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 /* 1203 * Unregister sysfs before destroying jbd2 journal. 1204 * Since we could still access attr_journal_task attribute via sysfs 1205 * path which could have sbi->s_journal->j_task as NULL 1206 * Unregister sysfs before flush sbi->s_error_work. 1207 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If 1208 * read metadata verify failed then will queue error work. 1209 * flush_stashed_error_work will call start_this_handle may trigger 1210 * BUG_ON. 1211 */ 1212 ext4_unregister_sysfs(sb); 1213 1214 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount")) 1215 ext4_msg(sb, KERN_INFO, "unmounting filesystem."); 1216 1217 ext4_unregister_li_request(sb); 1218 ext4_quota_off_umount(sb); 1219 1220 flush_work(&sbi->s_error_work); 1221 destroy_workqueue(sbi->rsv_conversion_wq); 1222 ext4_release_orphan_info(sb); 1223 1224 if (sbi->s_journal) { 1225 aborted = is_journal_aborted(sbi->s_journal); 1226 err = jbd2_journal_destroy(sbi->s_journal); 1227 sbi->s_journal = NULL; 1228 if ((err < 0) && !aborted) { 1229 ext4_abort(sb, -err, "Couldn't clean up the journal"); 1230 } 1231 } 1232 1233 ext4_es_unregister_shrinker(sbi); 1234 del_timer_sync(&sbi->s_err_report); 1235 ext4_release_system_zone(sb); 1236 ext4_mb_release(sb); 1237 ext4_ext_release(sb); 1238 1239 if (!sb_rdonly(sb) && !aborted) { 1240 ext4_clear_feature_journal_needs_recovery(sb); 1241 ext4_clear_feature_orphan_present(sb); 1242 es->s_state = cpu_to_le16(sbi->s_mount_state); 1243 } 1244 if (!sb_rdonly(sb)) 1245 ext4_commit_super(sb); 1246 1247 rcu_read_lock(); 1248 group_desc = rcu_dereference(sbi->s_group_desc); 1249 for (i = 0; i < sbi->s_gdb_count; i++) 1250 brelse(group_desc[i]); 1251 kvfree(group_desc); 1252 flex_groups = rcu_dereference(sbi->s_flex_groups); 1253 if (flex_groups) { 1254 for (i = 0; i < sbi->s_flex_groups_allocated; i++) 1255 kvfree(flex_groups[i]); 1256 kvfree(flex_groups); 1257 } 1258 rcu_read_unlock(); 1259 percpu_counter_destroy(&sbi->s_freeclusters_counter); 1260 percpu_counter_destroy(&sbi->s_freeinodes_counter); 1261 percpu_counter_destroy(&sbi->s_dirs_counter); 1262 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 1263 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit); 1264 percpu_free_rwsem(&sbi->s_writepages_rwsem); 1265 #ifdef CONFIG_QUOTA 1266 for (i = 0; i < EXT4_MAXQUOTAS; i++) 1267 kfree(get_qf_name(sb, sbi, i)); 1268 #endif 1269 1270 /* Debugging code just in case the in-memory inode orphan list 1271 * isn't empty. The on-disk one can be non-empty if we've 1272 * detected an error and taken the fs readonly, but the 1273 * in-memory list had better be clean by this point. */ 1274 if (!list_empty(&sbi->s_orphan)) 1275 dump_orphan_list(sb, sbi); 1276 ASSERT(list_empty(&sbi->s_orphan)); 1277 1278 sync_blockdev(sb->s_bdev); 1279 invalidate_bdev(sb->s_bdev); 1280 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) { 1281 /* 1282 * Invalidate the journal device's buffers. We don't want them 1283 * floating about in memory - the physical journal device may 1284 * hotswapped, and it breaks the `ro-after' testing code. 1285 */ 1286 sync_blockdev(sbi->s_journal_bdev); 1287 invalidate_bdev(sbi->s_journal_bdev); 1288 ext4_blkdev_remove(sbi); 1289 } 1290 1291 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 1292 sbi->s_ea_inode_cache = NULL; 1293 1294 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 1295 sbi->s_ea_block_cache = NULL; 1296 1297 ext4_stop_mmpd(sbi); 1298 1299 brelse(sbi->s_sbh); 1300 sb->s_fs_info = NULL; 1301 /* 1302 * Now that we are completely done shutting down the 1303 * superblock, we need to actually destroy the kobject. 1304 */ 1305 kobject_put(&sbi->s_kobj); 1306 wait_for_completion(&sbi->s_kobj_unregister); 1307 if (sbi->s_chksum_driver) 1308 crypto_free_shash(sbi->s_chksum_driver); 1309 kfree(sbi->s_blockgroup_lock); 1310 fs_put_dax(sbi->s_daxdev); 1311 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy); 1312 #if IS_ENABLED(CONFIG_UNICODE) 1313 utf8_unload(sb->s_encoding); 1314 #endif 1315 kfree(sbi); 1316 } 1317 1318 static struct kmem_cache *ext4_inode_cachep; 1319 1320 /* 1321 * Called inside transaction, so use GFP_NOFS 1322 */ 1323 static struct inode *ext4_alloc_inode(struct super_block *sb) 1324 { 1325 struct ext4_inode_info *ei; 1326 1327 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS); 1328 if (!ei) 1329 return NULL; 1330 1331 inode_set_iversion(&ei->vfs_inode, 1); 1332 spin_lock_init(&ei->i_raw_lock); 1333 INIT_LIST_HEAD(&ei->i_prealloc_list); 1334 atomic_set(&ei->i_prealloc_active, 0); 1335 spin_lock_init(&ei->i_prealloc_lock); 1336 ext4_es_init_tree(&ei->i_es_tree); 1337 rwlock_init(&ei->i_es_lock); 1338 INIT_LIST_HEAD(&ei->i_es_list); 1339 ei->i_es_all_nr = 0; 1340 ei->i_es_shk_nr = 0; 1341 ei->i_es_shrink_lblk = 0; 1342 ei->i_reserved_data_blocks = 0; 1343 spin_lock_init(&(ei->i_block_reservation_lock)); 1344 ext4_init_pending_tree(&ei->i_pending_tree); 1345 #ifdef CONFIG_QUOTA 1346 ei->i_reserved_quota = 0; 1347 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); 1348 #endif 1349 ei->jinode = NULL; 1350 INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 1351 spin_lock_init(&ei->i_completed_io_lock); 1352 ei->i_sync_tid = 0; 1353 ei->i_datasync_tid = 0; 1354 atomic_set(&ei->i_unwritten, 0); 1355 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 1356 ext4_fc_init_inode(&ei->vfs_inode); 1357 mutex_init(&ei->i_fc_lock); 1358 return &ei->vfs_inode; 1359 } 1360 1361 static int ext4_drop_inode(struct inode *inode) 1362 { 1363 int drop = generic_drop_inode(inode); 1364 1365 if (!drop) 1366 drop = fscrypt_drop_inode(inode); 1367 1368 trace_ext4_drop_inode(inode, drop); 1369 return drop; 1370 } 1371 1372 static void ext4_free_in_core_inode(struct inode *inode) 1373 { 1374 fscrypt_free_inode(inode); 1375 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) { 1376 pr_warn("%s: inode %ld still in fc list", 1377 __func__, inode->i_ino); 1378 } 1379 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 1380 } 1381 1382 static void ext4_destroy_inode(struct inode *inode) 1383 { 1384 if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 1385 ext4_msg(inode->i_sb, KERN_ERR, 1386 "Inode %lu (%p): orphan list check failed!", 1387 inode->i_ino, EXT4_I(inode)); 1388 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 1389 EXT4_I(inode), sizeof(struct ext4_inode_info), 1390 true); 1391 dump_stack(); 1392 } 1393 1394 if (EXT4_I(inode)->i_reserved_data_blocks) 1395 ext4_msg(inode->i_sb, KERN_ERR, 1396 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!", 1397 inode->i_ino, EXT4_I(inode), 1398 EXT4_I(inode)->i_reserved_data_blocks); 1399 } 1400 1401 static void init_once(void *foo) 1402 { 1403 struct ext4_inode_info *ei = foo; 1404 1405 INIT_LIST_HEAD(&ei->i_orphan); 1406 init_rwsem(&ei->xattr_sem); 1407 init_rwsem(&ei->i_data_sem); 1408 inode_init_once(&ei->vfs_inode); 1409 ext4_fc_init_inode(&ei->vfs_inode); 1410 } 1411 1412 static int __init init_inodecache(void) 1413 { 1414 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache", 1415 sizeof(struct ext4_inode_info), 0, 1416 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD| 1417 SLAB_ACCOUNT), 1418 offsetof(struct ext4_inode_info, i_data), 1419 sizeof_field(struct ext4_inode_info, i_data), 1420 init_once); 1421 if (ext4_inode_cachep == NULL) 1422 return -ENOMEM; 1423 return 0; 1424 } 1425 1426 static void destroy_inodecache(void) 1427 { 1428 /* 1429 * Make sure all delayed rcu free inodes are flushed before we 1430 * destroy cache. 1431 */ 1432 rcu_barrier(); 1433 kmem_cache_destroy(ext4_inode_cachep); 1434 } 1435 1436 void ext4_clear_inode(struct inode *inode) 1437 { 1438 ext4_fc_del(inode); 1439 invalidate_inode_buffers(inode); 1440 clear_inode(inode); 1441 ext4_discard_preallocations(inode, 0); 1442 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 1443 dquot_drop(inode); 1444 if (EXT4_I(inode)->jinode) { 1445 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 1446 EXT4_I(inode)->jinode); 1447 jbd2_free_inode(EXT4_I(inode)->jinode); 1448 EXT4_I(inode)->jinode = NULL; 1449 } 1450 fscrypt_put_encryption_info(inode); 1451 fsverity_cleanup_inode(inode); 1452 } 1453 1454 static struct inode *ext4_nfs_get_inode(struct super_block *sb, 1455 u64 ino, u32 generation) 1456 { 1457 struct inode *inode; 1458 1459 /* 1460 * Currently we don't know the generation for parent directory, so 1461 * a generation of 0 means "accept any" 1462 */ 1463 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE); 1464 if (IS_ERR(inode)) 1465 return ERR_CAST(inode); 1466 if (generation && inode->i_generation != generation) { 1467 iput(inode); 1468 return ERR_PTR(-ESTALE); 1469 } 1470 1471 return inode; 1472 } 1473 1474 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 1475 int fh_len, int fh_type) 1476 { 1477 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1478 ext4_nfs_get_inode); 1479 } 1480 1481 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 1482 int fh_len, int fh_type) 1483 { 1484 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1485 ext4_nfs_get_inode); 1486 } 1487 1488 static int ext4_nfs_commit_metadata(struct inode *inode) 1489 { 1490 struct writeback_control wbc = { 1491 .sync_mode = WB_SYNC_ALL 1492 }; 1493 1494 trace_ext4_nfs_commit_metadata(inode); 1495 return ext4_write_inode(inode, &wbc); 1496 } 1497 1498 #ifdef CONFIG_QUOTA 1499 static const char * const quotatypes[] = INITQFNAMES; 1500 #define QTYPE2NAME(t) (quotatypes[t]) 1501 1502 static int ext4_write_dquot(struct dquot *dquot); 1503 static int ext4_acquire_dquot(struct dquot *dquot); 1504 static int ext4_release_dquot(struct dquot *dquot); 1505 static int ext4_mark_dquot_dirty(struct dquot *dquot); 1506 static int ext4_write_info(struct super_block *sb, int type); 1507 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1508 const struct path *path); 1509 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1510 size_t len, loff_t off); 1511 static ssize_t ext4_quota_write(struct super_block *sb, int type, 1512 const char *data, size_t len, loff_t off); 1513 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 1514 unsigned int flags); 1515 1516 static struct dquot **ext4_get_dquots(struct inode *inode) 1517 { 1518 return EXT4_I(inode)->i_dquot; 1519 } 1520 1521 static const struct dquot_operations ext4_quota_operations = { 1522 .get_reserved_space = ext4_get_reserved_space, 1523 .write_dquot = ext4_write_dquot, 1524 .acquire_dquot = ext4_acquire_dquot, 1525 .release_dquot = ext4_release_dquot, 1526 .mark_dirty = ext4_mark_dquot_dirty, 1527 .write_info = ext4_write_info, 1528 .alloc_dquot = dquot_alloc, 1529 .destroy_dquot = dquot_destroy, 1530 .get_projid = ext4_get_projid, 1531 .get_inode_usage = ext4_get_inode_usage, 1532 .get_next_id = dquot_get_next_id, 1533 }; 1534 1535 static const struct quotactl_ops ext4_qctl_operations = { 1536 .quota_on = ext4_quota_on, 1537 .quota_off = ext4_quota_off, 1538 .quota_sync = dquot_quota_sync, 1539 .get_state = dquot_get_state, 1540 .set_info = dquot_set_dqinfo, 1541 .get_dqblk = dquot_get_dqblk, 1542 .set_dqblk = dquot_set_dqblk, 1543 .get_nextdqblk = dquot_get_next_dqblk, 1544 }; 1545 #endif 1546 1547 static const struct super_operations ext4_sops = { 1548 .alloc_inode = ext4_alloc_inode, 1549 .free_inode = ext4_free_in_core_inode, 1550 .destroy_inode = ext4_destroy_inode, 1551 .write_inode = ext4_write_inode, 1552 .dirty_inode = ext4_dirty_inode, 1553 .drop_inode = ext4_drop_inode, 1554 .evict_inode = ext4_evict_inode, 1555 .put_super = ext4_put_super, 1556 .sync_fs = ext4_sync_fs, 1557 .freeze_fs = ext4_freeze, 1558 .unfreeze_fs = ext4_unfreeze, 1559 .statfs = ext4_statfs, 1560 .show_options = ext4_show_options, 1561 #ifdef CONFIG_QUOTA 1562 .quota_read = ext4_quota_read, 1563 .quota_write = ext4_quota_write, 1564 .get_dquots = ext4_get_dquots, 1565 #endif 1566 }; 1567 1568 static const struct export_operations ext4_export_ops = { 1569 .fh_to_dentry = ext4_fh_to_dentry, 1570 .fh_to_parent = ext4_fh_to_parent, 1571 .get_parent = ext4_get_parent, 1572 .commit_metadata = ext4_nfs_commit_metadata, 1573 }; 1574 1575 enum { 1576 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1577 Opt_resgid, Opt_resuid, Opt_sb, 1578 Opt_nouid32, Opt_debug, Opt_removed, 1579 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 1580 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1581 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1582 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1583 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1584 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 1585 Opt_inlinecrypt, 1586 Opt_usrjquota, Opt_grpjquota, Opt_quota, 1587 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1588 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, 1589 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never, 1590 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error, 1591 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize, 1592 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 1593 Opt_inode_readahead_blks, Opt_journal_ioprio, 1594 Opt_dioread_nolock, Opt_dioread_lock, 1595 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1596 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, 1597 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan, 1598 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type, 1599 #ifdef CONFIG_EXT4_DEBUG 1600 Opt_fc_debug_max_replay, Opt_fc_debug_force 1601 #endif 1602 }; 1603 1604 static const struct constant_table ext4_param_errors[] = { 1605 {"continue", EXT4_MOUNT_ERRORS_CONT}, 1606 {"panic", EXT4_MOUNT_ERRORS_PANIC}, 1607 {"remount-ro", EXT4_MOUNT_ERRORS_RO}, 1608 {} 1609 }; 1610 1611 static const struct constant_table ext4_param_data[] = { 1612 {"journal", EXT4_MOUNT_JOURNAL_DATA}, 1613 {"ordered", EXT4_MOUNT_ORDERED_DATA}, 1614 {"writeback", EXT4_MOUNT_WRITEBACK_DATA}, 1615 {} 1616 }; 1617 1618 static const struct constant_table ext4_param_data_err[] = { 1619 {"abort", Opt_data_err_abort}, 1620 {"ignore", Opt_data_err_ignore}, 1621 {} 1622 }; 1623 1624 static const struct constant_table ext4_param_jqfmt[] = { 1625 {"vfsold", QFMT_VFS_OLD}, 1626 {"vfsv0", QFMT_VFS_V0}, 1627 {"vfsv1", QFMT_VFS_V1}, 1628 {} 1629 }; 1630 1631 static const struct constant_table ext4_param_dax[] = { 1632 {"always", Opt_dax_always}, 1633 {"inode", Opt_dax_inode}, 1634 {"never", Opt_dax_never}, 1635 {} 1636 }; 1637 1638 /* String parameter that allows empty argument */ 1639 #define fsparam_string_empty(NAME, OPT) \ 1640 __fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL) 1641 1642 /* 1643 * Mount option specification 1644 * We don't use fsparam_flag_no because of the way we set the 1645 * options and the way we show them in _ext4_show_options(). To 1646 * keep the changes to a minimum, let's keep the negative options 1647 * separate for now. 1648 */ 1649 static const struct fs_parameter_spec ext4_param_specs[] = { 1650 fsparam_flag ("bsddf", Opt_bsd_df), 1651 fsparam_flag ("minixdf", Opt_minix_df), 1652 fsparam_flag ("grpid", Opt_grpid), 1653 fsparam_flag ("bsdgroups", Opt_grpid), 1654 fsparam_flag ("nogrpid", Opt_nogrpid), 1655 fsparam_flag ("sysvgroups", Opt_nogrpid), 1656 fsparam_u32 ("resgid", Opt_resgid), 1657 fsparam_u32 ("resuid", Opt_resuid), 1658 fsparam_u32 ("sb", Opt_sb), 1659 fsparam_enum ("errors", Opt_errors, ext4_param_errors), 1660 fsparam_flag ("nouid32", Opt_nouid32), 1661 fsparam_flag ("debug", Opt_debug), 1662 fsparam_flag ("oldalloc", Opt_removed), 1663 fsparam_flag ("orlov", Opt_removed), 1664 fsparam_flag ("user_xattr", Opt_user_xattr), 1665 fsparam_flag ("nouser_xattr", Opt_nouser_xattr), 1666 fsparam_flag ("acl", Opt_acl), 1667 fsparam_flag ("noacl", Opt_noacl), 1668 fsparam_flag ("norecovery", Opt_noload), 1669 fsparam_flag ("noload", Opt_noload), 1670 fsparam_flag ("bh", Opt_removed), 1671 fsparam_flag ("nobh", Opt_removed), 1672 fsparam_u32 ("commit", Opt_commit), 1673 fsparam_u32 ("min_batch_time", Opt_min_batch_time), 1674 fsparam_u32 ("max_batch_time", Opt_max_batch_time), 1675 fsparam_u32 ("journal_dev", Opt_journal_dev), 1676 fsparam_bdev ("journal_path", Opt_journal_path), 1677 fsparam_flag ("journal_checksum", Opt_journal_checksum), 1678 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum), 1679 fsparam_flag ("journal_async_commit",Opt_journal_async_commit), 1680 fsparam_flag ("abort", Opt_abort), 1681 fsparam_enum ("data", Opt_data, ext4_param_data), 1682 fsparam_enum ("data_err", Opt_data_err, 1683 ext4_param_data_err), 1684 fsparam_string_empty 1685 ("usrjquota", Opt_usrjquota), 1686 fsparam_string_empty 1687 ("grpjquota", Opt_grpjquota), 1688 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt), 1689 fsparam_flag ("grpquota", Opt_grpquota), 1690 fsparam_flag ("quota", Opt_quota), 1691 fsparam_flag ("noquota", Opt_noquota), 1692 fsparam_flag ("usrquota", Opt_usrquota), 1693 fsparam_flag ("prjquota", Opt_prjquota), 1694 fsparam_flag ("barrier", Opt_barrier), 1695 fsparam_u32 ("barrier", Opt_barrier), 1696 fsparam_flag ("nobarrier", Opt_nobarrier), 1697 fsparam_flag ("i_version", Opt_i_version), 1698 fsparam_flag ("dax", Opt_dax), 1699 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax), 1700 fsparam_u32 ("stripe", Opt_stripe), 1701 fsparam_flag ("delalloc", Opt_delalloc), 1702 fsparam_flag ("nodelalloc", Opt_nodelalloc), 1703 fsparam_flag ("warn_on_error", Opt_warn_on_error), 1704 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error), 1705 fsparam_u32 ("debug_want_extra_isize", 1706 Opt_debug_want_extra_isize), 1707 fsparam_flag ("mblk_io_submit", Opt_removed), 1708 fsparam_flag ("nomblk_io_submit", Opt_removed), 1709 fsparam_flag ("block_validity", Opt_block_validity), 1710 fsparam_flag ("noblock_validity", Opt_noblock_validity), 1711 fsparam_u32 ("inode_readahead_blks", 1712 Opt_inode_readahead_blks), 1713 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio), 1714 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc), 1715 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc), 1716 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc), 1717 fsparam_flag ("dioread_nolock", Opt_dioread_nolock), 1718 fsparam_flag ("nodioread_nolock", Opt_dioread_lock), 1719 fsparam_flag ("dioread_lock", Opt_dioread_lock), 1720 fsparam_flag ("discard", Opt_discard), 1721 fsparam_flag ("nodiscard", Opt_nodiscard), 1722 fsparam_u32 ("init_itable", Opt_init_itable), 1723 fsparam_flag ("init_itable", Opt_init_itable), 1724 fsparam_flag ("noinit_itable", Opt_noinit_itable), 1725 #ifdef CONFIG_EXT4_DEBUG 1726 fsparam_flag ("fc_debug_force", Opt_fc_debug_force), 1727 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay), 1728 #endif 1729 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb), 1730 fsparam_flag ("test_dummy_encryption", 1731 Opt_test_dummy_encryption), 1732 fsparam_string ("test_dummy_encryption", 1733 Opt_test_dummy_encryption), 1734 fsparam_flag ("inlinecrypt", Opt_inlinecrypt), 1735 fsparam_flag ("nombcache", Opt_nombcache), 1736 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */ 1737 fsparam_flag ("prefetch_block_bitmaps", 1738 Opt_removed), 1739 fsparam_flag ("no_prefetch_block_bitmaps", 1740 Opt_no_prefetch_block_bitmaps), 1741 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan), 1742 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */ 1743 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */ 1744 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */ 1745 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */ 1746 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */ 1747 {} 1748 }; 1749 1750 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1751 1752 static const char deprecated_msg[] = 1753 "Mount option \"%s\" will be removed by %s\n" 1754 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1755 1756 #define MOPT_SET 0x0001 1757 #define MOPT_CLEAR 0x0002 1758 #define MOPT_NOSUPPORT 0x0004 1759 #define MOPT_EXPLICIT 0x0008 1760 #ifdef CONFIG_QUOTA 1761 #define MOPT_Q 0 1762 #define MOPT_QFMT 0x0010 1763 #else 1764 #define MOPT_Q MOPT_NOSUPPORT 1765 #define MOPT_QFMT MOPT_NOSUPPORT 1766 #endif 1767 #define MOPT_NO_EXT2 0x0020 1768 #define MOPT_NO_EXT3 0x0040 1769 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1770 #define MOPT_SKIP 0x0080 1771 #define MOPT_2 0x0100 1772 1773 static const struct mount_opts { 1774 int token; 1775 int mount_opt; 1776 int flags; 1777 } ext4_mount_opts[] = { 1778 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 1779 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 1780 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 1781 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 1782 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 1783 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 1784 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 1785 MOPT_EXT4_ONLY | MOPT_SET}, 1786 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 1787 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1788 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 1789 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 1790 {Opt_delalloc, EXT4_MOUNT_DELALLOC, 1791 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1792 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 1793 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1794 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET}, 1795 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR}, 1796 {Opt_commit, 0, MOPT_NO_EXT2}, 1797 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1798 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1799 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1800 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1801 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 1802 EXT4_MOUNT_JOURNAL_CHECKSUM), 1803 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1804 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 1805 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2}, 1806 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 1807 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 1808 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 1809 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 1810 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 1811 {Opt_dax_type, 0, MOPT_EXT4_ONLY}, 1812 {Opt_journal_dev, 0, MOPT_NO_EXT2}, 1813 {Opt_journal_path, 0, MOPT_NO_EXT2}, 1814 {Opt_journal_ioprio, 0, MOPT_NO_EXT2}, 1815 {Opt_data, 0, MOPT_NO_EXT2}, 1816 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 1817 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 1818 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1819 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 1820 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 1821 #else 1822 {Opt_acl, 0, MOPT_NOSUPPORT}, 1823 {Opt_noacl, 0, MOPT_NOSUPPORT}, 1824 #endif 1825 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 1826 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1827 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 1828 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 1829 MOPT_SET | MOPT_Q}, 1830 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 1831 MOPT_SET | MOPT_Q}, 1832 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, 1833 MOPT_SET | MOPT_Q}, 1834 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 1835 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), 1836 MOPT_CLEAR | MOPT_Q}, 1837 {Opt_usrjquota, 