4358 set_opt(sb, GRPID); 4359 if (def_mount_opts & EXT4_DEFM_UID16) 4360 set_opt(sb, NO_UID32); 4361 /* xattr user namespace & acls are now defaulted on */ 4362 set_opt(sb, XATTR_USER); 4363 #ifdef CONFIG_EXT4_FS_POSIX_ACL 4364 set_opt(sb, POSIX_ACL); 4365 #endif 4366 if (ext4_has_feature_fast_commit(sb)) 4367 set_opt2(sb, JOURNAL_FAST_COMMIT); 4368 /* don't forget to enable journal_csum when metadata_csum is enabled. */ 4369 if (ext4_has_metadata_csum(sb)) 4370 set_opt(sb, JOURNAL_CHECKSUM); 4371 4372 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 4373 set_opt(sb, JOURNAL_DATA); 4374 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 4375 set_opt(sb, ORDERED_DATA); 4376 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 4377 set_opt(sb, WRITEBACK_DATA); 4378 4379 if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_PANIC) 4380 set_opt(sb, ERRORS_PANIC); 4381 else if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_CONTINUE) 4382 set_opt(sb, ERRORS_CONT); 4383 else 4384 set_opt(sb, ERRORS_RO); 4385 /* block_validity enabled by default; disable with noblock_validity */ 4386 set_opt(sb, BLOCK_VALIDITY); 4387 if (def_mount_opts & EXT4_DEFM_DISCARD) 4388 set_opt(sb, DISCARD); 4389 4390 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 4391 set_opt(sb, BARRIER); 4392 4393 /* 4394 * enable delayed allocation by default 4395 * Use -o nodelalloc to turn it off 4396 */ 4397 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 4398 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 4399 set_opt(sb, DELALLOC); 4400 4401 if (sb->s_blocksize == PAGE_SIZE) 4402 set_opt(sb, DIOREAD_NOLOCK); 4403 } 4404 4405 static int ext4_handle_clustersize(struct super_block *sb) 4406 { 4407 struct ext4_sb_info *sbi = EXT4_SB(sb); 4408 struct ext4_super_block *es = sbi->s_es; 4409 int clustersize; 4410 4411 /* Handle clustersize */ 4412 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 4413 if (ext4_has_feature_bigalloc(sb)) { 4414 if (clustersize < sb->s_blocksize) { 4415 ext4_msg(sb, KERN_ERR, 4416 "cluster size (%d) smaller than " 4417 "block size (%lu)", clustersize, sb->s_blocksize); 4418 return -EINVAL; 4419 } 4420 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 4421 le32_to_cpu(es->s_log_block_size); 4422 sbi->s_clusters_per_group = 4423 le32_to_cpu(es->s_clusters_per_group); 4424 if (sbi->s_clusters_per_group > sb->s_blocksize * 8) { 4425 ext4_msg(sb, KERN_ERR, 4426 "#clusters per group too big: %lu", 4427 sbi->s_clusters_per_group); 4428 return -EINVAL; 4429 } 4430 if (sbi->s_blocks_per_group != 4431 (sbi->s_clusters_per_group * (clustersize / sb->s_blocksize))) { 4432 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 4433 "clusters per group (%lu) inconsistent", 4434 sbi->s_blocks_per_group, 4435 sbi->s_clusters_per_group); 4436 return -EINVAL; 4437 } 4438 } else { 4439 if (clustersize != sb->s_blocksize) { 4440 ext4_msg(sb, KERN_ERR, 4441 "fragment/cluster size (%d) != " 4442 "block size (%lu)", clustersize, sb->s_blocksize); 4443 return -EINVAL; 4444 } 4445 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) { 4446 ext4_msg(sb, KERN_ERR, 4447 "#blocks per group too big: %lu", 4448 sbi->s_blocks_per_group); 4449 return -EINVAL; 4450 } 4451 sbi->s_clusters_per_group = sbi->s_blocks_per_group; 4452 sbi->s_cluster_bits = 0; 4453 } 4454 sbi->s_cluster_ratio = clustersize / sb->s_blocksize; 4455 4456 /* Do we have standard group size of clustersize * 8 blocks ? */ 4457 if (sbi->s_blocks_per_group == clustersize << 3) 4458 set_opt2(sb, STD_GROUP_SIZE); 4459 4460 return 0; 4461 } 4462 4463 static void ext4_fast_commit_init(struct super_block *sb) 4464 { 4465 struct ext4_sb_info *sbi = EXT4_SB(sb); 4466 4467 /* Initialize fast commit stuff */ 4468 atomic_set(&sbi->s_fc_subtid, 0); 4469 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]); 4470 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]); 4471 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]); 4472 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]); 4473 sbi->s_fc_bytes = 0; 4474 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE); 4475 sbi->s_fc_ineligible_tid = 0; 4476 spin_lock_init(&sbi->s_fc_lock); 4477 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats)); 4478 sbi->s_fc_replay_state.fc_regions = NULL; 4479 sbi->s_fc_replay_state.fc_regions_size = 0; 4480 sbi->s_fc_replay_state.fc_regions_used = 0; 4481 sbi->s_fc_replay_state.fc_regions_valid = 0; 4482 sbi->s_fc_replay_state.fc_modified_inodes = NULL; 4483 sbi->s_fc_replay_state.fc_modified_inodes_size = 0; 4484 sbi->s_fc_replay_state.fc_modified_inodes_used = 0; 4485 } 4486 4487 static int ext4_inode_info_init(struct super_block *sb, 4488 struct ext4_super_block *es) 4489 { 4490 struct ext4_sb_info *sbi = EXT4_SB(sb); 4491 4492 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 4493 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 4494 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 4495 } else { 4496 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 4497 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 4498 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) { 4499 ext4_msg(sb, KERN_ERR, "invalid first ino: %u", 4500 sbi->s_first_ino); 4501 return -EINVAL; 4502 } 4503 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 4504 (!is_power_of_2(sbi->s_inode_size)) || 4505 (sbi->s_inode_size > sb->s_blocksize)) { 4506 ext4_msg(sb, KERN_ERR, 4507 "unsupported inode size: %d", 4508 sbi->s_inode_size); 4509 ext4_msg(sb, KERN_ERR, "blocksize: %lu", sb->s_blocksize); 4510 return -EINVAL; 4511 } 4512 /* 4513 * i_atime_extra is the last extra field available for 4514 * [acm]times in struct ext4_inode. Checking for that 4515 * field should suffice to ensure we have extra space 4516 * for all three. 4517 */ 4518 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) + 4519 sizeof(((struct ext4_inode *)0)->i_atime_extra)) { 4520 sb->s_time_gran = 1; 4521 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX; 4522 } else { 4523 sb->s_time_gran = NSEC_PER_SEC; 4524 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX; 4525 } 4526 sb->s_time_min = EXT4_TIMESTAMP_MIN; 4527 } 4528 4529 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 4530 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 4531 EXT4_GOOD_OLD_INODE_SIZE; 4532 if (ext4_has_feature_extra_isize(sb)) { 4533 unsigned v, max = (sbi->s_inode_size - 4534 EXT4_GOOD_OLD_INODE_SIZE); 4535 4536 v = le16_to_cpu(es->s_want_extra_isize); 4537 if (v > max) { 4538 ext4_msg(sb, KERN_ERR, 4539 "bad s_want_extra_isize: %d", v); 4540 return -EINVAL; 4541 } 4542 if (sbi->s_want_extra_isize < v) 4543 sbi->s_want_extra_isize = v; 4544 4545 v = le16_to_cpu(es->s_min_extra_isize); 4546 if (v > max) { 4547 ext4_msg(sb, KERN_ERR, 4548 "bad s_min_extra_isize: %d", v); 4549 return -EINVAL; 4550 } 4551 if (sbi->s_want_extra_isize < v) 4552 sbi->s_want_extra_isize = v; 4553 } 4554 } 4555 4556 return 0; 4557 } 4558 4559 #if IS_ENABLED(CONFIG_UNICODE) 4560 static int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es) 4561 { 4562 const struct ext4_sb_encodings *encoding_info; 4563 struct unicode_map *encoding; 4564 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags); 4565 4566 if (!ext4_has_feature_casefold(sb) || sb->s_encoding) 4567 return 0; 4568 4569 encoding_info = ext4_sb_read_encoding(es); 4570 if (!encoding_info) { 4571 ext4_msg(sb, KERN_ERR, 4572 "Encoding requested by superblock is unknown"); 4573 return -EINVAL; 4574 } 4575 4576 encoding = utf8_load(encoding_info->version); 4577 if (IS_ERR(encoding)) { 4578 ext4_msg(sb, KERN_ERR, 4579 "can't mount with superblock charset: %s-%u.%u.%u " 4580 "not supported by the kernel. flags: 0x%x.", 4581 encoding_info->name, 4582 unicode_major(encoding_info->version), 4583 unicode_minor(encoding_info->version), 4584 unicode_rev(encoding_info->version), 4585 encoding_flags); 4586 return -EINVAL; 4587 } 4588 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: " 4589 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name, 4590 unicode_major(encoding_info->version), 4591 unicode_minor(encoding_info->version), 4592 unicode_rev(encoding_info->version), 4593 encoding_flags); 4594 4595 sb->s_encoding = encoding; 4596 sb->s_encoding_flags = encoding_flags; 4597 4598 return 0; 4599 } 4600 #else 4601 static inline int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es) 4602 { 4603 return 0; 4604 } 4605 #endif 4606 4607 static int ext4_init_metadata_csum(struct super_block *sb, struct ext4_super_block *es) 4608 { 4609 struct ext4_sb_info *sbi = EXT4_SB(sb); 4610 4611 /* Warn if metadata_csum and gdt_csum are both set. */ 4612 if (ext4_has_feature_metadata_csum(sb) && 4613 ext4_has_feature_gdt_csum(sb)) 4614 ext4_warning(sb, "metadata_csum and uninit_bg are " 4615 "redundant flags; please run fsck."); 4616 4617 /* Check for a known checksum algorithm */ 4618 if (!ext4_verify_csum_type(sb, es)) { 4619 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 4620 "unknown checksum algorithm."); 4621 return -EINVAL; 4622 } 4623 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE, 4624 ext4_orphan_file_block_trigger); 4625 4626 /* Load the checksum driver */ 4627 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 4628 if (IS_ERR(sbi->s_chksum_driver)) { 4629 int ret = PTR_ERR(sbi->s_chksum_driver); 4630 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 4631 sbi->s_chksum_driver = NULL; 4632 return ret; 4633 } 4634 4635 /* Check superblock checksum */ 4636 if (!ext4_superblock_csum_verify(sb, es)) { 4637 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 4638 "invalid superblock checksum. Run e2fsck?"); 4639 return -EFSBADCRC; 4640 } 4641 4642 /* Precompute checksum seed for all metadata */ 4643 if (ext4_has_feature_csum_seed(sb)) 4644 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 4645 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb)) 4646 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 4647 sizeof(es->s_uuid)); 4648 return 0; 4649 } 4650 4651 static int ext4_check_feature_compatibility(struct super_block *sb, 4652 struct ext4_super_block *es, 4653 int silent) 4654 { 4655 struct ext4_sb_info *sbi = EXT4_SB(sb); 4656 4657 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 4658 (ext4_has_compat_features(sb) || 4659 ext4_has_ro_compat_features(sb) || 4660 ext4_has_incompat_features(sb))) 4661 ext4_msg(sb, KERN_WARNING, 4662 "feature flags set on rev 0 fs, " 4663 "running e2fsck is recommended"); 4664 4665 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 4666 set_opt2(sb, HURD_COMPAT); 4667 if (ext4_has_feature_64bit(sb)) { 4668 ext4_msg(sb, KERN_ERR, 4669 "The Hurd can't support 64-bit file systems"); 4670 return -EINVAL; 4671 } 4672 4673 /* 4674 * ea_inode feature uses l_i_version field which is not 4675 * available in HURD_COMPAT mode. 4676 */ 4677 if (ext4_has_feature_ea_inode(sb)) { 4678 ext4_msg(sb, KERN_ERR, 4679 "ea_inode feature is not supported for Hurd"); 4680 return -EINVAL; 4681 } 4682 } 4683 4684 if (IS_EXT2_SB(sb)) { 4685 if (ext2_feature_set_ok(sb)) 4686 ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 4687 "using the ext4 subsystem"); 4688 else { 4689 /* 4690 * If we're probing be silent, if this looks like 4691 * it's actually an ext[34] filesystem. 4692 */ 4693 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 4694 return -EINVAL; 4695 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 4696 "to feature incompatibilities"); 4697 return -EINVAL; 4698 } 4699 } 4700 4701 if (IS_EXT3_SB(sb)) { 4702 if (ext3_feature_set_ok(sb)) 4703 ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 4704 "using the ext4 subsystem"); 4705 else { 4706 /* 4707 * If we're probing be silent, if this looks like 4708 * it's actually an ext4 filesystem. 