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/cleancache.h> 43 #include <linux/uaccess.h> 44 #include <linux/iversion.h> 45 #include <linux/unicode.h> 46 #include <linux/part_stat.h> 47 #include <linux/kthread.h> 48 #include <linux/freezer.h> 49 #include <linux/fsnotify.h> 50 51 #include "ext4.h" 52 #include "ext4_extents.h" /* Needed for trace points definition */ 53 #include "ext4_jbd2.h" 54 #include "xattr.h" 55 #include "acl.h" 56 #include "mballoc.h" 57 #include "fsmap.h" 58 59 #define CREATE_TRACE_POINTS 60 #include <trace/events/ext4.h> 61 62 static struct ext4_lazy_init *ext4_li_info; 63 static DEFINE_MUTEX(ext4_li_mtx); 64 static struct ratelimit_state ext4_mount_msg_ratelimit; 65 66 static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 67 unsigned long journal_devnum); 68 static int ext4_show_options(struct seq_file *seq, struct dentry *root); 69 static void ext4_update_super(struct super_block *sb); 70 static int ext4_commit_super(struct super_block *sb); 71 static int ext4_mark_recovery_complete(struct super_block *sb, 72 struct ext4_super_block *es); 73 static int ext4_clear_journal_err(struct super_block *sb, 74 struct ext4_super_block *es); 75 static int ext4_sync_fs(struct super_block *sb, int wait); 76 static int ext4_remount(struct super_block *sb, int *flags, char *data); 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 struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 81 const char *dev_name, void *data); 82 static inline int ext2_feature_set_ok(struct super_block *sb); 83 static inline int ext3_feature_set_ok(struct super_block *sb); 84 static void ext4_destroy_lazyinit_thread(void); 85 static void ext4_unregister_li_request(struct super_block *sb); 86 static void ext4_clear_request_list(void); 87 static struct inode *ext4_get_journal_inode(struct super_block *sb, 88 unsigned int journal_inum); 89 90 /* 91 * Lock ordering 92 * 93 * page fault path: 94 * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start 95 * -> page lock -> i_data_sem (rw) 96 * 97 * buffered write path: 98 * sb_start_write -> i_mutex -> mmap_lock 99 * sb_start_write -> i_mutex -> transaction start -> page lock -> 100 * i_data_sem (rw) 101 * 102 * truncate: 103 * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) -> 104 * page lock 105 * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start -> 106 * i_data_sem (rw) 107 * 108 * direct IO: 109 * sb_start_write -> i_mutex -> mmap_lock 110 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw) 111 * 112 * writepages: 113 * transaction start -> page lock(s) -> i_data_sem (rw) 114 */ 115 116 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 117 static struct file_system_type ext2_fs_type = { 118 .owner = THIS_MODULE, 119 .name = "ext2", 120 .mount = ext4_mount, 121 .kill_sb = kill_block_super, 122 .fs_flags = FS_REQUIRES_DEV, 123 }; 124 MODULE_ALIAS_FS("ext2"); 125 MODULE_ALIAS("ext2"); 126 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) 127 #else 128 #define IS_EXT2_SB(sb) (0) 129 #endif 130 131 132 static struct file_system_type ext3_fs_type = { 133 .owner = THIS_MODULE, 134 .name = "ext3", 135 .mount = ext4_mount, 136 .kill_sb = kill_block_super, 137 .fs_flags = FS_REQUIRES_DEV, 138 }; 139 MODULE_ALIAS_FS("ext3"); 140 MODULE_ALIAS("ext3"); 141 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) 142 143 144 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags, 145 bh_end_io_t *end_io) 146 { 147 /* 148 * buffer's verified bit is no longer valid after reading from 149 * disk again due to write out error, clear it to make sure we 150 * recheck the buffer contents. 151 */ 152 clear_buffer_verified(bh); 153 154 bh->b_end_io = end_io ? end_io : end_buffer_read_sync; 155 get_bh(bh); 156 submit_bh(REQ_OP_READ, op_flags, bh); 157 } 158 159 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags, 160 bh_end_io_t *end_io) 161 { 162 BUG_ON(!buffer_locked(bh)); 163 164 if (ext4_buffer_uptodate(bh)) { 165 unlock_buffer(bh); 166 return; 167 } 168 __ext4_read_bh(bh, op_flags, end_io); 169 } 170 171 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io) 172 { 173 BUG_ON(!buffer_locked(bh)); 174 175 if (ext4_buffer_uptodate(bh)) { 176 unlock_buffer(bh); 177 return 0; 178 } 179 180 __ext4_read_bh(bh, op_flags, end_io); 181 182 wait_on_buffer(bh); 183 if (buffer_uptodate(bh)) 184 return 0; 185 return -EIO; 186 } 187 188 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait) 189 { 190 if (trylock_buffer(bh)) { 191 if (wait) 192 return ext4_read_bh(bh, op_flags, NULL); 193 ext4_read_bh_nowait(bh, op_flags, NULL); 194 return 0; 195 } 196 if (wait) { 197 wait_on_buffer(bh); 198 if (buffer_uptodate(bh)) 199 return 0; 200 return -EIO; 201 } 202 return 0; 203 } 204 205 /* 206 * This works like __bread_gfp() except it uses ERR_PTR for error 207 * returns. Currently with sb_bread it's impossible to distinguish 208 * between ENOMEM and EIO situations (since both result in a NULL 209 * return. 210 */ 211 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb, 212 sector_t block, int op_flags, 213 gfp_t gfp) 214 { 215 struct buffer_head *bh; 216 int ret; 217 218 bh = sb_getblk_gfp(sb, block, gfp); 219 if (bh == NULL) 220 return ERR_PTR(-ENOMEM); 221 if (ext4_buffer_uptodate(bh)) 222 return bh; 223 224 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true); 225 if (ret) { 226 put_bh(bh); 227 return ERR_PTR(ret); 228 } 229 return bh; 230 } 231 232 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block, 233 int op_flags) 234 { 235 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE); 236 } 237 238 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb, 239 sector_t block) 240 { 241 return __ext4_sb_bread_gfp(sb, block, 0, 0); 242 } 243 244 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block) 245 { 246 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0); 247 248 if (likely(bh)) { 249 ext4_read_bh_lock(bh, REQ_RAHEAD, false); 250 brelse(bh); 251 } 252 } 253 254 static int ext4_verify_csum_type(struct super_block *sb, 255 struct ext4_super_block *es) 256 { 257 if (!ext4_has_feature_metadata_csum(sb)) 258 return 1; 259 260 return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 261 } 262 263 static __le32 ext4_superblock_csum(struct super_block *sb, 264 struct ext4_super_block *es) 265 { 266 struct ext4_sb_info *sbi = EXT4_SB(sb); 267 int offset = offsetof(struct ext4_super_block, s_checksum); 268 __u32 csum; 269 270 csum = ext4_chksum(sbi, ~0, (char *)es, offset); 271 272 return cpu_to_le32(csum); 273 } 274 275 static int ext4_superblock_csum_verify(struct super_block *sb, 276 struct ext4_super_block *es) 277 { 278 if (!ext4_has_metadata_csum(sb)) 279 return 1; 280 281 return es->s_checksum == ext4_superblock_csum(sb, es); 282 } 283 284 void ext4_superblock_csum_set(struct super_block *sb) 285 { 286 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 287 288 if (!ext4_has_metadata_csum(sb)) 289 return; 290 291 es->s_checksum = ext4_superblock_csum(sb, es); 292 } 293 294 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 295 struct ext4_group_desc *bg) 296 { 297 return le32_to_cpu(bg->bg_block_bitmap_lo) | 298 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 299 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 300 } 301 302 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 303 struct ext4_group_desc *bg) 304 { 305 return le32_to_cpu(bg->bg_inode_bitmap_lo) | 306 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 307 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 308 } 309 310 ext4_fsblk_t ext4_inode_table(struct super_block *sb, 311 struct ext4_group_desc *bg) 312 { 313 return le32_to_cpu(bg->bg_inode_table_lo) | 314 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 315 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 316 } 317 318 __u32 ext4_free_group_clusters(struct super_block *sb, 319 struct ext4_group_desc *bg) 320 { 321 return le16_to_cpu(bg->bg_free_blocks_count_lo) | 322 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 323 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 324 } 325 326 __u32 ext4_free_inodes_count(struct super_block *sb, 327 struct ext4_group_desc *bg) 328 { 329 return le16_to_cpu(bg->bg_free_inodes_count_lo) | 330 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 331 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); 332 } 333 334 __u32 ext4_used_dirs_count(struct super_block *sb, 335 struct ext4_group_desc *bg) 336 { 337 return le16_to_cpu(bg->bg_used_dirs_count_lo) | 338 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 339 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 340 } 341 342 __u32 ext4_itable_unused_count(struct super_block *sb, 343 struct ext4_group_desc *bg) 344 { 345 return le16_to_cpu(bg->bg_itable_unused_lo) | 346 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 347 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 348 } 349 350 void ext4_block_bitmap_set(struct super_block *sb, 351 struct ext4_group_desc *bg, ext4_fsblk_t blk) 352 { 353 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 354 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 355 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 356 } 357 358 void ext4_inode_bitmap_set(struct super_block *sb, 359 struct ext4_group_desc *bg, ext4_fsblk_t blk) 360 { 361 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 362 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 363 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 364 } 365 366 void ext4_inode_table_set(struct super_block *sb, 367 struct ext4_group_desc *bg, ext4_fsblk_t blk) 368 { 369 bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 370 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 371 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 372 } 373 374 void ext4_free_group_clusters_set(struct super_block *sb, 375 struct ext4_group_desc *bg, __u32 count) 376 { 377 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 378 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 379 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 380 } 381 382 void ext4_free_inodes_set(struct super_block *sb, 383 struct ext4_group_desc *bg, __u32 count) 384 { 385 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); 386 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 387 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); 388 } 389 390 void ext4_used_dirs_set(struct super_block *sb, 391 struct ext4_group_desc *bg, __u32 count) 392 { 393 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 394 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 395 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 396 } 397 398 void ext4_itable_unused_set(struct super_block *sb, 399 struct ext4_group_desc *bg, __u32 count) 400 { 401 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 402 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 403 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 404 } 405 406 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now) 407 { 408 now = clamp_val(now, 0, (1ull << 40) - 1); 409 410 *lo = cpu_to_le32(lower_32_bits(now)); 411 *hi = upper_32_bits(now); 412 } 413 414 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi) 415 { 416 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo); 417 } 418 #define ext4_update_tstamp(es, tstamp) \ 419 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \ 420 ktime_get_real_seconds()) 421 #define ext4_get_tstamp(es, tstamp) \ 422 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 423 424 /* 425 * The del_gendisk() function uninitializes the disk-specific data 426 * structures, including the bdi structure, without telling anyone 427 * else. Once this happens, any attempt to call mark_buffer_dirty() 428 * (for example, by ext4_commit_super), will cause a kernel OOPS. 429 * This is a kludge to prevent these oops until we can put in a proper 430 * hook in del_gendisk() to inform the VFS and file system layers. 431 */ 432 static int block_device_ejected(struct super_block *sb) 433 { 434 struct inode *bd_inode = sb->s_bdev->bd_inode; 435 struct backing_dev_info *bdi = inode_to_bdi(bd_inode); 436 437 return bdi->dev == NULL; 438 } 439 440 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 441 { 442 struct super_block *sb = journal->j_private; 443 struct ext4_sb_info *sbi = EXT4_SB(sb); 444 int error = is_journal_aborted(journal); 445 struct ext4_journal_cb_entry *jce; 446 447 BUG_ON(txn->t_state == T_FINISHED); 448 449 ext4_process_freed_data(sb, txn->t_tid); 450 451 spin_lock(&sbi->s_md_lock); 452 while (!list_empty(&txn->t_private_list)) { 453 jce = list_entry(txn->t_private_list.next, 454 struct ext4_journal_cb_entry, jce_list); 455 list_del_init(&jce->jce_list); 456 spin_unlock(&sbi->s_md_lock); 457 jce->jce_func(sb, jce, error); 458 spin_lock(&sbi->s_md_lock); 459 } 460 spin_unlock(&sbi->s_md_lock); 461 } 462 463 /* 464 * This writepage callback for write_cache_pages() 465 * takes care of a few cases after page cleaning. 466 * 467 * write_cache_pages() already checks for dirty pages 468 * and calls clear_page_dirty_for_io(), which we want, 469 * to write protect the pages. 470 * 471 * However, we may have to redirty a page (see below.) 472 */ 473 static int ext4_journalled_writepage_callback(struct page *page, 474 struct writeback_control *wbc, 475 void *data) 476 { 477 transaction_t *transaction = (transaction_t *) data; 478 struct buffer_head *bh, *head; 479 struct journal_head *jh; 480 481 bh = head = page_buffers(page); 482 do { 483 /* 484 * We have to redirty a page in these cases: 485 * 1) If buffer is dirty, it means the page was dirty because it 486 * contains a buffer that needs checkpointing. So the dirty bit 487 * needs to be preserved so that checkpointing writes the buffer 488 * properly. 489 * 2) If buffer is not part of the committing transaction 490 * (we may have just accidentally come across this buffer because 491 * inode range tracking is not exact) or if the currently running 492 * transaction already contains this buffer as well, dirty bit 493 * needs to be preserved so that the buffer gets writeprotected 494 * properly on running transaction's commit. 495 */ 496 jh = bh2jh(bh); 497 if (buffer_dirty(bh) || 498 (jh && (jh->b_transaction != transaction || 499 jh->b_next_transaction))) { 500 redirty_page_for_writepage(wbc, page); 501 goto out; 502 } 503 } while ((bh = bh->b_this_page) != head); 504 505 out: 506 return AOP_WRITEPAGE_ACTIVATE; 507 } 508 509 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode) 510 { 511 struct address_space *mapping = jinode->i_vfs_inode->i_mapping; 512 struct writeback_control wbc = { 513 .sync_mode = WB_SYNC_ALL, 514 .nr_to_write = LONG_MAX, 515 .range_start = jinode->i_dirty_start, 516 .range_end = jinode->i_dirty_end, 517 }; 518 519 return write_cache_pages(mapping, &wbc, 520 ext4_journalled_writepage_callback, 521 jinode->i_transaction); 522 } 523 524 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode) 525 { 526 int ret; 527 528 if (ext4_should_journal_data(jinode->i_vfs_inode)) 529 ret = ext4_journalled_submit_inode_data_buffers(jinode); 530 else 531 ret = jbd2_journal_submit_inode_data_buffers(jinode); 532 533 return ret; 534 } 535 536 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode) 537 { 538 int ret = 0; 539 540 if (!ext4_should_journal_data(jinode->i_vfs_inode)) 541 ret = jbd2_journal_finish_inode_data_buffers(jinode); 542 543 return ret; 544 } 545 546 static bool system_going_down(void) 547 { 548 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 549 || system_state == SYSTEM_RESTART; 550 } 551 552 struct ext4_err_translation { 553 int code; 554 int errno; 555 }; 556 557 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err } 558 559 static struct ext4_err_translation err_translation[] = { 560 EXT4_ERR_TRANSLATE(EIO), 561 EXT4_ERR_TRANSLATE(ENOMEM), 562 EXT4_ERR_TRANSLATE(EFSBADCRC), 563 EXT4_ERR_TRANSLATE(EFSCORRUPTED), 564 EXT4_ERR_TRANSLATE(ENOSPC), 565 EXT4_ERR_TRANSLATE(ENOKEY), 566 EXT4_ERR_TRANSLATE(EROFS), 567 EXT4_ERR_TRANSLATE(EFBIG), 568 EXT4_ERR_TRANSLATE(EEXIST), 569 EXT4_ERR_TRANSLATE(ERANGE), 570 EXT4_ERR_TRANSLATE(EOVERFLOW), 571 EXT4_ERR_TRANSLATE(EBUSY), 572 EXT4_ERR_TRANSLATE(ENOTDIR), 573 EXT4_ERR_TRANSLATE(ENOTEMPTY), 574 EXT4_ERR_TRANSLATE(ESHUTDOWN), 575 EXT4_ERR_TRANSLATE(EFAULT), 576 }; 577 578 static int ext4_errno_to_code(int errno) 579 { 580 int i; 581 582 for (i = 0; i < ARRAY_SIZE(err_translation); i++) 583 if (err_translation[i].errno == errno) 584 return err_translation[i].code; 585 return EXT4_ERR_UNKNOWN; 586 } 587 588 static void save_error_info(struct super_block *sb, int error, 589 __u32 ino, __u64 block, 590 const char *func, unsigned int line) 591 { 592 struct ext4_sb_info *sbi = EXT4_SB(sb); 593 594 /* We default to EFSCORRUPTED error... */ 595 if (error == 0) 596 error = EFSCORRUPTED; 597 598 spin_lock(&sbi->s_error_lock); 599 sbi->s_add_error_count++; 600 sbi->s_last_error_code = error; 601 sbi->s_last_error_line = line; 602 sbi->s_last_error_ino = ino; 603 sbi->s_last_error_block = block; 604 sbi->s_last_error_func = func; 605 sbi->s_last_error_time = ktime_get_real_seconds(); 606 if (!sbi->s_first_error_time) { 607 sbi->s_first_error_code = error; 608 sbi->s_first_error_line = line; 609 sbi->s_first_error_ino = ino; 610 sbi->s_first_error_block = block; 611 sbi->s_first_error_func = func; 612 sbi->s_first_error_time = sbi->s_last_error_time; 613 } 614 spin_unlock(&sbi->s_error_lock); 615 } 616 617 /* Deal with the reporting of failure conditions on a filesystem such as 618 * inconsistencies detected or read IO failures. 619 * 620 * On ext2, we can store the error state of the filesystem in the 621 * superblock. That is not possible on ext4, because we may have other 622 * write ordering constraints on the superblock which prevent us from 623 * writing it out straight away; and given that the journal is about to 624 * be aborted, we can't rely on the current, or future, transactions to 625 * write out the superblock safely. 626 * 627 * We'll just use the jbd2_journal_abort() error code to record an error in 628 * the journal instead. On recovery, the journal will complain about 629 * that error until we've noted it down and cleared it. 630 * 631 * If force_ro is set, we unconditionally force the filesystem into an 632 * ABORT|READONLY state, unless the error response on the fs has been set to 633 * panic in which case we take the easy way out and panic immediately. This is 634 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM 635 * at a critical moment in log management. 636 */ 637 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error, 638 __u32 ino, __u64 block, 639 const char *func, unsigned int line) 640 { 641 journal_t *journal = EXT4_SB(sb)->s_journal; 642 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT); 643 644 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 645 if (test_opt(sb, WARN_ON_ERROR)) 646 WARN_ON_ONCE(1); 647 648 if (!continue_fs && !sb_rdonly(sb)) { 649 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED); 650 if (journal) 651 jbd2_journal_abort(journal, -EIO); 652 } 653 654 if (!bdev_read_only(sb->s_bdev)) { 655 save_error_info(sb, error, ino, block, func, line); 656 /* 657 * In case the fs should keep running, we need to writeout 658 * superblock through the journal. Due to lock ordering 659 * constraints, it may not be safe to do it right here so we 660 * defer superblock flushing to a workqueue. 661 */ 662 if (continue_fs && journal) 663 schedule_work(&EXT4_SB(sb)->s_error_work); 664 else 665 ext4_commit_super(sb); 666 } 667 668 /* 669 * We force ERRORS_RO behavior when system is rebooting. Otherwise we 670 * could panic during 'reboot -f' as the underlying device got already 671 * disabled. 672 */ 673 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) { 674 panic("EXT4-fs (device %s): panic forced after error\n", 675 sb->s_id); 676 } 677 678 if (sb_rdonly(sb) || continue_fs) 679 return; 680 681 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 682 /* 683 * Make sure updated value of ->s_mount_flags will be visible before 684 * ->s_flags update 685 */ 686 smp_wmb(); 687 sb->s_flags |= SB_RDONLY; 688 } 689 690 static void flush_stashed_error_work(struct work_struct *work) 691 { 692 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info, 693 s_error_work); 694 journal_t *journal = sbi->s_journal; 695 handle_t *handle; 696 697 /* 698 * If the journal is still running, we have to write out superblock 699 * through the journal to avoid collisions of other journalled sb 700 * updates. 701 * 702 * We use directly jbd2 functions here to avoid recursing back into 703 * ext4 error handling code during handling of previous errors. 704 */ 705 if (!sb_rdonly(sbi->s_sb) && journal) { 706 struct buffer_head *sbh = sbi->s_sbh; 707 handle = jbd2_journal_start(journal, 1); 708 if (IS_ERR(handle)) 709 goto write_directly; 710 if (jbd2_journal_get_write_access(handle, sbh)) { 711 jbd2_journal_stop(handle); 712 goto write_directly; 713 } 714 ext4_update_super(sbi->s_sb); 715 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 716 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to " 717 "superblock detected"); 718 clear_buffer_write_io_error(sbh); 719 set_buffer_uptodate(sbh); 720 } 721 722 if (jbd2_journal_dirty_metadata(handle, sbh)) { 723 jbd2_journal_stop(handle); 724 goto write_directly; 725 } 726 jbd2_journal_stop(handle); 727 ext4_notify_error_sysfs(sbi); 728 return; 729 } 730 write_directly: 731 /* 732 * Write through journal failed. Write sb directly to get error info 733 * out and hope for the best. 734 */ 735 ext4_commit_super(sbi->s_sb); 736 ext4_notify_error_sysfs(sbi); 737 } 738 739 #define ext4_error_ratelimit(sb) \ 740 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ 741 "EXT4-fs error") 742 743 void __ext4_error(struct super_block *sb, const char *function, 744 unsigned int line, bool force_ro, int error, __u64 block, 745 const char *fmt, ...) 746 { 747 struct va_format vaf; 748 va_list args; 749 750 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 751 return; 752 753 trace_ext4_error(sb, function, line); 754 if (ext4_error_ratelimit(sb)) { 755 va_start(args, fmt); 756 vaf.fmt = fmt; 757 vaf.va = &args; 758 printk(KERN_CRIT 759 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 760 sb->s_id, function, line, current->comm, &vaf); 761 va_end(args); 762 } 763 fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED); 764 765 ext4_handle_error(sb, force_ro, error, 0, block, function, line); 766 } 767 768 void __ext4_error_inode(struct inode *inode, const char *function, 769 unsigned int line, ext4_fsblk_t block, int error, 770 const char *fmt, ...) 771 { 772 va_list args; 773 struct va_format vaf; 774 775 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 776 return; 777 778 trace_ext4_error(inode->i_sb, function, line); 779 if (ext4_error_ratelimit(inode->i_sb)) { 780 va_start(args, fmt); 781 vaf.fmt = fmt; 782 vaf.va = &args; 783 if (block) 784 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 785 "inode #%lu: block %llu: comm %s: %pV\n", 786 inode->i_sb->s_id, function, line, inode->i_ino, 787 block, current->comm, &vaf); 788 else 789 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 790 "inode #%lu: comm %s: %pV\n", 791 inode->i_sb->s_id, function, line, inode->i_ino, 792 current->comm, &vaf); 793 va_end(args); 794 } 795 fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED); 796 797 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block, 798 function, line); 799 } 800 801 void __ext4_error_file(struct file *file, const char *function, 802 unsigned int line, ext4_fsblk_t block, 803 const char *fmt, ...) 804 { 805 va_list args; 806 struct va_format vaf; 807 struct inode *inode = file_inode(file); 808 char pathname[80], *path; 809 810 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 811 return; 812 813 trace_ext4_error(inode->i_sb, function, line); 814 if (ext4_error_ratelimit(inode->i_sb)) { 815 path = file_path(file, pathname, sizeof(pathname)); 816 if (IS_ERR(path)) 817 path = "(unknown)"; 818 va_start(args, fmt); 819 vaf.fmt = fmt; 820 vaf.va = &args; 821 if (block) 822 printk(KERN_CRIT 823 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 824 "block %llu: comm %s: path %s: %pV\n", 825 inode->i_sb->s_id, function, line, inode->i_ino, 826 block, current->comm, path, &vaf); 827 else 828 printk(KERN_CRIT 829 "EXT4-fs error (device %s): %s:%d: inode #%lu: " 830 "comm %s: path %s: %pV\n", 831 inode->i_sb->s_id, function, line, inode->i_ino, 832 current->comm, path, &vaf); 833 va_end(args); 834 } 835 fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED); 836 837 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block, 838 function, line); 839 } 840 841 const char *ext4_decode_error(struct super_block *sb, int errno, 842 char nbuf[16]) 843 { 844 char *errstr = NULL; 845 846 switch (errno) { 847 case -EFSCORRUPTED: 848 errstr = "Corrupt filesystem"; 849 break; 850 case -EFSBADCRC: 851 errstr = "Filesystem failed CRC"; 852 break; 853 case -EIO: 854 errstr = "IO failure"; 855 break; 856 case -ENOMEM: 857 errstr = "Out of memory"; 858 break; 859 case -EROFS: 860 if (!sb || (EXT4_SB(sb)->s_journal && 861 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 862 errstr = "Journal has aborted"; 863 else 864 errstr = "Readonly filesystem"; 865 break; 866 default: 867 /* If the caller passed in an extra buffer for unknown 868 * errors, textualise them now. Else we just return 869 * NULL. */ 870 if (nbuf) { 871 /* Check for truncated error codes... */ 872 if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 873 errstr = nbuf; 874 } 875 break; 876 } 877 878 return errstr; 879 } 880 881 /* __ext4_std_error decodes expected errors from journaling functions 882 * automatically and invokes the appropriate error response. */ 883 884 void __ext4_std_error(struct super_block *sb, const char *function, 885 unsigned int line, int errno) 886 { 887 char nbuf[16]; 888 const char *errstr; 889 890 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 891 return; 892 893 /* Special case: if the error is EROFS, and we're not already 894 * inside a transaction, then there's really no point in logging 895 * an error. */ 896 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb)) 897 return; 898 899 if (ext4_error_ratelimit(sb)) { 900 errstr = ext4_decode_error(sb, errno, nbuf); 901 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 902 sb->s_id, function, line, errstr); 903 } 904 fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED); 905 906 ext4_handle_error(sb, false, -errno, 0, 0, function, line); 907 } 908 909 void __ext4_msg(struct super_block *sb, 910 const char *prefix, const char *fmt, ...) 