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 #include <linux/fserror.h> 52 53 #include "ext4.h" 54 #include "ext4_extents.h" /* Needed for trace points definition */ 55 #include "ext4_jbd2.h" 56 #include "xattr.h" 57 #include "acl.h" 58 #include "mballoc.h" 59 #include "fsmap.h" 60 61 #define CREATE_TRACE_POINTS 62 #include <trace/events/ext4.h> 63 64 static struct ext4_lazy_init *ext4_li_info; 65 static DEFINE_MUTEX(ext4_li_mtx); 66 static struct ratelimit_state ext4_mount_msg_ratelimit; 67 68 static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 69 unsigned long journal_devnum); 70 static int ext4_show_options(struct seq_file *seq, struct dentry *root); 71 static void ext4_update_super(struct super_block *sb); 72 static int ext4_commit_super(struct super_block *sb); 73 static int ext4_mark_recovery_complete(struct super_block *sb, 74 struct ext4_super_block *es); 75 static int ext4_clear_journal_err(struct super_block *sb, 76 struct ext4_super_block *es); 77 static int ext4_sync_fs(struct super_block *sb, int wait); 78 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 79 static int ext4_unfreeze(struct super_block *sb); 80 static int ext4_freeze(struct super_block *sb); 81 static inline int ext2_feature_set_ok(struct super_block *sb); 82 static inline int ext3_feature_set_ok(struct super_block *sb); 83 static void ext4_unregister_li_request(struct super_block *sb); 84 static void ext4_clear_request_list(void); 85 static struct inode *ext4_get_journal_inode(struct super_block *sb, 86 unsigned int journal_inum); 87 static int ext4_validate_options(struct fs_context *fc); 88 static int ext4_check_opt_consistency(struct fs_context *fc, 89 struct super_block *sb); 90 static void ext4_apply_options(struct fs_context *fc, struct super_block *sb); 91 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param); 92 static int ext4_get_tree(struct fs_context *fc); 93 static int ext4_reconfigure(struct fs_context *fc); 94 static void ext4_fc_free(struct fs_context *fc); 95 static int ext4_init_fs_context(struct fs_context *fc); 96 static void ext4_kill_sb(struct super_block *sb); 97 static const struct fs_parameter_spec ext4_param_specs[]; 98 99 /* 100 * Lock ordering 101 * 102 * page fault path: 103 * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start 104 * -> page lock -> i_data_sem (rw) 105 * 106 * buffered write path: 107 * sb_start_write -> i_mutex -> mmap_lock 108 * sb_start_write -> i_mutex -> transaction start -> page lock -> 109 * i_data_sem (rw) 110 * 111 * truncate: 112 * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) -> 113 * page lock 114 * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start -> 115 * i_data_sem (rw) 116 * 117 * direct IO: 118 * sb_start_write -> i_mutex -> mmap_lock 119 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw) 120 * 121 * writepages: 122 * transaction start -> page lock(s) -> i_data_sem (rw) 123 */ 124 125 static const struct fs_context_operations ext4_context_ops = { 126 .parse_param = ext4_parse_param, 127 .get_tree = ext4_get_tree, 128 .reconfigure = ext4_reconfigure, 129 .free = ext4_fc_free, 130 }; 131 132 133 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 134 static struct file_system_type ext2_fs_type = { 135 .owner = THIS_MODULE, 136 .name = "ext2", 137 .init_fs_context = ext4_init_fs_context, 138 .parameters = ext4_param_specs, 139 .kill_sb = ext4_kill_sb, 140 .fs_flags = FS_REQUIRES_DEV, 141 }; 142 MODULE_ALIAS_FS("ext2"); 143 MODULE_ALIAS("ext2"); 144 #define IS_EXT2_SB(sb) ((sb)->s_type == &ext2_fs_type) 145 #else 146 #define IS_EXT2_SB(sb) (0) 147 #endif 148 149 150 static struct file_system_type ext3_fs_type = { 151 .owner = THIS_MODULE, 152 .name = "ext3", 153 .init_fs_context = ext4_init_fs_context, 154 .parameters = ext4_param_specs, 155 .kill_sb = ext4_kill_sb, 156 .fs_flags = FS_REQUIRES_DEV, 157 }; 158 MODULE_ALIAS_FS("ext3"); 159 MODULE_ALIAS("ext3"); 160 #define IS_EXT3_SB(sb) ((sb)->s_type == &ext3_fs_type) 161 162 163 static inline void __ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags, 164 bh_end_io_t *end_io, bool simu_fail) 165 { 166 if (simu_fail) { 167 clear_buffer_uptodate(bh); 168 unlock_buffer(bh); 169 return; 170 } 171 172 /* 173 * buffer's verified bit is no longer valid after reading from 174 * disk again due to write out error, clear it to make sure we 175 * recheck the buffer contents. 176 */ 177 clear_buffer_verified(bh); 178 179 bh->b_end_io = end_io ? end_io : end_buffer_read_sync; 180 get_bh(bh); 181 submit_bh(REQ_OP_READ | op_flags, bh); 182 } 183 184 void ext4_read_bh_nowait(struct buffer_head *bh, blk_opf_t op_flags, 185 bh_end_io_t *end_io, bool simu_fail) 186 { 187 BUG_ON(!buffer_locked(bh)); 188 189 if (ext4_buffer_uptodate(bh)) { 190 unlock_buffer(bh); 191 return; 192 } 193 __ext4_read_bh(bh, op_flags, end_io, simu_fail); 194 } 195 196 int ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags, 197 bh_end_io_t *end_io, bool simu_fail) 198 { 199 BUG_ON(!buffer_locked(bh)); 200 201 if (ext4_buffer_uptodate(bh)) { 202 unlock_buffer(bh); 203 return 0; 204 } 205 206 __ext4_read_bh(bh, op_flags, end_io, simu_fail); 207 208 wait_on_buffer(bh); 209 if (buffer_uptodate(bh)) 210 return 0; 211 return -EIO; 212 } 213 214 int ext4_read_bh_lock(struct buffer_head *bh, blk_opf_t op_flags, bool wait) 215 { 216 lock_buffer(bh); 217 if (!wait) { 218 ext4_read_bh_nowait(bh, op_flags, NULL, false); 219 return 0; 220 } 221 return ext4_read_bh(bh, op_flags, NULL, false); 222 } 223 224 /* 225 * This works like __bread_gfp() except it uses ERR_PTR for error 226 * returns. Currently with sb_bread it's impossible to distinguish 227 * between ENOMEM and EIO situations (since both result in a NULL 228 * return. 229 */ 230 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb, 231 sector_t block, 232 blk_opf_t op_flags, gfp_t gfp) 233 { 234 struct buffer_head *bh; 235 int ret; 236 237 bh = sb_getblk_gfp(sb, block, gfp); 238 if (bh == NULL) 239 return ERR_PTR(-ENOMEM); 240 if (ext4_buffer_uptodate(bh)) 241 return bh; 242 243 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true); 244 if (ret) { 245 put_bh(bh); 246 return ERR_PTR(ret); 247 } 248 return bh; 249 } 250 251 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block, 252 blk_opf_t op_flags) 253 { 254 gfp_t gfp = mapping_gfp_constraint(sb->s_bdev->bd_mapping, 255 ~__GFP_FS) | __GFP_MOVABLE; 256 257 return __ext4_sb_bread_gfp(sb, block, op_flags, gfp); 258 } 259 260 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb, 261 sector_t block) 262 { 263 gfp_t gfp = mapping_gfp_constraint(sb->s_bdev->bd_mapping, 264 ~__GFP_FS); 265 266 return __ext4_sb_bread_gfp(sb, block, 0, gfp); 267 } 268 269 struct buffer_head *ext4_sb_bread_nofail(struct super_block *sb, 270 sector_t block) 271 { 272 gfp_t gfp = mapping_gfp_constraint(sb->s_bdev->bd_mapping, 273 ~__GFP_FS) | __GFP_MOVABLE | __GFP_NOFAIL; 274 275 return __ext4_sb_bread_gfp(sb, block, 0, gfp); 276 } 277 278 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block) 279 { 280 struct buffer_head *bh = bdev_getblk(sb->s_bdev, block, 281 sb->s_blocksize, GFP_NOWAIT); 282 283 if (likely(bh)) { 284 if (trylock_buffer(bh)) 285 ext4_read_bh_nowait(bh, REQ_RAHEAD, NULL, false); 286 brelse(bh); 287 } 288 } 289 290 static int ext4_verify_csum_type(struct super_block *sb, 291 struct ext4_super_block *es) 292 { 293 if (!ext4_has_feature_metadata_csum(sb)) 294 return 1; 295 296 return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 297 } 298 299 __le32 ext4_superblock_csum(struct ext4_super_block *es) 300 { 301 int offset = offsetof(struct ext4_super_block, s_checksum); 302 __u32 csum; 303 304 csum = ext4_chksum(~0, (char *)es, offset); 305 306 return cpu_to_le32(csum); 307 } 308 309 static int ext4_superblock_csum_verify(struct super_block *sb, 310 struct ext4_super_block *es) 311 { 312 if (!ext4_has_feature_metadata_csum(sb)) 313 return 1; 314 315 return es->s_checksum == ext4_superblock_csum(es); 316 } 317 318 void ext4_superblock_csum_set(struct super_block *sb) 319 { 320 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 321 322 if (!ext4_has_feature_metadata_csum(sb)) 323 return; 324 325 es->s_checksum = ext4_superblock_csum(es); 326 } 327 328 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 329 struct ext4_group_desc *bg) 330 { 331 return le32_to_cpu(bg->bg_block_bitmap_lo) | 332 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 333 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 334 } 335 336 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 337 struct ext4_group_desc *bg) 338 { 339 return le32_to_cpu(bg->bg_inode_bitmap_lo) | 340 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 341 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 342 } 343 344 ext4_fsblk_t ext4_inode_table(struct super_block *sb, 345 struct ext4_group_desc *bg) 346 { 347 return le32_to_cpu(bg->bg_inode_table_lo) | 348 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 349 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 350 } 351 352 __u32 ext4_free_group_clusters(struct super_block *sb, 353 struct ext4_group_desc *bg) 354 { 355 return le16_to_cpu(bg->bg_free_blocks_count_lo) | 356 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 357 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 358 } 359 360 __u32 ext4_free_inodes_count(struct super_block *sb, 361 struct ext4_group_desc *bg) 362 { 363 return le16_to_cpu(READ_ONCE(bg->bg_free_inodes_count_lo)) | 364 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 365 (__u32)le16_to_cpu(READ_ONCE(bg->bg_free_inodes_count_hi)) << 16 : 0); 366 } 367 368 __u32 ext4_used_dirs_count(struct super_block *sb, 369 struct ext4_group_desc *bg) 370 { 371 return le16_to_cpu(bg->bg_used_dirs_count_lo) | 372 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 373 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 374 } 375 376 __u32 ext4_itable_unused_count(struct super_block *sb, 377 struct ext4_group_desc *bg) 378 { 379 return le16_to_cpu(bg->bg_itable_unused_lo) | 380 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 381 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 382 } 383 384 void ext4_block_bitmap_set(struct super_block *sb, 385 struct ext4_group_desc *bg, ext4_fsblk_t blk) 386 { 387 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 388 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 389 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 390 } 391 392 void ext4_inode_bitmap_set(struct super_block *sb, 393 struct ext4_group_desc *bg, ext4_fsblk_t blk) 394 { 395 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 396 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 397 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 398 } 399 400 void ext4_inode_table_set(struct super_block *sb, 401 struct ext4_group_desc *bg, ext4_fsblk_t blk) 402 { 403 bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 404 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 405 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 406 } 407 408 void ext4_free_group_clusters_set(struct super_block *sb, 409 struct ext4_group_desc *bg, __u32 count) 410 { 411 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 412 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 413 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 414 } 415 416 void ext4_free_inodes_set(struct super_block *sb, 417 struct ext4_group_desc *bg, __u32 count) 418 { 419 WRITE_ONCE(bg->bg_free_inodes_count_lo, cpu_to_le16((__u16)count)); 420 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 421 WRITE_ONCE(bg->bg_free_inodes_count_hi, cpu_to_le16(count >> 16)); 422 } 423 424 void ext4_used_dirs_set(struct super_block *sb, 425 struct ext4_group_desc *bg, __u32 count) 426 { 427 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 428 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 429 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 430 } 431 432 void ext4_itable_unused_set(struct super_block *sb, 433 struct ext4_group_desc *bg, __u32 count) 434 { 435 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 436 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 437 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 438 } 439 440 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now) 441 { 442 now = clamp_val(now, 0, (1ull << 40) - 1); 443 444 *lo = cpu_to_le32(lower_32_bits(now)); 445 *hi = upper_32_bits(now); 446 } 447 448 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi) 449 { 450 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo); 451 } 452 #define ext4_update_tstamp(es, tstamp) \ 453 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \ 454 ktime_get_real_seconds()) 455 #define ext4_get_tstamp(es, tstamp) \ 456 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 457 458 /* 459 * The ext4_maybe_update_superblock() function checks and updates the 460 * superblock if needed. 461 * 462 * This function is designed to update the on-disk superblock only under 463 * certain conditions to prevent excessive disk writes and unnecessary 464 * waking of the disk from sleep. The superblock will be updated if: 465 * 1. More than sbi->s_sb_update_sec (def: 1 hour) has passed since the last 466 * superblock update 467 * 2. More than sbi->s_sb_update_kb (def: 16MB) kbs have been written since the 468 * last superblock update. 469 * 470 * @sb: The superblock 471 */ 472 static void ext4_maybe_update_superblock(struct super_block *sb) 473 { 474 struct ext4_sb_info *sbi = EXT4_SB(sb); 475 struct ext4_super_block *es = sbi->s_es; 476 journal_t *journal = sbi->s_journal; 477 time64_t now; 478 __u64 last_update; 479 __u64 lifetime_write_kbytes; 480 __u64 diff_size; 481 482 if (ext4_emergency_state(sb) || sb_rdonly(sb) || 483 !(sb->s_flags & SB_ACTIVE) || !journal || 484 journal->j_flags & JBD2_UNMOUNT) 485 return; 486 487 now = ktime_get_real_seconds(); 488 last_update = ext4_get_tstamp(es, s_wtime); 489 490 if (likely(now - last_update < sbi->s_sb_update_sec)) 491 return; 492 493 lifetime_write_kbytes = sbi->s_kbytes_written + 494 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) - 495 sbi->s_sectors_written_start) >> 1); 496 497 /* Get the number of kilobytes not written to disk to account 498 * for statistics and compare with a multiple of 16 MB. This 499 * is used to determine when the next superblock commit should 500 * occur (i.e. not more often than once per 16MB if there was 501 * less written in an hour). 502 */ 503 diff_size = lifetime_write_kbytes - le64_to_cpu(es->s_kbytes_written); 504 505 if (diff_size > sbi->s_sb_update_kb) 506 schedule_work(&EXT4_SB(sb)->s_sb_upd_work); 507 } 508 509 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 510 { 511 struct super_block *sb = journal->j_private; 512 513 BUG_ON(txn->t_state == T_FINISHED); 514 515 ext4_process_freed_data(sb, txn->t_tid); 516 ext4_maybe_update_superblock(sb); 517 } 518 519 static bool ext4_journalled_writepage_needs_redirty(struct jbd2_inode *jinode, 520 struct folio *folio) 521 { 522 struct buffer_head *bh, *head; 523 struct journal_head *jh; 524 transaction_t *trans = READ_ONCE(jinode->i_transaction); 525 526 bh = head = folio_buffers(folio); 527 do { 528 /* 529 * We have to redirty a page in these cases: 530 * 1) If buffer is dirty, it means the page was dirty because it 531 * contains a buffer that needs checkpointing. So the dirty bit 532 * needs to be preserved so that checkpointing writes the buffer 533 * properly. 534 * 2) If buffer is not part of the committing transaction 535 * (we may have just accidentally come across this buffer because 536 * inode range tracking is not exact) or if the currently running 537 * transaction already contains this buffer as well, dirty bit 538 * needs to be preserved so that the buffer gets writeprotected 539 * properly on running transaction's commit. 540 */ 541 jh = bh2jh(bh); 542 if (buffer_dirty(bh) || 543 (jh && (jh->b_transaction != trans || 544 jh->b_next_transaction))) 545 return true; 546 } while ((bh = bh->b_this_page) != head); 547 548 return false; 549 } 550 551 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode) 552 { 553 struct address_space *mapping = jinode->i_vfs_inode->i_mapping; 554 loff_t range_start, range_end; 555 struct writeback_control wbc = { 556 .sync_mode = WB_SYNC_ALL, 557 .nr_to_write = LONG_MAX, 558 }; 559 struct folio *folio = NULL; 560 int error; 561 562 if (!jbd2_jinode_get_dirty_range(jinode, &range_start, &range_end)) 563 return 0; 564 565 wbc.range_start = range_start; 566 wbc.range_end = range_end; 567 568 /* 569 * writeback_iter() already checks for dirty pages and calls 570 * folio_clear_dirty_for_io(), which we want to write protect the 571 * folios. 572 * 573 * However, we may have to redirty a folio sometimes. 574 */ 575 while ((folio = writeback_iter(mapping, &wbc, folio, &error))) { 576 if (ext4_journalled_writepage_needs_redirty(jinode, folio)) 577 folio_redirty_for_writepage(&wbc, folio); 578 folio_unlock(folio); 579 } 580 581 return error; 582 } 583 584 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode) 585 { 586 int ret; 587 588 if (ext4_should_journal_data(jinode->i_vfs_inode)) 589 ret = ext4_journalled_submit_inode_data_buffers(jinode); 590 else 591 ret = ext4_normal_submit_inode_data_buffers(jinode); 592 return ret; 593 } 594 595 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode) 596 { 597 int ret = 0; 598 599 if (!ext4_should_journal_data(jinode->i_vfs_inode)) 600 ret = jbd2_journal_finish_inode_data_buffers(jinode); 601 602 return ret; 603 } 604 605 static bool system_going_down(void) 606 { 607 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 608 || system_state == SYSTEM_RESTART; 609 } 610 611 struct ext4_err_translation { 612 int code; 613 int errno; 614 }; 615 616 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err } 617 618 static struct ext4_err_translation err_translation[] = { 619 EXT4_ERR_TRANSLATE(EIO), 620 EXT4_ERR_TRANSLATE(ENOMEM), 621 EXT4_ERR_TRANSLATE(EFSBADCRC), 622 EXT4_ERR_TRANSLATE(EFSCORRUPTED), 623 EXT4_ERR_TRANSLATE(ENOSPC), 624 EXT4_ERR_TRANSLATE(ENOKEY), 625 EXT4_ERR_TRANSLATE(EROFS), 626 EXT4_ERR_TRANSLATE(EFBIG), 627 EXT4_ERR_TRANSLATE(EEXIST), 628 EXT4_ERR_TRANSLATE(ERANGE), 629 EXT4_ERR_TRANSLATE(EOVERFLOW), 630 EXT4_ERR_TRANSLATE(EBUSY), 631 EXT4_ERR_TRANSLATE(ENOTDIR), 632 EXT4_ERR_TRANSLATE(ENOTEMPTY), 633 EXT4_ERR_TRANSLATE(ESHUTDOWN), 634 EXT4_ERR_TRANSLATE(EFAULT), 635 }; 636 637 static int ext4_errno_to_code(int errno) 638 { 639 int i; 640 641 for (i = 0; i < ARRAY_SIZE(err_translation); i++) 642 if (err_translation[i].errno == errno) 643 return err_translation[i].code; 644 return EXT4_ERR_UNKNOWN; 645 } 646 647 static void save_error_info(struct super_block *sb, int error, 648 __u32 ino, __u64 block, 649 const char *func, unsigned int line) 650 { 651 struct ext4_sb_info *sbi = EXT4_SB(sb); 652 653 /* We default to EFSCORRUPTED error... */ 654 if (error == 0) 655 error = EFSCORRUPTED; 656 657 spin_lock(&sbi->s_error_lock); 658 sbi->s_add_error_count++; 659 sbi->s_last_error_code = error; 660 sbi->s_last_error_line = line; 661 sbi->s_last_error_ino = ino; 662 sbi->s_last_error_block = block; 663 sbi->s_last_error_func = func; 664 sbi->s_last_error_time = ktime_get_real_seconds(); 665 if (!sbi->s_first_error_time) { 666 sbi->s_first_error_code = error; 667 sbi->s_first_error_line = line; 668 sbi->s_first_error_ino = ino; 669 sbi->s_first_error_block = block; 670 sbi->s_first_error_func = func; 671 sbi->s_first_error_time = sbi->s_last_error_time; 672 } 673 spin_unlock(&sbi->s_error_lock); 674 } 675 676 /* Deal with the reporting of failure conditions on a filesystem such as 677 * inconsistencies detected or read IO failures. 678 * 679 * On ext2, we can store the error state of the filesystem in the 680 * superblock. That is not possible on ext4, because we may have other 681 * write ordering constraints on the superblock which prevent us from 682 * writing it out straight away; and given that the journal is about to 683 * be aborted, we can't rely on the current, or future, transactions to 684 * write out the superblock safely. 685 * 686 * We'll just use the jbd2_journal_abort() error code to record an error in 687 * the journal instead. On recovery, the journal will complain about 688 * that error until we've noted it down and cleared it. 689 * 690 * If force_ro is set, we unconditionally force the filesystem into an 691 * ABORT|READONLY state, unless the error response on the fs has been set to 692 * panic in which case we take the easy way out and panic immediately. This is 693 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM 694 * at a critical moment in log management. 695 */ 696 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error, 697 __u32 ino, __u64 block, 698 const char *func, unsigned int line) 699 { 700 journal_t *journal = EXT4_SB(sb)->s_journal; 701 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT); 702 703 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 704 if (test_opt(sb, WARN_ON_ERROR)) 705 WARN_ON_ONCE(1); 706 707 if (!continue_fs && !ext4_emergency_ro(sb) && journal) 708 jbd2_journal_abort(journal, -error); 709 710 if (!bdev_read_only(sb->s_bdev)) { 711 save_error_info(sb, error, ino, block, func, line); 712 /* 713 * In case the fs should keep running, we need to writeout 714 * superblock through the journal. Due to lock ordering 715 * constraints, it may not be safe to do it right here so we 716 * defer superblock flushing to a workqueue. We just need to be 717 * careful when the journal is already shutting down. If we get 718 * here in that case, just update the sb directly as the last 719 * transaction won't commit anyway. 720 */ 721 if (continue_fs && journal && 722 !ext4_test_mount_flag(sb, EXT4_MF_JOURNAL_DESTROY)) 723 schedule_work(&EXT4_SB(sb)->s_sb_upd_work); 724 else 725 ext4_commit_super(sb); 726 } 727 728 /* 729 * We force ERRORS_RO behavior when system is rebooting. Otherwise we 730 * could panic during 'reboot -f' as the underlying device got already 731 * disabled. 732 */ 733 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) { 734 panic("EXT4-fs (device %s): panic forced after error\n", 735 sb->s_id); 736 } 737 738 if (ext4_emergency_ro(sb) || continue_fs) 739 return; 740 741 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 742 /* 743 * We don't set SB_RDONLY because that requires sb->s_umount 744 * semaphore and setting it without proper remount procedure is 745 * confusing code such as freeze_super() leading to deadlocks 746 * and other problems. 747 */ 748 set_bit(EXT4_FLAGS_EMERGENCY_RO, &EXT4_SB(sb)->s_ext4_flags); 749 } 750 751 static void update_super_work(struct work_struct *work) 752 { 753 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info, 754 s_sb_upd_work); 755 journal_t *journal = sbi->s_journal; 756 handle_t *handle; 757 758 /* 759 * If the journal is still running, we have to write out superblock 760 * through the journal to avoid collisions of other journalled sb 761 * updates. 762 * 763 * We use directly jbd2 functions here to avoid recursing back into 764 * ext4 error handling code during handling of previous errors. 765 */ 766 if (!ext4_emergency_state(sbi->s_sb) && 767 !sb_rdonly(sbi->s_sb) && journal) { 768 struct buffer_head *sbh = sbi->s_sbh; 769 bool call_notify_err = false; 770 771 handle = jbd2_journal_start(journal, 1); 772 if (IS_ERR(handle)) 773 goto write_directly; 774 if (jbd2_journal_get_write_access(handle, sbh)) { 775 jbd2_journal_stop(handle); 776 goto write_directly; 777 } 778 779 if (sbi->s_add_error_count > 0) 780 call_notify_err = true; 781 782 ext4_update_super(sbi->s_sb); 783 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 784 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to " 785 "superblock detected"); 786 clear_buffer_write_io_error(sbh); 787 set_buffer_uptodate(sbh); 788 } 789 790 if (jbd2_journal_dirty_metadata(handle, sbh)) { 791 jbd2_journal_stop(handle); 792 goto write_directly; 793 } 794 jbd2_journal_stop(handle); 795 796 if (call_notify_err) 797 ext4_notify_error_sysfs(sbi); 798 799 return; 800 } 801 write_directly: 802 /* 803 * Write through journal failed. Write sb directly to get error info 804 * out and hope for the best. 805 */ 806 ext4_commit_super(sbi->s_sb); 807 ext4_notify_error_sysfs(sbi); 808 } 809 810 #define ext4_error_ratelimit(sb) \ 811 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ 812 "EXT4-fs error") 813 814 void __ext4_error(struct super_block *sb, const char *function, 815 unsigned int line, bool force_ro, int error, __u64 block, 816 const char *fmt, ...) 817 { 818 struct va_format vaf; 819 va_list args; 820 821 if (unlikely(ext4_emergency_state(sb))) 822 return; 823 824 trace_ext4_error(sb, function, line); 825 if (ext4_error_ratelimit(sb)) { 826 va_start(args, fmt); 827 vaf.fmt = fmt; 828 vaf.va = &args; 829 printk(KERN_CRIT 830 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 831 sb->s_id, function, line, current->comm, &vaf); 832 va_end(args); 833 } 834 fserror_report_metadata(sb, error ? -abs(error) : -EFSCORRUPTED, 835 GFP_ATOMIC); 836 837 ext4_handle_error(sb, force_ro, error, 0, block, function, line); 838 } 839 840 void __ext4_error_inode(struct inode *inode, const char *function, 841 unsigned int line, ext4_fsblk_t block, int error, 842 const char *fmt, ...) 843 { 844 va_list args; 845 struct va_format vaf; 846 847 if (unlikely(ext4_emergency_state(inode->i_sb))) 848 return; 849 850 trace_ext4_error(inode->i_sb, function, line); 851 if (ext4_error_ratelimit(inode->i_sb)) { 852 va_start(args, fmt); 853 vaf.fmt = fmt; 854 vaf.va = &args; 855 if (block) 856 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 857 "inode #%llu: block %llu: comm %s: %pV\n", 858 inode->i_sb->s_id, function, line, inode->i_ino, 859 block, current->comm, &vaf); 860 else 861 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 862 "inode #%llu: comm %s: %pV\n", 863 inode->i_sb->s_id, function, line, inode->i_ino, 864 current->comm, &vaf); 865 va_end(args); 866 } 867 fserror_report_file_metadata(inode, 868 error ? -abs(error) : -EFSCORRUPTED, 869 GFP_ATOMIC); 870 871 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block, 872 function, line); 873 } 874 875 void __ext4_error_file(struct file *file, const char *function, 876 unsigned int line, ext4_fsblk_t block, 877 const char *fmt, ...) 878 { 879 va_list args; 880 struct va_format vaf; 881 struct inode *inode = file_inode(file); 882 char pathname[80], *path; 883 884 if (unlikely(ext4_emergency_state(inode->i_sb))) 885 return; 886 887 trace_ext4_error(inode->i_sb, function, line); 888 if (ext4_error_ratelimit(inode->i_sb)) { 889 path = file_path(file, pathname, sizeof(pathname)); 890 if (IS_ERR(path)) 891 path = "(unknown)"; 892 va_start(args, fmt); 893 vaf.fmt = fmt; 894 vaf.va = &args; 895 if (block) 896 printk(KERN_CRIT 897 "EXT4-fs error (device %s): %s:%d: inode #%llu: " 898 "block %llu: comm %s: path %s: %pV\n", 899 inode->i_sb->s_id, function, line, inode->i_ino, 900 block, current->comm, path, &vaf); 901 else 902 printk(KERN_CRIT 903 "EXT4-fs error (device %s): %s:%d: inode #%llu: " 904 "comm %s: path %s: %pV\n", 905 inode->i_sb->s_id, function, line, inode->i_ino, 906 current->comm, path, &vaf); 907 va_end(args); 908 } 909 fserror_report_file_metadata(inode, -EFSCORRUPTED, GFP_ATOMIC); 910 911 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block, 912 function, line); 913 } 914 915 const char *ext4_decode_error(struct super_block *sb, int errno, 916 char nbuf[16]) 917 { 918 char *errstr = NULL; 919 920 switch (errno) { 921 case -EFSCORRUPTED: 922 errstr = "Corrupt filesystem"; 923 break; 924 case -EFSBADCRC: 925 errstr = "Filesystem failed CRC"; 926 break; 927 case -EIO: 928 errstr = "IO failure"; 929 break; 930 case -ENOMEM: 931 errstr = "Out of memory"; 932 break; 933 case -EROFS: 934 if (!sb || (EXT4_SB(sb)->s_journal && 935 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 936 errstr = "Journal has aborted"; 937 else 938 errstr = "Readonly filesystem"; 939 break; 940 default: 941 /* If the caller passed in an extra buffer for unknown 942 * errors, textualise them now. Else we just return 943 * NULL. */ 944 if (nbuf) { 945 /* Check for truncated error codes... */ 946 if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 947 errstr = nbuf; 948 } 949 break; 950 } 951 952 return errstr; 953 } 954 955 /* __ext4_std_error decodes expected errors from journaling functions 956 * automatically and invokes the appropriate error response. */ 957 958 void __ext4_std_error(struct super_block *sb, const char *function, 959 unsigned int line, int errno) 960 { 961 char nbuf[16]; 962 const char *errstr; 963 964 if (unlikely(ext4_emergency_state(sb))) 965 return; 966 967 /* Special case: if the error is EROFS, and we're not already 968 * inside a transaction, then there's really no point in logging 969 * an error. */ 970 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb)) 971 return; 972 973 if (ext4_error_ratelimit(sb)) { 974 errstr = ext4_decode_error(sb, errno, nbuf); 975 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 976 sb->s_id, function, line, errstr); 977 } 978 fserror_report_metadata(sb, errno ? -abs(errno) : -EFSCORRUPTED, 979 GFP_ATOMIC); 980 981 ext4_handle_error(sb, false, -errno, 0, 0, function, line); 982 } 983 984 void __ext4_msg(struct super_block *sb, 985 const char *prefix, const char *fmt, ...) 