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