911 { 912 struct va_format vaf; 913 va_list args; 914 915 atomic_inc(&EXT4_SB(sb)->s_msg_count); 916 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs")) 917 return; 918 919 va_start(args, fmt); 920 vaf.fmt = fmt; 921 vaf.va = &args; 922 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 923 va_end(args); 924 } 925 926 static int ext4_warning_ratelimit(struct super_block *sb) 927 { 928 atomic_inc(&EXT4_SB(sb)->s_warning_count); 929 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), 930 "EXT4-fs warning"); 931 } 932 933 void __ext4_warning(struct super_block *sb, const char *function, 934 unsigned int line, const char *fmt, ...) 935 { 936 struct va_format vaf; 937 va_list args; 938 939 if (!ext4_warning_ratelimit(sb)) 940 return; 941 942 va_start(args, fmt); 943 vaf.fmt = fmt; 944 vaf.va = &args; 945 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 946 sb->s_id, function, line, &vaf); 947 va_end(args); 948 } 949 950 void __ext4_warning_inode(const struct inode *inode, const char *function, 951 unsigned int line, const char *fmt, ...) 952 { 953 struct va_format vaf; 954 va_list args; 955 956 if (!ext4_warning_ratelimit(inode->i_sb)) 957 return; 958 959 va_start(args, fmt); 960 vaf.fmt = fmt; 961 vaf.va = &args; 962 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: " 963 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id, 964 function, line, inode->i_ino, current->comm, &vaf); 965 va_end(args); 966 } 967 968 void __ext4_grp_locked_error(const char *function, unsigned int line, 969 struct super_block *sb, ext4_group_t grp, 970 unsigned long ino, ext4_fsblk_t block, 971 const char *fmt, ...) 972 __releases(bitlock) 973 __acquires(bitlock) 974 { 975 struct va_format vaf; 976 va_list args; 977 978 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 979 return; 980 981 trace_ext4_error(sb, function, line); 982 if (ext4_error_ratelimit(sb)) { 983 va_start(args, fmt); 984 vaf.fmt = fmt; 985 vaf.va = &args; 986 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 987 sb->s_id, function, line, grp); 988 if (ino) 989 printk(KERN_CONT "inode %lu: ", ino); 990 if (block) 991 printk(KERN_CONT "block %llu:", 992 (unsigned long long) block); 993 printk(KERN_CONT "%pV\n", &vaf); 994 va_end(args); 995 } 996 997 if (test_opt(sb, ERRORS_CONT)) { 998 if (test_opt(sb, WARN_ON_ERROR)) 999 WARN_ON_ONCE(1); 1000 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 1001 if (!bdev_read_only(sb->s_bdev)) { 1002 save_error_info(sb, EFSCORRUPTED, ino, block, function, 1003 line); 1004 schedule_work(&EXT4_SB(sb)->s_error_work); 1005 } 1006 return; 1007 } 1008 ext4_unlock_group(sb, grp); 1009 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line); 1010 /* 1011 * We only get here in the ERRORS_RO case; relocking the group 1012 * may be dangerous, but nothing bad will happen since the 1013 * filesystem will have already been marked read/only and the 1014 * journal has been aborted. We return 1 as a hint to callers 1015 * who might what to use the return value from 1016 * ext4_grp_locked_error() to distinguish between the 1017 * ERRORS_CONT and ERRORS_RO case, and perhaps return more 1018 * aggressively from the ext4 function in question, with a 1019 * more appropriate error code. 1020 */ 1021 ext4_lock_group(sb, grp); 1022 return; 1023 } 1024 1025 void ext4_mark_group_bitmap_corrupted(struct super_block *sb, 1026 ext4_group_t group, 1027 unsigned int flags) 1028 { 1029 struct ext4_sb_info *sbi = EXT4_SB(sb); 1030 struct ext4_group_info *grp = ext4_get_group_info(sb, group); 1031 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL); 1032 int ret; 1033 1034 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) { 1035 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, 1036 &grp->bb_state); 1037 if (!ret) 1038 percpu_counter_sub(&sbi->s_freeclusters_counter, 1039 grp->bb_free); 1040 } 1041 1042 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) { 1043 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, 1044 &grp->bb_state); 1045 if (!ret && gdp) { 1046 int count; 1047 1048 count = ext4_free_inodes_count(sb, gdp); 1049 percpu_counter_sub(&sbi->s_freeinodes_counter, 1050 count); 1051 } 1052 } 1053 } 1054 1055 void ext4_update_dynamic_rev(struct super_block *sb) 1056 { 1057 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 1058 1059 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 1060 return; 1061 1062 ext4_warning(sb, 1063 "updating to rev %d because of new feature flag, " 1064 "running e2fsck is recommended", 1065 EXT4_DYNAMIC_REV); 1066 1067 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 1068 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 1069 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 1070 /* leave es->s_feature_*compat flags alone */ 1071 /* es->s_uuid will be set by e2fsck if empty */ 1072 1073 /* 1074 * The rest of the superblock fields should be zero, and if not it 1075 * means they are likely already in use, so leave them alone. We 1076 * can leave it up to e2fsck to clean up any inconsistencies there. 1077 */ 1078 } 1079 1080 /* 1081 * Open the external journal device 1082 */ 1083 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 1084 { 1085 struct block_device *bdev; 1086 1087 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); 1088 if (IS_ERR(bdev)) 1089 goto fail; 1090 return bdev; 1091 1092 fail: 1093 ext4_msg(sb, KERN_ERR, 1094 "failed to open journal device unknown-block(%u,%u) %ld", 1095 MAJOR(dev), MINOR(dev), PTR_ERR(bdev)); 1096 return NULL; 1097 } 1098 1099 /* 1100 * Release the journal device 1101 */ 1102 static void ext4_blkdev_put(struct block_device *bdev) 1103 { 1104 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 1105 } 1106 1107 static void ext4_blkdev_remove(struct ext4_sb_info *sbi) 1108 { 1109 struct block_device *bdev; 1110 bdev = sbi->s_journal_bdev; 1111 if (bdev) { 1112 ext4_blkdev_put(bdev); 1113 sbi->s_journal_bdev = NULL; 1114 } 1115 } 1116 1117 static inline struct inode *orphan_list_entry(struct list_head *l) 1118 { 1119 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 1120 } 1121 1122 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 1123 { 1124 struct list_head *l; 1125 1126 ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 1127 le32_to_cpu(sbi->s_es->s_last_orphan)); 1128 1129 printk(KERN_ERR "sb_info orphan list:\n"); 1130 list_for_each(l, &sbi->s_orphan) { 1131 struct inode *inode = orphan_list_entry(l); 1132 printk(KERN_ERR " " 1133 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 1134 inode->i_sb->s_id, inode->i_ino, inode, 1135 inode->i_mode, inode->i_nlink, 1136 NEXT_ORPHAN(inode)); 1137 } 1138 } 1139 1140 #ifdef CONFIG_QUOTA 1141 static int ext4_quota_off(struct super_block *sb, int type); 1142 1143 static inline void ext4_quota_off_umount(struct super_block *sb) 1144 { 1145 int type; 1146 1147 /* Use our quota_off function to clear inode flags etc. */ 1148 for (type = 0; type < EXT4_MAXQUOTAS; type++) 1149 ext4_quota_off(sb, type); 1150 } 1151 1152 /* 1153 * This is a helper function which is used in the mount/remount 1154 * codepaths (which holds s_umount) to fetch the quota file name. 1155 */ 1156 static inline char *get_qf_name(struct super_block *sb, 1157 struct ext4_sb_info *sbi, 1158 int type) 1159 { 1160 return rcu_dereference_protected(sbi->s_qf_names[type], 1161 lockdep_is_held(&sb->s_umount)); 1162 } 1163 #else 1164 static inline void ext4_quota_off_umount(struct super_block *sb) 1165 { 1166 } 1167 #endif 1168 1169 static void ext4_put_super(struct super_block *sb) 1170 { 1171 struct ext4_sb_info *sbi = EXT4_SB(sb); 1172 struct ext4_super_block *es = sbi->s_es; 1173 struct buffer_head **group_desc; 1174 struct flex_groups **flex_groups; 1175 int aborted = 0; 1176 int i, err; 1177 1178 ext4_unregister_li_request(sb); 1179 ext4_quota_off_umount(sb); 1180 1181 flush_work(&sbi->s_error_work); 1182 destroy_workqueue(sbi->rsv_conversion_wq); 1183 ext4_release_orphan_info(sb); 1184 1185 /* 1186 * Unregister sysfs before destroying jbd2 journal. 1187 * Since we could still access attr_journal_task attribute via sysfs 1188 * path which could have sbi->s_journal->j_task as NULL 1189 */ 1190 ext4_unregister_sysfs(sb); 1191 1192 if (sbi->s_journal) { 1193 aborted = is_journal_aborted(sbi->s_journal); 1194 err = jbd2_journal_destroy(sbi->s_journal); 1195 sbi->s_journal = NULL; 1196 if ((err < 0) && !aborted) { 1197 ext4_abort(sb, -err, "Couldn't clean up the journal"); 1198 } 1199 } 1200 1201 ext4_es_unregister_shrinker(sbi); 1202 del_timer_sync(&sbi->s_err_report); 1203 ext4_release_system_zone(sb); 1204 ext4_mb_release(sb); 1205 ext4_ext_release(sb); 1206 1207 if (!sb_rdonly(sb) && !aborted) { 1208 ext4_clear_feature_journal_needs_recovery(sb); 1209 ext4_clear_feature_orphan_present(sb); 1210 es->s_state = cpu_to_le16(sbi->s_mount_state); 1211 } 1212 if (!sb_rdonly(sb)) 1213 ext4_commit_super(sb); 1214 1215 rcu_read_lock(); 1216 group_desc = rcu_dereference(sbi->s_group_desc); 1217 for (i = 0; i < sbi->s_gdb_count; i++) 1218 brelse(group_desc[i]); 1219 kvfree(group_desc); 1220 flex_groups = rcu_dereference(sbi->s_flex_groups); 1221 if (flex_groups) { 1222 for (i = 0; i < sbi->s_flex_groups_allocated; i++) 1223 kvfree(flex_groups[i]); 1224 kvfree(flex_groups); 1225 } 1226 rcu_read_unlock(); 1227 percpu_counter_destroy(&sbi->s_freeclusters_counter); 1228 percpu_counter_destroy(&sbi->s_freeinodes_counter); 1229 percpu_counter_destroy(&sbi->s_dirs_counter); 1230 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 1231 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit); 1232 percpu_free_rwsem(&sbi->s_writepages_rwsem); 1233 #ifdef CONFIG_QUOTA 1234 for (i = 0; i < EXT4_MAXQUOTAS; i++) 1235 kfree(get_qf_name(sb, sbi, i)); 1236 #endif 1237 1238 /* Debugging code just in case the in-memory inode orphan list 1239 * isn't empty. The on-disk one can be non-empty if we've 1240 * detected an error and taken the fs readonly, but the 1241 * in-memory list had better be clean by this point. */ 1242 if (!list_empty(&sbi->s_orphan)) 1243 dump_orphan_list(sb, sbi); 1244 ASSERT(list_empty(&sbi->s_orphan)); 1245 1246 sync_blockdev(sb->s_bdev); 1247 invalidate_bdev(sb->s_bdev); 1248 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) { 1249 /* 1250 * Invalidate the journal device's buffers. We don't want them 1251 * floating about in memory - the physical journal device may 1252 * hotswapped, and it breaks the `ro-after' testing code. 1253 */ 1254 sync_blockdev(sbi->s_journal_bdev); 1255 invalidate_bdev(sbi->s_journal_bdev); 1256 ext4_blkdev_remove(sbi); 1257 } 1258 1259 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 1260 sbi->s_ea_inode_cache = NULL; 1261 1262 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 1263 sbi->s_ea_block_cache = NULL; 1264 1265 ext4_stop_mmpd(sbi); 1266 1267 brelse(sbi->s_sbh); 1268 sb->s_fs_info = NULL; 1269 /* 1270 * Now that we are completely done shutting down the 1271 * superblock, we need to actually destroy the kobject. 1272 */ 1273 kobject_put(&sbi->s_kobj); 1274 wait_for_completion(&sbi->s_kobj_unregister); 1275 if (sbi->s_chksum_driver) 1276 crypto_free_shash(sbi->s_chksum_driver); 1277 kfree(sbi->s_blockgroup_lock); 1278 fs_put_dax(sbi->s_daxdev); 1279 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy); 1280 #ifdef CONFIG_UNICODE 1281 utf8_unload(sb->s_encoding); 1282 #endif 1283 kfree(sbi); 1284 } 1285 1286 static struct kmem_cache *ext4_inode_cachep; 1287 1288 /* 1289 * Called inside transaction, so use GFP_NOFS 1290 */ 1291 static struct inode *ext4_alloc_inode(struct super_block *sb) 1292 { 1293 struct ext4_inode_info *ei; 1294 1295 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); 1296 if (!ei) 1297 return NULL; 1298 1299 inode_set_iversion(&ei->vfs_inode, 1); 1300 spin_lock_init(&ei->i_raw_lock); 1301 INIT_LIST_HEAD(&ei->i_prealloc_list); 1302 atomic_set(&ei->i_prealloc_active, 0); 1303 spin_lock_init(&ei->i_prealloc_lock); 1304 ext4_es_init_tree(&ei->i_es_tree); 1305 rwlock_init(&ei->i_es_lock); 1306 INIT_LIST_HEAD(&ei->i_es_list); 1307 ei->i_es_all_nr = 0; 1308 ei->i_es_shk_nr = 0; 1309 ei->i_es_shrink_lblk = 0; 1310 ei->i_reserved_data_blocks = 0; 1311 spin_lock_init(&(ei->i_block_reservation_lock)); 1312 ext4_init_pending_tree(&ei->i_pending_tree); 1313 #ifdef CONFIG_QUOTA 1314 ei->i_reserved_quota = 0; 1315 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); 1316 #endif 1317 ei->jinode = NULL; 1318 INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 1319 spin_lock_init(&ei->i_completed_io_lock); 1320 ei->i_sync_tid = 0; 1321 ei->i_datasync_tid = 0; 1322 atomic_set(&ei->i_unwritten, 0); 1323 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 1324 ext4_fc_init_inode(&ei->vfs_inode); 1325 mutex_init(&ei->i_fc_lock); 1326 return &ei->vfs_inode; 1327 } 1328 1329 static int ext4_drop_inode(struct inode *inode) 1330 { 1331 int drop = generic_drop_inode(inode); 1332 1333 if (!drop) 1334 drop = fscrypt_drop_inode(inode); 1335 1336 trace_ext4_drop_inode(inode, drop); 1337 return drop; 1338 } 1339 1340 static void ext4_free_in_core_inode(struct inode *inode) 1341 { 1342 fscrypt_free_inode(inode); 1343 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) { 1344 pr_warn("%s: inode %ld still in fc list", 1345 __func__, inode->i_ino); 1346 } 1347 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 1348 } 1349 1350 static void ext4_destroy_inode(struct inode *inode) 1351 { 1352 if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 1353 ext4_msg(inode->i_sb, KERN_ERR, 1354 "Inode %lu (%p): orphan list check failed!", 1355 inode->i_ino, EXT4_I(inode)); 1356 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 1357 EXT4_I(inode), sizeof(struct ext4_inode_info), 1358 true); 1359 dump_stack(); 1360 } 1361 1362 if (EXT4_I(inode)->i_reserved_data_blocks) 1363 ext4_msg(inode->i_sb, KERN_ERR, 1364 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!", 1365 inode->i_ino, EXT4_I(inode), 1366 EXT4_I(inode)->i_reserved_data_blocks); 1367 } 1368 1369 static void init_once(void *foo) 1370 { 1371 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 1372 1373 INIT_LIST_HEAD(&ei->i_orphan); 1374 init_rwsem(&ei->xattr_sem); 1375 init_rwsem(&ei->i_data_sem); 1376 inode_init_once(&ei->vfs_inode); 1377 ext4_fc_init_inode(&ei->vfs_inode); 1378 } 1379 1380 static int __init init_inodecache(void) 1381 { 1382 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache", 1383 sizeof(struct ext4_inode_info), 0, 1384 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD| 1385 SLAB_ACCOUNT), 1386 offsetof(struct ext4_inode_info, i_data), 1387 sizeof_field(struct ext4_inode_info, i_data), 1388 init_once); 1389 if (ext4_inode_cachep == NULL) 1390 return -ENOMEM; 1391 return 0; 1392 } 1393 1394 static void destroy_inodecache(void) 1395 { 1396 /* 1397 * Make sure all delayed rcu free inodes are flushed before we 1398 * destroy cache. 1399 */ 1400 rcu_barrier(); 1401 kmem_cache_destroy(ext4_inode_cachep); 1402 } 1403 1404 void ext4_clear_inode(struct inode *inode) 1405 { 1406 ext4_fc_del(inode); 1407 invalidate_inode_buffers(inode); 1408 clear_inode(inode); 1409 ext4_discard_preallocations(inode, 0); 1410 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 1411 dquot_drop(inode); 1412 if (EXT4_I(inode)->jinode) { 1413 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 1414 EXT4_I(inode)->jinode); 1415 jbd2_free_inode(EXT4_I(inode)->jinode); 1416 EXT4_I(inode)->jinode = NULL; 1417 } 1418 fscrypt_put_encryption_info(inode); 1419 fsverity_cleanup_inode(inode); 1420 } 1421 1422 static struct inode *ext4_nfs_get_inode(struct super_block *sb, 1423 u64 ino, u32 generation) 1424 { 1425 struct inode *inode; 1426 1427 /* 1428 * Currently we don't know the generation for parent directory, so 1429 * a generation of 0 means "accept any" 1430 */ 1431 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE); 1432 if (IS_ERR(inode)) 1433 return ERR_CAST(inode); 1434 if (generation && inode->i_generation != generation) { 1435 iput(inode); 1436 return ERR_PTR(-ESTALE); 1437 } 1438 1439 return inode; 1440 } 1441 1442 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 1443 int fh_len, int fh_type) 1444 { 1445 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1446 ext4_nfs_get_inode); 1447 } 1448 1449 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 1450 int fh_len, int fh_type) 1451 { 1452 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1453 ext4_nfs_get_inode); 1454 } 1455 1456 static int ext4_nfs_commit_metadata(struct inode *inode) 1457 { 1458 struct writeback_control wbc = { 1459 .sync_mode = WB_SYNC_ALL 1460 }; 1461 1462 trace_ext4_nfs_commit_metadata(inode); 1463 return ext4_write_inode(inode, &wbc); 1464 } 1465 1466 #ifdef CONFIG_FS_ENCRYPTION 1467 static int ext4_get_context(struct inode *inode, void *ctx, size_t len) 1468 { 1469 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION, 1470 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len); 1471 } 1472 1473 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len, 1474 void *fs_data) 1475 { 1476 handle_t *handle = fs_data; 1477 int res, res2, credits, retries = 0; 1478 1479 /* 1480 * Encrypting the root directory is not allowed because e2fsck expects 1481 * lost+found to exist and be unencrypted, and encrypting the root 1482 * directory would imply encrypting the lost+found directory as well as 1483 * the filename "lost+found" itself. 1484 */ 1485 if (inode->i_ino == EXT4_ROOT_INO) 1486 return -EPERM; 1487 1488 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode))) 1489 return -EINVAL; 1490 1491 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX)) 1492 return -EOPNOTSUPP; 1493 1494 res = ext4_convert_inline_data(inode); 1495 if (res) 1496 return res; 1497 1498 /* 1499 * If a journal handle was specified, then the encryption context is 1500 * being set on a new inode via inheritance and is part of a larger 1501 * transaction to create the inode. Otherwise the encryption context is 1502 * being set on an existing inode in its own transaction. Only in the 1503 * latter case should the "retry on ENOSPC" logic be used. 1504 */ 1505 1506 if (handle) { 1507 res = ext4_xattr_set_handle(handle, inode, 1508 EXT4_XATTR_INDEX_ENCRYPTION, 1509 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 1510 ctx, len, 0); 1511 if (!res) { 1512 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1513 ext4_clear_inode_state(inode, 1514 EXT4_STATE_MAY_INLINE_DATA); 1515 /* 1516 * Update inode->i_flags - S_ENCRYPTED will be enabled, 1517 * S_DAX may be disabled 1518 */ 1519 ext4_set_inode_flags(inode, false); 1520 } 1521 return res; 1522 } 1523 1524 res = dquot_initialize(inode); 1525 if (res) 1526 return res; 1527 retry: 1528 res = ext4_xattr_set_credits(inode, len, false /* is_create */, 1529 &credits); 1530 if (res) 1531 return res; 1532 1533 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits); 1534 if (IS_ERR(handle)) 1535 return PTR_ERR(handle); 1536 1537 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION, 1538 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 1539 ctx, len, 0); 1540 if (!res) { 1541 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1542 /* 1543 * Update inode->i_flags - S_ENCRYPTED will be enabled, 1544 * S_DAX may be disabled 1545 */ 1546 ext4_set_inode_flags(inode, false); 1547 res = ext4_mark_inode_dirty(handle, inode); 1548 if (res) 1549 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty"); 1550 } 1551 res2 = ext4_journal_stop(handle); 1552 1553 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) 1554 goto retry; 1555 if (!res) 1556 res = res2; 1557 return res; 1558 } 1559 1560 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb) 1561 { 1562 return EXT4_SB(sb)->s_dummy_enc_policy.policy; 1563 } 1564 1565 static bool ext4_has_stable_inodes(struct super_block *sb) 1566 { 1567 return ext4_has_feature_stable_inodes(sb); 1568 } 1569 1570 static void ext4_get_ino_and_lblk_bits(struct super_block *sb, 1571 int *ino_bits_ret, int *lblk_bits_ret) 1572 { 1573 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count); 1574 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t); 1575 } 1576 1577 static const struct fscrypt_operations ext4_cryptops = { 1578 .key_prefix = "ext4:", 1579 .get_context = ext4_get_context, 1580 .set_context = ext4_set_context, 1581 .get_dummy_policy = ext4_get_dummy_policy, 1582 .empty_dir = ext4_empty_dir, 1583 .has_stable_inodes = ext4_has_stable_inodes, 1584 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits, 1585 }; 1586 #endif 1587 1588 #ifdef CONFIG_QUOTA 1589 static const char * const quotatypes[] = INITQFNAMES; 1590 #define QTYPE2NAME(t) (quotatypes[t]) 1591 1592 static int ext4_write_dquot(struct dquot *dquot); 1593 static int ext4_acquire_dquot(struct dquot *dquot); 1594 static int ext4_release_dquot(struct dquot *dquot); 1595 static int ext4_mark_dquot_dirty(struct dquot *dquot); 1596 static int ext4_write_info(struct super_block *sb, int type); 1597 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1598 const struct path *path); 1599 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1600 size_t len, loff_t off); 1601 static ssize_t ext4_quota_write(struct super_block *sb, int type, 1602 const char *data, size_t len, loff_t off); 1603 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 1604 unsigned int flags); 1605 1606 static struct dquot **ext4_get_dquots(struct inode *inode) 1607 { 1608 return EXT4_I(inode)->i_dquot; 1609 } 1610 1611 static const struct dquot_operations ext4_quota_operations = { 1612 .get_reserved_space = ext4_get_reserved_space, 1613 .write_dquot = ext4_write_dquot, 1614 .acquire_dquot = ext4_acquire_dquot, 1615 .release_dquot = ext4_release_dquot, 1616 .mark_dirty = ext4_mark_dquot_dirty, 1617 .write_info = ext4_write_info, 1618 .alloc_dquot = dquot_alloc, 1619 .destroy_dquot = dquot_destroy, 1620 .get_projid = ext4_get_projid, 1621 .get_inode_usage = ext4_get_inode_usage, 1622 .get_next_id = dquot_get_next_id, 1623 }; 1624 1625 static const struct quotactl_ops ext4_qctl_operations = { 1626 .quota_on = ext4_quota_on, 1627 .quota_off = ext4_quota_off, 1628 .quota_sync = dquot_quota_sync, 1629 .get_state = dquot_get_state, 1630 .set_info = dquot_set_dqinfo, 1631 .get_dqblk = dquot_get_dqblk, 1632 .set_dqblk = dquot_set_dqblk, 1633 .get_nextdqblk = dquot_get_next_dqblk, 1634 }; 1635 #endif 1636 1637 static const struct super_operations ext4_sops = { 1638 .alloc_inode = ext4_alloc_inode, 1639 .free_inode = ext4_free_in_core_inode, 1640 .destroy_inode = ext4_destroy_inode, 1641 .write_inode = ext4_write_inode, 1642 .dirty_inode = ext4_dirty_inode, 1643 .drop_inode = ext4_drop_inode, 1644 .evict_inode = ext4_evict_inode, 1645 .put_super = ext4_put_super, 1646 .sync_fs = ext4_sync_fs, 1647 .freeze_fs = ext4_freeze, 1648 .unfreeze_fs = ext4_unfreeze, 1649 .statfs = ext4_statfs, 1650 .remount_fs = ext4_remount, 1651 .show_options = ext4_show_options, 1652 #ifdef CONFIG_QUOTA 1653 .quota_read = ext4_quota_read, 1654 .quota_write = ext4_quota_write, 1655 .get_dquots = ext4_get_dquots, 1656 #endif 1657 }; 1658 1659 static const struct export_operations ext4_export_ops = { 1660 .fh_to_dentry = ext4_fh_to_dentry, 1661 .fh_to_parent = ext4_fh_to_parent, 1662 .get_parent = ext4_get_parent, 1663 .commit_metadata = ext4_nfs_commit_metadata, 1664 }; 1665 1666 enum { 1667 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1668 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 1669 Opt_nouid32, Opt_debug, Opt_removed, 1670 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 1671 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1672 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1673 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1674 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1675 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 1676 Opt_inlinecrypt, 1677 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 1678 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1679 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1680 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, 1681 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never, 1682 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error, 1683 Opt_nowarn_on_error, Opt_mblk_io_submit, 1684 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize, 1685 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 1686 Opt_inode_readahead_blks, Opt_journal_ioprio, 1687 Opt_dioread_nolock, Opt_dioread_lock, 1688 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1689 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, 1690 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan, 1691 #ifdef CONFIG_EXT4_DEBUG 1692 Opt_fc_debug_max_replay, Opt_fc_debug_force 1693 #endif 1694 }; 1695 1696 static const match_table_t tokens = { 1697 {Opt_bsd_df, "bsddf"}, 1698 {Opt_minix_df, "minixdf"}, 1699 {Opt_grpid, "grpid"}, 1700 {Opt_grpid, "bsdgroups"}, 1701 {Opt_nogrpid, "nogrpid"}, 1702 {Opt_nogrpid, "sysvgroups"}, 1703 {Opt_resgid, "resgid=%u"}, 1704 {Opt_resuid, "resuid=%u"}, 1705 {Opt_sb, "sb=%u"}, 1706 {Opt_err_cont, "errors=continue"}, 1707 {Opt_err_panic, "errors=panic"}, 1708 {Opt_err_ro, "errors=remount-ro"}, 1709 {Opt_nouid32, "nouid32"}, 1710 {Opt_debug, "debug"}, 1711 {Opt_removed, "oldalloc"}, 1712 {Opt_removed, "orlov"}, 1713 {Opt_user_xattr, "user_xattr"}, 1714 {Opt_nouser_xattr, "nouser_xattr"}, 1715 {Opt_acl, "acl"}, 1716 {Opt_noacl, "noacl"}, 1717 {Opt_noload, "norecovery"}, 1718 {Opt_noload, "noload"}, 1719 {Opt_removed, "nobh"}, 1720 {Opt_removed, "bh"}, 1721 {Opt_commit, "commit=%u"}, 1722 {Opt_min_batch_time, "min_batch_time=%u"}, 1723 {Opt_max_batch_time, "max_batch_time=%u"}, 1724 {Opt_journal_dev, "journal_dev=%u"}, 1725 {Opt_journal_path, "journal_path=%s"}, 1726 {Opt_journal_checksum, "journal_checksum"}, 1727 {Opt_nojournal_checksum, "nojournal_checksum"}, 1728 {Opt_journal_async_commit, "journal_async_commit"}, 1729 {Opt_abort, "abort"}, 1730 {Opt_data_journal, "data=journal"}, 1731 {Opt_data_ordered, "data=ordered"}, 1732 {Opt_data_writeback, "data=writeback"}, 1733 {Opt_data_err_abort, "data_err=abort"}, 