0, MOPT_Q}, 1838 {Opt_grpjquota, 0, MOPT_Q}, 1839 {Opt_jqfmt, 0, MOPT_QFMT}, 1840 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, 1841 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS, 1842 MOPT_SET}, 1843 #ifdef CONFIG_EXT4_DEBUG 1844 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT, 1845 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY}, 1846 #endif 1847 {Opt_err, 0, 0} 1848 }; 1849 1850 #if IS_ENABLED(CONFIG_UNICODE) 1851 static const struct ext4_sb_encodings { 1852 __u16 magic; 1853 char *name; 1854 unsigned int version; 1855 } ext4_sb_encoding_map[] = { 1856 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)}, 1857 }; 1858 1859 static const struct ext4_sb_encodings * 1860 ext4_sb_read_encoding(const struct ext4_super_block *es) 1861 { 1862 __u16 magic = le16_to_cpu(es->s_encoding); 1863 int i; 1864 1865 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++) 1866 if (magic == ext4_sb_encoding_map[i].magic) 1867 return &ext4_sb_encoding_map[i]; 1868 1869 return NULL; 1870 } 1871 #endif 1872 1873 static int ext4_set_test_dummy_encryption(struct super_block *sb, char *arg) 1874 { 1875 #ifdef CONFIG_FS_ENCRYPTION 1876 struct ext4_sb_info *sbi = EXT4_SB(sb); 1877 int err; 1878 1879 err = fscrypt_set_test_dummy_encryption(sb, arg, 1880 &sbi->s_dummy_enc_policy); 1881 if (err) { 1882 ext4_msg(sb, KERN_WARNING, 1883 "Error while setting test dummy encryption [%d]", err); 1884 return err; 1885 } 1886 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled"); 1887 #endif 1888 return 0; 1889 } 1890 1891 #define EXT4_SPEC_JQUOTA (1 << 0) 1892 #define EXT4_SPEC_JQFMT (1 << 1) 1893 #define EXT4_SPEC_DATAJ (1 << 2) 1894 #define EXT4_SPEC_SB_BLOCK (1 << 3) 1895 #define EXT4_SPEC_JOURNAL_DEV (1 << 4) 1896 #define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5) 1897 #define EXT4_SPEC_DUMMY_ENCRYPTION (1 << 6) 1898 #define EXT4_SPEC_s_want_extra_isize (1 << 7) 1899 #define EXT4_SPEC_s_max_batch_time (1 << 8) 1900 #define EXT4_SPEC_s_min_batch_time (1 << 9) 1901 #define EXT4_SPEC_s_inode_readahead_blks (1 << 10) 1902 #define EXT4_SPEC_s_li_wait_mult (1 << 11) 1903 #define EXT4_SPEC_s_max_dir_size_kb (1 << 12) 1904 #define EXT4_SPEC_s_stripe (1 << 13) 1905 #define EXT4_SPEC_s_resuid (1 << 14) 1906 #define EXT4_SPEC_s_resgid (1 << 15) 1907 #define EXT4_SPEC_s_commit_interval (1 << 16) 1908 #define EXT4_SPEC_s_fc_debug_max_replay (1 << 17) 1909 #define EXT4_SPEC_s_sb_block (1 << 18) 1910 #define EXT4_SPEC_mb_optimize_scan (1 << 19) 1911 1912 struct ext4_fs_context { 1913 char *s_qf_names[EXT4_MAXQUOTAS]; 1914 char *test_dummy_enc_arg; 1915 int s_jquota_fmt; /* Format of quota to use */ 1916 #ifdef CONFIG_EXT4_DEBUG 1917 int s_fc_debug_max_replay; 1918 #endif 1919 unsigned short qname_spec; 1920 unsigned long vals_s_flags; /* Bits to set in s_flags */ 1921 unsigned long mask_s_flags; /* Bits changed in s_flags */ 1922 unsigned long journal_devnum; 1923 unsigned long s_commit_interval; 1924 unsigned long s_stripe; 1925 unsigned int s_inode_readahead_blks; 1926 unsigned int s_want_extra_isize; 1927 unsigned int s_li_wait_mult; 1928 unsigned int s_max_dir_size_kb; 1929 unsigned int journal_ioprio; 1930 unsigned int vals_s_mount_opt; 1931 unsigned int mask_s_mount_opt; 1932 unsigned int vals_s_mount_opt2; 1933 unsigned int mask_s_mount_opt2; 1934 unsigned long vals_s_mount_flags; 1935 unsigned long mask_s_mount_flags; 1936 unsigned int opt_flags; /* MOPT flags */ 1937 unsigned int spec; 1938 u32 s_max_batch_time; 1939 u32 s_min_batch_time; 1940 kuid_t s_resuid; 1941 kgid_t s_resgid; 1942 ext4_fsblk_t s_sb_block; 1943 }; 1944 1945 static void ext4_fc_free(struct fs_context *fc) 1946 { 1947 struct ext4_fs_context *ctx = fc->fs_private; 1948 int i; 1949 1950 if (!ctx) 1951 return; 1952 1953 for (i = 0; i < EXT4_MAXQUOTAS; i++) 1954 kfree(ctx->s_qf_names[i]); 1955 1956 kfree(ctx->test_dummy_enc_arg); 1957 kfree(ctx); 1958 } 1959 1960 int ext4_init_fs_context(struct fs_context *fc) 1961 { 1962 struct ext4_fs_context *ctx; 1963 1964 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL); 1965 if (!ctx) 1966 return -ENOMEM; 1967 1968 fc->fs_private = ctx; 1969 fc->ops = &ext4_context_ops; 1970 1971 return 0; 1972 } 1973 1974 #ifdef CONFIG_QUOTA 1975 /* 1976 * Note the name of the specified quota file. 1977 */ 1978 static int note_qf_name(struct fs_context *fc, int qtype, 1979 struct fs_parameter *param) 1980 { 1981 struct ext4_fs_context *ctx = fc->fs_private; 1982 char *qname; 1983 1984 if (param->size < 1) { 1985 ext4_msg(NULL, KERN_ERR, "Missing quota name"); 1986 return -EINVAL; 1987 } 1988 if (strchr(param->string, '/')) { 1989 ext4_msg(NULL, KERN_ERR, 1990 "quotafile must be on filesystem root"); 1991 return -EINVAL; 1992 } 1993 if (ctx->s_qf_names[qtype]) { 1994 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) { 1995 ext4_msg(NULL, KERN_ERR, 1996 "%s quota file already specified", 1997 QTYPE2NAME(qtype)); 1998 return -EINVAL; 1999 } 2000 return 0; 2001 } 2002 2003 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL); 2004 if (!qname) { 2005 ext4_msg(NULL, KERN_ERR, 2006 "Not enough memory for storing quotafile name"); 2007 return -ENOMEM; 2008 } 2009 ctx->s_qf_names[qtype] = qname; 2010 ctx->qname_spec |= 1 << qtype; 2011 ctx->spec |= EXT4_SPEC_JQUOTA; 2012 return 0; 2013 } 2014 2015 /* 2016 * Clear the name of the specified quota file. 2017 */ 2018 static int unnote_qf_name(struct fs_context *fc, int qtype) 2019 { 2020 struct ext4_fs_context *ctx = fc->fs_private; 2021 2022 if (ctx->s_qf_names[qtype]) 2023 kfree(ctx->s_qf_names[qtype]); 2024 2025 ctx->s_qf_names[qtype] = NULL; 2026 ctx->qname_spec |= 1 << qtype; 2027 ctx->spec |= EXT4_SPEC_JQUOTA; 2028 return 0; 2029 } 2030 #endif 2031 2032 #define EXT4_SET_CTX(name) \ 2033 static inline void ctx_set_##name(struct ext4_fs_context *ctx, \ 2034 unsigned long flag) \ 2035 { \ 2036 ctx->mask_s_##name |= flag; \ 2037 ctx->vals_s_##name |= flag; \ 2038 } 2039 2040 #define EXT4_CLEAR_CTX(name) \ 2041 static inline void ctx_clear_##name(struct ext4_fs_context *ctx, \ 2042 unsigned long flag) \ 2043 { \ 2044 ctx->mask_s_##name |= flag; \ 2045 ctx->vals_s_##name &= ~flag; \ 2046 } 2047 2048 #define EXT4_TEST_CTX(name) \ 2049 static inline unsigned long \ 2050 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \ 2051 { \ 2052 return (ctx->vals_s_##name & flag); \ 2053 } 2054 2055 EXT4_SET_CTX(flags); /* set only */ 2056 EXT4_SET_CTX(mount_opt); 2057 EXT4_CLEAR_CTX(mount_opt); 2058 EXT4_TEST_CTX(mount_opt); 2059 EXT4_SET_CTX(mount_opt2); 2060 EXT4_CLEAR_CTX(mount_opt2); 2061 EXT4_TEST_CTX(mount_opt2); 2062 2063 static inline void ctx_set_mount_flag(struct ext4_fs_context *ctx, int bit) 2064 { 2065 set_bit(bit, &ctx->mask_s_mount_flags); 2066 set_bit(bit, &ctx->vals_s_mount_flags); 2067 } 2068 2069 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param) 2070 { 2071 struct ext4_fs_context *ctx = fc->fs_private; 2072 struct fs_parse_result result; 2073 const struct mount_opts *m; 2074 int is_remount; 2075 kuid_t uid; 2076 kgid_t gid; 2077 int token; 2078 2079 token = fs_parse(fc, ext4_param_specs, param, &result); 2080 if (token < 0) 2081 return token; 2082 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE; 2083 2084 for (m = ext4_mount_opts; m->token != Opt_err; m++) 2085 if (token == m->token) 2086 break; 2087 2088 ctx->opt_flags |= m->flags; 2089 2090 if (m->flags & MOPT_EXPLICIT) { 2091 if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 2092 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC); 2093 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 2094 ctx_set_mount_opt2(ctx, 2095 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM); 2096 } else 2097 return -EINVAL; 2098 } 2099 2100 if (m->flags & MOPT_NOSUPPORT) { 2101 ext4_msg(NULL, KERN_ERR, "%s option not supported", 2102 param->key); 2103 return 0; 2104 } 2105 2106 switch (token) { 2107 #ifdef CONFIG_QUOTA 2108 case Opt_usrjquota: 2109 if (!*param->string) 2110 return unnote_qf_name(fc, USRQUOTA); 2111 else 2112 return note_qf_name(fc, USRQUOTA, param); 2113 case Opt_grpjquota: 2114 if (!*param->string) 2115 return unnote_qf_name(fc, GRPQUOTA); 2116 else 2117 return note_qf_name(fc, GRPQUOTA, param); 2118 #endif 2119 case Opt_noacl: 2120 case Opt_nouser_xattr: 2121 ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "3.5"); 2122 break; 2123 case Opt_sb: 2124 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) { 2125 ext4_msg(NULL, KERN_WARNING, 2126 "Ignoring %s option on remount", param->key); 2127 } else { 2128 ctx->s_sb_block = result.uint_32; 2129 ctx->spec |= EXT4_SPEC_s_sb_block; 2130 } 2131 return 0; 2132 case Opt_removed: 2133 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option", 2134 param->key); 2135 return 0; 2136 case Opt_abort: 2137 ctx_set_mount_flag(ctx, EXT4_MF_FS_ABORTED); 2138 return 0; 2139 case Opt_i_version: 2140 ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "5.20"); 2141 ext4_msg(NULL, KERN_WARNING, "Use iversion instead\n"); 2142 ctx_set_flags(ctx, SB_I_VERSION); 2143 return 0; 2144 case Opt_inlinecrypt: 2145 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 2146 ctx_set_flags(ctx, SB_INLINECRYPT); 2147 #else 2148 ext4_msg(NULL, KERN_ERR, "inline encryption not supported"); 2149 #endif 2150 return 0; 2151 case Opt_errors: 2152 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK); 2153 ctx_set_mount_opt(ctx, result.uint_32); 2154 return 0; 2155 #ifdef CONFIG_QUOTA 2156 case Opt_jqfmt: 2157 ctx->s_jquota_fmt = result.uint_32; 2158 ctx->spec |= EXT4_SPEC_JQFMT; 2159 return 0; 2160 #endif 2161 case Opt_data: 2162 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS); 2163 ctx_set_mount_opt(ctx, result.uint_32); 2164 ctx->spec |= EXT4_SPEC_DATAJ; 2165 return 0; 2166 case Opt_commit: 2167 if (result.uint_32 == 0) 2168 ctx->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE; 2169 else if (result.uint_32 > INT_MAX / HZ) { 2170 ext4_msg(NULL, KERN_ERR, 2171 "Invalid commit interval %d, " 2172 "must be smaller than %d", 2173 result.uint_32, INT_MAX / HZ); 2174 return -EINVAL; 2175 } 2176 ctx->s_commit_interval = HZ * result.uint_32; 2177 ctx->spec |= EXT4_SPEC_s_commit_interval; 2178 return 0; 2179 case Opt_debug_want_extra_isize: 2180 if ((result.uint_32 & 1) || (result.uint_32 < 4)) { 2181 ext4_msg(NULL, KERN_ERR, 2182 "Invalid want_extra_isize %d", result.uint_32); 2183 return -EINVAL; 2184 } 2185 ctx->s_want_extra_isize = result.uint_32; 2186 ctx->spec |= EXT4_SPEC_s_want_extra_isize; 2187 return 0; 2188 case Opt_max_batch_time: 2189 ctx->s_max_batch_time = result.uint_32; 2190 ctx->spec |= EXT4_SPEC_s_max_batch_time; 2191 return 0; 2192 case Opt_min_batch_time: 2193 ctx->s_min_batch_time = result.uint_32; 2194 ctx->spec |= EXT4_SPEC_s_min_batch_time; 2195 return 0; 2196 case Opt_inode_readahead_blks: 2197 if (result.uint_32 && 2198 (result.uint_32 > (1 << 30) || 2199 !is_power_of_2(result.uint_32))) { 2200 ext4_msg(NULL, KERN_ERR, 2201 "EXT4-fs: inode_readahead_blks must be " 2202 "0 or a power of 2 smaller than 2^31"); 2203 return -EINVAL; 2204 } 2205 ctx->s_inode_readahead_blks = result.uint_32; 2206 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks; 2207 return 0; 2208 case Opt_init_itable: 2209 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE); 2210 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 2211 if (param->type == fs_value_is_string) 2212 ctx->s_li_wait_mult = result.uint_32; 2213 ctx->spec |= EXT4_SPEC_s_li_wait_mult; 2214 return 0; 2215 case Opt_max_dir_size_kb: 2216 ctx->s_max_dir_size_kb = result.uint_32; 2217 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb; 2218 return 0; 2219 #ifdef CONFIG_EXT4_DEBUG 2220 case Opt_fc_debug_max_replay: 2221 ctx->s_fc_debug_max_replay = result.uint_32; 2222 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay; 2223 return 0; 2224 #endif 2225 case Opt_stripe: 2226 ctx->s_stripe = result.uint_32; 2227 ctx->spec |= EXT4_SPEC_s_stripe; 2228 return 0; 2229 case Opt_resuid: 2230 uid = make_kuid(current_user_ns(), result.uint_32); 2231 if (!uid_valid(uid)) { 2232 ext4_msg(NULL, KERN_ERR, "Invalid uid value %d", 2233 result.uint_32); 2234 return -EINVAL; 2235 } 2236 ctx->s_resuid = uid; 2237 ctx->spec |= EXT4_SPEC_s_resuid; 2238 return 0; 2239 case Opt_resgid: 2240 gid = make_kgid(current_user_ns(), result.uint_32); 2241 if (!gid_valid(gid)) { 2242 ext4_msg(NULL, KERN_ERR, "Invalid gid value %d", 2243 result.uint_32); 2244 return -EINVAL; 2245 } 2246 ctx->s_resgid = gid; 2247 ctx->spec |= EXT4_SPEC_s_resgid; 2248 return 0; 2249 case Opt_journal_dev: 2250 if (is_remount) { 2251 ext4_msg(NULL, KERN_ERR, 2252 "Cannot specify journal on remount"); 2253 return -EINVAL; 2254 } 2255 ctx->journal_devnum = result.uint_32; 2256 ctx->spec |= EXT4_SPEC_JOURNAL_DEV; 2257 return 0; 2258 case Opt_journal_path: 2259 { 2260 struct inode *journal_inode; 2261 struct path path; 2262 int error; 2263 2264 if (is_remount) { 2265 ext4_msg(NULL, KERN_ERR, 2266 "Cannot specify journal on remount"); 2267 return -EINVAL; 2268 } 2269 2270 error = fs_lookup_param(fc, param, 1, &path); 2271 if (error) { 2272 ext4_msg(NULL, KERN_ERR, "error: could not find " 2273 "journal device path"); 2274 return -EINVAL; 2275 } 2276 2277 journal_inode = d_inode(path.dentry); 2278 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev); 2279 ctx->spec |= EXT4_SPEC_JOURNAL_DEV; 2280 path_put(&path); 2281 return 0; 2282 } 2283 case Opt_journal_ioprio: 2284 if (result.uint_32 > 7) { 2285 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority" 2286 " (must be 0-7)"); 2287 return -EINVAL; 2288 } 2289 ctx->journal_ioprio = 2290 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32); 2291 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO; 2292 return 0; 2293 case Opt_test_dummy_encryption: 2294 #ifdef CONFIG_FS_ENCRYPTION 2295 if (param->type == fs_value_is_flag) { 2296 ctx->spec |= EXT4_SPEC_DUMMY_ENCRYPTION; 2297 ctx->test_dummy_enc_arg = NULL; 2298 return 0; 2299 } 2300 if (*param->string && 2301 !(!strcmp(param->string, "v1") || 2302 !strcmp(param->string, "v2"))) { 2303 ext4_msg(NULL, KERN_WARNING, 2304 "Value of option \"%s\" is unrecognized", 2305 param->key); 2306 return -EINVAL; 2307 } 2308 ctx->spec |= EXT4_SPEC_DUMMY_ENCRYPTION; 2309 ctx->test_dummy_enc_arg = kmemdup_nul(param->string, param->size, 2310 GFP_KERNEL); 2311 return 0; 2312 #else 2313 ext4_msg(NULL, KERN_WARNING, 2314 "test_dummy_encryption option not supported"); 2315 return -EINVAL; 2316 #endif 2317 case Opt_dax: 2318 case Opt_dax_type: 2319 #ifdef CONFIG_FS_DAX 2320 { 2321 int type = (token == Opt_dax) ? 2322 Opt_dax : result.uint_32; 2323 2324 switch (type) { 2325 case Opt_dax: 2326 case Opt_dax_always: 2327 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); 2328 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); 2329 break; 2330 case Opt_dax_never: 2331 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); 2332 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); 2333 break; 2334 case Opt_dax_inode: 2335 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); 2336 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); 2337 /* Strictly for printing options */ 2338 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE); 2339 break; 2340 } 2341 return 0; 2342 } 2343 #else 2344 ext4_msg(NULL, KERN_INFO, "dax option not supported"); 2345 return -EINVAL; 2346 #endif 2347 case Opt_data_err: 2348 if (result.uint_32 == Opt_data_err_abort) 2349 ctx_set_mount_opt(ctx, m->mount_opt); 2350 else if (result.uint_32 == Opt_data_err_ignore) 2351 ctx_clear_mount_opt(ctx, m->mount_opt); 2352 return 0; 2353 case Opt_mb_optimize_scan: 2354 if (result.int_32 == 1) { 2355 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN); 2356 ctx->spec |= EXT4_SPEC_mb_optimize_scan; 2357 } else if (result.int_32 == 0) { 2358 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN); 2359 ctx->spec |= EXT4_SPEC_mb_optimize_scan; 2360 } else { 2361 ext4_msg(NULL, KERN_WARNING, 2362 "mb_optimize_scan should be set to 0 or 1."); 2363 return -EINVAL; 2364 } 2365 return 0; 2366 } 2367 2368 /* 2369 * At this point we should only be getting options requiring MOPT_SET, 2370 * or MOPT_CLEAR. Anything else is a bug 2371 */ 2372 if (m->token == Opt_err) { 2373 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s", 2374 param->key); 2375 WARN_ON(1); 2376 return -EINVAL; 2377 } 2378 2379 else { 2380 unsigned int set = 0; 2381 2382 if ((param->type == fs_value_is_flag) || 2383 result.uint_32 > 0) 2384 set = 1; 2385 2386 if (m->flags & MOPT_CLEAR) 2387 set = !set; 2388 else if (unlikely(!(m->flags & MOPT_SET))) { 2389 ext4_msg(NULL, KERN_WARNING, 2390 "buggy handling of option %s", 2391 param->key); 2392 WARN_ON(1); 2393 return -EINVAL; 2394 } 2395 if (m->flags & MOPT_2) { 2396 if (set != 0) 2397 ctx_set_mount_opt2(ctx, m->mount_opt); 2398 else 2399 ctx_clear_mount_opt2(ctx, m->mount_opt); 2400 } else { 2401 if (set != 0) 2402 ctx_set_mount_opt(ctx, m->mount_opt); 2403 else 2404 ctx_clear_mount_opt(ctx, m->mount_opt); 2405 } 2406 } 2407 2408 return 0; 2409 } 2410 2411 static int parse_options(struct fs_context *fc, char *options) 2412 { 2413 struct fs_parameter param; 2414 int ret; 2415 char *key; 2416 2417 if (!options) 2418 return 0; 2419 2420 while ((key = strsep(&options, ",")) != NULL) { 2421 if (*key) { 2422 size_t v_len = 0; 2423 char *value = strchr(key, '='); 2424 2425 param.type = fs_value_is_flag; 2426 param.string = NULL; 2427 2428 if (value) { 2429 if (value == key) 2430 continue; 2431 2432 *value++ = 0; 2433 v_len = strlen(value); 2434 param.string = kmemdup_nul(value, v_len, 2435 GFP_KERNEL); 2436 if (!param.string) 2437 return -ENOMEM; 2438 param.type = fs_value_is_string; 2439 } 2440 2441 param.key = key; 2442 param.size = v_len; 2443 2444 ret = ext4_parse_param(fc, ¶m); 2445 if (param.string) 2446 kfree(param.string); 2447 if (ret < 0) 2448 return ret; 2449 } 2450 } 2451 2452 ret = ext4_validate_options(fc); 2453 if (ret < 0) 2454 return ret; 2455 2456 return 0; 2457 } 2458 2459 static int parse_apply_sb_mount_options(struct super_block *sb, 2460 struct ext4_fs_context *m_ctx) 2461 { 2462 struct ext4_sb_info *sbi = EXT4_SB(sb); 2463 char *s_mount_opts = NULL; 2464 struct ext4_fs_context *s_ctx = NULL; 2465 struct fs_context *fc = NULL; 2466 int ret = -ENOMEM; 2467 2468 if (!sbi->s_es->s_mount_opts[0]) 2469 return 0; 2470 2471 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts, 2472 sizeof(sbi->s_es->s_mount_opts), 2473 GFP_KERNEL); 2474 if (!s_mount_opts) 2475 return ret; 2476 2477 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL); 2478 if (!fc) 2479 goto out_free; 2480 2481 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL); 2482 if (!s_ctx) 2483 goto out_free; 2484 2485 fc->fs_private = s_ctx; 2486 fc->s_fs_info = sbi; 2487 2488 ret = parse_options(fc, s_mount_opts); 2489 if (ret < 0) 2490 goto parse_failed; 2491 2492 ret = ext4_check_opt_consistency(fc, sb); 2493 if (ret < 0) { 2494 parse_failed: 2495 ext4_msg(sb, KERN_WARNING, 2496 "failed to parse options in superblock: %s", 2497 s_mount_opts); 2498 ret = 0; 2499 goto out_free; 2500 } 2501 2502 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV) 2503 m_ctx->journal_devnum = s_ctx->journal_devnum; 2504 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO) 2505 m_ctx->journal_ioprio = s_ctx->journal_ioprio; 2506 2507 ret = ext4_apply_options(fc, sb); 2508 2509 out_free: 2510 if (fc) { 2511 ext4_fc_free(fc); 2512 kfree(fc); 2513 } 2514 kfree(s_mount_opts); 2515 return ret; 2516 } 2517 2518 static void ext4_apply_quota_options(struct fs_context *fc, 2519 struct super_block *sb) 2520 { 2521 #ifdef CONFIG_QUOTA 2522 bool quota_feature = ext4_has_feature_quota(sb); 2523 struct ext4_fs_context *ctx = fc->fs_private; 2524 struct ext4_sb_info *sbi = EXT4_SB(sb); 2525 char *qname; 2526 int i; 2527 2528 if (quota_feature) 2529 return; 2530 2531 if (ctx->spec & EXT4_SPEC_JQUOTA) { 2532 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2533 if (!(ctx->qname_spec & (1 << i))) 2534 continue; 2535 2536 qname = ctx->s_qf_names[i]; /* May be NULL */ 2537 if (qname) 2538 set_opt(sb, QUOTA); 2539 ctx->s_qf_names[i] = NULL; 2540 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname, 2541 lockdep_is_held(&sb->s_umount)); 2542 if (qname) 2543 kfree_rcu(qname); 2544 } 2545 } 2546 2547 if (ctx->spec & EXT4_SPEC_JQFMT) 2548 sbi->s_jquota_fmt = ctx->s_jquota_fmt; 2549 #endif 2550 } 2551 2552 /* 2553 * Check quota settings consistency. 2554 */ 2555 static int ext4_check_quota_consistency(struct fs_context *fc, 2556 struct super_block *sb) 2557 { 2558 #ifdef CONFIG_QUOTA 2559 struct ext4_fs_context *ctx = fc->fs_private; 2560 struct ext4_sb_info *sbi = EXT4_SB(sb); 2561 bool quota_feature = ext4_has_feature_quota(sb); 2562 bool quota_loaded = sb_any_quota_loaded(sb); 2563 bool usr_qf_name, grp_qf_name, usrquota, grpquota; 2564 int quota_flags, i; 2565 2566 /* 2567 * We do the test below only for project quotas. 'usrquota' and 2568 * 'grpquota' mount options are allowed even without quota feature 2569 * to support legacy quotas in quota files. 2570 */ 2571 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) && 2572 !ext4_has_feature_project(sb)) { 2573 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. " 2574 "Cannot enable project quota enforcement."); 2575 return -EINVAL; 2576 } 2577 2578 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 2579 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA; 2580 if (quota_loaded && 2581 ctx->mask_s_mount_opt & quota_flags && 2582 !ctx_test_mount_opt(ctx, quota_flags)) 2583 goto err_quota_change; 2584 2585 if (ctx->spec & EXT4_SPEC_JQUOTA) { 2586 2587 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2588 if (!(ctx->qname_spec & (1 << i))) 2589 continue; 2590 2591 if (quota_loaded && 2592 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i]) 2593 goto err_jquota_change; 2594 2595 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] && 2596 strcmp(get_qf_name(sb, sbi, i), 2597 ctx->s_qf_names[i]) != 0) 2598 goto err_jquota_specified; 2599 } 2600 2601 if (quota_feature) { 2602 ext4_msg(NULL, KERN_INFO, 2603 "Journaled quota options ignored when " 2604 "QUOTA feature is enabled"); 2605 return 0; 2606 } 2607 } 2608 2609 if (ctx->spec & EXT4_SPEC_JQFMT) { 2610 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded) 2611 goto err_jquota_change; 2612 if (quota_feature) { 2613 ext4_msg(NULL, KERN_INFO, "Quota format mount options " 2614 "ignored when QUOTA feature is enabled"); 2615 return 0; 2616 } 2617 } 2618 2619 /* Make sure we don't mix old and new quota format */ 2620 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) || 2621 ctx->s_qf_names[USRQUOTA]); 2622 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) || 2623 ctx->s_qf_names[GRPQUOTA]); 2624 2625 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) || 2626 test_opt(sb, USRQUOTA)); 2627 2628 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) || 2629 test_opt(sb, GRPQUOTA)); 2630 2631 if (usr_qf_name) { 2632 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA); 2633 usrquota = false; 2634 } 2635 if (grp_qf_name) { 2636 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA); 2637 grpquota = false; 2638 } 2639 2640 if (usr_qf_name || grp_qf_name) { 2641 if (usrquota || grpquota) { 2642 ext4_msg(NULL, KERN_ERR, "old and new quota " 2643 "format mixing"); 2644 return -EINVAL; 2645 } 2646 2647 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) { 2648 ext4_msg(NULL, KERN_ERR, "journaled quota format " 2649 "not specified"); 2650 return -EINVAL; 2651 } 2652 } 2653 2654 return 0; 2655 2656 err_quota_change: 2657 ext4_msg(NULL, KERN_ERR, 2658 "Cannot change quota options when quota turned on"); 2659 return -EINVAL; 2660 err_jquota_change: 2661 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota " 2662 "options when quota turned on"); 2663 return -EINVAL; 2664 err_jquota_specified: 2665 ext4_msg(NULL, KERN_ERR, "%s quota file already specified", 2666 QTYPE2NAME(i)); 2667 return -EINVAL; 2668 #else 2669 return 0; 2670 #endif 2671 } 2672 2673 static int ext4_check_test_dummy_encryption(const struct fs_context *fc, 2674 struct super_block *sb) 2675 { 2676 #ifdef CONFIG_FS_ENCRYPTION 2677 const struct ext4_fs_context *ctx = fc->fs_private; 2678 const struct ext4_sb_info *sbi = EXT4_SB(sb); 2679 2680 if (!