986 { 987 struct va_format vaf; 988 va_list args; 989 990 if (sb) { 991 atomic_inc(&EXT4_SB(sb)->s_msg_count); 992 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), 993 "EXT4-fs")) 994 return; 995 } 996 997 va_start(args, fmt); 998 vaf.fmt = fmt; 999 vaf.va = &args; 1000 if (sb) 1001 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 1002 else 1003 printk("%sEXT4-fs: %pV\n", prefix, &vaf); 1004 va_end(args); 1005 } 1006 1007 static int ext4_warning_ratelimit(struct super_block *sb) 1008 { 1009 atomic_inc(&EXT4_SB(sb)->s_warning_count); 1010 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), 1011 "EXT4-fs warning"); 1012 } 1013 1014 void __ext4_warning(struct super_block *sb, const char *function, 1015 unsigned int line, const char *fmt, ...) 1016 { 1017 struct va_format vaf; 1018 va_list args; 1019 1020 if (!ext4_warning_ratelimit(sb)) 1021 return; 1022 1023 va_start(args, fmt); 1024 vaf.fmt = fmt; 1025 vaf.va = &args; 1026 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 1027 sb->s_id, function, line, &vaf); 1028 va_end(args); 1029 } 1030 1031 void __ext4_warning_inode(const struct inode *inode, const char *function, 1032 unsigned int line, const char *fmt, ...) 1033 { 1034 struct va_format vaf; 1035 va_list args; 1036 1037 if (!ext4_warning_ratelimit(inode->i_sb)) 1038 return; 1039 1040 va_start(args, fmt); 1041 vaf.fmt = fmt; 1042 vaf.va = &args; 1043 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: " 1044 "inode #%llu: comm %s: %pV\n", inode->i_sb->s_id, 1045 function, line, inode->i_ino, current->comm, &vaf); 1046 va_end(args); 1047 } 1048 1049 void __ext4_grp_locked_error(const char *function, unsigned int line, 1050 struct super_block *sb, ext4_group_t grp, 1051 u64 ino, ext4_fsblk_t block, 1052 const char *fmt, ...) 1053 __releases(bitlock) 1054 __acquires(bitlock) 1055 { 1056 struct va_format vaf; 1057 va_list args; 1058 1059 if (unlikely(ext4_emergency_state(sb))) 1060 return; 1061 1062 trace_ext4_error(sb, function, line); 1063 if (ext4_error_ratelimit(sb)) { 1064 va_start(args, fmt); 1065 vaf.fmt = fmt; 1066 vaf.va = &args; 1067 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 1068 sb->s_id, function, line, grp); 1069 if (ino) 1070 printk(KERN_CONT "inode %llu: ", ino); 1071 if (block) 1072 printk(KERN_CONT "block %llu:", 1073 (unsigned long long) block); 1074 printk(KERN_CONT "%pV\n", &vaf); 1075 va_end(args); 1076 } 1077 1078 if (test_opt(sb, ERRORS_CONT)) { 1079 if (test_opt(sb, WARN_ON_ERROR)) 1080 WARN_ON_ONCE(1); 1081 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 1082 if (!bdev_read_only(sb->s_bdev)) { 1083 save_error_info(sb, EFSCORRUPTED, ino, block, function, 1084 line); 1085 schedule_work(&EXT4_SB(sb)->s_sb_upd_work); 1086 } 1087 return; 1088 } 1089 ext4_unlock_group(sb, grp); 1090 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line); 1091 /* 1092 * We only get here in the ERRORS_RO case; relocking the group 1093 * may be dangerous, but nothing bad will happen since the 1094 * filesystem will have already been marked read/only and the 1095 * journal has been aborted. We return 1 as a hint to callers 1096 * who might what to use the return value from 1097 * ext4_grp_locked_error() to distinguish between the 1098 * ERRORS_CONT and ERRORS_RO case, and perhaps return more 1099 * aggressively from the ext4 function in question, with a 1100 * more appropriate error code. 1101 */ 1102 ext4_lock_group(sb, grp); 1103 return; 1104 } 1105 1106 void ext4_mark_group_bitmap_corrupted(struct super_block *sb, 1107 ext4_group_t group, 1108 unsigned int flags) 1109 { 1110 struct ext4_sb_info *sbi = EXT4_SB(sb); 1111 struct ext4_group_info *grp = ext4_get_group_info(sb, group); 1112 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL); 1113 int ret; 1114 1115 if (!grp || !gdp) 1116 return; 1117 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) { 1118 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, 1119 &grp->bb_state); 1120 if (!ret) 1121 percpu_counter_sub(&sbi->s_freeclusters_counter, 1122 grp->bb_free); 1123 } 1124 1125 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) { 1126 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, 1127 &grp->bb_state); 1128 if (!ret && gdp) { 1129 int count; 1130 1131 count = ext4_free_inodes_count(sb, gdp); 1132 percpu_counter_sub(&sbi->s_freeinodes_counter, 1133 count); 1134 } 1135 } 1136 } 1137 1138 void ext4_update_dynamic_rev(struct super_block *sb) 1139 { 1140 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 1141 1142 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 1143 return; 1144 1145 ext4_warning(sb, 1146 "updating to rev %d because of new feature flag, " 1147 "running e2fsck is recommended", 1148 EXT4_DYNAMIC_REV); 1149 1150 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 1151 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 1152 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 1153 /* leave es->s_feature_*compat flags alone */ 1154 /* es->s_uuid will be set by e2fsck if empty */ 1155 1156 /* 1157 * The rest of the superblock fields should be zero, and if not it 1158 * means they are likely already in use, so leave them alone. We 1159 * can leave it up to e2fsck to clean up any inconsistencies there. 1160 */ 1161 } 1162 1163 static inline struct inode *orphan_list_entry(struct list_head *l) 1164 { 1165 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 1166 } 1167 1168 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 1169 { 1170 struct list_head *l; 1171 1172 ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 1173 le32_to_cpu(sbi->s_es->s_last_orphan)); 1174 1175 printk(KERN_ERR "sb_info orphan list:\n"); 1176 list_for_each(l, &sbi->s_orphan) { 1177 struct inode *inode = orphan_list_entry(l); 1178 printk(KERN_ERR " " 1179 "inode %s:%llu at %p: mode %o, nlink %d, next %d\n", 1180 inode->i_sb->s_id, inode->i_ino, inode, 1181 inode->i_mode, inode->i_nlink, 1182 NEXT_ORPHAN(inode)); 1183 } 1184 } 1185 1186 #ifdef CONFIG_QUOTA 1187 static int ext4_quota_off(struct super_block *sb, int type); 1188 1189 static inline void ext4_quotas_off(struct super_block *sb, int type) 1190 { 1191 BUG_ON(type > EXT4_MAXQUOTAS); 1192 1193 /* Use our quota_off function to clear inode flags etc. */ 1194 for (type--; type >= 0; type--) 1195 ext4_quota_off(sb, type); 1196 } 1197 1198 /* 1199 * This is a helper function which is used in the mount/remount 1200 * codepaths (which holds s_umount) to fetch the quota file name. 1201 */ 1202 static inline char *get_qf_name(struct super_block *sb, 1203 struct ext4_sb_info *sbi, 1204 int type) 1205 { 1206 return rcu_dereference_protected(sbi->s_qf_names[type], 1207 lockdep_is_held(&sb->s_umount)); 1208 } 1209 #else 1210 static inline void ext4_quotas_off(struct super_block *sb, int type) 1211 { 1212 } 1213 #endif 1214 1215 static int ext4_percpu_param_init(struct ext4_sb_info *sbi) 1216 { 1217 ext4_fsblk_t block; 1218 int err; 1219 1220 block = ext4_count_free_clusters(sbi->s_sb); 1221 ext4_free_blocks_count_set(sbi->s_es, EXT4_C2B(sbi, block)); 1222 err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 1223 GFP_KERNEL); 1224 if (!err) { 1225 unsigned long freei = ext4_count_free_inodes(sbi->s_sb); 1226 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 1227 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 1228 GFP_KERNEL); 1229 } 1230 if (!err) 1231 err = percpu_counter_init(&sbi->s_dirs_counter, 1232 ext4_count_dirs(sbi->s_sb), GFP_KERNEL); 1233 if (!err) 1234 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 1235 GFP_KERNEL); 1236 if (!err) 1237 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0, 1238 GFP_KERNEL); 1239 if (!err) 1240 err = percpu_init_rwsem(&sbi->s_writepages_rwsem); 1241 1242 if (err) 1243 ext4_msg(sbi->s_sb, KERN_ERR, "insufficient memory"); 1244 1245 return err; 1246 } 1247 1248 static void ext4_percpu_param_destroy(struct ext4_sb_info *sbi) 1249 { 1250 percpu_counter_destroy(&sbi->s_freeclusters_counter); 1251 percpu_counter_destroy(&sbi->s_freeinodes_counter); 1252 percpu_counter_destroy(&sbi->s_dirs_counter); 1253 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 1254 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit); 1255 percpu_free_rwsem(&sbi->s_writepages_rwsem); 1256 } 1257 1258 static void ext4_group_desc_free(struct ext4_sb_info *sbi) 1259 { 1260 struct buffer_head **group_desc; 1261 int i; 1262 1263 group_desc = rcu_access_pointer(sbi->s_group_desc); 1264 for (i = 0; i < sbi->s_gdb_count; i++) 1265 brelse(group_desc[i]); 1266 kvfree(group_desc); 1267 } 1268 1269 static void ext4_flex_groups_free(struct ext4_sb_info *sbi) 1270 { 1271 struct flex_groups **flex_groups; 1272 int i; 1273 1274 flex_groups = rcu_access_pointer(sbi->s_flex_groups); 1275 if (flex_groups) { 1276 for (i = 0; i < sbi->s_flex_groups_allocated; i++) 1277 kvfree(flex_groups[i]); 1278 kvfree(flex_groups); 1279 } 1280 } 1281 1282 static void ext4_put_super(struct super_block *sb) 1283 { 1284 struct ext4_sb_info *sbi = EXT4_SB(sb); 1285 struct ext4_super_block *es = sbi->s_es; 1286 int aborted = 0; 1287 int err; 1288 1289 /* 1290 * Unregister sysfs before destroying jbd2 journal. 1291 * Since we could still access attr_journal_task attribute via sysfs 1292 * path which could have sbi->s_journal->j_task as NULL 1293 * Unregister sysfs before flush sbi->s_sb_upd_work. 1294 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If 1295 * read metadata verify failed then will queue error work. 1296 * update_super_work will call start_this_handle may trigger 1297 * BUG_ON. 1298 */ 1299 ext4_unregister_sysfs(sb); 1300 1301 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount")) 1302 ext4_msg(sb, KERN_INFO, "unmounting filesystem %pU.", 1303 &sb->s_uuid); 1304 1305 ext4_unregister_li_request(sb); 1306 ext4_quotas_off(sb, EXT4_MAXQUOTAS); 1307 1308 destroy_workqueue(sbi->rsv_conversion_wq); 1309 ext4_release_orphan_info(sb); 1310 1311 if (sbi->s_journal) { 1312 aborted = is_journal_aborted(sbi->s_journal); 1313 err = ext4_journal_destroy(sbi, sbi->s_journal); 1314 if ((err < 0) && !aborted) { 1315 ext4_abort(sb, -err, "Couldn't clean up the journal"); 1316 } 1317 } else 1318 flush_work(&sbi->s_sb_upd_work); 1319 1320 ext4_es_unregister_shrinker(sbi); 1321 timer_shutdown_sync(&sbi->s_err_report); 1322 ext4_release_system_zone(sb); 1323 ext4_mb_release(sb); 1324 ext4_ext_release(sb); 1325 1326 if (!ext4_emergency_state(sb) && !sb_rdonly(sb)) { 1327 if (!aborted) { 1328 ext4_clear_feature_journal_needs_recovery(sb); 1329 ext4_clear_feature_orphan_present(sb); 1330 es->s_state = cpu_to_le16(sbi->s_mount_state); 1331 } 1332 ext4_commit_super(sb); 1333 } 1334 1335 ext4_group_desc_free(sbi); 1336 ext4_flex_groups_free(sbi); 1337 1338 WARN_ON_ONCE(!(sbi->s_mount_state & EXT4_ERROR_FS) && 1339 percpu_counter_sum(&sbi->s_dirtyclusters_counter)); 1340 ext4_percpu_param_destroy(sbi); 1341 #ifdef CONFIG_QUOTA 1342 for (int i = 0; i < EXT4_MAXQUOTAS; i++) 1343 kfree(get_qf_name(sb, sbi, i)); 1344 #endif 1345 1346 /* Debugging code just in case the in-memory inode orphan list 1347 * isn't empty. The on-disk one can be non-empty if we've 1348 * detected an error and taken the fs readonly, but the 1349 * in-memory list had better be clean by this point. */ 1350 if (!list_empty(&sbi->s_orphan)) 1351 dump_orphan_list(sb, sbi); 1352 ASSERT(list_empty(&sbi->s_orphan)); 1353 1354 sync_blockdev(sb->s_bdev); 1355 invalidate_bdev(sb->s_bdev); 1356 if (sbi->s_journal_bdev_file) { 1357 /* 1358 * Invalidate the journal device's buffers. We don't want them 1359 * floating about in memory - the physical journal device may 1360 * hotswapped, and it breaks the `ro-after' testing code. 1361 */ 1362 sync_blockdev(file_bdev(sbi->s_journal_bdev_file)); 1363 invalidate_bdev(file_bdev(sbi->s_journal_bdev_file)); 1364 } 1365 1366 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 1367 sbi->s_ea_inode_cache = NULL; 1368 1369 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 1370 sbi->s_ea_block_cache = NULL; 1371 1372 ext4_stop_mmpd(sbi); 1373 1374 brelse(sbi->s_sbh); 1375 sb->s_fs_info = NULL; 1376 /* 1377 * Now that we are completely done shutting down the 1378 * superblock, we need to actually destroy the kobject. 1379 */ 1380 kobject_put(&sbi->s_kobj); 1381 wait_for_completion(&sbi->s_kobj_unregister); 1382 kfree(sbi->s_blockgroup_lock); 1383 fs_put_dax(sbi->s_daxdev, NULL); 1384 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy); 1385 #if IS_ENABLED(CONFIG_UNICODE) 1386 utf8_unload(sb->s_encoding); 1387 #endif 1388 kfree(sbi); 1389 } 1390 1391 static struct kmem_cache *ext4_inode_cachep; 1392 1393 /* 1394 * Called inside transaction, so use GFP_NOFS 1395 */ 1396 static struct inode *ext4_alloc_inode(struct super_block *sb) 1397 { 1398 struct ext4_inode_info *ei; 1399 1400 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS); 1401 if (!ei) 1402 return NULL; 1403 1404 inode_set_iversion(&ei->vfs_inode, 1); 1405 ei->i_flags = 0; 1406 ext4_clear_state_flags(ei); /* Only relevant on 32-bit archs */ 1407 spin_lock_init(&ei->i_raw_lock); 1408 ei->i_prealloc_node = RB_ROOT; 1409 atomic_set(&ei->i_prealloc_active, 0); 1410 rwlock_init(&ei->i_prealloc_lock); 1411 ext4_es_init_tree(&ei->i_es_tree); 1412 rwlock_init(&ei->i_es_lock); 1413 INIT_LIST_HEAD(&ei->i_es_list); 1414 ei->i_es_all_nr = 0; 1415 ei->i_es_shk_nr = 0; 1416 ei->i_es_shrink_lblk = 0; 1417 ei->i_es_seq = 0; 1418 ei->i_reserved_data_blocks = 0; 1419 spin_lock_init(&(ei->i_block_reservation_lock)); 1420 ext4_init_pending_tree(&ei->i_pending_tree); 1421 #ifdef CONFIG_QUOTA 1422 ei->i_reserved_quota = 0; 1423 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); 1424 #endif 1425 ei->jinode = NULL; 1426 INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 1427 spin_lock_init(&ei->i_completed_io_lock); 1428 ei->i_sync_tid = 0; 1429 ei->i_datasync_tid = 0; 1430 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 1431 ext4_fc_init_inode(&ei->vfs_inode); 1432 spin_lock_init(&ei->i_fc_lock); 1433 mmb_init(&ei->i_metadata_bhs, &ei->vfs_inode.i_data); 1434 return &ei->vfs_inode; 1435 } 1436 1437 static int ext4_drop_inode(struct inode *inode) 1438 { 1439 int drop = inode_generic_drop(inode); 1440 1441 if (!drop) 1442 drop = fscrypt_drop_inode(inode); 1443 1444 trace_ext4_drop_inode(inode, drop); 1445 return drop; 1446 } 1447 1448 static void ext4_free_in_core_inode(struct inode *inode) 1449 { 1450 fscrypt_free_inode(inode); 1451 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) { 1452 pr_warn("%s: inode %llu still in fc list", 1453 __func__, inode->i_ino); 1454 } 1455 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 1456 } 1457 1458 static void ext4_destroy_inode(struct inode *inode) 1459 { 1460 if (ext4_inode_orphan_tracked(inode)) { 1461 ext4_msg(inode->i_sb, KERN_ERR, 1462 "Inode %llu (%p): inode tracked as orphan!", 1463 inode->i_ino, EXT4_I(inode)); 1464 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 1465 EXT4_I(inode), sizeof(struct ext4_inode_info), 1466 true); 1467 dump_stack(); 1468 } 1469 1470 if (!(EXT4_SB(inode->i_sb)->s_mount_state & EXT4_ERROR_FS) && 1471 WARN_ON_ONCE(EXT4_I(inode)->i_reserved_data_blocks)) 1472 ext4_msg(inode->i_sb, KERN_ERR, 1473 "Inode %llu (%p): i_reserved_data_blocks (%u) not cleared!", 1474 inode->i_ino, EXT4_I(inode), 1475 EXT4_I(inode)->i_reserved_data_blocks); 1476 } 1477 1478 static void ext4_shutdown(struct super_block *sb) 1479 { 1480 ext4_force_shutdown(sb, EXT4_GOING_FLAGS_NOLOGFLUSH); 1481 } 1482 1483 static void init_once(void *foo) 1484 { 1485 struct ext4_inode_info *ei = foo; 1486 1487 INIT_LIST_HEAD(&ei->i_orphan); 1488 init_rwsem(&ei->xattr_sem); 1489 init_rwsem(&ei->i_data_sem); 1490 inode_init_once(&ei->vfs_inode); 1491 ext4_fc_init_inode(&ei->vfs_inode); 1492 #ifdef CONFIG_FS_ENCRYPTION 1493 ei->i_crypt_info = NULL; 1494 #endif 1495 } 1496 1497 static int __init init_inodecache(void) 1498 { 1499 struct kmem_cache_args args = { 1500 .useroffset = offsetof(struct ext4_inode_info, i_data), 1501 .usersize = sizeof_field(struct ext4_inode_info, i_data), 1502 .use_freeptr_offset = true, 1503 .freeptr_offset = offsetof(struct ext4_inode_info, i_flags), 1504 .ctor = init_once, 1505 }; 1506 1507 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache", 1508 sizeof(struct ext4_inode_info), 1509 &args, 1510 SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT); 1511 1512 if (ext4_inode_cachep == NULL) 1513 return -ENOMEM; 1514 return 0; 1515 } 1516 1517 static void destroy_inodecache(void) 1518 { 1519 /* 1520 * Make sure all delayed rcu free inodes are flushed before we 1521 * destroy cache. 1522 */ 1523 rcu_barrier(); 1524 kmem_cache_destroy(ext4_inode_cachep); 1525 } 1526 1527 void ext4_clear_inode(struct inode *inode) 1528 { 1529 ext4_fc_del(inode); 1530 if (!EXT4_SB(inode->i_sb)->s_journal) 1531 mmb_invalidate(&EXT4_I(inode)->i_metadata_bhs); 1532 clear_inode(inode); 1533 ext4_discard_preallocations(inode); 1534 /* 1535 * We must remove the inode from the hash before ext4_free_inode() 1536 * clears the bit in inode bitmap as otherwise another process reusing 1537 * the inode will block in insert_inode_hash() waiting for inode 1538 * eviction to complete while holding transaction handle open, but 1539 * ext4_evict_inode() still running for that inode could block waiting 1540 * for transaction commit if the inode is marked as IS_SYNC => deadlock. 1541 * 1542 * Removing the inode from the hash here is safe. There are two cases 1543 * to consider: 1544 * 1) The inode still has references to it (i_nlink > 0). In that case 1545 * we are keeping the inode and once we remove the inode from the hash, 1546 * iget() can create the new inode structure for the same inode number 1547 * and we are fine with that as all IO on behalf of the inode is 1548 * finished. 1549 * 2) We are deleting the inode (i_nlink == 0). In that case inode 1550 * number cannot be reused until ext4_free_inode() clears the bit in 1551 * the inode bitmap, at which point all IO is done and reuse is fine 1552 * again. 1553 */ 1554 remove_inode_hash(inode); 1555 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 1556 dquot_drop(inode); 1557 if (EXT4_I(inode)->jinode) { 1558 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 1559 EXT4_I(inode)->jinode); 1560 jbd2_free_inode(EXT4_I(inode)->jinode); 1561 EXT4_I(inode)->jinode = NULL; 1562 } 1563 fscrypt_put_encryption_info(inode); 1564 } 1565 1566 static struct inode *ext4_nfs_get_inode(struct super_block *sb, 1567 u64 ino, u32 generation) 1568 { 1569 struct inode *inode; 1570 1571 /* 1572 * Currently we don't know the generation for parent directory, so 1573 * a generation of 0 means "accept any" 1574 */ 1575 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE); 1576 if (IS_ERR(inode)) 1577 return ERR_CAST(inode); 1578 if (generation && inode->i_generation != generation) { 1579 iput(inode); 1580 return ERR_PTR(-ESTALE); 1581 } 1582 1583 return inode; 1584 } 1585 1586 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 1587 int fh_len, int fh_type) 1588 { 1589 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1590 ext4_nfs_get_inode); 1591 } 1592 1593 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 1594 int fh_len, int fh_type) 1595 { 1596 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1597 ext4_nfs_get_inode); 1598 } 1599 1600 static int ext4_nfs_commit_metadata(struct inode *inode) 1601 { 1602 struct writeback_control wbc = { 1603 .sync_mode = WB_SYNC_ALL 1604 }; 1605 1606 trace_ext4_nfs_commit_metadata(inode); 1607 return ext4_write_inode(inode, &wbc); 1608 } 1609 1610 #ifdef CONFIG_QUOTA 1611 static const char * const quotatypes[] = INITQFNAMES; 1612 #define QTYPE2NAME(t) (quotatypes[t]) 1613 1614 static int ext4_write_dquot(struct dquot *dquot); 1615 static int ext4_acquire_dquot(struct dquot *dquot); 1616 static int ext4_release_dquot(struct dquot *dquot); 1617 static int ext4_mark_dquot_dirty(struct dquot *dquot); 1618 static int ext4_write_info(struct super_block *sb, int type); 1619 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1620 const struct path *path); 1621 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1622 size_t len, loff_t off); 1623 static ssize_t ext4_quota_write(struct super_block *sb, int type, 1624 const char *data, size_t len, loff_t off); 1625 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 1626 unsigned int flags); 1627 1628 static struct dquot __rcu **ext4_get_dquots(struct inode *inode) 1629 { 1630 return EXT4_I(inode)->i_dquot; 1631 } 1632 1633 static const struct dquot_operations ext4_quota_operations = { 1634 .get_reserved_space = ext4_get_reserved_space, 1635 .write_dquot = ext4_write_dquot, 1636 .acquire_dquot = ext4_acquire_dquot, 1637 .release_dquot = ext4_release_dquot, 1638 .mark_dirty = ext4_mark_dquot_dirty, 1639 .write_info = ext4_write_info, 1640 .alloc_dquot = dquot_alloc, 1641 .destroy_dquot = dquot_destroy, 1642 .get_projid = ext4_get_projid, 1643 .get_inode_usage = ext4_get_inode_usage, 1644 .get_next_id = dquot_get_next_id, 1645 }; 1646 1647 static const struct quotactl_ops ext4_qctl_operations = { 1648 .quota_on = ext4_quota_on, 1649 .quota_off = ext4_quota_off, 1650 .quota_sync = dquot_quota_sync, 1651 .get_state = dquot_get_state, 1652 .set_info = dquot_set_dqinfo, 1653 .get_dqblk = dquot_get_dqblk, 1654 .set_dqblk = dquot_set_dqblk, 1655 .get_nextdqblk = dquot_get_next_dqblk, 1656 }; 1657 #endif 1658 1659 static const struct super_operations ext4_sops = { 1660 .alloc_inode = ext4_alloc_inode, 1661 .free_inode = ext4_free_in_core_inode, 1662 .destroy_inode = ext4_destroy_inode, 1663 .write_inode = ext4_write_inode, 1664 .dirty_inode = ext4_dirty_inode, 1665 .drop_inode = ext4_drop_inode, 1666 .evict_inode = ext4_evict_inode, 1667 .put_super = ext4_put_super, 1668 .sync_fs = ext4_sync_fs, 1669 .freeze_fs = ext4_freeze, 1670 .unfreeze_fs = ext4_unfreeze, 1671 .statfs = ext4_statfs, 1672 .show_options = ext4_show_options, 1673 .shutdown = ext4_shutdown, 1674 #ifdef CONFIG_QUOTA 1675 .quota_read = ext4_quota_read, 1676 .quota_write = ext4_quota_write, 1677 .get_dquots = ext4_get_dquots, 1678 #endif 1679 }; 1680 1681 static const struct export_operations ext4_export_ops = { 1682 .encode_fh = generic_encode_ino32_fh, 1683 .fh_to_dentry = ext4_fh_to_dentry, 1684 .fh_to_parent = ext4_fh_to_parent, 1685 .get_parent = ext4_get_parent, 1686 .commit_metadata = ext4_nfs_commit_metadata, 1687 }; 1688 1689 enum { 1690 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1691 Opt_resgid, Opt_resuid, Opt_sb, 1692 Opt_nouid32, Opt_debug, Opt_removed, 1693 Opt_user_xattr, Opt_acl, 1694 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1695 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1696 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1697 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1698 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 1699 Opt_inlinecrypt, 1700 Opt_usrjquota, Opt_grpjquota, Opt_quota, 1701 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1702 Opt_usrquota, Opt_grpquota, Opt_prjquota, 1703 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never, 1704 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error, 1705 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize, 1706 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 1707 Opt_inode_readahead_blks, Opt_journal_ioprio, 1708 Opt_dioread_nolock, Opt_dioread_lock, 1709 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1710 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, 1711 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan, 1712 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type, 1713 #ifdef CONFIG_EXT4_DEBUG 1714 Opt_fc_debug_max_replay, Opt_fc_debug_force 1715 #endif 1716 }; 1717 1718 static const struct constant_table ext4_param_errors[] = { 1719 {"continue", EXT4_MOUNT_ERRORS_CONT}, 1720 {"panic", EXT4_MOUNT_ERRORS_PANIC}, 1721 {"remount-ro", EXT4_MOUNT_ERRORS_RO}, 1722 {} 1723 }; 1724 1725 static const struct constant_table ext4_param_data[] = { 1726 {"journal", EXT4_MOUNT_JOURNAL_DATA}, 1727 {"ordered", EXT4_MOUNT_ORDERED_DATA}, 1728 {"writeback", EXT4_MOUNT_WRITEBACK_DATA}, 1729 {} 1730 }; 1731 1732 static const struct constant_table ext4_param_data_err[] = { 1733 {"abort", Opt_data_err_abort}, 1734 {"ignore", Opt_data_err_ignore}, 1735 {} 1736 }; 1737 1738 static const struct constant_table ext4_param_jqfmt[] = { 1739 {"vfsold", QFMT_VFS_OLD}, 1740 {"vfsv0", QFMT_VFS_V0}, 1741 {"vfsv1", QFMT_VFS_V1}, 1742 {} 1743 }; 1744 1745 static const struct constant_table ext4_param_dax[] = { 1746 {"always", Opt_dax_always}, 1747 {"inode", Opt_dax_inode}, 1748 {"never", Opt_dax_never}, 1749 {} 1750 }; 1751 1752 /* 1753 * Mount option specification 1754 * We don't use fsparam_flag_no because of the way we set the 1755 * options and the way we show them in _ext4_show_options(). To 1756 * keep the changes to a minimum, let's keep the negative options 1757 * separate for now. 1758 */ 1759 static const struct fs_parameter_spec ext4_param_specs[] = { 1760 fsparam_flag ("bsddf", Opt_bsd_df), 1761 fsparam_flag ("minixdf", Opt_minix_df), 1762 fsparam_flag ("grpid", Opt_grpid), 1763 fsparam_flag ("bsdgroups", Opt_grpid), 1764 fsparam_flag ("nogrpid", Opt_nogrpid), 1765 fsparam_flag ("sysvgroups", Opt_nogrpid), 1766 fsparam_gid ("resgid", Opt_resgid), 1767 fsparam_uid ("resuid", Opt_resuid), 1768 fsparam_u32 ("sb", Opt_sb), 1769 fsparam_enum ("errors", Opt_errors, ext4_param_errors), 1770 fsparam_flag ("nouid32", Opt_nouid32), 1771 fsparam_flag ("debug", Opt_debug), 1772 fsparam_flag ("oldalloc", Opt_removed), 1773 fsparam_flag ("orlov", Opt_removed), 1774 fsparam_flag ("user_xattr", Opt_user_xattr), 1775 fsparam_flag ("acl", Opt_acl), 1776 fsparam_flag ("norecovery", Opt_noload), 1777 fsparam_flag ("noload", Opt_noload), 1778 fsparam_flag ("bh", Opt_removed), 1779 fsparam_flag ("nobh", Opt_removed), 1780 fsparam_u32 ("commit", Opt_commit), 1781 fsparam_u32 ("min_batch_time", Opt_min_batch_time), 1782 fsparam_u32 ("max_batch_time", Opt_max_batch_time), 1783 fsparam_u32 ("journal_dev", Opt_journal_dev), 1784 fsparam_bdev ("journal_path", Opt_journal_path), 1785 fsparam_flag ("journal_checksum", Opt_journal_checksum), 1786 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum), 1787 fsparam_flag ("journal_async_commit",Opt_journal_async_commit), 1788 fsparam_flag ("abort", Opt_abort), 1789 fsparam_enum ("data", Opt_data, ext4_param_data), 1790 fsparam_enum ("data_err", Opt_data_err, 1791 ext4_param_data_err), 1792 fsparam_string_empty 1793 ("usrjquota", Opt_usrjquota), 1794 fsparam_string_empty 1795 ("grpjquota", Opt_grpjquota), 1796 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt), 1797 fsparam_flag ("grpquota", Opt_grpquota), 1798 fsparam_flag ("quota", Opt_quota), 1799 fsparam_flag ("noquota", Opt_noquota), 1800 fsparam_flag ("usrquota", Opt_usrquota), 1801 fsparam_flag ("prjquota", Opt_prjquota), 1802 fsparam_flag ("barrier", Opt_barrier), 1803 fsparam_u32 ("barrier", Opt_barrier), 1804 fsparam_flag ("nobarrier", Opt_nobarrier), 1805 fsparam_flag ("i_version", Opt_removed), 1806 fsparam_flag ("dax", Opt_dax), 1807 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax), 1808 fsparam_u32 ("stripe", Opt_stripe), 1809 fsparam_flag ("delalloc", Opt_delalloc), 1810 fsparam_flag ("nodelalloc", Opt_nodelalloc), 1811 fsparam_flag ("warn_on_error", Opt_warn_on_error), 1812 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error), 1813 fsparam_u32 ("debug_want_extra_isize", 1814 Opt_debug_want_extra_isize), 1815 fsparam_flag ("mblk_io_submit", Opt_removed), 1816 fsparam_flag ("nomblk_io_submit", Opt_removed), 1817 fsparam_flag ("block_validity", Opt_block_validity), 1818 fsparam_flag ("noblock_validity", Opt_noblock_validity), 1819 fsparam_u32 ("inode_readahead_blks", 1820 Opt_inode_readahead_blks), 1821 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio), 1822 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc), 1823 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc), 1824 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc), 1825 fsparam_flag ("dioread_nolock", Opt_dioread_nolock), 1826 fsparam_flag ("nodioread_nolock", Opt_dioread_lock), 1827 fsparam_flag ("dioread_lock", Opt_dioread_lock), 1828 fsparam_flag ("discard", Opt_discard), 1829 fsparam_flag ("nodiscard", Opt_nodiscard), 1830 fsparam_u32 ("init_itable", Opt_init_itable), 1831 fsparam_flag ("init_itable", Opt_init_itable), 1832 fsparam_flag ("noinit_itable", Opt_noinit_itable), 1833 #ifdef CONFIG_EXT4_DEBUG 1834 fsparam_flag ("fc_debug_force", Opt_fc_debug_force), 1835 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay), 1836 #endif 1837 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb), 1838 fsparam_flag ("test_dummy_encryption", 1839 Opt_test_dummy_encryption), 1840 fsparam_string ("test_dummy_encryption", 1841 Opt_test_dummy_encryption), 1842 fsparam_flag ("inlinecrypt", Opt_inlinecrypt), 1843 fsparam_flag ("nombcache", Opt_nombcache), 1844 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */ 1845 fsparam_flag ("prefetch_block_bitmaps", 1846 Opt_removed), 1847 fsparam_flag ("no_prefetch_block_bitmaps", 1848 Opt_no_prefetch_block_bitmaps), 1849 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan), 1850 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */ 1851 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */ 1852 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */ 1853 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */ 1854 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */ 1855 {} 1856 }; 1857 1858 1859 #define MOPT_SET 0x0001 1860 #define MOPT_CLEAR 0x0002 1861 #define MOPT_NOSUPPORT 0x0004 1862 #define MOPT_EXPLICIT 0x0008 1863 #ifdef CONFIG_QUOTA 1864 #define MOPT_Q 0 1865 #define MOPT_QFMT 0x0010 1866 #else 1867 #define MOPT_Q MOPT_NOSUPPORT 1868 #define MOPT_QFMT MOPT_NOSUPPORT 1869 #endif 1870 #define MOPT_NO_EXT2 0x0020 1871 #define MOPT_NO_EXT3 0x0040 1872 