1734 {Opt_data_err_ignore, "data_err=ignore"}, 1735 {Opt_offusrjquota, "usrjquota="}, 1736 {Opt_usrjquota, "usrjquota=%s"}, 1737 {Opt_offgrpjquota, "grpjquota="}, 1738 {Opt_grpjquota, "grpjquota=%s"}, 1739 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1740 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 1741 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1742 {Opt_grpquota, "grpquota"}, 1743 {Opt_noquota, "noquota"}, 1744 {Opt_quota, "quota"}, 1745 {Opt_usrquota, "usrquota"}, 1746 {Opt_prjquota, "prjquota"}, 1747 {Opt_barrier, "barrier=%u"}, 1748 {Opt_barrier, "barrier"}, 1749 {Opt_nobarrier, "nobarrier"}, 1750 {Opt_i_version, "i_version"}, 1751 {Opt_dax, "dax"}, 1752 {Opt_dax_always, "dax=always"}, 1753 {Opt_dax_inode, "dax=inode"}, 1754 {Opt_dax_never, "dax=never"}, 1755 {Opt_stripe, "stripe=%u"}, 1756 {Opt_delalloc, "delalloc"}, 1757 {Opt_warn_on_error, "warn_on_error"}, 1758 {Opt_nowarn_on_error, "nowarn_on_error"}, 1759 {Opt_lazytime, "lazytime"}, 1760 {Opt_nolazytime, "nolazytime"}, 1761 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"}, 1762 {Opt_nodelalloc, "nodelalloc"}, 1763 {Opt_removed, "mblk_io_submit"}, 1764 {Opt_removed, "nomblk_io_submit"}, 1765 {Opt_block_validity, "block_validity"}, 1766 {Opt_noblock_validity, "noblock_validity"}, 1767 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1768 {Opt_journal_ioprio, "journal_ioprio=%u"}, 1769 {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 1770 {Opt_auto_da_alloc, "auto_da_alloc"}, 1771 {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1772 {Opt_dioread_nolock, "dioread_nolock"}, 1773 {Opt_dioread_lock, "nodioread_nolock"}, 1774 {Opt_dioread_lock, "dioread_lock"}, 1775 {Opt_discard, "discard"}, 1776 {Opt_nodiscard, "nodiscard"}, 1777 {Opt_init_itable, "init_itable=%u"}, 1778 {Opt_init_itable, "init_itable"}, 1779 {Opt_noinit_itable, "noinit_itable"}, 1780 #ifdef CONFIG_EXT4_DEBUG 1781 {Opt_fc_debug_force, "fc_debug_force"}, 1782 {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"}, 1783 #endif 1784 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, 1785 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 1786 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 1787 {Opt_inlinecrypt, "inlinecrypt"}, 1788 {Opt_nombcache, "nombcache"}, 1789 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */ 1790 {Opt_removed, "prefetch_block_bitmaps"}, 1791 {Opt_no_prefetch_block_bitmaps, "no_prefetch_block_bitmaps"}, 1792 {Opt_mb_optimize_scan, "mb_optimize_scan=%d"}, 1793 {Opt_removed, "check=none"}, /* mount option from ext2/3 */ 1794 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ 1795 {Opt_removed, "reservation"}, /* mount option from ext2/3 */ 1796 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ 1797 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ 1798 {Opt_err, NULL}, 1799 }; 1800 1801 static ext4_fsblk_t get_sb_block(void **data) 1802 { 1803 ext4_fsblk_t sb_block; 1804 char *options = (char *) *data; 1805 1806 if (!options || strncmp(options, "sb=", 3) != 0) 1807 return 1; /* Default location */ 1808 1809 options += 3; 1810 /* TODO: use simple_strtoll with >32bit ext4 */ 1811 sb_block = simple_strtoul(options, &options, 0); 1812 if (*options && *options != ',') { 1813 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1814 (char *) *data); 1815 return 1; 1816 } 1817 if (*options == ',') 1818 options++; 1819 *data = (void *) options; 1820 1821 return sb_block; 1822 } 1823 1824 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1825 #define DEFAULT_MB_OPTIMIZE_SCAN (-1) 1826 1827 static const char deprecated_msg[] = 1828 "Mount option \"%s\" will be removed by %s\n" 1829 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1830 1831 #ifdef CONFIG_QUOTA 1832 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 1833 { 1834 struct ext4_sb_info *sbi = EXT4_SB(sb); 1835 char *qname, *old_qname = get_qf_name(sb, sbi, qtype); 1836 int ret = -1; 1837 1838 if (sb_any_quota_loaded(sb) && !old_qname) { 1839 ext4_msg(sb, KERN_ERR, 1840 "Cannot change journaled " 1841 "quota options when quota turned on"); 1842 return -1; 1843 } 1844 if (ext4_has_feature_quota(sb)) { 1845 ext4_msg(sb, KERN_INFO, "Journaled quota options " 1846 "ignored when QUOTA feature is enabled"); 1847 return 1; 1848 } 1849 qname = match_strdup(args); 1850 if (!qname) { 1851 ext4_msg(sb, KERN_ERR, 1852 "Not enough memory for storing quotafile name"); 1853 return -1; 1854 } 1855 if (old_qname) { 1856 if (strcmp(old_qname, qname) == 0) 1857 ret = 1; 1858 else 1859 ext4_msg(sb, KERN_ERR, 1860 "%s quota file already specified", 1861 QTYPE2NAME(qtype)); 1862 goto errout; 1863 } 1864 if (strchr(qname, '/')) { 1865 ext4_msg(sb, KERN_ERR, 1866 "quotafile must be on filesystem root"); 1867 goto errout; 1868 } 1869 rcu_assign_pointer(sbi->s_qf_names[qtype], qname); 1870 set_opt(sb, QUOTA); 1871 return 1; 1872 errout: 1873 kfree(qname); 1874 return ret; 1875 } 1876 1877 static int clear_qf_name(struct super_block *sb, int qtype) 1878 { 1879 1880 struct ext4_sb_info *sbi = EXT4_SB(sb); 1881 char *old_qname = get_qf_name(sb, sbi, qtype); 1882 1883 if (sb_any_quota_loaded(sb) && old_qname) { 1884 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 1885 " when quota turned on"); 1886 return -1; 1887 } 1888 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL); 1889 synchronize_rcu(); 1890 kfree(old_qname); 1891 return 1; 1892 } 1893 #endif 1894 1895 #define MOPT_SET 0x0001 1896 #define MOPT_CLEAR 0x0002 1897 #define MOPT_NOSUPPORT 0x0004 1898 #define MOPT_EXPLICIT 0x0008 1899 #define MOPT_CLEAR_ERR 0x0010 1900 #define MOPT_GTE0 0x0020 1901 #ifdef CONFIG_QUOTA 1902 #define MOPT_Q 0 1903 #define MOPT_QFMT 0x0040 1904 #else 1905 #define MOPT_Q MOPT_NOSUPPORT 1906 #define MOPT_QFMT MOPT_NOSUPPORT 1907 #endif 1908 #define MOPT_DATAJ 0x0080 1909 #define MOPT_NO_EXT2 0x0100 1910 #define MOPT_NO_EXT3 0x0200 1911 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1912 #define MOPT_STRING 0x0400 1913 #define MOPT_SKIP 0x0800 1914 #define MOPT_2 0x1000 1915 1916 static const struct mount_opts { 1917 int token; 1918 int mount_opt; 1919 int flags; 1920 } ext4_mount_opts[] = { 1921 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 1922 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 1923 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 1924 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 1925 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 1926 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 1927 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 1928 MOPT_EXT4_ONLY | MOPT_SET}, 1929 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 1930 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1931 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 1932 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 1933 {Opt_delalloc, EXT4_MOUNT_DELALLOC, 1934 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1935 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 1936 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1937 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET}, 1938 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR}, 1939 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1940 MOPT_EXT4_ONLY | MOPT_CLEAR}, 1941 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1942 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1943 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 1944 EXT4_MOUNT_JOURNAL_CHECKSUM), 1945 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 1946 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 1947 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, 1948 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, 1949 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, 1950 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, 1951 MOPT_NO_EXT2}, 1952 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, 1953 MOPT_NO_EXT2}, 1954 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 1955 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 1956 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 1957 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 1958 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 1959 {Opt_commit, 0, MOPT_GTE0}, 1960 {Opt_max_batch_time, 0, MOPT_GTE0}, 1961 {Opt_min_batch_time, 0, MOPT_GTE0}, 1962 {Opt_inode_readahead_blks, 0, MOPT_GTE0}, 1963 {Opt_init_itable, 0, MOPT_GTE0}, 1964 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP}, 1965 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS, 1966 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, 1967 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE, 1968 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, 1969 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER, 1970 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, 1971 {Opt_stripe, 0, MOPT_GTE0}, 1972 {Opt_resuid, 0, MOPT_GTE0}, 1973 {Opt_resgid, 0, MOPT_GTE0}, 1974 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 1975 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING}, 1976 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 1977 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1978 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 1979 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, 1980 MOPT_NO_EXT2 | MOPT_DATAJ}, 1981 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 1982 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 1983 #ifdef CONFIG_EXT4_FS_POSIX_ACL 1984 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 1985 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 1986 #else 1987 {Opt_acl, 0, MOPT_NOSUPPORT}, 1988 {Opt_noacl, 0, MOPT_NOSUPPORT}, 1989 #endif 1990 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 1991 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1992 {Opt_debug_want_extra_isize, 0, MOPT_GTE0}, 1993 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 1994 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 1995 MOPT_SET | MOPT_Q}, 1996 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 1997 MOPT_SET | MOPT_Q}, 1998 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, 1999 MOPT_SET | MOPT_Q}, 2000 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 2001 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), 2002 MOPT_CLEAR | MOPT_Q}, 2003 {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING}, 2004 {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING}, 2005 {Opt_offusrjquota, 0, MOPT_Q}, 2006 {Opt_offgrpjquota, 0, MOPT_Q}, 2007 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, 2008 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, 2009 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, 2010 {Opt_max_dir_size_kb, 0, MOPT_GTE0}, 2011 {Opt_test_dummy_encryption, 0, MOPT_STRING}, 2012 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, 2013 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS, 2014 MOPT_SET}, 2015 {Opt_mb_optimize_scan, EXT4_MOUNT2_MB_OPTIMIZE_SCAN, MOPT_GTE0}, 2016 #ifdef CONFIG_EXT4_DEBUG 2017 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT, 2018 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY}, 2019 {Opt_fc_debug_max_replay, 0, MOPT_GTE0}, 2020 #endif 2021 {Opt_err, 0, 0} 2022 }; 2023 2024 #ifdef CONFIG_UNICODE 2025 static const struct ext4_sb_encodings { 2026 __u16 magic; 2027 char *name; 2028 char *version; 2029 } ext4_sb_encoding_map[] = { 2030 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"}, 2031 }; 2032 2033 static int ext4_sb_read_encoding(const struct ext4_super_block *es, 2034 const struct ext4_sb_encodings **encoding, 2035 __u16 *flags) 2036 { 2037 __u16 magic = le16_to_cpu(es->s_encoding); 2038 int i; 2039 2040 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++) 2041 if (magic == ext4_sb_encoding_map[i].magic) 2042 break; 2043 2044 if (i >= ARRAY_SIZE(ext4_sb_encoding_map)) 2045 return -EINVAL; 2046 2047 *encoding = &ext4_sb_encoding_map[i]; 2048 *flags = le16_to_cpu(es->s_encoding_flags); 2049 2050 return 0; 2051 } 2052 #endif 2053 2054 static int ext4_set_test_dummy_encryption(struct super_block *sb, 2055 const char *opt, 2056 const substring_t *arg, 2057 bool is_remount) 2058 { 2059 #ifdef CONFIG_FS_ENCRYPTION 2060 struct ext4_sb_info *sbi = EXT4_SB(sb); 2061 int err; 2062 2063 /* 2064 * This mount option is just for testing, and it's not worthwhile to 2065 * implement the extra complexity (e.g. RCU protection) that would be 2066 * needed to allow it to be set or changed during remount. We do allow 2067 * it to be specified during remount, but only if there is no change. 2068 */ 2069 if (is_remount && !sbi->s_dummy_enc_policy.policy) { 2070 ext4_msg(sb, KERN_WARNING, 2071 "Can't set test_dummy_encryption on remount"); 2072 return -1; 2073 } 2074 err = fscrypt_set_test_dummy_encryption(sb, arg->from, 2075 &sbi->s_dummy_enc_policy); 2076 if (err) { 2077 if (err == -EEXIST) 2078 ext4_msg(sb, KERN_WARNING, 2079 "Can't change test_dummy_encryption on remount"); 2080 else if (err == -EINVAL) 2081 ext4_msg(sb, KERN_WARNING, 2082 "Value of option \"%s\" is unrecognized", opt); 2083 else 2084 ext4_msg(sb, KERN_WARNING, 2085 "Error processing option \"%s\" [%d]", 2086 opt, err); 2087 return -1; 2088 } 2089 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled"); 2090 #else 2091 ext4_msg(sb, KERN_WARNING, 2092 "Test dummy encryption mount option ignored"); 2093 #endif 2094 return 1; 2095 } 2096 2097 struct ext4_parsed_options { 2098 unsigned long journal_devnum; 2099 unsigned int journal_ioprio; 2100 int mb_optimize_scan; 2101 }; 2102 2103 static int handle_mount_opt(struct super_block *sb, char *opt, int token, 2104 substring_t *args, struct ext4_parsed_options *parsed_opts, 2105 int is_remount) 2106 { 2107 struct ext4_sb_info *sbi = EXT4_SB(sb); 2108 const struct mount_opts *m; 2109 kuid_t uid; 2110 kgid_t gid; 2111 int arg = 0; 2112 2113 #ifdef CONFIG_QUOTA 2114 if (token == Opt_usrjquota) 2115 return set_qf_name(sb, USRQUOTA, &args[0]); 2116 else if (token == Opt_grpjquota) 2117 return set_qf_name(sb, GRPQUOTA, &args[0]); 2118 else if (token == Opt_offusrjquota) 2119 return clear_qf_name(sb, USRQUOTA); 2120 else if (token == Opt_offgrpjquota) 2121 return clear_qf_name(sb, GRPQUOTA); 2122 #endif 2123 switch (token) { 2124 case Opt_noacl: 2125 case Opt_nouser_xattr: 2126 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); 2127 break; 2128 case Opt_sb: 2129 return 1; /* handled by get_sb_block() */ 2130 case Opt_removed: 2131 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt); 2132 return 1; 2133 case Opt_abort: 2134 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED); 2135 return 1; 2136 case Opt_i_version: 2137 sb->s_flags |= SB_I_VERSION; 2138 return 1; 2139 case Opt_lazytime: 2140 sb->s_flags |= SB_LAZYTIME; 2141 return 1; 2142 case Opt_nolazytime: 2143 sb->s_flags &= ~SB_LAZYTIME; 2144 return 1; 2145 case Opt_inlinecrypt: 2146 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 2147 sb->s_flags |= SB_INLINECRYPT; 2148 #else 2149 ext4_msg(sb, KERN_ERR, "inline encryption not supported"); 2150 #endif 2151 return 1; 2152 } 2153 2154 for (m = ext4_mount_opts; m->token != Opt_err; m++) 2155 if (token == m->token) 2156 break; 2157 2158 if (m->token == Opt_err) { 2159 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " 2160 "or missing value", opt); 2161 return -1; 2162 } 2163 2164 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 2165 ext4_msg(sb, KERN_ERR, 2166 "Mount option \"%s\" incompatible with ext2", opt); 2167 return -1; 2168 } 2169 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 2170 ext4_msg(sb, KERN_ERR, 2171 "Mount option \"%s\" incompatible with ext3", opt); 2172 return -1; 2173 } 2174 2175 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg)) 2176 return -1; 2177 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) 2178 return -1; 2179 if (m->flags & MOPT_EXPLICIT) { 2180 if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 2181 set_opt2(sb, EXPLICIT_DELALLOC); 2182 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 2183 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM); 2184 } else 2185 return -1; 2186 } 2187 if (m->flags & MOPT_CLEAR_ERR) 2188 clear_opt(sb, ERRORS_MASK); 2189 if (token == Opt_noquota && sb_any_quota_loaded(sb)) { 2190 ext4_msg(sb, KERN_ERR, "Cannot change quota " 2191 "options when quota turned on"); 2192 return -1; 2193 } 2194 2195 if (m->flags & MOPT_NOSUPPORT) { 2196 ext4_msg(sb, KERN_ERR, "%s option not supported", opt); 2197 } else if (token == Opt_commit) { 2198 if (arg == 0) 2199 arg = JBD2_DEFAULT_MAX_COMMIT_AGE; 2200 else if (arg > INT_MAX / HZ) { 2201 ext4_msg(sb, KERN_ERR, 2202 "Invalid commit interval %d, " 2203 "must be smaller than %d", 2204 arg, INT_MAX / HZ); 2205 return -1; 2206 } 2207 sbi->s_commit_interval = HZ * arg; 2208 } else if (token == Opt_debug_want_extra_isize) { 2209 if ((arg & 1) || 2210 (arg < 4) || 2211 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) { 2212 ext4_msg(sb, KERN_ERR, 2213 "Invalid want_extra_isize %d", arg); 2214 return -1; 2215 } 2216 sbi->s_want_extra_isize = arg; 2217 } else if (token == Opt_max_batch_time) { 2218 sbi->s_max_batch_time = arg; 2219 } else if (token == Opt_min_batch_time) { 2220 sbi->s_min_batch_time = arg; 2221 } else if (token == Opt_inode_readahead_blks) { 2222 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) { 2223 ext4_msg(sb, KERN_ERR, 2224 "EXT4-fs: inode_readahead_blks must be " 2225 "0 or a power of 2 smaller than 2^31"); 2226 return -1; 2227 } 2228 sbi->s_inode_readahead_blks = arg; 2229 } else if (token == Opt_init_itable) { 2230 set_opt(sb, INIT_INODE_TABLE); 2231 if (!args->from) 2232 arg = EXT4_DEF_LI_WAIT_MULT; 2233 sbi->s_li_wait_mult = arg; 2234 } else if (token == Opt_max_dir_size_kb) { 2235 sbi->s_max_dir_size_kb = arg; 2236 #ifdef CONFIG_EXT4_DEBUG 2237 } else if (token == Opt_fc_debug_max_replay) { 2238 sbi->s_fc_debug_max_replay = arg; 2239 #endif 2240 } else if (token == Opt_stripe) { 2241 sbi->s_stripe = arg; 2242 } else if (token == Opt_resuid) { 2243 uid = make_kuid(current_user_ns(), arg); 2244 if (!uid_valid(uid)) { 2245 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); 2246 return -1; 2247 } 2248 sbi->s_resuid = uid; 2249 } else if (token == Opt_resgid) { 2250 gid = make_kgid(current_user_ns(), arg); 2251 if (!gid_valid(gid)) { 2252 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); 2253 return -1; 2254 } 2255 sbi->s_resgid = gid; 2256 } else if (token == Opt_journal_dev) { 2257 if (is_remount) { 2258 ext4_msg(sb, KERN_ERR, 2259 "Cannot specify journal on remount"); 2260 return -1; 2261 } 2262 parsed_opts->journal_devnum = arg; 2263 } else if (token == Opt_journal_path) { 2264 char *journal_path; 2265 struct inode *journal_inode; 2266 struct path path; 2267 int error; 2268 2269 if (is_remount) { 2270 ext4_msg(sb, KERN_ERR, 2271 "Cannot specify journal on remount"); 2272 return -1; 2273 } 2274 journal_path = match_strdup(&args[0]); 2275 if (!journal_path) { 2276 ext4_msg(sb, KERN_ERR, "error: could not dup " 2277 "journal device string"); 2278 return -1; 2279 } 2280 2281 error = kern_path(journal_path, LOOKUP_FOLLOW, &path); 2282 if (error) { 2283 ext4_msg(sb, KERN_ERR, "error: could not find " 2284 "journal device path: error %d", error); 2285 kfree(journal_path); 2286 return -1; 2287 } 2288 2289 journal_inode = d_inode(path.dentry); 2290 if (!S_ISBLK(journal_inode->i_mode)) { 2291 ext4_msg(sb, KERN_ERR, "error: journal path %s " 2292 "is not a block device", journal_path); 2293 path_put(&path); 2294 kfree(journal_path); 2295 return -1; 2296 } 2297 2298 parsed_opts->journal_devnum = new_encode_dev(journal_inode->i_rdev); 2299 path_put(&path); 2300 kfree(journal_path); 2301 } else if (token == Opt_journal_ioprio) { 2302 if (arg > 7) { 2303 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority" 2304 " (must be 0-7)"); 2305 return -1; 2306 } 2307 parsed_opts->journal_ioprio = 2308 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); 2309 } else if (token == Opt_test_dummy_encryption) { 2310 return ext4_set_test_dummy_encryption(sb, opt, &args[0], 2311 is_remount); 2312 } else if (m->flags & MOPT_DATAJ) { 2313 if (is_remount) { 2314 if (!sbi->s_journal) 2315 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); 2316 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) { 2317 ext4_msg(sb, KERN_ERR, 2318 "Cannot change data mode on remount"); 2319 return -1; 2320 } 2321 } else { 2322 clear_opt(sb, DATA_FLAGS); 2323 sbi->s_mount_opt |= m->mount_opt; 2324 } 2325 #ifdef CONFIG_QUOTA 2326 } else if (m->flags & MOPT_QFMT) { 2327 if (sb_any_quota_loaded(sb) && 2328 sbi->s_jquota_fmt != m->mount_opt) { 2329 ext4_msg(sb, KERN_ERR, "Cannot change journaled " 2330 "quota options when quota turned on"); 2331 return -1; 2332 } 2333 if (ext4_has_feature_quota(sb)) { 2334 ext4_msg(sb, KERN_INFO, 2335 "Quota format mount options ignored " 2336 "when QUOTA feature is enabled"); 2337 return 1; 2338 } 2339 sbi->s_jquota_fmt = m->mount_opt; 2340 #endif 2341 } else if (token == Opt_dax || token == Opt_dax_always || 2342 token == Opt_dax_inode || token == Opt_dax_never) { 2343 #ifdef CONFIG_FS_DAX 2344 switch (token) { 2345 case Opt_dax: 2346 case Opt_dax_always: 2347 if (is_remount && 2348 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2349 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) { 2350 fail_dax_change_remount: 2351 ext4_msg(sb, KERN_ERR, "can't change " 2352 "dax mount option while remounting"); 2353 return -1; 2354 } 2355 if (is_remount && 2356 (test_opt(sb, DATA_FLAGS) == 2357 EXT4_MOUNT_JOURNAL_DATA)) { 2358 ext4_msg(sb, KERN_ERR, "can't mount with " 2359 "both data=journal and dax"); 2360 return -1; 2361 } 2362 ext4_msg(sb, KERN_WARNING, 2363 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); 2364 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS; 2365 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER; 2366 break; 2367 case Opt_dax_never: 2368 if (is_remount && 2369 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2370 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) 2371 goto fail_dax_change_remount; 2372 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER; 2373 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS; 2374 break; 2375 case Opt_dax_inode: 2376 if (is_remount && 2377 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2378 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2379 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) 2380 goto fail_dax_change_remount; 2381 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS; 2382 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER; 2383 /* Strictly for printing options */ 2384 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE; 2385 break; 2386 } 2387 #else 2388 ext4_msg(sb, KERN_INFO, "dax option not supported"); 2389 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER; 2390 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS; 2391 return -1; 2392 #endif 2393 } else if (token == Opt_data_err_abort) { 2394 sbi->s_mount_opt |= m->mount_opt; 2395 } else if (token == Opt_data_err_ignore) { 2396 sbi->s_mount_opt &= ~m->mount_opt; 2397 } else if (token == Opt_mb_optimize_scan) { 2398 if (arg != 0 && arg != 1) { 2399 ext4_msg(sb, KERN_WARNING, 2400 "mb_optimize_scan should be set to 0 or 1."); 2401 return -1; 2402 } 2403 parsed_opts->mb_optimize_scan = arg; 2404 } else { 2405 if (!args->from) 2406 arg = 1; 2407 if (m->flags & MOPT_CLEAR) 2408 arg = !arg; 2409 else if (unlikely(!(m->flags & MOPT_SET))) { 2410 ext4_msg(sb, KERN_WARNING, 2411 "buggy handling of option %s", opt); 2412 WARN_ON(1); 2413 return -1; 2414 } 2415 if (m->flags & MOPT_2) { 2416 if (arg != 0) 2417 sbi->s_mount_opt2 |= m->mount_opt; 2418 else 2419 sbi->s_mount_opt2 &= ~m->mount_opt; 2420 } else { 2421 if (arg != 0) 2422 sbi->s_mount_opt |= m->mount_opt; 2423 else 2424 sbi->s_mount_opt &= ~m->mount_opt; 2425 } 2426 } 2427 return 1; 2428 } 2429 2430 static int parse_options(char *options, struct super_block *sb, 2431 struct ext4_parsed_options *ret_opts, 2432 int is_remount) 2433 { 2434 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb); 2435 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name; 2436 substring_t args[MAX_OPT_ARGS]; 2437 int token; 2438 2439 if (!options) 2440 return 1; 2441 2442 while ((p = strsep(&options, ",")) != NULL) { 2443 if (!*p) 2444 continue; 2445 /* 2446 * Initialize args struct so we know whether arg was 2447 * found; some options take optional arguments. 2448 */ 2449 args[0].to = args[0].from = NULL; 2450 token = match_token(p, tokens, args); 2451 if (handle_mount_opt(sb, p, token, args, ret_opts, 2452 is_remount) < 0) 2453 return 0; 2454 } 2455 #ifdef CONFIG_QUOTA 2456 /* 2457 * We do the test below only for project quotas. 'usrquota' and 2458 * 'grpquota' mount options are allowed even without quota feature 2459 * to support legacy quotas in quota files. 