1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * linux/fs/ext4/super.c 4 * 5 * Copyright (C) 1992, 1993, 1994, 1995 6 * Remy Card (card@masi.ibp.fr) 7 * Laboratoire MASI - Institut Blaise Pascal 8 * Universite Pierre et Marie Curie (Paris VI) 9 * 10 * from 11 * 12 * linux/fs/minix/inode.c 13 * 14 * Copyright (C) 1991, 1992 Linus Torvalds 15 * 16 * Big-endian to little-endian byte-swapping/bitmaps by 17 * David S. Miller (davem@caip.rutgers.edu), 1995 18 */ 19 20 #include <linux/module.h> 21 #include <linux/string.h> 22 #include <linux/fs.h> 23 #include <linux/time.h> 24 #include <linux/vmalloc.h> 25 #include <linux/slab.h> 26 #include <linux/init.h> 27 #include <linux/blkdev.h> 28 #include <linux/backing-dev.h> 29 #include <linux/parser.h> 30 #include <linux/buffer_head.h> 31 #include <linux/exportfs.h> 32 #include <linux/vfs.h> 33 #include <linux/random.h> 34 #include <linux/mount.h> 35 #include <linux/namei.h> 36 #include <linux/quotaops.h> 37 #include <linux/seq_file.h> 38 #include <linux/ctype.h> 39 #include <linux/log2.h> 40 #include <linux/crc16.h> 41 #include <linux/dax.h> 42 #include <linux/uaccess.h> 43 #include <linux/iversion.h> 44 #include <linux/unicode.h> 45 #include <linux/part_stat.h> 46 #include <linux/kthread.h> 47 #include <linux/freezer.h> 48 #include <linux/fsnotify.h> 49 #include <linux/fs_context.h> 50 #include <linux/fs_parser.h> 51 52 #include "ext4.h" 53 #include "ext4_extents.h" /* Needed for trace points definition */ 54 #include "ext4_jbd2.h" 55 #include "xattr.h" 56 #include "acl.h" 57 #include "mballoc.h" 58 #include "fsmap.h" 59 60 #define CREATE_TRACE_POINTS 61 #include <trace/events/ext4.h> 62 63 static struct ext4_lazy_init *ext4_li_info; 64 static DEFINE_MUTEX(ext4_li_mtx); 65 static struct ratelimit_state ext4_mount_msg_ratelimit; 66 67 static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 68 unsigned long journal_devnum); 69 static int ext4_show_options(struct seq_file *seq, struct dentry *root); 70 static void ext4_update_super(struct super_block *sb); 71 static int ext4_commit_super(struct super_block *sb); 72 static int ext4_mark_recovery_complete(struct super_block *sb, 73 struct ext4_super_block *es); 74 static int ext4_clear_journal_err(struct super_block *sb, 75 struct ext4_super_block *es); 76 static int ext4_sync_fs(struct super_block *sb, int wait); 77 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 78 static int ext4_unfreeze(struct super_block *sb); 79 static int ext4_freeze(struct super_block *sb); 80 static inline int ext2_feature_set_ok(struct super_block *sb); 81 static inline int ext3_feature_set_ok(struct super_block *sb); 82 static void ext4_unregister_li_request(struct super_block *sb); 83 static void ext4_clear_request_list(void); 84 static struct inode *ext4_get_journal_inode(struct super_block *sb, 85 unsigned int journal_inum); 86 static int ext4_validate_options(struct fs_context *fc); 87 static int ext4_check_opt_consistency(struct fs_context *fc, 88 struct super_block *sb); 89 static void ext4_apply_options(struct fs_context *fc, struct super_block *sb); 90 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param); 91 static int ext4_get_tree(struct fs_context *fc); 92 static int ext4_reconfigure(struct fs_context *fc); 93 static void ext4_fc_free(struct fs_context *fc); 94 static int ext4_init_fs_context(struct fs_context *fc); 95 static void ext4_kill_sb(struct super_block *sb); 96 static const struct fs_parameter_spec ext4_param_specs[]; 97 98 /* 99 * Lock ordering 100 * 101 * page fault path: 102 * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start 103 * -> page lock -> i_data_sem (rw) 104 * 105 * buffered write path: 106 * sb_start_write -> i_mutex -> mmap_lock 107 * sb_start_write -> i_mutex -> transaction start -> page lock -> 108 * i_data_sem (rw) 109 * 110 * truncate: 111 * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) -> 112 * page lock 113 * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start -> 114 * i_data_sem (rw) 115 * 116 * direct IO: 117 * sb_start_write -> i_mutex -> mmap_lock 118 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw) 119 * 120 * writepages: 121 * transaction start -> page lock(s) -> i_data_sem (rw) 122 */ 123 124 static const struct fs_context_operations ext4_context_ops = { 125 .parse_param = ext4_parse_param, 126 .get_tree = ext4_get_tree, 127 .reconfigure = ext4_reconfigure, 128 .free = ext4_fc_free, 129 }; 130 131 132 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 133 static struct file_system_type ext2_fs_type = { 134 .owner = THIS_MODULE, 135 .name = "ext2", 136 .init_fs_context = ext4_init_fs_context, 137 .parameters = ext4_param_specs, 138 .kill_sb = ext4_kill_sb, 139 .fs_flags = FS_REQUIRES_DEV, 140 }; 141 MODULE_ALIAS_FS("ext2"); 142 MODULE_ALIAS("ext2"); 143 #define IS_EXT2_SB(sb) ((sb)->s_type == &ext2_fs_type) 144 #else 145 #define IS_EXT2_SB(sb) (0) 146 #endif 147 148 149 static struct file_system_type ext3_fs_type = { 150 .owner = THIS_MODULE, 151 .name = "ext3", 152 .init_fs_context = ext4_init_fs_context, 153 .parameters = ext4_param_specs, 154 .kill_sb = ext4_kill_sb, 155 .fs_flags = FS_REQUIRES_DEV, 156 }; 157 MODULE_ALIAS_FS("ext3"); 158 MODULE_ALIAS("ext3"); 159 #define IS_EXT3_SB(sb) ((sb)->s_type == &ext3_fs_type) 160 161 162 static inline void __ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags, 163 bh_end_io_t *end_io, bool simu_fail) 164 { 165 if (simu_fail) { 166 clear_buffer_uptodate(bh); 167 unlock_buffer(bh); 168 return; 169 } 170 171 /* 172 * buffer's verified bit is no longer valid after reading from 173 * disk again due to write out error, clear it to make sure we 174 * recheck the buffer contents. 175 */ 176 clear_buffer_verified(bh); 177 178 bh->b_end_io = end_io ? end_io : end_buffer_read_sync; 179 get_bh(bh); 180 submit_bh(REQ_OP_READ | op_flags, bh); 181 } 182 183 void ext4_read_bh_nowait(struct buffer_head *bh, blk_opf_t op_flags, 184 bh_end_io_t *end_io, bool simu_fail) 185 { 186 BUG_ON(!buffer_locked(bh)); 187 188 if (ext4_buffer_uptodate(bh)) { 189 unlock_buffer(bh); 190 return; 191 } 192 __ext4_read_bh(bh, op_flags, end_io, simu_fail); 193 } 194 195 int ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags, 196 bh_end_io_t *end_io, bool simu_fail) 197 { 198 BUG_ON(!buffer_locked(bh)); 199 200 if (ext4_buffer_uptodate(bh)) { 201 unlock_buffer(bh); 202 return 0; 203 } 204 205 __ext4_read_bh(bh, op_flags, end_io, simu_fail); 206 207 wait_on_buffer(bh); 208 if (buffer_uptodate(bh)) 209 return 0; 210 return -EIO; 211 } 212 213 int ext4_read_bh_lock(struct buffer_head *bh, blk_opf_t op_flags, bool wait) 214 { 215 lock_buffer(bh); 216 if (!wait) { 217 ext4_read_bh_nowait(bh, op_flags, NULL, false); 218 return 0; 219 } 220 return ext4_read_bh(bh, op_flags, NULL, false); 221 } 222 223 /* 224 * This works like __bread_gfp() except it uses ERR_PTR for error 225 * returns. Currently with sb_bread it's impossible to distinguish 226 * between ENOMEM and EIO situations (since both result in a NULL 227 * return. 228 */ 229 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb, 230 sector_t block, 231 blk_opf_t op_flags, gfp_t gfp) 232 { 233 struct buffer_head *bh; 234 int ret; 235 236 bh = sb_getblk_gfp(sb, block, gfp); 237 if (bh == NULL) 238 return ERR_PTR(-ENOMEM); 239 if (ext4_buffer_uptodate(bh)) 240 return bh; 241 242 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true); 243 if (ret) { 244 put_bh(bh); 245 return ERR_PTR(ret); 246 } 247 return bh; 248 } 249 250 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block, 251 blk_opf_t op_flags) 252 { 253 gfp_t gfp = mapping_gfp_constraint(sb->s_bdev->bd_mapping, 254 ~__GFP_FS) | __GFP_MOVABLE; 255 256 return __ext4_sb_bread_gfp(sb, block, op_flags, gfp); 257 } 258 259 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb, 260 sector_t block) 261 { 262 gfp_t gfp = mapping_gfp_constraint(sb->s_bdev->bd_mapping, 263 ~__GFP_FS); 264 265 return __ext4_sb_bread_gfp(sb, block, 0, gfp); 266 } 267 268 struct buffer_head *ext4_sb_bread_nofail(struct super_block *sb, 269 sector_t block) 270 { 271 gfp_t gfp = mapping_gfp_constraint(sb->s_bdev->bd_mapping, 272 ~__GFP_FS) | __GFP_MOVABLE | __GFP_NOFAIL; 273 274 return __ext4_sb_bread_gfp(sb, block, 0, gfp); 275 } 276 277 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block) 278 { 279 struct buffer_head *bh = bdev_getblk(sb->s_bdev, block, 280 sb->s_blocksize, GFP_NOWAIT); 281 282 if (likely(bh)) { 283 if (trylock_buffer(bh)) 284 ext4_read_bh_nowait(bh, REQ_RAHEAD, NULL, false); 285 brelse(bh); 286 } 287 } 288 289 static int ext4_verify_csum_type(struct super_block *sb, 290 struct ext4_super_block *es) 291 { 292 if (!ext4_has_feature_metadata_csum(sb)) 293 return 1; 294 295 return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 296 } 297 298 __le32 ext4_superblock_csum(struct ext4_super_block *es) 299 { 300 int offset = offsetof(struct ext4_super_block, s_checksum); 301 __u32 csum; 302 303 csum = ext4_chksum(~0, (char *)es, offset); 304 305 return cpu_to_le32(csum); 306 } 307 308 static int ext4_superblock_csum_verify(struct super_block *sb, 309 struct ext4_super_block *es) 310 { 311 if (!ext4_has_feature_metadata_csum(sb)) 312 return 1; 313 314 return es->s_checksum == ext4_superblock_csum(es); 315 } 316 317 void ext4_superblock_csum_set(struct super_block *sb) 318 { 319 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 320 321 if (!ext4_has_feature_metadata_csum(sb)) 322 return; 323 324 es->s_checksum = ext4_superblock_csum(es); 325 } 326 327 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 328 struct ext4_group_desc *bg) 329 { 330 return le32_to_cpu(bg->bg_block_bitmap_lo) | 331 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 332 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 333 } 334 335 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 336 struct ext4_group_desc *bg) 337 { 338 return le32_to_cpu(bg->bg_inode_bitmap_lo) | 339 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 340 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 341 } 342 343 ext4_fsblk_t ext4_inode_table(struct super_block *sb, 344 struct ext4_group_desc *bg) 345 { 346 return le32_to_cpu(bg->bg_inode_table_lo) | 347 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 348 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 349 } 350 351 __u32 ext4_free_group_clusters(struct super_block *sb, 352 struct ext4_group_desc *bg) 353 { 354 return le16_to_cpu(bg->bg_free_blocks_count_lo) | 355 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 356 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 357 } 358 359 __u32 ext4_free_inodes_count(struct super_block *sb, 360 struct ext4_group_desc *bg) 361 { 362 return le16_to_cpu(READ_ONCE(bg->bg_free_inodes_count_lo)) | 363 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 364 (__u32)le16_to_cpu(READ_ONCE(bg->bg_free_inodes_count_hi)) << 16 : 0); 365 } 366 367 __u32 ext4_used_dirs_count(struct super_block *sb, 368 struct ext4_group_desc *bg) 369 { 370 return le16_to_cpu(bg->bg_used_dirs_count_lo) | 371 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 372 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 373 } 374 375 __u32 ext4_itable_unused_count(struct super_block *sb, 376 struct ext4_group_desc *bg) 377 { 378 return le16_to_cpu(bg->bg_itable_unused_lo) | 379 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 380 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 381 } 382 383 void ext4_block_bitmap_set(struct super_block *sb, 384 struct ext4_group_desc *bg, ext4_fsblk_t blk) 385 { 386 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 387 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 388 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 389 } 390 391 void ext4_inode_bitmap_set(struct super_block *sb, 392 struct ext4_group_desc *bg, ext4_fsblk_t blk) 393 { 394 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 395 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 396 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 397 } 398 399 void ext4_inode_table_set(struct super_block *sb, 400 struct ext4_group_desc *bg, ext4_fsblk_t blk) 401 { 402 bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 403 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 404 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 405 } 406 407 void ext4_free_group_clusters_set(struct super_block *sb, 408 struct ext4_group_desc *bg, __u32 count) 409 { 410 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 411 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 412 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 413 } 414 415 void ext4_free_inodes_set(struct super_block *sb, 416 struct ext4_group_desc *bg, __u32 count) 417 { 418 WRITE_ONCE(bg->bg_free_inodes_count_lo, cpu_to_le16((__u16)count)); 419 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 420 WRITE_ONCE(bg->bg_free_inodes_count_hi, cpu_to_le16(count >> 16)); 421 } 422 423 void ext4_used_dirs_set(struct super_block *sb, 424 struct ext4_group_desc *bg, __u32 count) 425 { 426 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 427 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 428 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 429 } 430 431 void ext4_itable_unused_set(struct super_block *sb, 432 struct ext4_group_desc *bg, __u32 count) 433 { 434 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 435 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 436 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 437 } 438 439 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now) 440 { 441 now = clamp_val(now, 0, (1ull << 40) - 1); 442 443 *lo = cpu_to_le32(lower_32_bits(now)); 444 *hi = upper_32_bits(now); 445 } 446 447 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi) 448 { 449 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo); 450 } 451 #define ext4_update_tstamp(es, tstamp) \ 452 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \ 453 ktime_get_real_seconds()) 454 #define ext4_get_tstamp(es, tstamp) \ 455 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 456 457 /* 458 * The ext4_maybe_update_superblock() function checks and updates the 459 * superblock if needed. 460 * 461 * This function is designed to update the on-disk superblock only under 462 * certain conditions to prevent excessive disk writes and unnecessary 463 * waking of the disk from sleep. The superblock will be updated if: 464 * 1. More than sbi->s_sb_update_sec (def: 1 hour) has passed since the last 465 * superblock update 466 * 2. More than sbi->s_sb_update_kb (def: 16MB) kbs have been written since the 467 * last superblock update. 468 * 469 * @sb: The superblock 470 */ 471 static void ext4_maybe_update_superblock(struct super_block *sb) 472 { 473 struct ext4_sb_info *sbi = EXT4_SB(sb); 474 struct ext4_super_block *es = sbi->s_es; 475 journal_t *journal = sbi->s_journal; 476 time64_t now; 477 __u64 last_update; 478 __u64 lifetime_write_kbytes; 479 __u64 diff_size; 480 481 if (ext4_emergency_state(sb) || sb_rdonly(sb) || 482 !(sb->s_flags & SB_ACTIVE) || !journal || 483 journal->j_flags & JBD2_UNMOUNT) 484 return; 485 486 now = ktime_get_real_seconds(); 487 last_update = ext4_get_tstamp(es, s_wtime); 488 489 if (likely(now - last_update < sbi->s_sb_update_sec)) 490 return; 491 492 lifetime_write_kbytes = sbi->s_kbytes_written + 493 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) - 494 sbi->s_sectors_written_start) >> 1); 495 496 /* Get the number of kilobytes not written to disk to account 497 * for statistics and compare with a multiple of 16 MB. This 498 * is used to determine when the next superblock commit should 499 * occur (i.e. not more often than once per 16MB if there was 500 * less written in an hour). 501 */ 502 diff_size = lifetime_write_kbytes - le64_to_cpu(es->s_kbytes_written); 503 504 if (diff_size > sbi->s_sb_update_kb) 505 schedule_work(&EXT4_SB(sb)->s_sb_upd_work); 506 } 507 508 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 509 { 510 struct super_block *sb = journal->j_private; 511 512 BUG_ON(txn->t_state == T_FINISHED); 513 514 ext4_process_freed_data(sb, txn->t_tid); 515 ext4_maybe_update_superblock(sb); 516 } 517 518 static bool ext4_journalled_writepage_needs_redirty(struct jbd2_inode *jinode, 519 struct folio *folio) 520 { 521 struct buffer_head *bh, *head; 522 struct journal_head *jh; 523 524 bh = head = folio_buffers(folio); 525 do { 526 /* 527 * We have to redirty a page in these cases: 528 * 1) If buffer is dirty, it means the page was dirty because it 529 * contains a buffer that needs checkpointing. So the dirty bit 530 * needs to be preserved so that checkpointing writes the buffer 531 * properly. 532 * 2) If buffer is not part of the committing transaction 533 * (we may have just accidentally come across this buffer because 534 * inode range tracking is not exact) or if the currently running 535 * transaction already contains this buffer as well, dirty bit 536 * needs to be preserved so that the buffer gets writeprotected 537 * properly on running transaction's commit. 538 */ 539 jh = bh2jh(bh); 540 if (buffer_dirty(bh) || 541 (jh && (jh->b_transaction != jinode->i_transaction || 542 jh->b_next_transaction))) 543 return true; 544 } while ((bh = bh->b_this_page) != head); 545 546 return false; 547 } 548 549 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode) 550 { 551 struct address_space *mapping = jinode->i_vfs_inode->i_mapping; 552 struct writeback_control wbc = { 553 .sync_mode = WB_SYNC_ALL, 554 .nr_to_write = LONG_MAX, 555 .range_start = jinode->i_dirty_start, 556 .range_end = jinode->i_dirty_end, 557 }; 558 struct folio *folio = NULL; 559 int error; 560 561 /* 562 * writeback_iter() already checks for dirty pages and calls 563 * folio_clear_dirty_for_io(), which we want to write protect the 564 * folios. 565 * 566 * However, we may have to redirty a folio sometimes. 567 */ 568 while ((folio = writeback_iter(mapping, &wbc, folio, &error))) { 569 if (ext4_journalled_writepage_needs_redirty(jinode, folio)) 570 folio_redirty_for_writepage(&wbc, folio); 571 folio_unlock(folio); 572 } 573 574 return error; 575 } 576 577 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode) 578 { 579 int ret; 580 581 if (ext4_should_journal_data(jinode->i_vfs_inode)) 582 ret = ext4_journalled_submit_inode_data_buffers(jinode); 583 else 584 ret = ext4_normal_submit_inode_data_buffers(jinode); 585 return ret; 586 } 587 588 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode) 589 { 590 int ret = 0; 591 592 if (!ext4_should_journal_data(jinode->i_vfs_inode)) 593 ret = jbd2_journal_finish_inode_data_buffers(jinode); 594 595 return ret; 596 } 597 598 static bool system_going_down(void) 599 { 600 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 601 || system_state == SYSTEM_RESTART; 602 } 603 604 struct ext4_err_translation { 605 int code; 606 int errno; 607 }; 608 609 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err } 610 611 static struct ext4_err_translation err_translation[] = { 612 EXT4_ERR_TRANSLATE(EIO), 613 EXT4_ERR_TRANSLATE(ENOMEM), 614 EXT4_ERR_TRANSLATE(EFSBADCRC), 615 EXT4_ERR_TRANSLATE(EFSCORRUPTED), 616 EXT4_ERR_TRANSLATE(ENOSPC), 617 EXT4_ERR_TRANSLATE(ENOKEY), 618 EXT4_ERR_TRANSLATE(EROFS), 619 EXT4_ERR_TRANSLATE(EFBIG), 620 EXT4_ERR_TRANSLATE(EEXIST), 621 EXT4_ERR_TRANSLATE(ERANGE), 622 EXT4_ERR_TRANSLATE(EOVERFLOW), 623 EXT4_ERR_TRANSLATE(EBUSY), 624 EXT4_ERR_TRANSLATE(ENOTDIR), 625 EXT4_ERR_TRANSLATE(ENOTEMPTY), 626 EXT4_ERR_TRANSLATE(ESHUTDOWN), 627 EXT4_ERR_TRANSLATE(EFAULT), 628 }; 629 630 static int ext4_errno_to_code(int errno) 631 { 632 int i; 633 634 for (i = 0; i < ARRAY_SIZE(err_translation); i++) 635 if (err_translation[i].errno == errno) 636 return err_translation[i].code; 637 return EXT4_ERR_UNKNOWN; 638 } 639 640 static void save_error_info(struct super_block *sb, int error, 641 __u32 ino, __u64 block, 642 const char *func, unsigned int line) 643 { 644 struct ext4_sb_info *sbi = EXT4_SB(sb); 645 646 /* We default to EFSCORRUPTED error... */ 647 if (error == 0) 648 error = EFSCORRUPTED; 649 650 spin_lock(&sbi->s_error_lock); 651 sbi->s_add_error_count++; 652 sbi->s_last_error_code = error; 653 sbi->s_last_error_line = line; 654 sbi->s_last_error_ino = ino; 655 sbi->s_last_error_block = block; 656 sbi->s_last_error_func = func; 657 sbi->s_last_error_time = ktime_get_real_seconds(); 658 if (!sbi->s_first_error_time) { 659 sbi->s_first_error_code = error; 660 sbi->s_first_error_line = line; 661 sbi->s_first_error_ino = ino; 662 sbi->s_first_error_block = block; 663 sbi->s_first_error_func = func; 664 sbi->s_first_error_time = sbi->s_last_error_time; 665 } 666 spin_unlock(&sbi->s_error_lock); 667 } 668 669 /* Deal with the reporting of failure conditions on a filesystem such as 670 * inconsistencies detected or read IO failures. 671 * 672 * On ext2, we can store the error state of the filesystem in the 673 * superblock. That is not possible on ext4, because we may have other 674 * write ordering constraints on the superblock which prevent us from 675 * writing it out straight away; and given that the journal is about to 676 * be aborted, we can't rely on the current, or future, transactions to 677 * write out the superblock safely. 678 * 679 * We'll just use the jbd2_journal_abort() error code to record an error in 680 * the journal instead. On recovery, the journal will complain about 681 * that error until we've noted it down and cleared it. 682 * 683 * If force_ro is set, we unconditionally force the filesystem into an 684 * ABORT|READONLY state, unless the error response on the fs has been set to 685 * panic in which case we take the easy way out and panic immediately. This is 686 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM 687 * at a critical moment in log management. 688 */ 689 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error, 690 __u32 ino, __u64 block, 691 const char *func, unsigned int line) 692 { 693 journal_t *journal = EXT4_SB(sb)->s_journal; 694 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT); 695 696 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 697 if (test_opt(sb, WARN_ON_ERROR)) 698 WARN_ON_ONCE(1); 699 700 if (!continue_fs && !ext4_emergency_ro(sb) && journal) 701 jbd2_journal_abort(journal, -error); 702 703 if (!bdev_read_only(sb->s_bdev)) { 704 save_error_info(sb, error, ino, block, func, line); 705 /* 706 * In case the fs should keep running, we need to writeout 707 * superblock through the journal. Due to lock ordering 708 * constraints, it may not be safe to do it right here so we 709 * defer superblock flushing to a workqueue. We just need to be 710 * careful when the journal is already shutting down. If we get 711 * here in that case, just update the sb directly as the last 712 * transaction won't commit anyway. 713 */ 714 if (continue_fs && journal && 715 !ext4_test_mount_flag(sb, EXT4_MF_JOURNAL_DESTROY)) 716 schedule_work(&EXT4_SB(sb)->s_sb_upd_work); 717 else 718 ext4_commit_super(sb); 719 } 720 721 /* 722 * We force ERRORS_RO behavior when system is rebooting. Otherwise we 723 * could panic during 'reboot -f' as the underlying device got already 724 * disabled. 725 */ 726 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) { 727 panic("EXT4-fs (device %s): panic forced after error\n", 728 sb->s_id); 729 } 730 731 if (ext4_emergency_ro(sb) || continue_fs) 732 return; 733 734 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 735 /* 736 * We don't set SB_RDONLY because that requires sb->s_umount 737 * semaphore and setting it without proper remount procedure is 738 * confusing code such as freeze_super() leading to deadlocks 739 * and other problems. 740 */ 741 set_bit(EXT4_FLAGS_EMERGENCY_RO, &EXT4_SB(sb)->s_ext4_flags); 742 } 743 744 static void update_super_work(struct work_struct *work) 745 { 746 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info, 747 s_sb_upd_work); 748 journal_t *journal = sbi->s_journal; 749 handle_t *handle; 750 751 /* 752 * If the journal is still running, we have to write out superblock 753 * through the journal to avoid collisions of other journalled sb 754 * updates. 755 * 756 * We use directly jbd2 functions here to avoid recursing back into 757 * ext4 error handling code during handling of previous errors. 758 */ 759 if (!ext4_emergency_state(sbi->s_sb) && 760 !sb_rdonly(sbi->s_sb) && journal) { 761 struct buffer_head *sbh = sbi->s_sbh; 762 bool call_notify_err = false; 763 764 handle = jbd2_journal_start(journal, 1); 765 if (IS_ERR(handle)) 766 goto write_directly; 767 if (jbd2_journal_get_write_access(handle, sbh)) { 768 jbd2_journal_stop(handle); 769 goto write_directly; 770 } 771 772 if (sbi->s_add_error_count > 0) 773 call_notify_err = true; 774 775 ext4_update_super(sbi->s_sb); 776 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 777 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to " 778 "superblock detected"); 779 clear_buffer_write_io_error(sbh); 780 set_buffer_uptodate(sbh); 781 } 782 783 if (jbd2_journal_dirty_metadata(handle, sbh)) { 784 jbd2_journal_stop(handle); 785 goto write_directly; 786 } 787 jbd2_journal_stop(handle); 788 789 if (call_notify_err) 790 ext4_notify_error_sysfs(sbi); 791 792 return; 793 } 794 write_directly: 795 /* 796 * Write through journal failed. Write sb directly to get error info 797 * out and hope for the best. 798 */ 799 ext4_commit_super(sbi->s_sb); 800 ext4_notify_error_sysfs(sbi); 801 } 802 803 #define ext4_error_ratelimit(sb) \ 804 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ 805 "EXT4-fs error") 806 807 void __ext4_error(struct super_block *sb, const char *function, 808 unsigned int line, bool force_ro, int error, __u64 block, 809 const char *fmt, ...) 810 { 811 struct va_format vaf; 812 va_list args; 813 814 if (unlikely(ext4_emergency_state(sb))) 815 return; 816 817 trace_ext4_error(sb, function, line); 818 if (ext4_error_ratelimit(sb)) { 819 va_start(args, fmt); 820 vaf.fmt = fmt; 821 vaf.va = &args; 822 printk(KERN_CRIT 823 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 824 sb->s_id, function, line, current->comm, &vaf); 825 va_end(args); 826 } 827 fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED); 828 829 ext4_handle_error(sb, force_ro, error, 0, block, function, line); 830 } 831 832 void __ext4_error_inode(struct inode *inode, const char *function, 833 unsigned int line, ext4_fsblk_t block, int error, 834 const char *fmt, ...) 835 { 836 va_list args; 837 struct va_format vaf; 838 839 if (unlikely(ext4_emergency_state(inode->i_sb))) 840 return; 841 842 trace_ext4_error(inode->i_sb, function, line); 843 if (ext4_error_ratelimit(inode->i_sb)) { 844 va_start(args, fmt); 845 vaf.fmt = fmt; 846 vaf.va = &args; 847 if (block) 848 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 849 "inode #%lu: block %llu: comm %s: %pV\n", 850 inode->i_sb->s_id, function, line, inode->i_ino, 851 block, current->comm, &vaf); 852 else 853 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 854 "inode #%lu: comm %s: %pV\n", 855 inode->i_sb->s_id, function, line, inode->i_ino, 856 current->comm, &vaf); 857 va_end(args); 858 } 859 fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED); 860 861 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block, 862 function, line); 863 } 864 865 void __ext4_error_file(struct file *file, const char *function, 866 unsigned int line, ext4_fsblk_t block, 867 const char *fmt, ...) 868 { 869 va_list args; 870 struct va_format vaf; 871 struct inode *inode = file_inode(file); 872 char pathname[80], *path; 873 874 if (unlikely(ext4_emergency_state(inode->i_sb))) 875 return; 876 877 trace_ext4_error(inode->i_sb, function, line); 878 if (ext4_error_ratelimit(inode->i_sb)) { 879 path = file_path(file, pathname, sizeof(pathname)); 880 if (IS_ERR(path)) 881 path = "(unknown)"; 882 va_start(args, fmt); 883 vaf.fmt = fmt; 884 vaf.va = &args; 885 if (block) 886 printk(KERN_CRIT 887 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 888 "block %llu: comm %s: path %s: %pV\n", 889 inode->i_sb->s_id, function, line, inode->i_ino, 890 block, current->comm, path, &vaf); 891 else 892 printk(KERN_CRIT 893 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 894 "comm %s: path %s: %pV\n", 895 inode->i_sb->s_id, function, line, inode->i_ino, 896 current->comm, path, &vaf); 897 va_end(args); 898 } 899 fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED); 900 901 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block, 902 function, line); 903 } 904 905 const char *ext4_decode_error(struct super_block *sb, int errno, 906 char nbuf[16]) 907 { 908 char *errstr = NULL; 909 910 switch (errno) { 911 case -EFSCORRUPTED: 912 errstr = "Corrupt filesystem"; 913 break; 914 case -EFSBADCRC: 915 errstr = "Filesystem failed CRC"; 916 break; 917 case -EIO: 918 errstr = "IO failure"; 919 break; 920 case -ENOMEM: 921 errstr = "Out of memory"; 922 break; 923 case -EROFS: 924 if (!sb || (EXT4_SB(sb)->s_journal && 925 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 926 errstr = "Journal has aborted"; 927 else 928 errstr = "Readonly filesystem"; 929 break; 930 default: 931 /* If the caller passed in an extra buffer for unknown 932 * errors, textualise them now. Else we just return 933 * NULL. */ 934 if (nbuf) { 935 /* Check for truncated error codes... */ 936 if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 937 errstr = nbuf; 938 } 939 break; 940 } 941 942 return errstr; 943 } 944 945 /* __ext4_std_error decodes expected errors from journaling functions 946 * automatically and invokes the appropriate error response. */ 947 948 void __ext4_std_error(struct super_block *sb, const char *function, 949 unsigned int line, int errno) 950 { 951 char nbuf[16]; 952 const char *errstr; 953 954 if (unlikely(ext4_emergency_state(sb))) 955 return; 956 957 /* Special case: if the error is EROFS, and we're not already 958 * inside a transaction, then there's really no point in logging 959 * an error. */ 960 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb)) 961 return; 962 963 if (ext4_error_ratelimit(sb)) { 964 errstr = ext4_decode_error(sb, errno, nbuf); 965 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 966 sb->s_id, function, line, errstr); 967 } 968 fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED); 969 970 ext4_handle_error(sb, false, -errno, 0, 0, function, line); 971 } 972 973 void __ext4_msg(struct super_block *sb, 974 const char *prefix, const char *fmt, ...) 