4709 */ 4710 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 4711 return -EINVAL; 4712 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 4713 "to feature incompatibilities"); 4714 return -EINVAL; 4715 } 4716 } 4717 4718 /* 4719 * Check feature flags regardless of the revision level, since we 4720 * previously didn't change the revision level when setting the flags, 4721 * so there is a chance incompat flags are set on a rev 0 filesystem. 4722 */ 4723 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) 4724 return -EINVAL; 4725 4726 if (sbi->s_daxdev) { 4727 if (sb->s_blocksize == PAGE_SIZE) 4728 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags); 4729 else 4730 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n"); 4731 } 4732 4733 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) { 4734 if (ext4_has_feature_inline_data(sb)) { 4735 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem" 4736 " that may contain inline data"); 4737 return -EINVAL; 4738 } 4739 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) { 4740 ext4_msg(sb, KERN_ERR, 4741 "DAX unsupported by block device."); 4742 return -EINVAL; 4743 } 4744 } 4745 4746 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 4747 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 4748 es->s_encryption_level); 4749 return -EINVAL; 4750 } 4751 4752 return 0; 4753 } 4754 4755 static int ext4_check_geometry(struct super_block *sb, 4756 struct ext4_super_block *es) 4757 { 4758 struct ext4_sb_info *sbi = EXT4_SB(sb); 4759 __u64 blocks_count; 4760 int err; 4761 4762 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (sb->s_blocksize / 4)) { 4763 ext4_msg(sb, KERN_ERR, 4764 "Number of reserved GDT blocks insanely large: %d", 4765 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); 4766 return -EINVAL; 4767 } 4768 /* 4769 * Test whether we have more sectors than will fit in sector_t, 4770 * and whether the max offset is addressable by the page cache. 4771 */ 4772 err = generic_check_addressable(sb->s_blocksize_bits, 4773 ext4_blocks_count(es)); 4774 if (err) { 4775 ext4_msg(sb, KERN_ERR, "filesystem" 4776 " too large to mount safely on this system"); 4777 return err; 4778 } 4779 4780 /* check blocks count against device size */ 4781 blocks_count = sb_bdev_nr_blocks(sb); 4782 if (blocks_count && ext4_blocks_count(es) > blocks_count) { 4783 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 4784 "exceeds size of device (%llu blocks)", 4785 ext4_blocks_count(es), blocks_count); 4786 return -EINVAL; 4787 } 4788 4789 /* 4790 * It makes no sense for the first data block to be beyond the end 4791 * of the filesystem. 4792 */ 4793 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 4794 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4795 "block %u is beyond end of filesystem (%llu)", 4796 le32_to_cpu(es->s_first_data_block), 4797 ext4_blocks_count(es)); 4798 return -EINVAL; 4799 } 4800 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) && 4801 (sbi->s_cluster_ratio == 1)) { 4802 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4803 "block is 0 with a 1k block and cluster size"); 4804 return -EINVAL; 4805 } 4806 4807 blocks_count = (ext4_blocks_count(es) - 4808 le32_to_cpu(es->s_first_data_block) + 4809 EXT4_BLOCKS_PER_GROUP(sb) - 1); 4810 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 4811 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 4812 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu " 4813 "(block count %llu, first data block %u, " 4814 "blocks per group %lu)", blocks_count, 4815 ext4_blocks_count(es), 4816 le32_to_cpu(es->s_first_data_block), 4817 EXT4_BLOCKS_PER_GROUP(sb)); 4818 return -EINVAL; 4819 } 4820 sbi->s_groups_count = blocks_count; 4821 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 4822 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 4823 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) != 4824 le32_to_cpu(es->s_inodes_count)) { 4825 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu", 4826 le32_to_cpu(es->s_inodes_count), 4827 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group)); 4828 return -EINVAL; 4829 } 4830 4831 return 0; 4832 } 4833 4834 static int ext4_group_desc_init(struct super_block *sb, 4835 struct ext4_super_block *es, 4836 ext4_fsblk_t logical_sb_block, 4837 ext4_group_t *first_not_zeroed) 4838 { 4839 struct ext4_sb_info *sbi = EXT4_SB(sb); 4840 unsigned int db_count; 4841 ext4_fsblk_t block; 4842 int i; 4843 4844 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 4845 EXT4_DESC_PER_BLOCK(sb); 4846 if (ext4_has_feature_meta_bg(sb)) { 4847 if (le32_to_cpu(es->s_first_meta_bg) > db_count) { 4848 ext4_msg(sb, KERN_WARNING, 4849 "first meta block group too large: %u " 4850 "(group descriptor block count %u)", 4851 le32_to_cpu(es->s_first_meta_bg), db_count); 4852 return -EINVAL; 4853 } 4854 } 4855 rcu_assign_pointer(sbi->s_group_desc, 4856 kvmalloc_array(db_count, 4857 sizeof(struct buffer_head *), 4858 GFP_KERNEL)); 4859 if (sbi->s_group_desc == NULL) { 4860 ext4_msg(sb, KERN_ERR, "not enough memory"); 4861 return -ENOMEM; 4862 } 4863 4864 bgl_lock_init(sbi->s_blockgroup_lock); 4865 4866 /* Pre-read the descriptors into the buffer cache */ 4867 for (i = 0; i < db_count; i++) { 4868 block = descriptor_loc(sb, logical_sb_block, i); 4869 ext4_sb_breadahead_unmovable(sb, block); 4870 } 4871 4872 for (i = 0; i < db_count; i++) { 4873 struct buffer_head *bh; 4874 4875 block = descriptor_loc(sb, logical_sb_block, i); 4876 bh = ext4_sb_bread_unmovable(sb, block); 4877 if (IS_ERR(bh)) { 4878 ext4_msg(sb, KERN_ERR, 4879 "can't read group descriptor %d", i); 4880 sbi->s_gdb_count = i; 4881 return PTR_ERR(bh); 4882 } 4883 rcu_read_lock(); 4884 rcu_dereference(sbi->s_group_desc)[i] = bh; 4885 rcu_read_unlock(); 4886 } 4887 sbi->s_gdb_count = db_count; 4888 if (!ext4_check_descriptors(sb, logical_sb_block, first_not_zeroed)) { 4889 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 4890 return -EFSCORRUPTED; 4891 } 4892 4893 return 0; 4894 } 4895 4896 static int ext4_load_and_init_journal(struct super_block *sb, 4897 struct ext4_super_block *es, 4898 struct ext4_fs_context *ctx) 4899 { 4900 struct ext4_sb_info *sbi = EXT4_SB(sb); 4901 int err; 4902 4903 err = ext4_load_journal(sb, es, ctx->journal_devnum); 4904 if (err) 4905 return err; 4906 4907 if (ext4_has_feature_64bit(sb) && 4908 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4909 JBD2_FEATURE_INCOMPAT_64BIT)) { 4910 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 4911 goto out; 4912 } 4913 4914 if (!set_journal_csum_feature_set(sb)) { 4915 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 4916 "feature set"); 4917 goto out; 4918 } 4919 4920 if (test_opt2(sb, JOURNAL_FAST_COMMIT) && 4921 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4922 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) { 4923 ext4_msg(sb, KERN_ERR, 4924 "Failed to set fast commit journal feature"); 4925 goto out; 4926 } 4927 4928 /* We have now updated the journal if required, so we can 4929 * validate the data journaling mode. */ 4930 switch (test_opt(sb, DATA_FLAGS)) { 4931 case 0: 4932 /* No mode set, assume a default based on the journal 4933 * capabilities: ORDERED_DATA if the journal can 4934 * cope, else JOURNAL_DATA 4935 */ 4936 if (jbd2_journal_check_available_features 4937 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4938 set_opt(sb, ORDERED_DATA); 4939 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA; 4940 } else { 4941 set_opt(sb, JOURNAL_DATA); 4942 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; 4943 } 4944 break; 4945 4946 case EXT4_MOUNT_ORDERED_DATA: 4947 case EXT4_MOUNT_WRITEBACK_DATA: 4948 if (!jbd2_journal_check_available_features 4949 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4950 ext4_msg(sb, KERN_ERR, "Journal does not support " 4951 "requested data journaling mode"); 4952 goto out; 4953 } 4954 break; 4955 default: 4956 break; 4957 } 4958 4959 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 4960 test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4961 ext4_msg(sb, KERN_ERR, "can't mount with " 4962 "journal_async_commit in data=ordered mode"); 4963 goto out; 4964 } 4965 4966 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio); 4967 4968 sbi->s_journal->j_submit_inode_data_buffers = 4969 ext4_journal_submit_inode_data_buffers; 4970 sbi->s_journal->j_finish_inode_data_buffers = 4971 ext4_journal_finish_inode_data_buffers; 4972 4973 return 0; 4974 4975 out: 4976 /* flush s_error_work before journal destroy. */ 4977 flush_work(&sbi->s_error_work); 4978 jbd2_journal_destroy(sbi->s_journal); 4979 sbi->s_journal = NULL; 4980 return -EINVAL; 4981 } 4982 4983 static int ext4_check_journal_data_mode(struct super_block *sb) 4984 { 4985 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 4986 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with " 4987 "data=journal disables delayed allocation, " 4988 "dioread_nolock, O_DIRECT and fast_commit support!\n"); 4989 /* can't mount with both data=journal and dioread_nolock. */ 4990 clear_opt(sb, DIOREAD_NOLOCK); 4991 clear_opt2(sb, JOURNAL_FAST_COMMIT); 4992 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 4993 ext4_msg(sb, KERN_ERR, "can't mount with " 4994 "both data=journal and delalloc"); 4995 return -EINVAL; 4996 } 4997 if (test_opt(sb, DAX_ALWAYS)) { 4998 ext4_msg(sb, KERN_ERR, "can't mount with " 4999 "both data=journal and dax"); 5000 return -EINVAL; 5001 } 5002 if (ext4_has_feature_encrypt(sb)) { 5003 ext4_msg(sb, KERN_WARNING, 5004 "encrypted files will use data=ordered " 5005 "instead of data journaling mode"); 5006 } 5007 if (test_opt(sb, DELALLOC)) 5008 clear_opt(sb, DELALLOC); 5009 } else { 5010 sb->s_iflags |= SB_I_CGROUPWB; 5011 } 5012 5013 return 0; 5014 } 5015 5016 static int ext4_load_super(struct super_block *sb, ext4_fsblk_t *lsb, 5017 int silent) 5018 { 5019 struct ext4_sb_info *sbi = EXT4_SB(sb); 5020 struct ext4_super_block *es; 5021 ext4_fsblk_t logical_sb_block; 5022 unsigned long offset = 0; 5023 struct buffer_head *bh; 5024 int ret = -EINVAL; 5025 int blocksize; 5026 5027 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 5028 if (!blocksize) { 5029 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 5030 return -EINVAL; 5031 } 5032 5033 /* 5034 * The ext4 superblock will not be buffer aligned for other than 1kB 5035 * block sizes. We need to calculate the offset from buffer start. 5036 */ 5037 if (blocksize != EXT4_MIN_BLOCK_SIZE) { 5038 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE; 5039 offset = do_div(logical_sb_block, blocksize); 5040 } else { 5041 logical_sb_block = sbi->s_sb_block; 5042 } 5043 5044 bh = ext4_sb_bread_unmovable(sb, logical_sb_block); 5045 if (IS_ERR(bh)) { 5046 ext4_msg(sb, KERN_ERR, "unable to read superblock"); 5047 return PTR_ERR(bh); 5048 } 5049 /* 5050 * Note: s_es must be initialized as soon as possible because 5051 * some ext4 macro-instructions depend on its value 5052 */ 5053 es = (struct ext4_super_block *) (bh->b_data + offset); 5054 sbi->s_es = es; 5055 sb->s_magic = le16_to_cpu(es->s_magic); 5056 if (sb->s_magic != EXT4_SUPER_MAGIC) { 5057 if (!silent) 5058 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 5059 goto out; 5060 } 5061 5062 if (le32_to_cpu(es->s_log_block_size) > 5063 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 5064 ext4_msg(sb, KERN_ERR, 5065 "Invalid log block size: %u", 5066 le32_to_cpu(es->s_log_block_size)); 5067 goto out; 5068 } 5069 if (le32_to_cpu(es->s_log_cluster_size) > 5070 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 5071 ext4_msg(sb, KERN_ERR, 5072 "Invalid log cluster size: %u", 5073 le32_to_cpu(es->s_log_cluster_size)); 5074 goto out; 5075 } 5076 5077 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 5078 5079 /* 5080 * If the default block size is not the same as the real block size, 5081 * we need to reload it. 5082 */ 5083 if (sb->s_blocksize == blocksize) { 5084 *lsb = logical_sb_block; 5085 sbi->s_sbh = bh; 5086 return 0; 5087 } 5088 5089 /* 5090 * bh must be released before kill_bdev(), otherwise 5091 * it won't be freed and its page also. kill_bdev() 5092 * is called by sb_set_blocksize(). 