2460 */ 2461 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) { 2462 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. " 2463 "Cannot enable project quota enforcement."); 2464 return 0; 2465 } 2466 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA); 2467 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA); 2468 if (usr_qf_name || grp_qf_name) { 2469 if (test_opt(sb, USRQUOTA) && usr_qf_name) 2470 clear_opt(sb, USRQUOTA); 2471 2472 if (test_opt(sb, GRPQUOTA) && grp_qf_name) 2473 clear_opt(sb, GRPQUOTA); 2474 2475 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 2476 ext4_msg(sb, KERN_ERR, "old and new quota " 2477 "format mixing"); 2478 return 0; 2479 } 2480 2481 if (!sbi->s_jquota_fmt) { 2482 ext4_msg(sb, KERN_ERR, "journaled quota format " 2483 "not specified"); 2484 return 0; 2485 } 2486 } 2487 #endif 2488 if (test_opt(sb, DIOREAD_NOLOCK)) { 2489 int blocksize = 2490 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 2491 if (blocksize < PAGE_SIZE) 2492 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an " 2493 "experimental mount option 'dioread_nolock' " 2494 "for blocksize < PAGE_SIZE"); 2495 } 2496 return 1; 2497 } 2498 2499 static inline void ext4_show_quota_options(struct seq_file *seq, 2500 struct super_block *sb) 2501 { 2502 #if defined(CONFIG_QUOTA) 2503 struct ext4_sb_info *sbi = EXT4_SB(sb); 2504 char *usr_qf_name, *grp_qf_name; 2505 2506 if (sbi->s_jquota_fmt) { 2507 char *fmtname = ""; 2508 2509 switch (sbi->s_jquota_fmt) { 2510 case QFMT_VFS_OLD: 2511 fmtname = "vfsold"; 2512 break; 2513 case QFMT_VFS_V0: 2514 fmtname = "vfsv0"; 2515 break; 2516 case QFMT_VFS_V1: 2517 fmtname = "vfsv1"; 2518 break; 2519 } 2520 seq_printf(seq, ",jqfmt=%s", fmtname); 2521 } 2522 2523 rcu_read_lock(); 2524 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); 2525 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); 2526 if (usr_qf_name) 2527 seq_show_option(seq, "usrjquota", usr_qf_name); 2528 if (grp_qf_name) 2529 seq_show_option(seq, "grpjquota", grp_qf_name); 2530 rcu_read_unlock(); 2531 #endif 2532 } 2533 2534 static const char *token2str(int token) 2535 { 2536 const struct match_token *t; 2537 2538 for (t = tokens; t->token != Opt_err; t++) 2539 if (t->token == token && !strchr(t->pattern, '=')) 2540 break; 2541 return t->pattern; 2542 } 2543 2544 /* 2545 * Show an option if 2546 * - it's set to a non-default value OR 2547 * - if the per-sb default is different from the global default 2548 */ 2549 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 2550 int nodefs) 2551 { 2552 struct ext4_sb_info *sbi = EXT4_SB(sb); 2553 struct ext4_super_block *es = sbi->s_es; 2554 int def_errors, def_mount_opt = sbi->s_def_mount_opt; 2555 const struct mount_opts *m; 2556 char sep = nodefs ? '\n' : ','; 2557 2558 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 2559 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 2560 2561 if (sbi->s_sb_block != 1) 2562 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 2563 2564 for (m = ext4_mount_opts; m->token != Opt_err; m++) { 2565 int want_set = m->flags & MOPT_SET; 2566 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 2567 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP) 2568 continue; 2569 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 2570 continue; /* skip if same as the default */ 2571 if ((want_set && 2572 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 2573 (!want_set && (sbi->s_mount_opt & m->mount_opt))) 2574 continue; /* select Opt_noFoo vs Opt_Foo */ 2575 SEQ_OPTS_PRINT("%s", token2str(m->token)); 2576 } 2577 2578 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 2579 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 2580 SEQ_OPTS_PRINT("resuid=%u", 2581 from_kuid_munged(&init_user_ns, sbi->s_resuid)); 2582 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 2583 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 2584 SEQ_OPTS_PRINT("resgid=%u", 2585 from_kgid_munged(&init_user_ns, sbi->s_resgid)); 2586 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 2587 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 2588 SEQ_OPTS_PUTS("errors=remount-ro"); 2589 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 2590 SEQ_OPTS_PUTS("errors=continue"); 2591 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 2592 SEQ_OPTS_PUTS("errors=panic"); 2593 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 2594 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 2595 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 2596 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 2597 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 2598 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 2599 if (sb->s_flags & SB_I_VERSION) 2600 SEQ_OPTS_PUTS("i_version"); 2601 if (nodefs || sbi->s_stripe) 2602 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 2603 if (nodefs || EXT4_MOUNT_DATA_FLAGS & 2604 (sbi->s_mount_opt ^ def_mount_opt)) { 2605 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 2606 SEQ_OPTS_PUTS("data=journal"); 2607 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 2608 SEQ_OPTS_PUTS("data=ordered"); 2609 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 2610 SEQ_OPTS_PUTS("data=writeback"); 2611 } 2612 if (nodefs || 2613 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 2614 SEQ_OPTS_PRINT("inode_readahead_blks=%u", 2615 sbi->s_inode_readahead_blks); 2616 2617 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs || 2618 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 2619 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 2620 if (nodefs || sbi->s_max_dir_size_kb) 2621 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 2622 if (test_opt(sb, DATA_ERR_ABORT)) 2623 SEQ_OPTS_PUTS("data_err=abort"); 2624 2625 fscrypt_show_test_dummy_encryption(seq, sep, sb); 2626 2627 if (sb->s_flags & SB_INLINECRYPT) 2628 SEQ_OPTS_PUTS("inlinecrypt"); 2629 2630 if (test_opt(sb, DAX_ALWAYS)) { 2631 if (IS_EXT2_SB(sb)) 2632 SEQ_OPTS_PUTS("dax"); 2633 else 2634 SEQ_OPTS_PUTS("dax=always"); 2635 } else if (test_opt2(sb, DAX_NEVER)) { 2636 SEQ_OPTS_PUTS("dax=never"); 2637 } else if (test_opt2(sb, DAX_INODE)) { 2638 SEQ_OPTS_PUTS("dax=inode"); 2639 } 2640 ext4_show_quota_options(seq, sb); 2641 return 0; 2642 } 2643 2644 static int ext4_show_options(struct seq_file *seq, struct dentry *root) 2645 { 2646 return _ext4_show_options(seq, root->d_sb, 0); 2647 } 2648 2649 int ext4_seq_options_show(struct seq_file *seq, void *offset) 2650 { 2651 struct super_block *sb = seq->private; 2652 int rc; 2653 2654 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); 2655 rc = _ext4_show_options(seq, sb, 1); 2656 seq_puts(seq, "\n"); 2657 return rc; 2658 } 2659 2660 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 2661 int read_only) 2662 { 2663 struct ext4_sb_info *sbi = EXT4_SB(sb); 2664 int err = 0; 2665 2666 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 2667 ext4_msg(sb, KERN_ERR, "revision level too high, " 2668 "forcing read-only mode"); 2669 err = -EROFS; 2670 goto done; 2671 } 2672 if (read_only) 2673 goto done; 2674 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 2675 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 2676 "running e2fsck is recommended"); 2677 else if (sbi->s_mount_state & EXT4_ERROR_FS) 2678 ext4_msg(sb, KERN_WARNING, 2679 "warning: mounting fs with errors, " 2680 "running e2fsck is recommended"); 2681 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 2682 le16_to_cpu(es->s_mnt_count) >= 2683 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 2684 ext4_msg(sb, KERN_WARNING, 2685 "warning: maximal mount count reached, " 2686 "running e2fsck is recommended"); 2687 else if (le32_to_cpu(es->s_checkinterval) && 2688 (ext4_get_tstamp(es, s_lastcheck) + 2689 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds())) 2690 ext4_msg(sb, KERN_WARNING, 2691 "warning: checktime reached, " 2692 "running e2fsck is recommended"); 2693 if (!sbi->s_journal) 2694 es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 2695 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 2696 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 2697 le16_add_cpu(&es->s_mnt_count, 1); 2698 ext4_update_tstamp(es, s_mtime); 2699 if (sbi->s_journal) { 2700 ext4_set_feature_journal_needs_recovery(sb); 2701 if (ext4_has_feature_orphan_file(sb)) 2702 ext4_set_feature_orphan_present(sb); 2703 } 2704 2705 err = ext4_commit_super(sb); 2706 done: 2707 if (test_opt(sb, DEBUG)) 2708 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 2709 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 2710 sb->s_blocksize, 2711 sbi->s_groups_count, 2712 EXT4_BLOCKS_PER_GROUP(sb), 2713 EXT4_INODES_PER_GROUP(sb), 2714 sbi->s_mount_opt, sbi->s_mount_opt2); 2715 2716 cleancache_init_fs(sb); 2717 return err; 2718 } 2719 2720 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 2721 { 2722 struct ext4_sb_info *sbi = EXT4_SB(sb); 2723 struct flex_groups **old_groups, **new_groups; 2724 int size, i, j; 2725 2726 if (!sbi->s_log_groups_per_flex) 2727 return 0; 2728 2729 size = ext4_flex_group(sbi, ngroup - 1) + 1; 2730 if (size <= sbi->s_flex_groups_allocated) 2731 return 0; 2732 2733 new_groups = kvzalloc(roundup_pow_of_two(size * 2734 sizeof(*sbi->s_flex_groups)), GFP_KERNEL); 2735 if (!new_groups) { 2736 ext4_msg(sb, KERN_ERR, 2737 "not enough memory for %d flex group pointers", size); 2738 return -ENOMEM; 2739 } 2740 for (i = sbi->s_flex_groups_allocated; i < size; i++) { 2741 new_groups[i] = kvzalloc(roundup_pow_of_two( 2742 sizeof(struct flex_groups)), 2743 GFP_KERNEL); 2744 if (!new_groups[i]) { 2745 for (j = sbi->s_flex_groups_allocated; j < i; j++) 2746 kvfree(new_groups[j]); 2747 kvfree(new_groups); 2748 ext4_msg(sb, KERN_ERR, 2749 "not enough memory for %d flex groups", size); 2750 return -ENOMEM; 2751 } 2752 } 2753 rcu_read_lock(); 2754 old_groups = rcu_dereference(sbi->s_flex_groups); 2755 if (old_groups) 2756 memcpy(new_groups, old_groups, 2757 (sbi->s_flex_groups_allocated * 2758 sizeof(struct flex_groups *))); 2759 rcu_read_unlock(); 2760 rcu_assign_pointer(sbi->s_flex_groups, new_groups); 2761 sbi->s_flex_groups_allocated = size; 2762 if (old_groups) 2763 ext4_kvfree_array_rcu(old_groups); 2764 return 0; 2765 } 2766 2767 static int ext4_fill_flex_info(struct super_block *sb) 2768 { 2769 struct ext4_sb_info *sbi = EXT4_SB(sb); 2770 struct ext4_group_desc *gdp = NULL; 2771 struct flex_groups *fg; 2772 ext4_group_t flex_group; 2773 int i, err; 2774 2775 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 2776 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 2777 sbi->s_log_groups_per_flex = 0; 2778 return 1; 2779 } 2780 2781 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 2782 if (err) 2783 goto failed; 2784 2785 for (i = 0; i < sbi->s_groups_count; i++) { 2786 gdp = ext4_get_group_desc(sb, i, NULL); 2787 2788 flex_group = ext4_flex_group(sbi, i); 2789 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group); 2790 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes); 2791 atomic64_add(ext4_free_group_clusters(sb, gdp), 2792 &fg->free_clusters); 2793 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs); 2794 } 2795 2796 return 1; 2797 failed: 2798 return 0; 2799 } 2800 2801 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 2802 struct ext4_group_desc *gdp) 2803 { 2804 int offset = offsetof(struct ext4_group_desc, bg_checksum); 2805 __u16 crc = 0; 2806 __le32 le_group = cpu_to_le32(block_group); 2807 struct ext4_sb_info *sbi = EXT4_SB(sb); 2808 2809 if (ext4_has_metadata_csum(sbi->s_sb)) { 2810 /* Use new metadata_csum algorithm */ 2811 __u32 csum32; 2812 __u16 dummy_csum = 0; 2813 2814 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 2815 sizeof(le_group)); 2816 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); 2817 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, 2818 sizeof(dummy_csum)); 2819 offset += sizeof(dummy_csum); 2820 if (offset < sbi->s_desc_size) 2821 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, 2822 sbi->s_desc_size - offset); 2823 2824 crc = csum32 & 0xFFFF; 2825 goto out; 2826 } 2827 2828 /* old crc16 code */ 2829 if (!ext4_has_feature_gdt_csum(sb)) 2830 return 0; 2831 2832 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 2833 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 2834 crc = crc16(crc, (__u8 *)gdp, offset); 2835 offset += sizeof(gdp->bg_checksum); /* skip checksum */ 2836 /* for checksum of struct ext4_group_desc do the rest...*/ 2837 if (ext4_has_feature_64bit(sb) && 2838 offset < le16_to_cpu(sbi->s_es->s_desc_size)) 2839 crc = crc16(crc, (__u8 *)gdp + offset, 2840 le16_to_cpu(sbi->s_es->s_desc_size) - 2841 offset); 2842 2843 out: 2844 return cpu_to_le16(crc); 2845 } 2846 2847 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 2848 struct ext4_group_desc *gdp) 2849 { 2850 if (ext4_has_group_desc_csum(sb) && 2851 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 2852 return 0; 2853 2854 return 1; 2855 } 2856 2857 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 2858 struct ext4_group_desc *gdp) 2859 { 2860 if (!ext4_has_group_desc_csum(sb)) 2861 return; 2862 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 2863 } 2864 2865 /* Called at mount-time, super-block is locked */ 2866 static int ext4_check_descriptors(struct super_block *sb, 2867 ext4_fsblk_t sb_block, 2868 ext4_group_t *first_not_zeroed) 2869 { 2870 struct ext4_sb_info *sbi = EXT4_SB(sb); 2871 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2872 ext4_fsblk_t last_block; 2873 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0); 2874 ext4_fsblk_t block_bitmap; 2875 ext4_fsblk_t inode_bitmap; 2876 ext4_fsblk_t inode_table; 2877 int flexbg_flag = 0; 2878 ext4_group_t i, grp = sbi->s_groups_count; 2879 2880 if (ext4_has_feature_flex_bg(sb)) 2881 flexbg_flag = 1; 2882 2883 ext4_debug("Checking group descriptors"); 2884 2885 for (i = 0; i < sbi->s_groups_count; i++) { 2886 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2887 2888 if (i == sbi->s_groups_count - 1 || flexbg_flag) 2889 last_block = ext4_blocks_count(sbi->s_es) - 1; 2890 else 2891 last_block = first_block + 2892 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2893 2894 if ((grp == sbi->s_groups_count) && 2895 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2896 grp = i; 2897 2898 block_bitmap = ext4_block_bitmap(sb, gdp); 2899 if (block_bitmap == sb_block) { 2900 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2901 "Block bitmap for group %u overlaps " 2902 "superblock", i); 2903 if (!sb_rdonly(sb)) 2904 return 0; 2905 } 2906 if (block_bitmap >= sb_block + 1 && 2907 block_bitmap <= last_bg_block) { 2908 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2909 "Block bitmap for group %u overlaps " 2910 "block group descriptors", i); 2911 if (!sb_rdonly(sb)) 2912 return 0; 2913 } 2914 if (block_bitmap < first_block || block_bitmap > last_block) { 2915 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2916 "Block bitmap for group %u not in group " 2917 "(block %llu)!", i, block_bitmap); 2918 return 0; 2919 } 2920 inode_bitmap = ext4_inode_bitmap(sb, gdp); 2921 if (inode_bitmap == sb_block) { 2922 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2923 "Inode bitmap for group %u overlaps " 2924 "superblock", i); 2925 if (!sb_rdonly(sb)) 2926 return 0; 2927 } 2928 if (inode_bitmap >= sb_block + 1 && 2929 inode_bitmap <= last_bg_block) { 2930 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2931 "Inode bitmap for group %u overlaps " 2932 "block group descriptors", i); 2933 if (!sb_rdonly(sb)) 2934 return 0; 2935 } 2936 if (inode_bitmap < first_block || inode_bitmap > last_block) { 2937 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2938 "Inode bitmap for group %u not in group " 2939 "(block %llu)!", i, inode_bitmap); 2940 return 0; 2941 } 2942 inode_table = ext4_inode_table(sb, gdp); 2943 if (inode_table == sb_block) { 2944 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2945 "Inode table for group %u overlaps " 2946 "superblock", i); 2947 if (!sb_rdonly(sb)) 2948 return 0; 2949 } 2950 if (inode_table >= sb_block + 1 && 2951 inode_table <= last_bg_block) { 2952 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2953 "Inode table for group %u overlaps " 2954 "block group descriptors", i); 2955 if (!sb_rdonly(sb)) 2956 return 0; 2957 } 2958 if (inode_table < first_block || 2959 inode_table + sbi->s_itb_per_group - 1 > last_block) { 2960 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2961 "Inode table for group %u not in group " 2962 "(block %llu)!", i, inode_table); 2963 return 0; 2964 } 2965 ext4_lock_group(sb, i); 2966 if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 2967 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2968 "Checksum for group %u failed (%u!=%u)", 2969 i, le16_to_cpu(ext4_group_desc_csum(sb, i, 2970 gdp)), le16_to_cpu(gdp->bg_checksum)); 2971 if (!sb_rdonly(sb)) { 2972 ext4_unlock_group(sb, i); 2973 return 0; 2974 } 2975 } 2976 ext4_unlock_group(sb, i); 2977 if (!flexbg_flag) 2978 first_block += EXT4_BLOCKS_PER_GROUP(sb); 2979 } 2980 if (NULL != first_not_zeroed) 2981 *first_not_zeroed = grp; 2982 return 1; 2983 } 2984 2985 /* 2986 * Maximal extent format file size. 2987 * Resulting logical blkno at s_maxbytes must fit in our on-disk 2988 * extent format containers, within a sector_t, and within i_blocks 2989 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 2990 * so that won't be a limiting factor. 2991 * 2992 * However there is other limiting factor. We do store extents in the form 2993 * of starting block and length, hence the resulting length of the extent 2994 * covering maximum file size must fit into on-disk format containers as 2995 * well. Given that length is always by 1 unit bigger than max unit (because 2996 * we count 0 as well) we have to lower the s_maxbytes by one fs block. 2997 * 2998 * Note, this does *not* consider any metadata overhead for vfs i_blocks. 2999 */ 3000 static loff_t ext4_max_size(int blkbits, int has_huge_files) 3001 { 3002 loff_t res; 3003 loff_t upper_limit = MAX_LFS_FILESIZE; 3004 3005 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64)); 3006 3007 if (!has_huge_files) { 3008 upper_limit = (1LL << 32) - 1; 3009 3010 /* total blocks in file system block size */ 3011 upper_limit >>= (blkbits - 9); 3012 upper_limit <<= blkbits; 3013 } 3014 3015 /* 3016 * 32-bit extent-start container, ee_block. We lower the maxbytes 3017 * by one fs block, so ee_len can cover the extent of maximum file 3018 * size 3019 */ 3020 res = (1LL << 32) - 1; 3021 res <<= blkbits; 3022 3023 /* Sanity check against vm- & vfs- imposed limits */ 3024 if (res > upper_limit) 3025 res = upper_limit; 3026 3027 return res; 3028 } 3029 3030 /* 3031 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 3032 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 3033 * We need to be 1 filesystem block less than the 2^48 sector limit. 3034 */ 3035 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 3036 { 3037 unsigned long long upper_limit, res = EXT4_NDIR_BLOCKS; 3038 int meta_blocks; 3039 3040 /* 3041 * This is calculated to be the largest file size for a dense, block 3042 * mapped file such that the file's total number of 512-byte sectors, 3043 * including data and all indirect blocks, does not exceed (2^48 - 1). 3044 * 3045 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 3046 * number of 512-byte sectors of the file. 3047 */ 3048 if (!has_huge_files) { 3049 /* 3050 * !has_huge_files or implies that the inode i_block field 3051 * represents total file blocks in 2^32 512-byte sectors == 3052 * size of vfs inode i_blocks * 8 3053 */ 3054 upper_limit = (1LL << 32) - 1; 3055 3056 /* total blocks in file system block size */ 3057 upper_limit >>= (bits - 9); 3058 3059 } else { 3060 /* 3061 * We use 48 bit ext4_inode i_blocks 3062 * With EXT4_HUGE_FILE_FL set the i_blocks 3063 * represent total number of blocks in 3064 * file system block size 3065 */ 3066 upper_limit = (1LL << 48) - 1; 3067 3068 } 3069 3070 /* indirect blocks */ 3071 meta_blocks = 1; 3072 /* double indirect blocks */ 3073 meta_blocks += 1 + (1LL << (bits-2)); 3074 /* tripple indirect blocks */ 3075 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 3076 3077 upper_limit -= meta_blocks; 3078 upper_limit <<= bits; 3079 3080 res += 1LL << (bits-2); 3081 res += 1LL << (2*(bits-2)); 3082 res += 1LL << (3*(bits-2)); 3083 res <<= bits; 3084 if (res > upper_limit) 3085 res = upper_limit; 3086 3087 if (res > MAX_LFS_FILESIZE) 3088 res = MAX_LFS_FILESIZE; 3089 3090 return (loff_t)res; 3091 } 3092 3093 static ext4_fsblk_t descriptor_loc(struct super_block *sb, 3094 ext4_fsblk_t logical_sb_block, int nr) 3095 { 3096 struct ext4_sb_info *sbi = EXT4_SB(sb); 3097 ext4_group_t bg, first_meta_bg; 3098 int has_super = 0; 3099 3100 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 3101 3102 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 3103 return logical_sb_block + nr + 1; 3104 bg = sbi->s_desc_per_block * nr; 3105 if (ext4_bg_has_super(sb, bg)) 3106 has_super = 1; 3107 3108 /* 3109 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 3110 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 3111 * on modern mke2fs or blksize > 1k on older mke2fs) then we must 3112 * compensate. 3113 */ 3114 if (sb->s_blocksize == 1024 && nr == 0 && 3115 le32_to_cpu(sbi->s_es->s_first_data_block) == 0) 3116 has_super++; 3117 3118 return (has_super + ext4_group_first_block_no(sb, bg)); 3119 } 3120 3121 /** 3122 * ext4_get_stripe_size: Get the stripe size. 3123 * @sbi: In memory super block info 3124 * 3125 * If we have specified it via mount option, then 3126 * use the mount option value. If the value specified at mount time is 3127 * greater than the blocks per group use the super block value. 3128 * If the super block value is greater than blocks per group return 0. 3129 * Allocator needs it be less than blocks per group. 3130 * 3131 */ 3132 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 3133 { 3134 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 3135 unsigned long stripe_width = 3136 le32_to_cpu(sbi->s_es->s_raid_stripe_width); 3137 int ret; 3138 3139 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 3140 ret = sbi->s_stripe; 3141 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) 3142 ret = stripe_width; 3143 else if (stride && stride <= sbi->s_blocks_per_group) 3144 ret = stride; 3145 else 3146 ret = 0; 3147 3148 /* 3149 * If the stripe width is 1, this makes no sense and 3150 * we set it to 0 to turn off stripe handling code. 3151 */ 3152 if (ret <= 1) 3153 ret = 0; 3154 3155 return ret; 3156 } 3157 3158 /* 3159 * Check whether this filesystem can be mounted based on 3160 * the features present and the RDONLY/RDWR mount requested. 3161 * Returns 1 if this filesystem can be mounted as requested, 3162 * 0 if it cannot be. 3163 */ 3164 int ext4_feature_set_ok(struct super_block *sb, int readonly) 3165 { 3166 if (ext4_has_unknown_ext4_incompat_features(sb)) { 3167 ext4_msg(sb, KERN_ERR, 3168 "Couldn't mount because of " 3169 "unsupported optional features (%x)", 3170 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 3171 ~EXT4_FEATURE_INCOMPAT_SUPP)); 3172 return 0; 3173 } 3174 3175 #ifndef CONFIG_UNICODE 3176 if (ext4_has_feature_casefold(sb)) { 3177 ext4_msg(sb, KERN_ERR, 3178 "Filesystem with casefold feature cannot be " 3179 "mounted without CONFIG_UNICODE"); 3180 return 0; 3181 } 3182 #endif 3183 3184 if (readonly) 3185 return 1; 3186 3187 if (ext4_has_feature_readonly(sb)) { 3188 ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 3189 sb->s_flags |= SB_RDONLY; 3190 return 1; 3191 } 3192 3193 /* Check that feature set is OK for a read-write mount */ 3194 if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 3195 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 3196 "unsupported optional features (%x)", 3197 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 3198 ~EXT4_FEATURE_RO_COMPAT_SUPP)); 3199 return 0; 3200 } 3201 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 3202 ext4_msg(sb, KERN_ERR, 3203 "Can't support bigalloc feature without " 3204 "extents feature\n"); 3205 return 0; 3206 } 3207 3208 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2) 3209 if (!readonly && (ext4_has_feature_quota(sb) || 3210 ext4_has_feature_project(sb))) { 3211 ext4_msg(sb, KERN_ERR, 3212 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2"); 3213 return 0; 3214 } 3215 #endif /* CONFIG_QUOTA */ 3216 return 1; 3217 } 3218 3219 /* 3220 * This function is called once a day if we have errors logged 3221 * on the file system 3222 */ 3223 static void print_daily_error_info(struct timer_list *t) 3224 { 3225 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report); 3226 struct super_block *sb = sbi->s_sb; 3227 struct ext4_super_block *es = sbi->s_es; 3228 3229 if (es->s_error_count) 3230 /* fsck newer than v1.41.13 is needed to clean this condition. */ 3231 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 3232 le32_to_cpu(es->s_error_count)); 3233 if (es->s_first_error_time) { 3234 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d", 3235 sb->s_id, 3236 ext4_get_tstamp(es, s_first_error_time), 3237 (int) sizeof(es->s_first_error_func), 3238 es->s_first_error_func, 3239 le32_to_cpu(es->s_first_error_line)); 3240 if (es->s_first_error_ino) 3241 printk(KERN_CONT ": inode %u", 3242 le32_to_cpu(es->s_first_error_ino)); 3243 if (es->s_first_error_block) 3244 printk(KERN_CONT ": block %llu", (unsigned long long) 3245 le64_to_cpu(es->s_first_error_block)); 3246 printk(KERN_CONT "\n"); 3247 } 3248 if (es->s_last_error_time) { 3249 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d", 3250 sb->s_id, 3251 ext4_get_tstamp(es, s_last_error_time), 3252 (int) sizeof(es->s_last_error_func), 3253 es->s_last_error_func, 3254 le32_to_cpu(es->s_last_error_line)); 3255 if (es->s_last_error_ino) 3256 printk(KERN_CONT ": inode %u", 3257 le32_to_cpu(es->s_last_error_ino)); 3258 if (es->s_last_error_block) 3259 printk(KERN_CONT ": block %llu", (unsigned long long) 3260 le64_to_cpu(es->s_last_error_block)); 3261 printk(KERN_CONT "\n"); 3262 } 3263 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 3264 } 3265 3266 /* Find next suitable group and run ext4_init_inode_table */ 3267 static int ext4_run_li_request(struct ext4_li_request *elr) 3268 { 3269 struct ext4_group_desc *gdp = NULL; 3270 struct super_block *sb = elr->lr_super; 3271 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 3272 ext4_group_t group = elr->lr_next_group; 3273 unsigned int prefetch_ios = 0; 3274 int ret = 0; 3275 u64 start_time; 3276 3277 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) { 3278 elr->lr_next_group = ext4_mb_prefetch(sb, group, 3279 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios); 3280 if (prefetch_ios) 3281 ext4_mb_prefetch_fini(sb, elr->lr_next_group, 3282 prefetch_ios); 3283 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, 3284 prefetch_ios); 3285 if (group >= elr->lr_next_group) { 3286 ret = 1; 3287 if (elr->lr_first_not_zeroed != ngroups && 3288 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) { 3289 elr->lr_next_group = elr->lr_first_not_zeroed; 3290 elr->lr_mode = EXT4_LI_MODE_ITABLE; 3291 ret = 0; 3292 } 3293 } 3294 return ret; 3295 } 3296 3297 for (; group < ngroups; group++) { 3298 gdp = ext4_get_group_desc(sb, group, NULL); 3299 if (!gdp) { 3300 ret = 1; 3301 break; 3302 } 3303 3304 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3305 break; 3306 } 3307 3308 if (group >= ngroups) 3309 ret = 1; 3310 3311 if (!ret) { 3312 start_time = ktime_get_real_ns(); 3313 ret = ext4_init_inode_table(sb, group, 3314 elr->lr_timeout ? 