975 { 976 struct va_format vaf; 977 va_list args; 978 979 if (sb) { 980 atomic_inc(&EXT4_SB(sb)->s_msg_count); 981 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), 982 "EXT4-fs")) 983 return; 984 } 985 986 va_start(args, fmt); 987 vaf.fmt = fmt; 988 vaf.va = &args; 989 if (sb) 990 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 991 else 992 printk("%sEXT4-fs: %pV\n", prefix, &vaf); 993 va_end(args); 994 } 995 996 static int ext4_warning_ratelimit(struct super_block *sb) 997 { 998 atomic_inc(&EXT4_SB(sb)->s_warning_count); 999 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), 1000 "EXT4-fs warning"); 1001 } 1002 1003 void __ext4_warning(struct super_block *sb, const char *function, 1004 unsigned int line, const char *fmt, ...) 1005 { 1006 struct va_format vaf; 1007 va_list args; 1008 1009 if (!ext4_warning_ratelimit(sb)) 1010 return; 1011 1012 va_start(args, fmt); 1013 vaf.fmt = fmt; 1014 vaf.va = &args; 1015 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 1016 sb->s_id, function, line, &vaf); 1017 va_end(args); 1018 } 1019 1020 void __ext4_warning_inode(const struct inode *inode, const char *function, 1021 unsigned int line, const char *fmt, ...) 1022 { 1023 struct va_format vaf; 1024 va_list args; 1025 1026 if (!ext4_warning_ratelimit(inode->i_sb)) 1027 return; 1028 1029 va_start(args, fmt); 1030 vaf.fmt = fmt; 1031 vaf.va = &args; 1032 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: " 1033 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id, 1034 function, line, inode->i_ino, current->comm, &vaf); 1035 va_end(args); 1036 } 1037 1038 void __ext4_grp_locked_error(const char *function, unsigned int line, 1039 struct super_block *sb, ext4_group_t grp, 1040 unsigned long ino, ext4_fsblk_t block, 1041 const char *fmt, ...) 1042 __releases(bitlock) 1043 __acquires(bitlock) 1044 { 1045 struct va_format vaf; 1046 va_list args; 1047 1048 if (unlikely(ext4_emergency_state(sb))) 1049 return; 1050 1051 trace_ext4_error(sb, function, line); 1052 if (ext4_error_ratelimit(sb)) { 1053 va_start(args, fmt); 1054 vaf.fmt = fmt; 1055 vaf.va = &args; 1056 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 1057 sb->s_id, function, line, grp); 1058 if (ino) 1059 printk(KERN_CONT "inode %lu: ", ino); 1060 if (block) 1061 printk(KERN_CONT "block %llu:", 1062 (unsigned long long) block); 1063 printk(KERN_CONT "%pV\n", &vaf); 1064 va_end(args); 1065 } 1066 1067 if (test_opt(sb, ERRORS_CONT)) { 1068 if (test_opt(sb, WARN_ON_ERROR)) 1069 WARN_ON_ONCE(1); 1070 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 1071 if (!bdev_read_only(sb->s_bdev)) { 1072 save_error_info(sb, EFSCORRUPTED, ino, block, function, 1073 line); 1074 schedule_work(&EXT4_SB(sb)->s_sb_upd_work); 1075 } 1076 return; 1077 } 1078 ext4_unlock_group(sb, grp); 1079 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line); 1080 /* 1081 * We only get here in the ERRORS_RO case; relocking the group 1082 * may be dangerous, but nothing bad will happen since the 1083 * filesystem will have already been marked read/only and the 1084 * journal has been aborted. We return 1 as a hint to callers 1085 * who might what to use the return value from 1086 * ext4_grp_locked_error() to distinguish between the 1087 * ERRORS_CONT and ERRORS_RO case, and perhaps return more 1088 * aggressively from the ext4 function in question, with a 1089 * more appropriate error code. 1090 */ 1091 ext4_lock_group(sb, grp); 1092 return; 1093 } 1094 1095 void ext4_mark_group_bitmap_corrupted(struct super_block *sb, 1096 ext4_group_t group, 1097 unsigned int flags) 1098 { 1099 struct ext4_sb_info *sbi = EXT4_SB(sb); 1100 struct ext4_group_info *grp = ext4_get_group_info(sb, group); 1101 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL); 1102 int ret; 1103 1104 if (!grp || !gdp) 1105 return; 1106 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) { 1107 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, 1108 &grp->bb_state); 1109 if (!ret) 1110 percpu_counter_sub(&sbi->s_freeclusters_counter, 1111 grp->bb_free); 1112 } 1113 1114 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) { 1115 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, 1116 &grp->bb_state); 1117 if (!ret && gdp) { 1118 int count; 1119 1120 count = ext4_free_inodes_count(sb, gdp); 1121 percpu_counter_sub(&sbi->s_freeinodes_counter, 1122 count); 1123 } 1124 } 1125 } 1126 1127 void ext4_update_dynamic_rev(struct super_block *sb) 1128 { 1129 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 1130 1131 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 1132 return; 1133 1134 ext4_warning(sb, 1135 "updating to rev %d because of new feature flag, " 1136 "running e2fsck is recommended", 1137 EXT4_DYNAMIC_REV); 1138 1139 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 1140 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 1141 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 1142 /* leave es->s_feature_*compat flags alone */ 1143 /* es->s_uuid will be set by e2fsck if empty */ 1144 1145 /* 1146 * The rest of the superblock fields should be zero, and if not it 1147 * means they are likely already in use, so leave them alone. We 1148 * can leave it up to e2fsck to clean up any inconsistencies there. 1149 */ 1150 } 1151 1152 static inline struct inode *orphan_list_entry(struct list_head *l) 1153 { 1154 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 1155 } 1156 1157 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 1158 { 1159 struct list_head *l; 1160 1161 ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 1162 le32_to_cpu(sbi->s_es->s_last_orphan)); 1163 1164 printk(KERN_ERR "sb_info orphan list:\n"); 1165 list_for_each(l, &sbi->s_orphan) { 1166 struct inode *inode = orphan_list_entry(l); 1167 printk(KERN_ERR " " 1168 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 1169 inode->i_sb->s_id, inode->i_ino, inode, 1170 inode->i_mode, inode->i_nlink, 1171 NEXT_ORPHAN(inode)); 1172 } 1173 } 1174 1175 #ifdef CONFIG_QUOTA 1176 static int ext4_quota_off(struct super_block *sb, int type); 1177 1178 static inline void ext4_quotas_off(struct super_block *sb, int type) 1179 { 1180 BUG_ON(type > EXT4_MAXQUOTAS); 1181 1182 /* Use our quota_off function to clear inode flags etc. */ 1183 for (type--; type >= 0; type--) 1184 ext4_quota_off(sb, type); 1185 } 1186 1187 /* 1188 * This is a helper function which is used in the mount/remount 1189 * codepaths (which holds s_umount) to fetch the quota file name. 1190 */ 1191 static inline char *get_qf_name(struct super_block *sb, 1192 struct ext4_sb_info *sbi, 1193 int type) 1194 { 1195 return rcu_dereference_protected(sbi->s_qf_names[type], 1196 lockdep_is_held(&sb->s_umount)); 1197 } 1198 #else 1199 static inline void ext4_quotas_off(struct super_block *sb, int type) 1200 { 1201 } 1202 #endif 1203 1204 static int ext4_percpu_param_init(struct ext4_sb_info *sbi) 1205 { 1206 ext4_fsblk_t block; 1207 int err; 1208 1209 block = ext4_count_free_clusters(sbi->s_sb); 1210 ext4_free_blocks_count_set(sbi->s_es, EXT4_C2B(sbi, block)); 1211 err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 1212 GFP_KERNEL); 1213 if (!err) { 1214 unsigned long freei = ext4_count_free_inodes(sbi->s_sb); 1215 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 1216 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 1217 GFP_KERNEL); 1218 } 1219 if (!err) 1220 err = percpu_counter_init(&sbi->s_dirs_counter, 1221 ext4_count_dirs(sbi->s_sb), GFP_KERNEL); 1222 if (!err) 1223 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 1224 GFP_KERNEL); 1225 if (!err) 1226 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0, 1227 GFP_KERNEL); 1228 if (!err) 1229 err = percpu_init_rwsem(&sbi->s_writepages_rwsem); 1230 1231 if (err) 1232 ext4_msg(sbi->s_sb, KERN_ERR, "insufficient memory"); 1233 1234 return err; 1235 } 1236 1237 static void ext4_percpu_param_destroy(struct ext4_sb_info *sbi) 1238 { 1239 percpu_counter_destroy(&sbi->s_freeclusters_counter); 1240 percpu_counter_destroy(&sbi->s_freeinodes_counter); 1241 percpu_counter_destroy(&sbi->s_dirs_counter); 1242 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 1243 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit); 1244 percpu_free_rwsem(&sbi->s_writepages_rwsem); 1245 } 1246 1247 static void ext4_group_desc_free(struct ext4_sb_info *sbi) 1248 { 1249 struct buffer_head **group_desc; 1250 int i; 1251 1252 rcu_read_lock(); 1253 group_desc = rcu_dereference(sbi->s_group_desc); 1254 for (i = 0; i < sbi->s_gdb_count; i++) 1255 brelse(group_desc[i]); 1256 kvfree(group_desc); 1257 rcu_read_unlock(); 1258 } 1259 1260 static void ext4_flex_groups_free(struct ext4_sb_info *sbi) 1261 { 1262 struct flex_groups **flex_groups; 1263 int i; 1264 1265 rcu_read_lock(); 1266 flex_groups = rcu_dereference(sbi->s_flex_groups); 1267 if (flex_groups) { 1268 for (i = 0; i < sbi->s_flex_groups_allocated; i++) 1269 kvfree(flex_groups[i]); 1270 kvfree(flex_groups); 1271 } 1272 rcu_read_unlock(); 1273 } 1274 1275 static void ext4_put_super(struct super_block *sb) 1276 { 1277 struct ext4_sb_info *sbi = EXT4_SB(sb); 1278 struct ext4_super_block *es = sbi->s_es; 1279 int aborted = 0; 1280 int err; 1281 1282 /* 1283 * Unregister sysfs before destroying jbd2 journal. 1284 * Since we could still access attr_journal_task attribute via sysfs 1285 * path which could have sbi->s_journal->j_task as NULL 1286 * Unregister sysfs before flush sbi->s_sb_upd_work. 1287 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If 1288 * read metadata verify failed then will queue error work. 1289 * update_super_work will call start_this_handle may trigger 1290 * BUG_ON. 1291 */ 1292 ext4_unregister_sysfs(sb); 1293 1294 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount")) 1295 ext4_msg(sb, KERN_INFO, "unmounting filesystem %pU.", 1296 &sb->s_uuid); 1297 1298 ext4_unregister_li_request(sb); 1299 ext4_quotas_off(sb, EXT4_MAXQUOTAS); 1300 1301 destroy_workqueue(sbi->rsv_conversion_wq); 1302 ext4_release_orphan_info(sb); 1303 1304 if (sbi->s_journal) { 1305 aborted = is_journal_aborted(sbi->s_journal); 1306 err = ext4_journal_destroy(sbi, sbi->s_journal); 1307 if ((err < 0) && !aborted) { 1308 ext4_abort(sb, -err, "Couldn't clean up the journal"); 1309 } 1310 } else 1311 flush_work(&sbi->s_sb_upd_work); 1312 1313 ext4_es_unregister_shrinker(sbi); 1314 timer_shutdown_sync(&sbi->s_err_report); 1315 ext4_release_system_zone(sb); 1316 ext4_mb_release(sb); 1317 ext4_ext_release(sb); 1318 1319 if (!ext4_emergency_state(sb) && !sb_rdonly(sb)) { 1320 if (!aborted) { 1321 ext4_clear_feature_journal_needs_recovery(sb); 1322 ext4_clear_feature_orphan_present(sb); 1323 es->s_state = cpu_to_le16(sbi->s_mount_state); 1324 } 1325 ext4_commit_super(sb); 1326 } 1327 1328 ext4_group_desc_free(sbi); 1329 ext4_flex_groups_free(sbi); 1330 1331 WARN_ON_ONCE(!(sbi->s_mount_state & EXT4_ERROR_FS) && 1332 percpu_counter_sum(&sbi->s_dirtyclusters_counter)); 1333 ext4_percpu_param_destroy(sbi); 1334 #ifdef CONFIG_QUOTA 1335 for (int i = 0; i < EXT4_MAXQUOTAS; i++) 1336 kfree(get_qf_name(sb, sbi, i)); 1337 #endif 1338 1339 /* Debugging code just in case the in-memory inode orphan list 1340 * isn't empty. The on-disk one can be non-empty if we've 1341 * detected an error and taken the fs readonly, but the 1342 * in-memory list had better be clean by this point. */ 1343 if (!list_empty(&sbi->s_orphan)) 1344 dump_orphan_list(sb, sbi); 1345 ASSERT(list_empty(&sbi->s_orphan)); 1346 1347 sync_blockdev(sb->s_bdev); 1348 invalidate_bdev(sb->s_bdev); 1349 if (sbi->s_journal_bdev_file) { 1350 /* 1351 * Invalidate the journal device's buffers. We don't want them 1352 * floating about in memory - the physical journal device may 1353 * hotswapped, and it breaks the `ro-after' testing code. 1354 */ 1355 sync_blockdev(file_bdev(sbi->s_journal_bdev_file)); 1356 invalidate_bdev(file_bdev(sbi->s_journal_bdev_file)); 1357 } 1358 1359 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 1360 sbi->s_ea_inode_cache = NULL; 1361 1362 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 1363 sbi->s_ea_block_cache = NULL; 1364 1365 ext4_stop_mmpd(sbi); 1366 1367 brelse(sbi->s_sbh); 1368 sb->s_fs_info = NULL; 1369 /* 1370 * Now that we are completely done shutting down the 1371 * superblock, we need to actually destroy the kobject. 1372 */ 1373 kobject_put(&sbi->s_kobj); 1374 wait_for_completion(&sbi->s_kobj_unregister); 1375 kfree(sbi->s_blockgroup_lock); 1376 fs_put_dax(sbi->s_daxdev, NULL); 1377 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy); 1378 #if IS_ENABLED(CONFIG_UNICODE) 1379 utf8_unload(sb->s_encoding); 1380 #endif 1381 kfree(sbi); 1382 } 1383 1384 static struct kmem_cache *ext4_inode_cachep; 1385 1386 /* 1387 * Called inside transaction, so use GFP_NOFS 1388 */ 1389 static struct inode *ext4_alloc_inode(struct super_block *sb) 1390 { 1391 struct ext4_inode_info *ei; 1392 1393 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS); 1394 if (!ei) 1395 return NULL; 1396 1397 inode_set_iversion(&ei->vfs_inode, 1); 1398 ei->i_flags = 0; 1399 ext4_clear_state_flags(ei); /* Only relevant on 32-bit archs */ 1400 spin_lock_init(&ei->i_raw_lock); 1401 ei->i_prealloc_node = RB_ROOT; 1402 atomic_set(&ei->i_prealloc_active, 0); 1403 rwlock_init(&ei->i_prealloc_lock); 1404 ext4_es_init_tree(&ei->i_es_tree); 1405 rwlock_init(&ei->i_es_lock); 1406 INIT_LIST_HEAD(&ei->i_es_list); 1407 ei->i_es_all_nr = 0; 1408 ei->i_es_shk_nr = 0; 1409 ei->i_es_shrink_lblk = 0; 1410 ei->i_es_seq = 0; 1411 ei->i_reserved_data_blocks = 0; 1412 spin_lock_init(&(ei->i_block_reservation_lock)); 1413 ext4_init_pending_tree(&ei->i_pending_tree); 1414 #ifdef CONFIG_QUOTA 1415 ei->i_reserved_quota = 0; 1416 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); 1417 #endif 1418 ei->jinode = NULL; 1419 INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 1420 spin_lock_init(&ei->i_completed_io_lock); 1421 ei->i_sync_tid = 0; 1422 ei->i_datasync_tid = 0; 1423 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 1424 ext4_fc_init_inode(&ei->vfs_inode); 1425 spin_lock_init(&ei->i_fc_lock); 1426 return &ei->vfs_inode; 1427 } 1428 1429 static int ext4_drop_inode(struct inode *inode) 1430 { 1431 int drop = inode_generic_drop(inode); 1432 1433 if (!drop) 1434 drop = fscrypt_drop_inode(inode); 1435 1436 trace_ext4_drop_inode(inode, drop); 1437 return drop; 1438 } 1439 1440 static void ext4_free_in_core_inode(struct inode *inode) 1441 { 1442 fscrypt_free_inode(inode); 1443 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) { 1444 pr_warn("%s: inode %ld still in fc list", 1445 __func__, inode->i_ino); 1446 } 1447 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 1448 } 1449 1450 static void ext4_destroy_inode(struct inode *inode) 1451 { 1452 if (ext4_inode_orphan_tracked(inode)) { 1453 ext4_msg(inode->i_sb, KERN_ERR, 1454 "Inode %lu (%p): inode tracked as orphan!", 1455 inode->i_ino, EXT4_I(inode)); 1456 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 1457 EXT4_I(inode), sizeof(struct ext4_inode_info), 1458 true); 1459 dump_stack(); 1460 } 1461 1462 if (!(EXT4_SB(inode->i_sb)->s_mount_state & EXT4_ERROR_FS) && 1463 WARN_ON_ONCE(EXT4_I(inode)->i_reserved_data_blocks)) 1464 ext4_msg(inode->i_sb, KERN_ERR, 1465 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!", 1466 inode->i_ino, EXT4_I(inode), 1467 EXT4_I(inode)->i_reserved_data_blocks); 1468 } 1469 1470 static void ext4_shutdown(struct super_block *sb) 1471 { 1472 ext4_force_shutdown(sb, EXT4_GOING_FLAGS_NOLOGFLUSH); 1473 } 1474 1475 static void init_once(void *foo) 1476 { 1477 struct ext4_inode_info *ei = foo; 1478 1479 INIT_LIST_HEAD(&ei->i_orphan); 1480 init_rwsem(&ei->xattr_sem); 1481 init_rwsem(&ei->i_data_sem); 1482 inode_init_once(&ei->vfs_inode); 1483 ext4_fc_init_inode(&ei->vfs_inode); 1484 #ifdef CONFIG_FS_ENCRYPTION 1485 ei->i_crypt_info = NULL; 1486 #endif 1487 #ifdef CONFIG_FS_VERITY 1488 ei->i_verity_info = NULL; 1489 #endif 1490 } 1491 1492 static int __init init_inodecache(void) 1493 { 1494 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache", 1495 sizeof(struct ext4_inode_info), 0, 1496 SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT, 1497 offsetof(struct ext4_inode_info, i_data), 1498 sizeof_field(struct ext4_inode_info, i_data), 1499 init_once); 1500 if (ext4_inode_cachep == NULL) 1501 return -ENOMEM; 1502 return 0; 1503 } 1504 1505 static void destroy_inodecache(void) 1506 { 1507 /* 1508 * Make sure all delayed rcu free inodes are flushed before we 1509 * destroy cache. 1510 */ 1511 rcu_barrier(); 1512 kmem_cache_destroy(ext4_inode_cachep); 1513 } 1514 1515 void ext4_clear_inode(struct inode *inode) 1516 { 1517 ext4_fc_del(inode); 1518 invalidate_inode_buffers(inode); 1519 clear_inode(inode); 1520 ext4_discard_preallocations(inode); 1521 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 1522 dquot_drop(inode); 1523 if (EXT4_I(inode)->jinode) { 1524 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 1525 EXT4_I(inode)->jinode); 1526 jbd2_free_inode(EXT4_I(inode)->jinode); 1527 EXT4_I(inode)->jinode = NULL; 1528 } 1529 fscrypt_put_encryption_info(inode); 1530 fsverity_cleanup_inode(inode); 1531 } 1532 1533 static struct inode *ext4_nfs_get_inode(struct super_block *sb, 1534 u64 ino, u32 generation) 1535 { 1536 struct inode *inode; 1537 1538 /* 1539 * Currently we don't know the generation for parent directory, so 1540 * a generation of 0 means "accept any" 1541 */ 1542 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE); 1543 if (IS_ERR(inode)) 1544 return ERR_CAST(inode); 1545 if (generation && inode->i_generation != generation) { 1546 iput(inode); 1547 return ERR_PTR(-ESTALE); 1548 } 1549 1550 return inode; 1551 } 1552 1553 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 1554 int fh_len, int fh_type) 1555 { 1556 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1557 ext4_nfs_get_inode); 1558 } 1559 1560 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 1561 int fh_len, int fh_type) 1562 { 1563 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1564 ext4_nfs_get_inode); 1565 } 1566 1567 static int ext4_nfs_commit_metadata(struct inode *inode) 1568 { 1569 struct writeback_control wbc = { 1570 .sync_mode = WB_SYNC_ALL 1571 }; 1572 1573 trace_ext4_nfs_commit_metadata(inode); 1574 return ext4_write_inode(inode, &wbc); 1575 } 1576 1577 #ifdef CONFIG_QUOTA 1578 static const char * const quotatypes[] = INITQFNAMES; 1579 #define QTYPE2NAME(t) (quotatypes[t]) 1580 1581 static int ext4_write_dquot(struct dquot *dquot); 1582 static int ext4_acquire_dquot(struct dquot *dquot); 1583 static int ext4_release_dquot(struct dquot *dquot); 1584 static int ext4_mark_dquot_dirty(struct dquot *dquot); 1585 static int ext4_write_info(struct super_block *sb, int type); 1586 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1587 const struct path *path); 1588 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1589 size_t len, loff_t off); 1590 static ssize_t ext4_quota_write(struct super_block *sb, int type, 1591 const char *data, size_t len, loff_t off); 1592 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 1593 unsigned int flags); 1594 1595 static struct dquot __rcu **ext4_get_dquots(struct inode *inode) 1596 { 1597 return EXT4_I(inode)->i_dquot; 1598 } 1599 1600 static const struct dquot_operations ext4_quota_operations = { 1601 .get_reserved_space = ext4_get_reserved_space, 1602 .write_dquot = ext4_write_dquot, 1603 .acquire_dquot = ext4_acquire_dquot, 1604 .release_dquot = ext4_release_dquot, 1605 .mark_dirty = ext4_mark_dquot_dirty, 1606 .write_info = ext4_write_info, 1607 .alloc_dquot = dquot_alloc, 1608 .destroy_dquot = dquot_destroy, 1609 .get_projid = ext4_get_projid, 1610 .get_inode_usage = ext4_get_inode_usage, 1611 .get_next_id = dquot_get_next_id, 1612 }; 1613 1614 static const struct quotactl_ops ext4_qctl_operations = { 1615 .quota_on = ext4_quota_on, 1616 .quota_off = ext4_quota_off, 1617 .quota_sync = dquot_quota_sync, 1618 .get_state = dquot_get_state, 1619 .set_info = dquot_set_dqinfo, 1620 .get_dqblk = dquot_get_dqblk, 1621 .set_dqblk = dquot_set_dqblk, 1622 .get_nextdqblk = dquot_get_next_dqblk, 1623 }; 1624 #endif 1625 1626 static const struct super_operations ext4_sops = { 1627 .alloc_inode = ext4_alloc_inode, 1628 .free_inode = ext4_free_in_core_inode, 1629 .destroy_inode = ext4_destroy_inode, 1630 .write_inode = ext4_write_inode, 1631 .dirty_inode = ext4_dirty_inode, 1632 .drop_inode = ext4_drop_inode, 1633 .evict_inode = ext4_evict_inode, 1634 .put_super = ext4_put_super, 1635 .sync_fs = ext4_sync_fs, 1636 .freeze_fs = ext4_freeze, 1637 .unfreeze_fs = ext4_unfreeze, 1638 .statfs = ext4_statfs, 1639 .show_options = ext4_show_options, 1640 .shutdown = ext4_shutdown, 1641 #ifdef CONFIG_QUOTA 1642 .quota_read = ext4_quota_read, 1643 .quota_write = ext4_quota_write, 1644 .get_dquots = ext4_get_dquots, 1645 #endif 1646 }; 1647 1648 static const struct export_operations ext4_export_ops = { 1649 .encode_fh = generic_encode_ino32_fh, 1650 .fh_to_dentry = ext4_fh_to_dentry, 1651 .fh_to_parent = ext4_fh_to_parent, 1652 .get_parent = ext4_get_parent, 1653 .commit_metadata = ext4_nfs_commit_metadata, 1654 }; 1655 1656 enum { 1657 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1658 Opt_resgid, Opt_resuid, Opt_sb, 1659 Opt_nouid32, Opt_debug, Opt_removed, 1660 Opt_user_xattr, Opt_acl, 1661 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1662 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1663 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1664 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1665 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 1666 Opt_inlinecrypt, 1667 Opt_usrjquota, Opt_grpjquota, Opt_quota, 1668 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1669 Opt_usrquota, Opt_grpquota, Opt_prjquota, 1670 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never, 1671 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error, 1672 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize, 1673 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 1674 Opt_inode_readahead_blks, Opt_journal_ioprio, 1675 Opt_dioread_nolock, Opt_dioread_lock, 1676 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1677 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, 1678 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan, 1679 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type, 1680 #ifdef CONFIG_EXT4_DEBUG 1681 Opt_fc_debug_max_replay, Opt_fc_debug_force 1682 #endif 1683 }; 1684 1685 static const struct constant_table ext4_param_errors[] = { 1686 {"continue", EXT4_MOUNT_ERRORS_CONT}, 1687 {"panic", EXT4_MOUNT_ERRORS_PANIC}, 1688 {"remount-ro", EXT4_MOUNT_ERRORS_RO}, 1689 {} 1690 }; 1691 1692 static const struct constant_table ext4_param_data[] = { 1693 {"journal", EXT4_MOUNT_JOURNAL_DATA}, 1694 {"ordered", EXT4_MOUNT_ORDERED_DATA}, 1695 {"writeback", EXT4_MOUNT_WRITEBACK_DATA}, 1696 {} 1697 }; 1698 1699 static const struct constant_table ext4_param_data_err[] = { 1700 {"abort", Opt_data_err_abort}, 1701 {"ignore", Opt_data_err_ignore}, 1702 {} 1703 }; 1704 1705 static const struct constant_table ext4_param_jqfmt[] = { 1706 {"vfsold", QFMT_VFS_OLD}, 1707 {"vfsv0", QFMT_VFS_V0}, 1708 {"vfsv1", QFMT_VFS_V1}, 1709 {} 1710 }; 1711 1712 static const struct constant_table ext4_param_dax[] = { 1713 {"always", Opt_dax_always}, 1714 {"inode", Opt_dax_inode}, 1715 {"never", Opt_dax_never}, 1716 {} 1717 }; 1718 1719 /* 1720 * Mount option specification 1721 * We don't use fsparam_flag_no because of the way we set the 1722 * options and the way we show them in _ext4_show_options(). To 1723 * keep the changes to a minimum, let's keep the negative options 1724 * separate for now. 1725 */ 1726 static const struct fs_parameter_spec ext4_param_specs[] = { 1727 fsparam_flag ("bsddf", Opt_bsd_df), 1728 fsparam_flag ("minixdf", Opt_minix_df), 1729 fsparam_flag ("grpid", Opt_grpid), 1730 fsparam_flag ("bsdgroups", Opt_grpid), 1731 fsparam_flag ("nogrpid", Opt_nogrpid), 1732 fsparam_flag ("sysvgroups", Opt_nogrpid), 1733 fsparam_gid ("resgid", Opt_resgid), 1734 fsparam_uid ("resuid", Opt_resuid), 1735 fsparam_u32 ("sb", Opt_sb), 1736 fsparam_enum ("errors", Opt_errors, ext4_param_errors), 1737 fsparam_flag ("nouid32", Opt_nouid32), 1738 fsparam_flag ("debug", Opt_debug), 1739 fsparam_flag ("oldalloc", Opt_removed), 1740 fsparam_flag ("orlov", Opt_removed), 1741 fsparam_flag ("user_xattr", Opt_user_xattr), 1742 fsparam_flag ("acl", Opt_acl), 1743 fsparam_flag ("norecovery", Opt_noload), 1744 fsparam_flag ("noload", Opt_noload), 1745 fsparam_flag ("bh", Opt_removed), 1746 fsparam_flag ("nobh", Opt_removed), 1747 fsparam_u32 ("commit", Opt_commit), 1748 fsparam_u32 ("min_batch_time", Opt_min_batch_time), 1749 fsparam_u32 ("max_batch_time", Opt_max_batch_time), 1750 fsparam_u32 ("journal_dev", Opt_journal_dev), 1751 fsparam_bdev ("journal_path", Opt_journal_path), 1752 fsparam_flag ("journal_checksum", Opt_journal_checksum), 1753 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum), 1754 fsparam_flag ("journal_async_commit",Opt_journal_async_commit), 1755 fsparam_flag ("abort", Opt_abort), 1756 fsparam_enum ("data", Opt_data, ext4_param_data), 1757 fsparam_enum ("data_err", Opt_data_err, 1758 ext4_param_data_err), 1759 fsparam_string_empty 1760 ("usrjquota", Opt_usrjquota), 1761 fsparam_string_empty 1762 ("grpjquota", Opt_grpjquota), 1763 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt), 1764 fsparam_flag ("grpquota", Opt_grpquota), 1765 fsparam_flag ("quota", Opt_quota), 1766 fsparam_flag ("noquota", Opt_noquota), 1767 fsparam_flag ("usrquota", Opt_usrquota), 1768 fsparam_flag ("prjquota", Opt_prjquota), 1769 fsparam_flag ("barrier", Opt_barrier), 1770 fsparam_u32 ("barrier", Opt_barrier), 1771 fsparam_flag ("nobarrier", Opt_nobarrier), 1772 fsparam_flag ("i_version", Opt_removed), 1773 fsparam_flag ("dax", Opt_dax), 1774 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax), 1775 fsparam_u32 ("stripe", Opt_stripe), 1776 fsparam_flag ("delalloc", Opt_delalloc), 1777 fsparam_flag ("nodelalloc", Opt_nodelalloc), 1778 fsparam_flag ("warn_on_error", Opt_warn_on_error), 1779 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error), 1780 fsparam_u32 ("debug_want_extra_isize", 1781 Opt_debug_want_extra_isize), 1782 fsparam_flag ("mblk_io_submit", Opt_removed), 1783 fsparam_flag ("nomblk_io_submit", Opt_removed), 1784 fsparam_flag ("block_validity", Opt_block_validity), 1785 fsparam_flag ("noblock_validity", Opt_noblock_validity), 1786 fsparam_u32 ("inode_readahead_blks", 1787 Opt_inode_readahead_blks), 1788 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio), 1789 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc), 1790 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc), 1791 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc), 1792 fsparam_flag ("dioread_nolock", Opt_dioread_nolock), 1793 fsparam_flag ("nodioread_nolock", Opt_dioread_lock), 1794 fsparam_flag ("dioread_lock", Opt_dioread_lock), 1795 fsparam_flag ("discard", Opt_discard), 1796 fsparam_flag ("nodiscard", Opt_nodiscard), 1797 fsparam_u32 ("init_itable", Opt_init_itable), 1798 fsparam_flag ("init_itable", Opt_init_itable), 1799 fsparam_flag ("noinit_itable", Opt_noinit_itable), 1800 #ifdef CONFIG_EXT4_DEBUG 1801 fsparam_flag ("fc_debug_force", Opt_fc_debug_force), 1802 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay), 1803 #endif 1804 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb), 1805 fsparam_flag ("test_dummy_encryption", 1806 Opt_test_dummy_encryption), 1807 fsparam_string ("test_dummy_encryption", 1808 Opt_test_dummy_encryption), 1809 fsparam_flag ("inlinecrypt", Opt_inlinecrypt), 1810 fsparam_flag ("nombcache", Opt_nombcache), 1811 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */ 1812 fsparam_flag ("prefetch_block_bitmaps", 1813 Opt_removed), 1814 fsparam_flag ("no_prefetch_block_bitmaps", 1815 Opt_no_prefetch_block_bitmaps), 1816 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan), 1817 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */ 1818 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */ 1819 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */ 1820 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */ 1821 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */ 1822 {} 1823 }; 1824 1825 1826 #define MOPT_SET 0x0001 1827 #define MOPT_CLEAR 0x0002 1828 #define MOPT_NOSUPPORT 0x0004 1829 #define MOPT_EXPLICIT 0x0008 1830 #ifdef CONFIG_QUOTA 1831 #define MOPT_Q 0 1832 #define MOPT_QFMT 0x0010 1833 #else 1834 #define MOPT_Q MOPT_NOSUPPORT 1835 #define MOPT_QFMT MOPT_NOSUPPORT 1836 #endif 1837 #define MOPT_NO_EXT2 0x0020 1838 #define MOPT_NO_EXT3 0x0040 1839 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1840 #define MOPT_SKIP 0x0080 1841 #define MOPT_2 0x0100 1842 1843 static const struct mount_opts { 1844 int token; 1845 int mount_opt; 1846 int flags; 1847 } ext4_mount_opts[] = { 1848 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 1849 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 1850 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 1851 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 1852 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 1853 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 1854 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 1855 MOPT_EXT4_ONLY | MOPT_SET}, 1856 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 1857 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1858 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 1859 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 1860 {Opt_delalloc, EXT4_MOUNT_DELALLOC, 1861 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1862 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 1863 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1864 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET}, 1865 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR}, 1866 {Opt_commit, 0, MOPT_NO_EXT2}, 1867 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1868 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1869 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1870 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1871 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 1872 EXT4_MOUNT_JOURNAL_CHECKSUM), 1873 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1874 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 1875 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2}, 1876 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 1877 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 1878 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 1879 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 1880 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 1881 {Opt_dax_type, 0, MOPT_EXT4_ONLY}, 1882 {Opt_journal_dev, 0, MOPT_NO_EXT2}, 1883 {Opt_journal_path, 0, MOPT_NO_EXT2}, 1884 {Opt_journal_ioprio, 0, MOPT_NO_EXT2}, 1885 {Opt_data, 0, MOPT_NO_EXT2}, 1886 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 1887 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1888 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 1889 #else 1890 {Opt_acl, 0, MOPT_NOSUPPORT}, 1891 #endif 1892 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 1893 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1894 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 1895 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 1896 MOPT_SET | MOPT_Q}, 1897 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 1898 MOPT_SET | MOPT_Q}, 1899 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, 1900 MOPT_SET | MOPT_Q}, 1901 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 1902 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), 1903 MOPT_CLEAR | MOPT_Q}, 1904 {Opt_usrjquota, 0, MOPT_Q}, 1905 {Opt_grpjquota, 0, MOPT_Q}, 1906 {Opt_jqfmt, 0, MOPT_QFMT}, 1907 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, 1908 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS, 1909 MOPT_SET}, 1910 #ifdef CONFIG_EXT4_DEBUG 1911 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT, 1912 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY}, 1913 #endif 1914 {Opt_abort, EXT4_MOUNT2_ABORT, MOPT_SET | MOPT_2}, 1915 {Opt_err, 0, 0} 1916 }; 1917 1918 #if IS_ENABLED(CONFIG_UNICODE) 1919 static const struct ext4_sb_encodings { 1920 __u16 magic; 1921 char *name; 1922 unsigned int version; 1923 } ext4_sb_encoding_map[] = { 1924 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)}, 1925 }; 1926 1927 static const struct ext4_sb_encodings * 1928 ext4_sb_read_encoding(const struct ext4_super_block *es) 1929 { 1930 __u16 magic = le16_to_cpu(es->s_encoding); 1931 int i; 1932 1933 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++) 1934 if (magic == ext4_sb_encoding_map[i].magic) 1935 return &ext4_sb_encoding_map[i]; 1936 1937 return NULL; 1938 } 1939 #endif 1940 1941 #define EXT4_SPEC_JQUOTA (1 << 0) 1942 #define EXT4_SPEC_JQFMT (1 << 1) 1943 #define EXT4_SPEC_DATAJ (1 << 2) 1944 #define EXT4_SPEC_SB_BLOCK (1 << 3) 1945 #define EXT4_SPEC_JOURNAL_DEV (1 << 4) 1946 #define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5) 1947 #define EXT4_SPEC_s_want_extra_isize (1 << 7) 1948 #define EXT4_SPEC_s_max_batch_time (1 << 8) 1949 #define EXT4_SPEC_s_min_batch_time (1 << 9) 1950 #define EXT4_SPEC_s_inode_readahead_blks (1 << 10) 1951 #define EXT4_SPEC_s_li_wait_mult (1 << 11) 1952 #define EXT4_SPEC_s_max_dir_size_kb (1 << 12) 1953 #define EXT4_SPEC_s_stripe (1 << 13) 1954 #define EXT4_SPEC_s_resuid (1 << 14) 1955 #define EXT4_SPEC_s_resgid (1 << 15) 1956 #define EXT4_SPEC_s_commit_interval (1 << 16) 1957 #define EXT4_SPEC_s_fc_debug_max_replay (1 << 17) 1958 #define EXT4_SPEC_s_sb_block (1 << 18) 1959 #define EXT4_SPEC_mb_optimize_scan (1 << 19) 1960 1961 struct ext4_fs_context { 1962 char *s_qf_names[EXT4_MAXQUOTAS]; 1963 struct fscrypt_dummy_policy dummy_enc_policy; 1964 int s_jquota_fmt; /* Format of quota to use */ 1965 #ifdef CONFIG_EXT4_DEBUG 1966 int s_fc_debug_max_replay; 1967 #endif 1968 unsigned short qname_spec; 1969 unsigned long vals_s_flags; /* Bits to set in s_flags */ 1970 unsigned long mask_s_flags; /* Bits changed in s_flags */ 1971 unsigned long journal_devnum; 1972 unsigned long s_commit_interval; 1973 unsigned long s_stripe; 1974 unsigned int s_inode_readahead_blks; 1975 unsigned int s_want_extra_isize; 1976 unsigned int s_li_wait_mult; 1977 unsigned int s_max_dir_size_kb; 1978 unsigned int journal_ioprio; 1979 unsigned int vals_s_mount_opt; 1980 unsigned int mask_s_mount_opt; 1981 unsigned int vals_s_mount_opt2; 1982 unsigned int mask_s_mount_opt2; 1983 unsigned int opt_flags; /* MOPT flags */ 1984 unsigned int spec; 1985 u32 s_max_batch_time; 1986 u32 s_min_batch_time; 1987 kuid_t s_resuid; 1988 kgid_t s_resgid; 1989 ext4_fsblk_t s_sb_block; 1990 }; 1991 1992 static void ext4_fc_free(struct fs_context *fc) 1993 { 1994 struct ext4_fs_context *ctx = fc->fs_private; 1995 int i; 1996 1997 if (!ctx) 1998 return; 1999 2000 for (i = 0; i < EXT4_MAXQUOTAS; i++) 2001 kfree(ctx->s_qf_names[i]); 2002 2003 fscrypt_free_dummy_policy(&ctx->dummy_enc_policy); 2004 kfree(ctx); 2005 } 2006 2007 int ext4_init_fs_context(struct fs_context *fc) 2008 { 2009 struct ext4_fs_context *ctx; 2010 2011 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL); 2012 if (!ctx) 2013 return -ENOMEM; 2014 2015 fc->fs_private = ctx; 2016 fc->ops = &ext4_context_ops; 2017 2018 /* i_version is always enabled now */ 2019 fc->sb_flags |= SB_I_VERSION; 2020 2021 return 0; 2022 } 2023 2024 #ifdef CONFIG_QUOTA 2025 /* 2026 * Note the name of the specified quota file. 2027 */ 2028 static int note_qf_name(struct fs_context *fc, int qtype, 2029 struct fs_parameter *param) 2030 { 2031 struct ext4_fs_context *ctx = fc->fs_private; 2032 char *qname; 2033 2034 if (param->size < 1) { 2035 ext4_msg(NULL, KERN_ERR, "Missing quota name"); 2036 return -EINVAL; 2037 } 2038 if (strchr(param->string, '/')) { 2039 ext4_msg(NULL, KERN_ERR, 2040 "quotafile must be on filesystem root"); 2041 return -EINVAL; 2042 } 2043 if (ctx->s_qf_names[qtype]) { 2044 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) { 2045 ext4_msg(NULL, KERN_ERR, 2046 "%s quota file already specified", 2047 QTYPE2NAME(qtype)); 2048 return -EINVAL; 2049 } 2050 return 0; 2051 } 2052 2053 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL); 2054 if (!qname) { 2055 ext4_msg(NULL, KERN_ERR, 2056 "Not enough memory for storing quotafile name"); 2057 return -ENOMEM; 2058 } 2059 ctx->s_qf_names[qtype] = qname; 2060 ctx->qname_spec |= 1 << qtype; 2061 ctx->spec |= EXT4_SPEC_JQUOTA; 2062 return 0; 2063 } 2064 2065 /* 2066 * Clear the name of the specified quota file. 2067 */ 2068 static int unnote_qf_name(struct fs_context *fc, int qtype) 2069 { 2070 struct ext4_fs_context *ctx = fc->fs_private; 2071 2072 kfree(ctx->s_qf_names[qtype]); 2073 2074 ctx->s_qf_names[qtype] = NULL; 2075 ctx->qname_spec |= 1 << qtype; 2076 ctx->spec |= EXT4_SPEC_JQUOTA; 2077 return 0; 2078 } 2079 #endif 2080 2081 static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param, 2082 struct ext4_fs_context *ctx) 2083 { 2084 int err; 2085 2086 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) { 2087 ext4_msg(NULL, KERN_WARNING, 2088 "test_dummy_encryption option not supported"); 2089 return -EINVAL; 2090 } 2091 err = fscrypt_parse_test_dummy_encryption(param, 2092 &ctx->dummy_enc_policy); 2093 if (err == -EINVAL) { 2094 ext4_msg(NULL, KERN_WARNING, 2095 "Value of option \"%s\" is unrecognized", param->key); 2096 } else if (err == -EEXIST) { 2097 ext4_msg(NULL, KERN_WARNING, 2098 "Conflicting test_dummy_encryption options"); 2099 return -EINVAL; 2100 } 2101 return err; 2102 } 2103 2104 #define EXT4_SET_CTX(name) \ 2105 static inline __maybe_unused \ 2106 void ctx_set_##name(struct ext4_fs_context *ctx, unsigned long flag) \ 2107 { \ 2108 ctx->mask_s_##name |= flag; \ 2109 ctx->vals_s_##name |= flag; \ 2110 } 2111 2112 #define EXT4_CLEAR_CTX(name) \ 2113 static inline __maybe_unused \ 2114 void ctx_clear_##name(struct ext4_fs_context *ctx, unsigned long flag) \ 2115 { \ 2116 ctx->mask_s_##name |= flag; \ 2117 ctx->vals_s_##name &= ~flag; \ 2118 } 2119 2120 #define EXT4_TEST_CTX(name) \ 2121 static inline unsigned long \ 2122 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \ 2123 { \ 2124 return (ctx->vals_s_##name & flag); \ 2125 } 2126 2127 EXT4_SET_CTX(flags); /* set only */ 2128 EXT4_SET_CTX(mount_opt); 2129 EXT4_CLEAR_CTX(mount_opt); 2130 EXT4_TEST_CTX(mount_opt); 2131 EXT4_SET_CTX(mount_opt2); 2132 EXT4_CLEAR_CTX(mount_opt2); 2133 EXT4_TEST_CTX(mount_opt2); 2134 2135 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param) 2136 { 2137 struct ext4_fs_context *ctx = fc->fs_private; 2138 struct fs_parse_result result; 2139 const struct mount_opts *m; 2140 int is_remount; 2141 int token; 2142 2143 token = fs_parse(fc, ext4_param_specs, param, &result); 2144 if (token < 0) 2145 return token; 2146 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE; 2147 2148 for (m = ext4_mount_opts; m->token != Opt_err; m++) 2149 if (token == m->token) 2150 break; 2151 2152 ctx->opt_flags |= m->flags; 2153 2154 if (m->flags & MOPT_EXPLICIT) { 2155 if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 2156 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC); 2157 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 2158 ctx_set_mount_opt2(ctx, 2159 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM); 2160 } else 2161 return -EINVAL; 2162 } 2163 2164 if (m->flags & MOPT_NOSUPPORT) { 2165 ext4_msg(NULL, KERN_ERR, "%s option not supported", 2166 param->key); 2167 return 0; 2168 } 2169 2170 switch (token) { 2171 #ifdef CONFIG_QUOTA 2172 case Opt_usrjquota: 2173 if (!*param->string) 2174 return unnote_qf_name(fc, USRQUOTA); 2175 else 2176 return note_qf_name(fc, USRQUOTA, param); 2177 case Opt_grpjquota: 2178 if (!*param->string) 2179 return unnote_qf_name(fc, GRPQUOTA); 2180 else 2181 return note_qf_name(fc, GRPQUOTA, param); 2182 #endif 2183 case Opt_sb: 2184 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) { 2185 ext4_msg(NULL, KERN_WARNING, 2186 "Ignoring %s option on remount", param->key); 2187 } else { 2188 ctx->s_sb_block = result.uint_32; 2189 ctx->spec |= EXT4_SPEC_s_sb_block; 2190 } 2191 return 0; 2192 case Opt_removed: 2193 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option", 2194 param->key); 2195 return 0; 2196 case Opt_inlinecrypt: 2197 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 2198 ctx_set_flags(ctx, SB_INLINECRYPT); 2199 #else 2200 ext4_msg(NULL, KERN_ERR, "inline encryption not supported"); 2201 #endif 2202 return 0; 2203 case Opt_errors: 2204 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK); 2205 ctx_set_mount_opt(ctx, result.uint_32); 2206 return 0; 2207 #ifdef CONFIG_QUOTA 2208 case Opt_jqfmt: 2209 ctx->s_jquota_fmt = result.uint_32; 2210 ctx->spec |= EXT4_SPEC_JQFMT; 2211 return 0; 2212 #endif 2213 case Opt_data: 2214 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS); 2215 ctx_set_mount_opt(ctx, result.uint_32); 2216 ctx->spec |= EXT4_SPEC_DATAJ; 2217 return 0; 2218 case Opt_commit: 2219 if (result.uint_32 == 0) 2220 result.uint_32 = JBD2_DEFAULT_MAX_COMMIT_AGE; 2221 else if (result.uint_32 > INT_MAX / HZ) { 2222 ext4_msg(NULL, KERN_ERR, 2223 "Invalid commit interval %d, " 2224 "must be smaller than %d", 2225 result.uint_32, INT_MAX / HZ); 2226 return -EINVAL; 2227 } 2228 ctx->s_commit_interval = HZ * result.uint_32; 2229 ctx->spec |= EXT4_SPEC_s_commit_interval; 2230 return 0; 2231 case Opt_debug_want_extra_isize: 2232 if ((result.uint_32 & 1) || (result.uint_32 < 4)) { 2233 ext4_msg(NULL, KERN_ERR, 2234 "Invalid want_extra_isize %d", result.uint_32); 2235 return -EINVAL; 2236 } 2237 ctx->s_want_extra_isize = result.uint_32; 2238 ctx->spec |= EXT4_SPEC_s_want_extra_isize; 2239 return 0; 2240 case Opt_max_batch_time: 2241 ctx->s_max_batch_time = result.uint_32; 2242 ctx->spec |= EXT4_SPEC_s_max_batch_time; 2243 return 0; 2244 case Opt_min_batch_time: 2245 ctx->s_min_batch_time = result.uint_32; 2246 ctx->spec |= EXT4_SPEC_s_min_batch_time; 2247 return 0; 2248 case Opt_inode_readahead_blks: 2249 if (result.uint_32 && 2250 (result.uint_32 > (1 << 30) || 2251 !is_power_of_2(result.uint_32))) { 2252 ext4_msg(NULL, KERN_ERR, 2253 "EXT4-fs: inode_readahead_blks must be " 2254 "0 or a power of 2 smaller than 2^31"); 2255 return -EINVAL; 2256 } 2257 ctx->s_inode_readahead_blks = result.uint_32; 2258 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks; 2259 return 0; 2260 case Opt_init_itable: 2261 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE); 2262 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 2263 if (param->type == fs_value_is_string) 2264 ctx->s_li_wait_mult = result.uint_32; 2265 ctx->spec |= EXT4_SPEC_s_li_wait_mult; 2266 return 0; 2267 case Opt_max_dir_size_kb: 2268 ctx->s_max_dir_size_kb = result.uint_32; 2269 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb; 2270 return 0; 2271 #ifdef CONFIG_EXT4_DEBUG 2272 case Opt_fc_debug_max_replay: 2273 ctx->s_fc_debug_max_replay = result.uint_32; 2274 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay; 2275 return 0; 2276 #endif 2277 case Opt_stripe: 2278 ctx->s_stripe = result.uint_32; 2279 ctx->spec |= EXT4_SPEC_s_stripe; 2280 return 0; 2281 case Opt_resuid: 2282 ctx->s_resuid = result.uid; 2283 ctx->spec |= EXT4_SPEC_s_resuid; 2284 return 0; 2285 case Opt_resgid: 2286 ctx->s_resgid = result.gid; 2287 ctx->spec |= EXT4_SPEC_s_resgid; 2288 return 0; 2289 case Opt_journal_dev: 2290 if (is_remount) { 2291 ext4_msg(NULL, KERN_ERR, 2292 "Cannot specify journal on remount"); 2293 return -EINVAL; 2294 } 2295 ctx->journal_devnum = result.uint_32; 2296 ctx->spec |= EXT4_SPEC_JOURNAL_DEV; 2297 return 0; 2298 case Opt_journal_path: 2299 { 2300 struct inode *journal_inode; 2301 struct path path; 2302 int error; 2303 2304 if (is_remount) { 2305 ext4_msg(NULL, KERN_ERR, 2306 "Cannot specify journal on remount"); 2307 return -EINVAL; 2308 } 2309 2310 error = fs_lookup_param(fc, param, 1, LOOKUP_FOLLOW, &path); 2311 if (error) { 2312 ext4_msg(NULL, KERN_ERR, "error: could not find " 2313 "journal device path"); 2314 return -EINVAL; 2315 } 2316 2317 journal_inode = d_inode(path.dentry); 2318 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev); 2319 ctx->spec |= EXT4_SPEC_JOURNAL_DEV; 2320 path_put(&path); 2321 return 0; 2322 } 2323 case Opt_journal_ioprio: 2324 if (result.uint_32 > 7) { 2325 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority" 2326 " (must be 0-7)"); 2327 return -EINVAL; 2328 } 2329 ctx->journal_ioprio = 2330 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32); 2331 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO; 2332 return 0; 2333 case Opt_test_dummy_encryption: 2334 return ext4_parse_test_dummy_encryption(param, ctx); 2335 case Opt_dax: 2336 case Opt_dax_type: 2337 #ifdef CONFIG_FS_DAX 2338 { 2339 int type = (token == Opt_dax) ? 2340 Opt_dax : result.uint_32; 2341 2342 switch (type) { 2343 case Opt_dax: 2344 case Opt_dax_always: 2345 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); 2346 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); 2347 break; 2348 case Opt_dax_never: 2349 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); 2350 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); 2351 break; 2352 case Opt_dax_inode: 2353 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS); 2354 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER); 2355 /* Strictly for printing options */ 2356 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE); 2357 break; 2358 } 2359 return 0; 2360 } 2361 #else 2362 ext4_msg(NULL, KERN_INFO, "dax option not supported"); 2363 return -EINVAL; 2364 #endif 2365 case Opt_data_err: 2366 if (result.uint_32 == Opt_data_err_abort) 2367 ctx_set_mount_opt(ctx, m->mount_opt); 2368 else if (result.uint_32 == Opt_data_err_ignore) 2369 ctx_clear_mount_opt(ctx, m->mount_opt); 2370 return 0; 2371 case Opt_mb_optimize_scan: 2372 if (result.int_32 == 1) { 2373 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN); 2374 ctx->spec |= EXT4_SPEC_mb_optimize_scan; 2375 } else if (result.int_32 == 0) { 2376 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN); 2377 ctx->spec |= EXT4_SPEC_mb_optimize_scan; 2378 } else { 2379 ext4_msg(NULL, KERN_WARNING, 2380 "mb_optimize_scan should be set to 0 or 1."); 2381 return -EINVAL; 2382 } 2383 return 0; 2384 } 2385 2386 /* 2387 * At this point we should only be getting options requiring MOPT_SET, 2388 * or MOPT_CLEAR. Anything else is a bug 2389 */ 2390 if (m->token == Opt_err) { 2391 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s", 2392 param->key); 2393 WARN_ON(1); 2394 return -EINVAL; 2395 } 2396 2397 else { 2398 unsigned int set = 0; 2399 2400 if ((param->type == fs_value_is_flag) || 2401 result.uint_32 > 0) 2402 set = 1; 2403 2404 if (m->flags & MOPT_CLEAR) 2405 set = !set; 2406 else if (unlikely(!(m->flags & MOPT_SET))) { 2407 ext4_msg(NULL, KERN_WARNING, 2408 "buggy handling of option %s", 2409 param->key); 2410 WARN_ON(1); 2411 return -EINVAL; 2412 } 2413 if (m->flags & MOPT_2) { 2414 if (set != 0) 2415 ctx_set_mount_opt2(ctx, m->mount_opt); 2416 else 2417 ctx_clear_mount_opt2(ctx, m->mount_opt); 2418 } else { 2419 if (set != 0) 2420 ctx_set_mount_opt(ctx, m->mount_opt); 2421 else 2422 ctx_clear_mount_opt(ctx, m->mount_opt); 2423 } 2424 } 2425 2426 return 0; 2427 } 2428 2429 static int parse_options(struct fs_context *fc, char *options) 2430 { 2431 struct fs_parameter param; 2432 int ret; 2433 char *key; 2434 2435 if (!options) 2436 return 0; 2437 2438 while ((key = strsep(&options, ",")) != NULL) { 2439 if (*key) { 2440 size_t v_len = 0; 2441 char *value = strchr(key, '='); 2442 2443 param.type = fs_value_is_flag; 2444 param.string = NULL; 2445 2446 if (value) { 2447 if (value == key) 2448 continue; 2449 2450 *value++ = 0; 2451 v_len = strlen(value); 2452 param.string = kmemdup_nul(value, v_len, 2453 GFP_KERNEL); 2454 if (!param.string) 2455 return -ENOMEM; 2456 param.type = fs_value_is_string; 2457 } 2458 2459 param.key = key; 2460 param.size = v_len; 2461 2462 ret = ext4_parse_param(fc, ¶m); 2463 kfree(param.string); 2464 if (ret < 0) 2465 return ret; 2466 } 2467 } 2468 2469 ret = ext4_validate_options(fc); 2470 if (ret < 0) 2471 return ret; 2472 2473 return 0; 2474 } 2475 2476 static int parse_apply_sb_mount_options(struct super_block *sb, 2477 struct ext4_fs_context *m_ctx) 2478 { 2479 struct ext4_sb_info *sbi = EXT4_SB(sb); 2480 char s_mount_opts[64]; 2481 struct ext4_fs_context *s_ctx = NULL; 2482 struct fs_context *fc = NULL; 2483 int ret = -ENOMEM; 2484 2485 if (!sbi->s_es->s_mount_opts[0]) 2486 return 0; 2487 2488 if (strscpy_pad(s_mount_opts, sbi->s_es->s_mount_opts) < 0) 2489 return -E2BIG; 2490 2491 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL); 2492 if (!fc) 2493 return -ENOMEM; 2494 2495 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL); 2496 if (!s_ctx) 2497 goto out_free; 2498 2499 fc->fs_private = s_ctx; 2500 fc->s_fs_info = sbi; 2501 2502 ret = parse_options(fc, s_mount_opts); 2503 if (ret < 0) 2504 goto parse_failed; 2505 2506 ret = ext4_check_opt_consistency(fc, sb); 2507 if (ret < 0) { 2508 parse_failed: 2509 ext4_msg(sb, KERN_WARNING, 2510 "failed to parse options in superblock: %s", 2511 s_mount_opts); 2512 ret = 0; 2513 goto out_free; 2514 } 2515 2516 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV) 2517 m_ctx->journal_devnum = s_ctx->journal_devnum; 2518 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO) 2519 m_ctx->journal_ioprio = s_ctx->journal_ioprio; 2520 2521 ext4_apply_options(fc, sb); 2522 ret = 0; 2523 2524 out_free: 2525 ext4_fc_free(fc); 2526 kfree(fc); 2527 return ret; 2528 } 2529 2530 static void ext4_apply_quota_options(struct fs_context *fc, 2531 struct super_block *sb) 2532 { 2533 #ifdef CONFIG_QUOTA 2534 bool quota_feature = ext4_has_feature_quota(sb); 2535 struct ext4_fs_context *ctx = fc->fs_private; 2536 struct ext4_sb_info *sbi = EXT4_SB(sb); 2537 char *qname; 2538 int i; 2539 2540 if (quota_feature) 2541 return; 2542 2543 if (ctx->spec & EXT4_SPEC_JQUOTA) { 2544 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2545 if (!(ctx->qname_spec & (1 << i))) 2546 continue; 2547 2548 qname = ctx->s_qf_names[i]; /* May be NULL */ 2549 if (qname) 2550 set_opt(sb, QUOTA); 2551 ctx->s_qf_names[i] = NULL; 2552 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname, 2553 lockdep_is_held(&sb->s_umount)); 2554 if (qname) 2555 kfree_rcu_mightsleep(qname); 2556 } 2557 } 2558 2559 if (ctx->spec & EXT4_SPEC_JQFMT) 2560 sbi->s_jquota_fmt = ctx->s_jquota_fmt; 2561 #endif 2562 } 2563 2564 /* 2565 * Check quota settings consistency. 2566 */ 2567 static int ext4_check_quota_consistency(struct fs_context *fc, 2568 struct super_block *sb) 2569 { 2570 #ifdef CONFIG_QUOTA 2571 struct ext4_fs_context *ctx = fc->fs_private; 2572 struct ext4_sb_info *sbi = EXT4_SB(sb); 2573 bool quota_feature = ext4_has_feature_quota(sb); 2574 bool quota_loaded = sb_any_quota_loaded(sb); 2575 bool usr_qf_name, grp_qf_name, usrquota, grpquota; 2576 int quota_flags, i; 2577 2578 /* 2579 * We do the test below only for project quotas. 'usrquota' and 2580 * 'grpquota' mount options are allowed even without quota feature 2581 * to support legacy quotas in quota files. 2582 */ 2583 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) && 2584 !ext4_has_feature_project(sb)) { 2585 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. " 2586 "Cannot enable project quota enforcement."); 2587 return -EINVAL; 2588 } 2589 2590 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 2591 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA; 2592 if (quota_loaded && 2593 ctx->mask_s_mount_opt & quota_flags && 2594 !ctx_test_mount_opt(ctx, quota_flags)) 2595 goto err_quota_change; 2596 2597 if (ctx->spec & EXT4_SPEC_JQUOTA) { 2598 2599 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2600 if (!(ctx->qname_spec & (1 << i))) 2601 continue; 2602 2603 if (quota_loaded && 2604 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i]) 2605 goto err_jquota_change; 2606 2607 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] && 2608 strcmp(get_qf_name(sb, sbi, i), 2609 ctx->s_qf_names[i]) != 0) 2610 goto err_jquota_specified; 2611 } 2612 2613 if (quota_feature) { 2614 ext4_msg(NULL, KERN_INFO, 2615 "Journaled quota options ignored when " 2616 "QUOTA feature is enabled"); 2617 return 0; 2618 } 2619 } 2620 2621 if (ctx->spec & EXT4_SPEC_JQFMT) { 2622 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded) 2623 goto err_jquota_change; 2624 if (quota_feature) { 2625 ext4_msg(NULL, KERN_INFO, "Quota format mount options " 2626 "ignored when QUOTA feature is enabled"); 2627 return 0; 2628 } 2629 } 2630 2631 /* Make sure we don't mix old and new quota format */ 2632 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) || 2633 ctx->s_qf_names[USRQUOTA]); 2634 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) || 2635 ctx->s_qf_names[GRPQUOTA]); 2636 2637 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) || 2638 test_opt(sb, USRQUOTA)); 2639 2640 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) || 2641 test_opt(sb, GRPQUOTA)); 2642 2643 if (usr_qf_name) { 2644 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA); 2645 usrquota = false; 2646 } 2647 if (grp_qf_name) { 2648 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA); 2649 grpquota = false; 2650 } 2651 2652 if (usr_qf_name || grp_qf_name) { 2653 if (usrquota || grpquota) { 2654 ext4_msg(NULL, KERN_ERR, "old and new quota " 2655 "format mixing"); 2656 return -EINVAL; 2657 } 2658 2659 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) { 2660 ext4_msg(NULL, KERN_ERR, "journaled quota format " 2661 "not specified"); 2662 return -EINVAL; 2663 } 2664 } 2665 2666 return 0; 2667 2668 err_quota_change: 2669 ext4_msg(NULL, KERN_ERR, 2670 "Cannot change quota options when quota turned on"); 2671 return -EINVAL; 2672 err_jquota_change: 2673 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota " 2674 "options when quota turned on"); 2675 return -EINVAL; 2676 err_jquota_specified: 2677 ext4_msg(NULL, KERN_ERR, "%s quota file already specified", 2678 QTYPE2NAME(i)); 2679 return -EINVAL; 2680 #else 2681 return 0; 2682 #endif 2683 } 2684 2685 static int ext4_check_test_dummy_encryption(const struct fs_context *fc, 2686 struct super_block *sb) 2687 { 2688 const struct ext4_fs_context *ctx = fc->fs_private; 2689 const struct ext4_sb_info *sbi = EXT4_SB(sb); 2690 2691 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy)) 2692 return 0; 2693 2694 if (!ext4_has_feature_encrypt(sb)) { 2695 ext4_msg(NULL, KERN_WARNING, 2696 "test_dummy_encryption requires encrypt feature"); 2697 return -EINVAL; 2698 } 2699 /* 2700 * This mount option is just for testing, and it's not worthwhile to 2701 * implement the extra complexity (e.g. RCU protection) that would be 2702 * needed to allow it to be set or changed during remount. We do allow 2703 * it to be specified during remount, but only if there is no change. 