(ctx->spec & EXT4_SPEC_DUMMY_ENCRYPTION)) 2681 return 0; 2682 2683 if (!ext4_has_feature_encrypt(sb)) { 2684 ext4_msg(NULL, KERN_WARNING, 2685 "test_dummy_encryption requires encrypt feature"); 2686 return -EINVAL; 2687 } 2688 /* 2689 * This mount option is just for testing, and it's not worthwhile to 2690 * implement the extra complexity (e.g. RCU protection) that would be 2691 * needed to allow it to be set or changed during remount. We do allow 2692 * it to be specified during remount, but only if there is no change. 2693 */ 2694 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE && 2695 !sbi->s_dummy_enc_policy.policy) { 2696 ext4_msg(NULL, KERN_WARNING, 2697 "Can't set test_dummy_encryption on remount"); 2698 return -EINVAL; 2699 } 2700 #endif /* CONFIG_FS_ENCRYPTION */ 2701 return 0; 2702 } 2703 2704 static int ext4_check_opt_consistency(struct fs_context *fc, 2705 struct super_block *sb) 2706 { 2707 struct ext4_fs_context *ctx = fc->fs_private; 2708 struct ext4_sb_info *sbi = fc->s_fs_info; 2709 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE; 2710 int err; 2711 2712 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 2713 ext4_msg(NULL, KERN_ERR, 2714 "Mount option(s) incompatible with ext2"); 2715 return -EINVAL; 2716 } 2717 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 2718 ext4_msg(NULL, KERN_ERR, 2719 "Mount option(s) incompatible with ext3"); 2720 return -EINVAL; 2721 } 2722 2723 if (ctx->s_want_extra_isize > 2724 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) { 2725 ext4_msg(NULL, KERN_ERR, 2726 "Invalid want_extra_isize %d", 2727 ctx->s_want_extra_isize); 2728 return -EINVAL; 2729 } 2730 2731 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DIOREAD_NOLOCK)) { 2732 int blocksize = 2733 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 2734 if (blocksize < PAGE_SIZE) 2735 ext4_msg(NULL, KERN_WARNING, "Warning: mounting with an " 2736 "experimental mount option 'dioread_nolock' " 2737 "for blocksize < PAGE_SIZE"); 2738 } 2739 2740 err = ext4_check_test_dummy_encryption(fc, sb); 2741 if (err) 2742 return err; 2743 2744 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) { 2745 if (!sbi->s_journal) { 2746 ext4_msg(NULL, KERN_WARNING, 2747 "Remounting file system with no journal " 2748 "so ignoring journalled data option"); 2749 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS); 2750 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) != 2751 test_opt(sb, DATA_FLAGS)) { 2752 ext4_msg(NULL, KERN_ERR, "Cannot change data mode " 2753 "on remount"); 2754 return -EINVAL; 2755 } 2756 } 2757 2758 if (is_remount) { 2759 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) && 2760 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) { 2761 ext4_msg(NULL, KERN_ERR, "can't mount with " 2762 "both data=journal and dax"); 2763 return -EINVAL; 2764 } 2765 2766 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) && 2767 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2768 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) { 2769 fail_dax_change_remount: 2770 ext4_msg(NULL, KERN_ERR, "can't change " 2771 "dax mount option while remounting"); 2772 return -EINVAL; 2773 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) && 2774 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2775 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) { 2776 goto fail_dax_change_remount; 2777 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) && 2778 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2779 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2780 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) { 2781 goto fail_dax_change_remount; 2782 } 2783 } 2784 2785 return ext4_check_quota_consistency(fc, sb); 2786 } 2787 2788 static int ext4_apply_options(struct fs_context *fc, struct super_block *sb) 2789 { 2790 struct ext4_fs_context *ctx = fc->fs_private; 2791 struct ext4_sb_info *sbi = fc->s_fs_info; 2792 int ret = 0; 2793 2794 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt; 2795 sbi->s_mount_opt |= ctx->vals_s_mount_opt; 2796 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2; 2797 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2; 2798 sbi->s_mount_flags &= ~ctx->mask_s_mount_flags; 2799 sbi->s_mount_flags |= ctx->vals_s_mount_flags; 2800 sb->s_flags &= ~ctx->mask_s_flags; 2801 sb->s_flags |= ctx->vals_s_flags; 2802 2803 /* 2804 * i_version differs from common mount option iversion so we have 2805 * to let vfs know that it was set, otherwise it would get cleared 2806 * on remount 2807 */ 2808 if (ctx->mask_s_flags & SB_I_VERSION) 2809 fc->sb_flags |= SB_I_VERSION; 2810 2811 #define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; }) 2812 APPLY(s_commit_interval); 2813 APPLY(s_stripe); 2814 APPLY(s_max_batch_time); 2815 APPLY(s_min_batch_time); 2816 APPLY(s_want_extra_isize); 2817 APPLY(s_inode_readahead_blks); 2818 APPLY(s_max_dir_size_kb); 2819 APPLY(s_li_wait_mult); 2820 APPLY(s_resgid); 2821 APPLY(s_resuid); 2822 2823 #ifdef CONFIG_EXT4_DEBUG 2824 APPLY(s_fc_debug_max_replay); 2825 #endif 2826 2827 ext4_apply_quota_options(fc, sb); 2828 2829 if (ctx->spec & EXT4_SPEC_DUMMY_ENCRYPTION) 2830 ret = ext4_set_test_dummy_encryption(sb, ctx->test_dummy_enc_arg); 2831 2832 return ret; 2833 } 2834 2835 2836 static int ext4_validate_options(struct fs_context *fc) 2837 { 2838 #ifdef CONFIG_QUOTA 2839 struct ext4_fs_context *ctx = fc->fs_private; 2840 char *usr_qf_name, *grp_qf_name; 2841 2842 usr_qf_name = ctx->s_qf_names[USRQUOTA]; 2843 grp_qf_name = ctx->s_qf_names[GRPQUOTA]; 2844 2845 if (usr_qf_name || grp_qf_name) { 2846 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name) 2847 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA); 2848 2849 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name) 2850 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA); 2851 2852 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) || 2853 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) { 2854 ext4_msg(NULL, KERN_ERR, "old and new quota " 2855 "format mixing"); 2856 return -EINVAL; 2857 } 2858 } 2859 #endif 2860 return 1; 2861 } 2862 2863 static inline void ext4_show_quota_options(struct seq_file *seq, 2864 struct super_block *sb) 2865 { 2866 #if defined(CONFIG_QUOTA) 2867 struct ext4_sb_info *sbi = EXT4_SB(sb); 2868 char *usr_qf_name, *grp_qf_name; 2869 2870 if (sbi->s_jquota_fmt) { 2871 char *fmtname = ""; 2872 2873 switch (sbi->s_jquota_fmt) { 2874 case QFMT_VFS_OLD: 2875 fmtname = "vfsold"; 2876 break; 2877 case QFMT_VFS_V0: 2878 fmtname = "vfsv0"; 2879 break; 2880 case QFMT_VFS_V1: 2881 fmtname = "vfsv1"; 2882 break; 2883 } 2884 seq_printf(seq, ",jqfmt=%s", fmtname); 2885 } 2886 2887 rcu_read_lock(); 2888 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); 2889 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); 2890 if (usr_qf_name) 2891 seq_show_option(seq, "usrjquota", usr_qf_name); 2892 if (grp_qf_name) 2893 seq_show_option(seq, "grpjquota", grp_qf_name); 2894 rcu_read_unlock(); 2895 #endif 2896 } 2897 2898 static const char *token2str(int token) 2899 { 2900 const struct fs_parameter_spec *spec; 2901 2902 for (spec = ext4_param_specs; spec->name != NULL; spec++) 2903 if (spec->opt == token && !spec->type) 2904 break; 2905 return spec->name; 2906 } 2907 2908 /* 2909 * Show an option if 2910 * - it's set to a non-default value OR 2911 * - if the per-sb default is different from the global default 2912 */ 2913 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 2914 int nodefs) 2915 { 2916 struct ext4_sb_info *sbi = EXT4_SB(sb); 2917 struct ext4_super_block *es = sbi->s_es; 2918 int def_errors, def_mount_opt = sbi->s_def_mount_opt; 2919 const struct mount_opts *m; 2920 char sep = nodefs ? '\n' : ','; 2921 2922 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 2923 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 2924 2925 if (sbi->s_sb_block != 1) 2926 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 2927 2928 for (m = ext4_mount_opts; m->token != Opt_err; m++) { 2929 int want_set = m->flags & MOPT_SET; 2930 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 2931 m->flags & MOPT_SKIP) 2932 continue; 2933 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 2934 continue; /* skip if same as the default */ 2935 if ((want_set && 2936 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 2937 (!want_set && (sbi->s_mount_opt & m->mount_opt))) 2938 continue; /* select Opt_noFoo vs Opt_Foo */ 2939 SEQ_OPTS_PRINT("%s", token2str(m->token)); 2940 } 2941 2942 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 2943 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 2944 SEQ_OPTS_PRINT("resuid=%u", 2945 from_kuid_munged(&init_user_ns, sbi->s_resuid)); 2946 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 2947 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 2948 SEQ_OPTS_PRINT("resgid=%u", 2949 from_kgid_munged(&init_user_ns, sbi->s_resgid)); 2950 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 2951 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 2952 SEQ_OPTS_PUTS("errors=remount-ro"); 2953 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 2954 SEQ_OPTS_PUTS("errors=continue"); 2955 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 2956 SEQ_OPTS_PUTS("errors=panic"); 2957 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 2958 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 2959 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 2960 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 2961 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 2962 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 2963 if (sb->s_flags & SB_I_VERSION) 2964 SEQ_OPTS_PUTS("i_version"); 2965 if (nodefs || sbi->s_stripe) 2966 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 2967 if (nodefs || EXT4_MOUNT_DATA_FLAGS & 2968 (sbi->s_mount_opt ^ def_mount_opt)) { 2969 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 2970 SEQ_OPTS_PUTS("data=journal"); 2971 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 2972 SEQ_OPTS_PUTS("data=ordered"); 2973 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 2974 SEQ_OPTS_PUTS("data=writeback"); 2975 } 2976 if (nodefs || 2977 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 2978 SEQ_OPTS_PRINT("inode_readahead_blks=%u", 2979 sbi->s_inode_readahead_blks); 2980 2981 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs || 2982 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 2983 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 2984 if (nodefs || sbi->s_max_dir_size_kb) 2985 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 2986 if (test_opt(sb, DATA_ERR_ABORT)) 2987 SEQ_OPTS_PUTS("data_err=abort"); 2988 2989 fscrypt_show_test_dummy_encryption(seq, sep, sb); 2990 2991 if (sb->s_flags & SB_INLINECRYPT) 2992 SEQ_OPTS_PUTS("inlinecrypt"); 2993 2994 if (test_opt(sb, DAX_ALWAYS)) { 2995 if (IS_EXT2_SB(sb)) 2996 SEQ_OPTS_PUTS("dax"); 2997 else 2998 SEQ_OPTS_PUTS("dax=always"); 2999 } else if (test_opt2(sb, DAX_NEVER)) { 3000 SEQ_OPTS_PUTS("dax=never"); 3001 } else if (test_opt2(sb, DAX_INODE)) { 3002 SEQ_OPTS_PUTS("dax=inode"); 3003 } 3004 ext4_show_quota_options(seq, sb); 3005 return 0; 3006 } 3007 3008 static int ext4_show_options(struct seq_file *seq, struct dentry *root) 3009 { 3010 return _ext4_show_options(seq, root->d_sb, 0); 3011 } 3012 3013 int ext4_seq_options_show(struct seq_file *seq, void *offset) 3014 { 3015 struct super_block *sb = seq->private; 3016 int rc; 3017 3018 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); 3019 rc = _ext4_show_options(seq, sb, 1); 3020 seq_puts(seq, "\n"); 3021 return rc; 3022 } 3023 3024 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 3025 int read_only) 3026 { 3027 struct ext4_sb_info *sbi = EXT4_SB(sb); 3028 int err = 0; 3029 3030 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 3031 ext4_msg(sb, KERN_ERR, "revision level too high, " 3032 "forcing read-only mode"); 3033 err = -EROFS; 3034 goto done; 3035 } 3036 if (read_only) 3037 goto done; 3038 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 3039 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 3040 "running e2fsck is recommended"); 3041 else if (sbi->s_mount_state & EXT4_ERROR_FS) 3042 ext4_msg(sb, KERN_WARNING, 3043 "warning: mounting fs with errors, " 3044 "running e2fsck is recommended"); 3045 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 3046 le16_to_cpu(es->s_mnt_count) >= 3047 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 3048 ext4_msg(sb, KERN_WARNING, 3049 "warning: maximal mount count reached, " 3050 "running e2fsck is recommended"); 3051 else if (le32_to_cpu(es->s_checkinterval) && 3052 (ext4_get_tstamp(es, s_lastcheck) + 3053 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds())) 3054 ext4_msg(sb, KERN_WARNING, 3055 "warning: checktime reached, " 3056 "running e2fsck is recommended"); 3057 if (!sbi->s_journal) 3058 es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 3059 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 3060 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 3061 le16_add_cpu(&es->s_mnt_count, 1); 3062 ext4_update_tstamp(es, s_mtime); 3063 if (sbi->s_journal) { 3064 ext4_set_feature_journal_needs_recovery(sb); 3065 if (ext4_has_feature_orphan_file(sb)) 3066 ext4_set_feature_orphan_present(sb); 3067 } 3068 3069 err = ext4_commit_super(sb); 3070 done: 3071 if (test_opt(sb, DEBUG)) 3072 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 3073 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 3074 sb->s_blocksize, 3075 sbi->s_groups_count, 3076 EXT4_BLOCKS_PER_GROUP(sb), 3077 EXT4_INODES_PER_GROUP(sb), 3078 sbi->s_mount_opt, sbi->s_mount_opt2); 3079 return err; 3080 } 3081 3082 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 3083 { 3084 struct ext4_sb_info *sbi = EXT4_SB(sb); 3085 struct flex_groups **old_groups, **new_groups; 3086 int size, i, j; 3087 3088 if (!sbi->s_log_groups_per_flex) 3089 return 0; 3090 3091 size = ext4_flex_group(sbi, ngroup - 1) + 1; 3092 if (size <= sbi->s_flex_groups_allocated) 3093 return 0; 3094 3095 new_groups = kvzalloc(roundup_pow_of_two(size * 3096 sizeof(*sbi->s_flex_groups)), GFP_KERNEL); 3097 if (!new_groups) { 3098 ext4_msg(sb, KERN_ERR, 3099 "not enough memory for %d flex group pointers", size); 3100 return -ENOMEM; 3101 } 3102 for (i = sbi->s_flex_groups_allocated; i < size; i++) { 3103 new_groups[i] = kvzalloc(roundup_pow_of_two( 3104 sizeof(struct flex_groups)), 3105 GFP_KERNEL); 3106 if (!new_groups[i]) { 3107 for (j = sbi->s_flex_groups_allocated; j < i; j++) 3108 kvfree(new_groups[j]); 3109 kvfree(new_groups); 3110 ext4_msg(sb, KERN_ERR, 3111 "not enough memory for %d flex groups", size); 3112 return -ENOMEM; 3113 } 3114 } 3115 rcu_read_lock(); 3116 old_groups = rcu_dereference(sbi->s_flex_groups); 3117 if (old_groups) 3118 memcpy(new_groups, old_groups, 3119 (sbi->s_flex_groups_allocated * 3120 sizeof(struct flex_groups *))); 3121 rcu_read_unlock(); 3122 rcu_assign_pointer(sbi->s_flex_groups, new_groups); 3123 sbi->s_flex_groups_allocated = size; 3124 if (old_groups) 3125 ext4_kvfree_array_rcu(old_groups); 3126 return 0; 3127 } 3128 3129 static int ext4_fill_flex_info(struct super_block *sb) 3130 { 3131 struct ext4_sb_info *sbi = EXT4_SB(sb); 3132 struct ext4_group_desc *gdp = NULL; 3133 struct flex_groups *fg; 3134 ext4_group_t flex_group; 3135 int i, err; 3136 3137 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 3138 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 3139 sbi->s_log_groups_per_flex = 0; 3140 return 1; 3141 } 3142 3143 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 3144 if (err) 3145 goto failed; 3146 3147 for (i = 0; i < sbi->s_groups_count; i++) { 3148 gdp = ext4_get_group_desc(sb, i, NULL); 3149 3150 flex_group = ext4_flex_group(sbi, i); 3151 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group); 3152 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes); 3153 atomic64_add(ext4_free_group_clusters(sb, gdp), 3154 &fg->free_clusters); 3155 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs); 3156 } 3157 3158 return 1; 3159 failed: 3160 return 0; 3161 } 3162 3163 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 3164 struct ext4_group_desc *gdp) 3165 { 3166 int offset = offsetof(struct ext4_group_desc, bg_checksum); 3167 __u16 crc = 0; 3168 __le32 le_group = cpu_to_le32(block_group); 3169 struct ext4_sb_info *sbi = EXT4_SB(sb); 3170 3171 if (ext4_has_metadata_csum(sbi->s_sb)) { 3172 /* Use new metadata_csum algorithm */ 3173 __u32 csum32; 3174 __u16 dummy_csum = 0; 3175 3176 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 3177 sizeof(le_group)); 3178 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); 3179 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, 3180 sizeof(dummy_csum)); 3181 offset += sizeof(dummy_csum); 3182 if (offset < sbi->s_desc_size) 3183 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, 3184 sbi->s_desc_size - offset); 3185 3186 crc = csum32 & 0xFFFF; 3187 goto out; 3188 } 3189 3190 /* old crc16 code */ 3191 if (!ext4_has_feature_gdt_csum(sb)) 3192 return 0; 3193 3194 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 3195 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 3196 crc = crc16(crc, (__u8 *)gdp, offset); 3197 offset += sizeof(gdp->bg_checksum); /* skip checksum */ 3198 /* for checksum of struct ext4_group_desc do the rest...*/ 3199 if (ext4_has_feature_64bit(sb) && 3200 offset < le16_to_cpu(sbi->s_es->s_desc_size)) 3201 crc = crc16(crc, (__u8 *)gdp + offset, 3202 le16_to_cpu(sbi->s_es->s_desc_size) - 3203 offset); 3204 3205 out: 3206 return cpu_to_le16(crc); 3207 } 3208 3209 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 3210 struct ext4_group_desc *gdp) 3211 { 3212 if (ext4_has_group_desc_csum(sb) && 3213 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 3214 return 0; 3215 3216 return 1; 3217 } 3218 3219 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 3220 struct ext4_group_desc *gdp) 3221 { 3222 if (!ext4_has_group_desc_csum(sb)) 3223 return; 3224 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 3225 } 3226 3227 /* Called at mount-time, super-block is locked */ 3228 static int ext4_check_descriptors(struct super_block *sb, 3229 ext4_fsblk_t sb_block, 3230 ext4_group_t *first_not_zeroed) 3231 { 3232 struct ext4_sb_info *sbi = EXT4_SB(sb); 3233 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 3234 ext4_fsblk_t last_block; 3235 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0); 3236 ext4_fsblk_t block_bitmap; 3237 ext4_fsblk_t inode_bitmap; 3238 ext4_fsblk_t inode_table; 3239 int flexbg_flag = 0; 3240 ext4_group_t i, grp = sbi->s_groups_count; 3241 3242 if (ext4_has_feature_flex_bg(sb)) 3243 flexbg_flag = 1; 3244 3245 ext4_debug("Checking group descriptors"); 3246 3247 for (i = 0; i < sbi->s_groups_count; i++) { 3248 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 3249 3250 if (i == sbi->s_groups_count - 1 || flexbg_flag) 3251 last_block = ext4_blocks_count(sbi->s_es) - 1; 3252 else 3253 last_block = first_block + 3254 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 3255 3256 if ((grp == sbi->s_groups_count) && 3257 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3258 grp = i; 3259 3260 block_bitmap = ext4_block_bitmap(sb, gdp); 3261 if (block_bitmap == sb_block) { 3262 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3263 "Block bitmap for group %u overlaps " 3264 "superblock", i); 3265 if (!sb_rdonly(sb)) 3266 return 0; 3267 } 3268 if (block_bitmap >= sb_block + 1 && 3269 block_bitmap <= last_bg_block) { 3270 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3271 "Block bitmap for group %u overlaps " 3272 "block group descriptors", i); 3273 if (!sb_rdonly(sb)) 3274 return 0; 3275 } 3276 if (block_bitmap < first_block || block_bitmap > last_block) { 3277 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3278 "Block bitmap for group %u not in group " 3279 "(block %llu)!", i, block_bitmap); 3280 return 0; 3281 } 3282 inode_bitmap = ext4_inode_bitmap(sb, gdp); 3283 if (inode_bitmap == sb_block) { 3284 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3285 "Inode bitmap for group %u overlaps " 3286 "superblock", i); 3287 if (!sb_rdonly(sb)) 3288 return 0; 3289 } 3290 if (inode_bitmap >= sb_block + 1 && 3291 inode_bitmap <= last_bg_block) { 3292 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3293 "Inode bitmap for group %u overlaps " 3294 "block group descriptors", i); 3295 if (!sb_rdonly(sb)) 3296 return 0; 3297 } 3298 if (inode_bitmap < first_block || inode_bitmap > last_block) { 3299 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3300 "Inode bitmap for group %u not in group " 3301 "(block %llu)!", i, inode_bitmap); 3302 return 0; 3303 } 3304 inode_table = ext4_inode_table(sb, gdp); 3305 if (inode_table == sb_block) { 3306 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3307 "Inode table for group %u overlaps " 3308 "superblock", i); 3309 if (!sb_rdonly(sb)) 3310 return 0; 3311 } 3312 if (inode_table >= sb_block + 1 && 3313 inode_table <= last_bg_block) { 3314 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3315 "Inode table for group %u overlaps " 3316 "block group descriptors", i); 3317 if (!sb_rdonly(sb)) 3318 return 0; 3319 } 3320 if (inode_table < first_block || 3321 inode_table + sbi->s_itb_per_group - 1 > last_block) { 3322 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3323 "Inode table for group %u not in group " 3324 "(block %llu)!", i, inode_table); 3325 return 0; 3326 } 3327 ext4_lock_group(sb, i); 3328 if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 3329 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3330 "Checksum for group %u failed (%u!=%u)", 3331 i, le16_to_cpu(ext4_group_desc_csum(sb, i, 3332 gdp)), le16_to_cpu(gdp->bg_checksum)); 3333 if (!sb_rdonly(sb)) { 3334 ext4_unlock_group(sb, i); 3335 return 0; 3336 } 3337 } 3338 ext4_unlock_group(sb, i); 3339 if (!flexbg_flag) 3340 first_block += EXT4_BLOCKS_PER_GROUP(sb); 3341 } 3342 if (NULL != first_not_zeroed) 3343 *first_not_zeroed = grp; 3344 return 1; 3345 } 3346 3347 /* 3348 * Maximal extent format file size. 3349 * Resulting logical blkno at s_maxbytes must fit in our on-disk 3350 * extent format containers, within a sector_t, and within i_blocks 3351 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 3352 * so that won't be a limiting factor. 3353 * 3354 * However there is other limiting factor. We do store extents in the form 3355 * of starting block and length, hence the resulting length of the extent 3356 * covering maximum file size must fit into on-disk format containers as 3357 * well. Given that length is always by 1 unit bigger than max unit (because 3358 * we count 0 as well) we have to lower the s_maxbytes by one fs block. 