5093 */ 5094 brelse(bh); 5095 /* Validate the filesystem blocksize */ 5096 if (!sb_set_blocksize(sb, blocksize)) { 5097 ext4_msg(sb, KERN_ERR, "bad block size %d", 5098 blocksize); 5099 bh = NULL; 5100 goto out; 5101 } 5102 5103 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE; 5104 offset = do_div(logical_sb_block, blocksize); 5105 bh = ext4_sb_bread_unmovable(sb, logical_sb_block); 5106 if (IS_ERR(bh)) { 5107 ext4_msg(sb, KERN_ERR, "Can't read superblock on 2nd try"); 5108 ret = PTR_ERR(bh); 5109 bh = NULL; 5110 goto out; 5111 } 5112 es = (struct ext4_super_block *)(bh->b_data + offset); 5113 sbi->s_es = es; 5114 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 5115 ext4_msg(sb, KERN_ERR, "Magic mismatch, very weird!"); 5116 goto out; 5117 } 5118 *lsb = logical_sb_block; 5119 sbi->s_sbh = bh; 5120 return 0; 5121 out: 5122 brelse(bh); 5123 return ret; 5124 } 5125 5126 static void ext4_hash_info_init(struct super_block *sb) 5127 { 5128 struct ext4_sb_info *sbi = EXT4_SB(sb); 5129 struct ext4_super_block *es = sbi->s_es; 5130 unsigned int i; 5131 5132 for (i = 0; i < 4; i++) 5133 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 5134 5135 sbi->s_def_hash_version = es->s_def_hash_version; 5136 if (ext4_has_feature_dir_index(sb)) { 5137 i = le32_to_cpu(es->s_flags); 5138 if (i & EXT2_FLAGS_UNSIGNED_HASH) 5139 sbi->s_hash_unsigned = 3; 5140 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 5141 #ifdef __CHAR_UNSIGNED__ 5142 if (!sb_rdonly(sb)) 5143 es->s_flags |= 5144 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 5145 sbi->s_hash_unsigned = 3; 5146 #else 5147 if (!sb_rdonly(sb)) 5148 es->s_flags |= 5149 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 5150 #endif 5151 } 5152 } 5153 } 5154 5155 static int ext4_block_group_meta_init(struct super_block *sb, int silent) 5156 { 5157 struct ext4_sb_info *sbi = EXT4_SB(sb); 5158 struct ext4_super_block *es = sbi->s_es; 5159 int has_huge_files; 5160 5161 has_huge_files = ext4_has_feature_huge_file(sb); 5162 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 5163 has_huge_files); 5164 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 5165 5166 sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 5167 if (ext4_has_feature_64bit(sb)) { 5168 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 5169 sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 5170 !is_power_of_2(sbi->s_desc_size)) { 5171 ext4_msg(sb, KERN_ERR, 5172 "unsupported descriptor size %lu", 5173 sbi->s_desc_size); 5174 return -EINVAL; 5175 } 5176 } else 5177 sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 5178 5179 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 5180 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 5181 5182 sbi->s_inodes_per_block = sb->s_blocksize / EXT4_INODE_SIZE(sb); 5183 if (sbi->s_inodes_per_block == 0 || sbi->s_blocks_per_group == 0) { 5184 if (!silent) 5185 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 5186 return -EINVAL; 5187 } 5188 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || 5189 sbi->s_inodes_per_group > sb->s_blocksize * 8) { 5190 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", 5191 sbi->s_inodes_per_group); 5192 return -EINVAL; 5193 } 5194 sbi->s_itb_per_group = sbi->s_inodes_per_group / 5195 sbi->s_inodes_per_block; 5196 sbi->s_desc_per_block = sb->s_blocksize / EXT4_DESC_SIZE(sb); 5197 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY; 5198 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 5199 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 5200 5201 return 0; 5202 } 5203 5204 static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb) 5205 { 5206 struct ext4_super_block *es = NULL; 5207 struct ext4_sb_info *sbi = EXT4_SB(sb); 5208 ext4_fsblk_t logical_sb_block; 5209 struct inode *root; 5210 int needs_recovery; 5211 int err; 5212 ext4_group_t first_not_zeroed; 5213 struct ext4_fs_context *ctx = fc->fs_private; 5214 int silent = fc->sb_flags & SB_SILENT; 5215 5216 /* Set defaults for the variables that will be set during parsing */ 5217 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) 5218 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 5219 5220 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 5221 sbi->s_sectors_written_start = 5222 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]); 5223 5224 err = ext4_load_super(sb, &logical_sb_block, silent); 5225 if (err) 5226 goto out_fail; 5227 5228 es = sbi->s_es; 5229 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 5230 5231 err = ext4_init_metadata_csum(sb, es); 5232 if (err) 5233 goto failed_mount; 5234 5235 ext4_set_def_opts(sb, es); 5236 5237 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 5238 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 5239 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 5240 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 5241 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 5242 5243 /* 5244 * set default s_li_wait_mult for lazyinit, for the case there is 5245 * no mount option specified. 5246 */ 5247 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 5248 5249 err = ext4_inode_info_init(sb, es); 5250 if (err) 5251 goto failed_mount; 5252 5253 err = parse_apply_sb_mount_options(sb, ctx); 5254 if (err < 0) 5255 goto failed_mount; 5256 5257 sbi->s_def_mount_opt = sbi->s_mount_opt; 5258 sbi->s_def_mount_opt2 = sbi->s_mount_opt2; 5259 5260 err = ext4_check_opt_consistency(fc, sb); 5261 if (err < 0) 5262 goto failed_mount; 5263 5264 ext4_apply_options(fc, sb); 5265 5266 err = ext4_encoding_init(sb, es); 5267 if (err) 5268 goto failed_mount; 5269 5270 err = ext4_check_journal_data_mode(sb); 5271 if (err) 5272 goto failed_mount; 5273 5274 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 5275 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 5276 5277 /* i_version is always enabled now */ 5278 sb->s_flags |= SB_I_VERSION; 5279 5280 err = ext4_check_feature_compatibility(sb, es, silent); 5281 if (err) 5282 goto failed_mount; 5283 5284 err = ext4_block_group_meta_init(sb, silent); 5285 if (err) 5286 goto failed_mount; 5287 5288 ext4_hash_info_init(sb); 5289 5290 err = ext4_handle_clustersize(sb); 5291 if (err) 5292 goto failed_mount; 5293 5294 err = ext4_check_geometry(sb, es); 5295 if (err) 5296 goto failed_mount; 5297 5298 timer_setup(&sbi->s_err_report, print_daily_error_info, 0); 5299 spin_lock_init(&sbi->s_error_lock); 5300 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work); 5301 5302 err = ext4_group_desc_init(sb, es, logical_sb_block, &first_not_zeroed); 5303 if (err) 5304 goto failed_mount3; 5305 5306 err = ext4_es_register_shrinker(sbi); 5307 if (err) 5308 goto failed_mount3; 5309 5310 sbi->s_stripe = ext4_get_stripe_size(sbi); 5311 sbi->s_extent_max_zeroout_kb = 32; 5312 5313 /* 5314 * set up enough so that it can read an inode 5315 */ 5316 sb->s_op = &ext4_sops; 5317 sb->s_export_op = &ext4_export_ops; 5318 sb->s_xattr = ext4_xattr_handlers; 5319 #ifdef CONFIG_FS_ENCRYPTION 5320 sb->s_cop = &ext4_cryptops; 5321 #endif 5322 #ifdef CONFIG_FS_VERITY 5323 sb->s_vop = &ext4_verityops; 5324 #endif 5325 #ifdef CONFIG_QUOTA 5326 sb->dq_op = &ext4_quota_operations; 5327 if (ext4_has_feature_quota(sb)) 5328 sb->s_qcop = &dquot_quotactl_sysfile_ops; 5329 else 5330 sb->s_qcop = &ext4_qctl_operations; 5331 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 5332 #endif 5333 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 5334 5335 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 5336 mutex_init(&sbi->s_orphan_lock); 5337 5338 ext4_fast_commit_init(sb); 5339 5340 sb->s_root = NULL; 5341 5342 needs_recovery = (es->s_last_orphan != 0 || 5343 ext4_has_feature_orphan_present(sb) || 5344 ext4_has_feature_journal_needs_recovery(sb)); 5345 5346 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) { 5347 err = ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)); 5348 if (err) 5349 goto failed_mount3a; 5350 } 5351 5352 err = -EINVAL; 5353 /* 5354 * The first inode we look at is the journal inode. Don't try 5355 * root first: it may be modified in the journal! 5356 */ 5357 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 5358 err = ext4_load_and_init_journal(sb, es, ctx); 5359 if (err) 5360 goto failed_mount3a; 5361 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && 5362 ext4_has_feature_journal_needs_recovery(sb)) { 5363 ext4_msg(sb, KERN_ERR, "required journal recovery " 5364 "suppressed and not mounted read-only"); 5365 goto failed_mount3a; 5366 } else { 5367 /* Nojournal mode, all journal mount options are illegal */ 5368 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 5369 ext4_msg(sb, KERN_ERR, "can't mount with " 5370 "journal_async_commit, fs mounted w/o journal"); 5371 goto failed_mount3a; 5372 } 5373 5374 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { 5375 ext4_msg(sb, KERN_ERR, "can't mount with " 5376 "journal_checksum, fs mounted w/o journal"); 5377 goto failed_mount3a; 5378 } 5379 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { 5380 ext4_msg(sb, KERN_ERR, "can't mount with " 5381 "commit=%lu, fs mounted w/o journal", 5382 sbi->s_commit_interval / HZ); 5383 goto failed_mount3a; 5384 } 5385 if (EXT4_MOUNT_DATA_FLAGS & 5386 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 5387 ext4_msg(sb, KERN_ERR, "can't mount with " 5388 "data=, fs mounted w/o journal"); 5389 goto failed_mount3a; 5390 } 5391 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM; 5392 clear_opt(sb, JOURNAL_CHECKSUM); 5393 clear_opt(sb, DATA_FLAGS); 5394 clear_opt2(sb, JOURNAL_FAST_COMMIT); 5395 sbi->s_journal = NULL; 5396 needs_recovery = 0; 5397 } 5398 5399 if (!test_opt(sb, NO_MBCACHE)) { 5400 sbi->s_ea_block_cache = ext4_xattr_create_cache(); 5401 if (!sbi->s_ea_block_cache) { 5402 ext4_msg(sb, KERN_ERR, 5403 "Failed to create ea_block_cache"); 5404 err = -EINVAL; 5405 goto failed_mount_wq; 5406 } 5407 5408 if (ext4_has_feature_ea_inode(sb)) { 5409 sbi->s_ea_inode_cache = ext4_xattr_create_cache(); 5410 if (!sbi->s_ea_inode_cache) { 5411 ext4_msg(sb, KERN_ERR, 5412 "Failed to create ea_inode_cache"); 5413 err = -EINVAL; 5414 goto failed_mount_wq; 5415 } 5416 } 5417 } 5418 5419 /* 5420 * Get the # of file system overhead blocks from the 5421 * superblock if present. 5422 */ 5423 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 5424 /* ignore the precalculated value if it is ridiculous */ 5425 if (sbi->s_overhead > ext4_blocks_count(es)) 5426 sbi->s_overhead = 0; 5427 /* 5428 * If the bigalloc feature is not enabled recalculating the 5429 * overhead doesn't take long, so we might as well just redo 5430 * it to make sure we are using the correct value. 