2704 */ 2705 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) { 2706 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy, 2707 &ctx->dummy_enc_policy)) 2708 return 0; 2709 ext4_msg(NULL, KERN_WARNING, 2710 "Can't set or change test_dummy_encryption on remount"); 2711 return -EINVAL; 2712 } 2713 /* Also make sure s_mount_opts didn't contain a conflicting value. */ 2714 if (fscrypt_is_dummy_policy_set(&sbi->s_dummy_enc_policy)) { 2715 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy, 2716 &ctx->dummy_enc_policy)) 2717 return 0; 2718 ext4_msg(NULL, KERN_WARNING, 2719 "Conflicting test_dummy_encryption options"); 2720 return -EINVAL; 2721 } 2722 return 0; 2723 } 2724 2725 static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx, 2726 struct super_block *sb) 2727 { 2728 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy) || 2729 /* if already set, it was already verified to be the same */ 2730 fscrypt_is_dummy_policy_set(&EXT4_SB(sb)->s_dummy_enc_policy)) 2731 return; 2732 EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy; 2733 memset(&ctx->dummy_enc_policy, 0, sizeof(ctx->dummy_enc_policy)); 2734 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled"); 2735 } 2736 2737 static int ext4_check_opt_consistency(struct fs_context *fc, 2738 struct super_block *sb) 2739 { 2740 struct ext4_fs_context *ctx = fc->fs_private; 2741 struct ext4_sb_info *sbi = fc->s_fs_info; 2742 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE; 2743 int err; 2744 2745 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 2746 ext4_msg(NULL, KERN_ERR, 2747 "Mount option(s) incompatible with ext2"); 2748 return -EINVAL; 2749 } 2750 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 2751 ext4_msg(NULL, KERN_ERR, 2752 "Mount option(s) incompatible with ext3"); 2753 return -EINVAL; 2754 } 2755 2756 if (ctx->s_want_extra_isize > 2757 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) { 2758 ext4_msg(NULL, KERN_ERR, 2759 "Invalid want_extra_isize %d", 2760 ctx->s_want_extra_isize); 2761 return -EINVAL; 2762 } 2763 2764 err = ext4_check_test_dummy_encryption(fc, sb); 2765 if (err) 2766 return err; 2767 2768 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) { 2769 if (!sbi->s_journal) { 2770 ext4_msg(NULL, KERN_WARNING, 2771 "Remounting file system with no journal " 2772 "so ignoring journalled data option"); 2773 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS); 2774 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) != 2775 test_opt(sb, DATA_FLAGS)) { 2776 ext4_msg(NULL, KERN_ERR, "Cannot change data mode " 2777 "on remount"); 2778 return -EINVAL; 2779 } 2780 } 2781 2782 if (is_remount) { 2783 if (!sbi->s_journal && 2784 ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_ERR_ABORT)) { 2785 ext4_msg(NULL, KERN_WARNING, 2786 "Remounting fs w/o journal so ignoring data_err option"); 2787 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_ERR_ABORT); 2788 } 2789 2790 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) && 2791 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) { 2792 ext4_msg(NULL, KERN_ERR, "can't mount with " 2793 "both data=journal and dax"); 2794 return -EINVAL; 2795 } 2796 2797 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) && 2798 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2799 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) { 2800 fail_dax_change_remount: 2801 ext4_msg(NULL, KERN_ERR, "can't change " 2802 "dax mount option while remounting"); 2803 return -EINVAL; 2804 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) && 2805 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2806 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) { 2807 goto fail_dax_change_remount; 2808 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) && 2809 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2810 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2811 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) { 2812 goto fail_dax_change_remount; 2813 } 2814 } 2815 2816 return ext4_check_quota_consistency(fc, sb); 2817 } 2818 2819 static void ext4_apply_options(struct fs_context *fc, struct super_block *sb) 2820 { 2821 struct ext4_fs_context *ctx = fc->fs_private; 2822 struct ext4_sb_info *sbi = fc->s_fs_info; 2823 2824 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt; 2825 sbi->s_mount_opt |= ctx->vals_s_mount_opt; 2826 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2; 2827 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2; 2828 sb->s_flags &= ~ctx->mask_s_flags; 2829 sb->s_flags |= ctx->vals_s_flags; 2830 2831 #define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; }) 2832 APPLY(s_commit_interval); 2833 APPLY(s_stripe); 2834 APPLY(s_max_batch_time); 2835 APPLY(s_min_batch_time); 2836 APPLY(s_want_extra_isize); 2837 APPLY(s_inode_readahead_blks); 2838 APPLY(s_max_dir_size_kb); 2839 APPLY(s_li_wait_mult); 2840 APPLY(s_resgid); 2841 APPLY(s_resuid); 2842 2843 #ifdef CONFIG_EXT4_DEBUG 2844 APPLY(s_fc_debug_max_replay); 2845 #endif 2846 2847 ext4_apply_quota_options(fc, sb); 2848 ext4_apply_test_dummy_encryption(ctx, sb); 2849 } 2850 2851 2852 static int ext4_validate_options(struct fs_context *fc) 2853 { 2854 #ifdef CONFIG_QUOTA 2855 struct ext4_fs_context *ctx = fc->fs_private; 2856 char *usr_qf_name, *grp_qf_name; 2857 2858 usr_qf_name = ctx->s_qf_names[USRQUOTA]; 2859 grp_qf_name = ctx->s_qf_names[GRPQUOTA]; 2860 2861 if (usr_qf_name || grp_qf_name) { 2862 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name) 2863 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA); 2864 2865 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name) 2866 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA); 2867 2868 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) || 2869 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) { 2870 ext4_msg(NULL, KERN_ERR, "old and new quota " 2871 "format mixing"); 2872 return -EINVAL; 2873 } 2874 } 2875 #endif 2876 return 1; 2877 } 2878 2879 static inline void ext4_show_quota_options(struct seq_file *seq, 2880 struct super_block *sb) 2881 { 2882 #if defined(CONFIG_QUOTA) 2883 struct ext4_sb_info *sbi = EXT4_SB(sb); 2884 char *usr_qf_name, *grp_qf_name; 2885 2886 if (sbi->s_jquota_fmt) { 2887 char *fmtname = ""; 2888 2889 switch (sbi->s_jquota_fmt) { 2890 case QFMT_VFS_OLD: 2891 fmtname = "vfsold"; 2892 break; 2893 case QFMT_VFS_V0: 2894 fmtname = "vfsv0"; 2895 break; 2896 case QFMT_VFS_V1: 2897 fmtname = "vfsv1"; 2898 break; 2899 } 2900 seq_printf(seq, ",jqfmt=%s", fmtname); 2901 } 2902 2903 rcu_read_lock(); 2904 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); 2905 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); 2906 if (usr_qf_name) 2907 seq_show_option(seq, "usrjquota", usr_qf_name); 2908 if (grp_qf_name) 2909 seq_show_option(seq, "grpjquota", grp_qf_name); 2910 rcu_read_unlock(); 2911 #endif 2912 } 2913 2914 static const char *token2str(int token) 2915 { 2916 const struct fs_parameter_spec *spec; 2917 2918 for (spec = ext4_param_specs; spec->name != NULL; spec++) 2919 if (spec->opt == token && !spec->type) 2920 break; 2921 return spec->name; 2922 } 2923 2924 /* 2925 * Show an option if 2926 * - it's set to a non-default value OR 2927 * - if the per-sb default is different from the global default 2928 */ 2929 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 2930 int nodefs) 2931 { 2932 struct ext4_sb_info *sbi = EXT4_SB(sb); 2933 struct ext4_super_block *es = sbi->s_es; 2934 int def_errors; 2935 const struct mount_opts *m; 2936 char sep = nodefs ? '\n' : ','; 2937 2938 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 2939 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 2940 2941 if (sbi->s_sb_block != 1) 2942 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 2943 2944 for (m = ext4_mount_opts; m->token != Opt_err; m++) { 2945 int want_set = m->flags & MOPT_SET; 2946 int opt_2 = m->flags & MOPT_2; 2947 unsigned int mount_opt, def_mount_opt; 2948 2949 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 2950 m->flags & MOPT_SKIP) 2951 continue; 2952 2953 if (opt_2) { 2954 mount_opt = sbi->s_mount_opt2; 2955 def_mount_opt = sbi->s_def_mount_opt2; 2956 } else { 2957 mount_opt = sbi->s_mount_opt; 2958 def_mount_opt = sbi->s_def_mount_opt; 2959 } 2960 /* skip if same as the default */ 2961 if (!nodefs && !(m->mount_opt & (mount_opt ^ def_mount_opt))) 2962 continue; 2963 /* select Opt_noFoo vs Opt_Foo */ 2964 if ((want_set && 2965 (mount_opt & m->mount_opt) != m->mount_opt) || 2966 (!want_set && (mount_opt & m->mount_opt))) 2967 continue; 2968 SEQ_OPTS_PRINT("%s", token2str(m->token)); 2969 } 2970 2971 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 2972 ext4_get_resuid(es) != EXT4_DEF_RESUID) 2973 SEQ_OPTS_PRINT("resuid=%u", 2974 from_kuid_munged(&init_user_ns, sbi->s_resuid)); 2975 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 2976 ext4_get_resgid(es) != EXT4_DEF_RESGID) 2977 SEQ_OPTS_PRINT("resgid=%u", 2978 from_kgid_munged(&init_user_ns, sbi->s_resgid)); 2979 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 2980 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 2981 SEQ_OPTS_PUTS("errors=remount-ro"); 2982 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 2983 SEQ_OPTS_PUTS("errors=continue"); 2984 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 2985 SEQ_OPTS_PUTS("errors=panic"); 2986 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 2987 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 2988 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 2989 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 2990 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 2991 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 2992 if (nodefs && sb->s_flags & SB_I_VERSION) 2993 SEQ_OPTS_PUTS("i_version"); 2994 if (nodefs || sbi->s_stripe) 2995 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 2996 if (nodefs || EXT4_MOUNT_DATA_FLAGS & 2997 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 2998 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 2999 SEQ_OPTS_PUTS("data=journal"); 3000 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 3001 SEQ_OPTS_PUTS("data=ordered"); 3002 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 3003 SEQ_OPTS_PUTS("data=writeback"); 3004 } 3005 if (nodefs || 3006 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 3007 SEQ_OPTS_PRINT("inode_readahead_blks=%u", 3008 sbi->s_inode_readahead_blks); 3009 3010 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs || 3011 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 3012 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 3013 if (nodefs || sbi->s_max_dir_size_kb) 3014 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 3015 if (test_opt(sb, DATA_ERR_ABORT)) 3016 SEQ_OPTS_PUTS("data_err=abort"); 3017 3018 fscrypt_show_test_dummy_encryption(seq, sep, sb); 3019 3020 if (sb->s_flags & SB_INLINECRYPT) 3021 SEQ_OPTS_PUTS("inlinecrypt"); 3022 3023 if (test_opt(sb, DAX_ALWAYS)) { 3024 if (IS_EXT2_SB(sb)) 3025 SEQ_OPTS_PUTS("dax"); 3026 else 3027 SEQ_OPTS_PUTS("dax=always"); 3028 } else if (test_opt2(sb, DAX_NEVER)) { 3029 SEQ_OPTS_PUTS("dax=never"); 3030 } else if (test_opt2(sb, DAX_INODE)) { 3031 SEQ_OPTS_PUTS("dax=inode"); 3032 } 3033 3034 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD && 3035 !test_opt2(sb, MB_OPTIMIZE_SCAN)) { 3036 SEQ_OPTS_PUTS("mb_optimize_scan=0"); 3037 } else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD && 3038 test_opt2(sb, MB_OPTIMIZE_SCAN)) { 3039 SEQ_OPTS_PUTS("mb_optimize_scan=1"); 3040 } 3041 3042 if (nodefs && !test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) 3043 SEQ_OPTS_PUTS("prefetch_block_bitmaps"); 3044 3045 if (ext4_emergency_ro(sb)) 3046 SEQ_OPTS_PUTS("emergency_ro"); 3047 3048 if (ext4_forced_shutdown(sb)) 3049 SEQ_OPTS_PUTS("shutdown"); 3050 3051 ext4_show_quota_options(seq, sb); 3052 return 0; 3053 } 3054 3055 static int ext4_show_options(struct seq_file *seq, struct dentry *root) 3056 { 3057 return _ext4_show_options(seq, root->d_sb, 0); 3058 } 3059 3060 int ext4_seq_options_show(struct seq_file *seq, void *offset) 3061 { 3062 struct super_block *sb = seq->private; 3063 int rc; 3064 3065 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); 3066 rc = _ext4_show_options(seq, sb, 1); 3067 seq_putc(seq, '\n'); 3068 return rc; 3069 } 3070 3071 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 3072 int read_only) 3073 { 3074 struct ext4_sb_info *sbi = EXT4_SB(sb); 3075 int err = 0; 3076 3077 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 3078 ext4_msg(sb, KERN_ERR, "revision level too high, " 3079 "forcing read-only mode"); 3080 err = -EROFS; 3081 goto done; 3082 } 3083 if (read_only) 3084 goto done; 3085 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 3086 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 3087 "running e2fsck is recommended"); 3088 else if (sbi->s_mount_state & EXT4_ERROR_FS) 3089 ext4_msg(sb, KERN_WARNING, 3090 "warning: mounting fs with errors, " 3091 "running e2fsck is recommended"); 3092 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 3093 le16_to_cpu(es->s_mnt_count) >= 3094 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 3095 ext4_msg(sb, KERN_WARNING, 3096 "warning: maximal mount count reached, " 3097 "running e2fsck is recommended"); 3098 else if (le32_to_cpu(es->s_checkinterval) && 3099 (ext4_get_tstamp(es, s_lastcheck) + 3100 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds())) 3101 ext4_msg(sb, KERN_WARNING, 3102 "warning: checktime reached, " 3103 "running e2fsck is recommended"); 3104 if (!sbi->s_journal) 3105 es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 3106 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 3107 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 3108 le16_add_cpu(&es->s_mnt_count, 1); 3109 ext4_update_tstamp(es, s_mtime); 3110 if (sbi->s_journal) { 3111 ext4_set_feature_journal_needs_recovery(sb); 3112 if (ext4_has_feature_orphan_file(sb)) 3113 ext4_set_feature_orphan_present(sb); 3114 } 3115 3116 err = ext4_commit_super(sb); 3117 done: 3118 if (test_opt(sb, DEBUG)) 3119 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 3120 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 3121 sb->s_blocksize, 3122 sbi->s_groups_count, 3123 EXT4_BLOCKS_PER_GROUP(sb), 3124 EXT4_INODES_PER_GROUP(sb), 3125 sbi->s_mount_opt, sbi->s_mount_opt2); 3126 return err; 3127 } 3128 3129 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 3130 { 3131 struct ext4_sb_info *sbi = EXT4_SB(sb); 3132 struct flex_groups **old_groups, **new_groups; 3133 int size, i, j; 3134 3135 if (!sbi->s_log_groups_per_flex) 3136 return 0; 3137 3138 size = ext4_flex_group(sbi, ngroup - 1) + 1; 3139 if (size <= sbi->s_flex_groups_allocated) 3140 return 0; 3141 3142 new_groups = kvzalloc(roundup_pow_of_two(size * 3143 sizeof(*sbi->s_flex_groups)), GFP_KERNEL); 3144 if (!new_groups) { 3145 ext4_msg(sb, KERN_ERR, 3146 "not enough memory for %d flex group pointers", size); 3147 return -ENOMEM; 3148 } 3149 for (i = sbi->s_flex_groups_allocated; i < size; i++) { 3150 new_groups[i] = kvzalloc(roundup_pow_of_two( 3151 sizeof(struct flex_groups)), 3152 GFP_KERNEL); 3153 if (!new_groups[i]) { 3154 for (j = sbi->s_flex_groups_allocated; j < i; j++) 3155 kvfree(new_groups[j]); 3156 kvfree(new_groups); 3157 ext4_msg(sb, KERN_ERR, 3158 "not enough memory for %d flex groups", size); 3159 return -ENOMEM; 3160 } 3161 } 3162 rcu_read_lock(); 3163 old_groups = rcu_dereference(sbi->s_flex_groups); 3164 if (old_groups) 3165 memcpy(new_groups, old_groups, 3166 (sbi->s_flex_groups_allocated * 3167 sizeof(struct flex_groups *))); 3168 rcu_read_unlock(); 3169 rcu_assign_pointer(sbi->s_flex_groups, new_groups); 3170 sbi->s_flex_groups_allocated = size; 3171 if (old_groups) 3172 ext4_kvfree_array_rcu(old_groups); 3173 return 0; 3174 } 3175 3176 static int ext4_fill_flex_info(struct super_block *sb) 3177 { 3178 struct ext4_sb_info *sbi = EXT4_SB(sb); 3179 struct ext4_group_desc *gdp = NULL; 3180 struct flex_groups *fg; 3181 ext4_group_t flex_group; 3182 int i, err; 3183 3184 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 3185 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 3186 sbi->s_log_groups_per_flex = 0; 3187 return 1; 3188 } 3189 3190 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 3191 if (err) 3192 goto failed; 3193 3194 for (i = 0; i < sbi->s_groups_count; i++) { 3195 gdp = ext4_get_group_desc(sb, i, NULL); 3196 3197 flex_group = ext4_flex_group(sbi, i); 3198 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group); 3199 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes); 3200 atomic64_add(ext4_free_group_clusters(sb, gdp), 3201 &fg->free_clusters); 3202 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs); 3203 } 3204 3205 return 1; 3206 failed: 3207 return 0; 3208 } 3209 3210 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 3211 struct ext4_group_desc *gdp) 3212 { 3213 int offset = offsetof(struct ext4_group_desc, bg_checksum); 3214 __u16 crc = 0; 3215 __le32 le_group = cpu_to_le32(block_group); 3216 struct ext4_sb_info *sbi = EXT4_SB(sb); 3217 3218 if (ext4_has_feature_metadata_csum(sbi->s_sb)) { 3219 /* Use new metadata_csum algorithm */ 3220 __u32 csum32; 3221 __u16 dummy_csum = 0; 3222 3223 csum32 = ext4_chksum(sbi->s_csum_seed, (__u8 *)&le_group, 3224 sizeof(le_group)); 3225 csum32 = ext4_chksum(csum32, (__u8 *)gdp, offset); 3226 csum32 = ext4_chksum(csum32, (__u8 *)&dummy_csum, 3227 sizeof(dummy_csum)); 3228 offset += sizeof(dummy_csum); 3229 if (offset < sbi->s_desc_size) 3230 csum32 = ext4_chksum(csum32, (__u8 *)gdp + offset, 3231 sbi->s_desc_size - offset); 3232 3233 crc = csum32 & 0xFFFF; 3234 goto out; 3235 } 3236 3237 /* old crc16 code */ 3238 if (!ext4_has_feature_gdt_csum(sb)) 3239 return 0; 3240 3241 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 3242 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 3243 crc = crc16(crc, (__u8 *)gdp, offset); 3244 offset += sizeof(gdp->bg_checksum); /* skip checksum */ 3245 /* for checksum of struct ext4_group_desc do the rest...*/ 3246 if (ext4_has_feature_64bit(sb) && offset < sbi->s_desc_size) 3247 crc = crc16(crc, (__u8 *)gdp + offset, 3248 sbi->s_desc_size - offset); 3249 3250 out: 3251 return cpu_to_le16(crc); 3252 } 3253 3254 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 3255 struct ext4_group_desc *gdp) 3256 { 3257 if (ext4_has_group_desc_csum(sb) && 3258 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 3259 return 0; 3260 3261 return 1; 3262 } 3263 3264 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 3265 struct ext4_group_desc *gdp) 3266 { 3267 if (!ext4_has_group_desc_csum(sb)) 3268 return; 3269 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 3270 } 3271 3272 /* Called at mount-time, super-block is locked */ 3273 static int ext4_check_descriptors(struct super_block *sb, 3274 ext4_fsblk_t sb_block, 3275 ext4_group_t *first_not_zeroed) 3276 { 3277 struct ext4_sb_info *sbi = EXT4_SB(sb); 3278 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 3279 ext4_fsblk_t last_block; 3280 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0); 3281 ext4_fsblk_t block_bitmap; 3282 ext4_fsblk_t inode_bitmap; 3283 ext4_fsblk_t inode_table; 3284 int flexbg_flag = 0; 3285 ext4_group_t i, grp = sbi->s_groups_count; 3286 3287 if (ext4_has_feature_flex_bg(sb)) 3288 flexbg_flag = 1; 3289 3290 ext4_debug("Checking group descriptors"); 3291 3292 for (i = 0; i < sbi->s_groups_count; i++) { 3293 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 3294 3295 if (i == sbi->s_groups_count - 1 || flexbg_flag) 3296 last_block = ext4_blocks_count(sbi->s_es) - 1; 3297 else 3298 last_block = first_block + 3299 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 3300 3301 if ((grp == sbi->s_groups_count) && 3302 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3303 grp = i; 3304 3305 block_bitmap = ext4_block_bitmap(sb, gdp); 3306 if (block_bitmap == sb_block) { 3307 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3308 "Block bitmap for group %u overlaps " 3309 "superblock", i); 3310 if (!sb_rdonly(sb)) 3311 return 0; 3312 } 3313 if (block_bitmap >= sb_block + 1 && 3314 block_bitmap <= last_bg_block) { 3315 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3316 "Block bitmap for group %u overlaps " 3317 "block group descriptors", i); 3318 if (!sb_rdonly(sb)) 3319 return 0; 3320 } 3321 if (block_bitmap < first_block || block_bitmap > last_block) { 3322 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3323 "Block bitmap for group %u not in group " 3324 "(block %llu)!", i, block_bitmap); 3325 return 0; 3326 } 3327 inode_bitmap = ext4_inode_bitmap(sb, gdp); 3328 if (inode_bitmap == sb_block) { 3329 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3330 "Inode bitmap for group %u overlaps " 3331 "superblock", i); 3332 if (!sb_rdonly(sb)) 3333 return 0; 3334 } 3335 if (inode_bitmap >= sb_block + 1 && 3336 inode_bitmap <= last_bg_block) { 3337 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3338 "Inode bitmap for group %u overlaps " 3339 "block group descriptors", i); 3340 if (!sb_rdonly(sb)) 3341 return 0; 3342 } 3343 if (inode_bitmap < first_block || inode_bitmap > last_block) { 3344 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3345 "Inode bitmap for group %u not in group " 3346 "(block %llu)!", i, inode_bitmap); 3347 return 0; 3348 } 3349 inode_table = ext4_inode_table(sb, gdp); 3350 if (inode_table == sb_block) { 3351 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3352 "Inode table for group %u overlaps " 3353 "superblock", i); 3354 if (!sb_rdonly(sb)) 3355 return 0; 3356 } 3357 if (inode_table >= sb_block + 1 && 3358 inode_table <= last_bg_block) { 3359 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3360 "Inode table for group %u overlaps " 3361 "block group descriptors", i); 3362 if (!sb_rdonly(sb)) 3363 return 0; 3364 } 3365 if (inode_table < first_block || 3366 inode_table + sbi->s_itb_per_group - 1 > last_block) { 3367 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3368 "Inode table for group %u not in group " 3369 "(block %llu)!", i, inode_table); 3370 return 0; 3371 } 3372 ext4_lock_group(sb, i); 3373 if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 3374 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 3375 "Checksum for group %u failed (%u!=%u)", 3376 i, le16_to_cpu(ext4_group_desc_csum(sb, i, 3377 gdp)), le16_to_cpu(gdp->bg_checksum)); 3378 if (!sb_rdonly(sb)) { 3379 ext4_unlock_group(sb, i); 3380 return 0; 3381 } 3382 } 3383 ext4_unlock_group(sb, i); 3384 if (!flexbg_flag) 3385 first_block += EXT4_BLOCKS_PER_GROUP(sb); 3386 } 3387 if (NULL != first_not_zeroed) 3388 *first_not_zeroed = grp; 3389 return 1; 3390 } 3391 3392 /* 3393 * Maximal extent format file size. 3394 * Resulting logical blkno at s_maxbytes must fit in our on-disk 3395 * extent format containers, within a sector_t, and within i_blocks 3396 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 3397 * so that won't be a limiting factor. 3398 * 3399 * However there is other limiting factor. We do store extents in the form 3400 * of starting block and length, hence the resulting length of the extent 3401 * covering maximum file size must fit into on-disk format containers as 3402 * well. Given that length is always by 1 unit bigger than max unit (because 3403 * we count 0 as well) we have to lower the s_maxbytes by one fs block. 3404 * 3405 * Note, this does *not* consider any metadata overhead for vfs i_blocks. 3406 */ 3407 static loff_t ext4_max_size(int blkbits, int has_huge_files) 3408 { 3409 loff_t res; 3410 loff_t upper_limit = MAX_LFS_FILESIZE; 3411 3412 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64)); 3413 3414 if (!has_huge_files) { 3415 upper_limit = (1LL << 32) - 1; 3416 3417 /* total blocks in file system block size */ 3418 upper_limit >>= (blkbits - 9); 3419 upper_limit <<= blkbits; 3420 } 3421 3422 /* 3423 * 32-bit extent-start container, ee_block. We lower the maxbytes 3424 * by one fs block, so ee_len can cover the extent of maximum file 3425 * size 3426 */ 3427 res = (1LL << 32) - 1; 3428 res <<= blkbits; 3429 3430 /* Sanity check against vm- & vfs- imposed limits */ 3431 if (res > upper_limit) 3432 res = upper_limit; 3433 3434 return res; 3435 } 3436 3437 /* 3438 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 3439 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 3440 * We need to be 1 filesystem block less than the 2^48 sector limit. 3441 */ 3442 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 3443 { 3444 loff_t upper_limit, res = EXT4_NDIR_BLOCKS; 3445 int meta_blocks; 3446 unsigned int ppb = 1 << (bits - 2); 3447 3448 /* 3449 * This is calculated to be the largest file size for a dense, block 3450 * mapped file such that the file's total number of 512-byte sectors, 3451 * including data and all indirect blocks, does not exceed (2^48 - 1). 3452 * 3453 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 3454 * number of 512-byte sectors of the file. 3455 */ 3456 if (!has_huge_files) { 3457 /* 3458 * !has_huge_files or implies that the inode i_block field 3459 * represents total file blocks in 2^32 512-byte sectors == 3460 * size of vfs inode i_blocks * 8 3461 */ 3462 upper_limit = (1LL << 32) - 1; 3463 3464 /* total blocks in file system block size */ 3465 upper_limit >>= (bits - 9); 3466 3467 } else { 3468 /* 3469 * We use 48 bit ext4_inode i_blocks 3470 * With EXT4_HUGE_FILE_FL set the i_blocks 3471 * represent total number of blocks in 3472 * file system block size 3473 */ 3474 upper_limit = (1LL << 48) - 1; 3475 3476 } 3477 3478 /* Compute how many blocks we can address by block tree */ 3479 res += ppb; 3480 res += ppb * ppb; 3481 res += ((loff_t)ppb) * ppb * ppb; 3482 /* Compute how many metadata blocks are needed */ 3483 meta_blocks = 1; 3484 meta_blocks += 1 + ppb; 3485 meta_blocks += 1 + ppb + ppb * ppb; 3486 /* Does block tree limit file size? */ 3487 if (res + meta_blocks <= upper_limit) 3488 goto check_lfs; 3489 3490 res = upper_limit; 3491 /* How many metadata blocks are needed for addressing upper_limit? */ 3492 upper_limit -= EXT4_NDIR_BLOCKS; 3493 /* indirect blocks */ 3494 meta_blocks = 1; 3495 upper_limit -= ppb; 3496 /* double indirect blocks */ 3497 if (upper_limit < ppb * ppb) { 3498 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb); 3499 res -= meta_blocks; 3500 goto check_lfs; 3501 } 3502 meta_blocks += 1 + ppb; 3503 upper_limit -= ppb * ppb; 3504 /* tripple indirect blocks for the rest */ 3505 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) + 3506 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb); 3507 res -= meta_blocks; 3508 check_lfs: 3509 res <<= bits; 3510 if (res > MAX_LFS_FILESIZE) 3511 res = MAX_LFS_FILESIZE; 3512 3513 return res; 3514 } 3515 3516 static ext4_fsblk_t descriptor_loc(struct super_block *sb, 3517 ext4_fsblk_t logical_sb_block, int nr) 3518 { 3519 struct ext4_sb_info *sbi = EXT4_SB(sb); 3520 ext4_group_t bg, first_meta_bg; 3521 int has_super = 0; 3522 3523 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 3524 3525 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 3526 return logical_sb_block + nr + 1; 3527 bg = sbi->s_desc_per_block * nr; 3528 if (ext4_bg_has_super(sb, bg)) 3529 has_super = 1; 3530 3531 /* 3532 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 3533 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 3534 * on modern mke2fs or blksize > 1k on older mke2fs) then we must 3535 * compensate. 3536 */ 3537 if (sb->s_blocksize == 1024 && nr == 0 && 3538 le32_to_cpu(sbi->s_es->s_first_data_block) == 0) 3539 has_super++; 3540 3541 return (has_super + ext4_group_first_block_no(sb, bg)); 3542 } 3543 3544 /** 3545 * ext4_get_stripe_size: Get the stripe size. 3546 * @sbi: In memory super block info 3547 * 3548 * If we have specified it via mount option, then 3549 * use the mount option value. If the value specified at mount time is 3550 * greater than the blocks per group use the super block value. 3551 * If the super block value is greater than blocks per group return 0. 3552 * Allocator needs it be less than blocks per group. 3553 * 3554 */ 3555 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 3556 { 3557 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 3558 unsigned long stripe_width = 3559 le32_to_cpu(sbi->s_es->s_raid_stripe_width); 3560 int ret; 3561 3562 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 3563 ret = sbi->s_stripe; 3564 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) 3565 ret = stripe_width; 3566 else if (stride && stride <= sbi->s_blocks_per_group) 3567 ret = stride; 3568 else 3569 ret = 0; 3570 3571 /* 3572 * If the stripe width is 1, this makes no sense and 3573 * we set it to 0 to turn off stripe handling code. 3574 */ 3575 if (ret <= 1) 3576 ret = 0; 3577 3578 return ret; 3579 } 3580 3581 /* 3582 * Check whether this filesystem can be mounted based on 3583 * the features present and the RDONLY/RDWR mount requested. 3584 * Returns 1 if this filesystem can be mounted as requested, 3585 * 0 if it cannot be. 3586 */ 3587 int ext4_feature_set_ok(struct super_block *sb, int readonly) 3588 { 3589 if (ext4_has_unknown_ext4_incompat_features(sb)) { 3590 ext4_msg(sb, KERN_ERR, 3591 "Couldn't mount because of " 3592 "unsupported optional features (%x)", 3593 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 3594 ~EXT4_FEATURE_INCOMPAT_SUPP)); 3595 return 0; 3596 } 3597 3598 if (!IS_ENABLED(CONFIG_UNICODE) && ext4_has_feature_casefold(sb)) { 3599 ext4_msg(sb, KERN_ERR, 3600 "Filesystem with casefold feature cannot be " 3601 "mounted without CONFIG_UNICODE"); 3602 return 0; 3603 } 3604 3605 if (readonly) 3606 return 1; 3607 3608 if (ext4_has_feature_readonly(sb)) { 3609 ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 3610 sb->s_flags |= SB_RDONLY; 3611 return 1; 3612 } 3613 3614 /* Check that feature set is OK for a read-write mount */ 3615 if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 3616 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 3617 "unsupported optional features (%x)", 3618 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 3619 ~EXT4_FEATURE_RO_COMPAT_SUPP)); 3620 return 0; 3621 } 3622 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 3623 ext4_msg(sb, KERN_ERR, 3624 "Can't support bigalloc feature without " 3625 "extents feature\n"); 3626 return 0; 3627 } 3628 3629 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2) 3630 if (!readonly && (ext4_has_feature_quota(sb) || 3631 ext4_has_feature_project(sb))) { 3632 ext4_msg(sb, KERN_ERR, 3633 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2"); 3634 return 0; 3635 } 3636 #endif /* CONFIG_QUOTA */ 3637 return 1; 3638 } 3639 3640 /* 3641 * This function is called once a day if we have errors logged 3642 * on the file system 3643 */ 3644 static void print_daily_error_info(struct timer_list *t) 3645 { 3646 struct ext4_sb_info *sbi = timer_container_of(sbi, t, s_err_report); 3647 struct super_block *sb = sbi->s_sb; 3648 struct ext4_super_block *es = sbi->s_es; 3649 3650 if (es->s_error_count) 3651 /* fsck newer than v1.41.13 is needed to clean this condition. */ 3652 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 3653 le32_to_cpu(es->s_error_count)); 3654 if (es->s_first_error_time) { 3655 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d", 3656 sb->s_id, 3657 ext4_get_tstamp(es, s_first_error_time), 3658 (int) sizeof(es->s_first_error_func), 3659 es->s_first_error_func, 3660 le32_to_cpu(es->s_first_error_line)); 3661 if (es->s_first_error_ino) 3662 printk(KERN_CONT ": inode %u", 3663 le32_to_cpu(es->s_first_error_ino)); 3664 if (es->s_first_error_block) 3665 printk(KERN_CONT ": block %llu", (unsigned long long) 3666 le64_to_cpu(es->s_first_error_block)); 3667 printk(KERN_CONT "\n"); 3668 } 3669 if (es->s_last_error_time) { 3670 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d", 3671 sb->s_id, 3672 ext4_get_tstamp(es, s_last_error_time), 3673 (int) sizeof(es->s_last_error_func), 3674 es->s_last_error_func, 3675 le32_to_cpu(es->s_last_error_line)); 3676 if (es->s_last_error_ino) 3677 printk(KERN_CONT ": inode %u", 3678 le32_to_cpu(es->s_last_error_ino)); 3679 if (es->s_last_error_block) 3680 printk(KERN_CONT ": block %llu", (unsigned long long) 3681 le64_to_cpu(es->s_last_error_block)); 3682 printk(KERN_CONT "\n"); 3683 } 3684 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 3685 } 3686 3687 /* Find next suitable group and run ext4_init_inode_table */ 3688 static int ext4_run_li_request(struct ext4_li_request *elr) 3689 { 3690 struct ext4_group_desc *gdp = NULL; 3691 struct super_block *sb = elr->lr_super; 3692 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 3693 ext4_group_t group = elr->lr_next_group; 3694 unsigned int prefetch_ios = 0; 3695 int ret = 0; 3696 int nr = EXT4_SB(sb)->s_mb_prefetch; 3697 u64 start_time; 3698 3699 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) { 3700 elr->lr_next_group = ext4_mb_prefetch(sb, group, nr, &prefetch_ios); 3701 ext4_mb_prefetch_fini(sb, elr->lr_next_group, nr); 3702 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, nr); 3703 if (group >= elr->lr_next_group) { 3704 ret = 1; 3705 if (elr->lr_first_not_zeroed != ngroups && 3706 !ext4_emergency_state(sb) && !sb_rdonly(sb) && 3707 test_opt(sb, INIT_INODE_TABLE)) { 3708 elr->lr_next_group = elr->lr_first_not_zeroed; 3709 elr->lr_mode = EXT4_LI_MODE_ITABLE; 3710 ret = 0; 3711 } 3712 } 3713 return ret; 3714 } 3715 3716 for (; group < ngroups; group++) { 3717 gdp = ext4_get_group_desc(sb, group, NULL); 3718 if (!gdp) { 3719 ret = 1; 3720 break; 3721 } 3722 3723 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3724 break; 3725 } 3726 3727 if (group >= ngroups) 3728 ret = 1; 3729 3730 if (!ret) { 3731 start_time = ktime_get_ns(); 3732 ret = ext4_init_inode_table(sb, group, 3733 elr->lr_timeout ? 