3359 * 3360 * Note, this does *not* consider any metadata overhead for vfs i_blocks. 3361 */ 3362 static loff_t ext4_max_size(int blkbits, int has_huge_files) 3363 { 3364 loff_t res; 3365 loff_t upper_limit = MAX_LFS_FILESIZE; 3366 3367 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64)); 3368 3369 if (!has_huge_files) { 3370 upper_limit = (1LL << 32) - 1; 3371 3372 /* total blocks in file system block size */ 3373 upper_limit >>= (blkbits - 9); 3374 upper_limit <<= blkbits; 3375 } 3376 3377 /* 3378 * 32-bit extent-start container, ee_block. We lower the maxbytes 3379 * by one fs block, so ee_len can cover the extent of maximum file 3380 * size 3381 */ 3382 res = (1LL << 32) - 1; 3383 res <<= blkbits; 3384 3385 /* Sanity check against vm- & vfs- imposed limits */ 3386 if (res > upper_limit) 3387 res = upper_limit; 3388 3389 return res; 3390 } 3391 3392 /* 3393 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 3394 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 3395 * We need to be 1 filesystem block less than the 2^48 sector limit. 3396 */ 3397 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 3398 { 3399 loff_t upper_limit, res = EXT4_NDIR_BLOCKS; 3400 int meta_blocks; 3401 unsigned int ppb = 1 << (bits - 2); 3402 3403 /* 3404 * This is calculated to be the largest file size for a dense, block 3405 * mapped file such that the file's total number of 512-byte sectors, 3406 * including data and all indirect blocks, does not exceed (2^48 - 1). 3407 * 3408 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 3409 * number of 512-byte sectors of the file. 3410 */ 3411 if (!has_huge_files) { 3412 /* 3413 * !has_huge_files or implies that the inode i_block field 3414 * represents total file blocks in 2^32 512-byte sectors == 3415 * size of vfs inode i_blocks * 8 3416 */ 3417 upper_limit = (1LL << 32) - 1; 3418 3419 /* total blocks in file system block size */ 3420 upper_limit >>= (bits - 9); 3421 3422 } else { 3423 /* 3424 * We use 48 bit ext4_inode i_blocks 3425 * With EXT4_HUGE_FILE_FL set the i_blocks 3426 * represent total number of blocks in 3427 * file system block size 3428 */ 3429 upper_limit = (1LL << 48) - 1; 3430 3431 } 3432 3433 /* Compute how many blocks we can address by block tree */ 3434 res += ppb; 3435 res += ppb * ppb; 3436 res += ((loff_t)ppb) * ppb * ppb; 3437 /* Compute how many metadata blocks are needed */ 3438 meta_blocks = 1; 3439 meta_blocks += 1 + ppb; 3440 meta_blocks += 1 + ppb + ppb * ppb; 3441 /* Does block tree limit file size? */ 3442 if (res + meta_blocks <= upper_limit) 3443 goto check_lfs; 3444 3445 res = upper_limit; 3446 /* How many metadata blocks are needed for addressing upper_limit? */ 3447 upper_limit -= EXT4_NDIR_BLOCKS; 3448 /* indirect blocks */ 3449 meta_blocks = 1; 3450 upper_limit -= ppb; 3451 /* double indirect blocks */ 3452 if (upper_limit < ppb * ppb) { 3453 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb); 3454 res -= meta_blocks; 3455 goto check_lfs; 3456 } 3457 meta_blocks += 1 + ppb; 3458 upper_limit -= ppb * ppb; 3459 /* tripple indirect blocks for the rest */ 3460 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) + 3461 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb); 3462 res -= meta_blocks; 3463 check_lfs: 3464 res <<= bits; 3465 if (res > MAX_LFS_FILESIZE) 3466 res = MAX_LFS_FILESIZE; 3467 3468 return res; 3469 } 3470 3471 static ext4_fsblk_t descriptor_loc(struct super_block *sb, 3472 ext4_fsblk_t logical_sb_block, int nr) 3473 { 3474 struct ext4_sb_info *sbi = EXT4_SB(sb); 3475 ext4_group_t bg, first_meta_bg; 3476 int has_super = 0; 3477 3478 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 3479 3480 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 3481 return logical_sb_block + nr + 1; 3482 bg = sbi->s_desc_per_block * nr; 3483 if (ext4_bg_has_super(sb, bg)) 3484 has_super = 1; 3485 3486 /* 3487 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 3488 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 3489 * on modern mke2fs or blksize > 1k on older mke2fs) then we must 3490 * compensate. 3491 */ 3492 if (sb->s_blocksize == 1024 && nr == 0 && 3493 le32_to_cpu(sbi->s_es->s_first_data_block) == 0) 3494 has_super++; 3495 3496 return (has_super + ext4_group_first_block_no(sb, bg)); 3497 } 3498 3499 /** 3500 * ext4_get_stripe_size: Get the stripe size. 3501 * @sbi: In memory super block info 3502 * 3503 * If we have specified it via mount option, then 3504 * use the mount option value. If the value specified at mount time is 3505 * greater than the blocks per group use the super block value. 3506 * If the super block value is greater than blocks per group return 0. 3507 * Allocator needs it be less than blocks per group. 3508 * 3509 */ 3510 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 3511 { 3512 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 3513 unsigned long stripe_width = 3514 le32_to_cpu(sbi->s_es->s_raid_stripe_width); 3515 int ret; 3516 3517 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 3518 ret = sbi->s_stripe; 3519 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) 3520 ret = stripe_width; 3521 else if (stride && stride <= sbi->s_blocks_per_group) 3522 ret = stride; 3523 else 3524 ret = 0; 3525 3526 /* 3527 * If the stripe width is 1, this makes no sense and 3528 * we set it to 0 to turn off stripe handling code. 3529 */ 3530 if (ret <= 1) 3531 ret = 0; 3532 3533 return ret; 3534 } 3535 3536 /* 3537 * Check whether this filesystem can be mounted based on 3538 * the features present and the RDONLY/RDWR mount requested. 3539 * Returns 1 if this filesystem can be mounted as requested, 3540 * 0 if it cannot be. 3541 */ 3542 int ext4_feature_set_ok(struct super_block *sb, int readonly) 3543 { 3544 if (ext4_has_unknown_ext4_incompat_features(sb)) { 3545 ext4_msg(sb, KERN_ERR, 3546 "Couldn't mount because of " 3547 "unsupported optional features (%x)", 3548 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 3549 ~EXT4_FEATURE_INCOMPAT_SUPP)); 3550 return 0; 3551 } 3552 3553 #if !IS_ENABLED(CONFIG_UNICODE) 3554 if (ext4_has_feature_casefold(sb)) { 3555 ext4_msg(sb, KERN_ERR, 3556 "Filesystem with casefold feature cannot be " 3557 "mounted without CONFIG_UNICODE"); 3558 return 0; 3559 } 3560 #endif 3561 3562 if (readonly) 3563 return 1; 3564 3565 if (ext4_has_feature_readonly(sb)) { 3566 ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 3567 sb->s_flags |= SB_RDONLY; 3568 return 1; 3569 } 3570 3571 /* Check that feature set is OK for a read-write mount */ 3572 if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 3573 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 3574 "unsupported optional features (%x)", 3575 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 3576 ~EXT4_FEATURE_RO_COMPAT_SUPP)); 3577 return 0; 3578 } 3579 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 3580 ext4_msg(sb, KERN_ERR, 3581 "Can't support bigalloc feature without " 3582 "extents feature\n"); 3583 return 0; 3584 } 3585 3586 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2) 3587 if (!readonly && (ext4_has_feature_quota(sb) || 3588 ext4_has_feature_project(sb))) { 3589 ext4_msg(sb, KERN_ERR, 3590 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2"); 3591 return 0; 3592 } 3593 #endif /* CONFIG_QUOTA */ 3594 return 1; 3595 } 3596 3597 /* 3598 * This function is called once a day if we have errors logged 3599 * on the file system 3600 */ 3601 static void print_daily_error_info(struct timer_list *t) 3602 { 3603 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report); 3604 struct super_block *sb = sbi->s_sb; 3605 struct ext4_super_block *es = sbi->s_es; 3606 3607 if (es->s_error_count) 3608 /* fsck newer than v1.41.13 is needed to clean this condition. */ 3609 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 3610 le32_to_cpu(es->s_error_count)); 3611 if (es->s_first_error_time) { 3612 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d", 3613 sb->s_id, 3614 ext4_get_tstamp(es, s_first_error_time), 3615 (int) sizeof(es->s_first_error_func), 3616 es->s_first_error_func, 3617 le32_to_cpu(es->s_first_error_line)); 3618 if (es->s_first_error_ino) 3619 printk(KERN_CONT ": inode %u", 3620 le32_to_cpu(es->s_first_error_ino)); 3621 if (es->s_first_error_block) 3622 printk(KERN_CONT ": block %llu", (unsigned long long) 3623 le64_to_cpu(es->s_first_error_block)); 3624 printk(KERN_CONT "\n"); 3625 } 3626 if (es->s_last_error_time) { 3627 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d", 3628 sb->s_id, 3629 ext4_get_tstamp(es, s_last_error_time), 3630 (int) sizeof(es->s_last_error_func), 3631 es->s_last_error_func, 3632 le32_to_cpu(es->s_last_error_line)); 3633 if (es->s_last_error_ino) 3634 printk(KERN_CONT ": inode %u", 3635 le32_to_cpu(es->s_last_error_ino)); 3636 if (es->s_last_error_block) 3637 printk(KERN_CONT ": block %llu", (unsigned long long) 3638 le64_to_cpu(es->s_last_error_block)); 3639 printk(KERN_CONT "\n"); 3640 } 3641 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 3642 } 3643 3644 /* Find next suitable group and run ext4_init_inode_table */ 3645 static int ext4_run_li_request(struct ext4_li_request *elr) 3646 { 3647 struct ext4_group_desc *gdp = NULL; 3648 struct super_block *sb = elr->lr_super; 3649 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 3650 ext4_group_t group = elr->lr_next_group; 3651 unsigned int prefetch_ios = 0; 3652 int ret = 0; 3653 u64 start_time; 3654 3655 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) { 3656 elr->lr_next_group = ext4_mb_prefetch(sb, group, 3657 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios); 3658 if (prefetch_ios) 3659 ext4_mb_prefetch_fini(sb, elr->lr_next_group, 3660 prefetch_ios); 3661 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, 3662 prefetch_ios); 3663 if (group >= elr->lr_next_group) { 3664 ret = 1; 3665 if (elr->lr_first_not_zeroed != ngroups && 3666 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) { 3667 elr->lr_next_group = elr->lr_first_not_zeroed; 3668 elr->lr_mode = EXT4_LI_MODE_ITABLE; 3669 ret = 0; 3670 } 3671 } 3672 return ret; 3673 } 3674 3675 for (; group < ngroups; group++) { 3676 gdp = ext4_get_group_desc(sb, group, NULL); 3677 if (!gdp) { 3678 ret = 1; 3679 break; 3680 } 3681 3682 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3683 break; 3684 } 3685 3686 if (group >= ngroups) 3687 ret = 1; 3688 3689 if (!ret) { 3690 start_time = ktime_get_real_ns(); 3691 ret = ext4_init_inode_table(sb, group, 3692 elr->lr_timeout ? 0 : 1); 3693 trace_ext4_lazy_itable_init(sb, group); 3694 if (elr->lr_timeout == 0) { 3695 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) * 3696 EXT4_SB(elr->lr_super)->s_li_wait_mult); 3697 } 3698 elr->lr_next_sched = jiffies + elr->lr_timeout; 3699 elr->lr_next_group = group + 1; 3700 } 3701 return ret; 3702 } 3703 3704 /* 3705 * Remove lr_request from the list_request and free the 3706 * request structure. Should be called with li_list_mtx held 3707 */ 3708 static void ext4_remove_li_request(struct ext4_li_request *elr) 3709 { 3710 if (!elr) 3711 return; 3712 3713 list_del(&elr->lr_request); 3714 EXT4_SB(elr->lr_super)->s_li_request = NULL; 3715 kfree(elr); 3716 } 3717 3718 static void ext4_unregister_li_request(struct super_block *sb) 3719 { 3720 mutex_lock(&ext4_li_mtx); 3721 if (!ext4_li_info) { 3722 mutex_unlock(&ext4_li_mtx); 3723 return; 3724 } 3725 3726 mutex_lock(&ext4_li_info->li_list_mtx); 3727 ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 3728 mutex_unlock(&ext4_li_info->li_list_mtx); 3729 mutex_unlock(&ext4_li_mtx); 3730 } 3731 3732 static struct task_struct *ext4_lazyinit_task; 3733 3734 /* 3735 * This is the function where ext4lazyinit thread lives. It walks 3736 * through the request list searching for next scheduled filesystem. 3737 * When such a fs is found, run the lazy initialization request 3738 * (ext4_rn_li_request) and keep track of the time spend in this 3739 * function. Based on that time we compute next schedule time of 3740 * the request. When walking through the list is complete, compute 3741 * next waking time and put itself into sleep. 3742 */ 3743 static int ext4_lazyinit_thread(void *arg) 3744 { 3745 struct ext4_lazy_init *eli = arg; 3746 struct list_head *pos, *n; 3747 struct ext4_li_request *elr; 3748 unsigned long next_wakeup, cur; 3749 3750 BUG_ON(NULL == eli); 3751 3752 cont_thread: 3753 while (true) { 3754 next_wakeup = MAX_JIFFY_OFFSET; 3755 3756 mutex_lock(&eli->li_list_mtx); 3757 if (list_empty(&eli->li_request_list)) { 3758 mutex_unlock(&eli->li_list_mtx); 3759 goto exit_thread; 3760 } 3761 list_for_each_safe(pos, n, &eli->li_request_list) { 3762 int err = 0; 3763 int progress = 0; 3764 elr = list_entry(pos, struct ext4_li_request, 3765 lr_request); 3766 3767 if (time_before(jiffies, elr->lr_next_sched)) { 3768 if (time_before(elr->lr_next_sched, next_wakeup)) 3769 next_wakeup = elr->lr_next_sched; 3770 continue; 3771 } 3772 if (down_read_trylock(&elr->lr_super->s_umount)) { 3773 if (sb_start_write_trylock(elr->lr_super)) { 3774 progress = 1; 3775 /* 3776 * We hold sb->s_umount, sb can not 3777 * be removed from the list, it is 3778 * now safe to drop li_list_mtx 3779 */ 3780 mutex_unlock(&eli->li_list_mtx); 3781 err = ext4_run_li_request(elr); 3782 sb_end_write(elr->lr_super); 3783 mutex_lock(&eli->li_list_mtx); 3784 n = pos->next; 3785 } 3786 up_read((&elr->lr_super->s_umount)); 3787 } 3788 /* error, remove the lazy_init job */ 3789 if (err) { 3790 ext4_remove_li_request(elr); 3791 continue; 3792 } 3793 if (!progress) { 3794 elr->lr_next_sched = jiffies + 3795 (prandom_u32() 3796 % (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3797 } 3798 if (time_before(elr->lr_next_sched, next_wakeup)) 3799 next_wakeup = elr->lr_next_sched; 3800 } 3801 mutex_unlock(&eli->li_list_mtx); 3802 3803 try_to_freeze(); 3804 3805 cur = jiffies; 3806 if ((time_after_eq(cur, next_wakeup)) || 3807 (MAX_JIFFY_OFFSET == next_wakeup)) { 3808 cond_resched(); 3809 continue; 3810 } 3811 3812 schedule_timeout_interruptible(next_wakeup - cur); 3813 3814 if (kthread_should_stop()) { 3815 ext4_clear_request_list(); 3816 goto exit_thread; 3817 } 3818 } 3819 3820 exit_thread: 3821 /* 3822 * It looks like the request list is empty, but we need 3823 * to check it under the li_list_mtx lock, to prevent any 3824 * additions into it, and of course we should lock ext4_li_mtx 3825 * to atomically free the list and ext4_li_info, because at 3826 * this point another ext4 filesystem could be registering 3827 * new one. 3828 */ 3829 mutex_lock(&ext4_li_mtx); 3830 mutex_lock(&eli->li_list_mtx); 3831 if (!list_empty(&eli->li_request_list)) { 3832 mutex_unlock(&eli->li_list_mtx); 3833 mutex_unlock(&ext4_li_mtx); 3834 goto cont_thread; 3835 } 3836 mutex_unlock(&eli->li_list_mtx); 3837 kfree(ext4_li_info); 3838 ext4_li_info = NULL; 3839 mutex_unlock(&ext4_li_mtx); 3840 3841 return 0; 3842 } 3843 3844 static void ext4_clear_request_list(void) 3845 { 3846 struct list_head *pos, *n; 3847 struct ext4_li_request *elr; 3848 3849 mutex_lock(&ext4_li_info->li_list_mtx); 3850 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3851 elr = list_entry(pos, struct ext4_li_request, 3852 lr_request); 3853 ext4_remove_li_request(elr); 3854 } 3855 mutex_unlock(&ext4_li_info->li_list_mtx); 3856 } 3857 3858 static int ext4_run_lazyinit_thread(void) 3859 { 3860 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 3861 ext4_li_info, "ext4lazyinit"); 3862 if (IS_ERR(ext4_lazyinit_task)) { 3863 int err = PTR_ERR(ext4_lazyinit_task); 3864 ext4_clear_request_list(); 3865 kfree(ext4_li_info); 3866 ext4_li_info = NULL; 3867 printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3868 "initialization thread\n", 3869 err); 3870 return err; 3871 } 3872 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3873 return 0; 3874 } 3875 3876 /* 3877 * Check whether it make sense to run itable init. thread or not. 3878 * If there is at least one uninitialized inode table, return 3879 * corresponding group number, else the loop goes through all 3880 * groups and return total number of groups. 3881 */ 3882 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3883 { 3884 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3885 struct ext4_group_desc *gdp = NULL; 3886 3887 if (!ext4_has_group_desc_csum(sb)) 3888 return ngroups; 3889 3890 for (group = 0; group < ngroups; group++) { 3891 gdp = ext4_get_group_desc(sb, group, NULL); 3892 if (!gdp) 3893 continue; 3894 3895 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3896 break; 3897 } 3898 3899 return group; 3900 } 3901 3902 static int ext4_li_info_new(void) 3903 { 3904 struct ext4_lazy_init *eli = NULL; 3905 3906 eli = kzalloc(sizeof(*eli), GFP_KERNEL); 3907 if (!eli) 3908 return -ENOMEM; 3909 3910 INIT_LIST_HEAD(&eli->li_request_list); 3911 mutex_init(&eli->li_list_mtx); 3912 3913 eli->li_state |= EXT4_LAZYINIT_QUIT; 3914 3915 ext4_li_info = eli; 3916 3917 return 0; 3918 } 3919 3920 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3921 ext4_group_t start) 3922 { 3923 struct ext4_li_request *elr; 3924 3925 elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3926 if (!elr) 3927 return NULL; 3928 3929 elr->lr_super = sb; 3930 elr->lr_first_not_zeroed = start; 3931 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) { 3932 elr->lr_mode = EXT4_LI_MODE_ITABLE; 3933 elr->lr_next_group = start; 3934 } else { 3935 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP; 3936 } 3937 3938 /* 3939 * Randomize first schedule time of the request to 3940 * spread the inode table initialization requests 3941 * better. 3942 */ 3943 elr->lr_next_sched = jiffies + (prandom_u32() % 3944 (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3945 return elr; 3946 } 3947 3948 int ext4_register_li_request(struct super_block *sb, 3949 ext4_group_t first_not_zeroed) 3950 { 3951 struct ext4_sb_info *sbi = EXT4_SB(sb); 3952 struct ext4_li_request *elr = NULL; 3953 ext4_group_t ngroups = sbi->s_groups_count; 3954 int ret = 0; 3955 3956 mutex_lock(&ext4_li_mtx); 3957 if (sbi->s_li_request != NULL) { 3958 /* 3959 * Reset timeout so it can be computed again, because 3960 * s_li_wait_mult might have changed. 3961 */ 3962 sbi->s_li_request->lr_timeout = 0; 3963 goto out; 3964 } 3965 3966 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) && 3967 (first_not_zeroed == ngroups || sb_rdonly(sb) || 3968 !test_opt(sb, INIT_INODE_TABLE))) 3969 goto out; 3970 3971 elr = ext4_li_request_new(sb, first_not_zeroed); 3972 if (!elr) { 3973 ret = -ENOMEM; 3974 goto out; 3975 } 3976 3977 if (NULL == ext4_li_info) { 3978 ret = ext4_li_info_new(); 3979 if (ret) 3980 goto out; 3981 } 3982 3983 mutex_lock(&ext4_li_info->li_list_mtx); 3984 list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3985 mutex_unlock(&ext4_li_info->li_list_mtx); 3986 3987 sbi->s_li_request = elr; 3988 /* 3989 * set elr to NULL here since it has been inserted to 3990 * the request_list and the removal and free of it is 3991 * handled by ext4_clear_request_list from now on. 3992 */ 3993 elr = NULL; 3994 3995 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3996 ret = ext4_run_lazyinit_thread(); 3997 if (ret) 3998 goto out; 3999 } 4000 out: 4001 mutex_unlock(&ext4_li_mtx); 4002 if (ret) 4003 kfree(elr); 4004 return ret; 4005 } 4006 4007 /* 4008 * We do not need to lock anything since this is called on 4009 * module unload. 4010 */ 4011 static void ext4_destroy_lazyinit_thread(void) 4012 { 4013 /* 4014 * If thread exited earlier 4015 * there's nothing to be done. 4016 */ 4017 if (!ext4_li_info || !ext4_lazyinit_task) 4018 return; 4019 4020 kthread_stop(ext4_lazyinit_task); 4021 } 4022 4023 static int set_journal_csum_feature_set(struct super_block *sb) 4024 { 4025 int ret = 1; 4026 int compat, incompat; 4027 struct ext4_sb_info *sbi = EXT4_SB(sb); 4028 4029 if (ext4_has_metadata_csum(sb)) { 4030 /* journal checksum v3 */ 4031 compat = 0; 4032 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 4033 } else { 4034 /* journal checksum v1 */ 4035 compat = JBD2_FEATURE_COMPAT_CHECKSUM; 4036 incompat = 0; 4037 } 4038 4039 jbd2_journal_clear_features(sbi->s_journal, 4040 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 4041 JBD2_FEATURE_INCOMPAT_CSUM_V3 | 4042 JBD2_FEATURE_INCOMPAT_CSUM_V2); 4043 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4044 ret = jbd2_journal_set_features(sbi->s_journal, 4045 compat, 0, 4046 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 4047 incompat); 4048 } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 4049 ret = jbd2_journal_set_features(sbi->s_journal, 4050 compat, 0, 4051 incompat); 4052 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 4053 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 4054 } else { 4055 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 4056 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 4057 } 4058 4059 return ret; 4060 } 4061 4062 /* 4063 * Note: calculating the overhead so we can be compatible with 4064 * historical BSD practice is quite difficult in the face of 4065 * clusters/bigalloc. This is because multiple metadata blocks from 4066 * different block group can end up in the same allocation cluster. 