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1873 #define MOPT_SKIP 0x0080 1874 #define MOPT_2 0x0100 1875 1876 static const struct mount_opts { 1877 int token; 1878 int mount_opt; 1879 int flags; 1880 } ext4_mount_opts[] = { 1881 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 1882 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 1883 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 1884 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 1885 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 1886 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 1887 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 1888 MOPT_EXT4_ONLY | MOPT_SET}, 1889 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 1890 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1891 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 1892 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 1893 {Opt_delalloc, EXT4_MOUNT_DELALLOC, 1894 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1895 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 1896 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1897 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET}, 1898 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR}, 1899 {Opt_commit, 0, MOPT_NO_EXT2}, 1900 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1901 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1902 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1903 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1904 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 1905 EXT4_MOUNT_JOURNAL_CHECKSUM), 1906 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1907 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 1908 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2}, 1909 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 1910 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 1911 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 1912 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 1913 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 1914 {Opt_dax_type, 0, MOPT_EXT4_ONLY}, 1915 {Opt_journal_dev, 0, MOPT_NO_EXT2}, 1916 {Opt_journal_path, 0, MOPT_NO_EXT2}, 1917 {Opt_journal_ioprio, 0, MOPT_NO_EXT2}, 1918 {Opt_data, 0, MOPT_NO_EXT2}, 1919 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 1920 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1921 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 1922 #else 1923 {Opt_acl, 0, MOPT_NOSUPPORT}, 1924 #endif 1925 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 1926 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1927 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 1928 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 1929 MOPT_SET | MOPT_Q}, 1930 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 1931 MOPT_SET | MOPT_Q}, 1932 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, 1933 MOPT_SET | MOPT_Q}, 1934 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 1935 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), 1936 MOPT_CLEAR | MOPT_Q}, 1937 {Opt_usrjquota, 0, MOPT_Q}, 1938 {Opt_grpjquota, 0, MOPT_Q}, 1939 {Opt_jqfmt, 0, MOPT_QFMT}, 1940 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, 1941 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS, 1942 MOPT_SET}, 1943 #ifdef CONFIG_EXT4_DEBUG 1944 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT, 1945 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY}, 1946 #endif 1947 {Opt_abort, EXT4_MOUNT2_ABORT, MOPT_SET | MOPT_2}, 1948 {Opt_err, 0, 0} 1949 }; 1950 1951 #if IS_ENABLED(CONFIG_UNICODE) 1952 static const struct ext4_sb_encodings { 1953 __u16 magic; 1954 char *name; 1955 unsigned int version; 1956 } ext4_sb_encoding_map[] = { 1957 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)}, 1958 }; 1959 1960 static const struct ext4_sb_encodings * 1961 ext4_sb_read_encoding(const struct ext4_super_block *es) 1962 { 1963 __u16 magic = le16_to_cpu(es->s_encoding); 1964 int i; 1965 1966 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++) 1967 if (magic == ext4_sb_encoding_map[i].magic) 1968 return &ext4_sb_encoding_map[i]; 1969 1970 return NULL; 1971 } 1972 #endif 1973 1974 #define EXT4_SPEC_JQUOTA (1 << 0) 1975 #define EXT4_SPEC_JQFMT (1 << 1) 1976 #define EXT4_SPEC_DATAJ (1 << 2) 1977 #define EXT4_SPEC_SB_BLOCK (1 << 3) 1978 #define EXT4_SPEC_JOURNAL_DEV (1 << 4) 1979 #define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5) 1980 #define EXT4_SPEC_s_want_extra_isize (1 << 7) 1981 #define EXT4_SPEC_s_max_batch_time (1 << 8) 1982 #define EXT4_SPEC_s_min_batch_time (1 << 9) 1983 #define EXT4_SPEC_s_inode_readahead_blks (1 << 10) 1984 #define EXT4_SPEC_s_li_wait_mult (1 << 11) 1985 #define EXT4_SPEC_s_max_dir_size_kb (1 << 12) 1986 #define EXT4_SPEC_s_stripe (1 << 13) 1987 #define EXT4_SPEC_s_resuid (1 << 14) 1988 #define EXT4_SPEC_s_resgid (1 << 15) 1989 #define EXT4_SPEC_s_commit_interval (1 << 16) 1990 #define EXT4_SPEC_s_fc_debug_max_replay (1 << 17) 1991 #define EXT4_SPEC_s_sb_block (1 << 18) 1992 #define EXT4_SPEC_mb_optimize_scan (1 << 19) 1993 1994 struct ext4_fs_context { 1995 char *s_qf_names[EXT4_MAXQUOTAS]; 1996 struct fscrypt_dummy_policy dummy_enc_policy; 1997 int s_jquota_fmt; /* Format of quota to use */ 1998 #ifdef CONFIG_EXT4_DEBUG 1999 int s_fc_debug_max_replay; 2000 #endif 2001 unsigned short qname_spec; 2002 unsigned long vals_s_flags; /* Bits to set in s_flags */ 2003 unsigned long mask_s_flags; /* Bits changed in s_flags */ 2004 unsigned long journal_devnum; 2005 unsigned long s_commit_interval; 2006 unsigned long s_stripe; 2007 unsigned int s_inode_readahead_blks; 2008 unsigned int s_want_extra_isize; 2009 unsigned int s_li_wait_mult; 2010 unsigned int s_max_dir_size_kb; 2011 unsigned int journal_ioprio; 2012 unsigned int vals_s_mount_opt; 2013 unsigned int mask_s_mount_opt; 2014 unsigned int vals_s_mount_opt2; 2015 unsigned int mask_s_mount_opt2; 2016 unsigned int opt_flags; /* MOPT flags */ 2017 unsigned int spec; 2018 u32 s_max_batch_time; 2019 u32 s_min_batch_time; 2020 kuid_t s_resuid; 2021 kgid_t s_resgid; 2022 ext4_fsblk_t s_sb_block; 2023 }; 2024 2025 static void ext4_fc_free(struct fs_context *fc) 2026 { 2027 struct ext4_fs_context *ctx = fc->fs_private; 2028 int i; 2029 2030 if (!ctx) 2031 return; 2032 2033 for (i = 0; i < EXT4_MAXQUOTAS; i++) 2034 kfree(ctx->s_qf_names[i]); 2035 2036 fscrypt_free_dummy_policy(&ctx->dummy_enc_policy); 2037 kfree(ctx); 2038 } 2039 2040 int ext4_init_fs_context(struct fs_context *fc) 2041 { 2042 struct ext4_fs_context *ctx; 2043 2044 ctx = kzalloc_obj(struct ext4_fs_context); 2045 if (!ctx) 2046 return -ENOMEM; 2047 2048 fc->fs_private = ctx; 2049 fc->ops = &ext4_context_ops; 2050 2051 /* i_version is always enabled now */ 2052 fc->sb_flags |= SB_I_VERSION; 2053 2054 return 0; 2055 } 2056 2057 #ifdef CONFIG_QUOTA 2058 /* 2059 * Note the name of the specified quota file. 2060 */ 2061 static int note_qf_name(struct fs_context *fc, int qtype, 2062 struct fs_parameter *param) 2063 { 2064 struct ext4_fs_context *ctx = fc->fs_private; 2065 char *qname; 2066 2067 if (param->size < 1) { 2068 ext4_msg(NULL, KERN_ERR, "Missing quota name"); 2069 return -EINVAL; 2070 } 2071 if (strchr(param->string, '/')) { 2072 ext4_msg(NULL, KERN_ERR, 2073 "quotafile must be on filesystem root"); 2074 return -EINVAL; 2075 } 2076 if (ctx->s_qf_names[qtype]) { 2077 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) { 2078 ext4_msg(NULL, KERN_ERR, 2079 "%s quota file already specified", 2080 QTYPE2NAME(qtype)); 2081 return -EINVAL; 2082 } 2083 return 0; 2084 } 2085 2086 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL); 2087 if (!qname) { 2088 ext4_msg(NULL, KERN_ERR, 2089 "Not enough memory for storing quotafile name"); 2090 return -ENOMEM; 2091 } 2092 ctx->s_qf_names[qtype] = qname; 2093 ctx->qname_spec |= 1 << qtype; 2094 ctx->spec |= EXT4_SPEC_JQUOTA; 2095 return 0; 2096 } 2097 2098 /* 2099 * Clear the name of the specified quota file. 2100 */ 2101 static int unnote_qf_name(struct fs_context *fc, int qtype) 2102 { 2103 struct ext4_fs_context *ctx = fc->fs_private; 2104 2105 kfree(ctx->s_qf_names[qtype]); 2106 2107 ctx->s_qf_names[qtype] = NULL; 2108 ctx->qname_spec |= 1 << qtype; 2109 ctx->spec |= EXT4_SPEC_JQUOTA; 2110 return 0; 2111 } 2112 #endif 2113 2114 static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param, 2115 struct ext4_fs_context *ctx) 2116 { 2117 int err; 2118 2119 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) { 2120 ext4_msg(NULL, KERN_WARNING, 2121 "test_dummy_encryption option not supported"); 2122 return -EINVAL; 2123 } 2124 err = fscrypt_parse_test_dummy_encryption(param, 2125 &ctx->dummy_enc_policy); 2126 if (err == -EINVAL) { 2127 ext4_msg(NULL, KERN_WARNING, 2128 "Value of option \"%s\" is unrecognized", param->key); 2129 } else if (err == -EEXIST) { 2130 ext4_msg(NULL, KERN_WARNING, 2131 "Conflicting test_dummy_encryption options"); 2132 return -EINVAL; 2133 } 2134 return err; 2135 } 2136 2137 #define EXT4_SET_CTX(name) \ 2138 static inline __maybe_unused \ 2139 void ctx_set_##name(struct ext4_fs_context *ctx, unsigned long flag) \ 2140 { \ 2141 ctx->mask_s_##name |= flag; \ 2142 ctx->vals_s_##name |= flag; \ 2143 } 2144 2145 #define EXT4_CLEAR_CTX(name) \ 2146 static inline __maybe_unused \ 2147 void ctx_clear_##name(struct ext4_fs_context *ctx, unsigned long flag) \ 2148 { \ 2149 ctx->mask_s_##name |= flag; \ 2150 ctx->vals_s_##name &= ~flag; \ 2151 } 2152 2153 #define EXT4_TEST_CTX(name) \ 2154 static inline unsigned long \ 2155 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \ 2156 { \ 2157 return (ctx->vals_s_##name & flag); \ 2158 } 2159 2160 EXT4_SET_CTX(flags); /* set only */ 2161 EXT4_SET_CTX(mount_opt); 2162 EXT4_CLEAR_CTX(mount_opt); 2163 EXT4_TEST_CTX(mount_opt); 2164 EXT4_SET_CTX(mount_opt2); 2165 EXT4_CLEAR_CTX(mount_opt2); 2166 EXT4_TEST_CTX(mount_opt2); 2167 2168 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param) 2169 { 2170 struct ext4_fs_context *ctx = fc->fs_private; 2171 struct fs_parse_result result; 2172 const struct mount_opts *m; 2173 int is_remount; 2174 int token; 2175 2176 token = fs_parse(fc, ext4_param_specs, param, &result); 2177 if (token < 0) 2178 return token; 2179 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE; 2180 2181 for (m = ext4_mount_opts; m->token != Opt_err; m++) 2182 if (token == m->token) 2183 break; 2184 2185 ctx->opt_flags |= m->flags; 2186 2187 if (m->flags & MOPT_EXPLICIT) { 2188 if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 2189 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC); 2190 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 2191 ctx_set_mount_opt2(ctx, 2192 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM); 2193 } else 2194 return -EINVAL; 2195 } 2196 2197 if (m->flags & MOPT_NOSUPPORT) { 2198 ext4_msg(NULL, KERN_ERR, "%s option not supported", 2199 param->key); 2200 return 0; 2201 } 2202 2203 switch (token) { 2204 #ifdef CONFIG_QUOTA 2205 case Opt_usrjquota: 2206 if (!*param->string) 2207 return unnote_qf_name(fc, USRQUOTA); 2208 else 2209 return note_qf_name(fc, USRQUOTA, param); 2210 case Opt_grpjquota: 2211 if (!*param->string) 2212 return unnote_qf_name(fc, GRPQUOTA); 2213 else 2214 return note_qf_name(fc, GRPQUOTA, param); 2215 #endif 2216 case Opt_sb: 2217 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) { 2218 ext4_msg(NULL, KERN_WARNING, 2219 "Ignoring %s option on remount", param->key); 2220 } else { 2221 ctx->s_sb_block = result.uint_32; 2222 ctx->spec |= EXT4_SPEC_s_sb_block; 2223 } 2224 return 0; 2225 case Opt_removed: 2226 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option", 2227 param->key); 2228 return 0; 2229 case Opt_inlinecrypt: 2230 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 2231 ctx_set_flags(ctx, SB_INLINECRYPT); 2232 #else 2233 ext4_msg(NULL, KERN_ERR, "inline encryption not supported"); 2234 #endif 2235 return 0; 2236 case Opt_errors: 2237 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK); 2238 ctx_set_mount_opt(ctx, result.uint_32); 2239 return 0; 2240 #ifdef CONFIG_QUOTA 2241 case Opt_jqfmt: 2242 ctx->s_jquota_fmt = result.uint_32; 2243 ctx->spec |= EXT4_SPEC_JQFMT; 2244 return 0; 2245 #endif 2246 case Opt_data: 2247 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS); 2248 ctx_set_mount_opt(ctx, result.uint_32); 2249 ctx->spec |= EXT4_SPEC_DATAJ; 2250 return 0; 2251 case Opt_commit: 2252 if (result.uint_32 == 0) 2253 result.uint_32 = JBD2_DEFAULT_MAX_COMMIT_AGE; 2254 else if (result.uint_32 > INT_MAX / HZ) { 2255 ext4_msg(NULL, KERN_ERR, 2256 "Invalid commit interval %d, " 2257 "must be smaller than %d", 2258 result.uint_32, INT_MAX / HZ); 2259 return -EINVAL; 2260 } 2261 ctx->s_commit_interval = HZ * result.uint_32; 2262 ctx->spec |= EXT4_SPEC_s_commit_interval; 2263 return 0; 2264 case Opt_debug_want_extra_isize: 2265 if ((result.uint_32 & 1) || (result.uint_32 < 4)) { 2266 ext4_msg(NULL, KERN_ERR, 2267 "Invalid want_extra_isize %d", result.uint_32); 2268 return -EINVAL; 2269 } 2270 ctx->s_want_extra_isize = result.uint_32; 2271 ctx->spec |= EXT4_SPEC_s_want_extra_isize; 2272 return 0; 2273 case Opt_max_batch_time: 2274 ctx->s_max_batch_time = result.uint_32; 2275 ctx->spec |= EXT4_SPEC_s_max_batch_time; 2276 return 0; 2277 case Opt_min_batch_time: 2278 ctx->s_min_batch_time = result.uint_32; 2279 ctx->spec |= EXT4_SPEC_s_min_batch_time; 2280 return 0; 2281 case Opt_inode_readahead_blks: 2282 if (result.uint_32 && 2283 (result.uint_32 > (1 << 30) || 2284 !is_power_of_2(result.uint_32))) { 2285 ext4_msg(NULL, KERN_ERR, 2286 "EXT4-fs: inode_readahead_blks must be " 2287 "0 or a power of 2 smaller than 2^31"); 2288 return -EINVAL; 2289 } 2290 ctx->s_inode_readahead_blks = result.uint_32; 2291 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks; 2292 return 0; 2293 case Opt_init_itable: 2294 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE); 2295 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 2296 if (param->type == fs_value_is_string) 2297 ctx->s_li_wait_mult = result.uint_32; 2298 ctx->spec |= EXT4_SPEC_s_li_wait_mult; 2299 return 0; 2300 case Opt_max_dir_size_kb: 2301 ctx->s_max_dir_size_kb = result.uint_32; 2302 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb; 2303 return 0; 2304 #ifdef CONFIG_EXT4_DEBUG 2305 case Opt_fc_debug_max_replay: 2306 ctx->s_fc_debug_max_replay = result.uint_32; 2307 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay; 2308 return 0; 2309 #endif 2310 case Opt_stripe: 2311 ctx->s_stripe = result.uint_32; 2312 ctx->spec |= EXT4_SPEC_s_stripe; 2313 return 0; 2314 case Opt_resuid: 2315 ctx->s_resuid = result.uid; 2316 ctx->spec |= EXT4_SPEC_s_resuid; 2317 return 0; 2318 case Opt_resgid: 2319 ctx->s_resgid = result.gid; 2320 ctx->spec |= EXT4_SPEC_s_resgid; 2321 return 0; 2322 case Opt_journal_dev: 2323 if (is_remount) { 2324 ext4_msg(NULL, KERN_ERR, 2325 "Cannot specify journal on remount"); 2326 return -EINVAL; 2327 } 2328 ctx->journal_devnum = result.uint_32; 2329 ctx->spec |= EXT4_SPEC_JOURNAL_DEV; 2330 return 0; 2331 case Opt_journal_path: 2332 { 2333 struct inode *journal_inode; 2334 struct path path; 2335 int error; 2336 2337 if (is_remount) { 2338 ext4_msg(NULL, KERN_ERR, 2339 "Cannot specify journal on remount"); 2340 return -EINVAL; 2341 } 2342 2343 error = fs_lookup_param(fc, param, 1, LOOKUP_FOLLOW, &path); 2344 if (error) { 2345 ext4_msg(NULL, KERN_ERR, "error: could not find " 2346 "journal device path"); 2347 return -EINVAL; 2348 } 2349 2350 journal_inode = d_inode(path.dentry); 2351 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev); 2352 ctx->spec |= EXT4_SPEC_JOURNAL_DEV; 2353 path_put(&path); 2354 return 0; 2355 } 2356 case Opt_journal_ioprio: 2357 if (result.uint_32 > 7) { 2358 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority" 2359 " (must be 0-7)"); 2360 return -EINVAL; 2361 } 2362 ctx->journal_ioprio = 2363 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32); 2364 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO; 2365 return 0; 2366 case Opt_test_dummy_encryption: 2367 return ext4_parse_test_dummy_encryption(param, ctx); 2368 case Opt_dax: 2369 case Opt_dax_type: 2370 #ifdef CONFIG_FS_DAX 2371 { 2372 int type = (token == Opt_dax) ? 2373 Opt_dax : result.uint_32; 2374 2375 switch (type) { 2376 case Opt_dax: 2377 case Opt_dax_always: 2378 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); 2379 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); 2380 break; 2381 case Opt_dax_never: 2382 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); 2383 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); 2384 break; 2385 case Opt_dax_inode: 2386 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); 2387 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); 2388 /* Strictly for printing options */ 2389 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE); 2390 break; 2391 } 2392 return 0; 2393 } 2394 #else 2395 ext4_msg(NULL, KERN_INFO, "dax option not supported"); 2396 return -EINVAL; 2397 #endif 2398 case Opt_data_err: 2399 if (result.uint_32 == Opt_data_err_abort) 2400 ctx_set_mount_opt(ctx, m->mount_opt); 2401 else if (result.uint_32 == Opt_data_err_ignore) 2402 ctx_clear_mount_opt(ctx, m->mount_opt); 2403 return 0; 2404 case Opt_mb_optimize_scan: 2405 if (result.int_32 == 1) { 2406 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN); 2407 ctx->spec |= EXT4_SPEC_mb_optimize_scan; 2408 } else if (result.int_32 == 0) { 2409 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN); 2410 ctx->spec |= EXT4_SPEC_mb_optimize_scan; 2411 } else { 2412 ext4_msg(NULL, KERN_WARNING, 2413 "mb_optimize_scan should be set to 0 or 1."); 2414 return -EINVAL; 2415 } 2416 return 0; 2417 } 2418 2419 /* 2420 * At this point we should only be getting options requiring MOPT_SET, 2421 * or MOPT_CLEAR. Anything else is a bug 2422 */ 2423 if (m->token == Opt_err) { 2424 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s", 2425 param->key); 2426 WARN_ON(1); 2427 return -EINVAL; 2428 } 2429 2430 else { 2431 unsigned int set = 0; 2432 2433 if ((param->type == fs_value_is_flag) || 2434 result.uint_32 > 0) 2435 set = 1; 2436 2437 if (m->flags & MOPT_CLEAR) 2438 set = !set; 2439 else if (unlikely(!(m->flags & MOPT_SET))) { 2440 ext4_msg(NULL, KERN_WARNING, 2441 "buggy handling of option %s", 2442 param->key); 2443 WARN_ON(1); 2444 return -EINVAL; 2445 } 2446 if (m->flags & MOPT_2) { 2447 if (set != 0) 2448 ctx_set_mount_opt2(ctx, m->mount_opt); 2449 else 2450 ctx_clear_mount_opt2(ctx, m->mount_opt); 2451 } else { 2452 if (set != 0) 2453 ctx_set_mount_opt(ctx, m->mount_opt); 2454 else 2455 ctx_clear_mount_opt(ctx, m->mount_opt); 2456 } 2457 } 2458 2459 return 0; 2460 } 2461 2462 static int parse_options(struct fs_context *fc, char *options) 2463 { 2464 struct fs_parameter param; 2465 int ret; 2466 char *key; 2467 2468 if (!options) 2469 return 0; 2470 2471 while ((key = strsep(&options, ",")) != NULL) { 2472 if (*key) { 2473 size_t v_len = 0; 2474 char *value = strchr(key, '='); 2475 2476 param.type = fs_value_is_flag; 2477 param.string = NULL; 2478 2479 if (value) { 2480 if (value == key) 2481 continue; 2482 2483 *value++ = 0; 2484 v_len = strlen(value); 2485 param.string = kmemdup_nul(value, v_len, 2486 GFP_KERNEL); 2487 if (!param.string) 2488 return -ENOMEM; 2489 param.type = fs_value_is_string; 2490 } 2491 2492 param.key = key; 2493 param.size = v_len; 2494 2495 ret = ext4_parse_param(fc, ¶m); 2496 kfree(param.string); 2497 if (ret < 0) 2498 return ret; 2499 } 2500 } 2501 2502 ret = ext4_validate_options(fc); 2503 if (ret < 0) 2504 return ret; 2505 2506 return 0; 2507 } 2508 2509 static int parse_apply_sb_mount_options(struct super_block *sb, 2510 struct ext4_fs_context *m_ctx) 2511 { 2512 struct ext4_sb_info *sbi = EXT4_SB(sb); 2513 char s_mount_opts[64]; 2514 struct ext4_fs_context *s_ctx = NULL; 2515 struct fs_context *fc = NULL; 2516 int ret = -ENOMEM; 2517 2518 if (!sbi->s_es->s_mount_opts[0]) 2519 return 0; 2520 2521 if (strscpy_pad(s_mount_opts, sbi->s_es->s_mount_opts) < 0) 2522 return -E2BIG; 2523 2524 fc = kzalloc_obj(struct fs_context); 2525 if (!fc) 2526 return -ENOMEM; 2527 2528 s_ctx = kzalloc_obj(struct ext4_fs_context); 2529 if (!s_ctx) 2530 goto out_free; 2531 2532 fc->fs_private = s_ctx; 2533 fc->s_fs_info = sbi; 2534 2535 ret = parse_options(fc, s_mount_opts); 2536 if (ret < 0) 2537 goto parse_failed; 2538 2539 ret = ext4_check_opt_consistency(fc, sb); 2540 if (ret < 0) { 2541 parse_failed: 2542 ext4_msg(sb, KERN_WARNING, 2543 "failed to parse options in superblock: %s", 2544 s_mount_opts); 2545 ret = 0; 2546 goto out_free; 2547 } 2548 2549 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV) 2550 m_ctx->journal_devnum = s_ctx->journal_devnum; 2551 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO) 2552 m_ctx->journal_ioprio = s_ctx->journal_ioprio; 2553 2554 ext4_apply_options(fc, sb); 2555 ret = 0; 2556 2557 out_free: 2558 ext4_fc_free(fc); 2559 kfree(fc); 2560 return ret; 2561 } 2562 2563 static void ext4_apply_quota_options(struct fs_context *fc, 2564 struct super_block *sb) 2565 { 2566 #ifdef CONFIG_QUOTA 2567 bool quota_feature = ext4_has_feature_quota(sb); 2568 struct ext4_fs_context *ctx = fc->fs_private; 2569 struct ext4_sb_info *sbi = EXT4_SB(sb); 2570 char *qname; 2571 int i; 2572 2573 if (quota_feature) 2574 return; 2575 2576 if (ctx->spec & EXT4_SPEC_JQUOTA) { 2577 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2578 if (!(ctx->qname_spec & (1 << i))) 2579 continue; 2580 2581 qname = ctx->s_qf_names[i]; /* May be NULL */ 2582 if (qname) 2583 set_opt(sb, QUOTA); 2584 ctx->s_qf_names[i] = NULL; 2585 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname, 2586 lockdep_is_held(&sb->s_umount)); 2587 if (qname) 2588 kfree_rcu_mightsleep(qname); 2589 } 2590 } 2591 2592 if (ctx->spec & EXT4_SPEC_JQFMT) 2593 sbi->s_jquota_fmt = ctx->s_jquota_fmt; 2594 #endif 2595 } 2596 2597 /* 2598 * Check quota settings consistency. 2599 */ 2600 static int ext4_check_quota_consistency(struct fs_context *fc, 2601 struct super_block *sb) 2602 { 2603 #ifdef CONFIG_QUOTA 2604 struct ext4_fs_context *ctx = fc->fs_private; 2605 struct ext4_sb_info *sbi = EXT4_SB(sb); 2606 bool quota_feature = ext4_has_feature_quota(sb); 2607 bool quota_loaded = sb_any_quota_loaded(sb); 2608 bool usr_qf_name, grp_qf_name, usrquota, grpquota; 2609 int quota_flags, i; 2610 2611 /* 2612 * We do the test below only for project quotas. 'usrquota' and 2613 * 'grpquota' mount options are allowed even without quota feature 2614 * to support legacy quotas in quota files. 2615 */ 2616 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) && 2617 !ext4_has_feature_project(sb)) { 2618 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. " 2619 "Cannot enable project quota enforcement."); 2620 return -EINVAL; 2621 } 2622 2623 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 2624 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA; 2625 if (quota_loaded && 2626 ctx->mask_s_mount_opt & quota_flags && 2627 !ctx_test_mount_opt(ctx, quota_flags)) 2628 goto err_quota_change; 2629 2630 if (ctx->spec & EXT4_SPEC_JQUOTA) { 2631 2632 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2633 if (!(ctx->qname_spec & (1 << i))) 2634 continue; 2635 2636 if (quota_loaded && 2637 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i]) 2638 goto err_jquota_change; 2639 2640 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] && 2641 strcmp(get_qf_name(sb, sbi, i), 2642 ctx->s_qf_names[i]) != 0) 2643 goto err_jquota_specified; 2644 } 2645 2646 if (quota_feature) { 2647 ext4_msg(NULL, KERN_INFO, 2648 "Journaled quota options ignored when " 2649 "QUOTA feature is enabled"); 2650 return 0; 2651 } 2652 } 2653 2654 if (ctx->spec & EXT4_SPEC_JQFMT) { 2655 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded) 2656 goto err_jquota_change; 2657 if (quota_feature) { 2658 ext4_msg(NULL, KERN_INFO, "Quota format mount options " 2659 "ignored when QUOTA feature is enabled"); 2660 return 0; 2661 } 2662 } 2663 2664 /* Make sure we don't mix old and new quota format */ 2665 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) || 2666 ctx->s_qf_names[USRQUOTA]); 2667 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) || 2668 ctx->s_qf_names[GRPQUOTA]); 2669 2670 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) || 2671 test_opt(sb, USRQUOTA)); 2672 2673 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) || 2674 test_opt(sb, GRPQUOTA)); 2675 2676 if (usr_qf_name) { 2677 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA); 2678 usrquota = false; 2679 } 2680 if (grp_qf_name) { 2681 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA); 2682 grpquota = false; 2683 } 2684 2685 if (usr_qf_name || grp_qf_name) { 2686 if (usrquota || grpquota) { 2687 ext4_msg(NULL, KERN_ERR, "old and new quota " 2688 "format mixing"); 2689 return -EINVAL; 2690 } 2691 2692 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) { 2693 ext4_msg(NULL, KERN_ERR, "journaled quota format " 2694 "not specified"); 2695 return -EINVAL; 2696 } 2697 } 2698 2699 return 0; 2700 2701 err_quota_change: 2702 ext4_msg(NULL, KERN_ERR, 2703 "Cannot change quota options when quota turned on"); 2704 return -EINVAL; 2705 err_jquota_change: 2706 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota " 2707 "options when quota turned on"); 2708 return -EINVAL; 2709 err_jquota_specified: 2710 ext4_msg(NULL, KERN_ERR, "%s quota file already specified", 2711 QTYPE2NAME(i)); 2712 return -EINVAL; 2713 #else 2714 return 0; 2715 #endif 2716 } 2717 2718 static int ext4_check_test_dummy_encryption(const struct fs_context *fc, 2719 struct super_block *sb) 2720 { 2721 const struct ext4_fs_context *ctx = fc->fs_private; 2722 const struct ext4_sb_info *sbi = EXT4_SB(sb); 2723 2724 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy)) 2725 return 0; 2726 2727 if (!ext4_has_feature_encrypt(sb)) { 2728 ext4_msg(NULL, KERN_WARNING, 2729 "test_dummy_encryption requires encrypt feature"); 2730 return -EINVAL; 2731 } 2732 /* 2733 * This mount option is just for testing, and it's not worthwhile to 2734 * implement the extra complexity (e.g. RCU protection) that would be 2735 * needed to allow it to be set or changed during remount. We do allow 2736 * it to be specified during remount, but only if there is no change. 