0 : 1); 3734 trace_ext4_lazy_itable_init(sb, group); 3735 if (elr->lr_timeout == 0) { 3736 elr->lr_timeout = nsecs_to_jiffies((ktime_get_ns() - start_time) * 3737 EXT4_SB(elr->lr_super)->s_li_wait_mult); 3738 } 3739 elr->lr_next_sched = jiffies + elr->lr_timeout; 3740 elr->lr_next_group = group + 1; 3741 } 3742 return ret; 3743 } 3744 3745 /* 3746 * Remove lr_request from the list_request and free the 3747 * request structure. Should be called with li_list_mtx held 3748 */ 3749 static void ext4_remove_li_request(struct ext4_li_request *elr) 3750 { 3751 if (!elr) 3752 return; 3753 3754 list_del(&elr->lr_request); 3755 EXT4_SB(elr->lr_super)->s_li_request = NULL; 3756 kfree(elr); 3757 } 3758 3759 static void ext4_unregister_li_request(struct super_block *sb) 3760 { 3761 mutex_lock(&ext4_li_mtx); 3762 if (!ext4_li_info) { 3763 mutex_unlock(&ext4_li_mtx); 3764 return; 3765 } 3766 3767 mutex_lock(&ext4_li_info->li_list_mtx); 3768 ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 3769 mutex_unlock(&ext4_li_info->li_list_mtx); 3770 mutex_unlock(&ext4_li_mtx); 3771 } 3772 3773 static struct task_struct *ext4_lazyinit_task; 3774 3775 /* 3776 * This is the function where ext4lazyinit thread lives. It walks 3777 * through the request list searching for next scheduled filesystem. 3778 * When such a fs is found, run the lazy initialization request 3779 * (ext4_rn_li_request) and keep track of the time spend in this 3780 * function. Based on that time we compute next schedule time of 3781 * the request. When walking through the list is complete, compute 3782 * next waking time and put itself into sleep. 3783 */ 3784 static int ext4_lazyinit_thread(void *arg) 3785 { 3786 struct ext4_lazy_init *eli = arg; 3787 struct list_head *pos, *n; 3788 struct ext4_li_request *elr; 3789 unsigned long next_wakeup, cur; 3790 3791 BUG_ON(NULL == eli); 3792 set_freezable(); 3793 3794 cont_thread: 3795 while (true) { 3796 bool next_wakeup_initialized = false; 3797 3798 next_wakeup = 0; 3799 mutex_lock(&eli->li_list_mtx); 3800 if (list_empty(&eli->li_request_list)) { 3801 mutex_unlock(&eli->li_list_mtx); 3802 goto exit_thread; 3803 } 3804 list_for_each_safe(pos, n, &eli->li_request_list) { 3805 int err = 0; 3806 int progress = 0; 3807 elr = list_entry(pos, struct ext4_li_request, 3808 lr_request); 3809 3810 if (time_before(jiffies, elr->lr_next_sched)) { 3811 if (!next_wakeup_initialized || 3812 time_before(elr->lr_next_sched, next_wakeup)) { 3813 next_wakeup = elr->lr_next_sched; 3814 next_wakeup_initialized = true; 3815 } 3816 continue; 3817 } 3818 if (down_read_trylock(&elr->lr_super->s_umount)) { 3819 if (sb_start_write_trylock(elr->lr_super)) { 3820 progress = 1; 3821 /* 3822 * We hold sb->s_umount, sb can not 3823 * be removed from the list, it is 3824 * now safe to drop li_list_mtx 3825 */ 3826 mutex_unlock(&eli->li_list_mtx); 3827 err = ext4_run_li_request(elr); 3828 sb_end_write(elr->lr_super); 3829 mutex_lock(&eli->li_list_mtx); 3830 n = pos->next; 3831 } 3832 up_read((&elr->lr_super->s_umount)); 3833 } 3834 /* error, remove the lazy_init job */ 3835 if (err) { 3836 ext4_remove_li_request(elr); 3837 continue; 3838 } 3839 if (!progress) { 3840 elr->lr_next_sched = jiffies + 3841 get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ); 3842 } 3843 if (!next_wakeup_initialized || 3844 time_before(elr->lr_next_sched, next_wakeup)) { 3845 next_wakeup = elr->lr_next_sched; 3846 next_wakeup_initialized = true; 3847 } 3848 } 3849 mutex_unlock(&eli->li_list_mtx); 3850 3851 try_to_freeze(); 3852 3853 cur = jiffies; 3854 if (!next_wakeup_initialized || time_after_eq(cur, next_wakeup)) { 3855 cond_resched(); 3856 continue; 3857 } 3858 3859 schedule_timeout_interruptible(next_wakeup - cur); 3860 3861 if (kthread_should_stop()) { 3862 ext4_clear_request_list(); 3863 goto exit_thread; 3864 } 3865 } 3866 3867 exit_thread: 3868 /* 3869 * It looks like the request list is empty, but we need 3870 * to check it under the li_list_mtx lock, to prevent any 3871 * additions into it, and of course we should lock ext4_li_mtx 3872 * to atomically free the list and ext4_li_info, because at 3873 * this point another ext4 filesystem could be registering 3874 * new one. 3875 */ 3876 mutex_lock(&ext4_li_mtx); 3877 mutex_lock(&eli->li_list_mtx); 3878 if (!list_empty(&eli->li_request_list)) { 3879 mutex_unlock(&eli->li_list_mtx); 3880 mutex_unlock(&ext4_li_mtx); 3881 goto cont_thread; 3882 } 3883 mutex_unlock(&eli->li_list_mtx); 3884 kfree(ext4_li_info); 3885 ext4_li_info = NULL; 3886 mutex_unlock(&ext4_li_mtx); 3887 3888 return 0; 3889 } 3890 3891 static void ext4_clear_request_list(void) 3892 { 3893 struct list_head *pos, *n; 3894 struct ext4_li_request *elr; 3895 3896 mutex_lock(&ext4_li_info->li_list_mtx); 3897 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3898 elr = list_entry(pos, struct ext4_li_request, 3899 lr_request); 3900 ext4_remove_li_request(elr); 3901 } 3902 mutex_unlock(&ext4_li_info->li_list_mtx); 3903 } 3904 3905 static int ext4_run_lazyinit_thread(void) 3906 { 3907 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 3908 ext4_li_info, "ext4lazyinit"); 3909 if (IS_ERR(ext4_lazyinit_task)) { 3910 int err = PTR_ERR(ext4_lazyinit_task); 3911 ext4_clear_request_list(); 3912 kfree(ext4_li_info); 3913 ext4_li_info = NULL; 3914 printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3915 "initialization thread\n", 3916 err); 3917 return err; 3918 } 3919 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3920 return 0; 3921 } 3922 3923 /* 3924 * Check whether it make sense to run itable init. thread or not. 3925 * If there is at least one uninitialized inode table, return 3926 * corresponding group number, else the loop goes through all 3927 * groups and return total number of groups. 3928 */ 3929 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3930 { 3931 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3932 struct ext4_group_desc *gdp = NULL; 3933 3934 if (!ext4_has_group_desc_csum(sb)) 3935 return ngroups; 3936 3937 for (group = 0; group < ngroups; group++) { 3938 gdp = ext4_get_group_desc(sb, group, NULL); 3939 if (!gdp) 3940 continue; 3941 3942 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3943 break; 3944 } 3945 3946 return group; 3947 } 3948 3949 static int ext4_li_info_new(void) 3950 { 3951 struct ext4_lazy_init *eli = NULL; 3952 3953 eli = kzalloc(sizeof(*eli), GFP_KERNEL); 3954 if (!eli) 3955 return -ENOMEM; 3956 3957 INIT_LIST_HEAD(&eli->li_request_list); 3958 mutex_init(&eli->li_list_mtx); 3959 3960 eli->li_state |= EXT4_LAZYINIT_QUIT; 3961 3962 ext4_li_info = eli; 3963 3964 return 0; 3965 } 3966 3967 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3968 ext4_group_t start) 3969 { 3970 struct ext4_li_request *elr; 3971 3972 elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3973 if (!elr) 3974 return NULL; 3975 3976 elr->lr_super = sb; 3977 elr->lr_first_not_zeroed = start; 3978 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) { 3979 elr->lr_mode = EXT4_LI_MODE_ITABLE; 3980 elr->lr_next_group = start; 3981 } else { 3982 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP; 3983 } 3984 3985 /* 3986 * Randomize first schedule time of the request to 3987 * spread the inode table initialization requests 3988 * better. 3989 */ 3990 elr->lr_next_sched = jiffies + get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ); 3991 return elr; 3992 } 3993 3994 int ext4_register_li_request(struct super_block *sb, 3995 ext4_group_t first_not_zeroed) 3996 { 3997 struct ext4_sb_info *sbi = EXT4_SB(sb); 3998 struct ext4_li_request *elr = NULL; 3999 ext4_group_t ngroups = sbi->s_groups_count; 4000 int ret = 0; 4001 4002 mutex_lock(&ext4_li_mtx); 4003 if (sbi->s_li_request != NULL) { 4004 /* 4005 * Reset timeout so it can be computed again, because 4006 * s_li_wait_mult might have changed. 4007 */ 4008 sbi->s_li_request->lr_timeout = 0; 4009 goto out; 4010 } 4011 4012 if (ext4_emergency_state(sb) || sb_rdonly(sb) || 4013 (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) && 4014 (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE)))) 4015 goto out; 4016 4017 elr = ext4_li_request_new(sb, first_not_zeroed); 4018 if (!elr) { 4019 ret = -ENOMEM; 4020 goto out; 4021 } 4022 4023 if (NULL == ext4_li_info) { 4024 ret = ext4_li_info_new(); 4025 if (ret) 4026 goto out; 4027 } 4028 4029 mutex_lock(&ext4_li_info->li_list_mtx); 4030 list_add(&elr->lr_request, &ext4_li_info->li_request_list); 4031 mutex_unlock(&ext4_li_info->li_list_mtx); 4032 4033 sbi->s_li_request = elr; 4034 /* 4035 * set elr to NULL here since it has been inserted to 4036 * the request_list and the removal and free of it is 4037 * handled by ext4_clear_request_list from now on. 4038 */ 4039 elr = NULL; 4040 4041 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 4042 ret = ext4_run_lazyinit_thread(); 4043 if (ret) 4044 goto out; 4045 } 4046 out: 4047 mutex_unlock(&ext4_li_mtx); 4048 if (ret) 4049 kfree(elr); 4050 return ret; 4051 } 4052 4053 /* 4054 * We do not need to lock anything since this is called on 4055 * module unload. 4056 */ 4057 static void ext4_destroy_lazyinit_thread(void) 4058 { 4059 /* 4060 * If thread exited earlier 4061 * there's nothing to be done. 4062 */ 4063 if (!ext4_li_info || !ext4_lazyinit_task) 4064 return; 4065 4066 kthread_stop(ext4_lazyinit_task); 4067 } 4068 4069 static int set_journal_csum_feature_set(struct super_block *sb) 4070 { 4071 int ret = 1; 4072 int compat, incompat; 4073 struct ext4_sb_info *sbi = EXT4_SB(sb); 4074 4075 if (ext4_has_feature_metadata_csum(sb)) { 4076 /* journal checksum v3 */ 4077 compat = 0; 4078 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 4079 } else { 4080 /* journal checksum v1 */ 4081 compat = JBD2_FEATURE_COMPAT_CHECKSUM; 4082 incompat = 0; 4083 } 4084 4085 jbd2_journal_clear_features(sbi->s_journal, 4086 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 4087 JBD2_FEATURE_INCOMPAT_CSUM_V3 | 4088 JBD2_FEATURE_INCOMPAT_CSUM_V2); 4089 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4090 ret = jbd2_journal_set_features(sbi->s_journal, 4091 compat, 0, 4092 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 4093 incompat); 4094 } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 4095 ret = jbd2_journal_set_features(sbi->s_journal, 4096 compat, 0, 4097 incompat); 4098 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 4099 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 4100 } else { 4101 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 4102 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 4103 } 4104 4105 return ret; 4106 } 4107 4108 /* 4109 * Note: calculating the overhead so we can be compatible with 4110 * historical BSD practice is quite difficult in the face of 4111 * clusters/bigalloc. This is because multiple metadata blocks from 4112 * different block group can end up in the same allocation cluster. 4113 * Calculating the exact overhead in the face of clustered allocation 4114 * requires either O(all block bitmaps) in memory or O(number of block 4115 * groups**2) in time. We will still calculate the superblock for 4116 * older file systems --- and if we come across with a bigalloc file 4117 * system with zero in s_overhead_clusters the estimate will be close to 4118 * correct especially for very large cluster sizes --- but for newer 4119 * file systems, it's better to calculate this figure once at mkfs 4120 * time, and store it in the superblock. If the superblock value is 4121 * present (even for non-bigalloc file systems), we will use it. 4122 */ 4123 static int count_overhead(struct super_block *sb, ext4_group_t grp, 4124 char *buf) 4125 { 4126 struct ext4_sb_info *sbi = EXT4_SB(sb); 4127 struct ext4_group_desc *gdp; 4128 ext4_fsblk_t first_block, last_block, b; 4129 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 4130 int s, j, count = 0; 4131 int has_super = ext4_bg_has_super(sb, grp); 4132 4133 if (!ext4_has_feature_bigalloc(sb)) 4134 return (has_super + ext4_bg_num_gdb(sb, grp) + 4135 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) + 4136 sbi->s_itb_per_group + 2); 4137 4138 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 4139 (grp * EXT4_BLOCKS_PER_GROUP(sb)); 4140 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 4141 for (i = 0; i < ngroups; i++) { 4142 gdp = ext4_get_group_desc(sb, i, NULL); 4143 b = ext4_block_bitmap(sb, gdp); 4144 if (b >= first_block && b <= last_block) { 4145 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 4146 count++; 4147 } 4148 b = ext4_inode_bitmap(sb, gdp); 4149 if (b >= first_block && b <= last_block) { 4150 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 4151 count++; 4152 } 4153 b = ext4_inode_table(sb, gdp); 4154 if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 4155 for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 4156 int c = EXT4_B2C(sbi, b - first_block); 4157 ext4_set_bit(c, buf); 4158 count++; 4159 } 4160 if (i != grp) 4161 continue; 4162 s = 0; 4163 if (ext4_bg_has_super(sb, grp)) { 4164 ext4_set_bit(s++, buf); 4165 count++; 4166 } 4167 j = ext4_bg_num_gdb(sb, grp); 4168 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { 4169 ext4_error(sb, "Invalid number of block group " 4170 "descriptor blocks: %d", j); 4171 j = EXT4_BLOCKS_PER_GROUP(sb) - s; 4172 } 4173 count += j; 4174 for (; j > 0; j--) 4175 ext4_set_bit(EXT4_B2C(sbi, s++), buf); 4176 } 4177 if (!count) 4178 return 0; 4179 return EXT4_CLUSTERS_PER_GROUP(sb) - 4180 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 4181 } 4182 4183 /* 4184 * Compute the overhead and stash it in sbi->s_overhead 4185 */ 4186 int ext4_calculate_overhead(struct super_block *sb) 4187 { 4188 struct ext4_sb_info *sbi = EXT4_SB(sb); 4189 struct ext4_super_block *es = sbi->s_es; 4190 struct inode *j_inode; 4191 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); 4192 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 4193 ext4_fsblk_t overhead = 0; 4194 char *buf = (char *) get_zeroed_page(GFP_NOFS); 4195 4196 if (!buf) 4197 return -ENOMEM; 4198 4199 /* 4200 * Compute the overhead (FS structures). This is constant 4201 * for a given filesystem unless the number of block groups 4202 * changes so we cache the previous value until it does. 4203 */ 4204 4205 /* 4206 * All of the blocks before first_data_block are overhead 4207 */ 4208 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 4209 4210 /* 4211 * Add the overhead found in each block group 4212 */ 4213 for (i = 0; i < ngroups; i++) { 4214 int blks; 4215 4216 blks = count_overhead(sb, i, buf); 4217 overhead += blks; 4218 if (blks) 4219 memset(buf, 0, PAGE_SIZE); 4220 cond_resched(); 4221 } 4222 4223 /* 4224 * Add the internal journal blocks whether the journal has been 4225 * loaded or not 4226 */ 4227 if (sbi->s_journal && !sbi->s_journal_bdev_file) 4228 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len); 4229 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) { 4230 /* j_inum for internal journal is non-zero */ 4231 j_inode = ext4_get_journal_inode(sb, j_inum); 4232 if (!IS_ERR(j_inode)) { 4233 j_blocks = j_inode->i_size >> sb->s_blocksize_bits; 4234 overhead += EXT4_NUM_B2C(sbi, j_blocks); 4235 iput(j_inode); 4236 } else { 4237 ext4_msg(sb, KERN_ERR, "can't get journal size"); 4238 } 4239 } 4240 sbi->s_overhead = overhead; 4241 smp_wmb(); 4242 free_page((unsigned long) buf); 4243 return 0; 4244 } 4245 4246 static void ext4_set_resv_clusters(struct super_block *sb) 4247 { 4248 ext4_fsblk_t resv_clusters; 4249 struct ext4_sb_info *sbi = EXT4_SB(sb); 4250 4251 /* 4252 * There's no need to reserve anything when we aren't using extents. 4253 * The space estimates are exact, there are no unwritten extents, 4254 * hole punching doesn't need new metadata... This is needed especially 4255 * to keep ext2/3 backward compatibility. 4256 */ 4257 if (!ext4_has_feature_extents(sb)) 4258 return; 4259 /* 4260 * By default we reserve 2% or 4096 clusters, whichever is smaller. 4261 * This should cover the situations where we can not afford to run 4262 * out of space like for example punch hole, or converting 4263 * unwritten extents in delalloc path. In most cases such 4264 * allocation would require 1, or 2 blocks, higher numbers are 4265 * very rare. 4266 */ 4267 resv_clusters = (ext4_blocks_count(sbi->s_es) >> 4268 sbi->s_cluster_bits); 4269 4270 do_div(resv_clusters, 50); 4271 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 4272 4273 atomic64_set(&sbi->s_resv_clusters, resv_clusters); 4274 } 4275 4276 static const char *ext4_quota_mode(struct super_block *sb) 4277 { 4278 #ifdef CONFIG_QUOTA 4279 if (!ext4_quota_capable(sb)) 4280 return "none"; 4281 4282 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb)) 4283 return "journalled"; 4284 else 4285 return "writeback"; 4286 #else 4287 return "disabled"; 4288 #endif 4289 } 4290 4291 static void ext4_setup_csum_trigger(struct super_block *sb, 4292 enum ext4_journal_trigger_type type, 4293 void (*trigger)( 4294 struct jbd2_buffer_trigger_type *type, 4295 struct buffer_head *bh, 4296 void *mapped_data, 4297 size_t size)) 4298 { 4299 struct ext4_sb_info *sbi = EXT4_SB(sb); 4300 4301 sbi->s_journal_triggers[type].sb = sb; 4302 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger; 4303 } 4304 4305 static void ext4_free_sbi(struct ext4_sb_info *sbi) 4306 { 4307 if (!sbi) 4308 return; 4309 4310 kfree(sbi->s_blockgroup_lock); 4311 fs_put_dax(sbi->s_daxdev, NULL); 4312 kfree(sbi); 4313 } 4314 4315 static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb) 4316 { 4317 struct ext4_sb_info *sbi; 4318 4319 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 4320 if (!sbi) 4321 return NULL; 4322 4323 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off, 4324 NULL, NULL); 4325 4326 sbi->s_blockgroup_lock = 4327 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 4328 4329 if (!sbi->s_blockgroup_lock) 4330 goto err_out; 4331 4332 sb->s_fs_info = sbi; 4333 sbi->s_sb = sb; 4334 return sbi; 4335 err_out: 4336 fs_put_dax(sbi->s_daxdev, NULL); 4337 kfree(sbi); 4338 return NULL; 4339 } 4340 4341 static void ext4_set_def_opts(struct super_block *sb, 4342 struct ext4_super_block *es) 4343 { 4344 unsigned long def_mount_opts; 4345 4346 /* Set defaults before we parse the mount options */ 4347 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 4348 set_opt(sb, INIT_INODE_TABLE); 4349 if (def_mount_opts & EXT4_DEFM_DEBUG) 4350 set_opt(sb, DEBUG); 4351 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 4352 set_opt(sb, GRPID); 4353 if (def_mount_opts & EXT4_DEFM_UID16) 4354 set_opt(sb, NO_UID32); 4355 /* xattr user namespace & acls are now defaulted on */ 4356 set_opt(sb, XATTR_USER); 4357 #ifdef CONFIG_EXT4_FS_POSIX_ACL 4358 set_opt(sb, POSIX_ACL); 4359 #endif 4360 if (ext4_has_feature_fast_commit(sb)) 4361 set_opt2(sb, JOURNAL_FAST_COMMIT); 4362 /* don't forget to enable journal_csum when metadata_csum is enabled. */ 4363 if (ext4_has_feature_metadata_csum(sb)) 4364 set_opt(sb, JOURNAL_CHECKSUM); 4365 4366 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 4367 set_opt(sb, JOURNAL_DATA); 4368 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 4369 set_opt(sb, ORDERED_DATA); 4370 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 4371 set_opt(sb, WRITEBACK_DATA); 4372 4373 if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_PANIC) 4374 set_opt(sb, ERRORS_PANIC); 4375 else if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_CONTINUE) 4376 set_opt(sb, ERRORS_CONT); 4377 else 4378 set_opt(sb, ERRORS_RO); 4379 /* block_validity enabled by default; disable with noblock_validity */ 4380 set_opt(sb, BLOCK_VALIDITY); 4381 if (def_mount_opts & EXT4_DEFM_DISCARD) 4382 set_opt(sb, DISCARD); 4383 4384 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 4385 set_opt(sb, BARRIER); 4386 4387 /* 4388 * enable delayed allocation by default 4389 * Use -o nodelalloc to turn it off 4390 */ 4391 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 4392 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 4393 set_opt(sb, DELALLOC); 4394 4395 if (sb->s_blocksize <= PAGE_SIZE) 4396 set_opt(sb, DIOREAD_NOLOCK); 4397 } 4398 4399 static int ext4_handle_clustersize(struct super_block *sb) 4400 { 4401 struct ext4_sb_info *sbi = EXT4_SB(sb); 4402 struct ext4_super_block *es = sbi->s_es; 4403 int clustersize; 4404 4405 /* Handle clustersize */ 4406 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 4407 if (ext4_has_feature_bigalloc(sb)) { 4408 if (clustersize < sb->s_blocksize) { 4409 ext4_msg(sb, KERN_ERR, 4410 "cluster size (%d) smaller than " 4411 "block size (%lu)", clustersize, sb->s_blocksize); 4412 return -EINVAL; 4413 } 4414 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 4415 le32_to_cpu(es->s_log_block_size); 4416 } else { 4417 if (clustersize != sb->s_blocksize) { 4418 ext4_msg(sb, KERN_ERR, 4419 "fragment/cluster size (%d) != " 4420 "block size (%lu)", clustersize, sb->s_blocksize); 4421 return -EINVAL; 4422 } 4423 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) { 4424 ext4_msg(sb, KERN_ERR, 4425 "#blocks per group too big: %lu", 4426 sbi->s_blocks_per_group); 4427 return -EINVAL; 4428 } 4429 sbi->s_cluster_bits = 0; 4430 } 4431 sbi->s_clusters_per_group = le32_to_cpu(es->s_clusters_per_group); 4432 if (sbi->s_clusters_per_group > sb->s_blocksize * 8) { 4433 ext4_msg(sb, KERN_ERR, "#clusters per group too big: %lu", 4434 sbi->s_clusters_per_group); 4435 return -EINVAL; 4436 } 4437 if (sbi->s_blocks_per_group != 4438 (sbi->s_clusters_per_group * (clustersize / sb->s_blocksize))) { 4439 ext4_msg(sb, KERN_ERR, 4440 "blocks per group (%lu) and clusters per group (%lu) inconsistent", 4441 sbi->s_blocks_per_group, sbi->s_clusters_per_group); 4442 return -EINVAL; 4443 } 4444 sbi->s_cluster_ratio = clustersize / sb->s_blocksize; 4445 4446 /* Do we have standard group size of clustersize * 8 blocks ? */ 4447 if (sbi->s_blocks_per_group == clustersize << 3) 4448 set_opt2(sb, STD_GROUP_SIZE); 4449 4450 return 0; 4451 } 4452 4453 /* 4454 * ext4_atomic_write_init: Initializes filesystem min & max atomic write units. 4455 * With non-bigalloc filesystem awu will be based upon filesystem blocksize 4456 * & bdev awu units. 4457 * With bigalloc it will be based upon bigalloc cluster size & bdev awu units. 4458 * @sb: super block 4459 */ 4460 static void ext4_atomic_write_init(struct super_block *sb) 4461 { 4462 struct ext4_sb_info *sbi = EXT4_SB(sb); 4463 struct block_device *bdev = sb->s_bdev; 4464 unsigned int clustersize = EXT4_CLUSTER_SIZE(sb); 4465 4466 if (!bdev_can_atomic_write(bdev)) 4467 return; 4468 4469 if (!ext4_has_feature_extents(sb)) 4470 return; 4471 4472 sbi->s_awu_min = max(sb->s_blocksize, 4473 bdev_atomic_write_unit_min_bytes(bdev)); 4474 sbi->s_awu_max = min(clustersize, 4475 bdev_atomic_write_unit_max_bytes(bdev)); 4476 if (sbi->s_awu_min && sbi->s_awu_max && 4477 sbi->s_awu_min <= sbi->s_awu_max) { 4478 ext4_msg(sb, KERN_NOTICE, "Supports (experimental) DIO atomic writes awu_min: %u, awu_max: %u", 4479 sbi->s_awu_min, sbi->s_awu_max); 4480 } else { 4481 sbi->s_awu_min = 0; 4482 sbi->s_awu_max = 0; 4483 } 4484 } 4485 4486 static void ext4_fast_commit_init(struct super_block *sb) 4487 { 4488 struct ext4_sb_info *sbi = EXT4_SB(sb); 4489 4490 /* Initialize fast commit stuff */ 4491 atomic_set(&sbi->s_fc_subtid, 0); 4492 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]); 4493 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]); 4494 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]); 4495 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]); 4496 sbi->s_fc_bytes = 0; 4497 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE); 4498 sbi->s_fc_ineligible_tid = 0; 4499 mutex_init(&sbi->s_fc_lock); 4500 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats)); 4501 sbi->s_fc_replay_state.fc_regions = NULL; 4502 sbi->s_fc_replay_state.fc_regions_size = 0; 4503 sbi->s_fc_replay_state.fc_regions_used = 0; 4504 sbi->s_fc_replay_state.fc_regions_valid = 0; 4505 sbi->s_fc_replay_state.fc_modified_inodes = NULL; 4506 sbi->s_fc_replay_state.fc_modified_inodes_size = 0; 4507 sbi->s_fc_replay_state.fc_modified_inodes_used = 0; 4508 } 4509 4510 static int ext4_inode_info_init(struct super_block *sb, 4511 struct ext4_super_block *es) 4512 { 4513 struct ext4_sb_info *sbi = EXT4_SB(sb); 4514 4515 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 4516 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 4517 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 4518 } else { 4519 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 4520 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 4521 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) { 4522 ext4_msg(sb, KERN_ERR, "invalid first ino: %u", 4523 sbi->s_first_ino); 4524 return -EINVAL; 4525 } 4526 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 4527 (!is_power_of_2(sbi->s_inode_size)) || 4528 (sbi->s_inode_size > sb->s_blocksize)) { 4529 ext4_msg(sb, KERN_ERR, 4530 "unsupported inode size: %d", 4531 sbi->s_inode_size); 4532 ext4_msg(sb, KERN_ERR, "blocksize: %lu", sb->s_blocksize); 4533 return -EINVAL; 4534 } 4535 /* 4536 * i_atime_extra is the last extra field available for 4537 * [acm]times in struct ext4_inode. Checking for that 4538 * field should suffice to ensure we have extra space 4539 * for all three. 4540 */ 4541 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) + 4542 sizeof(((struct ext4_inode *)0)->i_atime_extra)) { 4543 sb->s_time_gran = 1; 4544 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX; 4545 } else { 4546 sb->s_time_gran = NSEC_PER_SEC; 4547 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX; 4548 } 4549 sb->s_time_min = EXT4_TIMESTAMP_MIN; 4550 } 4551 4552 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 4553 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 4554 EXT4_GOOD_OLD_INODE_SIZE; 4555 if (ext4_has_feature_extra_isize(sb)) { 4556 unsigned v, max = (sbi->s_inode_size - 4557 EXT4_GOOD_OLD_INODE_SIZE); 4558 4559 v = le16_to_cpu(es->s_want_extra_isize); 4560 if (v > max) { 4561 ext4_msg(sb, KERN_ERR, 4562 "bad s_want_extra_isize: %d", v); 4563 return -EINVAL; 4564 } 4565 if (sbi->s_want_extra_isize < v) 4566 sbi->s_want_extra_isize = v; 4567 4568 v = le16_to_cpu(es->s_min_extra_isize); 4569 if (v > max) { 4570 ext4_msg(sb, KERN_ERR, 4571 "bad s_min_extra_isize: %d", v); 4572 return -EINVAL; 4573 } 4574 if (sbi->s_want_extra_isize < v) 4575 sbi->s_want_extra_isize = v; 4576 } 4577 } 4578 4579 return 0; 4580 } 4581 4582 #if IS_ENABLED(CONFIG_UNICODE) 4583 static int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es) 4584 { 4585 const struct ext4_sb_encodings *encoding_info; 4586 struct unicode_map *encoding; 4587 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags); 4588 4589 if (!ext4_has_feature_casefold(sb) || sb->s_encoding) 4590 return 0; 4591 4592 encoding_info = ext4_sb_read_encoding(es); 4593 if (!encoding_info) { 4594 ext4_msg(sb, KERN_ERR, 4595 "Encoding requested by superblock is unknown"); 4596 return -EINVAL; 4597 } 4598 4599 encoding = utf8_load(encoding_info->version); 4600 if (IS_ERR(encoding)) { 4601 ext4_msg(sb, KERN_ERR, 4602 "can't mount with superblock charset: %s-%u.%u.