4067 * Calculating the exact overhead in the face of clustered allocation 4068 * requires either O(all block bitmaps) in memory or O(number of block 4069 * groups**2) in time. We will still calculate the superblock for 4070 * older file systems --- and if we come across with a bigalloc file 4071 * system with zero in s_overhead_clusters the estimate will be close to 4072 * correct especially for very large cluster sizes --- but for newer 4073 * file systems, it's better to calculate this figure once at mkfs 4074 * time, and store it in the superblock. If the superblock value is 4075 * present (even for non-bigalloc file systems), we will use it. 4076 */ 4077 static int count_overhead(struct super_block *sb, ext4_group_t grp, 4078 char *buf) 4079 { 4080 struct ext4_sb_info *sbi = EXT4_SB(sb); 4081 struct ext4_group_desc *gdp; 4082 ext4_fsblk_t first_block, last_block, b; 4083 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 4084 int s, j, count = 0; 4085 int has_super = ext4_bg_has_super(sb, grp); 4086 4087 if (!ext4_has_feature_bigalloc(sb)) 4088 return (has_super + ext4_bg_num_gdb(sb, grp) + 4089 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) + 4090 sbi->s_itb_per_group + 2); 4091 4092 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 4093 (grp * EXT4_BLOCKS_PER_GROUP(sb)); 4094 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 4095 for (i = 0; i < ngroups; i++) { 4096 gdp = ext4_get_group_desc(sb, i, NULL); 4097 b = ext4_block_bitmap(sb, gdp); 4098 if (b >= first_block && b <= last_block) { 4099 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 4100 count++; 4101 } 4102 b = ext4_inode_bitmap(sb, gdp); 4103 if (b >= first_block && b <= last_block) { 4104 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 4105 count++; 4106 } 4107 b = ext4_inode_table(sb, gdp); 4108 if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 4109 for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 4110 int c = EXT4_B2C(sbi, b - first_block); 4111 ext4_set_bit(c, buf); 4112 count++; 4113 } 4114 if (i != grp) 4115 continue; 4116 s = 0; 4117 if (ext4_bg_has_super(sb, grp)) { 4118 ext4_set_bit(s++, buf); 4119 count++; 4120 } 4121 j = ext4_bg_num_gdb(sb, grp); 4122 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { 4123 ext4_error(sb, "Invalid number of block group " 4124 "descriptor blocks: %d", j); 4125 j = EXT4_BLOCKS_PER_GROUP(sb) - s; 4126 } 4127 count += j; 4128 for (; j > 0; j--) 4129 ext4_set_bit(EXT4_B2C(sbi, s++), buf); 4130 } 4131 if (!count) 4132 return 0; 4133 return EXT4_CLUSTERS_PER_GROUP(sb) - 4134 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 4135 } 4136 4137 /* 4138 * Compute the overhead and stash it in sbi->s_overhead 4139 */ 4140 int ext4_calculate_overhead(struct super_block *sb) 4141 { 4142 struct ext4_sb_info *sbi = EXT4_SB(sb); 4143 struct ext4_super_block *es = sbi->s_es; 4144 struct inode *j_inode; 4145 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); 4146 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 4147 ext4_fsblk_t overhead = 0; 4148 char *buf = (char *) get_zeroed_page(GFP_NOFS); 4149 4150 if (!buf) 4151 return -ENOMEM; 4152 4153 /* 4154 * Compute the overhead (FS structures). This is constant 4155 * for a given filesystem unless the number of block groups 4156 * changes so we cache the previous value until it does. 4157 */ 4158 4159 /* 4160 * All of the blocks before first_data_block are overhead 4161 */ 4162 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 4163 4164 /* 4165 * Add the overhead found in each block group 4166 */ 4167 for (i = 0; i < ngroups; i++) { 4168 int blks; 4169 4170 blks = count_overhead(sb, i, buf); 4171 overhead += blks; 4172 if (blks) 4173 memset(buf, 0, PAGE_SIZE); 4174 cond_resched(); 4175 } 4176 4177 /* 4178 * Add the internal journal blocks whether the journal has been 4179 * loaded or not 4180 */ 4181 if (sbi->s_journal && !sbi->s_journal_bdev) 4182 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len); 4183 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) { 4184 /* j_inum for internal journal is non-zero */ 4185 j_inode = ext4_get_journal_inode(sb, j_inum); 4186 if (j_inode) { 4187 j_blocks = j_inode->i_size >> sb->s_blocksize_bits; 4188 overhead += EXT4_NUM_B2C(sbi, j_blocks); 4189 iput(j_inode); 4190 } else { 4191 ext4_msg(sb, KERN_ERR, "can't get journal size"); 4192 } 4193 } 4194 sbi->s_overhead = overhead; 4195 smp_wmb(); 4196 free_page((unsigned long) buf); 4197 return 0; 4198 } 4199 4200 static void ext4_set_resv_clusters(struct super_block *sb) 4201 { 4202 ext4_fsblk_t resv_clusters; 4203 struct ext4_sb_info *sbi = EXT4_SB(sb); 4204 4205 /* 4206 * There's no need to reserve anything when we aren't using extents. 4207 * The space estimates are exact, there are no unwritten extents, 4208 * hole punching doesn't need new metadata... This is needed especially 4209 * to keep ext2/3 backward compatibility. 4210 */ 4211 if (!ext4_has_feature_extents(sb)) 4212 return; 4213 /* 4214 * By default we reserve 2% or 4096 clusters, whichever is smaller. 4215 * This should cover the situations where we can not afford to run 4216 * out of space like for example punch hole, or converting 4217 * unwritten extents in delalloc path. In most cases such 4218 * allocation would require 1, or 2 blocks, higher numbers are 4219 * very rare. 4220 */ 4221 resv_clusters = (ext4_blocks_count(sbi->s_es) >> 4222 sbi->s_cluster_bits); 4223 4224 do_div(resv_clusters, 50); 4225 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 4226 4227 atomic64_set(&sbi->s_resv_clusters, resv_clusters); 4228 } 4229 4230 static const char *ext4_quota_mode(struct super_block *sb) 4231 { 4232 #ifdef CONFIG_QUOTA 4233 if (!ext4_quota_capable(sb)) 4234 return "none"; 4235 4236 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb)) 4237 return "journalled"; 4238 else 4239 return "writeback"; 4240 #else 4241 return "disabled"; 4242 #endif 4243 } 4244 4245 static void ext4_setup_csum_trigger(struct super_block *sb, 4246 enum ext4_journal_trigger_type type, 4247 void (*trigger)( 4248 struct jbd2_buffer_trigger_type *type, 4249 struct buffer_head *bh, 4250 void *mapped_data, 4251 size_t size)) 4252 { 4253 struct ext4_sb_info *sbi = EXT4_SB(sb); 4254 4255 sbi->s_journal_triggers[type].sb = sb; 4256 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger; 4257 } 4258 4259 static void ext4_free_sbi(struct ext4_sb_info *sbi) 4260 { 4261 if (!sbi) 4262 return; 4263 4264 kfree(sbi->s_blockgroup_lock); 4265 fs_put_dax(sbi->s_daxdev); 4266 kfree(sbi); 4267 } 4268 4269 static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb) 4270 { 4271 struct ext4_sb_info *sbi; 4272 4273 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 4274 if (!sbi) 4275 return NULL; 4276 4277 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off); 4278 4279 sbi->s_blockgroup_lock = 4280 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 4281 4282 if (!sbi->s_blockgroup_lock) 4283 goto err_out; 4284 4285 sb->s_fs_info = sbi; 4286 sbi->s_sb = sb; 4287 return sbi; 4288 err_out: 4289 fs_put_dax(sbi->s_daxdev); 4290 kfree(sbi); 4291 return NULL; 4292 } 4293 4294 static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb) 4295 { 4296 struct buffer_head *bh, **group_desc; 4297 struct ext4_super_block *es = NULL; 4298 struct ext4_sb_info *sbi = EXT4_SB(sb); 4299 struct flex_groups **flex_groups; 4300 ext4_fsblk_t block; 4301 ext4_fsblk_t logical_sb_block; 4302 unsigned long offset = 0; 4303 unsigned long def_mount_opts; 4304 struct inode *root; 4305 int ret = -ENOMEM; 4306 int blocksize, clustersize; 4307 unsigned int db_count; 4308 unsigned int i; 4309 int needs_recovery, has_huge_files; 4310 __u64 blocks_count; 4311 int err = 0; 4312 ext4_group_t first_not_zeroed; 4313 struct ext4_fs_context *ctx = fc->fs_private; 4314 int silent = fc->sb_flags & SB_SILENT; 4315 4316 /* Set defaults for the variables that will be set during parsing */ 4317 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) 4318 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 4319 4320 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 4321 sbi->s_sectors_written_start = 4322 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]); 4323 4324 /* -EINVAL is default */ 4325 ret = -EINVAL; 4326 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 4327 if (!blocksize) { 4328 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 4329 goto out_fail; 4330 } 4331 4332 /* 4333 * The ext4 superblock will not be buffer aligned for other than 1kB 4334 * block sizes. We need to calculate the offset from buffer start. 4335 */ 4336 if (blocksize != EXT4_MIN_BLOCK_SIZE) { 4337 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE; 4338 offset = do_div(logical_sb_block, blocksize); 4339 } else { 4340 logical_sb_block = sbi->s_sb_block; 4341 } 4342 4343 bh = ext4_sb_bread_unmovable(sb, logical_sb_block); 4344 if (IS_ERR(bh)) { 4345 ext4_msg(sb, KERN_ERR, "unable to read superblock"); 4346 ret = PTR_ERR(bh); 4347 goto out_fail; 4348 } 4349 /* 4350 * Note: s_es must be initialized as soon as possible because 4351 * some ext4 macro-instructions depend on its value 4352 */ 4353 es = (struct ext4_super_block *) (bh->b_data + offset); 4354 sbi->s_es = es; 4355 sb->s_magic = le16_to_cpu(es->s_magic); 4356 if (sb->s_magic != EXT4_SUPER_MAGIC) 4357 goto cantfind_ext4; 4358 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 4359 4360 /* Warn if metadata_csum and gdt_csum are both set. */ 4361 if (ext4_has_feature_metadata_csum(sb) && 4362 ext4_has_feature_gdt_csum(sb)) 4363 ext4_warning(sb, "metadata_csum and uninit_bg are " 4364 "redundant flags; please run fsck."); 4365 4366 /* Check for a known checksum algorithm */ 4367 if (!ext4_verify_csum_type(sb, es)) { 4368 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 4369 "unknown checksum algorithm."); 4370 silent = 1; 4371 goto cantfind_ext4; 4372 } 4373 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE, 4374 ext4_orphan_file_block_trigger); 4375 4376 /* Load the checksum driver */ 4377 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 4378 if (IS_ERR(sbi->s_chksum_driver)) { 4379 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 4380 ret = PTR_ERR(sbi->s_chksum_driver); 4381 sbi->s_chksum_driver = NULL; 4382 goto failed_mount; 4383 } 4384 4385 /* Check superblock checksum */ 4386 if (!ext4_superblock_csum_verify(sb, es)) { 4387 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 4388 "invalid superblock checksum. Run e2fsck?"); 4389 silent = 1; 4390 ret = -EFSBADCRC; 4391 goto cantfind_ext4; 4392 } 4393 4394 /* Precompute checksum seed for all metadata */ 4395 if (ext4_has_feature_csum_seed(sb)) 4396 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 4397 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb)) 4398 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 4399 sizeof(es->s_uuid)); 4400 4401 /* Set defaults before we parse the mount options */ 4402 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 4403 set_opt(sb, INIT_INODE_TABLE); 4404 if (def_mount_opts & EXT4_DEFM_DEBUG) 4405 set_opt(sb, DEBUG); 4406 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 4407 set_opt(sb, GRPID); 4408 if (def_mount_opts & EXT4_DEFM_UID16) 4409 set_opt(sb, NO_UID32); 4410 /* xattr user namespace & acls are now defaulted on */ 4411 set_opt(sb, XATTR_USER); 4412 #ifdef CONFIG_EXT4_FS_POSIX_ACL 4413 set_opt(sb, POSIX_ACL); 4414 #endif 4415 if (ext4_has_feature_fast_commit(sb)) 4416 set_opt2(sb, JOURNAL_FAST_COMMIT); 4417 /* don't forget to enable journal_csum when metadata_csum is enabled. */ 4418 if (ext4_has_metadata_csum(sb)) 4419 set_opt(sb, JOURNAL_CHECKSUM); 4420 4421 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 4422 set_opt(sb, JOURNAL_DATA); 4423 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 4424 set_opt(sb, ORDERED_DATA); 4425 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 4426 set_opt(sb, WRITEBACK_DATA); 4427 4428 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 4429 set_opt(sb, ERRORS_PANIC); 4430 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 4431 set_opt(sb, ERRORS_CONT); 4432 else 4433 set_opt(sb, ERRORS_RO); 4434 /* block_validity enabled by default; disable with noblock_validity */ 4435 set_opt(sb, BLOCK_VALIDITY); 4436 if (def_mount_opts & EXT4_DEFM_DISCARD) 4437 set_opt(sb, DISCARD); 4438 4439 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 4440 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 4441 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 4442 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 4443 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 4444 4445 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 4446 set_opt(sb, BARRIER); 4447 4448 /* 4449 * enable delayed allocation by default 4450 * Use -o nodelalloc to turn it off 4451 */ 4452 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 4453 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 4454 set_opt(sb, DELALLOC); 4455 4456 /* 4457 * set default s_li_wait_mult for lazyinit, for the case there is 4458 * no mount option specified. 4459 */ 4460 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 4461 4462 if (le32_to_cpu(es->s_log_block_size) > 4463 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 4464 ext4_msg(sb, KERN_ERR, 4465 "Invalid log block size: %u", 4466 le32_to_cpu(es->s_log_block_size)); 4467 goto failed_mount; 4468 } 4469 if (le32_to_cpu(es->s_log_cluster_size) > 4470 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 4471 ext4_msg(sb, KERN_ERR, 4472 "Invalid log cluster size: %u", 4473 le32_to_cpu(es->s_log_cluster_size)); 4474 goto failed_mount; 4475 } 4476 4477 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 4478 4479 if (blocksize == PAGE_SIZE) 4480 set_opt(sb, DIOREAD_NOLOCK); 4481 4482 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 4483 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 4484 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 4485 } else { 4486 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 4487 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 4488 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) { 4489 ext4_msg(sb, KERN_ERR, "invalid first ino: %u", 4490 sbi->s_first_ino); 4491 goto failed_mount; 4492 } 4493 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 4494 (!is_power_of_2(sbi->s_inode_size)) || 4495 (sbi->s_inode_size > blocksize)) { 4496 ext4_msg(sb, KERN_ERR, 4497 "unsupported inode size: %d", 4498 sbi->s_inode_size); 4499 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize); 4500 goto failed_mount; 4501 } 4502 /* 4503 * i_atime_extra is the last extra field available for 4504 * [acm]times in struct ext4_inode. Checking for that 4505 * field should suffice to ensure we have extra space 4506 * for all three. 4507 */ 4508 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) + 4509 sizeof(((struct ext4_inode *)0)->i_atime_extra)) { 4510 sb->s_time_gran = 1; 4511 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX; 4512 } else { 4513 sb->s_time_gran = NSEC_PER_SEC; 4514 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX; 4515 } 4516 sb->s_time_min = EXT4_TIMESTAMP_MIN; 4517 } 4518 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 4519 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 4520 EXT4_GOOD_OLD_INODE_SIZE; 4521 if (ext4_has_feature_extra_isize(sb)) { 4522 unsigned v, max = (sbi->s_inode_size - 4523 EXT4_GOOD_OLD_INODE_SIZE); 4524 4525 v = le16_to_cpu(es->s_want_extra_isize); 4526 if (v > max) { 4527 ext4_msg(sb, KERN_ERR, 4528 "bad s_want_extra_isize: %d", v); 4529 goto failed_mount; 4530 } 4531 if (sbi->s_want_extra_isize < v) 4532 sbi->s_want_extra_isize = v; 4533 4534 v = le16_to_cpu(es->s_min_extra_isize); 4535 if (v > max) { 4536 ext4_msg(sb, KERN_ERR, 4537 "bad s_min_extra_isize: %d", v); 4538 goto failed_mount; 4539 } 4540 if (sbi->s_want_extra_isize < v) 4541 sbi->s_want_extra_isize = v; 4542 } 4543 } 4544 4545 err = parse_apply_sb_mount_options(sb, ctx); 4546 if (err < 0) 4547 goto failed_mount; 4548 4549 sbi->s_def_mount_opt = sbi->s_mount_opt; 4550 4551 err = ext4_check_opt_consistency(fc, sb); 4552 if (err < 0) 4553 goto failed_mount; 4554 4555 err = ext4_apply_options(fc, sb); 4556 if (err < 0) 4557 goto failed_mount; 4558 4559 #if IS_ENABLED(CONFIG_UNICODE) 4560 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) { 4561 const struct ext4_sb_encodings *encoding_info; 4562 struct unicode_map *encoding; 4563 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags); 4564 4565 encoding_info = ext4_sb_read_encoding(es); 4566 if (!encoding_info) { 4567 ext4_msg(sb, KERN_ERR, 4568 "Encoding requested by superblock is unknown"); 4569 goto failed_mount; 4570 } 4571 4572 encoding = utf8_load(encoding_info->version); 4573 if (IS_ERR(encoding)) { 4574 ext4_msg(sb, KERN_ERR, 4575 "can't mount with superblock charset: %s-%u.%u.%u " 4576 "not supported by the kernel. flags: 0x%x.", 4577 encoding_info->name, 4578 unicode_major(encoding_info->version), 4579 unicode_minor(encoding_info->version), 4580 unicode_rev(encoding_info->version), 4581 encoding_flags); 4582 goto failed_mount; 4583 } 4584 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: " 4585 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name, 4586 unicode_major(encoding_info->version), 4587 unicode_minor(encoding_info->version), 4588 unicode_rev(encoding_info->version), 4589 encoding_flags); 4590 4591 sb->s_encoding = encoding; 4592 sb->s_encoding_flags = encoding_flags; 4593 } 4594 #endif 4595 4596 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 4597 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n"); 4598 /* can't mount with both data=journal and dioread_nolock. */ 4599 clear_opt(sb, DIOREAD_NOLOCK); 4600 clear_opt2(sb, JOURNAL_FAST_COMMIT); 4601 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 4602 ext4_msg(sb, KERN_ERR, "can't mount with " 4603 "both data=journal and delalloc"); 4604 goto failed_mount; 4605 } 4606 if (test_opt(sb, DAX_ALWAYS)) { 4607 ext4_msg(sb, KERN_ERR, "can't mount with " 4608 "both data=journal and dax"); 4609 goto failed_mount; 4610 } 4611 if (ext4_has_feature_encrypt(sb)) { 4612 ext4_msg(sb, KERN_WARNING, 4613 "encrypted files will use data=ordered " 4614 "instead of data journaling mode"); 4615 } 4616 if (test_opt(sb, DELALLOC)) 4617 clear_opt(sb, DELALLOC); 4618 } else { 4619 sb->s_iflags |= SB_I_CGROUPWB; 4620 } 4621 4622 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 4623 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 4624 4625 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 4626 (ext4_has_compat_features(sb) || 4627 ext4_has_ro_compat_features(sb) || 4628 ext4_has_incompat_features(sb))) 4629 ext4_msg(sb, KERN_WARNING, 4630 "feature flags set on rev 0 fs, " 4631 "running e2fsck is recommended"); 4632 4633 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 4634 set_opt2(sb, HURD_COMPAT); 4635 if (ext4_has_feature_64bit(sb)) { 4636 ext4_msg(sb, KERN_ERR, 4637 "The Hurd can't support 64-bit file systems"); 4638 goto failed_mount; 4639 } 4640 4641 /* 4642 * ea_inode feature uses l_i_version field which is not 4643 * available in HURD_COMPAT mode. 4644 */ 4645 if (ext4_has_feature_ea_inode(sb)) { 4646 ext4_msg(sb, KERN_ERR, 4647 "ea_inode feature is not supported for Hurd"); 4648 goto failed_mount; 4649 } 4650 } 4651 4652 if (IS_EXT2_SB(sb)) { 4653 if (ext2_feature_set_ok(sb)) 4654 ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 4655 "using the ext4 subsystem"); 4656 else { 4657 /* 4658 * If we're probing be silent, if this looks like 4659 * it's actually an ext[34] filesystem. 4660 */ 4661 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 4662 goto failed_mount; 4663 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 4664 "to feature incompatibilities"); 4665 goto failed_mount; 4666 } 4667 } 4668 4669 if (IS_EXT3_SB(sb)) { 4670 if (ext3_feature_set_ok(sb)) 4671 ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 4672 "using the ext4 subsystem"); 4673 else { 4674 /* 4675 * If we're probing be silent, if this looks like 4676 * it's actually an ext4 filesystem. 4677 */ 4678 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 4679 goto failed_mount; 4680 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 4681 "to feature incompatibilities"); 4682 goto failed_mount; 4683 } 4684 } 4685 4686 /* 4687 * Check feature flags regardless of the revision level, since we 4688 * previously didn't change the revision level when setting the flags, 4689 * so there is a chance incompat flags are set on a rev 0 filesystem. 4690 */ 4691 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) 4692 goto failed_mount; 4693 4694 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) { 4695 ext4_msg(sb, KERN_ERR, 4696 "Number of reserved GDT blocks insanely large: %d", 4697 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); 4698 goto failed_mount; 4699 } 4700 4701 if (sbi->s_daxdev) { 4702 if (blocksize == PAGE_SIZE) 4703 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags); 4704 else 4705 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n"); 4706 } 4707 4708 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) { 4709 if (ext4_has_feature_inline_data(sb)) { 4710 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem" 4711 " that may contain inline data"); 4712 goto failed_mount; 4713 } 4714 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) { 4715 ext4_msg(sb, KERN_ERR, 4716 "DAX unsupported by block device."); 4717 goto failed_mount; 4718 } 4719 } 4720 4721 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 4722 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 4723 es->s_encryption_level); 4724 goto failed_mount; 4725 } 4726 4727 if (sb->s_blocksize != blocksize) { 4728 /* 4729 * bh must be released before kill_bdev(), otherwise 4730 * it won't be freed and its page also. kill_bdev() 4731 * is called by sb_set_blocksize(). 4732 */ 4733 brelse(bh); 4734 /* Validate the filesystem blocksize */ 4735 if (!sb_set_blocksize(sb, blocksize)) { 4736 ext4_msg(sb, KERN_ERR, "bad block size %d", 4737 blocksize); 4738 bh = NULL; 4739 goto failed_mount; 4740 } 4741 4742 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE; 4743 offset = do_div(logical_sb_block, blocksize); 4744 bh = ext4_sb_bread_unmovable(sb, logical_sb_block); 4745 if (IS_ERR(bh)) { 4746 ext4_msg(sb, KERN_ERR, 4747 "Can't read superblock on 2nd try"); 4748 ret = PTR_ERR(bh); 4749 bh = NULL; 4750 goto failed_mount; 4751 } 4752 es = (struct ext4_super_block *)(bh->b_data + offset); 4753 sbi->s_es = es; 4754 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 4755 ext4_msg(sb, KERN_ERR, 4756 "Magic mismatch, very weird!"); 4757 goto failed_mount; 4758 } 4759 } 4760 4761 has_huge_files = ext4_has_feature_huge_file(sb); 4762 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 4763 has_huge_files); 4764 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 4765 4766 sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 4767 if (ext4_has_feature_64bit(sb)) { 4768 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 4769 sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 4770 !is_power_of_2(sbi->s_desc_size)) { 4771 ext4_msg(sb, KERN_ERR, 4772 "unsupported descriptor size %lu", 4773 sbi->s_desc_size); 4774 goto failed_mount; 4775 } 4776 } else 4777 sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 4778 4779 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 4780 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 4781 4782 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 4783 if (sbi->s_inodes_per_block == 0) 4784 goto cantfind_ext4; 4785 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || 4786 sbi->s_inodes_per_group > blocksize * 8) { 4787 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", 4788 sbi->s_inodes_per_group); 4789 goto failed_mount; 4790 } 4791 sbi->s_itb_per_group = sbi->s_inodes_per_group / 4792 sbi->s_inodes_per_block; 4793 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 4794 sbi->s_sbh = bh; 4795 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY; 4796 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 4797 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 4798 4799 for (i = 0; i < 4; i++) 4800 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 4801 sbi->s_def_hash_version = es->s_def_hash_version; 4802 if (ext4_has_feature_dir_index(sb)) { 4803 i = le32_to_cpu(es->s_flags); 4804 if (i & EXT2_FLAGS_UNSIGNED_HASH) 4805 sbi->s_hash_unsigned = 3; 4806 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 4807 #ifdef __CHAR_UNSIGNED__ 4808 if (!sb_rdonly(sb)) 4809 es->s_flags |= 4810 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 4811 sbi->s_hash_unsigned = 3; 4812 #else 4813 if (!sb_rdonly(sb)) 4814 es->s_flags |= 4815 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 4816 #endif 4817 } 4818 } 4819 4820 /* Handle clustersize */ 4821 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 4822 if (ext4_has_feature_bigalloc(sb)) { 4823 if (clustersize < blocksize) { 4824 ext4_msg(sb, KERN_ERR, 4825 "cluster size (%d) smaller than " 4826 "block size (%d)", clustersize, blocksize); 4827 goto failed_mount; 4828 } 4829 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 4830 le32_to_cpu(es->s_log_block_size); 4831 sbi->s_clusters_per_group = 4832 le32_to_cpu(es->s_clusters_per_group); 4833 if (sbi->s_clusters_per_group > blocksize * 8) { 4834 ext4_msg(sb, KERN_ERR, 4835 "#clusters per group too big: %lu", 4836 sbi->s_clusters_per_group); 4837 goto failed_mount; 4838 } 4839 if (sbi->s_blocks_per_group != 4840 (sbi->s_clusters_per_group * (clustersize / blocksize))) { 4841 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 4842 "clusters per group (%lu) inconsistent", 4843 sbi->s_blocks_per_group, 4844 sbi->s_clusters_per_group); 4845 goto failed_mount; 4846 } 4847 } else { 4848 if (clustersize != blocksize) { 4849 ext4_msg(sb, KERN_ERR, 4850 "fragment/cluster size (%d) != " 4851 "block size (%d)", clustersize, blocksize); 4852 goto failed_mount; 4853 } 4854 if (sbi->s_blocks_per_group > blocksize * 8) { 4855 ext4_msg(sb, KERN_ERR, 4856 "#blocks per group too big: %lu", 4857 sbi->s_blocks_per_group); 4858 goto failed_mount; 4859 } 4860 sbi->s_clusters_per_group = sbi->s_blocks_per_group; 4861 sbi->s_cluster_bits = 0; 4862 } 4863 sbi->s_cluster_ratio = clustersize / blocksize; 4864 4865 /* Do we have standard group size of clustersize * 8 blocks ? */ 4866 if (sbi->s_blocks_per_group == clustersize << 3) 4867 set_opt2(sb, STD_GROUP_SIZE); 4868 4869 /* 4870 * Test whether we have more sectors than will fit in sector_t, 4871 * and whether the max offset is addressable by the page cache. 