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