2737 */ 2738 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) { 2739 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy, 2740 &ctx->dummy_enc_policy)) 2741 return 0; 2742 ext4_msg(NULL, KERN_WARNING, 2743 "Can't set or change test_dummy_encryption on remount"); 2744 return -EINVAL; 2745 } 2746 /* Also make sure s_mount_opts didn't contain a conflicting value. */ 2747 if (fscrypt_is_dummy_policy_set(&sbi->s_dummy_enc_policy)) { 2748 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy, 2749 &ctx->dummy_enc_policy)) 2750 return 0; 2751 ext4_msg(NULL, KERN_WARNING, 2752 "Conflicting test_dummy_encryption options"); 2753 return -EINVAL; 2754 } 2755 return 0; 2756 } 2757 2758 static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx, 2759 struct super_block *sb) 2760 { 2761 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy) || 2762 /* if already set, it was already verified to be the same */ 2763 fscrypt_is_dummy_policy_set(&EXT4_SB(sb)->s_dummy_enc_policy)) 2764 return; 2765 EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy; 2766 memset(&ctx->dummy_enc_policy, 0, sizeof(ctx->dummy_enc_policy)); 2767 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled"); 2768 } 2769 2770 static int ext4_check_opt_consistency(struct fs_context *fc, 2771 struct super_block *sb) 2772 { 2773 struct ext4_fs_context *ctx = fc->fs_private; 2774 struct ext4_sb_info *sbi = fc->s_fs_info; 2775 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE; 2776 int err; 2777 2778 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 2779 ext4_msg(NULL, KERN_ERR, 2780 "Mount option(s) incompatible with ext2"); 2781 return -EINVAL; 2782 } 2783 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 2784 ext4_msg(NULL, KERN_ERR, 2785 "Mount option(s) incompatible with ext3"); 2786 return -EINVAL; 2787 } 2788 2789 if (ctx->s_want_extra_isize > 2790 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) { 2791 ext4_msg(NULL, KERN_ERR, 2792 "Invalid want_extra_isize %d", 2793 ctx->s_want_extra_isize); 2794 return -EINVAL; 2795 } 2796 2797 err = ext4_check_test_dummy_encryption(fc, sb); 2798 if (err) 2799 return err; 2800 2801 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) { 2802 if (!sbi->s_journal) { 2803 ext4_msg(NULL, KERN_WARNING, 2804 "Remounting file system with no journal " 2805 "so ignoring journalled data option"); 2806 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS); 2807 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) != 2808 test_opt(sb, DATA_FLAGS)) { 2809 ext4_msg(NULL, KERN_ERR, "Cannot change data mode " 2810 "on remount"); 2811 return -EINVAL; 2812 } 2813 } 2814 2815 if (is_remount) { 2816 if (!sbi->s_journal && 2817 ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_ERR_ABORT)) { 2818 ext4_msg(NULL, KERN_WARNING, 2819 "Remounting fs w/o journal so ignoring data_err option"); 2820 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_ERR_ABORT); 2821 } 2822 2823 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) && 2824 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) { 2825 ext4_msg(NULL, KERN_ERR, "can't mount with " 2826 "both data=journal and dax"); 2827 return -EINVAL; 2828 } 2829 2830 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) && 2831 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2832 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) { 2833 fail_dax_change_remount: 2834 ext4_msg(NULL, KERN_ERR, "can't change " 2835 "dax mount option while remounting"); 2836 return -EINVAL; 2837 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) && 2838 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2839 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) { 2840 goto fail_dax_change_remount; 2841 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) && 2842 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2843 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2844 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) { 2845 goto fail_dax_change_remount; 2846 } 2847 } 2848 2849 return ext4_check_quota_consistency(fc, sb); 2850 } 2851 2852 static void ext4_apply_options(struct fs_context *fc, struct super_block *sb) 2853 { 2854 struct ext4_fs_context *ctx = fc->fs_private; 2855 struct ext4_sb_info *sbi = fc->s_fs_info; 2856 2857 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt; 2858 sbi->s_mount_opt |= ctx->vals_s_mount_opt; 2859 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2; 2860 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2; 2861 sb->s_flags &= ~ctx->mask_s_flags; 2862 sb->s_flags |= ctx->vals_s_flags; 2863 2864 #define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; }) 2865 APPLY(s_commit_interval); 2866 APPLY(s_stripe); 2867 APPLY(s_max_batch_time); 2868 APPLY(s_min_batch_time); 2869 APPLY(s_want_extra_isize); 2870 APPLY(s_inode_readahead_blks); 2871 APPLY(s_max_dir_size_kb); 2872 APPLY(s_li_wait_mult); 2873 APPLY(s_resgid); 2874 APPLY(s_resuid); 2875 2876 #ifdef CONFIG_EXT4_DEBUG 2877 APPLY(s_fc_debug_max_replay); 2878 #endif 2879 2880 ext4_apply_quota_options(fc, sb); 2881 ext4_apply_test_dummy_encryption(ctx, sb); 2882 } 2883 2884 2885 static int ext4_validate_options(struct fs_context *fc) 2886 { 2887 #ifdef CONFIG_QUOTA 2888 struct ext4_fs_context *ctx = fc->fs_private; 2889 char *usr_qf_name, *grp_qf_name; 2890 2891 usr_qf_name = ctx->s_qf_names[USRQUOTA]; 2892 grp_qf_name = ctx->s_qf_names[GRPQUOTA]; 2893 2894 if (usr_qf_name || grp_qf_name) { 2895 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name) 2896 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA); 2897 2898 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name) 2899 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA); 2900 2901 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) || 2902 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) { 2903 ext4_msg(NULL, KERN_ERR, "old and new quota " 2904 "format mixing"); 2905 return -EINVAL; 2906 } 2907 } 2908 #endif 2909 return 1; 2910 } 2911 2912 static inline void ext4_show_quota_options(struct seq_file *seq, 2913 struct super_block *sb) 2914 { 2915 #if defined(CONFIG_QUOTA) 2916 struct ext4_sb_info *sbi = EXT4_SB(sb); 2917 char *usr_qf_name, *grp_qf_name; 2918 2919 if (sbi->s_jquota_fmt) { 2920 char *fmtname = ""; 2921 2922 switch (sbi->s_jquota_fmt) { 2923 case QFMT_VFS_OLD: 2924 fmtname = "vfsold"; 2925 break; 2926 case QFMT_VFS_V0: 2927 fmtname = "vfsv0"; 2928 break; 2929 case QFMT_VFS_V1: 2930 fmtname = "vfsv1"; 2931 break; 2932 } 2933 seq_printf(seq, ",jqfmt=%s", fmtname); 2934 } 2935 2936 rcu_read_lock(); 2937 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); 2938 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); 2939 if (usr_qf_name) 2940 seq_show_option(seq, "usrjquota", usr_qf_name); 2941 if (grp_qf_name) 2942 seq_show_option(seq, "grpjquota", grp_qf_name); 2943 rcu_read_unlock(); 2944 #endif 2945 } 2946 2947 static const char *token2str(int token) 2948 { 2949 const struct fs_parameter_spec *spec; 2950 2951 for (spec = ext4_param_specs; spec->name != NULL; spec++) 2952 if (spec->opt == token && !spec->type) 2953 break; 2954 return spec->name; 2955 } 2956 2957 /* 2958 * Show an option if 2959 * - it's set to a non-default value OR 2960 * - if the per-sb default is different from the global default 2961 */ 2962 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 2963 int nodefs) 2964 { 2965 struct ext4_sb_info *sbi = EXT4_SB(sb); 2966 struct ext4_super_block *es = sbi->s_es; 2967 int def_errors; 2968 const struct mount_opts *m; 2969 char sep = nodefs ? '\n' : ','; 2970 2971 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 2972 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 2973 2974 if (sbi->s_sb_block != 1) 2975 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 2976 2977 for (m = ext4_mount_opts; m->token != Opt_err; m++) { 2978 int want_set = m->flags & MOPT_SET; 2979 int opt_2 = m->flags & MOPT_2; 2980 unsigned int mount_opt, def_mount_opt; 2981 2982 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 2983 m->flags & MOPT_SKIP) 2984 continue; 2985 2986 if (opt_2) { 2987 mount_opt = sbi->s_mount_opt2; 2988 def_mount_opt = sbi->s_def_mount_opt2; 2989 } else { 2990 mount_opt = sbi->s_mount_opt; 2991 def_mount_opt = sbi->s_def_mount_opt; 2992 } 2993 /* skip if same as the default */ 2994 if (!nodefs && !(m->mount_opt & (mount_opt ^ def_mount_opt))) 2995 continue; 2996 /* select Opt_noFoo vs Opt_Foo */ 2997 if ((want_set && 2998 (mount_opt & m->mount_opt) != m->mount_opt) || 2999 (!want_set && (mount_opt & m->mount_opt))) 3000 continue; 3001 SEQ_OPTS_PRINT("%s", token2str(m->token)); 3002 } 3003 3004 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 3005 ext4_get_resuid(es) != EXT4_DEF_RESUID) 3006 SEQ_OPTS_PRINT("resuid=%u", 3007 from_kuid_munged(&init_user_ns, sbi->s_resuid)); 3008 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 3009 ext4_get_resgid(es) != EXT4_DEF_RESGID) 3010 SEQ_OPTS_PRINT("resgid=%u", 3011 from_kgid_munged(&init_user_ns, sbi->s_resgid)); 3012 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 3013 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 3014 SEQ_OPTS_PUTS("errors=remount-ro"); 3015 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 3016 SEQ_OPTS_PUTS("errors=continue"); 3017 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 3018 SEQ_OPTS_PUTS("errors=panic"); 3019 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 3020 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 3021 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 3022 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 3023 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 3024 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 3025 if (nodefs && sb->s_flags & SB_I_VERSION) 3026 SEQ_OPTS_PUTS("i_version"); 3027 if (nodefs || sbi->s_stripe) 3028 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 3029 if (nodefs || EXT4_MOUNT_DATA_FLAGS & 3030 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 3031 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 3032 SEQ_OPTS_PUTS("data=journal"); 3033 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 3034 SEQ_OPTS_PUTS("data=ordered"); 3035 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 3036 SEQ_OPTS_PUTS("data=writeback"); 3037 } 3038 if (nodefs || 3039 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 3040 SEQ_OPTS_PRINT("inode_readahead_blks=%u", 3041 sbi->s_inode_readahead_blks); 3042 3043 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs || 3044 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 3045 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 3046 if (nodefs || sbi->s_max_dir_size_kb) 3047 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 3048 if (test_opt(sb, DATA_ERR_ABORT)) 3049 SEQ_OPTS_PUTS("data_err=abort"); 3050 3051 fscrypt_show_test_dummy_encryption(seq, sep, sb); 3052 3053 if (sb->s_flags & SB_INLINECRYPT) 3054 SEQ_OPTS_PUTS("inlinecrypt"); 3055 3056 if (test_opt(sb, DAX_ALWAYS)) { 3057 if (IS_EXT2_SB(sb)) 3058 SEQ_OPTS_PUTS("dax"); 3059 else 3060 SEQ_OPTS_PUTS("dax=always"); 3061 } else if (test_opt2(sb, DAX_NEVER)) { 3062 SEQ_OPTS_PUTS("dax=never"); 3063 } else if (test_opt2(sb, DAX_INODE)) { 3064 SEQ_OPTS_PUTS("dax=inode"); 3065 } 3066 3067 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD && 3068 !test_opt2(sb, MB_OPTIMIZE_SCAN)) { 3069 SEQ_OPTS_PUTS("mb_optimize_scan=0"); 3070 } else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD && 3071 test_opt2(sb, MB_OPTIMIZE_SCAN)) { 3072 SEQ_OPTS_PUTS("mb_optimize_scan=1"); 3073 } 3074 3075 if (nodefs && !test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) 3076 SEQ_OPTS_PUTS("prefetch_block_bitmaps"); 3077 3078 if (ext4_emergency_ro(sb)) 3079 SEQ_OPTS_PUTS("emergency_ro"); 3080 3081 if (ext4_forced_shutdown(sb)) 3082 SEQ_OPTS_PUTS("shutdown"); 3083 3084 ext4_show_quota_options(seq, sb); 3085 return 0; 3086 } 3087 3088 static int ext4_show_options(struct seq_file *seq, struct dentry *root) 3089 { 3090 return _ext4_show_options(seq, root->d_sb, 0); 3091 } 3092 3093 int ext4_seq_options_show(struct seq_file *seq, void *offset) 3094 { 3095 struct super_block *sb = seq->private; 3096 int rc; 3097 3098 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); 3099 rc = _ext4_show_options(seq, sb, 1); 3100 seq_putc(seq, '\n'); 3101 return rc; 3102 } 3103 3104 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 3105 int read_only) 3106 { 3107 struct ext4_sb_info *sbi = EXT4_SB(sb); 3108 int err = 0; 3109 3110 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 3111 ext4_msg(sb, KERN_ERR, "revision level too high, " 3112 "forcing read-only mode"); 3113 err = -EROFS; 3114 goto done; 3115 } 3116 if (read_only) 3117 goto done; 3118 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 3119 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 3120 "running e2fsck is recommended"); 3121 else if (sbi->s_mount_state & EXT4_ERROR_FS) 3122 ext4_msg(sb, KERN_WARNING, 3123 "warning: mounting fs with errors, " 3124 "running e2fsck is recommended"); 3125 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 3126 le16_to_cpu(es->s_mnt_count) >= 3127 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 3128 ext4_msg(sb, KERN_WARNING, 3129 "warning: maximal mount count reached, " 3130 "running e2fsck is recommended"); 3131 else if (le32_to_cpu(es->s_checkinterval) && 3132 (ext4_get_tstamp(es, s_lastcheck) + 3133 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds())) 3134 ext4_msg(sb, KERN_WARNING, 3135 "warning: checktime reached, " 3136 "running e2fsck is recommended"); 3137 if (!sbi->s_journal) 3138 es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 3139 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 3140 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 3141 le16_add_cpu(&es->s_mnt_count, 1); 3142 ext4_update_tstamp(es, s_mtime); 3143 if (sbi->s_journal) { 3144 ext4_set_feature_journal_needs_recovery(sb); 3145 if (ext4_has_feature_orphan_file(sb)) 3146 ext4_set_feature_orphan_present(sb); 3147 } 3148 3149 err = ext4_commit_super(sb); 3150 done: 3151 if (test_opt(sb, DEBUG)) 3152 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 3153 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 3154 sb->s_blocksize, 3155 sbi->s_groups_count, 3156 EXT4_BLOCKS_PER_GROUP(sb), 3157 EXT4_INODES_PER_GROUP(sb), 3158 sbi->s_mount_opt, sbi->s_mount_opt2); 3159 return err; 3160 } 3161 3162 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 3163 { 3164 struct ext4_sb_info *sbi = EXT4_SB(sb); 3165 struct flex_groups **old_groups, **new_groups; 3166 int size, i, j; 3167 3168 if (!sbi->s_log_groups_per_flex) 3169 return 0; 3170 3171 size = ext4_flex_group(sbi, ngroup - 1) + 1; 3172 if (size <= sbi->s_flex_groups_allocated) 3173 return 0; 3174 3175 new_groups = kvzalloc(roundup_pow_of_two(size * 3176 sizeof(*sbi->s_flex_groups)), GFP_KERNEL); 3177 if (!new_groups) { 3178 ext4_msg(sb, KERN_ERR, 3179 "not enough memory for %d flex group pointers", size); 3180 return -ENOMEM; 3181 } 3182 for (i = sbi->s_flex_groups_allocated; i < size; i++) { 3183 new_groups[i] = kvzalloc(roundup_pow_of_two( 3184 sizeof(struct flex_groups)), 3185 GFP_KERNEL); 3186 if (!new_groups[i]) { 3187 for (j = sbi->s_flex_groups_allocated; j < i; j++) 3188 kvfree(new_groups[j]); 3189 kvfree(new_groups); 3190 ext4_msg(sb, KERN_ERR, 3191 "not enough memory for %d flex groups", size); 3192 return -ENOMEM; 3193 } 3194 } 3195 rcu_read_lock(); 3196 old_groups = rcu_dereference(sbi->s_flex_groups); 3197 if (old_groups) 3198 memcpy(new_groups, old_groups, 3199 (sbi->s_flex_groups_allocated * 3200 sizeof(struct flex_groups *))); 3201 rcu_read_unlock(); 3202 rcu_assign_pointer(sbi->s_flex_groups, new_groups); 3203 sbi->s_flex_groups_allocated = size; 3204 if (old_groups) 3205 ext4_kvfree_array_rcu(old_groups); 3206 return 0; 3207 } 3208 3209 static int ext4_fill_flex_info(struct super_block *sb) 3210 { 3211 struct ext4_sb_info *sbi = EXT4_SB(sb); 3212 struct ext4_group_desc *gdp = NULL; 3213 struct flex_groups *fg; 3214 ext4_group_t flex_group; 3215 int i, err; 3216 3217 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 3218 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 3219 sbi->s_log_groups_per_flex = 0; 3220 return 1; 3221 } 3222 3223 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 3224 if (err) 3225 goto failed; 3226 3227 for (i = 0; i < sbi->s_groups_count; i++) { 3228 gdp = ext4_get_group_desc(sb, i, NULL); 3229 3230 flex_group = ext4_flex_group(sbi, i); 3231 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group); 3232 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes); 3233 atomic64_add(ext4_free_group_clusters(sb, gdp), 3234 &fg->free_clusters); 3235 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs); 3236 } 3237 3238 return 1; 3239 failed: 3240 return 0; 3241 } 3242 3243 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 3244 struct ext4_group_desc *gdp) 3245 { 3246 int offset = offsetof(struct ext4_group_desc, bg_checksum); 3247 __u16 crc = 0; 3248 __le32 le_group = cpu_to_le32(block_group); 3249 struct ext4_sb_info *sbi = EXT4_SB(sb); 3250 3251 if (ext4_has_feature_metadata_csum(sbi->s_sb)) { 3252 /* Use new metadata_csum algorithm */ 3253 __u32 csum32; 3254 __u16 dummy_csum = 0; 3255 3256 csum32 = ext4_chksum(sbi->s_csum_seed, (__u8 *)&le_group, 3257 sizeof(le_group)); 3258 csum32 = ext4_chksum(csum32, (__u8 *)gdp, offset); 3259 csum32 = ext4_chksum(csum32, (__u8 *)&dummy_csum, 3260 sizeof(dummy_csum)); 3261 offset += sizeof(dummy_csum); 3262 if (offset < sbi->s_desc_size) 3263 csum32 = ext4_chksum(csum32, (__u8 *)gdp + offset, 3264 sbi->s_desc_size - offset); 3265 3266 crc = csum32 & 0xFFFF; 3267 goto out; 3268 } 3269 3270 /* old crc16 code */ 3271 if (!ext4_has_feature_gdt_csum(sb)) 3272 return 0; 3273 3274 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 3275 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 3276 crc = crc16(crc, (__u8 *)gdp, offset); 3277 offset += sizeof(gdp->bg_checksum); /* skip checksum */ 3278 /* for checksum of struct ext4_group_desc do the rest...*/ 3279 if (ext4_has_feature_64bit(sb) && offset < sbi->s_desc_size) 3280 crc = crc16(crc, (__u8 *)gdp + offset, 3281 sbi->s_desc_size - offset); 3282 3283 out: 3284 return cpu_to_le16(crc); 3285 } 3286 3287 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 3288 struct ext4_group_desc *gdp) 3289 { 3290 if (ext4_has_group_desc_csum(sb) && 3291 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 3292 return 0; 3293 3294 return 1; 3295 } 3296 3297 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 3298 struct ext4_group_desc *gdp) 3299 { 3300 if (!ext4_has_group_desc_csum(sb)) 3301 return; 3302 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 3303 } 3304 3305 /* Called at mount-time, super-block is locked */ 3306 static int ext4_check_descriptors(struct super_block *sb, 3307 ext4_fsblk_t sb_block, 3308 ext4_group_t *first_not_zeroed) 3309 { 3310 struct ext4_sb_info *sbi = EXT4_SB(sb); 3311 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 3312 ext4_fsblk_t last_block; 3313 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0); 3314 ext4_fsblk_t block_bitmap; 3315 ext4_fsblk_t inode_bitmap; 3316 ext4_fsblk_t inode_table; 3317 int flexbg_flag = 0; 3318 ext4_group_t i, grp = sbi->s_groups_count; 3319 3320 if (ext4_has_feature_flex_bg(sb)) 3321 flexbg_flag = 1; 3322 3323 ext4_debug("Checking group descriptors"); 3324 3325 for (i = 0; i < sbi->s_groups_count; i++) { 3326 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 3327 3328 if (i == sbi->s_groups_count - 1 || flexbg_flag) 3329 last_block = ext4_blocks_count(sbi->s_es) - 1; 3330 else 3331 last_block = first_block + 3332 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 3333 3334 if ((grp == sbi->s_groups_count) && 3335 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3336 grp = i; 3337 3338 block_bitmap = ext4_block_bitmap(sb, gdp); 3339 if (block_bitmap == sb_block) { 3340 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3341 "Block bitmap for group %u overlaps " 3342 "superblock", i); 3343 if (!sb_rdonly(sb)) 3344 return 0; 3345 } 3346 if (block_bitmap >= sb_block + 1 && 3347 block_bitmap <= last_bg_block) { 3348 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3349 "Block bitmap for group %u overlaps " 3350 "block group descriptors", i); 3351 if (!sb_rdonly(sb)) 3352 return 0; 3353 } 3354 if (block_bitmap < first_block || block_bitmap > last_block) { 3355 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3356 "Block bitmap for group %u not in group " 3357 "(block %llu)!", i, block_bitmap); 3358 return 0; 3359 } 3360 inode_bitmap = ext4_inode_bitmap(sb, gdp); 3361 if (inode_bitmap == sb_block) { 3362 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3363 "Inode bitmap for group %u overlaps " 3364 "superblock", i); 3365 if (!sb_rdonly(sb)) 3366 return 0; 3367 } 3368 if (inode_bitmap >= sb_block + 1 && 3369 inode_bitmap <= last_bg_block) { 3370 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3371 "Inode bitmap for group %u overlaps " 3372 "block group descriptors", i); 3373 if (!sb_rdonly(sb)) 3374 return 0; 3375 } 3376 if (inode_bitmap < first_block || inode_bitmap > last_block) { 3377 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3378 "Inode bitmap for group %u not in group " 3379 "(block %llu)!", i, inode_bitmap); 3380 return 0; 3381 } 3382 inode_table = ext4_inode_table(sb, gdp); 3383 if (inode_table == sb_block) { 3384 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3385 "Inode table for group %u overlaps " 3386 "superblock", i); 3387 if (!sb_rdonly(sb)) 3388 return 0; 3389 } 3390 if (inode_table >= sb_block + 1 && 3391 inode_table <= last_bg_block) { 3392 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3393 "Inode table for group %u overlaps " 3394 "block group descriptors", i); 3395 if (!sb_rdonly(sb)) 3396 return 0; 3397 } 3398 if (inode_table < first_block || 3399 inode_table + sbi->s_itb_per_group - 1 > last_block) { 3400 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3401 "Inode table for group %u not in group " 3402 "(block %llu)!", i, inode_table); 3403 return 0; 3404 } 3405 ext4_lock_group(sb, i); 3406 if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 3407 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3408 "Checksum for group %u failed (%u!=%u)", 3409 i, le16_to_cpu(ext4_group_desc_csum(sb, i, 3410 gdp)), le16_to_cpu(gdp->bg_checksum)); 3411 if (!sb_rdonly(sb)) { 3412 ext4_unlock_group(sb, i); 3413 return 0; 3414 } 3415 } 3416 ext4_unlock_group(sb, i); 3417 if (!flexbg_flag) 3418 first_block += EXT4_BLOCKS_PER_GROUP(sb); 3419 } 3420 if (NULL != first_not_zeroed) 3421 *first_not_zeroed = grp; 3422 return 1; 3423 } 3424 3425 /* 3426 * Maximal extent format file size. 3427 * Resulting logical blkno at s_maxbytes must fit in our on-disk 3428 * extent format containers, within a sector_t, and within i_blocks 3429 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 3430 * so that won't be a limiting factor. 3431 * 3432 * However there is other limiting factor. We do store extents in the form 3433 * of starting block and length, hence the resulting length of the extent 3434 * covering maximum file size must fit into on-disk format containers as 3435 * well. Given that length is always by 1 unit bigger than max unit (because 3436 * we count 0 as well) we have to lower the s_maxbytes by one fs block. 3437 * 3438 * Note, this does *not* consider any metadata overhead for vfs i_blocks. 3439 */ 3440 static loff_t ext4_max_size(int blkbits, int has_huge_files) 3441 { 3442 loff_t res; 3443 loff_t upper_limit = MAX_LFS_FILESIZE; 3444 3445 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64)); 3446 3447 if (!has_huge_files) { 3448 upper_limit = (1LL << 32) - 1; 3449 3450 /* total blocks in file system block size */ 3451 upper_limit >>= (blkbits - 9); 3452 upper_limit <<= blkbits; 3453 } 3454 3455 /* 3456 * 32-bit extent-start container, ee_block. We lower the maxbytes 3457 * by one fs block, so ee_len can cover the extent of maximum file 3458 * size 3459 */ 3460 res = (1LL << 32) - 1; 3461 res <<= blkbits; 3462 3463 /* Sanity check against vm- & vfs- imposed limits */ 3464 if (res > upper_limit) 3465 res = upper_limit; 3466 3467 return res; 3468 } 3469 3470 /* 3471 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 3472 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 3473 * We need to be 1 filesystem block less than the 2^48 sector limit. 3474 */ 3475 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 3476 { 3477 loff_t upper_limit, res = EXT4_NDIR_BLOCKS; 3478 int meta_blocks; 3479 unsigned int ppb = 1 << (bits - 2); 3480 3481 /* 3482 * This is calculated to be the largest file size for a dense, block 3483 * mapped file such that the file's total number of 512-byte sectors, 3484 * including data and all indirect blocks, does not exceed (2^48 - 1). 3485 * 3486 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 3487 * number of 512-byte sectors of the file. 3488 */ 3489 if (!has_huge_files) { 3490 /* 3491 * !has_huge_files or implies that the inode i_block field 3492 * represents total file blocks in 2^32 512-byte sectors == 3493 * size of vfs inode i_blocks * 8 3494 */ 3495 upper_limit = (1LL << 32) - 1; 3496 3497 /* total blocks in file system block size */ 3498 upper_limit >>= (bits - 9); 3499 3500 } else { 3501 /* 3502 * We use 48 bit ext4_inode i_blocks 3503 * With EXT4_HUGE_FILE_FL set the i_blocks 3504 * represent total number of blocks in 3505 * file system block size 3506 */ 3507 upper_limit = (1LL << 48) - 1; 3508 3509 } 3510 3511 /* Compute how many blocks we can address by block tree */ 3512 res += ppb; 3513 res += ppb * ppb; 3514 res += ((loff_t)ppb) * ppb * ppb; 3515 /* Compute how many metadata blocks are needed */ 3516 meta_blocks = 1; 3517 meta_blocks += 1 + ppb; 3518 meta_blocks += 1 + ppb + ppb * ppb; 3519 /* Does block tree limit file size? */ 3520 if (res + meta_blocks <= upper_limit) 3521 goto check_lfs; 3522 3523 res = upper_limit; 3524 /* How many metadata blocks are needed for addressing upper_limit? */ 3525 upper_limit -= EXT4_NDIR_BLOCKS; 3526 /* indirect blocks */ 3527 meta_blocks = 1; 3528 upper_limit -= ppb; 3529 /* double indirect blocks */ 3530 if (upper_limit < ppb * ppb) { 3531 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb); 3532 res -= meta_blocks; 3533 goto check_lfs; 3534 } 3535 meta_blocks += 1 + ppb; 3536 upper_limit -= ppb * ppb; 3537 /* tripple indirect blocks for the rest */ 3538 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) + 3539 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb); 3540 res -= meta_blocks; 3541 check_lfs: 3542 res <<= bits; 3543 if (res > MAX_LFS_FILESIZE) 3544 res = MAX_LFS_FILESIZE; 3545 3546 return res; 3547 } 3548 3549 static ext4_fsblk_t descriptor_loc(struct super_block *sb, 3550 ext4_fsblk_t logical_sb_block, int nr) 3551 { 3552 struct ext4_sb_info *sbi = EXT4_SB(sb); 3553 ext4_group_t bg, first_meta_bg; 3554 int has_super = 0; 3555 3556 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 3557 3558 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 3559 return logical_sb_block + nr + 1; 3560 bg = sbi->s_desc_per_block * nr; 3561 if (ext4_bg_has_super(sb, bg)) 3562 has_super = 1; 3563 3564 /* 3565 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 3566 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 3567 * on modern mke2fs or blksize > 1k on older mke2fs) then we must 3568 * compensate. 3569 */ 3570 if (sb->s_blocksize == 1024 && nr == 0 && 3571 le32_to_cpu(sbi->s_es->s_first_data_block) == 0) 3572 has_super++; 3573 3574 return (has_super + ext4_group_first_block_no(sb, bg)); 3575 } 3576 3577 /** 3578 * ext4_get_stripe_size: Get the stripe size. 3579 * @sbi: In memory super block info 3580 * 3581 * If we have specified it via mount option, then 3582 * use the mount option value. If the value specified at mount time is 3583 * greater than the blocks per group use the super block value. 3584 * If the super block value is greater than blocks per group return 0. 3585 * Allocator needs it be less than blocks per group. 3586 * 3587 */ 3588 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 3589 { 3590 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 3591 unsigned long stripe_width = 3592 le32_to_cpu(sbi->s_es->s_raid_stripe_width); 3593 int ret; 3594 3595 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 3596 ret = sbi->s_stripe; 3597 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) 3598 ret = stripe_width; 3599 else if (stride && stride <= sbi->s_blocks_per_group) 3600 ret = stride; 3601 else 3602 ret = 0; 3603 3604 /* 3605 * If the stripe width is 1, this makes no sense and 3606 * we set it to 0 to turn off stripe handling code. 3607 */ 3608 if (ret <= 1) 3609 ret = 0; 3610 3611 return ret; 3612 } 3613 3614 /* 3615 * Check whether this filesystem can be mounted based on 3616 * the features present and the RDONLY/RDWR mount requested. 3617 * Returns 1 if this filesystem can be mounted as requested, 3618 * 0 if it cannot be. 3619 */ 3620 int ext4_feature_set_ok(struct super_block *sb, int readonly) 3621 { 3622 if (ext4_has_unknown_ext4_incompat_features(sb)) { 3623 ext4_msg(sb, KERN_ERR, 3624 "Couldn't mount because of " 3625 "unsupported optional features (%x)", 3626 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 3627 ~EXT4_FEATURE_INCOMPAT_SUPP)); 3628 return 0; 3629 } 3630 3631 if (!IS_ENABLED(CONFIG_UNICODE) && ext4_has_feature_casefold(sb)) { 3632 ext4_msg(sb, KERN_ERR, 3633 "Filesystem with casefold feature cannot be " 3634 "mounted without CONFIG_UNICODE"); 3635 return 0; 3636 } 3637 3638 if (readonly) 3639 return 1; 3640 3641 if (ext4_has_feature_readonly(sb)) { 3642 ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 3643 sb->s_flags |= SB_RDONLY; 3644 return 1; 3645 } 3646 3647 /* Check that feature set is OK for a read-write mount */ 3648 if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 3649 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 3650 "unsupported optional features (%x)", 3651 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 3652 ~EXT4_FEATURE_RO_COMPAT_SUPP)); 3653 return 0; 3654 } 3655 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 3656 ext4_msg(sb, KERN_ERR, 3657 "Can't support bigalloc feature without " 3658 "extents feature\n"); 3659 return 0; 3660 } 3661 if (ext4_has_feature_bigalloc(sb) && 3662 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block)) { 3663 ext4_msg(sb, KERN_WARNING, 3664 "bad geometry: bigalloc file system with non-zero " 3665 "first_data_block\n"); 3666 return 0; 3667 } 3668 3669 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2) 3670 if (!readonly && (ext4_has_feature_quota(sb) || 3671 ext4_has_feature_project(sb))) { 3672 ext4_msg(sb, KERN_ERR, 3673 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2"); 3674 return 0; 3675 } 3676 #endif /* CONFIG_QUOTA */ 3677 return 1; 3678 } 3679 3680 /* 3681 * This function is called once a day by default if we have errors logged 3682 * on the file system. 3683 * Use the err_report_sec sysfs attribute to disable or adjust its call 3684 * freequency. 3685 */ 3686 void print_daily_error_info(struct timer_list *t) 3687 { 3688 struct ext4_sb_info *sbi = timer_container_of(sbi, t, s_err_report); 3689 struct super_block *sb = sbi->s_sb; 3690 struct ext4_super_block *es = sbi->s_es; 3691 3692 if (es->s_error_count) 3693 /* fsck newer than v1.41.13 is needed to clean this condition. */ 3694 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 3695 le32_to_cpu(es->s_error_count)); 3696 if (es->s_first_error_time) { 3697 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d", 3698 sb->s_id, 3699 ext4_get_tstamp(es, s_first_error_time), 3700 (int) sizeof(es->s_first_error_func), 3701 es->s_first_error_func, 3702 le32_to_cpu(es->s_first_error_line)); 3703 if (es->s_first_error_ino) 3704 printk(KERN_CONT ": inode %u", 3705 le32_to_cpu(es->s_first_error_ino)); 3706 if (es->s_first_error_block) 3707 printk(KERN_CONT ": block %llu", (unsigned long long) 3708 le64_to_cpu(es->s_first_error_block)); 3709 printk(KERN_CONT "\n"); 3710 } 3711 if (es->s_last_error_time) { 3712 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d", 3713 sb->s_id, 3714 ext4_get_tstamp(es, s_last_error_time), 3715 (int) sizeof(es->s_last_error_func), 3716 es->s_last_error_func, 3717 le32_to_cpu(es->s_last_error_line)); 3718 if (es->s_last_error_ino) 3719 printk(KERN_CONT ": inode %u", 3720 le32_to_cpu(es->s_last_error_ino)); 3721 if (es->s_last_error_block) 3722 printk(KERN_CONT ": block %llu", (unsigned long long) 3723 le64_to_cpu(es->s_last_error_block)); 3724 printk(KERN_CONT "\n"); 3725 } 3726 3727 if (sbi->s_err_report_sec) 3728 mod_timer(&sbi->s_err_report, jiffies + secs_to_jiffies(sbi->s_err_report_sec)); 3729 } 3730 3731 /* Find next suitable group and run ext4_init_inode_table */ 3732 static int ext4_run_li_request(struct ext4_li_request *elr) 3733 { 3734 struct ext4_group_desc *gdp = NULL; 3735 struct super_block *sb = elr->lr_super; 3736 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 3737 ext4_group_t group = elr->lr_next_group; 3738 unsigned int prefetch_ios = 0; 3739 int ret = 0; 3740 int nr = EXT4_SB(sb)->s_mb_prefetch; 3741 u64 start_time; 3742 3743 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) { 3744 elr->lr_next_group = ext4_mb_prefetch(sb, group, nr, &prefetch_ios); 3745 ext4_mb_prefetch_fini(sb, elr->lr_next_group, nr); 3746 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, nr); 3747 if (group >= elr->lr_next_group) { 3748 ret = 1; 3749 if (elr->lr_first_not_zeroed != ngroups && 3750 !ext4_emergency_state(sb) && !sb_rdonly(sb) && 3751 test_opt(sb, INIT_INODE_TABLE)) { 3752 elr->lr_next_group = elr->lr_first_not_zeroed; 3753 elr->lr_mode = EXT4_LI_MODE_ITABLE; 3754 ret = 0; 3755 } 3756 } 3757 return ret; 3758 } 3759 3760 for (; group < ngroups; group++) { 3761 gdp = ext4_get_group_desc(sb, group, NULL); 3762 if (!gdp) { 3763 ret = 1; 3764 break; 3765 } 3766 3767 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3768 break; 3769 } 3770 3771 if (group >= ngroups) 3772 ret = 1; 3773 3774 if (!ret) { 3775 start_time = ktime_get_ns(); 3776 ret = ext4_init_inode_table(sb, group, 3777 elr->lr_timeout ? 