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