%u " 4603 "not supported by the kernel. flags: 0x%x.", 4604 encoding_info->name, 4605 unicode_major(encoding_info->version), 4606 unicode_minor(encoding_info->version), 4607 unicode_rev(encoding_info->version), 4608 encoding_flags); 4609 return -EINVAL; 4610 } 4611 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: " 4612 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name, 4613 unicode_major(encoding_info->version), 4614 unicode_minor(encoding_info->version), 4615 unicode_rev(encoding_info->version), 4616 encoding_flags); 4617 4618 sb->s_encoding = encoding; 4619 sb->s_encoding_flags = encoding_flags; 4620 4621 return 0; 4622 } 4623 #else 4624 static inline int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es) 4625 { 4626 return 0; 4627 } 4628 #endif 4629 4630 static int ext4_init_metadata_csum(struct super_block *sb, struct ext4_super_block *es) 4631 { 4632 struct ext4_sb_info *sbi = EXT4_SB(sb); 4633 4634 /* Warn if metadata_csum and gdt_csum are both set. */ 4635 if (ext4_has_feature_metadata_csum(sb) && 4636 ext4_has_feature_gdt_csum(sb)) 4637 ext4_warning(sb, "metadata_csum and uninit_bg are " 4638 "redundant flags; please run fsck."); 4639 4640 /* Check for a known checksum algorithm */ 4641 if (!ext4_verify_csum_type(sb, es)) { 4642 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 4643 "unknown checksum algorithm."); 4644 return -EINVAL; 4645 } 4646 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE, 4647 ext4_orphan_file_block_trigger); 4648 4649 /* Check superblock checksum */ 4650 if (!ext4_superblock_csum_verify(sb, es)) { 4651 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 4652 "invalid superblock checksum. Run e2fsck?"); 4653 return -EFSBADCRC; 4654 } 4655 4656 /* Precompute checksum seed for all metadata */ 4657 if (ext4_has_feature_csum_seed(sb)) 4658 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 4659 else if (ext4_has_feature_metadata_csum(sb) || 4660 ext4_has_feature_ea_inode(sb)) 4661 sbi->s_csum_seed = ext4_chksum(~0, es->s_uuid, 4662 sizeof(es->s_uuid)); 4663 return 0; 4664 } 4665 4666 static int ext4_check_feature_compatibility(struct super_block *sb, 4667 struct ext4_super_block *es, 4668 int silent) 4669 { 4670 struct ext4_sb_info *sbi = EXT4_SB(sb); 4671 4672 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 4673 (ext4_has_compat_features(sb) || 4674 ext4_has_ro_compat_features(sb) || 4675 ext4_has_incompat_features(sb))) 4676 ext4_msg(sb, KERN_WARNING, 4677 "feature flags set on rev 0 fs, " 4678 "running e2fsck is recommended"); 4679 4680 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 4681 set_opt2(sb, HURD_COMPAT); 4682 if (ext4_has_feature_64bit(sb)) { 4683 ext4_msg(sb, KERN_ERR, 4684 "The Hurd can't support 64-bit file systems"); 4685 return -EINVAL; 4686 } 4687 4688 /* 4689 * ea_inode feature uses l_i_version field which is not 4690 * available in HURD_COMPAT mode. 4691 */ 4692 if (ext4_has_feature_ea_inode(sb)) { 4693 ext4_msg(sb, KERN_ERR, 4694 "ea_inode feature is not supported for Hurd"); 4695 return -EINVAL; 4696 } 4697 } 4698 4699 if (IS_EXT2_SB(sb)) { 4700 if (ext2_feature_set_ok(sb)) 4701 ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 4702 "using the ext4 subsystem"); 4703 else { 4704 /* 4705 * If we're probing be silent, if this looks like 4706 * it's actually an ext[34] filesystem. 4707 */ 4708 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 4709 return -EINVAL; 4710 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 4711 "to feature incompatibilities"); 4712 return -EINVAL; 4713 } 4714 } 4715 4716 if (IS_EXT3_SB(sb)) { 4717 if (ext3_feature_set_ok(sb)) 4718 ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 4719 "using the ext4 subsystem"); 4720 else { 4721 /* 4722 * If we're probing be silent, if this looks like 4723 * it's actually an ext4 filesystem. 4724 */ 4725 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 4726 return -EINVAL; 4727 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 4728 "to feature incompatibilities"); 4729 return -EINVAL; 4730 } 4731 } 4732 4733 /* 4734 * Check feature flags regardless of the revision level, since we 4735 * previously didn't change the revision level when setting the flags, 4736 * so there is a chance incompat flags are set on a rev 0 filesystem. 4737 */ 4738 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) 4739 return -EINVAL; 4740 4741 if (sbi->s_daxdev) { 4742 if (sb->s_blocksize == PAGE_SIZE) 4743 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags); 4744 else 4745 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n"); 4746 } 4747 4748 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) { 4749 if (ext4_has_feature_inline_data(sb)) { 4750 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem" 4751 " that may contain inline data"); 4752 return -EINVAL; 4753 } 4754 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) { 4755 ext4_msg(sb, KERN_ERR, 4756 "DAX unsupported by block device."); 4757 return -EINVAL; 4758 } 4759 } 4760 4761 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 4762 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 4763 es->s_encryption_level); 4764 return -EINVAL; 4765 } 4766 4767 return 0; 4768 } 4769 4770 static int ext4_check_geometry(struct super_block *sb, 4771 struct ext4_super_block *es) 4772 { 4773 struct ext4_sb_info *sbi = EXT4_SB(sb); 4774 __u64 blocks_count; 4775 int err; 4776 4777 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (sb->s_blocksize / 4)) { 4778 ext4_msg(sb, KERN_ERR, 4779 "Number of reserved GDT blocks insanely large: %d", 4780 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); 4781 return -EINVAL; 4782 } 4783 /* 4784 * Test whether we have more sectors than will fit in sector_t, 4785 * and whether the max offset is addressable by the page cache. 4786 */ 4787 err = generic_check_addressable(sb->s_blocksize_bits, 4788 ext4_blocks_count(es)); 4789 if (err) { 4790 ext4_msg(sb, KERN_ERR, "filesystem" 4791 " too large to mount safely on this system"); 4792 return err; 4793 } 4794 4795 /* check blocks count against device size */ 4796 blocks_count = sb_bdev_nr_blocks(sb); 4797 if (blocks_count && ext4_blocks_count(es) > blocks_count) { 4798 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 4799 "exceeds size of device (%llu blocks)", 4800 ext4_blocks_count(es), blocks_count); 4801 return -EINVAL; 4802 } 4803 4804 /* 4805 * It makes no sense for the first data block to be beyond the end 4806 * of the filesystem. 4807 */ 4808 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 4809 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4810 "block %u is beyond end of filesystem (%llu)", 4811 le32_to_cpu(es->s_first_data_block), 4812 ext4_blocks_count(es)); 4813 return -EINVAL; 4814 } 4815 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) && 4816 (sbi->s_cluster_ratio == 1)) { 4817 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4818 "block is 0 with a 1k block and cluster size"); 4819 return -EINVAL; 4820 } 4821 4822 blocks_count = (ext4_blocks_count(es) - 4823 le32_to_cpu(es->s_first_data_block) + 4824 EXT4_BLOCKS_PER_GROUP(sb) - 1); 4825 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 4826 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 4827 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu " 4828 "(block count %llu, first data block %u, " 4829 "blocks per group %lu)", blocks_count, 4830 ext4_blocks_count(es), 4831 le32_to_cpu(es->s_first_data_block), 4832 EXT4_BLOCKS_PER_GROUP(sb)); 4833 return -EINVAL; 4834 } 4835 sbi->s_groups_count = blocks_count; 4836 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 4837 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 4838 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) != 4839 le32_to_cpu(es->s_inodes_count)) { 4840 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu", 4841 le32_to_cpu(es->s_inodes_count), 4842 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group)); 4843 return -EINVAL; 4844 } 4845 4846 return 0; 4847 } 4848 4849 static int ext4_group_desc_init(struct super_block *sb, 4850 struct ext4_super_block *es, 4851 ext4_fsblk_t logical_sb_block, 4852 ext4_group_t *first_not_zeroed) 4853 { 4854 struct ext4_sb_info *sbi = EXT4_SB(sb); 4855 unsigned int db_count; 4856 ext4_fsblk_t block; 4857 int i; 4858 4859 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 4860 EXT4_DESC_PER_BLOCK(sb); 4861 if (ext4_has_feature_meta_bg(sb)) { 4862 if (le32_to_cpu(es->s_first_meta_bg) > db_count) { 4863 ext4_msg(sb, KERN_WARNING, 4864 "first meta block group too large: %u " 4865 "(group descriptor block count %u)", 4866 le32_to_cpu(es->s_first_meta_bg), db_count); 4867 return -EINVAL; 4868 } 4869 } 4870 rcu_assign_pointer(sbi->s_group_desc, 4871 kvmalloc_array(db_count, 4872 sizeof(struct buffer_head *), 4873 GFP_KERNEL)); 4874 if (sbi->s_group_desc == NULL) { 4875 ext4_msg(sb, KERN_ERR, "not enough memory"); 4876 return -ENOMEM; 4877 } 4878 4879 bgl_lock_init(sbi->s_blockgroup_lock); 4880 4881 /* Pre-read the descriptors into the buffer cache */ 4882 for (i = 0; i < db_count; i++) { 4883 block = descriptor_loc(sb, logical_sb_block, i); 4884 ext4_sb_breadahead_unmovable(sb, block); 4885 } 4886 4887 for (i = 0; i < db_count; i++) { 4888 struct buffer_head *bh; 4889 4890 block = descriptor_loc(sb, logical_sb_block, i); 4891 bh = ext4_sb_bread_unmovable(sb, block); 4892 if (IS_ERR(bh)) { 4893 ext4_msg(sb, KERN_ERR, 4894 "can't read group descriptor %d", i); 4895 sbi->s_gdb_count = i; 4896 return PTR_ERR(bh); 4897 } 4898 rcu_read_lock(); 4899 rcu_dereference(sbi->s_group_desc)[i] = bh; 4900 rcu_read_unlock(); 4901 } 4902 sbi->s_gdb_count = db_count; 4903 if (!ext4_check_descriptors(sb, logical_sb_block, first_not_zeroed)) { 4904 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 4905 return -EFSCORRUPTED; 4906 } 4907 4908 return 0; 4909 } 4910 4911 static int ext4_load_and_init_journal(struct super_block *sb, 4912 struct ext4_super_block *es, 4913 struct ext4_fs_context *ctx) 4914 { 4915 struct ext4_sb_info *sbi = EXT4_SB(sb); 4916 int err; 4917 4918 err = ext4_load_journal(sb, es, ctx->journal_devnum); 4919 if (err) 4920 return err; 4921 4922 if (ext4_has_feature_64bit(sb) && 4923 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4924 JBD2_FEATURE_INCOMPAT_64BIT)) { 4925 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 4926 goto out; 4927 } 4928 4929 if (!set_journal_csum_feature_set(sb)) { 4930 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 4931 "feature set"); 4932 goto out; 4933 } 4934 4935 if (test_opt2(sb, JOURNAL_FAST_COMMIT) && 4936 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4937 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) { 4938 ext4_msg(sb, KERN_ERR, 4939 "Failed to set fast commit journal feature"); 4940 goto out; 4941 } 4942 4943 /* We have now updated the journal if required, so we can 4944 * validate the data journaling mode. */ 4945 switch (test_opt(sb, DATA_FLAGS)) { 4946 case 0: 4947 /* No mode set, assume a default based on the journal 4948 * capabilities: ORDERED_DATA if the journal can 4949 * cope, else JOURNAL_DATA 4950 */ 4951 if (jbd2_journal_check_available_features 4952 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4953 set_opt(sb, ORDERED_DATA); 4954 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA; 4955 } else { 4956 set_opt(sb, JOURNAL_DATA); 4957 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; 4958 } 4959 break; 4960 4961 case EXT4_MOUNT_ORDERED_DATA: 4962 case EXT4_MOUNT_WRITEBACK_DATA: 4963 if (!jbd2_journal_check_available_features 4964 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4965 ext4_msg(sb, KERN_ERR, "Journal does not support " 4966 "requested data journaling mode"); 4967 goto out; 4968 } 4969 break; 4970 default: 4971 break; 4972 } 4973 4974 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 4975 test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4976 ext4_msg(sb, KERN_ERR, "can't mount with " 4977 "journal_async_commit in data=ordered mode"); 4978 goto out; 4979 } 4980 4981 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio); 4982 4983 sbi->s_journal->j_submit_inode_data_buffers = 4984 ext4_journal_submit_inode_data_buffers; 4985 sbi->s_journal->j_finish_inode_data_buffers = 4986 ext4_journal_finish_inode_data_buffers; 4987 4988 return 0; 4989 4990 out: 4991 ext4_journal_destroy(sbi, sbi->s_journal); 4992 return -EINVAL; 4993 } 4994 4995 static int ext4_check_journal_data_mode(struct super_block *sb) 4996 { 4997 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 4998 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with " 4999 "data=journal disables delayed allocation, " 5000 "dioread_nolock, O_DIRECT and fast_commit support!\n"); 5001 /* can't mount with both data=journal and dioread_nolock. */ 5002 clear_opt(sb, DIOREAD_NOLOCK); 5003 clear_opt2(sb, JOURNAL_FAST_COMMIT); 5004 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 5005 ext4_msg(sb, KERN_ERR, "can't mount with " 5006 "both data=journal and delalloc"); 5007 return -EINVAL; 5008 } 5009 if (test_opt(sb, DAX_ALWAYS)) { 5010 ext4_msg(sb, KERN_ERR, "can't mount with " 5011 "both data=journal and dax"); 5012 return -EINVAL; 5013 } 5014 if (ext4_has_feature_encrypt(sb)) { 5015 ext4_msg(sb, KERN_WARNING, 5016 "encrypted files will use data=ordered " 5017 "instead of data journaling mode"); 5018 } 5019 if (test_opt(sb, DELALLOC)) 5020 clear_opt(sb, DELALLOC); 5021 } else { 5022 sb->s_iflags |= SB_I_CGROUPWB; 5023 } 5024 5025 return 0; 5026 } 5027 5028 static const char *ext4_has_journal_option(struct super_block *sb) 5029 { 5030 struct ext4_sb_info *sbi = EXT4_SB(sb); 5031 5032 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) 5033 return "journal_async_commit"; 5034 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) 5035 return "journal_checksum"; 5036 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 5037 return "commit="; 5038 if (EXT4_MOUNT_DATA_FLAGS & 5039 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) 5040 return "data="; 5041 if (test_opt(sb, DATA_ERR_ABORT)) 5042 return "data_err=abort"; 5043 return NULL; 5044 } 5045 5046 static int ext4_load_super(struct super_block *sb, ext4_fsblk_t *lsb, 5047 int silent) 5048 { 5049 struct ext4_sb_info *sbi = EXT4_SB(sb); 5050 struct ext4_super_block *es; 5051 ext4_fsblk_t logical_sb_block; 5052 unsigned long offset = 0; 5053 struct buffer_head *bh; 5054 int ret = -EINVAL; 5055 int blocksize; 5056 5057 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 5058 if (!blocksize) { 5059 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 5060 return -EINVAL; 5061 } 5062 5063 /* 5064 * The ext4 superblock will not be buffer aligned for other than 1kB 5065 * block sizes. We need to calculate the offset from buffer start. 5066 */ 5067 if (blocksize != EXT4_MIN_BLOCK_SIZE) { 5068 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE; 5069 offset = do_div(logical_sb_block, blocksize); 5070 } else { 5071 logical_sb_block = sbi->s_sb_block; 5072 } 5073 5074 bh = ext4_sb_bread_unmovable(sb, logical_sb_block); 5075 if (IS_ERR(bh)) { 5076 ext4_msg(sb, KERN_ERR, "unable to read superblock"); 5077 return PTR_ERR(bh); 5078 } 5079 /* 5080 * Note: s_es must be initialized as soon as possible because 5081 * some ext4 macro-instructions depend on its value 5082 */ 5083 es = (struct ext4_super_block *) (bh->b_data + offset); 5084 sbi->s_es = es; 5085 sb->s_magic = le16_to_cpu(es->s_magic); 5086 if (sb->s_magic != EXT4_SUPER_MAGIC) { 5087 if (!silent) 5088 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 5089 goto out; 5090 } 5091 5092 if (le32_to_cpu(es->s_log_block_size) > 5093 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 5094 ext4_msg(sb, KERN_ERR, 5095 "Invalid log block size: %u", 5096 le32_to_cpu(es->s_log_block_size)); 5097 goto out; 5098 } 5099 if (le32_to_cpu(es->s_log_cluster_size) > 5100 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 5101 ext4_msg(sb, KERN_ERR, 5102 "Invalid log cluster size: %u", 5103 le32_to_cpu(es->s_log_cluster_size)); 5104 goto out; 5105 } 5106 5107 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 5108 5109 /* 5110 * If the default block size is not the same as the real block size, 5111 * we need to reload it. 5112 */ 5113 if (sb->s_blocksize == blocksize) { 5114 *lsb = logical_sb_block; 5115 sbi->s_sbh = bh; 5116 return 0; 5117 } 5118 5119 /* 5120 * bh must be released before kill_bdev(), otherwise 5121 * it won't be freed and its page also. kill_bdev() 5122 * is called by sb_set_blocksize(). 5123 */ 5124 brelse(bh); 5125 /* Validate the filesystem blocksize */ 5126 if (!sb_set_blocksize(sb, blocksize)) { 5127 ext4_msg(sb, KERN_ERR, "bad block size %d", 5128 blocksize); 5129 bh = NULL; 5130 goto out; 5131 } 5132 5133 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE; 5134 offset = do_div(logical_sb_block, blocksize); 5135 bh = ext4_sb_bread_unmovable(sb, logical_sb_block); 5136 if (IS_ERR(bh)) { 5137 ext4_msg(sb, KERN_ERR, "Can't read superblock on 2nd try"); 5138 ret = PTR_ERR(bh); 5139 bh = NULL; 5140 goto out; 5141 } 5142 es = (struct ext4_super_block *)(bh->b_data + offset); 5143 sbi->s_es = es; 5144 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 5145 ext4_msg(sb, KERN_ERR, "Magic mismatch, very weird!"); 5146 goto out; 5147 } 5148 *lsb = logical_sb_block; 5149 sbi->s_sbh = bh; 5150 return 0; 5151 out: 5152 brelse(bh); 5153 return ret; 5154 } 5155 5156 static int ext4_hash_info_init(struct super_block *sb) 5157 { 5158 struct ext4_sb_info *sbi = EXT4_SB(sb); 5159 struct ext4_super_block *es = sbi->s_es; 5160 unsigned int i; 5161 5162 sbi->s_def_hash_version = es->s_def_hash_version; 5163 5164 if (sbi->s_def_hash_version > DX_HASH_LAST) { 5165 ext4_msg(sb, KERN_ERR, 5166 "Invalid default hash set in the superblock"); 5167 return -EINVAL; 5168 } else if (sbi->s_def_hash_version == DX_HASH_SIPHASH) { 5169 ext4_msg(sb, KERN_ERR, 5170 "SIPHASH is not a valid default hash value"); 5171 return -EINVAL; 5172 } 5173 5174 for (i = 0; i < 4; i++) 5175 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 5176 5177 if (ext4_has_feature_dir_index(sb)) { 5178 i = le32_to_cpu(es->s_flags); 5179 if (i & EXT2_FLAGS_UNSIGNED_HASH) 5180 sbi->s_hash_unsigned = 3; 5181 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 5182 #ifdef __CHAR_UNSIGNED__ 5183 if (!sb_rdonly(sb)) 5184 es->s_flags |= 5185 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 5186 sbi->s_hash_unsigned = 3; 5187 #else 5188 if (!sb_rdonly(sb)) 5189 es->s_flags |= 5190 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 5191 #endif 5192 } 5193 } 5194 return 0; 5195 } 5196 5197 static int ext4_block_group_meta_init(struct super_block *sb, int silent) 5198 { 5199 struct ext4_sb_info *sbi = EXT4_SB(sb); 5200 struct ext4_super_block *es = sbi->s_es; 5201 int has_huge_files; 5202 5203 has_huge_files = ext4_has_feature_huge_file(sb); 5204 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 5205 has_huge_files); 5206 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 5207 5208 sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 5209 if (ext4_has_feature_64bit(sb)) { 5210 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 5211 sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 5212 !is_power_of_2(sbi->s_desc_size)) { 5213 ext4_msg(sb, KERN_ERR, 5214 "unsupported descriptor size %lu", 5215 sbi->s_desc_size); 5216 return -EINVAL; 5217 } 5218 } else 5219 sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 5220 5221 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 5222 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 5223 5224 sbi->s_inodes_per_block = sb->s_blocksize / EXT4_INODE_SIZE(sb); 5225 if (sbi->s_inodes_per_block == 0 || sbi->s_blocks_per_group == 0) { 5226 if (!silent) 5227 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 5228 return -EINVAL; 5229 } 5230 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || 5231 sbi->s_inodes_per_group > sb->s_blocksize * 8) { 5232 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", 5233 sbi->s_inodes_per_group); 5234 return -EINVAL; 5235 } 5236 sbi->s_itb_per_group = sbi->s_inodes_per_group / 5237 sbi->s_inodes_per_block; 5238 sbi->s_desc_per_block = sb->s_blocksize / EXT4_DESC_SIZE(sb); 5239 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY; 5240 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 5241 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 5242 5243 return 0; 5244 } 5245 5246 /* 5247 * It's hard to get stripe aligned blocks if stripe is not aligned with 5248 * cluster, just disable stripe and alert user to simplify code and avoid 5249 * stripe aligned allocation which will rarely succeed. 5250 */ 5251 static bool ext4_is_stripe_incompatible(struct super_block *sb, unsigned long stripe) 5252 { 5253 struct ext4_sb_info *sbi = EXT4_SB(sb); 5254 return (stripe > 0 && sbi->s_cluster_ratio > 1 && 5255 stripe % sbi->s_cluster_ratio != 0); 5256 } 5257 5258 static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb) 5259 { 5260 struct ext4_super_block *es = NULL; 5261 struct ext4_sb_info *sbi = EXT4_SB(sb); 5262 ext4_fsblk_t logical_sb_block; 5263 struct inode *root; 5264 int needs_recovery; 5265 int err; 5266 ext4_group_t first_not_zeroed; 5267 struct ext4_fs_context *ctx = fc->fs_private; 5268 int silent = fc->sb_flags & SB_SILENT; 5269 5270 /* Set defaults for the variables that will be set during parsing */ 5271 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) 5272 ctx->journal_ioprio = EXT4_DEF_JOURNAL_IOPRIO; 5273 5274 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 5275 sbi->s_sectors_written_start = 5276 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]); 5277 5278 err = ext4_load_super(sb, &logical_sb_block, silent); 5279 if (err) 5280 goto out_fail; 5281 5282 es = sbi->s_es; 5283 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 5284 5285 err = ext4_init_metadata_csum(sb, es); 5286 if (err) 5287 goto failed_mount; 5288 5289 ext4_set_def_opts(sb, es); 5290 5291 sbi->s_resuid = make_kuid(&init_user_ns, ext4_get_resuid(es)); 5292 sbi->s_resgid = make_kgid(&init_user_ns, ext4_get_resuid(es)); 5293 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 5294 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 5295 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 5296 sbi->s_sb_update_kb = EXT4_DEF_SB_UPDATE_INTERVAL_KB; 5297 sbi->s_sb_update_sec = EXT4_DEF_SB_UPDATE_INTERVAL_SEC; 5298 5299 /* 5300 * set default s_li_wait_mult for lazyinit, for the case there is 5301 * no mount option specified. 5302 */ 5303 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 5304 5305 err = ext4_inode_info_init(sb, es); 5306 if (err) 5307 goto failed_mount; 5308 5309 err = parse_apply_sb_mount_options(sb, ctx); 5310 if (err < 0) 5311 goto failed_mount; 5312 5313 sbi->s_def_mount_opt = sbi->s_mount_opt; 5314 sbi->s_def_mount_opt2 = sbi->s_mount_opt2; 5315 5316 err = ext4_check_opt_consistency(fc, sb); 5317 if (err < 0) 5318 goto failed_mount; 5319 5320 ext4_apply_options(fc, sb); 5321 5322 err = ext4_encoding_init(sb, es); 5323 if (err) 5324 goto failed_mount; 5325 5326 err = ext4_check_journal_data_mode(sb); 5327 if (err) 5328 goto failed_mount; 5329 5330 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 5331 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 5332 5333 /* HSM events are allowed by default. */ 5334 sb->s_iflags |= SB_I_ALLOW_HSM; 5335 5336 err = ext4_check_feature_compatibility(sb, es, silent); 5337 if (err) 5338 goto failed_mount; 5339 5340 err = ext4_block_group_meta_init(sb, silent); 5341 if (err) 5342 goto failed_mount; 5343 5344 err = ext4_hash_info_init(sb); 5345 if (err) 5346 goto failed_mount; 5347 5348 err = ext4_handle_clustersize(sb); 5349 if (err) 5350 goto failed_mount; 5351 5352 err = ext4_check_geometry(sb, es); 5353 if (err) 5354 goto failed_mount; 5355 5356 timer_setup(&sbi->s_err_report, print_daily_error_info, 0); 5357 spin_lock_init(&sbi->s_error_lock); 5358 INIT_WORK(&sbi->s_sb_upd_work, update_super_work); 5359 5360 err = ext4_group_desc_init(sb, es, logical_sb_block, &first_not_zeroed); 5361 if (err) 5362 goto failed_mount3; 5363 5364 err = ext4_es_register_shrinker(sbi); 5365 if (err) 5366 goto failed_mount3; 5367 5368 sbi->s_stripe = ext4_get_stripe_size(sbi); 5369 if (ext4_is_stripe_incompatible(sb, sbi->s_stripe)) { 5370 ext4_msg(sb, KERN_WARNING, 5371 "stripe (%lu) is not aligned with cluster size (%u), " 5372 "stripe is disabled", 5373 sbi->s_stripe, sbi->s_cluster_ratio); 5374 sbi->s_stripe = 0; 5375 } 5376 sbi->s_extent_max_zeroout_kb = 32; 5377 5378 /* 5379 * set up enough so that it can read an inode 5380 */ 5381 sb->s_op = &ext4_sops; 5382 sb->s_export_op = &ext4_export_ops; 5383 sb->s_xattr = ext4_xattr_handlers; 5384 #ifdef CONFIG_FS_ENCRYPTION 5385 sb->s_cop = &ext4_cryptops; 5386 #endif 5387 #ifdef CONFIG_FS_VERITY 5388 sb->s_vop = &ext4_verityops; 5389 #endif 5390 #ifdef CONFIG_QUOTA 5391 sb->dq_op = &ext4_quota_operations; 5392 if (ext4_has_feature_quota(sb)) 5393 sb->s_qcop = &dquot_quotactl_sysfile_ops; 5394 else 5395 sb->s_qcop = &ext4_qctl_operations; 5396 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 5397 #endif 5398 super_set_uuid(sb, es->s_uuid, sizeof(es->s_uuid)); 5399 super_set_sysfs_name_bdev(sb); 5400 5401 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 5402 mutex_init(&sbi->s_orphan_lock); 5403 5404 spin_lock_init(&sbi->s_bdev_wb_lock); 5405 5406 ext4_atomic_write_init(sb); 5407 ext4_fast_commit_init(sb); 5408 5409 sb->s_root = NULL; 5410 5411 needs_recovery = (es->s_last_orphan != 0 || 5412 ext4_has_feature_orphan_present(sb) || 5413 ext4_has_feature_journal_needs_recovery(sb)); 5414 5415 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) { 5416 err = ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)); 5417 if (err) 5418 goto failed_mount3a; 5419 } 5420 5421 err = -EINVAL; 5422 /* 5423 * The first inode we look at is the journal inode. Don't try 5424 * root first: it may be modified in the journal! 5425 */ 5426 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 5427 err = ext4_load_and_init_journal(sb, es, ctx); 5428 if (err) 5429 goto failed_mount3a; 5430 if (bdev_read_only(sb->s_bdev)) 5431 needs_recovery = 0; 5432 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && 5433 ext4_has_feature_journal_needs_recovery(sb)) { 5434 ext4_msg(sb, KERN_ERR, "required journal recovery " 5435 "suppressed and not mounted read-only"); 5436 goto failed_mount3a; 5437 } else { 5438 const char *journal_option; 5439 5440 /* Nojournal mode, all journal mount options are illegal */ 5441 journal_option = ext4_has_journal_option(sb); 5442 if (journal_option != NULL) { 5443 ext4_msg(sb, KERN_ERR, 5444 "can't mount with %s, fs mounted w/o journal", 5445 journal_option); 5446 goto failed_mount3a; 5447 } 5448 5449 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM; 5450 clear_opt(sb, JOURNAL_CHECKSUM); 5451 clear_opt(sb, DATA_FLAGS); 5452 clear_opt2(sb, JOURNAL_FAST_COMMIT); 5453 sbi->s_journal = NULL; 5454 needs_recovery = 0; 5455 } 5456 5457 if (!test_opt(sb, NO_MBCACHE)) { 5458 sbi->s_ea_block_cache = ext4_xattr_create_cache(); 5459 if (!sbi->s_ea_block_cache) { 5460 ext4_msg(sb, KERN_ERR, 5461 "Failed to create ea_block_cache"); 5462 err = -EINVAL; 5463 goto failed_mount_wq; 5464 } 5465 5466 if (ext4_has_feature_ea_inode(sb)) { 5467 sbi->s_ea_inode_cache = ext4_xattr_create_cache(); 5468 if (!sbi->s_ea_inode_cache) { 5469 ext4_msg(sb, KERN_ERR, 5470 "Failed to create ea_inode_cache"); 5471 err = -EINVAL; 5472 goto failed_mount_wq; 5473 } 5474 } 5475 } 5476 5477 /* 5478 * Get the # of file system overhead blocks from the 5479 * superblock if present. 5480 */ 5481 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 5482 /* ignore the precalculated value if it is ridiculous */ 5483 if (sbi->s_overhead > ext4_blocks_count(es)) 5484 sbi->s_overhead = 0; 5485 /* 5486 * If the bigalloc feature is not enabled recalculating the 5487 * overhead doesn't take long, so we might as well just redo 5488 * it to make sure we are using the correct value. 5489 */ 5490 if (!ext4_has_feature_bigalloc(sb)) 5491 sbi->s_overhead = 0; 5492 if (sbi->s_overhead == 0) { 5493 err = ext4_calculate_overhead(sb); 5494 if (err) 5495 goto failed_mount_wq; 5496 } 5497 5498 /* 5499 * The maximum number of concurrent works can be high and 5500 * concurrency isn't really necessary. Limit it to 1. 5501 */ 5502 EXT4_SB(sb)->rsv_conversion_wq = 5503 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 5504 if (!EXT4_SB(sb)->rsv_conversion_wq) { 5505 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 5506 err = -ENOMEM; 5507 goto failed_mount4; 5508 } 5509 5510 /* 5511 * The jbd2_journal_load will have done any necessary log recovery, 5512 * so we can safely mount the rest of the filesystem now. 5513 */ 5514 5515 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL); 5516 if (IS_ERR(root)) { 5517 ext4_msg(sb, KERN_ERR, "get root inode failed"); 5518 err = PTR_ERR(root); 5519 root = NULL; 5520 goto failed_mount4; 5521 } 5522 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 5523 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 5524 iput(root); 5525 err = -EFSCORRUPTED; 5526 goto failed_mount4; 5527 } 5528 5529 generic_set_sb_d_ops(sb); 5530 sb->s_root = d_make_root(root); 5531 if (!sb->s_root) { 5532 ext4_msg(sb, KERN_ERR, "get root dentry failed"); 5533 err = -ENOMEM; 5534 goto failed_mount4; 5535 } 5536 5537 err = ext4_setup_super(sb, es, sb_rdonly(sb)); 5538 if (err == -EROFS) { 5539 sb->s_flags |= SB_RDONLY; 5540 } else if (err) 5541 goto failed_mount4a; 5542 5543 ext4_set_resv_clusters(sb); 5544 5545 if (test_opt(sb, BLOCK_VALIDITY)) { 5546 err = ext4_setup_system_zone(sb); 5547 if (err) { 5548 ext4_msg(sb, KERN_ERR, "failed to initialize system " 5549 "zone (%d)", err); 5550 goto failed_mount4a; 5551 } 5552 } 5553 ext4_fc_replay_cleanup(sb); 5554 5555 ext4_ext_init(sb); 5556 5557 /* 5558 * Enable optimize_scan if number of groups is > threshold. This can be 5559 * turned off by passing "mb_optimize_scan=0". This can also be 5560 * turned on forcefully by passing "mb_optimize_scan=1". 