4872 */ 4873 err = generic_check_addressable(sb->s_blocksize_bits, 4874 ext4_blocks_count(es)); 4875 if (err) { 4876 ext4_msg(sb, KERN_ERR, "filesystem" 4877 " too large to mount safely on this system"); 4878 goto failed_mount; 4879 } 4880 4881 if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 4882 goto cantfind_ext4; 4883 4884 /* check blocks count against device size */ 4885 blocks_count = sb_bdev_nr_blocks(sb); 4886 if (blocks_count && ext4_blocks_count(es) > blocks_count) { 4887 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 4888 "exceeds size of device (%llu blocks)", 4889 ext4_blocks_count(es), blocks_count); 4890 goto failed_mount; 4891 } 4892 4893 /* 4894 * It makes no sense for the first data block to be beyond the end 4895 * of the filesystem. 4896 */ 4897 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 4898 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4899 "block %u is beyond end of filesystem (%llu)", 4900 le32_to_cpu(es->s_first_data_block), 4901 ext4_blocks_count(es)); 4902 goto failed_mount; 4903 } 4904 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) && 4905 (sbi->s_cluster_ratio == 1)) { 4906 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4907 "block is 0 with a 1k block and cluster size"); 4908 goto failed_mount; 4909 } 4910 4911 blocks_count = (ext4_blocks_count(es) - 4912 le32_to_cpu(es->s_first_data_block) + 4913 EXT4_BLOCKS_PER_GROUP(sb) - 1); 4914 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 4915 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 4916 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu " 4917 "(block count %llu, first data block %u, " 4918 "blocks per group %lu)", blocks_count, 4919 ext4_blocks_count(es), 4920 le32_to_cpu(es->s_first_data_block), 4921 EXT4_BLOCKS_PER_GROUP(sb)); 4922 goto failed_mount; 4923 } 4924 sbi->s_groups_count = blocks_count; 4925 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 4926 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 4927 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) != 4928 le32_to_cpu(es->s_inodes_count)) { 4929 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu", 4930 le32_to_cpu(es->s_inodes_count), 4931 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group)); 4932 ret = -EINVAL; 4933 goto failed_mount; 4934 } 4935 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 4936 EXT4_DESC_PER_BLOCK(sb); 4937 if (ext4_has_feature_meta_bg(sb)) { 4938 if (le32_to_cpu(es->s_first_meta_bg) > db_count) { 4939 ext4_msg(sb, KERN_WARNING, 4940 "first meta block group too large: %u " 4941 "(group descriptor block count %u)", 4942 le32_to_cpu(es->s_first_meta_bg), db_count); 4943 goto failed_mount; 4944 } 4945 } 4946 rcu_assign_pointer(sbi->s_group_desc, 4947 kvmalloc_array(db_count, 4948 sizeof(struct buffer_head *), 4949 GFP_KERNEL)); 4950 if (sbi->s_group_desc == NULL) { 4951 ext4_msg(sb, KERN_ERR, "not enough memory"); 4952 ret = -ENOMEM; 4953 goto failed_mount; 4954 } 4955 4956 bgl_lock_init(sbi->s_blockgroup_lock); 4957 4958 /* Pre-read the descriptors into the buffer cache */ 4959 for (i = 0; i < db_count; i++) { 4960 block = descriptor_loc(sb, logical_sb_block, i); 4961 ext4_sb_breadahead_unmovable(sb, block); 4962 } 4963 4964 for (i = 0; i < db_count; i++) { 4965 struct buffer_head *bh; 4966 4967 block = descriptor_loc(sb, logical_sb_block, i); 4968 bh = ext4_sb_bread_unmovable(sb, block); 4969 if (IS_ERR(bh)) { 4970 ext4_msg(sb, KERN_ERR, 4971 "can't read group descriptor %d", i); 4972 db_count = i; 4973 ret = PTR_ERR(bh); 4974 goto failed_mount2; 4975 } 4976 rcu_read_lock(); 4977 rcu_dereference(sbi->s_group_desc)[i] = bh; 4978 rcu_read_unlock(); 4979 } 4980 sbi->s_gdb_count = db_count; 4981 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) { 4982 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 4983 ret = -EFSCORRUPTED; 4984 goto failed_mount2; 4985 } 4986 4987 timer_setup(&sbi->s_err_report, print_daily_error_info, 0); 4988 spin_lock_init(&sbi->s_error_lock); 4989 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work); 4990 4991 /* Register extent status tree shrinker */ 4992 if (ext4_es_register_shrinker(sbi)) 4993 goto failed_mount3; 4994 4995 sbi->s_stripe = ext4_get_stripe_size(sbi); 4996 sbi->s_extent_max_zeroout_kb = 32; 4997 4998 /* 4999 * set up enough so that it can read an inode 5000 */ 5001 sb->s_op = &ext4_sops; 5002 sb->s_export_op = &ext4_export_ops; 5003 sb->s_xattr = ext4_xattr_handlers; 5004 #ifdef CONFIG_FS_ENCRYPTION 5005 sb->s_cop = &ext4_cryptops; 5006 #endif 5007 #ifdef CONFIG_FS_VERITY 5008 sb->s_vop = &ext4_verityops; 5009 #endif 5010 #ifdef CONFIG_QUOTA 5011 sb->dq_op = &ext4_quota_operations; 5012 if (ext4_has_feature_quota(sb)) 5013 sb->s_qcop = &dquot_quotactl_sysfile_ops; 5014 else 5015 sb->s_qcop = &ext4_qctl_operations; 5016 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 5017 #endif 5018 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 5019 5020 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 5021 mutex_init(&sbi->s_orphan_lock); 5022 5023 /* Initialize fast commit stuff */ 5024 atomic_set(&sbi->s_fc_subtid, 0); 5025 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]); 5026 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]); 5027 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]); 5028 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]); 5029 sbi->s_fc_bytes = 0; 5030 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE); 5031 sbi->s_fc_ineligible_tid = 0; 5032 spin_lock_init(&sbi->s_fc_lock); 5033 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats)); 5034 sbi->s_fc_replay_state.fc_regions = NULL; 5035 sbi->s_fc_replay_state.fc_regions_size = 0; 5036 sbi->s_fc_replay_state.fc_regions_used = 0; 5037 sbi->s_fc_replay_state.fc_regions_valid = 0; 5038 sbi->s_fc_replay_state.fc_modified_inodes = NULL; 5039 sbi->s_fc_replay_state.fc_modified_inodes_size = 0; 5040 sbi->s_fc_replay_state.fc_modified_inodes_used = 0; 5041 5042 sb->s_root = NULL; 5043 5044 needs_recovery = (es->s_last_orphan != 0 || 5045 ext4_has_feature_orphan_present(sb) || 5046 ext4_has_feature_journal_needs_recovery(sb)); 5047 5048 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) 5049 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 5050 goto failed_mount3a; 5051 5052 /* 5053 * The first inode we look at is the journal inode. Don't try 5054 * root first: it may be modified in the journal! 5055 */ 5056 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 5057 err = ext4_load_journal(sb, es, ctx->journal_devnum); 5058 if (err) 5059 goto failed_mount3a; 5060 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && 5061 ext4_has_feature_journal_needs_recovery(sb)) { 5062 ext4_msg(sb, KERN_ERR, "required journal recovery " 5063 "suppressed and not mounted read-only"); 5064 goto failed_mount_wq; 5065 } else { 5066 /* Nojournal mode, all journal mount options are illegal */ 5067 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { 5068 ext4_msg(sb, KERN_ERR, "can't mount with " 5069 "journal_checksum, fs mounted w/o journal"); 5070 goto failed_mount_wq; 5071 } 5072 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 5073 ext4_msg(sb, KERN_ERR, "can't mount with " 5074 "journal_async_commit, fs mounted w/o journal"); 5075 goto failed_mount_wq; 5076 } 5077 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { 5078 ext4_msg(sb, KERN_ERR, "can't mount with " 5079 "commit=%lu, fs mounted w/o journal", 5080 sbi->s_commit_interval / HZ); 5081 goto failed_mount_wq; 5082 } 5083 if (EXT4_MOUNT_DATA_FLAGS & 5084 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 5085 ext4_msg(sb, KERN_ERR, "can't mount with " 5086 "data=, fs mounted w/o journal"); 5087 goto failed_mount_wq; 5088 } 5089 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM; 5090 clear_opt(sb, JOURNAL_CHECKSUM); 5091 clear_opt(sb, DATA_FLAGS); 5092 clear_opt2(sb, JOURNAL_FAST_COMMIT); 5093 sbi->s_journal = NULL; 5094 needs_recovery = 0; 5095 goto no_journal; 5096 } 5097 5098 if (ext4_has_feature_64bit(sb) && 5099 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 5100 JBD2_FEATURE_INCOMPAT_64BIT)) { 5101 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 5102 goto failed_mount_wq; 5103 } 5104 5105 if (!set_journal_csum_feature_set(sb)) { 5106 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 5107 "feature set"); 5108 goto failed_mount_wq; 5109 } 5110 5111 if (test_opt2(sb, JOURNAL_FAST_COMMIT) && 5112 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 5113 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) { 5114 ext4_msg(sb, KERN_ERR, 5115 "Failed to set fast commit journal feature"); 5116 goto failed_mount_wq; 5117 } 5118 5119 /* We have now updated the journal if required, so we can 5120 * validate the data journaling mode. */ 5121 switch (test_opt(sb, DATA_FLAGS)) { 5122 case 0: 5123 /* No mode set, assume a default based on the journal 5124 * capabilities: ORDERED_DATA if the journal can 5125 * cope, else JOURNAL_DATA 5126 */ 5127 if (jbd2_journal_check_available_features 5128 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 5129 set_opt(sb, ORDERED_DATA); 5130 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA; 5131 } else { 5132 set_opt(sb, JOURNAL_DATA); 5133 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; 5134 } 5135 break; 5136 5137 case EXT4_MOUNT_ORDERED_DATA: 5138 case EXT4_MOUNT_WRITEBACK_DATA: 5139 if (!jbd2_journal_check_available_features 5140 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 5141 ext4_msg(sb, KERN_ERR, "Journal does not support " 5142 "requested data journaling mode"); 5143 goto failed_mount_wq; 5144 } 5145 break; 5146 default: 5147 break; 5148 } 5149 5150 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 5151 test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 5152 ext4_msg(sb, KERN_ERR, "can't mount with " 5153 "journal_async_commit in data=ordered mode"); 5154 goto failed_mount_wq; 5155 } 5156 5157 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio); 5158 5159 sbi->s_journal->j_submit_inode_data_buffers = 5160 ext4_journal_submit_inode_data_buffers; 5161 sbi->s_journal->j_finish_inode_data_buffers = 5162 ext4_journal_finish_inode_data_buffers; 5163 5164 no_journal: 5165 if (!test_opt(sb, NO_MBCACHE)) { 5166 sbi->s_ea_block_cache = ext4_xattr_create_cache(); 5167 if (!sbi->s_ea_block_cache) { 5168 ext4_msg(sb, KERN_ERR, 5169 "Failed to create ea_block_cache"); 5170 goto failed_mount_wq; 5171 } 5172 5173 if (ext4_has_feature_ea_inode(sb)) { 5174 sbi->s_ea_inode_cache = ext4_xattr_create_cache(); 5175 if (!sbi->s_ea_inode_cache) { 5176 ext4_msg(sb, KERN_ERR, 5177 "Failed to create ea_inode_cache"); 5178 goto failed_mount_wq; 5179 } 5180 } 5181 } 5182 5183 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) { 5184 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity"); 5185 goto failed_mount_wq; 5186 } 5187 5188 /* 5189 * Get the # of file system overhead blocks from the 5190 * superblock if present. 5191 */ 5192 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 5193 /* ignore the precalculated value if it is ridiculous */ 5194 if (sbi->s_overhead > ext4_blocks_count(es)) 5195 sbi->s_overhead = 0; 5196 /* 5197 * If the bigalloc feature is not enabled recalculating the 5198 * overhead doesn't take long, so we might as well just redo 5199 * it to make sure we are using the correct value. 5200 */ 5201 if (!ext4_has_feature_bigalloc(sb)) 5202 sbi->s_overhead = 0; 5203 if (sbi->s_overhead == 0) { 5204 err = ext4_calculate_overhead(sb); 5205 if (err) 5206 goto failed_mount_wq; 5207 } 5208 5209 /* 5210 * The maximum number of concurrent works can be high and 5211 * concurrency isn't really necessary. Limit it to 1. 5212 */ 5213 EXT4_SB(sb)->rsv_conversion_wq = 5214 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 5215 if (!EXT4_SB(sb)->rsv_conversion_wq) { 5216 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 5217 ret = -ENOMEM; 5218 goto failed_mount4; 5219 } 5220 5221 /* 5222 * The jbd2_journal_load will have done any necessary log recovery, 5223 * so we can safely mount the rest of the filesystem now. 5224 */ 5225 5226 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL); 5227 if (IS_ERR(root)) { 5228 ext4_msg(sb, KERN_ERR, "get root inode failed"); 5229 ret = PTR_ERR(root); 5230 root = NULL; 5231 goto failed_mount4; 5232 } 5233 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 5234 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 5235 iput(root); 5236 goto failed_mount4; 5237 } 5238 5239 sb->s_root = d_make_root(root); 5240 if (!sb->s_root) { 5241 ext4_msg(sb, KERN_ERR, "get root dentry failed"); 5242 ret = -ENOMEM; 5243 goto failed_mount4; 5244 } 5245 5246 ret = ext4_setup_super(sb, es, sb_rdonly(sb)); 5247 if (ret == -EROFS) { 5248 sb->s_flags |= SB_RDONLY; 5249 ret = 0; 5250 } else if (ret) 5251 goto failed_mount4a; 5252 5253 ext4_set_resv_clusters(sb); 5254 5255 if (test_opt(sb, BLOCK_VALIDITY)) { 5256 err = ext4_setup_system_zone(sb); 5257 if (err) { 5258 ext4_msg(sb, KERN_ERR, "failed to initialize system " 5259 "zone (%d)", err); 5260 goto failed_mount4a; 5261 } 5262 } 5263 ext4_fc_replay_cleanup(sb); 5264 5265 ext4_ext_init(sb); 5266 5267 /* 5268 * Enable optimize_scan if number of groups is > threshold. This can be 5269 * turned off by passing "mb_optimize_scan=0". This can also be 5270 * turned on forcefully by passing "mb_optimize_scan=1". 5271 */ 5272 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) { 5273 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD) 5274 set_opt2(sb, MB_OPTIMIZE_SCAN); 5275 else 5276 clear_opt2(sb, MB_OPTIMIZE_SCAN); 5277 } 5278 5279 err = ext4_mb_init(sb); 5280 if (err) { 5281 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 5282 err); 5283 goto failed_mount5; 5284 } 5285 5286 /* 5287 * We can only set up the journal commit callback once 5288 * mballoc is initialized 5289 */ 5290 if (sbi->s_journal) 5291 sbi->s_journal->j_commit_callback = 5292 ext4_journal_commit_callback; 5293 5294 block = ext4_count_free_clusters(sb); 5295 ext4_free_blocks_count_set(sbi->s_es, 5296 EXT4_C2B(sbi, block)); 5297 err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 5298 GFP_KERNEL); 5299 if (!err) { 5300 unsigned long freei = ext4_count_free_inodes(sb); 5301 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 5302 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 5303 GFP_KERNEL); 5304 } 5305 /* 5306 * Update the checksum after updating free space/inode 5307 * counters. Otherwise the superblock can have an incorrect 5308 * checksum in the buffer cache until it is written out and 5309 * e2fsprogs programs trying to open a file system immediately 5310 * after it is mounted can fail. 5311 */ 5312 ext4_superblock_csum_set(sb); 5313 if (!err) 5314 err = percpu_counter_init(&sbi->s_dirs_counter, 5315 ext4_count_dirs(sb), GFP_KERNEL); 5316 if (!err) 5317 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 5318 GFP_KERNEL); 5319 if (!err) 5320 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0, 5321 GFP_KERNEL); 5322 if (!err) 5323 err = percpu_init_rwsem(&sbi->s_writepages_rwsem); 5324 5325 if (err) { 5326 ext4_msg(sb, KERN_ERR, "insufficient memory"); 5327 goto failed_mount6; 5328 } 5329 5330 if (ext4_has_feature_flex_bg(sb)) 5331 if (!ext4_fill_flex_info(sb)) { 5332 ext4_msg(sb, KERN_ERR, 5333 "unable to initialize " 5334 "flex_bg meta info!"); 5335 ret = -ENOMEM; 5336 goto failed_mount6; 5337 } 5338 5339 err = ext4_register_li_request(sb, first_not_zeroed); 5340 if (err) 5341 goto failed_mount6; 5342 5343 err = ext4_register_sysfs(sb); 5344 if (err) 5345 goto failed_mount7; 5346 5347 err = ext4_init_orphan_info(sb); 5348 if (err) 5349 goto failed_mount8; 5350 #ifdef CONFIG_QUOTA 5351 /* Enable quota usage during mount. */ 5352 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { 5353 err = ext4_enable_quotas(sb); 5354 if (err) 5355 goto failed_mount9; 5356 } 5357 #endif /* CONFIG_QUOTA */ 5358 5359 /* 5360 * Save the original bdev mapping's wb_err value which could be 5361 * used to detect the metadata async write error. 5362 */ 5363 spin_lock_init(&sbi->s_bdev_wb_lock); 5364 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err, 5365 &sbi->s_bdev_wb_err); 5366 sb->s_bdev->bd_super = sb; 5367 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 5368 ext4_orphan_cleanup(sb, es); 5369 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 5370 if (needs_recovery) { 5371 ext4_msg(sb, KERN_INFO, "recovery complete"); 5372 err = ext4_mark_recovery_complete(sb, es); 5373 if (err) 5374 goto failed_mount9; 5375 } 5376 5377 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev)) 5378 ext4_msg(sb, KERN_WARNING, 5379 "mounting with \"discard\" option, but the device does not support discard"); 5380 5381 if (es->s_error_count) 5382 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 5383 5384 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 5385 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 5386 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 5387 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 5388 atomic_set(&sbi->s_warning_count, 0); 5389 atomic_set(&sbi->s_msg_count, 0); 5390 5391 return 0; 5392 5393 cantfind_ext4: 5394 if (!silent) 5395 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 5396 goto failed_mount; 5397 5398 failed_mount9: 5399 ext4_release_orphan_info(sb); 5400 failed_mount8: 5401 ext4_unregister_sysfs(sb); 5402 kobject_put(&sbi->s_kobj); 5403 failed_mount7: 5404 ext4_unregister_li_request(sb); 5405 failed_mount6: 5406 ext4_mb_release(sb); 5407 rcu_read_lock(); 5408 flex_groups = rcu_dereference(sbi->s_flex_groups); 5409 if (flex_groups) { 5410 for (i = 0; i < sbi->s_flex_groups_allocated; i++) 5411 kvfree(flex_groups[i]); 5412 kvfree(flex_groups); 5413 } 5414 rcu_read_unlock(); 5415 percpu_counter_destroy(&sbi->s_freeclusters_counter); 5416 percpu_counter_destroy(&sbi->s_freeinodes_counter); 5417 percpu_counter_destroy(&sbi->s_dirs_counter); 5418 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 5419 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit); 5420 percpu_free_rwsem(&sbi->s_writepages_rwsem); 5421 failed_mount5: 5422 ext4_ext_release(sb); 5423 ext4_release_system_zone(sb); 5424 failed_mount4a: 5425 dput(sb->s_root); 5426 sb->s_root = NULL; 5427 failed_mount4: 5428 ext4_msg(sb, KERN_ERR, "mount failed"); 5429 if (EXT4_SB(sb)->rsv_conversion_wq) 5430 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 5431 failed_mount_wq: 5432 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 5433 sbi->s_ea_inode_cache = NULL; 5434 5435 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 5436 sbi->s_ea_block_cache = NULL; 5437 5438 if (sbi->s_journal) { 5439 /* flush s_error_work before journal destroy. */ 5440 flush_work(&sbi->s_error_work); 5441 jbd2_journal_destroy(sbi->s_journal); 5442 sbi->s_journal = NULL; 5443 } 5444 failed_mount3a: 5445 ext4_es_unregister_shrinker(sbi); 5446 failed_mount3: 5447 /* flush s_error_work before sbi destroy */ 5448 flush_work(&sbi->s_error_work); 5449 del_timer_sync(&sbi->s_err_report); 5450 ext4_stop_mmpd(sbi); 5451 failed_mount2: 5452 rcu_read_lock(); 5453 group_desc = rcu_dereference(sbi->s_group_desc); 5454 for (i = 0; i < db_count; i++) 5455 brelse(group_desc[i]); 5456 kvfree(group_desc); 5457 rcu_read_unlock(); 5458 failed_mount: 5459 if (sbi->s_chksum_driver) 5460 crypto_free_shash(sbi->s_chksum_driver); 5461 5462 #if IS_ENABLED(CONFIG_UNICODE) 5463 utf8_unload(sb->s_encoding); 5464 #endif 5465 5466 #ifdef CONFIG_QUOTA 5467 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5468 kfree(get_qf_name(sb, sbi, i)); 5469 #endif 5470 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy); 5471 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */ 5472 brelse(bh); 5473 ext4_blkdev_remove(sbi); 5474 out_fail: 5475 sb->s_fs_info = NULL; 5476 return err ? err : ret; 5477 } 5478 5479 static int ext4_fill_super(struct super_block *sb, struct fs_context *fc) 5480 { 5481 struct ext4_fs_context *ctx = fc->fs_private; 5482 struct ext4_sb_info *sbi; 5483 const char *descr; 5484 int ret; 5485 5486 sbi = ext4_alloc_sbi(sb); 5487 if (!sbi) 5488 return -ENOMEM; 5489 5490 fc->s_fs_info = sbi; 5491 5492 /* Cleanup superblock name */ 5493 strreplace(sb->s_id, '/', '!'); 5494 5495 sbi->s_sb_block = 1; /* Default super block location */ 5496 if (ctx->spec & EXT4_SPEC_s_sb_block) 5497 sbi->s_sb_block = ctx->s_sb_block; 5498 5499 ret = __ext4_fill_super(fc, sb); 5500 if (ret < 0) 5501 goto free_sbi; 5502 5503 if (sbi->s_journal) { 5504 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 5505 descr = " journalled data mode"; 5506 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 5507 descr = " ordered data mode"; 5508 else 5509 descr = " writeback data mode"; 5510 } else 5511 descr = "out journal"; 5512 5513 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 5514 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 5515 "Quota mode: %s.", descr, ext4_quota_mode(sb)); 5516 5517 /* Update the s_overhead_clusters if necessary */ 5518 ext4_update_overhead(sb); 5519 return 0; 5520 5521 free_sbi: 5522 ext4_free_sbi(sbi); 5523 fc->s_fs_info = NULL; 5524 return ret; 5525 } 5526 5527 static int ext4_get_tree(struct fs_context *fc) 5528 { 5529 return get_tree_bdev(fc, ext4_fill_super); 5530 } 5531 5532 /* 5533 * Setup any per-fs journal parameters now. We'll do this both on 5534 * initial mount, once the journal has been initialised but before we've 5535 * done any recovery; and again on any subsequent remount. 