5431 */ 5432 if (!ext4_has_feature_bigalloc(sb)) 5433 sbi->s_overhead = 0; 5434 if (sbi->s_overhead == 0) { 5435 err = ext4_calculate_overhead(sb); 5436 if (err) 5437 goto failed_mount_wq; 5438 } 5439 5440 /* 5441 * The maximum number of concurrent works can be high and 5442 * concurrency isn't really necessary. Limit it to 1. 5443 */ 5444 EXT4_SB(sb)->rsv_conversion_wq = 5445 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 5446 if (!EXT4_SB(sb)->rsv_conversion_wq) { 5447 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 5448 err = -ENOMEM; 5449 goto failed_mount4; 5450 } 5451 5452 /* 5453 * The jbd2_journal_load will have done any necessary log recovery, 5454 * so we can safely mount the rest of the filesystem now. 5455 */ 5456 5457 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL); 5458 if (IS_ERR(root)) { 5459 ext4_msg(sb, KERN_ERR, "get root inode failed"); 5460 err = PTR_ERR(root); 5461 root = NULL; 5462 goto failed_mount4; 5463 } 5464 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 5465 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 5466 iput(root); 5467 err = -EFSCORRUPTED; 5468 goto failed_mount4; 5469 } 5470 5471 sb->s_root = d_make_root(root); 5472 if (!sb->s_root) { 5473 ext4_msg(sb, KERN_ERR, "get root dentry failed"); 5474 err = -ENOMEM; 5475 goto failed_mount4; 5476 } 5477 5478 err = ext4_setup_super(sb, es, sb_rdonly(sb)); 5479 if (err == -EROFS) { 5480 sb->s_flags |= SB_RDONLY; 5481 } else if (err) 5482 goto failed_mount4a; 5483 5484 ext4_set_resv_clusters(sb); 5485 5486 if (test_opt(sb, BLOCK_VALIDITY)) { 5487 err = ext4_setup_system_zone(sb); 5488 if (err) { 5489 ext4_msg(sb, KERN_ERR, "failed to initialize system " 5490 "zone (%d)", err); 5491 goto failed_mount4a; 5492 } 5493 } 5494 ext4_fc_replay_cleanup(sb); 5495 5496 ext4_ext_init(sb); 5497 5498 /* 5499 * Enable optimize_scan if number of groups is > threshold. This can be 5500 * turned off by passing "mb_optimize_scan=0". This can also be 5501 * turned on forcefully by passing "mb_optimize_scan=1". 5502 */ 5503 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) { 5504 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD) 5505 set_opt2(sb, MB_OPTIMIZE_SCAN); 5506 else 5507 clear_opt2(sb, MB_OPTIMIZE_SCAN); 5508 } 5509 5510 err = ext4_mb_init(sb); 5511 if (err) { 5512 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 5513 err); 5514 goto failed_mount5; 5515 } 5516 5517 /* 5518 * We can only set up the journal commit callback once 5519 * mballoc is initialized 5520 */ 5521 if (sbi->s_journal) 5522 sbi->s_journal->j_commit_callback = 5523 ext4_journal_commit_callback; 5524 5525 err = ext4_percpu_param_init(sbi); 5526 if (err) 5527 goto failed_mount6; 5528 5529 if (ext4_has_feature_flex_bg(sb)) 5530 if (!ext4_fill_flex_info(sb)) { 5531 ext4_msg(sb, KERN_ERR, 5532 "unable to initialize " 5533 "flex_bg meta info!"); 5534 err = -ENOMEM; 5535 goto failed_mount6; 5536 } 5537 5538 err = ext4_register_li_request(sb, first_not_zeroed); 5539 if (err) 5540 goto failed_mount6; 5541 5542 err = ext4_register_sysfs(sb); 5543 if (err) 5544 goto failed_mount7; 5545 5546 err = ext4_init_orphan_info(sb); 5547 if (err) 5548 goto failed_mount8; 5549 #ifdef CONFIG_QUOTA 5550 /* Enable quota usage during mount. */ 5551 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { 5552 err = ext4_enable_quotas(sb); 5553 if (err) 5554 goto failed_mount9; 5555 } 5556 #endif /* CONFIG_QUOTA */ 5557 5558 /* 5559 * Save the original bdev mapping's wb_err value which could be 5560 * used to detect the metadata async write error. 5561 */ 5562 spin_lock_init(&sbi->s_bdev_wb_lock); 5563 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err, 5564 &sbi->s_bdev_wb_err); 5565 sb->s_bdev->bd_super = sb; 5566 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 5567 ext4_orphan_cleanup(sb, es); 5568 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 5569 /* 5570 * Update the checksum after updating free space/inode counters and 5571 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect 5572 * checksum in the buffer cache until it is written out and 5573 * e2fsprogs programs trying to open a file system immediately 5574 * after it is mounted can fail. 5575 */ 5576 ext4_superblock_csum_set(sb); 5577 if (needs_recovery) { 5578 ext4_msg(sb, KERN_INFO, "recovery complete"); 5579 err = ext4_mark_recovery_complete(sb, es); 5580 if (err) 5581 goto failed_mount9; 5582 } 5583 5584 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev)) 5585 ext4_msg(sb, KERN_WARNING, 5586 "mounting with \"discard\" option, but the device does not support discard"); 5587 5588 if (es->s_error_count) 5589 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 5590 5591 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 5592 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 5593 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 5594 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 5595 atomic_set(&sbi->s_warning_count, 0); 5596 atomic_set(&sbi->s_msg_count, 0); 5597 5598 return 0; 5599 5600 failed_mount9: 5601 ext4_release_orphan_info(sb); 5602 failed_mount8: 5603 ext4_unregister_sysfs(sb); 5604 kobject_put(&sbi->s_kobj); 5605 failed_mount7: 5606 ext4_unregister_li_request(sb); 5607 failed_mount6: 5608 ext4_mb_release(sb); 5609 ext4_flex_groups_free(sbi); 5610 ext4_percpu_param_destroy(sbi); 5611 failed_mount5: 5612 ext4_ext_release(sb); 5613 ext4_release_system_zone(sb); 5614 failed_mount4a: 5615 dput(sb->s_root); 5616 sb->s_root = NULL; 5617 failed_mount4: 5618 ext4_msg(sb, KERN_ERR, "mount failed"); 5619 if (EXT4_SB(sb)->rsv_conversion_wq) 5620 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 5621 failed_mount_wq: 5622 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 5623 sbi->s_ea_inode_cache = NULL; 5624 5625 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 5626 sbi->s_ea_block_cache = NULL; 5627 5628 if (sbi->s_journal) { 5629 /* flush s_error_work before journal destroy. */ 5630 flush_work(&sbi->s_error_work); 5631 jbd2_journal_destroy(sbi->s_journal); 5632 sbi->s_journal = NULL; 5633 } 5634 failed_mount3a: 5635 ext4_es_unregister_shrinker(sbi); 5636 failed_mount3: 5637 /* flush s_error_work before sbi destroy */ 5638 flush_work(&sbi->s_error_work); 5639 del_timer_sync(&sbi->s_err_report); 5640 ext4_stop_mmpd(sbi); 5641 ext4_group_desc_free(sbi); 5642 failed_mount: 5643 if (sbi->s_chksum_driver) 5644 crypto_free_shash(sbi->s_chksum_driver); 5645 5646 #if IS_ENABLED(CONFIG_UNICODE) 5647 utf8_unload(sb->s_encoding); 5648 #endif 5649 5650 #ifdef CONFIG_QUOTA 5651 for (unsigned int i = 0; i < EXT4_MAXQUOTAS; i++) 5652 kfree(get_qf_name(sb, sbi, i)); 5653 #endif 5654 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy); 5655 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */ 5656 brelse(sbi->s_sbh); 5657 ext4_blkdev_remove(sbi); 5658 out_fail: 5659 sb->s_fs_info = NULL; 5660 return err; 5661 } 5662 5663 static int ext4_fill_super(struct super_block *sb, struct fs_context *fc) 5664 { 5665 struct ext4_fs_context *ctx = fc->fs_private; 5666 struct ext4_sb_info *sbi; 5667 const char *descr; 5668 int ret; 5669 5670 sbi = ext4_alloc_sbi(sb); 5671 if (!sbi) 5672 return -ENOMEM; 5673 5674 fc->s_fs_info = sbi; 5675 5676 /* Cleanup superblock name */ 5677 strreplace(sb->s_id, '/', '!'); 5678 5679 sbi->s_sb_block = 1; /* Default super block location */ 5680 if (ctx->spec & EXT4_SPEC_s_sb_block) 5681 sbi->s_sb_block = ctx->s_sb_block; 5682 5683 ret = __ext4_fill_super(fc, sb); 5684 if (ret < 0) 5685 goto free_sbi; 5686 5687 if (sbi->s_journal) { 5688 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 5689 descr = " journalled data mode"; 5690 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 5691 descr = " ordered data mode"; 5692 else 5693 descr = " writeback data mode"; 5694 } else 5695 descr = "out journal"; 5696 5697 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 5698 ext4_msg(sb, KERN_INFO, "mounted filesystem %pU %s with%s. " 5699 "Quota mode: %s.", &sb->s_uuid, 5700 sb_rdonly(sb) ? "ro" : "r/w", descr, 5701 ext4_quota_mode(sb)); 5702 5703 /* Update the s_overhead_clusters if necessary */ 5704 ext4_update_overhead(sb, false); 5705 return 0; 5706 5707 free_sbi: 5708 ext4_free_sbi(sbi); 5709 fc->s_fs_info = NULL; 5710 return ret; 5711 } 5712 5713 static int ext4_get_tree(struct fs_context *fc) 5714 { 5715 return get_tree_bdev(fc, ext4_fill_super); 5716 } 5717 5718 /* 5719 * Setup any per-fs journal parameters now. We'll do this both on 5720 * initial mount, once the journal has been initialised but before we've 5721 * done any recovery; and again on any subsequent remount. 5722 */ 5723 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 5724 { 5725 struct ext4_sb_info *sbi = EXT4_SB(sb); 5726 5727 journal->j_commit_interval = sbi->s_commit_interval; 5728 journal->j_min_batch_time = sbi->s_min_batch_time; 5729 journal->j_max_batch_time = sbi->s_max_batch_time; 5730 ext4_fc_init(sb, journal); 5731 5732 write_lock(&journal->j_state_lock); 5733 if (test_opt(sb, BARRIER)) 5734 journal->j_flags |= JBD2_BARRIER; 5735 else 5736 journal->j_flags &= ~JBD2_BARRIER; 5737 if (test_opt(sb, DATA_ERR_ABORT)) 5738 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 5739 else 5740 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 5741 write_unlock(&journal->j_state_lock); 5742 } 5743 5744 static struct inode *ext4_get_journal_inode(struct super_block *sb, 5745 unsigned int journal_inum) 5746 { 5747 struct inode *journal_inode; 5748 5749 /* 5750 * Test for the existence of a valid inode on disk. Bad things 5751 * happen if we iget() an unused inode, as the subsequent iput() 5752 * will try to delete it. 5753 */ 5754 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL); 5755 if (IS_ERR(journal_inode)) { 5756 ext4_msg(sb, KERN_ERR, "no journal found"); 5757 return NULL; 5758 } 5759 if (!journal_inode->i_nlink) { 5760 make_bad_inode(journal_inode); 5761 iput(journal_inode); 5762 ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 5763 return NULL; 5764 } 5765 5766 ext4_debug("Journal inode found at %p: %lld bytes\n", 5767 journal_inode, journal_inode->i_size); 5768 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) { 5769 ext4_msg(sb, KERN_ERR, "invalid journal inode"); 5770 iput(journal_inode); 5771 return NULL; 5772 } 5773 return journal_inode; 5774 } 5775 5776 static int ext4_journal_bmap(journal_t *journal, sector_t *block) 5777 { 5778 struct ext4_map_blocks map; 5779 int ret; 5780 5781 if (journal->j_inode == NULL) 5782 return 0; 5783 5784 map.m_lblk = *block; 5785 map.m_len = 1; 5786 ret = ext4_map_blocks(NULL, journal->j_inode, &map, 0); 5787 if (ret <= 0) { 5788 ext4_msg(journal->j_inode->i_sb, KERN_CRIT, 5789 "journal bmap failed: block %llu ret %d\n", 5790 *block, ret); 5791 jbd2_journal_abort(journal, ret ? ret : -EIO); 5792 return ret; 5793 } 5794 *block = map.m_pblk; 5795 return 0; 5796 } 5797 5798 static journal_t *ext4_get_journal(struct super_block *sb, 5799 unsigned int journal_inum) 5800 { 5801 struct inode *journal_inode; 5802 journal_t *journal; 5803 5804 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 5805 return NULL; 5806 5807 journal_inode = ext4_get_journal_inode(sb, journal_inum); 5808 if (!journal_inode) 5809 return NULL; 5810 5811 journal = jbd2_journal_init_inode(journal_inode); 5812 if (!journal) { 5813 ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 5814 iput(journal_inode); 5815 return NULL; 5816 } 5817 journal->j_private = sb; 5818 journal->j_bmap = ext4_journal_bmap; 5819 ext4_init_journal_params(sb, journal); 5820 return journal; 5821 } 5822 5823 static journal_t *ext4_get_dev_journal(struct super_block *sb, 5824 dev_t j_dev) 5825 { 5826 struct buffer_head *bh; 5827 journal_t *journal; 5828 ext4_fsblk_t start; 5829 ext4_fsblk_t len; 5830 int hblock, blocksize; 5831 ext4_fsblk_t sb_block; 5832 unsigned long offset; 5833 struct ext4_super_block *es; 5834 struct block_device *bdev; 5835 5836 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 5837 return NULL; 5838 5839 bdev = ext4_blkdev_get(j_dev, sb); 5840 if (bdev == NULL) 5841 return NULL; 5842 5843 blocksize = sb->s_blocksize; 5844 hblock = bdev_logical_block_size(bdev); 5845 if (blocksize < hblock) { 5846 ext4_msg(sb, KERN_ERR, 5847 "blocksize too small for journal device"); 5848 goto out_bdev; 5849 } 5850 5851 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 5852 offset = EXT4_MIN_BLOCK_SIZE % blocksize; 5853 set_blocksize(bdev, blocksize); 5854 if (!