0 : 1); 3778 trace_ext4_lazy_itable_init(sb, group); 3779 if (elr->lr_timeout == 0) { 3780 elr->lr_timeout = nsecs_to_jiffies((ktime_get_ns() - start_time) * 3781 EXT4_SB(elr->lr_super)->s_li_wait_mult); 3782 } 3783 elr->lr_next_sched = jiffies + elr->lr_timeout; 3784 elr->lr_next_group = group + 1; 3785 } 3786 return ret; 3787 } 3788 3789 /* 3790 * Remove lr_request from the list_request and free the 3791 * request structure. Should be called with li_list_mtx held 3792 */ 3793 static void ext4_remove_li_request(struct ext4_li_request *elr) 3794 { 3795 if (!elr) 3796 return; 3797 3798 list_del(&elr->lr_request); 3799 EXT4_SB(elr->lr_super)->s_li_request = NULL; 3800 kfree(elr); 3801 } 3802 3803 static void ext4_unregister_li_request(struct super_block *sb) 3804 { 3805 mutex_lock(&ext4_li_mtx); 3806 if (!ext4_li_info) { 3807 mutex_unlock(&ext4_li_mtx); 3808 return; 3809 } 3810 3811 mutex_lock(&ext4_li_info->li_list_mtx); 3812 ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 3813 mutex_unlock(&ext4_li_info->li_list_mtx); 3814 mutex_unlock(&ext4_li_mtx); 3815 } 3816 3817 static struct task_struct *ext4_lazyinit_task; 3818 3819 /* 3820 * This is the function where ext4lazyinit thread lives. It walks 3821 * through the request list searching for next scheduled filesystem. 3822 * When such a fs is found, run the lazy initialization request 3823 * (ext4_rn_li_request) and keep track of the time spend in this 3824 * function. Based on that time we compute next schedule time of 3825 * the request. When walking through the list is complete, compute 3826 * next waking time and put itself into sleep. 3827 */ 3828 static int ext4_lazyinit_thread(void *arg) 3829 { 3830 struct ext4_lazy_init *eli = arg; 3831 struct list_head *pos, *n; 3832 struct ext4_li_request *elr; 3833 unsigned long next_wakeup, cur; 3834 3835 BUG_ON(NULL == eli); 3836 set_freezable(); 3837 3838 cont_thread: 3839 while (true) { 3840 bool next_wakeup_initialized = false; 3841 3842 next_wakeup = 0; 3843 mutex_lock(&eli->li_list_mtx); 3844 if (list_empty(&eli->li_request_list)) { 3845 mutex_unlock(&eli->li_list_mtx); 3846 goto exit_thread; 3847 } 3848 list_for_each_safe(pos, n, &eli->li_request_list) { 3849 int err = 0; 3850 int progress = 0; 3851 elr = list_entry(pos, struct ext4_li_request, 3852 lr_request); 3853 3854 if (time_before(jiffies, elr->lr_next_sched)) { 3855 if (!next_wakeup_initialized || 3856 time_before(elr->lr_next_sched, next_wakeup)) { 3857 next_wakeup = elr->lr_next_sched; 3858 next_wakeup_initialized = true; 3859 } 3860 continue; 3861 } 3862 if (down_read_trylock(&elr->lr_super->s_umount)) { 3863 if (sb_start_write_trylock(elr->lr_super)) { 3864 progress = 1; 3865 /* 3866 * We hold sb->s_umount, sb can not 3867 * be removed from the list, it is 3868 * now safe to drop li_list_mtx 3869 */ 3870 mutex_unlock(&eli->li_list_mtx); 3871 err = ext4_run_li_request(elr); 3872 sb_end_write(elr->lr_super); 3873 mutex_lock(&eli->li_list_mtx); 3874 n = pos->next; 3875 } 3876 up_read((&elr->lr_super->s_umount)); 3877 } 3878 /* error, remove the lazy_init job */ 3879 if (err) { 3880 ext4_remove_li_request(elr); 3881 continue; 3882 } 3883 if (!progress) { 3884 elr->lr_next_sched = jiffies + 3885 get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ); 3886 } 3887 if (!next_wakeup_initialized || 3888 time_before(elr->lr_next_sched, next_wakeup)) { 3889 next_wakeup = elr->lr_next_sched; 3890 next_wakeup_initialized = true; 3891 } 3892 } 3893 mutex_unlock(&eli->li_list_mtx); 3894 3895 try_to_freeze(); 3896 3897 cur = jiffies; 3898 if (!next_wakeup_initialized || time_after_eq(cur, next_wakeup)) { 3899 cond_resched(); 3900 continue; 3901 } 3902 3903 schedule_timeout_interruptible(next_wakeup - cur); 3904 3905 if (kthread_should_stop()) { 3906 ext4_clear_request_list(); 3907 goto exit_thread; 3908 } 3909 } 3910 3911 exit_thread: 3912 /* 3913 * It looks like the request list is empty, but we need 3914 * to check it under the li_list_mtx lock, to prevent any 3915 * additions into it, and of course we should lock ext4_li_mtx 3916 * to atomically free the list and ext4_li_info, because at 3917 * this point another ext4 filesystem could be registering 3918 * new one. 3919 */ 3920 mutex_lock(&ext4_li_mtx); 3921 mutex_lock(&eli->li_list_mtx); 3922 if (!list_empty(&eli->li_request_list)) { 3923 mutex_unlock(&eli->li_list_mtx); 3924 mutex_unlock(&ext4_li_mtx); 3925 goto cont_thread; 3926 } 3927 mutex_unlock(&eli->li_list_mtx); 3928 kfree(ext4_li_info); 3929 ext4_li_info = NULL; 3930 mutex_unlock(&ext4_li_mtx); 3931 3932 return 0; 3933 } 3934 3935 static void ext4_clear_request_list(void) 3936 { 3937 struct list_head *pos, *n; 3938 struct ext4_li_request *elr; 3939 3940 mutex_lock(&ext4_li_info->li_list_mtx); 3941 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3942 elr = list_entry(pos, struct ext4_li_request, 3943 lr_request); 3944 ext4_remove_li_request(elr); 3945 } 3946 mutex_unlock(&ext4_li_info->li_list_mtx); 3947 } 3948 3949 static int ext4_run_lazyinit_thread(void) 3950 { 3951 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 3952 ext4_li_info, "ext4lazyinit"); 3953 if (IS_ERR(ext4_lazyinit_task)) { 3954 int err = PTR_ERR(ext4_lazyinit_task); 3955 ext4_clear_request_list(); 3956 kfree(ext4_li_info); 3957 ext4_li_info = NULL; 3958 printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3959 "initialization thread\n", 3960 err); 3961 return err; 3962 } 3963 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3964 return 0; 3965 } 3966 3967 /* 3968 * Check whether it make sense to run itable init. thread or not. 3969 * If there is at least one uninitialized inode table, return 3970 * corresponding group number, else the loop goes through all 3971 * groups and return total number of groups. 3972 */ 3973 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3974 { 3975 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3976 struct ext4_group_desc *gdp = NULL; 3977 3978 if (!ext4_has_group_desc_csum(sb)) 3979 return ngroups; 3980 3981 for (group = 0; group < ngroups; group++) { 3982 gdp = ext4_get_group_desc(sb, group, NULL); 3983 if (!gdp) 3984 continue; 3985 3986 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3987 break; 3988 } 3989 3990 return group; 3991 } 3992 3993 static int ext4_li_info_new(void) 3994 { 3995 struct ext4_lazy_init *eli = NULL; 3996 3997 eli = kzalloc_obj(*eli); 3998 if (!eli) 3999 return -ENOMEM; 4000 4001 INIT_LIST_HEAD(&eli->li_request_list); 4002 mutex_init(&eli->li_list_mtx); 4003 4004 eli->li_state |= EXT4_LAZYINIT_QUIT; 4005 4006 ext4_li_info = eli; 4007 4008 return 0; 4009 } 4010 4011 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 4012 ext4_group_t start) 4013 { 4014 struct ext4_li_request *elr; 4015 4016 elr = kzalloc_obj(*elr); 4017 if (!elr) 4018 return NULL; 4019 4020 elr->lr_super = sb; 4021 elr->lr_first_not_zeroed = start; 4022 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) { 4023 elr->lr_mode = EXT4_LI_MODE_ITABLE; 4024 elr->lr_next_group = start; 4025 } else { 4026 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP; 4027 } 4028 4029 /* 4030 * Randomize first schedule time of the request to 4031 * spread the inode table initialization requests 4032 * better. 4033 */ 4034 elr->lr_next_sched = jiffies + get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ); 4035 return elr; 4036 } 4037 4038 int ext4_register_li_request(struct super_block *sb, 4039 ext4_group_t first_not_zeroed) 4040 { 4041 struct ext4_sb_info *sbi = EXT4_SB(sb); 4042 struct ext4_li_request *elr = NULL; 4043 ext4_group_t ngroups = sbi->s_groups_count; 4044 int ret = 0; 4045 4046 mutex_lock(&ext4_li_mtx); 4047 if (sbi->s_li_request != NULL) { 4048 /* 4049 * Reset timeout so it can be computed again, because 4050 * s_li_wait_mult might have changed. 4051 */ 4052 sbi->s_li_request->lr_timeout = 0; 4053 goto out; 4054 } 4055 4056 if (ext4_emergency_state(sb) || sb_rdonly(sb) || 4057 (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) && 4058 (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE)))) 4059 goto out; 4060 4061 elr = ext4_li_request_new(sb, first_not_zeroed); 4062 if (!elr) { 4063 ret = -ENOMEM; 4064 goto out; 4065 } 4066 4067 if (NULL == ext4_li_info) { 4068 ret = ext4_li_info_new(); 4069 if (ret) 4070 goto out; 4071 } 4072 4073 mutex_lock(&ext4_li_info->li_list_mtx); 4074 list_add(&elr->lr_request, &ext4_li_info->li_request_list); 4075 mutex_unlock(&ext4_li_info->li_list_mtx); 4076 4077 sbi->s_li_request = elr; 4078 /* 4079 * set elr to NULL here since it has been inserted to 4080 * the request_list and the removal and free of it is 4081 * handled by ext4_clear_request_list from now on. 4082 */ 4083 elr = NULL; 4084 4085 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 4086 ret = ext4_run_lazyinit_thread(); 4087 if (ret) 4088 goto out; 4089 } 4090 out: 4091 mutex_unlock(&ext4_li_mtx); 4092 if (ret) 4093 kfree(elr); 4094 return ret; 4095 } 4096 4097 /* 4098 * We do not need to lock anything since this is called on 4099 * module unload. 4100 */ 4101 static void ext4_destroy_lazyinit_thread(void) 4102 { 4103 /* 4104 * If thread exited earlier 4105 * there's nothing to be done. 4106 */ 4107 if (!ext4_li_info || !ext4_lazyinit_task) 4108 return; 4109 4110 kthread_stop(ext4_lazyinit_task); 4111 } 4112 4113 static int set_journal_csum_feature_set(struct super_block *sb) 4114 { 4115 int ret = 1; 4116 int compat, incompat; 4117 struct ext4_sb_info *sbi = EXT4_SB(sb); 4118 4119 if (ext4_has_feature_metadata_csum(sb)) { 4120 /* journal checksum v3 */ 4121 compat = 0; 4122 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 4123 } else { 4124 /* journal checksum v1 */ 4125 compat = JBD2_FEATURE_COMPAT_CHECKSUM; 4126 incompat = 0; 4127 } 4128 4129 jbd2_journal_clear_features(sbi->s_journal, 4130 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 4131 JBD2_FEATURE_INCOMPAT_CSUM_V3 | 4132 JBD2_FEATURE_INCOMPAT_CSUM_V2); 4133 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4134 ret = jbd2_journal_set_features(sbi->s_journal, 4135 compat, 0, 4136 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 4137 incompat); 4138 } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 4139 ret = jbd2_journal_set_features(sbi->s_journal, 4140 compat, 0, 4141 incompat); 4142 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 4143 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 4144 } else { 4145 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 4146 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 4147 } 4148 4149 return ret; 4150 } 4151 4152 /* 4153 * Note: calculating the overhead so we can be compatible with 4154 * historical BSD practice is quite difficult in the face of 4155 * clusters/bigalloc. This is because multiple metadata blocks from 4156 * different block group can end up in the same allocation cluster. 4157 * Calculating the exact overhead in the face of clustered allocation 4158 * requires either O(all block bitmaps) in memory or O(number of block 4159 * groups**2) in time. We will still calculate the superblock for 4160 * older file systems --- and if we come across with a bigalloc file 4161 * system with zero in s_overhead_clusters the estimate will be close to 4162 * correct especially for very large cluster sizes --- but for newer 4163 * file systems, it's better to calculate this figure once at mkfs 4164 * time, and store it in the superblock. If the superblock value is 4165 * present (even for non-bigalloc file systems), we will use it. 4166 */ 4167 static int count_overhead(struct super_block *sb, ext4_group_t grp, 4168 char *buf) 4169 { 4170 struct ext4_sb_info *sbi = EXT4_SB(sb); 4171 struct ext4_group_desc *gdp; 4172 ext4_fsblk_t first_block, last_block, b; 4173 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 4174 int s, j, count = 0; 4175 int has_super = ext4_bg_has_super(sb, grp); 4176 4177 if (!ext4_has_feature_bigalloc(sb)) 4178 return (has_super + ext4_bg_num_gdb(sb, grp) + 4179 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) + 4180 sbi->s_itb_per_group + 2); 4181 4182 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 4183 (grp * EXT4_BLOCKS_PER_GROUP(sb)); 4184 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 4185 for (i = 0; i < ngroups; i++) { 4186 gdp = ext4_get_group_desc(sb, i, NULL); 4187 b = ext4_block_bitmap(sb, gdp); 4188 if (b >= first_block && b <= last_block) { 4189 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 4190 count++; 4191 } 4192 b = ext4_inode_bitmap(sb, gdp); 4193 if (b >= first_block && b <= last_block) { 4194 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 4195 count++; 4196 } 4197 b = ext4_inode_table(sb, gdp); 4198 if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 4199 for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 4200 int c = EXT4_B2C(sbi, b - first_block); 4201 ext4_set_bit(c, buf); 4202 count++; 4203 } 4204 if (i != grp) 4205 continue; 4206 s = 0; 4207 if (ext4_bg_has_super(sb, grp)) { 4208 ext4_set_bit(s++, buf); 4209 count++; 4210 } 4211 j = ext4_bg_num_gdb(sb, grp); 4212 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { 4213 ext4_error(sb, "Invalid number of block group " 4214 "descriptor blocks: %d", j); 4215 j = EXT4_BLOCKS_PER_GROUP(sb) - s; 4216 } 4217 count += j; 4218 for (; j > 0; j--) 4219 ext4_set_bit(EXT4_B2C(sbi, s++), buf); 4220 } 4221 if (!count) 4222 return 0; 4223 return EXT4_CLUSTERS_PER_GROUP(sb) - 4224 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 4225 } 4226 4227 /* 4228 * Compute the overhead and stash it in sbi->s_overhead 4229 */ 4230 int ext4_calculate_overhead(struct super_block *sb) 4231 { 4232 struct ext4_sb_info *sbi = EXT4_SB(sb); 4233 struct ext4_super_block *es = sbi->s_es; 4234 struct inode *j_inode; 4235 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); 4236 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 4237 ext4_fsblk_t overhead = 0; 4238 char *buf = kvmalloc(sb->s_blocksize, GFP_NOFS | __GFP_ZERO); 4239 4240 if (!buf) 4241 return -ENOMEM; 4242 4243 /* 4244 * Compute the overhead (FS structures). This is constant 4245 * for a given filesystem unless the number of block groups 4246 * changes so we cache the previous value until it does. 4247 */ 4248 4249 /* 4250 * All of the blocks before first_data_block are overhead 4251 */ 4252 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 4253 4254 /* 4255 * Add the overhead found in each block group 4256 */ 4257 for (i = 0; i < ngroups; i++) { 4258 int blks; 4259 4260 blks = count_overhead(sb, i, buf); 4261 overhead += blks; 4262 if (blks) 4263 memset(buf, 0, sb->s_blocksize); 4264 cond_resched(); 4265 } 4266 4267 /* 4268 * Add the internal journal blocks whether the journal has been 4269 * loaded or not 4270 */ 4271 if (sbi->s_journal && !sbi->s_journal_bdev_file) 4272 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len); 4273 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) { 4274 /* j_inum for internal journal is non-zero */ 4275 j_inode = ext4_get_journal_inode(sb, j_inum); 4276 if (!IS_ERR(j_inode)) { 4277 j_blocks = j_inode->i_size >> sb->s_blocksize_bits; 4278 overhead += EXT4_NUM_B2C(sbi, j_blocks); 4279 iput(j_inode); 4280 } else { 4281 ext4_msg(sb, KERN_ERR, "can't get journal size"); 4282 } 4283 } 4284 sbi->s_overhead = overhead; 4285 smp_wmb(); 4286 kvfree(buf); 4287 return 0; 4288 } 4289 4290 static void ext4_set_resv_clusters(struct super_block *sb) 4291 { 4292 ext4_fsblk_t resv_clusters; 4293 struct ext4_sb_info *sbi = EXT4_SB(sb); 4294 4295 /* 4296 * There's no need to reserve anything when we aren't using extents. 4297 * The space estimates are exact, there are no unwritten extents, 4298 * hole punching doesn't need new metadata... This is needed especially 4299 * to keep ext2/3 backward compatibility. 4300 */ 4301 if (!ext4_has_feature_extents(sb)) 4302 return; 4303 /* 4304 * By default we reserve 2% or 4096 clusters, whichever is smaller. 4305 * This should cover the situations where we can not afford to run 4306 * out of space like for example punch hole, or converting 4307 * unwritten extents in delalloc path. In most cases such 4308 * allocation would require 1, or 2 blocks, higher numbers are 4309 * very rare. 4310 */ 4311 resv_clusters = (ext4_blocks_count(sbi->s_es) >> 4312 sbi->s_cluster_bits); 4313 4314 do_div(resv_clusters, 50); 4315 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 4316 4317 atomic64_set(&sbi->s_resv_clusters, resv_clusters); 4318 } 4319 4320 static const char *ext4_quota_mode(struct super_block *sb) 4321 { 4322 #ifdef CONFIG_QUOTA 4323 if (!ext4_quota_capable(sb)) 4324 return "none"; 4325 4326 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb)) 4327 return "journalled"; 4328 else 4329 return "writeback"; 4330 #else 4331 return "disabled"; 4332 #endif 4333 } 4334 4335 static void ext4_setup_csum_trigger(struct super_block *sb, 4336 enum ext4_journal_trigger_type type, 4337 void (*trigger)( 4338 struct jbd2_buffer_trigger_type *type, 4339 struct buffer_head *bh, 4340 void *mapped_data, 4341 size_t size)) 4342 { 4343 struct ext4_sb_info *sbi = EXT4_SB(sb); 4344 4345 sbi->s_journal_triggers[type].sb = sb; 4346 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger; 4347 } 4348 4349 static void ext4_free_sbi(struct ext4_sb_info *sbi) 4350 { 4351 if (!sbi) 4352 return; 4353 4354 kfree(sbi->s_blockgroup_lock); 4355 fs_put_dax(sbi->s_daxdev, NULL); 4356 kfree(sbi); 4357 } 4358 4359 static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb) 4360 { 4361 struct ext4_sb_info *sbi; 4362 4363 sbi = kzalloc_obj(*sbi); 4364 if (!sbi) 4365 return NULL; 4366 4367 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off, 4368 NULL, NULL); 4369 4370 sbi->s_blockgroup_lock = 4371 kzalloc_obj(struct blockgroup_lock); 4372 4373 if (!sbi->s_blockgroup_lock) 4374 goto err_out; 4375 4376 sb->s_fs_info = sbi; 4377 sbi->s_sb = sb; 4378 return sbi; 4379 err_out: 4380 fs_put_dax(sbi->s_daxdev, NULL); 4381 kfree(sbi); 4382 return NULL; 4383 } 4384 4385 static void ext4_set_def_opts(struct super_block *sb, 4386 struct ext4_super_block *es) 4387 { 4388 unsigned long def_mount_opts; 4389 4390 /* Set defaults before we parse the mount options */ 4391 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 4392 set_opt(sb, INIT_INODE_TABLE); 4393 if (def_mount_opts & EXT4_DEFM_DEBUG) 4394 set_opt(sb, DEBUG); 4395 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 4396 set_opt(sb, GRPID); 4397 if (def_mount_opts & EXT4_DEFM_UID16) 4398 set_opt(sb, NO_UID32); 4399 /* xattr user namespace & acls are now defaulted on */ 4400 set_opt(sb, XATTR_USER); 4401 #ifdef CONFIG_EXT4_FS_POSIX_ACL 4402 set_opt(sb, POSIX_ACL); 4403 #endif 4404 if (ext4_has_feature_fast_commit(sb)) 4405 set_opt2(sb, JOURNAL_FAST_COMMIT); 4406 /* don't forget to enable journal_csum when metadata_csum is enabled. */ 4407 if (ext4_has_feature_metadata_csum(sb)) 4408 set_opt(sb, JOURNAL_CHECKSUM); 4409 4410 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 4411 set_opt(sb, JOURNAL_DATA); 4412 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 4413 set_opt(sb, ORDERED_DATA); 4414 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 4415 set_opt(sb, WRITEBACK_DATA); 4416 4417 if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_PANIC) 4418 set_opt(sb, ERRORS_PANIC); 4419 else if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_CONTINUE) 4420 set_opt(sb, ERRORS_CONT); 4421 else 4422 set_opt(sb, ERRORS_RO); 4423 /* block_validity enabled by default; disable with noblock_validity */ 4424 set_opt(sb, BLOCK_VALIDITY); 4425 if (def_mount_opts & EXT4_DEFM_DISCARD) 4426 set_opt(sb, DISCARD); 4427 4428 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 4429 set_opt(sb, BARRIER); 4430 4431 /* 4432 * enable delayed allocation by default 4433 * Use -o nodelalloc to turn it off 4434 */ 4435 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 4436 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 4437 set_opt(sb, DELALLOC); 4438 4439 set_opt(sb, DIOREAD_NOLOCK); 4440 } 4441 4442 static int ext4_handle_clustersize(struct super_block *sb) 4443 { 4444 struct ext4_sb_info *sbi = EXT4_SB(sb); 4445 struct ext4_super_block *es = sbi->s_es; 4446 int clustersize; 4447 4448 /* Handle clustersize */ 4449 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 4450 if (ext4_has_feature_bigalloc(sb)) { 4451 if (clustersize < sb->s_blocksize) { 4452 ext4_msg(sb, KERN_ERR, 4453 "cluster size (%d) smaller than " 4454 "block size (%lu)", clustersize, sb->s_blocksize); 4455 return -EINVAL; 4456 } 4457 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 4458 le32_to_cpu(es->s_log_block_size); 4459 } else { 4460 if (clustersize != sb->s_blocksize) { 4461 ext4_msg(sb, KERN_ERR, 4462 "fragment/cluster size (%d) != " 4463 "block size (%lu)", clustersize, sb->s_blocksize); 4464 return -EINVAL; 4465 } 4466 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) { 4467 ext4_msg(sb, KERN_ERR, 4468 "#blocks per group too big: %lu", 4469 sbi->s_blocks_per_group); 4470 return -EINVAL; 4471 } 4472 sbi->s_cluster_bits = 0; 4473 } 4474 sbi->s_clusters_per_group = le32_to_cpu(es->s_clusters_per_group); 4475 if (sbi->s_clusters_per_group > sb->s_blocksize * 8) { 4476 ext4_msg(sb, KERN_ERR, "#clusters per group too big: %lu", 4477 sbi->s_clusters_per_group); 4478 return -EINVAL; 4479 } 4480 if (sbi->s_blocks_per_group != 4481 (sbi->s_clusters_per_group * (clustersize / sb->s_blocksize))) { 4482 ext4_msg(sb, KERN_ERR, 4483 "blocks per group (%lu) and clusters per group (%lu) inconsistent", 4484 sbi->s_blocks_per_group, sbi->s_clusters_per_group); 4485 return -EINVAL; 4486 } 4487 sbi->s_cluster_ratio = clustersize / sb->s_blocksize; 4488 4489 /* Do we have standard group size of clustersize * 8 blocks ? */ 4490 if (sbi->s_blocks_per_group == clustersize << 3) 4491 set_opt2(sb, STD_GROUP_SIZE); 4492 4493 return 0; 4494 } 4495 4496 /* 4497 * ext4_atomic_write_init: Initializes filesystem min & max atomic write units. 4498 * With non-bigalloc filesystem awu will be based upon filesystem blocksize 4499 * & bdev awu units. 4500 * With bigalloc it will be based upon bigalloc cluster size & bdev awu units. 4501 * @sb: super block 4502 */ 4503 static void ext4_atomic_write_init(struct super_block *sb) 4504 { 4505 struct ext4_sb_info *sbi = EXT4_SB(sb); 4506 struct block_device *bdev = sb->s_bdev; 4507 unsigned int clustersize = EXT4_CLUSTER_SIZE(sb); 4508 4509 if (!bdev_can_atomic_write(bdev)) 4510 return; 4511 4512 if (!ext4_has_feature_extents(sb)) 4513 return; 4514 4515 sbi->s_awu_min = max(sb->s_blocksize, 4516 bdev_atomic_write_unit_min_bytes(bdev)); 4517 sbi->s_awu_max = min(clustersize, 4518 bdev_atomic_write_unit_max_bytes(bdev)); 4519 if (sbi->s_awu_min && sbi->s_awu_max && 4520 sbi->s_awu_min <= sbi->s_awu_max) { 4521 ext4_msg(sb, KERN_NOTICE, "Supports (experimental) DIO atomic writes awu_min: %u, awu_max: %u", 4522 sbi->s_awu_min, sbi->s_awu_max); 4523 } else { 4524 sbi->s_awu_min = 0; 4525 sbi->s_awu_max = 0; 4526 } 4527 } 4528 4529 static void ext4_fast_commit_init(struct super_block *sb) 4530 { 4531 struct ext4_sb_info *sbi = EXT4_SB(sb); 4532 4533 /* Initialize fast commit stuff */ 4534 atomic_set(&sbi->s_fc_subtid, 0); 4535 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]); 4536 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]); 4537 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]); 4538 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]); 4539 sbi->s_fc_bytes = 0; 4540 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE); 4541 sbi->s_fc_ineligible_tid = 0; 4542 mutex_init(&sbi->s_fc_lock); 4543 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats)); 4544 sbi->s_fc_replay_state.fc_regions = NULL; 4545 sbi->s_fc_replay_state.fc_regions_size = 0; 4546 sbi->s_fc_replay_state.fc_regions_used = 0; 4547 sbi->s_fc_replay_state.fc_regions_valid = 0; 4548 sbi->s_fc_replay_state.fc_modified_inodes = NULL; 4549 sbi->s_fc_replay_state.fc_modified_inodes_size = 0; 4550 sbi->s_fc_replay_state.fc_modified_inodes_used = 0; 4551 } 4552 4553 static int ext4_inode_info_init(struct super_block *sb, 4554 struct ext4_super_block *es) 4555 { 4556 struct ext4_sb_info *sbi = EXT4_SB(sb); 4557 4558 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 4559 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 4560 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 4561 } else { 4562 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 4563 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 4564 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) { 4565 ext4_msg(sb, KERN_ERR, "invalid first ino: %u", 4566 sbi->s_first_ino); 4567 return -EINVAL; 4568 } 4569 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 4570 (!is_power_of_2(sbi->s_inode_size)) || 4571 (sbi->s_inode_size > sb->s_blocksize)) { 4572 ext4_msg(sb, KERN_ERR, 4573 "unsupported inode size: %d", 4574 sbi->s_inode_size); 4575 ext4_msg(sb, KERN_ERR, "blocksize: %lu", sb->s_blocksize); 4576 return -EINVAL; 4577 } 4578 /* 4579 * i_atime_extra is the last extra field available for 4580 * [acm]times in struct ext4_inode. Checking for that 4581 * field should suffice to ensure we have extra space 4582 * for all three. 4583 */ 4584 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) + 4585 sizeof(((struct ext4_inode *)0)->i_atime_extra)) { 4586 sb->s_time_gran = 1; 4587 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX; 4588 } else { 4589 sb->s_time_gran = NSEC_PER_SEC; 4590 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX; 4591 } 4592 sb->s_time_min = EXT4_TIMESTAMP_MIN; 4593 } 4594 4595 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 4596 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 4597 EXT4_GOOD_OLD_INODE_SIZE; 4598 if (ext4_has_feature_extra_isize(sb)) { 4599 unsigned v, max = (sbi->s_inode_size - 4600 EXT4_GOOD_OLD_INODE_SIZE); 4601 4602 v = le16_to_cpu(es->s_want_extra_isize); 4603 if (v > max) { 4604 ext4_msg(sb, KERN_ERR, 4605 "bad s_want_extra_isize: %d", v); 4606 return -EINVAL; 4607 } 4608 if (sbi->s_want_extra_isize < v) 4609 sbi->s_want_extra_isize = v; 4610 4611 v = le16_to_cpu(es->s_min_extra_isize); 4612 if (v > max) { 4613 ext4_msg(sb, KERN_ERR, 4614 "bad s_min_extra_isize: %d", v); 4615 return -EINVAL; 4616 } 4617 if (sbi->s_want_extra_isize < v) 4618 sbi->s_want_extra_isize = v; 4619 } 4620 } 4621 4622 return 0; 4623 } 4624 4625 #if IS_ENABLED(CONFIG_UNICODE) 4626 static int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es) 4627 { 4628 const struct ext4_sb_encodings *encoding_info; 4629 struct unicode_map *encoding; 4630 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags); 4631 4632 if (!ext4_has_feature_casefold(sb) || sb->s_encoding) 4633 return 0; 4634 4635 encoding_info = ext4_sb_read_encoding(es); 4636 if (!encoding_info) { 4637 ext4_msg(sb, KERN_ERR, 4638 "Encoding requested by superblock is unknown"); 4639 return -EINVAL; 4640 } 4641 4642 encoding = utf8_load(encoding_info->version); 4643 if (IS_ERR(encoding)) { 4644 ext4_msg(sb, KERN_ERR, 4645 "can't mount with superblock charset: %s-%u.%u.