0 : 1); 3315 trace_ext4_lazy_itable_init(sb, group); 3316 if (elr->lr_timeout == 0) { 3317 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) * 3318 EXT4_SB(elr->lr_super)->s_li_wait_mult); 3319 } 3320 elr->lr_next_sched = jiffies + elr->lr_timeout; 3321 elr->lr_next_group = group + 1; 3322 } 3323 return ret; 3324 } 3325 3326 /* 3327 * Remove lr_request from the list_request and free the 3328 * request structure. Should be called with li_list_mtx held 3329 */ 3330 static void ext4_remove_li_request(struct ext4_li_request *elr) 3331 { 3332 if (!elr) 3333 return; 3334 3335 list_del(&elr->lr_request); 3336 EXT4_SB(elr->lr_super)->s_li_request = NULL; 3337 kfree(elr); 3338 } 3339 3340 static void ext4_unregister_li_request(struct super_block *sb) 3341 { 3342 mutex_lock(&ext4_li_mtx); 3343 if (!ext4_li_info) { 3344 mutex_unlock(&ext4_li_mtx); 3345 return; 3346 } 3347 3348 mutex_lock(&ext4_li_info->li_list_mtx); 3349 ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 3350 mutex_unlock(&ext4_li_info->li_list_mtx); 3351 mutex_unlock(&ext4_li_mtx); 3352 } 3353 3354 static struct task_struct *ext4_lazyinit_task; 3355 3356 /* 3357 * This is the function where ext4lazyinit thread lives. It walks 3358 * through the request list searching for next scheduled filesystem. 3359 * When such a fs is found, run the lazy initialization request 3360 * (ext4_rn_li_request) and keep track of the time spend in this 3361 * function. Based on that time we compute next schedule time of 3362 * the request. When walking through the list is complete, compute 3363 * next waking time and put itself into sleep. 3364 */ 3365 static int ext4_lazyinit_thread(void *arg) 3366 { 3367 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 3368 struct list_head *pos, *n; 3369 struct ext4_li_request *elr; 3370 unsigned long next_wakeup, cur; 3371 3372 BUG_ON(NULL == eli); 3373 3374 cont_thread: 3375 while (true) { 3376 next_wakeup = MAX_JIFFY_OFFSET; 3377 3378 mutex_lock(&eli->li_list_mtx); 3379 if (list_empty(&eli->li_request_list)) { 3380 mutex_unlock(&eli->li_list_mtx); 3381 goto exit_thread; 3382 } 3383 list_for_each_safe(pos, n, &eli->li_request_list) { 3384 int err = 0; 3385 int progress = 0; 3386 elr = list_entry(pos, struct ext4_li_request, 3387 lr_request); 3388 3389 if (time_before(jiffies, elr->lr_next_sched)) { 3390 if (time_before(elr->lr_next_sched, next_wakeup)) 3391 next_wakeup = elr->lr_next_sched; 3392 continue; 3393 } 3394 if (down_read_trylock(&elr->lr_super->s_umount)) { 3395 if (sb_start_write_trylock(elr->lr_super)) { 3396 progress = 1; 3397 /* 3398 * We hold sb->s_umount, sb can not 3399 * be removed from the list, it is 3400 * now safe to drop li_list_mtx 3401 */ 3402 mutex_unlock(&eli->li_list_mtx); 3403 err = ext4_run_li_request(elr); 3404 sb_end_write(elr->lr_super); 3405 mutex_lock(&eli->li_list_mtx); 3406 n = pos->next; 3407 } 3408 up_read((&elr->lr_super->s_umount)); 3409 } 3410 /* error, remove the lazy_init job */ 3411 if (err) { 3412 ext4_remove_li_request(elr); 3413 continue; 3414 } 3415 if (!progress) { 3416 elr->lr_next_sched = jiffies + 3417 (prandom_u32() 3418 % (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3419 } 3420 if (time_before(elr->lr_next_sched, next_wakeup)) 3421 next_wakeup = elr->lr_next_sched; 3422 } 3423 mutex_unlock(&eli->li_list_mtx); 3424 3425 try_to_freeze(); 3426 3427 cur = jiffies; 3428 if ((time_after_eq(cur, next_wakeup)) || 3429 (MAX_JIFFY_OFFSET == next_wakeup)) { 3430 cond_resched(); 3431 continue; 3432 } 3433 3434 schedule_timeout_interruptible(next_wakeup - cur); 3435 3436 if (kthread_should_stop()) { 3437 ext4_clear_request_list(); 3438 goto exit_thread; 3439 } 3440 } 3441 3442 exit_thread: 3443 /* 3444 * It looks like the request list is empty, but we need 3445 * to check it under the li_list_mtx lock, to prevent any 3446 * additions into it, and of course we should lock ext4_li_mtx 3447 * to atomically free the list and ext4_li_info, because at 3448 * this point another ext4 filesystem could be registering 3449 * new one. 3450 */ 3451 mutex_lock(&ext4_li_mtx); 3452 mutex_lock(&eli->li_list_mtx); 3453 if (!list_empty(&eli->li_request_list)) { 3454 mutex_unlock(&eli->li_list_mtx); 3455 mutex_unlock(&ext4_li_mtx); 3456 goto cont_thread; 3457 } 3458 mutex_unlock(&eli->li_list_mtx); 3459 kfree(ext4_li_info); 3460 ext4_li_info = NULL; 3461 mutex_unlock(&ext4_li_mtx); 3462 3463 return 0; 3464 } 3465 3466 static void ext4_clear_request_list(void) 3467 { 3468 struct list_head *pos, *n; 3469 struct ext4_li_request *elr; 3470 3471 mutex_lock(&ext4_li_info->li_list_mtx); 3472 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3473 elr = list_entry(pos, struct ext4_li_request, 3474 lr_request); 3475 ext4_remove_li_request(elr); 3476 } 3477 mutex_unlock(&ext4_li_info->li_list_mtx); 3478 } 3479 3480 static int ext4_run_lazyinit_thread(void) 3481 { 3482 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 3483 ext4_li_info, "ext4lazyinit"); 3484 if (IS_ERR(ext4_lazyinit_task)) { 3485 int err = PTR_ERR(ext4_lazyinit_task); 3486 ext4_clear_request_list(); 3487 kfree(ext4_li_info); 3488 ext4_li_info = NULL; 3489 printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3490 "initialization thread\n", 3491 err); 3492 return err; 3493 } 3494 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3495 return 0; 3496 } 3497 3498 /* 3499 * Check whether it make sense to run itable init. thread or not. 3500 * If there is at least one uninitialized inode table, return 3501 * corresponding group number, else the loop goes through all 3502 * groups and return total number of groups. 3503 */ 3504 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3505 { 3506 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3507 struct ext4_group_desc *gdp = NULL; 3508 3509 if (!ext4_has_group_desc_csum(sb)) 3510 return ngroups; 3511 3512 for (group = 0; group < ngroups; group++) { 3513 gdp = ext4_get_group_desc(sb, group, NULL); 3514 if (!gdp) 3515 continue; 3516 3517 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3518 break; 3519 } 3520 3521 return group; 3522 } 3523 3524 static int ext4_li_info_new(void) 3525 { 3526 struct ext4_lazy_init *eli = NULL; 3527 3528 eli = kzalloc(sizeof(*eli), GFP_KERNEL); 3529 if (!eli) 3530 return -ENOMEM; 3531 3532 INIT_LIST_HEAD(&eli->li_request_list); 3533 mutex_init(&eli->li_list_mtx); 3534 3535 eli->li_state |= EXT4_LAZYINIT_QUIT; 3536 3537 ext4_li_info = eli; 3538 3539 return 0; 3540 } 3541 3542 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3543 ext4_group_t start) 3544 { 3545 struct ext4_li_request *elr; 3546 3547 elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3548 if (!elr) 3549 return NULL; 3550 3551 elr->lr_super = sb; 3552 elr->lr_first_not_zeroed = start; 3553 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) { 3554 elr->lr_mode = EXT4_LI_MODE_ITABLE; 3555 elr->lr_next_group = start; 3556 } else { 3557 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP; 3558 } 3559 3560 /* 3561 * Randomize first schedule time of the request to 3562 * spread the inode table initialization requests 3563 * better. 3564 */ 3565 elr->lr_next_sched = jiffies + (prandom_u32() % 3566 (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3567 return elr; 3568 } 3569 3570 int ext4_register_li_request(struct super_block *sb, 3571 ext4_group_t first_not_zeroed) 3572 { 3573 struct ext4_sb_info *sbi = EXT4_SB(sb); 3574 struct ext4_li_request *elr = NULL; 3575 ext4_group_t ngroups = sbi->s_groups_count; 3576 int ret = 0; 3577 3578 mutex_lock(&ext4_li_mtx); 3579 if (sbi->s_li_request != NULL) { 3580 /* 3581 * Reset timeout so it can be computed again, because 3582 * s_li_wait_mult might have changed. 3583 */ 3584 sbi->s_li_request->lr_timeout = 0; 3585 goto out; 3586 } 3587 3588 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) && 3589 (first_not_zeroed == ngroups || sb_rdonly(sb) || 3590 !test_opt(sb, INIT_INODE_TABLE))) 3591 goto out; 3592 3593 elr = ext4_li_request_new(sb, first_not_zeroed); 3594 if (!elr) { 3595 ret = -ENOMEM; 3596 goto out; 3597 } 3598 3599 if (NULL == ext4_li_info) { 3600 ret = ext4_li_info_new(); 3601 if (ret) 3602 goto out; 3603 } 3604 3605 mutex_lock(&ext4_li_info->li_list_mtx); 3606 list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3607 mutex_unlock(&ext4_li_info->li_list_mtx); 3608 3609 sbi->s_li_request = elr; 3610 /* 3611 * set elr to NULL here since it has been inserted to 3612 * the request_list and the removal and free of it is 3613 * handled by ext4_clear_request_list from now on. 3614 */ 3615 elr = NULL; 3616 3617 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3618 ret = ext4_run_lazyinit_thread(); 3619 if (ret) 3620 goto out; 3621 } 3622 out: 3623 mutex_unlock(&ext4_li_mtx); 3624 if (ret) 3625 kfree(elr); 3626 return ret; 3627 } 3628 3629 /* 3630 * We do not need to lock anything since this is called on 3631 * module unload. 3632 */ 3633 static void ext4_destroy_lazyinit_thread(void) 3634 { 3635 /* 3636 * If thread exited earlier 3637 * there's nothing to be done. 3638 */ 3639 if (!ext4_li_info || !ext4_lazyinit_task) 3640 return; 3641 3642 kthread_stop(ext4_lazyinit_task); 3643 } 3644 3645 static int set_journal_csum_feature_set(struct super_block *sb) 3646 { 3647 int ret = 1; 3648 int compat, incompat; 3649 struct ext4_sb_info *sbi = EXT4_SB(sb); 3650 3651 if (ext4_has_metadata_csum(sb)) { 3652 /* journal checksum v3 */ 3653 compat = 0; 3654 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 3655 } else { 3656 /* journal checksum v1 */ 3657 compat = JBD2_FEATURE_COMPAT_CHECKSUM; 3658 incompat = 0; 3659 } 3660 3661 jbd2_journal_clear_features(sbi->s_journal, 3662 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 3663 JBD2_FEATURE_INCOMPAT_CSUM_V3 | 3664 JBD2_FEATURE_INCOMPAT_CSUM_V2); 3665 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 3666 ret = jbd2_journal_set_features(sbi->s_journal, 3667 compat, 0, 3668 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 3669 incompat); 3670 } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 3671 ret = jbd2_journal_set_features(sbi->s_journal, 3672 compat, 0, 3673 incompat); 3674 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3675 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3676 } else { 3677 jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3678 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 3679 } 3680 3681 return ret; 3682 } 3683 3684 /* 3685 * Note: calculating the overhead so we can be compatible with 3686 * historical BSD practice is quite difficult in the face of 3687 * clusters/bigalloc. This is because multiple metadata blocks from 3688 * different block group can end up in the same allocation cluster. 3689 * Calculating the exact overhead in the face of clustered allocation 3690 * requires either O(all block bitmaps) in memory or O(number of block 3691 * groups**2) in time. We will still calculate the superblock for 3692 * older file systems --- and if we come across with a bigalloc file 3693 * system with zero in s_overhead_clusters the estimate will be close to 3694 * correct especially for very large cluster sizes --- but for newer 3695 * file systems, it's better to calculate this figure once at mkfs 3696 * time, and store it in the superblock. If the superblock value is 3697 * present (even for non-bigalloc file systems), we will use it. 3698 */ 3699 static int count_overhead(struct super_block *sb, ext4_group_t grp, 3700 char *buf) 3701 { 3702 struct ext4_sb_info *sbi = EXT4_SB(sb); 3703 struct ext4_group_desc *gdp; 3704 ext4_fsblk_t first_block, last_block, b; 3705 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3706 int s, j, count = 0; 3707 3708 if (!ext4_has_feature_bigalloc(sb)) 3709 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + 3710 sbi->s_itb_per_group + 2); 3711 3712 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 3713 (grp * EXT4_BLOCKS_PER_GROUP(sb)); 3714 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 3715 for (i = 0; i < ngroups; i++) { 3716 gdp = ext4_get_group_desc(sb, i, NULL); 3717 b = ext4_block_bitmap(sb, gdp); 3718 if (b >= first_block && b <= last_block) { 3719 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3720 count++; 3721 } 3722 b = ext4_inode_bitmap(sb, gdp); 3723 if (b >= first_block && b <= last_block) { 3724 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3725 count++; 3726 } 3727 b = ext4_inode_table(sb, gdp); 3728 if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 3729 for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 3730 int c = EXT4_B2C(sbi, b - first_block); 3731 ext4_set_bit(c, buf); 3732 count++; 3733 } 3734 if (i != grp) 3735 continue; 3736 s = 0; 3737 if (ext4_bg_has_super(sb, grp)) { 3738 ext4_set_bit(s++, buf); 3739 count++; 3740 } 3741 j = ext4_bg_num_gdb(sb, grp); 3742 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { 3743 ext4_error(sb, "Invalid number of block group " 3744 "descriptor blocks: %d", j); 3745 j = EXT4_BLOCKS_PER_GROUP(sb) - s; 3746 } 3747 count += j; 3748 for (; j > 0; j--) 3749 ext4_set_bit(EXT4_B2C(sbi, s++), buf); 3750 } 3751 if (!count) 3752 return 0; 3753 return EXT4_CLUSTERS_PER_GROUP(sb) - 3754 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 3755 } 3756 3757 /* 3758 * Compute the overhead and stash it in sbi->s_overhead 3759 */ 3760 int ext4_calculate_overhead(struct super_block *sb) 3761 { 3762 struct ext4_sb_info *sbi = EXT4_SB(sb); 3763 struct ext4_super_block *es = sbi->s_es; 3764 struct inode *j_inode; 3765 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); 3766 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3767 ext4_fsblk_t overhead = 0; 3768 char *buf = (char *) get_zeroed_page(GFP_NOFS); 3769 3770 if (!buf) 3771 return -ENOMEM; 3772 3773 /* 3774 * Compute the overhead (FS structures). This is constant 3775 * for a given filesystem unless the number of block groups 3776 * changes so we cache the previous value until it does. 3777 */ 3778 3779 /* 3780 * All of the blocks before first_data_block are overhead 3781 */ 3782 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 3783 3784 /* 3785 * Add the overhead found in each block group 3786 */ 3787 for (i = 0; i < ngroups; i++) { 3788 int blks; 3789 3790 blks = count_overhead(sb, i, buf); 3791 overhead += blks; 3792 if (blks) 3793 memset(buf, 0, PAGE_SIZE); 3794 cond_resched(); 3795 } 3796 3797 /* 3798 * Add the internal journal blocks whether the journal has been 3799 * loaded or not 3800 */ 3801 if (sbi->s_journal && !sbi->s_journal_bdev) 3802 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len); 3803 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) { 3804 /* j_inum for internal journal is non-zero */ 3805 j_inode = ext4_get_journal_inode(sb, j_inum); 3806 if (j_inode) { 3807 j_blocks = j_inode->i_size >> sb->s_blocksize_bits; 3808 overhead += EXT4_NUM_B2C(sbi, j_blocks); 3809 iput(j_inode); 3810 } else { 3811 ext4_msg(sb, KERN_ERR, "can't get journal size"); 3812 } 3813 } 3814 sbi->s_overhead = overhead; 3815 smp_wmb(); 3816 free_page((unsigned long) buf); 3817 return 0; 3818 } 3819 3820 static void ext4_set_resv_clusters(struct super_block *sb) 3821 { 3822 ext4_fsblk_t resv_clusters; 3823 struct ext4_sb_info *sbi = EXT4_SB(sb); 3824 3825 /* 3826 * There's no need to reserve anything when we aren't using extents. 3827 * The space estimates are exact, there are no unwritten extents, 3828 * hole punching doesn't need new metadata... This is needed especially 3829 * to keep ext2/3 backward compatibility. 3830 */ 3831 if (!ext4_has_feature_extents(sb)) 3832 return; 3833 /* 3834 * By default we reserve 2% or 4096 clusters, whichever is smaller. 3835 * This should cover the situations where we can not afford to run 3836 * out of space like for example punch hole, or converting 3837 * unwritten extents in delalloc path. In most cases such 3838 * allocation would require 1, or 2 blocks, higher numbers are 3839 * very rare. 3840 */ 3841 resv_clusters = (ext4_blocks_count(sbi->s_es) >> 3842 sbi->s_cluster_bits); 3843 3844 do_div(resv_clusters, 50); 3845 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 3846 3847 atomic64_set(&sbi->s_resv_clusters, resv_clusters); 3848 } 3849 3850 static const char *ext4_quota_mode(struct super_block *sb) 3851 { 3852 #ifdef CONFIG_QUOTA 3853 if (!ext4_quota_capable(sb)) 3854 return "none"; 3855 3856 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb)) 3857 return "journalled"; 3858 else 3859 return "writeback"; 3860 #else 3861 return "disabled"; 3862 #endif 3863 } 3864 3865 static void ext4_setup_csum_trigger(struct super_block *sb, 3866 enum ext4_journal_trigger_type type, 3867 void (*trigger)( 3868 struct jbd2_buffer_trigger_type *type, 3869 struct buffer_head *bh, 3870 void *mapped_data, 3871 size_t size)) 3872 { 3873 struct ext4_sb_info *sbi = EXT4_SB(sb); 3874 3875 sbi->s_journal_triggers[type].sb = sb; 3876 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger; 3877 } 3878 3879 static int ext4_fill_super(struct super_block *sb, void *data, int silent) 3880 { 3881 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev); 3882 char *orig_data = kstrdup(data, GFP_KERNEL); 3883 struct buffer_head *bh, **group_desc; 3884 struct ext4_super_block *es = NULL; 3885 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 3886 struct flex_groups **flex_groups; 3887 ext4_fsblk_t block; 3888 ext4_fsblk_t sb_block = get_sb_block(&data); 3889 ext4_fsblk_t logical_sb_block; 3890 unsigned long offset = 0; 3891 unsigned long def_mount_opts; 3892 struct inode *root; 3893 const char *descr; 3894 int ret = -ENOMEM; 3895 int blocksize, clustersize; 3896 unsigned int db_count; 3897 unsigned int i; 3898 int needs_recovery, has_huge_files; 3899 __u64 blocks_count; 3900 int err = 0; 3901 ext4_group_t first_not_zeroed; 3902 struct ext4_parsed_options parsed_opts; 3903 3904 /* Set defaults for the variables that will be set during parsing */ 3905 parsed_opts.journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3906 parsed_opts.journal_devnum = 0; 3907 parsed_opts.mb_optimize_scan = DEFAULT_MB_OPTIMIZE_SCAN; 3908 3909 if ((data && !orig_data) || !sbi) 3910 goto out_free_base; 3911 3912 sbi->s_daxdev = dax_dev; 3913 sbi->s_blockgroup_lock = 3914 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 3915 if (!sbi->s_blockgroup_lock) 3916 goto out_free_base; 3917 3918 sb->s_fs_info = sbi; 3919 sbi->s_sb = sb; 3920 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3921 sbi->s_sb_block = sb_block; 3922 sbi->s_sectors_written_start = 3923 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]); 3924 3925 /* Cleanup superblock name */ 3926 strreplace(sb->s_id, '/', '!'); 3927 3928 /* -EINVAL is default */ 3929 ret = -EINVAL; 3930 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3931 if (!blocksize) { 3932 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3933 goto out_fail; 3934 } 3935 3936 /* 3937 * The ext4 superblock will not be buffer aligned for other than 1kB 3938 * block sizes. We need to calculate the offset from buffer start. 3939 */ 3940 if (blocksize != EXT4_MIN_BLOCK_SIZE) { 3941 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 3942 offset = do_div(logical_sb_block, blocksize); 3943 } else { 3944 logical_sb_block = sb_block; 3945 } 3946 3947 bh = ext4_sb_bread_unmovable(sb, logical_sb_block); 3948 if (IS_ERR(bh)) { 3949 ext4_msg(sb, KERN_ERR, "unable to read superblock"); 3950 ret = PTR_ERR(bh); 3951 goto out_fail; 3952 } 3953 /* 3954 * Note: s_es must be initialized as soon as possible because 3955 * some ext4 macro-instructions depend on its value 3956 */ 3957 es = (struct ext4_super_block *) (bh->b_data + offset); 3958 sbi->s_es = es; 3959 sb->s_magic = le16_to_cpu(es->s_magic); 3960 if (sb->s_magic != EXT4_SUPER_MAGIC) 3961 goto cantfind_ext4; 3962 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 3963 3964 /* Warn if metadata_csum and gdt_csum are both set. */ 3965 if (ext4_has_feature_metadata_csum(sb) && 3966 ext4_has_feature_gdt_csum(sb)) 3967 ext4_warning(sb, "metadata_csum and uninit_bg are " 3968 "redundant flags; please run fsck."); 3969 3970 /* Check for a known checksum algorithm */ 3971 if (!ext4_verify_csum_type(sb, es)) { 3972 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3973 "unknown checksum algorithm."); 3974 silent = 1; 3975 goto cantfind_ext4; 3976 } 3977 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE, 3978 ext4_orphan_file_block_trigger); 3979 3980 /* Load the checksum driver */ 3981 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 3982 if (IS_ERR(sbi->s_chksum_driver)) { 3983 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 3984 ret = PTR_ERR(sbi->s_chksum_driver); 3985 sbi->s_chksum_driver = NULL; 3986 goto failed_mount; 3987 } 3988 3989 /* Check superblock checksum */ 3990 if (!ext4_superblock_csum_verify(sb, es)) { 3991 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3992 "invalid superblock checksum. Run e2fsck?"); 3993 silent = 1; 3994 ret = -EFSBADCRC; 3995 goto cantfind_ext4; 3996 } 3997 3998 /* Precompute checksum seed for all metadata */ 3999 if (ext4_has_feature_csum_seed(sb)) 4000 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 4001 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb)) 4002 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 4003 sizeof(es->s_uuid)); 4004 4005 /* Set defaults before we parse the mount options */ 4006 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 4007 set_opt(sb, INIT_INODE_TABLE); 4008 if (def_mount_opts & EXT4_DEFM_DEBUG) 4009 set_opt(sb, DEBUG); 4010 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 4011 set_opt(sb, GRPID); 4012 if (def_mount_opts & EXT4_DEFM_UID16) 4013 set_opt(sb, NO_UID32); 4014 /* xattr user namespace & acls are now defaulted on */ 4015 set_opt(sb, XATTR_USER); 4016 #ifdef CONFIG_EXT4_FS_POSIX_ACL 4017 set_opt(sb, POSIX_ACL); 4018 #endif 4019 if (ext4_has_feature_fast_commit(sb)) 4020 set_opt2(sb, JOURNAL_FAST_COMMIT); 4021 /* don't forget to enable journal_csum when metadata_csum is enabled. */ 4022 if (ext4_has_metadata_csum(sb)) 4023 set_opt(sb, JOURNAL_CHECKSUM); 4024 4025 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 4026 set_opt(sb, JOURNAL_DATA); 4027 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 4028 set_opt(sb, ORDERED_DATA); 4029 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 4030 set_opt(sb, WRITEBACK_DATA); 4031 4032 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 4033 set_opt(sb, ERRORS_PANIC); 4034 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 4035 set_opt(sb, ERRORS_CONT); 4036 else 4037 set_opt(sb, ERRORS_RO); 4038 /* block_validity enabled by default; disable with noblock_validity */ 4039 set_opt(sb, BLOCK_VALIDITY); 4040 if (def_mount_opts & EXT4_DEFM_DISCARD) 4041 set_opt(sb, DISCARD); 4042 4043 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 4044 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 4045 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 4046 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 4047 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 4048 4049 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 4050 set_opt(sb, BARRIER); 4051 4052 /* 4053 * enable delayed allocation by default 4054 * Use -o nodelalloc to turn it off 4055 */ 4056 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 4057 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 4058 set_opt(sb, DELALLOC); 4059 4060 /* 4061 * set default s_li_wait_mult for lazyinit, for the case there is 4062 * no mount option specified. 4063 */ 4064 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 4065 4066 if (le32_to_cpu(es->s_log_block_size) > 4067 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 4068 ext4_msg(sb, KERN_ERR, 4069 "Invalid log block size: %u", 4070 le32_to_cpu(es->s_log_block_size)); 4071 goto failed_mount; 4072 } 4073 if (le32_to_cpu(es->s_log_cluster_size) > 4074 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 4075 ext4_msg(sb, KERN_ERR, 4076 "Invalid log cluster size: %u", 4077 le32_to_cpu(es->s_log_cluster_size)); 4078 goto failed_mount; 4079 } 4080 4081 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 4082 4083 if (blocksize == PAGE_SIZE) 4084 set_opt(sb, DIOREAD_NOLOCK); 4085 4086 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 4087 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 4088 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 4089 } else { 4090 sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 4091 sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 4092 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) { 4093 ext4_msg(sb, KERN_ERR, "invalid first ino: %u", 4094 sbi->s_first_ino); 4095 goto failed_mount; 4096 } 4097 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 4098 (!is_power_of_2(sbi->s_inode_size)) || 4099 (sbi->s_inode_size > blocksize)) { 4100 ext4_msg(sb, KERN_ERR, 4101 "unsupported inode size: %d", 4102 sbi->s_inode_size); 4103 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize); 4104 goto failed_mount; 4105 } 4106 /* 4107 * i_atime_extra is the last extra field available for 4108 * [acm]times in struct ext4_inode. Checking for that 4109 * field should suffice to ensure we have extra space 4110 * for all three. 