(bh = __bread(bdev, sb_block, blocksize))) { 5855 ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 5856 "external journal"); 5857 goto out_bdev; 5858 } 5859 5860 es = (struct ext4_super_block *) (bh->b_data + offset); 5861 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 5862 !(le32_to_cpu(es->s_feature_incompat) & 5863 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 5864 ext4_msg(sb, KERN_ERR, "external journal has " 5865 "bad superblock"); 5866 brelse(bh); 5867 goto out_bdev; 5868 } 5869 5870 if ((le32_to_cpu(es->s_feature_ro_compat) & 5871 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 5872 es->s_checksum != ext4_superblock_csum(sb, es)) { 5873 ext4_msg(sb, KERN_ERR, "external journal has " 5874 "corrupt superblock"); 5875 brelse(bh); 5876 goto out_bdev; 5877 } 5878 5879 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 5880 ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 5881 brelse(bh); 5882 goto out_bdev; 5883 } 5884 5885 len = ext4_blocks_count(es); 5886 start = sb_block + 1; 5887 brelse(bh); /* we're done with the superblock */ 5888 5889 journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 5890 start, len, blocksize); 5891 if (!journal) { 5892 ext4_msg(sb, KERN_ERR, "failed to create device journal"); 5893 goto out_bdev; 5894 } 5895 journal->j_private = sb; 5896 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) { 5897 ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 5898 goto out_journal; 5899 } 5900 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 5901 ext4_msg(sb, KERN_ERR, "External journal has more than one " 5902 "user (unsupported) - %d", 5903 be32_to_cpu(journal->j_superblock->s_nr_users)); 5904 goto out_journal; 5905 } 5906 EXT4_SB(sb)->s_journal_bdev = bdev; 5907 ext4_init_journal_params(sb, journal); 5908 return journal; 5909 5910 out_journal: 5911 jbd2_journal_destroy(journal); 5912 out_bdev: 5913 blkdev_put(bdev, sb); 5914 return NULL; 5915 } 5916 5917 static int ext4_load_journal(struct super_block *sb, 5918 struct ext4_super_block *es, 5919 unsigned long journal_devnum) 5920 { 5921 journal_t *journal; 5922 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 5923 dev_t journal_dev; 5924 int err = 0; 5925 int really_read_only; 5926 int journal_dev_ro; 5927 5928 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 5929 return -EFSCORRUPTED; 5930 5931 if (journal_devnum && 5932 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 5933 ext4_msg(sb, KERN_INFO, "external journal device major/minor " 5934 "numbers have changed"); 5935 journal_dev = new_decode_dev(journal_devnum); 5936 } else 5937 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 5938 5939 if (journal_inum && journal_dev) { 5940 ext4_msg(sb, KERN_ERR, 5941 "filesystem has both journal inode and journal device!"); 5942 return -EINVAL; 5943 } 5944 5945 if (journal_inum) { 5946 journal = ext4_get_journal(sb, journal_inum); 5947 if (!journal) 5948 return -EINVAL; 5949 } else { 5950 journal = ext4_get_dev_journal(sb, journal_dev); 5951 if (!journal) 5952 return -EINVAL; 5953 } 5954 5955 journal_dev_ro = bdev_read_only(journal->j_dev); 5956 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro; 5957 5958 if (journal_dev_ro && !sb_rdonly(sb)) { 5959 ext4_msg(sb, KERN_ERR, 5960 "journal device read-only, try mounting with '-o ro'"); 5961 err = -EROFS; 5962 goto err_out; 5963 } 5964 5965 /* 5966 * Are we loading a blank journal or performing recovery after a 5967 * crash? For recovery, we need to check in advance whether we 5968 * can get read-write access to the device. 5969 */ 5970 if (ext4_has_feature_journal_needs_recovery(sb)) { 5971 if (sb_rdonly(sb)) { 5972 ext4_msg(sb, KERN_INFO, "INFO: recovery " 5973 "required on readonly filesystem"); 5974 if (really_read_only) { 5975 ext4_msg(sb, KERN_ERR, "write access " 5976 "unavailable, cannot proceed " 5977 "(try mounting with noload)"); 5978 err = -EROFS; 5979 goto err_out; 5980 } 5981 ext4_msg(sb, KERN_INFO, "write access will " 5982 "be enabled during recovery"); 5983 } 5984 } 5985 5986 if (!(journal->j_flags & JBD2_BARRIER)) 5987 ext4_msg(sb, KERN_INFO, "barriers disabled"); 5988 5989 if (!ext4_has_feature_journal_needs_recovery(sb)) 5990 err = jbd2_journal_wipe(journal, !really_read_only); 5991 if (!err) { 5992 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 5993 5994 if (save) 5995 memcpy(save, ((char *) es) + 5996 EXT4_S_ERR_START, EXT4_S_ERR_LEN); 5997 err = jbd2_journal_load(journal); 5998 if (save) 5999 memcpy(((char *) es) + EXT4_S_ERR_START, 6000 save, EXT4_S_ERR_LEN); 6001 kfree(save); 6002 es->s_state |= cpu_to_le16(EXT4_SB(sb)->s_mount_state & 6003 EXT4_ERROR_FS); 6004 /* Write out restored error information to the superblock */ 6005 if (!bdev_read_only(sb->s_bdev)) { 6006 int err2; 6007 err2 = ext4_commit_super(sb); 6008 err = err ? : err2; 6009 } 6010 } 6011 6012 if (err) { 6013 ext4_msg(sb, KERN_ERR, "error loading journal"); 6014 goto err_out; 6015 } 6016 6017 EXT4_SB(sb)->s_journal = journal; 6018 err = ext4_clear_journal_err(sb, es); 6019 if (err) { 6020 EXT4_SB(sb)->s_journal = NULL; 6021 jbd2_journal_destroy(journal); 6022 return err; 6023 } 6024 6025 if (!really_read_only && journal_devnum && 6026 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 6027 es->s_journal_dev = cpu_to_le32(journal_devnum); 6028 ext4_commit_super(sb); 6029 } 6030 if (!really_read_only && journal_inum && 6031 journal_inum != le32_to_cpu(es->s_journal_inum)) { 6032 es->s_journal_inum = cpu_to_le32(journal_inum); 6033 ext4_commit_super(sb); 6034 } 6035 6036 return 0; 6037 6038 err_out: 6039 jbd2_journal_destroy(journal); 6040 return err; 6041 } 6042 6043 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */ 6044 static void ext4_update_super(struct super_block *sb) 6045 { 6046 struct ext4_sb_info *sbi = EXT4_SB(sb); 6047 struct ext4_super_block *es = sbi->s_es; 6048 struct buffer_head *sbh = sbi->s_sbh; 6049 6050 lock_buffer(sbh); 6051 /* 6052 * If the file system is mounted read-only, don't update the 6053 * superblock write time. This avoids updating the superblock 6054 * write time when we are mounting the root file system 6055 * read/only but we need to replay the journal; at that point, 6056 * for people who are east of GMT and who make their clock 6057 * tick in localtime for Windows bug-for-bug compatibility, 6058 * the clock is set in the future, and this will cause e2fsck 6059 * to complain and force a full file system check. 6060 */ 6061 if (!(sb->s_flags & SB_RDONLY)) 6062 ext4_update_tstamp(es, s_wtime); 6063 es->s_kbytes_written = 6064 cpu_to_le64(sbi->s_kbytes_written + 6065 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) - 6066 sbi->s_sectors_written_start) >> 1)); 6067 if (percpu_counter_initialized(&sbi->s_freeclusters_counter)) 6068 ext4_free_blocks_count_set(es, 6069 EXT4_C2B(sbi, percpu_counter_sum_positive( 6070 &sbi->s_freeclusters_counter))); 6071 if (percpu_counter_initialized(&sbi->s_freeinodes_counter)) 6072 es->s_free_inodes_count = 6073 cpu_to_le32(percpu_counter_sum_positive( 6074 &sbi->s_freeinodes_counter)); 6075 /* Copy error information to the on-disk superblock */ 6076 spin_lock(&sbi->s_error_lock); 6077 if (sbi->s_add_error_count > 0) { 6078 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 6079 if (!es->s_first_error_time && !es->s_first_error_time_hi) { 6080 __ext4_update_tstamp(&es->s_first_error_time, 6081 &es->s_first_error_time_hi, 6082 sbi->s_first_error_time); 6083 strncpy(es->s_first_error_func, sbi->s_first_error_func, 6084 sizeof(es->s_first_error_func)); 6085 es->s_first_error_line = 6086 cpu_to_le32(sbi->s_first_error_line); 6087 es->s_first_error_ino = 6088 cpu_to_le32(sbi->s_first_error_ino); 6089 es->s_first_error_block = 6090 cpu_to_le64(sbi->s_first_error_block); 6091 es->s_first_error_errcode = 6092 ext4_errno_to_code(sbi->s_first_error_code); 6093 } 6094 __ext4_update_tstamp(&es->s_last_error_time, 6095 &es->s_last_error_time_hi, 6096 sbi->s_last_error_time); 6097 strncpy(es->s_last_error_func, sbi->s_last_error_func, 6098 sizeof(es->s_last_error_func)); 6099 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line); 6100 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino); 6101 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block); 6102 es->s_last_error_errcode = 6103 ext4_errno_to_code(sbi->s_last_error_code); 6104 /* 6105 * Start the daily error reporting function if it hasn't been 6106 * started already 6107 */ 6108 if (!es->s_error_count) 6109 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); 6110 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count); 6111 sbi->s_add_error_count = 0; 6112 } 6113 spin_unlock(&sbi->s_error_lock); 6114 6115 ext4_superblock_csum_set(sb); 6116 unlock_buffer(sbh); 6117 } 6118 6119 static int ext4_commit_super(struct super_block *sb) 6120 { 6121 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 6122 6123 if (!sbh) 6124 return -EINVAL; 6125 if (block_device_ejected(sb)) 6126 return -ENODEV; 6127 6128 ext4_update_super(sb); 6129 6130 lock_buffer(sbh); 6131 /* Buffer got discarded which means block device got invalidated */ 6132 if (!buffer_mapped(sbh)) { 6133 unlock_buffer(sbh); 6134 return -EIO; 6135 } 6136 6137 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 6138 /* 6139 * Oh, dear. A previous attempt to write the 6140 * superblock failed. This could happen because the 6141 * USB device was yanked out. Or it could happen to 6142 * be a transient write error and maybe the block will 6143 * be remapped. Nothing we can do but to retry the 6144 * write and hope for the best. 6145 */ 6146 ext4_msg(sb, KERN_ERR, "previous I/O error to " 6147 "superblock detected"); 6148 clear_buffer_write_io_error(sbh); 6149 set_buffer_uptodate(sbh); 6150 } 6151 get_bh(sbh); 6152 /* Clear potential dirty bit if it was journalled update */ 6153 clear_buffer_dirty(sbh); 6154 sbh->b_end_io = end_buffer_write_sync; 6155 submit_bh(REQ_OP_WRITE | REQ_SYNC | 6156 (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh); 6157 wait_on_buffer(sbh); 6158 if (buffer_write_io_error(sbh)) { 6159 ext4_msg(sb, KERN_ERR, "I/O error while writing " 6160 "superblock"); 6161 clear_buffer_write_io_error(sbh); 6162 set_buffer_uptodate(sbh); 6163 return -EIO; 6164 } 6165 return 0; 6166 } 6167 6168 /* 6169 * Have we just finished recovery? If so, and if we are mounting (or 6170 * remounting) the filesystem readonly, then we will end up with a 6171 * consistent fs on disk. Record that fact. 6172 */ 6173 static int ext4_mark_recovery_complete(struct super_block *sb, 6174 struct ext4_super_block *es) 6175 { 6176 int err; 6177 journal_t *journal = EXT4_SB(sb)->s_journal; 6178 6179 if (!ext4_has_feature_journal(sb)) { 6180 if (journal != NULL) { 6181 ext4_error(sb, "Journal got removed while the fs was " 6182 "mounted!"); 6183 return -EFSCORRUPTED; 6184 } 6185 return 0; 6186 } 6187 jbd2_journal_lock_updates(journal); 6188 err = jbd2_journal_flush(journal, 0); 6189 if (err < 0) 6190 goto out; 6191 6192 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) || 6193 ext4_has_feature_orphan_present(sb))) { 6194 if (!ext4_orphan_file_empty(sb)) { 6195 ext4_error(sb, "Orphan file not empty on read-only fs."); 6196 err = -EFSCORRUPTED; 6197 goto out; 6198 } 6199 ext4_clear_feature_journal_needs_recovery(sb); 6200 ext4_clear_feature_orphan_present(sb); 6201 ext4_commit_super(sb); 6202 } 6203 out: 6204 jbd2_journal_unlock_updates(journal); 6205 return err; 6206 } 6207 6208 /* 6209 * If we are mounting (or read-write remounting) a filesystem whose journal 6210 * has recorded an error from a previous lifetime, move that error to the 6211 * main filesystem now. 6212 */ 6213 static int ext4_clear_journal_err(struct super_block *sb, 6214 struct ext4_super_block *es) 6215 { 6216 journal_t *journal; 6217 int j_errno; 6218 const char *errstr; 6219 6220 if (!ext4_has_feature_journal(sb)) { 6221 ext4_error(sb, "Journal got removed while the fs was mounted!"); 6222 return -EFSCORRUPTED; 6223 } 6224 6225 journal = EXT4_SB(sb)->s_journal; 6226 6227 /* 6228 * Now check for any error status which may have been recorded in the 6229 * journal by a prior ext4_error() or ext4_abort() 6230 */ 6231 6232 j_errno = jbd2_journal_errno(journal); 6233 if (j_errno) { 6234 char nbuf[16]; 6235 6236 errstr = ext4_decode_error(sb, j_errno, nbuf); 6237 ext4_warning(sb, "Filesystem error recorded " 6238 "from previous mount: %s", errstr); 6239 6240 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 6241 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 6242 j_errno = ext4_commit_super(sb); 6243 if (j_errno) 6244 return j_errno; 6245 ext4_warning(sb, "Marked fs in need of filesystem check."); 6246 6247 jbd2_journal_clear_err(journal); 6248 jbd2_journal_update_sb_errno(journal); 6249 } 6250 return 0; 6251 } 6252 6253 /* 6254 * Force the running and committing transactions to commit, 6255 * and wait on the commit. 6256 */ 6257 int ext4_force_commit(struct super_block *sb) 6258 { 6259 journal_t *journal; 6260 6261 if (sb_rdonly(sb)) 6262 return 0; 6263 6264 journal = EXT4_SB(sb)->s_journal; 6265 return ext4_journal_force_commit(journal); 6266 } 6267 6268 static int ext4_sync_fs(struct super_block *sb, int wait) 6269 { 6270 int ret = 0; 6271 tid_t target; 6272 bool needs_barrier = false; 6273 struct ext4_sb_info *sbi = EXT4_SB(sb); 6274 6275 if (unlikely(ext4_forced_shutdown(sbi))) 6276 return 0; 6277 6278 trace_ext4_sync_fs(sb, wait); 6279 flush_workqueue(sbi->rsv_conversion_wq); 6280 /* 6281 * Writeback quota in non-journalled quota case - journalled quota has 6282 * no dirty dquots 6283 */ 6284 dquot_writeback_dquots(sb, -1); 6285 /* 6286 * Data writeback is possible w/o journal transaction, so barrier must 6287 * being sent at the end of the function. But we can skip it if 6288 * transaction_commit will do it for us. 6289 */ 6290 if (sbi->s_journal) { 6291 target = jbd2_get_latest_transaction(sbi->s_journal); 6292 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 6293 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 6294 needs_barrier = true; 6295 6296 if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 6297 if (wait) 6298 ret = jbd2_log_wait_commit(sbi->s_journal, 6299 target); 6300 } 6301 } else if (wait && test_opt(sb, BARRIER)) 6302 needs_barrier = true; 6303 if (needs_barrier) { 6304 int err; 6305 err = blkdev_issue_flush(sb->s_bdev); 6306 if (!ret) 6307 ret = err; 6308 } 6309 6310 return ret; 6311 } 6312 6313 /* 6314 * LVM calls this function before a (read-only) snapshot is created. This 6315 * gives us a chance to flush the journal completely and mark the fs clean. 6316 * 6317 * Note that only this function cannot bring a filesystem to be in a clean 6318 * state independently. It relies on upper layer to stop all data & metadata 6319 * modifications. 6320 */ 6321 static int ext4_freeze(struct super_block *sb) 6322 { 6323 int error = 0; 6324 journal_t *journal; 6325 6326 if (sb_rdonly(sb)) 6327 return 0; 6328 6329 journal = EXT4_SB(sb)->s_journal; 6330 6331 if (journal) { 6332 /* Now we set up the journal barrier. */ 6333 jbd2_journal_lock_updates(journal); 6334 6335 /* 6336 * Don't clear the needs_recovery flag if we failed to 6337 * flush the journal. 6338 */ 6339 error = jbd2_journal_flush(journal, 0); 6340 if (error < 0) 6341 goto out; 6342 6343 /* Journal blocked and flushed, clear needs_recovery flag. */ 6344 ext4_clear_feature_journal_needs_recovery(sb); 6345 if (ext4_orphan_file_empty(sb)) 6346 ext4_clear_feature_orphan_present(sb); 6347 } 6348 6349 error = ext4_commit_super(sb); 6350 out: 6351 if (journal) 6352 /* we rely on upper layer to stop further updates */ 6353 jbd2_journal_unlock_updates(journal); 6354 return error; 6355 } 6356 6357 /* 6358 * Called by LVM after the snapshot is done. We need to reset the RECOVER 6359 * flag here, even though the filesystem is not technically dirty yet. 6360 */ 6361 static int ext4_unfreeze(struct super_block *sb) 6362 { 6363 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb))) 6364 return 0; 6365 6366 if (EXT4_SB(sb)->s_journal) { 6367 /* Reset the needs_recovery flag before the fs is unlocked. */ 6368 ext4_set_feature_journal_needs_recovery(sb); 6369 if (ext4_has_feature_orphan_file(sb)) 6370 ext4_set_feature_orphan_present(sb); 6371 } 6372 6373 ext4_commit_super(sb); 6374 return 0; 6375 } 6376 6377 /* 6378 * Structure to save mount options for ext4_remount's benefit 6379 */ 6380 struct ext4_mount_options { 6381 unsigned long s_mount_opt; 6382 unsigned long s_mount_opt2; 6383 kuid_t s_resuid; 6384 kgid_t s_resgid; 6385 unsigned long s_commit_interval; 6386 u32 s_min_batch_time, s_max_batch_time; 6387 #ifdef CONFIG_QUOTA 6388 int s_jquota_fmt; 6389 char *s_qf_names[EXT4_MAXQUOTAS]; 6390 #endif 6391 }; 6392 6393 static int __ext4_remount(struct fs_context *fc, struct super_block *sb) 6394 { 6395 struct ext4_fs_context *ctx = fc->fs_private; 6396 struct ext4_super_block *es; 6397 struct ext4_sb_info *sbi = EXT4_SB(sb); 6398 unsigned long old_sb_flags; 6399 struct ext4_mount_options old_opts; 6400 ext4_group_t g; 6401 int err = 0; 6402 #ifdef CONFIG_QUOTA 6403 int enable_quota = 0; 6404 int i, j; 6405 char *to_free[EXT4_MAXQUOTAS]; 6406 #endif 6407 6408 6409 /* Store the original options */ 6410 old_sb_flags = sb->s_flags; 6411 old_opts.s_mount_opt = sbi->s_mount_opt; 6412 old_opts.s_mount_opt2 = sbi->s_mount_opt2; 6413 old_opts.s_resuid = sbi->s_resuid; 6414 old_opts.s_resgid = sbi->s_resgid; 6415 old_opts.s_commit_interval = sbi->s_commit_interval; 6416 old_opts.s_min_batch_time = sbi->s_min_batch_time; 6417 old_opts.s_max_batch_time = sbi->s_max_batch_time; 6418 #ifdef CONFIG_QUOTA 6419 old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 6420 for (i = 0; i < EXT4_MAXQUOTAS; i++) 6421 if (sbi->s_qf_names[i]) { 6422 char *qf_name = get_qf_name(sb, sbi, i); 6423 6424 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); 6425 if (!old_opts.s_qf_names[i]) { 6426 for (j = 0; j < i; j++) 6427 kfree(old_opts.s_qf_names[j]); 6428 return -ENOMEM; 6429 } 6430 } else 6431 old_opts.s_qf_names[i] = NULL; 6432 #endif 6433 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) { 6434 if (sbi->s_journal && sbi->s_journal->j_task->io_context) 6435 ctx->journal_ioprio = 6436 sbi->s_journal->j_task->io_context->ioprio; 6437 else 6438 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 6439 6440 } 6441 6442 ext4_apply_options(fc, sb); 6443 6444 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 6445 test_opt(sb, JOURNAL_CHECKSUM)) { 6446 ext4_msg(sb, KERN_ERR, "changing journal_checksum " 6447 "during remount not supported; ignoring"); 6448 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 6449 } 6450 6451 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 6452 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 6453 ext4_msg(sb, KERN_ERR, "can't mount with " 6454 "both data=journal and delalloc"); 6455 err = -EINVAL; 6456 goto restore_opts; 6457 } 6458 if (test_opt(sb, DIOREAD_NOLOCK)) { 6459 ext4_msg(sb, KERN_ERR, "can't mount with " 6460 "both data=journal and dioread_nolock"); 6461 err = -EINVAL; 6462 goto restore_opts; 6463 } 6464 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { 6465 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 6466 ext4_msg(sb, KERN_ERR, "can't mount with " 6467 "journal_async_commit in data=ordered mode"); 6468 err = -EINVAL; 6469 goto restore_opts; 6470 } 6471 } 6472 6473 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { 6474 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); 6475 err = -EINVAL; 6476 goto restore_opts; 6477 } 6478 6479 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) 6480 ext4_abort(sb, ESHUTDOWN, "Abort forced by user"); 6481 6482 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 6483 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 6484 6485 es = sbi->s_es; 6486 6487 if (sbi->s_journal) { 6488 ext4_init_journal_params(sb, sbi->s_journal); 6489 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio); 6490 } 6491 6492 /* Flush outstanding errors before changing fs state */ 6493 flush_work(&sbi->s_error_work); 6494 6495 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) { 6496 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) { 6497 err = -EROFS; 6498 goto restore_opts; 6499 } 6500 6501 if (fc->sb_flags & SB_RDONLY) { 6502 err = sync_filesystem(sb); 6503 if (err < 0) 6504 goto restore_opts; 6505 err = dquot_suspend(sb, -1); 6506 if (err < 0) 6507 goto restore_opts; 6508 6509 /* 6510 * First of all, the unconditional stuff we have to do 6511 * to disable replay of the journal when we next remount 6512 */ 6513 sb->s_flags |= SB_RDONLY; 6514 6515 /* 6516 * OK, test if we are remounting a valid rw partition 6517 * readonly, and if so set the rdonly flag and then 6518 * mark the partition as valid again. 6519 */ 6520 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 6521 (sbi->s_mount_state & EXT4_VALID_FS)) 6522 es->s_state = cpu_to_le16(sbi->s_mount_state); 6523 6524 if (sbi->s_journal) { 6525 /* 6526 * We let remount-ro finish even if marking fs 6527 * as clean failed... 6528 */ 6529 ext4_mark_recovery_complete(sb, es); 6530 } 6531 } else { 6532 /* Make sure we can mount this feature set readwrite */ 6533 if (ext4_has_feature_readonly(sb) || 6534 !ext4_feature_set_ok(sb, 0)) { 6535 err = -EROFS; 6536 goto restore_opts; 6537 } 6538 /* 6539 * Make sure the group descriptor checksums 6540 * are sane. If they aren't, refuse to remount r/w. 6541 */ 6542 for (g = 0; g < sbi->s_groups_count; g++) { 6543 struct ext4_group_desc *gdp = 6544 ext4_get_group_desc(sb, g, NULL); 6545 6546 if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 6547 ext4_msg(sb, KERN_ERR, 6548 "ext4_remount: Checksum for group %u failed (%u!=%u)", 6549 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 6550 le16_to_cpu(gdp->bg_checksum)); 6551 err = -EFSBADCRC; 6552 goto restore_opts; 6553 } 6554 } 6555 6556 /* 6557 * If we have an unprocessed orphan list hanging 6558 * around from a previously readonly bdev mount, 6559 * require a full umount/remount for now. 6560 */ 6561 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) { 6562 ext4_msg(sb, KERN_WARNING, "Couldn't " 6563 "remount RDWR because of unprocessed " 6564 "orphan inode list. Please " 6565 "umount/remount instead"); 6566 err = -EINVAL; 6567 goto restore_opts; 6568 } 6569 6570 /* 6571 * Mounting a RDONLY partition read-write, so reread 6572 * and store the current valid flag. (It may have 6573 * been changed by e2fsck since we originally mounted 6574 * the partition.) 