5561 */ 5562 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) { 5563 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD) 5564 set_opt2(sb, MB_OPTIMIZE_SCAN); 5565 else 5566 clear_opt2(sb, MB_OPTIMIZE_SCAN); 5567 } 5568 5569 err = ext4_mb_init(sb); 5570 if (err) { 5571 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 5572 err); 5573 goto failed_mount5; 5574 } 5575 5576 /* 5577 * We can only set up the journal commit callback once 5578 * mballoc is initialized 5579 */ 5580 if (sbi->s_journal) 5581 sbi->s_journal->j_commit_callback = 5582 ext4_journal_commit_callback; 5583 5584 err = ext4_percpu_param_init(sbi); 5585 if (err) 5586 goto failed_mount6; 5587 5588 if (ext4_has_feature_flex_bg(sb)) 5589 if (!ext4_fill_flex_info(sb)) { 5590 ext4_msg(sb, KERN_ERR, 5591 "unable to initialize " 5592 "flex_bg meta info!"); 5593 err = -ENOMEM; 5594 goto failed_mount6; 5595 } 5596 5597 err = ext4_register_li_request(sb, first_not_zeroed); 5598 if (err) 5599 goto failed_mount6; 5600 5601 err = ext4_init_orphan_info(sb); 5602 if (err) 5603 goto failed_mount7; 5604 #ifdef CONFIG_QUOTA 5605 /* Enable quota usage during mount. */ 5606 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { 5607 err = ext4_enable_quotas(sb); 5608 if (err) 5609 goto failed_mount8; 5610 } 5611 #endif /* CONFIG_QUOTA */ 5612 5613 /* 5614 * Save the original bdev mapping's wb_err value which could be 5615 * used to detect the metadata async write error. 5616 */ 5617 errseq_check_and_advance(&sb->s_bdev->bd_mapping->wb_err, 5618 &sbi->s_bdev_wb_err); 5619 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 5620 ext4_orphan_cleanup(sb, es); 5621 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 5622 /* 5623 * Update the checksum after updating free space/inode counters and 5624 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect 5625 * checksum in the buffer cache until it is written out and 5626 * e2fsprogs programs trying to open a file system immediately 5627 * after it is mounted can fail. 5628 */ 5629 ext4_superblock_csum_set(sb); 5630 if (needs_recovery) { 5631 ext4_msg(sb, KERN_INFO, "recovery complete"); 5632 err = ext4_mark_recovery_complete(sb, es); 5633 if (err) 5634 goto failed_mount9; 5635 } 5636 5637 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev)) { 5638 ext4_msg(sb, KERN_WARNING, 5639 "mounting with \"discard\" option, but the device does not support discard"); 5640 clear_opt(sb, DISCARD); 5641 } 5642 5643 if (es->s_error_count) 5644 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 5645 5646 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 5647 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 5648 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 5649 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 5650 atomic_set(&sbi->s_warning_count, 0); 5651 atomic_set(&sbi->s_msg_count, 0); 5652 5653 /* Register sysfs after all initializations are complete. */ 5654 err = ext4_register_sysfs(sb); 5655 if (err) 5656 goto failed_mount9; 5657 5658 return 0; 5659 5660 failed_mount9: 5661 ext4_quotas_off(sb, EXT4_MAXQUOTAS); 5662 failed_mount8: __maybe_unused 5663 ext4_release_orphan_info(sb); 5664 failed_mount7: 5665 ext4_unregister_li_request(sb); 5666 failed_mount6: 5667 ext4_mb_release(sb); 5668 ext4_flex_groups_free(sbi); 5669 ext4_percpu_param_destroy(sbi); 5670 failed_mount5: 5671 ext4_ext_release(sb); 5672 ext4_release_system_zone(sb); 5673 failed_mount4a: 5674 dput(sb->s_root); 5675 sb->s_root = NULL; 5676 failed_mount4: 5677 ext4_msg(sb, KERN_ERR, "mount failed"); 5678 if (EXT4_SB(sb)->rsv_conversion_wq) 5679 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 5680 failed_mount_wq: 5681 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 5682 sbi->s_ea_inode_cache = NULL; 5683 5684 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 5685 sbi->s_ea_block_cache = NULL; 5686 5687 if (sbi->s_journal) { 5688 ext4_journal_destroy(sbi, sbi->s_journal); 5689 } 5690 failed_mount3a: 5691 ext4_es_unregister_shrinker(sbi); 5692 failed_mount3: 5693 /* flush s_sb_upd_work before sbi destroy */ 5694 flush_work(&sbi->s_sb_upd_work); 5695 ext4_stop_mmpd(sbi); 5696 timer_delete_sync(&sbi->s_err_report); 5697 ext4_group_desc_free(sbi); 5698 failed_mount: 5699 #if IS_ENABLED(CONFIG_UNICODE) 5700 utf8_unload(sb->s_encoding); 5701 #endif 5702 5703 #ifdef CONFIG_QUOTA 5704 for (unsigned int i = 0; i < EXT4_MAXQUOTAS; i++) 5705 kfree(get_qf_name(sb, sbi, i)); 5706 #endif 5707 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy); 5708 brelse(sbi->s_sbh); 5709 if (sbi->s_journal_bdev_file) { 5710 invalidate_bdev(file_bdev(sbi->s_journal_bdev_file)); 5711 bdev_fput(sbi->s_journal_bdev_file); 5712 } 5713 out_fail: 5714 invalidate_bdev(sb->s_bdev); 5715 sb->s_fs_info = NULL; 5716 return err; 5717 } 5718 5719 static int ext4_fill_super(struct super_block *sb, struct fs_context *fc) 5720 { 5721 struct ext4_fs_context *ctx = fc->fs_private; 5722 struct ext4_sb_info *sbi; 5723 const char *descr; 5724 int ret; 5725 5726 sbi = ext4_alloc_sbi(sb); 5727 if (!sbi) 5728 return -ENOMEM; 5729 5730 fc->s_fs_info = sbi; 5731 5732 /* Cleanup superblock name */ 5733 strreplace(sb->s_id, '/', '!'); 5734 5735 sbi->s_sb_block = 1; /* Default super block location */ 5736 if (ctx->spec & EXT4_SPEC_s_sb_block) 5737 sbi->s_sb_block = ctx->s_sb_block; 5738 5739 ret = __ext4_fill_super(fc, sb); 5740 if (ret < 0) 5741 goto free_sbi; 5742 5743 if (sbi->s_journal) { 5744 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 5745 descr = " journalled data mode"; 5746 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 5747 descr = " ordered data mode"; 5748 else 5749 descr = " writeback data mode"; 5750 } else 5751 descr = "out journal"; 5752 5753 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 5754 ext4_msg(sb, KERN_INFO, "mounted filesystem %pU %s with%s. " 5755 "Quota mode: %s.", &sb->s_uuid, 5756 sb_rdonly(sb) ? "ro" : "r/w", descr, 5757 ext4_quota_mode(sb)); 5758 5759 /* Update the s_overhead_clusters if necessary */ 5760 ext4_update_overhead(sb, false); 5761 return 0; 5762 5763 free_sbi: 5764 ext4_free_sbi(sbi); 5765 fc->s_fs_info = NULL; 5766 return ret; 5767 } 5768 5769 static int ext4_get_tree(struct fs_context *fc) 5770 { 5771 return get_tree_bdev(fc, ext4_fill_super); 5772 } 5773 5774 /* 5775 * Setup any per-fs journal parameters now. We'll do this both on 5776 * initial mount, once the journal has been initialised but before we've 5777 * done any recovery; and again on any subsequent remount. 5778 */ 5779 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 5780 { 5781 struct ext4_sb_info *sbi = EXT4_SB(sb); 5782 5783 journal->j_commit_interval = sbi->s_commit_interval; 5784 journal->j_min_batch_time = sbi->s_min_batch_time; 5785 journal->j_max_batch_time = sbi->s_max_batch_time; 5786 ext4_fc_init(sb, journal); 5787 5788 write_lock(&journal->j_state_lock); 5789 if (test_opt(sb, BARRIER)) 5790 journal->j_flags |= JBD2_BARRIER; 5791 else 5792 journal->j_flags &= ~JBD2_BARRIER; 5793 /* 5794 * Always enable journal cycle record option, letting the journal 5795 * records log transactions continuously between each mount. 5796 */ 5797 journal->j_flags |= JBD2_CYCLE_RECORD; 5798 write_unlock(&journal->j_state_lock); 5799 } 5800 5801 static struct inode *ext4_get_journal_inode(struct super_block *sb, 5802 unsigned int journal_inum) 5803 { 5804 struct inode *journal_inode; 5805 5806 /* 5807 * Test for the existence of a valid inode on disk. Bad things 5808 * happen if we iget() an unused inode, as the subsequent iput() 5809 * will try to delete it. 5810 */ 5811 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL); 5812 if (IS_ERR(journal_inode)) { 5813 ext4_msg(sb, KERN_ERR, "no journal found"); 5814 return ERR_CAST(journal_inode); 5815 } 5816 if (!journal_inode->i_nlink) { 5817 make_bad_inode(journal_inode); 5818 iput(journal_inode); 5819 ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 5820 return ERR_PTR(-EFSCORRUPTED); 5821 } 5822 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) { 5823 ext4_msg(sb, KERN_ERR, "invalid journal inode"); 5824 iput(journal_inode); 5825 return ERR_PTR(-EFSCORRUPTED); 5826 } 5827 5828 ext4_debug("Journal inode found at %p: %lld bytes\n", 5829 journal_inode, journal_inode->i_size); 5830 return journal_inode; 5831 } 5832 5833 static int ext4_journal_bmap(journal_t *journal, sector_t *block) 5834 { 5835 struct ext4_map_blocks map; 5836 int ret; 5837 5838 if (journal->j_inode == NULL) 5839 return 0; 5840 5841 map.m_lblk = *block; 5842 map.m_len = 1; 5843 ret = ext4_map_blocks(NULL, journal->j_inode, &map, 0); 5844 if (ret <= 0) { 5845 ext4_msg(journal->j_inode->i_sb, KERN_CRIT, 5846 "journal bmap failed: block %llu ret %d\n", 5847 *block, ret); 5848 jbd2_journal_abort(journal, ret ? ret : -EFSCORRUPTED); 5849 return ret; 5850 } 5851 *block = map.m_pblk; 5852 return 0; 5853 } 5854 5855 static journal_t *ext4_open_inode_journal(struct super_block *sb, 5856 unsigned int journal_inum) 5857 { 5858 struct inode *journal_inode; 5859 journal_t *journal; 5860 5861 journal_inode = ext4_get_journal_inode(sb, journal_inum); 5862 if (IS_ERR(journal_inode)) 5863 return ERR_CAST(journal_inode); 5864 5865 journal = jbd2_journal_init_inode(journal_inode); 5866 if (IS_ERR(journal)) { 5867 ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 5868 iput(journal_inode); 5869 return ERR_CAST(journal); 5870 } 5871 journal->j_private = sb; 5872 journal->j_bmap = ext4_journal_bmap; 5873 ext4_init_journal_params(sb, journal); 5874 return journal; 5875 } 5876 5877 static struct file *ext4_get_journal_blkdev(struct super_block *sb, 5878 dev_t j_dev, ext4_fsblk_t *j_start, 5879 ext4_fsblk_t *j_len) 5880 { 5881 struct buffer_head *bh; 5882 struct block_device *bdev; 5883 struct file *bdev_file; 5884 int hblock, blocksize; 5885 ext4_fsblk_t sb_block; 5886 unsigned long offset; 5887 struct ext4_super_block *es; 5888 int errno; 5889 5890 bdev_file = bdev_file_open_by_dev(j_dev, 5891 BLK_OPEN_READ | BLK_OPEN_WRITE | BLK_OPEN_RESTRICT_WRITES, 5892 sb, &fs_holder_ops); 5893 if (IS_ERR(bdev_file)) { 5894 ext4_msg(sb, KERN_ERR, 5895 "failed to open journal device unknown-block(%u,%u) %ld", 5896 MAJOR(j_dev), MINOR(j_dev), PTR_ERR(bdev_file)); 5897 return bdev_file; 5898 } 5899 5900 bdev = file_bdev(bdev_file); 5901 blocksize = sb->s_blocksize; 5902 hblock = bdev_logical_block_size(bdev); 5903 if (blocksize < hblock) { 5904 ext4_msg(sb, KERN_ERR, 5905 "blocksize too small for journal device"); 5906 errno = -EINVAL; 5907 goto out_bdev; 5908 } 5909 5910 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 5911 offset = EXT4_MIN_BLOCK_SIZE % blocksize; 5912 set_blocksize(bdev_file, blocksize); 5913 bh = __bread(bdev, sb_block, blocksize); 5914 if (!bh) { 5915 ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 5916 "external journal"); 5917 errno = -EINVAL; 5918 goto out_bdev; 5919 } 5920 5921 es = (struct ext4_super_block *) (bh->b_data + offset); 5922 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 5923 !(le32_to_cpu(es->s_feature_incompat) & 5924 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 5925 ext4_msg(sb, KERN_ERR, "external journal has bad superblock"); 5926 errno = -EFSCORRUPTED; 5927 goto out_bh; 5928 } 5929 5930 if ((le32_to_cpu(es->s_feature_ro_compat) & 5931 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 5932 es->s_checksum != ext4_superblock_csum(es)) { 5933 ext4_msg(sb, KERN_ERR, "external journal has corrupt superblock"); 5934 errno = -EFSCORRUPTED; 5935 goto out_bh; 5936 } 5937 5938 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 5939 ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 5940 errno = -EFSCORRUPTED; 5941 goto out_bh; 5942 } 5943 5944 *j_start = sb_block + 1; 5945 *j_len = ext4_blocks_count(es); 5946 brelse(bh); 5947 return bdev_file; 5948 5949 out_bh: 5950 brelse(bh); 5951 out_bdev: 5952 bdev_fput(bdev_file); 5953 return ERR_PTR(errno); 5954 } 5955 5956 static journal_t *ext4_open_dev_journal(struct super_block *sb, 5957 dev_t j_dev) 5958 { 5959 journal_t *journal; 5960 ext4_fsblk_t j_start; 5961 ext4_fsblk_t j_len; 5962 struct file *bdev_file; 5963 int errno = 0; 5964 5965 bdev_file = ext4_get_journal_blkdev(sb, j_dev, &j_start, &j_len); 5966 if (IS_ERR(bdev_file)) 5967 return ERR_CAST(bdev_file); 5968 5969 journal = jbd2_journal_init_dev(file_bdev(bdev_file), sb->s_bdev, j_start, 5970 j_len, sb->s_blocksize); 5971 if (IS_ERR(journal)) { 5972 ext4_msg(sb, KERN_ERR, "failed to create device journal"); 5973 errno = PTR_ERR(journal); 5974 goto out_bdev; 5975 } 5976 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 5977 ext4_msg(sb, KERN_ERR, "External journal has more than one " 5978 "user (unsupported) - %d", 5979 be32_to_cpu(journal->j_superblock->s_nr_users)); 5980 errno = -EINVAL; 5981 goto out_journal; 5982 } 5983 journal->j_private = sb; 5984 EXT4_SB(sb)->s_journal_bdev_file = bdev_file; 5985 ext4_init_journal_params(sb, journal); 5986 return journal; 5987 5988 out_journal: 5989 ext4_journal_destroy(EXT4_SB(sb), journal); 5990 out_bdev: 5991 bdev_fput(bdev_file); 5992 return ERR_PTR(errno); 5993 } 5994 5995 static int ext4_load_journal(struct super_block *sb, 5996 struct ext4_super_block *es, 5997 unsigned long journal_devnum) 5998 { 5999 journal_t *journal; 6000 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 6001 dev_t journal_dev; 6002 int err = 0; 6003 int really_read_only; 6004 int journal_dev_ro; 6005 6006 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 6007 return -EFSCORRUPTED; 6008 6009 if (journal_devnum && 6010 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 6011 ext4_msg(sb, KERN_INFO, "external journal device major/minor " 6012 "numbers have changed"); 6013 journal_dev = new_decode_dev(journal_devnum); 6014 } else 6015 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 6016 6017 if (journal_inum && journal_dev) { 6018 ext4_msg(sb, KERN_ERR, 6019 "filesystem has both journal inode and journal device!"); 6020 return -EINVAL; 6021 } 6022 6023 if (journal_inum) { 6024 journal = ext4_open_inode_journal(sb, journal_inum); 6025 if (IS_ERR(journal)) 6026 return PTR_ERR(journal); 6027 } else { 6028 journal = ext4_open_dev_journal(sb, journal_dev); 6029 if (IS_ERR(journal)) 6030 return PTR_ERR(journal); 6031 } 6032 6033 journal_dev_ro = bdev_read_only(journal->j_dev); 6034 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro; 6035 6036 if (journal_dev_ro && !sb_rdonly(sb)) { 6037 ext4_msg(sb, KERN_ERR, 6038 "journal device read-only, try mounting with '-o ro'"); 6039 err = -EROFS; 6040 goto err_out; 6041 } 6042 6043 /* 6044 * Are we loading a blank journal or performing recovery after a 6045 * crash? For recovery, we need to check in advance whether we 6046 * can get read-write access to the device. 6047 */ 6048 if (ext4_has_feature_journal_needs_recovery(sb)) { 6049 if (sb_rdonly(sb)) { 6050 ext4_msg(sb, KERN_INFO, "INFO: recovery " 6051 "required on readonly filesystem"); 6052 if (really_read_only) { 6053 ext4_msg(sb, KERN_ERR, "write access " 6054 "unavailable, cannot proceed " 6055 "(try mounting with noload)"); 6056 err = -EROFS; 6057 goto err_out; 6058 } 6059 ext4_msg(sb, KERN_INFO, "write access will " 6060 "be enabled during recovery"); 6061 } 6062 } 6063 6064 if (!(journal->j_flags & JBD2_BARRIER)) 6065 ext4_msg(sb, KERN_INFO, "barriers disabled"); 6066 6067 if (!ext4_has_feature_journal_needs_recovery(sb)) 6068 err = jbd2_journal_wipe(journal, !really_read_only); 6069 if (!err) { 6070 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 6071 __le16 orig_state; 6072 bool changed = false; 6073 6074 if (save) 6075 memcpy(save, ((char *) es) + 6076 EXT4_S_ERR_START, EXT4_S_ERR_LEN); 6077 err = jbd2_journal_load(journal); 6078 if (save && memcmp(((char *) es) + EXT4_S_ERR_START, 6079 save, EXT4_S_ERR_LEN)) { 6080 memcpy(((char *) es) + EXT4_S_ERR_START, 6081 save, EXT4_S_ERR_LEN); 6082 changed = true; 6083 } 6084 kfree(save); 6085 orig_state = es->s_state; 6086 es->s_state |= cpu_to_le16(EXT4_SB(sb)->s_mount_state & 6087 EXT4_ERROR_FS); 6088 if (orig_state != es->s_state) 6089 changed = true; 6090 /* Write out restored error information to the superblock */ 6091 if (changed && !really_read_only) { 6092 int err2; 6093 err2 = ext4_commit_super(sb); 6094 err = err ? : err2; 6095 } 6096 } 6097 6098 if (err) { 6099 ext4_msg(sb, KERN_ERR, "error loading journal"); 6100 goto err_out; 6101 } 6102 6103 EXT4_SB(sb)->s_journal = journal; 6104 err = ext4_clear_journal_err(sb, es); 6105 if (err) { 6106 ext4_journal_destroy(EXT4_SB(sb), journal); 6107 return err; 6108 } 6109 6110 if (!really_read_only && journal_devnum && 6111 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 6112 es->s_journal_dev = cpu_to_le32(journal_devnum); 6113 ext4_commit_super(sb); 6114 } 6115 if (!really_read_only && journal_inum && 6116 journal_inum != le32_to_cpu(es->s_journal_inum)) { 6117 es->s_journal_inum = cpu_to_le32(journal_inum); 6118 ext4_commit_super(sb); 6119 } 6120 6121 return 0; 6122 6123 err_out: 6124 ext4_journal_destroy(EXT4_SB(sb), journal); 6125 return err; 6126 } 6127 6128 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */ 6129 static void ext4_update_super(struct super_block *sb) 6130 { 6131 struct ext4_sb_info *sbi = EXT4_SB(sb); 6132 struct ext4_super_block *es = sbi->s_es; 6133 struct buffer_head *sbh = sbi->s_sbh; 6134 6135 lock_buffer(sbh); 6136 /* 6137 * If the file system is mounted read-only, don't update the 6138 * superblock write time. This avoids updating the superblock 6139 * write time when we are mounting the root file system 6140 * read/only but we need to replay the journal; at that point, 6141 * for people who are east of GMT and who make their clock 6142 * tick in localtime for Windows bug-for-bug compatibility, 6143 * the clock is set in the future, and this will cause e2fsck 6144 * to complain and force a full file system check. 6145 */ 6146 if (!sb_rdonly(sb)) 6147 ext4_update_tstamp(es, s_wtime); 6148 es->s_kbytes_written = 6149 cpu_to_le64(sbi->s_kbytes_written + 6150 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) - 6151 sbi->s_sectors_written_start) >> 1)); 6152 if (percpu_counter_initialized(&sbi->s_freeclusters_counter)) 6153 ext4_free_blocks_count_set(es, 6154 EXT4_C2B(sbi, percpu_counter_sum_positive( 6155 &sbi->s_freeclusters_counter))); 6156 if (percpu_counter_initialized(&sbi->s_freeinodes_counter)) 6157 es->s_free_inodes_count = 6158 cpu_to_le32(percpu_counter_sum_positive( 6159 &sbi->s_freeinodes_counter)); 6160 /* Copy error information to the on-disk superblock */ 6161 spin_lock(&sbi->s_error_lock); 6162 if (sbi->s_add_error_count > 0) { 6163 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 6164 if (!es->s_first_error_time && !es->s_first_error_time_hi) { 6165 __ext4_update_tstamp(&es->s_first_error_time, 6166 &es->s_first_error_time_hi, 6167 sbi->s_first_error_time); 6168 strtomem_pad(es->s_first_error_func, 6169 sbi->s_first_error_func, 0); 6170 es->s_first_error_line = 6171 cpu_to_le32(sbi->s_first_error_line); 6172 es->s_first_error_ino = 6173 cpu_to_le32(sbi->s_first_error_ino); 6174 es->s_first_error_block = 6175 cpu_to_le64(sbi->s_first_error_block); 6176 es->s_first_error_errcode = 6177 ext4_errno_to_code(sbi->s_first_error_code); 6178 } 6179 __ext4_update_tstamp(&es->s_last_error_time, 6180 &es->s_last_error_time_hi, 6181 sbi->s_last_error_time); 6182 strtomem_pad(es->s_last_error_func, sbi->s_last_error_func, 0); 6183 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line); 6184 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino); 6185 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block); 6186 es->s_last_error_errcode = 6187 ext4_errno_to_code(sbi->s_last_error_code); 6188 /* 6189 * Start the daily error reporting function if it hasn't been 6190 * started already 6191 */ 6192 if (!es->s_error_count) 6193 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); 6194 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count); 6195 sbi->s_add_error_count = 0; 6196 } 6197 spin_unlock(&sbi->s_error_lock); 6198 6199 ext4_superblock_csum_set(sb); 6200 unlock_buffer(sbh); 6201 } 6202 6203 static int ext4_commit_super(struct super_block *sb) 6204 { 6205 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 6206 6207 if (!sbh) 6208 return -EINVAL; 6209 6210 ext4_update_super(sb); 6211 6212 lock_buffer(sbh); 6213 /* Buffer got discarded which means block device got invalidated */ 6214 if (!buffer_mapped(sbh)) { 6215 unlock_buffer(sbh); 6216 return -EIO; 6217 } 6218 6219 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 6220 /* 6221 * Oh, dear. A previous attempt to write the 6222 * superblock failed. This could happen because the 6223 * USB device was yanked out. Or it could happen to 6224 * be a transient write error and maybe the block will 6225 * be remapped. Nothing we can do but to retry the 6226 * write and hope for the best. 6227 */ 6228 ext4_msg(sb, KERN_ERR, "previous I/O error to " 6229 "superblock detected"); 6230 clear_buffer_write_io_error(sbh); 6231 set_buffer_uptodate(sbh); 6232 } 6233 get_bh(sbh); 6234 /* Clear potential dirty bit if it was journalled update */ 6235 clear_buffer_dirty(sbh); 6236 sbh->b_end_io = end_buffer_write_sync; 6237 submit_bh(REQ_OP_WRITE | REQ_SYNC | 6238 (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh); 6239 wait_on_buffer(sbh); 6240 if (buffer_write_io_error(sbh)) { 6241 ext4_msg(sb, KERN_ERR, "I/O error while writing " 6242 "superblock"); 6243 clear_buffer_write_io_error(sbh); 6244 set_buffer_uptodate(sbh); 6245 return -EIO; 6246 } 6247 return 0; 6248 } 6249 6250 /* 6251 * Have we just finished recovery? If so, and if we are mounting (or 6252 * remounting) the filesystem readonly, then we will end up with a 6253 * consistent fs on disk. Record that fact. 6254 */ 6255 static int ext4_mark_recovery_complete(struct super_block *sb, 6256 struct ext4_super_block *es) 6257 { 6258 int err; 6259 journal_t *journal = EXT4_SB(sb)->s_journal; 6260 6261 if (!ext4_has_feature_journal(sb)) { 6262 if (journal != NULL) { 6263 ext4_error(sb, "Journal got removed while the fs was " 6264 "mounted!"); 6265 return -EFSCORRUPTED; 6266 } 6267 return 0; 6268 } 6269 jbd2_journal_lock_updates(journal); 6270 err = jbd2_journal_flush(journal, 0); 6271 if (err < 0) 6272 goto out; 6273 6274 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) || 6275 ext4_has_feature_orphan_present(sb))) { 6276 if (!ext4_orphan_file_empty(sb)) { 6277 ext4_error(sb, "Orphan file not empty on read-only fs."); 6278 err = -EFSCORRUPTED; 6279 goto out; 6280 } 6281 ext4_clear_feature_journal_needs_recovery(sb); 6282 ext4_clear_feature_orphan_present(sb); 6283 ext4_commit_super(sb); 6284 } 6285 out: 6286 jbd2_journal_unlock_updates(journal); 6287 return err; 6288 } 6289 6290 /* 6291 * If we are mounting (or read-write remounting) a filesystem whose journal 6292 * has recorded an error from a previous lifetime, move that error to the 6293 * main filesystem now. 6294 */ 6295 static int ext4_clear_journal_err(struct super_block *sb, 6296 struct ext4_super_block *es) 6297 { 6298 journal_t *journal; 6299 int j_errno; 6300 const char *errstr; 6301 6302 if (!ext4_has_feature_journal(sb)) { 6303 ext4_error(sb, "Journal got removed while the fs was mounted!"); 6304 return -EFSCORRUPTED; 6305 } 6306 6307 journal = EXT4_SB(sb)->s_journal; 6308 6309 /* 6310 * Now check for any error status which may have been recorded in the 6311 * journal by a prior ext4_error() or ext4_abort() 6312 */ 6313 6314 j_errno = jbd2_journal_errno(journal); 6315 if (j_errno) { 6316 char nbuf[16]; 6317 6318 errstr = ext4_decode_error(sb, j_errno, nbuf); 6319 ext4_warning(sb, "Filesystem error recorded " 6320 "from previous mount: %s", errstr); 6321 6322 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 6323 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 6324 j_errno = ext4_commit_super(sb); 6325 if (j_errno) 6326 return j_errno; 6327 ext4_warning(sb, "Marked fs in need of filesystem check."); 6328 6329 jbd2_journal_clear_err(journal); 6330 jbd2_journal_update_sb_errno(journal); 6331 } 6332 return 0; 6333 } 6334 6335 /* 6336 * Force the running and committing transactions to commit, 6337 * and wait on the commit. 6338 */ 6339 int ext4_force_commit(struct super_block *sb) 6340 { 6341 return ext4_journal_force_commit(EXT4_SB(sb)->s_journal); 6342 } 6343 6344 static int ext4_sync_fs(struct super_block *sb, int wait) 6345 { 6346 int ret = 0; 6347 tid_t target; 6348 bool needs_barrier = false; 6349 struct ext4_sb_info *sbi = EXT4_SB(sb); 6350 6351 ret = ext4_emergency_state(sb); 6352 if (unlikely(ret)) 6353 return ret; 6354 6355 trace_ext4_sync_fs(sb, wait); 6356 flush_workqueue(sbi->rsv_conversion_wq); 6357 /* 6358 * Writeback quota in non-journalled quota case - journalled quota has 6359 * no dirty dquots 6360 */ 6361 dquot_writeback_dquots(sb, -1); 6362 /* 6363 * Data writeback is possible w/o journal transaction, so barrier must 6364 * being sent at the end of the function. But we can skip it if 6365 * transaction_commit will do it for us. 6366 */ 6367 if (sbi->s_journal) { 6368 target = jbd2_get_latest_transaction(sbi->s_journal); 6369 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 6370 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 6371 needs_barrier = true; 6372 6373 if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 6374 if (wait) 6375 ret = jbd2_log_wait_commit(sbi->s_journal, 6376 target); 6377 } 6378 } else if (wait && test_opt(sb, BARRIER)) 6379 needs_barrier = true; 6380 if (needs_barrier) { 6381 int err; 6382 err = blkdev_issue_flush(sb->s_bdev); 6383 if (!ret) 6384 ret = err; 6385 } 6386 6387 return ret; 6388 } 6389 6390 /* 6391 * LVM calls this function before a (read-only) snapshot is created. This 6392 * gives us a chance to flush the journal completely and mark the fs clean. 6393 * 6394 * Note that only this function cannot bring a filesystem to be in a clean 6395 * state independently. It relies on upper layer to stop all data & metadata 6396 * modifications. 6397 */ 6398 static int ext4_freeze(struct super_block *sb) 6399 { 6400 int error = 0; 6401 journal_t *journal = EXT4_SB(sb)->s_journal; 6402 6403 if (journal) { 6404 /* Now we set up the journal barrier. */ 6405 jbd2_journal_lock_updates(journal); 6406 6407 /* 6408 * Don't clear the needs_recovery flag if we failed to 6409 * flush the journal. 6410 */ 6411 error = jbd2_journal_flush(journal, 0); 6412 if (error < 0) 6413 goto out; 6414 6415 /* Journal blocked and flushed, clear needs_recovery flag. */ 6416 ext4_clear_feature_journal_needs_recovery(sb); 6417 if (ext4_orphan_file_empty(sb)) 6418 ext4_clear_feature_orphan_present(sb); 6419 } 6420 6421 error = ext4_commit_super(sb); 6422 out: 6423 if (journal) 6424 /* we rely on upper layer to stop further updates */ 6425 jbd2_journal_unlock_updates(journal); 6426 return error; 6427 } 6428 6429 /* 6430 * Called by LVM after the snapshot is done. We need to reset the RECOVER 6431 * flag here, even though the filesystem is not technically dirty yet. 6432 */ 6433 static int ext4_unfreeze(struct super_block *sb) 6434 { 6435 if (ext4_emergency_state(sb)) 6436 return 0; 6437 6438 if (EXT4_SB(sb)->s_journal) { 6439 /* Reset the needs_recovery flag before the fs is unlocked. */ 6440 ext4_set_feature_journal_needs_recovery(sb); 6441 if (ext4_has_feature_orphan_file(sb)) 6442 ext4_set_feature_orphan_present(sb); 6443 } 6444 6445 ext4_commit_super(sb); 6446 return 0; 6447 } 6448 6449 /* 6450 * Structure to save mount options for ext4_remount's benefit 6451 */ 6452 struct ext4_mount_options { 6453 unsigned long s_mount_opt; 6454 unsigned long s_mount_opt2; 6455 kuid_t s_resuid; 6456 kgid_t s_resgid; 6457 unsigned long s_commit_interval; 6458 u32 s_min_batch_time, s_max_batch_time; 6459 #ifdef CONFIG_QUOTA 6460 int s_jquota_fmt; 6461 char *s_qf_names[EXT4_MAXQUOTAS]; 6462 #endif 6463 }; 6464 6465 static int __ext4_remount(struct fs_context *fc, struct super_block *sb) 6466 { 6467 struct ext4_fs_context *ctx = fc->fs_private; 6468 struct ext4_super_block *es; 6469 struct ext4_sb_info *sbi = EXT4_SB(sb); 6470 unsigned long old_sb_flags; 6471 struct ext4_mount_options old_opts; 6472 ext4_group_t g; 6473 int err = 0; 6474 int alloc_ctx; 6475 #ifdef CONFIG_QUOTA 6476 int enable_quota = 0; 6477 int i, j; 6478 char *to_free[EXT4_MAXQUOTAS]; 6479 #endif 6480 6481 6482 /* Store the original options */ 6483 old_sb_flags = sb->s_flags; 6484 old_opts.s_mount_opt = sbi->s_mount_opt; 6485 old_opts.s_mount_opt2 = sbi->s_mount_opt2; 6486 old_opts.s_resuid = sbi->s_resuid; 6487 old_opts.s_resgid = sbi->s_resgid; 6488 old_opts.s_commit_interval = sbi->s_commit_interval; 6489 old_opts.s_min_batch_time = sbi->s_min_batch_time; 6490 old_opts.s_max_batch_time = sbi->s_max_batch_time; 6491 #ifdef CONFIG_QUOTA 6492 old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 6493 for (i = 0; i < EXT4_MAXQUOTAS; i++) 6494 if (sbi->s_qf_names[i]) { 6495 char *qf_name = get_qf_name(sb, sbi, i); 6496 6497 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); 6498 if (!old_opts.s_qf_names[i]) { 6499 for (j = 0; j < i; j++) 6500 kfree(old_opts.s_qf_names[j]); 6501 return -ENOMEM; 6502 } 6503 } else 6504 old_opts.s_qf_names[i] = NULL; 6505 #endif 6506 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) { 6507 if (sbi->s_journal && sbi->s_journal->j_task->io_context) 6508 ctx->journal_ioprio = 6509 sbi->s_journal->j_task->io_context->ioprio; 6510 else 6511 ctx->journal_ioprio = EXT4_DEF_JOURNAL_IOPRIO; 6512 6513 } 6514 6515 if ((ctx->spec & EXT4_SPEC_s_stripe) && 6516 ext4_is_stripe_incompatible(sb, ctx->s_stripe)) { 6517 ext4_msg(sb, KERN_WARNING, 6518 "stripe (%lu) is not aligned with cluster size (%u), " 6519 "stripe is disabled", 6520 ctx->s_stripe, sbi->s_cluster_ratio); 6521 ctx->s_stripe = 0; 6522 } 6523 6524 /* 6525 * Changing the DIOREAD_NOLOCK or DELALLOC mount options may cause 6526 * two calls to ext4_should_dioread_nolock() to return inconsistent 6527 * values, triggering WARN_ON in ext4_add_complete_io(). we grab 6528 * here s_writepages_rwsem to avoid race between writepages ops and 6529 * remount. 