4111 */ 4112 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) + 4113 sizeof(((struct ext4_inode *)0)->i_atime_extra)) { 4114 sb->s_time_gran = 1; 4115 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX; 4116 } else { 4117 sb->s_time_gran = NSEC_PER_SEC; 4118 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX; 4119 } 4120 sb->s_time_min = EXT4_TIMESTAMP_MIN; 4121 } 4122 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 4123 sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 4124 EXT4_GOOD_OLD_INODE_SIZE; 4125 if (ext4_has_feature_extra_isize(sb)) { 4126 unsigned v, max = (sbi->s_inode_size - 4127 EXT4_GOOD_OLD_INODE_SIZE); 4128 4129 v = le16_to_cpu(es->s_want_extra_isize); 4130 if (v > max) { 4131 ext4_msg(sb, KERN_ERR, 4132 "bad s_want_extra_isize: %d", v); 4133 goto failed_mount; 4134 } 4135 if (sbi->s_want_extra_isize < v) 4136 sbi->s_want_extra_isize = v; 4137 4138 v = le16_to_cpu(es->s_min_extra_isize); 4139 if (v > max) { 4140 ext4_msg(sb, KERN_ERR, 4141 "bad s_min_extra_isize: %d", v); 4142 goto failed_mount; 4143 } 4144 if (sbi->s_want_extra_isize < v) 4145 sbi->s_want_extra_isize = v; 4146 } 4147 } 4148 4149 if (sbi->s_es->s_mount_opts[0]) { 4150 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts, 4151 sizeof(sbi->s_es->s_mount_opts), 4152 GFP_KERNEL); 4153 if (!s_mount_opts) 4154 goto failed_mount; 4155 if (!parse_options(s_mount_opts, sb, &parsed_opts, 0)) { 4156 ext4_msg(sb, KERN_WARNING, 4157 "failed to parse options in superblock: %s", 4158 s_mount_opts); 4159 } 4160 kfree(s_mount_opts); 4161 } 4162 sbi->s_def_mount_opt = sbi->s_mount_opt; 4163 if (!parse_options((char *) data, sb, &parsed_opts, 0)) 4164 goto failed_mount; 4165 4166 #ifdef CONFIG_UNICODE 4167 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) { 4168 const struct ext4_sb_encodings *encoding_info; 4169 struct unicode_map *encoding; 4170 __u16 encoding_flags; 4171 4172 if (ext4_sb_read_encoding(es, &encoding_info, 4173 &encoding_flags)) { 4174 ext4_msg(sb, KERN_ERR, 4175 "Encoding requested by superblock is unknown"); 4176 goto failed_mount; 4177 } 4178 4179 encoding = utf8_load(encoding_info->version); 4180 if (IS_ERR(encoding)) { 4181 ext4_msg(sb, KERN_ERR, 4182 "can't mount with superblock charset: %s-%s " 4183 "not supported by the kernel. flags: 0x%x.", 4184 encoding_info->name, encoding_info->version, 4185 encoding_flags); 4186 goto failed_mount; 4187 } 4188 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: " 4189 "%s-%s with flags 0x%hx", encoding_info->name, 4190 encoding_info->version?:"\b", encoding_flags); 4191 4192 sb->s_encoding = encoding; 4193 sb->s_encoding_flags = encoding_flags; 4194 } 4195 #endif 4196 4197 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 4198 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n"); 4199 /* can't mount with both data=journal and dioread_nolock. */ 4200 clear_opt(sb, DIOREAD_NOLOCK); 4201 clear_opt2(sb, JOURNAL_FAST_COMMIT); 4202 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 4203 ext4_msg(sb, KERN_ERR, "can't mount with " 4204 "both data=journal and delalloc"); 4205 goto failed_mount; 4206 } 4207 if (test_opt(sb, DAX_ALWAYS)) { 4208 ext4_msg(sb, KERN_ERR, "can't mount with " 4209 "both data=journal and dax"); 4210 goto failed_mount; 4211 } 4212 if (ext4_has_feature_encrypt(sb)) { 4213 ext4_msg(sb, KERN_WARNING, 4214 "encrypted files will use data=ordered " 4215 "instead of data journaling mode"); 4216 } 4217 if (test_opt(sb, DELALLOC)) 4218 clear_opt(sb, DELALLOC); 4219 } else { 4220 sb->s_iflags |= SB_I_CGROUPWB; 4221 } 4222 4223 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 4224 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 4225 4226 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 4227 (ext4_has_compat_features(sb) || 4228 ext4_has_ro_compat_features(sb) || 4229 ext4_has_incompat_features(sb))) 4230 ext4_msg(sb, KERN_WARNING, 4231 "feature flags set on rev 0 fs, " 4232 "running e2fsck is recommended"); 4233 4234 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 4235 set_opt2(sb, HURD_COMPAT); 4236 if (ext4_has_feature_64bit(sb)) { 4237 ext4_msg(sb, KERN_ERR, 4238 "The Hurd can't support 64-bit file systems"); 4239 goto failed_mount; 4240 } 4241 4242 /* 4243 * ea_inode feature uses l_i_version field which is not 4244 * available in HURD_COMPAT mode. 4245 */ 4246 if (ext4_has_feature_ea_inode(sb)) { 4247 ext4_msg(sb, KERN_ERR, 4248 "ea_inode feature is not supported for Hurd"); 4249 goto failed_mount; 4250 } 4251 } 4252 4253 if (IS_EXT2_SB(sb)) { 4254 if (ext2_feature_set_ok(sb)) 4255 ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 4256 "using the ext4 subsystem"); 4257 else { 4258 /* 4259 * If we're probing be silent, if this looks like 4260 * it's actually an ext[34] filesystem. 4261 */ 4262 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 4263 goto failed_mount; 4264 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 4265 "to feature incompatibilities"); 4266 goto failed_mount; 4267 } 4268 } 4269 4270 if (IS_EXT3_SB(sb)) { 4271 if (ext3_feature_set_ok(sb)) 4272 ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 4273 "using the ext4 subsystem"); 4274 else { 4275 /* 4276 * If we're probing be silent, if this looks like 4277 * it's actually an ext4 filesystem. 4278 */ 4279 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 4280 goto failed_mount; 4281 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 4282 "to feature incompatibilities"); 4283 goto failed_mount; 4284 } 4285 } 4286 4287 /* 4288 * Check feature flags regardless of the revision level, since we 4289 * previously didn't change the revision level when setting the flags, 4290 * so there is a chance incompat flags are set on a rev 0 filesystem. 4291 */ 4292 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) 4293 goto failed_mount; 4294 4295 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) { 4296 ext4_msg(sb, KERN_ERR, 4297 "Number of reserved GDT blocks insanely large: %d", 4298 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); 4299 goto failed_mount; 4300 } 4301 4302 if (dax_supported(dax_dev, sb->s_bdev, blocksize, 0, 4303 bdev_nr_sectors(sb->s_bdev))) 4304 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags); 4305 4306 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) { 4307 if (ext4_has_feature_inline_data(sb)) { 4308 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem" 4309 " that may contain inline data"); 4310 goto failed_mount; 4311 } 4312 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) { 4313 ext4_msg(sb, KERN_ERR, 4314 "DAX unsupported by block device."); 4315 goto failed_mount; 4316 } 4317 } 4318 4319 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 4320 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 4321 es->s_encryption_level); 4322 goto failed_mount; 4323 } 4324 4325 if (sb->s_blocksize != blocksize) { 4326 /* 4327 * bh must be released before kill_bdev(), otherwise 4328 * it won't be freed and its page also. kill_bdev() 4329 * is called by sb_set_blocksize(). 4330 */ 4331 brelse(bh); 4332 /* Validate the filesystem blocksize */ 4333 if (!sb_set_blocksize(sb, blocksize)) { 4334 ext4_msg(sb, KERN_ERR, "bad block size %d", 4335 blocksize); 4336 bh = NULL; 4337 goto failed_mount; 4338 } 4339 4340 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 4341 offset = do_div(logical_sb_block, blocksize); 4342 bh = ext4_sb_bread_unmovable(sb, logical_sb_block); 4343 if (IS_ERR(bh)) { 4344 ext4_msg(sb, KERN_ERR, 4345 "Can't read superblock on 2nd try"); 4346 ret = PTR_ERR(bh); 4347 bh = NULL; 4348 goto failed_mount; 4349 } 4350 es = (struct ext4_super_block *)(bh->b_data + offset); 4351 sbi->s_es = es; 4352 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 4353 ext4_msg(sb, KERN_ERR, 4354 "Magic mismatch, very weird!"); 4355 goto failed_mount; 4356 } 4357 } 4358 4359 has_huge_files = ext4_has_feature_huge_file(sb); 4360 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 4361 has_huge_files); 4362 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 4363 4364 sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 4365 if (ext4_has_feature_64bit(sb)) { 4366 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 4367 sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 4368 !is_power_of_2(sbi->s_desc_size)) { 4369 ext4_msg(sb, KERN_ERR, 4370 "unsupported descriptor size %lu", 4371 sbi->s_desc_size); 4372 goto failed_mount; 4373 } 4374 } else 4375 sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 4376 4377 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 4378 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 4379 4380 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 4381 if (sbi->s_inodes_per_block == 0) 4382 goto cantfind_ext4; 4383 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || 4384 sbi->s_inodes_per_group > blocksize * 8) { 4385 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", 4386 sbi->s_inodes_per_group); 4387 goto failed_mount; 4388 } 4389 sbi->s_itb_per_group = sbi->s_inodes_per_group / 4390 sbi->s_inodes_per_block; 4391 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 4392 sbi->s_sbh = bh; 4393 sbi->s_mount_state = le16_to_cpu(es->s_state); 4394 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 4395 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 4396 4397 for (i = 0; i < 4; i++) 4398 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 4399 sbi->s_def_hash_version = es->s_def_hash_version; 4400 if (ext4_has_feature_dir_index(sb)) { 4401 i = le32_to_cpu(es->s_flags); 4402 if (i & EXT2_FLAGS_UNSIGNED_HASH) 4403 sbi->s_hash_unsigned = 3; 4404 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 4405 #ifdef __CHAR_UNSIGNED__ 4406 if (!sb_rdonly(sb)) 4407 es->s_flags |= 4408 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 4409 sbi->s_hash_unsigned = 3; 4410 #else 4411 if (!sb_rdonly(sb)) 4412 es->s_flags |= 4413 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 4414 #endif 4415 } 4416 } 4417 4418 /* Handle clustersize */ 4419 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 4420 if (ext4_has_feature_bigalloc(sb)) { 4421 if (clustersize < blocksize) { 4422 ext4_msg(sb, KERN_ERR, 4423 "cluster size (%d) smaller than " 4424 "block size (%d)", clustersize, blocksize); 4425 goto failed_mount; 4426 } 4427 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 4428 le32_to_cpu(es->s_log_block_size); 4429 sbi->s_clusters_per_group = 4430 le32_to_cpu(es->s_clusters_per_group); 4431 if (sbi->s_clusters_per_group > blocksize * 8) { 4432 ext4_msg(sb, KERN_ERR, 4433 "#clusters per group too big: %lu", 4434 sbi->s_clusters_per_group); 4435 goto failed_mount; 4436 } 4437 if (sbi->s_blocks_per_group != 4438 (sbi->s_clusters_per_group * (clustersize / blocksize))) { 4439 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 4440 "clusters per group (%lu) inconsistent", 4441 sbi->s_blocks_per_group, 4442 sbi->s_clusters_per_group); 4443 goto failed_mount; 4444 } 4445 } else { 4446 if (clustersize != blocksize) { 4447 ext4_msg(sb, KERN_ERR, 4448 "fragment/cluster size (%d) != " 4449 "block size (%d)", clustersize, blocksize); 4450 goto failed_mount; 4451 } 4452 if (sbi->s_blocks_per_group > blocksize * 8) { 4453 ext4_msg(sb, KERN_ERR, 4454 "#blocks per group too big: %lu", 4455 sbi->s_blocks_per_group); 4456 goto failed_mount; 4457 } 4458 sbi->s_clusters_per_group = sbi->s_blocks_per_group; 4459 sbi->s_cluster_bits = 0; 4460 } 4461 sbi->s_cluster_ratio = clustersize / blocksize; 4462 4463 /* Do we have standard group size of clustersize * 8 blocks ? */ 4464 if (sbi->s_blocks_per_group == clustersize << 3) 4465 set_opt2(sb, STD_GROUP_SIZE); 4466 4467 /* 4468 * Test whether we have more sectors than will fit in sector_t, 4469 * and whether the max offset is addressable by the page cache. 4470 */ 4471 err = generic_check_addressable(sb->s_blocksize_bits, 4472 ext4_blocks_count(es)); 4473 if (err) { 4474 ext4_msg(sb, KERN_ERR, "filesystem" 4475 " too large to mount safely on this system"); 4476 goto failed_mount; 4477 } 4478 4479 if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 4480 goto cantfind_ext4; 4481 4482 /* check blocks count against device size */ 4483 blocks_count = sb_bdev_nr_blocks(sb); 4484 if (blocks_count && ext4_blocks_count(es) > blocks_count) { 4485 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 4486 "exceeds size of device (%llu blocks)", 4487 ext4_blocks_count(es), blocks_count); 4488 goto failed_mount; 4489 } 4490 4491 /* 4492 * It makes no sense for the first data block to be beyond the end 4493 * of the filesystem. 4494 */ 4495 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 4496 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4497 "block %u is beyond end of filesystem (%llu)", 4498 le32_to_cpu(es->s_first_data_block), 4499 ext4_blocks_count(es)); 4500 goto failed_mount; 4501 } 4502 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) && 4503 (sbi->s_cluster_ratio == 1)) { 4504 ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4505 "block is 0 with a 1k block and cluster size"); 4506 goto failed_mount; 4507 } 4508 4509 blocks_count = (ext4_blocks_count(es) - 4510 le32_to_cpu(es->s_first_data_block) + 4511 EXT4_BLOCKS_PER_GROUP(sb) - 1); 4512 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 4513 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 4514 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu " 4515 "(block count %llu, first data block %u, " 4516 "blocks per group %lu)", blocks_count, 4517 ext4_blocks_count(es), 4518 le32_to_cpu(es->s_first_data_block), 4519 EXT4_BLOCKS_PER_GROUP(sb)); 4520 goto failed_mount; 4521 } 4522 sbi->s_groups_count = blocks_count; 4523 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 4524 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 4525 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) != 4526 le32_to_cpu(es->s_inodes_count)) { 4527 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu", 4528 le32_to_cpu(es->s_inodes_count), 4529 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group)); 4530 ret = -EINVAL; 4531 goto failed_mount; 4532 } 4533 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 4534 EXT4_DESC_PER_BLOCK(sb); 4535 if (ext4_has_feature_meta_bg(sb)) { 4536 if (le32_to_cpu(es->s_first_meta_bg) > db_count) { 4537 ext4_msg(sb, KERN_WARNING, 4538 "first meta block group too large: %u " 4539 "(group descriptor block count %u)", 4540 le32_to_cpu(es->s_first_meta_bg), db_count); 4541 goto failed_mount; 4542 } 4543 } 4544 rcu_assign_pointer(sbi->s_group_desc, 4545 kvmalloc_array(db_count, 4546 sizeof(struct buffer_head *), 4547 GFP_KERNEL)); 4548 if (sbi->s_group_desc == NULL) { 4549 ext4_msg(sb, KERN_ERR, "not enough memory"); 4550 ret = -ENOMEM; 4551 goto failed_mount; 4552 } 4553 4554 bgl_lock_init(sbi->s_blockgroup_lock); 4555 4556 /* Pre-read the descriptors into the buffer cache */ 4557 for (i = 0; i < db_count; i++) { 4558 block = descriptor_loc(sb, logical_sb_block, i); 4559 ext4_sb_breadahead_unmovable(sb, block); 4560 } 4561 4562 for (i = 0; i < db_count; i++) { 4563 struct buffer_head *bh; 4564 4565 block = descriptor_loc(sb, logical_sb_block, i); 4566 bh = ext4_sb_bread_unmovable(sb, block); 4567 if (IS_ERR(bh)) { 4568 ext4_msg(sb, KERN_ERR, 4569 "can't read group descriptor %d", i); 4570 db_count = i; 4571 ret = PTR_ERR(bh); 4572 goto failed_mount2; 4573 } 4574 rcu_read_lock(); 4575 rcu_dereference(sbi->s_group_desc)[i] = bh; 4576 rcu_read_unlock(); 4577 } 4578 sbi->s_gdb_count = db_count; 4579 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) { 4580 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 4581 ret = -EFSCORRUPTED; 4582 goto failed_mount2; 4583 } 4584 4585 timer_setup(&sbi->s_err_report, print_daily_error_info, 0); 4586 spin_lock_init(&sbi->s_error_lock); 4587 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work); 4588 4589 /* Register extent status tree shrinker */ 4590 if (ext4_es_register_shrinker(sbi)) 4591 goto failed_mount3; 4592 4593 sbi->s_stripe = ext4_get_stripe_size(sbi); 4594 sbi->s_extent_max_zeroout_kb = 32; 4595 4596 /* 4597 * set up enough so that it can read an inode 4598 */ 4599 sb->s_op = &ext4_sops; 4600 sb->s_export_op = &ext4_export_ops; 4601 sb->s_xattr = ext4_xattr_handlers; 4602 #ifdef CONFIG_FS_ENCRYPTION 4603 sb->s_cop = &ext4_cryptops; 4604 #endif 4605 #ifdef CONFIG_FS_VERITY 4606 sb->s_vop = &ext4_verityops; 4607 #endif 4608 #ifdef CONFIG_QUOTA 4609 sb->dq_op = &ext4_quota_operations; 4610 if (ext4_has_feature_quota(sb)) 4611 sb->s_qcop = &dquot_quotactl_sysfile_ops; 4612 else 4613 sb->s_qcop = &ext4_qctl_operations; 4614 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4615 #endif 4616 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 4617 4618 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 4619 mutex_init(&sbi->s_orphan_lock); 4620 4621 /* Initialize fast commit stuff */ 4622 atomic_set(&sbi->s_fc_subtid, 0); 4623 atomic_set(&sbi->s_fc_ineligible_updates, 0); 4624 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]); 4625 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]); 4626 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]); 4627 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]); 4628 sbi->s_fc_bytes = 0; 4629 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE); 4630 ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING); 4631 spin_lock_init(&sbi->s_fc_lock); 4632 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats)); 4633 sbi->s_fc_replay_state.fc_regions = NULL; 4634 sbi->s_fc_replay_state.fc_regions_size = 0; 4635 sbi->s_fc_replay_state.fc_regions_used = 0; 4636 sbi->s_fc_replay_state.fc_regions_valid = 0; 4637 sbi->s_fc_replay_state.fc_modified_inodes = NULL; 4638 sbi->s_fc_replay_state.fc_modified_inodes_size = 0; 4639 sbi->s_fc_replay_state.fc_modified_inodes_used = 0; 4640 4641 sb->s_root = NULL; 4642 4643 needs_recovery = (es->s_last_orphan != 0 || 4644 ext4_has_feature_orphan_present(sb) || 4645 ext4_has_feature_journal_needs_recovery(sb)); 4646 4647 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) 4648 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 4649 goto failed_mount3a; 4650 4651 /* 4652 * The first inode we look at is the journal inode. Don't try 4653 * root first: it may be modified in the journal! 4654 */ 4655 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 4656 err = ext4_load_journal(sb, es, parsed_opts.journal_devnum); 4657 if (err) 4658 goto failed_mount3a; 4659 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && 4660 ext4_has_feature_journal_needs_recovery(sb)) { 4661 ext4_msg(sb, KERN_ERR, "required journal recovery " 4662 "suppressed and not mounted read-only"); 4663 goto failed_mount_wq; 4664 } else { 4665 /* Nojournal mode, all journal mount options are illegal */ 4666 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { 4667 ext4_msg(sb, KERN_ERR, "can't mount with " 4668 "journal_checksum, fs mounted w/o journal"); 4669 goto failed_mount_wq; 4670 } 4671 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4672 ext4_msg(sb, KERN_ERR, "can't mount with " 4673 "journal_async_commit, fs mounted w/o journal"); 4674 goto failed_mount_wq; 4675 } 4676 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { 4677 ext4_msg(sb, KERN_ERR, "can't mount with " 4678 "commit=%lu, fs mounted w/o journal", 4679 sbi->s_commit_interval / HZ); 4680 goto failed_mount_wq; 4681 } 4682 if (EXT4_MOUNT_DATA_FLAGS & 4683 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 4684 ext4_msg(sb, KERN_ERR, "can't mount with " 4685 "data=, fs mounted w/o journal"); 4686 goto failed_mount_wq; 4687 } 4688 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM; 4689 clear_opt(sb, JOURNAL_CHECKSUM); 4690 clear_opt(sb, DATA_FLAGS); 4691 clear_opt2(sb, JOURNAL_FAST_COMMIT); 4692 sbi->s_journal = NULL; 4693 needs_recovery = 0; 4694 goto no_journal; 4695 } 4696 4697 if (ext4_has_feature_64bit(sb) && 4698 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4699 JBD2_FEATURE_INCOMPAT_64BIT)) { 4700 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 4701 goto failed_mount_wq; 4702 } 4703 4704 if (!set_journal_csum_feature_set(sb)) { 4705 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 4706 "feature set"); 4707 goto failed_mount_wq; 4708 } 4709 4710 if (test_opt2(sb, JOURNAL_FAST_COMMIT) && 4711 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4712 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) { 4713 ext4_msg(sb, KERN_ERR, 4714 "Failed to set fast commit journal feature"); 4715 goto failed_mount_wq; 4716 } 4717 4718 /* We have now updated the journal if required, so we can 4719 * validate the data journaling mode. */ 4720 switch (test_opt(sb, DATA_FLAGS)) { 4721 case 0: 4722 /* No mode set, assume a default based on the journal 4723 * capabilities: ORDERED_DATA if the journal can 4724 * cope, else JOURNAL_DATA 4725 */ 4726 if (jbd2_journal_check_available_features 4727 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4728 set_opt(sb, ORDERED_DATA); 4729 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA; 4730 } else { 4731 set_opt(sb, JOURNAL_DATA); 4732 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; 4733 } 4734 break; 4735 4736 case EXT4_MOUNT_ORDERED_DATA: 4737 case EXT4_MOUNT_WRITEBACK_DATA: 4738 if (!jbd2_journal_check_available_features 4739 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4740 ext4_msg(sb, KERN_ERR, "Journal does not support " 4741 "requested data journaling mode"); 4742 goto failed_mount_wq; 4743 } 4744 break; 4745 default: 4746 break; 4747 } 4748 4749 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 4750 test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4751 ext4_msg(sb, KERN_ERR, "can't mount with " 4752 "journal_async_commit in data=ordered mode"); 4753 goto failed_mount_wq; 4754 } 4755 4756 set_task_ioprio(sbi->s_journal->j_task, parsed_opts.journal_ioprio); 4757 4758 sbi->s_journal->j_submit_inode_data_buffers = 4759 ext4_journal_submit_inode_data_buffers; 4760 sbi->s_journal->j_finish_inode_data_buffers = 4761 ext4_journal_finish_inode_data_buffers; 4762 4763 no_journal: 4764 if (!test_opt(sb, NO_MBCACHE)) { 4765 sbi->s_ea_block_cache = ext4_xattr_create_cache(); 4766 if (!sbi->s_ea_block_cache) { 4767 ext4_msg(sb, KERN_ERR, 4768 "Failed to create ea_block_cache"); 4769 goto failed_mount_wq; 4770 } 4771 4772 if (ext4_has_feature_ea_inode(sb)) { 4773 sbi->s_ea_inode_cache = ext4_xattr_create_cache(); 4774 if (!sbi->s_ea_inode_cache) { 4775 ext4_msg(sb, KERN_ERR, 4776 "Failed to create ea_inode_cache"); 4777 goto failed_mount_wq; 4778 } 4779 } 4780 } 4781 4782 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) { 4783 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity"); 4784 goto failed_mount_wq; 4785 } 4786 4787 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) && 4788 !ext4_has_feature_encrypt(sb)) { 4789 ext4_set_feature_encrypt(sb); 4790 ext4_commit_super(sb); 4791 } 4792 4793 /* 4794 * Get the # of file system overhead blocks from the 4795 * superblock if present. 4796 */ 4797 if (es->s_overhead_clusters) 4798 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 4799 else { 4800 err = ext4_calculate_overhead(sb); 4801 if (err) 4802 goto failed_mount_wq; 4803 } 4804 4805 /* 4806 * The maximum number of concurrent works can be high and 4807 * concurrency isn't really necessary. Limit it to 1. 4808 */ 4809 EXT4_SB(sb)->rsv_conversion_wq = 4810 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 4811 if (!EXT4_SB(sb)->rsv_conversion_wq) { 4812 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 4813 ret = -ENOMEM; 4814 goto failed_mount4; 4815 } 4816 4817 /* 4818 * The jbd2_journal_load will have done any necessary log recovery, 4819 * so we can safely mount the rest of the filesystem now. 4820 */ 4821 4822 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL); 4823 if (IS_ERR(root)) { 4824 ext4_msg(sb, KERN_ERR, "get root inode failed"); 4825 ret = PTR_ERR(root); 4826 root = NULL; 4827 goto failed_mount4; 4828 } 4829 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 4830 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 4831 iput(root); 4832 goto failed_mount4; 4833 } 4834 4835 sb->s_root = d_make_root(root); 4836 if (!sb->s_root) { 4837 ext4_msg(sb, KERN_ERR, "get root dentry failed"); 4838 ret = -ENOMEM; 4839 goto failed_mount4; 4840 } 4841 4842 ret = ext4_setup_super(sb, es, sb_rdonly(sb)); 4843 if (ret == -EROFS) { 4844 sb->s_flags |= SB_RDONLY; 4845 ret = 0; 4846 } else if (ret) 4847 goto failed_mount4a; 4848 4849 ext4_set_resv_clusters(sb); 4850 4851 if (test_opt(sb, BLOCK_VALIDITY)) { 4852 err = ext4_setup_system_zone(sb); 4853 if (err) { 4854 ext4_msg(sb, KERN_ERR, "failed to initialize system " 4855 "zone (%d)", err); 4856 goto failed_mount4a; 4857 } 4858 } 4859 ext4_fc_replay_cleanup(sb); 4860 4861 ext4_ext_init(sb); 4862 4863 /* 4864 * Enable optimize_scan if number of groups is > threshold. This can be 4865 * turned off by passing "mb_optimize_scan=0". This can also be 4866 * turned on forcefully by passing "mb_optimize_scan=1". 4867 */ 4868 if (parsed_opts.mb_optimize_scan == 1) 4869 set_opt2(sb, MB_OPTIMIZE_SCAN); 4870 else if (parsed_opts.mb_optimize_scan == 0) 4871 clear_opt2(sb, MB_OPTIMIZE_SCAN); 4872 else if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD) 4873 set_opt2(sb, MB_OPTIMIZE_SCAN); 4874 4875 err = ext4_mb_init(sb); 4876 if (err) { 4877 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 4878 err); 4879 goto failed_mount5; 4880 } 4881 4882 /* 4883 * We can only set up the journal commit callback once 4884 * mballoc is initialized 4885 */ 4886 if (sbi->s_journal) 4887 sbi->s_journal->j_commit_callback = 4888 ext4_journal_commit_callback; 4889 4890 block = ext4_count_free_clusters(sb); 4891 ext4_free_blocks_count_set(sbi->s_es, 4892 EXT4_C2B(sbi, block)); 4893 err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 4894 GFP_KERNEL); 4895 if (!err) { 4896 unsigned long freei = ext4_count_free_inodes(sb); 4897 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 4898 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 4899 GFP_KERNEL); 4900 } 4901 /* 4902 * Update the checksum after updating free space/inode 4903 * counters. Otherwise the superblock can have an incorrect 4904 * checksum in the buffer cache until it is written out and 4905 * e2fsprogs programs trying to open a file system immediately 4906 * after it is mounted can fail. 4907 */ 4908 ext4_superblock_csum_set(sb); 4909 if (!err) 4910 err = percpu_counter_init(&sbi->s_dirs_counter, 4911 ext4_count_dirs(sb), GFP_KERNEL); 4912 if (!err) 4913 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 4914 GFP_KERNEL); 4915 if (!err) 4916 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0, 4917 GFP_KERNEL); 4918 if (!err) 4919 err = percpu_init_rwsem(&sbi->s_writepages_rwsem); 4920 4921 if (err) { 4922 ext4_msg(sb, KERN_ERR, "insufficient memory"); 4923 goto failed_mount6; 4924 } 4925 4926 if (ext4_has_feature_flex_bg(sb)) 4927 if (!ext4_fill_flex_info(sb)) { 4928 ext4_msg(sb, KERN_ERR, 4929 "unable to initialize " 4930 "flex_bg meta info!"); 4931 ret = -ENOMEM; 4932 goto failed_mount6; 4933 } 4934 4935 err = ext4_register_li_request(sb, first_not_zeroed); 4936 if (err) 4937 goto failed_mount6; 4938 4939 err = ext4_register_sysfs(sb); 4940 if (err) 4941 goto failed_mount7; 4942 4943 err = ext4_init_orphan_info(sb); 4944 if (err) 4945 goto failed_mount8; 4946 #ifdef CONFIG_QUOTA 4947 /* Enable quota usage during mount. */ 4948 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { 4949 err = ext4_enable_quotas(sb); 4950 if (err) 4951 goto failed_mount9; 4952 } 4953 #endif /* CONFIG_QUOTA */ 4954 4955 /* 4956 * Save the original bdev mapping's wb_err value which could be 4957 * used to detect the metadata async write error. 