6575 */ 6576 if (sbi->s_journal) { 6577 err = ext4_clear_journal_err(sb, es); 6578 if (err) 6579 goto restore_opts; 6580 } 6581 sbi->s_mount_state = (le16_to_cpu(es->s_state) & 6582 ~EXT4_FC_REPLAY); 6583 6584 err = ext4_setup_super(sb, es, 0); 6585 if (err) 6586 goto restore_opts; 6587 6588 sb->s_flags &= ~SB_RDONLY; 6589 if (ext4_has_feature_mmp(sb)) { 6590 err = ext4_multi_mount_protect(sb, 6591 le64_to_cpu(es->s_mmp_block)); 6592 if (err) 6593 goto restore_opts; 6594 } 6595 #ifdef CONFIG_QUOTA 6596 enable_quota = 1; 6597 #endif 6598 } 6599 } 6600 6601 /* 6602 * Handle creation of system zone data early because it can fail. 6603 * Releasing of existing data is done when we are sure remount will 6604 * succeed. 6605 */ 6606 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) { 6607 err = ext4_setup_system_zone(sb); 6608 if (err) 6609 goto restore_opts; 6610 } 6611 6612 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) { 6613 err = ext4_commit_super(sb); 6614 if (err) 6615 goto restore_opts; 6616 } 6617 6618 #ifdef CONFIG_QUOTA 6619 if (enable_quota) { 6620 if (sb_any_quota_suspended(sb)) 6621 dquot_resume(sb, -1); 6622 else if (ext4_has_feature_quota(sb)) { 6623 err = ext4_enable_quotas(sb); 6624 if (err) 6625 goto restore_opts; 6626 } 6627 } 6628 /* Release old quota file names */ 6629 for (i = 0; i < EXT4_MAXQUOTAS; i++) 6630 kfree(old_opts.s_qf_names[i]); 6631 #endif 6632 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 6633 ext4_release_system_zone(sb); 6634 6635 /* 6636 * Reinitialize lazy itable initialization thread based on 6637 * current settings 6638 */ 6639 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) 6640 ext4_unregister_li_request(sb); 6641 else { 6642 ext4_group_t first_not_zeroed; 6643 first_not_zeroed = ext4_has_uninit_itable(sb); 6644 ext4_register_li_request(sb, first_not_zeroed); 6645 } 6646 6647 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb)) 6648 ext4_stop_mmpd(sbi); 6649 6650 return 0; 6651 6652 restore_opts: 6653 /* 6654 * If there was a failing r/w to ro transition, we may need to 6655 * re-enable quota 6656 */ 6657 if ((sb->s_flags & SB_RDONLY) && !(old_sb_flags & SB_RDONLY) && 6658 sb_any_quota_suspended(sb)) 6659 dquot_resume(sb, -1); 6660 sb->s_flags = old_sb_flags; 6661 sbi->s_mount_opt = old_opts.s_mount_opt; 6662 sbi->s_mount_opt2 = old_opts.s_mount_opt2; 6663 sbi->s_resuid = old_opts.s_resuid; 6664 sbi->s_resgid = old_opts.s_resgid; 6665 sbi->s_commit_interval = old_opts.s_commit_interval; 6666 sbi->s_min_batch_time = old_opts.s_min_batch_time; 6667 sbi->s_max_batch_time = old_opts.s_max_batch_time; 6668 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 6669 ext4_release_system_zone(sb); 6670 #ifdef CONFIG_QUOTA 6671 sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 6672 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 6673 to_free[i] = get_qf_name(sb, sbi, i); 6674 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); 6675 } 6676 synchronize_rcu(); 6677 for (i = 0; i < EXT4_MAXQUOTAS; i++) 6678 kfree(to_free[i]); 6679 #endif 6680 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb)) 6681 ext4_stop_mmpd(sbi); 6682 return err; 6683 } 6684 6685 static int ext4_reconfigure(struct fs_context *fc) 6686 { 6687 struct super_block *sb = fc->root->d_sb; 6688 int ret; 6689 6690 fc->s_fs_info = EXT4_SB(sb); 6691 6692 ret = ext4_check_opt_consistency(fc, sb); 6693 if (ret < 0) 6694 return ret; 6695 6696 ret = __ext4_remount(fc, sb); 6697 if (ret < 0) 6698 return ret; 6699 6700 ext4_msg(sb, KERN_INFO, "re-mounted %pU %s. Quota mode: %s.", 6701 &sb->s_uuid, sb_rdonly(sb) ? "ro" : "r/w", 6702 ext4_quota_mode(sb)); 6703 6704 return 0; 6705 } 6706 6707 #ifdef CONFIG_QUOTA 6708 static int ext4_statfs_project(struct super_block *sb, 6709 kprojid_t projid, struct kstatfs *buf) 6710 { 6711 struct kqid qid; 6712 struct dquot *dquot; 6713 u64 limit; 6714 u64 curblock; 6715 6716 qid = make_kqid_projid(projid); 6717 dquot = dqget(sb, qid); 6718 if (IS_ERR(dquot)) 6719 return PTR_ERR(dquot); 6720 spin_lock(&dquot->dq_dqb_lock); 6721 6722 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 6723 dquot->dq_dqb.dqb_bhardlimit); 6724 limit >>= sb->s_blocksize_bits; 6725 6726 if (limit && buf->f_blocks > limit) { 6727 curblock = (dquot->dq_dqb.dqb_curspace + 6728 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 6729 buf->f_blocks = limit; 6730 buf->f_bfree = buf->f_bavail = 6731 (buf->f_blocks > curblock) ? 6732 (buf->f_blocks - curblock) : 0; 6733 } 6734 6735 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 6736 dquot->dq_dqb.dqb_ihardlimit); 6737 if (limit && buf->f_files > limit) { 6738 buf->f_files = limit; 6739 buf->f_ffree = 6740 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 6741 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 6742 } 6743 6744 spin_unlock(&dquot->dq_dqb_lock); 6745 dqput(dquot); 6746 return 0; 6747 } 6748 #endif 6749 6750 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 6751 { 6752 struct super_block *sb = dentry->d_sb; 6753 struct ext4_sb_info *sbi = EXT4_SB(sb); 6754 struct ext4_super_block *es = sbi->s_es; 6755 ext4_fsblk_t overhead = 0, resv_blocks; 6756 s64 bfree; 6757 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 6758 6759 if (!test_opt(sb, MINIX_DF)) 6760 overhead = sbi->s_overhead; 6761 6762 buf->f_type = EXT4_SUPER_MAGIC; 6763 buf->f_bsize = sb->s_blocksize; 6764 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 6765 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 6766 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 6767 /* prevent underflow in case that few free space is available */ 6768 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 6769 buf->f_bavail = buf->f_bfree - 6770 (ext4_r_blocks_count(es) + resv_blocks); 6771 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 6772 buf->f_bavail = 0; 6773 buf->f_files = le32_to_cpu(es->s_inodes_count); 6774 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 6775 buf->f_namelen = EXT4_NAME_LEN; 6776 buf->f_fsid = uuid_to_fsid(es->s_uuid); 6777 6778 #ifdef CONFIG_QUOTA 6779 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 6780 sb_has_quota_limits_enabled(sb, PRJQUOTA)) 6781 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 6782 #endif 6783 return 0; 6784 } 6785 6786 6787 #ifdef CONFIG_QUOTA 6788 6789 /* 6790 * Helper functions so that transaction is started before we acquire dqio_sem 6791 * to keep correct lock ordering of transaction > dqio_sem 6792 */ 6793 static inline struct inode *dquot_to_inode(struct dquot *dquot) 6794 { 6795 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 6796 } 6797 6798 static int ext4_write_dquot(struct dquot *dquot) 6799 { 6800 int ret, err; 6801 handle_t *handle; 6802 struct inode *inode; 6803 6804 inode = dquot_to_inode(dquot); 6805 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 6806 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 6807 if (IS_ERR(handle)) 6808 return PTR_ERR(handle); 6809 ret = dquot_commit(dquot); 6810 err = ext4_journal_stop(handle); 6811 if (!ret) 6812 ret = err; 6813 return ret; 6814 } 6815 6816 static int ext4_acquire_dquot(struct dquot *dquot) 6817 { 6818 int ret, err; 6819 handle_t *handle; 6820 6821 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 6822 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 6823 if (IS_ERR(handle)) 6824 return PTR_ERR(handle); 6825 ret = dquot_acquire(dquot); 6826 err = ext4_journal_stop(handle); 6827 if (!ret) 6828 ret = err; 6829 return ret; 6830 } 6831 6832 static int ext4_release_dquot(struct dquot *dquot) 6833 { 6834 int ret, err; 6835 handle_t *handle; 6836 6837 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 6838 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 6839 if (IS_ERR(handle)) { 6840 /* Release dquot anyway to avoid endless cycle in dqput() */ 6841 dquot_release(dquot); 6842 return PTR_ERR(handle); 6843 } 6844 ret = dquot_release(dquot); 6845 err = ext4_journal_stop(handle); 6846 if (!ret) 6847 ret = err; 6848 return ret; 6849 } 6850 6851 static int ext4_mark_dquot_dirty(struct dquot *dquot) 6852 { 6853 struct super_block *sb = dquot->dq_sb; 6854 6855 if (ext4_is_quota_journalled(sb)) { 6856 dquot_mark_dquot_dirty(dquot); 6857 return ext4_write_dquot(dquot); 6858 } else { 6859 return dquot_mark_dquot_dirty(dquot); 6860 } 6861 } 6862 6863 static int ext4_write_info(struct super_block *sb, int type) 6864 { 6865 int ret, err; 6866 handle_t *handle; 6867 6868 /* Data block + inode block */ 6869 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2); 6870 if (IS_ERR(handle)) 6871 return PTR_ERR(handle); 6872 ret = dquot_commit_info(sb, type); 6873 err = ext4_journal_stop(handle); 6874 if (!ret) 6875 ret = err; 6876 return ret; 6877 } 6878 6879 static void lockdep_set_quota_inode(struct inode *inode, int subclass) 6880 { 6881 struct ext4_inode_info *ei = EXT4_I(inode); 6882 6883 /* The first argument of lockdep_set_subclass has to be 6884 * *exactly* the same as the argument to init_rwsem() --- in 6885 * this case, in init_once() --- or lockdep gets unhappy 6886 * because the name of the lock is set using the 6887 * stringification of the argument to init_rwsem(). 6888 */ 6889 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 6890 lockdep_set_subclass(&ei->i_data_sem, subclass); 6891 } 6892 6893 /* 6894 * Standard function to be called on quota_on 6895 */ 6896 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 6897 const struct path *path) 6898 { 6899 int err; 6900 6901 if (!test_opt(sb, QUOTA)) 6902 return -EINVAL; 6903 6904 /* Quotafile not on the same filesystem? */ 6905 if (path->dentry->d_sb != sb) 6906 return -EXDEV; 6907 6908 /* Quota already enabled for this file? */ 6909 if (IS_NOQUOTA(d_inode(path->dentry))) 6910 return -EBUSY; 6911 6912 /* Journaling quota? */ 6913 if (EXT4_SB(sb)->s_qf_names[type]) { 6914 /* Quotafile not in fs root? */ 6915 if (path->dentry->d_parent != sb->s_root) 6916 ext4_msg(sb, KERN_WARNING, 6917 "Quota file not on filesystem root. " 6918 "Journaled quota will not work"); 6919 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY; 6920 } else { 6921 /* 6922 * Clear the flag just in case mount options changed since 6923 * last time. 6924 */ 6925 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY; 6926 } 6927 6928 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 6929 err = dquot_quota_on(sb, type, format_id, path); 6930 if (!err) { 6931 struct inode *inode = d_inode(path->dentry); 6932 handle_t *handle; 6933 6934 /* 6935 * Set inode flags to prevent userspace from messing with quota 6936 * files. If this fails, we return success anyway since quotas 6937 * are already enabled and this is not a hard failure. 