%u " 4646 "not supported by the kernel. flags: 0x%x.", 4647 encoding_info->name, 4648 unicode_major(encoding_info->version), 4649 unicode_minor(encoding_info->version), 4650 unicode_rev(encoding_info->version), 4651 encoding_flags); 4652 return -EINVAL; 4653 } 4654 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: " 4655 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name, 4656 unicode_major(encoding_info->version), 4657 unicode_minor(encoding_info->version), 4658 unicode_rev(encoding_info->version), 4659 encoding_flags); 4660 4661 sb->s_encoding = encoding; 4662 sb->s_encoding_flags = encoding_flags; 4663 4664 return 0; 4665 } 4666 #else 4667 static inline int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es) 4668 { 4669 return 0; 4670 } 4671 #endif 4672 4673 static int ext4_init_metadata_csum(struct super_block *sb, struct ext4_super_block *es) 4674 { 4675 struct ext4_sb_info *sbi = EXT4_SB(sb); 4676 4677 /* Warn if metadata_csum and gdt_csum are both set. */ 4678 if (ext4_has_feature_metadata_csum(sb) && 4679 ext4_has_feature_gdt_csum(sb)) 4680 ext4_warning(sb, "metadata_csum and uninit_bg are " 4681 "redundant flags; please run fsck."); 4682 4683 /* Check for a known checksum algorithm */ 4684 if (!ext4_verify_csum_type(sb, es)) { 4685 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 4686 "unknown checksum algorithm."); 4687 return -EINVAL; 4688 } 4689 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE, 4690 ext4_orphan_file_block_trigger); 4691 4692 /* Check superblock checksum */ 4693 if (!ext4_superblock_csum_verify(sb, es)) { 4694 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 4695 "invalid superblock checksum. Run e2fsck?"); 4696 return -EFSBADCRC; 4697 } 4698 4699 /* Precompute checksum seed for all metadata */ 4700 if (ext4_has_feature_csum_seed(sb)) 4701 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 4702 else if (ext4_has_feature_metadata_csum(sb) || 4703 ext4_has_feature_ea_inode(sb)) 4704 sbi->s_csum_seed = ext4_chksum(~0, es->s_uuid, 4705 sizeof(es->s_uuid)); 4706 return 0; 4707 } 4708 4709 static int ext4_check_feature_compatibility(struct super_block *sb, 4710 struct ext4_super_block *es, 4711 int silent) 4712 { 4713 struct ext4_sb_info *sbi = EXT4_SB(sb); 4714 4715 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 4716 (ext4_has_compat_features(sb) || 4717 ext4_has_ro_compat_features(sb) || 4718 ext4_has_incompat_features(sb))) 4719 ext4_msg(sb, KERN_WARNING, 4720 "feature flags set on rev 0 fs, " 4721 "running e2fsck is recommended"); 4722 4723 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 4724 set_opt2(sb, HURD_COMPAT); 4725 if (ext4_has_feature_64bit(sb)) { 4726 ext4_msg(sb, KERN_ERR, 4727 "The Hurd can't support 64-bit file systems"); 4728 return -EINVAL; 4729 } 4730 4731 /* 4732 * ea_inode feature uses l_i_version field which is not 4733 * available in HURD_COMPAT mode. 4734 */ 4735 if (ext4_has_feature_ea_inode(sb)) { 4736 ext4_msg(sb, KERN_ERR, 4737 "ea_inode feature is not supported for Hurd"); 4738 return -EINVAL; 4739 } 4740 } 4741 4742 if (IS_EXT2_SB(sb)) { 4743 if (ext2_feature_set_ok(sb)) 4744 ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 4745 "using the ext4 subsystem"); 4746 else { 4747 /* 4748 * If we're probing be silent, if this looks like 4749 * it's actually an ext[34] filesystem. 4750 */ 4751 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 4752 return -EINVAL; 4753 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 4754 "to feature incompatibilities"); 4755 return -EINVAL; 4756 } 4757 } 4758 4759 if (IS_EXT3_SB(sb)) { 4760 if (ext3_feature_set_ok(sb)) 4761 ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 4762 "using the ext4 subsystem"); 4763 else { 4764 /* 4765 * If we're probing be silent, if this looks like 4766 * it's actually an ext4 filesystem. 4767 */ 4768 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 4769 return -EINVAL; 4770 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 4771 "to feature incompatibilities"); 4772 return -EINVAL; 4773 } 4774 } 4775 4776 /* 4777 * Check feature flags regardless of the revision level, since we 4778 * previously didn't change the revision level when setting the flags, 4779 * so there is a chance incompat flags are set on a rev 0 filesystem. 4780 */ 4781 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) 4782 return -EINVAL; 4783 4784 if (sbi->s_daxdev) { 4785 if (sb->s_blocksize == PAGE_SIZE) 4786 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags); 4787 else 4788 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n"); 4789 } 4790 4791 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) { 4792 if (ext4_has_feature_inline_data(sb)) { 4793 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem" 4794 " that may contain inline data"); 4795 return -EINVAL; 4796 } 4797 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) { 4798 ext4_msg(sb, KERN_ERR, 4799 "DAX unsupported by block device."); 4800 return -EINVAL; 4801 } 4802 } 4803 4804 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 4805 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 4806 es->s_encryption_level); 4807 return -EINVAL; 4808 } 4809 4810 return 0; 4811 } 4812 4813 static int ext4_check_geometry(struct super_block *sb, 4814 struct ext4_super_block *es) 4815 { 4816 struct ext4_sb_info *sbi = EXT4_SB(sb); 4817 __u64 blocks_count; 4818 int err; 4819 4820 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (sb->s_blocksize / 4)) { 4821 ext4_msg(sb, KERN_ERR, 4822 "Number of reserved GDT blocks insanely large: %d", 4823 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); 4824 return -EINVAL; 4825 } 4826 /* 4827 * Test whether we have more sectors than will fit in sector_t, 4828 * and whether the max offset is addressable by the page cache. 4829 */ 4830 err = generic_check_addressable(sb->s_blocksize_bits, 4831 ext4_blocks_count(es)); 4832 if (err) { 4833 ext4_msg(sb, KERN_ERR, "filesystem" 4834 " too large to mount safely on this system"); 4835 return err; 4836 } 4837 4838 /* check blocks count against device size */ 4839 blocks_count = sb_bdev_nr_blocks(sb); 4840 if (blocks_count && ext4_blocks_count(es) > blocks_count) { 4841 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 4842 "exceeds size of device (%llu blocks)", 4843 ext4_blocks_count(es), blocks_count); 4844 return -EINVAL; 4845 } 4846 4847 /* 4848 * It makes no sense for the first data block to be beyond the end 4849 * of the filesystem. 4850 */ 4851 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 4852 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4853 "block %u is beyond end of filesystem (%llu)", 4854 le32_to_cpu(es->s_first_data_block), 4855 ext4_blocks_count(es)); 4856 return -EINVAL; 4857 } 4858 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) && 4859 (sbi->s_cluster_ratio == 1)) { 4860 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4861 "block is 0 with a 1k block and cluster size"); 4862 return -EINVAL; 4863 } 4864 4865 blocks_count = (ext4_blocks_count(es) - 4866 le32_to_cpu(es->s_first_data_block) + 4867 EXT4_BLOCKS_PER_GROUP(sb) - 1); 4868 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 4869 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 4870 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu " 4871 "(block count %llu, first data block %u, " 4872 "blocks per group %lu)", blocks_count, 4873 ext4_blocks_count(es), 4874 le32_to_cpu(es->s_first_data_block), 4875 EXT4_BLOCKS_PER_GROUP(sb)); 4876 return -EINVAL; 4877 } 4878 sbi->s_groups_count = blocks_count; 4879 sbi->s_blockfile_groups = min(sbi->s_groups_count, 4880 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 4881 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) != 4882 le32_to_cpu(es->s_inodes_count)) { 4883 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu", 4884 le32_to_cpu(es->s_inodes_count), 4885 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group)); 4886 return -EINVAL; 4887 } 4888 4889 return 0; 4890 } 4891 4892 static int ext4_group_desc_init(struct super_block *sb, 4893 struct ext4_super_block *es, 4894 ext4_fsblk_t logical_sb_block, 4895 ext4_group_t *first_not_zeroed) 4896 { 4897 struct ext4_sb_info *sbi = EXT4_SB(sb); 4898 unsigned int db_count; 4899 ext4_fsblk_t block; 4900 int i; 4901 4902 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 4903 EXT4_DESC_PER_BLOCK(sb); 4904 if (ext4_has_feature_meta_bg(sb)) { 4905 if (le32_to_cpu(es->s_first_meta_bg) > db_count) { 4906 ext4_msg(sb, KERN_WARNING, 4907 "first meta block group too large: %u " 4908 "(group descriptor block count %u)", 4909 le32_to_cpu(es->s_first_meta_bg), db_count); 4910 return -EINVAL; 4911 } 4912 } 4913 rcu_assign_pointer(sbi->s_group_desc, 4914 kvmalloc_objs(struct buffer_head *, db_count)); 4915 if (sbi->s_group_desc == NULL) { 4916 ext4_msg(sb, KERN_ERR, "not enough memory"); 4917 return -ENOMEM; 4918 } 4919 4920 bgl_lock_init(sbi->s_blockgroup_lock); 4921 4922 /* Pre-read the descriptors into the buffer cache */ 4923 for (i = 0; i < db_count; i++) { 4924 block = descriptor_loc(sb, logical_sb_block, i); 4925 ext4_sb_breadahead_unmovable(sb, block); 4926 } 4927 4928 for (i = 0; i < db_count; i++) { 4929 struct buffer_head *bh; 4930 4931 block = descriptor_loc(sb, logical_sb_block, i); 4932 bh = ext4_sb_bread_unmovable(sb, block); 4933 if (IS_ERR(bh)) { 4934 ext4_msg(sb, KERN_ERR, 4935 "can't read group descriptor %d", i); 4936 sbi->s_gdb_count = i; 4937 return PTR_ERR(bh); 4938 } 4939 rcu_read_lock(); 4940 rcu_dereference(sbi->s_group_desc)[i] = bh; 4941 rcu_read_unlock(); 4942 } 4943 sbi->s_gdb_count = db_count; 4944 if (!ext4_check_descriptors(sb, logical_sb_block, first_not_zeroed)) { 4945 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 4946 return -EFSCORRUPTED; 4947 } 4948 4949 return 0; 4950 } 4951 4952 static int ext4_load_and_init_journal(struct super_block *sb, 4953 struct ext4_super_block *es, 4954 struct ext4_fs_context *ctx) 4955 { 4956 struct ext4_sb_info *sbi = EXT4_SB(sb); 4957 int err; 4958 4959 err = ext4_load_journal(sb, es, ctx->journal_devnum); 4960 if (err) 4961 return err; 4962 4963 if (ext4_has_feature_64bit(sb) && 4964 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4965 JBD2_FEATURE_INCOMPAT_64BIT)) { 4966 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 4967 goto out; 4968 } 4969 4970 if (!set_journal_csum_feature_set(sb)) { 4971 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 4972 "feature set"); 4973 goto out; 4974 } 4975 4976 if (test_opt2(sb, JOURNAL_FAST_COMMIT) && 4977 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4978 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) { 4979 ext4_msg(sb, KERN_ERR, 4980 "Failed to set fast commit journal feature"); 4981 goto out; 4982 } 4983 4984 /* We have now updated the journal if required, so we can 4985 * validate the data journaling mode. */ 4986 switch (test_opt(sb, DATA_FLAGS)) { 4987 case 0: 4988 /* No mode set, assume a default based on the journal 4989 * capabilities: ORDERED_DATA if the journal can 4990 * cope, else JOURNAL_DATA 4991 */ 4992 if (jbd2_journal_check_available_features 4993 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4994 set_opt(sb, ORDERED_DATA); 4995 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA; 4996 } else { 4997 set_opt(sb, JOURNAL_DATA); 4998 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; 4999 } 5000 break; 5001 5002 case EXT4_MOUNT_ORDERED_DATA: 5003 case EXT4_MOUNT_WRITEBACK_DATA: 5004 if (!jbd2_journal_check_available_features 5005 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 5006 ext4_msg(sb, KERN_ERR, "Journal does not support " 5007 "requested data journaling mode"); 5008 goto out; 5009 } 5010 break; 5011 default: 5012 break; 5013 } 5014 5015 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 5016 test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 5017 ext4_msg(sb, KERN_ERR, "can't mount with " 5018 "journal_async_commit in data=ordered mode"); 5019 goto out; 5020 } 5021 5022 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio); 5023 5024 sbi->s_journal->j_submit_inode_data_buffers = 5025 ext4_journal_submit_inode_data_buffers; 5026 sbi->s_journal->j_finish_inode_data_buffers = 5027 ext4_journal_finish_inode_data_buffers; 5028 5029 return 0; 5030 5031 out: 5032 ext4_journal_destroy(sbi, sbi->s_journal); 5033 return -EINVAL; 5034 } 5035 5036 static int ext4_check_journal_data_mode(struct super_block *sb) 5037 { 5038 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 5039 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with " 5040 "data=journal disables delayed allocation, " 5041 "dioread_nolock, O_DIRECT and fast_commit support!\n"); 5042 /* can't mount with both data=journal and dioread_nolock. */ 5043 clear_opt(sb, DIOREAD_NOLOCK); 5044 clear_opt2(sb, JOURNAL_FAST_COMMIT); 5045 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 5046 ext4_msg(sb, KERN_ERR, "can't mount with " 5047 "both data=journal and delalloc"); 5048 return -EINVAL; 5049 } 5050 if (test_opt(sb, DAX_ALWAYS)) { 5051 ext4_msg(sb, KERN_ERR, "can't mount with " 5052 "both data=journal and dax"); 5053 return -EINVAL; 5054 } 5055 if (ext4_has_feature_encrypt(sb)) { 5056 ext4_msg(sb, KERN_WARNING, 5057 "encrypted files will use data=ordered " 5058 "instead of data journaling mode"); 5059 } 5060 if (test_opt(sb, DELALLOC)) 5061 clear_opt(sb, DELALLOC); 5062 } else { 5063 sb->s_iflags |= SB_I_CGROUPWB; 5064 } 5065 5066 return 0; 5067 } 5068 5069 static const char *ext4_has_journal_option(struct super_block *sb) 5070 { 5071 struct ext4_sb_info *sbi = EXT4_SB(sb); 5072 5073 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) 5074 return "journal_async_commit"; 5075 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) 5076 return "journal_checksum"; 5077 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 5078 return "commit="; 5079 if (EXT4_MOUNT_DATA_FLAGS & 5080 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) 5081 return "data="; 5082 if (test_opt(sb, DATA_ERR_ABORT)) 5083 return "data_err=abort"; 5084 return NULL; 5085 } 5086 5087 /* 5088 * Limit the maximum folio order to 2048 blocks to prevent overestimation 5089 * of reserve handle credits during the folio writeback in environments 5090 * where the PAGE_SIZE exceeds 4KB. 5091 */ 5092 #define EXT4_MAX_PAGECACHE_ORDER(sb) \ 5093 umin(MAX_PAGECACHE_ORDER, (11 + (sb)->s_blocksize_bits - PAGE_SHIFT)) 5094 static void ext4_set_max_mapping_order(struct super_block *sb) 5095 { 5096 struct ext4_sb_info *sbi = EXT4_SB(sb); 5097 5098 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 5099 sbi->s_max_folio_order = sbi->s_min_folio_order; 5100 else 5101 sbi->s_max_folio_order = EXT4_MAX_PAGECACHE_ORDER(sb); 5102 } 5103 5104 static int ext4_check_large_folio(struct super_block *sb) 5105 { 5106 const char *err_str = NULL; 5107 5108 if (ext4_has_feature_encrypt(sb)) 5109 err_str = "encrypt"; 5110 5111 if (!err_str) { 5112 ext4_set_max_mapping_order(sb); 5113 } else if (sb->s_blocksize > PAGE_SIZE) { 5114 ext4_msg(sb, KERN_ERR, "bs(%lu) > ps(%lu) unsupported for %s", 5115 sb->s_blocksize, PAGE_SIZE, err_str); 5116 return -EINVAL; 5117 } 5118 5119 return 0; 5120 } 5121 5122 static int ext4_load_super(struct super_block *sb, ext4_fsblk_t *lsb, 5123 int silent) 5124 { 5125 struct ext4_sb_info *sbi = EXT4_SB(sb); 5126 struct ext4_super_block *es; 5127 ext4_fsblk_t logical_sb_block; 5128 unsigned long offset = 0; 5129 struct buffer_head *bh; 5130 int ret = -EINVAL; 5131 int blocksize; 5132 5133 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 5134 if (!blocksize) { 5135 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 5136 return -EINVAL; 5137 } 5138 5139 /* 5140 * The ext4 superblock will not be buffer aligned for other than 1kB 5141 * block sizes. We need to calculate the offset from buffer start. 5142 */ 5143 if (blocksize != EXT4_MIN_BLOCK_SIZE) { 5144 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE; 5145 offset = do_div(logical_sb_block, blocksize); 5146 } else { 5147 logical_sb_block = sbi->s_sb_block; 5148 } 5149 5150 bh = ext4_sb_bread_unmovable(sb, logical_sb_block); 5151 if (IS_ERR(bh)) { 5152 ext4_msg(sb, KERN_ERR, "unable to read superblock"); 5153 return PTR_ERR(bh); 5154 } 5155 /* 5156 * Note: s_es must be initialized as soon as possible because 5157 * some ext4 macro-instructions depend on its value 5158 */ 5159 es = (struct ext4_super_block *) (bh->b_data + offset); 5160 sbi->s_es = es; 5161 sb->s_magic = le16_to_cpu(es->s_magic); 5162 if (sb->s_magic != EXT4_SUPER_MAGIC) { 5163 if (!silent) 5164 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 5165 goto out; 5166 } 5167 5168 if (le32_to_cpu(es->s_log_block_size) > 5169 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 5170 ext4_msg(sb, KERN_ERR, 5171 "Invalid log block size: %u", 5172 le32_to_cpu(es->s_log_block_size)); 5173 goto out; 5174 } 5175 if (le32_to_cpu(es->s_log_cluster_size) > 5176 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 5177 ext4_msg(sb, KERN_ERR, 5178 "Invalid log cluster size: %u", 5179 le32_to_cpu(es->s_log_cluster_size)); 5180 goto out; 5181 } 5182 5183 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 5184 5185 /* 5186 * If the default block size is not the same as the real block size, 5187 * we need to reload it. 5188 */ 5189 if (sb->s_blocksize == blocksize) 5190 goto success; 5191 5192 /* 5193 * bh must be released before kill_bdev(), otherwise 5194 * it won't be freed and its page also. kill_bdev() 5195 * is called by sb_set_blocksize(). 5196 */ 5197 brelse(bh); 5198 /* Validate the filesystem blocksize */ 5199 if (!sb_set_blocksize(sb, blocksize)) { 5200 ext4_msg(sb, KERN_ERR, "bad block size %d", 5201 blocksize); 5202 bh = NULL; 5203 goto out; 5204 } 5205 5206 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE; 5207 offset = do_div(logical_sb_block, blocksize); 5208 bh = ext4_sb_bread_unmovable(sb, logical_sb_block); 5209 if (IS_ERR(bh)) { 5210 ext4_msg(sb, KERN_ERR, "Can't read superblock on 2nd try"); 5211 ret = PTR_ERR(bh); 5212 bh = NULL; 5213 goto out; 5214 } 5215 es = (struct ext4_super_block *)(bh->b_data + offset); 5216 sbi->s_es = es; 5217 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 5218 ext4_msg(sb, KERN_ERR, "Magic mismatch, very weird!"); 5219 goto out; 5220 } 5221 5222 success: 5223 sbi->s_min_folio_order = get_order(blocksize); 5224 *lsb = logical_sb_block; 5225 sbi->s_sbh = bh; 5226 return 0; 5227 out: 5228 brelse(bh); 5229 return ret; 5230 } 5231 5232 static int ext4_hash_info_init(struct super_block *sb) 5233 { 5234 struct ext4_sb_info *sbi = EXT4_SB(sb); 5235 struct ext4_super_block *es = sbi->s_es; 5236 unsigned int i; 5237 5238 sbi->s_def_hash_version = es->s_def_hash_version; 5239 5240 if (sbi->s_def_hash_version > DX_HASH_LAST) { 5241 ext4_msg(sb, KERN_ERR, 5242 "Invalid default hash set in the superblock"); 5243 return -EINVAL; 5244 } else if (sbi->s_def_hash_version == DX_HASH_SIPHASH) { 5245 ext4_msg(sb, KERN_ERR, 5246 "SIPHASH is not a valid default hash value"); 5247 return -EINVAL; 5248 } 5249 5250 for (i = 0; i < 4; i++) 5251 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 5252 5253 if (ext4_has_feature_dir_index(sb)) { 5254 i = le32_to_cpu(es->s_flags); 5255 if (i & EXT2_FLAGS_UNSIGNED_HASH) 5256 sbi->s_hash_unsigned = 3; 5257 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 5258 #ifdef __CHAR_UNSIGNED__ 5259 if (!sb_rdonly(sb)) 5260 es->s_flags |= 5261 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 5262 sbi->s_hash_unsigned = 3; 5263 #else 5264 if (!sb_rdonly(sb)) 5265 es->s_flags |= 5266 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 5267 #endif 5268 } 5269 } 5270 return 0; 5271 } 5272 5273 static int ext4_block_group_meta_init(struct super_block *sb, int silent) 5274 { 5275 struct ext4_sb_info *sbi = EXT4_SB(sb); 5276 struct ext4_super_block *es = sbi->s_es; 5277 int has_huge_files; 5278 5279 has_huge_files = ext4_has_feature_huge_file(sb); 5280 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 5281 has_huge_files); 5282 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 5283 5284 sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 5285 if (ext4_has_feature_64bit(sb)) { 5286 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 5287 sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 5288 !is_power_of_2(sbi->s_desc_size)) { 5289 ext4_msg(sb, KERN_ERR, 5290 "unsupported descriptor size %lu", 5291 sbi->s_desc_size); 5292 return -EINVAL; 5293 } 5294 } else 5295 sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 5296 5297 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 5298 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 5299 5300 sbi->s_inodes_per_block = sb->s_blocksize / EXT4_INODE_SIZE(sb); 5301 if (sbi->s_inodes_per_block == 0 || sbi->s_blocks_per_group == 0) { 5302 if (!silent) 5303 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 5304 return -EINVAL; 5305 } 5306 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || 5307 sbi->s_inodes_per_group > sb->s_blocksize * 8) { 5308 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", 5309 sbi->s_inodes_per_group); 5310 return -EINVAL; 5311 } 5312 sbi->s_itb_per_group = sbi->s_inodes_per_group / 5313 sbi->s_inodes_per_block; 5314 sbi->s_desc_per_block = sb->s_blocksize / EXT4_DESC_SIZE(sb); 5315 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY; 5316 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 5317 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 5318 5319 return 0; 5320 } 5321 5322 /* 5323 * It's hard to get stripe aligned blocks if stripe is not aligned with 5324 * cluster, just disable stripe and alert user to simplify code and avoid 5325 * stripe aligned allocation which will rarely succeed. 5326 */ 5327 static bool ext4_is_stripe_incompatible(struct super_block *sb, unsigned long stripe) 5328 { 5329 struct ext4_sb_info *sbi = EXT4_SB(sb); 5330 return (stripe > 0 && sbi->s_cluster_ratio > 1 && 5331 stripe % sbi->s_cluster_ratio != 0); 5332 } 5333 5334 static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb) 5335 { 5336 struct ext4_super_block *es = NULL; 5337 struct ext4_sb_info *sbi = EXT4_SB(sb); 5338 ext4_fsblk_t logical_sb_block; 5339 struct inode *root; 5340 int needs_recovery; 5341 int err; 5342 ext4_group_t first_not_zeroed; 5343 struct ext4_fs_context *ctx = fc->fs_private; 5344 int silent = fc->sb_flags & SB_SILENT; 5345 5346 /* Set defaults for the variables that will be set during parsing */ 5347 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) 5348 ctx->journal_ioprio = EXT4_DEF_JOURNAL_IOPRIO; 5349 5350 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 5351 sbi->s_sectors_written_start = 5352 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]); 5353 5354 err = ext4_load_super(sb, &logical_sb_block, silent); 5355 if (err) 5356 goto out_fail; 5357 5358 es = sbi->s_es; 5359 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 5360 5361 err = ext4_init_metadata_csum(sb, es); 5362 if (err) 5363 goto failed_mount; 5364 5365 ext4_set_def_opts(sb, es); 5366 5367 sbi->s_resuid = make_kuid(&init_user_ns, ext4_get_resuid(es)); 5368 sbi->s_resgid = make_kgid(&init_user_ns, ext4_get_resuid(es)); 5369 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 5370 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 5371 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 5372 sbi->s_sb_update_kb = EXT4_DEF_SB_UPDATE_INTERVAL_KB; 5373 sbi->s_sb_update_sec = EXT4_DEF_SB_UPDATE_INTERVAL_SEC; 5374 5375 /* 5376 * set default s_li_wait_mult for lazyinit, for the case there is 5377 * no mount option specified. 5378 */ 5379 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 5380 5381 err = ext4_inode_info_init(sb, es); 5382 if (err) 5383 goto failed_mount; 5384 5385 err = parse_apply_sb_mount_options(sb, ctx); 5386 if (err < 0) 5387 goto failed_mount; 5388 5389 sbi->s_def_mount_opt = sbi->s_mount_opt; 5390 sbi->s_def_mount_opt2 = sbi->s_mount_opt2; 5391 5392 err = ext4_check_opt_consistency(fc, sb); 5393 if (err < 0) 5394 goto failed_mount; 5395 5396 ext4_apply_options(fc, sb); 5397 5398 err = ext4_check_large_folio(sb); 5399 if (err < 0) 5400 goto failed_mount; 5401 5402 err = ext4_encoding_init(sb, es); 5403 if (err) 5404 goto failed_mount; 5405 5406 err = ext4_check_journal_data_mode(sb); 5407 if (err) 5408 goto failed_mount; 5409 5410 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 5411 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 5412 5413 /* HSM events are allowed by default. */ 5414 sb->s_iflags |= SB_I_ALLOW_HSM; 5415 5416 err = ext4_check_feature_compatibility(sb, es, silent); 5417 if (err) 5418 goto failed_mount; 5419 5420 err = ext4_block_group_meta_init(sb, silent); 5421 if (err) 5422 goto failed_mount; 5423 5424 err = ext4_hash_info_init(sb); 5425 if (err) 5426 goto failed_mount; 5427 5428 err = ext4_handle_clustersize(sb); 5429 if (err) 5430 goto failed_mount; 5431 5432 err = ext4_check_geometry(sb, es); 5433 if (err) 5434 goto failed_mount; 5435 5436 timer_setup(&sbi->s_err_report, print_daily_error_info, 0); 5437 spin_lock_init(&sbi->s_error_lock); 5438 mutex_init(&sbi->s_error_notify_mutex); 5439 INIT_WORK(&sbi->s_sb_upd_work, update_super_work); 5440 5441 err = ext4_group_desc_init(sb, es, logical_sb_block, &first_not_zeroed); 5442 if (err) 5443 goto failed_mount3; 5444 5445 err = ext4_es_register_shrinker(sbi); 5446 if (err) 5447 goto failed_mount3; 5448 5449 sbi->s_stripe = ext4_get_stripe_size(sbi); 5450 if (ext4_is_stripe_incompatible(sb, sbi->s_stripe)) { 5451 ext4_msg(sb, KERN_WARNING, 5452 "stripe (%lu) is not aligned with cluster size (%u), " 5453 "stripe is disabled", 5454 sbi->s_stripe, sbi->s_cluster_ratio); 5455 sbi->s_stripe = 0; 5456 } 5457 sbi->s_extent_max_zeroout_kb = 32; 5458 5459 /* 5460 * set up enough so that it can read an inode 5461 */ 5462 sb->s_op = &ext4_sops; 5463 sb->s_export_op = &ext4_export_ops; 5464 sb->s_xattr = ext4_xattr_handlers; 5465 #ifdef CONFIG_FS_ENCRYPTION 5466 sb->s_cop = &ext4_cryptops; 5467 #endif 5468 #ifdef CONFIG_FS_VERITY 5469 sb->s_vop = &ext4_verityops; 5470 #endif 5471 #ifdef CONFIG_QUOTA 5472 sb->dq_op = &ext4_quota_operations; 5473 if (ext4_has_feature_quota(sb)) 5474 sb->s_qcop = &dquot_quotactl_sysfile_ops; 5475 else 5476 sb->s_qcop = &ext4_qctl_operations; 5477 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 5478 #endif 5479 super_set_uuid(sb, es->s_uuid, sizeof(es->s_uuid)); 5480 super_set_sysfs_name_bdev(sb); 5481 5482 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 5483 mutex_init(&sbi->s_orphan_lock); 5484 5485 spin_lock_init(&sbi->s_bdev_wb_lock); 5486 5487 ext4_atomic_write_init(sb); 5488 ext4_fast_commit_init(sb); 5489 5490 sb->s_root = NULL; 5491 5492 needs_recovery = (es->s_last_orphan != 0 || 5493 ext4_has_feature_orphan_present(sb) || 5494 ext4_has_feature_journal_needs_recovery(sb)); 5495 5496 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) { 5497 err = ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)); 5498 if (err) 5499 goto failed_mount3a; 5500 } 5501 5502 err = -EINVAL; 5503 /* 5504 * The first inode we look at is the journal inode. Don't try 5505 * root first: it may be modified in the journal! 5506 */ 5507 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 5508 err = ext4_load_and_init_journal(sb, es, ctx); 5509 if (err) 5510 goto failed_mount3a; 5511 if (bdev_read_only(sb->s_bdev)) 5512 needs_recovery = 0; 5513 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && 5514 ext4_has_feature_journal_needs_recovery(sb)) { 5515 ext4_msg(sb, KERN_ERR, "required journal recovery " 5516 "suppressed and not mounted read-only"); 5517 goto failed_mount3a; 5518 } else { 5519 const char *journal_option; 5520 5521 /* Nojournal mode, all journal mount options are illegal */ 5522 journal_option = ext4_has_journal_option(sb); 5523 if (journal_option != NULL) { 5524 ext4_msg(sb, KERN_ERR, 5525 "can't mount with %s, fs mounted w/o journal", 5526 journal_option); 5527 goto failed_mount3a; 5528 } 5529 5530 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM; 5531 clear_opt(sb, JOURNAL_CHECKSUM); 5532 clear_opt(sb, DATA_FLAGS); 5533 clear_opt2(sb, JOURNAL_FAST_COMMIT); 5534 sbi->s_journal = NULL; 5535 needs_recovery = 0; 5536 } 5537 5538 if (!test_opt(sb, NO_MBCACHE)) { 5539 sbi->s_ea_block_cache = ext4_xattr_create_cache(); 5540 if (!sbi->s_ea_block_cache) { 5541 ext4_msg(sb, KERN_ERR, 5542 "Failed to create ea_block_cache"); 5543 err = -EINVAL; 5544 goto failed_mount_wq; 5545 } 5546 5547 if (ext4_has_feature_ea_inode(sb)) { 5548 sbi->s_ea_inode_cache = ext4_xattr_create_cache(); 5549 if (!sbi->s_ea_inode_cache) { 5550 ext4_msg(sb, KERN_ERR, 5551 "Failed to create ea_inode_cache"); 5552 err = -EINVAL; 5553 goto failed_mount_wq; 5554 } 5555 } 5556 } 5557 5558 /* 5559 * Get the # of file system overhead blocks from the 5560 * superblock if present. 5561 */ 5562 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 5563 /* ignore the precalculated value if it is ridiculous */ 5564 if (sbi->s_overhead > ext4_blocks_count(es)) 5565 sbi->s_overhead = 0; 5566 /* 5567 * If the bigalloc feature is not enabled recalculating the 5568 * overhead doesn't take long, so we might as well just redo 5569 * it to make sure we are using the correct value. 5570 */ 5571 if (!ext4_has_feature_bigalloc(sb)) 5572 sbi->s_overhead = 0; 5573 if (sbi->s_overhead == 0) { 5574 err = ext4_calculate_overhead(sb); 5575 if (err) 5576 goto failed_mount_wq; 5577 } 5578 5579 /* 5580 * The maximum number of concurrent works can be high and 5581 * concurrency isn't really necessary. Limit it to 1. 5582 */ 5583 EXT4_SB(sb)->rsv_conversion_wq = 5584 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 5585 if (!EXT4_SB(sb)->rsv_conversion_wq) { 5586 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 5587 err = -ENOMEM; 5588 goto failed_mount4; 5589 } 5590 5591 /* 5592 * The jbd2_journal_load will have done any necessary log recovery, 5593 * so we can safely mount the rest of the filesystem now. 5594 */ 5595 5596 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL); 5597 if (IS_ERR(root)) { 5598 ext4_msg(sb, KERN_ERR, "get root inode failed"); 5599 err = PTR_ERR(root); 5600 root = NULL; 5601 goto failed_mount4; 5602 } 5603 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 5604 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 5605 iput(root); 5606 err = -EFSCORRUPTED; 5607 goto failed_mount4; 5608 } 5609 5610 generic_set_sb_d_ops(sb); 5611 sb->s_root = d_make_root(root); 5612 if (!sb->s_root) { 5613 ext4_msg(sb, KERN_ERR, "get root dentry failed"); 5614 err = -ENOMEM; 5615 goto failed_mount4; 5616 } 5617 5618 err = ext4_setup_super(sb, es, sb_rdonly(sb)); 5619 if (err == -EROFS) { 5620 sb->s_flags |= SB_RDONLY; 5621 } else if (err) 5622 goto failed_mount4a; 5623 5624 ext4_set_resv_clusters(sb); 5625 5626 if (test_opt(sb, BLOCK_VALIDITY)) { 5627 err = ext4_setup_system_zone(sb); 5628 if (err) { 5629 ext4_msg(sb, KERN_ERR, "failed to initialize system " 5630 "zone (%d)", err); 5631 goto failed_mount4a; 5632 } 5633 } 5634 ext4_fc_replay_cleanup(sb); 5635 5636 ext4_ext_init(sb); 5637 5638 /* 5639 * Enable optimize_scan if number of groups is > threshold. This can be 5640 * turned off by passing "mb_optimize_scan=0". This can also be 5641 * turned on forcefully by passing "mb_optimize_scan=1". 