4958 */ 4959 spin_lock_init(&sbi->s_bdev_wb_lock); 4960 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err, 4961 &sbi->s_bdev_wb_err); 4962 sb->s_bdev->bd_super = sb; 4963 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 4964 ext4_orphan_cleanup(sb, es); 4965 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 4966 if (needs_recovery) { 4967 ext4_msg(sb, KERN_INFO, "recovery complete"); 4968 err = ext4_mark_recovery_complete(sb, es); 4969 if (err) 4970 goto failed_mount9; 4971 } 4972 if (EXT4_SB(sb)->s_journal) { 4973 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 4974 descr = " journalled data mode"; 4975 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 4976 descr = " ordered data mode"; 4977 else 4978 descr = " writeback data mode"; 4979 } else 4980 descr = "out journal"; 4981 4982 if (test_opt(sb, DISCARD)) { 4983 struct request_queue *q = bdev_get_queue(sb->s_bdev); 4984 if (!blk_queue_discard(q)) 4985 ext4_msg(sb, KERN_WARNING, 4986 "mounting with \"discard\" option, but " 4987 "the device does not support discard"); 4988 } 4989 4990 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 4991 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 4992 "Opts: %.*s%s%s. Quota mode: %s.", descr, 4993 (int) sizeof(sbi->s_es->s_mount_opts), 4994 sbi->s_es->s_mount_opts, 4995 *sbi->s_es->s_mount_opts ? "; " : "", orig_data, 4996 ext4_quota_mode(sb)); 4997 4998 if (es->s_error_count) 4999 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 5000 5001 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 5002 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 5003 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 5004 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 5005 atomic_set(&sbi->s_warning_count, 0); 5006 atomic_set(&sbi->s_msg_count, 0); 5007 5008 kfree(orig_data); 5009 return 0; 5010 5011 cantfind_ext4: 5012 if (!silent) 5013 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 5014 goto failed_mount; 5015 5016 failed_mount9: 5017 ext4_release_orphan_info(sb); 5018 failed_mount8: 5019 ext4_unregister_sysfs(sb); 5020 kobject_put(&sbi->s_kobj); 5021 failed_mount7: 5022 ext4_unregister_li_request(sb); 5023 failed_mount6: 5024 ext4_mb_release(sb); 5025 rcu_read_lock(); 5026 flex_groups = rcu_dereference(sbi->s_flex_groups); 5027 if (flex_groups) { 5028 for (i = 0; i < sbi->s_flex_groups_allocated; i++) 5029 kvfree(flex_groups[i]); 5030 kvfree(flex_groups); 5031 } 5032 rcu_read_unlock(); 5033 percpu_counter_destroy(&sbi->s_freeclusters_counter); 5034 percpu_counter_destroy(&sbi->s_freeinodes_counter); 5035 percpu_counter_destroy(&sbi->s_dirs_counter); 5036 percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 5037 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit); 5038 percpu_free_rwsem(&sbi->s_writepages_rwsem); 5039 failed_mount5: 5040 ext4_ext_release(sb); 5041 ext4_release_system_zone(sb); 5042 failed_mount4a: 5043 dput(sb->s_root); 5044 sb->s_root = NULL; 5045 failed_mount4: 5046 ext4_msg(sb, KERN_ERR, "mount failed"); 5047 if (EXT4_SB(sb)->rsv_conversion_wq) 5048 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 5049 failed_mount_wq: 5050 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 5051 sbi->s_ea_inode_cache = NULL; 5052 5053 ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 5054 sbi->s_ea_block_cache = NULL; 5055 5056 if (sbi->s_journal) { 5057 /* flush s_error_work before journal destroy. */ 5058 flush_work(&sbi->s_error_work); 5059 jbd2_journal_destroy(sbi->s_journal); 5060 sbi->s_journal = NULL; 5061 } 5062 failed_mount3a: 5063 ext4_es_unregister_shrinker(sbi); 5064 failed_mount3: 5065 /* flush s_error_work before sbi destroy */ 5066 flush_work(&sbi->s_error_work); 5067 del_timer_sync(&sbi->s_err_report); 5068 ext4_stop_mmpd(sbi); 5069 failed_mount2: 5070 rcu_read_lock(); 5071 group_desc = rcu_dereference(sbi->s_group_desc); 5072 for (i = 0; i < db_count; i++) 5073 brelse(group_desc[i]); 5074 kvfree(group_desc); 5075 rcu_read_unlock(); 5076 failed_mount: 5077 if (sbi->s_chksum_driver) 5078 crypto_free_shash(sbi->s_chksum_driver); 5079 5080 #ifdef CONFIG_UNICODE 5081 utf8_unload(sb->s_encoding); 5082 #endif 5083 5084 #ifdef CONFIG_QUOTA 5085 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5086 kfree(get_qf_name(sb, sbi, i)); 5087 #endif 5088 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy); 5089 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */ 5090 brelse(bh); 5091 ext4_blkdev_remove(sbi); 5092 out_fail: 5093 sb->s_fs_info = NULL; 5094 kfree(sbi->s_blockgroup_lock); 5095 out_free_base: 5096 kfree(sbi); 5097 kfree(orig_data); 5098 fs_put_dax(dax_dev); 5099 return err ? err : ret; 5100 } 5101 5102 /* 5103 * Setup any per-fs journal parameters now. We'll do this both on 5104 * initial mount, once the journal has been initialised but before we've 5105 * done any recovery; and again on any subsequent remount. 5106 */ 5107 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 5108 { 5109 struct ext4_sb_info *sbi = EXT4_SB(sb); 5110 5111 journal->j_commit_interval = sbi->s_commit_interval; 5112 journal->j_min_batch_time = sbi->s_min_batch_time; 5113 journal->j_max_batch_time = sbi->s_max_batch_time; 5114 ext4_fc_init(sb, journal); 5115 5116 write_lock(&journal->j_state_lock); 5117 if (test_opt(sb, BARRIER)) 5118 journal->j_flags |= JBD2_BARRIER; 5119 else 5120 journal->j_flags &= ~JBD2_BARRIER; 5121 if (test_opt(sb, DATA_ERR_ABORT)) 5122 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 5123 else 5124 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 5125 write_unlock(&journal->j_state_lock); 5126 } 5127 5128 static struct inode *ext4_get_journal_inode(struct super_block *sb, 5129 unsigned int journal_inum) 5130 { 5131 struct inode *journal_inode; 5132 5133 /* 5134 * Test for the existence of a valid inode on disk. Bad things 5135 * happen if we iget() an unused inode, as the subsequent iput() 5136 * will try to delete it. 5137 */ 5138 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL); 5139 if (IS_ERR(journal_inode)) { 5140 ext4_msg(sb, KERN_ERR, "no journal found"); 5141 return NULL; 5142 } 5143 if (!journal_inode->i_nlink) { 5144 make_bad_inode(journal_inode); 5145 iput(journal_inode); 5146 ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 5147 return NULL; 5148 } 5149 5150 jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 5151 journal_inode, journal_inode->i_size); 5152 if (!S_ISREG(journal_inode->i_mode)) { 5153 ext4_msg(sb, KERN_ERR, "invalid journal inode"); 5154 iput(journal_inode); 5155 return NULL; 5156 } 5157 return journal_inode; 5158 } 5159 5160 static journal_t *ext4_get_journal(struct super_block *sb, 5161 unsigned int journal_inum) 5162 { 5163 struct inode *journal_inode; 5164 journal_t *journal; 5165 5166 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 5167 return NULL; 5168 5169 journal_inode = ext4_get_journal_inode(sb, journal_inum); 5170 if (!journal_inode) 5171 return NULL; 5172 5173 journal = jbd2_journal_init_inode(journal_inode); 5174 if (!journal) { 5175 ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 5176 iput(journal_inode); 5177 return NULL; 5178 } 5179 journal->j_private = sb; 5180 ext4_init_journal_params(sb, journal); 5181 return journal; 5182 } 5183 5184 static journal_t *ext4_get_dev_journal(struct super_block *sb, 5185 dev_t j_dev) 5186 { 5187 struct buffer_head *bh; 5188 journal_t *journal; 5189 ext4_fsblk_t start; 5190 ext4_fsblk_t len; 5191 int hblock, blocksize; 5192 ext4_fsblk_t sb_block; 5193 unsigned long offset; 5194 struct ext4_super_block *es; 5195 struct block_device *bdev; 5196 5197 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 5198 return NULL; 5199 5200 bdev = ext4_blkdev_get(j_dev, sb); 5201 if (bdev == NULL) 5202 return NULL; 5203 5204 blocksize = sb->s_blocksize; 5205 hblock = bdev_logical_block_size(bdev); 5206 if (blocksize < hblock) { 5207 ext4_msg(sb, KERN_ERR, 5208 "blocksize too small for journal device"); 5209 goto out_bdev; 5210 } 5211 5212 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 5213 offset = EXT4_MIN_BLOCK_SIZE % blocksize; 5214 set_blocksize(bdev, blocksize); 5215 if (!(bh = __bread(bdev, sb_block, blocksize))) { 5216 ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 5217 "external journal"); 5218 goto out_bdev; 5219 } 5220 5221 es = (struct ext4_super_block *) (bh->b_data + offset); 5222 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 5223 !(le32_to_cpu(es->s_feature_incompat) & 5224 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 5225 ext4_msg(sb, KERN_ERR, "external journal has " 5226 "bad superblock"); 5227 brelse(bh); 5228 goto out_bdev; 5229 } 5230 5231 if ((le32_to_cpu(es->s_feature_ro_compat) & 5232 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 5233 es->s_checksum != ext4_superblock_csum(sb, es)) { 5234 ext4_msg(sb, KERN_ERR, "external journal has " 5235 "corrupt superblock"); 5236 brelse(bh); 5237 goto out_bdev; 5238 } 5239 5240 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 5241 ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 5242 brelse(bh); 5243 goto out_bdev; 5244 } 5245 5246 len = ext4_blocks_count(es); 5247 start = sb_block + 1; 5248 brelse(bh); /* we're done with the superblock */ 5249 5250 journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 5251 start, len, blocksize); 5252 if (!journal) { 5253 ext4_msg(sb, KERN_ERR, "failed to create device journal"); 5254 goto out_bdev; 5255 } 5256 journal->j_private = sb; 5257 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) { 5258 ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 5259 goto out_journal; 5260 } 5261 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 5262 ext4_msg(sb, KERN_ERR, "External journal has more than one " 5263 "user (unsupported) - %d", 5264 be32_to_cpu(journal->j_superblock->s_nr_users)); 5265 goto out_journal; 5266 } 5267 EXT4_SB(sb)->s_journal_bdev = bdev; 5268 ext4_init_journal_params(sb, journal); 5269 return journal; 5270 5271 out_journal: 5272 jbd2_journal_destroy(journal); 5273 out_bdev: 5274 ext4_blkdev_put(bdev); 5275 return NULL; 5276 } 5277 5278 static int ext4_load_journal(struct super_block *sb, 5279 struct ext4_super_block *es, 5280 unsigned long journal_devnum) 5281 { 5282 journal_t *journal; 5283 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 5284 dev_t journal_dev; 5285 int err = 0; 5286 int really_read_only; 5287 int journal_dev_ro; 5288 5289 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 5290 return -EFSCORRUPTED; 5291 5292 if (journal_devnum && 5293 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 5294 ext4_msg(sb, KERN_INFO, "external journal device major/minor " 5295 "numbers have changed"); 5296 journal_dev = new_decode_dev(journal_devnum); 5297 } else 5298 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 5299 5300 if (journal_inum && journal_dev) { 5301 ext4_msg(sb, KERN_ERR, 5302 "filesystem has both journal inode and journal device!"); 5303 return -EINVAL; 5304 } 5305 5306 if (journal_inum) { 5307 journal = ext4_get_journal(sb, journal_inum); 5308 if (!journal) 5309 return -EINVAL; 5310 } else { 5311 journal = ext4_get_dev_journal(sb, journal_dev); 5312 if (!journal) 5313 return -EINVAL; 5314 } 5315 5316 journal_dev_ro = bdev_read_only(journal->j_dev); 5317 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro; 5318 5319 if (journal_dev_ro && !sb_rdonly(sb)) { 5320 ext4_msg(sb, KERN_ERR, 5321 "journal device read-only, try mounting with '-o ro'"); 5322 err = -EROFS; 5323 goto err_out; 5324 } 5325 5326 /* 5327 * Are we loading a blank journal or performing recovery after a 5328 * crash? For recovery, we need to check in advance whether we 5329 * can get read-write access to the device. 5330 */ 5331 if (ext4_has_feature_journal_needs_recovery(sb)) { 5332 if (sb_rdonly(sb)) { 5333 ext4_msg(sb, KERN_INFO, "INFO: recovery " 5334 "required on readonly filesystem"); 5335 if (really_read_only) { 5336 ext4_msg(sb, KERN_ERR, "write access " 5337 "unavailable, cannot proceed " 5338 "(try mounting with noload)"); 5339 err = -EROFS; 5340 goto err_out; 5341 } 5342 ext4_msg(sb, KERN_INFO, "write access will " 5343 "be enabled during recovery"); 5344 } 5345 } 5346 5347 if (!(journal->j_flags & JBD2_BARRIER)) 5348 ext4_msg(sb, KERN_INFO, "barriers disabled"); 5349 5350 if (!ext4_has_feature_journal_needs_recovery(sb)) 5351 err = jbd2_journal_wipe(journal, !really_read_only); 5352 if (!err) { 5353 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 5354 if (save) 5355 memcpy(save, ((char *) es) + 5356 EXT4_S_ERR_START, EXT4_S_ERR_LEN); 5357 err = jbd2_journal_load(journal); 5358 if (save) 5359 memcpy(((char *) es) + EXT4_S_ERR_START, 5360 save, EXT4_S_ERR_LEN); 5361 kfree(save); 5362 } 5363 5364 if (err) { 5365 ext4_msg(sb, KERN_ERR, "error loading journal"); 5366 goto err_out; 5367 } 5368 5369 EXT4_SB(sb)->s_journal = journal; 5370 err = ext4_clear_journal_err(sb, es); 5371 if (err) { 5372 EXT4_SB(sb)->s_journal = NULL; 5373 jbd2_journal_destroy(journal); 5374 return err; 5375 } 5376 5377 if (!really_read_only && journal_devnum && 5378 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 5379 es->s_journal_dev = cpu_to_le32(journal_devnum); 5380 5381 /* Make sure we flush the recovery flag to disk. */ 5382 ext4_commit_super(sb); 5383 } 5384 5385 return 0; 5386 5387 err_out: 5388 jbd2_journal_destroy(journal); 5389 return err; 5390 } 5391 5392 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */ 5393 static void ext4_update_super(struct super_block *sb) 5394 { 5395 struct ext4_sb_info *sbi = EXT4_SB(sb); 5396 struct ext4_super_block *es = sbi->s_es; 5397 struct buffer_head *sbh = sbi->s_sbh; 5398 5399 lock_buffer(sbh); 5400 /* 5401 * If the file system is mounted read-only, don't update the 5402 * superblock write time. This avoids updating the superblock 5403 * write time when we are mounting the root file system 5404 * read/only but we need to replay the journal; at that point, 5405 * for people who are east of GMT and who make their clock 5406 * tick in localtime for Windows bug-for-bug compatibility, 5407 * the clock is set in the future, and this will cause e2fsck 5408 * to complain and force a full file system check. 5409 */ 5410 if (!(sb->s_flags & SB_RDONLY)) 5411 ext4_update_tstamp(es, s_wtime); 5412 es->s_kbytes_written = 5413 cpu_to_le64(sbi->s_kbytes_written + 5414 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) - 5415 sbi->s_sectors_written_start) >> 1)); 5416 if (percpu_counter_initialized(&sbi->s_freeclusters_counter)) 5417 ext4_free_blocks_count_set(es, 5418 EXT4_C2B(sbi, percpu_counter_sum_positive( 5419 &sbi->s_freeclusters_counter))); 5420 if (percpu_counter_initialized(&sbi->s_freeinodes_counter)) 5421 es->s_free_inodes_count = 5422 cpu_to_le32(percpu_counter_sum_positive( 5423 &sbi->s_freeinodes_counter)); 5424 /* Copy error information to the on-disk superblock */ 5425 spin_lock(&sbi->s_error_lock); 5426 if (sbi->s_add_error_count > 0) { 5427 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 5428 if (!es->s_first_error_time && !es->s_first_error_time_hi) { 5429 __ext4_update_tstamp(&es->s_first_error_time, 5430 &es->s_first_error_time_hi, 5431 sbi->s_first_error_time); 5432 strncpy(es->s_first_error_func, sbi->s_first_error_func, 5433 sizeof(es->s_first_error_func)); 5434 es->s_first_error_line = 5435 cpu_to_le32(sbi->s_first_error_line); 5436 es->s_first_error_ino = 5437 cpu_to_le32(sbi->s_first_error_ino); 5438 es->s_first_error_block = 5439 cpu_to_le64(sbi->s_first_error_block); 5440 es->s_first_error_errcode = 5441 ext4_errno_to_code(sbi->s_first_error_code); 5442 } 5443 __ext4_update_tstamp(&es->s_last_error_time, 5444 &es->s_last_error_time_hi, 5445 sbi->s_last_error_time); 5446 strncpy(es->s_last_error_func, sbi->s_last_error_func, 5447 sizeof(es->s_last_error_func)); 5448 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line); 5449 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino); 5450 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block); 5451 es->s_last_error_errcode = 5452 ext4_errno_to_code(sbi->s_last_error_code); 5453 /* 5454 * Start the daily error reporting function if it hasn't been 5455 * started already 5456 */ 5457 if (!es->s_error_count) 5458 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); 5459 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count); 5460 sbi->s_add_error_count = 0; 5461 } 5462 spin_unlock(&sbi->s_error_lock); 5463 5464 ext4_superblock_csum_set(sb); 5465 unlock_buffer(sbh); 5466 } 5467 5468 static int ext4_commit_super(struct super_block *sb) 5469 { 5470 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 5471 int error = 0; 5472 5473 if (!sbh) 5474 return -EINVAL; 5475 if (block_device_ejected(sb)) 5476 return -ENODEV; 5477 5478 ext4_update_super(sb); 5479 5480 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 5481 /* 5482 * Oh, dear. A previous attempt to write the 5483 * superblock failed. This could happen because the 5484 * USB device was yanked out. Or it could happen to 5485 * be a transient write error and maybe the block will 5486 * be remapped. Nothing we can do but to retry the 5487 * write and hope for the best. 5488 */ 5489 ext4_msg(sb, KERN_ERR, "previous I/O error to " 5490 "superblock detected"); 5491 clear_buffer_write_io_error(sbh); 5492 set_buffer_uptodate(sbh); 5493 } 5494 BUFFER_TRACE(sbh, "marking dirty"); 5495 mark_buffer_dirty(sbh); 5496 error = __sync_dirty_buffer(sbh, 5497 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0)); 5498 if (buffer_write_io_error(sbh)) { 5499 ext4_msg(sb, KERN_ERR, "I/O error while writing " 5500 "superblock"); 5501 clear_buffer_write_io_error(sbh); 5502 set_buffer_uptodate(sbh); 5503 } 5504 return error; 5505 } 5506 5507 /* 5508 * Have we just finished recovery? If so, and if we are mounting (or 5509 * remounting) the filesystem readonly, then we will end up with a 5510 * consistent fs on disk. Record that fact. 5511 */ 5512 static int ext4_mark_recovery_complete(struct super_block *sb, 5513 struct ext4_super_block *es) 5514 { 5515 int err; 5516 journal_t *journal = EXT4_SB(sb)->s_journal; 5517 5518 if (!ext4_has_feature_journal(sb)) { 5519 if (journal != NULL) { 5520 ext4_error(sb, "Journal got removed while the fs was " 5521 "mounted!"); 5522 return -EFSCORRUPTED; 5523 } 5524 return 0; 5525 } 5526 jbd2_journal_lock_updates(journal); 5527 err = jbd2_journal_flush(journal, 0); 5528 if (err < 0) 5529 goto out; 5530 5531 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) || 5532 ext4_has_feature_orphan_present(sb))) { 5533 if (!ext4_orphan_file_empty(sb)) { 5534 ext4_error(sb, "Orphan file not empty on read-only fs."); 5535 err = -EFSCORRUPTED; 5536 goto out; 5537 } 5538 ext4_clear_feature_journal_needs_recovery(sb); 5539 ext4_clear_feature_orphan_present(sb); 5540 ext4_commit_super(sb); 5541 } 5542 out: 5543 jbd2_journal_unlock_updates(journal); 5544 return err; 5545 } 5546 5547 /* 5548 * If we are mounting (or read-write remounting) a filesystem whose journal 5549 * has recorded an error from a previous lifetime, move that error to the 5550 * main filesystem now. 5551 */ 5552 static int ext4_clear_journal_err(struct super_block *sb, 5553 struct ext4_super_block *es) 5554 { 5555 journal_t *journal; 5556 int j_errno; 5557 const char *errstr; 5558 5559 if (!ext4_has_feature_journal(sb)) { 5560 ext4_error(sb, "Journal got removed while the fs was mounted!"); 5561 return -EFSCORRUPTED; 5562 } 5563 5564 journal = EXT4_SB(sb)->s_journal; 5565 5566 /* 5567 * Now check for any error status which may have been recorded in the 5568 * journal by a prior ext4_error() or ext4_abort() 5569 */ 5570 5571 j_errno = jbd2_journal_errno(journal); 5572 if (j_errno) { 5573 char nbuf[16]; 5574 5575 errstr = ext4_decode_error(sb, j_errno, nbuf); 5576 ext4_warning(sb, "Filesystem error recorded " 5577 "from previous mount: %s", errstr); 5578 ext4_warning(sb, "Marking fs in need of filesystem check."); 5579 5580 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 5581 es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 5582 ext4_commit_super(sb); 5583 5584 jbd2_journal_clear_err(journal); 5585 jbd2_journal_update_sb_errno(journal); 5586 } 5587 return 0; 5588 } 5589 5590 /* 5591 * Force the running and committing transactions to commit, 5592 * and wait on the commit. 5593 */ 5594 int ext4_force_commit(struct super_block *sb) 5595 { 5596 journal_t *journal; 5597 5598 if (sb_rdonly(sb)) 5599 return 0; 5600 5601 journal = EXT4_SB(sb)->s_journal; 5602 return ext4_journal_force_commit(journal); 5603 } 5604 5605 static int ext4_sync_fs(struct super_block *sb, int wait) 5606 { 5607 int ret = 0; 5608 tid_t target; 5609 bool needs_barrier = false; 5610 struct ext4_sb_info *sbi = EXT4_SB(sb); 5611 5612 if (unlikely(ext4_forced_shutdown(sbi))) 5613 return 0; 5614 5615 trace_ext4_sync_fs(sb, wait); 5616 flush_workqueue(sbi->rsv_conversion_wq); 5617 /* 5618 * Writeback quota in non-journalled quota case - journalled quota has 5619 * no dirty dquots 5620 */ 5621 dquot_writeback_dquots(sb, -1); 5622 /* 5623 * Data writeback is possible w/o journal transaction, so barrier must 5624 * being sent at the end of the function. But we can skip it if 5625 * transaction_commit will do it for us. 5626 */ 5627 if (sbi->s_journal) { 5628 target = jbd2_get_latest_transaction(sbi->s_journal); 5629 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 5630 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 5631 needs_barrier = true; 5632 5633 if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 5634 if (wait) 5635 ret = jbd2_log_wait_commit(sbi->s_journal, 5636 target); 5637 } 5638 } else if (wait && test_opt(sb, BARRIER)) 5639 needs_barrier = true; 5640 if (needs_barrier) { 5641 int err; 5642 err = blkdev_issue_flush(sb->s_bdev); 5643 if (!ret) 5644 ret = err; 5645 } 5646 5647 return ret; 5648 } 5649 5650 /* 5651 * LVM calls this function before a (read-only) snapshot is created. This 5652 * gives us a chance to flush the journal completely and mark the fs clean. 5653 * 5654 * Note that only this function cannot bring a filesystem to be in a clean 5655 * state independently. It relies on upper layer to stop all data & metadata 5656 * modifications. 5657 */ 5658 static int ext4_freeze(struct super_block *sb) 5659 { 5660 int error = 0; 5661 journal_t *journal; 5662 5663 if (sb_rdonly(sb)) 5664 return 0; 5665 5666 journal = EXT4_SB(sb)->s_journal; 5667 5668 if (journal) { 5669 /* Now we set up the journal barrier. */ 5670 jbd2_journal_lock_updates(journal); 5671 5672 /* 5673 * Don't clear the needs_recovery flag if we failed to 5674 * flush the journal. 5675 */ 5676 error = jbd2_journal_flush(journal, 0); 5677 if (error < 0) 5678 goto out; 5679 5680 /* Journal blocked and flushed, clear needs_recovery flag. */ 5681 ext4_clear_feature_journal_needs_recovery(sb); 5682 if (ext4_orphan_file_empty(sb)) 5683 ext4_clear_feature_orphan_present(sb); 5684 } 5685 5686 error = ext4_commit_super(sb); 5687 out: 5688 if (journal) 5689 /* we rely on upper layer to stop further updates */ 5690 jbd2_journal_unlock_updates(journal); 5691 return error; 5692 } 5693 5694 /* 5695 * Called by LVM after the snapshot is done. We need to reset the RECOVER 5696 * flag here, even though the filesystem is not technically dirty yet. 5697 */ 5698 static int ext4_unfreeze(struct super_block *sb) 5699 { 5700 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb))) 5701 return 0; 5702 5703 if (EXT4_SB(sb)->s_journal) { 5704 /* Reset the needs_recovery flag before the fs is unlocked. */ 5705 ext4_set_feature_journal_needs_recovery(sb); 5706 if (ext4_has_feature_orphan_file(sb)) 5707 ext4_set_feature_orphan_present(sb); 5708 } 5709 5710 ext4_commit_super(sb); 5711 return 0; 5712 } 5713 5714 /* 5715 * Structure to save mount options for ext4_remount's benefit 5716 */ 5717 struct ext4_mount_options { 5718 unsigned long s_mount_opt; 5719 unsigned long s_mount_opt2; 5720 kuid_t s_resuid; 5721 kgid_t s_resgid; 5722 unsigned long s_commit_interval; 5723 u32 s_min_batch_time, s_max_batch_time; 5724 #ifdef CONFIG_QUOTA 5725 int s_jquota_fmt; 5726 char *s_qf_names[EXT4_MAXQUOTAS]; 5727 #endif 5728 }; 5729 5730 static int ext4_remount(struct super_block *sb, int *flags, char *data) 5731 { 5732 struct ext4_super_block *es; 5733 struct ext4_sb_info *sbi = EXT4_SB(sb); 5734 unsigned long old_sb_flags, vfs_flags; 5735 struct ext4_mount_options old_opts; 5736 ext4_group_t g; 5737 int err = 0; 5738 #ifdef CONFIG_QUOTA 5739 int enable_quota = 0; 5740 int i, j; 5741 char *to_free[EXT4_MAXQUOTAS]; 5742 #endif 5743 char *orig_data = kstrdup(data, GFP_KERNEL); 5744 struct ext4_parsed_options parsed_opts; 5745 5746 parsed_opts.journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 5747 parsed_opts.journal_devnum = 0; 5748 5749 if (data && !orig_data) 5750 return -ENOMEM; 5751 5752 /* Store the original options */ 5753 old_sb_flags = sb->s_flags; 5754 old_opts.s_mount_opt = sbi->s_mount_opt; 5755 old_opts.s_mount_opt2 = sbi->s_mount_opt2; 5756 old_opts.s_resuid = sbi->s_resuid; 5757 old_opts.s_resgid = sbi->s_resgid; 5758 old_opts.s_commit_interval = sbi->s_commit_interval; 5759 old_opts.s_min_batch_time = sbi->s_min_batch_time; 5760 old_opts.s_max_batch_time = sbi->s_max_batch_time; 5761 #ifdef CONFIG_QUOTA 5762 old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 5763 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5764 if (sbi->s_qf_names[i]) { 5765 char *qf_name = get_qf_name(sb, sbi, i); 5766 5767 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); 5768 if (!old_opts.s_qf_names[i]) { 5769 for (j = 0; j < i; j++) 5770 kfree(old_opts.s_qf_names[j]); 5771 kfree(orig_data); 5772 return -ENOMEM; 5773 } 5774 } else 5775 old_opts.s_qf_names[i] = NULL; 5776 #endif 5777 if (sbi->s_journal && sbi->s_journal->j_task->io_context) 5778 parsed_opts.journal_ioprio = 5779 sbi->s_journal->j_task->io_context->ioprio; 5780 5781 /* 5782 * Some options can be enabled by ext4 and/or by VFS mount flag 5783 * either way we need to make sure it matches in both *flags and 5784 * s_flags. Copy those selected flags from *flags to s_flags 5785 */ 5786 vfs_flags = SB_LAZYTIME | SB_I_VERSION; 5787 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags); 5788 5789 if (!parse_options(data, sb, &parsed_opts, 1)) { 5790 err = -EINVAL; 5791 goto restore_opts; 5792 } 5793 5794 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 5795 test_opt(sb, JOURNAL_CHECKSUM)) { 5796 ext4_msg(sb, KERN_ERR, "changing journal_checksum " 5797 "during remount not supported; ignoring"); 5798 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 5799 } 5800 5801 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 5802 if (test_opt2(sb, EXPLICIT_DELALLOC)) { 5803 ext4_msg(sb, KERN_ERR, "can't mount with " 5804 "both data=journal and delalloc"); 5805 err = -EINVAL; 5806 goto restore_opts; 5807 } 5808 if (test_opt(sb, DIOREAD_NOLOCK)) { 5809 ext4_msg(sb, KERN_ERR, "can't mount with " 5810 "both data=journal and dioread_nolock"); 5811 err = -EINVAL; 5812 goto restore_opts; 5813 } 5814 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { 5815 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 5816 ext4_msg(sb, KERN_ERR, "can't mount with " 5817 "journal_async_commit in data=ordered mode"); 5818 err = -EINVAL; 5819 goto restore_opts; 5820 } 5821 } 5822 5823 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { 5824 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); 5825 err = -EINVAL; 5826 goto restore_opts; 5827 } 5828 5829 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) 5830 ext4_abort(sb, ESHUTDOWN, "Abort forced by user"); 5831 5832 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 5833 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 5834 5835 es = sbi->s_es; 5836 5837 if (sbi->s_journal) { 5838 ext4_init_journal_params(sb, sbi->s_journal); 5839 set_task_ioprio(sbi->s_journal->j_task, parsed_opts.journal_ioprio); 5840 } 5841 5842 /* Flush outstanding errors before changing fs state */ 5843 flush_work(&sbi->s_error_work); 5844 5845 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) { 5846 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) { 5847 err = -EROFS; 5848 goto restore_opts; 5849 } 5850 5851 if (*flags & SB_RDONLY) { 5852 err = sync_filesystem(sb); 5853 if (err < 0) 5854 goto restore_opts; 5855 err = dquot_suspend(sb, -1); 5856 if (err < 0) 5857 goto restore_opts; 5858 5859 /* 5860 * First of all, the unconditional stuff we have to do 5861 * to disable replay of the journal when we next remount 5862 */ 5863 sb->s_flags |= SB_RDONLY; 5864 5865 /* 5866 * OK, test if we are remounting a valid rw partition 5867 * readonly, and if so set the rdonly flag and then 5868 * mark the partition as valid again. 