6938 */ 6939 inode_lock(inode); 6940 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 6941 if (IS_ERR(handle)) 6942 goto unlock_inode; 6943 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; 6944 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 6945 S_NOATIME | S_IMMUTABLE); 6946 err = ext4_mark_inode_dirty(handle, inode); 6947 ext4_journal_stop(handle); 6948 unlock_inode: 6949 inode_unlock(inode); 6950 if (err) 6951 dquot_quota_off(sb, type); 6952 } 6953 if (err) 6954 lockdep_set_quota_inode(path->dentry->d_inode, 6955 I_DATA_SEM_NORMAL); 6956 return err; 6957 } 6958 6959 static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum) 6960 { 6961 switch (type) { 6962 case USRQUOTA: 6963 return qf_inum == EXT4_USR_QUOTA_INO; 6964 case GRPQUOTA: 6965 return qf_inum == EXT4_GRP_QUOTA_INO; 6966 case PRJQUOTA: 6967 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO; 6968 default: 6969 BUG(); 6970 } 6971 } 6972 6973 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 6974 unsigned int flags) 6975 { 6976 int err; 6977 struct inode *qf_inode; 6978 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 6979 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 6980 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 6981 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 6982 }; 6983 6984 BUG_ON(!ext4_has_feature_quota(sb)); 6985 6986 if (!qf_inums[type]) 6987 return -EPERM; 6988 6989 if (!ext4_check_quota_inum(type, qf_inums[type])) { 6990 ext4_error(sb, "Bad quota inum: %lu, type: %d", 6991 qf_inums[type], type); 6992 return -EUCLEAN; 6993 } 6994 6995 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL); 6996 if (IS_ERR(qf_inode)) { 6997 ext4_error(sb, "Bad quota inode: %lu, type: %d", 6998 qf_inums[type], type); 6999 return PTR_ERR(qf_inode); 7000 } 7001 7002 /* Don't account quota for quota files to avoid recursion */ 7003 qf_inode->i_flags |= S_NOQUOTA; 7004 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 7005 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 7006 if (err) 7007 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 7008 iput(qf_inode); 7009 7010 return err; 7011 } 7012 7013 /* Enable usage tracking for all quota types. */ 7014 int ext4_enable_quotas(struct super_block *sb) 7015 { 7016 int type, err = 0; 7017 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 7018 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 7019 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 7020 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 7021 }; 7022 bool quota_mopt[EXT4_MAXQUOTAS] = { 7023 test_opt(sb, USRQUOTA), 7024 test_opt(sb, GRPQUOTA), 7025 test_opt(sb, PRJQUOTA), 7026 }; 7027 7028 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 7029 for (type = 0; type < EXT4_MAXQUOTAS; type++) { 7030 if (qf_inums[type]) { 7031 err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 7032 DQUOT_USAGE_ENABLED | 7033 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 7034 if (err) { 7035 ext4_warning(sb, 7036 "Failed to enable quota tracking " 7037 "(type=%d, err=%d, ino=%lu). " 7038 "Please run e2fsck to fix.", type, 7039 err, qf_inums[type]); 7040 for (type--; type >= 0; type--) { 7041 struct inode *inode; 7042 7043 inode = sb_dqopt(sb)->files[type]; 7044 if (inode) 7045 inode = igrab(inode); 7046 dquot_quota_off(sb, type); 7047 if (inode) { 7048 lockdep_set_quota_inode(inode, 7049 I_DATA_SEM_NORMAL); 7050 iput(inode); 7051 } 7052 } 7053 7054 return err; 7055 } 7056 } 7057 } 7058 return 0; 7059 } 7060 7061 static int ext4_quota_off(struct super_block *sb, int type) 7062 { 7063 struct inode *inode = sb_dqopt(sb)->files[type]; 7064 handle_t *handle; 7065 int err; 7066 7067 /* Force all delayed allocation blocks to be allocated. 7068 * Caller already holds s_umount sem */ 7069 if (test_opt(sb, DELALLOC)) 7070 sync_filesystem(sb); 7071 7072 if (!inode || !igrab(inode)) 7073 goto out; 7074 7075 err = dquot_quota_off(sb, type); 7076 if (err || ext4_has_feature_quota(sb)) 7077 goto out_put; 7078 7079 inode_lock(inode); 7080 /* 7081 * Update modification times of quota files when userspace can 7082 * start looking at them. If we fail, we return success anyway since 7083 * this is not a hard failure and quotas are already disabled. 7084 */ 7085 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 7086 if (IS_ERR(handle)) { 7087 err = PTR_ERR(handle); 7088 goto out_unlock; 7089 } 7090 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); 7091 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 7092 inode->i_mtime = inode->i_ctime = current_time(inode); 7093 err = ext4_mark_inode_dirty(handle, inode); 7094 ext4_journal_stop(handle); 7095 out_unlock: 7096 inode_unlock(inode); 7097 out_put: 7098 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); 7099 iput(inode); 7100 return err; 7101 out: 7102 return dquot_quota_off(sb, type); 7103 } 7104 7105 /* Read data from quotafile - avoid pagecache and such because we cannot afford 7106 * acquiring the locks... As quota files are never truncated and quota code 7107 * itself serializes the operations (and no one else should touch the files) 7108 * we don't have to be afraid of races */ 7109 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 7110 size_t len, loff_t off) 7111 { 7112 struct inode *inode = sb_dqopt(sb)->files[type]; 7113 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 7114 int offset = off & (sb->s_blocksize - 1); 7115 int tocopy; 7116 size_t toread; 7117 struct buffer_head *bh; 7118 loff_t i_size = i_size_read(inode); 7119 7120 if (off > i_size) 7121 return 0; 7122 if (off+len > i_size) 7123 len = i_size-off; 7124 toread = len; 7125 while (toread > 0) { 7126 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 7127 bh = ext4_bread(NULL, inode, blk, 0); 7128 if (IS_ERR(bh)) 7129 return PTR_ERR(bh); 7130 if (!bh) /* A hole? */ 7131 memset(data, 0, tocopy); 7132 else 7133 memcpy(data, bh->b_data+offset, tocopy); 7134 brelse(bh); 7135 offset = 0; 7136 toread -= tocopy; 7137 data += tocopy; 7138 blk++; 7139 } 7140 return len; 7141 } 7142 7143 /* Write to quotafile (we know the transaction is already started and has 7144 * enough credits) */ 7145 static ssize_t ext4_quota_write(struct super_block *sb, int type, 7146 const char *data, size_t len, loff_t off) 7147 { 7148 struct inode *inode = sb_dqopt(sb)->files[type]; 7149 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 7150 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1); 7151 int retries = 0; 7152 struct buffer_head *bh; 7153 handle_t *handle = journal_current_handle(); 7154 7155 if (!handle) { 7156 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 7157 " cancelled because transaction is not started", 7158 (unsigned long long)off, (unsigned long long)len); 7159 return -EIO; 7160 } 7161 /* 7162 * Since we account only one data block in transaction credits, 7163 * then it is impossible to cross a block boundary. 7164 */ 7165 if (sb->s_blocksize - offset < len) { 7166 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 7167 " cancelled because not block aligned", 7168 (unsigned long long)off, (unsigned long long)len); 7169 return -EIO; 7170 } 7171 7172 do { 7173 bh = ext4_bread(handle, inode, blk, 7174 EXT4_GET_BLOCKS_CREATE | 7175 EXT4_GET_BLOCKS_METADATA_NOFAIL); 7176 } while (PTR_ERR(bh) == -ENOSPC && 7177 ext4_should_retry_alloc(inode->i_sb, &retries)); 7178 if (IS_ERR(bh)) 7179 return PTR_ERR(bh); 7180 if (!bh) 7181 goto out; 7182 BUFFER_TRACE(bh, "get write access"); 7183 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE); 7184 if (err) { 7185 brelse(bh); 7186 return err; 7187 } 7188 lock_buffer(bh); 7189 memcpy(bh->b_data+offset, data, len); 7190 flush_dcache_page(bh->b_page); 7191 unlock_buffer(bh); 7192 err = ext4_handle_dirty_metadata(handle, NULL, bh); 7193 brelse(bh); 7194 out: 7195 if (inode->i_size < off + len) { 7196 i_size_write(inode, off + len); 7197 EXT4_I(inode)->i_disksize = inode->i_size; 7198 err2 = ext4_mark_inode_dirty(handle, inode); 7199 if (unlikely(err2 && !err)) 7200 err = err2; 7201 } 7202 return err ? err : len; 7203 } 7204 #endif 7205 7206 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 7207 static inline void register_as_ext2(void) 7208 { 7209 int err = register_filesystem(&ext2_fs_type); 7210 if (err) 7211 printk(KERN_WARNING 7212 "EXT4-fs: Unable to register as ext2 (%d)\n", err); 7213 } 7214 7215 static inline void unregister_as_ext2(void) 7216 { 7217 unregister_filesystem(&ext2_fs_type); 7218 } 7219 7220 static inline int ext2_feature_set_ok(struct super_block *sb) 7221 { 7222 if (ext4_has_unknown_ext2_incompat_features(sb)) 7223 return 0; 7224 if (sb_rdonly(sb)) 7225 return 1; 7226 if (ext4_has_unknown_ext2_ro_compat_features(sb)) 7227 return 0; 7228 return 1; 7229 } 7230 #else 7231 static inline void register_as_ext2(void) { } 7232 static inline void unregister_as_ext2(void) { } 7233 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 7234 #endif 7235 7236 static inline void register_as_ext3(void) 7237 { 7238 int err = register_filesystem(&ext3_fs_type); 7239 if (err) 7240 printk(KERN_WARNING 7241 "EXT4-fs: Unable to register as ext3 (%d)\n", err); 7242 } 7243 7244 static inline void unregister_as_ext3(void) 7245 { 7246 unregister_filesystem(&ext3_fs_type); 7247 } 7248 7249 static inline int ext3_feature_set_ok(struct super_block *sb) 7250 { 7251 if (ext4_has_unknown_ext3_incompat_features(sb)) 7252 return 0; 7253 if (!ext4_has_feature_journal(sb)) 7254 return 0; 7255 if (sb_rdonly(sb)) 7256 return 1; 7257 if (ext4_has_unknown_ext3_ro_compat_features(sb)) 7258 return 0; 7259 return 1; 7260 } 7261 7262 static struct file_system_type ext4_fs_type = { 7263 .owner = THIS_MODULE, 7264 .name = "ext4", 7265 .init_fs_context = ext4_init_fs_context, 7266 .parameters = ext4_param_specs, 7267 .kill_sb = kill_block_super, 7268 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP, 7269 }; 7270 MODULE_ALIAS_FS("ext4"); 7271 7272 /* Shared across all ext4 file systems */ 7273 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 7274 7275 static int __init ext4_init_fs(void) 7276 { 7277 int i, err; 7278 7279 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 7280 ext4_li_info = NULL; 7281 7282 /* Build-time check for flags consistency */ 7283 ext4_check_flag_values(); 7284 7285 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) 7286 init_waitqueue_head(&ext4__ioend_wq[i]); 7287 7288 err = ext4_init_es(); 7289 if (err) 7290 return err; 7291 7292 err = ext4_init_pending(); 7293 if (err) 7294 goto out7; 7295 7296 err = ext4_init_post_read_processing(); 7297 if (err) 7298 goto out6; 7299 7300 err = ext4_init_pageio(); 7301 if (err) 7302 goto out5; 7303 7304 err = ext4_init_system_zone(); 7305 if (err) 7306 goto out4; 7307 7308 err = ext4_init_sysfs(); 7309 if (err) 7310 goto out3; 7311 7312 err = ext4_init_mballoc(); 7313 if (err) 7314 goto out2; 7315 err = init_inodecache(); 7316 if (err) 7317 goto out1; 7318 7319 err = ext4_fc_init_dentry_cache(); 7320 if (err) 7321 goto out05; 7322 7323 register_as_ext3(); 7324 register_as_ext2(); 7325 err = register_filesystem(&ext4_fs_type); 7326 if (err) 7327 goto out; 7328 7329 return 0; 7330 out: 7331 unregister_as_ext2(); 7332 unregister_as_ext3(); 7333 ext4_fc_destroy_dentry_cache(); 7334 out05: 7335 destroy_inodecache(); 7336 out1: 7337 ext4_exit_mballoc(); 7338 out2: 7339 ext4_exit_sysfs(); 7340 out3: 7341 ext4_exit_system_zone(); 7342 out4: 7343 ext4_exit_pageio(); 7344 out5: 7345 ext4_exit_post_read_processing(); 7346 out6: 7347 ext4_exit_pending(); 7348 out7: 7349 ext4_exit_es(); 7350 7351 return err; 7352 } 7353 7354 static void __exit ext4_exit_fs(void) 7355 { 7356 ext4_destroy_lazyinit_thread(); 7357 unregister_as_ext2(); 7358 unregister_as_ext3(); 7359 unregister_filesystem(&ext4_fs_type); 7360 ext4_fc_destroy_dentry_cache(); 7361 destroy_inodecache(); 7362 ext4_exit_mballoc(); 7363 ext4_exit_sysfs(); 7364 ext4_exit_system_zone(); 7365 ext4_exit_pageio(); 7366 ext4_exit_post_read_processing(); 7367 ext4_exit_es(); 7368 ext4_exit_pending(); 7369 } 7370 7371 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 7372 MODULE_DESCRIPTION("Fourth Extended Filesystem"); 7373 MODULE_LICENSE("GPL"); 7374 MODULE_SOFTDEP("pre: crc32c"); 7375 module_init(ext4_init_fs) 7376 module_exit(ext4_exit_fs) 7377