5536 */ 5537 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 5538 { 5539 struct ext4_sb_info *sbi = EXT4_SB(sb); 5540 5541 journal->j_commit_interval = sbi->s_commit_interval; 5542 journal->j_min_batch_time = sbi->s_min_batch_time; 5543 journal->j_max_batch_time = sbi->s_max_batch_time; 5544 ext4_fc_init(sb, journal); 5545 5546 write_lock(&journal->j_state_lock); 5547 if (test_opt(sb, BARRIER)) 5548 journal->j_flags |= JBD2_BARRIER; 5549 else 5550 journal->j_flags &= ~JBD2_BARRIER; 5551 if (test_opt(sb, DATA_ERR_ABORT)) 5552 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 5553 else 5554 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 5555 write_unlock(&journal->j_state_lock); 5556 } 5557 5558 static struct inode *ext4_get_journal_inode(struct super_block *sb, 5559 unsigned int journal_inum) 5560 { 5561 struct inode *journal_inode; 5562 5563 /* 5564 * Test for the existence of a valid inode on disk. Bad things 5565 * happen if we iget() an unused inode, as the subsequent iput() 5566 * will try to delete it. 5567 */ 5568 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL); 5569 if (IS_ERR(journal_inode)) { 5570 ext4_msg(sb, KERN_ERR, "no journal found"); 5571 return NULL; 5572 } 5573 if (!journal_inode->i_nlink) { 5574 make_bad_inode(journal_inode); 5575 iput(journal_inode); 5576 ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 5577 return NULL; 5578 } 5579 5580 jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 5581 journal_inode, journal_inode->i_size); 5582 if (!S_ISREG(journal_inode->i_mode)) { 5583 ext4_msg(sb, KERN_ERR, "invalid journal inode"); 5584 iput(journal_inode); 5585 return NULL; 5586 } 5587 return journal_inode; 5588 } 5589 5590 static journal_t *ext4_get_journal(struct super_block *sb, 5591 unsigned int journal_inum) 5592 { 5593 struct inode *journal_inode; 5594 journal_t *journal; 5595 5596 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 5597 return NULL; 5598 5599 journal_inode = ext4_get_journal_inode(sb, journal_inum); 5600 if (!journal_inode) 5601 return NULL; 5602 5603 journal = jbd2_journal_init_inode(journal_inode); 5604 if (!journal) { 5605 ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 5606 iput(journal_inode); 5607 return NULL; 5608 } 5609 journal->j_private = sb; 5610 ext4_init_journal_params(sb, journal); 5611 return journal; 5612 } 5613 5614 static journal_t *ext4_get_dev_journal(struct super_block *sb, 5615 dev_t j_dev) 5616 { 5617 struct buffer_head *bh; 5618 journal_t *journal; 5619 ext4_fsblk_t start; 5620 ext4_fsblk_t len; 5621 int hblock, blocksize; 5622 ext4_fsblk_t sb_block; 5623 unsigned long offset; 5624 struct ext4_super_block *es; 5625 struct block_device *bdev; 5626 5627 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 5628 return NULL; 5629 5630 bdev = ext4_blkdev_get(j_dev, sb); 5631 if (bdev == NULL) 5632 return NULL; 5633 5634 blocksize = sb->s_blocksize; 5635 hblock = bdev_logical_block_size(bdev); 5636 if (blocksize < hblock) { 5637 ext4_msg(sb, KERN_ERR, 5638 "blocksize too small for journal device"); 5639 goto out_bdev; 5640 } 5641 5642 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 5643 offset = EXT4_MIN_BLOCK_SIZE % blocksize; 5644 set_blocksize(bdev, blocksize); 5645 if (!(bh = __bread(bdev, sb_block, blocksize))) { 5646 ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 5647 "external journal"); 5648 goto out_bdev; 5649 } 5650 5651 es = (struct ext4_super_block *) (bh->b_data + offset); 5652 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 5653 !(le32_to_cpu(es->s_feature_incompat) & 5654 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 5655 ext4_msg(sb, KERN_ERR, "external journal has " 5656 "bad superblock"); 5657 brelse(bh); 5658 goto out_bdev; 5659 } 5660 5661 if ((le32_to_cpu(es->s_feature_ro_compat) & 5662 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 5663 es->s_checksum != ext4_superblock_csum(sb, es)) { 5664 ext4_msg(sb, KERN_ERR, "external journal has " 5665 "corrupt superblock"); 5666 brelse(bh); 5667 goto out_bdev; 5668 } 5669 5670 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 5671 ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 5672 brelse(bh); 5673 goto out_bdev; 5674 } 5675 5676 len = ext4_blocks_count(es); 5677 start = sb_block + 1; 5678 brelse(bh); /* we're done with the superblock */ 5679 5680 journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 5681 start, len, blocksize); 5682 if (!journal) { 5683 ext4_msg(sb, KERN_ERR, "failed to create device journal"); 5684 goto out_bdev; 5685 } 5686 journal->j_private = sb; 5687 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) { 5688 ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 5689 goto out_journal; 5690 } 5691 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 5692 ext4_msg(sb, KERN_ERR, "External journal has more than one " 5693 "user (unsupported) - %d", 5694 be32_to_cpu(journal->j_superblock->s_nr_users)); 5695 goto out_journal; 5696 } 5697 EXT4_SB(sb)->s_journal_bdev = bdev; 5698 ext4_init_journal_params(sb, journal); 5699 return journal; 5700 5701 out_journal: 5702 jbd2_journal_destroy(journal); 5703 out_bdev: 5704 ext4_blkdev_put(bdev); 5705 return NULL; 5706 } 5707 5708 static int ext4_load_journal(struct super_block *sb, 5709 struct ext4_super_block *es, 5710 unsigned long journal_devnum) 5711 { 5712 journal_t *journal; 5713 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 5714 dev_t journal_dev; 5715 int err = 0; 5716 int really_read_only; 5717 int journal_dev_ro; 5718 5719 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 5720 return -EFSCORRUPTED; 5721 5722 if (journal_devnum && 5723 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 5724 ext4_msg(sb, KERN_INFO, "external journal device major/minor " 5725 "numbers have changed"); 5726 journal_dev = new_decode_dev(journal_devnum); 5727 } else 5728 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 5729 5730 if (journal_inum && journal_dev) { 5731 ext4_msg(sb, KERN_ERR, 5732 "filesystem has both journal inode and journal device!"); 5733 return -EINVAL; 5734 } 5735 5736 if (journal_inum) { 5737 journal = ext4_get_journal(sb, journal_inum); 5738 if (!journal) 5739 return -EINVAL; 5740 } else { 5741 journal = ext4_get_dev_journal(sb, journal_dev); 5742 if (!journal) 5743 return -EINVAL; 5744 } 5745 5746 journal_dev_ro = bdev_read_only(journal->j_dev); 5747 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro; 5748 5749 if (journal_dev_ro && !sb_rdonly(sb)) { 5750 ext4_msg(sb, KERN_ERR, 5751 "journal device read-only, try mounting with '-o ro'"); 5752 err = -EROFS; 5753 goto err_out; 5754 } 5755 5756 /* 5757 * Are we loading a blank journal or performing recovery after a 5758 * crash? For recovery, we need to check in advance whether we 5759 * can get read-write access to the device. 5760 */ 5761 if (ext4_has_feature_journal_needs_recovery(sb)) { 5762 if (sb_rdonly(sb)) { 5763 ext4_msg(sb, KERN_INFO, "INFO: recovery " 5764 "required on readonly filesystem"); 5765 if (really_read_only) { 5766 ext4_msg(sb, KERN_ERR, "write access " 5767 "unavailable, cannot proceed " 5768 "(try mounting with noload)"); 5769 err = -EROFS; 5770 goto err_out; 5771 } 5772 ext4_msg(sb, KERN_INFO, "write access will " 5773 "be enabled during recovery"); 5774 } 5775 } 5776 5777 if (!(journal->j_flags & JBD2_BARRIER)) 5778 ext4_msg(sb, KERN_INFO, "barriers disabled"); 5779 5780 if (!ext4_has_feature_journal_needs_recovery(sb)) 5781 err = jbd2_journal_wipe(journal, !really_read_only); 5782 if (!err) { 5783 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 5784 if (save) 5785 memcpy(save, ((char *) es) + 5786 EXT4_S_ERR_START, EXT4_S_ERR_LEN); 5787 err = jbd2_journal_load(journal); 5788 if (save) 5789 memcpy(((char *) es) + EXT4_S_ERR_START, 5790 save, EXT4_S_ERR_LEN); 5791 kfree(save); 5792 } 5793 5794 if (err) { 5795 ext4_msg(sb, KERN_ERR, "error loading journal"); 5796 goto err_out; 5797 } 5798 5799 EXT4_SB(sb)->s_journal = journal; 5800 err = ext4_clear_journal_err(sb, es); 5801 if (err) { 5802 EXT4_SB(sb)->s_journal = NULL; 5803 jbd2_journal_destroy(journal); 5804 return err; 5805 } 5806 5807 if (!really_read_only && journal_devnum && 5808 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 5809 es->s_journal_dev = cpu_to_le32(journal_devnum); 5810 5811 /* Make sure we flush the recovery flag to disk. */ 5812 ext4_commit_super(sb); 5813 } 5814 5815 return 0; 5816 5817 err_out: 5818 jbd2_journal_destroy(journal); 5819 return err; 5820 } 5821 5822 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */ 5823 static void ext4_update_super(struct super_block *sb) 5824 { 5825 struct ext4_sb_info *sbi = EXT4_SB(sb); 5826 struct ext4_super_block *es = sbi->s_es; 5827 struct buffer_head *sbh = sbi->s_sbh; 5828 5829 lock_buffer(sbh); 5830 /* 5831 * If the file system is mounted read-only, don't update the 5832 * superblock write time. This avoids updating the superblock 5833 * write time when we are mounting the root file system 5834 * read/only but we need to replay the journal; at that point, 5835 * for people who are east of GMT and who make their clock 5836 * tick in localtime for Windows bug-for-bug compatibility, 5837 * the clock is set in the future, and this will cause e2fsck 5838 * to complain and force a full file system check. 5839 */ 5840 if (!(sb->s_flags & SB_RDONLY)) 5841 ext4_update_tstamp(es, s_wtime); 5842 es->s_kbytes_written = 5843 cpu_to_le64(sbi->s_kbytes_written + 5844 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) - 5845 sbi->s_sectors_written_start) >> 1)); 5846 if (percpu_counter_initialized(&sbi->s_freeclusters_counter)) 5847 ext4_free_blocks_count_set(es, 5848 EXT4_C2B(sbi, percpu_counter_sum_positive( 5849 &sbi->s_freeclusters_counter))); 5850 if (percpu_counter_initialized(&sbi->s_freeinodes_counter)) 5851 es->s_free_inodes_count = 5852 cpu_to_le32(percpu_counter_sum_positive( 5853 &sbi->s_freeinodes_counter)); 5854 /* Copy error information to the on-disk superblock */ 5855 spin_lock(&sbi->s_error_lock); 5856 if (sbi->s_add_error_count > 0) { 5857 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 5858 if (!es->s_first_error_time && !es->s_first_error_time_hi) { 5859 __ext4_update_tstamp(&es->s_first_error_time, 5860 &es->s_first_error_time_hi, 5861 sbi->s_first_error_time); 5862 strncpy(es->s_first_error_func, sbi->s_first_error_func, 5863 sizeof(es->s_first_error_func)); 5864 es->s_first_error_line = 5865 cpu_to_le32(sbi->s_first_error_line); 5866 es->s_first_error_ino = 5867 cpu_to_le32(sbi->s_first_error_ino); 5868 es->s_first_error_block = 5869 cpu_to_le64(sbi->s_first_error_block); 5870 es->s_first_error_errcode = 5871 ext4_errno_to_code(sbi->s_first_error_code); 5872 } 5873 __ext4_update_tstamp(&es->s_last_error_time, 5874 &es->s_last_error_time_hi, 5875 sbi->s_last_error_time); 5876 strncpy(es->s_last_error_func, sbi->s_last_error_func, 5877 sizeof(es->s_last_error_func)); 5878 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line); 5879 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino); 5880 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block); 5881 es->s_last_error_errcode = 5882 ext4_errno_to_code(sbi->s_last_error_code); 5883 /* 5884 * Start the daily error reporting function if it hasn't been 5885 * started already 5886 */ 5887 if (!es->s_error_count) 5888 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); 5889 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count); 5890 sbi->s_add_error_count = 0; 5891 } 5892 spin_unlock(&sbi->s_error_lock); 5893 5894 ext4_superblock_csum_set(sb); 5895 unlock_buffer(sbh); 5896 } 5897 5898 static int ext4_commit_super(struct super_block *sb) 5899 { 5900 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 5901 int error = 0; 5902 5903 if (!sbh) 5904 return -EINVAL; 5905 if (block_device_ejected(sb)) 5906 return -ENODEV; 5907 5908 ext4_update_super(sb); 5909 5910 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 5911 /* 5912 * Oh, dear. A previous attempt to write the 5913 * superblock failed. This could happen because the 5914 * USB device was yanked out. Or it could happen to 5915 * be a transient write error and maybe the block will 5916 * be remapped. Nothing we can do but to retry the 5917 * write and hope for the best. 5918 */ 5919 ext4_msg(sb, KERN_ERR, "previous I/O error to " 5920 "superblock detected"); 5921 clear_buffer_write_io_error(sbh); 5922 set_buffer_uptodate(sbh); 5923 } 5924 BUFFER_TRACE(sbh, "marking dirty"); 5925 mark_buffer_dirty(sbh); 5926 error = __sync_dirty_buffer(sbh, 5927 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0)); 5928 if (buffer_write_io_error(sbh)) { 5929 ext4_msg(sb, KERN_ERR, "I/O error while writing " 5930 "superblock"); 5931 clear_buffer_write_io_error(sbh); 5932 set_buffer_uptodate(sbh); 5933 } 5934 return error; 5935 } 5936 5937 /* 5938 * Have we just finished recovery? If so, and if we are mounting (or 5939 * remounting) the filesystem readonly, then we will end up with a 5940 * consistent fs on disk. Record that fact. 5941 */ 5942 static int ext4_mark_recovery_complete(struct super_block *sb, 5943 struct ext4_super_block *es) 5944 { 5945 int err; 5946 journal_t *journal = EXT4_SB(sb)->s_journal; 5947 5948 if (!ext4_has_feature_journal(sb)) { 5949 if (journal != NULL) { 5950 ext4_error(sb, "Journal got removed while the fs was " 5951 "mounted!"); 5952 return -EFSCORRUPTED; 5953 } 5954 return 0; 5955 } 5956 jbd2_journal_lock_updates(journal); 5957 err = jbd2_journal_flush(journal, 0); 5958 if (err < 0) 5959 goto out; 5960 5961 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) || 5962 ext4_has_feature_orphan_present(sb))) { 5963 if (!ext4_orphan_file_empty(sb)) { 5964 ext4_error(sb, "Orphan file not empty on read-only fs."); 5965 err = -EFSCORRUPTED; 5966 goto out; 5967 } 5968 ext4_clear_feature_journal_needs_recovery(sb); 5969 ext4_clear_feature_orphan_present(sb); 5970 ext4_commit_super(sb); 5971 } 5972 out: 5973 jbd2_journal_unlock_updates(journal); 5974 return err; 5975 } 5976 5977 /* 5978 * If we are mounting (or read-write remounting) a filesystem whose journal 5979 * has recorded an error from a previous lifetime, move that error to the 5980 * main filesystem now. 5981 */ 5982 static int ext4_clear_journal_err(struct super_block *sb, 5983 struct ext4_super_block *es) 5984 { 5985 journal_t *journal; 5986 int j_errno; 5987 const char *errstr; 5988 5989 if (!ext4_has_feature_journal(sb)) { 5990 ext4_error(sb, "Journal got removed while the fs was mounted!"); 5991 return -EFSCORRUPTED; 5992 } 5993 5994 journal = EXT4_SB(sb)->s_journal; 5995 5996 /* 5997 * Now check for any error status which may have been recorded in the 5998 * journal by a prior ext4_error() or ext4_abort() 5999 */ 6000 6001 j_errno = jbd2_journal_errno(journal); 6002 if (j_errno) { 6003 char nbuf[16]; 6004 6005 errstr = ext4_decode_error(sb, j_errno, nbuf); 6006 ext4_warning(sb, "Filesystem error recorded " 6007 "from previous mount: %s", errstr); 6008 ext4_warning(sb, "Marking fs in need of filesystem check."); 6009 6010 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 6011 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 6012 ext4_commit_super(sb); 6013 6014 jbd2_journal_clear_err(journal); 6015 jbd2_journal_update_sb_errno(journal); 6016 } 6017 return 0; 6018 } 6019 6020 /* 6021 * Force the running and committing transactions to commit, 6022 * and wait on the commit. 6023 */ 6024 int ext4_force_commit(struct super_block *sb) 6025 { 6026 journal_t *journal; 6027 6028 if (sb_rdonly(sb)) 6029 return 0; 6030 6031 journal = EXT4_SB(sb)->s_journal; 6032 return ext4_journal_force_commit(journal); 6033 } 6034 6035 static int ext4_sync_fs(struct super_block *sb, int wait) 6036 { 6037 int ret = 0; 6038 tid_t target; 6039 bool needs_barrier = false; 6040 struct ext4_sb_info *sbi = EXT4_SB(sb); 6041 6042 if (unlikely(ext4_forced_shutdown(sbi))) 6043 return 0; 6044 6045 trace_ext4_sync_fs(sb, wait); 6046 flush_workqueue(sbi->rsv_conversion_wq); 6047 /* 6048 * Writeback quota in non-journalled quota case - journalled quota has 6049 * no dirty dquots 6050 */ 6051 dquot_writeback_dquots(sb, -1); 6052 /* 6053 * Data writeback is possible w/o journal transaction, so barrier must 6054 * being sent at the end of the function. But we can skip it if 6055 * transaction_commit will do it for us. 6056 */ 6057 if (sbi->s_journal) { 6058 target = jbd2_get_latest_transaction(sbi->s_journal); 6059 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 6060 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 6061 needs_barrier = true; 6062 6063 if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 6064 if (wait) 6065 ret = jbd2_log_wait_commit(sbi->s_journal, 6066 target); 6067 } 6068 } else if (wait && test_opt(sb, BARRIER)) 6069 needs_barrier = true; 6070 if (needs_barrier) { 6071 int err; 6072 err = blkdev_issue_flush(sb->s_bdev); 6073 if (!ret) 6074 ret = err; 6075 } 6076 6077 return ret; 6078 } 6079 6080 /* 6081 * LVM calls this function before a (read-only) snapshot is created. This 6082 * gives us a chance to flush the journal completely and mark the fs clean. 6083 * 6084 * Note that only this function cannot bring a filesystem to be in a clean 6085 * state independently. It relies on upper layer to stop all data & metadata 6086 * modifications. 6087 */ 6088 static int ext4_freeze(struct super_block *sb) 6089 { 6090 int error = 0; 6091 journal_t *journal; 6092 6093 if (sb_rdonly(sb)) 6094 return 0; 6095 6096 journal = EXT4_SB(sb)->s_journal; 6097 6098 if (journal) { 6099 /* Now we set up the journal barrier. */ 6100 jbd2_journal_lock_updates(journal); 6101 6102 /* 6103 * Don't clear the needs_recovery flag if we failed to 6104 * flush the journal. 6105 */ 6106 error = jbd2_journal_flush(journal, 0); 6107 if (error < 0) 6108 goto out; 6109 6110 /* Journal blocked and flushed, clear needs_recovery flag. */ 6111 ext4_clear_feature_journal_needs_recovery(sb); 6112 if (ext4_orphan_file_empty(sb)) 6113 ext4_clear_feature_orphan_present(sb); 6114 } 6115 6116 error = ext4_commit_super(sb); 6117 out: 6118 if (journal) 6119 /* we rely on upper layer to stop further updates */ 6120 jbd2_journal_unlock_updates(journal); 6121 return error; 6122 } 6123 6124 /* 6125 * Called by LVM after the snapshot is done. We need to reset the RECOVER 6126 * flag here, even though the filesystem is not technically dirty yet. 6127 */ 6128 static int ext4_unfreeze(struct super_block *sb) 6129 { 6130 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb))) 6131 return 0; 6132 6133 if (EXT4_SB(sb)->s_journal) { 6134 /* Reset the needs_recovery flag before the fs is unlocked. */ 6135 ext4_set_feature_journal_needs_recovery(sb); 6136 if (ext4_has_feature_orphan_file(sb)) 6137 ext4_set_feature_orphan_present(sb); 6138 } 6139 6140 ext4_commit_super(sb); 6141 return 0; 6142 } 6143 6144 /* 6145 * Structure to save mount options for ext4_remount's benefit 6146 */ 6147 struct ext4_mount_options { 6148 unsigned long s_mount_opt; 6149 unsigned long s_mount_opt2; 6150 kuid_t s_resuid; 6151 kgid_t s_resgid; 6152 unsigned long s_commit_interval; 6153 u32 s_min_batch_time, s_max_batch_time; 6154 #ifdef CONFIG_QUOTA 6155 int s_jquota_fmt; 6156 char *s_qf_names[EXT4_MAXQUOTAS]; 6157 #endif 6158 }; 6159 6160 static int __ext4_remount(struct fs_context *fc, struct super_block *sb) 6161 { 6162 struct ext4_fs_context *ctx = fc->fs_private; 6163 struct ext4_super_block *es; 6164 struct ext4_sb_info *sbi = EXT4_SB(sb); 6165 unsigned long old_sb_flags; 6166 struct ext4_mount_options old_opts; 6167 ext4_group_t g; 6168 int err = 0; 6169 #ifdef CONFIG_QUOTA 6170 int enable_quota = 0; 6171 int i, j; 6172 char *to_free[EXT4_MAXQUOTAS]; 6173 #endif 6174 6175 6176 /* Store the original options */ 6177 old_sb_flags = sb->s_flags; 6178 old_opts.s_mount_opt = sbi->s_mount_opt; 6179 old_opts.s_mount_opt2 = sbi->s_mount_opt2; 6180 old_opts.s_resuid = sbi->s_resuid; 6181 old_opts.s_resgid = sbi->s_resgid; 6182 old_opts.s_commit_interval = sbi->s_commit_interval; 6183 old_opts.s_min_batch_time = sbi->s_min_batch_time; 6184 old_opts.s_max_batch_time = sbi->s_max_batch_time; 6185 #ifdef CONFIG_QUOTA 6186 old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 6187 for (i = 0; i < EXT4_MAXQUOTAS; i++) 6188 if (sbi->s_qf_names[i]) { 6189 char *qf_name = get_qf_name(sb, sbi, i); 6190 6191 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); 6192 if (!old_opts.s_qf_names[i]) { 6193 for (j = 0; j < i; j++) 6194 kfree(old_opts.s_qf_names[j]); 6195 return -ENOMEM; 6196 } 6197 } else 6198 old_opts.s_qf_names[i] = NULL; 6199 #endif 6200 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) { 6201 if (sbi->s_journal && sbi->s_journal->j_task->io_context) 6202 ctx->journal_ioprio = 6203 sbi->s_journal->j_task->io_context->ioprio; 6204 else 6205 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 6206 6207 } 6208 6209 ext4_apply_options(fc, sb); 6210 6211 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 6212 test_opt(sb, JOURNAL_CHECKSUM)) { 6213 ext4_msg(sb, KERN_ERR, "changing journal_checksum " 6214 "during remount not supported; ignoring"); 6215 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 6216 } 6217 6218 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 6219 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 6220 ext4_msg(sb, KERN_ERR, "can't mount with " 6221 "both data=journal and delalloc"); 6222 err = -EINVAL; 6223 goto restore_opts; 6224 } 6225 if (test_opt(sb, DIOREAD_NOLOCK)) { 6226 ext4_msg(sb, KERN_ERR, "can't mount with " 6227 "both data=journal and dioread_nolock"); 6228 err = -EINVAL; 6229 goto restore_opts; 6230 } 6231 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { 6232 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 6233 ext4_msg(sb, KERN_ERR, "can't mount with " 6234 "journal_async_commit in data=ordered mode"); 6235 err = -EINVAL; 6236 goto restore_opts; 6237 } 6238 } 6239 6240 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { 6241 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); 6242 err = -EINVAL; 6243 goto restore_opts; 6244 } 6245 6246 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) 6247 ext4_abort(sb, ESHUTDOWN, "Abort forced by user"); 6248 6249 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 6250 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 6251 6252 es = sbi->s_es; 6253 6254 if (sbi->s_journal) { 6255 ext4_init_journal_params(sb, sbi->s_journal); 6256 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio); 6257 } 6258 6259 /* Flush outstanding errors before changing fs state */ 6260 flush_work(&sbi->s_error_work); 6261 6262 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) { 6263 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) { 6264 err = -EROFS; 6265 goto restore_opts; 6266 } 6267 6268 if (fc->sb_flags & SB_RDONLY) { 6269 err = sync_filesystem(sb); 6270 if (err < 0) 6271 goto restore_opts; 6272 err = dquot_suspend(sb, -1); 6273 if (err < 0) 6274 goto restore_opts; 6275 6276 /* 6277 * First of all, the unconditional stuff we have to do 6278 * to disable replay of the journal when we next remount 6279 */ 6280 sb->s_flags |= SB_RDONLY; 6281 6282 /* 6283 * OK, test if we are remounting a valid rw partition 6284 * readonly, and if so set the rdonly flag and then 6285 * mark the partition as valid again. 6286 */ 6287 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 6288 (sbi->s_mount_state & EXT4_VALID_FS)) 6289 es->s_state = cpu_to_le16(sbi->s_mount_state); 6290 6291 if (sbi->s_journal) { 6292 /* 6293 * We let remount-ro finish even if marking fs 6294 * as clean failed... 6295 */ 6296 ext4_mark_recovery_complete(sb, es); 6297 } 6298 } else { 6299 /* Make sure we can mount this feature set readwrite */ 6300 if (ext4_has_feature_readonly(sb) || 6301 !ext4_feature_set_ok(sb, 0)) { 6302 err = -EROFS; 6303 goto restore_opts; 6304 } 6305 /* 6306 * Make sure the group descriptor checksums 6307 * are sane. If they aren't, refuse to remount r/w. 6308 */ 6309 for (g = 0; g < sbi->s_groups_count; g++) { 6310 struct ext4_group_desc *gdp = 6311 ext4_get_group_desc(sb, g, NULL); 6312 6313 if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 6314 ext4_msg(sb, KERN_ERR, 6315 "ext4_remount: Checksum for group %u failed (%u!