5869 */ 5870 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 5871 (sbi->s_mount_state & EXT4_VALID_FS)) 5872 es->s_state = cpu_to_le16(sbi->s_mount_state); 5873 5874 if (sbi->s_journal) { 5875 /* 5876 * We let remount-ro finish even if marking fs 5877 * as clean failed... 5878 */ 5879 ext4_mark_recovery_complete(sb, es); 5880 } 5881 } else { 5882 /* Make sure we can mount this feature set readwrite */ 5883 if (ext4_has_feature_readonly(sb) || 5884 !ext4_feature_set_ok(sb, 0)) { 5885 err = -EROFS; 5886 goto restore_opts; 5887 } 5888 /* 5889 * Make sure the group descriptor checksums 5890 * are sane. If they aren't, refuse to remount r/w. 5891 */ 5892 for (g = 0; g < sbi->s_groups_count; g++) { 5893 struct ext4_group_desc *gdp = 5894 ext4_get_group_desc(sb, g, NULL); 5895 5896 if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 5897 ext4_msg(sb, KERN_ERR, 5898 "ext4_remount: Checksum for group %u failed (%u!=%u)", 5899 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 5900 le16_to_cpu(gdp->bg_checksum)); 5901 err = -EFSBADCRC; 5902 goto restore_opts; 5903 } 5904 } 5905 5906 /* 5907 * If we have an unprocessed orphan list hanging 5908 * around from a previously readonly bdev mount, 5909 * require a full umount/remount for now. 5910 */ 5911 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) { 5912 ext4_msg(sb, KERN_WARNING, "Couldn't " 5913 "remount RDWR because of unprocessed " 5914 "orphan inode list. Please " 5915 "umount/remount instead"); 5916 err = -EINVAL; 5917 goto restore_opts; 5918 } 5919 5920 /* 5921 * Mounting a RDONLY partition read-write, so reread 5922 * and store the current valid flag. (It may have 5923 * been changed by e2fsck since we originally mounted 5924 * the partition.) 5925 */ 5926 if (sbi->s_journal) { 5927 err = ext4_clear_journal_err(sb, es); 5928 if (err) 5929 goto restore_opts; 5930 } 5931 sbi->s_mount_state = le16_to_cpu(es->s_state); 5932 5933 err = ext4_setup_super(sb, es, 0); 5934 if (err) 5935 goto restore_opts; 5936 5937 sb->s_flags &= ~SB_RDONLY; 5938 if (ext4_has_feature_mmp(sb)) 5939 if (ext4_multi_mount_protect(sb, 5940 le64_to_cpu(es->s_mmp_block))) { 5941 err = -EROFS; 5942 goto restore_opts; 5943 } 5944 #ifdef CONFIG_QUOTA 5945 enable_quota = 1; 5946 #endif 5947 } 5948 } 5949 5950 /* 5951 * Reinitialize lazy itable initialization thread based on 5952 * current settings 5953 */ 5954 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) 5955 ext4_unregister_li_request(sb); 5956 else { 5957 ext4_group_t first_not_zeroed; 5958 first_not_zeroed = ext4_has_uninit_itable(sb); 5959 ext4_register_li_request(sb, first_not_zeroed); 5960 } 5961 5962 /* 5963 * Handle creation of system zone data early because it can fail. 5964 * Releasing of existing data is done when we are sure remount will 5965 * succeed. 5966 */ 5967 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) { 5968 err = ext4_setup_system_zone(sb); 5969 if (err) 5970 goto restore_opts; 5971 } 5972 5973 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) { 5974 err = ext4_commit_super(sb); 5975 if (err) 5976 goto restore_opts; 5977 } 5978 5979 #ifdef CONFIG_QUOTA 5980 /* Release old quota file names */ 5981 for (i = 0; i < EXT4_MAXQUOTAS; i++) 5982 kfree(old_opts.s_qf_names[i]); 5983 if (enable_quota) { 5984 if (sb_any_quota_suspended(sb)) 5985 dquot_resume(sb, -1); 5986 else if (ext4_has_feature_quota(sb)) { 5987 err = ext4_enable_quotas(sb); 5988 if (err) 5989 goto restore_opts; 5990 } 5991 } 5992 #endif 5993 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 5994 ext4_release_system_zone(sb); 5995 5996 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb)) 5997 ext4_stop_mmpd(sbi); 5998 5999 /* 6000 * Some options can be enabled by ext4 and/or by VFS mount flag 6001 * either way we need to make sure it matches in both *flags and 6002 * s_flags. Copy those selected flags from s_flags to *flags 6003 */ 6004 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags); 6005 6006 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s. Quota mode: %s.", 6007 orig_data, ext4_quota_mode(sb)); 6008 kfree(orig_data); 6009 return 0; 6010 6011 restore_opts: 6012 sb->s_flags = old_sb_flags; 6013 sbi->s_mount_opt = old_opts.s_mount_opt; 6014 sbi->s_mount_opt2 = old_opts.s_mount_opt2; 6015 sbi->s_resuid = old_opts.s_resuid; 6016 sbi->s_resgid = old_opts.s_resgid; 6017 sbi->s_commit_interval = old_opts.s_commit_interval; 6018 sbi->s_min_batch_time = old_opts.s_min_batch_time; 6019 sbi->s_max_batch_time = old_opts.s_max_batch_time; 6020 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 6021 ext4_release_system_zone(sb); 6022 #ifdef CONFIG_QUOTA 6023 sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 6024 for (i = 0; i < EXT4_MAXQUOTAS; i++) { 6025 to_free[i] = get_qf_name(sb, sbi, i); 6026 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); 6027 } 6028 synchronize_rcu(); 6029 for (i = 0; i < EXT4_MAXQUOTAS; i++) 6030 kfree(to_free[i]); 6031 #endif 6032 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb)) 6033 ext4_stop_mmpd(sbi); 6034 kfree(orig_data); 6035 return err; 6036 } 6037 6038 #ifdef CONFIG_QUOTA 6039 static int ext4_statfs_project(struct super_block *sb, 6040 kprojid_t projid, struct kstatfs *buf) 6041 { 6042 struct kqid qid; 6043 struct dquot *dquot; 6044 u64 limit; 6045 u64 curblock; 6046 6047 qid = make_kqid_projid(projid); 6048 dquot = dqget(sb, qid); 6049 if (IS_ERR(dquot)) 6050 return PTR_ERR(dquot); 6051 spin_lock(&dquot->dq_dqb_lock); 6052 6053 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 6054 dquot->dq_dqb.dqb_bhardlimit); 6055 limit >>= sb->s_blocksize_bits; 6056 6057 if (limit && buf->f_blocks > limit) { 6058 curblock = (dquot->dq_dqb.dqb_curspace + 6059 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 6060 buf->f_blocks = limit; 6061 buf->f_bfree = buf->f_bavail = 6062 (buf->f_blocks > curblock) ? 6063 (buf->f_blocks - curblock) : 0; 6064 } 6065 6066 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 6067 dquot->dq_dqb.dqb_ihardlimit); 6068 if (limit && buf->f_files > limit) { 6069 buf->f_files = limit; 6070 buf->f_ffree = 6071 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 6072 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 6073 } 6074 6075 spin_unlock(&dquot->dq_dqb_lock); 6076 dqput(dquot); 6077 return 0; 6078 } 6079 #endif 6080 6081 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 6082 { 6083 struct super_block *sb = dentry->d_sb; 6084 struct ext4_sb_info *sbi = EXT4_SB(sb); 6085 struct ext4_super_block *es = sbi->s_es; 6086 ext4_fsblk_t overhead = 0, resv_blocks; 6087 s64 bfree; 6088 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 6089 6090 if (!test_opt(sb, MINIX_DF)) 6091 overhead = sbi->s_overhead; 6092 6093 buf->f_type = EXT4_SUPER_MAGIC; 6094 buf->f_bsize = sb->s_blocksize; 6095 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 6096 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 6097 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 6098 /* prevent underflow in case that few free space is available */ 6099 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 6100 buf->f_bavail = buf->f_bfree - 6101 (ext4_r_blocks_count(es) + resv_blocks); 6102 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 6103 buf->f_bavail = 0; 6104 buf->f_files = le32_to_cpu(es->s_inodes_count); 6105 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 6106 buf->f_namelen = EXT4_NAME_LEN; 6107 buf->f_fsid = uuid_to_fsid(es->s_uuid); 6108 6109 #ifdef CONFIG_QUOTA 6110 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 6111 sb_has_quota_limits_enabled(sb, PRJQUOTA)) 6112 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 6113 #endif 6114 return 0; 6115 } 6116 6117 6118 #ifdef CONFIG_QUOTA 6119 6120 /* 6121 * Helper functions so that transaction is started before we acquire dqio_sem 6122 * to keep correct lock ordering of transaction > dqio_sem 6123 */ 6124 static inline struct inode *dquot_to_inode(struct dquot *dquot) 6125 { 6126 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 6127 } 6128 6129 static int ext4_write_dquot(struct dquot *dquot) 6130 { 6131 int ret, err; 6132 handle_t *handle; 6133 struct inode *inode; 6134 6135 inode = dquot_to_inode(dquot); 6136 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 6137 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 6138 if (IS_ERR(handle)) 6139 return PTR_ERR(handle); 6140 ret = dquot_commit(dquot); 6141 err = ext4_journal_stop(handle); 6142 if (!ret) 6143 ret = err; 6144 return ret; 6145 } 6146 6147 static int ext4_acquire_dquot(struct dquot *dquot) 6148 { 6149 int ret, err; 6150 handle_t *handle; 6151 6152 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 6153 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 6154 if (IS_ERR(handle)) 6155 return PTR_ERR(handle); 6156 ret = dquot_acquire(dquot); 6157 err = ext4_journal_stop(handle); 6158 if (!ret) 6159 ret = err; 6160 return ret; 6161 } 6162 6163 static int ext4_release_dquot(struct dquot *dquot) 6164 { 6165 int ret, err; 6166 handle_t *handle; 6167 6168 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 6169 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 6170 if (IS_ERR(handle)) { 6171 /* Release dquot anyway to avoid endless cycle in dqput() */ 6172 dquot_release(dquot); 6173 return PTR_ERR(handle); 6174 } 6175 ret = dquot_release(dquot); 6176 err = ext4_journal_stop(handle); 6177 if (!ret) 6178 ret = err; 6179 return ret; 6180 } 6181 6182 static int ext4_mark_dquot_dirty(struct dquot *dquot) 6183 { 6184 struct super_block *sb = dquot->dq_sb; 6185 6186 if (ext4_is_quota_journalled(sb)) { 6187 dquot_mark_dquot_dirty(dquot); 6188 return ext4_write_dquot(dquot); 6189 } else { 6190 return dquot_mark_dquot_dirty(dquot); 6191 } 6192 } 6193 6194 static int ext4_write_info(struct super_block *sb, int type) 6195 { 6196 int ret, err; 6197 handle_t *handle; 6198 6199 /* Data block + inode block */ 6200 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2); 6201 if (IS_ERR(handle)) 6202 return PTR_ERR(handle); 6203 ret = dquot_commit_info(sb, type); 6204 err = ext4_journal_stop(handle); 6205 if (!ret) 6206 ret = err; 6207 return ret; 6208 } 6209 6210 static void lockdep_set_quota_inode(struct inode *inode, int subclass) 6211 { 6212 struct ext4_inode_info *ei = EXT4_I(inode); 6213 6214 /* The first argument of lockdep_set_subclass has to be 6215 * *exactly* the same as the argument to init_rwsem() --- in 6216 * this case, in init_once() --- or lockdep gets unhappy 6217 * because the name of the lock is set using the 6218 * stringification of the argument to init_rwsem(). 6219 */ 6220 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 6221 lockdep_set_subclass(&ei->i_data_sem, subclass); 6222 } 6223 6224 /* 6225 * Standard function to be called on quota_on 6226 */ 6227 static int ext4_quota_on(struct super_block *sb, int type, int format_id, 6228 const struct path *path) 6229 { 6230 int err; 6231 6232 if (!test_opt(sb, QUOTA)) 6233 return -EINVAL; 6234 6235 /* Quotafile not on the same filesystem? */ 6236 if (path->dentry->d_sb != sb) 6237 return -EXDEV; 6238 6239 /* Quota already enabled for this file? */ 6240 if (IS_NOQUOTA(d_inode(path->dentry))) 6241 return -EBUSY; 6242 6243 /* Journaling quota? */ 6244 if (EXT4_SB(sb)->s_qf_names[type]) { 6245 /* Quotafile not in fs root? */ 6246 if (path->dentry->d_parent != sb->s_root) 6247 ext4_msg(sb, KERN_WARNING, 6248 "Quota file not on filesystem root. " 6249 "Journaled quota will not work"); 6250 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY; 6251 } else { 6252 /* 6253 * Clear the flag just in case mount options changed since 6254 * last time. 6255 */ 6256 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY; 6257 } 6258 6259 /* 6260 * When we journal data on quota file, we have to flush journal to see 6261 * all updates to the file when we bypass pagecache... 6262 */ 6263 if (EXT4_SB(sb)->s_journal && 6264 ext4_should_journal_data(d_inode(path->dentry))) { 6265 /* 6266 * We don't need to lock updates but journal_flush() could 6267 * otherwise be livelocked... 6268 */ 6269 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 6270 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0); 6271 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 6272 if (err) 6273 return err; 6274 } 6275 6276 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 6277 err = dquot_quota_on(sb, type, format_id, path); 6278 if (err) { 6279 lockdep_set_quota_inode(path->dentry->d_inode, 6280 I_DATA_SEM_NORMAL); 6281 } else { 6282 struct inode *inode = d_inode(path->dentry); 6283 handle_t *handle; 6284 6285 /* 6286 * Set inode flags to prevent userspace from messing with quota 6287 * files. If this fails, we return success anyway since quotas 6288 * are already enabled and this is not a hard failure. 6289 */ 6290 inode_lock(inode); 6291 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 6292 if (IS_ERR(handle)) 6293 goto unlock_inode; 6294 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; 6295 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 6296 S_NOATIME | S_IMMUTABLE); 6297 err = ext4_mark_inode_dirty(handle, inode); 6298 ext4_journal_stop(handle); 6299 unlock_inode: 6300 inode_unlock(inode); 6301 } 6302 return err; 6303 } 6304 6305 static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 6306 unsigned int flags) 6307 { 6308 int err; 6309 struct inode *qf_inode; 6310 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 6311 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 6312 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 6313 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 6314 }; 6315 6316 BUG_ON(!ext4_has_feature_quota(sb)); 6317 6318 if (!qf_inums[type]) 6319 return -EPERM; 6320 6321 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL); 6322 if (IS_ERR(qf_inode)) { 6323 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 6324 return PTR_ERR(qf_inode); 6325 } 6326 6327 /* Don't account quota for quota files to avoid recursion */ 6328 qf_inode->i_flags |= S_NOQUOTA; 6329 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 6330 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 6331 if (err) 6332 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 6333 iput(qf_inode); 6334 6335 return err; 6336 } 6337 6338 /* Enable usage tracking for all quota types. */ 6339 int ext4_enable_quotas(struct super_block *sb) 6340 { 6341 int type, err = 0; 6342 unsigned long qf_inums[EXT4_MAXQUOTAS] = { 6343 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 6344 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 6345 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 6346 }; 6347 bool quota_mopt[EXT4_MAXQUOTAS] = { 6348 test_opt(sb, USRQUOTA), 6349 test_opt(sb, GRPQUOTA), 6350 test_opt(sb, PRJQUOTA), 6351 }; 6352 6353 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 6354 for (type = 0; type < EXT4_MAXQUOTAS; type++) { 6355 if (qf_inums[type]) { 6356 err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 6357 DQUOT_USAGE_ENABLED | 6358 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 6359 if (err) { 6360 ext4_warning(sb, 6361 "Failed to enable quota tracking " 6362 "(type=%d, err=%d). Please run " 6363 "e2fsck to fix.", type, err); 6364 for (type--; type >= 0; type--) 6365 dquot_quota_off(sb, type); 6366 6367 return err; 6368 } 6369 } 6370 } 6371 return 0; 6372 } 6373 6374 static int ext4_quota_off(struct super_block *sb, int type) 6375 { 6376 struct inode *inode = sb_dqopt(sb)->files[type]; 6377 handle_t *handle; 6378 int err; 6379 6380 /* Force all delayed allocation blocks to be allocated. 6381 * Caller already holds s_umount sem */ 6382 if (test_opt(sb, DELALLOC)) 6383 sync_filesystem(sb); 6384 6385 if (!inode || !igrab(inode)) 6386 goto out; 6387 6388 err = dquot_quota_off(sb, type); 6389 if (err || ext4_has_feature_quota(sb)) 6390 goto out_put; 6391 6392 inode_lock(inode); 6393 /* 6394 * Update modification times of quota files when userspace can 6395 * start looking at them. If we fail, we return success anyway since 6396 * this is not a hard failure and quotas are already disabled. 6397 */ 6398 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 6399 if (IS_ERR(handle)) { 6400 err = PTR_ERR(handle); 6401 goto out_unlock; 6402 } 6403 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); 6404 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 6405 inode->i_mtime = inode->i_ctime = current_time(inode); 6406 err = ext4_mark_inode_dirty(handle, inode); 6407 ext4_journal_stop(handle); 6408 out_unlock: 6409 inode_unlock(inode); 6410 out_put: 6411 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); 6412 iput(inode); 6413 return err; 6414 out: 6415 return dquot_quota_off(sb, type); 6416 } 6417 6418 /* Read data from quotafile - avoid pagecache and such because we cannot afford 6419 * acquiring the locks... As quota files are never truncated and quota code 6420 * itself serializes the operations (and no one else should touch the files) 6421 * we don't have to be afraid of races */ 6422 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 6423 size_t len, loff_t off) 6424 { 6425 struct inode *inode = sb_dqopt(sb)->files[type]; 6426 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 6427 int offset = off & (sb->s_blocksize - 1); 6428 int tocopy; 6429 size_t toread; 6430 struct buffer_head *bh; 6431 loff_t i_size = i_size_read(inode); 6432 6433 if (off > i_size) 6434 return 0; 6435 if (off+len > i_size) 6436 len = i_size-off; 6437 toread = len; 6438 while (toread > 0) { 6439 tocopy = sb->s_blocksize - offset < toread ? 6440 sb->s_blocksize - offset : toread; 6441 bh = ext4_bread(NULL, inode, blk, 0); 6442 if (IS_ERR(bh)) 6443 return PTR_ERR(bh); 6444 if (!bh) /* A hole? */ 6445 memset(data, 0, tocopy); 6446 else 6447 memcpy(data, bh->b_data+offset, tocopy); 6448 brelse(bh); 6449 offset = 0; 6450 toread -= tocopy; 6451 data += tocopy; 6452 blk++; 6453 } 6454 return len; 6455 } 6456 6457 /* Write to quotafile (we know the transaction is already started and has 6458 * enough credits) */ 6459 static ssize_t ext4_quota_write(struct super_block *sb, int type, 6460 const char *data, size_t len, loff_t off) 6461 { 6462 struct inode *inode = sb_dqopt(sb)->files[type]; 6463 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 6464 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1); 6465 int retries = 0; 6466 struct buffer_head *bh; 6467 handle_t *handle = journal_current_handle(); 6468 6469 if (EXT4_SB(sb)->s_journal && !handle) { 6470 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 6471 " cancelled because transaction is not started", 6472 (unsigned long long)off, (unsigned long long)len); 6473 return -EIO; 6474 } 6475 /* 6476 * Since we account only one data block in transaction credits, 6477 * then it is impossible to cross a block boundary. 6478 */ 6479 if (sb->s_blocksize - offset < len) { 6480 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 6481 " cancelled because not block aligned", 6482 (unsigned long long)off, (unsigned long long)len); 6483 return -EIO; 6484 } 6485 6486 do { 6487 bh = ext4_bread(handle, inode, blk, 6488 EXT4_GET_BLOCKS_CREATE | 6489 EXT4_GET_BLOCKS_METADATA_NOFAIL); 6490 } while (PTR_ERR(bh) == -ENOSPC && 6491 ext4_should_retry_alloc(inode->i_sb, &retries)); 6492 if (IS_ERR(bh)) 6493 return PTR_ERR(bh); 6494 if (!bh) 6495 goto out; 6496 BUFFER_TRACE(bh, "get write access"); 6497 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE); 6498 if (err) { 6499 brelse(bh); 6500 return err; 6501 } 6502 lock_buffer(bh); 6503 memcpy(bh->b_data+offset, data, len); 6504 flush_dcache_page(bh->b_page); 6505 unlock_buffer(bh); 6506 err = ext4_handle_dirty_metadata(handle, NULL, bh); 6507 brelse(bh); 6508 out: 6509 if (inode->i_size < off + len) { 6510 i_size_write(inode, off + len); 6511 EXT4_I(inode)->i_disksize = inode->i_size; 6512 err2 = ext4_mark_inode_dirty(handle, inode); 6513 if (unlikely(err2 && !err)) 6514 err = err2; 6515 } 6516 return err ? err : len; 6517 } 6518 #endif 6519 6520 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 6521 const char *dev_name, void *data) 6522 { 6523 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 6524 } 6525 6526 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 6527 static inline void register_as_ext2(void) 6528 { 6529 int err = register_filesystem(&ext2_fs_type); 6530 if (err) 6531 printk(KERN_WARNING 6532 "EXT4-fs: Unable to register as ext2 (%d)\n", err); 6533 } 6534 6535 static inline void unregister_as_ext2(void) 6536 { 6537 unregister_filesystem(&ext2_fs_type); 6538 } 6539 6540 static inline int ext2_feature_set_ok(struct super_block *sb) 6541 { 6542 if (ext4_has_unknown_ext2_incompat_features(sb)) 6543 return 0; 6544 if (sb_rdonly(sb)) 6545 return 1; 6546 if (ext4_has_unknown_ext2_ro_compat_features(sb)) 6547 return 0; 6548 return 1; 6549 } 6550 #else 6551 static inline void register_as_ext2(void) { } 6552 static inline void unregister_as_ext2(void) { } 6553 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 6554 #endif 6555 6556 static inline void register_as_ext3(void) 6557 { 6558 int err = register_filesystem(&ext3_fs_type); 6559 if (err) 6560 printk(KERN_WARNING 6561 "EXT4-fs: Unable to register as ext3 (%d)\n", err); 6562 } 6563 6564 static inline void unregister_as_ext3(void) 6565 { 6566 unregister_filesystem(&ext3_fs_type); 6567 } 6568 6569 static inline int ext3_feature_set_ok(struct super_block *sb) 6570 { 6571 if (ext4_has_unknown_ext3_incompat_features(sb)) 6572 return 0; 6573 if (!ext4_has_feature_journal(sb)) 6574 return 0; 6575 if (sb_rdonly(sb)) 6576 return 1; 6577 if (ext4_has_unknown_ext3_ro_compat_features(sb)) 6578 return 0; 6579 return 1; 6580 } 6581 6582 static struct file_system_type ext4_fs_type = { 6583 .owner = THIS_MODULE, 6584 .name = "ext4", 6585 .mount = ext4_mount, 6586 .kill_sb = kill_block_super, 6587 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP, 6588 }; 6589 MODULE_ALIAS_FS("ext4"); 6590 6591 /* Shared across all ext4 file systems */ 6592 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 6593 6594 static int __init ext4_init_fs(void) 6595 { 6596 int i, err; 6597 6598 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 6599 ext4_li_info = NULL; 6600 6601 /* Build-time check for flags consistency */ 6602 ext4_check_flag_values(); 6603 6604 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) 6605 init_waitqueue_head(&ext4__ioend_wq[i]); 6606 6607 err = ext4_init_es(); 6608 if (err) 6609 return err; 6610 6611 err = ext4_init_pending(); 6612 if (err) 6613 goto out7; 6614 6615 err = ext4_init_post_read_processing(); 6616 if (err) 6617 goto out6; 6618 6619 err = ext4_init_pageio(); 6620 if (err) 6621 goto out5; 6622 6623 err = ext4_init_system_zone(); 6624 if (err) 6625 goto out4; 6626 6627 err = ext4_init_sysfs(); 6628 if (err) 6629 goto out3; 6630 6631 err = ext4_init_mballoc(); 6632 if (err) 6633 goto out2; 6634 err = init_inodecache(); 6635 if (err) 6636 goto out1; 6637 6638 err = ext4_fc_init_dentry_cache(); 6639 if (err) 6640 goto out05; 6641 6642 register_as_ext3(); 6643 register_as_ext2(); 6644 err = register_filesystem(&ext4_fs_type); 6645 if (err) 6646 goto out; 6647 6648 return 0; 6649 out: 6650 unregister_as_ext2(); 6651 unregister_as_ext3(); 6652 out05: 6653 destroy_inodecache(); 6654 out1: 6655 ext4_exit_mballoc(); 6656 out2: 6657 ext4_exit_sysfs(); 6658 out3: 6659 ext4_exit_system_zone(); 6660 out4: 6661 ext4_exit_pageio(); 6662 out5: 6663 ext4_exit_post_read_processing(); 6664 out6: 6665 ext4_exit_pending(); 6666 out7: 6667 ext4_exit_es(); 6668 6669 return err; 6670 } 6671 6672 static void __exit ext4_exit_fs(void) 6673 { 6674 ext4_destroy_lazyinit_thread(); 6675 unregister_as_ext2(); 6676 unregister_as_ext3(); 6677 unregister_filesystem(&ext4_fs_type); 6678 destroy_inodecache(); 6679 ext4_exit_mballoc(); 6680 ext4_exit_sysfs(); 6681 ext4_exit_system_zone(); 6682 ext4_exit_pageio(); 6683 ext4_exit_post_read_processing(); 6684 ext4_exit_es(); 6685 ext4_exit_pending(); 6686 } 6687 6688 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 6689 MODULE_DESCRIPTION("Fourth Extended Filesystem"); 6690 MODULE_LICENSE("GPL"); 6691 MODULE_SOFTDEP("pre: crc32c"); 6692 module_init(ext4_init_fs) 6693 module_exit(ext4_exit_fs) 6694