5642 */ 5643 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) { 5644 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD) 5645 set_opt2(sb, MB_OPTIMIZE_SCAN); 5646 else 5647 clear_opt2(sb, MB_OPTIMIZE_SCAN); 5648 } 5649 5650 err = ext4_percpu_param_init(sbi); 5651 if (err) 5652 goto failed_mount5; 5653 5654 err = ext4_mb_init(sb); 5655 if (err) { 5656 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 5657 err); 5658 goto failed_mount5; 5659 } 5660 5661 /* 5662 * We can only set up the journal commit callback once 5663 * mballoc is initialized 5664 */ 5665 if (sbi->s_journal) 5666 sbi->s_journal->j_commit_callback = 5667 ext4_journal_commit_callback; 5668 5669 if (ext4_has_feature_flex_bg(sb)) 5670 if (!ext4_fill_flex_info(sb)) { 5671 ext4_msg(sb, KERN_ERR, 5672 "unable to initialize " 5673 "flex_bg meta info!"); 5674 err = -ENOMEM; 5675 goto failed_mount6; 5676 } 5677 5678 err = ext4_register_li_request(sb, first_not_zeroed); 5679 if (err) 5680 goto failed_mount6; 5681 5682 err = ext4_init_orphan_info(sb); 5683 if (err) 5684 goto failed_mount7; 5685 #ifdef CONFIG_QUOTA 5686 /* Enable quota usage during mount. */ 5687 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { 5688 err = ext4_enable_quotas(sb); 5689 if (err) 5690 goto failed_mount8; 5691 } 5692 #endif /* CONFIG_QUOTA */ 5693 5694 /* 5695 * Save the original bdev mapping's wb_err value which could be 5696 * used to detect the metadata async write error. 5697 */ 5698 errseq_check_and_advance(&sb->s_bdev->bd_mapping->wb_err, 5699 &sbi->s_bdev_wb_err); 5700 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 5701 ext4_orphan_cleanup(sb, es); 5702 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 5703 /* 5704 * Update the checksum after updating free space/inode counters and 5705 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect 5706 * checksum in the buffer cache until it is written out and 5707 * e2fsprogs programs trying to open a file system immediately 5708 * after it is mounted can fail. 5709 */ 5710 ext4_superblock_csum_set(sb); 5711 if (needs_recovery) { 5712 ext4_msg(sb, KERN_INFO, "recovery complete"); 5713 err = ext4_mark_recovery_complete(sb, es); 5714 if (err) 5715 goto failed_mount9; 5716 } 5717 5718 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev)) { 5719 ext4_msg(sb, KERN_WARNING, 5720 "mounting with \"discard\" option, but the device does not support discard"); 5721 clear_opt(sb, DISCARD); 5722 } 5723 5724 if (es->s_error_count) { 5725 sbi->s_err_report_sec = 5*60; /* first time 5 minutes */ 5726 mod_timer(&sbi->s_err_report, 5727 jiffies + secs_to_jiffies(sbi->s_err_report_sec)); 5728 } 5729 sbi->s_err_report_sec = 24*60*60; /* Once a day */ 5730 5731 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 5732 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 5733 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 5734 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 5735 atomic_set(&sbi->s_warning_count, 0); 5736 atomic_set(&sbi->s_msg_count, 0); 5737 5738 /* Register sysfs after all initializations are complete. */ 5739 err = ext4_register_sysfs(sb); 5740 if (err) 5741 goto failed_mount9; 5742 5743 return 0; 5744 5745 failed_mount9: 5746 ext4_quotas_off(sb, EXT4_MAXQUOTAS); 5747 failed_mount8: __maybe_unused 5748 ext4_release_orphan_info(sb); 5749 failed_mount7: 5750 ext4_unregister_li_request(sb); 5751 failed_mount6: 5752 ext4_mb_release(sb); 5753 ext4_flex_groups_free(sbi); 5754 failed_mount5: 5755 ext4_percpu_param_destroy(sbi); 5756 ext4_ext_release(sb); 5757 ext4_release_system_zone(sb); 5758 failed_mount4a: 5759 dput(sb->s_root); 5760 sb->s_root = NULL; 5761 failed_mount4: 5762 ext4_msg(sb, KERN_ERR, "mount failed"); 5763 if (EXT4_SB(sb)->rsv_conversion_wq) 5764 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 5765 failed_mount_wq: 5766 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 5767 sbi->s_ea_inode_cache = NULL; 5768 5769 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 5770 sbi->s_ea_block_cache = NULL; 5771 5772 if (sbi->s_journal) { 5773 ext4_journal_destroy(sbi, sbi->s_journal); 5774 } 5775 failed_mount3a: 5776 ext4_es_unregister_shrinker(sbi); 5777 failed_mount3: 5778 /* flush s_sb_upd_work before sbi destroy */ 5779 flush_work(&sbi->s_sb_upd_work); 5780 ext4_stop_mmpd(sbi); 5781 timer_delete_sync(&sbi->s_err_report); 5782 ext4_group_desc_free(sbi); 5783 failed_mount: 5784 #if IS_ENABLED(CONFIG_UNICODE) 5785 utf8_unload(sb->s_encoding); 5786 #endif 5787 5788 #ifdef CONFIG_QUOTA 5789 for (unsigned int i = 0; i < EXT4_MAXQUOTAS; i++) 5790 kfree(get_qf_name(sb, sbi, i)); 5791 #endif 5792 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy); 5793 brelse(sbi->s_sbh); 5794 if (sbi->s_journal_bdev_file) { 5795 invalidate_bdev(file_bdev(sbi->s_journal_bdev_file)); 5796 bdev_fput(sbi->s_journal_bdev_file); 5797 } 5798 out_fail: 5799 invalidate_bdev(sb->s_bdev); 5800 sb->s_fs_info = NULL; 5801 return err; 5802 } 5803 5804 static int ext4_fill_super(struct super_block *sb, struct fs_context *fc) 5805 { 5806 struct ext4_fs_context *ctx = fc->fs_private; 5807 struct ext4_sb_info *sbi; 5808 const char *descr; 5809 int ret; 5810 5811 sbi = ext4_alloc_sbi(sb); 5812 if (!sbi) 5813 return -ENOMEM; 5814 5815 fc->s_fs_info = sbi; 5816 5817 /* Cleanup superblock name */ 5818 strreplace(sb->s_id, '/', '!'); 5819 5820 sbi->s_sb_block = 1; /* Default super block location */ 5821 if (ctx->spec & EXT4_SPEC_s_sb_block) 5822 sbi->s_sb_block = ctx->s_sb_block; 5823 5824 ret = __ext4_fill_super(fc, sb); 5825 if (ret < 0) 5826 goto free_sbi; 5827 5828 if (sbi->s_journal) { 5829 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 5830 descr = " journalled data mode"; 5831 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 5832 descr = " ordered data mode"; 5833 else 5834 descr = " writeback data mode"; 5835 } else 5836 descr = "out journal"; 5837 5838 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 5839 ext4_msg(sb, KERN_INFO, "mounted filesystem %pU %s with%s. " 5840 "Quota mode: %s.", &sb->s_uuid, 5841 sb_rdonly(sb) ? "ro" : "r/w", descr, 5842 ext4_quota_mode(sb)); 5843 5844 /* Update the s_overhead_clusters if necessary */ 5845 ext4_update_overhead(sb, false); 5846 return 0; 5847 5848 free_sbi: 5849 ext4_free_sbi(sbi); 5850 fc->s_fs_info = NULL; 5851 return ret; 5852 } 5853 5854 static int ext4_get_tree(struct fs_context *fc) 5855 { 5856 return get_tree_bdev(fc, ext4_fill_super); 5857 } 5858 5859 /* 5860 * Setup any per-fs journal parameters now. We'll do this both on 5861 * initial mount, once the journal has been initialised but before we've 5862 * done any recovery; and again on any subsequent remount. 5863 */ 5864 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 5865 { 5866 struct ext4_sb_info *sbi = EXT4_SB(sb); 5867 5868 journal->j_commit_interval = sbi->s_commit_interval; 5869 journal->j_min_batch_time = sbi->s_min_batch_time; 5870 journal->j_max_batch_time = sbi->s_max_batch_time; 5871 ext4_fc_init(sb, journal); 5872 5873 write_lock(&journal->j_state_lock); 5874 if (test_opt(sb, BARRIER)) 5875 journal->j_flags |= JBD2_BARRIER; 5876 else 5877 journal->j_flags &= ~JBD2_BARRIER; 5878 /* 5879 * Always enable journal cycle record option, letting the journal 5880 * records log transactions continuously between each mount. 5881 */ 5882 journal->j_flags |= JBD2_CYCLE_RECORD; 5883 write_unlock(&journal->j_state_lock); 5884 } 5885 5886 static struct inode *ext4_get_journal_inode(struct super_block *sb, 5887 unsigned int journal_inum) 5888 { 5889 struct inode *journal_inode; 5890 5891 /* 5892 * Test for the existence of a valid inode on disk. Bad things 5893 * happen if we iget() an unused inode, as the subsequent iput() 5894 * will try to delete it. 5895 */ 5896 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL); 5897 if (IS_ERR(journal_inode)) { 5898 ext4_msg(sb, KERN_ERR, "no journal found"); 5899 return ERR_CAST(journal_inode); 5900 } 5901 if (!journal_inode->i_nlink) { 5902 make_bad_inode(journal_inode); 5903 iput(journal_inode); 5904 ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 5905 return ERR_PTR(-EFSCORRUPTED); 5906 } 5907 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) { 5908 ext4_msg(sb, KERN_ERR, "invalid journal inode"); 5909 iput(journal_inode); 5910 return ERR_PTR(-EFSCORRUPTED); 5911 } 5912 5913 ext4_debug("Journal inode found at %p: %lld bytes\n", 5914 journal_inode, journal_inode->i_size); 5915 return journal_inode; 5916 } 5917 5918 static int ext4_journal_bmap(journal_t *journal, sector_t *block) 5919 { 5920 struct ext4_map_blocks map; 5921 int ret; 5922 5923 if (journal->j_inode == NULL) 5924 return 0; 5925 5926 map.m_lblk = *block; 5927 map.m_len = 1; 5928 ret = ext4_map_blocks(NULL, journal->j_inode, &map, 0); 5929 if (ret <= 0) { 5930 ext4_msg(journal->j_inode->i_sb, KERN_CRIT, 5931 "journal bmap failed: block %llu ret %d\n", 5932 *block, ret); 5933 jbd2_journal_abort(journal, ret ? ret : -EFSCORRUPTED); 5934 return ret; 5935 } 5936 *block = map.m_pblk; 5937 return 0; 5938 } 5939 5940 static journal_t *ext4_open_inode_journal(struct super_block *sb, 5941 unsigned int journal_inum) 5942 { 5943 struct inode *journal_inode; 5944 journal_t *journal; 5945 5946 journal_inode = ext4_get_journal_inode(sb, journal_inum); 5947 if (IS_ERR(journal_inode)) 5948 return ERR_CAST(journal_inode); 5949 5950 journal = jbd2_journal_init_inode(journal_inode); 5951 if (IS_ERR(journal)) { 5952 ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 5953 iput(journal_inode); 5954 return ERR_CAST(journal); 5955 } 5956 journal->j_private = sb; 5957 journal->j_bmap = ext4_journal_bmap; 5958 ext4_init_journal_params(sb, journal); 5959 return journal; 5960 } 5961 5962 static struct file *ext4_get_journal_blkdev(struct super_block *sb, 5963 dev_t j_dev, ext4_fsblk_t *j_start, 5964 ext4_fsblk_t *j_len) 5965 { 5966 struct buffer_head *bh; 5967 struct block_device *bdev; 5968 struct file *bdev_file; 5969 int hblock, blocksize; 5970 ext4_fsblk_t sb_block; 5971 unsigned long offset; 5972 struct ext4_super_block *es; 5973 int errno; 5974 5975 bdev_file = bdev_file_open_by_dev(j_dev, 5976 BLK_OPEN_READ | BLK_OPEN_WRITE | BLK_OPEN_RESTRICT_WRITES, 5977 sb, &fs_holder_ops); 5978 if (IS_ERR(bdev_file)) { 5979 ext4_msg(sb, KERN_ERR, 5980 "failed to open journal device unknown-block(%u,%u) %ld", 5981 MAJOR(j_dev), MINOR(j_dev), PTR_ERR(bdev_file)); 5982 return bdev_file; 5983 } 5984 5985 bdev = file_bdev(bdev_file); 5986 blocksize = sb->s_blocksize; 5987 hblock = bdev_logical_block_size(bdev); 5988 if (blocksize < hblock) { 5989 ext4_msg(sb, KERN_ERR, 5990 "blocksize too small for journal device"); 5991 errno = -EINVAL; 5992 goto out_bdev; 5993 } 5994 5995 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 5996 offset = EXT4_MIN_BLOCK_SIZE % blocksize; 5997 set_blocksize(bdev_file, blocksize); 5998 bh = __bread(bdev, sb_block, blocksize); 5999 if (!bh) { 6000 ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 6001 "external journal"); 6002 errno = -EINVAL; 6003 goto out_bdev; 6004 } 6005 6006 es = (struct ext4_super_block *) (bh->b_data + offset); 6007 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 6008 !(le32_to_cpu(es->s_feature_incompat) & 6009 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 6010 ext4_msg(sb, KERN_ERR, "external journal has bad superblock"); 6011 errno = -EFSCORRUPTED; 6012 goto out_bh; 6013 } 6014 6015 if ((le32_to_cpu(es->s_feature_ro_compat) & 6016 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 6017 es->s_checksum != ext4_superblock_csum(es)) { 6018 ext4_msg(sb, KERN_ERR, "external journal has corrupt superblock"); 6019 errno = -EFSCORRUPTED; 6020 goto out_bh; 6021 } 6022 6023 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 6024 ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 6025 errno = -EFSCORRUPTED; 6026 goto out_bh; 6027 } 6028 6029 *j_start = sb_block + 1; 6030 *j_len = ext4_blocks_count(es); 6031 brelse(bh); 6032 return bdev_file; 6033 6034 out_bh: 6035 brelse(bh); 6036 out_bdev: 6037 bdev_fput(bdev_file); 6038 return ERR_PTR(errno); 6039 } 6040 6041 static journal_t *ext4_open_dev_journal(struct super_block *sb, 6042 dev_t j_dev) 6043 { 6044 journal_t *journal; 6045 ext4_fsblk_t j_start; 6046 ext4_fsblk_t j_len; 6047 struct file *bdev_file; 6048 int errno = 0; 6049 6050 bdev_file = ext4_get_journal_blkdev(sb, j_dev, &j_start, &j_len); 6051 if (IS_ERR(bdev_file)) 6052 return ERR_CAST(bdev_file); 6053 6054 journal = jbd2_journal_init_dev(file_bdev(bdev_file), sb->s_bdev, j_start, 6055 j_len, sb->s_blocksize); 6056 if (IS_ERR(journal)) { 6057 ext4_msg(sb, KERN_ERR, "failed to create device journal"); 6058 errno = PTR_ERR(journal); 6059 goto out_bdev; 6060 } 6061 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 6062 ext4_msg(sb, KERN_ERR, "External journal has more than one " 6063 "user (unsupported) - %d", 6064 be32_to_cpu(journal->j_superblock->s_nr_users)); 6065 errno = -EINVAL; 6066 goto out_journal; 6067 } 6068 journal->j_private = sb; 6069 EXT4_SB(sb)->s_journal_bdev_file = bdev_file; 6070 ext4_init_journal_params(sb, journal); 6071 return journal; 6072 6073 out_journal: 6074 ext4_journal_destroy(EXT4_SB(sb), journal); 6075 out_bdev: 6076 bdev_fput(bdev_file); 6077 return ERR_PTR(errno); 6078 } 6079 6080 static int ext4_load_journal(struct super_block *sb, 6081 struct ext4_super_block *es, 6082 unsigned long journal_devnum) 6083 { 6084 journal_t *journal; 6085 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 6086 dev_t journal_dev; 6087 int err = 0; 6088 int really_read_only; 6089 int journal_dev_ro; 6090 6091 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 6092 return -EFSCORRUPTED; 6093 6094 if (journal_devnum && 6095 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 6096 ext4_msg(sb, KERN_INFO, "external journal device major/minor " 6097 "numbers have changed"); 6098 journal_dev = new_decode_dev(journal_devnum); 6099 } else 6100 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 6101 6102 if (journal_inum && journal_dev) { 6103 ext4_msg(sb, KERN_ERR, 6104 "filesystem has both journal inode and journal device!"); 6105 return -EINVAL; 6106 } 6107 6108 if (journal_inum) { 6109 journal = ext4_open_inode_journal(sb, journal_inum); 6110 if (IS_ERR(journal)) 6111 return PTR_ERR(journal); 6112 } else { 6113 journal = ext4_open_dev_journal(sb, journal_dev); 6114 if (IS_ERR(journal)) 6115 return PTR_ERR(journal); 6116 } 6117 6118 journal_dev_ro = bdev_read_only(journal->j_dev); 6119 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro; 6120 6121 if (journal_dev_ro && !sb_rdonly(sb)) { 6122 ext4_msg(sb, KERN_ERR, 6123 "journal device read-only, try mounting with '-o ro'"); 6124 err = -EROFS; 6125 goto err_out; 6126 } 6127 6128 /* 6129 * Are we loading a blank journal or performing recovery after a 6130 * crash? For recovery, we need to check in advance whether we 6131 * can get read-write access to the device. 6132 */ 6133 if (ext4_has_feature_journal_needs_recovery(sb)) { 6134 if (sb_rdonly(sb)) { 6135 ext4_msg(sb, KERN_INFO, "INFO: recovery " 6136 "required on readonly filesystem"); 6137 if (really_read_only) { 6138 ext4_msg(sb, KERN_ERR, "write access " 6139 "unavailable, cannot proceed " 6140 "(try mounting with noload)"); 6141 err = -EROFS; 6142 goto err_out; 6143 } 6144 ext4_msg(sb, KERN_INFO, "write access will " 6145 "be enabled during recovery"); 6146 } 6147 } 6148 6149 if (!(journal->j_flags & JBD2_BARRIER)) 6150 ext4_msg(sb, KERN_INFO, "barriers disabled"); 6151 6152 if (!ext4_has_feature_journal_needs_recovery(sb)) 6153 err = jbd2_journal_wipe(journal, !really_read_only); 6154 if (!err) { 6155 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 6156 __le16 orig_state; 6157 bool changed = false; 6158 6159 if (save) 6160 memcpy(save, ((char *) es) + 6161 EXT4_S_ERR_START, EXT4_S_ERR_LEN); 6162 err = jbd2_journal_load(journal); 6163 if (save && memcmp(((char *) es) + EXT4_S_ERR_START, 6164 save, EXT4_S_ERR_LEN)) { 6165 memcpy(((char *) es) + EXT4_S_ERR_START, 6166 save, EXT4_S_ERR_LEN); 6167 changed = true; 6168 } 6169 kfree(save); 6170 orig_state = es->s_state; 6171 es->s_state |= cpu_to_le16(EXT4_SB(sb)->s_mount_state & 6172 EXT4_ERROR_FS); 6173 if (orig_state != es->s_state) 6174 changed = true; 6175 /* Write out restored error information to the superblock */ 6176 if (changed && !really_read_only) { 6177 int err2; 6178 err2 = ext4_commit_super(sb); 6179 err = err ? : err2; 6180 } 6181 } 6182 6183 if (err) { 6184 ext4_msg(sb, KERN_ERR, "error loading journal"); 6185 goto err_out; 6186 } 6187 6188 EXT4_SB(sb)->s_journal = journal; 6189 err = ext4_clear_journal_err(sb, es); 6190 if (err) { 6191 ext4_journal_destroy(EXT4_SB(sb), journal); 6192 return err; 6193 } 6194 6195 if (!really_read_only && journal_devnum && 6196 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 6197 es->s_journal_dev = cpu_to_le32(journal_devnum); 6198 ext4_commit_super(sb); 6199 } 6200 if (!really_read_only && journal_inum && 6201 journal_inum != le32_to_cpu(es->s_journal_inum)) { 6202 es->s_journal_inum = cpu_to_le32(journal_inum); 6203 ext4_commit_super(sb); 6204 } 6205 6206 return 0; 6207 6208 err_out: 6209 ext4_journal_destroy(EXT4_SB(sb), journal); 6210 return err; 6211 } 6212 6213 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */ 6214 static void ext4_update_super(struct super_block *sb) 6215 { 6216 struct ext4_sb_info *sbi = EXT4_SB(sb); 6217 struct ext4_super_block *es = sbi->s_es; 6218 struct buffer_head *sbh = sbi->s_sbh; 6219 6220 lock_buffer(sbh); 6221 /* 6222 * If the file system is mounted read-only, don't update the 6223 * superblock write time. This avoids updating the superblock 6224 * write time when we are mounting the root file system 6225 * read/only but we need to replay the journal; at that point, 6226 * for people who are east of GMT and who make their clock 6227 * tick in localtime for Windows bug-for-bug compatibility, 6228 * the clock is set in the future, and this will cause e2fsck 6229 * to complain and force a full file system check. 6230 */ 6231 if (!sb_rdonly(sb)) 6232 ext4_update_tstamp(es, s_wtime); 6233 es->s_kbytes_written = 6234 cpu_to_le64(sbi->s_kbytes_written + 6235 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) - 6236 sbi->s_sectors_written_start) >> 1)); 6237 if (percpu_counter_initialized(&sbi->s_freeclusters_counter)) 6238 ext4_free_blocks_count_set(es, 6239 EXT4_C2B(sbi, percpu_counter_sum_positive( 6240 &sbi->s_freeclusters_counter))); 6241 if (percpu_counter_initialized(&sbi->s_freeinodes_counter)) 6242 es->s_free_inodes_count = 6243 cpu_to_le32(percpu_counter_sum_positive( 6244 &sbi->s_freeinodes_counter)); 6245 /* Copy error information to the on-disk superblock */ 6246 spin_lock(&sbi->s_error_lock); 6247 if (sbi->s_add_error_count > 0) { 6248 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 6249 if (!es->s_first_error_time && !es->s_first_error_time_hi) { 6250 __ext4_update_tstamp(&es->s_first_error_time, 6251 &es->s_first_error_time_hi, 6252 sbi->s_first_error_time); 6253 strtomem_pad(es->s_first_error_func, 6254 sbi->s_first_error_func, 0); 6255 es->s_first_error_line = 6256 cpu_to_le32(sbi->s_first_error_line); 6257 es->s_first_error_ino = 6258 cpu_to_le32(sbi->s_first_error_ino); 6259 es->s_first_error_block = 6260 cpu_to_le64(sbi->s_first_error_block); 6261 es->s_first_error_errcode = 6262 ext4_errno_to_code(sbi->s_first_error_code); 6263 } 6264 __ext4_update_tstamp(&es->s_last_error_time, 6265 &es->s_last_error_time_hi, 6266 sbi->s_last_error_time); 6267 strtomem_pad(es->s_last_error_func, sbi->s_last_error_func, 0); 6268 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line); 6269 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino); 6270 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block); 6271 es->s_last_error_errcode = 6272 ext4_errno_to_code(sbi->s_last_error_code); 6273 /* 6274 * Start the daily error reporting function if it hasn't been 6275 * started already and sbi->s_err_report_sec is not zero 6276 */ 6277 if (!es->s_error_count && !sbi->s_err_report_sec) 6278 mod_timer(&sbi->s_err_report, 6279 jiffies + secs_to_jiffies(sbi->s_err_report_sec)); 6280 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count); 6281 sbi->s_add_error_count = 0; 6282 } 6283 spin_unlock(&sbi->s_error_lock); 6284 6285 ext4_superblock_csum_set(sb); 6286 unlock_buffer(sbh); 6287 } 6288 6289 static int ext4_commit_super(struct super_block *sb) 6290 { 6291 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 6292 6293 if (!sbh) 6294 return -EINVAL; 6295 6296 ext4_update_super(sb); 6297 6298 lock_buffer(sbh); 6299 /* Buffer got discarded which means block device got invalidated */ 6300 if (!buffer_mapped(sbh)) { 6301 unlock_buffer(sbh); 6302 return -EIO; 6303 } 6304 6305 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 6306 /* 6307 * Oh, dear. A previous attempt to write the 6308 * superblock failed. This could happen because the 6309 * USB device was yanked out. Or it could happen to 6310 * be a transient write error and maybe the block will 6311 * be remapped. Nothing we can do but to retry the 6312 * write and hope for the best. 6313 */ 6314 ext4_msg(sb, KERN_ERR, "previous I/O error to " 6315 "superblock detected"); 6316 clear_buffer_write_io_error(sbh); 6317 set_buffer_uptodate(sbh); 6318 } 6319 get_bh(sbh); 6320 /* Clear potential dirty bit if it was journalled update */ 6321 clear_buffer_dirty(sbh); 6322 sbh->b_end_io = end_buffer_write_sync; 6323 submit_bh(REQ_OP_WRITE | REQ_SYNC | 6324 (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh); 6325 wait_on_buffer(sbh); 6326 if (buffer_write_io_error(sbh)) { 6327 ext4_msg(sb, KERN_ERR, "I/O error while writing " 6328 "superblock"); 6329 clear_buffer_write_io_error(sbh); 6330 set_buffer_uptodate(sbh); 6331 return -EIO; 6332 } 6333 return 0; 6334 } 6335 6336 /* 6337 * Have we just finished recovery? If so, and if we are mounting (or 6338 * remounting) the filesystem readonly, then we will end up with a 6339 * consistent fs on disk. Record that fact. 6340 */ 6341 static int ext4_mark_recovery_complete(struct super_block *sb, 6342 struct ext4_super_block *es) 6343 { 6344 int err; 6345 journal_t *journal = EXT4_SB(sb)->s_journal; 6346 6347 if (!ext4_has_feature_journal(sb)) { 6348 if (journal != NULL) { 6349 ext4_error(sb, "Journal got removed while the fs was " 6350 "mounted!"); 6351 return -EFSCORRUPTED; 6352 } 6353 return 0; 6354 } 6355 jbd2_journal_lock_updates(journal); 6356 err = jbd2_journal_flush(journal, 0); 6357 if (err < 0) 6358 goto out; 6359 6360 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) || 6361 ext4_has_feature_orphan_present(sb))) { 6362 if (!ext4_orphan_file_empty(sb)) { 6363 ext4_error(sb, "Orphan file not empty on read-only fs."); 6364 err = -EFSCORRUPTED; 6365 goto out; 6366 } 6367 ext4_clear_feature_journal_needs_recovery(sb); 6368 ext4_clear_feature_orphan_present(sb); 6369 ext4_commit_super(sb); 6370 } 6371 out: 6372 jbd2_journal_unlock_updates(journal); 6373 return err; 6374 } 6375 6376 /* 6377 * If we are mounting (or read-write remounting) a filesystem whose journal 6378 * has recorded an error from a previous lifetime, move that error to the 6379 * main filesystem now. 6380 */ 6381 static int ext4_clear_journal_err(struct super_block *sb, 6382 struct ext4_super_block *es) 6383 { 6384 journal_t *journal; 6385 int j_errno; 6386 const char *errstr; 6387 6388 if (!ext4_has_feature_journal(sb)) { 6389 ext4_error(sb, "Journal got removed while the fs was mounted!"); 6390 return -EFSCORRUPTED; 6391 } 6392 6393 journal = EXT4_SB(sb)->s_journal; 6394 6395 /* 6396 * Now check for any error status which may have been recorded in the 6397 * journal by a prior ext4_error() or ext4_abort() 6398 */ 6399 6400 j_errno = jbd2_journal_errno(journal); 6401 if (j_errno) { 6402 char nbuf[16]; 6403 6404 errstr = ext4_decode_error(sb, j_errno, nbuf); 6405 ext4_warning(sb, "Filesystem error recorded " 6406 "from previous mount: %s", errstr); 6407 6408 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 6409 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 6410 j_errno = ext4_commit_super(sb); 6411 if (j_errno) 6412 return j_errno; 6413 ext4_warning(sb, "Marked fs in need of filesystem check."); 6414 6415 jbd2_journal_clear_err(journal); 6416 jbd2_journal_update_sb_errno(journal); 6417 } 6418 return 0; 6419 } 6420 6421 /* 6422 * Force the running and committing transactions to commit, 6423 * and wait on the commit. 6424 */ 6425 int ext4_force_commit(struct super_block *sb) 6426 { 6427 return ext4_journal_force_commit(EXT4_SB(sb)->s_journal); 6428 } 6429 6430 static int ext4_sync_fs(struct super_block *sb, int wait) 6431 { 6432 int ret = 0; 6433 tid_t target; 6434 bool needs_barrier = false; 6435 struct ext4_sb_info *sbi = EXT4_SB(sb); 6436 6437 ret = ext4_emergency_state(sb); 6438 if (unlikely(ret)) 6439 return ret; 6440 6441 trace_ext4_sync_fs(sb, wait); 6442 flush_workqueue(sbi->rsv_conversion_wq); 6443 /* 6444 * Writeback quota in non-journalled quota case - journalled quota has 6445 * no dirty dquots 6446 */ 6447 dquot_writeback_dquots(sb, -1); 6448 /* 6449 * Data writeback is possible w/o journal transaction, so barrier must 6450 * being sent at the end of the function. But we can skip it if 6451 * transaction_commit will do it for us. 6452 */ 6453 if (sbi->s_journal) { 6454 target = jbd2_get_latest_transaction(sbi->s_journal); 6455 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 6456 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 6457 needs_barrier = true; 6458 6459 if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 6460 if (wait) 6461 ret = jbd2_log_wait_commit(sbi->s_journal, 6462 target); 6463 } 6464 } else if (wait && test_opt(sb, BARRIER)) 6465 needs_barrier = true; 6466 if (needs_barrier) { 6467 int err; 6468 err = blkdev_issue_flush(sb->s_bdev); 6469 if (!ret) 6470 ret = err; 6471 } 6472 6473 return ret; 6474 } 6475 6476 /* 6477 * LVM calls this function before a (read-only) snapshot is created. This 6478 * gives us a chance to flush the journal completely and mark the fs clean. 6479 * 6480 * Note that only this function cannot bring a filesystem to be in a clean 6481 * state independently. It relies on upper layer to stop all data & metadata 6482 * modifications. 6483 */ 6484 static int ext4_freeze(struct super_block *sb) 6485 { 6486 int error = 0; 6487 journal_t *journal = EXT4_SB(sb)->s_journal; 6488 6489 if (journal) { 6490 /* Now we set up the journal barrier. */ 6491 jbd2_journal_lock_updates(journal); 6492 6493 /* 6494 * Don't clear the needs_recovery flag if we failed to 6495 * flush the journal. 6496 */ 6497 error = jbd2_journal_flush(journal, 0); 6498 if (error < 0) 6499 goto out; 6500 6501 /* Journal blocked and flushed, clear needs_recovery flag. */ 6502 ext4_clear_feature_journal_needs_recovery(sb); 6503 if (ext4_orphan_file_empty(sb)) 6504 ext4_clear_feature_orphan_present(sb); 6505 } 6506 6507 error = ext4_commit_super(sb); 6508 out: 6509 if (journal) 6510 /* we rely on upper layer to stop further updates */ 6511 jbd2_journal_unlock_updates(journal); 6512 return error; 6513 } 6514 6515 /* 6516 * Called by LVM after the snapshot is done. We need to reset the RECOVER 6517 * flag here, even though the filesystem is not technically dirty yet. 6518 */ 6519 static int ext4_unfreeze(struct super_block *sb) 6520 { 6521 if (ext4_emergency_state(sb)) 6522 return 0; 6523 6524 if (EXT4_SB(sb)->s_journal) { 6525 /* Reset the needs_recovery flag before the fs is unlocked. */ 6526 ext4_set_feature_journal_needs_recovery(sb); 6527 if (ext4_has_feature_orphan_file(sb)) 6528 ext4_set_feature_orphan_present(sb); 6529 } 6530 6531 ext4_commit_super(sb); 6532 return 0; 6533 } 6534 6535 /* 6536 * Structure to save mount options for ext4_remount's benefit 6537 */ 6538 struct ext4_mount_options { 6539 unsigned long s_mount_opt; 6540 unsigned long s_mount_opt2; 6541 kuid_t s_resuid; 6542 kgid_t s_resgid; 6543 unsigned long s_commit_interval; 6544 u32 s_min_batch_time, s_max_batch_time; 6545 #ifdef CONFIG_QUOTA 6546 int s_jquota_fmt; 6547 char *s_qf_names[EXT4_MAXQUOTAS]; 6548 #endif 6549 }; 6550 6551 static int __ext4_remount(struct fs_context *fc, struct super_block *sb) 6552 { 6553 struct ext4_fs_context *ctx = fc->fs_private; 6554 struct ext4_super_block *es; 6555 struct ext4_sb_info *sbi = EXT4_SB(sb); 6556 unsigned long old_sb_flags; 6557 struct ext4_mount_options old_opts; 6558 ext4_group_t g; 6559 int err = 0; 6560 int alloc_ctx; 6561 #ifdef CONFIG_QUOTA 6562 int enable_quota = 0; 6563 int i, j; 6564 char *to_free[EXT4_MAXQUOTAS]; 6565 #endif 6566 6567 6568 /* Store the original options */ 6569 old_sb_flags = sb->s_flags; 6570 old_opts.s_mount_opt = sbi->s_mount_opt; 6571 old_opts.s_mount_opt2 = sbi->s_mount_opt2; 6572 old_opts.s_resuid = sbi->s_resuid; 6573 old_opts.s_resgid = sbi->s_resgid; 6574 old_opts.s_commit_interval = sbi->s_commit_interval; 6575 old_opts.s_min_batch_time = sbi->s_min_batch_time; 6576 old_opts.s_max_batch_time = sbi->s_max_batch_time; 6577 #ifdef CONFIG_QUOTA 6578 old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 6579 for (i = 0; i < EXT4_MAXQUOTAS; i++) 6580 if (sbi->s_qf_names[i]) { 6581 char *qf_name = get_qf_name(sb, sbi, i); 6582 6583 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); 6584 if (!old_opts.s_qf_names[i]) { 6585 for (j = 0; j < i; j++) 6586 kfree(old_opts.s_qf_names[j]); 6587 return -ENOMEM; 6588 } 6589 } else 6590 old_opts.s_qf_names[i] = NULL; 6591 #endif 6592 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) { 6593 if (sbi->s_journal && sbi->s_journal->j_task->io_context) 6594 ctx->journal_ioprio = 6595 sbi->s_journal->j_task->io_context->ioprio; 6596 else 6597 ctx->journal_ioprio = EXT4_DEF_JOURNAL_IOPRIO; 6598 6599 } 6600 6601 if ((ctx->spec & EXT4_SPEC_s_stripe) && 6602 ext4_is_stripe_incompatible(sb, ctx->s_stripe)) { 6603 ext4_msg(sb, KERN_WARNING, 6604 "stripe (%lu) is not aligned with cluster size (%u), " 6605 "stripe is disabled", 6606 ctx->s_stripe, sbi->s_cluster_ratio); 6607 ctx->s_stripe = 0; 6608 } 6609 6610 /* 6611 * Changing the DIOREAD_NOLOCK or DELALLOC mount options may cause 6612 * two calls to ext4_should_dioread_nolock() to return inconsistent 6613 * values, triggering WARN_ON in ext4_add_complete_io(). we grab 6614 * here s_writepages_rwsem to avoid race between writepages ops and 6615 * remount. 