=%u)", 6316 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 6317 le16_to_cpu(gdp->bg_checksum)); 6318 err = -EFSBADCRC; 6319 goto restore_opts; 6320 } 6321 } 6322 6323 /* 6324 * If we have an unprocessed orphan list hanging 6325 * around from a previously readonly bdev mount, 6326 * require a full umount/remount for now. 6327 */ 6328 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) { 6329 ext4_msg(sb, KERN_WARNING, "Couldn't " 6330 "remount RDWR because of unprocessed " 6331 "orphan inode list. Please " 6332 "umount/remount instead"); 6333 err = -EINVAL; 6334 goto restore_opts; 6335 } 6336 6337 /* 6338 * Mounting a RDONLY partition read-write, so reread 6339 * and store the current valid flag. (It may have 6340 * been changed by e2fsck since we originally mounted 6341 * the partition.) 6342 */ 6343 if (sbi->s_journal) { 6344 err = ext4_clear_journal_err(sb, es); 6345 if (err) 6346 goto restore_opts; 6347 } 6348 sbi->s_mount_state = (le16_to_cpu(es->s_state) & 6349 ~EXT4_FC_REPLAY); 6350 6351 err = ext4_setup_super(sb, es, 0); 6352 if (err) 6353 goto restore_opts; 6354 6355 sb->s_flags &= ~SB_RDONLY; 6356 if (ext4_has_feature_mmp(sb)) 6357 if (ext4_multi_mount_protect(sb, 6358 le64_to_cpu(es->s_mmp_block))) { 6359 err = -EROFS; 6360 goto restore_opts; 6361 } 6362 #ifdef CONFIG_QUOTA 6363 enable_quota = 1; 6364 #endif 6365 } 6366 } 6367 6368 /* 6369 * Reinitialize lazy itable initialization thread based on 6370 * current settings 6371 */ 6372 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) 6373 ext4_unregister_li_request(sb); 6374 else { 6375 ext4_group_t first_not_zeroed; 6376 first_not_zeroed = ext4_has_uninit_itable(sb); 6377 ext4_register_li_request(sb, first_not_zeroed); 6378 } 6379 6380 /* 6381 * Handle creation of system zone data early because it can fail. 6382 * Releasing of existing data is done when we are sure remount will 6383 * succeed. 6384 */ 6385 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) { 6386 err = ext4_setup_system_zone(sb); 6387 if (err) 6388 goto restore_opts; 6389 } 6390 6391 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) { 6392 err = ext4_commit_super(sb); 6393 if (err) 6394 goto restore_opts; 6395 } 6396 6397 #ifdef CONFIG_QUOTA 6398 /* Release old quota file names */ 6399 for (i = 0; i < EXT4_MAXQUOTAS; i++) 6400 kfree(old_opts.s_qf_names[i]); 6401 if (enable_quota) { 6402 if (sb_any_quota_suspended(sb)) 6403 dquot_resume(sb, -1); 6404 else if (ext4_has_feature_quota(sb)) { 6405 err = ext4_enable_quotas(sb); 6406 if (err) 6407 goto restore_opts; 6408 } 6409 } 6410 #endif 6411 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 6412 ext4_release_system_zone(sb); 6413 6414 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb)) 6415 ext4_stop_mmpd(sbi); 6416 6417 return 0; 6418 6419 restore_opts: 6420 sb->s_flags = old_sb_flags; 6421 sbi->s_mount_opt = old_opts.s_mount_opt; 6422 sbi->s_mount_opt2 = old_opts.s_mount_opt2; 6423 sbi->s_resuid = old_opts.s_resuid; 6424 sbi->s_resgid = old_opts.s_resgid; 6425 sbi->s_commit_interval = old_opts.s_commit_interval; 6426 sbi->s_min_batch_time = old_opts.s_min_batch_time; 6427 sbi->s_max_batch_time = old_opts.s_max_batch_time; 6428 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 6429 ext4_release_system_zone(sb); 6430 #ifdef CONFIG_QUOTA 6431 sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 6432 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 6433 to_free[i] = get_qf_name(sb, sbi, i); 6434 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); 6435 } 6436 synchronize_rcu(); 6437 for (i = 0; i < EXT4_MAXQUOTAS; i++) 6438 kfree(to_free[i]); 6439 #endif 6440 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb)) 6441 ext4_stop_mmpd(sbi); 6442 return err; 6443 } 6444 6445 static int ext4_reconfigure(struct fs_context *fc) 6446 { 6447 struct super_block *sb = fc->root->d_sb; 6448 int ret; 6449 6450 fc->s_fs_info = EXT4_SB(sb); 6451 6452 ret = ext4_check_opt_consistency(fc, sb); 6453 if (ret < 0) 6454 return ret; 6455 6456 ret = __ext4_remount(fc, sb); 6457 if (ret < 0) 6458 return ret; 6459 6460 ext4_msg(sb, KERN_INFO, "re-mounted. Quota mode: %s.", 6461 ext4_quota_mode(sb)); 6462 6463 return 0; 6464 } 6465 6466 #ifdef CONFIG_QUOTA 6467 static int ext4_statfs_project(struct super_block *sb, 6468 kprojid_t projid, struct kstatfs *buf) 6469 { 6470 struct kqid qid; 6471 struct dquot *dquot; 6472 u64 limit; 6473 u64 curblock; 6474 6475 qid = make_kqid_projid(projid); 6476 dquot = dqget(sb, qid); 6477 if (IS_ERR(dquot)) 6478 return PTR_ERR(dquot); 6479 spin_lock(&dquot->dq_dqb_lock); 6480 6481 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 6482 dquot->dq_dqb.dqb_bhardlimit); 6483 limit >>= sb->s_blocksize_bits; 6484 6485 if (limit && buf->f_blocks > limit) { 6486 curblock = (dquot->dq_dqb.dqb_curspace + 6487 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 6488 buf->f_blocks = limit; 6489 buf->f_bfree = buf->f_bavail = 6490 (buf->f_blocks > curblock) ? 6491 (buf->f_blocks - curblock) : 0; 6492 } 6493 6494 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 6495 dquot->dq_dqb.dqb_ihardlimit); 6496 if (limit && buf->f_files > limit) { 6497 buf->f_files = limit; 6498 buf->f_ffree = 6499 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 6500 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 6501 } 6502 6503 spin_unlock(&dquot->dq_dqb_lock); 6504 dqput(dquot); 6505 return 0; 6506 } 6507 #endif 6508 6509 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 6510 { 6511 struct super_block *sb = dentry->d_sb; 6512 struct ext4_sb_info *sbi = EXT4_SB(sb); 6513 struct ext4_super_block *es = sbi->s_es; 6514 ext4_fsblk_t overhead = 0, resv_blocks; 6515 s64 bfree; 6516 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 6517 6518 if (!test_opt(sb, MINIX_DF)) 6519 overhead = sbi->s_overhead; 6520 6521 buf->f_type = EXT4_SUPER_MAGIC; 6522 buf->f_bsize = sb->s_blocksize; 6523 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 6524 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 6525 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 6526 /* prevent underflow in case that few free space is available */ 6527 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 6528 buf->f_bavail = buf->f_bfree - 6529 (ext4_r_blocks_count(es) + resv_blocks); 6530 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 6531 buf->f_bavail = 0; 6532 buf->f_files = le32_to_cpu(es->s_inodes_count); 6533 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 6534 buf->f_namelen = EXT4_NAME_LEN; 6535 buf->f_fsid = uuid_to_fsid(es->s_uuid); 6536 6537 #ifdef CONFIG_QUOTA 6538 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 6539 sb_has_quota_limits_enabled(sb, PRJQUOTA)) 6540 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 6541 #endif 6542 return 0; 6543 } 6544 6545 6546 #ifdef CONFIG_QUOTA 6547 6548 /* 6549 * Helper functions so that transaction is started before we acquire dqio_sem 6550 * to keep correct lock ordering of transaction > dqio_sem 6551 */ 6552 static inline struct inode *dquot_to_inode(struct dquot *dquot) 6553 { 6554 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 6555 } 6556 6557 static int ext4_write_dquot(struct dquot *dquot) 6558 { 6559 int ret, err; 6560 handle_t *handle; 6561 struct inode *inode; 6562 6563 inode = dquot_to_inode(dquot); 6564 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 6565 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 6566 if (IS_ERR(handle)) 6567 return PTR_ERR(handle); 6568 ret = dquot_commit(dquot); 6569 err = ext4_journal_stop(handle); 6570 if (!ret) 6571 ret = err; 6572 return ret; 6573 } 6574 6575 static int ext4_acquire_dquot(struct dquot *dquot) 6576 { 6577 int ret, err; 6578 handle_t *handle; 6579 6580 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 6581 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 6582 if (IS_ERR(handle)) 6583 return PTR_ERR(handle); 6584 ret = dquot_acquire(dquot); 6585 err = ext4_journal_stop(handle); 6586 if (!ret) 6587 ret = err; 6588 return ret; 6589 } 6590 6591 static int ext4_release_dquot(struct dquot *dquot) 6592 { 6593 int ret, err; 6594 handle_t *handle; 6595 6596 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 6597 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 6598 if (IS_ERR(handle)) { 6599 /* Release dquot anyway to avoid endless cycle in dqput() */ 6600 dquot_release(dquot); 6601 return PTR_ERR(handle); 6602 } 6603 ret = dquot_release(dquot); 6604 err = ext4_journal_stop(handle); 6605 if (!ret) 6606 ret = err; 6607 return ret; 6608 } 6609 6610 static int ext4_mark_dquot_dirty(struct dquot *dquot) 6611 { 6612 struct super_block *sb = dquot->dq_sb; 6613 6614 if (ext4_is_quota_journalled(sb)) { 6615 dquot_mark_dquot_dirty(dquot); 6616 return ext4_write_dquot(dquot); 6617 } else { 6618 return dquot_mark_dquot_dirty(dquot); 6619 } 6620 } 6621 6622 static int ext4_write_info(struct super_block *sb, int type) 6623 { 6624 int ret, err; 6625 handle_t *handle; 6626 6627 /* Data block + inode block */ 6628 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2); 6629 if (IS_ERR(handle)) 6630 return PTR_ERR(handle); 6631 ret = dquot_commit_info(sb, type); 6632 err = ext4_journal_stop(handle); 6633 if (!ret) 6634 ret = err; 6635 return ret; 6636 } 6637 6638 static void lockdep_set_quota_inode(struct inode *inode, int subclass) 6639 { 6640 struct ext4_inode_info *ei = EXT4_I(inode); 6641 6642 /* The first argument of lockdep_set_subclass has to be 6643 * *exactly* the same as the argument to init_rwsem() --- in 6644 * this case, in init_once() --- or lockdep gets unhappy 6645 * because the name of the lock is set using the 6646 * stringification of the argument to init_rwsem(). 6647 */ 6648 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 6649 lockdep_set_subclass(&ei->i_data_sem, subclass); 6650 } 6651 6652 /* 6653 * Standard function to be called on quota_on 6654 */ 6655 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 6656 const struct path *path) 6657 { 6658 int err; 6659 6660 if (!test_opt(sb, QUOTA)) 6661 return -EINVAL; 6662 6663 /* Quotafile not on the same filesystem? */ 6664 if (path->dentry->d_sb != sb) 6665 return -EXDEV; 6666 6667 /* Quota already enabled for this file? */ 6668 if (IS_NOQUOTA(d_inode(path->dentry))) 6669 return -EBUSY; 6670 6671 /* Journaling quota? */ 6672 if (EXT4_SB(sb)->s_qf_names[type]) { 6673 /* Quotafile not in fs root? */ 6674 if (path->dentry->d_parent != sb->s_root) 6675 ext4_msg(sb, KERN_WARNING, 6676 "Quota file not on filesystem root. " 6677 "Journaled quota will not work"); 6678 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY; 6679 } else { 6680 /* 6681 * Clear the flag just in case mount options changed since 6682 * last time. 6683 */ 6684 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY; 6685 } 6686 6687 /* 6688 * When we journal data on quota file, we have to flush journal to see 6689 * all updates to the file when we bypass pagecache... 6690 */ 6691 if (EXT4_SB(sb)->s_journal && 6692 ext4_should_journal_data(d_inode(path->dentry))) { 6693 /* 6694 * We don't need to lock updates but journal_flush() could 6695 * otherwise be livelocked... 6696 */ 6697 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 6698 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0); 6699 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 6700 if (err) 6701 return err; 6702 } 6703 6704 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 6705 err = dquot_quota_on(sb, type, format_id, path); 6706 if (!err) { 6707 struct inode *inode = d_inode(path->dentry); 6708 handle_t *handle; 6709 6710 /* 6711 * Set inode flags to prevent userspace from messing with quota 6712 * files. If this fails, we return success anyway since quotas 6713 * are already enabled and this is not a hard failure. 6714 */ 6715 inode_lock(inode); 6716 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 6717 if (IS_ERR(handle)) 6718 goto unlock_inode; 6719 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; 6720 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 6721 S_NOATIME | S_IMMUTABLE); 6722 err = ext4_mark_inode_dirty(handle, inode); 6723 ext4_journal_stop(handle); 6724 unlock_inode: 6725 inode_unlock(inode); 6726 if (err) 6727 dquot_quota_off(sb, type); 6728 } 6729 if (err) 6730 lockdep_set_quota_inode(path->dentry->d_inode, 6731 I_DATA_SEM_NORMAL); 6732 return err; 6733 } 6734 6735 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 6736 unsigned int flags) 6737 { 6738 int err; 6739 struct inode *qf_inode; 6740 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 6741 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 6742 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 6743 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 6744 }; 6745 6746 BUG_ON(!ext4_has_feature_quota(sb)); 6747 6748 if (!qf_inums[type]) 6749 return -EPERM; 6750 6751 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL); 6752 if (IS_ERR(qf_inode)) { 6753 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 6754 return PTR_ERR(qf_inode); 6755 } 6756 6757 /* Don't account quota for quota files to avoid recursion */ 6758 qf_inode->i_flags |= S_NOQUOTA; 6759 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 6760 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 6761 if (err) 6762 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 6763 iput(qf_inode); 6764 6765 return err; 6766 } 6767 6768 /* Enable usage tracking for all quota types. */ 6769 int ext4_enable_quotas(struct super_block *sb) 6770 { 6771 int type, err = 0; 6772 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 6773 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 6774 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 6775 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 6776 }; 6777 bool quota_mopt[EXT4_MAXQUOTAS] = { 6778 test_opt(sb, USRQUOTA), 6779 test_opt(sb, GRPQUOTA), 6780 test_opt(sb, PRJQUOTA), 6781 }; 6782 6783 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 6784 for (type = 0; type < EXT4_MAXQUOTAS; type++) { 6785 if (qf_inums[type]) { 6786 err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 6787 DQUOT_USAGE_ENABLED | 6788 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 6789 if (err) { 6790 ext4_warning(sb, 6791 "Failed to enable quota tracking " 6792 "(type=%d, err=%d). Please run " 6793 "e2fsck to fix.", type, err); 6794 for (type--; type >= 0; type--) { 6795 struct inode *inode; 6796 6797 inode = sb_dqopt(sb)->files[type]; 6798 if (inode) 6799 inode = igrab(inode); 6800 dquot_quota_off(sb, type); 6801 if (inode) { 6802 lockdep_set_quota_inode(inode, 6803 I_DATA_SEM_NORMAL); 6804 iput(inode); 6805 } 6806 } 6807 6808 return err; 6809 } 6810 } 6811 } 6812 return 0; 6813 } 6814 6815 static int ext4_quota_off(struct super_block *sb, int type) 6816 { 6817 struct inode *inode = sb_dqopt(sb)->files[type]; 6818 handle_t *handle; 6819 int err; 6820 6821 /* Force all delayed allocation blocks to be allocated. 6822 * Caller already holds s_umount sem */ 6823 if (test_opt(sb, DELALLOC)) 6824 sync_filesystem(sb); 6825 6826 if (!inode || !igrab(inode)) 6827 goto out; 6828 6829 err = dquot_quota_off(sb, type); 6830 if (err || ext4_has_feature_quota(sb)) 6831 goto out_put; 6832 6833 inode_lock(inode); 6834 /* 6835 * Update modification times of quota files when userspace can 6836 * start looking at them. If we fail, we return success anyway since 6837 * this is not a hard failure and quotas are already disabled. 6838 */ 6839 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 6840 if (IS_ERR(handle)) { 6841 err = PTR_ERR(handle); 6842 goto out_unlock; 6843 } 6844 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); 6845 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 6846 inode->i_mtime = inode->i_ctime = current_time(inode); 6847 err = ext4_mark_inode_dirty(handle, inode); 6848 ext4_journal_stop(handle); 6849 out_unlock: 6850 inode_unlock(inode); 6851 out_put: 6852 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); 6853 iput(inode); 6854 return err; 6855 out: 6856 return dquot_quota_off(sb, type); 6857 } 6858 6859 /* Read data from quotafile - avoid pagecache and such because we cannot afford 6860 * acquiring the locks... As quota files are never truncated and quota code 6861 * itself serializes the operations (and no one else should touch the files) 6862 * we don't have to be afraid of races */ 6863 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 6864 size_t len, loff_t off) 6865 { 6866 struct inode *inode = sb_dqopt(sb)->files[type]; 6867 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 6868 int offset = off & (sb->s_blocksize - 1); 6869 int tocopy; 6870 size_t toread; 6871 struct buffer_head *bh; 6872 loff_t i_size = i_size_read(inode); 6873 6874 if (off > i_size) 6875 return 0; 6876 if (off+len > i_size) 6877 len = i_size-off; 6878 toread = len; 6879 while (toread > 0) { 6880 tocopy = sb->s_blocksize - offset < toread ? 6881 sb->s_blocksize - offset : toread; 6882 bh = ext4_bread(NULL, inode, blk, 0); 6883 if (IS_ERR(bh)) 6884 return PTR_ERR(bh); 6885 if (!bh) /* A hole? */ 6886 memset(data, 0, tocopy); 6887 else 6888 memcpy(data, bh->b_data+offset, tocopy); 6889 brelse(bh); 6890 offset = 0; 6891 toread -= tocopy; 6892 data += tocopy; 6893 blk++; 6894 } 6895 return len; 6896 } 6897 6898 /* Write to quotafile (we know the transaction is already started and has 6899 * enough credits) */ 6900 static ssize_t ext4_quota_write(struct super_block *sb, int type, 6901 const char *data, size_t len, loff_t off) 6902 { 6903 struct inode *inode = sb_dqopt(sb)->files[type]; 6904 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 6905 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1); 6906 int retries = 0; 6907 struct buffer_head *bh; 6908 handle_t *handle = journal_current_handle(); 6909 6910 if (!handle) { 6911 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 6912 " cancelled because transaction is not started", 6913 (unsigned long long)off, (unsigned long long)len); 6914 return -EIO; 6915 } 6916 /* 6917 * Since we account only one data block in transaction credits, 6918 * then it is impossible to cross a block boundary. 6919 */ 6920 if (sb->s_blocksize - offset < len) { 6921 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 6922 " cancelled because not block aligned", 6923 (unsigned long long)off, (unsigned long long)len); 6924 return -EIO; 6925 } 6926 6927 do { 6928 bh = ext4_bread(handle, inode, blk, 6929 EXT4_GET_BLOCKS_CREATE | 6930 EXT4_GET_BLOCKS_METADATA_NOFAIL); 6931 } while (PTR_ERR(bh) == -ENOSPC && 6932 ext4_should_retry_alloc(inode->i_sb, &retries)); 6933 if (IS_ERR(bh)) 6934 return PTR_ERR(bh); 6935 if (!bh) 6936 goto out; 6937 BUFFER_TRACE(bh, "get write access"); 6938 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE); 6939 if (err) { 6940 brelse(bh); 6941 return err; 6942 } 6943 lock_buffer(bh); 6944 memcpy(bh->b_data+offset, data, len); 6945 flush_dcache_page(bh->b_page); 6946 unlock_buffer(bh); 6947 err = ext4_handle_dirty_metadata(handle, NULL, bh); 6948 brelse(bh); 6949 out: 6950 if (inode->i_size < off + len) { 6951 i_size_write(inode, off + len); 6952 EXT4_I(inode)->i_disksize = inode->i_size; 6953 err2 = ext4_mark_inode_dirty(handle, inode); 6954 if (unlikely(err2 && !err)) 6955 err = err2; 6956 } 6957 return err ? err : len; 6958 } 6959 #endif 6960 6961 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 6962 static inline void register_as_ext2(void) 6963 { 6964 int err = register_filesystem(&ext2_fs_type); 6965 if (err) 6966 printk(KERN_WARNING 6967 "EXT4-fs: Unable to register as ext2 (%d)\n", err); 6968 } 6969 6970 static inline void unregister_as_ext2(void) 6971 { 6972 unregister_filesystem(&ext2_fs_type); 6973 } 6974 6975 static inline int ext2_feature_set_ok(struct super_block *sb) 6976 { 6977 if (ext4_has_unknown_ext2_incompat_features(sb)) 6978 return 0; 6979 if (sb_rdonly(sb)) 6980 return 1; 6981 if (ext4_has_unknown_ext2_ro_compat_features(sb)) 6982 return 0; 6983 return 1; 6984 } 6985 #else 6986 static inline void register_as_ext2(void) { } 6987 static inline void unregister_as_ext2(void) { } 6988 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 6989 #endif 6990 6991 static inline void register_as_ext3(void) 6992 { 6993 int err = register_filesystem(&ext3_fs_type); 6994 if (err) 6995 printk(KERN_WARNING 6996 "EXT4-fs: Unable to register as ext3 (%d)\n", err); 6997 } 6998 6999 static inline void unregister_as_ext3(void) 7000 { 7001 unregister_filesystem(&ext3_fs_type); 7002 } 7003 7004 static inline int ext3_feature_set_ok(struct super_block *sb) 7005 { 7006 if (ext4_has_unknown_ext3_incompat_features(sb)) 7007 return 0; 7008 if (!ext4_has_feature_journal(sb)) 7009 return 0; 7010 if (sb_rdonly(sb)) 7011 return 1; 7012 if (ext4_has_unknown_ext3_ro_compat_features(sb)) 7013 return 0; 7014 return 1; 7015 } 7016 7017 static struct file_system_type ext4_fs_type = { 7018 .owner = THIS_MODULE, 7019 .name = "ext4", 7020 .init_fs_context = ext4_init_fs_context, 7021 .parameters = ext4_param_specs, 7022 .kill_sb = kill_block_super, 7023 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP, 7024 }; 7025 MODULE_ALIAS_FS("ext4"); 7026 7027 /* Shared across all ext4 file systems */ 7028 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 7029 7030 static int __init ext4_init_fs(void) 7031 { 7032 int i, err; 7033 7034 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 7035 ext4_li_info = NULL; 7036 7037 /* Build-time check for flags consistency */ 7038 ext4_check_flag_values(); 7039 7040 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) 7041 init_waitqueue_head(&ext4__ioend_wq[i]); 7042 7043 err = ext4_init_es(); 7044 if (err) 7045 return err; 7046 7047 err = ext4_init_pending(); 7048 if (err) 7049 goto out7; 7050 7051 err = ext4_init_post_read_processing(); 7052 if (err) 7053 goto out6; 7054 7055 err = ext4_init_pageio(); 7056 if (err) 7057 goto out5; 7058 7059 err = ext4_init_system_zone(); 7060 if (err) 7061 goto out4; 7062 7063 err = ext4_init_sysfs(); 7064 if (err) 7065 goto out3; 7066 7067 err = ext4_init_mballoc(); 7068 if (err) 7069 goto out2; 7070 err = init_inodecache(); 7071 if (err) 7072 goto out1; 7073 7074 err = ext4_fc_init_dentry_cache(); 7075 if (err) 7076 goto out05; 7077 7078 register_as_ext3(); 7079 register_as_ext2(); 7080 err = register_filesystem(&ext4_fs_type); 7081 if (err) 7082 goto out; 7083 7084 return 0; 7085 out: 7086 unregister_as_ext2(); 7087 unregister_as_ext3(); 7088 ext4_fc_destroy_dentry_cache(); 7089 out05: 7090 destroy_inodecache(); 7091 out1: 7092 ext4_exit_mballoc(); 7093 out2: 7094 ext4_exit_sysfs(); 7095 out3: 7096 ext4_exit_system_zone(); 7097 out4: 7098 ext4_exit_pageio(); 7099 out5: 7100 ext4_exit_post_read_processing(); 7101 out6: 7102 ext4_exit_pending(); 7103 out7: 7104 ext4_exit_es(); 7105 7106 return err; 7107 } 7108 7109 static void __exit ext4_exit_fs(void) 7110 { 7111 ext4_destroy_lazyinit_thread(); 7112 unregister_as_ext2(); 7113 unregister_as_ext3(); 7114 unregister_filesystem(&ext4_fs_type); 7115 ext4_fc_destroy_dentry_cache(); 7116 destroy_inodecache(); 7117 ext4_exit_mballoc(); 7118 ext4_exit_sysfs(); 7119 ext4_exit_system_zone(); 7120 ext4_exit_pageio(); 7121 ext4_exit_post_read_processing(); 7122 ext4_exit_es(); 7123 ext4_exit_pending(); 7124 } 7125 7126 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 7127 MODULE_DESCRIPTION("Fourth Extended Filesystem"); 7128 MODULE_LICENSE("GPL"); 7129 MODULE_SOFTDEP("pre: crc32c"); 7130 module_init(ext4_init_fs) 7131 module_exit(ext4_exit_fs) 7132