6530 */ 6531 alloc_ctx = ext4_writepages_down_write(sb); 6532 ext4_apply_options(fc, sb); 6533 ext4_writepages_up_write(sb, alloc_ctx); 6534 6535 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 6536 test_opt(sb, JOURNAL_CHECKSUM)) { 6537 ext4_msg(sb, KERN_ERR, "changing journal_checksum " 6538 "during remount not supported; ignoring"); 6539 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 6540 } 6541 6542 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 6543 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 6544 ext4_msg(sb, KERN_ERR, "can't mount with " 6545 "both data=journal and delalloc"); 6546 err = -EINVAL; 6547 goto restore_opts; 6548 } 6549 if (test_opt(sb, DIOREAD_NOLOCK)) { 6550 ext4_msg(sb, KERN_ERR, "can't mount with " 6551 "both data=journal and dioread_nolock"); 6552 err = -EINVAL; 6553 goto restore_opts; 6554 } 6555 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { 6556 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 6557 ext4_msg(sb, KERN_ERR, "can't mount with " 6558 "journal_async_commit in data=ordered mode"); 6559 err = -EINVAL; 6560 goto restore_opts; 6561 } 6562 } 6563 6564 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { 6565 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); 6566 err = -EINVAL; 6567 goto restore_opts; 6568 } 6569 6570 if ((old_opts.s_mount_opt & EXT4_MOUNT_DELALLOC) && 6571 !test_opt(sb, DELALLOC)) { 6572 ext4_msg(sb, KERN_ERR, "can't disable delalloc during remount"); 6573 err = -EINVAL; 6574 goto restore_opts; 6575 } 6576 6577 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 6578 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 6579 6580 es = sbi->s_es; 6581 6582 if (sbi->s_journal) { 6583 ext4_init_journal_params(sb, sbi->s_journal); 6584 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio); 6585 } 6586 6587 /* Flush outstanding errors before changing fs state */ 6588 flush_work(&sbi->s_sb_upd_work); 6589 6590 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) { 6591 if (ext4_emergency_state(sb)) { 6592 err = -EROFS; 6593 goto restore_opts; 6594 } 6595 6596 if (fc->sb_flags & SB_RDONLY) { 6597 err = sync_filesystem(sb); 6598 if (err < 0) 6599 goto restore_opts; 6600 err = dquot_suspend(sb, -1); 6601 if (err < 0) 6602 goto restore_opts; 6603 6604 /* 6605 * First of all, the unconditional stuff we have to do 6606 * to disable replay of the journal when we next remount 6607 */ 6608 sb->s_flags |= SB_RDONLY; 6609 6610 /* 6611 * OK, test if we are remounting a valid rw partition 6612 * readonly, and if so set the rdonly flag and then 6613 * mark the partition as valid again. 6614 */ 6615 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 6616 (sbi->s_mount_state & EXT4_VALID_FS)) 6617 es->s_state = cpu_to_le16(sbi->s_mount_state); 6618 6619 if (sbi->s_journal) { 6620 /* 6621 * We let remount-ro finish even if marking fs 6622 * as clean failed... 6623 */ 6624 ext4_mark_recovery_complete(sb, es); 6625 } 6626 } else { 6627 /* Make sure we can mount this feature set readwrite */ 6628 if (ext4_has_feature_readonly(sb) || 6629 !ext4_feature_set_ok(sb, 0)) { 6630 err = -EROFS; 6631 goto restore_opts; 6632 } 6633 /* 6634 * Make sure the group descriptor checksums 6635 * are sane. If they aren't, refuse to remount r/w. 6636 */ 6637 for (g = 0; g < sbi->s_groups_count; g++) { 6638 struct ext4_group_desc *gdp = 6639 ext4_get_group_desc(sb, g, NULL); 6640 6641 if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 6642 ext4_msg(sb, KERN_ERR, 6643 "ext4_remount: Checksum for group %u failed (%u!=%u)", 6644 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 6645 le16_to_cpu(gdp->bg_checksum)); 6646 err = -EFSBADCRC; 6647 goto restore_opts; 6648 } 6649 } 6650 6651 /* 6652 * If we have an unprocessed orphan list hanging 6653 * around from a previously readonly bdev mount, 6654 * require a full umount/remount for now. 6655 */ 6656 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) { 6657 ext4_msg(sb, KERN_WARNING, "Couldn't " 6658 "remount RDWR because of unprocessed " 6659 "orphan inode list. Please " 6660 "umount/remount instead"); 6661 err = -EINVAL; 6662 goto restore_opts; 6663 } 6664 6665 /* 6666 * Mounting a RDONLY partition read-write, so reread 6667 * and store the current valid flag. (It may have 6668 * been changed by e2fsck since we originally mounted 6669 * the partition.) 6670 */ 6671 if (sbi->s_journal) { 6672 err = ext4_clear_journal_err(sb, es); 6673 if (err) 6674 goto restore_opts; 6675 } 6676 sbi->s_mount_state = (le16_to_cpu(es->s_state) & 6677 ~EXT4_FC_REPLAY); 6678 6679 err = ext4_setup_super(sb, es, 0); 6680 if (err) 6681 goto restore_opts; 6682 6683 sb->s_flags &= ~SB_RDONLY; 6684 if (ext4_has_feature_mmp(sb)) { 6685 err = ext4_multi_mount_protect(sb, 6686 le64_to_cpu(es->s_mmp_block)); 6687 if (err) 6688 goto restore_opts; 6689 } 6690 #ifdef CONFIG_QUOTA 6691 enable_quota = 1; 6692 #endif 6693 } 6694 } 6695 6696 /* 6697 * Handle creation of system zone data early because it can fail. 6698 * Releasing of existing data is done when we are sure remount will 6699 * succeed. 6700 */ 6701 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) { 6702 err = ext4_setup_system_zone(sb); 6703 if (err) 6704 goto restore_opts; 6705 } 6706 6707 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) { 6708 err = ext4_commit_super(sb); 6709 if (err) 6710 goto restore_opts; 6711 } 6712 6713 #ifdef CONFIG_QUOTA 6714 if (enable_quota) { 6715 if (sb_any_quota_suspended(sb)) 6716 dquot_resume(sb, -1); 6717 else if (ext4_has_feature_quota(sb)) { 6718 err = ext4_enable_quotas(sb); 6719 if (err) 6720 goto restore_opts; 6721 } 6722 } 6723 /* Release old quota file names */ 6724 for (i = 0; i < EXT4_MAXQUOTAS; i++) 6725 kfree(old_opts.s_qf_names[i]); 6726 #endif 6727 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 6728 ext4_release_system_zone(sb); 6729 6730 /* 6731 * Reinitialize lazy itable initialization thread based on 6732 * current settings 6733 */ 6734 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) 6735 ext4_unregister_li_request(sb); 6736 else { 6737 ext4_group_t first_not_zeroed; 6738 first_not_zeroed = ext4_has_uninit_itable(sb); 6739 ext4_register_li_request(sb, first_not_zeroed); 6740 } 6741 6742 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb)) 6743 ext4_stop_mmpd(sbi); 6744 6745 /* 6746 * Handle aborting the filesystem as the last thing during remount to 6747 * avoid obsure errors during remount when some option changes fail to 6748 * apply due to shutdown filesystem. 6749 */ 6750 if (test_opt2(sb, ABORT)) 6751 ext4_abort(sb, ESHUTDOWN, "Abort forced by user"); 6752 6753 return 0; 6754 6755 restore_opts: 6756 /* 6757 * If there was a failing r/w to ro transition, we may need to 6758 * re-enable quota 6759 */ 6760 if (sb_rdonly(sb) && !(old_sb_flags & SB_RDONLY) && 6761 sb_any_quota_suspended(sb)) 6762 dquot_resume(sb, -1); 6763 6764 alloc_ctx = ext4_writepages_down_write(sb); 6765 sb->s_flags = old_sb_flags; 6766 sbi->s_mount_opt = old_opts.s_mount_opt; 6767 sbi->s_mount_opt2 = old_opts.s_mount_opt2; 6768 sbi->s_resuid = old_opts.s_resuid; 6769 sbi->s_resgid = old_opts.s_resgid; 6770 sbi->s_commit_interval = old_opts.s_commit_interval; 6771 sbi->s_min_batch_time = old_opts.s_min_batch_time; 6772 sbi->s_max_batch_time = old_opts.s_max_batch_time; 6773 ext4_writepages_up_write(sb, alloc_ctx); 6774 6775 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 6776 ext4_release_system_zone(sb); 6777 #ifdef CONFIG_QUOTA 6778 sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 6779 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 6780 to_free[i] = get_qf_name(sb, sbi, i); 6781 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); 6782 } 6783 synchronize_rcu(); 6784 for (i = 0; i < EXT4_MAXQUOTAS; i++) 6785 kfree(to_free[i]); 6786 #endif 6787 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb)) 6788 ext4_stop_mmpd(sbi); 6789 return err; 6790 } 6791 6792 static int ext4_reconfigure(struct fs_context *fc) 6793 { 6794 struct super_block *sb = fc->root->d_sb; 6795 int ret; 6796 bool old_ro = sb_rdonly(sb); 6797 6798 fc->s_fs_info = EXT4_SB(sb); 6799 6800 ret = ext4_check_opt_consistency(fc, sb); 6801 if (ret < 0) 6802 return ret; 6803 6804 ret = __ext4_remount(fc, sb); 6805 if (ret < 0) 6806 return ret; 6807 6808 ext4_msg(sb, KERN_INFO, "re-mounted %pU%s.", 6809 &sb->s_uuid, 6810 (old_ro != sb_rdonly(sb)) ? (sb_rdonly(sb) ? " ro" : " r/w") : ""); 6811 6812 return 0; 6813 } 6814 6815 #ifdef CONFIG_QUOTA 6816 static int ext4_statfs_project(struct super_block *sb, 6817 kprojid_t projid, struct kstatfs *buf) 6818 { 6819 struct kqid qid; 6820 struct dquot *dquot; 6821 u64 limit; 6822 u64 curblock; 6823 6824 qid = make_kqid_projid(projid); 6825 dquot = dqget(sb, qid); 6826 if (IS_ERR(dquot)) 6827 return PTR_ERR(dquot); 6828 spin_lock(&dquot->dq_dqb_lock); 6829 6830 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 6831 dquot->dq_dqb.dqb_bhardlimit); 6832 limit >>= sb->s_blocksize_bits; 6833 6834 if (limit) { 6835 uint64_t remaining = 0; 6836 6837 curblock = (dquot->dq_dqb.dqb_curspace + 6838 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 6839 if (limit > curblock) 6840 remaining = limit - curblock; 6841 6842 buf->f_blocks = min(buf->f_blocks, limit); 6843 buf->f_bfree = min(buf->f_bfree, remaining); 6844 buf->f_bavail = min(buf->f_bavail, remaining); 6845 } 6846 6847 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 6848 dquot->dq_dqb.dqb_ihardlimit); 6849 if (limit) { 6850 uint64_t remaining = 0; 6851 6852 if (limit > dquot->dq_dqb.dqb_curinodes) 6853 remaining = limit - dquot->dq_dqb.dqb_curinodes; 6854 6855 buf->f_files = min(buf->f_files, limit); 6856 buf->f_ffree = min(buf->f_ffree, remaining); 6857 } 6858 6859 spin_unlock(&dquot->dq_dqb_lock); 6860 dqput(dquot); 6861 return 0; 6862 } 6863 #endif 6864 6865 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 6866 { 6867 struct super_block *sb = dentry->d_sb; 6868 struct ext4_sb_info *sbi = EXT4_SB(sb); 6869 struct ext4_super_block *es = sbi->s_es; 6870 ext4_fsblk_t overhead = 0, resv_blocks; 6871 s64 bfree; 6872 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 6873 6874 if (!test_opt(sb, MINIX_DF)) 6875 overhead = sbi->s_overhead; 6876 6877 buf->f_type = EXT4_SUPER_MAGIC; 6878 buf->f_bsize = sb->s_blocksize; 6879 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 6880 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 6881 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 6882 /* prevent underflow in case that few free space is available */ 6883 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 6884 buf->f_bavail = buf->f_bfree - 6885 (ext4_r_blocks_count(es) + resv_blocks); 6886 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 6887 buf->f_bavail = 0; 6888 buf->f_files = le32_to_cpu(es->s_inodes_count); 6889 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 6890 buf->f_namelen = EXT4_NAME_LEN; 6891 buf->f_fsid = uuid_to_fsid(es->s_uuid); 6892 6893 #ifdef CONFIG_QUOTA 6894 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 6895 sb_has_quota_limits_enabled(sb, PRJQUOTA)) 6896 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 6897 #endif 6898 return 0; 6899 } 6900 6901 6902 #ifdef CONFIG_QUOTA 6903 6904 /* 6905 * Helper functions so that transaction is started before we acquire dqio_sem 6906 * to keep correct lock ordering of transaction > dqio_sem 6907 */ 6908 static inline struct inode *dquot_to_inode(struct dquot *dquot) 6909 { 6910 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 6911 } 6912 6913 static int ext4_write_dquot(struct dquot *dquot) 6914 { 6915 int ret, err; 6916 handle_t *handle; 6917 struct inode *inode; 6918 6919 inode = dquot_to_inode(dquot); 6920 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 6921 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 6922 if (IS_ERR(handle)) 6923 return PTR_ERR(handle); 6924 ret = dquot_commit(dquot); 6925 if (ret < 0) 6926 ext4_error_err(dquot->dq_sb, -ret, 6927 "Failed to commit dquot type %d", 6928 dquot->dq_id.type); 6929 err = ext4_journal_stop(handle); 6930 if (!ret) 6931 ret = err; 6932 return ret; 6933 } 6934 6935 static int ext4_acquire_dquot(struct dquot *dquot) 6936 { 6937 int ret, err; 6938 handle_t *handle; 6939 6940 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 6941 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 6942 if (IS_ERR(handle)) 6943 return PTR_ERR(handle); 6944 ret = dquot_acquire(dquot); 6945 if (ret < 0) 6946 ext4_error_err(dquot->dq_sb, -ret, 6947 "Failed to acquire dquot type %d", 6948 dquot->dq_id.type); 6949 err = ext4_journal_stop(handle); 6950 if (!ret) 6951 ret = err; 6952 return ret; 6953 } 6954 6955 static int ext4_release_dquot(struct dquot *dquot) 6956 { 6957 int ret, err; 6958 handle_t *handle; 6959 bool freeze_protected = false; 6960 6961 /* 6962 * Trying to sb_start_intwrite() in a running transaction 6963 * can result in a deadlock. Further, running transactions 6964 * are already protected from freezing. 6965 */ 6966 if (!ext4_journal_current_handle()) { 6967 sb_start_intwrite(dquot->dq_sb); 6968 freeze_protected = true; 6969 } 6970 6971 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 6972 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 6973 if (IS_ERR(handle)) { 6974 /* Release dquot anyway to avoid endless cycle in dqput() */ 6975 dquot_release(dquot); 6976 if (freeze_protected) 6977 sb_end_intwrite(dquot->dq_sb); 6978 return PTR_ERR(handle); 6979 } 6980 ret = dquot_release(dquot); 6981 if (ret < 0) 6982 ext4_error_err(dquot->dq_sb, -ret, 6983 "Failed to release dquot type %d", 6984 dquot->dq_id.type); 6985 err = ext4_journal_stop(handle); 6986 if (!ret) 6987 ret = err; 6988 6989 if (freeze_protected) 6990 sb_end_intwrite(dquot->dq_sb); 6991 6992 return ret; 6993 } 6994 6995 static int ext4_mark_dquot_dirty(struct dquot *dquot) 6996 { 6997 struct super_block *sb = dquot->dq_sb; 6998 6999 if (ext4_is_quota_journalled(sb)) { 7000 dquot_mark_dquot_dirty(dquot); 7001 return ext4_write_dquot(dquot); 7002 } else { 7003 return dquot_mark_dquot_dirty(dquot); 7004 } 7005 } 7006 7007 static int ext4_write_info(struct super_block *sb, int type) 7008 { 7009 int ret, err; 7010 handle_t *handle; 7011 7012 /* Data block + inode block */ 7013 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2); 7014 if (IS_ERR(handle)) 7015 return PTR_ERR(handle); 7016 ret = dquot_commit_info(sb, type); 7017 err = ext4_journal_stop(handle); 7018 if (!ret) 7019 ret = err; 7020 return ret; 7021 } 7022 7023 static void lockdep_set_quota_inode(struct inode *inode, int subclass) 7024 { 7025 struct ext4_inode_info *ei = EXT4_I(inode); 7026 7027 /* The first argument of lockdep_set_subclass has to be 7028 * *exactly* the same as the argument to init_rwsem() --- in 7029 * this case, in init_once() --- or lockdep gets unhappy 7030 * because the name of the lock is set using the 7031 * stringification of the argument to init_rwsem(). 7032 */ 7033 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 7034 lockdep_set_subclass(&ei->i_data_sem, subclass); 7035 } 7036 7037 /* 7038 * Standard function to be called on quota_on 7039 */ 7040 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 7041 const struct path *path) 7042 { 7043 int err; 7044 7045 if (!test_opt(sb, QUOTA)) 7046 return -EINVAL; 7047 7048 /* Quotafile not on the same filesystem? */ 7049 if (path->dentry->d_sb != sb) 7050 return -EXDEV; 7051 7052 /* Quota already enabled for this file? */ 7053 if (IS_NOQUOTA(d_inode(path->dentry))) 7054 return -EBUSY; 7055 7056 /* Journaling quota? */ 7057 if (EXT4_SB(sb)->s_qf_names[type]) { 7058 /* Quotafile not in fs root? */ 7059 if (path->dentry->d_parent != sb->s_root) 7060 ext4_msg(sb, KERN_WARNING, 7061 "Quota file not on filesystem root. " 7062 "Journaled quota will not work"); 7063 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY; 7064 } else { 7065 /* 7066 * Clear the flag just in case mount options changed since 7067 * last time. 7068 */ 7069 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY; 7070 } 7071 7072 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 7073 err = dquot_quota_on(sb, type, format_id, path); 7074 if (!err) { 7075 struct inode *inode = d_inode(path->dentry); 7076 handle_t *handle; 7077 7078 /* 7079 * Set inode flags to prevent userspace from messing with quota 7080 * files. If this fails, we return success anyway since quotas 7081 * are already enabled and this is not a hard failure. 7082 */ 7083 inode_lock(inode); 7084 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 7085 if (IS_ERR(handle)) 7086 goto unlock_inode; 7087 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; 7088 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 7089 S_NOATIME | S_IMMUTABLE); 7090 err = ext4_mark_inode_dirty(handle, inode); 7091 ext4_journal_stop(handle); 7092 unlock_inode: 7093 inode_unlock(inode); 7094 if (err) 7095 dquot_quota_off(sb, type); 7096 } 7097 if (err) 7098 lockdep_set_quota_inode(path->dentry->d_inode, 7099 I_DATA_SEM_NORMAL); 7100 return err; 7101 } 7102 7103 static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum) 7104 { 7105 switch (type) { 7106 case USRQUOTA: 7107 return qf_inum == EXT4_USR_QUOTA_INO; 7108 case GRPQUOTA: 7109 return qf_inum == EXT4_GRP_QUOTA_INO; 7110 case PRJQUOTA: 7111 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO; 7112 default: 7113 BUG(); 7114 } 7115 } 7116 7117 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 7118 unsigned int flags) 7119 { 7120 int err; 7121 struct inode *qf_inode; 7122 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 7123 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 7124 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 7125 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 7126 }; 7127 7128 BUG_ON(!ext4_has_feature_quota(sb)); 7129 7130 if (!qf_inums[type]) 7131 return -EPERM; 7132 7133 if (!ext4_check_quota_inum(type, qf_inums[type])) { 7134 ext4_error(sb, "Bad quota inum: %lu, type: %d", 7135 qf_inums[type], type); 7136 return -EUCLEAN; 7137 } 7138 7139 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL); 7140 if (IS_ERR(qf_inode)) { 7141 ext4_error(sb, "Bad quota inode: %lu, type: %d", 7142 qf_inums[type], type); 7143 return PTR_ERR(qf_inode); 7144 } 7145 7146 /* Don't account quota for quota files to avoid recursion */ 7147 qf_inode->i_flags |= S_NOQUOTA; 7148 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 7149 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 7150 if (err) 7151 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 7152 iput(qf_inode); 7153 7154 return err; 7155 } 7156 7157 /* Enable usage tracking for all quota types. */ 7158 int ext4_enable_quotas(struct super_block *sb) 7159 { 7160 int type, err = 0; 7161 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 7162 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 7163 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 7164 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 7165 }; 7166 bool quota_mopt[EXT4_MAXQUOTAS] = { 7167 test_opt(sb, USRQUOTA), 7168 test_opt(sb, GRPQUOTA), 7169 test_opt(sb, PRJQUOTA), 7170 }; 7171 7172 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 7173 for (type = 0; type < EXT4_MAXQUOTAS; type++) { 7174 if (qf_inums[type]) { 7175 err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 7176 DQUOT_USAGE_ENABLED | 7177 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 7178 if (err) { 7179 ext4_warning(sb, 7180 "Failed to enable quota tracking " 7181 "(type=%d, err=%d, ino=%lu). " 7182 "Please run e2fsck to fix.", type, 7183 err, qf_inums[type]); 7184 7185 ext4_quotas_off(sb, type); 7186 return err; 7187 } 7188 } 7189 } 7190 return 0; 7191 } 7192 7193 static int ext4_quota_off(struct super_block *sb, int type) 7194 { 7195 struct inode *inode = sb_dqopt(sb)->files[type]; 7196 handle_t *handle; 7197 int err; 7198 7199 /* Force all delayed allocation blocks to be allocated. 7200 * Caller already holds s_umount sem */ 7201 if (test_opt(sb, DELALLOC)) 7202 sync_filesystem(sb); 7203 7204 if (!inode || !igrab(inode)) 7205 goto out; 7206 7207 err = dquot_quota_off(sb, type); 7208 if (err || ext4_has_feature_quota(sb)) 7209 goto out_put; 7210 /* 7211 * When the filesystem was remounted read-only first, we cannot cleanup 7212 * inode flags here. Bad luck but people should be using QUOTA feature 7213 * these days anyway. 7214 */ 7215 if (sb_rdonly(sb)) 7216 goto out_put; 7217 7218 inode_lock(inode); 7219 /* 7220 * Update modification times of quota files when userspace can 7221 * start looking at them. If we fail, we return success anyway since 7222 * this is not a hard failure and quotas are already disabled. 7223 */ 7224 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 7225 if (IS_ERR(handle)) { 7226 err = PTR_ERR(handle); 7227 goto out_unlock; 7228 } 7229 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); 7230 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 7231 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 7232 err = ext4_mark_inode_dirty(handle, inode); 7233 ext4_journal_stop(handle); 7234 out_unlock: 7235 inode_unlock(inode); 7236 out_put: 7237 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); 7238 iput(inode); 7239 return err; 7240 out: 7241 return dquot_quota_off(sb, type); 7242 } 7243 7244 /* Read data from quotafile - avoid pagecache and such because we cannot afford 7245 * acquiring the locks... As quota files are never truncated and quota code 7246 * itself serializes the operations (and no one else should touch the files) 7247 * we don't have to be afraid of races */ 7248 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 7249 size_t len, loff_t off) 7250 { 7251 struct inode *inode = sb_dqopt(sb)->files[type]; 7252 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 7253 int offset = off & (sb->s_blocksize - 1); 7254 int tocopy; 7255 size_t toread; 7256 struct buffer_head *bh; 7257 loff_t i_size = i_size_read(inode); 7258 7259 if (off > i_size) 7260 return 0; 7261 if (off+len > i_size) 7262 len = i_size-off; 7263 toread = len; 7264 while (toread > 0) { 7265 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 7266 bh = ext4_bread(NULL, inode, blk, 0); 7267 if (IS_ERR(bh)) 7268 return PTR_ERR(bh); 7269 if (!bh) /* A hole? */ 7270 memset(data, 0, tocopy); 7271 else 7272 memcpy(data, bh->b_data+offset, tocopy); 7273 brelse(bh); 7274 offset = 0; 7275 toread -= tocopy; 7276 data += tocopy; 7277 blk++; 7278 } 7279 return len; 7280 } 7281 7282 /* Write to quotafile (we know the transaction is already started and has 7283 * enough credits) */ 7284 static ssize_t ext4_quota_write(struct super_block *sb, int type, 7285 const char *data, size_t len, loff_t off) 7286 { 7287 struct inode *inode = sb_dqopt(sb)->files[type]; 7288 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 7289 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1); 7290 int retries = 0; 7291 struct buffer_head *bh; 7292 handle_t *handle = journal_current_handle(); 7293 7294 if (!handle) { 7295 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 7296 " cancelled because transaction is not started", 7297 (unsigned long long)off, (unsigned long long)len); 7298 return -EIO; 7299 } 7300 /* 7301 * Since we account only one data block in transaction credits, 7302 * then it is impossible to cross a block boundary. 7303 */ 7304 if (sb->s_blocksize - offset < len) { 7305 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 7306 " cancelled because not block aligned", 7307 (unsigned long long)off, (unsigned long long)len); 7308 return -EIO; 7309 } 7310 7311 do { 7312 bh = ext4_bread(handle, inode, blk, 7313 EXT4_GET_BLOCKS_CREATE | 7314 EXT4_GET_BLOCKS_METADATA_NOFAIL); 7315 } while (PTR_ERR(bh) == -ENOSPC && 7316 ext4_should_retry_alloc(inode->i_sb, &retries)); 7317 if (IS_ERR(bh)) 7318 return PTR_ERR(bh); 7319 if (!bh) 7320 goto out; 7321 BUFFER_TRACE(bh, "get write access"); 7322 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE); 7323 if (err) { 7324 brelse(bh); 7325 return err; 7326 } 7327 lock_buffer(bh); 7328 memcpy(bh->b_data+offset, data, len); 7329 flush_dcache_folio(bh->b_folio); 7330 unlock_buffer(bh); 7331 err = ext4_handle_dirty_metadata(handle, NULL, bh); 7332 brelse(bh); 7333 out: 7334 if (inode->i_size < off + len) { 7335 i_size_write(inode, off + len); 7336 EXT4_I(inode)->i_disksize = inode->i_size; 7337 err2 = ext4_mark_inode_dirty(handle, inode); 7338 if (unlikely(err2 && !err)) 7339 err = err2; 7340 } 7341 return err ? err : len; 7342 } 7343 #endif 7344 7345 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 7346 static inline void register_as_ext2(void) 7347 { 7348 int err = register_filesystem(&ext2_fs_type); 7349 if (err) 7350 printk(KERN_WARNING 7351 "EXT4-fs: Unable to register as ext2 (%d)\n", err); 7352 } 7353 7354 static inline void unregister_as_ext2(void) 7355 { 7356 unregister_filesystem(&ext2_fs_type); 7357 } 7358 7359 static inline int ext2_feature_set_ok(struct super_block *sb) 7360 { 7361 if (ext4_has_unknown_ext2_incompat_features(sb)) 7362 return 0; 7363 if (sb_rdonly(sb)) 7364 return 1; 7365 if (ext4_has_unknown_ext2_ro_compat_features(sb)) 7366 return 0; 7367 return 1; 7368 } 7369 #else 7370 static inline void register_as_ext2(void) { } 7371 static inline void unregister_as_ext2(void) { } 7372 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 7373 #endif 7374 7375 static inline void register_as_ext3(void) 7376 { 7377 int err = register_filesystem(&ext3_fs_type); 7378 if (err) 7379 printk(KERN_WARNING 7380 "EXT4-fs: Unable to register as ext3 (%d)\n", err); 7381 } 7382 7383 static inline void unregister_as_ext3(void) 7384 { 7385 unregister_filesystem(&ext3_fs_type); 7386 } 7387 7388 static inline int ext3_feature_set_ok(struct super_block *sb) 7389 { 7390 if (ext4_has_unknown_ext3_incompat_features(sb)) 7391 return 0; 7392 if (!ext4_has_feature_journal(sb)) 7393 return 0; 7394 if (sb_rdonly(sb)) 7395 return 1; 7396 if (ext4_has_unknown_ext3_ro_compat_features(sb)) 7397 return 0; 7398 return 1; 7399 } 7400 7401 static void ext4_kill_sb(struct super_block *sb) 7402 { 7403 struct ext4_sb_info *sbi = EXT4_SB(sb); 7404 struct file *bdev_file = sbi ? sbi->s_journal_bdev_file : NULL; 7405 7406 kill_block_super(sb); 7407 7408 if (bdev_file) 7409 bdev_fput(bdev_file); 7410 } 7411 7412 static struct file_system_type ext4_fs_type = { 7413 .owner = THIS_MODULE, 7414 .name = "ext4", 7415 .init_fs_context = ext4_init_fs_context, 7416 .parameters = ext4_param_specs, 7417 .kill_sb = ext4_kill_sb, 7418 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP | FS_MGTIME, 7419 }; 7420 MODULE_ALIAS_FS("ext4"); 7421 7422 static int __init ext4_init_fs(void) 7423 { 7424 int err; 7425 7426 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 7427 ext4_li_info = NULL; 7428 7429 /* Build-time check for flags consistency */ 7430 ext4_check_flag_values(); 7431 7432 err = ext4_init_es(); 7433 if (err) 7434 return err; 7435 7436 err = ext4_init_pending(); 7437 if (err) 7438 goto out7; 7439 7440 err = ext4_init_post_read_processing(); 7441 if (err) 7442 goto out6; 7443 7444 err = ext4_init_pageio(); 7445 if (err) 7446 goto out5; 7447 7448 err = ext4_init_system_zone(); 7449 if (err) 7450 goto out4; 7451 7452 err = ext4_init_sysfs(); 7453 if (err) 7454 goto out3; 7455 7456 err = ext4_init_mballoc(); 7457 if (err) 7458 goto out2; 7459 err = init_inodecache(); 7460 if (err) 7461 goto out1; 7462 7463 err = ext4_fc_init_dentry_cache(); 7464 if (err) 7465 goto out05; 7466 7467 register_as_ext3(); 7468 register_as_ext2(); 7469 err = register_filesystem(&ext4_fs_type); 7470 if (err) 7471 goto out; 7472 7473 return 0; 7474 out: 7475 unregister_as_ext2(); 7476 unregister_as_ext3(); 7477 ext4_fc_destroy_dentry_cache(); 7478 out05: 7479 destroy_inodecache(); 7480 out1: 7481 ext4_exit_mballoc(); 7482 out2: 7483 ext4_exit_sysfs(); 7484 out3: 7485 ext4_exit_system_zone(); 7486 out4: 7487 ext4_exit_pageio(); 7488 out5: 7489 ext4_exit_post_read_processing(); 7490 out6: 7491 ext4_exit_pending(); 7492 out7: 7493 ext4_exit_es(); 7494 7495 return err; 7496 } 7497 7498 static void __exit ext4_exit_fs(void) 7499 { 7500 ext4_destroy_lazyinit_thread(); 7501 unregister_as_ext2(); 7502 unregister_as_ext3(); 7503 unregister_filesystem(&ext4_fs_type); 7504 ext4_fc_destroy_dentry_cache(); 7505 destroy_inodecache(); 7506 ext4_exit_mballoc(); 7507 ext4_exit_sysfs(); 7508 ext4_exit_system_zone(); 7509 ext4_exit_pageio(); 7510 ext4_exit_post_read_processing(); 7511 ext4_exit_es(); 7512 ext4_exit_pending(); 7513 } 7514 7515 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 7516 MODULE_DESCRIPTION("Fourth Extended Filesystem"); 7517 MODULE_LICENSE("GPL"); 7518 module_init(ext4_init_fs) 7519 module_exit(ext4_exit_fs) 7520