6616 */ 6617 alloc_ctx = ext4_writepages_down_write(sb); 6618 ext4_apply_options(fc, sb); 6619 ext4_writepages_up_write(sb, alloc_ctx); 6620 6621 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 6622 test_opt(sb, JOURNAL_CHECKSUM)) { 6623 ext4_msg(sb, KERN_ERR, "changing journal_checksum " 6624 "during remount not supported; ignoring"); 6625 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 6626 } 6627 6628 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 6629 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 6630 ext4_msg(sb, KERN_ERR, "can't mount with " 6631 "both data=journal and delalloc"); 6632 err = -EINVAL; 6633 goto restore_opts; 6634 } 6635 if (test_opt(sb, DIOREAD_NOLOCK)) { 6636 ext4_msg(sb, KERN_ERR, "can't mount with " 6637 "both data=journal and dioread_nolock"); 6638 err = -EINVAL; 6639 goto restore_opts; 6640 } 6641 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { 6642 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 6643 ext4_msg(sb, KERN_ERR, "can't mount with " 6644 "journal_async_commit in data=ordered mode"); 6645 err = -EINVAL; 6646 goto restore_opts; 6647 } 6648 } 6649 6650 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { 6651 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); 6652 err = -EINVAL; 6653 goto restore_opts; 6654 } 6655 6656 if ((old_opts.s_mount_opt & EXT4_MOUNT_DELALLOC) && 6657 !test_opt(sb, DELALLOC)) { 6658 ext4_msg(sb, KERN_ERR, "can't disable delalloc during remount"); 6659 err = -EINVAL; 6660 goto restore_opts; 6661 } 6662 6663 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 6664 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 6665 6666 es = sbi->s_es; 6667 6668 if (sbi->s_journal) { 6669 ext4_init_journal_params(sb, sbi->s_journal); 6670 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio); 6671 } 6672 6673 /* Flush outstanding errors before changing fs state */ 6674 flush_work(&sbi->s_sb_upd_work); 6675 6676 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) { 6677 if (ext4_emergency_state(sb)) { 6678 err = -EROFS; 6679 goto restore_opts; 6680 } 6681 6682 if (fc->sb_flags & SB_RDONLY) { 6683 err = sync_filesystem(sb); 6684 if (err < 0) 6685 goto restore_opts; 6686 err = dquot_suspend(sb, -1); 6687 if (err < 0) 6688 goto restore_opts; 6689 6690 /* 6691 * First of all, the unconditional stuff we have to do 6692 * to disable replay of the journal when we next remount 6693 */ 6694 sb->s_flags |= SB_RDONLY; 6695 6696 /* 6697 * OK, test if we are remounting a valid rw partition 6698 * readonly, and if so set the rdonly flag and then 6699 * mark the partition as valid again. 6700 */ 6701 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 6702 (sbi->s_mount_state & EXT4_VALID_FS)) 6703 es->s_state = cpu_to_le16(sbi->s_mount_state); 6704 6705 if (sbi->s_journal) { 6706 /* 6707 * We let remount-ro finish even if marking fs 6708 * as clean failed... 6709 */ 6710 ext4_mark_recovery_complete(sb, es); 6711 } 6712 } else { 6713 /* Make sure we can mount this feature set readwrite */ 6714 if (ext4_has_feature_readonly(sb) || 6715 !ext4_feature_set_ok(sb, 0)) { 6716 err = -EROFS; 6717 goto restore_opts; 6718 } 6719 /* 6720 * Make sure the group descriptor checksums 6721 * are sane. If they aren't, refuse to remount r/w. 6722 */ 6723 for (g = 0; g < sbi->s_groups_count; g++) { 6724 struct ext4_group_desc *gdp = 6725 ext4_get_group_desc(sb, g, NULL); 6726 6727 if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 6728 ext4_msg(sb, KERN_ERR, 6729 "ext4_remount: Checksum for group %u failed (%u!=%u)", 6730 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 6731 le16_to_cpu(gdp->bg_checksum)); 6732 err = -EFSBADCRC; 6733 goto restore_opts; 6734 } 6735 } 6736 6737 /* 6738 * If we have an unprocessed orphan list hanging 6739 * around from a previously readonly bdev mount, 6740 * require a full umount/remount for now. 6741 */ 6742 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) { 6743 ext4_msg(sb, KERN_WARNING, "Couldn't " 6744 "remount RDWR because of unprocessed " 6745 "orphan inode list. Please " 6746 "umount/remount instead"); 6747 err = -EINVAL; 6748 goto restore_opts; 6749 } 6750 6751 /* 6752 * Mounting a RDONLY partition read-write, so reread 6753 * and store the current valid flag. (It may have 6754 * been changed by e2fsck since we originally mounted 6755 * the partition.) 6756 */ 6757 if (sbi->s_journal) { 6758 err = ext4_clear_journal_err(sb, es); 6759 if (err) 6760 goto restore_opts; 6761 } 6762 sbi->s_mount_state = (le16_to_cpu(es->s_state) & 6763 ~EXT4_FC_REPLAY); 6764 6765 err = ext4_setup_super(sb, es, 0); 6766 if (err) 6767 goto restore_opts; 6768 6769 sb->s_flags &= ~SB_RDONLY; 6770 if (ext4_has_feature_mmp(sb)) { 6771 err = ext4_multi_mount_protect(sb, 6772 le64_to_cpu(es->s_mmp_block)); 6773 if (err) 6774 goto restore_opts; 6775 } 6776 #ifdef CONFIG_QUOTA 6777 enable_quota = 1; 6778 #endif 6779 } 6780 } 6781 6782 /* 6783 * Handle creation of system zone data early because it can fail. 6784 * Releasing of existing data is done when we are sure remount will 6785 * succeed. 6786 */ 6787 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) { 6788 err = ext4_setup_system_zone(sb); 6789 if (err) 6790 goto restore_opts; 6791 } 6792 6793 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) { 6794 err = ext4_commit_super(sb); 6795 if (err) 6796 goto restore_opts; 6797 } 6798 6799 #ifdef CONFIG_QUOTA 6800 if (enable_quota) { 6801 if (sb_any_quota_suspended(sb)) 6802 dquot_resume(sb, -1); 6803 else if (ext4_has_feature_quota(sb)) { 6804 err = ext4_enable_quotas(sb); 6805 if (err) 6806 goto restore_opts; 6807 } 6808 } 6809 /* Release old quota file names */ 6810 for (i = 0; i < EXT4_MAXQUOTAS; i++) 6811 kfree(old_opts.s_qf_names[i]); 6812 #endif 6813 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 6814 ext4_release_system_zone(sb); 6815 6816 /* 6817 * Reinitialize lazy itable initialization thread based on 6818 * current settings 6819 */ 6820 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) 6821 ext4_unregister_li_request(sb); 6822 else { 6823 ext4_group_t first_not_zeroed; 6824 first_not_zeroed = ext4_has_uninit_itable(sb); 6825 ext4_register_li_request(sb, first_not_zeroed); 6826 } 6827 6828 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb)) 6829 ext4_stop_mmpd(sbi); 6830 6831 /* 6832 * Handle aborting the filesystem as the last thing during remount to 6833 * avoid obsure errors during remount when some option changes fail to 6834 * apply due to shutdown filesystem. 6835 */ 6836 if (test_opt2(sb, ABORT)) 6837 ext4_abort(sb, ESHUTDOWN, "Abort forced by user"); 6838 6839 return 0; 6840 6841 restore_opts: 6842 /* 6843 * If there was a failing r/w to ro transition, we may need to 6844 * re-enable quota 6845 */ 6846 if (sb_rdonly(sb) && !(old_sb_flags & SB_RDONLY) && 6847 sb_any_quota_suspended(sb)) 6848 dquot_resume(sb, -1); 6849 6850 alloc_ctx = ext4_writepages_down_write(sb); 6851 sb->s_flags = old_sb_flags; 6852 sbi->s_mount_opt = old_opts.s_mount_opt; 6853 sbi->s_mount_opt2 = old_opts.s_mount_opt2; 6854 sbi->s_resuid = old_opts.s_resuid; 6855 sbi->s_resgid = old_opts.s_resgid; 6856 sbi->s_commit_interval = old_opts.s_commit_interval; 6857 sbi->s_min_batch_time = old_opts.s_min_batch_time; 6858 sbi->s_max_batch_time = old_opts.s_max_batch_time; 6859 ext4_writepages_up_write(sb, alloc_ctx); 6860 6861 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 6862 ext4_release_system_zone(sb); 6863 #ifdef CONFIG_QUOTA 6864 sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 6865 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 6866 to_free[i] = get_qf_name(sb, sbi, i); 6867 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); 6868 } 6869 synchronize_rcu(); 6870 for (i = 0; i < EXT4_MAXQUOTAS; i++) 6871 kfree(to_free[i]); 6872 #endif 6873 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb)) 6874 ext4_stop_mmpd(sbi); 6875 return err; 6876 } 6877 6878 static int ext4_reconfigure(struct fs_context *fc) 6879 { 6880 struct super_block *sb = fc->root->d_sb; 6881 int ret; 6882 bool old_ro = sb_rdonly(sb); 6883 6884 fc->s_fs_info = EXT4_SB(sb); 6885 6886 ret = ext4_check_opt_consistency(fc, sb); 6887 if (ret < 0) 6888 return ret; 6889 6890 ret = __ext4_remount(fc, sb); 6891 if (ret < 0) 6892 return ret; 6893 6894 ext4_msg(sb, KERN_INFO, "re-mounted %pU%s.", 6895 &sb->s_uuid, 6896 (old_ro != sb_rdonly(sb)) ? (sb_rdonly(sb) ? " ro" : " r/w") : ""); 6897 6898 return 0; 6899 } 6900 6901 #ifdef CONFIG_QUOTA 6902 static int ext4_statfs_project(struct super_block *sb, 6903 kprojid_t projid, struct kstatfs *buf) 6904 { 6905 struct kqid qid; 6906 struct dquot *dquot; 6907 u64 limit; 6908 u64 curblock; 6909 6910 qid = make_kqid_projid(projid); 6911 dquot = dqget(sb, qid); 6912 if (IS_ERR(dquot)) 6913 return PTR_ERR(dquot); 6914 spin_lock(&dquot->dq_dqb_lock); 6915 6916 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 6917 dquot->dq_dqb.dqb_bhardlimit); 6918 limit >>= sb->s_blocksize_bits; 6919 6920 if (limit) { 6921 uint64_t remaining = 0; 6922 6923 curblock = (dquot->dq_dqb.dqb_curspace + 6924 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 6925 if (limit > curblock) 6926 remaining = limit - curblock; 6927 6928 buf->f_blocks = min(buf->f_blocks, limit); 6929 buf->f_bfree = min(buf->f_bfree, remaining); 6930 buf->f_bavail = min(buf->f_bavail, remaining); 6931 } 6932 6933 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 6934 dquot->dq_dqb.dqb_ihardlimit); 6935 if (limit) { 6936 uint64_t remaining = 0; 6937 6938 if (limit > dquot->dq_dqb.dqb_curinodes) 6939 remaining = limit - dquot->dq_dqb.dqb_curinodes; 6940 6941 buf->f_files = min(buf->f_files, limit); 6942 buf->f_ffree = min(buf->f_ffree, remaining); 6943 } 6944 6945 spin_unlock(&dquot->dq_dqb_lock); 6946 dqput(dquot); 6947 return 0; 6948 } 6949 #endif 6950 6951 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 6952 { 6953 struct super_block *sb = dentry->d_sb; 6954 struct ext4_sb_info *sbi = EXT4_SB(sb); 6955 struct ext4_super_block *es = sbi->s_es; 6956 ext4_fsblk_t overhead = 0, resv_blocks; 6957 s64 bfree; 6958 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 6959 6960 if (!test_opt(sb, MINIX_DF)) 6961 overhead = sbi->s_overhead; 6962 6963 buf->f_type = EXT4_SUPER_MAGIC; 6964 buf->f_bsize = sb->s_blocksize; 6965 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 6966 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 6967 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 6968 /* prevent underflow in case that few free space is available */ 6969 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 6970 buf->f_bavail = buf->f_bfree - 6971 (ext4_r_blocks_count(es) + resv_blocks); 6972 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 6973 buf->f_bavail = 0; 6974 buf->f_files = le32_to_cpu(es->s_inodes_count); 6975 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 6976 buf->f_namelen = EXT4_NAME_LEN; 6977 buf->f_fsid = uuid_to_fsid(es->s_uuid); 6978 6979 #ifdef CONFIG_QUOTA 6980 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 6981 sb_has_quota_limits_enabled(sb, PRJQUOTA)) 6982 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 6983 #endif 6984 return 0; 6985 } 6986 6987 6988 #ifdef CONFIG_QUOTA 6989 6990 /* 6991 * Helper functions so that transaction is started before we acquire dqio_sem 6992 * to keep correct lock ordering of transaction > dqio_sem 6993 */ 6994 static inline struct inode *dquot_to_inode(struct dquot *dquot) 6995 { 6996 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 6997 } 6998 6999 static int ext4_write_dquot(struct dquot *dquot) 7000 { 7001 int ret, err; 7002 handle_t *handle; 7003 struct inode *inode; 7004 7005 inode = dquot_to_inode(dquot); 7006 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 7007 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 7008 if (IS_ERR(handle)) 7009 return PTR_ERR(handle); 7010 ret = dquot_commit(dquot); 7011 if (ret < 0) 7012 ext4_error_err(dquot->dq_sb, -ret, 7013 "Failed to commit dquot type %d", 7014 dquot->dq_id.type); 7015 err = ext4_journal_stop(handle); 7016 if (!ret) 7017 ret = err; 7018 return ret; 7019 } 7020 7021 static int ext4_acquire_dquot(struct dquot *dquot) 7022 { 7023 int ret, err; 7024 handle_t *handle; 7025 7026 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 7027 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 7028 if (IS_ERR(handle)) 7029 return PTR_ERR(handle); 7030 ret = dquot_acquire(dquot); 7031 if (ret < 0) 7032 ext4_error_err(dquot->dq_sb, -ret, 7033 "Failed to acquire dquot type %d", 7034 dquot->dq_id.type); 7035 err = ext4_journal_stop(handle); 7036 if (!ret) 7037 ret = err; 7038 return ret; 7039 } 7040 7041 static int ext4_release_dquot(struct dquot *dquot) 7042 { 7043 int ret, err; 7044 handle_t *handle; 7045 bool freeze_protected = false; 7046 7047 /* 7048 * Trying to sb_start_intwrite() in a running transaction 7049 * can result in a deadlock. Further, running transactions 7050 * are already protected from freezing. 7051 */ 7052 if (!ext4_journal_current_handle()) { 7053 sb_start_intwrite(dquot->dq_sb); 7054 freeze_protected = true; 7055 } 7056 7057 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 7058 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 7059 if (IS_ERR(handle)) { 7060 /* Release dquot anyway to avoid endless cycle in dqput() */ 7061 dquot_release(dquot); 7062 if (freeze_protected) 7063 sb_end_intwrite(dquot->dq_sb); 7064 return PTR_ERR(handle); 7065 } 7066 ret = dquot_release(dquot); 7067 if (ret < 0) 7068 ext4_error_err(dquot->dq_sb, -ret, 7069 "Failed to release dquot type %d", 7070 dquot->dq_id.type); 7071 err = ext4_journal_stop(handle); 7072 if (!ret) 7073 ret = err; 7074 7075 if (freeze_protected) 7076 sb_end_intwrite(dquot->dq_sb); 7077 7078 return ret; 7079 } 7080 7081 static int ext4_mark_dquot_dirty(struct dquot *dquot) 7082 { 7083 struct super_block *sb = dquot->dq_sb; 7084 7085 if (ext4_is_quota_journalled(sb)) { 7086 dquot_mark_dquot_dirty(dquot); 7087 return ext4_write_dquot(dquot); 7088 } else { 7089 return dquot_mark_dquot_dirty(dquot); 7090 } 7091 } 7092 7093 static int ext4_write_info(struct super_block *sb, int type) 7094 { 7095 int ret, err; 7096 handle_t *handle; 7097 7098 /* Data block + inode block */ 7099 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2); 7100 if (IS_ERR(handle)) 7101 return PTR_ERR(handle); 7102 ret = dquot_commit_info(sb, type); 7103 err = ext4_journal_stop(handle); 7104 if (!ret) 7105 ret = err; 7106 return ret; 7107 } 7108 7109 static void lockdep_set_quota_inode(struct inode *inode, int subclass) 7110 { 7111 struct ext4_inode_info *ei = EXT4_I(inode); 7112 7113 /* The first argument of lockdep_set_subclass has to be 7114 * *exactly* the same as the argument to init_rwsem() --- in 7115 * this case, in init_once() --- or lockdep gets unhappy 7116 * because the name of the lock is set using the 7117 * stringification of the argument to init_rwsem(). 7118 */ 7119 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 7120 lockdep_set_subclass(&ei->i_data_sem, subclass); 7121 } 7122 7123 /* 7124 * Standard function to be called on quota_on 7125 */ 7126 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 7127 const struct path *path) 7128 { 7129 int err; 7130 7131 if (!test_opt(sb, QUOTA)) 7132 return -EINVAL; 7133 7134 /* Quotafile not on the same filesystem? */ 7135 if (path->dentry->d_sb != sb) 7136 return -EXDEV; 7137 7138 /* Quota already enabled for this file? */ 7139 if (IS_NOQUOTA(d_inode(path->dentry))) 7140 return -EBUSY; 7141 7142 /* Journaling quota? */ 7143 if (EXT4_SB(sb)->s_qf_names[type]) { 7144 /* Quotafile not in fs root? */ 7145 if (path->dentry->d_parent != sb->s_root) 7146 ext4_msg(sb, KERN_WARNING, 7147 "Quota file not on filesystem root. " 7148 "Journaled quota will not work"); 7149 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY; 7150 } else { 7151 /* 7152 * Clear the flag just in case mount options changed since 7153 * last time. 7154 */ 7155 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY; 7156 } 7157 7158 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 7159 err = dquot_quota_on(sb, type, format_id, path); 7160 if (!err) { 7161 struct inode *inode = d_inode(path->dentry); 7162 handle_t *handle; 7163 7164 /* 7165 * Set inode flags to prevent userspace from messing with quota 7166 * files. If this fails, we return success anyway since quotas 7167 * are already enabled and this is not a hard failure. 7168 */ 7169 inode_lock(inode); 7170 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 7171 if (IS_ERR(handle)) 7172 goto unlock_inode; 7173 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; 7174 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 7175 S_NOATIME | S_IMMUTABLE); 7176 err = ext4_mark_inode_dirty(handle, inode); 7177 ext4_journal_stop(handle); 7178 unlock_inode: 7179 inode_unlock(inode); 7180 if (err) 7181 dquot_quota_off(sb, type); 7182 } 7183 if (err) 7184 lockdep_set_quota_inode(path->dentry->d_inode, 7185 I_DATA_SEM_NORMAL); 7186 return err; 7187 } 7188 7189 static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum) 7190 { 7191 switch (type) { 7192 case USRQUOTA: 7193 return qf_inum == EXT4_USR_QUOTA_INO; 7194 case GRPQUOTA: 7195 return qf_inum == EXT4_GRP_QUOTA_INO; 7196 case PRJQUOTA: 7197 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO; 7198 default: 7199 BUG(); 7200 } 7201 } 7202 7203 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 7204 unsigned int flags) 7205 { 7206 int err; 7207 struct inode *qf_inode; 7208 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 7209 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 7210 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 7211 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 7212 }; 7213 7214 BUG_ON(!ext4_has_feature_quota(sb)); 7215 7216 if (!qf_inums[type]) 7217 return -EPERM; 7218 7219 if (!ext4_check_quota_inum(type, qf_inums[type])) { 7220 ext4_error(sb, "Bad quota inum: %lu, type: %d", 7221 qf_inums[type], type); 7222 return -EUCLEAN; 7223 } 7224 7225 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL); 7226 if (IS_ERR(qf_inode)) { 7227 ext4_error(sb, "Bad quota inode: %lu, type: %d", 7228 qf_inums[type], type); 7229 return PTR_ERR(qf_inode); 7230 } 7231 7232 /* Don't account quota for quota files to avoid recursion */ 7233 qf_inode->i_flags |= S_NOQUOTA; 7234 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 7235 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 7236 if (err) 7237 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 7238 iput(qf_inode); 7239 7240 return err; 7241 } 7242 7243 /* Enable usage tracking for all quota types. */ 7244 int ext4_enable_quotas(struct super_block *sb) 7245 { 7246 int type, err = 0; 7247 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 7248 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 7249 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 7250 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 7251 }; 7252 bool quota_mopt[EXT4_MAXQUOTAS] = { 7253 test_opt(sb, USRQUOTA), 7254 test_opt(sb, GRPQUOTA), 7255 test_opt(sb, PRJQUOTA), 7256 }; 7257 7258 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 7259 for (type = 0; type < EXT4_MAXQUOTAS; type++) { 7260 if (qf_inums[type]) { 7261 err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 7262 DQUOT_USAGE_ENABLED | 7263 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 7264 if (err) { 7265 ext4_warning(sb, 7266 "Failed to enable quota tracking " 7267 "(type=%d, err=%d, ino=%lu). " 7268 "Please run e2fsck to fix.", type, 7269 err, qf_inums[type]); 7270 7271 ext4_quotas_off(sb, type); 7272 return err; 7273 } 7274 } 7275 } 7276 return 0; 7277 } 7278 7279 static int ext4_quota_off(struct super_block *sb, int type) 7280 { 7281 struct inode *inode = sb_dqopt(sb)->files[type]; 7282 handle_t *handle; 7283 int err; 7284 7285 /* Force all delayed allocation blocks to be allocated. 7286 * Caller already holds s_umount sem */ 7287 if (test_opt(sb, DELALLOC)) 7288 sync_filesystem(sb); 7289 7290 if (!inode || !igrab(inode)) 7291 goto out; 7292 7293 err = dquot_quota_off(sb, type); 7294 if (err || ext4_has_feature_quota(sb)) 7295 goto out_put; 7296 /* 7297 * When the filesystem was remounted read-only first, we cannot cleanup 7298 * inode flags here. Bad luck but people should be using QUOTA feature 7299 * these days anyway. 7300 */ 7301 if (sb_rdonly(sb)) 7302 goto out_put; 7303 7304 inode_lock(inode); 7305 /* 7306 * Update modification times of quota files when userspace can 7307 * start looking at them. If we fail, we return success anyway since 7308 * this is not a hard failure and quotas are already disabled. 7309 */ 7310 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 7311 if (IS_ERR(handle)) { 7312 err = PTR_ERR(handle); 7313 goto out_unlock; 7314 } 7315 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); 7316 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 7317 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 7318 err = ext4_mark_inode_dirty(handle, inode); 7319 ext4_journal_stop(handle); 7320 out_unlock: 7321 inode_unlock(inode); 7322 out_put: 7323 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); 7324 iput(inode); 7325 return err; 7326 out: 7327 return dquot_quota_off(sb, type); 7328 } 7329 7330 /* Read data from quotafile - avoid pagecache and such because we cannot afford 7331 * acquiring the locks... As quota files are never truncated and quota code 7332 * itself serializes the operations (and no one else should touch the files) 7333 * we don't have to be afraid of races */ 7334 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 7335 size_t len, loff_t off) 7336 { 7337 struct inode *inode = sb_dqopt(sb)->files[type]; 7338 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 7339 int offset = off & (sb->s_blocksize - 1); 7340 int tocopy; 7341 size_t toread; 7342 struct buffer_head *bh; 7343 loff_t i_size = i_size_read(inode); 7344 7345 if (off > i_size) 7346 return 0; 7347 if (off+len > i_size) 7348 len = i_size-off; 7349 toread = len; 7350 while (toread > 0) { 7351 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 7352 bh = ext4_bread(NULL, inode, blk, 0); 7353 if (IS_ERR(bh)) 7354 return PTR_ERR(bh); 7355 if (!bh) /* A hole? */ 7356 memset(data, 0, tocopy); 7357 else 7358 memcpy(data, bh->b_data+offset, tocopy); 7359 brelse(bh); 7360 offset = 0; 7361 toread -= tocopy; 7362 data += tocopy; 7363 blk++; 7364 } 7365 return len; 7366 } 7367 7368 /* Write to quotafile (we know the transaction is already started and has 7369 * enough credits) */ 7370 static ssize_t ext4_quota_write(struct super_block *sb, int type, 7371 const char *data, size_t len, loff_t off) 7372 { 7373 struct inode *inode = sb_dqopt(sb)->files[type]; 7374 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 7375 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1); 7376 int retries = 0; 7377 struct buffer_head *bh; 7378 handle_t *handle = journal_current_handle(); 7379 7380 if (!handle) { 7381 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 7382 " cancelled because transaction is not started", 7383 (unsigned long long)off, (unsigned long long)len); 7384 return -EIO; 7385 } 7386 /* 7387 * Since we account only one data block in transaction credits, 7388 * then it is impossible to cross a block boundary. 7389 */ 7390 if (sb->s_blocksize - offset < len) { 7391 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 7392 " cancelled because not block aligned", 7393 (unsigned long long)off, (unsigned long long)len); 7394 return -EIO; 7395 } 7396 7397 do { 7398 bh = ext4_bread(handle, inode, blk, 7399 EXT4_GET_BLOCKS_CREATE | 7400 EXT4_GET_BLOCKS_METADATA_NOFAIL); 7401 } while (PTR_ERR(bh) == -ENOSPC && 7402 ext4_should_retry_alloc(inode->i_sb, &retries)); 7403 if (IS_ERR(bh)) 7404 return PTR_ERR(bh); 7405 if (!bh) 7406 goto out; 7407 BUFFER_TRACE(bh, "get write access"); 7408 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE); 7409 if (err) { 7410 brelse(bh); 7411 return err; 7412 } 7413 lock_buffer(bh); 7414 memcpy(bh->b_data+offset, data, len); 7415 flush_dcache_folio(bh->b_folio); 7416 unlock_buffer(bh); 7417 err = ext4_handle_dirty_metadata(handle, NULL, bh); 7418 brelse(bh); 7419 out: 7420 if (inode->i_size < off + len) { 7421 i_size_write(inode, off + len); 7422 EXT4_I(inode)->i_disksize = inode->i_size; 7423 err2 = ext4_mark_inode_dirty(handle, inode); 7424 if (unlikely(err2 && !err)) 7425 err = err2; 7426 } 7427 return err ? err : len; 7428 } 7429 #endif 7430 7431 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 7432 static inline void register_as_ext2(void) 7433 { 7434 int err = register_filesystem(&ext2_fs_type); 7435 if (err) 7436 printk(KERN_WARNING 7437 "EXT4-fs: Unable to register as ext2 (%d)\n", err); 7438 } 7439 7440 static inline void unregister_as_ext2(void) 7441 { 7442 unregister_filesystem(&ext2_fs_type); 7443 } 7444 7445 static inline int ext2_feature_set_ok(struct super_block *sb) 7446 { 7447 if (ext4_has_unknown_ext2_incompat_features(sb)) 7448 return 0; 7449 if (sb_rdonly(sb)) 7450 return 1; 7451 if (ext4_has_unknown_ext2_ro_compat_features(sb)) 7452 return 0; 7453 return 1; 7454 } 7455 #else 7456 static inline void register_as_ext2(void) { } 7457 static inline void unregister_as_ext2(void) { } 7458 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 7459 #endif 7460 7461 static inline void register_as_ext3(void) 7462 { 7463 int err = register_filesystem(&ext3_fs_type); 7464 if (err) 7465 printk(KERN_WARNING 7466 "EXT4-fs: Unable to register as ext3 (%d)\n", err); 7467 } 7468 7469 static inline void unregister_as_ext3(void) 7470 { 7471 unregister_filesystem(&ext3_fs_type); 7472 } 7473 7474 static inline int ext3_feature_set_ok(struct super_block *sb) 7475 { 7476 if (ext4_has_unknown_ext3_incompat_features(sb)) 7477 return 0; 7478 if (!ext4_has_feature_journal(sb)) 7479 return 0; 7480 if (sb_rdonly(sb)) 7481 return 1; 7482 if (ext4_has_unknown_ext3_ro_compat_features(sb)) 7483 return 0; 7484 return 1; 7485 } 7486 7487 static void ext4_kill_sb(struct super_block *sb) 7488 { 7489 struct ext4_sb_info *sbi = EXT4_SB(sb); 7490 struct file *bdev_file = sbi ? sbi->s_journal_bdev_file : NULL; 7491 7492 kill_block_super(sb); 7493 7494 if (bdev_file) 7495 bdev_fput(bdev_file); 7496 } 7497 7498 static struct file_system_type ext4_fs_type = { 7499 .owner = THIS_MODULE, 7500 .name = "ext4", 7501 .init_fs_context = ext4_init_fs_context, 7502 .parameters = ext4_param_specs, 7503 .kill_sb = ext4_kill_sb, 7504 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP | FS_MGTIME | 7505 FS_LBS, 7506 }; 7507 MODULE_ALIAS_FS("ext4"); 7508 7509 static int __init ext4_init_fs(void) 7510 { 7511 int err; 7512 7513 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 7514 ext4_li_info = NULL; 7515 7516 /* Build-time check for flags consistency */ 7517 ext4_check_flag_values(); 7518 7519 err = ext4_init_es(); 7520 if (err) 7521 return err; 7522 7523 err = ext4_init_pending(); 7524 if (err) 7525 goto out7; 7526 7527 err = ext4_init_post_read_processing(); 7528 if (err) 7529 goto out6; 7530 7531 err = ext4_init_pageio(); 7532 if (err) 7533 goto out5; 7534 7535 err = ext4_init_system_zone(); 7536 if (err) 7537 goto out4; 7538 7539 err = ext4_init_sysfs(); 7540 if (err) 7541 goto out3; 7542 7543 err = ext4_init_mballoc(); 7544 if (err) 7545 goto out2; 7546 err = init_inodecache(); 7547 if (err) 7548 goto out1; 7549 7550 err = ext4_fc_init_dentry_cache(); 7551 if (err) 7552 goto out05; 7553 7554 register_as_ext3(); 7555 register_as_ext2(); 7556 err = register_filesystem(&ext4_fs_type); 7557 if (err) 7558 goto out; 7559 7560 return 0; 7561 out: 7562 unregister_as_ext2(); 7563 unregister_as_ext3(); 7564 ext4_fc_destroy_dentry_cache(); 7565 out05: 7566 destroy_inodecache(); 7567 out1: 7568 ext4_exit_mballoc(); 7569 out2: 7570 ext4_exit_sysfs(); 7571 out3: 7572 ext4_exit_system_zone(); 7573 out4: 7574 ext4_exit_pageio(); 7575 out5: 7576 ext4_exit_post_read_processing(); 7577 out6: 7578 ext4_exit_pending(); 7579 out7: 7580 ext4_exit_es(); 7581 7582 return err; 7583 } 7584 7585 static void __exit ext4_exit_fs(void) 7586 { 7587 ext4_destroy_lazyinit_thread(); 7588 unregister_as_ext2(); 7589 unregister_as_ext3(); 7590 unregister_filesystem(&ext4_fs_type); 7591 ext4_fc_destroy_dentry_cache(); 7592 destroy_inodecache(); 7593 ext4_exit_mballoc(); 7594 ext4_exit_sysfs(); 7595 ext4_exit_system_zone(); 7596 ext4_exit_pageio(); 7597 ext4_exit_post_read_processing(); 7598 ext4_exit_es(); 7599 ext4_exit_pending(); 7600 } 7601 7602 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 7603 MODULE_DESCRIPTION("Fourth Extended Filesystem"); 7604 MODULE_LICENSE("GPL"); 7605 module_init(ext4_init_fs) 7606 module_exit(ext4_exit_fs) 7607